Annals of the Rheumatic Diseases publishes original work on all aspects of rheumatology ARDThe EULAR Journal Editor Josef S Smolen (Austria) and disorders of connective tissue. Laboratory Associate Editors Francis Berenbaum (France) and clinical studies are equally welcome Dimitrios Boumpas (Greece) Gerd Burmester (Germany) Editorial Board Mary Crow (USA) Daniel Aletaha (Austria) Ingrid Lundberg (Sweden) Iain McInnes (UK) Johan Askling (Sweden) Gary MacFarlane (UK) Contact Details Thomas Pap (Germany) Sang-Cheol Bae (Korea) Xavier Mariette (France) David Pisetsky (USA) Xenofon Baraliakos (Germany) Alberto Martini (Italy) Editorial Office Désirée van der Heijde Anne Barton (UK) Marco Mattuci Cerinic (Italy) Annals of the Rheumatic Diseases Maarten Boers (The Netherlands) Dennis McGonagle (UK) (The Netherlands) BMJ Journals, BMA House, Tavistock Square Kazuhiko Yamamoto (Japan) Matthew Brown (Australia) Fred Miller (USA) Maya Buch (UK) Peter Nash (Australia) London WCIH 9JR,UK Methodological and Statistical Loreto Carmona (Spain) Michael Nurmohamed (The E: [email protected] Advisor Carlo Chizzolini (Switzerland) Netherlands) Stian Lydersen (Norway) Bernard Combe (France) Caroline Ospelt (Switzerland) Production Editor Social Media Advisors Philip Conaghan (UK) Monika Østensen (Norway) Teresa Jobson Maurizio Cutolo (Italy) Constatino Pitzalis (UK) Alessia Alunno (Italy) E: [email protected] Mary Canavan (Ireland) Nicola Dalbeth (Australia) Jane Salmon (USA) Christian Dejaco (Austria) Georg Schett (Germany) Meghna Jani (UK) Oliver Distler (Switzerland) José da Silva (Portugal) EULAR Elena Nikiphorou (UK) Thomas Dörner (Germany) Hendrik Schulze-Koops (Germany) EULAR Executive Secretariat Caroline Ospelt (Switzerland) Dirk Elewaut (Belgium) Nan Shen (China) Seestrasse 240, 8802 Kilchberg, Switzerland Christophe Richez (France) Axel Finckh (Switzerland) Alexander So (Switzerland) E: [email protected] Paul Studenic (Austria) Roy Fleischmann (USA) Hiroshi Takayanagi (Japan) www.eular.org Mary Goldring (USA) Tsutomu Takeuchi (Japan) Guidelines for Authors and Yoshiya Tanaka (Japan) Laure Gossec (France) Customer support Reviewers Walter Grassi (Italy) Dimitrios Vassilopoulos (Greece) For general queries and support with existing and Full instructions are available Ahmet Gül (Turkey) Douglas Veale (Ireland) Jiri Vencovsky (Czech Republic) new subscriptions: online at http://ard.bmj.com/ Frederic Houssiau (Belgium) Tom Huizinga (The Netherlands) Ronald van Vollenhoven (Sweden) W: support.bmj.com pages/authors. Articles must Arthur Kavanaugh (USA) Erwin Wagner (Spain) T: +44 (0)20 7111 1105 be submitted electronically at Michael Ward (USA) Robert Landewé (The Netherlands) E: [email protected] http://mc.manuscriptcentral. Zhan-Gou Li (China) Kevin Winthrop (USA) Rik Lories (Belgium) Huji Xu (China) com/ard. Authors retain copyright Self-archiving and permissions but are required to grant ARD Chairman of Advisory W: bmj.com/company/products-services/ an exclusive licence to publish. Committee rights-and-licensing/ (http://ard.bmj.com/pages/authors/). Johannes Bijlsma E: [email protected] (The Netherlands) Advertising Advisory Committee W: bmj.com/company/for-advertisers- Ferry Breedveld (The Netherlands) Zhan-guo Li (China) Marco Matucci Cerinic (Italy) Peter Lipsky (USA) and-sponsor/ Michael Doherty (UK) Sir Ravinder Maini (UK) Display Advertising ROW Maxime Dougados (France) Emilio Martín-Mola (Spain) Paul Emery (UK) Haralampos Moutsopoulos Sophie Fitzsimmons Daniel Furst (USA) (Greece) T: +44 (0)20 3655 5612 Steffen Gay (Switzerland) Karel Pavelka (Czech Republic) E: [email protected] Marc Hochberg (USA) Yehuda Shoenfeld (Israel) Joachim Kalden (Germany) Leo van de Putte (The Netherlands) Online Advertising ROW Edward Keystone (Canada) Frank Wollheim (Sweden) Marc Clifford Lars Klareskog (Sweden) Anthony Woolf (UK) T: +44 (0) 20 3655 5610 Tore Kvien (Norway) E: [email protected] Display & Online Advertising Americas American Medical Communications (AMC) T: +1 973 214 4374 Subscription Information E: [email protected] ARD is published monthly; subscribers receive all supplements Reprints ISSN 0003-4967 (print); 1468-2060 (online) Author Reprints Institutional Rates 2019 Personal Rates 2019 BMJ Reprints Team E: [email protected] Print Print (includes online access at no additional cost) Commercial Reprints ROW £1036 £428 Nadia Gurney-Randall M: +44 07866 262344 Online Online only £182 E: [email protected] Site licences are priced on FTE basis and allow access by the whole institution. Details available Eular congress delegates Commercial Reprints Americas online at http://journals.bmj.com/content/subscribers Delegates receive a Continuous Professional Ray Thibodeau or contact the Subscription Manager in the UK (see Development package that includes a 12 month T: +1 267 895 1758 above right) complimentary subscription to ARD in print M: +1 215 933 8484 and/or online E: [email protected] Personal print or online only and institutional print subscriptions may be purchased online at http://journals.bmj.com/content/ For all other journal contacts subscribers (payment by Visa/Mastercard only) ard.bmj.com/contact-us Residents of some EC countries must pay VAT; for details, call us or visit http://journals.bmj.com/content/subscribers For more information on subscription rates or to subscribe online please visit ard/bmj.com/pages/contact-us/ Contents Volume 78 Issue 6 | ARD June 2019

ARDThe Eular Journal June 2019 Volume 78 Issue 6 754 Effects of BI 655064, an antagonistic anti-CD40 78 Editorials 6 80th Year of Publication oue7 Ise6 Pages 715–866 Issue 6 Volume 78 715 Greetings from the editor antibody, on clinical and biomarker variables Impact Factor 12.350 J S Smolen in patients with active rheumatoid arthritis: a randomised, double-blind, placebo-controlled,

AnnAls of RheumAtIc DIseAses 716 Pathogenic effector functions of ACPA: Where phase IIa study do we stand? S Visvanathan, S Daniluk, R Ptaszyński, R Toes, D S Pisetsky U Müller-Ladner, M Ramanujam, B Rosenstock, A G Eleftheraki, R Vinisko, A Petříková, H Kellner, E Dokoupilova, B Kwiatkowska, R Alten,

June 2019 C Schwabe, P Baum, D Joseph, J S Fine, S J Padula, Views on news J Steffgen 723 Unexpected link between mitochondrial DNA and T cell help in systemic lupus 761 Synovial cellular and molecular signatures erythematosus stratify clinical response to csDMARD therapy Editor D S Pisetsky Josef S Smolen and predict radiographic progression in early Associate Editors rheumatoid arthritis patients Francis Berenbaum Dimitrios Boumpas F Humby, M Lewis, N Ramamoorthi, J A Hackney, Gerd Burmester Heroes and pillars of rheumatology M R Barnes, M Bombardieri, A F Setiadi, S Kelly, Mary Crow 725 Barbara Ansell: a person ahead of her time F Bene, M DiCicco, S Riahi, V Rocher, N Ng, Iain McInnes Thomas Pap P Woo, N M Wulffraat, A Tyndall I Lazarou, R Hands, D van der Heijde, David Pisetsky R B M Landewé, A van der Helm-van Mil, A Cauli, Désirée van der Heijde Kazuhiko Yamamoto I McInnes, C D Buckley, E H Choy, P C Taylor, Viewpoint M J Townsend, C Pitzalis Editorial office Annals of the Rheumatic Diseases 729 Current challenges in the development of new BMJ Publishing Group Ltd 773 Sequencing of the MHC region defines HLA- BMA House treatments for lupus Tavistock Square M Dall'Era, I N Bruce, C Gordon, S Manzi, DQA1 as the major genetic risk for seropositive London WCIH 9JR,UK J McCaffrey, P E Lipsky rheumatoid arthritis in Han Chinese population T: +44 (0)20 3655 5889 E: [email protected] J Guo, T Zhang, H Cao, X Li, H Liang, M Liu, Twitter: @ARD_BMJ Y Zou, Y Zhang, Y Wang, X Sun, F Hu, Y Du, ISSN: 0003-4967 (print) X Mo, X Liu, Y Yang, H Yang, X Wu, X Zhang, ISSN: 1468-2060 (online) Recommendation H Jia, H Jiang, Y Hou, X Liu, Y Su, M Zhang, Impact Factor: 12.350 736 2019 update of the EULAR recommendations H Yang, J Wang, L Sun, L Liu, L Padyukov, L Lai, Disclaimer: ARD is owned and published by for the management of systemic lupus BMJ Publishing Group Ltd (a wholly owned K Yamamoto, X Zhang, L Klareskog, X Xu, Z Li subsidiary of the British Medical Association) erythematosus and the European League Against Rheumatism. The owners grant editorial freedom tothe Editor of ARD. A Fanouriakis, M Kostopoulou, A Alunno, ARD follows guidelines on editorial independence M Aringer, I Bajema, J N Boletis, R Cervera, produced by the World Association of Medical Editors and the code on good publication practice of the A Doria, C Gordon, M Govoni, F Houssiau, Committee on Publication Ethics. D Jayne, M Kouloumas, A Kuhn, J L Larsen, ARD is intended for medical professionals and is provided without warranty, express or implied. K Lerstrøm, G Moroni, M Mosca, M Schneider, Statements in the journal are the responsibility J S Smolen, E Svenungsson, V Tesar, A Tincani, of their authors and advertisers and not authors’ institutions the BMJ Publishing Group, the A Troldborg, R van Vollenhoven, J Wenzel, European League Against Rheumatism or the G Bertsias, D T Boumpas BMA unless otherwise specified or determined by law. Acceptance of advertising does not imply endorsement. To the fullest extent permitted by law, the BMJ Publishing Group shall not be liable for any loss, injury or damage resulting from the use of ARD or any Rheumatoid arthritis information in it whether based on contract, tort, or 746 Gradual tapering TNF inhibitors versus otherwise. Readers are advised to verify any MORE CONTENTS ► information they choose to rely on. conventional synthetic DMARDs after achieving Copyright: © 2019 BMJ Publishing Group and European League Against Rheumatism. All rights controlled disease in patients with rheumatoid This article has been chosen by the Editor to be of special interest­ reserved; no part of this publication may be reproduced stored in a retrieval system, or arthritis: first-year results of the randomised or importance and is freely available online. transmitted in any form or by any means, electronic, controlled TARA study This article has been made freely available online under the BMJ mechanical, photocopying recording, or otherwise without prior permission E van Mulligen, P H P de Jong, T M Kuijper, Journals open access scheme. ARD is published by BMJ Publishing Group Ltd M van der Ven, C Appels, C Bijkerk, J B Harbers, See http://authors.bmj.com/open-access/ typeset by Exeter Premedia Services Private Ltd, Chennai, India and printed in the UK on acid-free Y de Man, T H E Molenaar, I Tchetverikov, paper. Y P M Goekoop-Ruiterman, J van Zeben, Annals of the Rheumatic Diseases (ISSN No: ­ Member since 2008 0003-4967) is published monthly by BMJ Publishing J M W Hazes, A E A M Weel, J J Luime JM00004 Group and distributed in the USA by Air Business Ltd. This journal is a member of and subscribes to the principles of the Periodicals postage paid at Jamaica NY 11431 POSTMASTER: send address changes to Annals of Committee on Publication Ethics the Rheumatic Diseases, Air Business Ltd, c/o WN http://publicationethics.org/ Shipping USA, 156-15, 146th Avenue, 2nd Floor, Jamaica, NY 11434, USA. Contents Volume 78 Issue 6 | ARD June 2019

Early arthritis Pain 781 MRI inflammation of the hand interosseous 837 Are corticosteroid injections needed after tendons occurs in anti-CCP-positive at-risk needling and lavage of calcific tendinitis? individuals and may precede the development Randomised, double-blind, non-inferiority trial of clinical synovitis C Darrieutort-Laffite, S Varin, G Coiffier, K Mankia, M-A D’Agostino, E Rowbotham, J-D Albert, L Planche, Y Maugars, G Cormier, E MA Hensor, L Hunt, I Möller, M Miguel, B Le Goff J R Mérida-Velasco, J Murillo-González, E Naredo, J L Nam, A L Tan, J E Freeston, A Grainger, P Emery Epidemiology 844 The world-wide burden of musculoskeletal 787 Running promotes chronicity of arthritis by diseases: a systematic analysis of the World local modulation of complement activators and Health Organization Burden of Diseases impairing T regulatory feedback loops Database I Cambré, D Gaublomme, N Schryvers, E Sebbag, R Felten, F Sagez, J Sibilia, H Devilliers, S Lambrecht, R Lories, K Venken, D Elewaut L Arnaud

Psoriatic arthritis 796 Serum-based soluble markers differentiate Letters psoriatic arthritis from osteoarthritis 849 Comparisons of hepatitis C viral replication in V Chandran, F Abji, A V Perruccio, R Gandhi, S Li, patients with rheumatoid arthritis receiving R J Cook, D D Gladman tocilizumab, abatacept and tofacitinib therapy Y-M Chen, W-N Huang, T-L Liao, J-P Chen, S-S Yang, H-H Chen, T-Y Hsieh, W-T Hung, Systemic lupus erythematosus Y-H Chen, D-Y Chen 802 All-cause, cause-specific and age-specific standardised mortality ratios of patients with 850 Reproductive health outcomes in women with systemic lupus erythematosus in Ontario, psoriatic arthritis Canada over 43 years (1971–2013) K Murray, L Moore, F McAuliffe, D J Veale K Tselios, D D Gladman, B J Sheane, J Su, M Urowitz 852 Genetic association of non-MHC region with ankylosing spondylitis in a Chinese population J Liu, W Pu, Y Li, Y Ma, Q Zhu, W Wan, C Yang, Systemic sclerosis X Wang, X Chen, X Zhou, J D Reveille, L Jin, 807 Development and validation of the Scleroderma H Zou, J Wang Clinical Trials Consortium Damage Index (SCTC-DI): a novel instrument to quantify organ 853 Revisiting the issue of how to assess damage in systemic sclerosis pneumococcal vaccine immunogenicity: a post N Ferdowsi, M Huq, W Stevens, M Hudson, hoc analysis of antipneumococcal antibody M Wang, T Tay, J L Burchell, S Mancuso, responses among adult patients with systemic C Rabusa, V Sundararajan, D Prior, lupus erythematosus previously immunised S M Proudman, M Baron, M Nikpour, The with 23-valent pneumococcal polysaccharide Scleroderma Clinical Trials Consortium Damage vaccine Index Working Group, The Australian Scleroderma R P Rezende, L E C Andrade, E M Klumb Interest Group, Canadian Scleroderma Research Group 855 Validity of the Swedish version of the systemic 817 Compendium of skin molecular signatures sclerosis quality of life questionnaire (SSCQoL): identifies key pathological features associated A novel measure of quality of life for patients with fibrosis in systemic sclerosis with systemic sclerosis S-J Moon, J M Bae, K-S Park, I Tagkopoulos, K-J Kim G Sandqvist, R Hesselstrand

857 Evaluation of retinal microvascular perfusion in Osteoarthritis systemic sclerosis: a case–control study 826 CircSERPINE2 protects against osteoarthritis by M Rothe, F Rommel, S Klapa, J Y Humrich, targeting miR-1271 and ETS-related gene R Nieberding, T Lange, J A M Sochurek, P Plöttner, S Grisanti, G Riemekasten, M Ranjbar S Shen, Y Wu, J Chen, Z Xie, K Huang, G Wang, Y Yang, W Ni, Z Chen, P Shi, Y Ma, S Fan Contents Volume 78 Issue 6 | ARD June 2019

858 Comparison of autoantibody specificities e49 Response to: ‘Variation in antinuclear tested by a line blot assay and antibody detection by automated indirect immunoprecipitation-based algorithm immunofluorescence analysis’ by van in patients with idiopathic inflammatory Hoovels et al myopathies D S Pisetsky, D M Spencer, P E Lipsky, B H Rovin F Espinosa-Ortega, M Holmqvist, H Alexanderson, H Storfors, T Mimori, e50 Pitfalls of antinuclear antibody detection in I E Lundberg, J Rönnelid systemic lupus erythematosus: the positive experience of a national multicentre study 860 Addressing immune-related adverse events F Pregnolato, M O Borghi, P L Meroni, Forum of cancer immunotherapy: how prepared are Interdisciplinare per la Ricerca sulle Malattie rheumatologists? Autoimmuni (FIRMA) Study Group M Kostine, L C Cappelli, C Calabrese, L H Calabrese, C O Bingham III, C Richez, e51 Response to: ‘Pitfalls of antinuclear antibody J-E Gottenberg, O Lambotte detection in systemic lupus erythematosus: the positive experience of a national multi-center 862 3D culture of Erdheim-Chester disease tissues study’ by Pregnalato et al unveils histiocyte metabolism as a new D S Pisetsky, D M Spencer, P E Lipsky, B H Rovin therapeutic target A Villa, D Belloni, B Vergani, S Cenci, G Cavalli, e52 Time to personalize the treatment of anti- R Biavasco, M Rodolfo, M G Cangi, C Doglioni, MDA-5 associated lung disease L Dagna, E Ferrero, M Ferrarini M Lake, G George, R Summer

e53 Response to: ‘Time to personalise the Miscellaneous treatment of anti-MDA-5 associated lung 865 Correction: Autoantibodies to citrullinated disease’ by Lake et al V Leclair, I E Lundberg, A Tjärnlund proteins induce joint pain independent of inflammation via a chemokine-dependent e54 Missing pebble in the mosaic of rheumatic mechanism diseases and mental health: younger does not 866 Correction: Identification of a novel always mean happier chemokine-dependent molecular A Alunno, P Studenic, D Wiek, P Balážová, D Aletaha mechanism underlying rheumatoid arthritis- associated autoantibody-mediated bone e55 Response to ‘Missing pebble in the mosaic of loss rheumatic diseases and mental health: younger does not always mean happier’ by Alunno et al I Redeker, F Hoffmann, J Callhoff, H Haibel, J Sieper, A Zink, D Poddubnyy Electronic pages e46 Unending story of the indirect e56 Forward to the past: ultrasound might be necessary immunofluorescence assay on HEp-2 cells: in some patients with rheumatoid arthritis old problems and new solutions? O Gadeholt P L Meroni, E K Chan, J Damoiseaux, L E C Andrade, X Bossuyt, K Conrad, X Mariette, e57 Cardiovascular events in ankylosing spondylitis: J Sheldon, J Rönnelid, M J Fritzler, On behalf of the a 2018 meta-analysis members of the committees S Mathieu, M Soubrier e47 Response to: 'Unending story of the indirect e58 Checkpoint inhibitors and arthritis immunofluorescence assay on HEp-2 cells: N Manolios, L Schrieber old problems and new solutions?' by Meroni et al e59 Correction: The development of assessment D S Pisetsky, D M Spencer, P E Lipsky, B H Rovin of SpondyloArthritis international society classification criteria for axial spondyloarthritis e48 Variation in antinuclear antibody detection (part II): validation and final selection by automated indirect immunofluorescence analysis e60 Correction: Visualisation of interstitial lung L Van Hoovels, S Schouwers, S Van den Bremt, disease by molecular imaging of integrin avb3 X Bossuyt and somatostatin receptor 2 Editorial

It is yet another privilege to receive and Greetings from the editor be able to count on these outstanding contributions. Josef S Smolen Let me please come to an end of these greetings by also thanking you, the readers, for your loyalty to ARD— please let us know where you see room 3 Since 20 years, June after June constitutes research in rheumatoid arthritis and a for improvement. Finally, I wish you a the ‘month of European rheumatology’, new look at the worldwide burden of successful EULAR 2019 Congress as well 4 the month in which the Annual EULAR rheumatic and musculoskeletal diseases. as much pleasure and lots of gain of new Congress is held. It was in June 2000 Further, many original papers on information when reading ARD. when this conference was organised for important studies in rheumatoid arthritis, Josef Smolen, Editor-in-Chief the first time in Nice and, therefore, this SLE, systemic sclerosis and osteoarthritis, year’s Congress marks its 20th Anniver- among others, are included in this issue. Competing interests None declared. sary. Also in 2000, the oldest rheuma- The ‘Heroes and Pillars’ section is devoted Patient consent for publication Not required. tology journal, Annals of the Rheumatic to Barbara Ansell, a legendary rheuma- Diseases, which was founded 80 years ago Provenance and peer review Commissioned; tologist who was a founder and broadly internally peer reviewed. in 1939, became the EULAR Journal, influenced the development of paediatric being part of the Congress educational 5 © Author(s) (or their employer(s)) 2019. No commercial rheumatology in the last century. Also, re-use. See rights and permissions. Published by BMJ. material ever since—every single one of when you look at the publications on clin- the more than 14 000 participants can ical trials, you may find our new editorial enjoy reading ARD for a full year between policy mostly fulfilled, namely a detailed two consecutive EULAR Congresses. presentation of all data points in respec- To cite Smolen JS. Ann Rheum Dis 2019;78:715. The EULAR Journal’s educational and tive figures for clearer insights into the Ann Rheum Dis 2019;78:715. scientific content was shaped by its first data sets and easier future meta-analysis doi:10.1136/annrheumdis-2019-215621 editor, Professor Leo van der Putte (editor work.6 7 until April 2008) and subsequently further Needless to say that, as always since References expanded by his successor, Professor Tore the foundation of the Annual EULAR 1. Fanouriakis A, Kostopoulou M, Alunno A, et al. 2019 Kvien (editor until August 2017). It is a Congress, you can read the abstracts update of the EULAR recommendations for the pleasure and obligation to pay tribute presented at the Congress in a supplement management of systemic lupus erythematosus. Ann to these two eminent personalities who Rheum Dis 2019;78:736–45. to ARD. 2. Dall’Era M, Bruce IN, Gordon C, et al. Current challenges have enlightened the rheumatological I would like to take this opportunity to in the development of new treatments for lupus. Ann community by their own scientific work thank the Associate Editors for all their Rheum Dis 2019;78:729–35. and as outstanding mentors within and support during the last year. Their rapid 3. Toes R, Pisetsky DS. Pathogenic effector functions beyond their institutions, and also by and thorough considerations when dealing of ACPA: where do we stand? Ann Rheum Dis their highly laudable and visionary activ- 2019;78:716–21. with first decision-making regarding 4. Sebbag E, Felten R, Sagez F, et al. The world-wide ities in the course of their ARD editor- submitted papers cannot be highlighted burden of musculoskeletal diseases: a systematic ships, importantly supporting the paths strongly enough. Further, several issues analysis of the World Health organization burden of into a flourishing and growing world of had arisen, such as on dual submission diseases database. Ann Rheum Dis 2019;78:844–8. 5. Woo P, Wulffraat NM, Tyndall A. Barbara Ansell: rheumatology. or corrections of manuscripts, and the It is also a privilege to focus your atten- a person ahead of her time. Ann Rheum Dis thoughtfulness and culture of discus- 2019;78:725–8. tion on a number of highlights published sion, the time provided during weeks of 6. van Mulligen E, Pieter de Jong PH, Kuijper TM et al. in the ARD issue you hold in your hands deliberations and the support received by Gradual tapering TNF inhibitors versus conventional right now: the update of the EULAR this group are exemplary. My gratitude synthetic DMARDs after achieving controlled disease in patients with rheumatoid arthritis: first-year results of recommendations for the management extends to the chair of ARD’s advisory of systemic lupus erythematosus (SLE),1 the randomised controlled TARA study. Ann Rheum Dis committee and the publisher for all their 2019;78:746–53. a viewpoint on current challenges in SLE 7. Visvanathan S, Daniluk S, Ptaszyński R. Effects of BI 2 help. And a big ‘Thank you!’ is also owed research , an editorial on autoantibody to all reviewers who support the journal 655064, an antagonistic anti-CD40 antibody, on clinical and biomarker variables in patients with active by their invaluable comments that allow Correspondence to Professor Josef S Smolen, rheumatoid arthritis: a randomised, double-blind, Rheumatology, Medical University of Vienna, Vienna to increase the quality of submitted papers placebo-controlled, phase IIa study. Ann Rheum Dis 1090, Austria; josef​ .​smolen.ard@​ ​meduniwien.​ac.​at and, thus, the quality of this Journal. 2019;78:754–60.

Smolen JS. Ann Rheum Dis June 2019 Vol 78 No 6 715 Editorial

Postulated mechanisms for ACPA Pathogenic effector functions of ACPA: pathogenicity For ACPA, two main mechanisms for Where do we stand? disease induction or exacerbation have been proposed.16–18 The first mechanism René Toes,‍ ‍ 1 David S Pisetsky2 involves the interaction of ACPA and a citrullinated protein to form immune complexes (IC). IC can activate the complement system and induce the release Introduction As shown in elegant biochemical studies, of chemotactic factors such as C3a and In the late 1990s, the discovery that citrul- ACPA target proteins that contain the C5a to allow recruitment of immune cells linated proteins are targeted by auto- amino acid citrulline, a post-translational to a local site. These cells can subsequently antibodies known as anti-citrullinated modification (PTM) of the amino acid be activated in a fragment crystallisable protein antibodies (ACPA) revolutionised arginine mediated by peptidyl arginine (Fc)γ-receptor-dependent manner. This 8 9 the study of rheumatoid arthritis (RA). deiminase enzymes ; these antibodies are activation can lead to the production of Because of the high specificity of ACPA also called anti-cyclic citrullinated peptide cytokines and other pro-inflammatory for RA, investigators began experiments (CCP) antibodies because of their binding mediators, culminating in inflammation to determine whether ACPA have a direct to CCP, a synthetic peptide used as a test in general, although arthritis could occur pathogenetic role in arthritis or whether antigen in immunoassays.8 9 ACPA provide preferentially if the citrullinated proteins these antibodies more simply reflect an important criterion in diagnosis and are located in the joint or if ACPA produc- underlying T-cell and B-cell responses classification.10 tion occurs primarily in that local envi- without themselves impacting on joint While the strong association of ACPA ronment. The ability of an ACPA to form inflammation and damage. The debate with RA suggests a role in joint inflam- an IC may depend on the specificity of on the role of ACPA has been extensive mation, proof of pathogenicity can be the antibody as well as the nature of the and has now entered a new phase with the challenging and requires both direct citrullinated protein that can be bound publication of correction notes appearing and indirect evidence. Analogous to 1 2 in complexed form. IC, including ACPA in this issue of Annals of the Rheumatic Koch’s postulates for infectious disease, containing IC, can also bind rheumatoid (ARD). These notes concern papers that Witebsky’s postulates for autoimmune 3 4 factor (RF), an autoantibody against IgG appeared in the journal in 2016 and disease systemise the experimental which very frequently co-occurs with follow a retraction note and correspon- 5–7 evidence needed to demonstrate that a ACPA; like that of ACPA, the expres- dence in other journals. clinical condition is autoimmune and that sion of RF can precede the clinical In these notes, the study authors indi- an autoimmune response (autoantibody manifestations of the disease and is also cate that the monoclonal antibodies used 11 or B/T cell) causes pathology. For ACPA, important for diagnostic and classification in their experiments on ACPA pathoge- 10 19–21 the challenge is especially great since purposes. nicity did not, in fact, react with citrulli- ACPA are highly cross-reactive and bind A more novel mechanism proposed for nated proteins and/or peptides. As such, a wide variety of citrullinated proteins; ACPA pathogenicity relates to the ability the putative functional activities of these these proteins include citrullinated of autoantibodies to serve as agonists for monoclonals must have resulted from versions of enolase, fibrinogen, vimentin, a receptor-mediated response. Indeed, mechanisms other than antibody binding collagen and histones, among others.12 13 these studies, which are the subject of to a citrullinated target. The implications The relevant target antigen in vivo and its the correction notes, presented intriguing of these corrections and retractions for tissue localisation are unknown, although observations indicating that ACPA can models of pathogenesis are wide ranging joint tissue in RA shows evidence of directly induce both pain and osteoclas- and call into question the nature of ACPA citrullination.14 togenesis; these are two cardinal features involvement in manifestations of RA. For an autoantibody-mediated disease, of RA that can impact on patient symp- Because of the importance of these issues pathogenicity is based on the presence of 3 4 for the scientific community, in this edito- tomatology as well as clinical outcomes. an autoantibody in affected patients; the rial, we will review the role of ACPA in These studies involved transfer into mice RA and the salient issues in establishing transfer of disease to an animal by a source of IgG purified from patient sera as well as the pathogenicity of these autoantibodies. of antibodies, including monoclonal anti- monoclonal ACPA to assess pain responses bodies derived from blood or tissue; and and bone loss; studies on osteoclastogen- induction of disease in an animal model esis involved in vitro culture systems as ACPA as biomarkers in RA by immunisation with the putative auto- well. While both polyclonal and mono- ACPA are unique biomarkers for RA, a antigen. In some situations, an in vitro clonal antibody preparations produced serious form of arthritis characterised by system can provide a model for a cellular similar findings in the ARD papers, the inflammation, pain and, unless treated response key to pathogenesis if the rela- authors have meanwhile revealed that effectively, damage of cartilage and bone. tionship to clinical findings is clear. The the monoclonal antibodies used in these use of animal models is frequently infor- experiments do not bind citrullinated 1Department of Rheumatology, Leiden University mative, although it is possible that an auto- antigens.5 Medical Center, Leiden, The Netherlands 2 antibody requires a particular setting (eg, In view of these issues, it is useful to Department of Medicine, Duke University Medical low level of tissue injury or inflammation, review studies on the pathogenicity of Center and Medical Research Service, VA Medical Center, Durham, North Carolina, USA the presence of other autoantibodies) for ACPA and determine how studies with the pathogenicity to be manifest; in this monoclonals without apparent ACPA Correspondence to Dr René Toes, Rheumatology, Leiden University Medical Center, Leiden 2333 ZA, case, the use of wild-type, unmanipulated activity could induce pathogenic events in 15 Netherlands; r.​ ​e.m.​ ​toes@​lumc.​nl mice may miss a pathogenic effect. either in vivo or in vitro studies.

716 Toes R, Pisetsky DS. Ann Rheum Dis June 2019 Vol 78 No 6 Editorial

The effects of ACPA on monomeric Fab fragments were reported osteoclastogenic and pain-inducing poten- inflammation, pain and bone loss to promote osteoclastogenesis,3 23 these tial notified the community that, in reas- As noted, one potential mechanism by effects appeared independent of the Fc sessing the original data, the antibodies in which ACPA can promote events in RA region of the antibodies and, hence, from question showed ‘no measurable affinity relates to their ability to form IC that can Fcγ-receptor triggering. This last notion for the tested citrullinated peptides’ as 7 activate cells. Indeed, several studies have is supported by the observation that only detailed in a recent correction and retrac- 5 demonstrated that the Fc tail of ACPA can monoclonal ACPA displaying a certain tion note ; in other words, the monoclonal activate both Fcγ-receptor expressing cells epitope-specificity profile could activate antibodies studied were not ACPA. This and the complement system.16 18 Since the osteoclasts, arguing that epitope specificity observation is in line with crystallographic expression of citrullinated antigens occurs of the ACPA is crucially involved in the studies of some of these monoclonal anti- 3 in the tissue in vivo, including the synovial induction of osteoclastogenesis. As now bodies that could not confirm the binding 25 compartment,14 these studies suggest that known, however, these monoclonal anti- of citrullinated peptides. While there is 5 ACPA-containing IC can induce inflam- bodies do not have citrulline specificity. a major unknown— the identity of the mation in RA. Importantly, other studies In view of the strong associations antigens that led to a positive signal on have provided evidence that RF can between ACPA and RA, the findings synovial tissue and/or osteoclasts—these enhance these effects in vitro, presumably published in ARD in 2016 attracted great antibodies had, nevertheless, in vivo and by promoting IC formation via multiva- interest as they could explain two prom- in vitro effects. As also acknowledged by lent binding.19–21 These observations link inent features of RA-joint pain and bone the authors in the two correction notes the two characteristic serological findings erosion in a disease-specific manner. As published in this issue of ARD, it is now in RA into one pathogenic pathway and such, these observations entailed a para- clear that ‘the functional results reported are consistent with clinical findings that digm shift in thinking about disease patho- for these monoclonal antibodies cannot arthritis is more severe in patients who genesis by suggesting that certain disease be attributed to reactivity against citrul- 22 manifestations in RA could arise directly linated proteins and/or peptides, but are express both ACPA and RF. 1 2 While ACPA could drive inflamma- from RA-specific autoantibodies and do due to other yet unknown mechanisms’. tion by complement system-mediated not entail IC formation. In this concep- In studies on pathogenicity, determi- tualisation of RA, ACPA would resemble nation of antibody specificity is crucial and Fcγ-receptor-mediated effects, other autoantibodies that directly activate or and requires appropriate assays to detect studies suggest a novel role of ACPA by functionally perturb cells such as autoan- binding at relevant concentrations. For direct binding to molecules expressed tibodies found in Graves’ disease which ACPA, several different antigens can be on the surface of cells; these molecules bind the thyrotropin receptor stimu- used for immunochemical determina- are presumably citrullinated. Indeed, the lating the production of thyroxine and tions. These include citrullinated proteins studies in ARD reported that ACPA can triiodothyronine. and citrullinated peptides especially directly induce osteoclastogenesis, thereby In ascribing pathogenicity to an auto- when presented in large arrays in which contributing to bone loss and joint erosion antibody in any disease, a relevant consid- a multitude of sequences can be assessed. that are typical for RA. Furthermore, eration is plausibility (ie, does clinical Other considerations in assessing antibody studies by Wigerblad et al in ARD indi- evidence support the proposed mecha- specificity relate to the detection method cated that ACPA can induce pain in an in nism?). In RA, a number of clinical situa- (eg, ELISA, surface plasmon resonance, vivo mouse model by activating osteoclasts tions impact on the plausibility that ACPA western blotting) that may, in turn, be to release interleukin 8 and potentially 3 4 are pathogenic either by IC formation or influenced by technical factors such as salt other mediators involved in pain. The agonistic activity. The first concerns the concentrations and even the nature of the effects observed in these studies appeared apparent lack of transfer of disease during buffer. to be specific for ACPA since human IgG pregnancy despite the placental transfer For monoclonal antibodies, polyreac- preparations enriched for ACPA, induced of autoantibodies. The expression of tivity and cross-reactivity (ie, ability to pain behaviour in mice; in contrast, IgG ACPA in asymptomatic individuals during bind to multiple different antigens or the preparations derived from ACPA-negative the phase of pre-disease autoimmunity ability to bind to multiple citrullinated patients or IgG preparations depleted of is another clinical scenario to consider. and/or carbamylated/acetylated antigens) ACPA were inactive. Another circumstance relates to the may be a complicating factor. Polyreac- Consistent with the principles continued expression of ACPA following tivity may result from an aberrant ‘pattern embodied in Witebsky’s postulates (which successful treatment of RA that can of antigen-driven selection’ in the setting long preceded monoclonal antibody tech- diminish pain and halt bone loss. While of autoimmunity. In autoimmune disease, nology), studies with purified IgG from some reduction in ACPA levels occurs the repertoires and signalling thresholds patient blood can provide at least circum- with therapy, this reduction is limited.24 for both T and B cells may be abnormal, stantial evidence for the pathogenicity In these various settings, why does not the skewing the binding properties of induced of a particular antibody specificity. In presence of ACPA lead to either pain or antibodies. In addition to disturbances in the ARD publications, the use of mono- bone loss? B-cell and T-cell function, antigen drive clonal antibodies derived from patient by multiple different antigens may lead sources provided perhaps more compel- to patterns of reactivity unlike those of ling evidence that the effects on pain and Re-evaluation of data on ACPA a conventional antibody response where osteoclastogenesis result from direct ACPA pathogenicity antigen binding is more specific and has binding. In these experiments, murinised While the concept that ACPA are agonistic higher avidity.26–28 In this regard, since monoclonal ACPA, but not monoclonal agents is intriguing, the new findings the nature of self-antigens in vivo is often control antibodies, elicited pain-like ask for a critical re-interpretation of this not known, extrapolating from in vitro behaviour that coincided with the ability possibility. Notably, the authors who assays with purified antigens to that which to activate osteoclasts. Furthermore, since described the monoclonal ACPA with occurs in vivo can be uncertain especially

Toes R, Pisetsky DS. Ann Rheum Dis June 2019 Vol 78 No 6 717 Editorial

while, in the control condition, they are retained. Still, the binding of antibodies to affinity columns provides presump- tive evidence of specificity, but does not prove it. Regardless, since the identifica- tion and molecular characterisation of any putative citrullinated receptor have not yet been accomplished, caution is needed in the interpretation of results with the polyclonal preparations as well as monoclonal antibodies. Of note, the avidity of ACPA appears to be relatively low, potentially limiting their ability to directly activate cellular receptors.29 30 While the relevant specificity of both the polyclonal and monoclonal antibody preparations remains an open question, the finding on the induction of pain and bone loss is nevertheless very interesting and would benefit from reasonable and plausible explanations to move the field forward. In retrospect, several possibil- ities could be at play. For example, it is possible that the monoclonal and poly- clonal antibody preparations contained contaminants that by themselves might have caused biological effects. Myeloid cells, for example, are highly sensitive to endotoxin and non-endotoxin pyro- gens; this sensitivity can differ from that of conventional endotoxin-detection kits, such as the Limulus Amebocyte Figure 1 Operational and postulated principles in defining the pathogenic activity of polyclonal Lysate test. Endotoxin and other pyro- or monoclonal antibodies. (A) Postulated mechanism for ACPA pathogenicity either by polyclonal gens can be co-purified during protein ACPA and/or monoclonal ACPA. Left side: confirmed autoantibody-mediated contributions to isolation due to hydrophobic interac- 16–18 arthritis facilitated by complement system activation and Fcγ receptor activation; right tions, for example.31 32 Such events can side: postulated contribution of ACPA to RA as now under debate indicated by correction notes occur variably among antibody prepara- 1 2 3 4 published in this issue to papers published in ARD as well as a preceding retraction and tions, explaining why some monoclonal 5–7 correspondence in other journals. The questions marks indicate the re-evaluation of the antibodies or their respective purified pathogenic activity of the monoclonal ACPA, as it is reassessed to be not specific for citrullination. Fab preparations activate osteoclasts, (B) Potential issues occurring during antibody isolation of either ACPA from patient sera or whereas others do not. supernatant containing recombinant monoclonal antibodies. Factors that can occur during Another contributing factor relates to purification: aggregation, immune complex formation, co-purification of RF in case of ACPA+ the immunochemical properties of an serum or synovial fluid and lastly introduction of endotoxin and non-endotoxin pyrogens. Light antibody molecule. Among the mono- blue antibody: monoclonal ACPA; dark blue: ACPA from patient IgG fraction enriched for ACPA; clonal antibodies that could induce asterisk: citrullinated protein; green antibody: RF IgG. ACPA, anti-citrullinated protein antibody; RA, osteoclastogenesis and joint pain, D10 rheumatoid arthritis; RF, rheumatoid factor. and B02 were reported to show ‘poly- reactivity’ towards at least two antigens out of three analysed (ie, to double- when an antibody can bind to structur- have contributed to the results obtained stranded DNA, insulin and/or lipopoly- ally similar molecules in a cross-reactive (figure 1). Whether this situation also saccharide (LPS). This activity contrasts manner. pertains to the effects of the affinity-pu- with that of at least one of the mono- rified polyclonal ACPA containing Ig clonal antibodies which did not appear preparations is more difficult to know. Issues in experimental design and to elicit these effects (RA1276:01:C07). performance Importantly, however, the authors of Likewise, because of aggregation that Given these considerations as well as the 2016 ARD papers on pain induc- can occur during affinity purification notices of correction and retraction, tion and osteoclastogenesis state that or during the isolation of monoclonal current data indicate that the biological ‘also these results have to be interpreted antibodies, certain monoclonal antibody effects observed with the monoclonal with caution waiting for additional preparations can act as ICs to activate 1 2 antibodies appear unrelated to the recog- mechanistic studies’. With polyclonal FcγR-positive cells; the production of nition of citrulline as a critical determi- antibodies, it is important to absorb reac- heavy chain dimers can lead to the same nant of specificity.5 6 25 Hence, other tivity with citrullinated versus control effect. In this situation, batch-to-batch mechanisms, unrelated to the recog- antigens to show that, in the first case, and IgG-to-IgG variations can occur.33–36 nition of citrullinated antigens, must the biological effects are eliminated Such variability could also explain why

718 Toes R, Pisetsky DS. Ann Rheum Dis June 2019 Vol 78 No 6 Editorial some (monoclonal) antibody prepara- tions induce effects in cellular systems, Box 1 Considerations for defining the pathogenic activity of either polyclonal whereas others do not. Similarly, the or monoclonal antibodies presence of IgG-RF in IgG preparations purified from serum could also lead to ►► Assess antibody activity using well-validated assays especially those that are clinically confounding results. useful. It is important to consider that ACPA ►► Validate assays sufficiently to ensure that they detect specific antibody binding and might be particularly prone to aggre- not non-specific interactions. gation and interactions with pyrogens ►► Assure antibody specificity by direct binding and inhibition studies of polyclonal and since they display some unique features monoclonal antibodies. including the presence of negatively ►► For affinity purification, consider use of different antigens (eg, different citrullinated charged, N-linked glycans in the variable proteins) for the affinity matrix and assess both bound and unbound fractions for domain.37 38 Also in this case, the compo- antibody binding and functional activity. sition of the N-linked glycans present on ►► Remove aggregates by ultracentrifugation or size-exclusion chromatography. recombinantly produced antibodies may ►► Assess for contamination by endotoxin and related material. vary substantially from batch-to-batch ►► For in vitro studies, perform dose–response curves and assess activity at relevant and/or between different antibodies antibody concentrations. when not controlled for and hence could ►► For in vivo studies, assess activity in different strains of mice including knockout mice impact on biological outcomes.39–41 to define mechanisms. The experience of one of us (RT) with ►► For in vivo studies, consider the use of models in which subclinical arthritis is induced. isolated ACPA, monoclonal ACPA, as ►► For both in vivo and in vitro studies, consider the effects of rheumatoid factor. well as control antibody preparations ►► In case of agonistic antibodies binding to cellular receptors, clone the receptor from is consistent with this possibility since the target cells for expression in reporter cells to mimic agonistic activity. we also obtained inconsistent and vari- able effects in similar studies addressing the osteoclastogenic effects of ACPA of the effects on investigators especially avidities, attributing biological effects (unpublished data). In this context, vari- when unforeseen technical issues have of monoclonal or polyclonal antibodies able effects seem to be present in the affected experiments. Beyond effects to the recognition of citrulline alone is published experiments as well since, in on investigators, however, retractions difficult and it can be debated whether a recent study, one of four monoclonal and corrections, especially if delayed, citrulline specificity and/or the ‘Peptidyl ACPA induced osteoclastogenesis slightly, also negatively impact the scientific Arginine Deiminase-Citrulline’ pathway two had no effect and the fourth inhib- community as a whole since investigators should be considered the predominant ited osteoclastogenesis; as such, when who were attracted by reported find- or indeed only auto-reactive phenom- antibodies are present together as may ings may have invested time and effort enon. In this respect, it is important to happen during disease, there may be no to explore these observations, leading note that recent epidemiological data net effect, with one effect on osteoclast them in a perhaps wrong direction or may not support the postulated connec- activation counteracted by another.42 deterring them from pursuing other tion between smoking and formation Similarly, it has been reported that potential mechanisms. Correcting the of citrullinated proteins and a breach the osteoclastogenic potential of ACPA literature will also take time especially as of tolerance to citrullinated antigens48; occurs only with cells from wild-type novel ideas often receive prominence in thus, smoking may be associated with the mice, but not with cells derived from reviews and other articles. concurrent presence of multiple autoan- Fcγ- or FcγRIII knockout mice,43 a notion While disappointing and regrettable, tibodies in RA rather than with ACPA per that seems to conflict with the observa- these circumstances nevertheless provide se.49–51 Ultimately, it will be essential to tion that Fab domains may be sufficient an opportunity to rethink concepts as delineate fully the autoantibody reper- to mediate the effects described. In this well as improve experimental designs. toire in RA and to clone, express and to respect, it might be relevant to also Given publication of several studies on define at the molecular level the possible re-evaluate the reported osteoclasto- agonist activities of ACPA, these consid- cellular receptors targeted by ACPA, genic potential of Fab fragment isolated erations do not exclude the possibility assessing the relevant PTM. In view of against citrullinated vimentin.23 The that ACPA (or even a related speci- the experience with ACPA, we would observations obtained with cells from ficity such as anti-malondialdehyde suggest the following operational princi- FcγR knockout mice are, however, in [monoclonal] antibodies as recently ples in defining the pathogenic activity line with the findings that (heat)-aggre- proposed)45 can directly activate osteo- of either polyclonal or monoclonal anti- gated IgGs display a strong osteoclas- clasts to induce bone resorption and bodies (Box 1). togenic potential; this potential may be pain; providing new evidence in support more pronounced for aggregated IgG of this phenomenon is crucial, especially preparations enriched for ACPA, perhaps to explain the activity of polyclonal anti- Conclusions related to the reduced sialylation of the body preparations. The recent information on the binding N-linked Fc glycans which enhances Adding to the complexity of the properties of monoclonal antibodies once FcγR binding.44 ACPA system is the extensive cross-re- considered to be ACPA will require a activity of these autoantibodies period of reflection and re-evaluation of towards different citrullinated antigens prior results on the potential pathogenicity Implications for the future as well as other PTMs expressed by of ACPA in RA. Nevertheless, the data The publication of retraction and correc- proteins.42 46 47 As these different modi- presented in a number of papers suggest tion notes is always unfortunate because fications are recognised with varying that autoantibodies can mediate events

Toes R, Pisetsky DS. Ann Rheum Dis June 2019 Vol 78 No 6 719 Editorial in RA by diverse mechanisms, including (pro)filaggrin by deimination of arginine residues. J dependent nuclear antigen. Arthritis Rheum IC formation and, possibly, direct agonist Immunol 1999;162:585–94. 1980;23:528–38. 9 Schellekens GA, de Jong BA, van den Hoogen 27 Pisetsky DS, Hoch SO, Klatt CL, et al. Specificity and activity. Clearly, as the field moves forward, FH, et al. Citrulline is an essential constituent of idiotypic analysis of a monoclonal anti-Sm antibody the claims that ACPA can activate cells antigenic determinants recognized by rheumatoid with anti-DNA activity. J Immunol 1985;135:4080–5. implicated in pathogenesis should receive arthritis-specific autoantibodies. J Clin Invest 28 Rubin RLet al. Multiple autoantigen binding more extensive and rigorous experimental 1998;101:273–81. capabilities of mouse monoclonal antibodies support before they can be accepted. None- 10 Aletaha D, Neogi T, Silman AJ, et al. 2010 rheumatoid selected for rheumatoid factor activity. J Exp Med arthritis classification criteria: an American College 1984;159:1429–40. theless, as our understanding of the unique of Rheumatology/European League against 29 Suwannalai P, Scherer HU, van der Woude D, et al. features of ACPA and other autoantibodies rheumatism collaborative initiative. Ann Rheum Dis Anti-citrullinated protein antibodies have a low avidity grows, these mechanisms will undoubtedly 2010;69:1580–8. compared with antibodies against recall antigens. Ann undergo further investigation. In the quest 11 Rose NR, Bona C. Defining criteria for autoimmune Rheum Dis 2011;70:373–9. diseases (Witebsky’s postulates revisited). Immunol 30 Usinger WR, Lucas AH. Avidity as a determinant to understand the possible contribution of Today 1993;14:426–30. of the protective efficacy of human antibodies to ACPA to disease pathogenesis, these efforts 12 Chandra PE, Sokolove J, Hipp BG, et al. Novel pneumococcal capsular polysaccharides. Infect Immun will provide new and stimulating insights multiplex technology for diagnostic characterization of 1999;67:2366–70. into the pathways contributing to RA patho- rheumatoid arthritis. Arthritis Res Ther 2011;13. 31 Dullah EC, Ongkudon CM. Current trends in 13 Ioan-Facsinay A, el-Bannoudi H, Scherer HU, et al. Anti- endotoxin detection and analysis of endotoxin-protein genesis and hopefully allow new approaches cyclic citrullinated peptide antibodies are a collection interactions. Crit Rev Biotechnol 2017;37:251–61. to diagnosis, staging and treatment. of anti-citrullinated protein antibodies and contain 32 Warger T, Hilf N, Rechtsteiner G, et al. Interaction of overlapping and non-overlapping reactivities. Ann TLR2 and TLR4 ligands with the N-terminal domain Handling editor Josef S Smolen Rheum Dis 2011;70:188–93. of Gp96 amplifies innate and adaptive immune 14 Vossenaar ER, Smeets TJM, Kraan MC, et al. The responses. J Biol Chem 2006;281:22545–53. Acknowledgements We thank Dr Hans Ulrich presence of citrullinated proteins is not specific 33 Bondos SE, Bicknell A. Detection and prevention Scherer for critically reading the manuscript and Ms Lise for rheumatoid synovial tissue. Arthritis Rheum of protein aggregation before, during, and after Hafkenscheid for drawing of figure 1. 2004;50:3485–94. purification. Anal Biochem 2003;316:223–31. Funding The authors have not declared a specific 15 Kuhn KA, Kulik L, Tomooka B, et al. Antibodies 34 van der Kant R, Karow-Zwick AR, Van Durme J, grant for this research from any funding agency in the against citrullinated proteins enhance tissue injury et al. Prediction and reduction of the aggregation of public, commercial or not-for-profit sectors. in experimental autoimmune arthritis. J Clin Invest monoclonal antibodies. J Mol Biol 2017;429:1244–61. 2006;116:961–73. 35 Li W, Prabakaran P, Chen W, et al. Antibody Competing interests None declared. 16 Clavel C, Nogueira L, Laurent L, et al. Induction aggregation: insights from sequence and structure. Patient consent for publication Not required. of macrophage secretion of tumor necrosis factor Antibodies 2016;5. alpha through Fcgamma receptor IIa engagement 36 Chung HH, Buck L, Daris K, et al. Investigation of Provenance and peer review Commissioned; by rheumatoid arthritis-specific autoantibodies to the free heavy chain homodimers of a monoclonal internally peer reviewed. citrullinated proteins complexed with fibrinogen. antibody. Biotechnol Prog 2018;34:738–45. © Author(s) (or their employer(s)) 2019. No Arthritis Rheum 2008;58:678–88. 37 Hafkenscheid L, Bondt A, Scherer HU, et al. Structural commercial re-use. See rights and permissions. 17 Ji H, Ohmura K, Mahmood U, et al. Arthritis critically analysis of variable domain glycosylation of Anti- Published by BMJ. dependent on innate immune system players. Citrullinated protein antibodies in rheumatoid arthritis Immunity 2002;16:157–68. reveals the presence of highly sialylated glycans. Mol 18 Trouw LA, Haisma EM, Levarht EWN, et al. Anti-cyclic Cell Proteomics 2017;16:278–87. citrullinated peptide antibodies from rheumatoid 38 Rombouts Y, Willemze A, van Beers JJBC, et al. To cite Toes R, Pisetsky DS. Ann Rheum Dis arthritis patients activate complement via both the Extensive glycosylation of ACPA-IgG variable domains 2019;78:716–721. classical and alternative pathways. Arthritis Rheum modulates binding to citrullinated antigens in 2009;60:1923–31. rheumatoid arthritis. Ann Rheum Dis 2016;75:578–85. Ann Rheum Dis 2019;78:716–721. 19 Anquetil F, Clavel C, Offer G, et al. IgM and IgA 39 Pacis E, Yu M, Autsen J, et al. Effects of cell culture doi:10.1136/annrheumdis-2019-215337 rheumatoid factors purified from rheumatoid arthritis conditions on antibody N-linked glycosylation--what sera boost the Fc receptor- and complement- affects high mannose 5 glycoform. Biotechnol Bioeng dependent effector functions of the disease-specific 2011;108:2348–58. References anti-citrullinated protein autoantibodies. J Immunol 40 Planinc A, Dejaegher B, Vander Heyden Y, et al. 1 Correction to the paper by Wigerblad. Ann Rheum Dis 2015;194:3664–74. Batch-to-batch N-glycosylation study of infliximab, 2016;75:730–8. 20 Laurent L, Anquetil F, Clavel C, et al. IgM rheumatoid trastuzumab and bevacizumab, and stability study of 2 Correction to the paper by Krishnamurty. Ann Rheum factor amplifies the inflammatory response of bevacizumab. Eur J Hosp Pharm 2017;24:286–92. Dis 2016;75:721–9. macrophages induced by the rheumatoid arthritis- 41 van Berkel PHC, Gerritsen J, Perdok G, et al. N-linked 3 Krishnamurthy A, Joshua V, Haj Hensvold A, et al. specific immune complexes containing anticitrullinated glycosylation is an important parameter for optimal Identification of a novel chemokine-dependent protein antibodies. Ann Rheum Dis 2015;74:1425–31. selection of cell lines producing biopharmaceutical molecular mechanism underlying rheumatoid arthritis- 21 Sokolove J, Johnson DS, Lahey LJ, et al. Rheumatoid human IgG. Biotechnol Prog 2009;25:244–51. associated autoantibody-mediated bone loss. Ann factor as a potentiator of anti-citrullinated protein 42 Steen J, Forsström B, Sahlström P, et al. Recognition Rheum Dis 2016;75:721–9. antibody-mediated inflammation in rheumatoid of amino acid motifs, rather than specific proteins, by 4 Wigerblad G, Bas DB, Fernades-Cerqueira C, et al. arthritis. Arthritis Rheumatol 2014;66:813–21. human plasma cell-derived monoclonal antibodies to Autoantibodies to citrullinated proteins may induce 22 Aletaha D, Alasti F, Smolen JS. Rheumatoid factor, not posttranslationally modified proteins in rheumatoid joint pain independent of inflammation. Ann Rheum antibodies against citrullinated proteins, is associated arthritis. Arthritis Rheumatol 2019;71:196–209. Dis 2016;75:730–8. with baseline disease activity in rheumatoid arthritis 43 Krishnamurthy A, Steen J, Gronwall C, et al. RA- 5 Amara K, Steen J, Murray F, et al. Retraction: clinical trials. Arthritis Res Ther 2015;17. Associated antibodies targeting post translational monoclonal IgG antibodies generated from joint- 23 Harre U, Georgess D, Bang H, et al. Induction modification have different Osteoclastogenetic derived B cells of RA patients have a strong bias of osteoclastogenesis and bone loss by human potential. Arthritis & Rheumatology 2017;69. toward citrullinated autoantigen recognition. J Exp autoantibodies against citrullinated vimentin. J Clin 44 Harre U, Lang SC, Pfeifle R, et al. Glycosylation Med 2019;216. Invest 2012;122:1791–802. of immunoglobulin G determines osteoclast 6 Holmdahl R. The specificity of monoclonal Anti- 24 Böhler C, Radner H, Smolen JS, et al. Serological differentiation and bone loss. Nat Commun 2015;6. Citrullinated protein antibodies: comment on changes in the course of traditional and biological 45 Grönwall C, Amara K, Hardt U, et al. Autoreactivity to the article by Steen et al. Arthritis Rheumatol disease modifying therapy of rheumatoid arthritis. Ann malondialdehyde-modifications in rheumatoid arthritis 2019;71:324–5. Rheum Dis 2013;72:241–4. is linked to disease activity and synovial pathogenesis. 7 Malmström V, Steen J, Klareskog L. Reply. Arthritis 25 Ge C, Xu B, Liang B, et al. Structural basis of cross- J Autoimmun 2017;84:29–45. Rheumatol 2019;71:325–7. reactivity of Anti-Citrullinated protein antibodies. 46 Lloyd KA, Wigerblad G, Sahlström P, et al. Differential 8 Girbal-Neuhauser E, Durieux JJ, Arnaud M, et al. Arthritis Rheumatol 2019;71:210–21. AcpA binding to nuclear antigens reveals a The epitopes targeted by the rheumatoid arthritis- 26 Aitcheson CT, Peebles C, Joslin F, et al. Characteristics PAD-Independent pathway and a distinct subset associated antifilaggrin autoantibodies are of antinuclear antibodies in rheumatoid arthritis. of acetylation cross-reactive autoantibodies in posttranslationally generated on various sites of Reactivity of rheumatoid factor with a histone- rheumatoid arthritis. Front Immunol 2018;9.

720 Toes R, Pisetsky DS. Ann Rheum Dis June 2019 Vol 78 No 6 Editorial

47 Shi J, Willemze A, Janssen GMC, et al. Recognition 49 Hedström AK, Rönnelid J, Klareskog L, et al. Smoking, primarily determined by rheumatoid factor titre and of citrullinated and carbamylated proteins by human HLA-genes and serology in rheumatoid arthritis; the shared epitope rather than smoking per se. PLoS antibodies: specificity, cross-reactivity and the ’AMC- complex relationships investigated in the Swedish One 2017;12:e0180655. Senshu’ method. Ann Rheum Dis 2013;72:148–50. EIRA case-control study. Arthritis Rheumatol 2019. 51 van Wesemael TJ, Ajeganova S, Humphreys J, et al. 48 Malmström V, Catrina AI, Klareskog L. The doi:10.1002/art.40852. [Epub ahead of print: 11 Feb Smoking is associated with the concurrent presence of immunopathogenesis of seropositive rheumatoid 2019]. multiple autoantibodies in rheumatoid arthritis rather arthritis: from triggering to targeting. Nat Rev Immunol 50 Murphy D, Mattey D, Hutchinson D. Anti-citrullinated than with anti-citrullinated protein antibodies per se: a 2017;17:60–75. protein antibody positive rheumatoid arthritis is multicenter cohort study. Arthritis Res Ther 2016;18.

Toes R, Pisetsky DS. Ann Rheum Dis June 2019 Vol 78 No 6 721 Heroes and pillars of rheumatology Barbara Ansell: a person ahead of her time Pat Woo,1 Nico M Wulffraat,‍ ‍ 2 Alan Tyndall3

Handling editor Josef S In the beginning enabled such an ambitious enterprise to flourish. In Smolen In 1951, Barbara Ansell, a young 28-year-old physi- this current age of effective targeted therapies, the 1 images of the children are especially moving; Cush- Immunology and Molecular cian and aspiring cardiologist, was encouraged by Pathology, University College Professor Eric Bywaters to join him in the Special ingnoid little faces, stunted growth, deformed joints London, London, UK Unit for Juvenile Rheumatism, recently created in splints and children using walking aids or tricycles 2 Pediatrics, University Medical by the Medical Research Council (MRC) of UK, to move around, all attesting to the limited success of Center Utrecht, Utrecht, The located at the Canadian Red Cross Memorial the available therapeutic options at the time. What Netherlands 3Department of Rheumatology, Hospital in Taplow, Buckinghamshire, UK. She had is completely absent from the text are references to Basel University, Basel, been advised that cardiology ‘was a man’s world’. her own seminal contributions to the whole enter- Swaziland Although initially aimed at research and treatment prise, a reflection of Barbara’s innate modesty. All of rheumatic fever, this unit of pathologists and members of the allied health professional team were Correspondence to clinicians morphed into a unit for juvenile chronic sent to congresses, junior doctors were encouraged Dr Nico M Wulffraat, Pediatrics, arthritis as the incidence of rheumatic fever grad- to perform clinical studies with experienced team University Medical Center Utrecht, Utrecht 3584, The ually waned. At that time, a child with chronic members or laboratory projects with established Netherlands; arthritis would mostly be confined to bed at home, scientists and had their names first on most publica- ​n.​wulffraat@​umcutrecht.​nl perhaps with aspirin and/or cortisone medication tions. Moreover, she was particularly welcoming to and bore a grim prognosis regarding both func- fellows from far and wide. Received 22 February 2019 Revised 30 March 2019 tion and survival. She began her pioneering work The public record of Barbara’s achievements is Accepted 1 April 2019 there with her MD thesis (1965) on the classifica- widely available on Google and rather than simply tion of children with chronic arthritis. Her clinical repeat this, we would like here to share those qual- skill and experience were legendary. Her major ities she exhibited which in our opinion made her clinical achievements are listed in box 1. She also so special.1–3 was actively engaged with scientists in the search for the causes of the diseases she encountered. She was ideally placed in the shared environment of Innate powers of observation and MRC facilities next to a general hospital. Box 2 is a synthesis of clinical patterns short summary of the main milestones and breadth Barbara’s memory was phenomenal. She could of her work. In 1988, the world-renowned paedi- ask a young patient about the new bicycle he had atric rheumatologist, Barbara Ansell CBE, FRCP, received for Christmas 6 months later at the next FRCS, FRCPCH, retired as director of the MRC visit. Kids (and parents) truly appreciated this ability. Division of Rheumatology at Northwick Park, Team members who had forgotten to carry out her Harrow, UK. She left as a legacy an army of former requests often did not get away with it, thus earning colleagues, protegées and friends from all areas of her a fearsome reputation. However, this trait was an healthcare, dedicated to children and adolescents important foundation of her ability for pattern recog- with rheumatic diseases. By then the outcome for nition, and her world renowned clinical skills. such young patients had improved dramatically. She was not afraid of calling something ‘atypical’. Gratifyingly, Barbara did see the fulfilment of her The little notebook kept in her white coat pocket long-held dream, expressed passionately at her during ward rounds and clinics was a familiar sight retirement Festschrift: the beginning of truly effec- to her visitors and trainees. tive therapies. Her MD thesis in 1965 was a classification of In her 1997 book ‘Looking Back. The Canadian juvenile chronic arthritis based on her observations Memorial Red Cross Hospital 1947–1985’, she of presentation and course. This remained essentially chronicled the development of the MRC Rheu- the basis of all later refinements of the classification. matism Unit at the Canadian Red Cross Memorial Her keen observations also led her to describe novel Hospital in Taplow. In the text she described the rare syndromes without the need for sophisticated introduction of multidisciplinary teams addressing laboratory data. One of many examples is the publica- all aspects of the child’s care including innovative tion with E. Bywaters and F.M. Elderkin, in the 1975 orthopaedic procedures, ophthalmology, dentistry, Proceedings of the Royal Society of Medicine.4 They podiatry, physiotherapy, occupational therapy, published the case of two siblings with an unknown © Author(s) (or their employer(s)) 2019. No psychiatry and nursing. In addition, from early syndrome, together with their normal sister. In the commercial re-use. See rights on, social development and education featured text they concluded: “because of these children’s and permissions. Published prominently. similar appearance and mental deficiency, rash, by BMJ. The photographs throughout the book are very occasional fever, uveitis, leucocytosis and epiphysial To cite: Woo P, impressive—children and team members engaged disorders we suspect this is a genetic disorder prob- Wulffraat NM, Tyndall A. in all aspects of daily life in the hospital. There were ably involving some biochemical defect in the inflam- Ann Rheum Dis extensive tributes made to secretaries, lay persons mation process, just as haemophilia is determined by 2019;78:725–728. involved in fund raising and all those individuals who biochemical defect in a defect in the clotting process.

Woo P, et al. Ann Rheum Dis 2019;78:725–728. doi:10.1136/annrheumdis-2019-215106 725 Heroes and pillars of rheumatology

Box 1 barbara Ansell: Key achievements Box 2 dr Barbara Ansell CBE (1923–2001)

►► Clinical skills in describing new phenotypes Born: 30 August 1923 ►► Proposal for Juvenile Idiopathic Arthritis classification (in her Awards: CBE, 1 January 1982 MD thesis) ►► Training of fellows, allied health professionals and senior staff Fellowships members ►► Royal College of Physicians 1967 ►► Commitment to initiate local national and international ►► Honorary fellow of the Royal College of Surgeons 1985 expertise centres ►► Honorary fellow of the Royal Society of Medicine ►► Started multidisciplinary clinical rounds ►► Involvement of parents and teachers Qualifications ►► 1946: MB ChB (Birmingham), MRCS, LRCP (London) ►► 1951: MRCP (London) So far we have no idea of its mechanism.” In her typically collab- ►► 1969: MD (Birmingham) The Classification of Juvenile Chronic orative fashion she shared this experience internationally,5 thus Arthritis began the era of autoinflammatory cryopyrin mutation disorders.6 Career synopsis Team building and collaboration ►► 1946–1951: junior posts in Birmingham, Northampton and Visiting clinical fellows arriving at Taplow from afar soon learnt London that their medical degrees or seniority was no automatic pass- ►► 1951–1953: registrar to Prof Bywaters at MRC Special Unit port to hierarchical importance. In fact they were slotted into for Juvenile Rheumatism, Canadian Red Cross Hospital, the extensive clinical team at a modest level, to learn about phys- Taplow, Buckinghamshire, UK ical therapy, the special emotional needs of sick children living a ►► 1953–1955: research fellow USA, and junior post at long way from home and so many other critical skills. Hammersmith Hospital, London Barbara supported all members of the team to attend conferences ►► 1955–1957: registrar to Bywaters at Hammersmith Hospital, and workshops with funds generated mostly by her through grants, London invited talks and side projects. We accepted this as normal and only ►► 1957–1962: senior registrar to Bywaters at MRC Rheumatism realised later as we assumed our own departmental responsibilities Unit, Taplow what a constant effort that requires. ►► 1962–1985: consultant, MRC Rheumatism Unit, Taplow Her collaboration with and the respect from colleagues in other ►► 1976–88: consultant and head of Division of Rheumatology disciplines were legion, and unusual in those days. She was elected at the MRC fellow of separate medical and surgical societies, and awarded ►► Clinical Research Centre and Northwick Park Hospital; from the Queen’s honour of CBE for her work. This respect was not 1985 consultant at Wexham Park Hospital, Berkshire, UK. based on her powerbase but purely due to the fact that her clinical opinion was unparalleled. She worked as a rheumatologist for chil- Committee chair of dren, adolescents and adults, thus allowing her to have a unique ►► British Heberden Society: secretary 1960, president 1976, long term evolutionary view of some diseases. It also opened doors treasurer 1978 for us junior colleagues to experience cross specialty wisdom long ►► British Society for Rheumatology, Clinical Affairs Committee before the expression ‘thinking out of the box’ was invented. 1984–1986 One of Barbara’s long cherished aims was the decentralised care ►► Royal College of Physicians specialist advisory, Committee on of children and young people with rheumatic diseases and she Higher Medical devoted enormous time and energy into achieving this. She set up ►► Training (rheumatology) 1977–1980; Royal College of regular peripatetic clinics in UK and abroad. Although Taplow was Physicians vice president 1987–1988 her showpiece, where national and international pioneers emerged ►► Arthritis and Rheumatism Council Education and Scientific from spending time there, she always recognised the trauma expe- Subcommittees rienced by children being so far from home. For example, when ►► European League Against Rheumatism (EULAR) Standing asked once why the functional scores in children in clinical studies Committee for Paediatric Rheumatology 1981–1991 on Friday could be so much better than on Monday, she simply said ►► Founding chair of the British Paediatric Rheumatology Group “of course- the family comes to visit on the weekend—and after 1981–1988 they are gone, the children are sad”. Holistic medicine and ‘Patient Reported Outcomes’ were evident long before their time. Member of UK and International Committees, (advisory and charities) Capacity for work including international ►► Committee on Safety of Medicine Advisory Committees ► Arthritis and Rheumatism Council Scientific and Education networks ► Committees A hapless junior colleague once pleaded fatigue as the reason for not ► EULAR Standing Committee on Drugs, Executive Committee delivering his draft manuscript on time: “I stayed up until midnight ► working on it!” “I didn’t go to bed at all last night” was the unsym- pathetic response. Indeed, after a busy clinical day, Barbara would then retire to her office to work on the many administrative and despite an intense travel and committee schedule. Her committee political aspects of running her large department, frequently after work as detailed in box 2, reflected her comment “I like to keep everyone had left for the day. She famously said that she would my fingers in many pies”. She was appointed as the first chair- ‘stay up all night once a month to keep up with her paperwork’. person of the European League Against Rheumatism (EULAR) It was always impressive to those around how much she achieved, standing committee on paediatric rheumatology in 1981. During

726 Woo P, et al. Ann Rheum Dis 2019;78:725–728. doi:10.1136/annrheumdis-2019-215106 Heroes and pillars of rheumatology

Figure 1 Picture taken at Loch Lomond, Scotland, during the inaugural Paediatric Rheumatology European Society Meeting held jointly with the European League Against Rheumatism conference organised by the British Society for Rheumatology, 1998. Barbara Ansell surrounded by some of her successors in paediatric rheumatology. Left to right: Patience White, Alberto Martini, Barbara Ansell, Taunton Southwood, Patricia Woo. her tenure, she included paediatricians who care for very young children with rheumatic diseases: “they see more systemic diseases in children’s hospitals” and was keen to reach out to all. By 1997, even though she was retired, she nonetheless threw her support and influence behind the proposals by her successors to provide Figure 2 Dame Professor Carol Black on paying tribute to Barbara education and research infrastructure in paediatric rheumatology, Ansell at the Royal College of Physicians (taken from https://www. and supported them in the creation of a professional society: the rcplondon.ac.uk/projects/women-medicine-carol-black-and-barbara- Paediatric Rheumatology European Society (PReS) (figure 1). ansell). Support for women in medicine Whether her initial exclusion from a career in cardiology on the Children’s Hospital. Together with Anne-Marie Prieur, Wietse grounds of gender, or simply that she had an innate sense of fair- Kuis, Alberto Martini, Tauny Southwood and other members ness, Barbara always supported women in building their careers. She strove to facilitate part-time specialty training to allow of the EULAR standing committee, she set up the European women to be both mothers and specialist physicians. However, Society for Paediatric Rheumatology (PReS) in her role as its this was not to the detriment of the male team members, many first president. Within this framework, continuation of the of whom built their further careers on the experiences under her ethos of friendship, research and collaboration with colleagues tutelage. throughout Europe and beyond has resulted in better training There were many notable examples of men and women who of physicians, allied healthcare workers and researchers in the have gone on to establish de novo centres for children with rheu- field of paediatric and adolescent rheumatology, bringing it into matic diseases in UK and abroad. Some in addition contributed the modern era of genomic medicine and effective biological to innovations in academia and healthcare. Of these Patience therapy. White is a good example of an international Rheumatology trainee at Northwick Park, set up a paediatric rheumatology centre in Washington, DC on return, and went on to become Lifelong friendships beyond work a pioneer and advocate for an orderly transition to adolescent Early on Barbara attracted a group of peers from all parts of the and adult services. She is working at the heart of government on globe who remained loyal friends and collaborators during and this subject. after her professionally active years. There are too many to name A former junior colleague, Carol Black (figure 2), paid tribute here, but many have been her colleagues, especially from the to Barbara in her 2017 article for the Royal College of Physicians EULAR Standing Committee that she chaired. ‘Women in Medicine’.7 Dame Professor Carol Black, an inter- In turn, these colleagues and friends often exchange the next national leader in scleroderma, former president of the Royal generation of fellows. College of Physicians, is also principal of Newnham College, So a network of collaborators and friends was established and Cambridge University and expert adviser on health and work continues to this day. Having ‘done time’ with Barbara became to National Health Service England and Public Health England. an unofficial passport to acceptance in both the paediatric and Barbara’s ‘unfinished business’ was taken up by Patricia Woo, rheumatology worlds. a former protégée, who established clinical and research units It must be mentioned that this ‘time’ was not always halcy- at Great Ormond Street Children’s Hospital, and University on—she expected the same rigorous standards from others as College London. Figure 3 depicts her among patients at the she set for herself and at times this could be quite stressful to opening of the rheumatology unit of the Great Ormond Street the recipient.

Woo P, et al. Ann Rheum Dis 2019;78:725–728. doi:10.1136/annrheumdis-2019-215106 727 Heroes and pillars of rheumatology

generations of practitioners. Those who knew her well under- stood that under her tough exterior, resided a truly committed and passionate person who dedicated herself to her work. But dedication alone is not enough to move a field forward—there must also be a degree of intuitive genius to achieve this and Barbara Ansell certainly had this. For those of us who knew her well, we also relish the many idiosyncratic aspects of her complex character which made her such a special person in the course of our training. Time often blurs the moments when game changing paradigm shifts occur, but not the fond personal memories of those special people who facilitated them. Contributors AT, NW and PW all contributed to the tekst and reviewed the final version. Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests None declared. Figure 3 Barbara as guest of honor at the official opening of the Patient consent for publication Not required. rheumatology ward in Great Ormond Street Children’s Hospital, London, Provenance and peer review Commissioned; externally peer reviewed. 1995. References Fun and hospitality 1 Ansell DB. Contemporary history. Arch Dis Child 1988;63:1509–10. Barbara enjoyed entertaining and for many of us fellows a long 2 Medallist JS. DR Barbara Mary Ansell. Arch Dis Child 1997;77. way from home, Sunday lunches at her Dumgoyne home in the 3 Hull R, Venning H. DR Barbara Mary Ansell, CBE, FRCP, FRCS, FRCPCH: 1923-2001. Thames Valley were wonderful social affairs. Many long-term Arch Dis Child 2003;88:185. 4 Proceedings of the first Ara conference on the rheumatic diseases of childhood. Park friendships blossomed from these gatherings. We juniors watched City, Utah, March 22-25, 1976. Arthritis Rheum 1977;20(2 Suppl):145–628. in bemusement as our tough demanding boss transformed into an 5 Prieur A-M, Griscelli C, Lampert F, et al. A chronic, infantile, neurological, cutaneous attentive, hostess and housewife under Angus’ affectionate, almost and articular (CINCA) syndrome. A specific entity analysed in 30 patients. Scand J paternal gaze. Heaven help the junior who transported this famil- Rheumatol 1987;16:57–68. 6 Hoffman HM, Mueller JL, Broide DH, et al. Mutation of a new gene encoding a putative iarity to the wards on Monday! pyrin-like protein causes familial cold autoinflammatory syndrome and Muckle–Wells syndrome. Nat Genet 2001;29:301–5. Postscript 7 Royal College of Physicians. Women in medicine: Carol black and Barbara Ansell. Barbara Ansell is remembered first for her unique contribution Available: https://www.​rcplondon.ac.​ ​uk/projects/./​ ​women-​medicine-​carol-black-​ ​and-​ to paediatric rheumatology which still echoes down several barbara-ansell​

728 Woo P, et al. Ann Rheum Dis 2019;78:725–728. doi:10.1136/annrheumdis-2019-215106 Viewpoint Current challenges in the development of new treatments for lupus Maria Dall’Era,1 Ian N Bruce,2 Caroline Gordon,3 Susan Manzi,4 Janis McCaffrey,5 Peter E Lipsky6

Handling editor Josef S Abstract stakeholders. The aim of this review is to highlight Smolen Systemic lupus erythematosus (SLE) is a chronic pitfalls in drug development for SLE and identify 1 areas where collaboration and consensus will be Division of Rheumatology and autoimmune disease with a considerable impact on Russell/Engleman Rheumatology patients’ quality of life. Despite the plethora of clinical important to facilitate progress. Research Center, University of trials for SLE since the turn of the millennium, only one California San Francisco, San new treatment has been approved for the condition, Francisco, California, USA Lessons learned from previous clinical 2Arthritis Research UK Centre and the overall pace of successful drug development trials in SLE for Epidemiology, Medicine remains slow. Nevertheless, the myriad of clinical studies Formal clinical trials in SLE have been carried out and Health, The University has yielded insights that have informed and refined our for nearly two decades, although many failed to of Manchester and NIHR understanding of eligibility criteria, outcome measures meet their primary endpoints.12 Post hoc analyses Manchester Biomedical and trial design in SLE. In this review, we highlight the Research Centre, Manchester of these studies have identified numerous issues University Hospitals NHS achievements of clinical trials as well as the major pitfalls related to trial design,12 17 prompting investigators, Foundation Trust, Manchester that have been identified in drug development for SLE pharmaceutical companies and regulatory agencies Academic Health Science and, in doing so, identify areas where collaboration and to standardise aspects, such as entry criteria (by Centre, Manchester, UK consensus will be important to facilitate progress. 3Institute of Inflammation defining thresholds for disease activity and autoan- and Ageing, University of tibody positivity), outcome measures, preferences Birmingham, Edgbaston, Introduction for trial duration and the choice of background Birmingham, UK Systemic lupus erythematosus (SLE) is a chronic therapy (including regimens for steroid use and 4Lupus Center of Excellence, Allegheny Health Network, autoimmune disease that follows a relapsing–remit- tapering). Pittsburgh, Pennsylvania, USA ting course. It is characterised by the production Many of these issues have been decided prag- 5Private Consultative Practice, of autoantibodies against a range of autoantigens matically, often in discussions between pharmaceu- West Vancouver, British including nuclear components, immune-mediated tical/biotech companies and regulatory agencies.12 Columbia, Canada 6 inflammation in a variety of organs and the accrual Whether the final negotiated approach is the most RILITE Research Institute, 1 Charlottesville, Virginia, USA of organ damage over time. Despite its prevalence scientifically valid has often not been determined, (approximately 70 cases per 100 000 people)2–6 and but operationally a ‘standard’ SLE therapeutic trial 7 8 9 10 Correspondence to considerable impact on quality of life, treatment design has emerged. Despite previous trial fail- Dr Maria Dall’Era, UCSF School options remain inadequate, and the development of ures, many companies continue to have a strong of Medicine, University of novel therapies has been slow. commitment towards developing new drugs for California San Francisco, San Francisco, CA, USA; Only one new drug, belimumab, has been SLE. In some cases, this may relate to promising Maria.​ ​DallEra@ucsf​ .​edu approved for the treatment of SLE in more than 60 trends in the trial data that differentiate potential years.9–11 While numerous other drugs and biolog- responders from non-responders.18 Other times, it Received 5 October 2018 icals have entered clinical trials for SLE since the is learning from another company’s successes and Revised 16 November 2018 14 turn of the millennium, development of most has failures. Regardless, the pace of clinical trials in Accepted 19 December 2018 12–14 Published Online First been halted at various stages. Although some SLE remains brisk. 12 January 2019 of these failures may be related to inefficacy of the Some of the failed trials in SLE might have tested investigational product or adverse events, there agents that were truly ineffective. However, in is general consensus in the community that prob- other cases, it was reasonable to expect success, lems with trial design and operation may have such as when testing agents related to compounds contributed to the unsuccessful outcomes. Notably, that had previously been successful (belimumab vs however, a number of recent phase IIb trials (each tabalumab), agents with strong evidence of effi- investigating agents with different modes of action) cacy from clinical experience (rituximab) or agents have reported positive outcomes,15 16 suggesting with extensive preclinical experience (abatacept that new treatment options may be on the horizon, plus cyclophosphamide [CYC]). These experiences though the recent failure of a phase III anifrolumab suggest that the current trial paradigm for SLE may trial to meet its primary endpoints, reported in the not be capable of convincingly testing the efficacy lay press, suggests that there is more to learn. of a compound, or definitively deciding whether to © Author(s) (or their Although the majority of trials in SLE have advance a compound into phase III. employer(s)) 2019. No commercial re-use. See rights failed to meet their primary endpoints, many of and permissions. Published these studies have offered valuable insights that Disease activity as an entry criterion by BMJ. have advanced the field considerably, including the One of the earliest trials in SLE examined the effi- 19 To cite: Dall’Era M, evolution of thinking about eligibility criteria and cacy of dehydroepiandrosterone in 381 patients. Bruce IN, Gordon C, outcome measures. Nevertheless, several important Although this study did not meet its primary et al. Ann Rheum Dis issues remain unresolved, and progress in these endpoint, valuable insights were obtained into the 2019;78:729–735. areas will require consensus among the relevant design of clinical trials in SLE. Specifically, a post

Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 729 Viewpoint hoc analysis indicated that the primary endpoint was only met results of a prespecified subgroup analysis suggested that ritux- in patients with a baseline SLE Disease Activity Index (SLEDAI) imab might be more efficacious in AA/Hispanic patients with score >2. As such, this was the first trial demonstrating the lupus nephritis.29 However, given the poor recruitment of AA potential utility of disease activity as an entry criterion. patients into SLE trials, these findings are at best speculative. Subsequent trials have confirmed the importance of recruiting Recent systematic reviews of potential predictors of response to patients with moderate-to-high disease activity in order to MMF and rituximab have concluded that the overall quality of discern a treatment effect. This has generally been achieved by evidence supporting differential effects related to ancestry and specifying disease activity beyond a certain threshold for inclu- ethnicity are low, and further confirmatory studies are needed sion (eg, SLEDAI ≥6 or a British Isles Lupus Assessment Group to discern whether such a difference exists.30 31 Other factors, (BILAG) score of B or A, indicating active organ involvement).14 including the small size of the trials and socioeconomic status, Post hoc analysis of the Belimumab in Subjects with Systemic may have contributed to the result. Lupus Erythematosus (BLISS) trials revealed additional benefits of belimumab treatment in patients with higher disease activity Heterogeneity of clinical manifestations (SLEDAI ≥10) at baseline.20 Notably, this increased benefit SLE is a heterogeneous disease, involving numerous organ seemed to rely on decreased responsiveness in patients receiving systems. In general, trials have focused either on generalised SLE standard of care only. Thus, the absolute difference in response or lupus nephritis. In trials of generalised SLE, it is common between the group receiving standard of care plus belimumab, to exclude patients with rapidly progressive kidney or central and the group receiving standard of care alone, was increased. nervous system (CNS) disease.9 10 As a result, studies have This observation suggests that standard of care therapy can be largely enrolled patients with mucocutaneous and musculoskel- effective in the setting of a clinical trial and could be especially etal involvement. This limits the heterogeneity of the enrolled effective in patients with mild disease on entry (despite having population and excludes patients with less common and more apparently failed standard of care before enrolment). Discerning severe disease manifestations. As a result, little information has a response to an experimental agent appears to be more likely emerged on the efficacy of agents in these patients. in patients with more active disease, in whom the response to Lupus nephritis trials have typically enrolled patients with background therapy is diminished. biopsy-proven nephritis, although there is considerable varia- tion in the timing of the biopsy relative to trial entry.32 As a Geographic location result, the exact nature of kidney involvement at the time of International SLE trials have demonstrated major differences in trial enrolment may not be certain, owing to the possible role treatment response according to geographical region. Specifi- of hypertensive kidney disease and progressive kidney fibrosis. cally, sites in developing countries generate data that indicate a Moreover, entry has usually been based on glomerular involve- better response to placebo (plus standard of care) compared with ment, despite evidence that interstitial inflammation may play a results obtained in the USA or Western Europe.21 The reason for more important role in progression to kidney failure.33 In addi- this discrepancy is uncertain but might relate to patients’ access tion, the level of proteinuria at baseline might be an important (or lack thereof) to standard of care therapies before entering the predictor of renal response. For example, in a trial comparing trial and the requirement to consistently take these medications the addition of abatacept or placebo to background MMF in in the trial setting. A second concern is the increased frequency of patients with lupus nephritis, those with baseline nephrotic infection22 as well as trial-related deaths23 24 in studies conducted syndrome had a greater response to abatacept as measured by in developing countries. Part of this may relate to higher rates reduction in proteinuria.34 of certain infections in these countries, coupled with the lack of access to standard and emergency medical care. Thus, while Outcome measures there are major attractions to recruiting patients from devel- Although numerous outcome measures were used in early SLE oping countries (including low expense, rapid enrolment and trials, the Food and Drug Administration (FDA) released draft the prevalence of active disease), there is a clear need to guard guidelines in 2005 to facilitate their standardisation.35 This against a greater placebo response and to mitigate the intrinsic included the recommendation of a composite endpoint that risk of serious adverse events in these regions. could measure clinically meaningful improvements in active organ systems without worsening in others and detect both early Ancestry/ethnicity and overall changes in disease activity. The FDA seemed to favour A number of SLE trials have suggested preferential response to the BILAG36; however, the success of the belimumab BLISS trials treatment in certain ancestral subsets. The comparative trial of resulted in widespread adoption of the SLE Responder Index mycophenolate mofetil (MMF) versus CYC in lupus nephritis (SRI)-4 as the primary outcome measure. The BILAG-Based suggested that MMF might be more efficacious in a primarily Composite Lupus Assessment (BICLA) and the SRI-5 were used African ancestry (AA)/Hispanic trial population.25 In contrast, the in unsuccessful phase III trials of epratuzumab and tabalumab, induction phase of the Aspreva Lupus Management Study trial, respectively.37–39 However, the BICLA was employed as an with patients of predominantly European Ancestry did not show outcome in successful phase II trials of anifrolumab15 and epratu- any significant difference between the two drugs.26 However, zumab40 and in a post hoc analysis of an interleukin-6 receptor post hoc analysis suggested an improved response to MMF antibody,24 both in generalised SLE. in the AA/Hispanic group.27 These findings were sufficiently The decision to adopt the SRI-4 was driven from a retrospec- convincing that the 2012 American College of Rheumatology tive analysis of the phase II belimumab data, which found that (ACR) Lupus Nephritis Treatment Guidelines recommend MMF this instrument distinguished active drug from placebo. This over CYC as first-line treatment for AA and Hispanic popula- hypothesis was borne out in subsequent trials with both intra- tions.28 Results from the pivotal trials of belimumab suggested venous and subcutaneous belimumab.41 42 In addition, the FDA that this agent was not effective in AA patients,9 10 although the believed it was important to determine whether an agent might numbers of AA patients in these trials were small. In contrast, the be effective in some domains while at the same time measuring

730 Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 Viewpoint whether deterioration occurs in others. However, subsequent SLE trials (both successful and unsuccessful), we firmly believe analyses have shown that the overwhelming contributor to the that it is time for the regulatory guidance to be revised through SRI-4 is improvement in SLEDAI alone, with very few patients collaboration between a range of stakeholders, including physi- manifesting improvement in SLEDAI but deterioration in BILAG cians, patients, members of industry and regulatory agencies who or Physician’s Global Assessment (PGA).9 10 Since the majority have expertise in this area. In addition, evidence-based guidance of patients in the belimumab trials had skin and joint involve- for outcomes in lupus nephritis are sorely needed. An even better ment, SLEDAI was sufficient to capture improvement in this option would be to propose and adopt new trial designs and case. However, SLEDAI cannot capture some types of organ outcome measures through an accelerated process that could involvement, and therefore may not be sufficient for assessing bypass a traditional guidance document. the outcomes of less common disease manifestations. In addi- tion, higher baseline SLEDAI scores may improve the ability to The role of serological assessments in entry criteria detect an SRI-4 response. The use of serological biomarkers for enrolment in SLE trials is a It is also important to note that while the SRI-4 continues to relatively recent development, having become standard practice be used as the preferred primary outcome measure in major SLE within the past decade. This approach was based largely on a clinical trials, it is not used as an assessment in routine clinical post hoc analysis of phase II belimumab data,51 which revealed a practice. This creates a disconnect between clinical trial data subset of autoantibody positive patients who responded to beli- and information that is meaningful to a practitioner or patient mumab better than their seronegative counterparts. Before this, and might delay the uptake of newly approved therapies into eligibility was based primarily on the ACR Classification Criteria practice, since physicians are unclear about the expected clinical for SLE (which do not require a serological criterion) and estab- response and are often inexperienced at using the component lished indices of disease activity, with no stand-alone serological instruments (SLEDAI, BILAG and PGA). Nevertheless, there is component.19 52 While SLEDAI includes measures of anti-DNA now a requirement to use BILAG and SLEDAI when assessing and complement levels, antinuclear antibody (ANA) positivity patients for treatment with rituximab or belimumab in the was not always required for trial entry. National Health Service in England.43–45 However, both of these Although ANA, anti-DNA, low complement C3 and C4 and/ instruments have significant limitations, including low interob- or anti-Smith autoantibodies are now commonly used as entry server correlation and high complexity in the case of the BILAG. criteria, there is no consensus on the specific autoantibody or Lastly, careful consideration should be given to the use of flare serological marker required, nor the method used to measure it. rather than response as an endpoint.46 As previously noted, the use of autoantibody positivity to clas- Collectively, these considerations suggest that a re-evaluation sify patients was based on phase II data from a B cell directed of outcome measures in clinical trials based on available data therapy, providing some rationale for its use in that context.51 might be necessary to develop more clinically meaningful ways However, the use of serological positivity to subset patients in to assess the impact of new therapies. Unfortunately, an initial trials of agents not directed toward B cell function may not be attempt to create a novel SLE response index from the belim- appropriate. umab trial data was not successful.47 However, a more recent In reality, this approach has been embraced in an attempt to effort based on the BLISS-76 data yielded the Lupus Multivari- confirm that patients enrolled in trials actually have SLE. It is able Lupus Outcome Score (LuMOS), a robust outcome measure thought that a positive ANA test increases that likelihood, and that outperformed the SRI-4 and was validated in the BLISS-52 therefore the test has been widely employed as an entry criterion. database.48 If recapitulated in trials for agents with other mech- Of note, the recent SLE classification criteria developed by the anisms of actions, LuMOS will be the first evidence-based SLE American College of Rheumatology and the European League outcome measure and may provide a better means to assess treat- Against Rheumatism require a positive ANA at 1:80 titre.53 ment response in clinical trials.48 As new outcome measures are However, given its poor specificity, variability between assays developed, it will be important to perform validation studies and the observation that patients with SLE may convert to a examining their association with longer term outcomes, quality negative ANA over time and with immunosuppressive therapy,54 of life and cost-effectiveness. the use of a positive ANA to classify patients and assess eligibility The aforementioned FDA guidelines have since been replaced in clinical trials is problematic. Using more specific tests, such as by a finalised document, which was issued in 2010 and is still a positive anti-DNA, imposes additional problems.55 Restricting operational.49 This includes guidance for a range of trial design trials to anti-DNA positive patients would exclude approximately issues, including statistical considerations. At the time of release, 40% of patients with SLE because of the lesser sensitivity of the FDA issued separate documents for SLE and lupus nephritis, these assays. Similarly, a requirement for low complement levels although the latter was eventually withdrawn and has not yet would exclude many patients, particularly those in non-renal been replaced. In the absence of specific guidelines for lupus lupus trials. In short, the utility of specific serological testing for nephritis, the most effective short-term renal response measures assessing eligibility in SLE trials has not been carefully examined remain elusive, with different instruments generating significant and will require thoughtful consideration in the future. variability in the data obtained and the conclusions drawn.50 Importantly, the process by which new outcome measures might be incorporated into SLE clinical trials has not been clearly Biomarkers articulated by the regulatory agencies. Some outcome measures, Apart from the use of autoantibodies, and possibly complement such as the Cutaneous Lupus Area and Severity Index to eval- levels, no biomarkers are currently accepted as informative in uate skin involvement have been developed by the academic SLE clinical trials. Recent data have indicated that anifrolumab, community but not accepted by regulatory authorities as a vali- a monoclonal antibody to the type 1 interferon receptor, might dated outcome measure in clinical trials. As a result, nearly all be effective in generalised SLE based on phase II data.15 These SLE trials rely on evidence derived from the standard composite results also suggested that the treatment was only effective in outcome measures described in the 2010 FDA guidance docu- individuals who had increased expression of certain interfer- ment. Given the recent developments and lessons learned from on-responsive gene products. There are phase III trials currently

Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 731 Viewpoint underway with secondary endpoints to assess whether this ‘inter- While understandable in terms of cost and efficiency, agents feron signature’ might facilitate stratification, and thus serve as such as tabalumab and epratuzumab61 may ultimately have failed a companion diagnostic to identify patients with a greater like- because of these drawbacks. Thus, the effectiveness of phase III lihood of responding to treatment.56 If successful, this would trials hinges critically on the data obtained in phase II, high- represent the first biomarker for selecting and monitoring the lighting the need to conduct more extensive studies at this stage. treatment of patients with SLE. Areas in need of collaboration and consensus Background therapy The authors believe that there are two major areas in which Most trials in SLE test new agents on the background of stan- collaboration and consensus would be useful. The first relates to dard of care therapy. This stems largely from investigators’ data that are already available from clinical trials in SLE. Over concerns about the development of irreversible organ damage 5000 patients have participated in these studies, providing a in the placebo group if adequate background medication were wealth of data that may help us to answer important questions not continued. However, there are a number of problems with about trial design and conduct. The second area relates to issues this approach. that cannot be addressed using the existing trial data alone. First, standard of care can be quite effective in many patients, Questions falling within each of these areas are listed in box 1. leaving a narrow window for improvement with the experi- mental agent, particularly if patients take their medications more The current landscape reliably in the trial setting.57 Second, there is no consensus on what constitutes ‘standard of care’, meaning that participants Various initiatives are already taking shape to address the within the same trial may be treated with different background outstanding issues in SLE clinical trials. Among these is the Defi- medications. Third, many trials permit variation in the dosing nitions of Remission In SLE task force, an international collab- of background medication for a period of time before or during oration between 60 specialists and patient representatives to the study, making it even more difficult to discern benefit from formulate a universal definition of ‘remission’ in SLE, in line the experimental agent. Finally, recent work from the Systemic with treat-to-target recommendations.62 The task force recently Lupus International Collaborating Clinics cohort has revealed published a framework for this purpose comprising eight key substantial variation in the dose of steroids used across centres. statements, as well as three principles to guide the refinement While patient and disease factors contribute to this variation, of this definition in future.62 Since remission is rarely a goal differences between centers suggest that physician-related in SLE trials, because it is too stringent and hard to achieve,63 factors also contribute. Such patient-independent heterogeneity optimal trial design will require the development of alternative in steroid use will contribute to ‘noise’ in multicentre trials and endpoints. To this end, one group of investigators is developing increase the likelihood of type 2 errors. A period of standardi- the Lupus Low Disease Activity State (LLDAS) for generalised sation in steroid use prior to randomization may be necessary to SLE,64 and another group has developed a low disease activity address such variation as was done in the APRIL-SLE atacicept outcome.65 These new endpoints could initially be incorpo- trial.23 58 rated in SLE trials as a secondary endpoint for comparison with Two recent studies that support the possibility of minimising existing instruments and might eventually replace SRI or SLEDAI background therapy through steroid tapering are the phase II as the primary outcome measure. For example, LLDAS has been anifrolumab and voclosporin trials,15 59 both of which reported incorporated as a secondary endpoint in the currently enrolling low placebo responses. These studies suggest, though do not BLISS-BELIEVE trial sponsored by GlaxoSmithKline.66 prove, that lowering standard of care therapy in a safe manner Further to these efforts, the Lupus Industry Counsel is might permit better discernment of the benefit of the experi- working with Outcome Measures in Rheumatology to develop mental agent. This would also be beneficial for the safety of trial new instruments by reanalysing existing trial data,67 and the participants by providing clearer criteria for rescue therapy or Lupus Foundation of America (LFA) is developing a system of study withdrawal. tools (dubbed LFA-REAL) for the measurement of both systemic and organ-specific disease activity.68 Another recent focus of collaboration in SLE is the identifica- Time and expense tion of genetic and/or phenotypic markers for patient stratifica- Clinical drug development is very costly and time consuming. tion. The Innovative Medicines Initiative is an European Union This has frequently prompted pharmaceutical companies to (EU)-based, public–private initiative to improve the drug devel- design clinical development programs that save time and use the opment process in Europe. Among the projects currently being fewest number of patients. Since drug approval requires data conducted under this initiative is Molecular Reclassification to 60 in two well-designed clinical trials, the standard phase I to III Find Clinically Useful Biomarkers for Systemic Autoimmune programs have often been truncated, typically by short-circuiting Diseases (PRECISESADS), which is bringing together academic, phase II trials or even treating them as pivotal in order to save clinical and industry partners to identify biomarkers that will time and resources. As a result, phase III trials are often initi- assist with developing a new taxonomy for systemic autoim- ated without proper dose-ranging or robust power calculations mune diseases, including SLE.69 PRECISESADS therefore aims and with little information on the optimal regimen and outcome to group patients with a shared pathogenesis, rather than using measures. In addition, the potential contribution of ancestry to standard diagnostic and clinical classification criteria.70 trial outcomes is frequently neglected (the FDA has requested In the UK, the Medical Research Council is funding MAxi- additional trials of belimumab in patients of African descent). mizing Sle ThERapeutic PotentiaL by Application of Novel and The impact of body composition may also be underestimated. Stratified approaches (MASTERPLANS).71 MASTERPLANS Most trials use fixed-dose drug regimens regardless of body mass seeks to identify patient-level variables, including genetic and index, which does not account for the potential pharmacokinetic other biomarkers, that define an endotype of response in SLE variation between patients. patients, with an initial focus on mycophenolate and rituximab.

732 Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 Viewpoint

Box 1 Aspects of SLE drug development that are in need Box 1 Continued of collaboration and consensus –– What is the role of PROs, and which is the most Questions that can be addressed using existing clinical appropriate to use? trial data: –– What is the role of PROs in clinical trial design and ►► Adverse events drug approval? Should drugs be approved based on –– What factors influence adverse events in SLE trials? improvement in PROs? What is the contribution of –– When do adverse events occur? comorbidities to PROs? –– Are adverse events more likely during the first few weeks ►► Background therapy of the trial when steroid doses tend to be higher? –– What is ‘standard of care’? Are standard of care regimens –– What is the evidence that adverse events are influenced by used for clinical or ethical reasons? steroid dose and/or background immunosuppressants? –– Should there be a mandatory steroid taper versus the –– Are adverse events more common in particular geographic commonly used ‘suggested’ steroid taper? Should steroid regions or certain types of research institution than others? taper to a certain dose be part of the primary endpoint? ►► Outcome measures ►► Other –– Is the SRI the most appropriate outcome measure? –– What does ‘treat to target’ really mean in lupus? –– Does it matter which version of the SRI is used? AUC, area under curve; PRO, patient-reported outcome; SLE, systemic –– Can we simplify this by focusing only on improvement in lupus erythematosus; SRI, SLE Responder Index. disease activity, or change in disease activity using a single instrument? –– Which outcome measures should be used in generalised The contribution of industry partners to these initiatives is SLE trials, and which in lupus nephritis trials? particularly valuable, given their large datasets and extensive –– Should we consider flare prevention as a primary outcome libraries of patient samples. Moreover, once new biomarkers are measure? identified, their uptake and use in clinical trials by industry part- –– Which outcome measures have performed best in lupus ners will be crucial for their validation. nephritis trials? Are these endpoints influenced by the duration of renal involvement, time since renal biopsy, the result of renal biopsy and/or a patient’s treatment history? Conclusion While most recent trials in SLE have failed to meet their primary ►► Background therapy –– Do trial outcomes differ with different background endpoints, many have offered valuable insights that have shaped regimens? the field as we know it today. Nevertheless, meaningful change –– What is the role of background steroids? from this point forward will require community-wide consensus, –– Should the target steroid dose depend on the patient’s including academia, industry and patient organisations, in baseline dose or body weight? dialogue and partnership with regulators and payers. In short, –– Should there be a ‘run-in’ period to standardise steroid use it is imperative that common goals are identified and that the prior to randomisation? community speaks with one voice. By forging new collaborations, reanalysing existing datasets ►► Regional differences –– Are there consistent regional differences in the use of and pooling data from multiple sources in a ‘pre-competitive’ background medications, or the development of adverse way, we will gain new insights to help improve trial design using events? evidence-based and data-driven approaches. The ability to iden- –– Should trials require a mandatory number of subjects of tify subsets of disease with a common pathogenic mechanism, or different ancestries? How should ancestry be determined? patient endotypes more likely to respond to a particular thera- How should mixed ethnicity be dealt with? peutic agent, will also allow us to offer patients the right drug for their condition in a timely manner and shorten the time from ►► Others –– What are the best biomarkers to judge efficacy?Are they diagnosis to disease control. Keeping the patient at the forefront specific to the agent’s mechanism of action, or related to of our efforts should spur us to achieve consensus on the way disease manifestations? forward and steer efforts to refine the guidance for trial design –– Is autoantibody stratification useful and/or necessary? and conduct. –– Will an AUC approach be better than a landmark analysis? Acknowledgements The authors conceived of and wrote the manuscript. The authors acknowledge Sam Fraser, PhD (Costello Medical, Cambridge, UK) for medical writing and editorial assistance in preparing an early draft of this manuscript, which Questions that require expertise and consensus beyond was funded by UCB Pharma and based on the authors’ input and direction. UCB what is available from clinical trial data Pharma was not involved in the development or review of the final version. ►► Trial design Contributors Substantial contributions to the development of this publication, or –– What is the most appropriate length of a trial? Should revising it critically for important intellectual content: MD, INB, CG, SM, JM and PEL; there be ‘induction’ and ‘maintenance’ trials? final approval of the publication: MD, INB, CG, SM, JM and PEL. –– What is the most appropriate design of a non-renal SLE Funding This article was funded by UCB Pharma. trial? Should we enroll active patients and measure Disclaimer The views expressed in this publication are those of the author(s) and improvement in disease activity, or should we be not necessarily those of the NHS, the NIHR, or the Department of Health. measuring ‘time to flare’ instead after initial treatment of Competing interests INB is a National Institute for Health Research (NIHR) flare? senior investigator and is supported by the NIHR Manchester Biomedical Research ►► Patient-reported outcomes Centre. He has also received grants from Genzyme Sanofi and GSK and undertaken consultancies and speakers’ bureau for GSK, AstraZeneca, UCB Pharma, Merck Continued Serono, Genzyme Sanofi, Eli Lilly and BMS. CG has undertaken consultancies and received honoraria from BMS, GSK, EMD Serono and UCB Pharma, related to

Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 733 Viewpoint clinical trial design and analysis. CG has also been a member of the speakers’ 21 Khamashta M, Merrill JT, Werth VP, et al. Sifalimumab, an anti-interferon-α bureau for GSK and UCB Pharma. She has received research grant support at monoclonal antibody, in moderate to severe systemic lupus erythematosus: Sandwell and West Birmingham Hospitals NHS Trust from UCB Pharma in the a randomised, double-blind, placebo-controlled study. Ann Rheum Dis past and currently, unrelated to the study of any specific drug and without any 2016;75:1909–16. personal payments to herself. She has participated in clinical trials sponsored by 22 Mysler EF, Spindler AJ, Guzman R, et al. Efficacy and safety of ocrelizumab in active UCB Pharma in the past and funded by Arthritis Research UK, with drug supplied proliferative lupus nephritis: results from a randomized, double-blind, phase III study. by GSK currently. She has no financial interests in any of these companies and does Arthritis Rheum 2013;65:2368–79. not and will not benefit from the sales of any drugs manufactured or developed by 23 Isenberg D, Gordon C, Licu D, et al. Efficacy and safety of atacicept for prevention of them. SM has a patent portfolio that was licenced and commercialised by Exagen, flares in patients with moderate-to-severe systemic lupus erythematosus (SLE): 52- Inc. She has served as a consultant or member of an advisory board for Exagen, Inc, week data (APRIL-SLE randomised trial). Ann Rheum Dis 2015;74:2006–15. GSK, UCB Pharma and AstraZeneca. She serves as the Chair, Board of Directors for 24 Wallace DJ, Strand V, Merrill JT, et al. Efficacy and safety of an interleukin 6 the Lupus Foundation of America. MD, JM and PEL have no conflicts of interest to monoclonal antibody for the treatment of systemic lupus erythematosus: a phase II declare. dose-ranging randomised controlled trial. Ann Rheum Dis 2017;76:534–42. Patient consent for publication Not required. 25 Ginzler EM, Dooley MA, Aranow C, et al. Mycophenolate mofetil or intravenous cyclophosphamide for lupus nephritis. N Engl J Med 2005;353:2219–28. Provenance and peer review Not commissioned; externally peer reviewed. 26 Appel GB, Contreras G, Dooley MA, et al. Mycophenolate mofetil versus cyclophosphamide for induction treatment of lupus nephritis. J Am Soc Nephrol References 2009;20:1103–12. 1 Bruce IN, O’Keeffe AG, Farewell V, et al. Factors associated with damage accrual 27 Isenberg D, Appel GB, Contreras G, et al. Influence of race/ethnicity on response to in patients with systemic lupus erythematosus: results from the Systemic Lupus lupus nephritis treatment: the ALMS study. Rheumatology 2010;49:128–40. International Collaborating Clinics (SLICC) Inception Cohort. Ann Rheum Dis 28 Hahn BH, McMahon MA, Wilkinson A, et al. American College of Rheumatology 2015;74:1706–13. guidelines for screening, treatment, and management of lupus nephritis. Arthritis Care 2 Dall’Era M, Cisternas MG, Snipes K, et al. The incidence and prevalence of systemic Res 2012;64:797–808. lupus erythematosus in san francisco county, california: the california lupus 29 Rovin BH, Furie R, Latinis K, et al. Efficacy and safety of rituximab in patients with surveillance project. Arthritis Rheumatol 2017;69:1996–2005. active proliferative lupus nephritis: the Lupus Nephritis Assessment with Rituximab 3 Ferucci ED, Johnston JM, Gaddy JR, et al. Prevalence and incidence of systemic lupus study. Arthritis Rheum 2012;64:1215–26. erythematosus in a population-based registry of American Indian and Alaska native 30 Mendoza-Pinto C, Pirone C, van der Windt DA, et al. Can we identify who gets benefit people, 2007-2009. Arthritis Rheumatol 2014;66:2494–502. or harm from mycophenolate mofetil in systemic lupus erythematosus? A systematic 4 Izmirly PM, Wan I, Sahl S, et al. The incidence and prevalence of systemic lupus review. Semin Arthritis Rheum 2017;47:65–78. erythematosus in New York county (Manhattan), New York: the manhattan lupus 31 Pirone C, Mendoza-Pinto C, van der Windt DA, et al. Predictive and prognostic factors surveillance program. Arthritis Rheumatol 2017;69:2006–17. influencing outcomes of rituximab therapy in systemic lupus erythematosus (SLE): a 5 Lim SS, Bayakly AR, Helmick CG, et al. The incidence and prevalence of systemic systematic review. Semin Arthritis Rheum 2017;47:384–96. lupus erythematosus, 2002-2004: the Georgia lupus registry. Arthritis Rheumatol 32 Anders HJ, Jayne DR, Rovin BH. Hurdles to the introduction of new therapies for 2014;66:357–68. immune-mediated kidney diseases. Nat Rev Nephrol 2016;12:205–16. 6 Somers EC, Marder W, Cagnoli P, et al. Population-based incidence and prevalence 33 Wilson PC, Kashgarian M, Moeckel G. Interstitial inflammation and interstitial of systemic lupus erythematosus: the Michigan lupus epidemiology and surveillance fibrosis and tubular atrophy predict renal survival in lupus nephritis. Clin Kidney J program. Arthritis Rheumatol 2014;66:369–78. 2018;11:207–18. 7 McElhone K, Abbott J, Sutton C, et al. Sensitivity to change and minimal important 34 Furie R, Nicholls K, Cheng TT, et al. Efficacy and safety of abatacept in lupus differences of the LupusQoL in patients with systemic lupus erythematosus. Arthritis nephritis: a twelve-month, randomized, double-blind study. Arthritis Rheumatol Care Res 2016;68:1505–13. 2014;66:379–89. 8 Mok CC, Ho LY, Cheung MY, et al. Effect of disease activity and damage on quality of 35 FDA. Guidance for industry. Systemic lupus erythematosus — developing drugs for life in patients with systemic lupus erythematosus: a 2-year prospective study. Scand J treatment. Draft Guidance 2005. Rheumatol 2009;38:121–7. 36 Murphy CL, Yee CS, Gordon C, et al. From BILAG to BILAG-based combined lupus 9 Furie R, Petri M, Zamani O, et al. A phase III, randomized, placebo-controlled study of assessment-30 years on. Rheumatology 2016;55:1357–63. belimumab, a monoclonal antibody that inhibits B lymphocyte stimulator, in patients 37 Clowse ME, Wallace DJ, Furie RA, et al. Efficacy and safety of epratuzumab in with systemic lupus erythematosus. Arthritis Rheum 2011;63:3918–30. moderately to severely active systemic lupus erythematosus: results from two 10 Navarra SV, Guzmán RM, Gallacher AE, et al. Efficacy and safety of belimumab in phase iii randomized, double-blind, placebo-controlled trials. Arthritis Rheumatol patients with active systemic lupus erythematosus: a randomised, placebo-controlled, 2017;69:362–75. phase 3 trial. The Lancet 2011;377:721–31. 38 Isenberg DA, Petri M, Kalunian K, et al. Efficacy and safety of subcutaneous 11 Questcor Pharmaceuticals Inc. H.P. Acthar® Gel (repository corticotropin injection) tabalumab in patients with systemic lupus erythematosus: results from ILLUMINATE-1, [prescribing information]. 1952 (Updated), 2012. a 52-week, phase III, multicentre, randomised, double-blind, placebo-controlled study. 12 Mahieu MA, Strand V, Simon LS, et al. A critical review of clinical trials in systemic Ann Rheum Dis 2016;75:323–31. lupus erythematosus. Lupus 2016;25:1122–40. 39 Merrill JT, van Vollenhoven RF, Buyon JP, et al. Efficacy and safety of subcutaneous 13 Narain S, Furie R. Update on clinical trials in systemic lupus erythematosus. Curr Opin tabalumab, a monoclonal antibody to B-cell activating factor, in patients with Rheumatol 2016;28:477–87. systemic lupus erythematosus: results from ILLUMINATE-2, a 52-week, phase III, 14 Wallace DJ. Improved strategies for designing lupus trials with targeted therapies: multicentre, randomised, double-blind, placebo-controlled study. Ann Rheum Dis learning from 65 years of experience. Lupus 2016;25:1141–9. 2016;75:332–40. 15 Furie R, Khamashta M, Merrill JT, et al. Anifrolumab, an anti-interferon-α receptor 40 Wallace DJ, Kalunian K, Petri MA, et al. Efficacy and safety of epratuzumab in patients monoclonal antibody, in moderate-to-severe systemic lupus erythematosus. Arthritis with moderate/severe active systemic lupus erythematosus: results from EMBLEM, Rheumatol 2017;69:376–86. a phase IIb, randomised, double-blind, placebo-controlled, multicentre study. Ann 16 van Vollenhoven R, Hahn B, Tsokos G. S7A:8 Efficacy and safety of ustekinumab, an Rheum Dis 2014;73:183–90. interleukin 12/23 inhibitor, in patients with active systemic lupus erythematosus: 41 Furie R, Petri MA, Strand V, et al. Clinical, laboratory and health-related quality of life results of a phase 2, randomised placebo-controlled study. Lupus Sci Med correlates of systemic lupus erythematosus responder index response: a post hoc 2018;5:A28–A29. analysis of the phase 3 belimumab trials. Lupus Sci Med 2014;1:e000031. 17 Touma Z, Gladman DD. Current and future therapies for SLE: obstacles and 42 Stohl W, Schwarting A, Okada M, et al. Efficacy and safety of subcutaneous recommendations for the development of novel treatments. Lupus Sci Med belimumab in systemic lupus erythematosus: a fifty-two-week randomized, double- 2017;4:e000239. blind, placebo-controlled study. Arthritis Rheumatol 2017;69:1016–27. 18 Merrill JT, Shanahan WR, Scheinberg M, et al. Phase III trial results with blisibimod, 43 Gordon C, Amissah-Arthur MB, Gayed M, et al. The British society for rheumatology a selective inhibitor of B-cell activating factor, in subjects with systemic lupus guideline for the management of systemic lupus erythematosus in adults. Rheumatol erythematosus (SLE): results from a randomised, double-blind, placebo-controlled trial. 2018;57:e1–e45. Ann Rheum Dis 2018;77:883–9. 44 McCarthy EM, Sutton E, Nesbit S, et al. The potential overlapping populations for 19 Petri MA, Mease PJ, Merrill JT, et al. Effects of prasterone on disease activity and treatment with belimumab and rituximab using current NHS England and national symptoms in women with active systemic lupus erythematosus. Arthritis Rheum institute for health and care excellence guidelines in England and Wales. Rheumatol 2004;50:2858–68. 2017;56:1049. 20 van Vollenhoven RF, Petri MA, Cervera R, et al. Belimumab in the treatment of 45 McCarthy EM, Sutton E, Nesbit S, et al. Short-term efficacy and safety of rituximab systemic lupus erythematosus: high disease activity predictors of response. Ann therapy in refractory systemic lupus erythematosus: results from the British isles lupus Rheum Dis 2012;71:1343–9. assessment group biologics register. Rheumatol 2018;57:470–9.

734 Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 Viewpoint

46 Dooley MA, Jayne D, Ginzler EM, et al. Mycophenolate versus azathioprine as 59 Aurinia Pharmaceuticals. Aurinia Pharmaceuticals announces voclosporin meets maintenance therapy for lupus nephritis. N Engl J Med 2011;365:1886–95. primary endpoint in phase IIb AURORA-LV study in lupus nephritis. 2016. Available: 47 Forbess LJ, Bresee C, Wallace DJ, et al. Failure of a systemic lupus erythematosus http://ir​ .​auriniapharma.com/​ ​press-releases/​ ​detail/49/​ ​aurinia-pharmaceuticals-​ ​ response index developed from clinical trial data: lessons examined and learned. announces-​voclosporin-​meets-​primary [Accessed 4 Oct 2018]. Lupus 2017;26:909–16. 60 FDA. Guidance for industry: providing clinical evidence of effectiveness for human 48 Abrahamowicz M, Esdaile JM, Ramsey-Goldman R, et al. Development and validation drug and biological products. 1998. Available: https://www.fda.​ ​gov/​downloads/​Drugs/​ of a novel evidence-based lupus multivariable outcome score for clinical trials. Guidance​ ​Complian​ ​ceRegula​ ​toryInfo​ ​rmation/Guidances/​ ​ucm072008.pdf​ [Accessed 4 Arthritis Rheumatol 2018;70:1450–8. Oct 2018]. 49 FDA. Guidance for industry: systemic lupus erythematosus — developing medical 61 Geh D, Gordon C. Epratuzumab for the treatment of systemic lupus erythematosus. products for treatment. 2010. Available: https://www.​fda.gov/​ ​downloads/Drugs/​ ​Guid​ Expert Rev Clin Immunol 2018;14:245–58. anceComp​ ​lianceRe​ ​gula​tory​Information/​ ​Guidances/ucm072063.​ ​pdf [Accessed 4 Oct 62 van Vollenhoven R, Voskuyl A, Bertsias G, et al. A framework for remission in SLE: 2018]. consensus findings from a large international task force on definitions of remission in 50 Wofsy D, Hillson JL, Diamond B. Abatacept for lupus nephritis: alternative SLE (DORIS). Ann Rheum Dis 2017;76:554–61. definitions of complete response support conflicting conclusions. Arthritis Rheum 63 Merrill JT, Neuwelt CM, Wallace DJ, et al. Efficacy and safety of rituximab in 2012;64:3660–5. moderately-to-severely active systemic lupus erythematosus: the randomized, double- 51 Wallace DJ, Stohl W, Furie RA, et al. A phase II, randomized, double-blind, placebo- blind, phase II/III systemic lupus erythematosus evaluation of rituximab trial. Arthritis controlled, dose-ranging study of belimumab in patients with active systemic lupus Rheum 2010;62:222–33. erythematosus. Arthritis Rheum 2009;61:1168–78. 64 Morand EF, Trasieva T, Berglind A, et al. Lupus Low Disease Activity State (LLDAS) 52 Merrill JT, Burgos-Vargas R, Westhovens R, et al. The efficacy and safety of abatacept attainment discriminates responders in a systemic lupus erythematosus trial: in patients with non-life-threatening manifestations of systemic lupus erythematosus: post-hoc analysis of the Phase IIb MUSE trial of anifrolumab. Ann Rheum Dis results of a twelve-month, multicenter, exploratory, phase IIb, randomized, double- 2018;77:706–13. blind, placebo-controlled trial. Arthritis Rheum 2010;62:3077–87. 65 Polachek A, Gladman DD, Su J, et al. Defining low disease activity in systemic lupus 53 Leuchten N, Hoyer A, Brinks R, et al. Performance of antinuclear antibodies for erythematosus. Arthritis Care Res 2017;69:997–1003. classifying systemic lupus erythematosus: a systematic literature review and meta- 66 ​ClinicalTrials.​gov. A study to evaluate the efficacy and safety of belimumab regression of diagnostic data. Arthritis Care Res 2018;70:428–38. administered in combination with rituximab to adult subjects With Systemic Lupus 54 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of Erythematosus (SLE) - BLISS-BELIEVE. 2017. Available: https://​clinicaltrials.​gov/​ct2/​ antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis show/​NCT03312907 [Accessed 16 Nov 2018]. 2018;365:annrheumdis-2017-212599–-2013. 67 Newswire PR. Lupus research institute launches industry council to accelerate new 55 Pisetsky DS. Anti-DNA antibodies--quintessential biomarkers of SLE. Nat Rev treatments for lupus. 2014. Available: https://www.​prnewswire.com/​ ​news/​Lupus+​ Rheumatol 2016;12:102–10. Research+​Institute [Accessed 4 Oct 2018]. 56 ​ClinicalTrials.​gov. Efficacy and safety of anifrolumab compared to placebo in adult 68 Lupus Foundation of America. LFA-REAL: Research. 2015. Available: http://www.​ subjects with active systemic lupus erythematosus (NCT02446899). 2018. Available: lupus.org/​ ​research/lfa-​ ​real [Accessed 4 Oct 2018]. https://clinicaltrials​ .​gov/​ct2/​show/​NCT02446899 [Accessed 4 Oct 2018]. 69 Innovative Medicines Initiative. IMI PRECISESADS. 2017. Available: http://www.​ 57 Thanou A, Merrill JT. New Trials in Lupus and where are we going. Curr Rheumatol precisesads.​eu/​imi/ [Accessed 4 Oct 2018]. Rep 2018;20:34. 70 Barturen G, Beretta L, Cervera R, et al. Moving towards a molecular taxonomy of 58 Little J, Parker B, Lunt M, et al. Glucocorticoid use and factors associated with autoimmune rheumatic diseases. Nat Rev Rheumatol 2018;14:75–93. variability in this use in the systemic lupus international collaborating clinics inception 71 MASTERPLANS Consortium. MASTERPLANS. 2016. Available: http://www.​ cohort. Rheumatol 2018;57:677–87. lupusmasterplans.​org/​home.​html [Accessed 4 Oct 2018].

Dall'Era M, et al. Ann Rheum Dis 2019;78:729–735. doi:10.1136/annrheumdis-2018-214530 735 Recommendation 2019 update of the EULAR recommendations for the management of systemic lupus erythematosus Antonis Fanouriakis,‍ ‍ 1 Myrto Kostopoulou,2 Alessia Alunno,‍ ‍ 3 Martin Aringer,4 Ingeborg Bajema,5 John N Boletis,6 Ricard Cervera,7 Andrea Doria,‍ ‍ 8 Caroline Gordon,9 Marcello Govoni,10 Frédéric Houssiau,11 David Jayne,12 Marios Kouloumas,13 Annegret Kuhn,14 Janni L Larsen,15 Kirsten Lerstrøm,16 Gabriella Moroni,17 Marta Mosca,18 Matthias Schneider,19 Josef S Smolen,20 Elisabet Svenungsson,21 Vladimir Tesar,22 Angela Tincani,23 Anne Troldborg,24 Ronald van Vollenhoven,25 Jörg Wenzel,26 George Bertsias,27 Dimitrios T Boumpas1,28,29

Handling editor David S Abstract the EULAR recommendations for lupus, capital- Pisetsky Our objective was to update the EULAR ising on the strengths of and experience from the recommendations for the management of systemic previous projects.6 ►► Additional material is published online only. To view lupus erythematosus (SLE), based on emerging new please visit the journal online evidence. We performed a systematic literature review Methods (http://dx.​ ​doi.org/​ ​10.​1136/​ (01/2007–12/2017), followed by modified Delphi After approval by the EULAR Executive annrheumdis-2019-​ ​215089). method, to form questions, elicit expert opinions and Committee, the convenor (DB) and methodolo- For numbered affiliations see reach consensus. Treatment in SLE aims at remission gist (GB) invited a Task Force to work on this end of article. or low disease activity and prevention of flares. update; two fellows (AF, MK) undertook the Hydroxychloroquine is recommended in all patients systematic literature review (SLR). The EULAR Correspondence to with lupus, at a dose not exceeding 5 mg/kg real standardised operating procedures7 and the Dr Antonis Fanouriakis, body weight. During chronic maintenance treatment, Appraisal of Guidelines Research and Evalu- Rheumatology and Clinical glucocorticoids (GC) should be minimised to less than ation instrument (AGREE II)8 were followed. Immunology Unit, “Attikon” University Hospital, Athens 7.5 mg/day (prednisone equivalent) and, when possible, Applying a Delphi-based methodology, 14 12462, Greece; withdrawn. Appropriate initiation of immunomodulatory research questions were selected for SLR (online afanour@​ ​med.uoa.​ ​gr agents (methotrexate, azathioprine, mycophenolate) supplementary table 1). PubMed was screened can expedite the tapering/discontinuation of GC. In using strings of relevant terms. Since this was an GB and DTB contributed equally. persistently active or flaring extrarenal disease, add-on update of the previous 2007 recommendations, Received 18 January 2019 belimumab should be considered; rituximab (RTX) may the SLR considered all English-language publi- Revised 7 March 2019 be considered in organ-threatening, refractory disease. cations from 01/2007 until 12/2017, with two Accepted 11 March 2019 Updated specific recommendations are also provided exceptions: (1) treatment of skin disease, where Published Online First for cutaneous, neuropsychiatric, haematological and an unrestricted date search was performed and 29 March 2019 renal disease. Patients with SLE should be assessed for (2) renal disease, where search was limited to their antiphospholipid antibody status, infectious and the period 01/2012–12/2017 (since the EULAR cardiovascular diseases risk profile and preventative recommendations for LN were published in strategies be tailored accordingly. The updated 2012). Pertinent articles, identified by manual recommendations provide physicians and patients with search within the reference list of the originally updated consensus guidance on the management of SLE, retrieved publications, were also included. All combining evidence-base and expert-opinion. retrieved items were refined based on article type, abstract, full-text content and number of included patients. The final level of evidence and grading of recommendations considered also the Introduction body of evidence that had informed the previous Systemic lupus erythematosus (SLE) has variable sets of EULAR recommendations for the manage- presentation, course and prognosis. The wide ment of SLE, as the convenor, methodologist and acceptance and popularity of the first EULAR several of the panellists had also participated in recommendations for its management, published the latter. A detailed presentation of the SLR in 2008,1 prompted the subsequent development results is given in online supplementary tables of specific recommendations regarding moni- 2 and 3. Evidence was categorised based on the toring, neuropsychiatric and renal disease, as well design and validity of available studies and the © Author(s) (or their 2–5 employer(s)) 2019. No as for pregnancy and women’s health in lupus. strength of the statements was graded (see online commercial re-use. See rights Since these publications, new data have emerged supplementary table 4). After rounds of discus- and permissions. Published on treatment strategies and validated goals of sions, the committee reached a consensus of 33 by BMJ. treatment, alternative regimens of glucocorticoids final statements, grouped in four broad cate- To cite: Fanouriakis A, (GC), ‘multitargeted’ therapy with the use of calci- gories (Goals of Treatment, Treatment of SLE, Kostopoulou M, Alunno A, neurin inhibitors (CNIs) in lupus nephritis (LN), Specific manifestations, Comorbidities—table 1). et al. Ann Rheum Dis and the approval of the first biological therapy Each Task Force member rated their agreement 2019;78:736–745. for SLE. These advances called for an update of with each statement.

736 Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 Recommendation

Table 1 Recommendations for the management of patients with systemic lupus erythematosus Overarching principles ►► SLE is a multisystem disease—occasionally limited to one or few organs—diagnosed on clinical grounds in the presence of characteristic serological abnormalities. ►► SLE care is multidisciplinary, based on a shared patient-physician decision, and should consider individual, medical and societal costs. ►► Treatment of organ-threatening/life-threatening SLE includes an initial period of high-intensity immunosuppressive therapy to control disease activity, followed by a longer period of less intensive therapy to consolidate response and prevent relapses. ►► Treatment goals include long-term patient survival, prevention of organ damage and optimisation of health-related quality of life.

Level of agreement, Recommendation/Statement mean (SD) 1. Goals of treatment 1.1 Treatment in SLE should aim at remission or low disease activity (2b/B) and prevention of flares (2b/B) in all organs, maintained with the lowest possible dose 10.0 (0) of glucocorticoids. 1.2 Flares of SLE can be treated according to the severity of organ(s) involvement by adjusting ongoing therapies (glucocorticoids, immunomodulating agents) to 9.95 (0.22) higher doses, switching or adding new therapies (2b/C). 2. Treatment of SLE 2.1 HCQ 2.1.1 HCQ is recommended for all patients with SLE (1b/A), unless contraindicated, at a dose not exceeding 5 mg/kg/real BW (3b/C). 9.65 (1.11) 2.1.2 In the absence of risk factors for retinal toxicity, ophthalmological screening (by visual fields examination and/or spectral domain-optical coherence 9.75 (0.70) tomography) should be performed at baseline, after 5 years, and yearly thereafter (2b/B). 2.2 GC 2.2.1 GC can be used at doses and route of administration that depend on the type and severity of organ involvement (2b/C). 9.95 (0.22) 2.2.2 Pulses of intravenous methylprednisolone (usually 250–1000 mg per day, for 1–3 days) provide immediate therapeutic effect and enable the use of lower 9.85 (0.36) starting dose of oral GC (3b/C). 2.2.3 For chronic maintenance treatment, GC should be minimised to less than 7.5 mg/day (prednisone equivalent) (1b/B) and, when possible, withdrawn. 9.65 (0.65) 2.2.4 Prompt initiation of immunomodulatory agents can expedite the tapering/discontinuation of GC (2b/B). 9.90 (0.30) 2.3 Immunosuppressive therapies 2.3.1 In patients not responding to HCQ (alone or in combination with GC) or patients unable to reduce GC below doses acceptable for chronic use, addition of 9.85 (0.48) immunomodulating/immunosuppressive agents such as methotrexate, (1b/B) azathioprine (2b/C) or mycophenolate (2a/B) should be considered. 2.3.2 Immunomodulating/immunosuppressive agents can be included in the initial therapy in cases of organ-threatening disease (2b/C). 9.85 (0.48) 2.3.3 Cyclophosphamide can be used for severe organ-threatening or life-threatening SLE as well as ‘rescue’ therapy in patients not responding to other 9.90 (0.30) immunosuppressive agents (2b/C). 2.4 Biologics 2.4.1 In patients with inadequate response to standard-of-care (combinations of HCQ and GC with or without immunosuppressive agents), defined as residual 9.20 (0.81) disease activity not allowing tapering of glucocorticoids and/or frequent relapses, add-on treatment with belimumab should be considered (1a/A). 2.4.2 In organ-threatening disease refractory or with intolerance/contraindications to standard immunosuppressive agents, rituximab can be considered (2b/C). 9.85 (0.48) 3 Specific manifestations 3.1 Skin disease 3.1.1 First-line treatment of skin disease in SLE includes topical agents (GC, calcineurin inhibitors) (2b/B), antimalarials (HCQ, quinacrine) (1a/A) and/or systemic 10.0 (0) GC (4/C). 3.1.2 In non-responsive cases or cases requiring high-dose GC, methotrexate (3a/B), retinoids (4/C), dapsone (4/C) or mycophenolate (4/C) can be added. 9.85 (0.48) 3.2 Neuropsychiatric disease 3.2.1 Attribution to SLE—as opposed to non-SLE—related neuropsychiatric manifestations, is essential and can be facilitated by neuroimaging, investigation 9.65 (0.85) of cerebrospinal fluid, consideration of risk factors (type and timing of the manifestation in relation to the onset of lupus, patient age, non-neurological lupus activity, presence of aPL) and exclusion of confounding factors (2b/C). 3.2.2 Treatment of SLE-related neuropsychiatric disease includes glucocorticoids/immunosuppressive agents for manifestations considered to reflect an 9.85 (0.48) inflammatory process (1b/A), and antiplatelet/anticoagulants for atherothrombotic/aPL-related manifestations (2b/C). 3.3 Haematological disease 3.3.1 Acute treatment of lupus thrombocytopenia includes high-dose GC (including pulses of intravenous methylprednisolone) (4/C) and/or intravenous 9.95 (0.22) immunoglobulin G (4/C). 3.3.2 For maintenance of response, immunosuppressive/GC-sparing agents such as mycophenolate (2b/C), azathioprine (2b/C) or cyclosporine (4/C) can be used. 9.75 (0.62) 3.3.3 Refractory cases can be treated with rituximab (3a/C) or cyclophosphamide (4/C). 9.65 (0.73) 3.4 Renal disease 3.4.1 Early recognition of signs of renal involvement and—when present—performance of a diagnostic renal biopsy are essential to ensure optimal outcomes 9.95 (0.22) (2b/B). 3.4.2 Mycophenolate (1a/A) or low-dose intravenous cyclophosphamide (2a/B) are recommended as initial (induction) treatment, as they have the best efficacy/ 9.85 (0.36) toxicity ratio. 3.4.3 In patients at high risk for renal failure (reduced glomerular filtration rate, histological presence of fibrous crescents or fibrinoid necrosis, or tubular atrophy/ 9.45 (0.80) interstitial fibrosis], similar regimens may be considered but high-dose intravenous cyclophosphamide can also be used (1b/A). 3.4.4 For maintenance therapy, mycophenolate (1a/A) or azathioprine (1a/A) should be used. 9.75 (0.62) Continued

Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 737 Recommendation

Table 1 Continued

Level of agreement, Recommendation/Statement mean (SD) 3.4.5 In cases with stable/improved renal function but incomplete renal response (persistent proteinuria >0.8–1 g/24 hours after at least 1 year of 9.85 (0.48) immunosuppressive treatment), repeat biopsy can distinguish chronic from active kidney lesions (4/C). 3.4.6 Mycophenolate may be combined with low dose of a calcineurin inhibitor in severe nephrotic syndrome (2b/C) or incomplete renal response (4/C), in the 9.50 (0.81) absence of uncontrolled hypertension, high chronicity index at kidney biopsy and/or reduced GFR. 4 Comorbidities 4.1 Antiphospholipid syndrome 4.1.1 All patients with SLE should be screened at diagnosis for aPL (1a/A). 10.0 (0) 4.1.2 Patients with SLE with high-risk aPL profile (persistently positive medium/high titres or multiple positivity) may receive primary prophylaxis with antiplatelet 9.45 (0.80) agents (2a/C), especially if other atherosclerotic/thrombophilic factors are present, after balancing the bleeding hazard. 4.1.3 For secondary prevention (thrombosis, pregnancy complication/loss), the therapeutic approach should be the same as for primary antiphospholipid 10.0 (0) syndrome (1b/B). 4.2 Infectious diseases 4.2.1 Patients with SLE should be assessed for general and disease-related risk factors for infections, such as advanced age/frailty (–/D), diabetes mellitus (–/D), 9.85 (0.65) renal involvement (2b/B), immunosuppressive/biological therapy (1b-2b/B-C) and use of GC (1a/A). 4.2.2 General preventative measures (including immunisations) and early recognition and treatment of infection/sepsis are recommended (–/D). 9.90 (0.44) 4.3 Cardiovascular disease 4.3.1 Patients with SLE should undergo regular assessment for traditional (1b/B-C) and disease-related risk factors for cardiovascular disease, including 9.85 (0.65) persistently active disease (1b/B), increased disease duration (1b/A), medium/high titres of aPL (1b/A), renal involvement (1b/B) (especially, persistent proteinuria and/or GFR <60 mL/min) and chronic use of GC (1b/B). 4.3.2 Based on their individual cardiovascular risk profile, patients with SLE may be candidates for preventative strategies as in the general population, including 9.85 (0.48) low-dose aspirin (2b/D) and/or lipid-lowering agents (2b/D). aPL, antiphospholipid antibodies; GC, glucocorticoids; GFR, glomerular filtration rate; HCQ, hydroxychloroquine; SLE, systemic lupus erythematosus.

Results and discussion In monitoring renal response, reduction of UPr (to less than Overarching principles 0.8 g/day) following treatment is more important than residual SLE represents a challenge for the treating physician in terms of haematuria.21 Patients with more severe proteinuria and longer- diagnosis and treatment. Its protean manifestations, often multi- standing disease are less likely to respond or show more delayed system but occasionally limited to a few or single organ, have led responses.22 23 some physicians to focus exclusively on evidence of serological Prevention of disease flares is an additional milestone of SLE autoimmunity (antinuclear and more specific autoantibodies), treatment. Although a universally accepted definition is lacking, for a disease where diagnosis is clinical after excluding competing most experts agree that a flare is a measurable increase in disease diagnoses. Monitoring of SLE through validated disease activity activity usually leading to change of treatment.24 Flares are and chronicity indices, including physician global assessment common in the disease course and contribute significantly to (PGA), is recommended. For patients with severe disease, multi- organ damage accrual and worse outcome.17 25 26 Consistently disciplinary care in dedicated lupus centres is desirable.9 Immu- reported risk factors for a higher disease flare rate include nosuppressive (IS) therapy (for induction and maintenance of younger age at disease onset, no use of antimalarials, persistent remission) is indicated in organ-threatening lupus. generalised disease activity and serological activity (anti-dsDNA, low complement).27–31 Assessment of adherence to drug treat- Recommendations ment, close monitoring and optimisation of disease control in Goals of treatment these patients may reduce the risk for a flare. To improve long-term patient outcomes, management should aim at remission of disease symptoms and signs, prevention of Treatment of SLE damage accrual and minimisation of drug side-effects, as well as Hydroxychloroquine improvement of quality of life.10 11 Complete remission (absence Hydroxychloroquine (HCQ) is recommended for all patients of clinical activity with no use of GC and IS drugs) is infre- with SLE. There is evidence for multiple beneficial effects of HCQ quent.12–16 To this end, newly defined low disease activity states in SLE,32 yet poor adherence to treatment is not uncommon.33–35 (based on a SLEDAI score ≤3 on antimalarials, or alternatively Drug blood levels can be used to assess compliance,33 35 but data SLEDAI ≤4, PGA≤1 with GC ≤7.5 mg of prednisone and well are currently insufficient to recommend routine monitoring of tolerated IS agents) have shown comparable rates with remission, drug levels. Concerns for retinal toxicity with long-term HCQ regarding halting of damage accrual (OR 0.5–0.7 for increase therapy led to the use of more sensitive screening techniques, in damage index) and prevention of flares.14 17–20 Accordingly, with a prevalence of retinal abnormalities exceeding 10% after treatment in SLE should aim at remission or, if this state cannot 20 years of continuous use.36 37 Major risk factors for retinop- be achieved, at low disease activity in all organ systems. In LN, athy include duration of treatment (OR 4.71 for every 5 years therapy should aim at least partial remission (defined as ≥50% of use), dose (OR 3.34 for every 100 mg daily dose), chronic reduction in proteinuria [UPr] to subnephrotic levels and serum kidney disease (adjusted OR 8.56) and pre-existing retinal or creatinine [SCr] within 10% from baseline) by 6–12 months; macular disease.37 Based on existing evidence suggesting that the complete renal remission (proteinuria <500 mg/24 hours and risk of toxicity is very low for doses below 5 mg/kg real body SCr within 10% from baseline), however, may require longer weight, the daily dose should not exceed this threshold. Of treatment duration, often more than 12 and until 24 months. note, efficacy of HCQ in lupus has been established in studies

738 Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 Recommendation with a prescribed dose of 6.5 mg/kg/day, thus it remains to be in extrarenal disease with inadequate control (ongoing disease confirmed whether a lower dose will have comparable clin- activity or frequent flares) to first-line treatments (typically ical effects. Patients in long-standing remission may have their including combination of HCQ and prednisone with or without dose lowered, although no studies have formally addressed this IS agents), and inability to taper GC daily dose to acceptable strategy. The choice of quinacrine, an alternative antimalarial, levels (ie, maximum 7.5 mg/day). Patients with persistent disease can be considered in patients with cutaneous manifestations and may benefit from belimumab; more likely to respond are patients HCQ-induced retinal toxicity. with high disease activity (eg, SLEDAI >10), prednisone dose >7.5 mg/day and serological activity (low C3/C4, high anti- Glucocorticoids dsDNA titres), with cutaneous, musculoskeletal and serological 67–69 GC can provide rapid symptom relief, but the medium to long- manifestations responding the most. term aim should be to minimise daily dose to ≤7.5 mg/day Due to the negative results of randomised controlled trials prednisone equivalent or to discontinue them, because long- (RCTs), RTX is currently only used off-label, in patients with term GC therapy can have various detrimental effects including severe renal or extrarenal (mainly haematological and neuropsy- irreversible organ damage.38–41 Risks are substantially increased chiatric) disease refractory to other IS agents and/or belimumab, at continuous GC doses above 7.5 mg/day, with some studies or in patients with contraindications to these drugs. As a general suggesting that also lower doses might be harmful.17 42–44 To rule, more than one IS drug need to have failed prior to RTX 70–73 this end, two approaches can be considered: (1) use of pulses administration, except perhaps for cases of severe auto- of intravenous methylprednisolone (MP) of various doses immune thrombocytopaenia and haemolytic anaemia, where (depending on severity and body weight), which take advantage RTX has demonstrated efficacy both in lupus and in patients 74–76 of the rapid non-genomic effects of GC45 and may allow for a with isolated immune thrombocytopaenia (ITP). In LN, lower starting dose and faster tapering of PO GC,46 47 and (2) RTX is typically considered following failure of first-line ther- 70 77 early initiation of IS agents, to facilitate tapering and eventual apies (CYC, MMF) or in relapsing disease. More recently, a discontinuation of oral GC (see below). High-dose intravenous posthoc analysis of the LUNAR trial showed that complete B-cell depletion following RTX treatment in LN was associated with MP (usually 250–1000 mg/day for 3 days) is often used in acute, 78 organ-threatening disease (eg, renal, neuropsychiatric) after higher odds for complete response at 78 weeks. excluding infections.48 Figure 1 summarises the various drugs used in the treatment of SLE, according to disease severity stratification. Online supple- mentary table 5 outlines the recommended doses of the drugs Immunosuppressive (IS) drugs mentioned in the manuscript. Consequent initiation of IS drugs facilitates a more rapid GC tapering and may prevent disease flares.49 The choice of agent depends on prevailing disease manifestation(s), patient age Specific manifestations and childbearing potential, safety concerns and cost. Metho- Skin disease trexate (MTX) and azathioprine (AZA) should be considered in A large body of evidence originates from studies in patients patients with poor symptom control after a trial with GC and with cutaneous lupus erythematosus (CLE). Effective protection HCQ or when HCQ alone is unlikely to be sufficient, due to from ultraviolet exposure with broad-spectrum sunscreens and the large experience gained with their use and their relatively 79–81 50 smoking cessation are strongly recommended. In atypical or safe profile. Published evidence is generally stronger for MTX refractory cases, a diagnostic skin biopsy should be considered. than AZA, yet the latter is compatible with pregnancy contem- First-line treatment of skin disease includes topical agents (GC plation. Mycophenolate mofetil (MMF) is a potent immunosup- and/or CNIs) and antimalarials, with or without systemic GC (the pressant with efficacy in renal and non-renal lupus (although latter at a starting dose depending on severity of skin involve- 51–53 not in neuropsychiatric disease). In a recent randomised, ment).82 83 HCQ is the antimalarial of choice over chloroquine open-label trial in extrarenal SLE, enteric-coated mycophenolate due to its multiple beneficial effects and possibly lower risk for sodium (EC-MPS) was superior to AZA in achieving remission retinal toxicity;84 in cases of inadequate response or evidence 54 and reducing flares. However, its teratogenic potential (needs of toxic retinopathy, quinacrine (mepacrine) may be used as an to be discontinued at least 6 weeks before conceiving), along add-on or sequential therapy, respectively.85–87 Although quin- with its higher cost compared with AZA or MTX, poses a limita- acrine is currently unavailable in several countries worldwide, tion towards universal recommendation in women of repro- it is a useful alternative when available. There are no studies ductive age with non-renal manifestations. Cyclophosphamide examining retinal toxicity of quinacrine with the newer, more (CYC) can be considered in organ-threatening disease (especially sensitive screening techniques (visual fields or optical coherence renal, cardiopulmonary or neuropsychiatric) and only as rescue tomography); however with current knowledge, retinopathy is therapy in refractory non-major organ manifestations; due to its not considered a side-effect of quinacrine. gonadotoxic effects, it should be used with caution in women A sizeable proportion (almost 40%) of patients will not 55–57 and men of fertile age. Concomitant use of GnRH analogues respond to first-line treatment.86 88 In such cases, MTX can be attenuates the depletion of ovarian reserve associated with CYC added.50 89 Other agents include retinoids, dapsone and MMF therapy and is recommended in premenopausal patients with or EC-mycophenolic acid.79 90 91 Belimumab and RTX have 4 58 59 SLE. Information about the possibility of ovarian cryopres- also shown efficacy in mucocutaneous manifestations of SLE, ervation should be offered ahead of treatment. Other risks of although these studies have not included a validated activity CYC therapy such as malignancy and infections should also be score for skin lesions; RTX may be less efficacious in chronic 60 61 considered. forms of skin lupus.62 92–94 Thalidomide is effective in various subtypes of cutaneous disease.95 96 Due to its strict contraindi- Biological agents cation in pregnancy, the risk for irreversible polyneuropathy, There is evidence to support beneficial effects of B-cell and the frequent relapses on drug discontinuation, it should be targeting agents in SLE.62–66 Belimumab should be considered considered only as a ‘rescue’ therapy in patients who have failed

Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 739 Recommendation

Figure 1 Treatment of non-renal SLE—recommended drugs with respective grading of recommendation. aPL, antiphospholipid antibodies; AZA, azathioprine; BEL, belimumab; BILAG: British Isles Lupus Assessment Group disease activity index; CNIs, calcineurin inhibitors; CYC, cyclophosphamide; GC, glucocorticoids; HCQ, hydroxychloroquine; IM, intramuscular; MMF, mycophenolate mofetil; MTX, methotrexate; Pre, prednisone; PO, per os; RTX, rituximab; PLTs: Platelets; SLEDAI, Systemic Lupus Erythematosus Disease Activity Index. multiple previous agents. A treatment algorithm for the various events.108 In this context, neuroimaging and/or CSF studies may subtypes of CLE has been published by a European group of provide additional supporting evidence for IS therapy. Targeted Dermatologists guided by the European Dermatology Forum in symptomatic therapy is indicated according to the type of mani- cooperation with the European Academy of Dermatology and festation (eg, antipsychotics for psychosis, anxiolytics for anxiety Venereology.79 disorder and so on).

Neuropsychiatric disease (NPSLE) Haematological disease Attribution of neuropsychiatric manifestations to SLE often Haematological manifestations frequently necessitating anti-in- requires a comprehensive, multidisciplinary approach to rule flammatory/IS treatment in patients with SLE include throm- out mimics (infections, malignancy and others), taking into bocytopaenia and autoimmune haemolytic anaemia (AIHA). account the presence of risk (‘favouring’) factors (type and First-line treatment of significant lupus thrombocytopaenia timing of manifestation, presence of generalised, non-neurolog- (platelet count below 30 000/mm3) consists of moderate/high ical disease activity, abnormal neuroimaging and cerebrospinal doses of GC in combination with IS agent (AZA, MMF or cyclo- fluid analysis, positive antiphospholipid antibodies [aPL]),97 as sporine; the latter having the least potential for myelotoxicity) well as confounding factors favouring alternative diagnoses.98 to facilitate GC-sparing. Initial therapy with pulses of intrave- The use of validated attribution models may aid in the diagnostic nous MP (1–3 days) is encouraged. Intravenous immunoglob- process.99 100 ulin (IVIG) may be considered in the acute phase, in cases of Treatment of NPSLE depends on whether the underlying inadequate response to high-dose GC or to avoid GC-related pathophysiological mechanism is presumed to be inflamma- infectious complications. tory or embolic/thrombotic/ischaemic.2 101 GC and/or IS agents Treatment of thrombocytopenia is typically lengthy and often should be considered in the former, while anticoagulant/anti- characterised by relapses during GC tapering.109 In patients with thrombotic treatment is favoured when aPL antibodies are no response to GC (ie, failure to reach a platelet count >50 000/ present.102–107 Distinction between the two pathophysiolog- mm3) or relapses, RTX should be considered, considering also its ical processes may not be easy in clinical practice, or the two efficacy in ITP.74 76 110 CYC may also be considered in such cases. processes may coexist in the same patient.2 Combination of IS Thrombopoietin agonists or splenectomy should be reserved as and anticoagulant/antithrombotic therapy may be considered in last options.111 112 Autoimmune haemolytic anaemia (AIHA) is these patients. Patients with SLE with cerebrovascular disease far less common than thrombocytopenia in SLE; its treatment should be managed like the general population in the acute follows the same principles regarding use of GC, IS drugs and phase; in addition to controlling extra-CNS lupus activity, IS RTX. Autoimmune leucopaenia is common in SLE but rarely therapy may be considered in the absence of aPL antibodies and needs treatment; careful work-up is recommended to exclude other atherosclerotic risk factors or in recurrent cerebrovascular other causes of leucopaenia (especially drug-induced).

740 Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 Recommendation

Renal disease warfarin in APS with triple aPL positivity (~21% of patients had Patients at high risk of developing renal involvement (males, SLE-APS) was prematurely terminated due to an excess of throm- juvenile lupus onset, serologically active including positivity for boembolic events in the rivaroxaban arm.139 Thus, in patients anti-C1q antibodies)113–115 should be under vigilant monitoring with SLE-APS, use of novel oral anticoagulants for secondary (eg, at least every 3 months) to detect early signs of kidney disease. prevention should be avoided; however, they could potentially Following diagnosis, secured with a kidney biopsy, treatment serve as an alternative option in selected patients (low-risk aPL of LN includes an initial induction phase, followed by a more profile, no history of arterial thrombotic events) with difficult prolonged maintenance phase. MMF and CYC are the IS agents to control international normalised ratio on warfarin, after of choice for induction treatment; low-dose CYC (Euro-Lupus balancing possible risks. regimen, online supplementary table 5) is preferred over high- dose CYC as it has comparable efficacy and lower risk of gonad- Infections 57 116 117 otoxicity. Published data support the use of MMF and Risk of infection in SLE is associated with both disease-related high-dose CYC (online supplementary table 5) in severe forms of and treatment-related factors; high-dose GC therapy, CYC, LN associated with increased risk of progression into end-stage MMF and RTX are all associated with an increased risk for renal disease (reduced glomerular filtration rate, histological infection, while high disease activity, severe leucopaenia and presence of fibrous crescents or fibrinoid necrosis, or tubular presence of renal involvement (±hypogammaglobulinaemia in 118 119 atrophy/interstitial fibrosis). An early significant drop in nephrotic syndrome) also contribute independently.48 140–143 UPr (to ≤1 g/day at 6 months or ≤0.8 g/day at 12 months) is a Protection against infections should be proactive, focusing both 21 117 120 predictor of favourable long-term renal outcome. MMF on primary prevention, as well as timely recognition and treat- or AZA may be used as maintenance therapy, with the former ment. Patients with lupus should receive vaccinations according 121 122 associated with fewer relapses; the choice depends on the to the EULAR recommendations for vaccination of patients with agent used for induction phase and on patient characteristics, autoimmune rheumatic diseases.144 145 Immunisation against including age, race and wish for pregnancy. In refractory or seasonal influenza and pneumococcal infection (both PCV13 relapsing disease, RTX may be considered. and PPSV23) should be strongly considered, preferably during Following the EULAR recommendations for LN in 2012, stable disease. Herpes zoster vaccination is now available for the several studies have been published regarding the use of CNIs general population, but a study in SLE has not been performed. to treat proliferative LN, either alone or in the form of a ‘multi- 123–127 Prompt diagnosis and treatment of sepsis is essential. To this end, target therapy’ (combination of tacrolimus with MMF). validated scores such as the quick SOFA ([systolic blood pressure These studies were performed almost exclusively in Asian popu- ≤100 mm Hg, respiratory rate ≥22/min, altered mental state lations and had short follow-up; hence, data have to be corrob- with Glasgow coma scale <15]: the presence of ≥2 points near orated with longer duration studies in multiethnic populations. the onset of infection is associated with a greater risk of death or To this end, at present, CNIs may be considered as second-line prolonged intensive care unit stay] may identify patients who are agents for induction or maintenance therapy mainly in membra- at greater risk for a poor outcome.146 nous LN, podocytopathy, or in proliferative disease with refrac- tory nephrotic syndrome, despite standard-of-care within 3–6 months;128 129 in the latter case, they may be used alone or in Cardiovascular disease SLE is an independent risk factor for cardiovascular disease combination with MMF, since small, observational studies have shown the CNI/MMF combination to be effective in disease (CVD), due to both traditional and disease-related risk factors, 130–132 such as persistent disease activity, LN, presence of aPL and use of refractory to standard therapy. Monitoring SCr and blood 147–149 levels of CNI to avoid chronic drug toxicity is essential. GC. Surrogate measures of atherosclerosis, such as carotid plaques, carotid intima media thickness (cIMT) and coronary artery calcium are frequently used to identify subclinical CVD in 150 Comorbidities SLE. Low-dose aspirin may be considered for primary preven- Antiphospholipid antibodies (aPL) and antiphospholipid syndrome tion of CVD, as it may reduce the risk for incident CVD in SLE 151 152 (APS) (HR 0.24 in one retrospective study). However, this has to The presence of aPL is associated with thrombotic and obstetric be viewed in light of recent large studies in diabetics and elderly complications and increased risk of damage accrual.133 134 In aPL showing that the benefits of aspirin for primary CVD prevention 136 153 carriers, a recent meta-analysis supported a protective role of are counterbalanced by the larger bleeding hazard. The low-dose aspirin for primary prophylaxis against thrombosis in value of statins in SLE has been tested in RCTs, which failed the subgroup of aPL carriers who had SLE;135 however, in view to show a clear benefit over placebo, when cIMT was used a 154 155 of the potential bleeding hazard,136 137 it is not clear whether this surrogate marker for CVD. Thus, routine use of statins is should be applied to patients with lupus with any aPL antibodies not recommended for all patients but should be considered on or only to those carrying a high risk aPL profile (ie, triple aPL the basis of lipid levels and the presence of other traditional risk positivity, lupus anticoagulant or high titres of anticardiolipin factors. Calculation of the 10-year CVD risk using the System- antibodies).138 Patients with SLE with aPL may also receive addi- atic Coronary Risk Evaluation (SCORE, https://www.​escardio.​ tional anticoagulant treatment, such as low-molecular weight org/​Education/​Practice-​Tools/​CVD-​prevention-​toolbox/​SCORE-​ heparin, during high-risk periods for thrombosis (pregnancy or Risk-​Charts) is recommended,156 although the actual risk is postoperatively), although no studies have formally addressed underestimated in patients with SLE. this question. Certain points to consider and the research agenda suggested No studies have been performed exclusively on patients with by the Task Force Members are reported in box 1. These points SLE-APS, with several studies excluding secondary APS due to aim to improve the design of clinical studies in order to answer lupus. Thus, with current knowledge, treatment of APS in the clinically important questions, for which the current ‘state-of- context of SLE should not differ from treatment of primary APS. the-art’ is insufficient. In particular, data regarding the optimal A recent randomised, open-label trial comparing rivaroxaban to duration and timing of discontinuation of therapy in both renal

Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 741 Recommendation

13Cyprus League Against Rheumatism, Aglantzia, Cyprus Box 1 Future research agenda in SLE 14University Hospital Muenster, Muenster, Germany 15Copenhagen Lupus and Vasculitis Clinic, Rheumatology and Spine Diseases Centre, Targets of therapy Rigshospitalet, Copenhagen, Copenhagen, Denmark 16Lupus Europe, Farum, Denmark ►► Exploration of a universally accepted level of residual disease 17Nephrology Unit, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, activity, if remission cannot be achieved. Milan, Italy Existing therapies and disease monitoring 18Rheumatology Unit, Department of Clinical and Experimental Medicine, University ►► Efficacy of calcineurin inhibitor-containing treatment of Pisa, Pisa, Italy regimens in lupus nephritis (LN) in different racial/ethnic 19Department of Rheumatology & Hiller Research Unit Rheumatology, UKD, Heinrich- groups and at longer time points. Heine University, Düsseldorf, Germany 20Division of Rheumatology, Department of Medicine 3, Medical University of Vienna, ►► Usefulness of measurements of drug blood levels Vienna, Austria (hydroxychloroquine [HCQ], mycophenolate mofetil and so 21Department of Medicine, Rheumatology Unit, Karolinska Institutet and Karolinska on). University Hospital, Stockholm, Sweden 22 ►► Efficacy of quinacrine as immunomodulator in patients with Department of Nephrology, 1st Faculty of Medicine and General University HCQ-induced retinal toxicity. Hospital, Charles University, Prague, Czech Republic 23Rheumatology and Clinical Immunology, University of Brescia, Brescia, Italy ►► Comparative trials of conventional immunosuppressive drugs 24Department of Rheumatology, Aarhus University Hospital, Aarhus, Denmark with global and organ-specific result reporting. 25Department of Rheumatology and Clinical Immunology, Amsterdam University ►► Randomised trials testing lower cumulative dose Medical Centers, Amsterdam, Netherlands 26 glucocorticoid regimens versus conventional regimens. Department of Dermatology and Allergy, University Hospital Bonn, Bonn, Germany 27Rheumatology, Clinical Immunology and Allergy, University Hospital of Heraklion, ►► Optimal treatment regimen of rituximab: regular versus Heraklion, Greece on-demand. 28Laboratory of Autoimmunity and Inflammation, Biomedical Research Foundation of ►► Optimal duration of therapy and timing of discontinuation the Academy of Athens, Athens, Greece (renal and extrarenal disease). 29Joint Academic Rheumatology Program, Medical School, National and Kapodestrian ►► Value of repeat kidney biopsy for monitoring LN and University of Athens, Athens, Greece and Medical School, University of Cyprus, determination of clinical versus histological response to Nicosia, Cyprus therapy. Acknowledgements The committee wishes to acknowledge the support of the Pathophysiology and Biomarkers EULAR Standing Committee on Clinical Affairs. The committee also expresses its sincere appreciation and gratitude to the EULAR Secretariat and especially to Patrizia ►► Susceptibility to develop systemic lupus erythematosus (SLE). Jud, executive assistant, for the outstanding organisation. ►► Involvement of particular organ systems over others, Contributors AF and MK performed the systematic literature review (SLR) and AF multisystem versus organ-dominant disease. drafted the manuscript. GB supervised the methodology of the SLR and edited the ►► Response to specific therapeutic agents over others manuscript. DTB convened and supervised the project and edited the manuscript. All (pharmacogenetics, transcriptomics and so on). authors edited the manuscript and accepted its final form. Clinical trial design and new drug development Funding AF was supported by an Articulum Fellowship and a grant from the ►► Optimisation of clinical trial design and study endpoints to Hellenic Society of Rheumatology during the completion of this work. maximise probability of new drug approval in SLE. Competing interests AF reports personal fees from GSK, Abbvie, Amgen, Enorasis ►► Handling of background medication to avoid polypharmacy and Genesis Pharma, outside the submitted work. MA reports fees from advisory and ‘dilution’ of positive effects of drugs under study. boards from Novartis, Pfizer, Roche. IB reports personal fees from consultant for GSK, outside the submitted work. JNB reports grants from GSK, personal fees from GSK, ►► Inclusion of organ-specific endpoints and disease activity personal fees from Abbvie, personal fees from UCB, personal fees from Enorasis, measures. grants from Pfizer, outside the submitted work. RC reports personal fees from GSK, ►► Increase in number of adequately trained trial sites personal fees from Astra Zeneca, personal fees from Rubió, outside the submitted (recruitment, infrastructure and training). work. DJ reports personal fees from Astra-Zeneca, Aurinia, Boehringer-Ingeleheim, ►► Academia versus industry-driven clinical trials. Celgene, BMS, Chemocentryx, grants and personal fees from GSK, from null, outside the submitted work. AK reports grants from Biogen, grants from Galderma, grants from GlaxoSmithKline, grants from Leo Pharma, personal fees from La Roche Posay, outside the submitted work. MM reports personal fees from GSK, Lilly and UCB. 157 and extrarenal disease are scarce; for the former, recent studies MS reports grants from GSK, UCB, Abbvie, outside the submitted work. JSS reports support the value of a repeat kidney biopsy for the management grants from AbbVie, Astra-Zeneca, Janssen, Lilly, MSD, Novartis, Pfizer and Roche, of maintenance therapy, but more data are needed.158 159 and personal fees from AbbVie, Amgen, Astra-Zeneca, Astro, BMS, Celgene, Celltrion, Chugai, Gilead, ILTOO, Janssen, Lilly, MSD, Novartis-Sandoz, Pfizer, Roche, Samsung, Sanofi and UCB, during the conduct of the study. AT reports personal fees from Author affiliations UCB, Pfizer, Abbvie, BMS, Sanofi, Roche, GSK, Alpha Sigma, Lilly, Jannsen, Cellgene 1Rheumatology and Clinical Immunology Unit, “Attikon” University Hospital, Athens, and Novartis, outside the submitted work. RvV reports grants from BMS, GSK, Lilly, Greece Pfizer, UCB Pharma, personal fees from AbbVie, AstraZeneca, Biotest, Celgene, GSK, 2Department of Nephrology, “G. Gennimatas” General Hospital, Athens, Greece Janssen, Lilly, Novartis, Pfizer, Servier, UCB, outside the submitted work. JW reports 3Rheumatology Unit, Department of Medicine, University of Perugia, Perugia, Italy grants from GSK, grants from Incyte, personal fees from Biogen, personal fees from 4Division of Rheumatology, Department of Medicine III, University Medical Center Leo, other from Novartis, during the conduct of the study. GB reports grants from & Faculty of Medicine Carl Gustav Carus, Technical University of Dresden, Dresden, GSK, Pfizer and personal fees from GSK, Abbvie, UCB and Enorasis, outside the Germany 5 submitted work. DTB reports unrestricted grant support/advisory board fees from Department of Pathology, Leiden University Medical Center, Leiden, Netherlands 6 Abbvie, BMS, Celgene, Enorasis, GSK, Pfizer, Novartis, UCB, Lilly, all deposited to the Nephrology Department and Renal Transplantation Unit, “Laikon” Hospital, National research account of the National and Kapodistrian University of Athens. and Kapodistrian University of Athens, Medical School, Athens, Greece 7Department of Autoimmune Diseases, Hospital Clinic, Barcelona, Spain Patient consent for publication Not required. 8Rheumatology Unit, Department of Medicine, University of Padova, Padova, Italy Provenance and peer review Not commissioned; externally peer reviewed. 9Rheumatology Research Group, Institute of Inflammation and Ageing, University of Birmingham, Birmingham, UK 10Department of Medical Sciences, Section of Rheumatology, University of Ferrara, References Azienda Ospedaliero-Universitaria Sant’Anna Ferrara, Ferrara, Italy 1 Bertsias G, Ioannidis JPA, Boletis J, et al. EULAR recommendations for the 11Cliniques Universitaires Saint-Luc, Université catholique de Louvain, Brussels, management of systemic lupus erythematosus. Report of a task Force of the EULAR Belgium standing Committee for international clinical studies including therapeutics. Ann 12Department of Medicine, Addenbrooke’s Hospital, Cambridge, UK Rheum Dis 2008;67:195–205.

742 Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 Recommendation

2 Bertsias GK, Ioannidis JPA, Aringer M, et al. EULAR recommendations for the 30 Steiman AJ, Gladman DD, Ibañez D, et al. Outcomes in patients with systemic lupus management of systemic lupus erythematosus with neuropsychiatric manifestations: erythematosus with and without a prolonged serologically active clinically quiescent report of a task Force of the EULAR standing Committee for clinical Affairs. Ann period. Arthritis Care Res 2012;64:511–8. Rheum Dis 2010;69:2074–82. 31 Weiss JE, Sison CP, Ilowite NT, et al. Flares in pediatric systemic lupus erythematosus. 3 Bertsias GK, Tektonidou M, Amoura Z, et al. Joint European League against J Rheumatol 2007;34:1341–4. rheumatism and European renal Association-European dialysis and Transplant 32 Ruiz-Irastorza G, Ramos-Casals M, Brito-Zeron P, et al. Clinical efficacy and side Association (EULAR/ERA-EDTA) recommendations for the management of adult and effects of antimalarials in systemic lupus erythematosus: a systematic review. Ann paediatric lupus nephritis. Ann Rheum Dis 2012;71:1771–82. Rheum Dis 2010;69:20–8. 4 Andreoli L, Bertsias GK, Agmon-Levin N, et al. EULAR recommendations for 33 Costedoat-Chalumeau N, Amoura Z, Hulot J-S, et al. Very low blood women’s health and the management of family planning, assisted reproduction, hydroxychloroquine concentration as an objective marker of poor adherence to pregnancy and menopause in patients with systemic lupus erythematosus and/or treatment of systemic lupus erythematosus. Ann Rheum Dis 2007;66:821–4. antiphospholipid syndrome. Ann Rheum Dis 2017;76:476–85. 34 Iudici M, Pantano I, Fasano S, et al. Health status and concomitant prescription 5 Mosca M, Tani C, Aringer M, et al. European League against rheumatism of immunosuppressants are risk factors for hydroxychloroquine non-adherence in recommendations for monitoring patients with systemic lupus erythematosus in systemic lupus patients with prolonged inactive disease. Lupus 2018;27:265–72. clinical practice and in observational studies. Ann Rheum Dis 2010;69:1269–74. 35 Mok CC, Penn HJ, Chan KL, et al. Hydroxychloroquine serum concentrations and 6 Boumpas DT, Bertsias GK, Fanouriakis A. 2008–2018: a decade of recommendations flares of systemic lupus erythematosus: a longitudinal cohort analysis. Arthritis Care for systemic lupus erythematosus. Ann Rheum Dis 2018;77:1547–8. Res 2016;68:1295–302. 7 van der Heijde D, Aletaha D, Carmona L, et al. 2014 update of the EULAR 36 Melles RB, Marmor MF. The risk of toxic retinopathy in patients on long-term standardised operating procedures for EULAR-endorsed recommendations. Ann hydroxychloroquine therapy. JAMA Ophthalmol 2014;132:1453–60. Rheum Dis 2015;74:8–13. 37 Kim J-W, Kim YY, Lee H, et al. Risk of retinal toxicity in longterm users of 8 Brouwers MC, Kho ME, Browman GP, et al. Agree II: advancing Guideline hydroxychloroquine. J Rheumatol 2017;44:1674–9. Development, reporting and evaluation in health care. CMAJ 2010;182:E839–42. 38 Bruce IN, O’Keeffe AG, Farewell V, et al. Factors associated with damage accrual 9 Ward MM. Hospital experience and mortality in patients with systemic lupus in patients with systemic lupus erythematosus: results from the systemic lupus erythematosus. Arthritis Rheum 1999;42:891–8. international collaborating clinics (SLICC) inception cohort. Ann Rheum Dis 10 van Vollenhoven R, Voskuyl A, Bertsias G, et al. A framework for remission in SLE: 2015;74:1706–13. consensus findings from a large international Task Force on definitions of remission 39 Chen H-L, Shen L-J, Hsu P-N, et al. Cumulative burden of Glucocorticoid-related in SLE (DORIS). Ann Rheum Dis 2017;76:554–61. adverse events in patients with systemic lupus erythematosus: findings from a 12- 11 van Vollenhoven RF, Mosca M, Bertsias G, et al. Treat-to-target in systemic lupus year longitudinal study. J Rheumatol 2018;45:83–9. erythematosus: recommendations from an international Task Force. Ann Rheum Dis 40 LSH L, Pullenayegum E, Lim L, et al. From childhood to adulthood: the trajectory of 2014;73:958–67. damage in patients with juvenile-onset systemic lupus erythematosus. Arthritis Care 12 Medina-Quiñones CV, Ramos-Merino L, Ruiz-Sada P, et al. Analysis of complete Res 2017;69:1627–35. remission in systemic lupus erythematosus patients over a 32-year period. Arthritis 41 Yee C-S, Su L, Toescu V, et al. Birmingham SLE cohort: outcomes of a large inception Care Res 2016;68:981–7. cohort followed for up to 21 years. Rheumatology 2015;54:836–43. 13 Steiman AJ, Urowitz MB, Ibañez D, et al. Prolonged clinical remission in patients with 42 Al Sawah S, Zhang X, Zhu B, et al. Effect of corticosteroid use by dose on the risk of systemic lupus erythematosus. J Rheumatol 2014;41:1808–16. developing organ damage over time in systemic lupus erythematosus--the Hopkins 14 Ugarte-Gil MF, Wojdyla D, Pons-Estel GJ, et al. Remission and low disease activity Lupus Cohort. Lupus Sci Med 2015;2:e000066. status (LDAS) protect lupus patients from damage occurrence: data from a 43 Ruiz-Arruza I, Barbosa C, Ugarte A, et al. Comparison of high versus low–medium multiethnic, multinational Latin American lupus cohort (GLADEL). Ann Rheum Dis prednisone doses for the treatment of systemic lupus erythematosus patients with 2017;76:2071–4. high activity at diagnosis. Autoimmun Rev 2015;14:875–9. 15 Urowitz MB, Feletar M, Bruce IN, et al. Prolonged remission in systemic lupus 44 Thamer MAE, Hernan MA, Zhang YI, et al. Prednisone, lupus activity, and permanent erythematosus. J Rheumatol 2005;32:1467–72. organ damage. J Rheumatol 2009;36:560–4. 16 Zen M, Iaccarino L, Gatto M, et al. Prolonged remission in Caucasian patients with 45 Buttgereit F, da Silva JAP, Boers M, et al. Standardised Nomenclature for SLE: prevalence and outcomes. Ann Rheum Dis 2015;74:2117–22. glucocorticoid dosages and glucocorticoid treatment regimens: current questions 17 Tsang-A-Sjoe MWP, Bultink IEM, Heslinga M, et al. Both prolonged remission and and tentative answers in rheumatology. Ann Rheum Dis 2002;61:718–22. lupus low disease activity state are associated with reduced damage accrual in 46 Ruiz-Arruza I, Lozano J, Cabezas-Rodriguez I, et al. Restrictive use of oral systemic lupus erythematosus. Rheumatology 2017;56:121–8. glucocorticoids in systemic lupus erythematosus and prevention of damage without 18 Zen M, Iaccarino L, Gatto M, et al. Lupus low disease activity state is associated with worsening long-term disease control: an observational study. Arthritis Care Res a decrease in damage progression in Caucasian patients with SLE, but overlaps with 2018;70:582–91. remission. Ann Rheum Dis 2018;77:104–10. 47 Ruiz-Irastorza G, Ugarte A, Saint-Pastou Terrier C, et al. Repeated pulses of methyl- 19 Polachek A, Gladman DD, Su J, et al. Defining low disease activity in systemic lupus prednisolone with reduced doses of prednisone improve the outcome of class III, IV erythematosus. Arthritis Care Res 2017;69:997–1003. and V lupus nephritis: an observational comparative study of the Lupus-Cruces and 20 Tselios K, Gladman DD, Touma Z, et al. Clinical remission and low disease activity lupus-Bordeaux cohorts. Autoimmun Rev 2017;16:826–32. have comparable outcomes over 10 years in systemic lupus erythematosus. Arthritis 48 Singh JA, Hossain A, Kotb A, et al. Risk of serious infections with immunosuppressive Care Res 2018. drugs and glucocorticoids for lupus nephritis: a systematic review and network 21 Dall’Era M, Cisternas MG, Smilek DE, et al. Predictors of long-term renal outcome in meta-analysis. BMC Med 2016;14. lupus nephritis trials: lessons learned from the Euro-Lupus nephritis cohort. Arthritis 49 Pego-Reigosa JM, Cobo-Ibáñez T, Calvo-Alén J, et al. Efficacy and safety of Rheumatol 2015;67:1305–13. nonbiologic immunosuppressants in the treatment of nonrenal systemic lupus 22 Touma Z, Urowitz MB, Ibañez D, et al. Time to recovery from proteinuria erythematosus: a systematic review. Arthritis Care Res 2013;65:1775–85. in patients with lupus nephritis receiving standard treatment. J Rheumatol 50 Sakthiswary R, Suresh E. Methotrexate in systemic lupus erythematosus: a systematic 2014;41:688–97. review of its efficacy. Lupus 2014;23:225–35. 23 Mackay M, Dall’Era M, Fishbein J, et al. Establishing surrogate kidney endpoints for 51 Mok CC. Mycophenolate mofetil for non‐renal manifestations of systemic lupus lupus nephritis clinical trials: development and validation of a novel approach to erythematosus: a systematic review. Scand J Rheumatol 2007;36:329–37. predict future kidney outcomes. Arthritis Rheumatol 2018. 52 Tselios K, Gladman DD, Su J, et al. Mycophenolate mofetil in nonrenal 24 Ruperto N, Hanrahan LM, Alarcón GS, et al. International consensus for a definition manifestations of systemic lupus erythematosus: an observational cohort study. J of disease flare in lupus. Lupus 2011;20:453–62. Rheumatol 2016;43:552–8. 25 Koutsonikoli A, Trachana M, Heidich A-B, et al. Dissecting the damage in northern 53 Touma Z, Gladman DD, Urowitz MB, et al. Mycophenolate mofetil for induction Greek patients with childhood-onset systemic lupus erythematosus: a retrospective treatment of lupus nephritis: a systematic review and metaanalysis. J Rheumatol cohort study. Rheumatol Int 2015;35:1225–32. 2011;38:69–78. 26 Ugarte-Gil MF, Acevedo-Vásquez E, Alarcón GS, et al. The number of flares patients 54 Ordi-Ros J, Sáez-Comet L, Pérez-Conesa M, et al. Enteric-coated mycophenolate experience impacts on damage accrual in systemic lupus erythematosus: data from a sodium versus azathioprine in patients with active systemic lupus erythematosus: a multiethnic Latin American cohort. Ann Rheum Dis 2015;74:1019–23. randomised clinical trial. Ann Rheum Dis 2017;76:1575–82. 27 Franklyn K, Lau CS, Navarra SV, et al. Definition and initial validation of a lupus low 55 Knight JH, Howards PP, Spencer JB, et al. Characteristics related to early secondary disease activity state (LLDAS). Ann Rheum Dis 2016;75:1615–21. amenorrhoea and pregnancy among women diagnosed with systemic lupus 28 Kasitanon N, Intaniwet T, Wangkaew S, et al. The clinically quiescent phase in early- erythematosus: an analysis using the goal study. Lupus Sci Med 2016;3:e000139. diagnosed SLE patients: inception cohort study. Rheumatology 2015;54:868–75. 56 Mok CC, Chan PT, To CH. Anti-Müllerian hormone and ovarian reserve in systemic 29 Petri MA, van Vollenhoven RF, Buyon J, et al. Baseline predictors of systemic lupus lupus erythematosus. Arthritis Rheum 2013;65:206–10. erythematosus flares: data from the combined placebo groups in the phase III 57 Tamirou F, Husson SN, Gruson D, et al. Brief report: the Euro-Lupus low-dose belimumab trials. Arthritis Rheum 2013;65:2143–53. intravenous cyclophosphamide regimen does not impact the ovarian reserve,

Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 743 Recommendation

as measured by serum levels of anti-müllerian hormone. Arthritis Rheumatol 83 Kuhn A, Gensch K, Haust M, et al. Efficacy of tacrolimus 0.1% ointment in cutaneous 2017;69:1267–71. lupus erythematosus: A multicenter, randomized, double-blind, vehicle-controlled 58 Blumenfeld Z, Mischari O, Schultz N, et al. Gonadotropin releasing hormone agonists trial. J Am Acad Dermatol 2011;65:54–64. may minimize cyclophosphamide associated gonadotoxicity in SLE and autoimmune 84 Costedoat-Chalumeau N, Dunogué B, Leroux G, et al. A critical review of the effects diseases. Semin Arthritis Rheum 2011;41:346–52. of hydroxychloroquine and chloroquine on the eye. Clin Rev Allergy Immunol 59 Marder W, McCune WJ, Wang L, et al. Adjunctive GnRH-a treatment attenuates 2015;49:317–26. depletion of ovarian reserve associated with cyclophosphamide therapy in 85 Chasset F, Arnaud L, Jachiet M, et al. Changing antimalarial agents after inefficacy premenopausal SLE patients. Gynecol Endocrinol 2012;28:624–7. or intolerance in patients with cutaneous lupus erythematosus: a multicenter 60 Bernatsky S, Ramsey-Goldman R, Joseph L, et al. Lymphoma risk in systemic lupus: observational study. J Am Acad Dermatol 2018;78:107–14. effects of disease activity versus treatment. Ann Rheum Dis 2014;73:138–42. 86 Chasset F, Bouaziz J-D, Costedoat-Chalumeau N, et al. Efficacy and comparison of 61 Hsu CY, Lin MS, YJ S, et al. Cumulative immunosuppressant exposure is associated antimalarials in cutaneous lupus erythematosus subtypes: a systematic review and with diversified cancer risk among 14 832 patients with systemic lupus meta-analysis. Br J Dermatol 2017;177:188–96. erythematosus: a nested case-control study. Rheumatology 2017;56:620–8. 87 Cavazzana I, Sala R, Bazzani C, et al. Treatment of lupus skin involvement with 62 Cobo-Ibáñez T, Loza-Santamaría E, Pego-Reigosa JM, et al. Efficacy and safety of quinacrine and hydroxychloroquine. Lupus 2009;18:735–9. rituximab in the treatment of non-renal systemic lupus erythematosus: a systematic 88 Fruchter R, Kurtzman DJB, Patel M, et al. Characteristics and alternative treatment review. Semin Arthritis Rheum 2014;44:175–85. outcomes of Antimalarial-Refractory cutaneous lupus erythematosus. JAMA 63 Merrill JT, Neuwelt CM, Wallace DJ, et al. Efficacy and safety of rituximab in Dermatol 2017;153:937–9. moderately-to-severely active systemic lupus erythematosus: The randomized, 89 Wenzel J, Brahler S, Bauer R, et al. Efficacy and safety of methotrexate in recalcitrant double-blind, phase II/III systemic lupus erythematosus evaluation of rituximab trial. cutaneous lupus erythematosus: results of a retrospective study in 43 patients. Br J Arthritis Rheum 2010;62:222–33. Dermatol 2005;153:157–62. 64 Rovin BH, Furie R, Latinis K, et al. Efficacy and safety of rituximab in patients with 90 Gammon B, Hansen C, Costner MI. Efficacy of mycophenolate mofetil in antimalarial- active proliferative lupus nephritis: the Lupus nephritis assessment with rituximab resistant cutaneous lupus erythematosus. J Am Acad Dermatol 2011;65:717–21. study. Arthritis Rheum 2012;64:1215–26. 91 Kreuter A, Tomi NS, Weiner SM, et al. Mycophenolate sodium for subacute cutaneous 65 Furie R, Petri M, Zamani O, et al. A phase III, randomized, placebo-controlled study lupus erythematosus resistant to standard therapy. Br J Dermatol 2007;156:1321–7. of belimumab, a monoclonal antibody that inhibits B lymphocyte stimulator, in 92 Md Yusof MY, Shaw D, El-Sherbiny YM, et al. Predicting and managing primary and patients with systemic lupus erythematosus. Arthritis Rheum 2011;63:3918–30. secondary non-response to rituximab using B-cell biomarkers in systemic lupus 66 Navarra SV, Guzmán RM, Gallacher AE, et al. Efficacy and safety of belimumab erythematosus. Ann Rheum Dis 2017;76:1829–36. in patients with active systemic lupus erythematosus: a randomised, placebo- 93 Vital EM, Wittmann M, Edward S, et al. Brief report: responses to rituximab suggest controlled, phase 3 trial. The Lancet 2011;377:721–31. B cell-independent inflammation in cutaneous systemic lupus erythematosus. 67 Iaccarino L, Andreoli L, Bocci EB, et al. Clinical predictors of response and Arthritis Rheumatol 2015;67:1586–91. discontinuation of belimumab in patients with systemic lupus erythematosus in 94 Fernández-Nebro A, de la Fuente JLM, Carreño L, et al. Multicenter longitudinal real life setting. Results of a large, multicentric, nationwide study. J Autoimmun study of B-lymphocyte depletion in refractory systemic lupus erythematosus: the 2018;86:1–8. LESIMAB study. Lupus 2012;21:1063–76. 68 Manzi S, Sánchez-Guerrero J, Merrill JT, et al. Effects of belimumab, a B lymphocyte 95 Chasset F, Tounsi T, Cesbron E, et al. Efficacy and tolerance profile of thalidomide in stimulator-specific inhibitor, on disease activity across multiple organ domains in cutaneous lupus erythematosus: a systematic review and meta-analysis. J Am Acad patients with systemic lupus erythematosus: combined results from two phase III Dermatol 2018;78:342–50. trials. Ann Rheum Dis 2012;71:1833–8. 96 Cortés-Hernández J, Torres-Salido M, Castro-Marrero J, et al. Thalidomide in the 69 van Vollenhoven RF, Petri MA, Cervera R, et al. Belimumab in the treatment of treatment of refractory cutaneous lupus erythematosus: prognostic factors of clinical systemic lupus erythematosus: high disease activity predictors of response. Ann outcome. Br J Dermatol 2012;166:616–23. Rheum Dis 2012;71:1343–9. 97 Bortoluzzi A, Scirè CA, Bombardieri S, et al. Development and validation of a new 70 Díaz-Lagares C, Croca S, Sangle S, et al. Efficacy of rituximab in 164 patients with algorithm for Attribution of neuropsychiatric events in systemic lupus erythematosus. biopsy-proven lupus nephritis: pooled data from European cohorts. Autoimmun Rev Rheumatology 2015;54:891–8. 2012;11:357–64. 98 Acr ad hoc committee on neuropsychiatric lupus nomenclature. The American 71 Duxbury B, Combescure C, Chizzolini C. Rituximab in systemic lupus erythematosus: College of rheumatology Nomenclature and case definitions for neuropsychiatric an updated systematic review and meta-analysis. Lupus 2013;22:1489–503. lupus syndromes. Arthritis Rheum 1999;42:599–608. 72 Iaccarino L, Bartoloni E, Carli L, et al. Efficacy and safety of off-label use of rituximab 99 Bortoluzzi A, Fanouriakis A, Appenzeller S, et al. Validity of the Italian algorithm in refractory lupus: data from the Italian multicentre registry. Clin Exp Rheumatol for the Attribution of neuropsychiatric events in systemic lupus erythematosus: 2015;33:449–56. a retrospective multicentre international diagnostic cohort study. BMJ Open 73 Ramos-Casals M, Soto MJ, Cuadrado MJ, et al. Rituximab in systemic lupus 2017;7:e015546. erythematosusA systematic review of off-label use in 188 cases. Lupus 100 Magro-Checa C, Zirkzee EJ, Beaart-van de Voorde LJJ, et al. Value of multidisciplinary 2009;18:767–76. reassessment in Attribution of neuropsychiatric events to systemic lupus 74 Olfat M, Silverman ED, Levy DM. Rituximab therapy has a rapid and durable erythematosus: prospective data from The Leiden NPSLE cohort. Rheumatology response for refractory cytopenia in childhood-onset systemic lupus erythematosus. 2017;56:1676–83. Lupus 2015;24:966–72. 101 Bertsias GK, Boumpas DT. Pathogenesis, diagnosis and management of 75 Terrier B, Amoura Z, Ravaud P, et al. Safety and efficacy of rituximab in systemic neuropsychiatric SLE manifestations. Nat Rev Rheumatol 2010;6:358–67. lupus erythematosus: results from 136 patients from the French autoimmunity and 102 Bortoluzzi A, Padovan M, Farina I, et al. Therapeutic strategies in severe rituximab registry. Arthritis Rheum 2010;62:2458–66. neuropsychiatric systemic lupus erythematosus: experience from a tertiary referral 76 Chugh S, Darvish-Kazem S, Lim W, et al. Rituximab plus standard of care for centre. Reumatismo 2012;64:350–9. treatment of primary immune thrombocytopenia: a systematic review and meta- 103 Dale RC, Brilot F, Duffy LV, et al. Utility and safety of rituximab in pediatric analysis. Lancet Haematol 2015;2:e75–81. autoimmune and inflammatory CNS disease. Neurology 2014;83:142–50. 77 Boletis JN, Marinaki S, Skalioti C, et al. Rituximab and mycophenolate mofetil for 104 Fanouriakis A, Pamfil C, Sidiropoulos P, et al. Cyclophosphamide in combination relapsing proliferative lupus nephritis: a long-term prospective study. Nephrol Dial with glucocorticoids for severe neuropsychiatric systemic lupus erythematosus: a Transplant 2009;24:2157–60. retrospective, observational two-centre study. Lupus 2016;25:627–36. 78 Gomez Mendez LM, Cascino MD, Garg J, et al. Peripheral blood B cell depletion 105 Gupta N, Ganpati A, Mandal S, et al. Mycophenolate mofetil and deflazacort after rituximab and complete response in lupus nephritis. Clin J Am Soc Nephrol combination in neuropsychiatric lupus: a decade of experience from a tertiary care 2018;13:1502–9. teaching hospital in southern India. Clin Rheumatol 2017;36:2273–9. 79 Kuhn A, Aberer E, Bata-Csörgő Z, et al. S2k guideline for treatment of cutaneous 106 Narváez J, Ríos-Rodriguez V, de la Fuente D, et al. Rituximab therapy in refractory lupus erythematosus - guided by the European Dermatology Forum (EDF) in neuropsychiatric lupus: current clinical evidence. Semin Arthritis Rheum cooperation with the European Academy of Dermatology and Venereology (EADV). J 2011;41:364–72. Eur Acad Dermatol Venereol 2017;31:389–404. 107 Reiner P, Galanaud D, Leroux G, et al. Long-term outcome of 32 patients with chorea 80 Jewell ML, McCauliffe DP. Patients with cutaneous lupus erythematosus who and systemic lupus erythematosus or antiphospholipid antibodies. Mov Disord smoke are less responsive to antimalarial treatment. J Am Acad Dermatol 2011;26:2422–7. 2000;42:983–7. 108 Pamfil C, Fanouriakis A, Damian L, et al. EULAR recommendations for 81 Kuhn A, Gensch K, Haust M, et al. Photoprotective effects of a broad-spectrum neuropsychiatric systemic lupus erythematosus vs usual care: results from two sunscreen in ultraviolet-induced cutaneous lupus erythematosus: A randomized, European centres. Rheumatology 2015;54:1270–8. vehicle-controlled, double-blind study. J Am Acad Dermatol 2011;64:37–48. 109 Jung J-H, Soh M-S, Ahn Y-H, et al. Thrombocytopenia in systemic lupus 82 Kreuter A, Gambichler T, Breuckmann F, et al. Pimecrolimus 1% cream for cutaneous erythematosus: clinical manifestations, treatment, and prognosis in 230 patients. lupus erythematosus. J Am Acad Dermatol 2004;51:407–10. Medicine 2016;95:e2818.

744 Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 Recommendation

110 Serris A, Amoura Z, Canouï-Poitrine F, et al. Efficacy and safety of rituximab for 135 Arnaud L, Mathian A, Ruffatti A, et al. Efficacy of aspirin for the primary prevention systemic lupus erythematosus-associated immune cytopenias: a multicenter of thrombosis in patients with antiphospholipid antibodies: an international and retrospective cohort study of 71 adults. Am J Hematol 2018;93:424–9. collaborative meta-analysis. Autoimmun Rev 2014;13:281–91. 111 Chaturvedi S, Arnold DM, McCrae KR. Splenectomy for immune thrombocytopenia: 136 Bowman L, Mafham M, Wallendszus K, et al. Effects of aspirin for primary prevention down but not out. Blood 2018;131:1172–82. in persons with diabetes mellitus. N Engl J Med 2018;379:1529–39. 112 You YN, Tefferi A, Nagorney DM. Outcome of splenectomy for thrombocytopenia 137 Ridker PM. Should aspirin be used for primary prevention in the Post-Statin era? N associated with systemic lupus erythematosus. Ann Surg 2004;240:286–92. Engl J Med 2018;379:1572–4. 113 Artim-Esen B, Çene E, Şahinkaya Y, et al. Cluster analysis of autoantibodies in 852 138 Pengo V, Ruffatti A, Legnani C, et al. Clinical course of high-risk patients diagnosed patients with systemic lupus erythematosus from a single center. J Rheumatol with antiphospholipid syndrome. J Thromb Haemost 2010;8:237–42. 2014;41:1304–10. 139 Pengo V, Denas G, Zoppellaro G, et al. Rivaroxaban vs warfarin in high-risk patients 114 Duarte-García A, Barr E, Magder LS, et al. Predictors of incident proteinuria among with antiphospholipid syndrome. Blood 2018;132:1365–71. patients with SLE. Lupus Sci Med 2017;4:e000200. 140 Chen D, Xie J, Chen H, et al. Infection in southern Chinese patients with systemic 115 Tang X, Huang Y, Deng W, et al. Clinical and serologic correlations and autoantibody lupus erythematosus: spectrum, drug resistance, outcomes, and risk factors. J clusters in systemic lupus erythematosus: a retrospective review of 917 patients in Rheumatol 2016;43:1650–6. South China. Medicine 2010;89:62–7. 141 Rúa-Figueroa Íñigo, López-Longo J, Galindo-Izquierdo M, et al. Incidence, associated 116 Houssiau FA, Vasconcelos C, D’Cruz D, et al. Immunosuppressive therapy in lupus factors and clinical impact of severe infections in a large, multicentric cohort of nephritis: the Euro-Lupus nephritis trial, a randomized trial of low-dose versus high- patients with systemic lupus erythematosus. Semin Arthritis Rheum 2017;47:38–45. dose intravenous cyclophosphamide. Arthritis Rheum 2002;46:2121–31. 142 Costa-Reis P, Nativ S, Isgro J, et al. Major infections in a cohort of 120 patients with 117 Houssiau FA, Vasconcelos C, D’Cruz D, et al. The 10-year follow-up data of juvenile-onset systemic lupus erythematosus. Clin Immunol 2013;149:442–9. the Euro-Lupus nephritis trial comparing low-dose and high-dose intravenous 143 Hiraki LT, Feldman CH, Marty FM, et al. Serious infection rates among children cyclophosphamide. Ann Rheum Dis 2010;69:61–4. with systemic lupus erythematosus enrolled in Medicaid. Arthritis Care Res 118 Rijnink EC, Teng YKO, Wilhelmus S, et al. Clinical and histopathologic characteristics 2017;69:1620–6. associated with renal outcomes in lupus nephritis. Clin J Am Soc Nephrol 144 Elkayam Oet al. Update of EULAR recommendations for vaccination of patients with 2017;12:734–43. autoimmune inflammatory rheumatic diseases. Ann Rheum Dis 2018;77. 119 Walsh M, Solomons N, Lisk L, et al. Mycophenolate mofetil or intravenous 145 van Assen S, Agmon-Levin N, Elkayam O, et al. EULAR recommendations for cyclophosphamide for lupus nephritis with poor kidney function: a vaccination in adult patients with autoimmune inflammatory rheumatic diseases. subgroup analysis of the Aspreva lupus management study. Am J Kidney Dis Ann Rheum Dis 2011;70:414–22. 2013;61:710–5. 146 Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of clinical criteria for sepsis: for 120 Houssiau FA, Vasconcelos C, D’Cruz D, et al. Early response to immunosuppressive the third International consensus definitions for sepsis and septic shock (Sepsis-3). JAMA 2016;315:762–74. therapy predicts good renal outcome in lupus nephritis: lessons from long- 147 Ballocca F, D’Ascenzo F, Moretti C, et al. Predictors of cardiovascular events in term followup of patients in the Euro-Lupus nephritis trial. Arthritis Rheum patients with systemic lupus erythematosus (SLE): a systematic review and meta- 2004;50:3934–40. analysis. Eur J Prev Cardiol 2015;22:1435–41. 121 Dooley MA, Jayne D, Ginzler EM, et al. Mycophenolate versus azathioprine as 148 Magder LS, Petri M. Incidence of and risk factors for adverse cardiovascular maintenance therapy for lupus nephritis. N Engl J Med 2011;365:1886–95. events among patients with systemic lupus erythematosus. Am J Epidemiol 122 Houssiau FA, D’Cruz D, Sangle S, et al. Azathioprine versus mycophenolate mofetil 2012;176:708–19. for long-term immunosuppression in lupus nephritis: results from the maintain 149 Wu G-C, Liu H-R, Leng R-X, et al. Subclinical atherosclerosis in patients with nephritis trial. Ann Rheum Dis 2010;69:2083–9. systemic lupus erythematosus: a systemic review and meta-analysis. Autoimmun Rev 123 Chen W, Tang X, Liu Q, et al. Short-term outcomes of induction therapy with 2016;15:22–37. tacrolimus versus cyclophosphamide for active lupus nephritis: a multicenter 150 Gustafsson JT, Svenungsson E. Definitions of and contributions to cardiovascular randomized clinical trial. Am J Kidney Dis 2011;57:235–44. disease in systemic lupus erythematosus. Autoimmunity 2014;47:67–76. 124 Lee YH, Lee HS, Choi SJ, et al. Efficacy and safety of tacrolimus therapy for lupus 151 Fasano S, Pierro L, Pantano I, et al. Longterm hydroxychloroquine therapy and nephritis: a systematic review of clinical trials. Lupus 2011;20:636–40. low-dose aspirin may have an additive effectiveness in the primary prevention of 125 Miyasaka N, Kawai S, Hashimoto H. Efficacy and safety of tacrolimus for lupus cardiovascular events in patients with systemic lupus erythematosus. J Rheumatol nephritis: a placebo-controlled double-blind multicenter study. Mod Rheumatol 2017;44:1032–8. 2009;19:606–15. 152 Iudici M, Fasano S, Gabriele Falcone L, et al. Low-dose aspirin as primary prophylaxis 126 Bao H, Liu Z-H, Xie H-L, et al. Successful treatment of class V+IV lupus nephritis with for cardiovascular events in systemic lupus erythematosus: a long-term retrospective multitarget therapy. J Am Soc Nephrol 2008;19:2001–10. cohort study. Rheumatology 2016;55:1623–30. 127 Liu Z, Zhang H, Liu Z, et al. Multitarget therapy for induction treatment of lupus 153 Gaziano JM, Brotons C, Coppolecchia R, et al. Use of aspirin to reduce risk of initial nephritis: a randomized trial. Ann Intern Med 2015;162:18–26. vascular events in patients at moderate risk of cardiovascular disease (ARRIVE): a 128 Szeto C-C, Kwan BC-H, Lai FM-M, et al. Tacrolimus for the treatment of systemic randomised, double-blind, placebo-controlled trial. Lancet 2018;392:1036–46. lupus erythematosus with pure class V nephritis. Rheumatology 2008;47:1678–81. 154 Petri MA, Kiani AN, Post W, et al. Lupus atherosclerosis prevention study (LAPS). Ann 129 Uchino A, Tsukamoto H, Nakashima H, et al. Tacrolimus is effective for lupus Rheum Dis 2011;70:760–5. nephritis patients with persistent proteinuria. Clin Exp Rheumatol 2010;28:6–12. 155 Schanberg LE, Sandborg C, Barnhart HX, et al. Use of atorvastatin in systemic lupus 130 Ikeuchi H, Hiromura K, Takahashi S, et al. Efficacy and safety of multi-target therapy erythematosus in children and adolescents. Arthritis Rheum 2012;64:285–96. using a combination of tacrolimus, mycophenolate mofetil and a steroid in patients 156 Piepoli MF, Hoes AW, Agewall S, et al. 2016 European Guidelines on cardiovascular with active lupus nephritis. Mod Rheumatol 2014;24:618–25. disease prevention in clinical practice: The Sixth Joint Task Force of the European 131 Kasitanon N, Boripatkosol P, Louthrenoo W. Response to combination of Society of Cardiology and Other Societies on Cardiovascular Disease Prevention mycophenolate mofetil, cyclosporin A and corticosteroid treatment in lupus nephritis in Clinical Practice (constituted by representatives of 10 societies and by invited patients with persistent proteinuria. Int J Rheum Dis 2018;21:200–7. experts)Developed with the special contribution of the European Association for 132 Mok CC, To CH, Yu KL, et al. Combined low-dose mycophenolate mofetil and Cardiovascular Prevention & Rehabilitation (EACPR). Eur Heart J 2016;37:2315–81. tacrolimus for lupus nephritis with suboptimal response to standard therapy: a 157 Moroni G, Raffiotta ,F Ponticelli C. Remission and withdrawal of therapy in lupus 12-month prospective study. Lupus 2013;22:1135–41. nephritis. J Nephrol 2016;29:559–65. 133 Conti F, Ceccarelli F, Perricone C, et al. The chronic damage in systemic lupus 158 Piñeiro GJ, Arrizabalaga P, Solé M, et al. Repeated Renal Biopsy - A Predictive erythematosus is driven by flares, glucocorticoids and antiphospholipid antibodies: Tool to Assess the Probability of Renal Flare in Lupus Nephritis. Am J Nephrol results from a monocentric cohort. Lupus 2016;25:719–26. 2016;44:439–46. 134 Taraborelli M, Leuenberger L, Lazzaroni MG, et al. The contribution of 159 De Rosa M, Azzato F, Toblli JE, et al. A prospective observational cohort study antiphospholipid antibodies to organ damage in systemic lupus erythematosus. highlights Kidney biopsy findings of lupus nephritis patients in remission who flare Lupus 2016;25:1365–8. following withdrawal of maintenance therapy. Kidney Int 2018;94:788–94.

Fanouriakis A, et al. Ann Rheum Dis 2019;78:736–745. doi:10.1136/annrheumdis-2019-215089 745 Rheumatoid arthritis

Clinical science Gradual tapering TNF inhibitors versus conventional synthetic DMARDs after achieving controlled disease in patients with rheumatoid arthritis: first-year results of the randomised controlled TARA study Elise van Mulligen,‍ ‍ 1 Pascal Hendrik Pieter de Jong,1 Tjallingius Martijn Kuijper,2 Myrthe van der Ven,1 Cathelijne Appels,3 Casper Bijkerk,4 Joop B Harbers,5 Yael de Man,6 T H Esmeralda Molenaar,7 Ilja Tchetverikov,8 Yvonne P M Goekoop-Ruiterman,9 Jende van Zeben,10 Johanna M W Hazes,1 Angelique E A M Weel,2 Jolanda J Luime1

Handling editor Josef S Abstract Key messages Smolen Objectives The aim of this study is to evaluate the effectiveness of two tapering strategies after achieving ►► Additional material is What is already known about this subject? published online only. To view controlled disease in patients with rheumatoid arthritis ►► With better treatment outcomes, it is nowadays please visit the journal online (RA), during 1 year of follow-up. common to taper medication in patients with (http://dx.​ ​doi.org/​ ​10.​1136/​ Methods in this multicentre single-blinded (research rheumatoid arthritis who are in sustained annrheumdis-2018-​ ​214970). nurses) randomised controlled trial, patients with RA remission. The optimal tapering approach still were included who achieved controlled disease, defined For numbered affiliations see has to be unravelled. end of article. as a Disease Activity Score (DAS) ≤ 2.4 and a Swollen Joint Count (SJC) ≤ 1, treated with both a conventional What does this study add? Correspondence to synthetic disease-modifying antirheumatic drugs ► This study compares the effectiveness of Elise van Mulligen, (csDMARD) and a TNF inhibitor. Eligible patients were ► gradual tapering the conventional synthetic Rheumatology, Erasmus MC, randomised into gradual tapering csDMARDs or TNF Rotterdam 3015, Netherlands; disease-modifying antirheumatic drugs inhibitors. Medication was tapered if the RA was still elise.​ ​vanmulligen@erasmusmc.​ ​ (csDMARD) or the TNF inhibitor in patients with under control, by cutting the dosage into half, a quarter nl rheumatoid arthritis with controlled disease and thereafter it was stopped. Primary outcome was treated with a combination of csDMARDs and a Received 21 December 2018 proportion of patients with a disease flare, defined as Revised 14 March 2019 TNF inhibitor. DAS > 2.4 and/or SJC > 1. Secondary outcomes were Accepted 16 March 2019 ► The TApering strategies in Rheumatoid Arthritis DAS, European Quality of Life-5 Dimensions (EQ5D) ► Published Online First (TARA) trial is one of the first trials which 6 April 2019 and functional ability (Health Assessment Questionnaire assesses differences in tapering strategies, and Disability Index [HAQ-DI]) after 1 year and over time. elaborates on current viewpoints concerning Results a total of 189 patients were randomly assigned tapering treatment, instead of only determining to tapering csDMARDs (n = 94) or tapering anti-TNF (n if tapering is feasible or not. = 95). The cumulative flare rates in the csDMARD and anti-TNF tapering group were, respectively, 33 % (95% How might this impact on clinical practice or CI,24% to 43 %) and 43 % (95% CI, 33% to 53 % (p future developments? = 0.17). Mean DAS, HAQ-DI and EQ-5D did not differ ► Tapering TNF inhibitors was not superior to between tapering groups after 1 year and over time. ► tapering csDMARDs. We, therefore, advise Conclusion Up to 9 months, flare atesr of tapering to taper the TNF inhibitor first. This supports csDMARDs or TNF inhibitors were similar. After 1 year, a current EULAR guidelines. non-significant difference asw found of 10 % favouring csDMARD tapering. Tapering TNF inhibitors was, therefore, not superior to tapering csDMARDs. From a societal perspective, it would be sensible to taper the conventional synthetic disease-modifying antirheu- TNF inhibitor first, because of possible cost reductions matic drugs (csDMARDs) and biological DMARDs and less long-term side effects. (bDMARDs). As a result, 50%–60% of patients Trial registration number NTR2754 with early RA are able to reach low disease activity or even sustained remission.1–4 Because of these © Author(s) (or their employer(s)) 2019. No improved outcomes, it is nowadays more common commercial re-use. See rights to taper medication in patients with RA, who are and permissions. Published Introduction in sustained remission. This is in accordance with by BMJ. Treatment outcomes of rheumatoid arthritis current treatment guidelines.4 However, an optimal To cite: van Mulligen E, (RA) have improved enormously during the past tapering approach, including in which order, still de Jong PHP, Kuijper TM, decades due to earlier detection of the disease, a has to be unravelled. et al. Ann Rheum Dis treat-to-target approach and intensified treat- The benefits of tapering treatment are: (1) a 2019;78:746–753. ment, especially combination therapy with decreased risk of long-term adverse events due to

746 van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 Rheumatoid arthritis immunosuppression, that is, increased infection risk and possi- flare, no further attempts were taken to taper medication during bility of malignancy development, (2) a reduction of healthcare the remainder of the first year of follow-up. costs, especially when biologicals are tapered and (3) a possibly 5 6 improved compliance. On the other hand, tapering treatment Outcomes may lead to more transient or persistent disease flares with The primary outcome was the proportion of patients with a 1 7 8 potential harmful consequences. disease flare within 1 year. Secondary endpoints were disease Previous studies have shown that it is possible to taper activity, functional ability, quality of life, medication usage and DMARDs in various ways, which has been extensively reviewed radiographic progression. 7 9–15 by several research groups. bDMARDs are most frequently Disease activity was measured with the DAS. Functional completely withdrawn. However, with this tapering strategy, ability was measured with the Health Assessment Questionnaire the risk of disease flares in the first year of follow-up is very Disability Index (HAQ-DI).16 Higher HAQ-DI scores indicate high. Other bDMARD-tapering studies used a dose-reduction poorer function. Quality of life was measured with the Euro- approach, which resulted in less disease flares. However, to pean Quality of Life-5 Dimensions (EQ-5D) and Short Form-36 our knowledge, no randomised trials have been performed that (SF-36).17–19 A higher EQ-5D index or SF-36 score indicates a investigate which DMARD should be tapered first. better quality of life. Radiographic progression was measured Therefore, the aim of this study is to compare the effectiveness with the modified total Sharp score (mTSS).20 Radiographs were of two tapering strategies, namely gradually tapering csDMARDs scored chronologically by two out of three qualified assessors, or tumor necrosis factor (TNF) inhibitors, in patients with RA who were blinded for study allocation and the identity of the with controlled disease under a combination of csDMARDs and patients.21 Median mTSS are reported.22 The weighted overall a TNF inhibitor. κ was 0.75 with >99% agreement. The percentage of patients with radiographic progression, defined as a change in mTSS 22 Patients and methods >0.5 and >0.9 (the smallest detectable change), are given. Study design Data were used from a clinical trial (NTR2754)—namely, Follow-up and assessments TApering strategies in Rheumatoid Arthritis (TARA). TARA, a Treatment strategies were tightly controlled, with patients being multicentre, single-blinded (research nurses) randomised trial, examined at baseline and every 3 months thereafter. At each was carried out in 12 rheumatology centres in the Southwestern time point, the DAS, medication usage, development of compli- part of the Netherlands. Inclusion started in September 2011 cations and self-reported questionnaires were collected, except and ended July 2016. for hand and foot radiographs, which were obtained at baseline and after 1 year of follow-up. Patients Adult patients with RA with controlled disease, defined as a Safety monitoring Disease Activity Score (DAS) ≤2.4 and a Swollen Joint Count Safety monitoring took place according to Dutch guidelines, and 23–25 (SJC) ≤1 at two consecutive time points within a 3-month included laboratory tests every 3 months. The medication interval, with a combination of a csDMARD and TNF inhib- was stopped or the dosage was lowered in case of adverse events itor, were included. Exclusion criteria were: (1) not being able to related to medication use. understand, speak and write in Dutch; (2) being diagnosed with a psychiatric or personality disorder and (3) tapering or stopping Statistical analysis therapy due to other reasons. The TARA study was a superiority trial, powered to detect a 20% difference in flare rates between both tapering strategies. Based on related prospective cohort studies from 2011 and before, Randomisation and blinding following assumptions were made: (1) 40% of the patients Patients were randomised using minimisation randomisation tapering their TNF inhibitors to half will have controlled disease stratified for centre. Trained research nurses, blinded to the allo- after 6 months and (2) 60% of the csDMARD tapering group cated treatment arm throughout the study, examined patients will have controlled disease after 6 months.26–28 Therefore, to and calculated the DAS. detect this 20% difference using a significance level of α=0.05 and a power of 80%, 107 patients were needed in each treat- Tapering schedule ment arm, also taking a 10% drop-out ratio into account. Patients were randomised into gradual tapering their csDMARD Outcomes were calculated in an intention-to-treat analysis, or TNF inhibitor. csDMARD tapering was realised by cutting using all available data. Differences in cumulative flare rates the dosage into half, a quarter and thereafter it was stopped. between groups were analysed with a logistic regression model. The TNF inhibitor was tapered by doubling the dose interval, To account for stratified randomisation by centre, intercepts for followed by cutting the dosage into half, and thereafter it was each centre were included. Flare-free survival was visualised stopped. The total tapering schedule took 6 months, with dose with Kaplan-Meier curves. Descriptive statistics were used to adjustments every 3 months as long as there was still a controlled assess the proportion of patients with a controlled disease after disease. At the start of the study, patients were asked to refrain 12 months of follow-up. A linear mixed model with maximum from glucocorticoids (GCs). There were no restrictions on likelihood optimisation was used to compare DAS, HAQ-DI the use of non-steroidal anti-inflammatory drugs (NSAIDs) or and EQ-5D over time. Random intercepts were included for intra-articular GC injections. both hospital and individual patients. Residual correlation was If a disease flare occurred, defined as DAS >2.4 and/or SJC modelled by inclusion of an autoregressive order correlation >1, tapering was stopped and the last effective treatment, when structure. In the final model, the differences in evolution over RA was under control, was restarted. In case of a flare, one intra- time for the outcome DAS, HAQ-DI and EQ-5D between the muscular GC injection was allowed as bridging therapy. After a two groups were assessed. van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 747 Rheumatoid arthritis

Table 1 Baseline characteristics of the csDMARD tapering group and the TNF inhibitor tapering group Tapering csDMARD Tapering TNF Characteristics (n=94) inhibitor (n=95) Demographic Age (years), mean (95% CI) 55.9 (53.0 to 58.8) 57.2 (55.0 to 59.4) Gender, female, n (%) 67 (71) 58 (61) Disease characteristics Symptom duration (years), median 6.0 (4.1–8.5) 6.4 (4.2–8.9) (IQR) RF positive, n (%) 50 (57) 59 (65) ACPA positive, n (%) 62 (71) 67 (75) Disease activity DAS44, mean (95% CI) 1.1 (0.9 to 1.2) 1.0 (0.9 to 1.1) DAS clinical remission, DAS44 76 (81) 87 (88) <1.6, n (%) TJC44, median (IQR) 0 (0–2) 0 (0–1) SJC44, median (IQR) 0 (0–0) 0 (0–0) VAS disease activity (0–100 mm), 20 (4–32) 12 (4–23) median (IQR) ESR in mm/hour, median (IQR) 8 (3–14) 8 (2–15) CRP in mg/L, median (IQR) 2.2 (1–5) 2 (1–6) Use of csDMARDs* MTX, n (%) 90 (96) 84 (88) SASP, n (%) 10 (11) 12 (13) HCQ, n (%) 24 (26) 37 (39) Leflunomide, n (%) 2 (2) 4 (4) Use of TNF inhibitor Etanercept, n (%) 51 (54) 52 (55) Adalimumab, n (%) 37 (39) 40 (42) Figure 1 Trial profile and patient participation. Results are shown Others, n (%)† 6 (6) 3 (3) as number of patients. csDMARDs, conventional synthetic disease- Radiographs (hand/foot) modifying antirheumatic drugs; TNF, tumor necrosis factor. mTSS (0–488), median (IQR) 2 (0–6.5) 1 (0–3.5) Erosion score (0–280), median 0 (0–2.5) 0 (0–2) (IQR) Statistical comparisons of the baseline characteristics and JSN score (0–168), median (IQR) 0.5 (0–2.5) 0 (0–2.5) outcomes were made by Student’s t-test, χ2 test or Wilcoxon Erosive disease, n(%)‡ 37 (39) 26 (27) rank-sum test, when appropriate. Patient-reported outcomes All data were analysed using STATA V.15. A p≤0.05 was HAQ-DI, mean (95% CI) 0.52 (0.42 to 0.62) 0.47 (0.35 to considered statistically significant. 0.58) SF-36, median (IQR) Results PCS 43 (29–48) 47 (39–51) Patients MCS 60 (56–63) 57 (51–62) A total of 330 patients were assessed for eligibility and 189 EQ-5D index, mean (95% CI) 0.86 (0.83 to 0.88) 0.87 (0.85 to of those were randomly assigned to tapering their csDMARD 0.89) (n=94) or tapering their TNF inhibitor (n=95). Most patients *Some patients used a combination of csDMARDs. who were not eligible did not meet the inclusion criteria for †Certolizumab or golimumab. remission or refused participation (figure 1). During the first ‡Erosive disease is characterised as having >1 erosion in three separate joints. ACPA, anticitrullinated protein antibody; CRP, C reactive protein;csDMARD, year of follow-up, 14 patients withdrew from the study, mainly conventional synthetic disease-modifying antirheumatic drug; DAS44, Disease because of refraining from further participation (figure 1). Activity Score Measured in 44 joints; EQ5D, European Quality of Life-5 Dimensions; Table 1 shows the baseline characteristics for both tapering ESR, erythrocyte sedimentation rate; HAQ-DI, Health Assessment Questionnaire strategies. Patients had an average symptom duration of 6.8 Disability Index; HCQ, hydroxychloroquine; JSN, joint space narrowing; MCS, years and were predominantly female (66.1%) with an average mental component summary; mTSS, modified Sharp/Van der Heijde score. MTX, age of 56.6 years. Baseline mean (SD) HAQ-DI was 0.52 (0.47) methotrexate; PCS, physical component summary; RF, rheumatoid factor; SASP, sulfasalazine; SF-36, Short Form-36; SJC, Swollen Joint Count; TJC, Tender Joint and 0.47 (0.53) and EQ-5D was 0.86 (0.12) and 0.87 (0.11) for, Count; VAS, Visual Analogue Scale; respectively, the csDMARD and TNF inhibitor tapering group. At baseline, 81% of the csDMARD tapering group and 88% of the TNF inhibitor tapering group were in remission (DAS <1.6) 55%) or adalimumab (respectively, 39% and 42%). Oral GCs (table 1). The majority of patients in the csDMARD and TNF were taken by 4 (4%) patients in the csDMARD tapering group inhibitor tapering group used methotrexate (respectively, 97% and 2 (2%) patients in the TNF inhibitor tapering group, while and 86%) in combination with etanercept (respectively, 54% and NSAIDs were taken by 14 (15%) and 20 (21%) patients (table 1).

748 van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 Rheumatoid arthritis

Table 2 Clinical response after 12 months for both tapering groups, according to intention-to-treat Clinical response after 12 Tapering csDMARD Tapering TNF months (n=85) inhibitor (n=89) Disease activity DAS44, mean (95% CI) 1.31 (1.17 to 1.46) 1.35 (1.19 to 1.51) TJC44, median (IQR) 0 (0–2) 0 (0–3) SJC44, median (IQR) 0 (0–0) 0 (0–1) VAS disease activity (0–100 17 (5–36) 19 (6–42) mm), median (IQR) ESR in mm/hour, median (IQR) 11 (5–21) 11 (4–19) CRP in mg/L, median (IQR) 2.9 (1–6) 4 (1–9) DAS clinical remission, DAS44 57 (69) 58 (66) <1.6, n (%) ΔDAS44 (T12–T0), mean (95% 0.28 (0.16 to 0.40) 0.40 (0.22 to 0.57) CI) Radiographic progression (hand/foot) mTSS (0–488), median (IQR) 2 (0–6.5) 1 (0–4) Erosion score (0–280), median 0.5 (0–2) 0 (0–2) (IQR) JSN score (0–168), median 0.5 (0–2.5) 0 (0–2.5) (IQR) ΔmTSS (T12–T0), median (IQR) 0 (0–0) 0 (0–0) Patients with progression >0.5, 4 (5) 5 (6) n (%) Patients with progression >0.9, 4 (5) 5 (6) n (%) Erosive disease, n(%)* 37 (44) 30 (34) Patient-reported outcomes HAQ-DI, mean (95% CI) 0.59 (0.46 to 0.73) 0.55 (0.43 to 0.66) ΔHAQ-DI (T12-T0), mean 0.05 (-0.05 to 0.13) 0.07 (-0.01 to 0.16) Figure 2 Percentages flares and Kaplan-Meier curves for maintenance (95% CI) of controlled disease in the first 12 months. Percentages with flare SF-36, median (IQR) indicates the cumulative number of patients with flares. Error bars PCS 43 (32–50) 44 (35–50) indicate 95% CIs. Kaplan-Meier curves indicate loss of controlled MCS 58 (53–62) 59 (51–62) disease (DAS >2.4 and/or SJC >1) over time. Numbers below the EQ-5D index, mean (95% CI) 0.80 (0.76 to 0.84) 0.82 (0.79 to 0.85) Kaplan-Meier curve indicate the number of patients at risk per ΔEQ-5D index (T12–T0), mean −0.06 (-0.09 to 0.02) −0.05 (-0.08 to 0.02) time point. csDMARD, conventional synthetic disease-modifying (95% CI) antirheumatic drug; DAS, Disease Activity Score Measured; SJC, Swollen *Erosive disease is characterised as having >1 erosion in three separate joints. Joint Count; TNF, tumor necrosis factor. . JSN; Joint space narrowing, CRP, C reactive protein; DAS44, Disease Activity Score Measured in 44 joints; EQ-5D, European Quality of Life-5 Dimensions; ESR, At baseline, respectively, 39% and 27% of patients within the erythrocyte sedimentation rate; HAQ-DI, Health Assessment Questionnaire Disability Index;MCS, mental component summary; PCS, physical component summary;SF-36, csDMARD or TNF inhibitor group had erosive disease. Short Form-36 ; SJC, Swollen Joint Count; TJC, Tender Joint Count;TNF, tumor necrosis factor; VAS, Visual Analogue Scale; csDMARD, conventional synthetic Outcomes disease-modifying antirheumatic drug; mTSS, modified Sharp/Van der Heijde score. After 1 year of follow-up, the cumulative flare rate was 33% (95% CI, 24% to 43%) in the csDMARD and 43% (95% CI, 33% to 53%) in the TNF inhibitor tapering group (figure 2). Over time, the patients with a disease flare increased and thus This means that 63/94 (67%) in the csDMARD tapering group the proportion of patients with a DAS <2.4 decreased in both and 54/95 (57%) in the TNF inhibitor tapering group still had tapering strategies. A similar trend was seen for the HAQ-DI and a well-controlled RA (p=0.17). Of the patients who flared and EQ-5D over time (figure 3). restarted the last effective treatment strategy, 46% regained Median mTSS scores were 2 (IQR 0–6.5) in the csDMARD a DAS <2.4 within 3 months, which increased to 67% by 6 and 1 (IQR 0–4) in the TNF inhibitor tapering group after 1 year months. Two patients (1%) were unable to get back in remission of follow-up (table 2). Radiographic progression was seen in 5% within the first year. of the csDMARD-tapering group and 6% of the TNF inhibitor No significant differences were seen in DAS (p=0.72), tapering group (p=0.82). Also, the cumulative probability plots HAQ-DI (p=0.63) and EQ-5D (p=0.58) after 1 year between were overlapping (figure 3B). both tapering strategies (table 2). Also over time, the DAS (p=0.49) and EQ-5D (p=0.35) were not significantly different Treatment between both tapering strategies (figure 3). Although the TNF After 12 months, 58 patients in the csDMARD-tapering group inhibitor tapering group seems to have lower HAQ-DI scores and 45 patients in the TNF inhibitor tapering group completely over time, this was not significantly different (p=0.15) (figure 3). tapered their medication (p=0.09). On the other hand, 8 and 16 van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 749 Rheumatoid arthritis

Figure 3 Disease activity, cumulative probability plot for radiological progression, functional ability and quality of life over time per tapering arm. (A, C, D) Error bars indicate 95% CIs. (B) Each point represents radiological progression (T12–T0) of an individual patient, measured with the mTSS score at 0 and 12 months. csDMARD, conventional synthetic disease-modifying antirheumatic drug; DAS: Disease Activity Score; EQ index, European Quality Index; HAQ-DI: Health Assessment Questionnaire disability index; mTSS, modified Sharp/Van Der Heijde score; TNF, tumor necrosis factor. patients were using the same dosage as at start of the trial. The comparable in both tapering groups over time and after 1 year remaining patients were able to taper their medication partially of follow-up. Also, no significant differences in adverse events (figure 4C). The course of the tapering schedule is visualised or radiological progression were seen between both tapering in figure 4A,B. There was an overall significant difference in strategies. tapering status after 12 months of follow-up between the two Nowadays, more patients with RA achieve a state of sustained tapering strategies (p=0.02). During the follow-up period, remission, which makes them eligible for tapering treatment. we found no significant differences in GC and NSAID usage This is reflected in current European league against rheumatism between both tapering groups (figure 4D). (EULAR) recommendations for the management for RA. The advice is to taper DMARD therapy in patients with RA who are in Adverse events sustained remission in the following ordering: GCs, bDMARDS In the csDMARD-tapering group, 82 adverse events were self-re- and csDMARDs.4 Our results and the fact that TNF-blockers ported versus 98 in the TNF inhibitor tapering group (online are more expensive than csDMARDs support aformentioned supplemental table S1). Serious adverse events (SAEs) were seen tapering order. in 10 (12%) patients tapering csDMARDs and 5 (6%) patients The majority of previous tapering trials focused on the with- tapering TNF inhibitors (p=0.3, online supplemental table S1). drawal of TNF inhibitors alone. Flare rates for tapering TNF Reported SAEs were hospitalisation, herpes zoster infection, inhibitors varied between 51% and 77%. The Potential Optimal- basal cell carcinoma, large cell lung carcinoma and a bruised isation of Expediency and Effectiveness of TNF-blockers (POET) rib. None of the SAEs were considered to be related to the trial study, for example, reported a 51.2% flare rate (DAS28 >3.2 or treatment. ΔDAS28 >0.6) after stopping the TNF inhibitor.7 The Spacing of TNF-blocker injections in Rheumatoid ArthritiS Study (STRASS) Discussion showed a 76.6% flare rate (DAS28 >2.6 or ΔDAS28 >0.6) Tapering csDMARDs resulted in a 33% (95% CI, 24% to 43%) when extending the dosage interval of the TNF inhibitor.14 The flare rate (DAS44 >2.4 and/or an SJC>1), while tapering TNF Dose REduction Strategy of Subcutaneous TNF inhibitors study inhibitors gave a 43% (95% CI, 33% to 53%) flare rate over a (DRESS) reported a 55% flare rate (ΔDAS28-CRP>0.6) after 1-year period in the randomised controlled TARA trial. At 12 a dose reduction of the TNF inbitor.13 Finally, the PRESERVE months, 103 (59%) patients were able to stop either their TNF trial (A Randomized, Double-Blind Study Comparing the Safety inhibitor or csDMARD, while 47 (37%) patients were using a & Efficacy of Once-Weekly Etanercept 50 mg, Etanercept 25 lower dosage. Clinical and patient-reported outcomes were mg, & Placebo in Combination With Methotrexate in Subjects

750 van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 Rheumatoid arthritis

Figure 4 Status of tapering in the first year of follow-up. (A, B) Overview of tapering status per time point. Results are shown as percentages of patients. According to protocol, the doses were halved every 3 months, starting at T0, and after 6 months, patients could stop their tapered medication when they were still in a controlled disease state. Detailed numbers per time point are given in online supplementary table S2. (C) Tapering status after 12 months. Columns indicate the percentage of patients that tapered medication until the indicated amount of the original dose. (D) Overview of glucocorticoids and NSAID use in the first year of follow-up. csDMARD, conventional synthetic disease-modifying antirheumatic drug; NSAIDs, non- steroidal anti-inflammatory drugs.

With Active Rheumatoid Arthritis) reported a 57.4% flare rate trials. This indicates that our criteria were more strict than other (DAS28 >3.2) when the TNF inhibitor was stopped, and a studies, but that flare rates are comparible between the tapering 20.9% flare rate (DAS28 >3.2) when the TNF inhibitor dose studies. was cut into half.15 Only few randomised controlled trials inves- Also, the flare duration was longer in the TARA trial tigated tapering of csDMARDs, but the majority looked at the compared with other trials, which could be due to the combined tapering of csDMARDs and biologicals. Flare rates measurement intervals of 3 months. If patients did not have within these studies varied between 35% and 56%.1 6 29–32 a controlled disease 3 months after flare, we assumed that Although flare rates of aforementioned studies are similar to or the duration of flare was 6 months. That might be a reason higher than our findings, direct comparison is difficult, because that our results seem to have a long flare duration compared of the differences in the study design. The most important study with the Dose Reduction or Discontinuation of Etanercept design differences are: (1) no common definition for relapse or in Methotrexate-Treated Rheumatoid Arthritis Patients Who flare, (2) no comparison between tapering of csDMARDs and Have Achieved a Stable Low Disease Activity-State (DOSERA) TNF inhibitors, and (3) DMARD therapy could only be tapered study or DRESS study, in which they knew the exact duration or stopped once during follow-up. If we would use other criteria of flare in weeks.12 13 to define a flare in the TARA population, we would observe In this study, there are several strengths and limitations. higher hypothetical flare rates. We would have encountered Strengths of the study are that we performed a randomised a 74.1% flare rate if using DAS28 >3.2 or ΔDAS28 >0.6, an controlled trial to asess tapering in patients with RA with a 80.5% flare rate if using DAS28 >2.6 or ΔDAS28 >0.6, a 52.3% controlled disease. The TARA trial is one of the first trials which flare rate if we use ΔDAS28-CRP>0.6, and a 39.1% flare rate assesses the differences in tapering strategies, and elaborates on if using DAS28 >3.2. Mostly, these flare rates are higher than current viewpoints concerning tapering treatment, instead of our reported flare rates, but are similar to previous mentioned only determining if tapering is feasible or not. van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 751 Rheumatoid arthritis

Some limitations should be noted as well. First of all, inclu- 4Rheumatology, Reinier de Graaf Gasthuis, Delft, The Netherlands 5 sion was terminated earlier due to difficulties with recruiting. Rheumatology, Franciscus ziekenhuis, Roosendaal, The Netherlands 6Rheumatology, Sint Antonius Ziekenhuis, Utrecht, The Netherlands This was due to the initial inclusion criteria being too strict 7Rheumatology, Groene Hart Ziekenhuis, Gouda, The Netherlands (DAS ≤1.6), and the start of another trial (POET study) which 8Rheumatology, Albert Schweitzer Ziekenhuis, Dordrecht, The Netherlands used the same pool of eligible patients.7 The study sample size 9Rheumatology, Haga Ziekenhuis, The Hague, The Netherlands 10 was based on a 20% difference between both tapering strat- Rheumatology, Franciscus Gasthuis, Rotterdam, The Netherlands egies resulting in 96 patients per arm. We found, however, Acknowledgements We especially thank the participating patients in the TARA a 10% difference with 85 patients in the csDMARD and 89 trial for their willingness to contribute to the study and for their cooperation. We patients in the TNF inhibitor arm. This resulted in a power of thank all rheumatologists from the following participating centres: Erasmus MC, Rotterdam; Maasstad ziekenhuis, Rotterdam; Sint Franciscus Gasthuis & Vlietland, 70% instead of 80%. For this reason, we performed a worst- Rotterdam and Schiedam; Amphia ziekenhuis, Breda; Reinier de Graaf ziekenhuis, case scenario analysis to see if our results were valid. We used Delft; Antonius ziekenhuis, Utrecht; Bravis ziekenhuis, Bergen op Zoom and the following assumptions: (1) all extra included patients in Roosendaal; Groene hart ziekenhuis, Gouda; Albert Schweitzer ziekenhuis, Dordrecht; the csDMARD tapering group had no flares and (2) all extra Haga ziekenhuis, Den Haag and Beatrix ziekenhuis, Gorinchem. We also thank all study-nurses, laboratory personnel, coinvestigators and others who were involved included patients in the TNF inhibitor tapering group flared. with the TARA study. This analysis showed an 18% difference in flares, which is still Contributors All authors contributed to the conception or design of the study; below the 20% difference on which our power calculation was or the acquisition, analysis or interpretation of data; drafting or revision of the based. Therefore, we think our current results and conclusions manuscript; and final approval of the manuscript for publication. are valid. Funding The study was supported by an unrestricted grant from ZonMW. Second, rheumatologist could have only referred patients Competing interests None declared. who achieved low disease activity quickly and had less severe disease and, therefore, creating selection bias. However, we Patient consent for publication Obtained. think that our target population is the same as the one we Ethics approval Medical ethics committees of each participating centre approved would apply our results to, because those are the patients who the protocol. are suitable for tapering and are willing to taper their medi- Provenance and peer review Not commissioned; externally peer reviewed. cation. Furthermore, only research nurses, who did the DAS assessment, were blinded. Rheumatologists, therefore, knew References the tapering strategy of their patients. This design was chosen 1 Kuijper TM, Luime JJ, de Jong PHP, et al. Tapering conventional synthetic DMARDs in patients with early arthritis in sustained remission: 2-year follow-up of the tREACH to mimic daily practise as much as possible. However, it could trial. Ann Rheum Dis 2016;75:2119–23. be a possible source of bias, since rheumatologist might prefer 2 scott DL, Wolfe F, Huizinga TWJ. Rheumatoid arthritis. Lancet 2010;376:1094–108. one of the two tapering strategies and would possibly treat 3 singh JA, Saag KG, Bridges SL, et al. 2015 American College of rheumatology patients differently depending on the tapering strategy. guideline for the treatment of rheumatoid arthritis. Arthritis Rheumatol 2016;68:1–26. Third, the time frame of follow-up was only 1 year. Although 4 smolen JS, Landewé R, Bijlsma J, et al. EULAR recommendations for the management the differences in flare rates were not significantly different of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic between both tapering strategies, the largest difference was drugs: 2016 update. Ann Rheum Dis 2017;76:960–77. seen at 12 months. Data of the second year are needed to 5 Fautrel B, Den Broeder AA. De-intensifying treatment in established rheumatoid investigate if this difference will increase. arthritis (rA): why, how, when and in whom can DMARDs be tapered? Best Pract Res Clin Rheumatol 2015;29:550–65. Last, we encountered 19% protocol violations, which could 6 Haschka J, Englbrecht M, Hueber AJ, et al. Relapse rates in patients with rheumatoid underestimate the effect of one of the two tapering strategies. arthritis in stable remission tapering or stopping antirheumatic therapy: interim We analysed the type of violations and we can conclude that results from the prospective randomised controlled retro study. Ann Rheum Dis most protocol violations were randomly distributed over the 2016;75:45–51. 7 Ghiti Moghadam M, Vonkeman HE, Ten Klooster PM, et al. Stopping tumor necrosis two treatment arms and were made due to a treat-to-target factor inhibitor treatment in patients with established rheumatoid arthritis in approach. remission or with stable low disease activity: a pragmatic multicenter, open-label To ensure optimal rheumatic care in the future, efficient use randomized controlled trial. Arthritis Rheumatol 2016;68:1810–7. of biological treatment is needed.33 By tapering medication, 8 schett G, Emery P, Tanaka Y, et al. Tapering biologic and conventional DMARD therapy costs can be reduced, especially when tapering bDMARDs. in rheumatoid arthritis: current evidence and future directions. Ann Rheum Dis 2016;75:1428–37. On the other hand, 38% of the patients in the TARA study 9 Kuijper TM, Lamers-Karnebeek FBG, Jacobs JWG, et al. Flare rate in patients with flared within the first year, which may have a direct impact on rheumatoid arthritis in low disease activity or remission when tapering or stopping patients’ lives (i.e, worker productivity and unemployment). synthetic or biologic DMARD: a systematic review. J Rheumatol 2015;42:2012–22. Therefore, it is important to know which tapering strategy is 10 edwards CJ, Fautrel B, Schulze-Koops H, et al. Dosing down with biologic therapies: a systematic review and clinicians’ perspective. Rheumatology 2017;56:1847–56. most cost-effective, which will be addressed in a follow-up 11 van Herwaarden N, den Broeder AA, Jacobs W, et al. Down-titration and analysis. discontinuation strategies of tumor necrosis factor-blocking agents for rheumatoid In conclusion, the TARA study showed that up to 9 months, arthritis in patients with low disease activity. Cochrane Database Syst Rev 2014;9. flare rates of tapering csDMARDs or TNF inhibitors were 12 van Vollenhoven RF, Østergaard M, Leirisalo-Repo M, et al. Full dose, reduced dose or discontinuation of etanercept in rheumatoid arthritis. Ann Rheum Dis 2016;75:52–8. similar. After 1 year, a non-significant difference in flare rates 13 van Herwaarden N, van der Maas A, Minten MJM, et al. Disease activity guided dose was found of 10% in favour of csDMARD tapering. Tapering reduction and withdrawal of adalimumab or etanercept compared with usual care TNF inhibitors was, therefore, not superior to tapering in rheumatoid arthritis: open label, randomised controlled, non-inferiority trial. BMJ csDMARDs. From a societal perspective, it would be sensible 2015;350:h1389. to taper the TNF inhibitor first, because of possible cost reduc- 14 Fautrel B, Pham T, Alfaiate T, et al. Step-down strategy of spacing TNF-blocker injections for established rheumatoid arthritis in remission: results of the multicentre tions and less long-term side effects. non-inferiority randomised open-label controlled trial (STRASS: spacing of TNF-blocker injections in rheumatoid arthritis study). Ann Rheum Dis 2016;75:59–67. Author affiliations 15 smolen JS, Nash P, Durez P, et al. Maintenance, reduction, or withdrawal of 1Rheumatology, Erasmus Medical Centre, Rotterdam, The Netherlands etanercept after treatment with etanercept and methotrexate in patients with 2Rheumatology, Maasstad Ziekenhuis, Rotterdam, The Netherlands moderate rheumatoid arthritis (preserve): a randomised controlled trial. Lancet 3Rheumatology, Amphia ziekenhuis, Breda, The Netherlands 2013;381:918–29.

752 van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 Rheumatoid arthritis

16 siegert CE, Vleming LJ, Vandenbroucke JP, et al. Measurement of disability in Dutch 26 Tanaka Y, Takeuchi T, Mimori T, et al. Discontinuation of infliximab after attaining low rheumatoid arthritis patients. Clin Rheumatol 1984;3:305–9. disease activity in patients with rheumatoid arthritis: RRR (remission induction by 17 Hurst NP, Kind P, Ruta D, et al. Measuring health-related quality of life in rheumatoid remicade in rA) study. Ann Rheum Dis 2010;69:1286–91. arthritis: validity, responsiveness and reliability of EuroQol (EQ-5D). Br J Rheumatol 27 saleem B, Brown AK, Keen H, et al. Disease remission state in patients treated with 1997;36:551–9. the combination of tumor necrosis factor blockade and methotrexate or with disease- 18 McHorney CA, Ware JE, Raczek AE. The mos 36-Item short-form health Survey (SF-36): modifying antirheumatic drugs: a clinical and imaging comparative study. Arthritis II. psychometric and clinical tests of validity in measuring physical and mental health Rheum 2009;60:1915–22. constructs. Med Care 1993;31:247–63. 28 laganà B, Picchianti Diamanti A, Ferlito C, et al. Imaging progression despite clinical 19 Ware JE, Sherbourne CD. The mos 36-item short-form health Survey (SF-36). I. remission in early rheumatoid arthritis patients after etanercept interruption. Int J conceptual framework and item selection. Med Care 1992;30:473–83. Immunopathol Pharmacol 2009;22:447–54. 20 van der Heijde D. How to read radiographs according to the Sharp/van Der Heijde 29 ten Wolde S, Breedveld FC, Hermans J, et al. Randomised placebo-controlled study of stopping second-line drugs in rheumatoid arthritis. Lancet 1996;347:347–52. method. J Rheumatol 2000;27:261–3. 30 Gøtzsche PC, Hansen M, Stoltenberg M, et al. Randomized, placebo controlled trial 21 van Tuyl LHD, van der Heijde D, Knol DL, et al. Chronological reading of radiographs of withdrawal of slow-acting antirheumatic drugs and of observer bias in rheumatoid in rheumatoid arthritis increases efficiency and does not lead to bias. Ann Rheum Dis arthritis. Scand J Rheumatol 1996;25:194–9. 2014;73:391–5. 31 Klarenbeek NB, van der Kooij SM, Güler-Yüksel M, et al. Discontinuing treatment in 22 van der Heijde D, Simon L, Smolen J, et al. How to report radiographic data in patients with rheumatoid arthritis in sustained clinical remission: exploratory analyses randomized clinical trials in rheumatoid arthritis: guidelines from a roundtable from the best study. Ann Rheum Dis 2011;70:315–9. discussion. Arthritis Rheum 2002;47:215–8. 32 allaart CF, Lems WF, Huizinga TWJ. The best way of withdrawing biologic agents. Clin 23 nederlandse Vereniging voor Reumatologie. NVR Richtlijnen medicijnen: biologicals, Exp Rheumatol 2013;31(4 Suppl 78):S14–18. 2014. Available: https://www.​nvr.nl/​ ​wp-content/​ ​uploads/2014/​ ​11/​NVR-​Medicijnen-​ 33 schoels M, Wong J, Scott DL, et al. Economic aspects of treatment options Update_Biologicals_​ ​richtlijn-23-​ ​6-​2014.​pdf [Accessed Jul 2018]. in rheumatoid arthritis: a systematic literature review Informing the EULAR 24 nederlandse Vereniging voor Reumatologie. NVR Richtlijnen medicijnen: sulfazalazine, recommendations for the management of rheumatoid arthritis. Ann Rheum Dis 2002. Available: https://www.​nvr.nl/​ ​wp-content/​ ​uploads/2014/​ ​11/​NVR-​Medicijnen-​ 2010;69:995–1003. richtlijn-​Sulfasalazine-​2002.​pdf [Accessed Jul 2018]. 34 Van Mulligen E, Weel A, Kuijper TM, et al. OP0113 Gradual tapering tnf blockers 25 nederlandse Vereniging voor Reumatologie. NVR Richtlijnen medicijnen: versus conventional synthetic dmards in patients with rheumatoid arthritis in methotrexaat, 2011. Available: https://www.​nvr.nl/​ ​wp-content/​ ​uploads/2014/​ ​11/​NVR-​ sustained remission: first year results of the randomised controlled tara-study Medicijnen-MTX-​ ​richtlijn-2009-​ ​update-​2011.​pdf [Accessed Jul 2018]. [abstract]. Ann Rheum Dis 2018;77.

van Mulligen E, et al. Ann Rheum Dis 2019;78:746–753. doi:10.1136/annrheumdis-2018-214970 753 Rheumatoid arthritis

Clinical science Effects of BI 655064, an antagonistic anti-CD40 antibody, on clinical and biomarker variables in patients with active rheumatoid arthritis: a randomised, double-blind, placebo-controlled, phase IIa study Sudha Visvanathan,‍ ‍ 1 Stefan Daniluk,2 Rafał Ptaszyński,3 Ulf Müller-Ladner,4 Meera Ramanujam,1 Bernd Rosenstock,5 Anastasia G Eleftheraki,5 Richard Vinisko,1 Alena Petříková,6 Herbert Kellner,7 Eva Dokoupilova,8,9 Brygida Kwiatkowska,10 Rieke Alten,11 Christian Schwabe,12 Patrick Baum,5 David Joseph,1 Jay S Fine,1 Steven J Padula,13 Jürgen Steffgen5

Handling editor Josef S Abstract Key messages Smolen Objective To evaluate the safety, efficacy and therapeutic mechanism of BI 655064, an antagonistic ►► Additional material is What is already known about this subject? published online only. To view anti-CD40 monoclonal antibody, in patients with ►► Hyperactivation of the CD40–CD40 ligand please visit the journal online rheumatoid arthritis (RA) and an inadequate response to (CD40L) pathway is associated with disease (http://dx.​ ​doi.org/​ ​10.​1136/​ methotrexate (MTX-IR). activity and pathogenesis of autoimmune annrheumdis-2018-​ ​214729). Methods in total, 67 patients were randomised to diseases and is a potential therapeutic target in receive weekly subcutaneous doses of 120 mg BI 655064 For numbered affiliations see rheumatoid arthritis (RA). end of article. (n=44) or placebo (n=23) for 12 weeks. The primary ► Former clinical trials of anti-CD40L antibodies endpoint was the proportion of patients who achieved ► were halted because of increased incidence of Correspondence to 20% improvement in American College of Rheumatology thromboembolism. Dr Sudha Visvanathan, criteria (ACR20) at week 12. Safety was assessed in Boehringer Ingelheim patients who received at least one dose of study drug. Pharmaceuticals Inc., Ridgefield, What does this study add? Results at week 12, the primary endpoint was not CT 06877, USA; ► In this small phase IIa study, treatment met, with 68.2% of patients treated with BI 655064 ► sudha.​ ​visvanathan@boehringer-​ ​ with BI 655064, an antagonistic anti-CD40 ingelheim.com​ achieving an ACR20 vs 45.5% with placebo (p=0.064); monoclonal antibody, in inadequate response using Bayesian analysis, the posterior probability of Received 8 November 2018 to methotrexate patients with RA resulted seeing a difference greater than 35% was 42.9%. BI Revised 20 February 2019 in marked changes in clinical and biological 655064 was associated with greater changes in CD40– Accepted 24 February 2019 parameters, including reductions in activated Published Online First CD40L pathway-related markers, including reductions in B-cells, autoantibody production and 22 March 2019 inflammatory and bone resorption markers (interleukin-6, inflammatory and bone resorption markers, matrix metalloproteinase-3, receptor activator of nuclear with a favourable safety profile; however, factor-κB ligand), concentration of autoantibodies the primary endpoint (20% improvement in (immunoglobulin [Ig]G rheumatoid factor [RF], IgM RF, American College of Rheumatology criteria IgA RF) and CD95+ activated B-cell subsets. No serious response at week 12) was not met. adverse events (AEs) related to BI 655064 treatment or thromboembolic events occurred; reported AEs were How might this impact on clinical practice or mainly of mild intensity. future developments? Conclusion although blockade of the CD40–CD40L ► The CD40–CD40L pathway has been shown pathway with BI 655064 in MTX-IR patients with RA ► to play an important role in the pathogenesis resulted in marked changes in clinical and biological of systemic lupus erythematosus and lupus parameters, including reductions in activated B-cells, nephritis (LN). autoantibody production and inflammatory and bone ► Given the favourable clinical safety profile of BI resorption markers, with a favourable safety profile, ► 655064 in this study, a clinical trial assessing clinical efficacy was not demonstrated in this small phase © Author(s) (or their the efficacy and safety of BI 655064 in LN is employer(s)) 2019. Re-use IIa study. ongoing. permitted under CC BY-NC. No Trial registration number NCT01751776 commercial re-use. See rights and permissions. Published by BMJ. CD40L (CD154) is expressed on the T-cell surface To cite: Visvanathan S, Introduction following activation. Binding of CD40L to CD40 Daniluk S, Ptaszyński R, Hyperactivation of the CD40–CD40 ligand expressed on B-cell surfaces mediates T-cell-de- et al. Ann Rheum Dis (CD40L) pathway is associated with disease activity pendent B-cell proliferation, maturation, antibody 2019;78:754–760. and pathogenesis of autoimmune diseases.1–7 formation and immunoglobulin isotype switch.

754 Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 Rheumatoid arthritis

Similarly, CD40L binding to CD40 expressed on dendritic defined as a Disease Activity Score in 28 joints (DAS28) using cells, monocytes and macrophages, endothelial cells and fibro- four variables and C reactive protein (CRP) ≥3.5, ≥6 tender blasts promotes cell differentiation, proinflammatory cytokine joints (on tender joint count, out of 68) and ≥6 swollen joints production and upregulated expression of costimulatory ligands. (on swollen joint count, out of 66) at screening and confirmed at Loss of CD40L expression or function results in X-linked hyper- randomisation; had a serum CRP level ≥0.8 mg/dL or erythro- immunoglobulin (Ig)M syndrome, which is characterised by cyte sedimentation rate (ESR) ≥28 mm/hour at screening (added recurrent infections due to impaired immunoglobulin isotype via protocol amendment to improve patient enrolment); and switching and somatic hypermutations.8 were seropositive for rheumatoid factor (RF) or anticyclic citrul- The CD40–CD40L pathway is a potentially attractive ther- linated protein antibodies (ACPAs). Patients were to continue apeutic target in the treatment of rheumatoid arthritis (RA). background treatment with a stable dose (for ≥6 weeks prior CD40 expression has been demonstrated on B-cells, dendritic to screening) of MTX (≥15 mg/week; for patients who were cells and monocytes in synovial fluid and on synovial fibroblasts not able to tolerate this dose due to side effects, a stable weekly in joints affected by RA, while enhanced CD40L expression has dose ≥7.5 mg was permitted); patients were required to have been found on T-cells in the periphery and synovial tissue in the been receiving MTX for ≥3 months prior to screening; changes RA disease setting.3 4 6 9 Anti-CD40 antibody treatment of syno- in MTX dosing were not permitted except in the event of an vial fibroblasts expressing CD40 from patients with RA resulted MTX-related safety issue. Low-dose systemic glucocorticoids in decreased levels of spontaneous tumour necrosis factor (TNF) (equivalent to ≤10 mg prednisolone/day), non-steroidal anti-in- α production in vitro.6 Furthermore, anti-TNFα treatment in flammatory drugs and analgesics were allowed at the discretion patients with RA resulted in a significant decrease of CD40L of the investigators; changes in dose were permitted. See online expression on T-cells.10 A similar effect was observed with syno- supplementary material for full inclusion/exclusion criteria. vial fluid macrophages, whereby CD40 ligation induced the secretion of interleukin (IL)-12, IL-10 and TNFα.11 12 CD40– Assessments CD40L pathway inhibition could impact multiple pathological All primary and secondary endpoint assessments were evaluated mechanisms in RA and dampen B-cell and T-cell responses and at week 12. The primary endpoint was the proportion of patients IL-12 production from dendritic cells and macrophages. who achieved a 20% improvement in the ACR criteria (ACR20). BI 655064 is a humanised, antagonistic anti-CD40 mono- Secondary endpoints included the proportion of patients clonal antibody that selectively binds CD40 and blocks the who achieved ACR50 or ACR70 response; European League CD40–CD40L interaction;13 two mutations in the Fc region Against Rheumatism (EULAR) response using DAS28-ESR and were introduced to prevent Fc-mediated antibody-dependent or DAS28-CRP, as defined by EULAR criteria; clinically mean- complement-mediated cellular cytotoxicity and platelet activa- ingful improvement in DAS28 (>1.2 decrease from baseline tion.13 BI 655064 demonstrated potent and comparable binding in DAS28-CRP); change in DAS28-CRP from baseline. Further properties in both human and cynomolgus monkey B-cells and endpoints included change from baseline in Simple Disease did not cause platelet activation, aggregation or function.13–15 Activity Index and Clinical Disease Activity Index. The objectives of this study were to characterise the efficacy, The safety and general tolerability of BI 655064 were evalu- safety and tolerability of weekly dosing of 120 mg BI 655064 in ated descriptively by monitoring adverse events (AEs), changes patients with RA with an inadequate response to methotrexate in vital signs, physical examinations, clinical laboratory tests (MTX-IR) and to elucidate the therapeutic mechanism of anti- and assessment of tolerability by the investigator. AE intensity CD40 antibody treatment. was graded according to the Rheumatology Common Toxicity Criteria V.2.0.17 Methods Blood samples for pharmacokinetic, immunogenicity, explor- Study design atory clinical biomarkers and pharmacogenomic analysis were This 12-week, multicentre, randomised, double-blind, paral- collected at prespecified visits (see online supplementary mate- lel-group, placebo-controlled, phase IIa study with an 8-week rials). Biomarkers associated with RA disease were assessed, follow-up period was conducted at 27 sites in six countries including CD95+ activated B-cell subpopulations, immunoglob- (Czech Republic, Germany, the Netherlands, New Zealand, ulins (IgG RF, IgM RF, IgA RF, IgG ACPA, total IgG, total IgM) Poland and Spain) between October 2013 and April 2015. and bone remodelling (RANKL, matrix metalloproteinase-3 Patients were randomised (via an Interactive Response System) [MMP3]) and inflammatory (IL-6) biomarkers. in a ratio of 2:1 to weekly subcutaneous 120 mg BI 655064 or placebo. Randomisation was stratified with respect to region Statistical analysis (Eastern Europe vs Western Europe/New Zealand). Sample size was determined based on a Bayesian approach The weekly subcutaneous 120 mg dose of BI 655064 was assuming an ACR20 response rate at week 12 of 65% in the BI selected based on safety, pharmacokinetic and pharmacodynamic 655064 group and 25% in the placebo group. data from the single and multiple dose studies.14 All subcuta- For binary endpoints, efficacy data were analysed using neous injections were given by staff at the study site. non-responder imputation; for non-binary endpoints, data were analysed using last observation carried forward. All effi- Patients cacy endpoints were analysed using the full analysis set (FAS) Patients aged 18–70 years were eligible if they had RA diagnosed and (one-sided) p values reported for comparisons of BI 655064 according to the American College of Rheumatology (ACR) 1987 versus placebo. The safety population included all patients who revised criteria;16 were MTX-IR and had received ≤2 anti-TNFα had received at least one dose of study drug. therapies (patients who were currently using or had previously The primary endpoint was analysed descriptively with a used a biological disease-modifying antirheumatic drug, such as Bayesian approach (details provided in the online supple- CTLA4-Ig, anti-IL-6 or anti-CD20 or an oral compound such mentary materials); additionally, a logistic regression model as Janus kinase inhibitors, were excluded); had active disease, was performed, including treatment, geographical region and

Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 755 Rheumatoid arthritis

Figure 1 Patient disposition. Disposition of the study patients treated with subcutaneous 120 mg BI 655064 or placebo administered once weekly for 12 weeks. Discontinuations due to AEs unrelated to RA were a fatal cerebral haemorrhage and one case of iron deficiency in the BI 655064 group and one case of pleural effusion, one case of elevated ALT and AST and one case of nasopharyngitis in the placebo group. *One patient excluded from full analysis set due to insufficient efficacy data; all 67 patients treated were included in the safety analysis set. AE, adverse event; ALT, alanine aminotransferase; AST, aspartate aminotransferase; RA, rheumatoid arthritis. previous use of anti-TNFα as covariates. Assessments of clinical to anti-TNFα therapy, who were all randomised to the BI 655064 biomarkers and pharmacogenomic analyses were exploratory. group, four achieved an ACR20 response and two achieved an Statistical analyses of secondary endpoints, clinical biomarkers ACR50 response. EULAR good/moderate responses (DAS28-CRP) and pharmacogenomic analyses are described in the online were achieved by 82.1% of patients in the BI 655064 group supplementary materials. compared with 70.0% in the placebo group (p=0.1894, figure 2B). Mean change in DAS28-CRP at week 12 was –1.61 in the BI 655064 Results group vs –1.45 in the placebo group (p=0.5463). In patients with Patient disposition and baseline characteristics CRP values above the median, the difference in DAS28-CRP A total of 67 patients were randomised and received treatment: between BI 655064 and placebo was more pronounced (–1.83 vs 44 patients were assigned to BI 655064 and 23 to placebo –1.38; p=0.2565; online supplementary table S1). (figure 1). All 67 patients were included in the safety analysis set; 66 patients were included in the FAS population (one patient receiving placebo was excluded from the FAS analysis due to insufficient efficacy data). Overall, eight patients discontinued Table 1 Baseline demographic and clinical characteristics treatment before week 12; the most common reason for discon- BI 655064 (n=44) Placebo (n=23) tinuation was AEs unrelated to RA, in five patients (7.5%). One patient in the placebo group discontinued because of worsening Age, years 53.7±13.3 55.1±8.3 of disease. Demographics at baseline were generally similar Female, n (%) 37 (84.1) 18 (78.3) across treatment groups (table 1), with the exception of baseline BMI 26.06±4.57 31.09±5.67 CRP values, which were lower in the BI 655064 group. Most Region, n (%) patients were anti-TNFα naive (n=61; 91%). Eastern Europe 28 (63.6) 15 (65.2) Western Europe 16 (36.4) 8 (34.8) Efficacy Ethnicity, n (%) An ACR20 response was seen for 30/44 patients (68.2%) in the BI White 43 (97.7) 23 (100) 655064 group and for 10/22 patients (45.5%) in the placebo group, Asian 1 (2.3) 0 yielding an observed difference of 22.7% (figure 2A). From the Disease duration, years 8.3±7.5 5.8±4.7 Bayesian analysis using an informative prior (an assumed response Anti-TNFα naive, n (%) 38 (86.4) 23 (100) rate of 25%) for the placebo group and a non-informative prior DAS28-CRP 5.39±0.85 5.53±0.79 for the BI 655064 group, the posterior mean difference between DAS28-ESR 6.21±0.73 6.16±0.80 the treatment groups was 33.0%; thus, the primary endpoint was CRP, mg/L 9.80±12.63 23.61±26.46 not met, as the observed difference was smaller than the expected ESR, mm/hour 37.45±18.67 44.45±23.76 difference. When the risk differences were adjusted for treatment, Duration of morning stiffness, mins 84.1±61.6 67.3±53.1 region and anti-TNFα history, the estimate of the adjusted differ- SJC 66 14.38±7.41 12.86±6.94 ence was 23.0% (95% CI –3.0% to 46.3%). TJC 68 21.25±11.87 19.77±10.84 ACR50 and ACR70 responses were achieved by 36.4% and 18.2% of patients in the BI 655064 group compared with 18.2% Except where indicated otherwise, values are mean±SD. BMI, body mass index; CRP, C-reactive protein; DAS28, Disease Activity Score in 28 and 13.6% of patients in the placebo group, respectively (differ- joints; ESR, erythrocyte sedimentation rate; SD, standard deviation; SJS 66, swollen ences were not statistically significant between BI 655064 and joint count based on 66 joints; TJC 68, tender joint count based on 68 joints; TNFα, placebo; figure 2A). Of the six patients with previous exposure tumour necrosis factor α.

756 Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 Rheumatoid arthritis

in median levels of total IgG and total IgM (figure 3); at weeks 8 and 12, the reductions were significant between BI 655064 vs placebo for both total IgG (–7.7% vs 0.2%; p=0.0026 and –8.6% vs –0.6%; p=0.0082, respectively) and total IgM (–10.2% vs 0.0%; p<0.0001 and –12.3% vs –0.3%; p=0.0002, respectively). At week 12, a greater reduction in the median levels of disease-spe- cific antibodies was observed in patients receiving BI 655064 vs placebo, with a significant reduction in IgG RF (–22.7% vs 14.3%; p=0.0018) and IgA RF (–13.9% vs 0.0%; p=0.0050); no change was observed in IgM RF (0.0% vs 0.0%; p=0.1743). IgG ACPA was reduced in both treatment groups at week 12 (–19.0% BI 655064 vs –11.9% placebo; p=0.2771). Levels of activated B-cells were modulated by BI 655064 (figure 4). BI 655064 was associated with greater reductions in the median percentage of circulating CD19+IgD−CD27−CD95+ B cells from baseline vs placebo over time and this change was statistically significant at week 12 (–14.6% vs 4.0%; p=0.0097). A reduction in the median percentage of circulating CD19+IgD− CD27+CD95+ B cells and CD19+IgD+CD27+CD95+ B cells from baseline was also observed for BI 655064 vs placebo, but the differences at week 12 did not reach statistical significance (–15.3% vs 2.3%; p=0.0765 and 16.8% vs –8.0%; p=0.0831, respectively).

Inflammatory and bone remodelling biomarkers The median changes in the levels of select biomarkers associated with inflammation (IL-6) and bone remodelling (MMP3 and RANKL) to week 12 are summarised in online supplementary figure S2 and table S4. BI 655064 was associated with a greater, but not statistically significant, decrease in IL-6 levels vs placebo at week 8 (–28.1% vs –2.75%; p=0.1332) and week 12 (–17.3% vs 18.7%; p=0.1001). Similarly, MMP3 levels were reduced in the BI 655064 group vs placebo at week 8 (–7.5% vs 4.6%; p=0.6176) and week 12 (–7.8% vs 2.3%; p=0.1574) but did not reach statis- tical significance. Levels of RANKL were significantly decreased Figure 2 Treatment response at week 12. Responses were defined following BI 655064 vs placebo at week 8 (–19.6% vs 1.8%; according to (A) ACR20/50/70 improvement criteria (FAS, non-responder p=0.0045) and week 12 (–29.4% vs 0.0%; p=0.0041). Analysis imputation) and (B) EULAR response (DAS28-CRP; FAS, observed). The of the ACR50 responder and non-responder subgroups identi- primary endpoint (ACR20 at week 12) was evaluated with a Bayesian fied a significant decrease in MMP3 levels versus non-responders approach. receiving BI 655064 at week 12 (–16.4% vs 13.6%; p=0.0005), ACR20/50/70, American College of Rheumatology 20/50/70% while a non-statistically significant reduction was observed in improvement criteria; DAS28-CRP, Disease Activity score in 28 joints IL-6 levels at week 12 in ACR50 responders vs non-responders based on C-reactive protein; EULAR, European League Against (–39.0% vs 3.9%; p=0.1215). The per cent change from baseline Rheumatism; FAS, full analysis set. in MMP3 and IL-6 in ACR20 responders versus non-responders at week 12 was –9.8% vs 3.4% (p=0.5292) and –7.4% vs –35.0% (p=0.2009), respectively. For ACR70 responders, the per cent A substantial decrease from baseline to week 12 in median ESR change from baseline in MMP3 and IL-6 levels was not calculated and pain-VAS in the BI 655064 group was observed compared as the number of patients was considered too small. with the placebo group (p=0.0288 and p=0.0456, respectively, online supplementary table S2). BI 655064 led to greater numer- ical, but not statistically significant, reductions across other ACR Pharmacogenomics core measures compared with the placebo group (online supple- The CD40 rs4810485 G/T polymorphism, which has previously mentary table S3). been linked to RA disease,18 was genotyped in all 46 patients who The impact of risk differences for the ACR20 response rate provided informed consent: 21 patients were homozygous for at week 12 based on baseline subcategories was evaluated and GG, 18 patients were heterozygous for GT and 5 patients were indicated a significant difference favouring BI 655064 treatment homozygous for TT. T-allele carriers treated with BI 655064 vs placebo for patients with a time since diagnosis <2.5 years showed a higher, but not statistically significant, ACR20 response (p=0.0091) (online supplementary figure S1). (72.7% vs 50.0%; p=0.1558) and significantly higher ACR50 response (54.5% vs 16.7%; p=0.0403) vs placebo at week 12. Biomarkers Similarly, BI 655064 led to a significant decrease in CD19+IgD− Changes in autoantibodies and B-cell subpopulations CD27+CD95+ (–24.2% vs 4.6%; p=0.0032) and CD19+IgD− As early as week 4, BI 655064 treatment led to greater, but CD27−CD95+ (–18.6% vs 7.0%; p=0.0022) B-cell subsets vs non-statistically significant, decreases from baseline vs placebo placebo in T-allele carriers. No significant reductions were observed

Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 757 Rheumatoid arthritis

Figure 3 Median per cent change from baseline to week 12 in autoantibody levels. Median per cent change from baseline to week 12 in levels of total IgG, IgM and IgG ACPA. FAS, observed. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ACPA, anticyclic citrullinated protein antibody; FAS, full analysis set; Ig, immunoglobulin. in CD19+IgD+CD27+CD95+ B-cell levels following BI 655064 steady-state was achieved at weeks 9–12; a correlation between treatment in T-allele carriers (online supplementary table S5). BI 655064 exposure and clinical response could not be clarified due to high pharmacokinetic variability. Overall, five patients Safety (22.7%) receiving placebo and six patients (13.3%) receiving BI There was a higher rate of AEs reported in patients receiving 655064 tested positive for anti-drug antibodies at any time (all placebo (78.3%) vs BI 655064 (65.9%; table 2). The most titres ≤8). frequently reported AEs in both groups were nasopharyngitis and headache. Discussion Of the 44 patients receiving BI 655064, serious AEs (SAEs) In this study, the primary endpoint was not met; BI 655064, a were reported in two patients: one case with pre-existing hyper- targeted therapy directed against CD40, led to improvements tension who developed a cerebral haemorrhage and cardiopul- over placebo in ACR responses at week 12, but these were monary failure leading to death and one case of myocardial not statistically significant. BI 655064 did, however, show infarction, which occurred 5 weeks after the last dose was some modulation of select inflammatory and bone resorp- administered. One other patient treated with BI 655064 discon- tion biomarkers and autoreactive activated B-cells linked tinued due to non-treatment-related iron deficiency anaemia. to the pathogenesis of RA,with significant changes in IgG AEs in all other patients (93.1%) receiving BI 655064 were of RF (p=0.0018), IgA RF (p=0.0050), RANKL (p=0.0041) grade 1 or 2. and activated CD19+IgD−CD27−CD95+ memory B-cells Of the 23 patients receiving placebo, SAEs were reported in (p=0.0097) at week 12 compared with placebo. Interest- two patients: one case of pleural effusion requiring hospital- ingly, the rs4810485 T-allele, which is linked to RA and isation and one case of medically significant anaemia. In total, systemic lupus erythematosus (SLE) susceptibility, was asso- four patients (17.4%) receiving placebo discontinued: one case ciated with larger reductions in select memory CD19+B- each of pleural effusion, nasopharyngitis, elevated liver enzymes cell subsets (CD19+IgD−CD27+CD95+; p=0.0032 and and worsening RA. There were no clinically relevant changes in CD19+IgD−CD27−CD95+; p=0.0022) post-treatment safety-related laboratory parameters between groups (table 2). with BI 655064.19–22 In this study, the GT/TT-allele was asso- With the exception of decreases in ESR, CRP and RF, there were ciated with greater improvements in ACR50 responses with BI no relevant changes in the frequencies of patients with values 655064 treatment vs placebo (p=0.0403). above or below normal limits. Assessments of local tolerability This study had several limitations, but there were three indicated similar signs and symptoms between groups. Baseline major issues that greatly affected the outcome of this study: the and postdose rates of findings were also similar, indicating no study size, characteristics of the patient population included in substantial change due to administration of study treatment. the study and the placebo response. Based on historical data, There was a three to fourfold accumulation of BI 655064 at the expected ACR20 placebo response rate was predicted to the end of treatment compared with that after a single dose. be 25%; the Bayesian approach used in this study estimated Although BI 655064 trough concentrations were highly variable, that a sample size of 66 patients would be effective. However,

Figure 4 Median per cent change from baseline to week 12 in levels of CD95+ activated CD19+ B-cell subsets. Median per cent change from baseline to week 12 in B-cell subsets CD19+IgD−CD27−CD95+, CD19+IgD−CD27+CD95+ and CD19+IgD+CD27+CD95+. FAS, observed. *p≤0.01. FAS, full analysis set.

758 Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 Rheumatoid arthritis

that it was not possible to determine whether there was a dose Table 2 Safety profile effect in this study. A single subcutaneous dose of 120 mg Placebo BI 655064 in healthy volunteers resulted in >90% receptor BI 655064 (n=44) (n=23) occupancy and >90% inhibition of CD54 upregulation, which Any AE 29 (65.9) 18 (78.3) was maintained for 1 week.14 Furthermore, multiple subcuta- Other significant AEs* 1 (2.3) 3 (13.0) neous dosing of 120 mg BI 655064 once weekly in healthy AEs leading to drug discontinuation 2 (4.5) 4 (17.4) volunteers resulted in about fourfold increase in exposure Serious AEs 2 (4.5) 2 (8.7) compared with a single dose.15 The 120 mg dose was, there- Death 1 (2.3)† 0 fore, anticipated to provide efficacy with a balanced safety Myocardial infarction 1 (2.3) 0 profile; however, data obtained after the initiation of this trial Pleural effusion 0 1 (4.3) suggest that it may take up to 12 weeks to reach steady-state 15 Anaemia 0 1 (4.3) when dosing 120 mg BI 655064 once weekly. This finding is RCTC AE grade ≥3 2 (4.5) 1 (4.3) supported in this study whereby pharmacokinetic steady-state 26 Common AEs‡ for BI 655064 was reached within 10–12 weeks and that a Nasopharyngitis 6 (13.6) 5 (21.7) loading dose is potentially needed in order to achieve steady- Diarrhoea 3 (6.8) 0 state more rapidly. Taking these considerations into account, it Fatigue 0 2 (8.7) is uncertain whether a study with a greater sample size, higher CRP and higher doses or a loading dose of BI 655064 may Headache 3 (6.8) 3 (13.0) have resulted in greater efficacy in this patient population. Arthralgia 2 (4.5) 2 (8.7) Former clinical trials of anti-CD40L antibodies were halted Liver disorder 2 (4.5) 2 (8.7) because of increased incidence of thromboembolism, initiated Laboratory AEs by activation and aggregation of platelets, possibly because of RCTC grade ≥3 0 0 the Fc region of anti-CD40L antibodies activating the Fc RIIa § γ RCTC grade 2 2 (4.5) 3 (13.0) (CD32a) platelet receptor;27 anti-CD40L antibodies lacking a ¶ RCTC grade 1 2 (4.5) 1 (4.3) functional Fc domain were shown to retain pharmacological Data are n (%). AEs were coded using MedDRA V.18.0. activity but were not associated with platelet activation.28–30 *Per ICH E3 Guideline; in the BI 655064 group there was one case of iron deficiency In this study, BI 655064 showed a good safety and tolerability anaemia and in the placebo group there was one case of elevated liver enzymes, one case of worsening RA and one case of nasopharyngitis. profile with no thromboembolic events, demonstrating that †Patient had pre-existing hypertension and developed cerebral haemorrhage and targeting CD40 alone in an autoimmune disease such as RA cardiopulmonary failure. produces a favourable safety profile. ‡Common AEs are reported by preferred term for ≥5% of patients in any treatment The biomarker results of this small study of BI 655064 in group (ie, for ≥3 patients in the BI 655064 group or ≥2 patients in the placebo patients with RA have provided insights into key aspects asso- group). § ciated with CD40 blockade and may have implications for In the BI 655064 group, there was one case of iron deficiency anaemia and one case of hypoglycaemia and in the placebo group there was one case of anaemia, other autoimmune diseases. The CD40–CD40L pathway has one case of elevated liver enzymes and one case of elevated uric acid. been shown to play an important role in the pathogenesis of ¶In the BI 655064 group, there was one case of anaemia and one case of elevated SLE and lupus nephritis (LN).31 Patients with LN have been liver enzymes and in the placebo group there was one case of hyperglycaemia. shown to have a high renal expression of CD40 and CD40L AE, adverse event; ICH, International Conference on Harmonisation; MedDRA, and increased expression of CD40L in autoreactive B-cells, Medical Dictionary for Regulatory Activities; RA, rheumatoid arthritis; RCTC, 32 Rheumatology Common Toxicity Criteria (V.2.0). resulting in spontaneous autoantibody production. Addi- tionally, increased CD40 renal expression correlates with LN activity.32 Consequently, targeting the CD40–CD40L pathway could be an effective approach for the treatment of autoim- this small sample size, combined with the BI 655064 group mune diseases, including RA, SLE and LN. Ongoing clinical having a considerably lower mean CRP than the placebo trials are assessing the safety and efficacy of BI 655064 in LN; group, may have influenced the higher than expected placebo however, further development for RA is not currently planned. response rates. This imbalance may have derived from a In conclusion, blockade of the CD40–CD40L pathway change in inclusion criteria based on a protocol amendment, with BI 655064 in MTX-IR patients with RA was associated from patients initially requiring a CRP ≥0.8 mg/dL (ULN of with a favourable clinical safety profile, however, the primary assay 0.6 mg/dL) at screening, to requiring a CRP ≥0.8 mg/ endpoint at week 12 was not met. dL or ESR ≥28 mm/hour. This amendment was made to aid recruitment into the study, however, this inadvertently led to Author affiliations more patients with lower or normal CRP levels being enrolled. 1Boehringer Ingelheim Pharmaceuticals Inc, Ridgefield, Connecticut, USA Normal CRP values at baseline were observed in 54.5% of 2ClinicMed Badurski and Partners, Białystok, Poland patients in the BI 655064 group, which, together with the 3Rheumatica, Warsaw, Poland 4Justus Liebig University Giessen, Bad Nauheim, Germany longer duration of disease, may have reduced the differentia- 5 tion observed in DAS28-CRP at week 12; this is supported by Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach, Germany 6CTCenter MaVe, Olomouc, Czech Republic improved DAS28-CRP differentiation in patients with baseline 7Centre for Inflammatory Joint Diseases, Munich, Germany CRP values above the median. Furthermore, the high ACR20 8Medical Plus, s.r.o, Uherské Hradiště, Czech Republic placebo response rate at week 12 may be a reflection of the 9Faculty of Pharmacy, University of Veterinary and Pharmaceutical Sciences, Brno, Czech Republic patient population, as similar results have been observed in RA 10 trials that enrolled MTX-IR patients.23–25 In the current study, Prof. Eleonora Reicher Memorial National Institute of Geriatrics, Rheumatology and Rehabilitation, Warsaw, Poland BI 655064 treatment led to a similar ACR20 response rate to 11Schlosspark-Klinik, Berlin, Germany that observed previously with TNFα inhibitors. Finally, only 12Auckland Clinical Studies, Auckland, New Zealand investigating a single active dose of BI 655064 (120 mg) meant 13Boehringer Ingelheim International GmbH, Ingelheim, Germany

Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 759 Rheumatoid arthritis

Acknowledgements We thank all patients and study investigators who 9 Rissoan MC, Van Kooten C, Chomarat P, et al. The functional CD40 antigen of participated in the clinical study described here, and Susanne Svoboda for facilitating fibroblasts may contribute to the proliferation of rheumatoid synovium. Clin Exp the analysis of protein and cellular biomarkers included in this study. Immunol 1996;106:481–90. 10 Tung C-H, Lu M-C, Lai N-S, et al. Tumor necrosis factor- blockade treatment Contributors SV, UM-L, MR, BR, AGE, RV, PB, DJ, JSF, SJP and JS contributed to α decreased CD154 (CD40-ligand) expression in rheumatoid arthritis. PLoS One study design. SD, RP, AP, HK, ED, BK, RA and CS contributed to data collection. SV, 2017;12:e0183726. AGE, RV, PB, DJ and JS contributed to data analysis. All authors had full access to the 11 Möttönen M, Isomäki P, Luukkainen R, et al. Regulation of CD154-induced study data, contributed to data analysis, data interpretation, writing and review of interleukin-12 production in synovial fluid macrophages. Arthritis Res 2002;4. the manuscript and approved the final version for publication. 12 Harigai M, Hara M, Kawamoto M, et al. Amplification of the synovial inflammatory Funding This study was funded by Boehringer Ingelheim. Professional medical response through activation of mitogen-activated protein kinases and nuclear writing support in the development of this manuscript, under the guidance of the factor kappaB using ligation of CD40 on CD14+ synovial cells from patients with authors, was provided by Tina Borg and Leigh Church of OPEN Health Medical rheumatoid arthritis. Arthritis Rheum 2004;50:2167–77. Communications (London, UK) and was funded by Boehringer Ingelheim. 13 Ralph K, Nicoletti A, Musvasva E, et al. THU0407 Preclinical Characterization Competing interests SD, RP, AP, HK, ED and BK report no disclosures; UM-L of a Highly Selective and Potent Antagonistic Anti-CD40 mAb. Ann Rheum Dis reports being an advisor for Boehringer Ingelheim; RA and CS report having received 2015;74(Suppl 2):344.1–344. research/grant support from Boehringer Ingelheim; SV, MR, BR, AGE, RV, PB, DJ, JSF, 14 albach FN, Wagner F, Hüser A, et al. Safety, pharmacokinetics and pharmacodynamics SJP and JS report being employed by Boehringer Ingelheim. of single rising doses of BI 655064, an antagonistic anti-CD40 antibody in healthy subjects: a potential novel treatment for autoimmune diseases. Eur J Clin Pharmacol Patient consent for publication Not required. 2018;74:161–9. Ethics approval The study protocol was approved by the institutional review 15 schwabe C, Rosenstock B, Doan T, et al. Safety, Pharmacokinetics, and board or ethics committee at each participating centre. The study was conducted Pharmacodynamics of Multiple Rising Doses of BI 655064, an Antagonistic Anti-CD40 according to the Declaration of Helsinki and the International Conference on Antibody, in Healthy Subjects: A Potential Novel Treatment for Autoimmune Diseases. Harmonisation guidelines. All patients provided written informed consent. J Clin Pharmacol 2018;58:1566–77. 16 arnett FC, Edworthy SM, Bloch DA, et al. The American rheumatism association Provenance and peer review Not commissioned; externally peer reviewed. 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum Data sharing statement To ensure independent interpretation of clinical study 1988;31:315–24. results, Boehringer Ingelheim grants all external authors access to all relevant 17 Woodworth T, Furst DE, Alten R, et al. Standardizing assessment and reporting of material, including participant-level clinical study data and relevant material as adverse effects in rheumatology clinical trials II: the rheumatology common toxicity needed by them to fulfil their role and obligations as authors under the ICMJE criteria v.2.0. J Rheumatol 2007;34:1401–14. criteria. Furthermore, clinical study documents (e.g. study report, study protocol, 18 lee YH, Bae S-C, Choi SJ, et al. Associations between the functional CD40 rs4810485 statistical analysis plan) and participant clinical study data are available to be G/T polymorphism and susceptibility to rheumatoid arthritis and systemic lupus shared after publication of the primary manuscript in a peer-reviewed journal erythematosus: a meta-analysis. Lupus 2015;24:1177–83. and if regulatory activities are complete and other criteria met per the BI Policy 19 Gaffney PM, Ortmann WA, Selby SA, et al. Genome screening in human systemic on Transparency and Publication of Clinical Study Data: https://​trials.​boehringer-​ lupus erythematosus: results from a second Minnesota cohort and combined analyses ingelheim.​com/​transparency_​policy.​html Prior to providing access, documents will be of 187 sib-pair families. Am J Hum Genet 2000;66:547–56. examined, and, if necessary, redacted and the data will be de-identified to protect 20 Rodríguez-Rodríguez L, Castañeda S, Vázquez-Rodríguez TR, et al. Influence of CD40 the personal data of study participants and personnel, and to respect the boundaries rs1883832 polymorphism in susceptibility to and clinical manifestations of biopsy- of the informed consent of the study participants. Clinical Study Reports and Related proven giant cell arteritis. J Rheumatol 2010;37:2076–80. Clinical Documents can be requested via this link:https://​trials.​boehringer-​ingelheim.​ 21 Vazgiourakis VM, Zervou MI, Choulaki C, et al. A common SNP in the CD40 region com/trial_​ ​results/clinical_​ ​submission_​documents.​html All such requests will be is associated with systemic lupus erythematosus and correlates with altered CD40 governed by a Document Sharing Agreement. Bona fide, qualified scientific and expression: implications for the pathogenesis. Ann Rheum Dis 2011;70:2184–90. medical researchers may request access to de-identified, analysable participant 22 Raychaudhuri S, Remmers EF, Lee AT, et al. Common variants at CD40 and other loci clinical study data with corresponding documentation describing the structure and confer risk of rheumatoid arthritis. Nat Genet 2008;40:1216–23. content of the datasets. Upon approval, and governed by a Data Sharing Agreement, 23 Huizinga TWJ, Fleischmann RM, Jasson M, et al. Sarilumab, a fully human monoclonal data are shared in a secured data-access system for a limited period of 1 year, which antibody against IL-6Rα in patients with rheumatoid arthritis and an inadequate may be extended upon request.Researchers should use https://​clin​ical​stud​ydat​ response to methotrexate: efficacy and safety results from the randomised SARIL-RA- arequest.​com to request access to study data. MOBILITY Part A trial. Ann Rheum Dis 2014;73:1626–34. 24 Genovese MC, Smolen JS, Weinblatt ME, et al. Efficacy and safety ofA BT-494, Open access This is an open access article distributed in accordance with the a selective JAK-1 inhibitor, in a phase IIb study in patients with rheumatoid Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which arthritis and an inadequate response to methotrexate. Arthritis Rheumatol permits others to distribute, remix, adapt, build upon this work non-commercially, 2016;68:2857–66. and license their derivative works on different terms, provided the original work is 25 Kivitz AJ, Gutierrez-Ureña SR, Poiley J, et al. Peficitinib, a AJ K inhibitor, in the properly cited, appropriate credit is given, any changes made indicated, and the use treatment of moderate-to-severe rheumatoid arthritis in patients with an inadequate is non-commercial. See: http://​creativecommons.org/​ ​licenses/​by-​nc/4.​ ​0/. response to methotrexate. Arthritis Rheumatol 2017;69:709–19. 26 Daniluk S, Ptaszynski R, Mueller-Ladner U, et al. SAT0147 Safety and Efficacy of BI 655064, An Antagonistic Anti-CD40 Antibody in Rheumatoid Arthritis (RA) Patients: References Table 1. Ann Rheum Dis 2016;75(Suppl 2):718.1–718. 1 Büchner K, Henn V, Gräfe M, et al. CD40 ligand is selectively expressed on CD4+ T 27 Boumpas DT, Furie R, Manzi S, et al. A short course of BG9588 (anti-CD40 ligand cells and platelets: implications for CD40-CD40L signalling in atherosclerosis. J antibody) improves serologic activity and decreases hematuria in patients with Pathol 2003;201:288–95. proliferative lupus glomerulonephritis. Arthritis Rheum 2003;48:719–27. 2 Desai-Mehta A, Lu L, Ramsey-Goldman R, et al. Hyperexpression of CD40 ligand by 28 shock A, Burkly L, WakefieldI , et al. CDP7657, an anti-CD40L antibody lacking B and T cells in human lupus and its role in pathogenic autoantibody production. J an Fc domain, inhibits CD40L-dependent immune responses without thrombotic Clin Invest 1996;97:2063–73. complications: an in vivo study. Arthritis Res Ther 2015;17. 3 Berner B, Wolf G, Hummel KM, et al. Increased expression of CD40 ligand (CD154) 29 Tocoian A, Buchan P, Kirby H, et al. First-in-human trial of the safety, on CD4+ T cells as a marker of disease activity in rheumatoid arthritis. Ann Rheum pharmacokinetics and immunogenicity of a PEGylated anti-CD40L antibody fragment Dis 2000;59:190–5. (CDP7657) in healthy individuals and patients with systemic lupus erythematosus. 4 MacDonald KP, Nishioka Y, Lipsky PE, et al. Functional CD40 ligand is expressed by T Lupus 2015;24:1045–56. cells in rheumatoid arthritis. J Clin Invest 1997;100:2404–14. 30 Chamberlain C, Colman PJ, Ranger AM, et al. Repeated administration of 5 Kyburz D, Corr M, Brinson DC, et al. Human rheumatoid factor production is dapirolizumab pegol in a randomised phase I study is well tolerated and accompanied dependent on CD40 signaling and autoantigen. J Immunol 1999;163:3116–22. by improvements in several composite measures of systemic lupus erythematosus 6 liu MF, Chao SC, Wang CR, et al. Expression of CD40 and CD40 ligand among disease activity and changes in whole blood transcriptomic profiles. Ann Rheum Dis cell populations within rheumatoid synovial compartment. Autoimmunity 2017;76:1837–44. 2001;34:107–13. 31 Peters AL, Stunz LL, Bishop GA. CD40 and autoimmunity: the dark side of a great 7 Koshy M, Berger D, Crow MK. Increased expression of CD40 ligand on systemic lupus activator. Semin Immunol 2009;21:293–300. erythematosus lymphocytes. J Clin Invest 1996;98:826–37. 32 Yellin MJ, D’Agati V, Parkinson G, et al. Immunohistologic analysis of renal CD40 and 8 Qamar N, Fuleihan RL. The hyper IgM syndromes. Clin Rev Allergy Immunol CD40L expression in lupus nephritis and other glomerulonephritides. Arthritis Rheum 2014;46:120–30. 1997;40:124–34.

760 Visvanathan S, et al. Ann Rheum Dis 2019;78:754–760. doi:10.1136/annrheumdis-2018-214729 Rheumatoid arthritis

Translational science Synovial cellular and molecular signatures stratify clinical response to csDMARD therapy and predict radiographic progression in early rheumatoid arthritis patients Frances Humby,‍ ‍ 1 Myles Lewis,‍ ‍ 1 Nandhini Ramamoorthi,2 Jason A Hackney,3 Michael R Barnes,1,4 Michele Bombardieri,1 A. Francesca Setiadi,2 Stephen Kelly,1 Fabiola Bene,1 Maria DiCicco,1 Sudeh Riahi,1 Vidalba Rocher,1 Nora Ng,1 Ilias Lazarou,1 Rebecca Hands,1 Désirée van der Heijde,‍ ‍ 5 Robert B M Landewé,6,7 Annette van der Helm-van Mil,5 Alberto Cauli,8 Iain McInnes,9 Christopher Dominic Buckley,10 Ernest H Choy,11 Peter C Taylor,‍ ‍ 12 Michael J Townsend,2 Costantino Pitzalis‍ ‍ 1

Handling editor Josef S Abstract Key messages Smolen Objectives To unravel the hierarchy of cellular/molecular pathways in the disease tissue of early, treatment-naïve For numbered affiliations see What is already known about this subject? end of article. rheumatoid arthritis (RA) patients and determine their ►► Disease tissue heterogeneity is postulated to relationship with clinical phenotypes and treatment play a major role in incomplete drug Correspondence to response/outcomes longitudinally. response related to diverse target Professor Costantino Pitzalis, Methods 144 consecutive treatment-naïve early RA Centre for Experimental expression levels and drug pharmacology. Medicine & Rheumatology, patients (<12 months symptoms duration) underwent In established/long-standing RA distinct William Harvey Research ultrasound-guided synovial biopsy before and 6 months histological and molecular subtypes have been Institute, Barts & The London after disease-modifying antirheumatic drug (DMARD) described. School of Medicine & Dentistry, initiation. Synovial biopsies were analysed for cellular Queen Mary University of ►► However, their relationship with clinical London, London, UK; (immunohistology) and molecular (NanoString) phenotypes (e.g. disease activity/ c.​ ​pitzalis@qmul.​ ​ac.​uk and Dr characteristics and results compared with clinical and progression) and treatment response remains Michael J Townsend, Biomarker imaging outcomes. Differential gene expression analysis controversial. Discovery OMNI, Genentech and logistic regression were applied to define variables Research and Early Development, South San correlating with treatment response and predicting What does this study add? Francisco, USA; radiographic progression. ►► This study provides: first-time evidence in townsend.​ ​michael@gene.​ ​com Results cellular and molecular analyses of synovial tissue early, treatment-naive RA patients of the demonstrated for the first time in early RA the presence existence in the synovium of three specific FH, ML, NR and JAH contributed equally. of three pathology groups: (1) lympho-myeloid dominated pathology subgroups (Lympho-myeloid, by the presence of B cells in addition to myeloid cells; (2) Diffuse-myeloid and Pauci-immune-fibroid), MJT and CP are joint senior diffuse-myeloid with myeloid lineage predominance but prior to the potential modification of authors. poor in B cells nd (3) pauci-immune characterised by scanty disease pathology by immune-modulatory Received 6 October 2018 immune cells and prevalent stromal cells. Longitudinal therapies. Revised 6 February 2019 correlation of molecular signatures demonstrated that ►► In addition, it demonstrates that specific Accepted 14 February 2019 elevation of myeloid- and lymphoid-associated gene histopathotypes and associated transcriptional Published Online First expression strongly correlated with disease activity, acute endotypes define distinct RA subtypes that are 16 March 2019 phase reactants and DMARD response at 6 months. linked to diverse clinical phenotypes, disease Furthermore, elevation of synovial lymphoid-associated activity/severity and treatment response and genes correlated with autoantibody positivity and elevation that gene expression signatures associated of osteoclast-targeting genes predicting radiographic with cellular infiltration in synovium strongly joint damage progression at 12 months. Patients with correlate with imaging modalities including predominant pauci-immune pathology showed less severe ultrasonographic measures of synovial disease activity and radiographic progression. thickness and power Doppler and erosive load on x-ray. © Author(s) (or their Conclusions We demonstrate at disease presentation, ► Moreover, longitudinal analysis demonstrated employer(s)) 2019. Re-use prior to pathology modulation by therapy, the presence ► permitted under CC BY. of specific cellular/molecular synovial signatures that that specific synovial signatures are associated Published by BMJ. delineate disease severity/progression and therapeutic with response to DMARD therapy, clinical outcome at 6 months and radiographic joint To cite: Humby F, response and may pave the way to more precise definition Lewis M, Ramamoorthi N, of RA taxonomy, therapeutic targeting and improved damage at 12 months. et al. Ann Rheum Dis outcomes. 2019;78:761–772.

Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 761 Rheumatoid arthritis

showed less severe disease activity and radiographic progression Key messages but, importantly, lower therapeutic response. How might this impact on clinical practice? ►► This study demonstrates that the cellular and molecular Methods signatures define pathobiological endotypes early in the Pathobiology of Early Arthritis Cohort disease process, prior to treatment modification, that have In all, 144 RA patients fulfilling 2010 American College of Rheu- a significant impact on disease prognosis and treatment matology/European League Against Rheumatism (EULAR) Clas- outcome. sification Criteria, with clinically defined synovitis and symptom ►► This offers the potential for a more accurate patient duration less than 12 months, were enrolled as part of the ‘Patho- stratification of this severe disabling disease where early biology of Early Arthritis Cohort’ (PEAC, http://www.peac-​ ​mrc.​ disease modification is crucial to life course outcome and mds.​qmul.​ac.uk)​ at three UK Academic Centres: Queen Mary quality of life. University of London/Barts Health NHS trust, University of Glasgow and University of Birmingham. All patients were naïve to steroid and DMARD therapy. Patients underwent a ultrasound (US)-guided synovial biopsy (figure 1A) procedure we pioneered6 of a clinically active joint selected according to a previously defined algorithm to ensure Introduction maximal synovial tissue retrieval6; a minimum of 12 synovial Better understanding of rheumatoid arthritis (RA) disease biopsies were stored for subsequent analysis at the William pathogenesis has led to the development of highly effective Harvey research Institute (six for histological analysis and six therapeutics inhibiting structural damage and improving prog- 1 for RNA extraction), and patients then commenced on stan- nosis ; however, a sizeable proportion of patients (~40%) do dard DMARD therapy and/or low-dose corticosteroid. A treat- not respond to these therapies and the mechanisms for response/ to-target approach to therapy escalation was followed aiming nonresponse remain unknown. Additionally, as currently it is not for low disease activity score-28 (DAS28) <3.2. Patients failing possible to predict which patients would benefit from individual DMARD therapy were commenced on biological therapy therapeutic modalities, this leaves a huge unmet clinical need. according to the UK National Institute for Clinical Excellence Disease tissue heterogeneity is postulated to play a major prescribing algorithm for RA patients if they continued to have a role in incomplete drug response related to diverse target DAS28 >5.1 at 6 months. expression levels and drug pharmacology. In established/long- Ultrasonographic images were collected at the time of biopsy standing RA, distinct histological and molecular subtypes have 2 3 for both the individual biopsied joint and the global joint score: been described. However, their relationship with clinical first to fifth metacarpophalangeal (MCP) joints and midline, phenotypes (eg, disease activity/progression) and treatment radial and ulnar views of both wrist joints. Images subsequently response remains controversial. The reasons for such discrep- 3 underwent semiquantitative assessment by a single blinded (to ancies (reviewed ) include multiple confounding factors clinical/histological data) assessor for both synovial thickening ranging from (i) small sample size (mostly 20–40 patient (ST) and power Doppler (PD) activity according to standard cohorts), (ii) reliance on cross-sectional rather than prospec- EULAR-OMERACT (Outcome Measures in Rheumatology) tive data and (iii) inclusion of patients at different disease US synovitis scores (grade 0–3).7 The mean global ST and PD stages/severity and treated with multiple therapies potentially 3 scores including the maximal score for the wrist joint were then influencing synovial pathobiology. determined. In addition, many studies have been performed in established/ Plain radiographs of the hands and feet performed at baseline late-stage disease and with a sampling bias due to prevalent repre- and 12-month follow-up were scored in time sequential order sentation of large joints (arthroscopy/joint replacement) while 3 according to the van der Heijde modified Sharp score (SHSS) only a handful of studies have included small joints. Further- by a single reader blinded to all clinical/histological data. The more, though molecular characterisation of synovial tissue has 4 5 study received local ethical approval (REC-05/Q0703/198) and been recently published in established/long-standing RA, here all participants gave written informed consent. we report the first systematic cellular (immunohistology) and molecular (NanoString) characterisation in early disease, treat- ment-naïve patients. Histology We demonstrate at disease presentation, prior to potential Paraffin-embedded H&E stained sections were graded as suit- modification by therapy, the presence of three distinct pathology able for further histopathological assessment if intact lining layer groups: (1) lympho-myeloid dominated by lymphoid lineage was identified. Following immunohistochemical staining for B cells (CD20), T cells (CD3), macrophages (CD68) and plasma infiltration (T cells, B cells, plasma cells) in addition to myeloid 8 cells; (2) diffuse-myeloid with myeloid lineage predominance cells (CD138) as previously reported, the degree of immune but poor in B cells/plasma cells and (3) Pauci-immune character- cell infiltration was determined semiquantitatively (0–4). Biop- ised by scanty immune cells and prevalent stromal cells. Longi- sies were then stratified into one of the three synovial groups tudinal correlation of molecular signatures demonstrated that according to the following criteria: (i) lympho-myeloid presence elevation of myeloid and lymphoid-associated gene expression of grades 2–3—CD20+ aggregates, (CD20 ≥2) and/or CD138 strongly correlated with disease activity, acute phase reactants ≥2; (ii) diffuse-myeloid—CD68 SL ≥2, CD20 ≤1 and/or CD3 and conventional synthetics disease-modifying antirheumatic ≥1, CD138 ≤2 and (iii) pauci-immune-fibroid—CD68 SL <2 drugs (DMARDs) response at 6 months. Furthermore, elevation and CD3, CD20, CD138 <1. of synovial lymphoid-associated genes correlated with autoan- tibody positivity and the elevation of osteoclast-targeting genes NanoString gene expression analysis and predicted radiographic joint damage progression at 12 NanoString Panels Construction. Pathotype-specific NanoString months. Patients with predominant pauci-immune pathology panels were developed based on a previous microarray study2

762 Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 Rheumatoid arthritis

Figure 1 (A) Representative image of an US-guided wrist biopsy. Inset: greyscale transverse US image of wrist joint demonstrating biopsy needle entering joint space under extensor tendon complex. Synovial tissue fragment (encircled) in biopsy needle. (B) Patient characteristics of cohort. (C) Number and type of joints biopsied. (D) Synovial pathotype according to joint biopsied. (E) H&E staining and immunohistochemistry of synovial biopsies for CD20+ B cells, CD3+ T cells, CD68+ macrophages in synovial lining/sublining layers and CD138+ plasma cells from treatment-naïve individuals with early rheumatoid arthritis. Synovial biopsies were categorised as lympho-myeloid, diffuse myeloid or pauci-immune fibroid. CCP, cyclic citrullinated peptide; CRP, C-reactive protein; DAS28, disease activity score-28; ESR, erythrocyte sedimentation rate; MTP, metatarsophalangeal; MCP, metacarpophalangeal; PIP, proximal interphalangeal; RF, rheumatoid factor; SHSS, van der Heijde modified Sharp score; SJ, swollen joints; TJ, tender joints; US, ultrasound; VAS, visual analogue scale.

(GEO accession number GSE48780). Differential expression diffuse-myeloid and lympho-myeloid groups. Genes were analysis was performed using the limma package.9 We performed selected for a pathotype if they showed differential expression all pairwise comparisons between samples in the pauci-immune, with each of the other two pathotypes at a Benjamini-Hochberg

Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 763 Rheumatoid arthritis adjusted p value <0.01 (online supplementary figure 1A–C and optimised by 10-fold cross-validation. Lambda tuned to the online supplementary tables 1–3). These genes were validated smallest mean cross-validated error was retained as final penalty in the baseline RNA-seq data (deposited in ArrayExpress acces- parameter in the model. sion code E-MTAB-6141) from the PEAC cohort. For each set Predictive performance evaluation: The predictive perfor- of pathotype-specific genes, we identified the 50 genes that best mance of the models with and without the genes (respec- correlated with the first principal component of the z-score tively, LASSO and backward stepwise models) was assessed by transformed expression data for that gene set and included computing the area under the receiver operating character- an additional set of 87 genes implicated in RA pathobiology istic curve (AUC). Both apparent and internally validated AUC (online supplementary table 3). We confirmed that expression were assessed. To compensate for overfitting inherent to the levels determined by NanoString were concordant with those apparent AUC, internal validation was performed to correct the measured by RNA sequencing, in samples where both measure- measure of predictive performance for optimism using bootstrap ments were available (online supplementary figure 2). (repeated 500 times) with the R-package boot V.1.3–18, as a NanoString Analysis. Total RNA was extracted from synovial valid method to generate unbiased optimism-adjusted estimates tissue using TRIzol Reagent (ThermoFisher Scientific, Life Tech- of the C statistic (AUC) with small absolute errors. nologies, Invitrogen Division, UK) as previously described.6 A custom CodeSet of capture and reporter probes was designed Results to target regions of 100 nucleotides of 242 genes. Hybridisa- Patient cohort tion of 50–100 ng of synovial RNA was carried out according Patient characteristics are summarised in figure 1B. Briefly, most to the manufacturer’s instructions. Raw expression data were patients had high disease activity (mean DAS28 5.6±SD 1.5), obtained using the NanoString nCounter MAX analysis system mean disease duration 5.6 months (SD 3.2), approximately 65% (NanoString Technologies, Seattle, WA, USA) according to the were positive for rheumatoid factor (RF) and/or anti-citrullinated manufacturer’s instructions. Analyses of NanoString expression peptide antibodies (ACPA), and 20% had at least one radiographic data were performed using R (V.3.3.2). For differential expres- erosion. The joint biopsied most frequently was the wrist (~65%), sion analyses, we used the limma Bioconductor package9 using 10 with good additional representation from MCP/proximal inter- default settings. We used the Benjamini-Hochberg method to phalangeal/metatarsophalangeal joints (~15%) and knee, elbow adjust for multiple testing, and considered genes to be differen- together ~20% (figure 1C). tially expressed if they had an adjusted p value <0.01. Eigengene scores were calculated as previously described,11 with statistical adjustment for biopsy joint position. The osteoclast gene set Identification of synovial pathotypes by histopathology utilised was described in the Harmonizome database (http://​ In total, 129 of the 144 recruited patients (89.6%) had synovial amp.​pharm.​mssm.​edu/​Harmonizome/). tissue suitable for subsequent histological analysis (failure rate 10.4%). Using the algorithm described in the Methods, we clas- sified patients into three distinct synovial pathological groups Serum biomarker assessments (figure 1D): (1) lympho-myeloid (n=51, 39%) dominated by Serum samples from 111 patients at baseline were assessed for lymphoid lineage infiltration (T cells, B cells, plasma cells) in levels of soluble intercellular adhesion molecule 1 (sICAM1), addition to myeloid cells; (2) diffuse-myeloid (n=44, 34%) with C-X-C motif-chemokine-13 (CXCL13), interleukin-8 (IL-8) and myeloid lineage predominance but poor in B cells/plasma-cells matrix metalloproteinase-3 (MMP-3) using customised electro- and (3) pauci-immune (n=34, 27%) characterised by fibroblasts chemiluminescence assays incorporating sample diluent blocking expansion but scanty immune cells. When evaluating prevalence of reagents to minimise interference from heterophilic antibodies. pathotype according to joint biopsied, we noted a higher propor- tion of pauci-immune pathotype in small- and medium-sized joints Statistical methods (figure 1D). Statistical analyses of histological and clinical parameters and predictive modelling of progression were run using R.3.0.2. P Identification of pathotype-specific gene expression markers values relating discrete variables to each other were calculated To determine the relationship of clinical characteristics with with the use of Fisher’s exact test. P values for correlation were molecular profiling in pretreatment and post-treatment biopsies, derived using Spearman’s correlation. To account for multiple we profiled gene expression of the 111 baseline and 68 post-treat- testing, we used the method of Benjamini and Hochberg to adjust ment samples using a custom NanoString panel. Unsupervised p values. For correlations between clinical characteristics and clustering of NanoString expression data showed strong grouping eigengene values with adjustment for biopsy joint position, all of pathotype-defined genes in concordance with their initial patho- p values from all comparisons were adjusted together. Likewise, type assigned by histology (figure 2A). Samples classified as either for correlations between serum biomarkers and clinical charac- lympho-myeloid or pauci-immune showed highest expression teristics and histological parameters, p values were adjusted alto- of the lymphoid or fibroid eigengene score, respectively, while gether. P values for comparison of means between three groups samples classified as diffuse-myeloid had myeloid eigengene scores were calculated using one-way analysis of variance with Bonfer- similar to the lympho-myeloid samples (figure 2B,C: Radar Plot). roni post hoc test. All p values reported are two sided. Furthermore, most genes measured significantly differed across the Selection of gene predictors: Backward stepwise selection three groups; all but 1 of 212 genes had an adjusted p value <0.01 method using logistic regression was performed using glm func- (figure 2D). Eigengene scores were then examined against synovial tion in R, with 16 baseline clinical covariates considered as aggregational scores demonstrating strong correlations with the candidates in the regression model. A subset of predictors was level of inflammatory infiltrate organisation (online supplemen- selected by imposing a L1 (least absolute shrinkage and selec- tary figure 3A,B). Comparing each pathotype to the others, we tion operator, LASSO) penalty on the regression parameters (46 found that pauci-immune-fibroid eigengene scores were negatively genes and 8 clinical covariates) of a sparse logistic regression correlated with lymphoid and myeloid eigengene scores, whereas using R package glmnet. The penalty parameter (lambda) was lymphoid and myeloid scores were positively correlated (online

764 Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 Rheumatoid arthritis

Figure 2 (A) Heatmap of NanoString gene expression data. Raw log2 NanoString counts for 212 genes and 111 patient samples were normalised per probe to give a mean of 0 and SD of 1. Normalised data were clustered by row and column using Euclidean distance and Ward’s linkage. Samples are coloured according to IHC-determined pathotype, with ungraded samples coloured grey. Rows are coloured according to the pathotype with which the gene was originally associated, with RA biology-associated genes coloured black. (B) Eigengene scores versus IHC-determined pathotypes. Individual eigengene scores are plotted for each sample, grouped and coloured by the pathotype as determined by IHC. Stars represent statistical significance as determined by linear regression across groups: *p<0.05, **p<0.01, ***p<0.001. (C) Radar plot of standardised eigengene scores. Eigengene values were normalised to give a mean 0 and SD of 1. Samples were grouped by pathotype, and the mean (solid lines) and SE of the mean (shaded region) were calculated for the normalised eigengenes. Spokes of the radar plot represent the distance along each normalised eigengene for each sample group. (D) Volcano plot of gene expression differences across pathotypes. For each gene, one-way ANOVA was performed comparing

expression across the three pathotypes. The –log10 p value from the one-way ANOVA is plotted against the root mean square of the log2 fold changes between each pair of eigengenes. Genes are coloured according to the pathotype in which it was initially identified, with RA biology-associated genes coloured black. ANOVA, analysis of variance; IHC, immunohistochemistry; RA, rheumatoid arthritis.

supplementary figure 4A). Comparing myeloid pathotype samples phase reactants, RF and ACPA positivity, and US scores despite with the other two yielded fewer differentially expressed genes, the presence of active disease, documented by high DAS28-ESR with lower expression of lymphoid-specific genes in these samples (mean 4.9), swollen joint counts, Health Assessment Questionnaire relative to the combined lympho-myeloid/pauci-immune-fibroid (HAQ) and patient global health visual analogue scale (VAS) scores group (online supplementary figure 4B). This is consistent with the as well as synovitis determined by ultrasonography (figure 3A). scoring algorithm used to determine the pathotypes: diffuse-my- The diffuse-myeloid group showed intermediate metrics. eloid samples mainly differ from lympho-myeloid samples by a relative lack of B- and T-cell infiltrate. Thus, the eigengene-specific Lymphoid/myeloid gene expression signatures correlate with gene sets show strong association with the expected pathotypes. RA disease activity Comparison of eigengene scores to features of disease activity Synovial pathotypes and clinical phenotypes (figure 3B) showed that the myeloid eigengene was highly asso- To establish the relationship between synovial pathotypes and ciated with disease activity, including acute phase reactant, clinical phenotypes, we compared a number of disease charac- DAS28-ESR, HAQ-Disability Index (DI) and overall PD ultraso- teristics with the three synovial pathotypes defined by immuno- nography score. The lymphoid eigengene was also correlated with histochemistry (IHC), with statistical adjustment for biopsy joint disease activity, but at a lower level, and was more associated with position. The lympho-myeloid pathotype had the highest levels of ultrasonographic scores. In keeping with the histopathological erythrocyte sedimentation rate (ESR), C-reactive protein, ACPA correlations, the fibroid eigengene was negatively associated with titre, swollen-joint-counts and DAS28-ESR scores (figure 3A). many aspects of disease activity (figure 3B). Assessment of joint damage by SHSS indicated an association of the lympho-myeloid group with worse erosive load and joint space narrowing (figure 3A). Furthermore, ultrasonographic assessment Serum biomarkers reflect synovial pathophysiology indicated that the lympho-myeloid group had significantly higher We next assessed if reported circulating serum biomarkers could levels of ST and PD scores both within the biopsied joint and overall function as surrogates of synovial tissue pathology2 and reliable scores. The pauci-immune group had the lowest levels of acute measures of disease activity. We selected CXCL13, sICAM-1,

Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 765 Rheumatoid arthritis

Figure 3 (A) Baseline clinical and histological parameters stratified according to three pathological subtypes, adjusted for joint type (n=129), *significant differences. (B) Correlation analysis of each Eigengene score with metrics of clinical disease activity, autoantibodies, acute phase reactants and ultrasonography. Values represent Spearman correlation coefficients between the clinical variables and the individual eigengene scores adjusted for joint type. Stars represent the significance of the correlation coefficient: *p<0.05, **p<0.01, ***p<0.001. ACPA, anti-citrullinated peptide antibodies; BJ, biopsied joint; CRP, C-reactive protein; DAS28, disease activity score-28; DI, Disability Index; ESR, erythrocyte sedimentation rate; HAQ, Health Assessment Questionnaire; RF, rheumatoid factor; SHSS, van der Heijde modified Sharp score; SJ, swollen joints; ST, synovial thickening; TJ, tender joints; US, ultrasound, VAS, visual analogue scale.

MMP3 and IL-8 due to their important inflammatory roles in (CCL19, BTLA, IL21R, CXCL13, LTA, LTB) and inflammatory RA pathophysiology. CXCL13 and sICAM1 have been previously cytokines (IL6) (figure 5B). In contrast, nonresponder patients shown to be elevated in different synovial phenotypes,2 whereas showed decrease in nonsignificant inflammatory gene expres- MMP3 is also elevated in synovitis and associated with radio- sion (figure 5C). graphic progression.12 IL-8 drives tissue recruitment of neutrophils We next assessed whether eigengene scores at baseline (polymorphonuclear neutrophils), which could play a role in syno- correlated with therapeutic response. Higher myeloid and vial autoimmunity through the exposure of citrullinated proteins lymphoid eigengene expression (but not fibroid) was associ- by neutrophil extracellular traps.13 Consistent with previous ated with larger decreases in DAS28-ESR scores post-treatment reports,14 15 we found that serum CXCL13 correlated with global (myeloid, p=0.003; lymphoid, p=0.044; figure 5D). Specific disease metrics, including DAS28, serological and ultrasonographic eigengene scores were then compared separately in good and measures of disease activity and synovial histology (figure 4A,B). nonresponders (figure 5E–G) determining that good DMARD Serum MMP-3 also showed modest yet significant correlation response showed more dynamic gene expression, with signifi- with acute phase reactants, DAS28 score and synovial histology cant decreases in both lymphoid and myeloid eigengenes, but (figure 4A,B). Intriguingly, CXCL13 and MMP3 were both a concomitant increase in fibroid eigengene expression. Nonre- elevated in patients with a lymphoid–myeloid pathotype, compared sponders exhibited a more muted change in gene expression, with with the other two pathotypes (figure 4C,D). In contrast, despite significant decrease in lymphoid eigengenes, but highly variable previous reports of elevated levels in RA patients,2 16 ICAM-1 changes in myeloid and fibroid eigengene expression indicating and IL-8 exhibited modest and variable correlations with clinical ongoing presence of myeloid gene expression in patients with indices such as tender joint scores, acute phase reactants and auto- continuing disease activity despite DMARD therapy, but strong antibody titres (figure 4A,B). Therefore, elevation of some but not downregulation in responder patients. all inflammatory proteins in the serum of RA patients tracks with synovitis and clinical disease activity. Synovial pathotypes and gene expression signatures predict Pretreatment synovial pathotypes gene expression signatures radiographic progression associate with response to DMARD treatment Next, we assessed whether baseline pathotypes or gene expres- Histologically defined pathotypes and gene expression signa- sion were associated with ongoing radiographic damage at 12 tures were examined for their response to DMARD therapy as months. Lympho-myeloid pathotype patients had a greater determined by change in DAS28-ESR at 6 months and EULAR change in SHSS compared with diffuse-myeloid and pauci-im- response criteria (figure 5). Overall, 90% of patients were treated mune-fibroid patients (figure 6A). Importantly, of the 14 with methotrexate either alone or in combination (figure 5A). patients subsequently commenced on biological therapy between Though histologically defined pathotypes did not associated 6 and 12 months of follow-up, a higher proportion with radio- with therapeutic outcome (figure 5A), differential gene expres- graphic progression at 12 months fell within the lympho-my- sion analysis in EULAR good-responder patients showed that eloid pathotype (26.5%, 9 progressors vs 25 nonprogressors), multiple inflammatory pathways were statistically significantly as compared with the diffuse-myeloid/pauci-immune-fibroid reduced including genes associated with lymphoid aggregates pathotypes (9.1%, 5 progressors vs 50 nonprogressors, Fishers

766 Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 Rheumatoid arthritis

Figure 4 (A) Correlation of pretreatment serum CXCL13, sICAM1, MMP3 and IL-8 with clinical disease metrics and ultrasonography scores. Values represent Spearman correlation coefficients between serum biomarkers and clinical variables. Stars represent the significance of the correlation coefficient: *p<0.05, **p<0.01, ***p<0.001. P values were corrected for multiple testing using Benjamini-Hochberg method. (B) Correlation of pretreatment serum CXCL13, sICAM1, MMP3 and IL-8 with synovial histology scores. Values represent Spearman correlation with individual eigengene scores adjusted for biopsy joint size, or histology semiquantitative scores. (C) Concentration of serum CXCL13 versus synovial pathotype status. (D) Concentration of serum MMP3 versus synovial pathotype status. P values were calculated using student t-test, with correction for multiple testing. ACPA, anti-citrullinated protein antibodies; BJ, biopsied joint; CRP, C-reactive protein; CXCL13, C-X-C motif-chemokine-13; DAS28, disease activity score-28; DI, Disability Index; ESR, erythrocyte sedimentation rate; HAQ, Health Assessment Questionnaire; IHC, immunohistochemistry; IL-8, interleukin-8; MMP-3, matrix metalloproteinase-3; PD, power Doppler; RF, rheumatoid factor; sICAM1, soluble intercellular adhesion molecule 1; ST, synovial thickening; US, ultrasound; VAS, visual analogue scale. exact test, p=0.029). Thus, despite more intensive treatment further noted that the pretreatment lymphoid eigengene score was regimens (including higher rates of biologic use), patients with significantly elevated in patients who had 12-month SHSS scores a lympho-myeloid pathotype were significantly more likely to increase versus those who did not (figure 6C), whereas in contrast develop joint damage progression. the myeloid and pauci-immune-fibroid eigengene scores were not Next, we compared pretreatment/baseline gene expression with significantly different between progressors versus non-progressors SHSS progression scores (figure 6B); 46 genes with p value <0.05 (figure 6D,E). were identified and these included B-cell-associated genes such as In order to explore mechanistically the link between the CD19, FCRL5 and BCMA. In agreement with these findings, we lympho-myeloid pathotype-associated gene signatures and

Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 767 Rheumatoid arthritis

Figure 5 (A) Clinical changes in disease activity and treatment regimens stratified according to pathotype. (B) Volcano plot showing changes in gene expression between baseline and 6 months in patients with a EULAR response. Individual points are coloured by the pathotype in which the gene was originally identified, with RA biology-associated genes coloured black. (C) Volcano plot showing changes in gene expression between baseline and 6 months in patients with EULAR nonresponse. Genes are coloured as above. (D) Correlation of pretreatment lymphoid, myeloid and fibroid eigengene scores with change in DAS28-ESR after 6 months of DMARD treatment. Spearman’s correlation coefficient is shown, along with the significance of this value. (E–G) Paired plots for baseline and 6 month lymphoid (E), myeloid (F) and fibroid (G) eigengene scores in patients who achieved good or poor clinical responses to DMARD treatment at 6 months by the EULAR response criteria. Patients who achieved a good response, or failed to achieve a moderate response, according to EULAR criteria are shown. For each patient, the pretreatment eigengene scores are connected to the post-treatment eigengene score, for each of the three eigengenes. Stars represent significance of the difference between pretreatment and post-treatment samples using a linear mixed effects model with sample date as a fixed effect and patient as a random effect: *p<0.05, **p<0.01, ***p<0.001. DAS28, disease activity score-28; DMARD, disease-modifying antirheumatic drug; ESR, erythrocyte sedimentation rate; EULAR, European League Against Rheumatism; MTX, methotrexate.

768 Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 Rheumatoid arthritis

Figure 6 (A) 12-month radiographic outcome of patients stratified according to pauci-immune-fibroid/diffuse-myeloid versus lympho-myeloid pathotypes. (B) Volcano plot of pretreatment genes differentially expressed between patients who progress radiographically after 1 year (ΔSHSS≥1); P values were from the two-sample t-test comparing the progressors and nonprogressors without adjustment. In all, 46 genes with p value <0.05 are highlighted in red. (C–E) Baseline eigengene values versus radiographic progression. Lymphoid (C), myeloid (D) and pauci-immune-fibroid (E) baseline scores are plotted against progression status (ΔSHSS≥1) at 1 year. **p<0.01 by t-test. (F–G) Identification of clinical and gene expression features predictive of radiographic progression at 1 year. Logistic regression, coupled with backward stepwise model selection, was applied to baseline clinical parameters against a dependent variable of radiographic progression or not at 12 months to select which clinical covariate contributed the most to the prediction; 16 baseline clinical covariates were considered as candidates in the regression model. Baseline variables included gender, age, disease duration, ESR, CRP, RF titre, ACPA titre (as continuous variables), VAS, tender and swollen joint number, baseline DAS28-ESR, EULAR response at 6 months (categorical), baseline HAQ, 12 max US ST and US PD scores and baseline pathotype (two categories: lympho-myeloid versus pauci-immune-fibroid/diffuse-myeloid). Stepwise variable selection yielded a model with eight clinical variables: baseline RF titre, disease duration, VAS, swollen joint number, DAS28-ESR, baseline pathotype, 12 max US ST and US PD scores. Selected covariates (46 genes plus 8 clinical covariates) were entered simultaneously into a logistic model with an L1 regularisation penalty (LASSO) in order to determine the optimal sparse prediction model. We have a better predictive performance of the model where clinical variables were penalised (F, blue-dashed line) than when they were not penalised (F, red-dotted line). (G) Nonzero weights associated with the final variables selected by the LASSO regression. ACPA, anti-citrullinated peptide antibodies; CRP, C-reactive protein; CXCL13, C-X-C motif-chemokine-13; DAS28-ESR, disease activity score-28-erythrocyte sedimentation rate; EULAR, European League Against Rheumatism; HAQ, Health Assessment Questionnaire; JSN, joint space narrowing; LASSO, least absolute shrinkage and selection operator; MMP-10, matrix metalloproteinase-10; RF, rheumatoid factor; SHSS, van der Heijde modified Sharp score; SJ, swollen joints; US ST, ultrasound synovial thickening; US PD, ultrasound power Doppler; VAS, visual analogue scale.

Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 769 Rheumatoid arthritis structural damage, we investigated the expression of a predefined and pauci-immune pathotypes are indeed present in early RA tissue osteoclast gene set derived from Harmonizome database (http://​ prior to potentially modifiable therapeutic-sensitive events. amp.​pharm.​mssm.​edu/​Harmonizome/). Lympho-myeloid patients had the highest levels of disease activity As expected, the osteoclast eigengene score strongly correlated and RF and ACPA seropositivity. Although previous definitions of with histology scores for both B cells, T cells, plasma cells and synovial pathotype have referred to lymphoid, myeloid and fibroid macrophages (online supplementary figure 5). However, when subgroups importantly, the data analysed herein demonstrate that patients were grouped according to pathotype, the demonstration lymphoid-rich patients also showed high expression of myeloid that osteoclast eigengene scores were most significantly elevated in genes hence the definition of ‘lympho-myeloid pathotype’ while the lympho-myeloid pathotype patients (p<0.01 lympho-myeloid another group of patients had a high expression of myeloid genes vs diffuse-myeloid and p<0.001 vs pauci-immune) strongly suggests but low expression of B-cell genes. This latter group also showed a that the lymphoid/B-cell component of the immune cell infiltrate diffuse inflammatory infiltrate lacking cellular aggregates and such is orchestrating osteoclast activation and therefore driving joint was defined as ‘diffuse myeloid’ while the ‘pauci-immune’ group erosion. showed a prevalence of stromal cells but almost complete absence Finally, to determine whether baseline clinical and gene expres- of immune cells. sion data could be combined into a model for predicting radio- Of note, the myeloid gene score was most strongly posi- graphic progression, we used two complementary approaches: (i) a tively correlated at baseline with clinical signs and symptoms logistic regression coupled with backward model selection to iden- including acute phase reactant, joint counts, DAS28-ESR, tify a minimal set of clinical predictors and (ii) a penalised method VAS, HAQ-DI. Importantly, subsequent treatment of these based on logistic regression with an L1 regularisation penalty patients with DMARDs indicated that the pretreatment levels (LASSO) to identify genes that improve the clinical model. of myeloid genes were most strongly correlated with treatment First, we used backward model selection on baseline clinical outcome at 6 months and, further, that significant reduction parameters to identify predictors of radiographic progression. The of myeloid gene scores was only observed in patients who final clinical covariates included baseline RF titre, disease duration, achieved a robust clinical response. This suggests that while VAS, swollen joint number, DAS28-ESR, baseline pathotype, 12 higher myeloid and lymphoid gene expression is associated max US ST and US PD scores. The predictive performance of the with increased disease activity, these are the patients more model evaluated by the optimism-corrected AUC was 0.75 (online likely to respond to broad DMARD immune suppression, supplementary figure 6A). as pretreatment levels correlated with therapeutic outcome. Next, a logistic regression with an L1 regularisation penalty These findings are consistent with previous reports indicating that the mean change in sublining macrophages is a strong (LASSO) was applied on the eight clinical covariates and 46 genes 21 (figure 6B) identified as being significantly differentially expressed biomarker of treatment response in RA. They also indicate that the myeloid lineage is a key driver in RA pathogenesis between progressors and nonprogressors. We found that a model and specific therapeutic targeting improves signs and symp- incorporating RF titre, and the expression of seven genes (SDC1, toms, although not specifically joint damage progression as CSF2, DENND1C, CD180, UBASH3A, CXCL1, MMP10, figure 6F), previously suggested in a small cohort of RA patients with yielded the optimal predictor of progression (lambda=0.0631, long-standing disease.22 It is worth reinforcing that the broad online supplementary figure 6B). After adjustment for optimism, anti-inflammatory activities of methotrexate via mechanisms our final prediction model was able to discriminate patients with and including activation of the adenosine pathway and down-mod- without radiographic progression with an optimism-corrected AUC ulation of adhesion molecules and proteases23 are supported of 0.88 (figure 6G). The predictive performance of the model was by our data indicating that multiple immune pathways are corrected for potential overfitting by computing the optimism-cor- suppressed by methotrexate treatment. rected AUC using a bootstrap method.17 18 These results suggest that Importantly, patients with high expression of fibroid genes but including both clinical covariates and genes in the model resulted in low expression of lymphoid and myeloid genes were found in 27% a superior predictive value compared with clinical covariates alone, of the cohort, suggesting that this fibroid/pauci-immune pheno- with improved discrimination between patients with and without type previously reported in joint replacement tissue appears to radiographic progression. be a defined disease endotype and not an end-stage (burned out) disease characteristic. Furthermore, the demonstration of a signif- icant difference in pathotype frequency between joints highlights Discussion the importance of techniques such as US-guided synovial biopsy 1–3 19 20 It is recognised that RA pathology is highly heterogeneous ; to enable unbiased recruitment of patients, through capacity to however, how such heterogeneity relates to disease activity, prog- sample a wide variety of joints, and thus ensure representative nosis and therapeutic response/outcome has been unclear. In sampling. While in our study we did not perform simultaneous addition, the reproducibility of defined histopathological cate- biopsies of different joints in the same patient, so cannot address gories, as well as their stability over time and/or relationship to per se whether synovial pathotype/molecular signatures are stable different disease stages also remains uncertain. The work described between joints of the same patient, it important to note that here tackles these problems investigating longitudinally a large, previous studies have documented stable cellular infiltrates24 and treatment-naïve, early RA cohort, with serial synovial sampling specific T-cell oligoclonal expansions25 between joints within the pre-DMARD and post-DMARD therapy and stringent clinical and same individuals, supporting the notion of a uniform pathology at imaging assessments. individual patient level across multiple joints. To address the issue of whether synovial heterogeneity is present Regarding the relationship of lineage-specific modules and clin- in early versus established RA, we designed gene panels linked ical/therapeutic outcomes, it is noteworthy that while fibroid gene to lymphoid, macrophage-myeloid and fibroblast cell-lineages, scores correlated with less severe disease, these patients still had trained on synovial subsets we reported in late-stage RA joint active disease (mean DAS28: 4.9). Moreover, these patients showed replacement synovial tissue,2 and tested them in the PEAC cohort. the poorest subsequent response to DMARD treatment while The data revealed that prevalent lympho-myeloid, diffuse-myeloid successful drug treatment of the patients with lympho-myeloid and

770 Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 Rheumatoid arthritis diffuse-myeloid pathotypes resulted in elevation of repair-response and Arthritis Research UK (ARUK) for their joint funding of Maximizing Therapeutic genes. These results indicate that DMARD treatment primarily Utility in Rheumatoid Arthritis (MATURA) [grant numbers MR/K015346/1, 20670, reduces inflammatory cell infiltrate, thus changing the cellular respectively]. composition of the synovium. Furthermore, as the fibroid genes Patient consent for publication Obtained. themselves are not reduced with current immune modulatory ther- Ethics approval The study has been approved by London – Dulwich Research apies, our data support the notion that new therapies are required Ethics Committee, REC reference: 05/Q0703/198. to target this specific cell-lineage known to play a critical role Provenance and peer review Not commissioned; externally peer reviewed. in pathogenesis/joint damage through epigenetic modifications Open access This is an open access article distributed in accordance with the leading to an aggressive phenotype even in the absence of immune/ Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits inflammatory cells.26 others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, The strong correlation between radiographic damage (base- and indication of whether changes were made. See: https://​creativecommons.​org/​ line and 12 months), lympho-myeloid pathotype and associated licenses/by/​ ​4.0/.​ osteoclast-targeting genes, suggests that active immunological processes drive inflammation and structural damage, accelerated References by high B-cell infiltration levels. This is concordant with previous 1 McInnes IB, Schett G. Pathogenetic insights from the treatment of rheumatoid reports showing that B cells and plasma cells produce RANKL arthritis. Lancet 2017;389:2328–37. and TWEAK27–29 while their number was elevated in patients 2 Dennis G, Holweg CTJ, Kummerfeld SK, et al. Synovial phenotypes in with MRI-determined bone oedema and increased RANKL rheumatoid arthritis correlate with response to biologic therapeutics. Arthritis Res Ther 2014;16. expression.30 In contrast, nonprogressors had elevated baseline 3 Pitzalis C, Kelly S, Humby F. New learnings on the pathophysiology of RA from synovial levels of osteoprotegerin (a decoy receptor/negative regulator biopsies. Curr Opin Rheumatol 2013;25:334–44. for RANKL-mediated osteoclastogenesis) and fibroid-associ- 4 Mandelin AM, Homan PJ, Shaffer AM, et al. Transcriptional profiling of synovial ated genes including fibroblast growth factor family members, macrophages using minimally invasive ultrasound-guided synovial biopsies in Noggin and cartilage intermediate layer protein (CILP). rheumatoid arthritis. Arthritis Rheumatol 2018;70:841–54. 5 orange DE, Agius P, DiCarlo EF, et al. Identification of three rheumatoid arthritis In summary, the results presented herein indicate that diverse disease subtypes by machine learning integration of synovial histologic features and synovial pathotypes are present at an early disease stage prior to RNA sequencing data. Arthritis Rheumatol 2018;70:690–701. potential drug-sensitive modification of disease pathology. Further- 6 Kelly S, Humby F, Filer A, et al. Ultrasound-guided synovial biopsy: a safe, well- more, we demonstrated that specific synovial pathotypes and tolerated and reliable technique for obtaining high-quality synovial tissue related molecular signatures are associated with clinical pheno- from both large and small joints in early arthritis patients. Ann Rheum Dis 2015;74:611–7. types, disease outcome/prognosis, that is, radiographic progression 7 naredo E, Bijlsma JWJ, Conaghan PG, et al. Recommendations for the content and and moreover that specific molecular signatures predict response conduct of European League against rheumatism (EULAR) musculoskeletal ultrasound to therapy. These data support the evaluation of such biomarkers courses. Ann Rheum Dis 2008;67:1017–22. within randomised clinical trials with the aim of enriching response 8 Humby F, Bombardieri M, Manzo A, et al. Ectopic lymphoid structures support ongoing production of class-switched autoantibodies in rheumatoid synovium. PLoS Med to current biological therapies by targeting the specific cognate 2009;6:e1. pathways expressed in some patients but not others, while offering 9 ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for the opportunity, similar to cancer medicine, of developing path- RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. way-driven/stratified approaches to patients with RA. 10 Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med 1990;9:811–8. 11 Bueno R, Stawiski EW, Goldstein LD, et al. Comprehensive genomic analysis of Author affiliations 1 malignant pleural mesothelioma identifies recurrent mutations, gene fusions and Centre for Experimental Medicine & Rheumatology, William Harvey Research splicing alterations. Nat Genet 2016;48:407–16. Institute, Barts & The London School of Medicine & Dentistry, Queen Mary University 12 Houseman M, Potter C, Marshall N, et al. Baseline serum MMP-3 levels in patients of London, London, UK with rheumatoid arthritis are still independently predictive of radiographic progression 2Biomarker Discovery OMNI, Genentech Inc, South San Francisco, California, USA 3 in a longitudinal observational cohort at 8 years follow up. Arthritis Res Ther 2012;14. Bioinformatics and Computational Biology, Genentech Inc, South San Francisco, 13 Khandpur R, Carmona-Rivera C, Vivekanandan-Giri A, et al. Nets are a source of California, USA 4 citrullinated autoantigens and stimulate inflammatory responses in rheumatoid Centre for Translational Bioinformatics, William Harvey Research Institute, London, arthritis. Sci Transl Med 2013;5. UK 5 14 Jones JD, Hamilton BJ, Challener GJ, et al. Serum C-X-C motif chemokine 13 is Department of Rheumatology, Leiden University Medical Center, Leiden, elevated in early and established rheumatoid arthritis and correlates with rheumatoid Netherlands factor levels. Arthritis Res Ther 2014;16. 6Amsterdam Rheumatology Center, AMC, Amsterdam, Netherlands 7 15 Greisen SR, Schelde KK, Rasmussen TK, et al. CXCL13 predicts disease activity in Department of Rheumatology, Zuyderland MC, Heerlen, Netherlands early rheumatoid arthritis and could be an indicator of the therapeutic ’window of 8Universita degli Studi di Cagliari, Cagliari, Italy 9 opportunity’. Arthritis Res Ther 2014;16. Institute of Infection, Immunity and Inflammation, University of Glasgow, Glasgow, 16 cascão R, Moura RA, Perpétuo I, et al. Identification of a cytokine network sustaining UK neutrophil and Th17 activation in untreated early rheumatoid arthritis. Arthritis Res 10University of Birmingham, Birmingham, UK 11 Ther 2010;12. Institute of Infection and Immunity, Cardiff University School of Medicine, Cardiff, 17 Davison A, Kuonen D. An introduction to the bootstrap with applications in R. Stat UK 12 Comput Stat Graph Newsl 2003;13:6–11. Nuffield Department ofO rthopaedics, Rheumatology and Musculoskeletal Sciences, 18 smith GCS, Seaman SR, Wood AM, et al. Correcting for optimistic prediction in small Kennedy Institute of Rheumatology, Oxford, UK data sets. Am J Epidemiol 2014;180:318–24. Acknowledgements We are grateful to all the patients who participated in this 19 van der Pouw Kraan TCTM, van Gaalen FA, Kasperkovitz PV, et al. Rheumatoid study and clinical staff who helped with recruitment. arthritis is a heterogeneous disease: evidence for differences in the activation of the STAT-1 pathway between rheumatoid tissues. Arthritis Rheum 2003;48:2132–45. Contributors All authors have contributed to different degrees to patients 20 van Baarsen LGM, Wijbrandts CA, Timmer TCG, et al. Synovial tissue heterogeneity in recruitment and or data generation and or data analysis and or writing the rheumatoid arthritis in relation to disease activity and biomarkers in peripheral blood. manuscript and or revising data and or manuscript. Arthritis Rheum 2010;62:1602–7. Funding This study was funded by Barts and The London School of Medicine 21 Haringman JJ, Gerlag DM, Zwinderman AH, et al. Synovial tissue macrophages: a and Dentistry Charity (grant number: 523/819); Arthritis Research UK (http://​dx.​ sensitive biomarker for response to treatment in patients with rheumatoid arthritis. doi.org/​ ​10.13039/​ ​501100000341), Experimental Arthritis Treatment Centre Grant Ann Rheum Dis 2005;64:834–8. n:20; Medical Research Council (http://​dx.doi.​ ​org/​10.​13039/501100000265),​ 22 Yanni G, Whelan A, Feighery C, et al. Synovial tissue macrophages and joint erosion in Pathobiology of Early Arthritis Cohort (PEAC) Grant; Medical Research Council (MRC) rheumatoid arthritis. Ann Rheum Dis 1994;53:39–44.

Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 771 Rheumatoid arthritis

23 Brown PM, Pratt AG, Isaacs JD. Mechanism of action of methotrexate in 27 Meednu N, Zhang H, Owen T, et al. Production of RANKL by memory B cells: a rheumatoid arthritis, and the search for biomarkers. Nat Rev Rheumatol link between B cells and bone erosion in rheumatoid arthritis. Arthritis Rheumatol 2016;12:731–42. 2016;68:805–16. 24 Kraan MC, Reece RJ, Smeets TJM, et al. Comparison of synovial tissues from the 28 Dharmapatni AASSK, Smith MD, Crotti TN, et al. TWEAK and Fn14 expression in the knee joints and the small joints of rheumatoid arthritis patients: implications for pathogenesis of joint inflammation and bone erosion in rheumatoid arthritis. Arthritis pathogenesis and evaluation of treatment. Arthritis Rheum 2002;46:2034–8. Res Ther 2011;13. 25 Musters A, Klarenbeek PL, Doorenspleet ME, et al. In rheumatoid arthritis, synovitis at 29 Gravallese EM. Bone wasn’t built in a day: destruction and formation of bone in the different inflammatory sites is dominated by shared but patient-specific T cell clones. J rheumatic diseases. Trans Am Clin Climatol Assoc 2017;128:24–43. Immunol 2018;201:417–22. 30 Dalbeth N, Smith T, Gray S, et al. Cellular characterisation of magnetic resonance 26 Bottini N, Firestein GS. Duality of fibroblast-like synoviocytes inRA : passive responders imaging bone oedema in rheumatoid arthritis; implications for pathogenesis of and imprinted aggressors. Nat Rev Rheumatol 2013;9:24–33. erosive disease. Ann Rheum Dis 2009;68:279–82.

772 Humby F, et al. Ann Rheum Dis 2019;78:761–772. doi:10.1136/annrheumdis-2018-214539 Rheumatoid arthritis

Epidemiological science Sequencing of the MHC region defines HLA-DQA1 as the major genetic risk for seropositive rheumatoid arthritis in Han Chinese population Jianping Guo,‍ ‍ 1,2 Tao Zhang,3,4 Hongzhi Cao,3,5,6 Xiaowei Li,3,4 Hao Liang,7 Mengru Liu,1 Yundong Zou,1 Yuanwei Zhang,3,4 Yuxuan Wang,1 Xiaolin Sun,1,2 Fanlei Hu,1,2 Yan Du,1 Xiaodong Mo,3,4 Xu Liu,1 Yue Yang,1 Huanjie Yang,3,4 Xinyu Wu,1 Xuewu Zhang,1 Huijue Jia,3,4 Hui Jiang,3,4 Yong Hou,3,4 Xin Liu,3,4 Yin Su,1,2 Mingrong Zhang,3,4 Huanming Yang,3,8 Jian Wang,3,8 Liangdan Sun,9 Liang Liu,10 Leonid Padyukov,11 Luhua Lai,7 Kazuhiko Yamamoto,‍ ‍ 12 Xuejun Zhang,9,13 11 3,4 1,2,14,15 Handling editor Josef S Lars Klareskog, Xun Xu, Zhanguo Li Smolen

►► Additional material is ABSTRACT Key messages published online only. To view Objective The strong genetic contribution of the major please visit the journal online histocompatibility complex (MHC) region to rheumatoid (http://dx.​ ​doi.org/​ ​10.​1136/​ What is already known about this subject? annrheumdis-2018-​ ​214725). arthritis (RA) has been generally attributed to human ► The strong genetic contribution of the major leukocyte antigen (HLA)-DRB1. However, due to the high ► histocompatibility complex (MHC) region to For numbered affiliations see polymorphisms and linkage disequilibrium within MHC, end of article. rheumatoid arthritis (RA) susceptibility has been it is difficult to define novel and/or independent genetic generally attributed to HLA-DRB1. risks using conventional HLA genotyping or chip-based Correspondence to ► However, due to the high linkage disequilibrium microarray technology. This study aimed to identify novel ► Dr Xuejun Zhang, Institute of in MHC, it is difficult to define novel and/or RA risk variants by performing deep sequencing for MHC. Dermatology and Department of independent RA risks using the conventional Dermatology, No 1 Hospital of Methods We first conducted target sequencing for HLA genotyping or chip-based microarray Anhui Medical University, Hefei the entire MHC region in 357 anticitrullinated protein 230032, China; technology. ayzxj@​ ​vip.sina.​ ​com, Professor antibodies (ACPA)-positive patients with RA and 1001 healthy controls, and then performed HLA typing in an Lars Klareskog, Rheumatology What does this study add? Unit, Department of Medicine, independent case–control cohort consisting of 1415 ► We first conducted target sequencing for the Karolinska Institutet, Stockholm samples for validation. All study subjects were Han ► entire MHC region and then performed HLA 171 76, Sweden; ​lars.​ Chinese. Genetic associations for RA susceptibility and klareskog@​ki.​se, Dr Xun Xu, typing in an independent cohort for validation. severity were analysed. Comparative modelling was Beijing Genomics Institute ►► We provide the first evidence that (1) HLA- (BGI)-Shenzhen, Shenzhen constructed to predict potential functions for the newly DQα1:160D is the strongest and independent 518000, China; discovered RA association variants. genetic risk for anticitrullinated protein xuxun@​ ​genomics.cn​ and Results HLA-DQα1:160D conferred the strongest Professor Jianping Guo and antibodies (ACPA)-positive RA in Han Chinese, and independent susceptibility to ACPA-positive Professor Zhanguo Li, −36 and (2) HLA-DRβ1:37N is a novel and RA (p=6.16×10 , OR=2.29). DRβ1:37N had an Department of Rheumatology independent protective factor for ACPA-positive and Immunology, Peking independent protective effect (p=5.81×10−16, OR=0.49). RA. University People’s Hospital, As predicted by comparative modelling, the negatively ► We next performed a structure-based functional Beijing 100044, China; charged DQα1:160D stabilises the dimer of dimers, thus ► jianping.​ ​guo@bjmu.​ ​edu.​cn, prediction for the two variants. may lead to an increased T cell activation. The negatively zli99@​ ​bjmu.edu.​ ​cn ► We showed that (1) negatively charged charged DRβ1:37N encoding alleles preferentially bind ► DQ 1:160D (D160 ) stabilised the structure JG, TZ, HC, XL and HL with epitope P9 arginine, thus may result in a decreased α α of DQ 1-DQ 1 dimer of dimers and may contributed equally. RA susceptibility. α β lead to an increased T cell activation, and (2) Conclusions We provide the first evidence that Received 8 November 2018 DRβ1:37N benefits electrostatic interaction and HLA-DQα1:160D, instead of HLA-DRB1*0405, is the Revised 6 March 2019 preferentially binds with epitope P9 arginine, Accepted 9 March 2019 strongest and independent genetic risk for ACPA-positive thus may result in a decreased RA susceptibility. Published Online First RA in Han Chinese. Our study also illustrates the value of 1 April 2019 deep sequencing for fine-mapping disease risk variants in the MHC region.

1–3 © Author(s) (or their environmental factors. The major histocompat- employer(s)) 2019. No ibility complex (MHC) genes represent the best commercial re-use. See rights Introduction described genetic risk loci linking to RA suscepti- and permissions. Published Rheumatoid arthritis (RA) is a chronic and systemic bility in many populations. HLA-DR variants were by BMJ. autoimmune disease primarily affecting the periph- mainly associated with anticitrullinated protein To cite: Guo J, Zhang T, eral joints. Studies indicate that RA is a hetero- antibodies (ACPA)-positive RA.4–7 Furthermore, the Cao H, et al. Ann Rheum Dis geneous disease where different subsets result RA-risk human leukocyte antigen (HLA) alleles are 2019;78:773–780. from complex interactions between genetic and heterogeneous among ethnic groups. For example,

Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 773 Rheumatoid arthritis

figure 1. The baseline demographic characteristics of the patients Key messages and healthy controls are summarised in table 1. All patients and healthy individuals were Han Chinese. How might this impact on clinical practice or future Patients satisfied the American College of Rheumatology 1987 developments? 14 revised criteria for a diagnosis of RA, and were recruited from ► A key finding of this study is the major genetic influence ► the Department of Rheumatology and Immunology at Peking of HLA-DQ 1:160D, instead of the well-known HLA-DRB1 α University People’s Hospital. All cases were ACPA-positive alleles, on ACPA-positive RA in Han ancestry. patients with RA. Among cases, a total of 558 X-ray sets of hands ► It may indicate that HLA-DQ 1:160D could be a potential ► α were available. All X-rays were chronologically scored for assess- therapeutic target for RA. 15 16 ment of bone destruction, as described previously.

HLA-DRB1*04:01 is the major RA-risk allele in Caucasians, MHC target sequencing and variant calling whereas DRB1*04:05 is the most frequent RA-risk allele in East The MHC region was sequenced using the targeted sequencing 17 Asians.5 8–10 In addition to HLA-DRB1, other HLA genes, such methodology, as described previously. The samples were as HLA-B and HLA-DPB1, have also been suggested to play a sequenced with standard 2×90 bp paired-end reads on the role in susceptibility to RA.6 11 However, due to the high linkage Illumina HiSeq 2000 sequencer. Sequencing data for the 1001 disequilibrium (LD) in the MHC region and high-density genes healthy controls were cited and selected by adjusting with 18 within the MHC region, it is difficult to identify the ‘real’ and/ age and sex from a recent publication of psoriasis project. or additional independent genetic risks by the conventional HLA Sequence reads were aligned to the National Center for Biotech- genotyping and chip-based microarray technology.12 13 nology Information (NCBI) human genome reference assembly To fine-map the MHC region and identify novel variants (Build 37) using Burrows-Wheeler Aligner (V.0.5.9). On average, contributing to RA susceptibility, we performed deep sequencing the MHC region was sequenced to a mean depth of 94 X, with for the entire MHC region and HLA typing for validation. Given 96.6% covered by at least one read and 93.1% covered by at that a particular amino acid(s) rather than its position may have least ten reads for cases, and a mean depth of 69 X for controls. potential biological function(s), in the present study we have Although the sequencing depth in cases and controls was mainly focused on the individual amino acid variants, classical differed, there was no bias for performance of variant calling 17 18 HLA alleles and single nucleotide polymorphism (SNPs). and HLA typing according to previous studies. The variant calling and quality control of the target sequencing are detailed Materials and methods in online supplementary materials and methods. Study design Two independent cohorts, including 1358 subjects for discovery HLA typing cohort (357 cases and 1001 controls) and 1415 subjects for vali- In the discovery cohort, a total of 26 highly polymorphic HLA dation cohort (604 cases and 811 controls), were enrolled in the genes were genotyped according to the Short Oligonucle- study. The schematic diagram of the study design is shown in otide Analysis Package (SOAP)-HLA. SOAP-HLA is a flow of

Figure 1 Schematic diagram of the study design for the discovery and validation of the MHC region in ACPA-positive RA. ACPA, anticitrullinated protein antibodies; BWA-GATK, Burrows-Wheeler Aligner-Genome Analysis Toolkit; HLA, human leukocyte antigen; MHC, major histocompatibility complex; NGS, next generation sequencing; RA, rheumatoid arthritis; SOAP-HLA, Short Oligonucleotide Analysis Package-human leukocyte antigen.

774 Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 Rheumatoid arthritis

Table 1 Demographic characteristics of the study cohorts Stage I (discovery) Stage II (validation) Number of patients/controls 357/1001 604/811 Demographic characteristics Female (%) 80.4/72.4 81.0/82.2 Age, mean±SD years 57.4±12.1/43.0±15.7 54.6±13.0/45.3±13.6 Clinical characteristics Age at onset, mean±SD years 47.1±13.9 45.3±14.6 Disease duration, mean±SD years 10.3±8.9 9.4±8.7 SHS, mean±SD (n)* 80.5±63.0 (336) 41.6±51.8 (212) *The case number when data were available. SHS, modified Sharp-van der Heijde Score. Figure 2 Plots of stepwise conditional analysis for ACPA-positive RA in the MHC region for the discovery cohort and the combined cohort. sequencing data analysis pipeline to type any HLA genes using (A) Each diamond represents −log10(p) of the variants, including SNPs, capture sequenced data based on IMGT/HLA database with Indels, classical HLA alleles and amino acid polymorphisms of HLA a high accuracy.17 The amino acid sequence was determined genes. The dotted horizontal line represents the suggestive significance −6 according to the IMGT/HLA database (V.3.22.0). threshold of p=1×10 . The bottom panel indicates the physical In the validation cohort, the subjects were genotyped for HLA- positions of the HLA genes on chromosome 6 (NCBI Build 37). (B) The A, HLA-B, HLA-DRB1, HLA-DPB1 and HLA-DQA1. HLA-A, association for each locus used for conditioning (HLA-DQA1, HLA- HLA-B, HLA-DRB1 and HLA-DPB1 were genotyped using the DRB1) is shown in red in each panel. For each panel, the horizontal axis next-generation sequencing method.19 HLA-DQA1 were geno- shows the position of amino acid for each HLA gene, and the vertical typed by sequencing of both exons 2 and 3 using the Sanger axis shows the negative log10-transformed p values for association. 20 The dashed horizontal line corresponds to the significance threshold sequencing method. of p=5×10−8. ACPA, anticitrullinated protein antibodies; HLA: human leukocyte antigen; MHC, major histocompatibility complex; NCBI: Statistical analyses National Center for Biotechnology Information; RA, rheumatoid arthritis; Using the same data processing and analysis,18 21 the logistic SNPs, single nucleotide polymorphisms. regression model was applied to test the association between disease and variants, adjusting by gender. We define the HLA Results variants by including the biallelic SNPs, Indels, classical HLA HLA-DQα1:160D is the strongest genetic risk for ACPA- alleles and individual HLA amino acid polymorphisms. For indi- positive RA, discovery by MHC sequencing vidual HLA allele and amino acid variants, the association was We first conducted a capture sequencing of the whole MHC determined after stratifying the data using the relative predispo- region in 357 patients with ACPA-positive RA and 1001 previ- 22 sitional effect method. To evaluate the overall significance of ously sequenced healthy individuals. After quality control, the multiallelic HLA amino acid polymorphisms for respective a total of 24 177 variants, 166 HLA types and 1283 amino positions, the omnibus association test was conducted using the acids were obtained (online supplementary tables 1 and 2), R statistics program (detailed in online supplementary M&M). in which 584 variants (including SNPs, HLA types and amino To assess the independent effects for the candidates identified acids) showed significant associations by a cut-off of p value less in logistic analysis, the stepwise conditional regression model than 1.0×10−6 (online supplementary table 3). We found that was applied by additionally including the most significant loci as the top association signal was mapped to HLA-DQA1, with a covariates, as described in online supplementary M&M. peak at HLA-DQα1 amino acid position 160 (DQα1:160D, In the discovery stage, a p value less than 1.0×10−6 was p=1.04×10−17, OR=2.36, 95% CI 1.94 to 2.87), followed by suggested as the significance threshold.23 24 In the validation and DQα1:160A (p=6.25×10−17, OR=2.27, 95% CI 1.87 to 2.75) combined stages, a genome-wide statistical significance threshold (figure 2A and online supplementary table 3). HLA-DRB1*0405 of p value less than 5.0×10−8 was applied. allele also showed a strong association with ACPA-positive RA, but fell out of the top 10 risk variants (p=7.94×10−14, OR=3.45, 95% CI 2.50 to 4.78) (online supplementary table Comparative modelling 3). These results indicated that by sequencing the entire MHC The comparative modelling was conducted using Modeller region, we discovered HLA-DQα1:160D may be the top genetic V.9.14.25 The construction for comparative modelling is risk for ACPA-positive RA in the Han population, instead of the detailed in online supplementary M&M. The crystal struc- well-known HLA-DRB1 *0405. ture of HLA-DR1 (DRA-DRB1*0101) (PDB ID: 1AQD) was employed as the template for comparative modelling of HLA-DRβ1:37N has an independent protective effect on ACPA- 26 DQA1*0303-DQB1*0401. The overall sequence identity and positive RA similarity between DQA1*0303-DQB1*0401 and HLA-DR1 By stepwise conditional analysis on HLA-DQα1:160D, the are 61.7% and 76.6%, respectively. The crystal structure of second signal was mapped to a list of SNPs, with a peak at HLA-DR3 (DRA-DRB1*03:01) (PDB ID: 1A6A) was employed rs9275376 (chr6_32668633_G_T) (online supplementary as the template for the comparative modelling for DRA- table 4). The SNPs are in high LD (r2=0.82), and majority of DRB1*13:01 and DRA-DRB1*13:02.27 The sequence identities the variants are located in the intergenic region between DQA2 of HLA-DR3 to DRA-DRB1*13:01 and DRA1-DRB*13:02 were and DQB1 according to the NCBI RefSeq database. An imme- 98.1% and 97.8%, respectively. diate significant association was observed for HLA-DRβ1:37N

Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 775 Rheumatoid arthritis following these intergenic SNPs (figure 2A and online supple- (p=1.05×10−35) and DQα1:160A (p=2.72×10−35). mentary table 4). Of note, HLA-DRβ1:37N has an indepen- DRB1*04:05 and DRβ1:11V also showed a strong association, dent protective effect on ACPA-positive RA (p=6.21×10−8, but out of the top 10 risks (p=7.75×10−30 and p=1.25×10−28, OR=0.48, 95% CI 0.37 to 0.63). DRβ1:96 hours also showed a respectively). significant association (p=6.27×10−7, OR=1.59, 95% CI 1.33 As HLA-DRB1 was well recognised as the strongest risk for to 1.91). Conditioning on DRβ1:37N, the intergenic SNPs and ACPA-positive RA in previous studies, we then investigated RA DRβ1:96 hours lost their association and no additional indepen- association at HLA-DRB1 separately. By stepwise conditional dent association(s) reached the suggestive statistical significance analysis, we found if DQA1 is excluded, the DRB1*0405, the threshold of p<1.0×10−6 (data not shown). amino acid variants at DRβ1:11, 13, 57, 74 and 71 could be To assess whether DQα1:160D and DRβ1:37N were indepen- strong risks for ACPA-positive RA (detailed in online supple- dent of each other, we performed the conditional analysis starting mentary results, supplementary figure 2 and supplementary from DRβ1:37N. As shown in online supplementary figure 1, table 11). We also compared the individual amino acid frequen- after conditioning on DRβ1:37N, DQα1:160D displayed an cies between Han Chinese and European populations. As even stronger association (p=6.38×10−23, OR=3.55, 95% CI shown in figure 3A, in the present study both DQα1:160D and 2.76 to 4.56). This indicates that DQα1:160D also has an inde- DQα1:160A are common amino acids in Han Chinese and were pendent effect on ACPA-positive RA. significantly increased in cases. However, the two amino acids have not been detected in European population.6 For amino acid positions 11, 13, 57, 71 and 74 in DRβ1, the amino acid The validation study confirms the findings discovered by frequencies are similar between the two ethnic groups or case capture sequencing and controls, except for a few amino acids (figure 3B,C,D,E,F). To validate the findings discovered by capture sequencing, we We next examined whether DQα1:160D and its encoding genotyped HLA-A, HLA-B, HLA-DRB1, HLA-DQA1 and HLA- alleles, that is, DQA1*0302 and DQA1*0303, confer a risk for DPB1 in an independent case–control cohort, consisting of 604 the severity of joint damage in ACPA-positive RA. As shown cases and 811 healthy subjects. In line with the findings in the in online supplementary figure 3, in the early disease stage, discovery stage, DQα1:160D showed consistent top associa- DQA1*0303 displayed high impact on radiographic scores in the −19 tion in the validation panel (p=7.14×10 , OR=2.23, 95% smoking group (p=3.02×10−5; online supplementary figure 3A CI 1.87 to 2.66) (online supplementary table 5). After condi- and detailed in online supplementary results). tioning to DQα1:160D, although DRβ1:96 hours became the second independent signal (p=2.80×10−10, OR=1.68, 95% CI −8 Additional negative charge of D160α enhances the 1.43 to 1.97), it was followed by DRβ1:37N (p=1.93×10 , interaction of DQ 1-DQ 1, leading to an increased T cell OR=0.51, 95% CI 0.40 to 0.65) (online supplementary table α β activation 6). When conditioning on DRβ1:96 hours, DRβ1:37N lost the association (p=5.96×10−3) and vice versa (p=1.18×10−4; data MHC class II molecules could present in the form of either heterodimers or the dimer of the heterodimers (dimer of not shown) according to the genome-wide statistical significance 26 29 30 −8 dimers). The dimer of dimers appears to play an important threshold of p<5.0×10 . role in T cell response to low affinity antigens by enhancing Joint analysis of discovery and replication panels provided overall affinity between MHC/peptide and T cell receptor compelling evidence that HLA-DQα1:160D conferred the 30 31 −36 (TCR). The electrostatic interactions between the interface highest risk on ACPA-positive RA (p=6.16×10 , OR=2.29, residues of the dimer of dimers are critical for maintaining struc- 95% CI 2.01 to 2.60) (figure 2B, online supplementary table tural stability as well as T cell activation.32 According to sequence 7). After conditioning to DQα1:160D, DRβ1:96 hours and analysis, D160α locates away from antigen binding groove and DRβ1:37N showed similar independent effects (DRβ1:96 hours: −16 should impose little influence on epitope binding. To investigate p=4.90×10 , OR=1.64, 95% CI 1.45 to 1.84; DRβ1:37N: the potential function of DQ 1:160D, we constructed the dimer p=5.81×10−16, OR=0.49, 95% CI 0.41 to 0.58; figure 2B and α of dimer structure for DQA1*0303-DQB1*0401 by compara- online supplementary table 8). After conditioning on DRβ1:96 −6 tive modelling, as it is the major D160α encoding haplotype in hours, the DRβ1:37N lost the association (p=4.85×10 ). Han Chinese (69.0% from our data and 56% in the Han MHC Similarly, when conditioning on DRβ1:37N, DRβ1:96 hours database).18 As shown in figure 4A, D160 is adjacent to the also lost the association (p=2.26×10−5), and there was no α dimer of dimer interface. Strong electrostatic interactions are additional independent association(s) that reached the genome- observed between negative DQ 1 interface residues (D161 , wide statistical significance of p<5.0×10−8 (data not shown). α α E182α and E184α) and positive DQβ1’ interface residues When conditioning on DQα1:160D, other DRβ1 variants also (R105β, H111β and H112β). Compared with non-charged showed significant associations, including the putative RA-risk −13 A160α or S160α coded by other DQA1 alleles, additional nega- variant DRβ1:11V (p=2.06×10 , OR=1.79, 95% CI 1.53 to −10 tive charge of D160α further enhances the interaction between 2.09) and DRβ1:11D (p=2.99×10 , OR=0.49, 95% CI 0.39 DQα1 and DQβ1’, which may lead to an increased T cell acti- to 0.61; online supplementary table 8). Notably, DRβ1:11D is vation (figure 4A). in strong LD with DRβ1:37N (r2=0.62; online supplementary table 9) and also showed a protective effect. Next, we performed the omnibus test to assess the overall The negatively charged P9 pockets from DRβ1:37N encoding significance of the multiallelic amino acids for respective posi- alleles benefit electrostatic interaction and epitope P9 tions. As shown in online supplementary table 10, position 13 arginine binding in DRβ1 showed the top association (p=4.03×10−45), followed It is well accepted that HLA-DRB1 susceptibility alleles are by position 11 (p=9.85×10−45), which is consistent with the strongly associated with ACPA-positive RA, and its encoded previous findings.6 28 However, if focusing on the individual molecules preferentially present the citrullinated autoantigens.33 amino acids or classical HLA alleles, DQα1:160D remained The susceptibility is determined by the electrostatic property of the top association (p=1.82×10−38), followed by DQA1*03:03 antigen binding pockets and the ability to differentially recognise

776 Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 Rheumatoid arthritis

Figure 3 Comparison of individual amino acid frequencies within DQα1:160 and DRβ1:11, 13, 57, 71 and 74 in Han Chinese and European populations. The individual amino acid frequencies are plotted in healthy controls (blue) and cases (red). The upper panel shows the amino acid frequencies in Han Chinese population (the data derived from the present study). The lower panel shows the amino acid frequencies in European population (the data cited from Raychaudhur et al‘s study6). (A) DQa1:160D and DQa1:160A are common amino acids in HanChinese and were significantly increased in patients with RA, but the two variantshave not been detected in European population. Similar frequencies of amino acids were observed at DRß1 position 11(B),13(C), 57(D), 71(E), and 74(F) between Han Chinese and European populations. citrullinated antigens. The positively charged antigen binding As DRβ1:96 hours and DRβ137N have similar effects and pocket in RA-susceptible alleles preferentially accommodate cannot be distinguished by statistical analysis, we then sought citrullinated antigens, whereas electronegative or electroneu- to predict the potential function for DRβ1:96 hours and tral pockets in RA-resistant alleles can bind both arginine and found the crystal structure of HLA-DR11 was available (DRA- citrullined antigens.34 The amino acid residues at position 37 DRB1*1101, one of major encoding alleles for DRβ1:96 hours in DRβ1 are located within the P9 pocket of antigen binding in Han Chinese). As shown in online supplementary figure 4, groove. By sequence analysis, we found four alleles containing unlike DRβ1:37N, DRβ1:96 hours (PDB code: 6CPL) locates an asparagine at position DRβ1:37 (Asn37β or 37N), including far away from either antigen binding groove or the dimer of DRB1*03:01, DRB1*09:01, DRB1*13:01 and DRB1*13:02. dimer interface. It may indicate that DRβ1:96 hours has little Although DRB1*0901 was reported to be the risk allele for RA,10 impact on either epitope binding or the structural stability of the it was associated with a low level of anti-cyclic citrullinated dimer of dimers. peptide (CCP) antibodies in patients with RA or anti-CCP-nega- tive RA.35 36 We then constructed the crystal structure for other allele-containing haplotypes by comparative modelling. As Discussion shown in the bottom panel of figure 4B, the DRA1-DRB1*03:01, A key finding of this study is the major genetic influence of DRA1-DRB1*13:01 and DRA1-DRB1*13:02 share identical HLA-DQα1:160D, instead of the well-described HLA-DRB1 P9 pocket, which is composed of Asn69α, Met73α, Tyr30β, alleles, on ACPA-positive RA in Han ancestry. HLA-DRβ1:37N Asn37β, Val38β and Asp57β. The electrostatic potential surface is an independent protective factor for ACPA-positive RA. The analysis as shown in the upper panel of figure 4B indicates that novel findings are confirmed in an independent case–control P9 pockets of the three haplotypes are negatively charged and cohort by classical HLA genotyping. could benefit electrostatic interaction with epitope P9 arginine. Previously, a number of studies have reported that the major Collectively, all three alleles containing negatively charged P9 genetic risk for ACPA-positive RA came from certain HLA-DRB1 pocket of Asn37 could bind the epitope P9 arginine and thus alleles.37–39 However, other HLA genes have also been linked may result in a decreased RA susceptibility. to ACPA-positive RA.6 11 Thus, it is difficult to draw definitive

Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 777 Rheumatoid arthritis

In the present study, we identified HLA-DQα1:160D as the strongest and independent genetic risk for ACPA-positive RA in Han population. Our novel finding could be rationally explained by the fact that DQα1:160D is encoded by two alleles, that is, DQA1*0302 and DQA1*0303. Although there were several large-scale RA–MHC association studies that included the DQA*03 alleles in the reference panels (European ancestry refer- ence panel and Pan-Asian reference panel, respectively),6 7 11 the DQA1*03 alleles were only presented in either two-digit resolu- tion (DQA1*03) and/or four-digit resolution of DQA1*03:01. The possible explanation for the absence of the DQA1*03:02 and DQA1*03:03 could be that the reference panels of previous large-scale RA-MHC studies were constructed by using the SNP2HLA, which inputs SNPs with allele frequencies ≥1% and removes the markers (eg, HLA alleles and amino acids) with rare allele frequencies (<0.01%).47 Although there was a report showing the frequencies of DQA1*03:02 and DQA1*03:03 were 0.017 and 0.082 in UK Europeans, the study had a very small sample size (n=254) and published after the SNP2HLA study.48 Thus, we speculate that the SNPs tagging DQA1*03:02 and DQA1*03:03 may not be included in the SNP2HLA impu- tation due to the low frequencies (minor allele frequency, MAF <1%). These could help to explain why DQα1:160D and its Figure 4 Dimer of dimer structure of DQA1*0303-DQB1*0401 (A) coding alleles *0302 and *0303 have not been detected or and the electrostatic potential surface and the pocket structure of DRA1- suggested to be risk factors for RA in previous studies, including DRB1*03:01, DQA1*0303-DQB1*0401 and DRA1-DRB1*13:02 (B). (A) a recent large-scale HLA imputation study of ACPA-positive RA 28 Overall dimer of dimer structure of DQA1*0303–DQB1*0401(left panel): in Japanese population. one dimer is composed of DQα1 (green) and DQβ1 (cyan), whereas the Since HLA-DRB1 was well recognised as the top risk for other dimer is composed of DQα1’ (purple) and DQβ1’ (yellow). The ACPA-positive RA in previous studies, we then tested the asso- interface of the dimer of dimer structure (right panel): DQα1 and DQβ1’ ciation at HLA-DRB1 separately. We showed that if DQA1 was residues involved in the interaction are displayed as sticks and their not included in analysis, the allele DRB1*0405, amino acid vari- carbon atoms are coloured in green and yellow, respectively, whereas ants at DRβ1:11, 13, 57, 74 and 71 could be strong risks for nitrogen and oxygen atoms are coloured in blue and red, respectively. ACPA-positive RA, which is consistent with the prevous find- 7 (B) The electrostatic potential surface of DRA1-DRB1*03:01, DRA1- ings. Furthermore, smoking has been reported as a risk factor DRB1*13:01 and DRA1-DRB1*13:02 (upper panel). All structure models linked to RA susceptibility and severity. This risk was increased are displayed by surface. The positive, neutral and negative regions are due to the gene–environment interaction between smoking and 9 coloured in blue, white and red, respectively. The P9 pocket structure of DRB1 alleles. Consistent with previous findings, we also found DRA1-DRB1*03:01, DRA1-DRB1*13:01 and DRA1-DRB1*13:02 (lower the smoking DRB1*0405 carriers had increased radiographic panel). HLA molecules and pocket residues are shown as grey cartoon damage in ACPA-positive patients with RA at an early stage of and green sticks, respectively. HLA, human leukocyte antigen. disease. We further showed that one of DQα1:160 coding alleles DQA1*0303 has high impact on radiographic severity, especially in smoking patients and in early disease. Our data thus support conclusion concerning different HLA genes in susceptibility to the notion that smoking, in the presence of RA-risk genetic back- ACPA-positive RA. Furthermore, several examples of RA genetic ground, may trigger immunity to citrullinated proteins and lead to RA development and accelerate joint damage. heterogeneity have been shown between Asian and Cauca- In the present study, we mainly focused on SNPs, classical sian populations. One of evidence is that the DRB1*0401 and HLA alleles and the individual amino acid variants rather than DRB1*0405 are considered as the top RA-risk alleles in Cauca- amino acid positions, since a particular amino acid(s) may have sians and East Asians, respectively.5 8 10 The non-HLA gene potential biological function(s). Furthermore, different amino PTPN22 has been confirmed to be a common genetic risk for acids at the same position may insert different functions. For multiple autoimmune diseases, including RA.40 The disease-as- example, the residues in DRB1*04:01 (Val11β and His13β) and sociated SNP in PTPN22 (rs2476601 C>T) results in a substi- DRB1*01:01 (Leu11β and Phe13β) would prevent the accom- tution at amino acid 620 (Arg620Trp), consequently leading to modation of arginine (Arg) at the P4 pocket.49 In contrast, a reduced TCR signalling.41 Interestingly, the T allele is almost 42 the P4-Arg is stabilised by H-bonds with Ser11β, Ser13β and absent in African–American and Asian populations, a clear Tyr30β.50 In Raychaudhuri et al’s6 study, they found the five evidence of genetic heterogeneity between Europeans and Asians. individual amino acids in three HLA proteins could explain PADI4 is another good example of genetic heterogeneity in RA most of the association between MHC and seropositive RA in susceptibility. PADI4 is one of several peptidylarginine deiminase Europeans. In the present study, we observed that, although enzymes that catalyse the conversion of arginine residues into DQα1:160 locates away from antigen binding groove and may citrulline, and thus may be linked to production of anticitrulline have little influence on epitope binding, it is adjacent to the antibodies. The association between the PADI4 polymorphism DQα1-DQβ1 dimer of dimer interface. A previous study has and ACPA-positive RA has been reported in several Asian popu- reported the amino acid substitutions in the dimer of dimer lations.43 44 In contrast, conflicting results were seen in European interface of HLA-DRβ1 inhibited CD4+ T cell activation.51 populations.45 46 Here we showed that the additional negative charge introduced

778 Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 Rheumatoid arthritis by D160α further enhances the interaction between DQα1 and Contributors JG, XX and ZL conceptualised and designed the study. XJZ and LK DQβ1 and stabilises the dimer of dimers, which may lead to an participated in the study design and supervised manuscript writing. JG, TZ and HC coordinated and supervised the study teams. JG, XWL, TZ, HL and HC conducted increased T cell activation. data management and manuscript preparation. TZ, XWL, YWZ, XM and HJY The DRβ1:37 residues are located within pocket P9 on the conducted the statistical analyses. HMY, HJJ, JW, LS, LP, LHL, LL and KY participated beta-sheet floor with their side chains oriented into the peptide- in data interpretation and manuscript writing. ML, YDZ, YW, XS, FH, YD, MZ, HJ, binding groove. Modification of DRβ1:37 residues is sufficient XuL, YH, XiL, YY, XW, XZ and YS conducted sample selection and participated in to alter the T cell receptor peptide recognition.52 53 Pocket P9 data management. All coauthors edited and reviewed the final version of the manuscript. of HLA-DRβ has been linked to autoimmune diseases, such as primary Sjogren’s syndrome and primary sclerosing cholan- Funding This work was supported in part by the National Key Basic Research 54 55 Program of China (973 Program) (No 2014CB541901), the National Natural gitis. In the present work, by electrostatic potential surface Science Foundation of China (No 81120108020, No 31711530023, No 31670915, analysis, we showed that three DRβ1:37N encoding alleles, No 31870913, No 31470875, No 31270914, No 31530020, No 31700794, No *03:01, *13:01 and *13:02, have negatively charged P9 pocket, 81401329, No 81771678, No 81471601, No 81671604), the National Key Research which benefits electrostatic interaction and could bind with and Development Program of China (No 2016YFA05022300), Beijing Natural Science Foundation (No 7162192), and Shenzhen Municipal of Government of China 56 also reported epitope P9 arginine. Previously, Kampstra et al (No CXB201108250094A). that DRB1*03:01 had no preference for citrulline in any of Competing interests None declared. the pockets, while it preferred arginine residues as anchors in pockets 6 and 9. Thus, it could at least partly explain the protec- Patient consent for publication Obtained. tive effect of DRβ1:37N for ACPA-positive RA. In contrast, Ethics approval The study was approved by the Medical Ethics Committee of DRβ1:96 hours locates far away from either the antigen binding Peking University People’s Hospital. groove or the dimer of dimer interface. It may indicate that Provenance and peer review Not commissioned; externally peer reviewed. DRβ1:96 hours has little impact on antigen presentation or Data sharing statement Data are available upon reasonable request. All data structural stability of the dimer of dimers. relevant to the study are included in the article or uploaded as online supplementary In summary, by sequencing of the entire MHC region for information. discovery and HLA genotyping for validation in two indepen- dent cohorts, our study demonstrates that HLA-DQα1:160D, References 1 Klareskog L, Padyukov L, Lorentzen J, et al. Mechanisms of disease: genetic instead of HLA-DRB1*0405, confers the greatest independent susceptibility and environmental triggers in the development of rheumatoid arthritis. genetic risk for ACPA-positive RA in Chinese Han. Our study Nat Clin Pract Rheumatol 2006;2:425–33. also illustrates the value of deep sequencing for fine-mapping 2 sparks JA, Chang S-C, Deane KD, et al. Associations of smoking and age with disease risk variants in MHC region. However, one should inflammatory joint signs among unaffected first-degree relatives of rheumatoid bear in mind that current findings are mainly built on statistical arthritis patients: results from studies of the etiology of rheumatoid arthritis. Arthritis Rheumatol 2016;68:1828–38. significance and the functional prediction. Further functional 3 Traylor M, Curtis C, Patel H, et al. Genetic and environmental risk factors for studies are needed to fully understand its contribution to RA rheumatoid arthritis in a UK African ancestry population: the GENRA case-control susceptibility. study. Rheumatology 2017;56:1282–92. 4 Klareskog L, Stolt P, Lundberg K, et al. A new model for an etiology of rheumatoid arthritis: smoking may trigger HLA-DR (shared epitope)-restricted immune reactions Author affiliations to autoantigens modified by citrullination. Arthritis Rheum 2006;54:38–46. 1Department of Rheumatology and Immunology, Peking University People’s Hospital, 5 liu X, Guo J, Jia Y, et al. HLA-DRB1 shared epitope-dependent DR-DQ haplotypes are Beijing, China associated with both anti-CCP-positive and -negative rheumatoid arthritis in Chinese 2Beijing Key Laboratory for Rheumatism Mechanism and Immune Diagnosis Han. PLoS One 2013;8:e71373. (BZ0135), Beijing, China 6 Raychaudhuri S, Sandor C, Stahl EA, et al. Five amino acids in three HLA proteins 3Beijing Genomics Institute (BGI)-Shenzhen, Shenzhen, China explain most of the association between MHC and seropositive rheumatoid arthritis. 4China National GeneBank-Shenzhen, BGI-Shenzhen, Shenzhen, China Nat Genet 2012;44:291–6. 5Shenzhen Digital Life Institute, Shenzhen, China 7 okada Y, Kim K, Han B, et al. Risk for ACPA-positive rheumatoid arthritis is driven by 6iCarbonX, Shenzhen, China shared HLA amino acid polymorphisms in Asian and European populations. Hum Mol 7BNLMS, State Key Laboratory for Structural Chemistry of Unstable and Stable Genet 2014;23:6916–26. Species, Peking-Tsinghua Center for Life Sciences at College of Chemistry and 8 lee H-S, Korman BD, Le JM, et al. Genetic risk factors for rheumatoid arthritis differ in Molecular Engineering, and Center for Quantitative Biology, Peking University, Caucasian and Korean populations. Arthritis Rheum 2009;60:364–71. Beijing, China 9 Terao C, Yano K, Ikari K, et al. Brief report: main contribution of DRB1*04:05 8James D Watson Institute of Genome Sciences, Hangzhou, China among the shared epitope alleles and involvement of Drb1 amino acid position 57 9Institute of Dermatology and Department of Dermatology, No 1 Hospital of Anhui in association with joint destruction in Anti-Citrullinated protein antibody-positive Medical University, Hefei, China rheumatoid arthritis. Arthritis Rheumatol 2015;67:1744–50. 10State Key Laboratory of Quality Research in Chinese Medicine, Macau Institute 10 lee H-S, Lee KW, Song GG, et al. Increased susceptibility to rheumatoid arthritis for Applied Research in Medicine and Health, Macau University of Science and in Koreans heterozygous for HLA-DRB1*0405 and *0901. Arthritis Rheum Technology, Macau, China 2004;50:3468–75. 11Rheumatology Unit, Department of Medicine, Karolinska Institutet, Stockholm, 11 Han B, Diogo D, Eyre S, et al. Fine mapping seronegative and seropositive Sweden rheumatoid arthritis to shared and distinct HLA alleles by adjusting for the effects of 12Laboratory for Autoimmune Diseases, RIKEN Center for Integrative Medical heterogeneity. Am J Hum Genet 2014;94:522–32. Sciences, Yokohama, Japan 12 chen P-L, Shih S-R, Wang P-W, et al. Genetic determinants of antithyroid drug- 13Institute of Dermatology and Department of Dermatology, Huashan Hospital, induced agranulocytosis by human leukocyte antigen genotyping and genome-wide Fudan University, Shanghai, China association study. Nat Commun 2015;6. 14Peking-Tsinghua Center for Life Sciences, Peking University, Beijing, China 13 de Bakker PIW, Raychaudhuri S. Interrogating the major histocompatibility complex 15State Key Laboratory of Natural and Biomimetic Drugs, School of Pharmaceutical with high-throughput genomics. Hum Mol Genet 2012;21:R29–R36. Sciences, Peking University, Beijing, China 14 arnett FC, Edworthy SM, Bloch DA, et al. The American rheumatism association Acknowledgements We thank the staff from the Department of Rheumatology 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum and Immunology, People’s Hospital, Peking University, for recruiting patients and 1988;31:315–24. healthy controls, the staff from the Computing Platform of the Center for Life 15 van der Heijde D. How to read radiographs according to the Sharp/van Der Heijde Sciences, Peking University, for assisting the functional prediction analysis, and method. J Rheumatol 1999;26:743–5. the staff from BGI-Shenzhen who contributed to the technical assistance. We wish 16 du Y, Cui Y, Liu X, et al. Contribution of functional LILRA3, but not nonfunctional to thank the patients and healthy volunteers for their cooperation and for giving LILRA3, to sex bias in susceptibility and severity of anti-citrullinated protein antibody- consent to participate in this study. positive rheumatoid arthritis. Arthritis Rheumatol 2014;66:822–30.

Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 779 Rheumatoid arthritis

17 cao H, Wu J, Wang Y, et al. An integrated tool to study MHC region: accurate SNV 38 Holoshitz J. The rheumatoid arthritis HLA-DRB1 shared epitope. Curr Opin Rheumatol detection and HLA genes typing in human MHC region using targeted high- 2010;22:293–8. throughput sequencing. PLoS One 2013;8:e69388. 39 okada Y, Terao C, Ikari K, et al. Meta-analysis identifies nine new loci associated with 18 Zhou F, Cao H, Zuo X, et al. Deep sequencing of the MHC region in the Chinese rheumatoid arthritis in the Japanese population. Nat Genet 2012;44:511–6. population contributes to studies of complex disease. Nat Genet 2016;48:740–6. 40 gregersen PK, Lee H-S, Batliwalla F, et al. PTPN22: setting thresholds for 19 cao H, Wang Y, Zhang W, et al. A short-read multiplex sequencing method for reliable, autoimmunity. Semin Immunol 2006;18:214–23. cost-effective and high-throughput genotyping in large-scale studies. Hum Mutat 41 Vang T, Congia M, Macis MD, et al. Autoimmune-associated lymphoid tyrosine 2013;34:1715–20. phosphatase is a gain-of-function variant. Nat Genet 2005;37:1317–9. 20 cordovado SK, Hancock LN, Simone AE, et al. High-resolution genotyping of HLA- 42 mori M, Yamada R, Kobayashi K, et al. Ethnic differences in allele frequency of DQA1 in the GoKinD study and identification of novel alleles HLA-DQA1*040102, autoimmune-disease-associated SNPs. J Hum Genet 2005;50:264–6. HLA-DQA1*0402 and HLA-DQA1*0404. Tissue Antigens 2005;65:448–58. 43 suzuki A, Yamada R, Chang X, et al. Functional haplotypes of PADI4, encoding 21 patsopoulos NA, Barcellos LF, Hintzen RQ, et al. Fine-mapping the genetic association citrullinating enzyme peptidylarginine deiminase 4, are associated with rheumatoid of the major histocompatibility complex in multiple sclerosis: HLA and non-HLA arthritis. Nat Genet 2003;34:395–402. effects. PLoS Genet 2013;9. 44 Kang CP, Lee H-S, Ju H, et al. A functional haplotype of thePADI4 gene associated 22 payami H, Joe S, Farid NR, et al. Relative predispositional effects (RPEs) of marker with increased rheumatoid arthritis susceptibility in Koreans. Arthritis Rheum alleles with disease: HLA-DR alleles and Graves disease. Am J Hum Genet 2006;54:90–6. 1989;45:541–6. 45 plenge RM, Padyukov L, Remmers EF, et al. Replication of putative candidate-gene 23 Hinks A, Cobb J, Marion MC, et al. Dense genotyping of immune-related disease associations with rheumatoid arthritis in >4,000 samples from North America and regions identifies 14 new susceptibility loci for juvenile idiopathic arthritis. Nat Genet Sweden: association of susceptibility with PTPN22, CTLA4, and PADI4. Am J Hum 2013;45:664–9. Genet 2005;77:1044–60. 24 imielinski M, Baldassano RN, GriffithsA , et al. Common variants at five new 46 Burr ML, Naseem H, Hinks A, et al. PADI4 genotype is not associated with loci associated with early-onset inflammatory bowel disease. Nat Genet rheumatoid arthritis in a large UK Caucasian population. Ann Rheum Dis 2009;41:1335–40. 2010;69:666–70. 25 Eswar N, Webb B, Marti-Renom MA, et al. Comparative protein structure modeling 47 Jia X, Han B, Onengut-Gumuscu S, et al. Imputing amino acid polymorphisms in using Modeller. Curr Protoc Bioinformatics 2006;15:5.6.1–5.6.30. human leukocyte antigens. PLoS One 2013;8:e64683. 26 Brown JH, Jardetzky TS, Gorga JC, et al. Three-dimensional structure of the human 48 lemin AJ, Darke C. Prevalence of HLA-DQA1 alleles and haplotypes in blood donors class II histocompatibility antigen HLA-DR1. Nature 1993;364:33–9. resident in Wales. Int J Immunogenet 2014;41:480–3. 27 ghosh P, Amaya M, Mellins E, et al. The structure of an intermediate in class II MHC 49 James EA, Rieck M, Pieper J, et al. Citrulline-specific h1T cells are increased in maturation: clip bound to HLA-DR3. Nature 1995;378:457–62. rheumatoid arthritis and their frequency is influenced by disease duration and 28 okada Y, Suzuki A, Ikari K, et al. Contribution of a non-classical HLA gene, HLA-DOA, therapy. Arthritis Rheumatol 2014;66:1712–22. to the risk of rheumatoid arthritis. Am J Hum Genet 2016;99:366–74. 50 scally SW, Law S-C, Ting YT, et al. Molecular basis for increased susceptibility of 29 cherry RJ, Wilson KM, Triantafilou K, et al. Detection of dimers of dimers of human leukocyte antigen (HLA)-DR on the surface of living cells by single-particle Indigenous North Americans to seropositive rheumatoid arthritis. Ann Rheum Dis fluorescence imaging. J Cell Biol 1998;140:71–9. 2017;76:1915–23. 30 schafer PH, Pierce SK. Evidence for dimers of MHC class II molecules in B lymphocytes 51 lindstedt R, Monk N, Lombardi G, et al. Amino acid substitutions in the putative and their role in low affinity T cell responses. Immunity 1994;1:699–707. MHC class II "dimer of dimers" interface inhibit CD4+ T cell activation. J Immunol 31 schafer PH, Pierce SK, Jardetzky TS. The structure of MHC class II: a role for dimer of 2001;166:800–8. dimers. Semin Immunol 1995;7:389–98. 52 chen YZ, Matsushita S, Nishimura Y. A single residue polymorphism at DR beta 37 32 nydam T, Wade TK, Yadati S, et al. Mutations in MHC class II dimer of dimers contact affects recognition of peptides by T cells. Hum Immunol 1997;54:30–9. residues: effects on antigen presentation. Int Immunol 1998;10:1237–49. 53 Kaneko T, Obata F. Allogeneic recognition of HLA-DRB1*0406 by T cells with HLA- 33 Viatte S, Plant D, Raychaudhuri S. Genetics and epigenetics of rheumatoid arthritis. DRB1*0403: role of amino acid residue 37 on the beta sheet in T cell recognition. Nat Rev Rheumatol 2013;9:141–53. Immunobiology 1996;195:261–70. 34 scally SW, Petersen J, Law SC, et al. A molecular basis for the association of the HLA- 54 Huang R, Yin J, Chen Y, et al. The amino acid variation within the binding pocket 7 DRB1 locus, citrullination, and rheumatoid arthritis. J Exp Med 2013;210:2569–82. and 9 of HLA-DRB1 molecules are associated with primary Sjögren’s syndrome. J 35 okada Y, Suzuki A, Yamada R, et al. HLA-DRB1*0901 lowers anti-cyclic citrullinated Autoimmun 2015;57:53–9. peptide antibody levels in Japanese patients with rheumatoid arthritis. Ann Rheum 55 Hov JR, Kosmoliaptsis V, Traherne JA, et al. Electrostatic modifications of the human Dis 2010;69:1569–70. leukocyte antigen-DR P9 peptide-binding pocket and susceptibility to primary 36 Furuya T, Hakoda M, Ichikawa N, et al. Differential association of HLA-DRB1 alleles in sclerosing cholangitis. Hepatology 2011;53:1967–76. Japanese patients with early rheumatoid arthritis in relationship to autoantibodies to 56 Kampstra ASB, van Heemst J, Moustakas AK, et al. The increased ability to present cyclic citrullinated peptide. Clin Exp Rheumatol 2007;25:219–24. citrullinated peptides is not unique to HLA-SE molecules: arginine-to-citrulline 37 orozco G, Rueda B, Martin J. Genetic basis of rheumatoid arthritis. Biomed conversion also enhances peptide affinity for HLA-DQ molecules. Arthritis Res Ther Pharmacother 2006;60:656–62. 2016;18.

780 Guo J, et al. Ann Rheum Dis 2019;78:773–780. doi:10.1136/annrheumdis-2018-214725 Early arthritis

Clinical science MRI inflammation of the hand interosseous tendons occurs in anti-CCP-positive at-risk individuals and may precede the development of clinical synovitis Kulveer Mankia,‍ ‍ 1,2 Maria-Antonietta D’Agostino,1,3 Emma Rowbotham,2,4 Elizabeth MA Hensor,1 Laura Hunt,1,2 Ingrid Möller,5 Misabel Miguel,6 José Ramon Mérida-Velasco,7 Jorge Murillo-González,7 Esperanza Naredo,8 Jacqueline Leong Nam,2 Ai Lyn Tan,1,2 Jane E Freeston,2 Andrew Grainger,2,4 Paul Emery‍ ‍ 1,2

Handling editor Josef S Abstract Key messages Smolen Interosseous tendon inflammation (ITI) has been described in rheumatoid arthritis (RA). Whether ITI occurs ►► Additional material is What is already known about this subject? published online only. To view in at-risk individuals before the onset of clinical synovitis ►► Individuals at risk of rheumatoid arthritis (RA), please visit the journal online is unknown. including those with anti-cyclic citrullinated (http://dx.​ ​doi.org/​ ​10.​1136/​ Objectives To investigate, by MRI, ITI in anti-cyclic protein (CCP) antibodies, often develop annrheumdis-2018-​ ​214331). citrullinated peptide (CCP)-positive at-risk individuals musculoskeletal symptoms before the onset of (CCP +at risk) and to describe the anatomy, prevalence For numbered affiliations see clinical or subclinical synovitis. end of article. and clinical associations across the RA continuum. ► MRI tenosynovitis of the flexor tendons is a Methods Hand MRI was performed in 93 CCP + ► prevalent finding in at-risk individuals. Correspondence to at risk, 47 early RA (ERA), 28 established ’late’ RA ► There is a high prevalence of interosseous Professor Paul Emery, Leeds (LRA) and 20 healthy controls (HC) and scored for ITI, ► tendon inflammation (ITI) in the hands of Institute of Rheumatic and flexor tenosynovitis (TSV) and RA MRI scoring at the Musculoskeletal Medicine, Leeds patients with established RA. LS2 9JT, UK; metacarpophalangeal joints (MCPJs). Cadaveric and histological studies were performed to explore the p.​ ​emery@leeds.​ ​ac.​uk What does this study add? anatomical basis for MRI ITI. ► MRI ITI may be associated with clinical Received 23 August 2018 Results The proportion of subjects with ITI and the ► Revised 1 February 2019 metacarpophalangeal joint tenderness and number of inflamed interosseous tendons (ITs) increased Accepted 4 March 2019 could explain symptoms in at-risk individuals along the disease continuum (p<0.001): 19% of CCP Published Online First who do not have clinical synovitis. 23 March 2019 +at risk, 49% of ERA and 57% of LRA had ≥1 IT ► ITI occurs in anti-CCP-positive at-risk individuals inflamed .I TI was not found in any HC. ITI was more ► without clinical synovitis. frequently identified in tender MCPJs compared with ► Cadaveric and histological studies show that nontender MCPJs (28% vs 12%, respectively). No ► the interosseous tendons do not have a tendon IT tenosynovial sheath was identified in cadavers on sheath and do not directly communicate with dissection or histological studies suggesting MRI findings the joint capsule. represent peritendonitis. Dye studies indicated no communication between the IT and the joint. Conclusions iTI occurs in CCP + at-risk individuals and can precede the onset of clinical synovitis. The ITs may tenosynovitis appears to be an early and prevalent MRI lesion in at-risk individuals, both in patients be important nonsynovial extracapsular targets in the 4 development and progression of RA. with clinically suspect arthralgia (CSA) and ACPA-positive at-risk individuals.5 Tenosynovitis of the wrist and finger flexor tendons is also the stron- gest predictor of progression to arthritis in patients Introduction with CSA.4 Furthermore, tenosynovitis is a frequent Rheumatoid arthritis (RA) is conventionally consid- presentation of RA, causes significant disability and ered as a disease of the synovial joints. However, occurs in remission, where it is predictive of flares.6 individuals at risk of RA (including those with Unlike tenosynovitis of the wrist and finger anti-citrullinated protein antibodies, ACPA) often flexor tendons, the involvement of other extra- © Author(s) (or their develop musculoskeletal (MSK) symptoms before capsular structures critical to hand function has employer(s)) 2019. No commercial re-use. See rights the onset of clinical synovitis (ie, synovitis deter- not been completely elucidated. For example, the and permissions. Published mined on physical examination) or subclinical syno- interosseous muscles of the hands are regarded by BMJ. vitis (ie, synovitis determined on high-resolution as the cornerstone of hand function7 and their imaging but not physical examination).1–3 Extra- tendons run adjacent to the metacarpophalan- To cite: Mankia K, D’Agostino M-A, capsular structures may be important early targets geal joints (MCPJs). Occupation-related overuse Rowbotham E, et al. of RA-related inflammation. This may involve the injuries of these muscles and their tendons have Ann Rheum Dis tenosynovium or potentially even tendons without been described.8 A recent study identified a high 2019;78:781–786. a tenosynovial sheath (ie, peritendonitis). Indeed, frequency of inflammation of the interosseous

Mankia K, et al. Ann Rheum Dis 2019;78:781–786. doi:10.1136/annrheumdis-2018-214331 781 Early arthritis tendons (ITs) in patients with RA.9 Whether these tendons details). For all CCP +at risk, patients with ERA and HC, MRI become inflamed in at-risk individuals and whether they may be scans were performed on a 3T Siemans Verio system (Erlangen, responsible for symptoms in these subjects is unknown. We there- Germany). Due to logistical issues, all patients with LRA were fore sought to comprehensively investigate interosseous tendon scanned on a 1.5T Siemans Avanto (see online supplementary inflammation (ITI) across the RA continuum by describing the materials for further details of MRI systems). All MRI scans were anatomy, histology, prevalence and clinical associations in both independently scored for the presence or absence of ITI by two anti-cyclic citrullinated peptide positive at-risk (CCP +at risk) MSK radiologists (AG and ER). The radiologists were blinded individuals and patients with RA. to all patient details and studies were presented to them in a The main aim of this study was to assess whether MRI ITI is random order with respect to disease status. Any discrepancies present in CCP +at-risk individuals without clinical synovitis. As in scoring between the two radiologists were reviewed at the end we hypothesised MRI ITI would be present in CCP +at risk, the of the reporting exercise and a consensus score was agreed. The secondary aim of the study was to determine whether MRI ITI consensus score was then used in the analysis. ITI was defined as is associated with clinical features in CCP +at-risk individuals. the presence of enhancing tissue surrounding the tendon which was evident in two planes.9 Eight tendons were assessed in each hand (ie, each MRI study): the tendons of the dorsal inter- Methods 1–4 1–3 Design ossei, the palmar interossei and the abductor digiti minimi. A retrospective analysis of MRI and clinical data of CCP +at-risk In addition to IT scoring, all MRI studies were scored for the individuals, patients with RA and healthy controls (HCs) were presence of synovitis, bone marrow oedema (BME) and erosions at MCPJs 2–5 according to the Outcome Measures in Rheuma- undertaken. 12 In order to understand whether ITI should be considered a tology RA MRI scoring (RAMRIS) system. Tenosynovitis of the tenosynovitis or a peritendonitis, cadaveric and histological flexor digitorum tendons was scored dichotomously as ‘present’ studies were performed to explore the anatomical basis for MRI or ‘absent’ in keeping with the dichotomous scoring of ITI. ITI. In the majority of studies (including all CCP +at risk studies), RAMRIS and tenosynovitis scoring were performed by the same radiologists that scored ITI (AG and ER), and a consensus score Clinical subjects was used in the analysis. However, in a small number of cases, Anti-CCP-positive at-risk individuals with MSK symptoms but only scores from a single radiologist were available and these no clinical synovitis (CCP +at risk), disease-modifying anti-rheu- were therefore used in the analysis. In all cases, the radiologists matic drug (DMARD)-naïve early RA patients (ERA), established were blinded to all patient details and studies were presented to ‘late’ RA patients (LRA) and asymptomatic HC were recruited them in a random order according to disease status. at Chapel Allerton Hospital, Leeds, UK. Informed consent was received from all subjects. The Leeds CCP +at risk cohort has been previously Anatomical study 3 10 described. Patients aged >18 years presenting to their general To assess the relationship between the ITs and the adjacent practitioners or other health professionals with new-onset MSK MCPJ, 20 fresh hand specimens from anonymous donors were symptoms (eg, arthralgia, epicondylitis, subacromial bursitis) obtained (Anatomy Department, Universitat de Barcelona). The but no clinical synovitis were invited to participate. Primary first dorsal interosseous muscle and tendon were identified by care recruitment was adopted by the UK Primary Care Clinical ultrasound (US) from the myotendinous junction to the capsular Research Network. Anti-CCP testing was performed centrally insertion. The adjacent second MCPJ was also identified on US. using the commercially available anti-CCP2 (immunocap Using US guidance, a green dye (green acrylic paint mixed 50:50 method; Phadia, Sweden). Those with a positive test were invited with water) was injected along the IT proximal to its capsular to attend a dedicated research clinic in Leeds where recruitment insertion. Similarly, a blue dye (blue acrylic paint mixed 50:50 for the current MRI study was undertaken. Patients from the with water) was injected into the second MCPJ space using US Leeds early arthritis clinic who were anti-CCP positive but did guidance. Specimens were then frozen and transverse sections not have clinical synovitis were also recruited. were made. Ethical approval for this study was obtained by the ERA patients had not received DMARDs at the time of imaging. Universitat de Barcelona, Spain. LRA patients had ≥1 year disease duration (total duration of To further confirm the absence of a tendon sheath around the reported symptoms), were anti-CCP and/or rheumatoid factor ITs, two embalmed adult human cadaver hands were dissected. positive and had active disease (disease activity score 28≥3.2). The distal attachment of the first dorsal IT was identified by These patients had received ≥1 DMARD and underwent MRI dissection and processed for histological analysis (Department of 11 prior to receiving rituximab as part of a separate study. Anatomy and Embryology, Complutense University of Madrid). HCs included staff members from the University of Leeds and The specimens were fixed in 10% formalin and embedded in members of the local community. None of the controls had any paraffin. Each specimen was sliced as a whole on a single plane. MSK symptoms at the time of recruitment. Sections of 7 µm were cut and stained with haematoxylin–eosin Demographic and clinical details were recorded at the time of and Mallory’s trichrome dye. Additionally, the first dorsal IT imaging assessments. All subjects (except HC) underwent MSK insertion was investigated in two human fetuses aged 14 weeks examination by a rheumatologist prior to the imaging assess- (greatest length 115 mm in both). The specimens belong to the ments and swollen and tender MCPJ counts were recorded. collection kept at the Department of Anatomy and Embryology Blood testing for C-reactive protein (CRP) was performed at the at UCM. Sections of 12 µm were cut and stained with haematox- clinical visit. ylin–eosin. Digital photomicrographs were taken with a Nikon DXM 1200 labour microscope (Nikon Corp., Tokyo, Japan) MRI and edited using Act One and Adobe Photoshop CS6 software. All subjects underwent MRI of the most symptomatic or domi- Ethical approval for this study was obtained by Complutense nant hand and wrist (see online supplementary materials for full University of Madrid.

782 Mankia K, et al. Ann Rheum Dis 2019;78:781–786. doi:10.1136/annrheumdis-2018-214331 Early arthritis

Table 1 Baseline characteristics of all participants HC n=20 CCP +at risk n=93 ERA n=47 LRA n=28 Age: mean (SD) 47.9 (10.3) 48.1 (12.3) 55.6 (15.3) 51.5 (13.3) Female 75% 69% 74% 93% Symptom duration, months: median (IQR) NA 14 (6, 40) 7 (3, 14) 96 (42, 216) Tender joint count 0 100% 83% 43% 25% 1 – 6% 26% 4% 2 – 4% 17% 32% 3 – 3% 9% 4% 4 – 3% 6% 36% Swollen joint count 0 100% 100% 43% 21% 1 – – 23% 14% 2 – – 28% 25% 3 – – 4% 21% 4 – – 2% 18% Anti-CCP+: n(%) NA 93 (100) 35 (74) 26/27 (96) Rheumatoid factor +: n(%) NA 44/92 (48) 28/46 (61) 19/24 (79) EMS (mins): median (IQR) NA 10 (0, 30) 120 (30, 180) 60 (30, 150) CRP (mg/L): median (IQR) NA <5 (<5,<5) 8 (<5, 19) 11 (5, 27) The frequency of tender and swollen joints and CRP level increased along the RA continuum with increasing disease duration. Joint counts refer to 2nd–5th metacarpophalangeal joints (MCPJs). CRP, C-reactive protein (mg/dL); EMS, early morning stiffness duration (minutes); ERA, early RA patients; HC, healthy controls;LRA, late RA patients; CCP +at risk, anti-CCP- positive at-risk individuals.

Statistical analysis LRA patients had inflammation of at least one IT. Of note, no ITI The number of inflamed ITs was compared between subject was identified in any HC. The number of inflamed IT per patient groups using a Jonckheere-Terpstra test in order to assess whether also increased along the RA continuum (Jonckheere-Terpstra the median number of sites increased according to disease along J=5.90, p<0.001). In all patient groups, ITs associated with the RA continuum. In all analyses where MRI-detected synovitis, MCPJ2 and MCPJ5 were most frequently affected (table 2), BME and erosions are reported at joint level, only those joints although trends with respect to increasing ITI prevalence across where inflammation scores were over and above that seen in the the groups were similar in all locations. Of note, ITI was iden- same anatomical location in symptom-free controls of the same tified in both anti-CCP-positive and anti-CCP-negative ERA age range were included (ie, true subclinical inflammation).13 patients: 17/34 (50%) of anti-CCP-positive ERA patients had ITI Adjustment for symptom-free controls was performed in order compared with 6/13 (46%) of anti-CCP-negative ERA patients. to account for the observation that MRI-detected inflammation The two radiologists showed an excellent level of agreement is prevalent in the general population, especially in older people for the identification of ITI in this study: kappa (k)=0.893 and at certain anatomical sites, for example, MCPJs 2 and 3.13 (SE=0.025), comparable to that demonstrated in the original True subclinical inflammation was considered to be present if description of MRI ITI (kappa, k=0.91; SE=0.03).9 that joint or bone was (i) scored positive for inflammation and (ii) the score obtained at that joint or bone was present in <5% Association between ITI and clinical features 13 of age-matched symptom-free persons. Tenosynovitis could not The prevalence of MCPJ tenderness increased with disease dura- be adjusted for in this way as this was scored dichotomously (ie, tion (tables 1 and 3). ITI was more frequently identified in tender present or absent) rather than semiquantitatively in this study. MCPJs compared with nontender MCPJs; 28% of all tender MCPJs had an associated ITI, whereas only 12% of nontender Results MCPJs had an associated ITI (table 3). Synovitis, BME, erosions Clinical study and tenosynovitis were also more frequently identified in tender Baseline characteristics MCPJs compared with nontender MCPJs (table 3). In all, 93 CCP +at risk, 47 ERA, 28 LRA and 20 HC were Early morning stiffness (EMS) duration (minutes) was not included. The frequency of tender and swollen MCPJs and CRP markedly different in CCP +at risk with ITI (ie, at least one level increased along the RA continuum with increasing disease inflamed IT at any site) compared with those without ITI duration (table 1). Overall, MRI inflammation (synovitis, tenosy- (median IQR 10 (0, 60) vs 5 (0, 10)). Similarly, the presence of novitis, BME, erosions) at the MCPJs also increased in a similar synovitis, BME, tenosynovitis or erosions at ≥1 MCPJ did not fashion (see online supplementary materials). appear to be associated with EMS duration (online supplemen- tary materials). The number of affected MCPJs was not associ- MRI Interosseous tendon inflammation ated with EMS duration for any of the MRI features. MRI ITI was observed as enhancing tissue around the midportion of the tendon, proximal to the enthesis (figure 1). The proportion Anatomical study of patients with ITI increased along the RA continuum (table 2); Detailed dissection of fresh hand specimens revealed no iden- 18/93 (19%) CCP +at risk, 23/47 (49%) ERA and 16/28 (57%) tifiable synovial tendon sheath around the IT of the cadaveric

Mankia K, et al. Ann Rheum Dis 2019;78:781–786. doi:10.1136/annrheumdis-2018-214331 783 Early arthritis

Table 2 Interosseous tendon inflammation according to RA status Interosseous HC n=20 CCP +at risk ERA n=47 LRA n=28 tendon n=93 inflammation present at MCP2 - (0) 11% (10) 34% (16) 39% (11) MCP3 - (0) 4% (4) 23% (11) 21% (6) MCP4 - (0) 1% (1) 15% (7) 25% (7) MCP5 - (0) 9% (8) 28% 32% (9) Dorsal tendons - (0) 12% (11) 45% (21) 50% (14) Palmar tendons - (0) 10% (9) 32% (15) 32% (9) Any site (patient - (0) 19% (18) 49% (23) 57% (16) level) Any site (site level) - (0) 3% (24/744) 15% (55/376) 18% (40/224) Number of sites per patient (n/8) 0 100% (20) 81% (85) 51% (24) 43% (12) 1 – 14% 17% (8) 21% (6) 2 – 4% (4) 15% (7) 18% (5) 3 – 1% (1) 6% (3) – 4 –– 2% (1) 11% (3) ≥5 –– 9% (4) 7% (2) Dorsal tendons refers to 1st–4th dorsal interosseous tendons and abductor digiti minimi. Palmar tendons refers to 1st–3rd palmar interosseous tendons. ERA, early RA patients; HC, healthy controls;LRA, late RA patients; MCP, metacarpophalangeal; RA, rheumatoid arthritis; CCP +at risk, anti-CCP-positive at-risk individuals.

examination of the first dorsal IT from a fetal hand (figure 2A, B) and adult hand (figure 2C–G) confirmed the presence of an epitendon, but no synovial sheath surrounding the tendon could be identified. Figure 1 IT inflammation in anti-CCP +at-risk individuals. (A) Inflammation of the first and second dorsal ITs and first palmar IT Discussion (arrows). MCPJ inflammation is present at the second and third MCPJs Individuals at risk of RA often experience a prodrome of joint and flexor tenosynovitis of the second finger is also observed (asterix). pain and stiffness before the onset of clinical synovitis. This may (B) Transverse section of cadaveric second MCPJ following ultrasound- reflect the earliest phase of RA inflammation, which in many guided dye injections. Green dye has accumulated around the superficial cases will progress to full-blown disease. However, the cause of margin of the interosseous muscle but remains outside the MCPJ these symptoms and their relevance to disease progression are capsule. Blue dye injected into the joint remains within the joint capsule. not clear. Understanding this phase of disease is likely to deliver (C) Inflammation of the third palmar IT. Flexor tenosynovitis is also important insights into pathogenesis and also inform future present (asterix). (D) Inflammation of the first dorsal IT (white arrow). preventative strategies. There is no inflammation in the adjacent MCPJ or of the second dorsal IT (black arrow). (E) Isolated inflammation of the second dorsal IT (arrow). No adjacent synovitis or tenosynovitis is seen. (F) Illustration Table 3 Clinical and MRI features according to clinical tenderness of the anatomy of the dorsal ITs of the hand. The interossei originate at MCP joints (MCPJs 2–5) for CCP+ and RA subjects (n=162) from the medial and lateral aspects of the metacarpals and attach into the extensor hood and proximal phalanx of each finger. (G) Illustration Tender MCPJs Nontender MCPJs n=507 n=141 of the anatomy of the palmar ITs. The three palmar interossei attach to the index, ring and little fingers. CCP, cyclic citrullinated peptide; ITs, Age: mean (SD) 50.6 (13.7) 52.2 (13.9) Interosseous tendons; MCPJ, metacarpophalangeal joint. Female 71% (358) 84% (118) Group CCP +at risk 66% (337) 18% (25) ERA 26% (132) 37% (52) specimens. Identification of the blue and green dye on transverse LRA 7% (38) 38% (54) sections was achieved in 13/20 (65%) specimens (figure 1b). Synovitis >0 20% (101) 37% (52) Blue dye was localised in the intra-articular space and was not Synovitis >1 6% (30) 19% (27) seen outside the joint capsule. In contrast, green dye was local- BME >0 4% (21) 14% (20) ised around the ITs and interosseous muscle and was not seen Tenosynovitis >0 30% (153) 49% (69) within the joint capsule. These findings suggest an absence of Erosions>0 3% (17) 13% (18) communication between intracapsular synovial membrane and ITI 12% (59) 28% (40) the adjacent IT. BME, bone marrow oedema; ERA, early RA patients; ITI, interosseous tendon Cadaveric dissection revealing the first dorsal IT and its attach- inflammation; LRA, late RA patients;MCPJS, metacarpophalangeal joints;RA, ment is shown in online supplementary figure 1. Histological rheumatoid arthritis; CCP +at risk, anti-CCP-positive at-risk individuals.

784 Mankia K, et al. Ann Rheum Dis 2019;78:781–786. doi:10.1136/annrheumdis-2018-214331 Early arthritis

ITI was identified in 19% of CCP +at-risk individuals with increasing prevalence with RA disease progression. The number of inflamed tendons in each patient also increased with disease progression. Of note, we did not identify any ITI in the HCs. This is in keeping with the observation that MRI tendon inflam- mation is rarely seen in healthy subjects.13 The association between specific clinical features (EMS dura- tion, MCPJ tenderness) and MRI inflammation in anti-CCP-pos- itive at-risk individuals has not previously been reported. However, Van Steenbergen et al have previously reported no association between EMS duration and MRI inflammation in individuals with CSA.14 While we found no association between ITI and EMS duration, there was an increased frequency of ITI in tender MCPJs compared with nontender MCPJs. This is an interesting finding, particularly as small joint tenderness is predictive of arthritis development in CCP +at-risk subjects.10 The potential association between clinical features and ITI should be further explored in a prospective study, where IT tenderness and flexor tendon tenderness at the MCPJs could be specifically assessed by clinical examination. In line with our findings, others have demonstrated that tendon inflammation is prevalent in the hands of patients with early arthritis15 and at-risk individuals.4 5 In these studies, tendon inflammation was mainly characterised by tenosynovitis, although isolated peritendinous inflammation of the digital extensor tendons was described in patients with ERA15; our current data extend the concept of early extracapsular inflam- mation and suggest nonsynovial tendon inflammation, that is, peritendonitis, may also be an important disease target in at-risk individuals who do not have clinical synovitis. Figure 2 Histological examination of the first dorsal IT. (A) rontalF While the dye injections in our cadaveric study revealed no section of the metacarpophalangeal joint of a human fetus (14 weeks communication between the IT and adjacent MCPJ, the speci- after conception) with haematoxylin–eosin staining. (B) High-power mens used were from anonymous donors who did not appear to magnification of black square in A: 1, superficial first dorsal IT; 2, deep have RA. It is not known if this would be different in the setting first IT inserting into the base of the proximal phalanx; 3, insertion of of established RA-related MCPJ inflammation. the deep tendon of the first interosseous into the joint capsule. MT, Although we have identified ITI as an early feature in the RA second metatarsal head; PH, proximal phalanx; PS, synovial plica; V, disease continuum, it is not clear how specific this lesion is for blood vessel. (C) Frontal section of the myotendinous junction of the anti-CCP-related inflammation (or autoimmune-related inflam- first dorsal IT in an adult cadaver with Mallory’s trichrome stain. The mation). Indeed, we identified a similar prevalence of ITI in tendon is formed by CF arranged longitudinally. The EPT surrounds anti-CCP-positive ERA patients compared with anti-CCP-nega- the tendon (illustrated by black line). (D) High-power magnification of tive ERA patients, suggesting ITI is unlikely to be an ACPA-spe- the square in C shows muscle fascicles separated by endomysium and cific phenomenon. Furthermore, although MRI peritendinous surrounded by PM. PM is continuous with peritendon that surrounds CF. inflammation has been reported in established RA, this finding (E) Frontal section of the distal insertion of the first dorsal IT in an adult is not necessarily disease specific. It is also possible that ITI may cadaver with Mallory’s trichrome stain. 1, Superficial first dorsal IT; 2, occur due to mechanical factors in some patients with hand deep first dorsal IT. (F) High-power magnification of the black square arthralgia. Future work should, therefore, investigate whether in E shows the EPT containing blood vessels. There is no tendon sheath ITI also occurs in symptomatic anti-CCP-negative patients with surrounding the EPT. (G) High-power magnification of the red square in arthralgia and also in other rheumatic diseases. E. Blood vessels are seen in the EPT with no surrounding tendon sheath. In conclusion, this study has identified MRI inflammation of CF, collagen fascicles; EPT, epitendon; IT, interosseous tendon; PM, the hand ITs in CCP +at-risk individuals who do not have clin- perimysium ical synovitis. This lesion was more frequently seen in tender MCPJs compared with nontender MCPJs. These data suggest that the ITs may be an early nonsynovial extracapsular target This study is the first to demonstrate that the IT of the hands of RA inflammation. Further longitudinal data are needed to are inflamed in anti-CCP-positive at-risk individuals and may investigate whether ITI predicts the development of clinical and represent an extracapsular cause for early symptoms in the subclinical synovitis in individuals at risk of RA. absence of clinical synovitis. We have also demonstrated that, on cadaveric dissection and histological examination, these tendons Author affiliations do not have a tenosynovial sheath and do not directly communi- 1Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds, cate with the joint capsule. This, alongside the observation that Leeds, UK 2 fluid is not seen around these tendons on MRI,9 suggests MRI NIHR Leeds Biomedical Research Centre, Leeds, UK 3Department of Rheumatology, Ambroise Paré Hospital, APHP, Université Versailles- ITI represents an additional nonsynovial target of inflamma- Saint-Quentin en Yvelines, Boulogne-Billancourt, France tion in the RA continuum (ie, peritendonitis rather than a true 4Department of Radiology, Leeds Teaching Hospitals NHS Trust, Leeds, UK tenosynovitis). 5Instituto Poal de Reumatología, University of Barcelona, Barcelona, Spain

Mankia K, et al. Ann Rheum Dis 2019;78:781–786. doi:10.1136/annrheumdis-2018-214331 785 Early arthritis

6Anatomy Department, Universitat de Barcelona, Barcelona, Spain 4 van Steenbergen HW, Mangnus L, Reijnierse M, et al. Clinical factors, anticitrullinated 7Department of Anatomy and Emryology, Universidad Complutense de Madrid, peptide antibodies and MRI-detected subclinical inflammation in relation Madrid, Spain to progression from clinically suspect arthralgia to arthritis. Ann Rheum Dis 8Department of Rheumatology, Bone and Joint Research Unit, Universidad Autonoma 2016;75:1824–30. de Madrid, Madrid, Spain 5 Kleyer A, Krieter M, Oliveira I, et al. High prevalence of tenosynovial inflammation before onset of rheumatoid arthritis and its link to progression to RA-A combined Acknowledgements We would like to thank Dr Daniel Glinatsi, Prof Mikkel MRI/CT study. Semin Arthritis Rheum 2016;46:143–50. Ostergaard and Dr Paul Bird for contributing clinical data. We would like to thank 6 Filippou G, Sakellariou G, Scirè CA, et al. The predictive role of ultrasound- Rob Evans and Brian Chaka for radiography support. detected Tenosynovitis and joint synovitis for flare in patients with rheumatoid Contributors KM recruited patients, collected and analysed the data and wrote arthritis in stable remission. Results of an Italian multicentre study of the Italian the manuscript. MADA designed the study and helped prepare the manuscript. ER Society for rheumatology group for ultrasound: the starter study. Ann Rheum Dis scored the MRI scans. EMAH led the statistical analysis. LH recruited patients and 2018;77:1283–9. collected data. IM, MP, JRM, JN and EN performed the cadaveric and histological 7 liss FE. The interosseous muscles: the foundation of hand function. Hand Clin studies. JLN was one of the study clinicians. ALT and JEF recruited patients and 2012;28:9–12. collected clinical data. AJG and PE designed and led the study. All coauthors read 8 sakai N. Interosseous muscle pain in the pianist’s hand: a description of 27 cases of and revised the manuscript. ’musician’s hand’. Med Probl Perform Art 2007;22:24–5. Funding The study was supported by the National Institute for Health Research 9 Rowbotham EL, Freeston JE, Emery P, et al. The prevalence of tenosynovitis of the (NIHR) Leeds Clinical Research Facility. interosseous tendons of the hand in patients with rheumatoid arthritis. Eur Radiol 2016;26:444–50. Competing interests None declared. 10 Rakieh C, Nam JL, Hunt L, et al. Predicting the development of clinical arthritis in anti- Patient consent for publication Not required. CCP positive individuals with non-specific musculoskeletal symptoms: a prospective observational cohort study. Ann Rheum Dis 2015;74:1659–66. Ethics approval NHS Health Research Authority National Research Ethics Service 11 Vital EM, Dass S, Buch MH, et al. An extra dose of rituximab improves clinical Committee Yorkshire & the Humber – Leeds West. response in rheumatoid arthritis patients with initial incomplete B cell depletion: a Provenance and peer review Not commissioned; externally peer reviewed. randomised controlled trial. Ann Rheum Dis 2015;74:1195–201. 12 Østergaard M, Peterfy C, Conaghan P, et al. OMERACT rheumatoid arthritis Data sharing statement No additional unpublished data are currently available. magnetic resonance imaging studies. Core set of MRI acquisitions, joint pathology definitions, and the OMERACT RA-MRI scoring system. J Rheumatol References 2003;30:1385–6. 1 stack RJ, van Tuyl LHD, Sloots M, et al. Symptom complexes in patients with 13 Mangnus L, van Steenbergen HW, Reijnierse M, et al. Magnetic resonance Imaging- seropositive arthralgia and in patients newly diagnosed with rheumatoid arthritis: a Detected features of inflammation and erosions in symptom-free persons from the qualitative exploration of symptom development. Rheumatology 2014;53:1646–53. general population. Arthritis Rheumatol 2016;68:2593–602. 2 Gerlag DM, Raza K, van Baarsen LGM, et al. EULAR recommendations for terminology 14 van Steenbergen HW, van Nies JAB, Huizinga TWJ, et al. Characterising arthralgia and research in individuals at risk of rheumatoid arthritis: report from the study Group in the preclinical phase of rheumatoid arthritis using MRI. Ann Rheum Dis for risk factors for rheumatoid arthritis. Ann Rheum Dis 2012;71:638–41. 2015;74:1225–32. 3 nam JL, Hensor EMA, Hunt L, et al. Ultrasound findings predict progression to 15 nieuwenhuis WP, Krabben A, Stomp W, et al. Evaluation of magnetic resonance inflammatory arthritis in anti-CCP antibody-positive patients without clinical synovitis. imaging-detected tenosynovitis in the hand and wrist in early arthritis. Arthritis Ann Rheum Dis 2016;75:2060–7. Rheumatol 2015;67:869–76.

786 Mankia K, et al. Ann Rheum Dis 2019;78:781–786. doi:10.1136/annrheumdis-2018-214331 Early arthritis

Translational science Running promotes chronicity of arthritis by local modulation of complement activators and impairing T regulatory feedback loops Isabelle Cambré,‍ ‍ 1,2 Djoere Gaublomme,1,2 Nadia Schryvers,1,2 Stijn Lambrecht,1 Rik Lories,3,4 Koen Venken,1,2 Dirk Elewaut2

Handling editor Josef S Abstract Key messages Smolen Objectives The mechanisms driving onset of joint inflammation in arthritides such as rheumatoid arthritis ►► Additional material is ► Voluntary running promotes disease chronicity and spondyloarthritis and the conversion to disease ► published online only. To view in experimental arthritis. please visit the journal online chronicity are poorly understood. We hypothesised ► Mechanostrain drives complement activation (http://dx.​ ​doi.org/​ ​10.​1136/​ mechanostrain could play an instrumental role herein ► thereby promoting disease chronicity. annrheumdis-2018-​ ​214627). by engaging local and/or systemic pathways, thereby ►► Arthritis resolution is regulated by T regulatory 1 attenuating disease course and outcome. Faculty of Medicine and cells but over-ruled by voluntary running Health Sciences, Department of Methods The development of collagen antibody- Internal Medicine and Pediatrics induced arthritis (CAIA) in C57BL/6 mice was evaluated (Rheumatology Unit), Ghent both clinically and histologically under different loading University, Ghent, Belgium interesting hallmark of these structures is that they 2 regimens: control, voluntary running or hindpaw unloading. Molecular Immunology and have to withstand high exposure to biomechan- Bone surface porosity was quantified by high-resolution Inflammation Unit,V IB Center of ical stresses. This suggests that biomechanical stress InflammatoryR esearch, Ghent, µ-CT. Gene expression analyses were conducted by 4 could impact the clinical features of these diseases. Belgium microarrays and qPCR on microdissected entheses, murine 3 In support thereof, recent clinical data suggest that Division of Rheumatology, and human synovial tissues (both normal and inflamed). Katholieke Universiteit Leuven highly demanding physical labour is linked to more Serum cytokines and chemokines were measured by ELISA. Hospital, Leuven, Belgium severe structural progression in ankylosing spondy- 4 The influence of complement activation and T regulatory 5 5–9 Laboratory of Tissue litis (AS) and PsA. Similar physical workload was Homeostasis and Disease, (Treg) cell function on the induction and resolution phase Skeletal Biology and associated with the risk of anti-citrullinated protein of disease was studied by respectively pharmacological 10 Engineering, Department of antibody (ACPA)-positive and ACPA-negative RA. modulation and conditional Treg depletion. Development and Regeneration, In addition, in certain patient populations, physical Results Voluntary running strongly impacts the 11 Katholieke Universiteit Leuven, exercise may lead to an increase of disease flares. Leuven, Belgium course of arthritis by impairing the resolution phase Thus, biomechanical stress may trigger the onset or of CAIA, leading to more persistent inflammation and influence the course of joint inflammation. Correspondence to bone surface porosity. Mechanical strain induced local Dr Dirk Elewaut, Faculty of In line with this, we previously showed in a complement activation, increased danger-associated ΔARE Medicine and Health Sciences, TNF-driven model of SpA (TNF ) that hindlimb molecular pattern expression, activating Fc receptors Department of Internal Medicine γ unloading prevented onset of SpA-like disease and as well as changes in fibroblast phenotype.I nterestingly, and Pediatrics (Rheumatology that enhanced loading induced by voluntary running Unit), Ghent University, complement C5a receptor blockade inhibited the impacts the progression of this model as well as the Molecular Immunology and enhanced joint pathology caused by voluntary running. Inflammation Unit,V IB Center onset of actively induced arthritis (collagen induced Moreover, Treg depletion led to a loss of disease 12 for Inflammation Research, arthritis (CIA)). The underlying mechanisms driving resolution in CAIA mice, which was not observed under Corneel Heymanslaan 10, Gent the impact of mechanical strain on joint inflammation voluntary running conditions. 9000, Belgium; remained poorly defined. We therefore investigated dirk.​ ​elewaut@ugent.​ ​be Conclusions running promotes onset and chronicity of the impact of different loading regimens (hindlimb arthritis by local upregulation of complement activators Received 21 October 2018 unloading vs voluntary running vs control conditions) and hampering regulatory T cell feedback loops. Revised 25 February 2019 in a passive arthritis model. We unravelled a crucial Accepted 28 February 2019 role for the complement pathway in the exacerbation Published Online First 30 March 2019 in onset of arthritis induced by voluntary running. Introduction Complement activators were locally induced within Rheumatic diseases such as rheumatoid arthritis (RA) joints and inhibition thereof markedly attenuated and spondyloarthritis (SpA) affect the musculoskel- disease. In addition, voluntary running also impacts etal system and are prototypic examples of arthritic naturally T regulatory (Treg) mediated feedback loops diseases characterised by chronic joint inflammation. that serve to dampen the increase in inflammation © Author(s) (or their induced by voluntary running in arthritic disease. employer(s)) 2019. No Disease chronicity and a high rate of severe disease commercial re-use. See rights flares are correlated with severe bone destruction and and permissions. Published deformation leading to important loss of function.1–3 Materials and methods by BMJ. A unifying feature of RA and SpA (including Mice arthritis experiments and biomechanical To cite: Cambré I, psoriatic arthritis [PsA]) is their impact on synovial conditions Gaublomme D, Schryvers N, joints (particularly those of the most mobile joints C57BL/6 mice (male, 10 weeks) (Envigo) were et al. Ann Rheum Dis of the body, including knee and ankle), and tendons, subjected to collagen antibody-induced arthritis 2019;78:787–795. which transmit forces from muscle to bone. An (CAIA) (MD Biosciences, ArthritoMap Antibody

Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 787 Early arthritis

Cocktail, 4 mg/20 g, intraperitoneal) with injection of 100 µg pores in the bone surface. The difference between the filled lipopolysaccharide (LPS) at day 3. Mice were kept, starting the surface and the normal non-filled bone surface was calculated, day of CAIA injection, in different biomechanical loading condi- giving an absolute erosion value. Dividing this value by the total tions for 4 weeks: tail suspended to prevent hindlimb loading bone value gives the percent erosion per bone. Surface porosity (CAIA unloaded, condition started 4 days before CAIA, reload quantification was done of the calcanei of CAIA C57BL/6 mice, 24 hours after LPS injection), mice with free access to a running 28 days after disease induction. Additionally, the trabeculae wheel (CAIA voluntary running, registration of running distance/ within the calcaneus were automatically classified using an algo- day), mice kept in normal cages without a running wheel (CAIA rithm similar to that of Buie et al.18 The average trabecular thick- control) (5–10 mice per group, two independent experiments). ness was measured using the Thickness plugin from BoneJ.19 RAG2-/- mice (on a C57BL/6 background; 9–10 weeks) were subjected to CAIA (Chondrex, Androgen-CIA 5-Clone Cocktail Gene expression in microdissected Achilles’ tendons and kit, 4 mg/20 g, intraperitoneal) in the absence of LPS (6–10 mice human synovial tissue samples per group, two independent experiments). RAG mice were treated Achilles’ tendons were dissected from C57BL/6 (Envigo) mice with a cocktail of five anti-collagen antibodies without the use of subjected to two different loading conditions (voluntary running LPS, because RAG2-/- mice are more susceptible for toxic shock [n=6], control [n=3]) for 14 days. Isolated tendons were stored due to their inability to clear the LPS from the bloodstream. This in RNAlater (Qiagen) before RNA isolation with microRNeasy has been linked to the lack of IgMs in RAG2-/- mice.13 Mice were Kit (Qiagen). Microarray assays (including quality controls) were kept either in voluntary running or control conditions during the run at the VIB Nucleomics Core (www.​ ​nucleomics.​be) with Chip experiment (11/14 days), starting 1 day before CAIA induction Mouse 430_2.0. (day −1) or at day 5 (once arthritis development has started). Achilles’ tendon samples isolated from TNF∆ARE mice (volun- Mice were treated with vehicle or PMX205, a complement C5a tary running and control) were processed similarly except that antagonist (R&D systems, 100 µg daily by oral gavage, starting preamplification was used to increase the cDNA yield (Quan- 1 day before CAIA induction or starting from day 5 by changing tiTect Whole Transcriptome kit, Qiagen). Next, real-time PCR regimen from vehicle treated to PMX205).14 Male DEpletion of was performed on a LightCycler 480 (Roche Molecular Diagnos- REGulatory T cell mice (DEREG; 11–12 weeks) were obtained tics) using SYBR-green Mastermix (SensiMix SYBR No Rox-kit, from Dr Sparwasser and bred in our facility. They were subjected Bioline). Expression values were normalised to TBP, B2M, to CAIA (MD Biosciences, 4 mg/20 g mice) with 100 µg LPS GCC1, GADPH and HPRT expression levels (housekeeping injection at day 3 (five to eight mice per group, two indepen- genes). Primers were obtained from BIORAD (PrimePCR assays), dent experiments). In these mice, Treg cells selectively express an Invitrogen or Qiagen. enhanced green fluorescent protein and diphtheria toxin (DT) Human RA and SpA synovial tissue samples were collected receptor gene containing bacterial artificial chromosome (BAC) by arthroscopy. This included samples from both patients with construct under an FOXP3 promoter, which permits to selec- RA (n=4, mean age 55.25±14.5 years; female/male count 3/1, tively deplete Treg cells by DT injection.15 Mice were kept in RF+/RF− 1/3, ACPA+/ACPA− 2/2) and patients with SpA voluntary running or control conditions for 4 weeks and were (n=4, mean age 36.5±12.3 years, female/male count 3/1). The treated with vehicle (phosphate buffered saline) or DT (Merck, latter group of patients were all HLAB27+ and were diagnosed 15 µg/kg) three times a week (intraperitoneal) starting from day as PsA (2), undifferentiated SpA (1) and AS (1). Diagnosis and 8. Clinical and histopathology scoring was done as mentioned classification of patients was made by staff rheumatologists in the online supplementary materials and methods. For some following the European League Against Rheumatism/American experiments inflamed synovial tissues were isolated from College of Rheumatology 2010 or Assessment in Ankylosing TNFΔARE/+ mice, a TNF-driven SpA-like disease model.16 Spondylitis 2010 criteria. None of the patients were treated with All mice were kept and bred in specific pathogen-free condi- disease-modifying antirheumatic drugs or biological agents at tions in accordance with the general recommendations for the moment of the synovial biopsy. Healthy synovial tissue was animal breeding and housing. Experiments were approved obtained in the context of diagnostic arthroscopic procedures by and conducted according to the guidelines of the Ethics following traumatic injuries or in patients with meniscus tears Committee of Laboratory Animals Welfare of Ghent University. (with no signs of synovitis). Used reference genes TBP, GCC1 and PIGO. Micro-CT acquisitions Healthy synovial fibroblasts originated from synovial tissue Micro-CT scans of sample hindpaws (n=40) were performed samples from individuals who underwent a diagnostic arthros- at the Centre for X-ray Tomography of the Ghent University copy following traumatic injuries or in patients with meniscus (UGCT, www.​ugct.​ugent.​be). The required voxel size defined the tears with no signs of synovitis (see online supplementary mate- dimensions of the scanning region, which included full hindfoot rials and methods). Isolation of human synovial fibroblasts was and mid-foot and part of forefoot. The resulting scan images conducted as previously described.12 have a voxel size of 4 µm. The obtained projection images of all scans were reconstructed using Octopus Reconstruction,17 a Statistical analysis commercial software licensed by XRE and originally developed Statistical analyses were performed using SPSS, GenStat by UGCT, based on the standard FDK algorithm for reconstruc- and GraphPad Prism. Experimental values are presented tion of CT data. 3D visualisations were made using the commer- as mean±SEM. Statistical differences were evaluated using cial rendering software VGStudio MAX (Volume Graphics). two-tailed unpaired non-parametric Mann-Whitney U tests if not mentioned otherwise. Gene expression data were checked Bone surface porosity quantification for normality distribution with Kolmogorov-Smirnov and Bone surface porosity was calculated on reconstructed micro-CT Shapiro-Wilk tests (>0.05) and subsequently analysed with a data with an in-house written Fiji-based script.12 The program parametric paired t-test (one tailed). Clinical scores were anal- determines the bone surface and bone inner space and fills the ysed as repeated measurements using the residual maximum

788 Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 Early arthritis likelihood approach as implemented in GenStat V.19. In analogy of disease (days 14–28), causing a significantly different disease with survival analysis, incidence data were defined as the time course in voluntary running mice compared with the control mice it takes for the arthritis to develop (score ≥1 for at least 2 (figure 1A). Furthermore, incidence of arthritis was comparable in consecutive days) in each mouse. Significant differences between both groups at the peak of inflammation (days 10–12), but in the experimental groups were represented as *p<0.05, **p<0.01, recovery phase (days 14–28) arthritis persisted in the voluntary ***p<0.001. running group but not in the control group (see online supplemen- tary figure S1). The data were corroborated by histological assess- Results ment of ankles (figure 1C,D, online supplementary figure S4,S5) Mechanical strain promotes chronicity by diminishing and knees (data not shown) with scoring of inflammation, carti- resolution of joint inflammation lage and bone erosions. By contrast, when mice were unloaded, a To examine the potential effect of biomechanical factors on the clear reduction in arthritis induction at the hindpaws was observed chronicity of arthritis we used a passive arthritis model, CAIA, in (online supplementary figure S2) which resolved in a similar way as the presence or absence of voluntary running conditions during in the control condition (figure 1B). For comparison, we observed 28 days. In this model, inflammation occurs usually 1 or 2 days no significant differences between the clinical scorings of the front after the initial inflammatory trigger (LPS, 3 days after anti-collagen paws of both groups in these experiments (online supplementary Ab Cocktail injection) and typically peaks by day 10 after which figure S3). Of note, the running data revealed that mice undergoing the inflammation resolves (by day 14). To determine the impact CAIA run comparable distances as wild-type mice in the first 3 days, of voluntary running on onset and peak of arthritis as well as on followed by a more limited motility pattern for the next 2–3 days the resolution of inflammation, disease features were monitored (probably because of the physical discomfort caused by the LPS chal- up until day 28 after disease induction. Curiously, the voluntary lenge). In the end, the diseased mice show an increase in running running group displayed a remarkable attenuated recovery phase distances (by day 14 reaching the levels measured at the start of

Figure 1 Voluntary running promotes chronicity in CAIA. (A) Clinical scores of C57BL/6 CAIA mice under voluntary running or control conditions show a significant different course of disease; repeated measures mixed models time*condition, n=15. (B) Hindpaw clinical scores of C57BL/6 CAIA mice in unloading versus control condition showing a later disease onset in unloaded mice with similar resolving capacities; repeated measures mixed models time*condition, n=10–15. (C) H&E histological data showing increased inflammation in ankles from C57BL/6 voluntary running mice compared with control and unloaded conditions; Mann-Whitney U test (MWU), n=5, results from one of two similar experiments are shown. (D) Micro-CT scans of CAIA in C57BL/6 mice under voluntary running (left), control (middle) and unloaded (right) conditions (showing a maximal clinical score of 2–2.5 at peak of disease). (E) Bone surface porosity quantification of the calcaneus shows significant increase of porosity by voluntary running compared with control and unloaded conditions for paws with a score >1.5, MWU, with n=3–8. (F) Quantification of the trabecular thickness of calcaneal bone shows significant decrease by voluntary running compared with control and unloaded mice (score >1.5), n same as for (E) unloaded mice with minimal or no scores (<1.5) already show a decrease in trabecular thickness compared with the control condition. (G) A significant lower level of MMP3 in serum of CAIA unloaded C57BL/6 mice versus voluntary running or control regimens, MWU, n=10–15. Mean values±SEM are shown, if not mentioned, otherwise, data are representative for two independent experiments.*p<0.05; **p<0.01; ***p<0.001. CAIA, collagen antibody-induced arthritis.

Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 789 Early arthritis the experiment), eventually comparable to those of wild-type mice in inflamed synovial and tendon samples (figure 2B). Moreover, by the time the inflammation has faded out (online supplementary previously we showed that voluntary running induced a more than figure S6). twofold increase in expression of CFP in the Achilles’ tendons of Quantitative analyses of µ-CT scans of calcaneus bones from arthritic mice, suggesting a link between the expression of this these CAIA mice (day 28) demonstrated significantly aggravated gene and mechanical strain in healthy but also under inflamma- porosity of the calcaneus cortex in the voluntary running versus tory conditions. Here we show that in inflamed synovium samples control and unloaded conditions (figure 1E,), particularly in from patients with RA, but curiously not patients with SpA, we mice with higher clinical scores (>1.5) which is not unexpected also observed increased levels of CFP expression (compared with as bone surface porosity is secondary to inflammation during healthy synovium levels), highlighting the importance of this arthritis. In addition, trabecular thickness of the calcaneus was finding in human disease (figure 2C). significantly diminished in voluntary running mice compared with CAIA control mice (clinical score >1.5) (figure 1F). The data also showed that in unloaded mice with a moderate disease Inflammation induced by voluntary running is controlled by state (>1.5) but also with a low to absent disease state (<1.5), complement activation the trabecular thickness is clearly diminished compared with We recently demonstrated that voluntary running is sufficient to wild-type mice, with a lower to similar decrease as seen in induce arthritis without the need of exogenous TLR4 stimulation 12 CAIA control mice, this is expected because unloading causes a in T and B cell-deficient mice (RAG2-/- mice). In this model, decrease in trabecular bone as well as increase in surface porosity mice under voluntary running conditions developed arthritis by which is reflected in a similar bone surface porosity as CAIA passive transfer of anticollagen II antibodies alone—so in the control mice (figure 1E) . Data from the histological assessment absence of LPS-mediated activation of immune (mainly myeloid) of the ankle were consistent with the µ-CT data (online supple- cells—whereas the control condition displayed only mild signs of mentary figure S4). arthritis in a minority of animals. This points to a crucial role for mechanotransduction in mediating this form of ‘sterile inflamma- tion’. Because the gene expression analyses (previous section) hinted Transition to disease chronicity caused by voluntary running towards changes in the complement activation pathways with strong is not linked to systemic responses induction of positive regulator CFP and initiator C1q, we evaluated Because the impact of biomechanical loading was comparable the impact of complement inhibition on mechanical strain-induced between different joints (ankle and knee) we assessed whether clinical arthritis symptoms. We reasoned that voluntary running voluntary running promotes the systemic secretion of proin- may induce complement activation thereby reducing the threshold flammatory cytokines and chemokines. Therefore, serum for arthritis development. To investigate this, we tested an inhibitor levels of IL6, CCL2, CXCL1, MMP3, IL17, TNF and IL1 β of the receptor for C5a, both in a preventive setting (prior to induc- were measured by ELISA in CAIA mice in control, voluntary tion of arthritis) as well as in a therapeutic setting (once initial signs running or unloading conditions, 14 days after disease induc- of arthritis developed). tion. No differences were found in levels of CCL2, CXCL1, IL6, So first, mice were treated daily with a C5aR inhibitor TNF and IL1 between the three groups (data not shown). In β (PMX205) or vehicle by oral gavage starting 1 day prior to contrast, MMP3 levels were found to be lower under unloaded CAIA induction. As expected, a clear induction of arthritis was conditions compared with voluntary running and control condi- apparent in the voluntary running versus control mice under tions (figure 1G). Furthermore, our data show that most of vehicle conditions (figure 3A). This was completely absent in the passively transferred antibodies were cleared at day 14 in mice treated with the C5aR inhibitor, observed both in terms of all conditions, and no differences in glycosylation patterns of incidence (figure 3A) and clinical arthritis severity (figure 3B). anticollagen antibodies, which are implicated in pathogenicity The findings were also confirmed by histopathology (figure 3C). of autoantibodies, could be documented12. The data suggest that Thus, prophylactic blocking a final element in the complement the impaired resolution cannot be directly linked to increased pathway, leads to a dramatically reduced induction of mechan- systemic levels of proinflammatory cytokines or an altered clear- ical loading dependent form of inflammation. ance of passively transferred autoantibodies. Second, we evaluated the role of the complement pathway on this mechanosensitive form of sterile inflammation once arthritis Positive regulator of complement activation (complement was present, thereby bypassing the initial induction phase. Here, factor properdin) locally induced by mechanical strain mice were treated with C5aR inhibitor or vehicle control either In our previously performed microarray on wild-type-mice-de- from the day before administration of anticollagen type II anti- rived Achilles’ tendon tissues,12 we discovered a nearly twofold body cocktail or from day 5 onwards. Incidence of arthritis increased expression of complement factor properdin (CFP), the under voluntary running conditions reached up until 70% of only known positive regulator of complement activation, induced the collagen Ab-injected animals in the vehicle group and this by voluntary running (as compared with control conditions). was prevented by C5aR inhibition (online supplemental figure Additionally, also other members of the complement pathway S6; comparable to our initial observations with the prophylactic were increased by voluntary running like the complement recep- regimen). The CAIA mice that initially received vehicle for 5 tors C5aR1 and C3aR1, the classical pathway initiator molecule days under control housing conditions (the arthritis incidence C1q which binds immune complexes and its C1 complex partner for this group was 40%; online supplemental figure S7) were subcomponent C1s (figure 2A). In contrast, the complement factor randomised at day 5 into four different groups: a control or H (CFH), which is known to be a negative complement regu- voluntary running group of mice in the presence (PMX205) or lator, was significantly downregulated. This complement activa- absence (vehicle) of C5Ra blockade. Incidence in the vehicle tion could in theory be responsible for an early disease onset but group under voluntary running conditions increased up to 80% also for impaired resolution and transition to chronicity caused by which was not the case in the presence of complement inhibition running. Interestingly, we observed in a TNF-driven arthritis mouse (50%) (figure 3D). In parallel, longitudinal scoring from start of models (TNF∆ARE mouse20) that CFP levels were also increased randomisation (day 5) until day 14 shows a significant difference

790 Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 Early arthritis

Figure 2 Local changes in tissue resident cells by mechanical strain. (A) Microarray data of C57BL/6 Achilles’ tendons from mice under voluntary running versus control conditions show significant increased fold changes for C5aR1, C1qa, C1qb, CFH and CFP (almost significant for C3aR1 and C1s), Mann-Whitney U test (MWU). (B) qPCR data showing that CFP gene expression levels are increased in both the Achilles’ tendon and synovial tissue of TNF∆ARE compared with C57BL/6, MWU, n=5. (C) Voluntary running induced an additional increase of CFP expression in Achilles’ tendons of arthritic mice (TNF∆ARE). MWU, n=5, 6. (D) Healthy (n=4) versus RA (n=4) and SpA (n=4) patient-derived synovial tissue shows a clear upregulation of CFP in RA patient samples. Mean values±SEM are shown. If not mentioned, otherwise, data are representative for two independent experiments. *p<0.05; **p<0.01. CFP, complement factor properdin; RA, rheumatoid arthritis; SpA, spondyloarthritis.

between control and voluntary running conditions, in vehicle to be upregulated under voluntary running conditions. They but not in C5aR antagonist-treated animals (figure 3E). Collec- include proteoglycans biglycan, the proteolytic enzyme of the tively, the data indicate that the effect of voluntary running extracellular matrix BMP1, the proinflammatory ECM glyco- on arthritis induction and aggravation is at least partly depen- protein TNC and the heat shock-related DAMP HSPH1. By dent on complement activation. These clinical data were also contrast, most of the autophagy and apoptosis regulatory confirmed by histopathology which shows comparable outcomes DAMPs showed no differences, except BCL2 which was found of vehicle-treated voluntary running mice on day −1 and day 5 to be reduced. Diseased (TNFΔARE) tendon samples showed the and diminished inflammatory responses when complement was most pronounced deregulation in the ECM and heat shock-re- inhibited (figure 3F,G,H). Because both experiments (figure 3C lated DAMPs. Likewise, most autophagy and apoptotic markers and F,G) were short term (14 days) there was overall limited were not altered, except BCL2 which was upregulated. In sum, cartilage damage and bone erosion present. Overall we can several DAMPs, especially ECM and heat shock-related ones, conclude that C5aR inhibition blocked the effect of voluntary are induced by voluntary running both in C57Bl/6 and TNF∆ARE running in disease progression (figure 3F,G,H). mice. These DAMPs can activate, directly or in interaction with pathogen-associated molecular patterns (PAMPs), the innate Lack of arthritis resolution is associated with differential immune system by binding to pattern recognition receptors like gene expression of several danger-associated molecular toll-like receptors and CFP. These data suggest that the increased patterns, FcγR and fibroblast markers presence of DAMPs as observed under voluntary running condi- In addition to the changes in complement factors, we observed tions might add to arthritis perpetuation by triggering of local that voluntary running also induces a diverse set of local innate immune responses. In addition to complement and danger-associated molecular pattern (DAMP) molecules in DAMPs we also observed an increase in expression of activating Achilles’ tendons from both healthy (C57BL/6) and diseased Fc gamma receptors (FCγRIII, FCγRIV, FCγRI) and patho- (TNFΔARE) mice (online supplemental figure S8). In healthy genic fibroblast markers FAP and THY1 which all can influ- tendon samples several of the extracellular matrix protein-re- ence the resolving phase of disease (online supplemental figure lated DAMPs, but also certain heat shock proteins were found S8-10). Altogether, these data indicate that many—potentially

Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 791 Early arthritis

Figure 3 Sterile inflammation caused by voluntary running is partly driven by complement activation. (A) Disease incidence of CAIA RAG2-/- mice under control and voluntary running conditions. Mice were treated with vehicle (H20) or PMX205 (C5aR antagonist) starting 1 day before CAIA induction. Data show a clear increased disease onset in voluntary running vehicle versus control conditions and PMX205-treated animals; survival analysis, n=8–10. (B) Clinical scoring of groups of CAIA RAG2-/- mice with vehicle or PMX205. Earlier onset and more severe arthritis in vehicle- treated voluntary running mice (n=6, 6) compared with control mice (n=5, 5). Reduced arthritis severity by PMX205 treatment; repeated measures mixed models condition. (C) H&E histological scoring of the ankle shows significantly increased inflammation in voluntary running vehicle compared with PMX205-treated mice. (D) Disease incidence after randomisation into the different conditions at day 5. Exaggerated onset of disease in voluntary running versus vehicle treatment or PMX205-treated mice, n=5–6. (E) Calculation of the disease progression of randomised group 3 mice between day 5 and day 14. Data show significant increase of CAIA RAG2-/- clinical scores by voluntary running versus controls. This is caused by the increase in the vehicle-treated group. GenStat variant analysis with delta values, n=5–6. (F) H&E histological scoring of ankle slides shows significantly more inflammation in voluntary running mice treated with vehicle versus PMX205, both in the groups treated from day −1 as well as from day 5 onwards. Control mice treated with vehicle or PMX205 did not show differences in inflammation. Exclusion of mice with score ≥2 before randomisation (=1 mouse in every group), Mann-Whitney U test (MWU), n=4–5. (G) H&E histological scoring of the toes is consistent with the ankle data. (H) Safranin-O staining of ankle slides of mice treated either with vehicle or PMX205 in voluntary running conditions from day −1 or day 5 shows a comparable therapeutic effect of complement inhibition. Mean values±SEM are shown.*P<0.05. CAIA, collagen antibody-induced arthritis. synergistically working—mechanisms underlie the changes were performed in DEREG mice, which allowed selective and observed in Achilles’ tendon tissues under mechanical strain, efficient depletion of Foxp3+ Treg cells by DT injection.15 We which could influence the course of arthritis. subjected these animals to control or voluntary running condi- tions for 28 days. Depletion of Tregs was achieved by admin- Resolution of arthritis is regulated by Treg cells but this effect istration of DT versus vehicle control from day 8, to avoid is over-ruled by voluntary running interference with the disease initiation phase. Treg depletion While induction of inflammation within joints appears to be at (verified by flow cytometry in blood and several organs including least partially dependent on tissue resident responses, resolution spleen or lymph nodes, data not shown) led to a delayed resolu- of inflammation may also depend on specific immunosuppres- tion phase in CAIA-induced animals. Under running conditions sive pathways. In this context, we speculated that Tregs could the vehicle group exerted an impaired resolution of inflamma- play a key role given their well-known capacity to suppress tion compared with control housing conditions. However, in effector immune cells. To investigate this, CAIA experiments contrast to the non-running situation, no additional effect by

792 Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 Early arthritis

Figure 4 T regulatory cells drive the resolution phase in CAIA, voluntary running can over-rule this effect. (A) Clinical scoring of CAIA-induced DEREG mice that were in control or voluntary running conditions. At day 3 lipopolysaccharide (LPS) was injected and at day 8 the treatment with DT or vehicle was started. In control conditions, DT induced aggravation of arthritis by more sustained inflammation. In the voluntary running group treated with vehicle, arthritis scores were increased compared with control conditions and achieved similar scores as observed after Treg depletion in the control group. Curiously, no additional effect of Treg depletion was observed in the voluntary running group; repeated measures mixed models with DEREG control versus DEREG voluntary running, n=5–6, similar results in two independent experiments, *p<0.05. (B) H&E histological data of the ankle show a significant increase in inflammation by control DT treatment versus control vehicle treatment. No difference was observed by Treg depletion in the voluntary running group, Mann-Whitney U test (MWU), n=4–6. Mean values±SEM are shown. If not mentioned, otherwise, data are representative for two independent experiments. *P<0.05. CAIA, collagen antibody-induced arthritis; DEREG, DEpletion of REGulatory T cell mice; DT, diphtheria toxin.

Treg depletion was noted (figure 4A). These observations suggest components both in healthy and TNF∆ARE mice compared with that voluntary running over-ruled the ability of Tregs to control control conditions. The data suggest that stimulation of comple- resolution of inflammation. In line with this, histology of the ment pathways takes place under running conditions, potentially ankle shows a significant increase in inflammation by Treg deple- by direct activation of both the classical pathway, by increase of tion in control DEREG mice but not under running conditions its initiating molecule C1q which binds Ab complexes, and the (figure 4B). alternative pathway. This was evidenced by an increase of the positive regulator CFP and decrease of the negative regulator Discussion CFH, but also by increase of the complement receptors (C5aR1 The impact of physical activity on rheumatic diseases such as RA and C3aR1) that enable to carry out the different effects of the and SpA has been a topic of several studies but its exact role in complement system on inflammation. To address whether this disease induction or persistence and bone remodelling remains increase in expression of complement factors had a functional unclear. In SpA there is evidence to underscore the importance role we blocked the complement receptor C5aR in a preventive- of a biomechanical component in onset of disease,11 21–23 specifi- and therapeutic setting in a modified—that is, sterile inflam- cally given the major impact of mechanical strain on induction of mation—form of the CAIA model12 under different loading enthesitis, a hallmark of SpA. In addition, very recent data from regimens. We observed a striking impact on arthritis onset by our group showed biomechanical loading to also be a key factor inhibition of the complement system, showing that the effect of in the induction of RA-like disease in experimental arthritis voluntary running on arthritis induction and aggravation is at models, particularly collagen-induced arthritis.12 The data from least partially dependent on complement activation. the current report provide substantially new additional insights Alternate factors that can contribute to the lack of resolu- into the underlying mechanisms by showing a profound effect of tion in voluntary running mice include the increased expression mechanical strain on the progression and chronicity of arthritis of activating FcγRs. The upregulation of FcγR expression as a and structural damage through an impaired resolution. This facil- consequence of biomechanical strain might result in antibodies itates chronicity, which leads to a longer period of severe inflam- (or Ab complexes) binding more efficiently to these receptors, mation. Moreover, this has a major impact on bone homeostasis causing a potential increase and persistent activation of innate as we observed more severe bone surface porosity and loss of and adaptive immune pathways. Furthermore, FcγRs are also trabecular bone mass in these animals. Interestingly, these results involved in osteoclastogenesis which might underlie the observed were not linked to systemic inflammatory responses like IL6, IL1, accelerated bone erosion and bone deformation. In addition, we CXCL1 and CCL2 as measured in the serum of CAIA animals. also measured increased expression levels of DAMP molecules. Furthermore, we also did not observe any changes in clear- DAMPs induce activation of innate immunity by binding PRR ance or altered glycosylation patterns of (passively transferred) like TLR, CFP, NOD2. A constant expression of these DAMPs anticollagen antibodies,12 suggesting that local tissue resident may favour chronicity in CAIA, and by this prolongs the initial responses on voluntary running are responsible for this effect. effect caused by LPS, a well-described PAMP, needed as an initial Indeed, the gene expression analyses of entheseal and synovial triggering event in this model. Moreover, direct and indirect tissue show a clear increase of several interesting target genes, interaction between PAMPs and DAMPs could modify activation including a number of complement regulating genes. The func- thresholds for these signalling molecules influencing the course tional significance of the induction of complement regulating of inflammatory responses. Recent data have pointed to addi- factors was further assessed in vivo. Voluntary running induced tional crucial roles for tissue resident and mesenchymal stromal a strong and significant upregulation of several complement cells in disease chronicity.24–27 In this regard, we observed a

Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 793 Early arthritis strong upregulation of FAP and THY1 expression in Achilles’ and an FWO Excellence of Science (EOS) grant. KV is supported by FWO-Flanders tendons under voluntary running conditions, both in healthy as and VLAIO. DG is supported by FWO-Flanders. well as diseased mice (TNF∆ARE). A similar increase could also Competing interests None declared. be observed in inflamed synovial tissue samples derived from Patient consent for publication Obtained. patients with SpA, and to a lesser extent also in RA samples Ethics approval The ethical committee of the University Hospital of Leuven (online supplemental figure S9), in line with previous obser- granted permission to perform this study with human samples. Animal experiments vations.25 28 The FAP and THY1 gene induction could also be were approved by and conducted according to the guidelines of the Ethics directly linked to mechanical strain (online supplemental figure Committee of Laboratory Animals Welfare of Ghent University S11). Provenance and peer review Not commissioned; externally peer reviewed. An interesting observation is that despite the ubiquitous role Data sharing statement The data that support the findings in this study are of mechanical strain across different arthritis models,12 the gene available from the corresponding author upon request. induction pattern of mechanosensitive genes in RA versus SpA seems to be somehow distinct. For example, the induction of References CFP in RA synovial tissue was clearly more abundant whereas the 1 Markusse IM, Dirven L, Gerards AH, et al. Disease flares in rheumatoid arthritis are induction of markers associated with a more pathogenic fibro- associated with joint damage progression and disability: 10-year results from the best study. Arthritis Res Ther 2015;17. blast phenotype was higher in SpA. This mirrors the well-de- 2 raheel S, Matteson EL, Crowson CS, et al. Improved flare and remission pattern in scribed role of autoantibodies in the pathogenesis of RA versus rheumatoid arthritis over recent decades: a population-based study. Rheumatology SpA,29 whereas stromal cells have been linked even more to SpA 2017;56:2154–61. pathogenesis.16 21 30 3 Portier A, Gossec L, Tubach F, et al. Patient-perceived flares in rheumatoid arthritis: a sub-analysis of the STRASS treatment tapering strategy trial. Joint Bone Spine Finally, we investigated whether naturally occurring immu- 2017;84:577–81. nomodulatory pathways, important in controlling and damp- 4 Debusschere K, Cambré I, Gracey E, et al. Born to run: the paradox of biomechanical ening of inflammatory responses, were affected under voluntary force in spondyloarthritis from an evolutionary perspective. Best Pract Res Clin running conditions. For this we focused on Tregs, best known Rheumatol 2017;31:887–94. for their control of effector immune cell function (including 5 Ward MM, Reveille JD, Learch TJ, et al. Occupational physical activities and long-term functional and radiographic outcomes in patients with ankylosing spondylitis. Arthritis antibody production) and suppression of local inflammation as Rheum 2008;59:822–32. they show the capacity to actively migrate towards local inflam- 6 Zhou W, Chandran V, Cook R, et al. The association between occupational-related matory sites.31 Our data in the CAIA model were clearly in line mechanical stress and radiographic damage in psoriatic arthritis. Semin Arthritis with this active role in disease suppression as on selective deple- Rheum 2019;48:638–43. 7 de Winter J, de Hooge M, van de Sande M, et al. Magnetic resonance imaging of the tion of Tregs (using DEREG mice) we observed an impaired Sacroiliac joints indicating sacroiliitis according to the assessment of spondyloarthritis resolution phase in diseased mice. Remarkably, this effect was International Society definition in healthy individuals, runners, and women with not seen under voluntary running conditions, suggesting that postpartum back pain. Arthritis Rheumatol 2018;70:1042–8. physical loading impacts naturally occurring immunosuppres- 8 ramiro S, Landewé R, van Tubergen A, et al. Lifestyle factors may modify the effect of sive mechanisms. Although the precise details of the underlying disease activity on radiographic progression in patients with ankylosing spondylitis: a longitudinal analysis. RMD Open 2015;1:e000153. mechanisms remain to be determined, it is reasonable to assume 9 Megna M, Gisonni P, Napolitano M, et al. The effect of smartphone addiction on hand that this is an indirect consequence of biomechanical loading. In joints in psoriatic patients: an ultrasound-based study. J Eur Acad Dermatol Venereol this regard, it is interesting to note that deficiency or blocking 2018;32:73–8. of C5aR signalling has been linked to de novo differentiation 10 Zeng P, Klareskog L, Alfredsson L, et al. Physical workload is associated with increased risk of rheumatoid arthritis: results from a Swedish population-based case-control and induction of Foxp3(+) Treg and C5aR signalling in Tregs 32–34 study. RMD Open 2017;3:e000324. impairs their function. This suggests that the impact of 11 ansell RC, Shuto T, Busquets-Perez N, et al. The role of biomechanical factors in mechanical strain on the complement pathway and Treg function ankylosing spondylitis: the patient’s perspective. Reumatismo 2015;67:91–6. may be intertwined. Thus, the increased complement activation 12 cambré I, Gaublomme D, Burssens A, et al. Mechanical strain determines the site- induced by voluntary running might play a role in switching off specific localization of inflammation and tissue damage in arthritis. Nat Commun 2018;9. the Treg-mediated resolution of arthritis. However, additional 13 reid RR, Prodeus AP, Khan W, et al. Endotoxin shock in antibody-deficient mice: mechanisms may also be in place as C5aR blockade clearly has unraveling the role of natural antibody and complement in the clearance of a protective effect on voluntary running-induced arthritis in the lipopolysaccharide. J Immunol 1997;159:970–5. absence of adaptive immunity. 14 Jain U, Woodruff TM, Stadnyk AW. The C5a receptor antagonist PMX205 ameliorates experimentally induced colitis associated with increased IL-4 and IL-10. Br J Collectively our data indicate that voluntary running has Pharmacol 2013;168:488–501. tremendous effects on the initiation and the resolution phase of 15 Mayer CT, Lahl K, Milanez-Almeida P, et al. Advantages of Foxp3(+) regulatory T cell disease, which has implications on physical activity recommen- depletion using DEREG mice. Immun Inflamm Dis 2014;2:162–5. dations to both patients with early and later stage RA and SpA. 16 armaka M, Apostolaki M, Jacques P, et al. Mesenchymal cell targeting by TNF as a common pathogenic principle in chronic inflammatory joint and intestinal diseases. J Correction notice This article has been corrected since it published Online First. Exp Med 2008;205:331–7. The affiliation details have been corrected. 17 Vlassenbroeck J, Dierick M, Masschaele B, et al. Software tools for quantification of Acknowledgements µ-CT scan protocols, devices and assistance were provided x-ray microtomography at the UGCT. Nuclear Instruments and Methods in Physics by the Center for X-ray Tomography of the Ghent University (UGCT, www.​ugct.​ugent.​ Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment be) for which we especially thank Amélie De Muynck and Luc Van Hoorebeke, and 2007;580:442–5. Tine Decruy for technical assistance. 18 Buie HR, Campbell GM, Klinck RJ, et al. Automatic segmentation of cortical and trabecular compartments based on a dual threshold technique for in vivo micro-CT Contributors The project was designed, conceived and planned by IC, SL, RL, DE bone analysis. Bone 2007;41:505–15. and KV and the experiments were performed by IC, NS, KV and DG. The data were 19 Doube M, Kłosowski MM, Arganda-Carreras I, et al. BoneJ: free and extensible bone analysed by IC, DG, KV and DE and the paper was written and edited by IC, DE and image analysis in ImageJ. Bone 2010;47:1076–9. KV. 20 Kontoyiannis D, Pasparakis M, Pizarro TT, et al. Impaired on/off regulation of TNF Funding DE is supported by FWO-Flanders, Fonds voor Wetenschappelijk biosynthesis in mice lacking TNF AU-rich elements: implications for joint and gut- Reumaonderzoek (FWRO), Council of Ghent University and interuniversity Attraction associated immunopathologies. Immunity 1999;10:387–98. Pole grant Devrepair from Belspo Agency (project P7/07) and an FWO Excellence 21 Jacques P, Lambrecht S, Verheugen E, et al. Proof of concept: enthesitis and new bone of Science (EOS) grant. RJUL is supported by FWO-Flanders, KU Leuven and formation in spondyloarthritis are driven by mechanical strain and stromal cells. Ann interuniversity Attraction Pole grant Devrepair from Belspo Agency (project P7/07) Rheum Dis 2014;73:437–45.

794 Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 Early arthritis

22 Benjamin M, McGonagle D. The enthesis organ concept and its relevance to the 29 Pfeifle R, Rothe T, Ipseiz N, et al. Regulation of autoantibody activity by the spondyloarthropathies. Adv Exp Med Biol 2009;649:57–70. IL-23-TH17 axis determines the onset of autoimmune disease. Nat Immunol 23 McGonagle D, Lories RJU, Tan AL, et al. The concept of a "synovio-entheseal complex" 2017;18:104–13. and its implications for understanding joint inflammation and damage in psoriatic 30 schett G, Lories RJ, D’Agostino M-A, et al. Enthesitis: From pathophysiology to arthritis and beyond. Arthritis Rheum 2007;56:2482–91. treatment. Nat Rev Rheumatol 2017;13:731–41. 24 Dakin SG, Coles M, Sherlock JP, et al. Pathogenic stromal cells as therapeutic targets 31 Ding Y, Xu J, Bromberg JS. Regulatory T cell migration during an immune response. in joint inflammation. Nat Rev Rheumatol 2018;14:714–26. Trends Immunol 2012;33:174–80. 25 Mizoguchi F, Slowikowski K, Wei K, et al. Functionally distinct disease-associated 32 van der Touw W, Cravedi P, Kwan W-hong, et al. Cutting edge: receptors for C3a and fibroblast subsets in rheumatoid arthritis. Nat Commun 2018;9. C5a modulate stability of alloantigen-reactive induced regulatory T cells. J Immunol 26 chen L, Qiu X, Wang X, et al. FAP positive fibroblasts induce immune checkpoint blockade resistance in colorectal cancer via promoting immunosuppression. Biochem 2013;190:5921–5. Biophys Res Commun 2017;487:8–14. 33 chen X-H, Ruan C-C, Ge Q, et al. Deficiency of complement C3a and C5a receptors 27 Jacob M, Chang L, Puré E. Fibroblast activation protein in remodeling tissues. Curr prevents angiotensin II-induced hypertension via regulatory T cells. Circ Res Mol Med 2012;12:1220–43. 2018;122:970–83. 28 choi IY, Karpus ON, Turner JD, et al. Stromal cell markers are differentially expressed in 34 Weaver DJ, Reis ES, Pandey MK, et al. C5a receptor-deficient dendritic cells promote the synovial tissue of patients with early arthritis. PLoS One 2017;12:e0182751. induction of Treg and Th17 cells. Eur J Immunol 2010;40:710–21.

Cambré I, et al. Ann Rheum Dis 2019;78:787–795. doi:10.1136/annrheumdis-2018-214627 795 Psoriatic arthritis

Translational science Serum-based soluble markers differentiate psoriatic arthritis from osteoarthritis Vinod Chandran,‍ ‍ 1 Fatima Abji,2 Anthony V Perruccio,3,4 Rajiv Gandhi,5 Suzanne Li,6 Richard J Cook,7 Dafna D Gladman‍ ‍ 1

Handling editor Josef S Abstract Key messages Smolen Objectives We aimed to identify soluble biomarkers that differentiate psoriatic arthritis (PsA) from ►► Additional material is What is already known about this subject? published online only. To view osteoarthritis (OA). ►► Patients with osteoarthritis (OA) may present please visit the journal online Methods Markers of cartilage metabolism (cartilage similar to patients with psoriatic arthritis (PsA) (http://dx.​ ​doi.org/​ ​10.​1136/​ oligomeric matrix protein [COMP], hyaluronan), with distal interphalangeal joint involvement annrheumdis-2018-​ ​214737). metabolic syndrome (adiponectin, adipsin, resistin, and back pain. For numbered affiliations see hepatocyte growth factor [HGF], insulin, leptin) and inflammation (C-reactive protein [CRP], interleukin-1β end of article. What does this study add? [IL-1β], IL-6, IL-8, tumour necrosis factor alpha [TNFα], ►► The study found that markers of cartilage Correspondence to monocyte chemoattractant protein-1 [MCP-1], nerve metabolism, metabolic syndrome and Professor Dafna D Gladman, growth factor [NGF]) were compared in serum samples inflammation are differentially expressed in Toronto Western Hospital, from 201 patients with OA, 77 patients with PsA and Toronto, ON M5T 2S8, Canada; patients with PsA, OA, and healthy controls. 76 controls. Levels across the groups were compared dafna.​ ​gladman@utoronto.​ ​ca A panel of four biomarkers, cartilage using the Kruskal-Wallis test. Pairwise comparisons were orligometric matrix protein, resistin, monocyte The abstract for this manuscript made with Wilcoxon rank-sum test. Multivariate logistic chemoartractant protein-1 and nerve growth (Abstract 1657) has been regression analyses were performed to identify markers presented at the American factors provdie addotopma; discriminatory that differentiate PsA from OA. Receiver operating College of Rheumatology (ACR)/ value in distinguishing psA from OA form that characteristic (ROC) curves were constructed based AHRP Annual Meeting Poster provided by age sex. Session B held on 14 November on multivariate models. The final model was further 2016 (https://​acrabstracts.org/​ ​ validated in an independent set of 73 PsA and 75 OA How might this impact on clinical practice or abstract/soluble-​ ​biomarkers-​ samples using predicted probabilities calculated with may-differentiate-​ ​psoriatic-​ future developments? coefficients of age, sex and biomarkers. arthritis-from-​ ​osteoarthritis/?​ ► These biomarkers taken together with other Results levels of the following markers were ► msg=fail&​ ​shared=email).​ features may provde a valuable tool for significantly different across the three groups diagnosis and management of patients with OA Received 9 November 2018 (p<0.001)—COMP, hyaluronan, resistin, HGF, insulin, Revised 24 February 2019 and PsA. Accepted 26 February 2019 leptin, CRP, IL-6, IL-8, TNFα, MCP-1, NGF. In multivariate Published Online First analysis, COMP (OR 1.24, 95% CI 1.06 to 1.46), resistin 25 March 2019 (OR 1.26, 95% CI 1.07 to 1.48), MCP-1 (OR 1.10, 95% CI 0.07 to 1.48) and NGF (OR<0.001, 95% CI (OA) due to the involvement of the large joints of <0.001 to 0.25) were found to be independently the lower extremities and proximal (PIP) and distal associated with PsA versus OA. The area under the ROC interphalangeal (DIP) joints of the hands as well as curve (AUROC) for this model was 0.99 compared with often normal blood levels of acute phase reactants model with only age and sex (AUROC 0.87, p<0.001). in both conditions.3 The identification of soluble Similar results were obtained using the validation sample. biomarkers that distinguish PsA from OA can aid Conclusion a panel of four biomarkers may distinguish in the development of tools for physicians to better PsA from OA. These results need further validation in identify patients with PsA. prospective studies. We hypothesised that soluble markers of carti- lage metabolism, metabolic syndrome and inflam- mation may differentiate patients with PsA from those with OA. We aimed to detect biomarkers Introduction associated with PsA by measuring serum levels Psoriatic arthritis (PsA) is an immune-mediated of 15 markers of cartilage metabolism (cartilage inflammatory musculoskeletal disease associated oligomeric matrix protein [COMP], hyaluronan), with joint, ligament and tendon pain, stiffness and © Author(s) (or their metabolic syndrome (adiponectin, adipsin, resistin, swelling that leads to damage to the peripheral, employer(s)) 2019. No hepatocyte growth factor [HGF], insulin, leptin) commercial re-use. See rights axial and entheseal structures, resulting in reduced and permissions. Published quality of life and life expectancy.1 2 The point of and inflammation/immune response (C-reactive by BMJ. entry into the healthcare system for individuals protein [CRP], interleukin-1β [IL-1β], IL-6, IL-8, tumour necrosis factor alpha [TNF ], monocyte To cite: Chandran V, affected by PsA is often through family physicians α Abji F, Perruccio AV, or dermatologists. Physicians with limited experi- chemoattractant protein [MCP-1], nerve growth et al. Ann Rheum Dis ence in this area may find it difficult to diagnose factor [NGF]) in patients with PsA, OA and healthy 2019;78:796–801. PsA, which can be confused with osteoarthritis controls (HC).

796 Chandran V, et al. Ann Rheum Dis 2019;78:796–801. doi:10.1136/annrheumdis-2018-214737 Psoriatic arthritis

Methods p≤0.1 were entered into a multivariate model in the second stage Study subjects and backward elimination was carried out to obtain a reduced Serum samples were obtained from the biobank of the University model. Further, internal cross-validation was performed using of Toronto PsA Program (PsA) and University Health Network the PsA and OA samples. The samples were divided into three Arthritis Program (OA). All patients with PsA had psoriasis subsamples within each set (PsA: 25, 26, 26; OA: 67, 67, 67) and confirmed by a dermatologist and they satisfied the Classifica- each pair of subsamples was pooled and used as the training set tion for Psoriatic Arthritis criteria.4 Patients with OA did not with the remaining subsamples used as the testing set. The same have inflammatory arthritis and HCs were healthy volunteers. procedure of variable selection was performed on each training All HCs completed screening questionnaires to rule out the pres- set: a logistic regression with age and sex was fitted for each ence of autoimmune conditions and family history of psoriasis biomarker; those with p≤0.1 were further considered in a multi- and PsA. The discovery set included 77 PsA, 201 OA and 76 HC. variate model adjusting for age and sex; the resulting multivariate An independent set of 73 PsA and 75 OA patient samples was model was then simplified using stepwise backward elimination. included for validation set. No patient with PsA was receiving In the validation set, predicted probabilities were estimated using treatment with a biological agent at the time of serum collec- the linear predictors from the model developed in the discovery tion, but may have received these agents after the sample was set. Logistic regression analyses were also conducted to generate drawn. Samples were obtained at the time of knee or hip joint models to compare the interaction between age and biomarkers replacement surgery (OA) or at the time of clinical assessment as well as determine the effects of age and sex alone in both the (PsA, HC), and stored at −80°C until laboratory assays were discovery and validation sets. Discriminative ability was assessed conducted. by way of receiver operating characteristic (ROC) curves based on the multivariate models. We compared the area under the Serum ELISA and multiplex assays curve (AUC) of the model based simply on age and sex with that Markers of cartilage metabolism (COMP and hyaluronan) were based on age, sex and marker levels using the DeLong’s test for measured using Quantikine immunoassays according to the two correlated ROC curves. manufacturer’s instructions (R&D Systems, Minneapolis, MN). Markers of metabolic syndrome (adiponectin, adipsin, resistin, Results HGF, insulin and leptin) and inflammation (CRP, IL-1β, IL-6, Patient characteristics IL-8, TNFα, MCP-1 and NGF) were measured using magnetic A summary of the demographics and disease characteristics of bead Luminex panels (EMD Millipore, Billerica, MA) according the study subjects can be found in table 1. The discovery set to the manufacturer’s instructions. Data were acquired using the included 77 patients with PsA, 201 OA and 76 HC. A signif- Luminex 200 system (Luminex, Austin, TX) and analysed with icant difference in age between the three groups was found the Bio-Plex Manager software (Bio-Rad Laboratories, Hercules, (p<0.0001). The validation set consisted of 73 patients with PsA CA). All samples were measured in duplicate and marker levels and 75 patients with OA, where patients with OA were signifi- were quantified in relation to a fivefold serially diluted standard cantly older than patients with PsA (p<0.0001). All patients with provided with the kits, using a five-parameter logistic regression PsA had active PsA and a total of nine patients (two in discovery curve. and seven in validation set) also had concomitant OA.

Statistical analysis Comparison of serum marker levels between PsA, OA and HC Marker levels in serum were compared across the three groups When we compared the levels of the 15 markers between the in the discovery set using the Kruskal-Wallis test. Additional pair- three groups (PsA, OA and HC), 12 markers were differen- wise comparisons were made with Wilcoxon rank-sum test to tially expressed. The marker levels and results of the compar- identify the source of differences. In order to identify markers ison across and between the subject groups are summarised in that could differentiate patients with PsA from patients with OA, table 2. Nine markers (CRP, resistin, HGF, IL-6, IL-8, insulin, we performed multivariate logistic regression analysis adjusting MCP-1, TNFα and COMP) were elevated in patients with PsA for age and sex. At the first stage, the regression models included compared with patients with OA. CRP, HGF, IL-6, leptin, NGF, age, sex and 1 of the 15 markers as covariates; markers having TNFα and COMP were also increased in PsA compared with

Table 1 Demographics and disease characteristics in study subjects Discovery set Validation set P Characteristics PsA (n=77) OA (n=201) HC (n=76) P value* PsA (n=73) OA (n=75) value* Sex (females), % 45.5 57.2 50 0.1767 41.1 49.3 0.3271 Age†, years 43.0 (34.0–54.5) 66.0 (58.0–72.0) 35.5 (27.0–44.0) <0.0001 49.0 (42.5–62.0) 67.0 (61.0–72.0) <0.0001 PsA duration‡, years 6.0 (6.8) – – 12.9 (11.0) – Swollen joint count‡ 3.3 (5.0) – – 0.3 (0.9) – Tender joint count‡ 6.9 (9.7) – – 3.4 (6.0) – Clinically damaged joint count‡ 2.9 (8.0) – – 6.4 (11.0) – PASI score‡ 5.1 (7.1) – – 3.8 (7.2) – Patients on systemic agents, % 57.1 – – 60.3 – *Mean (SD). †Median (IQR). ‡Kruskal-Wallis test (age) or Pearson’s χ2 test (sex). HC, healthy control; OA, osteoarthritis; PASI, Psoriasis Area and Severity Index; PsA, psoriatic arthritis.

Chandran V, et al. Ann Rheum Dis 2019;78:796–801. doi:10.1136/annrheumdis-2018-214737 797 Psoriatic arthritis

Table 2 Comparison of biomarker levels between PsA, OA and HC in the discovery set Pairwise comparison P value OA Overall PsA versus PsA versus versus Biomarkers PsA (n=77)* OA (n=201)* HC (n=76)* P value OA HC HC CRP, µg/mL 29.7 (9.8–115.2) 15.0 (4.8–43.8) 16.5 (6.5–36.9) 0.0005 0.0002 0.0023 0.706 Adiponectin, µg/mL 17.6 (9.9–24.5) 15.9 (9.8–25.2) 17.4 (10.6–24.2) 0.8899 0.9277 0.8281 0.6045 Adipsin, µg/mL 3.9 (3.1–5.2) 4.2 (3.4–5.0) 3.9 (3.1–5.2) 0.4641 0.3179 0.9143 0.3283 Resistin, ng/mL 37.5 (28.3–51.3) 26.9 (19.8–34.5) 40.3 (30.2–55.1) <0.0001 <0.0001 0.4446 <0.0001 HGF, ng/mL 0.8 (0.5–1.1) 0.3 (0.2–0.4) 0.5 (0.3–1.1) <0.0001 <0.0001 0.0191 <0.0001 IL-1β, pg/mL 1.3 (0.4–1.3) 1.3 (0.7–1.3) 1.3 (0.4–1.3) 0.1394 0.5825 0.2877 0.0424 IL-6, pg/mL 7.5 (3.3–13.8) 3.9 (1.9–7.6) 1.0 (0.7–1.9) <0.0001 0.0003 <0.0001 <0.0001 IL-8, pg/mL 8.6 (5.5–15.7) 4.2 (3.0–6.1) 6.7 (4.5–18.5) <0.0001 <0.0001 0.258 <0.0001 Insulin, pg/mL 384.8 (262.5–860.4) 276.7 (181.2–426.6) 430.7 (233.3–1065.3 <0.0001 <0.0001 0.9782 <0.0001 Leptin, ng/mL 14.7 (6.7–25.8) 17.8 (7.1–30.4) 6.3 (2.7–16.3) <0.0001 0.1737 0.0002 <0.0001 MCP-1, pg/mL 301.7 (233.0–413.7) 157.8 (118.3–199.2) 279.5 (305.2–372.8) <0.0001 <0.0001 0.2306 <0.0001 NGF, pg/mL 6.1 (3.8–8.1) 8.3 (5.2–15.0) 2.2 (1.6–3.7) <0.0001 <0.0001 <0.0001 <0.0001 TNFα, pg/mL 5.4 (3.7–7.7) 3.6 (2.5–5.5) 4.5 (3.4–5.6) <0.0001 <0.0001 0.0083 0.0723 COMP, ng/mL 193.8 (168.6–252.8) 148.2 (103.5–197.3) 160.7 (125.0–186.0) <0.0001 <0.0001 <0.0001 0.1725 Hyaluronan, ng/mL 33.2 (20.4–54.5) 67.8 (36.9–125.0) 23.3 (14.9–54.4) <0.0001 <0.0001 0.0709 <0.0001 *Median (IQR). COMP, cartilage oligomeric matrix protein; CRP, C-reactive protein; HC, healthy control; HGF, hepatocyte growth factor; IL, interleukin; MCP-1, monocyte chemoattractant protein-1; NGF, nerve growth factor; OA, osteoarthritis; PsA, psoriatic arthritis; TNFα, tumour necrosis factor alpha.

HC. Four markers (IL-6, leptin, NGF and hyaluronan) were MCP-1 and NGF. The resulting ROC curve had an AUC of present in higher levels in patients with OA as compared with 0.9896 (figure 1B). The model containing age and sex alone had HC and two of these markers (NGF and hyaluronan) were also an AUC of 0.8296. There was a significant difference in the AUC high in patients with OA compared with patients with PsA. In of the two ROC curves (p<0.001). Thus, we confirmed that the contrast, resistin, HGF, IL-8, insulin and MCP-1 were reduced biomarkers provide additional discriminatory value for discrim- in OA serum compared with HC. inating PsA from OA over demographic features (age and sex).

Four markers differentiate patients with PsA from patients with OA The results of multivariate logistic regression analysis comparing Table 3 Results of logistic regression analyses comparing serum PsA and OA serum levels in the discovery set are found in table 3. levels of PsA to OA in the discovery set COMP, resistin, MCP-1 and NGF were identified as markers that Univariate† Multivariate† are significantly different between patients with PsA and OA. P The ROC curve constructed using the model with age, sex and Biomarkers* OR† (95% CI) P value OR† (95% CI) value the four biomarkers (COMP, resistin, MCP-1, NGF) had an AUC COMP‡ 1.10 (1.07 to 1.14) <0.0001 1.24 (1.06 to 1.46) 0.0062 of 0.9984 (figure 1A). The model containing age and sex alone Hyaluronan 0.79 (0.25 to 2.51) 0.6896 had an AUC of 0.8727. There was a significant difference in Adiponectin 2.49 (0.85 to 7.30) 0.0974 the AUC of the two ROC curves (p<0.001). MCP-1 was consis- Adipsin 17.87 (1.42 to 225.3) 0.0258 tently identified as a PsA marker during cross-validation. The Resistin‡ 1.07 (1.04 to 1.10) <0.0001 1.26 (1.07 to 1.48) 0.0056 first fold identified HGF, MCP-1 and COMP and the AUC was HGF 139.37 (29.95 to <0.0001 0.9896, the second identified resistin, MCP-1, NGF and COMP 648.61) (AUC 1.000) and the third internal validation identified MCP-1 Insulin 12.74 (3.85 to 42.16) <0.0001 and NGF (AUC 0.9719), respectively. Leptin 1.88 (0.83 to 4.25) 0.1325 CRP 1.25 (0.35 to 4.45) 0.7293 Biomarker validation in an independent set IL-1β 0.93 (0.49 to 1.77) 0.8317 To further validate these findings, the four biomarkers iden- IL-6 6.41 (2.68 to 15.33) <0.0001 tified in the multivariate model of the discovery set (COMP, resistin, MCP-1 and NGF) were measured in an independent IL-8‡ 1.05 (1.04 to 1.07) <0.0001 set of patients with OA (n=75) and PsA (n=73) (validation set, TNFα 11.72 (2.65 to 56.75) 0.0012 table 4). HGF was also measured because it satisfied the criteria MCP-1‡ 1.09 (1.06 to 1.12) <0.0001 1.28 (1.10 to 1.48) 0.0014 for backward elimination in part of the cross-validation of the NGF 0.12 (0.03 to 0.39) 0.0005 <0.001 (<0.001 to 0.0016 discovery set. No differences in COMP were observed. Resistin, 0.25) HGF and MCP-1 were elevated in patients with PsA compared *With log10 transformation. with patients with OA while NGF was reduced, similar to the †Adjusted for age and sex. ‡Rescaled by multiplying by 100. discovery set. The performance of the multivariate model devel- COMP, cartilage oligomeric matrix protein; CRP, C-reactive protein; HGF, hepatocyte oped in the discovery set was assessed by calculating predicted growth factor; IL, interleukin; MCP-1, monocyte chemoattractant protein-1; NGF, probabilities with coefficients of age, sex, COMP, resistin, nerve growth factor; OA, osteoarthritis;

798 Chandran V, et al. Ann Rheum Dis 2019;78:796–801. doi:10.1136/annrheumdis-2018-214737 Psoriatic arthritis

Figure 1 ROC curves for the discovery (A) and validation (B) sets. ROC curves for the models including age and sex alone (dashed lines) and with the inclusion of COMP, resistin, MCP-1 and NGF (solid lines) are shown. AUC, area under the curve; COMP, cartilage oligomeric matrix protein; MCP-1, monocyte chemoattractant protein-1; NGF, nerve growth factor; ROC, receiver operating characteristic.

Because of the significant difference in age between the PsA and OA also share common risk factors and pathological trig- patients with PsA and OA in this study, we computed the logistic gers. Trauma and injury are well known to trigger OA and the rela- regression models to compare the interaction between each of tionship between trauma and PsA, known as the Koebner’s effect, is the four biomarkers (COMP, resistin, MCP-1 and NGF) and age well documented.7 Obesity is a primary risk factor for disease onset to predict disease status. There were no significant interactions in OA and is associated with increased risk and severity of arthritis in between age and the four biomarkers observed in the discovery PsA.8 9 OA and PsA typically commence around age 50, particularly and validation sets (online supplementary table 1). for women affected by OA coinciding with the drop in oestrogen levels.10 11 Imaging studies have revealed an age-related thickening Discussion of the normal enthesis and ligaments after age 40.3 12 The enthesis OA is the most commonly occurring form of arthritis. Due to the is one of the earliest distinguishable sites of hand OA and enthesitis similarities in the distribution of joints involved as PsA and the has been hypothesised to be the initiation site of pathogenesis in lack of biomarkers, it may be difficult to distinguish between the PsA.13 14 These overlapping aetiological processes in PsA and OA two arthritides leading to delayed diagnosis and inappropriate may contribute to a difficulty in distinguishing these disorders in treatment. Both OA and PsA can affect similar joint regions the clinic such as the DIP and PIP joints in the hands, cervical and lumbar Biomarkers can serve as objective measurements that are indi- spine and large joints of the lower limbs.3 Inflammation of the cators of normal biological or pathogenic processes. In this study, small joints of the hands and feet was in fact difficult to discrim- the aim was to identify soluble biomarkers that would aid in the inate between PsA and OA using high-resolution MRI.5 New distinction between PsA and OA. A panel of four markers (COMP, bone formation as osteophytes in OA and syndesmophytes in resistin, MCP-1 and NGF) was found to discriminate patients with PsA and joint erosions are typically found in both conditions PsA from patients with OA. Previous studies have also supported and are usually distinguished by imaging. However, a variant of these findings. Ross and colleagues have observed high expression OA called diffuse idiopathic skeletal hyperostosis involves the of MCP-1 in the synovium of patients with PsA supported by a formation of osteophytes that are similar to syndesmophytes in positive correlation between T cell numbers in synovial fluid (SF) some patients with PsA.6 and MCP-1 levels.15 MCP-1 was also elevated in PsA compared with OA serum,16 which may reflect the chronic inflammatory burden in PsA. NGF has been found in the SF of patients with Table 4 Summary of serum levels of COMP, resistin, MCP-1, HGF OA, spondyloarthritis and rheumatoid arthritis.17 Inhibition of and NGF in validation cohort of patients with PsA (n=73) and OA NGF has been shown to provide extensive pain relief in OA and (n=75) two anti-NGF antibodies, tanezumab and fasinumab, have shown P tremendous promise in clinical trials for the treatment of pain in Biomarkers PsA (n=73)* OA (n=75)* value† OA.18 19 NGF plays a crucial role in the chronic pain of OA20–22 COMP, ng/mL 217.3 (182.9–278.3) 210.3 (166.6–264.7) 0.3435 and this was reflected in the consistently high levels of NGF in OA Resistin, ng/mL 36.0 (28.3–51.9) 24.6 (19.8–34.7) <0.0001 compared with PsA serum in this study. COMP has previously been HGF, ng/mL 0.6 (0.5–1.0) 0.2 (0.2–0.3) <0.0001 found to be elevated in PsA compared with control serum and may 23–25 MCP-1, pg/mL 370.0 (269.8–458.4) 127.6 (106.3–162.7) <0.0001 be a prognostic marker in the preclinical diagnosis of PsA. In OA, previous studies have pointed to the use of serum COMP as NGF, pg/mL 3.0 (2.0–4.3) 4.5 (3.3–6.8) <0.0001 a marker for early cartilage lesions in the knee as it is negatively *Median (IQR). 26 27 †Wilcoxon rank-sum test. correlated with OA disease duration. A recent meta-analysis COMP, cartilage oligomeric matrix protein; HGF, hepatocyte growth factor; MCP-1, determined that high levels of COMP increase the probability 28 monocyte chemoattractant protein-1; NGF, nerve growth factor; OA, osteoarthritis. of developing knee OA. Resistin was previously shown to be

Chandran V, et al. Ann Rheum Dis 2019;78:796–801. doi:10.1136/annrheumdis-2018-214737 799 Psoriatic arthritis elevated in PsA compared with control serum and released from References OA cartilage, where it correlated with SF expression of IL-6, 1. ritchlin CT, Colbert RA, Gladman DD. Psoriatic arthritis. N Engl J Med matrix metalloproteinase-1 (MMP-1) and MMP-3.29 30 Obesity is 2017;376:957–70. 2. Taylor WJ. Impact of psoriatic arthritis on the patient: through the lens of the WHO a common predisposing factor for PsA and OA, although the role International Classification of Functioning, health, and disability. Curr Rheumatol Rep of resistin is not well understood. We observed consistently high 2012;14:369–74. levels of resistin in patients with PsA compared with patients with 3. Tan AL, Grainger AJ, Tanner SF, et al. A high-resolution magnetic resonance OA but not controls, consistent with previous reports.31–33 imaging study of distal interphalangeal joint arthropathy in psoriatic arthritis and The large sets of patients with PsA and OA available at the same osteoarthritis: are they the same? Arthritis Rheum 2006;54:1328–33. 4. Taylor W, Gladman D, Helliwell P, et al. Classification criteria for psoriatic arthritis: institution provided a unique opportunity for this study which development of new criteria from a large international study. Arthritis Rheum cannot be easily obtained. The validation of the results in two inde- 2006;54:2665–73. pendent groups of patients and consistently high sensitivity and 5. Braum LS, McGonagle D, Bruns A, et al. Characterisation of hand small joints specificity of the biomarker panel is promising. However, this study arthropathy using high-resolution MRI--limited discrimination between osteoarthritis is not without limitations. The patients with OA included in this and psoriatic arthritis. Eur Radiol 2013;23:1686–93. 6. Mader R, Sarzi-Puttini P, Atzeni F, et al. Extraspinal manifestations of diffuse idiopathic study were those undergoing joint replacement surgeries and may skeletal hyperostosis. Rheumatology 2009;48:1478–81. not be representative of the general population of patients with OA. 7. olivieri I, Padula A, D’Angelo S, et al. Role of trauma in psoriatic arthritis. J Rheumatol Thus, it is necessary to replicate these results using patients with OA 2008;35:2085–7. seen in family practice or rheumatology clinics. Also, the effect of 8. Blagojevic M, Jinks C, Jeffery A, et al. Risk factors for onset of osteoarthritis of the knee in older adults: a systematic review and meta-analysis. Osteoarthritis Cartilage other variables such as measurements of cardiovascular health and 2010;18:24–33. the use of medications was not accounted for in these findings. 9. eder L, Abji F, Rosen CF, et al. The association between obesity and clinical features of With the development of biological therapies, the inclusion of psoriatic arthritis: a case-control study. J Rheumatol 2017;44:437–43. soluble biomarkers in the diagnosis and treatment of PsA and OA 10. oliveria SA, Felson DT, Reed JI, et al. Incidence of symptomatic hand, hip, and knee would facilitate a personalised medicine approach to patient care. osteoarthritis among patients in a health maintenance organization. Arthritis Rheum 1995;38:1134–41. Anti-TNF treatments have been highly efficacious in PsA but do not 11. Gladman DD, Antoni C, Mease P, et al. Psoriatic arthritis: epidemiology, clinical 34–36 relieve symptoms of OA. The failure of these anti-TNF agents features, course, and outcome. Ann Rheum Dis 2005;64(Suppl 2):ii14–17. in some patients with PsA may in part be due to a misdiagnosis of 12. Yu T-Y, Tsai W-C, Cheng J-W, et al. The effects of aging on quantitative sonographic OA. Eventually, the availability of a biomarker panel could facilitate features of rotator cuff tendons. J Clin Ultrasound 2012;40:471–8. more accurate diagnosis to target appropriate therapeutic strategies. 13. McGonagle D, Khan MA, Marzo-Ortega H, et al. Enthesitis in spondyloarthropathy. Curr Opin Rheumatol 1999;11:244–50. However, the clinical feasibility of this biomarker panel needs to be 14. Tan AL, Grainger AJ, Tanner SF, et al. High-resolution magnetic resonance imaging for determined. Future studies are needed to evaluate this panel in a the assessment of hand osteoarthritis. Arthritis Rheum 2005;52:2355–65. prospective cohort of patients. 15. ross EL, D’Cruz D, Morrow WJ. Localized monocyte chemotactic protein-1 production In conclusion, PsA and OA sometimes affect the same anatom- correlates with T cell infiltration of synovium in patients with psoriatic arthritis. J ical regions which make them difficult to distinguish in a subgroup Rheumatol 2000;27:2432–43. 16. scanzello CR. Chemokines and inflammation in osteoarthritis: insights from patients of patients. We identified a panel of four biomarkers (COMP, and animal models. J Orthop Res 2017;35:735–9. resistin, MCP-1 and NGF) that when combined provide additional 17. Xia L, Shen H, Xiao W, et al. Increased serum TWEAK levels in psoriatic arthritis: discriminatory value to that provided by age and sex alone. These relationship with disease activity and matrix metalloproteinase-3 serum levels. markers were consistently validated internally within the discovery Cytokine 2011;53:289–91. 18. aloe L, Tuveri MA, Carcassi U, et al. Nerve growth factor in the synovial fluid of set as well through an independent set of patients. After further patients with chronic arthritis. Arthritis Rheum 1992;35:351–5. verification and validation, these markers could provide a valuable 19. schnitzer TJ, Marks JA. A systematic review of the efficacy and general safety of tool for the improved diagnosis and management of both OA and antibodies to NGF in the treatment of oa of the hip or knee. Osteoarthritis Cartilage PsA thus resulting in improved patient outcomes. 2015;23(Suppl 1):S8–S17. 20. Miller RE, Malfait A-M, Block JA. Current status of nerve growth factor antibodies for the treatment of osteoarthritis pain. Clin Exp Rheumatol 2017;35(Suppl 107):85–7. Author affiliations 21. Blaney Davidson EN, van Caam APM, Vitters EL, et al. TGF-β is a potent inducer of 1Division of Rheumatology, Department of Medicine, University of Toronto, Toronto nerve growth factor in articular cartilage via the ALK5-Smad2/3 pathway. Potential Western Hospital, Toronto, Ontario, Canada role in oa related pain? Osteoarthritis Cartilage 2015;23:478–86. 2Rheumatology Research, University of Toronto, University Health Network, Toronto, 22. raychaudhuri SP, Jiang W-Y, Raychaudhuri SK. Revisiting the Koebner phenomenon: Ontario, Canada role of NGF and its receptor system in the pathogenesis of psoriasis. Am J Pathol 3Arthritis Program, University Health Network, Toronto, Ontario, Canada 2008;172:961–71. 4Surgery, University of Toronto, Toronto, Ontario, Canada 23. Bartosińska J, Michalak-Stoma A, Juszkiewicz-Borowiec M, et al. The assessment of 5Division of Orthopaedic Surgery, Department of Surgery, Toronto Western Hospital, selected bone and cartilage biomarkers in psoriatic patients from Poland. Mediators Toronto, Ontario, Canada Inflamm 2015;2015:1–8. 6Centre for Prognosis Studies in the Rheumatic Diseases, Toronto Western Hospital, University Health Network, Toronto, Ontario, Canada 24. Farouk HM, Mostafa AAA, Youssef SS, et al. Value of entheseal ultrasonography 7Statistics and Actuarial Science, University of Waterloo, Waterloo, Ontario, Canada and serum cartilage oligomeric matrix protein in the preclinical diagnosis of psoriatic arthritis. Clin Med Insights Arthritis Musculoskelet Disord 2010;3:CMAMD. Contributors All authors were involved in drafting the article or revising it critically S4461–14. for important intellectual content, and all authors approved the final version to be 25. Verma P, Dalal K. Serum cartilage oligomeric matrix protein (COMP) in knee published. All authors were likewise involved in the study conception and design, osteoarthritis: a novel diagnostic and prognostic biomarker. J Orthop Res acquisition of data as well as analysis and interpretation of data. DDG had full 2013;31:999–1006. access to all of the data in the study and takes responsibility for the integrity of the 26. Jiao Q, Wei L, Chen C, et al. Cartilage oligomeric matrix protein and hyaluronic acid data and the accuracy of the data analysis. are sensitive serum biomarkers for early cartilage lesions in the knee joint. Biomarkers Funding The University of Toronto Psoriatic Arthritis Program is supported by a 2016;21:146–51. grant from the Krembil Foundation. 27. Zhang J. Meta-analysis of serum C-reactive protein and cartilage oligomeric matrix protein levels as biomarkers for clinical knee osteoarthritis. BMC Musculoskelet Disord Competing interests None declared. 2018;19. Patient consent for publication Obtained. 28. Park HK, Ahima RS. Resistin in rodents and humans. Diabetes Metab J 2013;37:404–14. Ethics approval This study was approved by the University Health Network 29. Dikbas O, Tosun M, Bes C, et al. Serum levels of visfatin, resistin and adiponectin in Research Ethics Board according to the principles of the Declaration of Helsinki. patients with psoriatic arthritis and associations with disease severity. Int J Rheum Dis Provenance and peer review Not commissioned; externally peer reviewed. 2016;19:672–7.

800 Chandran V, et al. Ann Rheum Dis 2019;78:796–801. doi:10.1136/annrheumdis-2018-214737 Psoriatic arthritis

30. Koskinen A, Vuolteenaho K, Moilanen T, et al. Resistin as a factor in osteoarthritis: 34. salvarani C, Cantini F, Olivieri I, et al. Efficacy of infliximab in resistant psoriatic synovial fluid resistin concentrations correlate positively with interleukin 6 and matrix arthritis. Arthritis Rheum 2003;49:541–5. metalloproteinases MMP-1 and MMP-3. Scand J Rheumatol 2014;43:249–53. 35. Verbruggen G, Wittoek R, Vander Cruyssen B, et al. Tumour necrosis factor 31. senolt L, Housa D, Vernerová Z, et al. Resistin in rheumatoid arthritis synovial tissue, blockade for the treatment of erosive osteoarthritis of the interphalangeal finger synovial fluid and serum. Ann Rheum Dis 2007;66:458–63. joints: a double blind, randomised trial on structure modification. Ann Rheum Dis 32. Xue Y, Jiang L, Cheng Q, et al. Adipokines in psoriatic arthritis patients: the 2012;71:891–8. correlations with osteoclast precursors and bone erosions. PLoS One 2012;7:e46740. 36. Glintborg B, Østergaard M, Dreyer L, et al. Treatment response, drug survival, and 33. Mease PJ, Gladman DD, Ritchlin CT, et al. Adalimumab for the treatment of patients predictors thereof in 764 patients with psoriatic arthritis treated with anti-tumor with moderately to severely active psoriatic arthritis: results of a double-blind, necrosis factor α therapy: results from the nationwide Danish DANBIO registry. randomized, placebo-controlled trial. Arthritis Rheum 2005;52:3279–89. Arthritis Rheum 2011;63:382–90.

Chandran V, et al. Ann Rheum Dis 2019;78:796–801. doi:10.1136/annrheumdis-2018-214737 801 Systemic lupus erythematosus

Translational science All-cause, cause-specific and age-specific standardised mortality ratios of patients with systemic lupus erythematosus in Ontario, Canada over 43 years (1971–2013) Konstantinos Tselios,1 Dafna D Gladman,‍ ‍ 2,3 Barry J Sheane,4 Jiandong Su,5 Murray Urowitz‍ ‍ 6

Handling editor Josef S Abstract Key messages Smolen Background survival in systemic lupus erythematosus 1 (SLE) has improved substantially in the last 50 years. Centre for Prognosis Studies in What is already known about this subject? the Rheumatic Diseases, Toronto The aim of the present study was to assess the evolution ►► Mortality has improved in systemic lupus Lupus Clinic, University Health of the all-cause, cause-specific and age-specific erythematosus (SLE) over the last 50 years. Network, University of Toronto, standardised mortality ratios (SMRs) of patients with However, the standardised mortality ratio (SMR) Toronto, Ontario, Canada lupus in Ontario, Canada. 2Centre for Prognosis Studies in is still approximately 3, indicating that patients Patients and methods Between 1971 and 2013, the Rheumatic Diseases, Toronto with SLE are three times more likely to die from 1732 patients were followed in the Toronto Lupus Lupus Clinic, University Health any cause as compared with patients without Network, University of Toronto, Clinic. Causes of death were retrieved from death lupus. Toronto, Ontario, Canada certificates, autopsy reports, hospital records or the 3Centre for Prognosis Studies records of the family physicians. They were categorised in the Rheumatic Diseases, What does this study add? University of Toronto Lupus as atherosclerotic, infectious, malignancy, active lupus ►► In this study, we showed that the all-cause Clinic, Toronto, Ontario, Canada and others. For the calculation of the SMR (indirect 4 and cause-specific (atherosclerosis, infections, Sports Surgery Clinic, Dublin, standardisation method), data from the general malignancies) SMR of patients with SLE have Ireland population of Ontario, Canada were used (Statistics 5Centre for Prognosis Studies in been significantly decreased from 1971 to 2013. the Rheumatic Diseases, Toronto Canada). Results Two hundred and forty-nine patients (205 Lupus Clinic, University Health How might this impact on clinical practice or Network, University of Toronto, women) died (infections 24.5%, atherosclerosis 15.7%, future developments? Toronto, Ontario, Canada active lupus 13.3%, malignancy 9.6%); mean age was 6Center for Prognosis Studies in ► SMR (all-cause and cause-specific) was 53.2±16.6 years and mean disease duration 15.2±11.7 ► the Rheumatic Diseases, Toronto particularly higher in patients younger than 40 years. The all-cause SMR was substantially decreased Lupus Clinic, University Health years old. These findings underline the need to Network, University of Toronto, from the 1970s (13.5, 95% CI 8.6 to 18.5) to recent optimise the management of the disease and its Toronto, Ontario, Canada years (2.2, 95% CI 1.4 to 3.1). Similar trends were major comorbidities, particularly in the younger observed for atherosclerosis, infections and malignancies age groups. Correspondence to over time. The all-cause age-specific SMR was particularly Dr Murray Urowitz, Center for Prognosis Studies in the high in younger (19–39 years old) patients (SMR=12.4, Rheumatic Diseases, Toronto 95% CI 9.7 to 15.1) as compared with individuals 9 Western Hospital, University of older than 40 years (SMR=3.1, 95% CI 2.6 to 3.6). The for the developed countries from 2008 to 2016. Toronto, Lupus Clinic, Toronto, cause-specificS MR was also higher in younger patients, However, their calculations were more modest for ON M5T2S8, Canada; particularly for infections and malignancies. the 10-year survival, which reached 89% for the m.​ ​urowitz@utoronto.​ ​ca Conclusions The all-cause and cause-specific SMR same period; the authors also reported that the 15-year survival was 82%.9 Received 21 November 2018 significantly decreased over time, likely reflecting Revised 19 March 2019 the advances in the management of SLE and certain Given that life expectancy has also greatly Accepted 23 March 2019 comorbidities. The all-cause and cause-specific SMR was improved in the general population for the same Published Online First 10 particularly high for younger patients (<40 years old). period, standardised indices that are age-ad- 16 April 2019 justed would better describe the real trends in SLE mortality. The standardised mortality ratio (SMR) represents the ratio of the number of deaths Introduction observed in a given cohort divided by the deaths Survival in systemic lupus erythematosus (SLE) reported in the general population for the same has improved considerably in the last 60 years. time period. The first population-based study of © Author(s) (or their employer(s)) 2019. No The 5-year survival rate was estimated at 50% in SMR in SLE was conducted in Sweden and reported 1 commercial re-use. See rights 1955, while it reached 64%–87% in the 1980s that the all-cause SMR was 3.63 (95% CI 3.49 to and permissions. Published and >95% in the early 2000s.2–5 Accordingly, the 3.78) from 1964 to 1995.11 Several studies were by BMJ. 10-year survival rate was reported at 87%–93% in published after that and were recently reviewed by 6 7 8 12 13 To cite: Tselios K, the 1990s and 96.3% in the contemporary years. two independent groups. In these meta-anal- Gladman DD, Sheane BJ, A recent systematic review and meta-analysis of 125 yses, Yurkovich et al12 and Lee et al13 reported an et al. Ann Rheum Dis studies in adult patients with lupus corroborated all-cause SMR of 2.98 (95% CI 2.32 to 3.83) and 2019;78:802–806. these findings with a pooled 5-year survival of 95% 2.66 (95% CI 2.09 to 3.39), respectively.

802 Tselios K, et al. Ann Rheum Dis 2019;78:802–806. doi:10.1136/annrheumdis-2018-214802 Systemic lupus erythematosus

Figure 1 Mean age at death of the UTLC patients through the decades (solid line). The life expectancy of the general Canadian population at the same time points is also depicted (dashed line) along with the potential years of life lost (lower dashed line).

However, both studies provided an overall estimate of the malignancies, atherosclerotic and others. Corresponding SMR throughout the observation period without analysing the ICD-10 code categories for infections included A00–B99, trend of SMR over time. The aim of the present study was to J09–J18 (respiratory infections), N30 (urinary tract infections) assess the evolution of the all-cause and cause-specific SMRs of and R65.2 (sepsis due to unidentified micro-organism). Corre- patients with lupus from 1971 to 2013 in Ontario, Canada. sponding codes for malignancies included C00–D49 whereas the respective codes for atherosclerosis included I20–25 (ischaemic Patients and methods coronary heart disease), I60–68 (cerebrovascular diseases), For the purposes of the present study, the long-term longitu- I70–75 (peripheral vascular disease) and I96 (gangrene). The dinal cohort of the University of Toronto Lupus Clinic (UTLC) same ICD-10 codes were applied to the UTLC cohort as well. was investigated from 1971 to 2013. During that period, 1732 The code M32 (SLE) was only applied to the UTLC cohort and patients with SLE were followed regularly in the clinic. Data not the general population. beyond 2013 were not used since, up to the time of writing, the cause-specific death counts by age/gender in Ontario, Canada Patient and public involvement were not available. Included patients are/were participating in the studies of the All patients fulfilled the revised American College of Rheu- University of Toronto Lupus Clinic after signed informed 14 matology criteria for the classification of SLE or had three consent. Relevant data for the general population of Ontario, criteria and a supportive biopsy (kidney). Patients are followed Canada were retrieved from Statistics Canada (as described regularly at intervals of 2–6 months according to a standardised above). research protocol, which is regularly updated.15 This protocol captures demographic, clinical, immunological and therapeutic variables as well as most comorbidities. Regarding mortality, the Statistical analysis protocol captures the cause (primary and secondary), location Descriptive statistics were represented by mean±SD for contin- and date of death. For the present study, only the primary causes uous variables and counts (per cent) for count variables. The of death were considered. Causes of death are recorded in the indirect standardisation method was used for the calculation of database based on autopsy reports, death certificates, hospital SMR. Ninety-five per cent CIs are reported for SMR where rele- discharge summaries or the records of the family physicians for vant. The procedure STDRATE in SAS V.9.4 was used to calcu- each patient. late all-cause and cause-specific SMR. In the context of this study, causes of death were categorised as atherosclerotic, infection, malignancy, active lupus and others. Results Patients were also categorised according to the age at death Two hundred and forty-nine patients (205 women) who had been in 10-year intervals (15–24, 25–34 etc up to 85–94) and in a enrolled in the UTLC cohort died between 1971 and 2013. The dichotomous manner (younger or older than 40 years). most common causes of death were infections (n=61, 24.5%), For the calculation of the SMR, mortality data from the atherosclerosis (n=39, 15.7%), active lupus (n=33, 13.3%) and general population of Ontario, Canada retrieved from Statis- malignancy (n=24, 9.6%). Concerning active lupus, the precise tics Canada for the same time period were used to derive the causes of death were multiple organ failure syndrome (in the expected number of events in our study population. The ratio of absence of sepsis, n=8), active lupus nephritis with rapidly observed to expected events provided the SMR. Primary causes progressive glomerulonephritis (n=7), central nervous system of death are available in Ontario since 1980. As such, cause-spe- vasculitis or cerebritis (n=6), lupus myocarditis (n=4), mesen- cific SMRs could be calculated after that time. From 1971 to teric vasculitis (n=3), diffuse alveolar haemorrhage/lupus pneu- 1979, only the all-cause SMR was calculated. Causes of death monitis (n=3), thrombotic thrombocytopenic purpura (n=1) were defined according to the 10th revision of the International and pericarditis/cardiac tamponade (n=1), all in the context Classification of Diseases (ICD-10) and grouped into infections, of active disease. Sixty-seven patients died from other causes

Tselios K, et al. Ann Rheum Dis 2019;78:802–806. doi:10.1136/annrheumdis-2018-214802 803 Systemic lupus erythematosus

Table 1 All-cause and cause-specific standardised mortality ratios* for SLE from 1971 to 2013 1971–1979 1980–1989 1990–1999 2000–2009 2010–2013 All cause 13.5 (8.6–18.5) 6.5 (4.9–8.2) 4.7 (3.6–5.8) 3.2 (2.4–4) 2.2 (1.4–3.1) No of deaths 29 59 72 64 25 Atherosclerosis 8.3 (3.8–12.8) 6.7 (3.8–9.5) 2.3 (0.8–3.8) 3.2 (0.1–6.3) No of deaths 3 9 16 8 3 Infection 14.2 (8–20.4) 6.5 (3.8–9.1) 2.5 (1.2–3.9) 0.9 (0–1.9) No of deaths 11 19 17 10 4 Malignancy 14.1 (4.3–23.9) 3 (1.1–4.8) 4.4 (1.5–7.3) 1.4 (0.2–2.7) No of deaths 0 6 9 5 4 *As compared to the general population of Ontario, Canada (1971–-2013). Values are given as standardised mortality ratioSMR (95% CIonfidence Interval).

(mainly renal failure, chronic obstructive pulmonary disease, (19–39 years) as compared with patients who were ≥40 years gastrointestinal bleeding, pulmonary embolism etc) whereas old (table 3). the cause was unknown in 25 (10%). The mean age at death was 53.2±16.6 years and mean disease duration 15.2±11.7 Discussion years. The mean age at death was significantly increased from In the present study, we showed that all-cause and cause-spe- 42.2±12.9 years in the 1970s to 58.8±14.6 years during the cific SMRs (for infections, atherosclerosis and malignancies) recent years 2010–2013 (details in figure 1). For the same time were progressively substantially decreased in Ontario, Canada span, the potential years of life lost were decreased from 33.5 in from 1971 to 2013. In addition, all-cause and cause-specific the 1970s to 23.6 between 2010 and 2013 (figure 1). SMRs were considerably higher for patients 19–39 years old as compared with patients 40 years of age and older. With regards All-cause and cause-specific SMRsstandardized mortality to age, the mortality rate was always higher than that of the ratios over time general population in all age groups except for the patients who The all-cause SMR over the entire period 1971–2013 was 4.1 were older than 85 years at the time of death. (95% CI 3.6 to 4.6). Regarding its evolution over time, there The significant decrease of the all-cause SMR over time was was a significant decrease from 13.5 (95% CI 8.6 to 18.5) in the first reported by Bjornadal et al in Sweden (from 1964 to 199511) 1970s to 2.2 (95% CI 1.4 to 3.1) during 2010–2013. Likewise, and then by Bernatsky et al in a multiethnic cohort (from 1970 the overall 1980–2013 cause-specific SMR was 4.7 (95% CI 3.4 to 2000).16 For the latter study, data were collected from 23 to 6.0) for atherosclerosis, 4.4 (95% CI 3.3 to 5.5) for infec- centres and 9547 patients (1255 deaths); the all-cause SMR was tions and 3.4 (95% CI 2.2 to 4.6) for malignancies. However, 2.4 (95% CI 2.3 to 2.5) from 1970 to 2001 with a substantial these cause-specific SMRs were considerably decreased from the decrease from the 1970s to the 1990s.16 Furthermore, Yurkovich 1980s to 2010–2013. The evolution of all-cause and cause-spe- et al reported a meta-SMR of 2.98 (95% CI 2.3 to 3.8) based on cific SMRs over the decades is shown in table 1. a meta-analysis of 12 studies (27 123 patients, 4993 deaths from 1950 to 2008),12 whereas that was 2.66 (95% CI 2.09 to 3.39) Age-specific SMR in a more recent meta-analysis (15 studies, 26 101 patients, 4640 13 The all-cause SMR by decades of age was the highest in the deaths). Our study yielded a significantly higher all-cause SMR age group 25–34 (19.7, 95% CI 13 to 26.4) and was gradually (4.1, 95% CI 3.6 to 4.6, from 1971 to 2013), although that decreased in older age groups. Details are given in table 2 and figure 2. The age-specific SMR was particularly high in younger (19–39 years old) patients (all-cause SMR=12.4, 95% CI 9.7 to 15.1) as compared with the patients who were older than 40 years (all-cause SMR=3.1, 95% CI 2.6 to 3.6). Likewise, cause-specific SMRs were significantly higher in this age group

Table 2 Age-specific SMR (all-cause) for UTLC patients in Ontario, Canada 1971–2013 Age (years) at death or last Number of Number of visit patients deaths SMR 95% CI 19–24 210 10 14.00 5.32 to 22.68 25–34 343 33 19.66 12.95 to 26.37 35–44 358 37 9.08 6.16 to 12.01 Figure 2 Age-specific standardised mortality ratio (SMR) for all 45–54 338 47 5.17 3.69 to 6.65 causes according to the age at death divided in decades. The SMR 55–64 251 47 3.24 2.31 to 4.17 was particularly high in young ages (19–44 years) while it remained 65–74 153 45 2.51 1.77 to 3.24 significantly higher than one even for patients who died between 75 75–84 71 27 2.44 1.52 to 3.36 and 84 years of age. For patients older than 85 years, the mortality 85–94 8 3 1.09 0.00 to 2.59 rate was not different than that of the general population of Ontario, SMR, standardised mortality ratio. Canada.

804 Tselios K, et al. Ann Rheum Dis 2019;78:802–806. doi:10.1136/annrheumdis-2018-214802 Systemic lupus erythematosus

extrapolation of the infection-related and malignancy-related Table 3 All-cause and cause-specific standardised mortality ratios* SMR.12 13 (SMRs) for patients 19–39 years old and ≥40 years 1980–2013 Interestingly, the all-cause SMR was four times higher in Deaths Deaths patients 20–39 years of age as compared with those over 40. (19–39 SMR (19–39 (≥40 SMR (≥40 For infections and malignancies, in particular, the relative risk Causes of death years), n years) years), n years) was 8.6 and 10.6, respectively, while atherosclerosis had a rela- All cause 58 12.4 (9.7–15.1) 191 3.1 (2.6–3.6) tive risk of 3.1. Furthermore, the all-cause SMR was decreasing Atherosclerosis 3 14.6 (0–43.3) 36 4.7 (3.3–6) for increasing age, and it was similar to that of the general Infection 30 30.2 (14.4–46) 31 3.5 (2.5–4.5) population only for the age group >85 years. These results are Malignancy 4 31.9 (0.6–63.1) 20 3 (1.9–4.1) in agreement with Bernatsky et al who reported an all-cause *As compared with the general population of Ontario, Canada 1971–2013. Values SMR of 10.7 for patients <40 years old and 1.4 for patients are given as SMR (95% CI). >60 years old.16 Similarly, Bjornadal et al described an all-cause SMR of 11.1 for patients 20–39 years old that decreased to 2.91 for the individuals who were older than 60.11 A more was comparable after 2000 (3.2, 95% CI 2.4 to 4). Concerning recent epidemiologic study of the entire female population the study of Bernatsky et al, the difference could be attributed of the USA showed that SLE was ranking 10th in the causes to the population diversity, with centres from Sweden, Iceland of death for the ages 15–24 and 14th for the ages 25–34 and and Scotland displaying an all-cause SMR less than 2 even from 35–44.18 the 1990s.16 On the contrary, Canadian centres had an all-cause Limitations of the present study include the relatively small SMR greater than 3. With regards to the meta-analyses, the number of deaths that does not allow for further categorisation range of the all-cause SMR was rather wide (from 0.95 to 7.23), of the patients (eg, men/women, history of renal involvement while a significant proportion of the patients were duplicated in or not, therapy etc). The medications used for disease manage- the included studies.12 ment are an important confounder in such outcomes. However, The noticeable improvement of the mortality rate in SLE a further subcategorisation of the patients according to their reflects the better identification of the patients, possibly related treatment would decrease the sample sizes significantly and to the introduction of the 1982 American College of Rheu- influence the statistical power of the analysis. In addition, the matology criteria for the classification of SLE,17 as well as the lack of ascertainment of the cause of death in 25 patients might improved management of the disease itself and the comorbidities have influenced the cause-specific SMRs. Such data (unknown that often complicate its course, such as atherosclerotic disease cause of death) are rarely provided in similar reports. In the and infections. In our study, atherosclerosis had an overall SMR study of Bernatsky et al,16 763/1255 deaths were attributed to of 4.7 with a substantial decrease from 8.3 in the 1980s to 3.2 the common causes of circulatory and renal diseases, infections in the most recent years 2010–2013. Bernatsky et al reported and malignancies. The precise number of deaths with unknown an SMR of 1.7 that remained practically unaltered from 1970 cause (of the 492) was not provided. In our study, the SMR for to 2001.16 Similarly, the SMR for cardiovascular causes of death deaths specifically due to active lupus could not be calculated was 2.72 (95% CI 1.83 to 4.04)12 and 2.25 (95% CI 1.30 to since this information was not available for the general popu- 3.89)13 in the two meta-analyses, which were heavily influenced lation. Moreover, the suboptimal reliability of the death certif- by the study of Bernatsky et al that contributed approximately icates must be considered. In a systematic review of 53 distinct 67% of the total patients. That can probably be attributed to autopsy serries, it was reported that a major error (clinically the different definitions applied. Indeed, Bernatsky et al used missed diagnosis involving a primary cause of death) occurred the term ‘circulatory’ diseases, which incorporates all diseases of in 23.5% (median).19 Of note, the cardiovascular, infectious the heart and pulmonary vasculature, even those entities without and neoplastic diseases have shown the highest rates of agree- an atherosclerotic background such as myocarditis or valvular ment between clinical diagnoses and autopsy reports (exceeding diseases. In contrast, we used only the ICD-10 codes that apply 90%).20 to atherosclerotic disease (either cardiac or cerebrovascular). In conclusion, the all-cause and cause-specific (atheroscle- Bjornadal et al reported an SMR of 2.97 (95% CI 2.78 to 3.16) rosis, infections, malignancies) SMR for patients with lupus has for coronary heart disease and stroke in Sweden for the period decreased substantially from the 1970s to recent years. The SMR 1964–1995.11 (all-cause and cause-specific) is significantly higher for younger Concerning the infection-related mortality, our findings are patients (19–39 years old) as compared with individuals older similar to those reported by previous studies. The overall SMR than 40. There is an excessive risk of all-cause mortality in SLE was 4.4, quite close to the 4.98 that was reported from both for all ages except for the age group >85 years. 12 13 meta-analyses. Of note, there was a substantial decrease of Contributors All authors were involved in drafting the article or revising it the SMR from 14.2 in the 1980s to 0.9 in 2010–2013, implying critically for important intellectual content, and all authors approved the final that infections overall did not confer an excessive risk for death version to be published. MU had full access to all of the data in the study and in patients with SLE in contemporary years. This finding should takes responsibility for the integrity of the data and the accuracy of the data analysis. be interpreted with caution since the number of deaths was small and the duration of that period (4 years, 2010–2013) was Funding KT is supported by the Lupus Program, Centre for Prognosis Studies in the Rheumatic Diseases. shorter than the other periods (10 years). Regarding malignan- cies, both meta-analyses reported that the risk of death was not Competing interests None declared. significantly increased in patients with lupus with an SMR of Patient consent for publication Obtained. 1.19 (95% CI 0.89 to 1.59) and 1.16 (95% CI 0.57 to 2.36), Ethics approval The study was approved by the University Health Network respectively.12 13 In contrast, we observed an overall SMR of 3.4 Research Ethics Board. from 1971 to 2013 that was decreased and became not signifi- Provenance and peer review Not commissioned; externally peer reviewed. cant in 2010–2013 (1.4, 95% CI 0.2 to 2.7). Of note, the two Data sharing statement All authors were involved in the study conception and meta-analyses were based on the same two studies11 16 for the design, acquisition of data, and analysis and interpretation of data.

Tselios K, et al. Ann Rheum Dis 2019;78:802–806. doi:10.1136/annrheumdis-2018-214802 805 Systemic lupus erythematosus

References 11 Björnådal L, Yin L, Granath F, et al. Cardiovascular disease a hazard despite improved 1 Merrell M, Shulman LE. Determination of prognosis in chronic disease, illustrated by prognosis in patients with systemic lupus erythematosus: results from a Swedish systemic lupus erythematosus. J Chronic Dis 1955;1:12–32. population based study 1964–95. J Rheumatol 2004;31:713–9. 2 Uramoto KM, Michet CJ, Thumboo J, et al. Trends in the incidence and mortality of 12 Yurkovich M, Vostretsova K, Chen W, et al. Overall and cause-specific mortality in systemic lupus erythematosus, 1950–1992. Arthritis Rheum 1999;42:46–50. patients with systemic lupus erythematosus: a meta-analysis of observational studies. 3 alamanos Y, Voulgari PV, Siozos C, et al. Epidemiology of systemic lupus Arthritis Care Res 2014;66:608–16. erythematosus in northwest Greece 1982–2001. J Rheumatol 2003;30:731–5. 13 lee YH, Choi SJ, Ji JD, et al. Overall and cause-specific mortality in systemic lupus 4 Urowitz MB, Gladman DD, Tom BDM, et al. Changing patterns in mortality and erythematosus: an updated meta-analysis. Lupus 2016;25:727–34. disease outcomes for patients with systemic lupus erythematosus. J Rheumatol 14 Hochberg MC. Updating the American College of Rheumatology revised criteria for 2008;35:2152–8. the classification of systemic lupus erythematosus. Arthritis Rheum 1997;40. 5 Borchers AT, Keen CL, Shoenfeld Y, et al. Surviving the butterfly and the wolf: mortality 15 Urowitz MB, Gladman DD. Contributions of observational cohort studies in systemic trends in systemic lupus erythematosus. Autoimmun Rev 2004;3:423–53. lupus erythematosus: the University of Toronto Lupus Clinic experience. Rheum Dis 6 swaak AJ, Nossent JC, Bronsveld W, et al. Systemic lupus erythematosus. I. Outcome Clin North Am 2005;31:211–21. and survival: Dutch experience with 110 patients studied prospectively. Ann Rheum 16 Bernatsky S, Boivin J-F, Joseph L, et al. Mortality in systemic lupus erythematosus. Dis 1989;48:447–54. Arthritis Rheum 2006;54:2550–7. 7 Pistiner M, Wallace DJ, Nessim S, et al. Lupus erythematosus in the 1980s: a survey of 17 Tan EM, Cohen AS, Fries JF, et al. The 1982 revised criteria for the classification of 570 patients. Semin Arthritis Rheum 1991;21:55–64. systemic lupus erythematosus. Arthritis Rheum 1982;25:1271–7. 8 Mahmoud GA, Shahin AA, Zayed HS, et al. Clinical and immunological pattern 18 Yen EY, Singh RR. Brief report: Lupus—an unrecognized leading cause of death and outcome of Egyptian systemic lupus erythematosus patients: a single center in young females: a population-based study using nationwide death certificates, experience. Lupus 2018;27:1562–9. 2000–2015. Arthritis Rheumatol 2018;70:1251–5. 9 Tektonidou MG, Lewandowski LB, Hu J, et al. Survival in adults and children with 19 shojania KG, Burton EC, McDonald KM, et al. Changes in rates of systemic lupus erythematosus: a systematic review and Bayesian meta-analysis of autopsy-detected diagnostic errors over time: a systematic review. JAMA studies from 1950 to 2016. Ann Rheum Dis 2017;76:2009–16. 2003;289:2849–56. 10 christensen K, Doblhammer G, Rau R, et al. Ageing populations: the challenges 20 schwanda-Burger S, Moch H, Muntwyler J, et al. Diagnostic errors in the new ahead. Lancet 2009;374:1196–208. millennium: a follow-up autopsy study. Mod Pathol 2012;25:777–83.

806 Tselios K, et al. Ann Rheum Dis 2019;78:802–806. doi:10.1136/annrheumdis-2018-214802 Systemic sclerosis

Clinical science Development and validation of the Scleroderma Clinical Trials Consortium Damage Index (SCTC-DI): a novel instrument to quantify organ damage in systemic sclerosis Nava Ferdowsi,‍ ‍ 1,2 Molla Huq,1,2 Wendy Stevens,2 Marie Hudson,3 Mianbo Wang,4 Tien Tay,1,2 Jodie L Burchell,1 Sam Mancuso,1 Candice Rabusa,2 Vijaya Sundararajan,5 David Prior,1 Susanna M Proudman,6 Murray Baron,7 Mandana Nikpour,1,2 The Scleroderma Clinical Trials Consortium Damage Index Working Group, The Australian Scleroderma Interest Group, Canadian Scleroderma Research Group

Handling editor Prof Josef S Abstract Key messages Smolen Objective We sought to develop the first Damage Index (DI) in systemic sclerosis (SSc). ►► Additional material is What is already known about this subject? published online only. To view Methods The conceptual definition of ’damage’ in ►► Organ damage is part of the natural history of please visit the journal online SSc was determined through consensus by a working systemic sclerosis (SSc). (http://dx.​ ​doi.org/​ ​10.​1136/​ group of the Scleroderma Clinical Trials Consortium ► Organ damage occurs early in the disease annrheumdis-2018-​ ​214764). (SCTC). Systematic literature review and consultation ► course of SSc. with patient partners and non-rheumatologist experts For numbered affiliations see ► Based on a systematic review of the literature, produced a list of potential items for inclusion in the DI. ► end of article. no tool currently exists to measure organ These steps were used to reduce the items: (1) Expert damage in SSc. Correspondence to members of the SCTC (n=331) were invited to rate What does this study add? Dr Mandana Nikpour, the appropriateness of each item for inclusion, using ► The first ever composite disease Damage Index Department of Medicine, a web-based survey. Items with >60% consensus ► University of Melbourne, Fitzroy in SSc has been developed. were retained; (2) Using a prospectively acquired 3065, Victoria, Australia; ► The methodology used to develop the index is Australian cohort data set of 1568 patients, the ► m.​ ​nikpour@unimelb.​ ​edu.​au somewhat unique, with the combination of both univariable relationships between the remaining items consensus methods and statistical analysis of MB and MN contributed equally. and the endpoints of mortality and morbidity (Physical patient data. Component Summary score of the Short Form 36) were Received 14 November 2018 How might this impact on clinical practice or analysed, and items with p<0.10 were retained; (3) Revised 11 March 2019 future developments? Accepted 12 March 2019 using multivariable regression analysis, coefficients were ► The Damage Index can potentially be used in Published Online First used to determine a weighted score for each item. The DI ► observational studies and clinical trials as an 30 March 2019 was externally validated in a Canadian cohort. endpoint, and for the enrichment of clinical Results ninety-three (28.1%) complete survey trials. responses were analysed; 58 of 83 items were retained. The univariable relationships with death and/or morbidity endpoints were statistically significant for 22 items, with one additional item forced into the multivariable model to define ‘organ damage’ in SSc, we found that by experts due to clinical importance, to create a 23-item irreversible organ damage accrues very early in the weighted SCTC DI (SCTC-DI). The SCTC-DI was predictive disease course, with 40% of patients having damage in one or more organ systems within 2 years of of morbidity and mortality in the external cohort. 2 Conclusions Through the combined use of consensus disease onset. and data-driven methods, a 23-item SCTC-DI was Despite the fundamental importance of organ developed and retrospectively validated. damage in SSc, there has been no clinical tool to identify and quantify damage in SSc, with current tools assessing disease severity, activity or combined treatment responses.3 The lack of an existing tool, Introduction together with the observation of damage accrual Systemic sclerosis (SSc; ‘scleroderma’) is a multi- early in the disease course of SSc, led to an initia- © Author(s) (or their employer(s)) 2019. No organ disease wherein the pathogenic processes tive to develop the multiorgan Scleroderma Clin- commercial re-use. See rights of inflammation, vasculopathy and fibrosis lead to ical Trials Consortium Damage Index (SCTC-DI). and permissions. Published multiple disease manifestations.1 Unlike other rheu- Potential applications of this instrument include use by BMJ. matological diseases that have a relapsing-remitting as an outcome measure in interventional drug trials To cite: Ferdowsi N, course, the course of SSc is often one of progres- and in observational studies, and as a selection tool Huq M, Stevens W, sive damage to multiple organs including the skin, to enrich clinical trials for patients at risk of devel- et al. Ann Rheum Dis joints, heart, lungs, gastrointestinal (GI) tract and oping organ damage who are most likely to benefit 2019;78:807–816. kidneys. Using a preliminary ad hoc set of criteria from therapy.

Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 807 Systemic sclerosis

The development of the SCTC-DI was an international the Australian Scleroderma Interest Group (ASIG) were invited collaboration using a combined approach of consensus and data- by email to take part in an online survey (Research Electronic driven methods applied to data from the prospective Australian Data Capture (REDCap), Vanderbilt University, Nashville Scleroderma Cohort Study (ASCS). External validation of the Tennessee; survey included as an online supplementary file) SCTC-DI was performed in the prospective Canadian Sclero- to rate the appropriateness (on a scale of 1–7 with 1=very derma Research Group (CSRG) registry. inappropriate, 2=inappropriate, 3=somewhat inappropriate, 4=neither inappropriate or appropriate, 5=somewhat appro- Methods priate, 6=appropriate, 7=very appropriate) in terms of both Investigators importance and feasibility of recording in a day-to-day setting, The SCTC-DI working group of including each item in an SSc DI. The survey items were cate- The SCTC is an international consortium representing researchers gorised according to organ systems and for each organ system, a and clinicians who have particular interest and expertise in care link was provided within the survey to a repository of 64 articles 4–66 and research of SSc. The goal of the SCTC is to conduct, sponsor from the systematic literature review. An a priori cut-off of or facilitate clinical research in SSc, in particular projects aimed 60% votes for ‘appropriate’ (score of 5 or above) was used to at improving and developing outcome measures for clinical trials reduce the items. Where applicable, responders were also asked and observational studies. to select an option regarding a time period for which the item In 2013, under the auspices of the SCTC, a working group needed to be present in order to constitute damage (present ever, comprised 22 international experts in SSc, representing Austral- 6 months, 12 months or 2 years). asia, the Americas and Europe, was assembled to develop a disease DI in SSc. The working group was led by a steering committee comprised 6 members (MN, MB, MH, WS, SMP, NF) Statistical analysis to achieve further item reduction and item overseeing the day-to-day development of the DI. weighting Data from 1568 patients in the ASCS fulfilling the European League Against Rheumatism (EULAR)/American College of Non-rheumatology expert advisors Rheumatology (ACR) criteria for SSc4 were used to retrospec- A panel of seven non-rheumatologist experts with an interest tively assess the univariable relationship of each damage item in SSc was assembled to give recommendations regarding what with the endpoint of mortality using time-dependent Cox constitutes significant damage in each organ system. This hazards regression models. ASCS is an ongoing longitudinal panel comprised two cardiologists, two respiratory physicians, study of patients with SSc recruited from nine sites in Australia two gastroenterologists and one nephrologist. Input from this since 2007, in whom standardised clinical (including the panel was sought for: (1) item generation for a survey of SSc health-related quality of life (HRQoL) questionnaire Short Form experts and (2) ensuring face and content validity of the final 36 (SF-36))67 and laboratory data are collected at each annual instrument. visit and recorded in a customised database. As there is no specific measure of morbidity in SSc per se, the Patient partners Physical Component Summary (PCS) score of the SF-36 ques- Three experienced SSc patient partners (from Australia, tionnaire was used as a surrogate for morbidity. SF-36 data Canada and the UK) provided insight from the patients’ were available for 1877 visits and panel logistic regression was perspective. Input from these patient partners was sought at used to determine the univariable relationship between each several stages in the process of instrument development: (1) damage item and the SF-36 PCS (dichotomised for analysis conceptual definition of organ damage; (2) item generation based on median value in the cohort) measured at each visit for for a survey of SSc experts; (3) ensuring face and content each patient. This method allows for the expected correlation validity of the final instrument. between repeated measures over time within an individual. As not all items in the ASCS were defined exactly according to the Development of the SCTC-DI definitions used in the DI, the closest possible equivalent of Conceptual definition of organ damage each item was used in item reduction and weighting. Items not The working group members participated in a two-round survey demonstrating a univariable relationship (p<0.10) with the to achieve consensus regarding the conceptual definition of mortality or morbidity endpoints were removed. In addition, organ damage in SSc. In the second round, responses from the items that were highly clinically collinear with other items first round were provided, with the survey focusing on resolu- or lacked feasibility were removed as per steering committee tion of contentious areas. The definition was then presented to consensus. each working group member to ensure further comments and In order to create a weighted score, the multivariable feedback were exhausted. All working group members had to coefficients were then calculated for the remaining items approve the final definition. against each of the mortality and morbidity end points. The following rules were used to combine the correlation coeffi- Item generation cients from each of the mortality and morbidity analyses in A systematic review of existing measures of disease status in SSc order to create a weighted score for each item: (1) if both was performed to generate items for potential inclusion in the coefficients were positive, an average was taken; (2) if one 3 DI. The working group members, expert advisors and patient coefficient was negative and one positive, we took the posi- partners, made further revisions to item definitions and provided tive only; (3) if both coefficients were negative, indicating further items for inclusion in a survey. protection from damage, the item was discarded. For scoring items, the following methods were used: (1) all the coeffi- Item reduction using consensus survey cients were divided by the smallest coefficient and multiplied Three hundred and thirty-one expert members of the SCTC, by 0.25. The multiplier of 0.25 was chosen to ensure that European Scleroderma Trials and Research Group, CSRG and if the smallest coefficient was very small (<0.10), the scores

808 Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 Systemic sclerosis did not become too large; (2) the result was rounded to the be reflective of morbidity and/or mortality. The duration that nearest integer. each abnormality must be present before constituting damage may depend on the particular item.’ Validation It must be noted that damage items caused by treatment of SSc The performance of the SCTC-DI in the derivation cohort were originally included in the definition, but later removed (see (internal validation) was determined using receiver operating below). characteristic (ROC) curves. Kaplan-Meier survival estimates against mortality and morbidity endpoints with patients divided into risk groups based on their SCTC-DI scores were generated. Item generation A total of 83 items for potential inclusion in the DI were gener- Risk groups were determined by looking for clear demarcation 3 of the cumulative frequency percentage of total scores within the ated through the systemic review of the literature and consulta- derivation cohort. Both of these analyses were used to demon- tion with expert advisors and patient partners (all items included strate criterion validity. Frequency of damage items in a subset of in the survey can be seen in online supplementary file 2). The patients with early disease at each year of follow-up was deter- variables come from six different organ system groupings, as mined to demonstrate construct validity and responsiveness to follows: musculoskeletal and skin, vascular, GI, respiratory, change over time. cardiovascular and renal. External validation was conducted using the CSRG patient registry. The CSRG is an ongoing longitudinal study of patients Item reduction using consensus survey with SSc fulfilling the EULAR/ACR criteria4 recruited from 14 A total of 93 of 331 (28.1%) international SSc experts provided sites in Canada and 1 site in Mexico since 2004, using a stan- complete responses. Of the 83 items presented, 58 items had dardised data collection protocol that is comparable to that of greater than 60% consensus with 36 items removed (online ASCS. Once again, ROC analyses were performed. Kaplan- supplementary table S3). Of these 58 items, nine items (bony Meier survival estimates were performed using the SCTC-DI erosions seen on radiographic imaging, acro-osteolysis on radio- risk scores generated in the derivation cohort. Damage accrual graphic imaging, sicca signs confirmed on objective testing, was assessed using the mean SCTC-DI score at each year of surgical limb amputation, insertion of permanent pacemaker follow-up in each organ domain and in the overall cohort. or implantable cardioverter defibrillator, histologically proven Barrett’s oesophagus, rectal prolapse, lung malignancy with SSc Results interstitial lung disease (ILD), left ventricular failure on clinical Development of the SCTC-DI assessment and confirmed on transthoracic echocardiogram Conceptual definition of organ damage [TTE]) were not collected in the ASCS database and could not be A total of 20 (of 22) responses from the working group analysed. Overall, the working group was comfortable with the members were received (for a summary of results see online removal of these items given the inclusion of similar alternatives, supplementary tables S1 and S2). The resulting definition of and in most cases, more clinically appropriate items. damage in SSc was as follows: ‘Damage in SSc is the perma- nent and irreversible loss of anatomical structure or physiolog- ical function, caused by SSc and not secondary to its treatment Removal of ‘damage due to treatment’ items or comorbidities. Damage should be differentiated from A preliminary conceptual definition of damage included ‘damage disease activity, which is potentially reversible, and disease caused by the treatment of SSc’. Inclusion of these items was severity which encompasses both activity and damage. The deemed unnecessary as most drug treatment clinical trials would DI will not include patient-reported outcomes. The DI should be collecting extensive data on drug-related adverse outcomes.

Figure 1 Scleroderma Clinical Trials Consortium-Damage Index (SCTC-DI)—item generation and reduction steps. *SF 36 PCS, Short Form 36 Physical Component Summary score.

Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 809 Systemic sclerosis

Furthermore, ascertaining attribution of damage to treatment Inclusion of renal items was deemed unreliable. The items in the renal domain, such as ‘scleroderma renal crisis’ (SRC) and ‘need for renal replacement therapy’ (dialysis), occur rarely in SSc, and therefore, their relationship with mortality did Removal of 6-minute walk distance not reach statistical significance in our derivation data set. Two The 6 min walk distance is used in some countries for various of the three renal items (SRC and SRC requiring renal replace- indications, including screening and monitoring of cardiopulmo- ment therapy) did, however, have a significant univariable nary disease in SSc, and access to drugs that treat right heart cath- relationship with morbidity. Given its clinical importance, the eterisation-proven pulmonary arterial hypertension. However, working group decided to also include the third renal item (SRC because of the lack of widespread use in clinical practice and and persistent renal impairment) despite the lack of a statistically accessibility across all centres, the working group members and significant relationship with either morbidity or mortality in the patient partners decided to remove this item. derivation data set.

Table 1 SCTC-DI final items—coefficient calculations and weighted score Mortality Morbidity Combined Weighted Item coeff coeff coeff score Musculoskeletal and skin 1 Joint contracture defined as any degree of contracture with the inability to reduce the joint to the anatomically −0.35 0.49 0.49 2 neutral position in any small joint of the fingers. 2 Joint contracture defined as any degree of contracture with the inability to reduce the joint to the anatomically −0.35 0.49 0.49 2 neutral position in the large joints, specifically elbows and knees. 3 Sicca symptoms defined as presence of patient-reported dry eyes and/or dry mouth requiring treatment on a −0.10 0.75 0.75 3 daily basis, for example, lubricant eye-drops, punctual plugs, saliva replacement. 4 Proximal muscle weakness on clinical examination defined as shoulder abduction and/or hip or knee flexion 0.60 0.92 0.76 3 less than 5/5 power (not due to contracture or pain). 5 Calcinosis complicated by infection or requiring surgery. −0.21 1.08 1.08 4 Vascular 6 Digital ulceration defined as loss of epithelialisation, of any degree, of the epidermis, the dermis and/or the 0.79 0.37 0.58 2 subcutaneous tissue, distal to or at the proximal interphalangeal joint of the hands or feet not thought to be due to trauma and refractory to therapy. 7 Digital amputation (surgical or autoamputation). 0.96 1.09 1.03 3 Gastrointestinal 8 Oesophageal dysmotility defined as distal dysphagia refractory to treatment with differential diagnoses (eg, −0.21 0.12 0.12 1 oesophageal stricture or malignancy) excluded by endoscopy. 9 Pseudo-obstruction with symptoms such as vomiting or constipation, with dilatation of the small and/or large 0.93 −0.82 0.93 3 bowel on imaging. 10 Low body mass index of <18.5 kg/m2 or weight loss of >10% in the last 12 months. 0.68 0.74 0.71 2 11 Gastric antral vascular ectasia confirmed on endoscopy. 0.67 0.32 0.50 2 12 Oesophageal stricture confirmed on testing such as endoscopy or barium swallow. 0.15 0.20 0.18 1 13 Symptoms of gastro-oesophageal reflux disease (heart burn) refractory to treatment and confirmed on 0.03 0.71 0.37 1 endoscopy. Respiratory 14 Moderate to severe interstitial lung disease with >20% extent on HRCT of the chest. 0.19 0.97 0.58 2 15 Moderate to severe interstitial lung disease with >20% extent on HRCT of the chest and forced vital capacity 1.11 2.47 1.79 6 <70% on lung function tests. 16 Dependence on home oxygen. 0.88 1.99 1.44 5 Cardiovascular 17 Pulmonary arterial hypertension (defined as mean pulmonary arterial pressure >25 mm Hg at rest and 0.75 0.59 0.67 2 pulmonary arterial wedge pressure <15 mm Hg on right heart catheterisation). 18 Moderate to severe right ventricular dysfunction noted on echocardiography report based on assessment of 1.81 2.49 2.15 7 any measure of RV function by experienced cardiologist. 19 Myocardial disease attributable to SSc based on a constellation of clinical features and supportive 0.85 0.77 0.81 3 investigations, for example, syncope secondary to conduction abnormality, arrhythmia requiring defibrillator, heart block requiring permanent pacemaker or ablation, systolic or diastolic dysfunction on TTE. 20 Presence of moderate to large pericardial effusion equivalent to greater than 1 cm on TTE. 0.45 0.09 0.27 1 Renal 21 History of scleroderma renal crisis (SRC), either hypertensive or normotensive, as defined by the International 0.46 1.54 1.00 3 Scleroderma Renal Crisis Study Investigators.45 22 History of SRC or other SSc-related kidney disease and persistent renal impairment with estimated glomerular 0.66 1.76 1.21 4 filtration rate <45 mL/min/1.73 m2. 23 SRC with stage 5 renal impairment defined as need for renal replacement therapy. 1.08 2.16 1.62 5 Coeff, coefficient; HRCT, high-resolution CT; RV, right ventricular;SCTC-DI, Scleroderma Clinical Trials Consortium-Damage Index; SSc, systemic sclerosis; TTE, transthoracic echocardiogram.

810 Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 Systemic sclerosis

Table 2 Scleroderma Clinical Trials Consortium Damage Index Table 2 Continued Item Score Item Score Musculoskeletal and skin History of scleroderma renal crisis (SRC), either hypertensive or 3 Joint contracture defined as any degree of contracture with the inability to 2 normotensive, as defined by the International Scleroderma Renal Crisis Study reduce the joint to the anatomically neutral position in any small joint of the Investigators.45 fingers.*† ►► Add 1 if history of SRC or other SSc-related kidney disease and Joint contracture defined as any degree of contracture with the inability 2 persistent renal impairment with estimated glomerular filtration rate 2 to reduce the joint to the anatomically neutral position in the large joints, <45 mL/min/1.73 m 1 specifically elbows and knees.*† ►► Add 2 if SRC with stage 5 renal impairment and need for renal replacement therapy Sicca symptoms defined as presence of dry eyes and/or dry mouth requiring 3 treatment on a daily basis, for example, lubricant eye-drops, punctual plugs, saliva replacement.* 2 Proximal muscle weakness on clinical examination defined as shoulder 3 TOTAL SCORE 55 abduction and/or hip or knee flexion less than 5/5 power (not due to contracture or pain).* Attribution to SSc required for all items. *Item must be present for a minimum of 6 months. Calcinosis complicated by infection or requiring surgery. 4 †See figure below for ‘anatomically neutral’ position. Vascular Digital ulceration defined as loss of epithelialisation, of any degree, of the 2 epidermis, the dermis and/or the subcutaneous tissue, distal to or at the proximal interphalangeal joint of the hands or feet not thought to be due to trauma and refractory to therapy* Add 1 if digital amputation required (surgical or autoamputation). HRCT, high-resolution CT; RV, right ventricular;SSc, systemic sclerosis; TTE, transthoracic echocardiogram. 1 Gastrointestinal Oesophageal dysmotility defined as distal dysphagia refractory to treatment, 1 with differential diagnoses (eg, oesophageal stricture or malignancy) Items removed due to clinical collinearity excluded by endoscopy. The following 12 items were removed as their definition was Oesophageal stricture confirmed on testing such as endoscopy or barium 1 very similar to that of other item(s) retained in the DI that swallow. were either more strongly associated with the mortality and Symptoms of gastro-oesophageal reflux disease (heart burn) refractory to 1 morbidity endpoints, or were more feasible to record reliably treatment (eg, proton pump inhibitors) and confirmed on endoscopy.* in a clinical care setting: joint contracture in any large and/or Gastric antral vascular ectasia confirmed on endoscopy. 2 small joint; joint contracture of the fingers with finger to palm Pseudo-obstruction with symptoms such as vomiting or constipation, with 3 distance >1 cm; calcinosis on physical examination; proximal dilatation of the small and/or large bowel on imaging. muscle atrophy; digital pitting scar; modified Rodnan Skin Score 2 Low body mass index of <18.5 kg/m or weight loss of >10% in the last 12 2 >20, left ventricular systolic dysfunction on TTE alone, right months. ventricular dilation and/or enlargement on TTE, right ventric- Respiratory ular failure on clinical assessment and confirmed on TTE, ILD Moderate to severe interstitial lung disease >20% extent on HRCT of the 2 of any extent on imaging, symptoms of gastroparesis and malab- chest sorption syndrome (online supplementary tables S4 and S5). Add 4 points if forced vital capacity <70% on lung function tests (not due to respiratory muscle weakness).* 4 Persistent erectile dysfunction was also removed given its lack of applicability to all patients. Dependence on home oxygen. 5 Cardiovascular Pulmonary arterial hypertension (defined as mean pulmonary arterial 2 Items removed due to insufficient observations pressure >25 mm Hg at rest and pulmonary arterial wedge pressure <15 The following three items were removed as they had insuffi- mm Hg on right heart catheterisation) Add 5 if moderate to severe right ventricular dysfunction noted on cient observations to allow for meaningful statistical analysis: echocardiography report based on assessment of any measure of RV SRC with proteinuria, need for intravenous hyperalimentation function by experienced cardiologist. 5 and need for enteral feeding. Please refer to figure 1, which Myocardial disease attributable to SSc based on a constellation of clinical 3 summarises the item generation and reduction steps. features and supportive investigations, for example, syncope secondary to conduction abnormality, arrhythmia requiring defibrillator, heartblock requiring permanent pacemaker or ablation, systolic or diastolic dysfunction Statistical analysis to achieve further item reduction and item on TTE. weighting Presence of moderate to large pericardial effusion equivalent to greater than 1 The univariable relationship with mortality and morbidity 1 cm on TTE.* (dichotomised using the cohort median SF-36 PCS=37.2) was Renal significant (p<0.10) in 17 and 22 items, respectively (online Continued supplementary tables S4 and S5). These items were then weighted and scored using multivariable regression (table 1). Four items were removed at this stage as they did not reach statistical signif- icance in both the mortality and morbidity analyses (late-stage nailfold capillaroscopic changes, moderate to severe left ventric- ular diastolic dysfunction on TTE, any conduction defect on ECG or Holter monitor and faecal incontinence).

Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 811 Systemic sclerosis

Figure 2 (A–D) ROC curves for mortality (A, B) and morbidity (C, D) in the ASCS (derivation) and CSRG (validation) cohorts. ASCS, Australian Scleroderma Cohort Study; CSRG, Canadian Scleroderma Research Group; ROC, receiver operating characteristic.

Scleroderma Clinical Trials Consortium Damage Index of follow-up shows a steady increase (figure 3A). Steady accrual The 23-item multiorgan weighted SCTC-DI (SCTC-DI) is of the mean DI score was also demonstrated in the CSRG cohort presented in table 2. The scores are additive with a maximum at each year of follow-up (figure 3B). When total scores were possible total score of 55. In the final product, items addressing categorised into risk groups (low damage score <5, moderate the same issue, for example SRC, are grouped together and damage score 6–12, high damage score ≥13), Kaplan-Meier scores are added if further clinical markers of severity are present. survival estimates showed a statistically significant graded rela- Members of the working group reviewed the final SCTC-DI for tionship of damage score and time to death or first reduction face and content validity and potential feasibility of application (ie, worsening) of SF-36 PCS below the median score of 37.2 of each item. (log rank p<0.001) in the derivation cohort. Similar results were seen when the SCTC-DI was assessed in the CSRG validation Validation of the SCTC-DI cohort (figure 4A and B). Characteristics of the ASCS and CSRG cohorts The characteristics of the ASCS (derivation data set) and CSRG Discussion (validation data set) cohorts and the frequency of items together Through use of consensus methods with data-driven item reduc- with percentage missing within each cohort are presented in tion and weighting, an international working group of the SCTC online supplementary tables 6-9. has produced the first instrument developed specifically for the purpose of quantifying organ damage in SSc. The 23-item Performance of the SCTC-DI in the derivation and external cohorts SCTC-DI has been designed to capture damage that is reflec- The mean (SD), median, minimum and maximum DI scores tive of both morbidity and mortality in each of the six major within the ASCS derivation data set were 6.65 (4.8), 6, 0 and organ systems affected. Although it was considered that sepa- 33, respectively. Within the external cohort of CSRG patients, rate damage indices each reflective of mortality or morbidity be the mean (SD), median, minimum and maximum DI score were created, the working group felt that the creative approach of 6.9 (4.7), 6, 0, 33, respectively. combining the two analyses as a single index would be more ROC analysis revealed a good discriminatory capability of practical in the research and clinical care setting. Relative the SCTC-DI with ROC area under the curve of 0.77 (95% CI to other damage indices such as the Vasculitis Damage Index 0.73 to 0.82) and 0.73 (95% CI 0.69 to 0.77) for mortality in (VDI)68 and the Systemic Lupus International Collaborative the ASCS and CSRG cohorts, respectively (figure 2A and B). Clinics (SLICC)/ACR DI in Systemic Lupus Erythematosus,69 For morbidity (SF-36 PCS), ROC area under the curve was 0.69 with 65 and 39 items, respectively, the SCTC-DI of 23 items is (95% CI 0.65 to 0.72) and 0.72 (95% CI 0.69 to 0.74) in the shorter and would be easy to use in clinical trials. While the total ASCS and CSRG cohorts, respectively (figure 2C and D). possible damage score is 55, the maximum score in the deriva- Damage accrual over time within the derivation cohort was tion data set was 33, indicating that the SCTC-DI is not limited demonstrated in a subset of patients recruited within 2 years of by a ceiling effect and that damage scores exceeding 33 are likely disease onset, where the percentage of items present at each year incompatible with life. In the derivation data set, the SCTC-DI

812 Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 Systemic sclerosis

Figure 3 (A) and (B) Accrual of organ damage in an incident subset of cohort. Graphs demonstrating the accrual of damage within the incident Figure 4 (A) and (B) Survival according to Organ Damage Score. ASCS, population of the ASCS (A) and CSRG (B) cohorts, recruited within 2 Australian Scleroderma Cohort Study; CSRG, Canadian Scleroderma years of SSc diagnosis. In addition to overall damage accrual, accrual of Research Group; DI, Damage Index. damage is also separated into organ systems. *Disease duration from first non-Raynaud disease manifestation. ASCS, Australian Scleroderma the Combined Response Index in Systemic Sclerosis (CRISS).70–72 Cohort Study; CSRG, Canadian Scleroderma Research Group; SSc, As the name suggests, the EScSG activity index measures disease systemic sclerosis. activity, which inherently implies that it measures potentially reversible components of SSc and is not cumulative in nature. The CRISS on the other hand is a set of outcome measures used demonstrated good discriminatory capability for mortality risk to measure the effect of treatments in SSc. While both are valu- and also morbidity. able clinical trial tools, they do not capture the concept of cumu- We acknowledge that we did not follow the exact consensus lative, irreversible organ damage which invariably occurs in the methodology prescribed by EULAR to generate comparable natural history of SSc. Experience with both the VDI and the indices, and this may be considered a limitation. However, the SLICC/ACR DI as endpoints in clinical trials also demonstrates international collaborative approach of the working group, the the potential applications of a DI in SSc. survey of a large number of SSc experts outside of the working The value of the patient perspective in the development of group, contributions of experts in non-rheumatology specialties disease indices has become increasingly recognised.73 Although and patient partners, together with robust statistical methods the conceptual definition of damage in SSc stipulated that used to reduce, weight and score the items make the develop- patient-reported outcomes not be included in the DI, the patient ment of the SCTC-DI rigorous and unique. While the SCTC perspective was still taken into account through inclusion of survey response rate was 28%, we received 93 complete survey patient partners who were consulted at all steps of instrument responses from international experts to work with. The devel- development. Furthermore, the use of the HRQoL measure SF-36 opment of the SCTC-DI also benefited from a recent mortality as an endpoint for weighting the scores allowed for a patient-re- study that we undertook using both ASIG and CSRG data where ported outcome to be included in instrument development. we traced lost to follow-up and confirmed all deaths.2 Our conceptual definition of ‘damage’ in SSc has inherent Existing multiorgan outcome measures in SSc include the limitations. However, this could be said of all complex multi- European Scleroderma Study Group (EScSG) activity index and organ connective tissue diseases where assignment of direct

Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 813 Systemic sclerosis attribution to the disease and irreversibility is based on physician Rischmueller, Adelaide, South Australia; Janet Roddy, Perth, Western Australia; judgement. Some items in the DI are required to be present for Joanne Sahhar, Melbourne, Victoria; Wendy Stevens, Melbourne, Victoria; Gemma Strickland, Geelong, Victoria; Jenny Walker, Adelaide, South Australia; Peter Youssef, a minimum of 6 months to ensure they are ‘irreversible’. The Sydney, New South Wales. Canadian Scleroderma Research Group (CSRG) members: future availability of truly disease modifying drugs may mean J. Pope, London, Ontario; M. Baron, Montreal, Quebec; J. Markland, Saskatoon, that some items in the DI may indeed be made reversible. In such Saskatchewan (deceased); D. Robinson, Winnipeg, Manitoba; N. Jones, Edmonton, a situation, an updated revision of the DI will be needed. The use Alberta; N. Khalidi, Hamilton, Ontario; P. Docherty, Moncton, New Brunswick; E. of an existing prospectively acquired clinical database for statis- Kaminska, Calgary, Alberta; A. Masetto, Sherbrooke, Quebec; E. Sutton, Halifax, Nova Scotia; J-P. Mathieu, Montreal, Quebec; M. Hudson, Montreal, Quebec; S. Ligier, tical analyses meant that a handful of items were not present or Montreal, Quebec; T. Grodzicky, Montreal, Quebec; S. LeClercq, Calgary, Alberta; C. did not have a definition identical to that presented in the survey Thorne, Newmarket, Ontario; G. Gyger, Montreal, Quebec; D. Smith, Ottawa, Ontario; of experts. However, where this was the case, a surrogate defini- P.R. Fortin, Quebec, Quebec; M. Larché, Hamilton, Ontario; M. Abu-Hakima, Calgary; tion, which was as close as possible to the original item, was used. TS Rodriguez-Reyna, Mexico City, Mexico; AR Cabral, Mexico City, Mexico; M. Fritzler, Despite this limitation, both internal and external validity were Mitogen Advanced Diagnostics Laboratory, Cumming School of Medicine, Calgary, Alberta. demonstrated against the endpoints of mortality and morbidity, as was responsiveness to change over time. To complete valida- Contributors NF: study design, data collection, data analysis, interpretation of results, preparation of manuscript; MoH (biostatistician): data analysis, interpretation tion of the SCTC-DI, future studies need to evaluate reliability/ of results, preparation of manuscript; WS: study design, data collection, interpretation reproducibility and feasibility of application of the SCTC-DI. of results, preparation of manuscript; MaH: study design, data collection, data The use of retrospective data in a prevalent cohort underscores analysis, interpretation of results, preparation of manuscript; TT: study design, the need for further validation of the SCTC-DI in a prospec- data collection, interpretation of results, preparation of manuscript; JLB: study design, data analysis, interpretation of results, preparation of manuscript; SM: tive data set where all items in the DI, together with those that study design, data analysis, interpretation of results, preparation of manuscript; did not make the final index but were deemed appropriate by CR: data collection, interpretation of results, preparation of manuscript; VS: study experts surveyed, are collected. In addition, the early mortality74 design, data analysis, interpretation of results, preparation of manuscript; DP: study and early accrual of damage in SSc has provided a rationale design, interpretation of results, preparation of manuscript; SMP: study design, data for the creation of an international inception cohort (Interna- collection, interpretation of results, preparation of manuscript; MB: study design, data collection, data analysis, interpretation of results, preparation of manuscript; tional Systemic Sclerosis Inception Cohort) of patients with SSc MN: study design, data collection, data analysis, interpretation of results, preparation in order to prospectively validate disease outcome measures, of manuscript. including the SCTC-DI. Funding The development of the SCTC-DI was supported by SCTC Working Group grants. The Australian Scleroderma Cohort Study is supported by Scleroderma Author affiliations Australia, Arthritis Australia, Actelion Australia, Bayer, CSL Biotherapies, 1Department of Medicine, University of Melbourne at St. Vincent Hospital, Melbourne, GlaxoSmithKline Australia and Pfizer. A/Prof Nikpour holds an NHMRC Fellowship Fitzroy, Victoria, Australia (APP1126370). The Canadian Scleroderma Research Group (CSRG) is funded by the 2Rheumatology, St. Vincent Hospital, Melbourne, Fitzroy, Victoria, Australia Canadian Institutes of Health Research (CIHR) (grant #FRN 83518), the Scleroderma 3Medicine, University of McGill, Montreal, Quebec, Canada Society of Canada and its provincial Chapters, Scleroderma Society of Ontario, 4Lady Davis Institute for Medical Research, Montreal, Quebec, Canada Scleroderma Society of Saskatchewan, Sclérodermie Québec, Cure Scleroderma 5Public Health, LaTrobe University, Melbourne, Victoria, Australia Foundation, INOVA Diagnostics (San Diego, CA), Fooke Laboratorien (Neuss, 6Rheumatology, Royal Adelaide Hospital, Adelaide, Victoria, Australia Germany), Euroimmun (Lubeck, Germany), Mikrogen (Neuried, Germany), Fonds 7Rheumatology, Jewish General Hospital, Montreal, Quebec, Canada de la rechercheen santé du Québec (FRSQ), the Canadian Arthritis Network (CAN) Acknowledgements We thank Kerry McKenna for assistance with the survey and the Lady Davis Institute of Medical Research of the Jewish General Hospital, of experts. In addition to members of the SCTC Damage Index Working Group, Montreal, QC. The CSRG has also received educational grants from Pfizer and we thank all members of the SCTC, EUSTAR, CSRG and ASIG who took part Actelion pharmaceuticals. in our survey. We thank the following non-rheumatology expert advisors for Competing interests None declared. their input: DP (cardiologist), St Vincent’s Hospital, Melbourne, Australia; David Celermajer (cardiologist), University of Sydney, Australia; Professor Robyn Langham Patient consent for publication Not required (nephrologist), Monash University, Melbourne, Australia; Eli Gabbay (respiratory Ethics approval Ethics approval for the survey of international experts was physician), University of Western Australia, Australia; Trevor Williams (respiratory obtained from the Human Research Ethics Committee of St. Vincent’s Hospital physician), Alfred Health, Australia; Michael Kamm (gastroenterologist), University Melbourne. Ethics approval for the ASCS was obtained from the Human Research of Melbourne, Australia and Peter De Cruz (gastroenterologist), Austin Health, Ethics Committees of each of the participating centres, led by St. Vincent’s Hospital Melbourne, Australia. We thank the following patient partners for their input: Kim Melbourne. Ethics approval for the CSRG registry was obtained from the Human Fligestone and Maureen Sauve. We also thank our third patient partner who wishes Research Ethics Committees of each of the participating centres, led by the Jewish to remain anonymous. General Hospital, Montreal. Collaborators Scleroderma Clinical Trials Consortium Damage Index (SCTC-DI) Provenance and peer review Not commissioned; externally peer reviewed. Working Group members: Mandana Nikpour (Co-chair), The University of Melbourne Data sharing statement All data that can be made available have been included at St. Vincent’s Hospital Melbourne, Australia; Murray Baron (Co-Chair), Lady Davis in online supplementary files. Institute for Medical Research and Jewish General Hospital Montreal, Canada; Nava Ferdowsi, The University of Melbourne at St. Vincent’s Hospital Melbourne, Australia; Tracy Frech, University of Utah, USA; Marie Hudson, Lady Davis Institute for Medical References Research and Jewish General Hospital Montreal, Canada; Susanna Proudman, 1 Denton CP. Systemic sclerosis: from pathogenesis to targeted therapy. Clin Exp The University of Adelaide at Royal Adelaide Hospital, Australia; Wendy Stevens, Rheumatol 2015;33(4 Suppl 92):S3–7. St Vincent’s Hospital, Melbourne, Australia; Tien Tay, The University of Melbourne, 2 Hao Y, Hudson M, Baron M, et al. Early mortality in a multinational systemic sclerosis Australia; Souyma Chatterjee, Cleveland Clinic, Ohio, USA; Lorinda Chung, Stanford inception cohort. Arthritis Rheumatol 2017;69:1067–77. University, USA; Jessica K Gordon, Hospital for Special Surgery New York, USA; Anna 3 Tay T, Ferdowsi N, Baron M, et al. On behalf of the damage index Working Group of Haemel, University of California San Francisco, USA; Sindhu R Johnson, University of the scleroderma clinical Trials Consortium. measures of disease status in systemic Toronto, Canada; Dinesh Khanna, University of Michigan, USA; Thomas A Medsger, sclerosis: a systematic review. Semin Arthritis Rheum 2016;S0049-0172:30125–1. University of Pittsburgh, USA; Peter Merkel, University of Pennsylvania, USA; John 4 van den Hoogen F, Khanna D, Fransen J, et al. 2013 classification criteria for Pauling, Royal National Hospital for Rheumatic Diseases, Bath, UK; Janet E Pope, systemic sclerosis: an American College of Rheumatology/European League against University of Ontario, Canada; Tatiana Rodriguez-Reyna, National Institute of rheumatism collaborative initiative. Arthritis Rheum 2013;65:2737–47. Medical Sciences, Mexico; Lesley Saketkoo, Louisiana State University, USA; James 5 allanore Y, Meune C, Vonk MC, et al. Prevalence and factors associated with left R Seibold, Scleroderma Research Consultants, USA; Ankoor Shah, Duke University, ventricular dysfunction in the EULAR scleroderma trial and research Group (EUSTAR) USA; Virginia Steen, Georgetown University, USA; Gemma Strickland, Barwon Health, database of patients with systemic sclerosis. Ann Rheum Dis 2010;69:218–21. Australia. Australian Scleroderma Interest Group (ASIG) members: Catherine Hill, 6 Bernardo P, Conforti ML, Bellando-Randone S, et al. Implantable cardioverter Adelaide, South Australia; Gene-Siew Ngian, Melbourne, Victoria; Mandana Nikpour, defibrillator prevents sudden cardiac death in systemic sclerosis. J Rheumatol Melbourne, Victoria; Susanna Proudman, Adelaide, South Australia; Maureen 2011;38:1617–21.

814 Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 Systemic sclerosis

7 au K, Singh MK, Bodukam V, et al. Atherosclerosis in systemic sclerosis: a systematic 33 Busquets J, Del Galdo F, Kissin EY, et al. Assessment of tissue fibrosis in skin review and meta-analysis. Arthritis Rheum 2011;63:2078–90. biopsies from patients with systemic sclerosis employing confocal laser scanning 8 Clements PJ, Lachenbruch PA, Furst DE, et al. Cardiac score. a semiquantitative microscopy: an objective outcome measure for clinical trials? Rheumatology measure of cardiac involvement that improves prediction of prognosis in systemic 2010;49:1069–75. sclerosis. Arthritis Rheum 1991;34:1371–80. 34 Clements PJ, Furst DE, Wong WK, et al. High-dose versus low-dose D-penicillamine in 9 Clements PJ, Furst DE, Cabeen W, et al. The relationship arrhythmias and conduction early diffuse systemic sclerosis. Arthritis Rheum 1999;42:1194–203. disturbances to other manifestations of cardiopulmonary disease in progressive 35 Clements P, Lachenbruch P, Siebold J, et al. Inter and intraobserver variability of systemic sclerosis (PSS). Am J Med 1981;71:38–46. total skin thickness score (modified Rodnan TSS) in systemic sclerosis. J Rheumatol 10 Follansbee WP, Curtiss EI, Rahko PS, et al. The electrocardiogram in systemic sclerosis 1995;22:1281–5. (scleroderma). study of 102 consecutive cases with functional correlations and review 36 Clements PJ, Allanore Y, Khanna D, et al. Arthritis in systemic sclerosis: systematic of the literature. Am J Med 1985;79:183–92. review of the literature and suggestions for the performance of future clinical trials in 11 Hinchcliff M, Desai CS, Varga J, et al. Prevalence, prognosis, and factors associated systemic sclerosis arthritis. Semin Arthritis Rheum 2012;41:801–14. with left ventricular diastolic dysfunction in systemic sclerosis. Clin Exp Rheumatol 37 Johnstone EM, Hutchinson CE, Vail A, et al. Acro-osteolysis in systemic sclerosis 2012;30(2 Suppl 71):S30–7. is associated with digital ischaemia and severe calcinosis. Rheumatology 12 nagueh SF. Echocardiographic assessment of left ventricular relaxation and cardiac 2012;51:2234–8. filling pressures. Curr Heart Fail Rep 2009;6:154–9. 38 la Montagna G, Sodano A, Capurro V, et al. The arthropathy of systemic sclerosis: a 13 lang RM, Bierig M, Devereux RB, et al. Recommendations for chamber quantification: 12 month prospective clinical and imaging study. Skeletal Radiol 2005;34:35–41. a report from the American Society of echocardiography’s guidelines and standards 39 la Montagna G, Baruffo A, Tirri R, et al. Foot involvement in systemic sclerosis: a Committee and the chamber quantification writing group, developed in conjunction longitudinal study of 100 patients. Semin Arthritis Rheum 2002;31:248–55. with the European association of echocardiography, a branch of the European Society 40 Medsger TA Jr, Bombardieri S, Czirjak L, et al. Assessment of disease severity and of cardiology. J Am Soc Echocardiogr 2005;18:1440–63. prognosis.. Clin Exp Rheumatol 2003;21(Suppl. 29):S42–6. 14 Rudski LG, Lai WW, Afilalo J, et al. Guidelines for the echocardiographic assessment 41 Medsger TA. Natural history of systemic sclerosis and the assessment of disease of the right heart in adults: a report from the American Society of echocardiography. J activity, severity, functional status, and psychologic well-being. Rheum Dis Clin North Am Soc Echocardiogr 2010;23:685–713. Am 2003;29:255–73. 15 lubitz SA, Goldbarg SH, Mehta D. Sudden cardiac death in infiltrative 42 Ranque B, Authier F-J, Le-Guern V, et al. A descriptive and prognostic study of systemic cardiomyopathies: sarcoidosis, scleroderma, amyloidosis, hemachromatosis. Prog sclerosis-associated myopathies. Ann Rheum Dis 2009;68:1474–7. Cardiovasc Dis 2008;51:58–73. 43 Torok KS, Baker NA, Lucas M, et al. Reliability and validity of the delta finger-to- 16 Man A, Zhu Y, Zhang Y, et al. The risk of cardiovascular disease in systemic sclerosis: a palm, a new measure of the finger range of motion in systemic sclerosis. Clin Exp population-based cohort study. Ann Rheum Dis 2013;72:1188–93. Rheumatol 2010;28(2 Suppl 58):S28–S36. 17 Mathai SC, Sibley CT, Forfia PR, et al. Tricuspid annular plane systolic excursion 44 au K, Mayes MD, Maranian P, et al. Course of dermal ulcers and musculoskeletal is a robust outcome measure in systemic sclerosis-associated pulmonary arterial involvement in systemic sclerosis patients in the scleroderma lung study. Arthritis Care hypertension. J Rheumatol 2011;38:2410–8. Res 2010;62:1772–8. 18 Mok MY, Chiu SSH, Lo Y, et al. Coronary atherosclerosis using computed tomography 45 Penn H, Howie AJ, Kingdon EJ, et al. Scleroderma renal crisis: patient characteristics coronary angiography in patients with systemic sclerosis. Scand J Rheumatol and long-term outcomes. QJM 2007;100:485–94. 2009;38:381–5. 46 Hudson M, Baron M, Tatibouet S, et al. Exposure to ACE inhibitors prior to the onset 19 ngian G-S, Sahhar J, Proudman SM, et al. Prevalence of coronary heart disease and of scleroderma renal crisis-results from the International scleroderma renal crisis cardiovascular risk factors in a national cross-sectional cohort study of systemic survey. Semin Arthritis Rheum 2014;43:666–72. sclerosis. Ann Rheum Dis 2012;71:1980–3. 47 Batal I, Domsic RT, Medsger TA, et al. Scleroderma renal crisis: a pathology 20 O’Gara PT, Kushner FG, Ascheim DD, et al. ACCF/AHA guideline for the management perspective. Int J Rheumatol 2010;2010:1–7. of ST-elevation myocardial infarction: a report of the American College of cardiology 48 siva B, McDonald SP, Hawley CM, et al. End-stage kidney disease due to scleroderma- Foundation/American Heart Association Task Force on practice guidelines. Circulation -outcomes in 127 consecutive ANZDATA registry cases. Nephrol Dial Transplant 2013;2013:e362–425. 2011;26:3165–71. 21 Roberts NKet al. The prevalence of conduction defects and cardiac arrhythmias in 49 steen VD, Medsger TA. Long-term outcomes of scleroderma renal crisis. Ann Intern progressive systemic sclerosis. Ann Intern Med 1981;94:38–40. Med 2000;133:600–3. 22 seferović PM, Ristić AD, Maksimović R, et al. Cardiac arrhythmias and conduction 50 Mathew T. On behalf of the Editorial Advisory Committee. In: Chronic Kidney Disease disturbances in autoimmune rheumatic diseases. Rheumatology 2006;45(Suppl (CKD). Kidney Health Australia. 3rd Edition, 2012. 4):iv39–42. 51 nsl G, Desai SR, Veeraraghavan S, et al. Interstitial lung disease in systemic sclerosis. 23 levine GN, Bates ER, Blankenship JC, et al. 2011 ACCF/AHA/SCAI guideline Am J Respir Crit Care Med 2008;177:1248–54. for percutaneous coronary intervention: a report of the American College of 52 De Cruz S, Ross D. Lung transplantation in patients with scleroderma. Curr Opin cardiology Foundation/American Heart Association Task Force on practice guidelines Rheumatol 2013;25:714–8. and the Society for cardiovascular angiography and interventions. Circulation 53 Khanna D, Tseng C-H, Farmani N, et al. Clinical course of lung physiology in patients 2011;124:e574–651. with scleroderma and interstitial lung disease: analysis of the scleroderma lung study 24 McLaughlin VV, Archer SL, Badesch DB, et al. ACCF/AHA 2009 expert consensus placebo group. Arthritis Rheum 2011;63:3078–85. document on pulmonary hypertension: a report of the American College of cardiology 54 Moore OA, Goh N, Corte T, et al. Extent of disease on high-resolution computed Foundation Task Force on expert consensus documents and the American Heart tomography lung is a predictor of decline and mortality in systemic sclerosis-related Association: developed in collaboration with the American College of chest physicians, interstitial lung disease. Rheumatology 2013;52:155–60. American Thoracic Society, Inc., and the pulmonary hypertension association. 55 sottile PD, Iturbe D, Katsumoto TR, et al. Outcomes in systemic Sclerosis–Related lung Circulation 2009;119:2250–94. disease after lung transplantation. Transplantation 2013;95:975–80. 25 Forbes A, Marie I. Gastrointestinal complications: the most frequent internal 56 steen VD, Graham G, Conte C, et al. Isolated diffusing capacity reduction in systemic complications of systemic sclerosis. Rheumatology 2009;48(Suppl 3):iii36–9. sclerosis. Arthritis Rheum 1992;35:765–70. 26 Khanna D, Nagaraja V, Gladue H, et al. Measuring response in the gastrointestinal 57 Wells AU, Behr J, Silver R. Outcome measures in the lung. Rheumatology tract in systemic sclerosis. Curr Opin Rheumatol 2013;25:700–6. 2008;47(Suppl 5):v48–50. 27 Khanna D, Hays RD, Maranian P, et al. Reliability and validity of the University of 58 assayag D, Kaduri S, Hudson M. High resolution computed tomography scoring California, Los Angeles scleroderma clinical trial Consortium gastrointestinal tract systems for evaluating interstitial lung disease in systemic sclerosis patients. instrument. Arthritis Rheum 2009;61:1257–63. Rheumatology 2012;01. 10.4172/2161-1149.S1-003. 28 sjogren RW. Gastrointestinal motility disorders in scleroderma. Arthritis Rheum 59 Foocharoen C, Tyndall A, Hachulla E, et al. Erectile dysfunction is frequent in systemic 1994;37:1265–82. sclerosis and associated with severe disease: a study of the EULAR scleroderma trial 29 Baron M, Bernier P, Cote LF, et al. Screening and management for malnutrition and and research Group. Arthritis Res Ther 2012;14. related gastro-intestinal disorders in systemic sclerosis: recommendations of a North 60 Caramaschi P, Biasi D, Caimmi C, et al. Digital amputation in systemic sclerosis: American expert panel. Clin Exp Rheumatol 2010;28(Suppl.58):S42–6. prevalence and clinical associations. A retrospective longitudinal study. J Rheumatol 30 avouac J, Walker U, Tyndall A, et al. Characteristics of joint involvement and 2012;39:1648–53. relationships with systemic inflammation in systemic sclerosis: results from the 61 Herrick AL, Oogarah PK, Freemont AJ, et al. Vasculitis in patients with systemic EULAR scleroderma trial and research Group (EUSTAR) database. J Rheumatol sclerosis and severe digital ischaemia requiring amputation. Ann Rheum Dis 2010;37:1488–501. 1994;53:323–6. 31 avouac J, Clements PJ, Khanna D, et al. Articular involvement in systemic sclerosis. 62 Medsger TA, Silman AJ, Steen VD, et al. A disease severity Scale for systemic sclerosis: Rheumatology 2012;51:1347–56. development and testing. J Rheumatol 1999;26:2159–67. 32 avouac J, Guerini H, Wipff J, et al. Radiological hand involvement in systemic sclerosis. 63 Masi AT, Rodnan GP, Medsger TA, et al. Preliminary criteria for the classification Ann Rheum Dis 2006;65:1088–92. of systemic sclerosis (scleroderma). Subcommittee for scleroderma criteria of the

Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 815 Systemic sclerosis

American rheumatism association diagnostic and therapeutic criteria Committee. 70 Gladman D, Ginzler E, Goldsmith C, et al. The development and initial validation of the Arthritis Rheum 1980;23:581–90. systemic lupus international collaborating Clinics/American College of rheumatology 64 nihtyanova SI, Brough GM, Black CM, et al. Clinical burden of digital vasculopathy in damage index for systemic lupus erythematosus. Arthritis Rheum 1996;39:363–9. limited and diffuse cutaneous systemic sclerosis. Ann Rheum Dis 2008;67:120–3. 71 Valentini G, Della Rossa A, Bombardieri S, et al. European Multicentre study to 65 smith V, Riccieri V, Pizzorni C, et al. Nailfold capillaroscopy for prediction of define disease activity criteria for systemic sclerosis. II. Identification of disease novel future severe organ involvement in systemic sclerosis. J Rheumatol activity variables and development of preliminary activity indexes. Ann Rheum Dis 2013;40:2023–8. 2001;60:592–8. 72 Valentini G, Iudici M, Walker UA, et al. The European scleroderma trials and research 66 steen V, Denton CP, Pope JE, et al. Digital ulcers: overt vascular disease in systemic Group (EUSTAR) Task Force for the development of revised activity criteria for systemic sclerosis. Rheumatology 2006;48(Suppl 3):iii19–24. sclerosis: derivation and validation of a preliminarily revised EUSTAR activity index. 67 amanzi L, Braschi F, Fiori G, et al. Digital ulcers in scleroderma: staging, characteristics Ann Rheum Dis 2017;76:270–6. and sub-setting through observation of 1614 digital lesions. Rheumatology 73 Khanna D, Berrocal VJ, Giannini EH, et al. The American College of rheumatology 2010;49:1374–82. provisional composite response index for clinical trials in early diffuse cutaneous 68 Ware JE, Kosinski M, Keller SD. SF-36 Physical and Mental Health Survey Summary systemic sclerosis. Arthritis Rheumatol 2016;68:299–311. Scales: A User’s manual. Boston, MA: The Health Institute, 1994. 74 saketkoo LA, Mittoo S, Frankel S, et al. Reconciling healthcare professional and 69 exley AR, Bacon PA, Luqmani RA, et al. Development and initial validation of the patient perspectives in the development of disease activity and response criteria vasculitis damage index for the standardized clinical assessment of damage in the in connective tissue disease-related interstitial lung diseases. J Rheumatol systemic vasculitides. Arthritis Rheum 1997;40:371–80. 2014;41:792–8.

816 Ferdowsi N, et al. Ann Rheum Dis 2019;78:807–816. doi:10.1136/annrheumdis-2018-214764 Systemic sclerosis

Translational science Compendium of skin molecular signatures identifies key pathological features associated with fibrosis in systemic sclerosis Su-Jin Moon,1 Jung Min Bae,2 Kyung-Su Park,3 Ilias Tagkopoulos,4 Ki-Jo Kim‍ ‍ 5

Handling editor Josef S ABSTRACT Key messages Smolen Objectives Treatment of patients with systemic sclerosis (SSc) can be challenging because of ►► Additional material is What is already known about this subject? published online only. To view clinical heterogeneity. Integration of genome-scale ►► Systemic sclerosis (SSc) is a clinically please visit the journal online transcriptomic profiling for patients withSS c can provide heterogeneous disease with a variable (http://dx.​ ​doi.org/​ ​10.​1136/​ insights on patient categorisation and novel drug targets. molecular signature and clinical course. annrheumdis-2018-​ ​214778). Methods a normalised compendium was created from ► The resolution of skin fibrosis remains an unmet 344 skin samples of 173 patients with SSc, covering ► For numbered affiliations see clinical need in patients with SSc. end of article. an intersection of 17 424 genes from eight data sets. Differentially expressed genes (DEGs) identified by three What does this study add? Correspondence to independent methods were subjected to functional ► Using unsupervised machine learning, a Dr Ki-Jo Kim, Division of network analysis, where samples were grouped ► pathway-driven matrix of skin gene expression Rheumatology, Department of using non-negative matrix factorisation. Finally, we Internal Medicine, St. Vincent’s profiles allows subdivision of SSc into four investigated the pathways and biomarkers associated Hospital, College of Medicine, subtypes with distinct activities of SSc-relevant with skin fibrosis using gene-set enrichment analysis. The Catholic University of Korea, pathways. 93, Jungbu-daero, Paldal-gu, Results We identified 1089 upregulated DEGs, ► The inflammatory subtype was related to Suwon-si, Gyeonggi-do, 16247, including 14 known genetic risk factors and five potential ► Republic of Korea; significant improvement in skin fibrosis drug targets. Pathway-based subgrouping revealed four md21c@​ ​catholic.ac.​ ​kr at follow-up. Increased activity of the distinct clusters of patients with SSc with distinct activity phosphoinositide-3-kinase-protein kinase B Received 16 November 2018 signatures for SSc-relevant pathways. The inflammatory (PI3K-Akt) pathway was closely correlated with Revised 26 February 2019 subtype was related to significant improvement in skin skin fibrosis in all patients and chronologically Accepted 23 March 2019 fibrosis at follow-up. The phosphoinositide-3-kinase- Published Online First and was driven by COMP, FN1, TNC, THBS1 and protein kinase B (PI3K-Akt) signalling pathway showed 5 April 2019 THBS4. both the closest correlation and temporal pattern to skin fibrosis score. COMP, THBS1, THBS4, FN1, and TNC were How might this impact on clinical practice or leading-edge genes of the PI3K-Akt pathway in skin future developments? fibrogenesis. ► Pathway-based clustering and integrative Conclusions construction and analysis of normalised ► analysis offer opportunities to understand skin transcriptomic compendia can provide useful the divergent mechanistic features of SSc and insights on pathway involvement by SSc subsets and shed new light on the PI3K-Akt pathway as a discovering viable biomarkers for a skin fibrosis index. promising biomarker or therapeutic target. Particularly, the PI3K-Akt pathway and its leading players are promising therapeutic targets.

measures, varying natural course of the disease and slow progression of the disease.1 3 As such, there Introduction are substantial gaps in our knowledge regarding the Systemic sclerosis (SSc) is a chronic autoimmune mechanistic basis of the SSc progression and prog- disease that is characterised by the distinctive patho- nosis stratification. genetic triad of microvascular damage, dysregulated In SSc, gene expression profiling has been used autoimmunity and generalised fibrosis of the skin to gain insights regarding the pathogenesis of the and multiple internal organs. SSc usually results in disease.4–6 Because all previous studies have been permanent functional impairment and is associated performed with small cohorts with varying patient with high morbidity and mortality.1 2 Studies using histories, technical protocols and different tech- tissue samples, diverse and sophisticated animal nologies, a direct comparison between data sets © Author(s) (or their employer(s)) 2019. No models and genome-wide analyses are providing and results is not possible. Therefore, we have commercial re-use. See rights insights regarding the pathogenesis of this disease. constructed a compendium that reduces data set and permissions. Published Current therapeutic approaches include treatments bias and provides opportunities to analyse features by BMJ. against vasculopathy, general immunosuppression missing from previous studies, bridging the gap To cite: Moon S-J, and affected organ-specific therapies, although between prior and current knowledge—a technique Bae JM, Park K-S, no single therapy has been proven to effectively that has been successfully applied in cancer and et al. Ann Rheum Dis reverse the fibrotic process in SSc.1 2 This is mainly infectious disease research.7–9 In this study, we aim to 2019;78:817–825. because of a lack of sensitive and specific outcome elucidate the various transcriptional and signalling

Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 817 Systemic sclerosis

final SSc compendium was constructed from eight studies with a total of 423 samples from 236 patients and covering 23 684 genes total (common core of 17 424 genes). After normalising the vectors for the matrix using quantile normalisation and correcting the batch effect, we performed differential expres- sion analysis and selected the gene candidates identified by three independent methods. We then performed functional gene-set enrichment analysis (GSEA) and protein–protein interaction network analysis, with subsequent grouping of the samples based on non-negative matrix factorisation (NMF). Finally, we inves- tigated the pathways and the biomarkers associated with skin fibrosis using GSEA. Detailed methods are described in supple- mentary methods.

Results DEGs in each subgroup of SSc To get a list of SSc-related DEGs, gene expression profiles of the involved forearm area of patients with SSc were compared with forearm samples from the NC groups. A total of 1089 upregulated DEGs were identified in dSSc and 192 in lSSc, respectively (supplementary file S1). Of these, 191 DEGs were shared by dSSc and lSSc. When compared with uninvolved back areas of patients with SSc and same-site samples from the NC groups, 31 upregulated DEGs were identified in dSSc and five in lSSc, respectively (supplementary file S1). A single DEG (IFI27) was shared by dSSc and lSSc. Enrichment of the type I interferon (IFN) signalling pathway (p=3.67×106) was observed in 31 DEGs of dSSc by performing functional enrich- ment analysis. The combined clinical characteristics of the patients with dSSc were summarised in supplementary figure S2B. In subsequent analyses, we focused the data of dSSc because there were more enriched gene expression signatures Figure 1 Overview of data processing steps. (A) A total of 20 studies with more samples. covering 23 684 genes were retrieved from the literature. (B) Eight data sets were selected for integrated analysis, including samples from 175 patients with SSc and 61 NC, covering 17 424 genes. (C) The merged Protein-to-protein interaction network and enriched data set was normalised using quantile normalisation, and its batch biological processes effect was further corrected. (D) DEGs of SSc compared with NC were Because identification of central attractors in the gene and obtained using three methods: eBayes, SAM and RP. The intersection of protein network can provide clues about novel disease-asso- three DEG sets was designated as significant. (E) Calculation of pathway ciated genes with high priority and hidden targets for further activation scores using the whole gene expression profiles. (F) A list of experimentation, we constructed a protein-to-protein interac- strategies for integrated analysis. DEG, differentially expressed genes; tion network of dSSc (figure 2A). We identified 1068 interac- eBayes, empirical Bayes; NC, normal controls; RP, rank products; SAM, tions of the 1089 DEGs, and 620 genes were isolated without significance analysis of microarray; SSc, systemic sclerosis. a link. Inter-relationships between genetic susceptibility, poten- tial drug targets and hub positions on the network were shown signatures of SSc by performing a comprehensive meta-analysis schematically in supplementary figure S4. Of these, 14 DEGs of the publicly available data sets that have been published so far. overlapped with SSc genetic susceptibility loci previously We have applied several preprocessing and normalisation steps discovered10–13 and five genes (CTGF, HCK, LYN, PDGFRB to create a cohesive, homogenised compendium of genome- and PPARG) were the potential target molecules for therapy.14 15 wide gene expression signatures for downstream analysis. We A total of 89 genes were ranked as hub molecules based on used this compendium to separate expression-driven subgroups, centrality analysis, and four target molecules, except PPARG, understand the key cellular components in each group and iden- have the hub position. Notably, CTGF was the unique gene tify genes and pathways with high information value to create with both molecular target potentiality and hub status among predictive models for drug responsiveness. the susceptibility genes.10 15–17 We performed GSEA for whole gene expression profiles using Methods the GSEA tool, from which 526 gene ontology processes were Overview of data processing and analysis identified (figure 2 and supplementary figure S5). As expected, We used the keywords ‘systemic sclerosis’, ‘skin’, ‘transcrip- immune related (cytokine and chemokine related), profibrotic tomics or microarray’ and ‘data set’ in Google Scholar and (extracellular matrix organisation, transforming growth factor PubMed to find relevant publications on genetic skin signatures beta (TGFβ) related, cell proliferation) and angiogenesis (vascular in patients with SSc (figure 1). The final number of patients with endothelial growth factor-related) processes were enriched and SSc was 175 (diffuse subset SSc (dSSc) 115, limited subset SSc adequately explained the current concept of dermato-patho- (lSSc) 60) and 61 healthy normal controls (NC). Ultimately, the physiology in SSc.

818 Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 Systemic sclerosis

Figure 2 Differentially expressed genes (DEGs) and their functional networks. (A) Protein–protein interaction network of upregulated DEG. Informative genes are coloured and guided at the right-side table. (B) Gene-set enrichment map for whole gene expression profiles. Nodes represent gene ontology (GO)-termed gene sets. Their colour intensity and size are proportional to the enrichment significance and the gene size, respectively. Edge thickness represents the degree of overlap between gene sets, and only edges with a Jaccard similarity coefficient larger than 0.25 were visualised. Conglomerates of functionally related gene-sets were manually curated based on the GO parent–child hierarchy and assigned a label.

Pathway activations describing SSc biology and subgrouping (PPAR) signalling pathway rarely shared with any one of them Genes, proteins and other chemical compounds in a living (Jaccard similarity coefficient <0.10) and exhibited the oppo- organism rarely act in isolation, but instead work coopera- site direction of activity against the other pathways (supple- tively to perform certain biological functions. In the same vein, mentary figure S7). disease is the summed result of aberrant activation of common Cluster 4 showed the strongest inclination towards extracel- pathways through dysregulated genes and aggregated activity lular matrix formation and was also highly active for TGFβ, of compounds.18 The advantage of pathway-based analysis has Apelin, PI3K-Akt and PDGF signalling pathways. The type I IFN been previously demonstrated in clinical stratification for cancer signalling pathway was more active in clusters 1 and 2. Cluster research.8 19 We curated 40 SSc-relevant pathways or processes 1 was more heavily weighted towards the Notch and Hedgehog from the literature20–24 and computed pathway activation scores signalling pathways, while cluster 2 showed a stronger predilec- using the gene sets from the KEGG and Reactome databases tion for the TGFβ signalling pathway. Cluster 3 was the weakest (supplementary file 2).25 26 We assessed whether patients with for most pathways, but was conspicuous for the PPAR signal- dSSc could be categorised into subgroups based on their pathway ling pathway. As matched with previous results by intrinsic gene activation profiles through consensus NMF clustering.27 To iden- subsets,5 6 clusters 1 and 2 and clusters 3 and 4 were conceptu- tify the optimal number of clusters and to assess the robustness of ally similar to inflammatory, normal like and fibroproliferative the clustering results, we computed the cophenetic coefficients types, respectively. and silhouette scores for different numbers of clusters from two To characterise the cell types responsible for gene expres- to six; we found that four clusters most optimally represent the sion differences among the skin tissue samples, we applied data (figure 3A and supplementary figure S6). Segregation of xCell software, the machine learning framework to estimate dSSc subgroups was also reproduced by t-distributed Stochastic cell type enrichment.28 Cluster 2 was the most enriched with Neighbor Embedding (t-SNE) analysis (figure 3B). Pathway immune cells such as B cells, T cells, dendritic cells and macro- elements matrix of four segregated subgroups were depicted phages (figure 4B). Fibroblast subsets predominated in cluster 4, by heatmap (figure 3C) and active pathways or processes were with strong activity for fibrosing signalling pathways (figures 3 distinct across four clusters. and 4A). Cluster 1 had a higher score for eosinophils and Th1 To validate the differences in activities of pathways among cells, while cluster 3 had a higher score for epithelial cells and the four clusters, we analysed the activation of individual path- keratinocytes. ways and filtered the 20 pathways of significant difference To validate the dSSc subgrouping scheme, a 20-pathway across the four clusters (figure 4A). The chord diagram in the classifier was developed using a naive Bayes machine learning centre shows the interconnectedness between the gene sets of algorithm. This classifier was applied to an independent dataset pathways or processes by the Jaccard similarity coefficient. consisting of 69 skin samples from 38 patients with dSSc In general, there was no remarkable link between cytokine (GSE106358).29 The 20-pathway classifier reliably categorised or chemokine pathways and processes of extracellular matrix dSSc samples in the GSE106358 into four subgroups, with an formation, but the platelet-derived growth factor (PDGF) average classification performance of 0.8372 (p=4.56×10−12) signalling pathway and collagen formation shared a propor- (supplementary figure S8). The four subgroups were identified tion of element genes (Jaccard similarity coefficient=0.144) and the proportions were comparable to the training data set and highly correlated (γ=0.928, p=2.2×10 -16) (supplemen- (supplementary figure S8), demonstrating proper functioning of tary figure S7). The peroxisome proliferator-activated receptor the classification model.

Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 819 Systemic sclerosis

Figure 3 Identification of systemic sclerosis (SSc) subgroups according to pathway-based skin signatures. (A) Reordered consensus matrices on SSc compendium. The samples were clustered using average linkage and 1-correlation distances. Deep-red colour indicates perfect agreement of the solution, while blue colour indicates no agreement (right-side colour bar). Basis and consensus represent clusters based on the basis and consensus matrices, respectively. The silhouette score is a similarity measure within its own cluster compared with other clusters. (B) t-distributed Stochastic Neighbor Embedding (t-SNE) reduces the dimensions of a multivariate dataset. Each data point is assigned a location in a three-dimensional map to illustrate potential clusters of neighbouring samples, which contain similar gene expression patterns. (C) Element matrix of pathways or processes by negative matrix factorisation. Heatmap with red gradient represents the level of fold enrichment for each pathway or processes. HIF, hypoxia- inducible factor; IL, interleukin; JAK, Janus kinase; MAPK, mitogen-activated protein kinase; mTOR,mammalian target of rapamycin; NOD-like, nucleotide-binding oligomerizationdomain-like; PDGF, platelet-derived growth factor; PI3K, phosphoinositide3-kinases; PPAR, peroxisome proliferator- activated receptors; RIG-I-like, retinoic acid-inducible gene-I-like; SEMA4D,semaphorin-4D; STAT, signal transducer and activator of transcription; TGF, transforming growth factor; VEGF, vascular endothelial growth factor.

Clinical implication of pathway-driven subgroups in dSSc pathways had the strongest correlation with mRSS (γ=0.553, We examined the relationships between identified four subgroups p=1.77×10 -10 and γ=0.501, p=1.39×10 -8, respectively), and pertinent clinical features based on the provided informa- and which were followed by mitogen-activated protein kinases tion. The highest correlation was age in cluster 4, and disease (MAPK) and TGFβ signalling pathways (γ=0.431, p=1.67×10 duration in cluster 1, but the range of values was comparable -6 and γ=0.418, p=3.55×10 -6, respectively). across the clusters (figure 5A). The modified Rodnan skin score The transcriptomic data set of GSE106358 includes 27 paired (mRSS) was significantly higher in cluster four compared with samples of baseline and 24-week follow-up patients in the place- clusters 1 and 3 (all p<0.05), but not compared with cluster 2 bo-treated arm from clinical trials, subcutaneous tocilizumab (p=0.31), which was a proxime accessit in the mRSS and TGFβ therapy in patients with dSSc.29 To investigate the correlation pathway activity (figure 4). Correlations between mRSS and between variation in mRSS and pathway enrichment score, SSc-relevant pathways were analysed, and significant correla- single sample GSEA was used to generate enrichment scores tions were plotted in figure 5B. PI3K-Akt and PDGF signalling for gene sets of 40 SSc-relevant pathways and the differentials

820 Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 Systemic sclerosis

Figure 4 (A) Pathway activation scores according to systemic sclerosis (SSc) subgroups. The chord diagram shows the interrelationship among pathways, and link thickness is proportional to the overlap between two pathways. These were calculated using the Jaccard similarity coefficients. Turkey box plots reveal pathway activation scores across the SSc subgroups, and analysis of variance (ANOVA) analysis was used to analyse the differences across four subgroups. (B) xCell-inferred enrichment score of cell types according to the four clusters. xCell scores predict relative enrichment for cell types, not the proportions. *P<0.05; **P<0.01; ***P<0.001 by ANOVA.

Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 821 Systemic sclerosis

Figure 5 (A) Levels of clinical variables by systemic sclerosis (SSc) subgroup. Comparison between the two subgroups was done by t-test. *P<0.05; **P<0.01; ***P<0.001. (B) Correlation between modified Rodnan skin score (mRSS) and pathway activation score. Correlation analysis was performed by Pearson’s method, and p value at the x-axis was log scaled for better visualisation. (C) Correlation of variations between mRSS and pathway enrichment scores. Correlation analyses for the 27 paired samples of baseline and 24-week follow-up by subcutaneous tocilizumab therapy in patients with diffuse subset SSc (GSE106358) were performed using Spearman’s method. ΔmRSS was positively correlated with Δenrichment score of phosphoinositide-3-kinase-protein kinase B (PI3K-Akt) (γ=0.407, p=0.034) and Janus kinase-signal transducers and activators of transcription (JAK-STAT) signalling pathways (γ=0.389, p=0.045) and dermatan sulfate biosynthesis (γ=0.466, p=0.014). (D) Evolution of the cluster (left panel) and change of mRSS by clusters (right panel) after 24 weeks.

(Δ) before and after treatment were calculated. The ΔmRSS was are five genes left except for the collagen family: COMP, THBS4, significantly correlated with Δenrichment score of PI3K-Akt THBS1, TNC and FN1. Expression values of these genes were signalling pathway (γ=0.407, p=0.034), Janus kinase-signal closely correlated with mRSS in patients with dSSc (all γ>0.4, transducers and activators of transcription (JAK-STAT) signalling p<0.001) (figure 6C). pathway (γ=0.389, p=0.045) and dermatan sulfate biosynthesis (γ=0.466, p=0.014) (figure 5C and supplementary figure S9). Discussion This implies that improvement in mRSS was accompanied by In the present study, we built the largest cohesive SSc tran- proportional mitigation of PI3K-Akt signalling, JAK-STAT signal- scriptomics compendium to understand differential expression ling and dermatan sulfate biosynthesis and vice versa. Next, we patterns and to identify key players, potential drug targets and compared the change in mRSS by cluster (figure 5D). The mRSS distinct patient clusters. Unbiased cluster analysis of the dSSc was significantly reduced in cluster 2 (p=0.040), while variable compendium resulted in meaningful categories of patients with responses were observed with no significant improvement in dSSc with distinct activities for relevant pathways. This path- mRSS in clusters 1 and 4 (p=0.677 and p=0.438, respectively). way-based analysis allowed refinement of our understanding of dSSc subgroups and allowed us to discover promising target Enrichment of PI3K-Akt signalling pathway and its leading- pathways and biomarkers that have a close correlation with the edge genes clinical index. In GSEA, the PI3K-Akt signalling pathway was identified as one To refine the SSc-unique genes, we compared SSc samples with of the most enriched gene sets (false discovery rate <0.001) in NC and adopted DEGs shared by three independent methods. dSSc (figure 6A). To identify which genes in the gene sets were We found that 14 of the DEGs are known SSc-associated genetic enriched in the PI3K-Akt signalling pathway, we focused on the loci and 89 DEGs occupy a central position in the molecular ‘leading edge’ of enrichment. Leading-edge genes in a GSEA skin network. Because functional implications of high-risk alleles are those that contribute most to the enrichment of a particular were often obscure, it was helpful to elucidate the biological gene set and include the most significantly upregulated genes in mechanisms in which risk alleles operate. CTGF, HCK, LYN and a given gene set.30 In total, 94 leading-edge genes were captured, PDGFRB were the targets, having hub positions in the network. and the top 10 genes are shown as a heatmap in figure 6B. There CTGF is a secreted matricellular protein contributing to tissue

822 Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 Systemic sclerosis

Figure 6 Enrichment of the phosphoinositide-3-kinase-protein kinase B (PI3K-Akt) signalling pathway and its leading-edge genes. (A) Enrichment plot generated by gene-set enrichment analysis (GSEA) analysis of ranked gene expression data (left, upregulated (red); right, downregulated (blue)) by the PI3K-Akt signalling pathway in diffuse subset systemic sclerosis (dSSc) versus normal controls (NC). (B) GSEA-derived heatmap of the top 10 leading-edge genes showing the strongest upregulation in dSSc versus NC based on a signal/noise ratio score. (C) Correlation between the expression values of the leading-edge genes and modified Rodnan skin score (mRSS) in patients with dSSc. Correlation analysis was performed using Pearson’s correlation as a similarity measure. remodelling and fibrosis in cooperation with TGFβ.31 32 Admin- clinical experiences of several tyrosine kinase inhibitors.15 We istration of anti-CTGF antibodies reduced or reversed fibrosis in also found enrichment of the type I IFN signature in unaffected animal models with various stages of fibrotic disease.31 32 HCK, skin areas of dSSc. This result corroborates systemic activation LYN and PDGFRB were all understood in terms of activation of the type I IFN signature, independent of overt fibrosis.35 36 of intracellular tyrosine kinases,15 which were proven to play We were able to identify four distinct subgroups through NMF key roles in the pathologic activation of fibroblasts during fibro- analysis of the dSSc compendium, and they differed in activation genesis.33 34 However, a discrepancy exists between the encour- levels of SSc-relevant signalling pathways. This was commen- aging preclinical results in animal models and the disappointing surate with the result by intrinsic gene subsets analysis,5 6 and

Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 823 Systemic sclerosis inflammatory types were further subdivided into two subgroups interruption of the PI3K-Akt signalling pathway is an ideal target (cluster 1 and 2) by the predominance of Notch/Hedgehog or for decelerating this feed-forward loop. TGFβ signalling pathways. Cluster 4 was most active in profi- There are some limitations to address in this study. First, brotic pathways (TGFβ, Apelin, PI3K-Akt and PDGF signalling removing batch effects is not ideal because it adds to the noise pathways) and extracellular matrix organisation and was graded in the compendium. Second, we did not address the association highest in mRSS on average. Given the short disease duration in of each SSc subgroup with other clinical factors, such as inter- cluster 2 (figure 5), cluster 2 was counted as the early inflamma- stitial lung disease, pulmonary hypertension and autoantibodies, tory phase and showed a better clinical outcome because it has because of lack of complete annotation for each SSc sample. a strong immunoinflammatory nature with reversibility, unlike Third, minority signatures by specific cell subsets might have irreversible fibrotic change. been diluted because the gene expression signature was at the However, it is not clear that these clusters are simply linked to tissue level. SSc is a major medical challenge with high unmet the stage of SSc clinically because the progression of skin fibrosis need, and no drug has been able to halt the fibrotic process or and thickening is variable.37 Disease duration varied widely across modify the clinical course. Efforts focusing on the application the clusters, although mean disease duration was the longest of advanced machine learning techniques, such as in this work, in cluster 1 and the shortest in cluster 2. This pathway-driven enable a better understanding of SSc at a systems level. clustering may provide a good starting point for understanding the divergent mechanistic features and clinical trajectory of Author affiliations 1 SSc. Future clinical trials can account for the categorisation by Division of Rheumatology, Department of Internal Medicine, Uijeongbu St Mary’s performing deep profiling on study participants and conducting Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea a differential analysis for prognosis. This approach can also be 2Department of Dermatology, St. Vincent’s Hospital, College of Medicine, The used in the clinical setting to determine whether certain groups Catholic University of Korea, Seoul, Republic of Korea of patients are more responsive to investigational drugs than 3Division of Rheumatology, Department of Internal Medicine, St Vincent’s Hospital, 38 College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea other types of patients. It would also be interesting to investi- 4 gate whether the distinct subgroups have different correlations Department of Computer Science & Genome Center, University of California, Davis, Davis, California, USA with occurrence or severity of other complications, such as inter- 5Division of Rheumatology, Department of Internal Medicine, St. Vincent’s Hospital, stitial lung disease and pulmonary arterial hypertension. College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea Dysregulation of multiple cellular and molecular components Acknowledgements We thank Katrina Krogh, MD, from Edanz Group (​www.​ and activation of profibrotic processes converge towards the edanzediting.​com/​ac) for editing a draft of this manuscript. progressive tissue accumulation of fibrous matrix, which is the Contributors K-JK and S-JM designed the study and carried out data collection. 1 23 distinguishing hallmark of SSc. Among the prevailing path- K-JK performed computational analysis and drafted the paper, but all authors were ways, the activity of PI3K-Akt pathway not only had the closest involved in critically revising its final preparation.A ll authors approved the final correlation with mRSS but also shifted in parallel with the version to be published. changes in mRSS. PI3K is a downstream molecule of the TGFβ Funding This work was supported by the National Research Foundation of Korea and PDGF signalling pathways,1 and PI3K inhibition was found (NRF) grant funded by the Korea government (MSIT) (No NRF-2018R1A2B6007291). to effectively reverse or alleviate organ fibrosis in vivo.39 40 Many Competing interests None declared. clinical trials have failed in the past for several reasons, including Patient consent for publication Not required. 15 41 choosing the wrong target pathways or molecules. In this Provenance and peer review Not commissioned; externally peer reviewed. context, inhibition of the PI3K-Akt pathway should be reap- praised as a promising antifibrotic therapy. A number of investi- References gational agents inhibiting the PI3K-Akt pathway in clinical trials 1 allanore Y, Simms R, Distler O, et al. Systemic sclerosis. Nat Rev Dis Primers 2015;1. for specific cancers can be adopted for SSc research.42 2 Denton CP, Khanna D. Systemic sclerosis. The Lancet 2017;390:1685–99. Not all members of a gene set participate in a specific biological 3 Medsger TA. Natural history of systemic sclerosis and the assessment of disease process at all levels; the core of a gene set accounts for the enrich- activity, severity, functional status, and psychologic well-being. Rheum Dis Clin North 30 Am 2003;29:255–73. ment signal. In this study, we filtered the top 10 leading-edge 4 assassi S, Swindell WR, Wu M, et al. Dissecting the heterogeneity of skin gene genes for enrichment of the PI3K-Akt pathway, and focused on expression patterns in systemic sclerosis. Arthritis & Rheumatology 2015;67:3016–26. five genes, excluding the collagen family: cartilage oligomeric 5 Milano A, Pendergrass SA, Sargent JL, et al. Molecular subsets in the gene expression matrix protein (COMP), thrombospondin-1 (THBS1), thrombo- signatures of scleroderma skin. PLoS ONE 2008;3:e2696. 6 Pendergrass SA, Lemaire R, Francis IP, et al. Intrinsic gene expression subsets of spondin-4 (THBS4), fibronectin 1 (FN1) and tenascin-C (TNC). diffuse cutaneous systemic sclerosis are stable in serial skin biopsies. J Invest All of these seem to represent the extent of skin fibrosis because Dermatol 2012;132:1363–73. they have a strong correlation with mRSS. Notably, COMP (also 7 sweeney TE, Braviak L, Tato CM, et al. Genome-wide expression for diagnosis of known as thrombospondin-5) is produced by dermal fibroblasts pulmonary tuberculosis: a multicohort analysis. Lancet Respir Med 2016;4:213–24. and plays a key role in assembly and maintenance of the fibrillar 8 You S, Knudsen BS, Erho N, et al. Integrated classification of prostate cancer reveals a 43 novel luminal subtype with poor outcome. Cancer Research 2016;76:4948–58. collagen extracellular network. Our results corroborated that 9 agrawal N, Akbani R, Aksoy BA, et al. Integrated genomic characterization of papillary COMP expression both in skin and blood is closely correlated thyroid carcinoma. Cell 2014;159:676–90. with mRSS, assessed by cross-sectional and longitudinal anal- 10 assassi S, Radstake TRDJ, Mayes MD, et al. Genetics of scleroderma: implications for yses.43 44 THBS1 is also an essential regulator in various fibrotic personalized medicine? BMC Med 2013;11. 11 Broen JCA, Radstake TRDJ, Rossato M. The role of genetics and epigenetics in the diseases by activating the latent TGF .45 FN1 and TNC are β pathogenesis of systemic sclerosis. Nat Rev Rheumatol 2014;10:671–81. connective tissue molecules, which are induced by TGFβ and 12 ramos PS, Silver RM, Feghali-Bostwick CA. Genetics of systemic sclerosis: recent 46 47 constitutively upregulated in SSc. They drive skin fibrosis by advances. Curr Opin Rheumatol 2015;27:521–9. stimulating collagen gene expression and myofibroblast transfor- 13 Murdaca G, Contatore M, Gulli R, et al. Genetic factors and systemic sclerosis. mation via toll like receptor 4 signalling.46 47 The leading-edge Autoimmun Rev 2016;15:427–32. 14 iwamoto N, Distler O. Molecular targets for therapy in systemic sclerosis. Fibrogenesis genes for enrichment of the PI3K-Akt pathways were the product Tissue Repair 2012;5. of activated fibroblasts, and in turn, they can further stimulate 15 Denton CP, Ong VH. Targeted therapies for systemic sclerosis. Nat Rev Rheumatol fibroblasts through the PI3K-Akt pathway.48 49 Therefore, the 2013;9:451–64.

824 Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 Systemic sclerosis

16 Fonseca C, Lindahl GE, Ponticos M, et al. A Polymorphism in the CTGF Promoter 34 Kanaan R, Strange C. Use of multitarget tyrosine kinase inhibitors to attenuate Region Associated with Systemic Sclerosis. N Engl J Med 2007;357:1210–20. platelet-derived growth factor signalling in lung disease. Eur Respir Rev 2017;26. 17 Makino K, Makino T, Stawski L, et al. Anti-connective tissue growth factor (CTGF/ 35 Brkic Z, van Bon L, Cossu M, et al. The interferon type I signature is present in CCN2) monoclonal antibody attenuates skin fibrosis in mice models of systemic systemic sclerosis before overt fibrosis and might contribute to its pathogenesis sclerosis. Arthritis Res Ther 2017;19. through high BAFF gene expression and high collagen synthesis. Ann Rheum Dis 18 Kim Y-A, Wuchty S, Przytycka TM. Identifying causal genes and dysregulated pathways 2016;75:1567–73. in complex diseases. PLoS Comput Biol 2011;7:e1001095. 36 eloranta M-L, Franck-Larsson K, Lövgren T, et al. Type I interferon system activation 19 Drier Y, Sheffer M, Domany E. Pathway-based personalized analysis of cancer. and association with disease manifestations in systemic sclerosis. Ann Rheum Dis Proceedings of the National Academy of Sciences 2013;110:6388–93. 2010;69:1396–402. 20 Yoshizaki A, Yanaba K, Yoshizaki A, et al. Treatment with rapamycin prevents fibrosis 37 amjadi S, Maranian P, Furst DE, et al. Course of the modifiedR odnan skin in tight-skin and bleomycin-induced mouse models of systemic sclerosis. Arthritis & thickness score in systemic sclerosis clinical trials: analysis of three large Rheumatism 2010;62:2476–87. multicenter, double-blind, randomized controlled trials. Arthritis Rheum 21 Besliu A, Banica L, Predeteanu D, et al. Peripheral blood lymphocytes analysis 2009;60:2490–8. detects CD100/SEMA4D alteration in systemic sclerosis patients. Autoimmunity 38 Gordon JK, Martyanov V, Magro C, et al. Nilotinib (Tasigna™) in the treatment of 2011;44:427–36. early diffuse systemic sclerosis: an open-label, pilot clinical trial. Arthritis Res Ther 22 Bhattacharyya S, Wei J, Varga J. Understanding fibrosis in systemic sclerosis: shifting 2015;17. paradigms, emerging opportunities. Nat Rev Rheumatol 2012;8:42–54. 39 liang M, Lv J, Chu H, et al. Vertical inhibition of PI3K/Akt/mTOR signaling 23 YY H, Lagares D, Tager AM, et al. Fibrosis-a lethal component of systemic sclerosis. demonstrates in vitro and in vivo anti-fibrotic activity. J Dermatol Sci Nat Rev Rheumatol 2014;10:390–402. 2014;76:104–11. 24 Yokoyama Y, Sekiguchi A, Fujiwara C, et al. Inhibitory regulation of skin fibrosis in 40 Wernig G, Chen S-Y, Cui L, et al. Unifying mechanism for different fibrotic diseases. systemic sclerosis by apelin/APJ signaling. Arthritis Rheumatol 2018;70:1661–72. Proc Natl Acad Sci USA 2017;114:4757–62. 25 Fabregat A, Sidiropoulos K, Garapati P, et al. The Reactome pathway Knowledgebase. 41 allanore Y, Matucci-Cerinic M, Distler O. Treatment of systemic sclerosis: is there any Nucleic Acids Res 2016;44:D481–D487. hope for the future?: Table 1. RMD Open 2016;2:e000260. 26 Kanehisa M, Furumichi M, Tanabe M, et al. KEGG: new perspectives on genomes, 42 Fruman DA, Chiu H, Hopkins BD, et al. The PI3K pathway in human disease. Cell pathways, diseases and drugs. Nucleic Acids Res 2017;45:D353–D361. 2017;170:605–35. 27 Brunet J-P, Tamayo P, Golub TR, et al. Metagenes and molecular pattern discovery 43 Farina G, Lemaire R, Pancari P, et al. Cartilage oligomeric matrix protein expression using matrix factorization. Proceedings of the National Academy of Sciences in systemic sclerosis reveals heterogeneity of dermal fibroblast responses to 2004;101:4164–9. transforming growth factor beta. Ann Rheum Dis 2009;68:435–41. 28 aran D, Hu Z, Butte AJ. xCell: digitally portraying the tissue cellular heterogeneity 44 Hesselstrand R, Kassner A, Heinegard D, et al. COMP: a candidate molecule in the landscape. Genome Biol 2017;18. pathogenesis of systemic sclerosis with a potential as a disease marker. Ann Rheum 29 stifano G, Sornasse T, Rice LM, et al. Skin gene expression is prognostic for the Dis 2008;67:1242–8. trajectory of skin disease in patients with diffuse cutaneous systemic sclerosis. 45 Murphy-Ullrich JE, Suto MJ. Thrombospondin-1 regulation of latent TGF-β activation: Arthritis Rheumatol 2018;70:912–9. a therapeutic target for fibrotic disease. Matrix Biology 2018;68-69:28–43. 30 subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a 46 Bhattacharyya S, Tamaki Z, Wang W, et al. FibronectinEDA promotes chronic knowledge-based approach for interpreting genome-wide expression profiles. Proc cutaneous fibrosis through oll-likeT receptor signaling. Sci Transl Med 2014;6. Natl Acad Sci U S A 2005;102:15545–50. 47 Bhattacharyya S, Wang W, Morales-Nebreda L, et al. Tenascin-C drives persistence of 31 lipson KE, Wong C, Teng Y, et al. CTGF is a central mediator of tissue remodeling and organ fibrosis. Nat Commun 2016;7. fibrosis and its inhibition can reverse the process of fibrosis. Fibrogenesis Tissue Repair 48 Goh FG, Piccinini AM, Krausgruber T, et al. Transcriptional regulation of the 2012;5. endogenous danger signal tenascin-C: a novel autocrine loop in inflammation. J 32 Wang Q, Usinger W, Nichols B, et al. Cooperative interaction of CTGF and TGF-β in Immunol 2010;184:2655–62. animal models of fibrotic disease. Fibrogenesis Tissue Repair 2011;4. 49 liao Y-X, Zhang Z-P, Zhao J, et al. Effects of fibronectin 1 on cell proliferation, 33 Beyer C, Distler JH. Tyrosine kinase signaling in fibrotic disorders: translation of basic senescence and apoptosis of human glioma cells through the PI3K/Akt signaling research to human disease. Biochim Biophys Acta 1832;2013:897–904. pathway. Cell Physiol Biochem 2018;48:1382–96.

Moon S-J, et al. Ann Rheum Dis 2019;78:817–825. doi:10.1136/annrheumdis-2018-214778 825 Osteoarthritis

Translational science CircSERPINE2 protects against osteoarthritis by targeting miR-1271 and ETS-related gene Shuying Shen, Yizheng Wu, Junxin Chen, Ziang Xie, Kangmao Huang, Gangliang Wang, Yute Yang, Weiyu Ni, Zhijun Chen, Peihua Shi, Yan Ma, Shunwu Fan

Handling editor Josef S Abstract Key messages Smolen Objectives circular RNAs (circRNA) expression aberration has been identified in ariousv human diseases. ►► Additional material is What is already known about this subject? published online only. To view In this study, we investigated whether circRNAs could ►► Circular RNAs (CircRNAs) have important roles please visit the journal online act as competing endogenous RNAs to regulate the in many diseases, including osteoarthritis (OA). (http://dx.​ ​doi.org/​ ​10.​1136/​ pathological process of osteoarthritis (OA). ► Targeting dysfunctional circRNA-micro-RNA- annrheumdis-2018-​ ​214786). Methods circRNA deep sequencing was performed ► mRNA interactions fulfil a critical therapeutic to the expression of circRNAs between OA and control Department of Orthopaedic promise. Surgery, Sir Run Run cartilage tissues. The regulatory and functional role of Shaw Hospital, Zhejiang CircSERPINE2 upregulation was examined in OA and What does this study add? University school of was validated in vitro and in vivo, downstream target of medicine & KeyLaboratory ► A powerful protective circRNA-CircSERPINE2 CircSERPINE2 was explored. RNA pull down, a luciferase ► of Musculoskeletal System and its ‘CircSERPINE2-miR-1271-5p-E26 reporter assay, biotin-coupled microRNA capture and Degeneration and Regeneration transformation-specific-related gene (ERG)’ axis Translational Research of fluorescence in situ hybridisation were used to evaluate involved in OA pathogenesis. Zhejiang, Hangzhou 310016, the interaction between CircSERPINE2 and miR-1271-5 China p, as well as the target mRNA, E26 transformation- How might this impact on clinical practice? specific-related gene (ERG). The role and mechanism of Correspondence to ► ‘CircSERPINE2-miR-1271-5p-ERG’ axis could be CircSERPINE2 in OA was also explored in rabbit models. ► Dr Shunwu Fan and Dr Yan Ma, a promising therapeutic target in the treatment Results The decreased expression of CircSERPINE2 in Department of Orthopaedic of OA. Surgery, Sir Run Run Shaw the OA cartilage tissues was directly associated with Hospital, Zhejiang University excessive apoptosis and imbalance between anabolic School of Medicine & Key and catabolic factors of extracellular matrix (ECM). Laboratory of Musculoskeletal Mechanistically, CircSERPINE2 acted as a sponge of miR- There is now overwhelming evidence that alter- System Degeneration and ations in microRNA (miRNA) expression levels are Regeneration Translational 1271-5 p and functioned in human chondrocytes (HCs) Research of Zhejiang Province, 3 linked to a variety of disease processes, including through targeting miR-1271-5 p and ERG. Intra-articular 7 8 East Qingchun Road, Hangzhou injection of adeno-associated virus-CircSERPINE2-wt OA progression. For example, expression of Zhejiang Province, 310016, alleviated OA in the rabbit model. miR-140 is reduced in OA tissue and miR-140 China; 0099203@​ ​zju.edu.​ ​cn, regulates ADAMTS-5 expression9 and MMP13,10 zjumayan@​ ​zju.edu.​ ​cn Conclusions our results reveal an important role for a novel circRNA-CircSERPINE2 in OA progression. which are involved in OA pathology in carti- SS, YW and JC contributed CircSERPINE2 overexpression could alleviate HCs lage. Importantly, in addition to miRNA, RNA equally. apoptosis and promote anabolism of ECM through miR- sequencing (RNAseq) has identified multiple fami- 1271-ERG pathway. It provides a potentially effective lies of noncoding RNAs, including antisense RNAs, Received 19 November 2018 long intergenic noncoding RNAs and circular RNAs Revised 26 February 2019 therapeutic strategy for OA progression. 11–13 Accepted 28 February 2019 (circRNAs). Of note, circRNAs are character- Published Online First ised by high tissue-specific expression and stable 28 March 2019 structure, and exist ubiquitously in eukaryotic Introduction cells.14 Compared with linear RNAs, circRNAs can Osteoarthritis (OA) is an age-related or post-trau- undergo noncanonical splicing without a free 3′ or matic degenerative joint disease that affects 12% of 5′ end.15 Recent reports show that some circRNAs people over the age of 25 years and 50% of popu- are enriched in miRNA-binding sites and can func- lation over the age of 65 years1. OA is characterised tion as miRNA sponges or competing endoge- with the progressive loss of hyaline articular carti- nous RNAs (ceRNAs) that naturally sequester and lage, concomitant sclerotic changes in the subchon- competitively inhibit miRNA activity.16 Of interest, dral bone and advancement in osteophytes.2 The some researches have indicated that circRNAs may identified risk factors include ageing, previous joint also be central regulators of biological processes.17 injury, obesity, genetics, sex and anatomical factors Currently, little is known about the functional role © Author(s) (or their employer(s)) 2019. Re-use related to joint shape and alignment; however, the of circRNAs in OA joint tissue. Their potential 3–6 permitted under CC BY-NC. No exact pathogenesis of OA remains undefined. importance is indicated in a recent report by Liu et commercial re-use. See rights Current therapy is limited and there is no effective al,18 who identified 104 circRNAs that were differ- and permissions. Published treatment for OA. Therefore, elucidation of the entially expressed in damage cartilage samples using by BMJ. mechanisms that regulate the expression of extra- a microarray-based approach. Although this prelim- To cite: Shen S, Wu Y, cellular matrix (ECM) degradation-related genes is inary study did not examine the mechanism and Chen J, et al. Ann Rheum Dis critical for the development of effective therapies function of these circRNAs in depth, we hypoth- 2019;78:826–836. for OA. esise that circRNAs may be central regulators in

826 Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 Osteoarthritis cartilage tissue. Therefore, the aim of this study was to investi- of means=4.135; 95% CI 0.412 to 7.857; *p<0.05), Circ- gate the functional circRNAs in OA progress. SERPINE2 expression was downregulated in OA samples In the current study, we identified a circRNA (hsa_ (figure 1E–F, differentials of means=0.407, 95% CI 0.043 circ_0008365, termed CircSERPINE2) in OA, and systemically to 0.772, *p<0.05). We used Sanger sequencing to confirm investigated its role in in vitro and animal models of OA. the head-to-tail splicing (a special splice reaction formed by a 5′-end splice site and the corresponding site at 3′-end of an Methods exon) in the RT-qPCR product of CircSERPINE2 identified by Detailed experimental procedures are described in the online its expected size (figure 1G). However, head-to-tail splicing may supplementary materials and methods and table S4. be produced by trans-splicing (a phenomenon naturally existing in different organisms, wherein the 5′-UTR of pre-mRNA is trimmed by a spliced leader RNA, followed by the attachment Results of a short sequence to the 5′-terminus) or genomic rearrange- CircRNA expression patterns in human OA and control tissues ment. Therefore, we took several steps to rule out these possi- To generate a circRNA profiling database, we performed RNAseq bilities as per the previously described methodology.25 26 We analyses of ribosomal RNA-depleted total RNA from three clin- first designed convergent primers to amplify SERPINE2 mRNA ical OA and three control tissues. Each sample was sequenced and divergent primers to amplify CircSERPINE2 using cDNA on an Illumina HiSeq yielding at least 30 million reads, which and genomic DNA (gDNA). CircSERPINE2 was amplified by mapped to reference genome (hg38, human genome）by BWA19 20 divergent primers in cDNA but not in gDNA (figure 1H). In the and STAR separately. A total of 12 738 circRNAs were iden- second step, we performed northern blot analysis, as shown in tified by both CIRI2 and CIRCexplorer, the majority of iden- figure 1I, a single band at expected size (707 nt) was detected tified circRNAs in the study were supported by more than 10 by CircSERPINE2-specific probe targeting the junction region reads (online supplementary figure S1A). We annotated these 21 22 in cells with or without RNase R digestion. We confirmed that identified candidates using the RefSeq database, most of CircSERPINE2 was resistant to RNase R, whereas SERPINE2 the circRNAs originated from protein-coding exons, and others mRNA level was significantly decreased after RNase R treatment aligned with 5′-UTR and 3′-UTR introns (online supplementary (figure 1J). FISH and RT-qPCR analysis indicated the abundant figure S1B). The majority of the identified exonic circRNAs were expression of cytoplasmic CircSERPINE2 in HCs (figure 1K,L). <1000 nucleotides (nt) in length (online supplementary figure S1C). The chromosome distribution of the identified circRNAs showed no obvious difference between OA and control groups CircSERPINE2 regulates apoptosis and ECM metabolism in (online supplementary figure S1D and E). Consistent with a chondrocytes previous report,23 analysis of the number of circRNAs in their To assess the involvement of CircSERPINE2 in the regulation of host genes indicated that one gene could produce multiple matrix-degrading enzymes, we transfected HCs with CircSER- circRNAs (online supplementary figure S1F). We normalised PINE2 small-interfering RNA (figure 2A). Knockdown of Circ- the back-spliced reads (support for circRNA) by read length and SERPINE2 expression had no effect on SERPINE2 mRNA level number of mapped reads (spliced reads per billion mapping,24 (figure 2B). We examined the effects of CircSERPINE2 inhibition which permits quantitative comparisons between back splicing or overexpression on IL-1β-induced HC apoptosis. As shown from different RNAseq data. There is often one predominantly in figure 2C, the knockdown of CircSERPINE2 expression expressed circRNA isoform from one gene locus (online supple- increased the rate of chondrocyte apoptosis while the increase mentary figure S1G). OA and control groups showed differ- in the apoptotic rate after IL-1β treatment markedly suppressed ential expression pattern of circRNAs (figure 1A). To validate in response to CircSERPINE2 overexpression (figure 2D,E). the RNA deep sequencing results, we performed real-time (RT) In addition, the inhibition of CircSERPINE2 expression quantitative (qPCR) analysis to identify the 25 most differen- promoted the expression of MMP3, MMP13 and ADAMTS4, tially expressed circRNAs (online supplementary table S1) and and decreased the levels of SOX9, COL2A1 and aggrecan, as confirmed 14 downregulated circRNAs (online supplementary observed in WB (figure 2F). The knockdown of CircSERPINE2 figure S2A). In the subsequent step, we used interleukin (IL)-1β- expression obviously affected MMP13, ADAMTS4, COL2A1 treated and tumour necrosis factor (TNF)-α-treated primary and aggrecan levels in HCs and SW1353 cells, as evident from human chondrocytes (HCs) and confirmed the downregulated immunofluorescence (IF) assay (figure 2G & supplementary expression of hsa_circ_00008667, hsa_circ_0072568, hsa_ figure 3A) and RT-qPCR results (figure 2H & supplementary circ_0008365, hsa_circ_0020093, hsa_circ_0110251 and hsa_ figure 3B). These data clearly indicate the proapoptotic and circ_0001103 in both IL-1β- and TNF-α-treated HCs (figure procatabolic effects of CircSERPINE2 knockdown in HCs and 1B & supplementary figure1B & S2B, *p<0.05). Moreover, SW1353 cells. western blotting (WB) for COL2A1, MMP3, MMP13 and SOX9 expression revealed the protective role of hsa_circ_0008365, CircSERPINE2 serves as sponge for miRNAs which is formed by the circularisation of exons 2–4 of Serpine2 As CircSERPINE2 level is abundant and stable in the cytoplasm, gene (hereafter referred to as CircSERPINE2) (online supple- we investigated the ability of CircSERPINE2 to bind to miRNAs. mentary figure S2C). A search in the Circular RNA Interactome database revealed that CircSERPINE2 has 11 conserved Argonaute 2 (AGO2)-binding CircSERPINE2 expression is relatively low in OA tissues and sites (figure 3A). We demonstrated the specific enrichment of cell lines and is predominantly localised in the cytoplasm endogenous CircSERPINE2 pulled down from Flag-AGO2 cells We detected the expression of CircSERPINE2 in 10 human OA by RNA immunoprecipitation (RIP) analysis (figure 3B). The cartilage and 10 control cartilage (figure 1C & online supple- differential expression of miRNAs in control and OA tissues was mentary figure1C & S2D-E). RT-qPCR and RNA fluorescence investigated using deep sequencing. In all, 191 miRNAs showed in situ hybridisation (FISH) showed that unlike the upreg- upregulated expression in OA tissues as compared with control ulated SERPINE2 gene expression (figure 1D, differentials tissues (figure 3C, supplementary figure 3S 3C, table S2, and

Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 827 Osteoarthritis

Figure 1 CircSERPINE2 validation and expression in OA cartilage tissue and cells. (A) Heat map of all differentially expressed circRNAs between OA and control cartilage samples. (B) The expression of circRNAs in the chondrocytes stimulated with TNF-α (10 ng/mL) and IL-1β (10 ng/mL) for 24 hours. n=3 (three different experiments). *p<0.05. (C) X-ray (upper) and safranin-O/fast green (lower) staining of the cartilage from OA or control patients. Scale bar: 50 µm. (D) SERPINE2 expression was higher in human OA cartilage than in control cartilage. n = 10. *p<0.05. (E) CircSERPINE2 expression was lower in human OA cartilage than in control cartilage tissue. n = 10. *p<0.05. (F) CircSERPINE2 expression was lower in human OA cartilage than in control cartilage tissue. Representative images are shown (scale bar, 50–200 µm). n=3 (three different donors). *p<0.05, mean with 95% CI. (G) Schematic illustration showing SERPINE2 exons 2–4 circularisation to form CircSERPINE2 (black arrow). The presence of CircSERPINE2 was validated by RT-PCR, followed by Sanger sequencing. Red arrow represents head-to-tail CircSERPINE2 splicing sites. (H) The presence of CircSERPINE2 was validated in HCs by RT-PCR. Divergent primers amplified CircSERPINE2 from cDNA, but not from genomic DNA. GAPDH was used as a negative control. (I) Northern blot analysis for the detection of CircSERPINE2 expression in HC and SW1353 cells treated with or without RNase R digestion. The panels show the blots probed with CircSERPINE2, and the red triangle indicates CircSERPINE2 band size (707 nt). (J) The expression of CircSERPINE2 and SERPINE2 mRNA in HC cells treated with or without RNase R was detected by RT-qPCR. The relative levels of CircSERPINE2 and SERPINE2 mRNA were normalised to the values measured for mock treatment. n=9 (three different donors for three different experiments). *p<0.05. (K) RNA FISH revealed the predominant localisation of CircSERPINE2 in the cytoplasm. CircSERPINE2 probes were labelled with Cy-3. Nuclei were stained with DAPI. Scale bar, 50 µm. Upper panel: FISH with junction-specific probes indicative of the cellular localisation of CircSERPINE2. n = 3 (three different donors). Scale bars=50 µM. Lower panel: CircSERPINE2 expression was detected in different cell fractions. Nuclear and cytoplasmic RNAs were extracted, and junction primers were used for CircSERPINE2 detection. U6 was used as an internal control for nuclear RNA, whereas GAPDH served as the control for cytoplasmic RNA. n=9 (three different donors for three different experiments). *p<0.05. CircRNAs, circular RNAs; DAPI, 4′,6-diamidino-2-phenylindole; FISH, fluorescence in situ hybridisation; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; HCs, human chondrocytes; IL, interleukin; OA, osteoarthritis; RT-qPCR, real-time quantitative PCR; TNF, tumour necrosis factor.

828 Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 Osteoarthritis

Figure 2 CircSERPINE2 expression knockdown induced the expression of matrix-degrading enzymes in chondrocytes. (A) HCs were transfected with CircSERPINE2 siRNA or negative control siRNA at a final concentration of 20 nM. After 48 hours of transfection, the expression level of CircSERPINE2 was measured by RT-qPCR and normalised to β-actin level. n=9 (three different donors for three different experiments). *p<0.05. (B) HCs were transfected with CircSERPINE2#2 siRNA or negative control siRNA at a final concentration of 20 nM. After 48 hours of transfection, the expression level of SERPINE2 was measured by RT-qPCR and normalised to β-actin level. n=9 (three different donors for three different experiments). *p<0.05. (C) HCs were transfected with CircSERPINE2#2 siRNA or negative control siRNA at a final concentration of 20 nM. After 48 hours, cell apoptosis was detected with annexin V-FITC/PI double staining using quantitative FACS analysis. n=3 (three different donors). *p<0.05. (D) Cells were infected with CircSERPINE2 lentivirus or control virus. Overexpression was detected by RT-qPCR. n=9 (three different donors for three different experiments). *p<0.05. (E) HCs were infected with CircSERPINE2 lentivirus or control virus and stimulated with IL-1β for 48 hours. Cell apoptosis was detected with annexin V-FITC/PI double staining using quantitative FACS analysis. n=3 (three different donors). *p<0.05. (F–H) HCs were transfected with CircSERPINE2#2 siRNA or negative control siRNA at a final concentration of 20 nM, followed by the treatment with or without IL-1β. After 48 hours, WB for (F, n=3) MMP13, MMP3, aggrecan, SOX9, COL2A1 and ADAMTS4 detection; IF (G, scale bar: 50 µm) for MMP13, aggrecan, COL2A1 and ADAMTS4 expression detection; and RT-qPCR (H, n=9; three different donors for three different experiments; *p<0.05) for MMP3, MMP13, ADAMTS4, ,COL2A1, SOX9 and aggrecan expression evaluation were performed. FACS, fuorescence-activated cell sorting; FITC, fluorescein isothiocyanate; HCs human chondrocytes; IF, immunofluorescence; IL; interleukin; PI, propidium iodide; RT-qPCR, real-time quantitative PCR; siRNA, small interfering RNA; WB, western blotting.

Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 829 Osteoarthritis

Figure 3 CircSERPINE2 serves as a sponge for miR-1271 in HC cells. (A) Number of RBP sites between RBPs and CircSERPINE2 is shown. (B) AGO2 RIP assay was performed to detect CircSERPINE2 levels in SW1353 cells stably expressing AGO2. n=3. *p<0.05. (C) miRNA-mediated differential dysregulation of miRNAs was detected by miRNA high-throughput sequencing in OA cartilage samples and control cartilage samples (n=3). (D) Schematic illustration to show the overlapping of the target miRNAs of CircSERPINE2, as predicted by Miranda, TargetScan and miRNAs upregulated in OA identified by sequencing analysis. (E) Lysates prepared from SW1353 cells were subjected to an RNA pull-down assay and tested with RT-qPCR. Relative levels of CircSERPINE2 were normalised to the levels of input. n = 3. *p<0.05 versus control (lac Z) probe. (F–G) miR-1271 expression was higher in human OA cartilage than in control cartilage tissue, as determined by FISH (F, n=3; three different donors; *p<0.05; mean with 95% CI) and RT-qPCR (G, n=10, *p<0.05). Representative images are shown (scale bar: 50–200 µm). *p<0.05. (H) HEK-293T cells were co-transfected with miR-1271 mimic or NC and a luciferase reporter construct containing WT or MUT CircSERPINE2. n = 6. *p<0.05. Mean with 95% CI. (I) Fish images showing the co-localisation of CircSERPINE2 and miR-1271 in HCs. miR-1271 probes were labelled with Alexa fluor 488, whereas CircSERPINE2 probes were tagged with Cy3. Nuclei were stained with DAPI. n = 3 (three different donors). Scale bar, 50 µm. (J–L) HCs were transfected with miR- 1271 mimic, inhibitor or NC at a final concentration of 20 nM. After 48 hours of transfection (J, n = 3; three different donors), protein levels of MMP3, MMP13, ADAMTS4, SOX9, COL2A1 and aggrecan in chondrocytes were detected; IF assay for MMP13, aggrecan, COL2A1 and ADAMTS4 expression detection in chondrocytes was performed (K, n = 3; three different donors; scale bar: 50 µm). mRNA levels of MMP3, MMP13, ADAMTS4, SOX9, COL2A1 and aggrecan in chondrocytes were evaluated (l, n=9; three different donors for three different experiments, *p<0.05). AGO2, Argonaute 2; DAPI, 4′,6-diamidino-2-phenylindole; FISH, fluorescence in situ hybridisation; HC, human chondrocyte; IF, immunofluorescence; miRNAs, microRNAs; MUT, mutant; NC, negative control; OA, osteoarthritis; RBP, RNA-binding protein; RIP, RNA immunoprecipitation; RT-qPCR; real-time quantitative PCR; WT, wild type.

830 Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 Osteoarthritis online supplementary figure S4A-C). We identified seven candi- figure S7D), as indicated in recent reports.27 28 ERG was there- date miRNAs by overlapping the predicted miRNA recognition fore selected for further investigation and was found to be mark- elements (MREs) in CircSERPINE2 sequence using miRanda, edly downregulated in OA-affected, damaged regions of human TargetScan and miRNA sequence (figure 3D). The levels of the cartilage (figure 5B,C, differentials of means=0.766, 95% CI seven candidate miRNAs were determined using a pull-down 0.213 to 1.319, *p<0.05), suggestive of the potential role of assay. As shown in figure 3E and supplementary figure S4D ERG in OA pathogenesis. Furthermore, bioinformatic analysis and E, miR-1271 and miR-93 were pulled down by CircSER- using miRNA target prediction software revealed that ERG is a PINE2. Furthermore, WB results demonstrated the association putative target of miR-1271 (online supplementary figure S7E). of miR-1271 with the protein expression of matrix-degrading Luciferase assay results showed that miR-1271 mimic signifi- and synthesising components (supplementary figure S4F). cantly inhibited the luciferase activity in the cells transfected with Figure 3F,G shows the upregulated expression of miR-1271 in the WT 3′-UTR of ERG, but not in those transfected with mutant 10 human OA cartilage samples (differentials of means=2.728; (MUT) 3′-UTR of ERG (figure 5D). Furthermore, figure 5E,F 95% CI 0.623 to 4.832, *p<0.05), consistent with the results demonstrated that miR-1271 mimic reduced the expression of deep sequencing (figure 3C). Therefore, miR-1271 was of ERG protein and mRNA while miR-1271 inhibitor exerted selected for further analysis. Luciferase assay confirmed the opposite effects. In addition, ERG overexpression decreased the binding of CircSERPINE2 to miR-1271 (figure 3H & supple- expression of MMP3, MMP13 and ADAMTS4 and increased the mentary figure3H & S4G). FISH experiment further confirmed levels of SOX9 and COL2A1 (figure 5G). These results suggest the co-localisation of CircSERPINE2 and miR-1271 (figure 3I). that miR-1271 targets ERG, and endogenous ERG expression is The effects of miR-1271 were investigated. miR-1271-overex- closely related to OA pathogenesis. pressing cells demonstrated the increase in the percentage of We designed a rescue experiment to verify whether the apoptotic cells (supplementary figure S4H-I), enhanced expres- effects of miR-1271 expression on chondrogenic phenotypes sion of MMP3, MMP13 and ADAMTS4, and decreased levels of were achieved through ERG. Figure 5H and online supplemen- SOX9, COL2A1 and aggrecan, as determined by WB (figure 3J), tary figure S7F and G showed that the infection of pre-miR- IF (figure 3K & supplementary figure3K & S5A-B) and RT-qPCR 1271 adenovirus resulted in the inhibition of ERG expression. (figure 3L & supplementary figure3L & S5C). On the contrary, Furthermore, as shown in figure 5H–J, the infection of pre-miR- miR-1271 expression inhibition exerted opposite effects on 1271 adenovirus remarkably decreased the expression levels of matrix-degrading and synthesising components (figure 3J–L & COL2A1, SOX9 and aggrecan and increased MMP3, MMP13 supplementary figure S5A-C). and ADAMTS4 levels in HCs as compared with the nega- tive control cells. However, these effects were abolished after ERG overexpression. Taken together, our results indicate that CircSERPINE2 functions in OA by targeting miR-1271 miR-1271 overexpression may impair chondrogenic phenotypes expression via ERG. To examine whether CircSERPINE2 functions in OA via inter- action with miR-1271, we co-infected cells with sh-CircSER- PINE2 and miR-1271 sponge adenovirus. The miR-1271 sponge Injection of CircSERPINE2 alleviates OA in a rabbit model adenovirus contains repeated complementary-binding sites to To determine whether CircSERPINE2 plays a role in OA progres- miR-1271 and has continuous miR-1271 loss of function in sion in vivo, WT or MUT adeno-associated virus (AAV) CircSER- cells. As shown in figure 4A, the knockdown of CircSERPINE2 PINE2 was intra-articularly administered to anterior cruciate expression increased the rate of HC apoptosis while this rate ligament transection (ACLT)-induced OA rabbits. Overexpres- was markedly rescued by the inhibition of the expression of both sion efficiency was confirmed by FISH and RT-qPCR (online miR-1271 and CircSERPINE2. The results showed that protein supplementary figure S8A and B). Safranin O and fast green and mRNA levels of MMP13, MMP3 and ADAMTS4 signifi- staining (figure 6A) showed cartilage surfaces in ACLT-induced cantly decreased while those of SOX9, COL2A1 and aggrecan OA rabbits improved after the injection of WT AAV CircSER- increased in the cells co-infected with sh-CircSERPINE2 and PINE2 but not MUT AAV CircSERPINE2. Quantitative anal- miR-1271 sponge adenovirus as compared with the cells infected ysis with Osteoarthritis Research Society International (OARSI) with sh-CircSERPINE2 alone (figure 4B,D & online supplemen- scoring showed that WT AAV CircSERPINE2 significantly tary figure S6A-C). IF analysis confirmed that the expression of lowered OARSI scores, whereas MUT AAV CircSERPINE2 treat- COL2A1 and aggrecan increased and MMP13 and ADAMTS4 ment increased OARSI scores (figure 6B). Three-dimensional levels decreased in the cells with CircSERPINE2 and miR-1271 reconstruction of micro-CT images demonstrated the increase expression inhibitors as compared with the cells treated with in osteophytes in OA rabbits, whereas CircSERPINE2 intra-ar- CircSERPINE2 inhibitor (figure 4C & supplementary figure4C ticular injection decreased osteophytes (figure 6C). The injection & S6D). Taken together, these results suggest that CircSER- of WT AAV CircSERPINE2 alleviated the degenerative changes PINE2 functions by targeting miR-1271 in vitro. in the cartilage matrix, such as enhanced apoptotic (TUNEL) and catabolic response and increased ECM composition in the rabbit model of OA, as indicated by immunohistochemistry and miR-1271 directly targets ETS-related gene WB results (figure 6D–E). Taken together, these results revealed We investigated the possible targets of miR-1271 and CircSER- the positive effects of elevated CircSERPINE2 on the inhibition PINE2 by overlapping the predicted results of TargetScan, miRDB of ECM catabolism to promote anabolism and alleviate OA in and RNA-seq (figure 5A, online supplementary figure S7A and vivo (figure 6F). B, and supplementary table S3). We first selected 10 significantly differentially expressed genes; RT-qPCR results showed that eight genes were regulated by CircSERPINE2 (supplementary Discussion figure S7C). Moreover, E26 transformation-specific (ETS)-re- At present, no effective therapeutic drugs have been approved lated gene (ERG) had the most obvious effect on COL2A1, for the clinical treatment of OA. To develop effective therapeu- aggrecan, MMP3 and MMP13 expression (online supplementary tics, it is important to investigate the underlying mechanisms

Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 831 Osteoarthritis

Figure 4 Knockdown of miR-1271 expression reverses the CircSERPINE2-induced cartilage degradation in HC cells. (A) Downregulation of both CircSERPINE2 and miR-1271 resulted in fewer apoptotic HCs than those observed with the inhibition of CircSERPINE2 alone. Apoptosis rates were determined with annexin V-FITC/PI staining and FACS. n = 3 (three different donors). *p<0.05. (B) The expression of MMP3, MMP13, ADAMTS4, SOX9, aggrecan and COL2A1 in HC cells was detected by WB. Cells were co-infected with sh-CircSERPINE2 and miR-1271 sponge adenovirus or control virus. n = 3 (three different donors). (C) The expression of MMP13, ADAMTS4, aggrecan and COL2A1 in HC cells was detected by IF analysis. Cells were infected with control virus or sh-CircSERPINE2 with or without miR-1271 sponge adenovirus. n = 3 (three different donors). Scale bars=50 µm. (D) mRNA expression of CircSERPINE2, MMP3, MMP13, ADAMTS4, SOX9, aggrecan and COL2A1 in HC cells was detected by RT-qPCR analysis. Cells were infected with control virus or sh-CircSERPINE2 with or without miR-1271 sponge adenovirus. n=9 (three different donors for three different experiments). *p<0.05, cells co-infected with miR-1271 sponge and sh-CircSERPINE2 adenovirus versus cells infected with sh-CircSERPINE2 ,adenovirus only. FACS, fuorescence-activated cell sorting; FITC, fluorescein isothiocyanate; HCs, human chondrocytes; IF, immunofluorescence; PI propidium iodide; RT-qPCR, real-time quantitative PCR; WB, western blotting.

832 Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 Osteoarthritis

Figure 5 ERG serves as a direct target of miR-1271 and mediates the function of miR-1271 in HC cells. (A) The differential dysregulation of mRNAs by CircSERPINE2 knockdown was detected by RNA-seq in HCs (n=3). (B) ERG expression levels were lower in human OA cartilage than in control tissue. n = 10. *p<0.05. (C) ERG expression levels were lower in human OA cartilage than in control tissue. Representative images are shown. n=3; *p<0.05; mean with 95% CI. (D) HEK-293T cells were co-transfected with miR-1271 mimics or NC and luciferase reporter constructs containing WT or MUT 3′-UTR of ERG. n = 6; *p<0.05; mean with 95% CI. (E–F) miR-1271 overexpression reduced ERG (E) protein and (F) mRNA levels while miR-1271 inhibition increased ERG (E) protein and (F) mRNA levels. Cells were transfected with NC or miR-1271 mimic/inhibitor, and mRNA or protein levels were evaluated. Protein expression was evaluated by WB; mRNA levels were evaluated by RT-qPCR. n=9 (three different donors for three different experiments). *p<0.05. (G–H) The expression of ERG, MMP3, MMP13, ADAMTS4, SOX9 and COL2A1 in HC cells was detected by WB. Cells were infected with ERG adenovirus (G) co-infected with pre-miR-1271 (H) or control virus. (I) Expression of MMP13, ADAMTS4, aggrecan and COL2A1 in HC cells was detected by IF. Cells were co-infected with ERG and pre-miR-1271 adenovirus or control virus. n = 3 (three different donors). Scale bars=50 µm. (J) mRNA expression of MMP3, MMP13, ADAMTS4, SOX9, aggrecan and COL2A1 in HC cells was detected by RT-qPCR analysis. Cells were co-infected with ERG and pre-miR-1271 adenovirus or control virus. n=9 (three different donors for three different experiments). *p<0.05, cells co-infected with ERG and pre-miR-1271 adenovirus versus cells infected with pre-miR-1271 adenovirus. ERG, E26 transformation-specific (ETS)- related gene; HC, human chondrocyte; MUT, mutant; NC, negative control; OA, osteoarthritis; RT-qPCR; real-time quantitative PCR; WB, western blotting; WT, wild type.

Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 833 Osteoarthritis

Figure 6 CircSERPINE2 alleviates OA in vivo. (A) Safranin-O/fast green staining of the cartilage in the indicated groups at 8 weeks after surgery. Scale bar=200 µm. ACLT-induced OA rabbits were injected with AAV negative control, WT AAV CircSERPINE2, or MUT AAV CircSERPINE2, and the degree of knee OA was evaluated by safranin-O/fast green staining (n=12). (B) OARSI scoring was performed according to staining results; *p<0.05. (C) Micro-CT images of OA rabbits. (D) FISH and IHC staining of ACLT-induced OA rabbits injected with AAV negative control, WT AAV CircSERPINE2 or MUT AAV CircSERPINE2. (E) WB for the expression of ERG, SOX9, catabolic enzymes (MMP-3 and MMP-13), and ECM composition (COL2A1) in rabbit cartilage tissues. the injection of WT AAV CircSERPINE2 alleviated the progression of OA and induced the expression of Col2a1 in the rabbit model of OA. Scale bar=100 µm. (F) Schematic of the working hypothesis. AAV, adeno-associated virus; ACLT, anterior cruciate ligament transection; ECM, extracellular matrix; ERG, E26 transformation-specific-related gene; FISH, fluorescence in situ hybridisation; IHC, immunohistochemistry; MUT mutant; OA, osteoarthritis; OARSI, Osteoarthritis Research Society International; TUNEL, terminal dexynucleotidyl transferase(TdT)-mediated dUTP nick end labeling; WB, western blotting; WT, wild type.

834 Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 Osteoarthritis of OA.29 30 CircRNAs are endogenous noncoding RNAs that been proposed that the lack of ERG expression may render joints are widespread, tissue specific and conserved in mammalian more susceptible to surgical challenge. In addition, the beneficial cells and have a great potential that remains to be explored. roles of ERG in OA have been verified in mice.43 The current Evidences have shown that certain circRNAs are related to the study also shows a distinctive decrease in ERG expression levels occurrence and progression of various diseases. However, little in the articular cartilage of patients with OA. Moreover, ERG is known about their functions in the development and progres- overexpression induced by miR-1271 knockdown was found sion of OA and their roles in the regulation of the vitality and to play a protective role in OA. Therefore, we propose the functions of HCs. Therefore, evaluation of the cellular mech- targeting of the CircSERPINE2/miR-1271/ERG axis as a preven- anisms of circRNAs in OA may provide a new perspective for tative strategy for OA treatment. However, we do not exclude the treatment of OA. In the present study, we first identified the possibility of the involvement of other critical gene targets CircSERPINE2 as a key downregulated circRNA involved in of CircSERPINE2 besides ERG that may play important roles in OA. In addition, we used gain-of-function and loss-of-function the progression of OA. approaches in vitro to demonstrate the participation of Circ- In conclusion, we have identified the CircSER- SERPINE2 in the proapoptotic response and protection of ECM PINE2-miR-1271-ERG axis as a novel target for the prevention components from degradation. We revealed a new function of and treatment of OA. CircSERPINE2 and ERG overexpression CircSERPINE2 in OA progression and identified it as a prom- may alleviate the catabolism of ECM. Besides, miR-1271 inhi- ising therapeutic target for OA treatment. bition may serve as an effective therapeutic strategy for OA CircSERPINE2 is generated by the back splicing of the exons treatment. 2 to 4 of SERPINE2 gene. It mainly comprises the protein- Contributors SF, YM and SS designed the experiments. SS, YM, YW and JC coding sequence of SERPINE2 mRNA; hence, its sequence is performed the experiments and acquired the data. SS, ZX, KH, GW, YY, WN, ZC relatively conserved between human and rabbit. Moreover, the and PS analysed the data. SF, YM and SS supervised the project and wrote the ERG-binding site of miR-1271 is highly conserved in rabbit, manuscript. justifying the use of a rabbit animal model. SERPINE2 is a Funding The study was sponsored by National Natural Science Foundation of subtype of the serpin superfamily and plays a role in the modi- China (81802680, 81871797, 81472064, 81873985, 81874015 and 81601925), Key Project of Zhejiang Medical Science and Technology Plan (2015145597 and fication of matrix metalloproteinase activity in ECM homeo- 2016145597), Key Project of Zhejiang Provincial Natural Science Foundation 31–33 stasis. In our study, SERPINE2 expression was upregulated (LZ15H06002), Medical Science and Technology Project of Zhejiang Province and CircSERPINE2 levels were downregulated in patients with (2017179447). No benefits in any form have been or will be received from a OA. However, both molecules played a protective role in OA commercial party related directly or indirectly to the subject of this study. progression. In addition, CircSERPINE2 knockdown or over- Competing interests None declared. expression had no effect on SERPINE2 expression, indicative of Patient consent for publication Obtained. the independent role of CircSERPINE2 in OA progression. Provenance and peer review Not commissioned; externally peer reviewed. Accumulating evidences have indicated the key mechanistic 34 Open access This is an open access article distributed in accordance with the role of ceRNA networks in many diseases, including OA. Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which Various types of RNAs, including circRNAs, pseudogenes, permits others to distribute, remix, adapt, build upon this work non-commercially, long noncoding RNAs and mRNAs, may function as ceRNAs and license their derivative works on different terms, provided the original work is in distinct physiological and pathophysiological conditions. properly cited, appropriate credit is given, any changes made indicated, and the use Computational models and analyses of clinical microarray data is non-commercial. See: http://​creativecommons.org/​ ​licenses/​by-​nc/4.​ ​0/. to explore ceRNA regulation network showed that ceRNA essen- tially operates in an miRNA-target titration mechanism,35 36 and References 1 Hayami T, Pickarski M, Zhuo Y, et al. Characterization of articular cartilage and the abundance of competing RNAs and miRNAs, the binding subchondral bone changes in the rat anterior cruciate ligament transection and energy of ceRNAs and miRNAs, and the number of MREs are meniscectomized models of osteoarthritis. Bone 2006;38:234–43. the key determinants of the range and strength of ceRNA regu- 2 Dieppe P, Kirwan J. The localization of osteoarthritis. Br J Rheumatol lation.37 38 Many validated ceRNA networks participate in the 1994;33:201–3. 39 3 Goldring MB. The role of the chondrocyte in osteoarthritis. Arthritis Rheum initiation and progression of various diseases. However, the 2000;43:1916–26. overall pathophysiological contributions of circRNAs to OA 4 Goldring MB. Update on the biology of the chondrocyte and new approaches to remain largely unknown. In the present study, CircSERPINE2 treating cartilage diseases. Best Pract Res Clin Rheumatol 2006;20:1003–25. contains a conserved miR-1271 target site that was validated by 5 conaghan PG, Kloppenburg M, Schett G, et al. Osteoarthritis research priorities: a pull-down, RIP, luciferase and FISH analyses. Furthermore, the report from a EULAR ad hoc expert Committee. Ann Rheum Dis 2014;73:1442–5. 6 lotz M. Cytokines in cartilage injury and repair. Clin Orthop Relat Res 2001;391(391 expression of ERG (miR-1271 target) was positively regulated Suppl):S108–15. by CircSERPINE2. Therefore, we propose a mechanism wherein 7 nugent M. MicroRNAs: exploring new horizons in osteoarthritis. Osteoarthritis CircSERPINE2 acts as an miR-1271 sponge to inhibit HC apop- Cartilage 2016;24:573–80. tosis, promote ECM anabolism and suppress ECM catabolism, 8 Vicente R, Noël D, Pers Y-M, et al. Deregulation and therapeutic potential of microRNAs in arthritic diseases. Nat Rev Rheumatol 2016;12. thereby delaying the progression of OA. 9 Tardif G, Hum D, Pelletier J-P, et al. Regulation of the IGFBP-5 and MMP-13 genes by Several miRNAs are known to exhibit aberrant expression the microRNAs miR-140 and miR-27a in human osteoarthritic chondrocytes. BMC levels in OA.40 41 With increasing evidences highlighting the Musculoskelet Disord 2009;10. link between miRNAs and OA pathogenesis, miRNAs have been 10 araldi E, Schipani E. MicroRNA-140 and the silencing of osteoarthritis. Genes Dev recognised as the potential therapeutic targets for OA treat- 2010;24:1075–80. 11 Batista PJ, Chang HY. Long noncoding RNAs: cellular address codes in development ment. In the present study, we found that miR-1271 expres- and disease. Cell 2013;152:1298–307. sion was significantly upregulated in OA and inversely related 12 cech TR, Steitz JA. The noncoding RNA revolution-trashing old rules to forge new to the expression of SOX9, aggrecan and COL2A1. Subsequent ones. Cell 2014;157:77–94. analyses showed that miR-1271 suppressed ERG expression, a 13 Tay Y, Rinn J, Pandolfi PP. The multilayered complexity of ceRNA crosstalk and competition. Nature 2014;505:344–52. mechanism reported herein for the first time. ERG was recently 14 salzman J, Gawad C, Wang PL, et al. Circular RNAs are the predominant shown to play an important role in inducing the permanent transcript isoform from hundreds of human genes in diverse cell types. PLoS One differentiation of chondrocytes into joint-forming cells.42 It has 2012;7:e30733.

Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 835 Osteoarthritis

15 Hentze MW, Preiss T. Circular RNAs: splicing’s enigma variations. Embo J 30 Wieland HA, Michaelis M, Kirschbaum BJ, et al. Osteoarthritis - an untreatable 2013;32:923–5. disease? Nat Rev Drug Discov 2005;4:331–44. 16 Hansen TB, Jensen TI, Clausen BH, et al. Natural RNA circles function as efficient 31 Boulaftali Y, François D, Venisse L, et al. Endothelial protease nexin-1 is a novel microRNA sponges. Nature 2013;495:384–8. regulator of a disintegrin and metalloproteinase 17 maturation and endothelial 17 chen YT, Rettig WJ, Yenamandra AK, et al. Cerebellar degeneration-related antigen: a protein C receptor shedding via furin inhibition. Arterioscler Thromb Vasc Biol highly conserved neuroectodermal marker mapped to chromosomes X in human and 2013;33:1647–54. mouse. Proc Natl Acad Sci U S A 1990;87:3077–81. 32 Pagliara V, Adornetto A, Mammì M, et al. Protease nexin-1 affects the migration and 18 liu Q, Zhang X, Hu X, et al. Emerging roles of circRNA related to the mechanical invasion of C6 glioma cells through the regulation of urokinase plasminogen activator stress in human cartilage degradation of osteoarthritis. Mol Ther Nucleic Acids and Matrix Metalloproteinase-9/2. Biochim Biophys Acta 2014;1843:2631–44. 2017;7:223–30. 33 rao JS, Kahler CB, Baker JB, et al. Protease nexin I, a serpin, inhibits plasminogen- 19 li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler dependent degradation of muscle extracellular matrix. Muscle Nerve 1989;12:640–6. transform. Bioinformatics 2009;25:1754–60. 34 Xiao K, Yang Y, Bian Y, et al. Identification of differentially expressed long noncoding 20 Dobin A, Davis CA, Schlesinger F, et al. StAR: ultrafast universal RNA-seq aligner. RNAs in human knee osteoarthritis. J Cell Biochem 2019;120. Bioinformatics 2013;29:15–21. 35 He J, Xie Q, Xu H, et al. Circular RNAs and cancer. Cancer Lett 2017;396:138–44. 21 Haft DH, DiCuccio M, Badretdin A, et al. RefSeq: an update on prokaryotic genome 36 Thomson DW, Dinger ME. Endogenous microRNA sponges: evidence and controversy. annotation and curation. Nucleic Acids Res 2018;46:D851–D60. Nat Rev Genet 2016;17:272–83. 22 o’Leary NA, Wright MW, Brister JR, et al. Reference sequence (RefSeq) database at 37 cheng X, Zhang L, Zhang K, et al. Circular RNA VMA21 protects against intervertebral NCBI: current status, taxonomic expansion, and functional annotation. Nucleic Acids disc degeneration through targeting miR-200c and X linked inhibitor-of-apoptosis Res 2016;44:D733–45. protein. Ann Rheum Dis 2018;77:770–9. 23 Zhang X-O, Wang H-B, Zhang Y, et al. Complementary sequence-mediated exon 38 Zhong Y, Du Y, Yang X, et al. Circular RNAs function as ceRNAs to regulate and control circularization. Cell 2014;159:134–47. human cancer progression. Mol Cancer 2018;17. 24 Zheng Q, Bao C, Guo W, et al. Circular RNA profiling reveals an abundant circHIPK3 39 Meng S, Zhou H, Feng Z, et al. CircRNA: functions and properties of a novel potential that regulates cell growth by sponging multiple miRNAs. Nat Commun 2016;7. biomarker for cancer. Mol Cancer 2017;16. 25 Memczak S, Jens M, Elefsinioti A, et al. Circular RNAs are a large class of animal RNAs 40 nakamura A, Rampersaud YR, Nakamura S, et al. microRNA-181a-5p antisense with regulatory potency. Nature 2013;495:333–8. oligonucleotides attenuate osteoarthritis in facet and knee joints. Ann Rheum Dis 26 Jeck WR, Sharpless NE. Detecting and characterizing circular RNAs. Nat Biotechnol 2019;78:111–21. 2014;32:453–61. 41 Wu C, Tian B, Qu X, et al. MicroRNAs play a role in chondrogenesis and osteoarthritis 27 li L, Yang C, Liu X, et al. Elevated expression of microRNA-30b in osteoarthritis and its (review). Int J Mol Med 2014;34:13–23. role in ERG regulation of chondrocyte. Biomed Pharmacother 2015;76:94–9. 42 Jones SW, Watkins G, Le Good N, et al. The identification of differentially expressed 28 ohta Y, Okabe T, Larmour C, et al. Articular cartilage endurance and resistance microRNA in osteoarthritic tissue that modulate the production of TNF-alpha and to osteoarthritic changes require transcription factor ERG. Arthritis Rheumatol MMP13. Osteoarthritis Cartilage 2009;17:464–72. 2015;67:2679–90. 43 iwamoto M, Tamamura Y, Koyama E, et al. Transcription factor ERG and joint and 29 Berenbaum F. Osteoarthritis year 2010 in review: pharmacological therapies. articular cartilage formation during mouse limb and spine skeletogenesis. Dev Biol Osteoarthritis Cartilage 2011;19:361–5. 2007;305:40–51.

836 Shen S, et al. Ann Rheum Dis 2019;78:826–836. doi:10.1136/annrheumdis-2018-214786 Pain

Clinical science Are corticosteroid injections needed after needling and lavage of calcific tendinitis? Randomised, double- blind, non-inferiority trial Christelle Darrieutort-Laffite,‍ ‍ 1 Stephane Varin,2 Guillaume Coiffier,3 Jean-David Albert,3 Lucie Planche,4 Yves Maugars,‍ ‍ 1 Grégoire Cormier,2 Benoit Le Goff‍ ‍ 1

Handling editor Josef S Abstract Key messages Smolen Objective steroid injections are common after an ultrasound-guided puncture and lavage (UGPL) of calcific ►► Additional material is What is already known about this subject? published online only. To view tendonitis of the rotator cuff. However, steroids may ►► Steroid injections are common after an please visit the journal online prevent calcification resorption and negatively affect ultrasound-guided puncture and lavage (UGPL) (http://dx.​ ​doi.org/​ ​10.​1136/​ tendon healing. Our study was designed to determine of calcific tendonitis of the rotator cuff. annrheumdis-2018-​ ​214971). whether saline solution was non-inferior to steroids ►► Steroids may prevent calcification resorption 1 in the prevention of acute pain reactions in the week Department of Rheumatology, and negatively affect tendon healing. CHU Nantes, Nantes, France following UGPL. 2Department of Rheumatology, Methods This was a randomised, double-blinded, What does this study add? CHD Vendée, La Roche sur Yon, controlled non-inferiority trial with 12-month follow- France ►► Steroid injection improved pain in the 6 weeks 3 up. We included 132 patients (66 in each group) Department of Rheumatology, following the procedure, and function in the 3 with symptomatic calcification measuring more than CHU de RENNES, Rennes, France months after. 4Biometrics and Statistic 5 mm. Patients received 1 mL of saline or steroid ► Steroids have no significant effect on Platform, CHU Nantes, Nantes, (methylprednisolone 40 mg) in the subacromial bursa ► calcification resorption. France at the end of UGPL. Primary outcome was the maximal ► Non-inferiority of saline when compared with pain during the week following the procedure with a ► Correspondence to steroids could not be established. Pr Benoit Le Goff, prespecified non-inferiority margin of 10 mm (0–100 visual analogue scale). Secondary outcomes included Rheumatology, CHU Nantes, How might this impact on clinical practice or Nantes 44000, France; pain at rest and during activity, function (disabilities future developments? benoit.​ ​legoff@chu-​ ​nantes.​fr of the arm, shoulder and hand score) and radiological ► This study supports the use of steroids evolution of the calcification over the 12-month follow- ► Received 21 December 2018 after puncture and lavage (UGPL) of calcific up. Revised 19 March 2019 tendonitis of the rotator cuff. Accepted 20 March 2019 Results The estimated mean difference in the first Published Online First week’s maximal pain between these two groups was 11 April 2019 11.76 (95% CI 3.78 to 19.75). Steroids significantly improved VAS pain at rest and during activities, as well UGPL may be the treatment of choice for calcific as function at 7 days and 6 weeks. They did not change tendinopathy compared with other non-sur- the rate of calcification resorption, which occurred in gical options, such as extracorporeal shock wave 83% and 74% of patients at 12 months in the saline therapy or subacromial bursa (SAB) steroid and steroid groups. injections.6 Arthroscopic removal of the calcific Conclusion non-inferiority of saline when compared deposit is considered to be a second-line therapy.7 with steroids could not be established. However, steroid Steroids are routinely injected into the SAB after injection improved pain in the 6 weeks following the procedure, and function in the 3 months after, with no UGPL to prevent the acute pain induced by the procedure. In a recent systematic review, Lanza significant effect on calcification resorption. 8 Trial registration number NTC02403856. et al found that this injection was performed in 14 out of the 15 studies they included in their analysis. However, certain authors argue that this injection may also have some deleterious effects. Introduction It has been shown both in vivo and in vitro that Calcific tendinitis of the rotator cuff is one of steroids may negatively affect the healing process © Author(s) (or their 9 the most common causes of shoulder conditions, of the tendon. They are known to lower the employer(s)) 2019. No local immune response, potentially favouring commercial re-use. See rights affecting about 10%–42% of patients referred for and permissions. Published a painful shoulder.1–4 Ultrasound-guided percuta- the occurrence of local infections with a recently by BMJ. neous lavage (UGPL or barbotage or irrigation) of published case report of SAB infection following 10 To cite: Darrieutort- calcific tendinopathy is indicated when conserva- UGPL. Finally, steroids may prevent the inflam- Laffite C, Varin S, Coiffier G, tive treatments (physiotherapy and non-steroidal matory reaction crucial to the disappearance of et al. Ann Rheum Dis anti-inflammatory drugs [NSAIDs]) have failed.5 the calcific deposit after the procedure.11 Of note, 2019;78:837–843. A recent network meta-analysis has shown that all these potential harms remain hypothetical

Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 837 Pain and we lack large prospective series of patients treated with procedure to prevent the needle from being clogged by calcific UGPL to better evaluate their prevalence. debris.15 When backflow of calcific material could be identi- It is therefore important to assess the benefit–risk balance of fied in the syringe, lavage of the deposit was performed with administering steroid injections after UGPL. Our hypothesis saline solution until the backflow became clear. At the end of was that pain after the procedure would not be significantly the procedure, patients received a blinded injection of either worse without steroid injections. However, the absence of 1 mL (40 mg) of methylprednisolone acetate or 1 mL of saline steroid injections may have significant advantages, such as a solution into the SAB depending on their group allocation. reduction in both tendon damage and the risk of infection, as Patients were systematically treated with diclofenac (75 mg LP well as accelerating the disappearance of the calcific deposit. twice daily) and paracetamol (1000 mg, four times a day) for Non-inferiority trials are usually chosen to determine whether 48 hours, and then only if needed in the five following days. a new treatment/procedure is not worse than a reference treat- Routine use of the shoulder was allowed without restriction ment on the premise that it has some other advantages such as and all patients had 1 week off work. reduced cost, less invasiveness, fewer adverse effects or greater ease of administration.12 We thus conducted a randomised, Outcomes and follow-up double-blinded, controlled non-inferiority trial comparing Patients were seen at the rheumatology clinic at screening, UGPL performed with or without subacromial corticosteroid baseline, 7 days, 3 months and 12 months. Additional phone injection. calls were made after 4 days, 6 weeks and 6 months to record any side effects and to remind patients to fill in their VAS Methods for pain and their disabilities of the arm, shoulder and hand Design overview (DASH scores) and post them back to the investigator. Data We conducted a 12-month double-blinded, controlled, multi- were collected using standardised case report forms. centric, non-inferiority trial comparing UGPL performed with Primary outcome was the maximal pain experienced by or without subacromial corticosteroid injections. All partic- the patient in the 7 days following the procedure.16 For this ipating patients were informed about the trial and signed purpose, patients were given a booklet and had to fill in a informed consent forms. No significant changes were made week-long daily diary that asked them to rate their worst pain after trial commencement (​ClinicalTrials.gov).​ intensity on a 0–100 visual analogue scale (VAS) (0=no pain; 100=worst pain imaginable). Secondary outcomes were: (1) Setting and participants average pain at rest and during activities in the last 48 hours The study was performed in three centres in France (Nantes, on a 0–100 NRS; (2) shoulder range of motion; (3) DASH 17 La Roche-sur-Yon and Rennes). Eligibility was assessed during score ; (4) surface of the calcific deposit evaluated on X-ray the screening visit after clinical, ultrasound (US) and X-ray at baseline, 7 days, 3 months and 12 months using a semi- evaluation of the affected shoulder. Eligible patients were quantitative evaluation as follows: no change or minimal aged 18 years or older, reported pain in the shoulder for more changes; decrease in size of the calcification of less than 50%; than 3 months with worsening of symptoms with activities decrease in the calcification of between 50% and 90%; and a above shoulder level; had at least one of the three following more than 90% decrease in size or disappearance of the calci- impingement positive clinical tests (Yocum, Hawkins and fication. X-ray were read in each centre by the physician in Neer); presented a calcification >5 mm in size on the standard charge of the patient, blinded from the treatment allocation anteroposterior radiographs; and were able and willing to give and in known time sequence. Occurrence of adverse events, consent and adhere to the study protocol. Exclusion criteria including infection and tendon tear, was recorded at each visit were other shoulder diseases (glenohumeral or acromioclavic- throughout the study. ular osteoarthritis, rotator cuff tear detected on the initial US evaluation, rheumatoid arthritis and frozen shoulder); previous Statistical analysis percutaneous or surgical irrigation of the same calcification; The study was designed and powered to detect whether SAB steroid injection in the previous month; type C calcifica- 13 UGPL performed without steroid injection was non-inferior tion on X-ray according to Molé’s classification (type A: well for the primary outcome (maximal VAS of pain during the 7 defined, dense and homogeneous; type B: well defined, dense days after the procedure) with regard to the same procedure but multiple/polylobular; type C: ill defined, non-homoge- performed with steroid injections. The non-inferiority margin nous; type D: at tendon insertion/enthesopathy) as this type was prespecified at 10 mm by choosing the minimal clinically of calcification is associated with calcification resorption and 18 14 important difference in the context of shoulder pain. The a favourable spontaneous outcome ; and contraindication for sample size of 136 patients was calculated to be enough (with the use of NSAIDs or paracetamol/acetaminophen. a one-sided 97.5% CI and 80% power) to establish non-in- feriority. The sample size calculation allowed for 5% loss to Randomisation and intervention follow-up. For primary outcome, estimation of the 95% CI Patients were randomly assigned to either steroids (reference) in the difference between the two groups was assessed using or saline (experimental) groups. Center-stratified block-per- a mixed linear model taking into account centre as random muted randomisation (1:1) was computer generated (CSon- effect. Non-inferiority would be accepted if the mean differ- line). Syringes were prepared in a different room by a nurse ence and the upper limit of this CI was less than the non-inferi- and covered with a large sticking plaster to prevent the ority margin set at 10 mm. The analysis of non-inferiority was patient and the investigator from knowing which product was based both on the modified intent to treat and on per protocol injected. All parties remained blinded to treatment allocation populations. If the lower bound of the CI of the difference throughout the trial. is greater than zero, saline would be considered significantly Patients were treated with a US-guided single needle tech- worse than steroids. Evolution in maximal pain during the nique. A 21 G paediatric spinal needle was used for the 7 days after the procedure, the VAS at rest and during daily

838 Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 Pain

Figure 1 CONSORT flow chart. CONSORT, Consolidated Standards of Reporting Trials; NSAIDs, non-steroidal anti-inflammatory drugs; SAB, subacromial bursa. activities and the DASH score were compared between the two in March 2018 after the 12-month follow-up visit of the last groups using a similar linear mixed model analysis. Resorp- patient. tion of the calcification was estimated in the two groups Baseline characteristics were balanced across treatment at 3 months and 12 months. Missing data for the primary groups (table 1). They were mainly female (n=69, 67.4%), outcome were managed as followed: if the pain was recorded mean age 49.8 years (SD ±9.7). Mean symptom duration was from baseline to at least day 3, then the maximal pain was 32 months (±37.9). The tendon treated was most commonly collected during this period. In the other cases, maximal VAS the supraspinatus (n=114; 86.4%), and the calcification was was imputed as a result of multiple imputations. Centre group type A in 54 cases (40.9%). All procedures were free from and pain at baseline were included in the imputation model. any immediate complications except for mild vagal reactions Five data sets were imputed. If not specified, results are given occurring in 12 patients (9%). The lavage was shortened in as mean (95% CI). two cases per group (3%) because of a vasovagal reaction, but the patients still received the allocated intervention.

Results Study population Primary outcome Overall, 134 patients (65%) were randomised (figure 1). No Mean maximal pain during the first week was 71.5 (63.9– patient was excluded because of tendon tear. Two randomised 79.2) in the saline group versus 59.8 (52.2–67.4) in the patients in the steroid group did not receive the procedure (no steroid group in the mITT population and 72.7 (65.4–79.99) calcification but enthesophyte on X-ray for one patient; type C in the saline group versus 60.01 (52.8–67.21) in the steroid calcification for the other) and were not included in the modi- group in the per-protocol population (figure 2). The esti- fied intention to treat analysis (mITT) analysis. Sixty-five and mated mean difference between these two groups was 11.76 63 patients in the steroid and saline groups, respectively, were (3.78–19.75) in the mITT population and 12.69 (5.59–19.79) included in the per-protocol analysis. Most of the participants in the per-protocol population (table 2). As the mean differ- completed the 12-month follow-up (63 [95%] in the steroid ence was greater than 10 mm, with a lower bound of the 95% group and 60 [90%] in the saline group). The trial stopped confidence below this prespecified margin, non-inferiority

Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 839 Pain

Table 1 Demographic and baseline characteristics of the patients included in the mITT population Characteristics Total (n=132) Steroid (n=66) Control (n=66) Mean age (SD) 49.8 (9.7) 47.3 (9.2) 52.2 (9.5) Sex, female, n (%) 89 (67.4) 49 (74.2) 40 (60.6) Physical occupational 55 (42.6) 29 (44.6) 26 (40.6) activities, n (%) Currently off work, n (%) 31 (25.4) 14 (23) 17 (27.9) Previous SAB steroid 84 (63.6) 45 (68.2) 39 (59.1) injection, n (%) Affected side, right/left, n 79/53 35/31 44/22 Duration of symptoms, 32.4 (37.9) 30.8 (34.4) 34.0 (41.3) month (SD) VAS (at rest), 0–100 mm, 32.2 (25.1) 32.3 (24.5) 32 (25.9) mean (SD) VAS (at motion), 0–100 72 (16.2) 69.8 (18.3) 70.2 (14) Figure 2 (A) Maximal pain experienced by the patient during the 7 mm, mean (SD) days following UGPL in the mITT (A) and per protocol population (B). (C) Nocturnal pain, yes, n (%) 108 (81.8) 53 (80.3) 55 (83.3) Non-inferiority plot of steroid versus serum saline injections regarding DASH score, mean (SD) 46.2 (15.6) 48.6 (15.7) 43.8 (15.1) the difference in maximal pain during the 7 days following UGPL X-ray (primary outcome). The non-inferiority margin is shown with dotted bold Tendons affected lines. The lower limit for each CI crosses the prespecified non-inferiority Supraspinatus, n (%) 114 (86.4) 56 (84.8) 58 (87.9) margin. Results of the mITT and per-protocol analysis are presented. Infraspinatus, n (%) 12 (9.1) 7 (10.6) 5 (7.6) Error bars represent the 95% CIs. mITT, modified intention to treat Subscapularis, n(%) 6 (4.5) 3 (4.5) 3 (4.5) analysis; UGPL, ultrasound-guided percutaneous lavage. Largest calcification 17.4±7.4 17.1±6.3 17.7±8.4 size, mm (SD) Molé classification (13) and 73.95 (64.28; 83.63) before the procedure to 32.71 Type A, n (%) 54 (40.9) 24 (36.4) 30 (45.5) (22.89; 42.53) and 20.37 (10.62; 30.13) 7 days after UGPL in the saline and steroid groups, respectively (see online supple- Type B, n (%) 78 (59.1) 42 (63.6) 36 (54.5) mentary table S1). Linear mixed model analysis demonstrated Ultrasound a significant difference between the two groups in favour of Ultrasound appearance steroids (p<0.001). Evolution in pain at rest and during daily Arc shaped with 72 (54.5) 34 (51.5) 38 (57.5) activity over the 12-month follow-up is shown in figure 4 and acoustic shadowing, n (%) summarised in table 2. After an initial decrease at day 7, pain at Fragmented with 37 (28) 23 (34.8) 14 (21.2) rest remained stable in the steroid group, whereas it increased acoustic shadowing, slightly at 6 weeks in the saline group. It eventually dropped n (%) significantly in the two groups at 12 months. Pain during activ- Fragmented without 11 (8.3) 3 (4.5) 8 (12.1) ities significantly decreased in the two groups at day 7 and acoustic shadowing, 6 weeks with a significant difference in favour of the steroid n (%) group at these two time points (p<0.001). After stabilisation at Nodular, n (%) 12 (9) 6 (9) 6 (9) 3 months, the pain gradually decreased in the two groups up to Bursitis, n (%) 37 (28) 18 (27) 19 (29) 12 months. The DASH score followed the same pattern with a Positive Doppler 28 (21.2) 13 (19.7) 15 (22.7) significant decrease in the steroid group at 6 weeks and 3 months signal, n (%) compared with the saline with a gradual decrease in the two Grade 1, n (%) 21 (75) 9 (69.2) 12 (80) groups up to 12 months (table 2). Evolution of shoulder range Grade 2, n (%) 7 (25) 4 (30.8) 3 (20) of motion in the two groups is available as supplementary data Grade 3, n (%) 0 0 0 (see online supplementary table S2). Ultrasound-guided puncture and lavage Radiological changes at 7 days, 3 months and 12 months are Consistence of the 54 (41.5) 29 (45) 25 (34) summarised in table 3. No difference was observed between the calcific deposit, hard, two groups (p=0.28, generalised linear model adjusted for the n (%) centre and time). Aspiration of calcific 107 (81) 53 (80) 54 (82) During follow-up, six patients underwent a further UGPL (two deposit, yes, n (%) in the steroid and four in the saline groups, a median (Q1–Q3) DASH, disabilities of the arm, shoulder and hand; SAB, subacromial bursa; VAS, 4.8 (2.8–11.9) and 4.1 (3.3–4.7) months after inclusion, respec- visual analogue scale; mITT, modified intention to treat. tively). Sixteen patients (6 in the steroid and 10 in the saline groups) needed surgery on their shoulder a median (Q1–Q3) 7.5 was not established. However, as the lower bound remained (4.1–11.2) and 6.7 (5.8–7.5) months after inclusion. One case above 0, we can conclude that saline was significantly inferior of frozen shoulder occurred in the saline group, but no case of compared with steroids. infection or tendon tear was recorded.

Secondary outcomes Discussion Evolution in maximal daily pain during the first week is shown Our study was a non-inferiority randomised trial designed to in figure 3. Maximal pain decreased from 73.97 (64.29; 83.66) assess the usefulness of steroid injection after UGPL of rotator

840 Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 Pain

Table 2 Estimated treatment differences in primary and secondary outcomes during the 12-month follow-up Steroids Saline Mean score Mean difference Analysis and follow-up Patients Mean score (95% CI) Patients (95% CI) (95% CI) Primary outcome Maximal pain during the first week mITT population n=66 59.8 (52.2 to 67.4) n=66 71.5 (63.9 to 79.2) 11.76 (3.8 to 19.7) Per -protocol population n=65 60.01 (52.80 to 67.21) n=63 72.70 (65.41 to 79.99) 12.69 (5.59 to 19.79) Secondary outcome VAS activity Inclusion n=66 72.78 (61.97 to 83.59) n=66 73.26 (62.44 to 84.07) 0.48 (-8.51 to 9.46) Day 7 n=65 27.39 (16.58 to 38.20) n=66 49.06 (38.21 to 59.92) 21.67 (12.65 to 30.69) W6 n=64 33.05 (22.09 to 44.01) n=65 52.28 (41.33 to 63.23) 19.23 (9.92 to 28.54) M3 n=64 44.03 (33.16 to 54.89) n=65 51.91 (40.99 to 62.82) 7.88 (-1.28 to 17.04) M6 n=64 38.49 (27.34 to 49.64) n=65 40.39 (29.10 to 51.67) 1.90 (-7.99 to 11.80) M12 n=60 27.57 (16.67 to 38.47) n=63 31.14 (20.12 to 42.17) 3.57 (-5.75 to 12.89) VAS at rest Inclusion n=66 35.62 (23.47 to 47.77) n=66 35.59 (23.44 to 47.75) −0.03 (-8.17 to 8.11) Day 7 n=65 14.78 (2.63 to 26.93) n=66 26.07 (13.89 to 38.26) 11.29 (3.12 to 19.47) W6 n=64 26.33 (14.07 to 38.59) n=65 43.04 (30.79 to 55.29) 16.71 (8.28 to 25.15) M3 n=64 21.55 (9.36 to 33.75) n=65 28.16 (15.93 to 40.38) 6.60 (-1.70 to 14.90) M6 n=64 31.63 (19.23 to 44.03) n=65 31.59 (19.09 to 44.10) −0.04 (-9.00 to 8.93) M12 n=60 12.55 (0.33 to 24.77) n=63 12.20 (-0.11 to 24.51) −0.35 (-8.79 to 8.09) DASH score Inclusion n=66 50.94(42.70 to 59.19) n=66 46.12 (37.88 to 54.36) −4.82 (-11.7 to 2.06) W6 n=64 30.32(22.04 to 38.61) n=65 42.26 (33.95 to 50.58) 11.94 (4.92 to 18.96) M3 n=64 31.95(23.64 to 40.26) n=65 39.91 (31.60 to 48.23) 7.96 (0.91 to 15.01) M6 n=64 33.02(24.60 to 41.43) n=65 33.36 (24.95 to 41.77) 0.35 (-6.92 to 7.61) M12 n=60 23.65(15.39 to 31.91) n=63 22.33 (13.95 to 30.71) −1.32 (-8.38 to 5.74) DASH, disabilities of the arm, shoulder and hand; M, month; VAS, visual analogue scale; W, week;miTT, modified intention to treat analysis. cuff calcifications. The potential harm caused by steroids with with steroids in the prevention of acute pain reactions after regard to tendon healing, as well as their negative effects on UGPL could not be established. On the contrary, saline turned calcific deposit resorption, raised questions about their system- out inferior. Indeed, the lower bound of the 95% CI was above atic use in daily practice. The non-inferiority of saline compared zero for the primary outcome as for other secondary outcomes such as pain or function. Whether that is clinically relevant is still under undetermined, as the lower boundary of the 95% CI falls within the prespecified non-inferiority margin selected for its clinical relevance. Finally, there was no difference in the radiographic evolution of the calcification at 7 days, 3 months and 12 months with the same rate of resorption in both groups. No study has demonstrated the usefulness of steroid injec- tions on the prevention of pain following UGPL. Our study is the first to focus on pain in the 7 days following UGPL. We observed a decrease in pain in both groups but with a signifi- cant difference in favour of steroids. Interestingly, there was no exacerbation of pain in the saline group. This shows that UGPL could also be performed without steroids if needed (contrain- dications, diabetes and patients’ will) without an increased risk of pain exacerbation. Of note, our patients were systematically treated with NSAIDs the 2 days following the procedure, then again, if needed, during the first week. This treatment may have decreased the risk of pain exacerbation and should be given in Figure 3 Evolution in maximal daily pain during the 7 days following case of no steroid use. UGPL (A) (p<0.001; linear mixed model analysis). Evolution in the VAS We observed a significant difference in pain intensity in the for pain at rest (B) and during activity (C) plus (D) the DASH score in first 6 weeks and improvement in shoulder function in the steroid the saline and steroid groups over the 12-month follow-up. Error bars group during the 3 months following the procedure. Interest- represent the 95% CIs. DASH, disabilities of the arm, shoulder and hand; ingly, a trend towards an exacerbation of pain was observed at 6 UGPL, ultrasound-guided percutaneous lavage; VAS, visual analogue weeks in the saline and in the steroid group. An increase in pain scale. after UGPL has already been described in other studies.5 19 del

Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 841 Pain

corresponding to the minimal relevant clinical threshold asso- Table 3 Radiological outcome at 7 days, 3 months and 12 months ciated with an improvement in shoulder pain.18 However, no Patient, n Steroids Patient, n Saline data are available to define what minimal threshold is clinically Day 7 meaningful in the context of acute postprocedural pain. Another No or minor changes, 65 24 (36.9) 63 17 (27.0) limitation is that steroids or saline were injected under real-time n (%) US guidance in the SAB. glucocorticoïd (GC) injections are often Less than 50% 22 (33.8) 23 (36.5) seen in the US as hyperechoic effusions that could have been decrease in size, n (%) differentiated from the saline by the investigator. However, the Between 50% and 14 (21.5) 16 (25.4) air contained in the syringes also leads to hyperechoic artefacts 90% decrease in size, when injected. Therefore, it was not possible to guess whether n (%) steroids or saline were injected. Finally, tendon tear and calcifi- Diseaperance or more 5 (7.7) 7 (11.1) cation can coexist. This has been described in 4% by US to 28% than 90% decrease in 21 size, n (%) of the patients with arthrography. We excluded patients with Month 3 tendon tear to avoid confounding as pain could be due to the tear or the calcific deposit itself. In our study, no patient has been No or minor changes, 63 10 (15.9) 60 10 (16.7) n (%) found with both diseases. This can be explained by the young Less than 50% 14 (22.2) 11 (18.3) age of our patients or the lesser sensitivity of US. Our study also decrease in size, n (%) has several strengths. The double-blinded design decreased the Between 50% and 10 (15.9) 15 (25.0) risk of placebo or Hawthorne effects, known to influence pain 90% decrease in size, assessment. The long-term follow-up and the low number of n (%) patients lost to follow-up allowed us to study the long-term effect Diseaperance or more 29 (46.0) 24 (40.0) of steroids on pain, function and also radiographic evolution. than 90% decrease in Overall, our study shows that steroids are beneficial after size, n (%) UGPL, decreasing pain and disability in the short term without Month 12 any effect on the rate of calcification resorption. Our findings No or minor changes, 62 4 (6.5) 59 2 (3.4) support the current practice of steroid injections in patients with n (%) calcification of the rotator cuff. Less than 50% 3 (4.8) 1 (1.7) decrease in size, n (%) Acknowledgements The authors would like to thank Dr Claire Daien and Pr Alain Saraux for their help in the preparation of the final manuscript. Between 50% and 9 (14.5) 7 (11.9) 90% decrease in size, Contributors CD-L designed the study, included patients, analysed the data and n (%) wrote the mansucript. SV, GC, J-DA, YM and GC included patients and analysed the data. LP designed the study and analysed the data. BLG designed the study, included Diseaperance or more 46 (74.2) 49 (83.1) patients, analysed the data and wrote the mansucript. than 90% decrease in size, n (%) Funding The study received funding and support from the French Ministry of Health PHRC-i (Programme Hospitalier de Recherche inter-régional). Disclaimer The funding organisations had no role in how the study was conducted, Cura et al showed that this recurrence occurred after a mean of manuscript preparation or the decision to submit the manuscript for publication. 15 weeks (5–28 weeks) and lasted for a mean of 6 weeks (2–20 Competing interests None declared. weeks). It is thought to be associated with the resorption of the Patient consent for publication Not required. calcification induced by the puncture of the calcific deposit. Ethics approval The research protocol was approved by the local ethics committee In our study, although steroids decreased the intensity of the (CPP) and the French National Agency for Medicines and Health Products Safety pain, they did not prevent this painful phase. Strategies making (Registration number RC15_0019). possible faster removal of the calcific deposit may be of interest Provenance and peer review Not commissioned; externally peer reviewed. for preventing or shortening this recurrence of pain. We observed the same rate of calcification resorption at 7 days, Data sharing statement Data are available upon reasonable request. 3 months and 12 months after UGPL in both groups. We found that 95% and 89% of patients had a more than 50% decrease References 1 Farin PU, Jaroma H. Sonographic findings of rotator cuff calcifications. J Ultrasound in the size of their calcification at 12 months in the saline and Med 1995;14:7–14. steroid groups, respectively. This shows that steroids do not 2 Friedman MS. Calcified tendinitis of the shoulder. Am J Surg 1957;94:56–61. interfere with the mechanisms leading to calcification resorp- 3 Harmon PH. Methods and results in the treatment of 2,580 painful shoulders, tion. Although the inflammatory reaction induced by apatite with special reference to calcific tendinitis and the frozen shoulder. Am J Surg crystals is thought to be important, one injection of steroids in 1958;95:527–44. 4 louwerens JKG, Sierevelt IN, van Hove RP, et al. Prevalence of calcific deposits the SAB seems to have no significant effect on this process. This within the rotator cuff tendons in adults with and without subacromial pain is of importance as some authors argued against steroid injec- syndrome: clinical and radiologic analysis of 1219 patients. J Shoulder Elbow Surg tions because of their potential negative effect on calcification 2015;24:1588–93. resorption.11 5 de Witte PB, Selten JW, Navas A, et al. Calcific tendinitis of the rotator cuff: a randomized controlled trial of ultrasound-guided needling and lavage versus Our study has several limitations. We hypothesised that steroid subacromial corticosteroids. Am J Sports Med 2013;41:1665–73. injections were not necessary after the procedure and could be 6 arirachakaran A, Boonard M, Yamaphai S, et al. Extracorporeal shock wave therapy, harmful to tendon healing and the resorption of the calcifica- ultrasound-guided percutaneous lavage, corticosteroid injection and combined tion. For this purpose, we designed a non-inferiority trial recom- treatment for the treatment of rotator cuff calcific tendinopathy: a network meta- mended when a new procedure is thought to be equivalent to analysis of RCTs. Eur J Orthop Surg Traumatol 2017;27:381–90. 20 7 louwerens JKG, Veltman ES, van Noort A, et al. The effectiveness of high-energy the reference and may have other advantages. We were unable extracorporeal shockwave therapy versus ultrasound-guided Needling versus to confirm this hypothesis as non-inferiority could not be estab- arthroscopic surgery in the management of Chronic calcific rotator cuff tendinopathy: lished. We chose a stringent non-inferiority margin of 10 mm, a systematic review. Arthroscopy 2016;32:165–75.

842 Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 Pain

8 lanza E, Banfi G,S erafini ,G et al. Ultrasound-guided percutaneous irrigation in 16 Gupta A, Kaur K, Sharma S, et al. Clinical aspects of acute post-operative pain rotator cuff calcific tendinopathy: what is the evidence?A systematic review with management & its assessment. J Adv Pharm Technol Res 2010;1:97–108. proposals for future reporting. Eur Radiol 2015;25:2176–83. 17 Dubert T, Voche P, Dumontier C, et al. [The DASH questionnaire. French translation of a 9 Maman E, Yehuda C, Pritsch T, et al. Detrimental effect of repeated and single trans-cultural adaptation]. Chir Main 2001;20:294–302. subacromial corticosteroid injections on the intact and injured rotator cuff: a 18 Hawker GA, Mian S, Kendzerska T, et al. Measures of adult pain: visual analog biomechanical and imaging study in rats. Am J Sports Med 2016;44:177–82. scale for pain (VAS pain), numeric rating scale for pain (NRS pain), McGill pain 10 sconfienza LM, Randelli F, Sdao S, et al. Septic bursitis after ultrasound-guided percutaneous treatment of rotator cuff calcific tendinopathy. Pm R 2014;6:746–8. questionnaire (MPQ), short-form McGill pain questionnaire (SF-MPQ), chronic pain 11 Re LP, Karzel RP. Management of rotator cuff calcifications. Orthop Clin North Am grade scale (CPGS), short Form-36 bodily Pain Scale (SF-36 BPS), and measure 1993;24:125–32. of intermittent and constant osteoarthritis pain (ICOAP). Arthritis Care Res 12 schumi J, Wittes JT. Through the looking glass: understanding non-inferiority. Trials 2011;63:S240–S252. 2011;12. 19 del Cura JL, Torre I, Zabala R, et al. Sonographically guided percutaneous needle 13 Molé D, Kempf JF, Gleyze P, et al. [Results of endoscopic treatment of non-broken lavage in calcific tendinitis of the shoulder: short- and long-term results. AJR Am J tendinopathies of the rotator cuff. 2. Calcifications of the rotator cuff]. Rev Chir Roentgenol 2007;189:W128–W134. Orthop Reparatrice Appar Mot 1993;79:532–41. 20 Piaggio G, Elbourne DR, Pocock SJ, et al. Reporting of noninferiority and 14 Rupp S, Seil R, Kohn D. [Tendinosis calcarea of the rotator cuff]. Orthopade 2000;29:852–67. equivalence randomized trials: extension of the CONSORT 2010 statement. JAMA 15 Jelsing EJ, Maida E, Smith J. A simple technique to restore needle patency during 2012;308:2594–604. percutaneous lavage and aspiration of calcific rotator cuff tendinopathy. Pm R 21 Hsu HC, Wu JJ, Jim YF, et al. Calcific tendinitis and rotator cuff tearing: a clinical and 2013;5:242–4. radiographic study. J Shoulder Elbow Surg 1994;3:159–64.

Darrieutort-Laffite C, et al. Ann Rheum Dis 2019;78:837–843. doi:10.1136/annrheumdis-2018-214971 843 Epidemiology

Epidemiological science The world-wide burden of musculoskeletal diseases: a systematic analysis of the World Health Organization Burden of Diseases Database Eden Sebbag,1 Renaud Felten,1 Flora Sagez,1 Jean Sibilia,1 Hervé Devilliers,2 Laurent Arnaud‍ ‍ 1

Handling editor Josef S Abstract Key messages Smolen Background musculoskeletal (MSK) diseases are expected to have a growing impact worldwide. ►► Additional material is What is already known about this subject? published online only. To view Objective To analyse the worldwide burden of MSK ►► The burden of musculoskeletal (MSK) diseases is please visit the journal online diseases from 2000 to 2015. expected to increase worldwide (http://​dx.​doi.​org/​10.​1136/​ Methods disability-adjusted life years (DALYs), annrheumdis-​2019-​215142). which combines the years of life lost (YLLs) and the What does this study add? 1 years lived with disability (YLDs), were extracted for Department of Rheumatology ► The burden of MSK diseases has increased 183 countries from the WHO Global Health Estimates ► - Centre National de Référence significantly between 2000 and 2015, with MSK des Maladies Systémiques Rares Database. We analysed the median proportion of diseases being the second cause of years lived Est Sud-Ouest (RESO), Hôpitaux DALYS, YLLs and YLDs for MSK diseases (ICD-10: Universitaires de Strasbourg, with disability (YLDs) worldwide M00–M99) among the 23 WHO categories of diseases. Strasbourg, France ►► The burden of MSK diseases is significantly 2 Mixed models were built to assess temporal changes. Internal Medicine and Systemic higher in Europe than in all other continents. Diseases, Hôpital François Results Worldwide, the total number of MSK DALYs ► MSK burden is strongly correlated with Mitterrand, CHU de Dijon, Dijon, increased significantly from 80,225,634.6 in 2000 to ► countries' gross domestic product per capita France 107,885,832.6 in 2015 (p < 0.001), with the total number of MSK YLDs increasing from 77,377,709.4 to Correspondence to How might this impact on clinical practice or 103,817,908.4 (p = 0.0008) and MSK diseases being Professor Laurent Arnaud, future developments? Department of Rheumatology - the second cause of YLDs worldwide. YLLs due to MSK ►► These results are crucial to inform health Centre National de Référence diseases increased from 2,847,925.2 to 4,067,924.2 professionals, policy makers and national des Maladies Systémiques Rares (p = 0.03). In 2015, the median proportion of DALYs healthcare systems for the implementation of Est Sud-Ouest (RESO), Hôpitaux attributed to MSK diseases was 6.66% (IQR: 5.30 – Universitaires de Strasbourg, future health-plan adjustments. Strasbourg 67085, France; 7.88) in Europe versus 4.66% (3.98 – 5.59) in the ​laurent.​arnaud@​chru-​ Americas (p < 0.0001 vs Europe), 4.17% (3.14 – 6.25) strasbourg.​fr in Asia (p < 0.0001), 4.14% (2.65 – 5.57) in Oceania (p = 0.0008) and 1.33% (1.03 – 1.92) in Africa (p Received 28 January 2019 osteoarthritis and low back pain are significantly Revised 26 February 2019 < 0.0001). We observed a significant correlation (r more common than most inflammatory diseases.9 Accepted 28 February 2019 = 0.85, p < 0.0001) between the proportion of MSK The economic burden of MSK diseases has been Published Online First DALYs and the gross domestic product per capita for well described,5 but national healthcare systems 15 April 2019 the year 2015. tend to underestimate the role of MSK diseases Conclusions The burden of MSK diseases increased owing to their low death rate.10 Importantly, significantly between 2000 and 2015 and is high in the overall MSK burden is expected to increase, Europe. These results are crucial to health professionals owing to the increase in average life expectancy11 and policy makers to implement future health plan and because many degenerative MSK diseases are adjustments for MSK diseases. 10 considered largely irreversible. The low death rate and irreversibility may reduce the relative importance of MSK diseases to national health- care systems and policy makers, compared with Introduction other categories of diseases such as cancers or Musculoskeletal (MSK) diseases, defined as cardiovascular diseases. diseases which affect the locomotor system This study aimed to analyse the global burden including muscles, bones, joints, tendons and liga- of MSK diseases compared with other causes ments, have a growing impact worldwide.1 This of morbidity-mortality, and its change over impact is measurable using disability-adjusted life time, using data from a large publicly available © Author(s) (or their 12 employer(s)) 2019. No years (DALYs), which combine the years lived with WHO database. We also sought to analyse the commercial re-use. See rights disability (YLDs) and the years of life lost (YLLs) socioeconomic determinants of this burden, by and permissions. Published through premature death. Previous studies have analysing its association with countries' gross by BMJ. 13 suggested a high prevalence and high disability domestic product (GDP) per capita. These To cite: Sebbag E, Felten R, linked to selected MSK diseases.2–8 However, results are crucial to rheumatologists, enabling Sagez F, et al. Ann Rheum Dis representative data for the impact of MSK a better understanding of the burden associated 2019;78:844–848. diseases as a whole are lacking, considering that with MSK diseases, but also vital for policy makers

844 Sebbag E, et al. Ann Rheum Dis 2019;78:844–848. doi:10.1136/annrheumdis-2019-215142 Epidemiology and national healthcare systems to implement adjustments to Table 1 Median proportions of DALYs, YLDs, and YLLs due to MSK future health plans. diseases more than all other causes in the world for the years 2000, 2005, 2010, 2015 Methods Type of DALYs, YLLs and YLDs burden Year 2000 Year 2005 Year 2010 Year 2015 The DALY Index is a summary measure which combines time lost through premature death, YLLs, and time lived in states of DALYs due to 3.3 (1.3–4.7) 3.7 (1.4–5.0) 3.9 (1.9–5.5) 4.3 (2.0–6.0) MSK diseases, less than optimal health, loosely referred to as 'disability', year 14 % median (IQR lived with disability (YLDs). One DALY can be thought of 25–75) as one lost year of 'healthy' life. DALYs for a specific cause are YLDs due to 11.8 (8.3–15.1) 12.6 (8.8–15.6) 13.1 (9.4–16.3) 13.5 (9.6–16.6) calculated as the sum of the YLLs from that cause and the YLDs MSK diseases, for people living in states of less than good health resulting from % median (IQR a specific cause. The methodological details for calculation of 25–75) the estimates used in this study are available online.15 YLLs due to 0.1 (0–0.3) 0.1 (0.1–0.2) 0.1 (0.1–0.3) 0.2 (0.1–0.4) MSK diseases, % median (IQR Data sources and comparison of MSK with other categories 25–75) of disease DALYs, disability-adjusted life years;MSK, musculoskeletal; YLDs, years lost due to We extracted the DALYs, YLDs and YLLs for the 183 countries disability; YLLs, years of life lost. from the WHO Global Health Estimates Database (publicly avail- able online12). This provides detailed information about those three indicators from the year 2000 onwards. In this database, proportion of MSK DALYs increasing significantly from 3.3 conditions are aggregated in 23 categories (eg, infectious and (1.3–4.7) in 2000 to 4.3 (2.0–6.0) in 2015, p<0.0001. The first parasitic diseases, cardiovascular diseases, malignant neoplasms, cause of DALYs in the world changed from infectious diseases musculoskeletal diseases, etc.) based on the International Clas- in 2000 (601 421 046 DALYs), 2005 (534,695,389 DALYs) and sification of Diseases 10th revision (ICD-10). MSK diseases are 15 2010 (420 621 169 DALYs) to cardiovascular diseases in 2015 defined as the ICD-10 codes M00 to M99. To analyse the socioeconomic determinants of MSK diseases, (407 636 500 DALYs). we correlated each country DALYs with its GDP per capita (in MSK diseases remained the second cause of YLDs behind US dollars), available from a publicly available source.13 'mental and substance use disorders' (table 3) from the years 2000 to 2015, but the total number of YLDs attributed to MSK Statistical analysis diseases worldwide increased significantly from 77,377,709.4 The data extracted from the WHO burden database were used to 103,817,908.4 (p=0.0008) during the same period. Between to compute the ratio of the proportion of DALYs, YLDs and 2000 and 2015, the YLLs due to MSK diseases increased from YLLs for MSK diseases to total DALYs, YLDs and YLLs, respec- 2,847,925.24,067,924.2 to (p=0.03) but remained ranked as tively, for each of the 183 countries. We then computed the the 19th disease category out of 23. median and 25–75th centiles interquartile range (IQR 25–75) proportion of DALYs, YLDs and YLLs, respectively, for all coun- D ALYs, YLDs and YLLs for MSK diseases (by country) tries and five continents. For the years 2000, 2005, 2010 and Detailed data for DALYs, YLDs, and YLLs by country are 2015, we ranked the proportion of DALYs, YLLs and YLDs for each category of conditions for each country and each conti- shown in table 4 and online supplementary appendix 1. In nent. To analyse temporal trends in the absolute number and 2015, the country with the highest proportion of DALYs due proportions of DALYs, YLLs and YLDs between 2000 and 2015, to MSK diseases was Australia, with 9.8% of total DALYs for we built mixed models using MSK DALYs, YLLs or YLDs (or the country, 19.6% of total YLDs and 0.6% of total YLLs due their proportions) as the dependent variable, taking the years to MSK diseases. Conversely, the country with the lowest and the continents as the independent variables. The interaction proportion of MSK DALYs in 2015 was Somalia, with 0.6% between year and continent was tested and kept in the model of total DALYs, 5.2% of YLLs and 0% of YLDs due to MSK if significant. Spearman's test was used to study the correlation diseases. We observed a significant correlation (r=0.85, between the proportion of MSK DALYs for each country and p<0.0001) between the proportion of MSK DALYs and coun- the GDP per capita of those countries. All statistical tests were tries' GDP per capita for the year 2015 (figure 1). two-sided with an α risk set at 0.05. Statistical analyses were performed using SAS 9.4 (Cary, North Carolina, USA). D ALYs, YLDs and YLLs for MSK diseases (by continent) Results Median proportions of DALYs, YLDs and YLLs for MSK DALYs, YLDs and YLLs for musculoskeletal diseases (world diseases by continent are shown in table 5. In 2015, the wide) median proportion of DALYs attributed to MSK diseases was The total number of DALYs attributed to MSK diseases world- 6.66% (IQR: 5.30–7.88) in Europe, a significantly higher wide increased from 80,225,634.6 in 2000107,885,832.6 to figure than in all other continents, with 4.66% (3.98–5.59) DALYs in 2015 (p<0.001). The detailed proportions of DALYs, in the Americas (p<0.0001 vs Europe), 4.17% (3.14–6.25) in YLLs and YLDs for MSK diseases for the years 2000, 2005, Asia (p<0.0001), 4.14% (2.65–5.57) in Oceania (p=0.0008) 2010, 2015 are shown in table 1. and 1.33% (1.03–1.92) in Africa (p<0.0001). Interaction with Among the 23 main categories of conditions reported by the time was significant for the proportions of DALYs, YLDs and WHO (see 'Methods'), MSK DALYs were ranked 10th in 2000 YLLs, indicating an increase in the difference between coun- and ninth in 2005, 2010 and 2015 (table 2), with the median tries across those years.

Sebbag E, et al. Ann Rheum Dis 2019;78:844–848. doi:10.1136/annrheumdis-2019-215142 845 Epidemiology

Table 2 Ranking of the 23 categories of DALYs in the world for the Table 3 Ranking of the 23 categories of DALYs, YLLs and YLDs years 2000, 2005, 2010 and 2015 (by decreasing proportion over all according to the WHO in the world, for the year 2015 (by decreasing causes) proportion) Year 2000 Year 2005 Year 2010 Year 2015 D ALYs YLDs YLLs 1.Infectious and 1.Infectious and 1.Infectious and 1.Cardiovascular 1.Cardiovascular diseases 1.Mental and substance use 1.Cardiovascular diseases parasitic diseases parasitic diseases parasitic diseases diseases disorders 2.Cardiovascular 2.Cardiovascular 2.Cardiovascular 2.Infectious and 2.Infectious and parasitic 2.Musculoskeletal diseases 2.Infectious and parasitic diseases diseases diseases parasitic diseases diseases diseases 3.Neonatal conditions 3.Neonatal 3.Neonatal 3.Malignant 3.Malignant neoplasms 3.Sense organ diseases 3.Malignant neoplasms conditions conditions neoplasms 4.Neonatal conditions 4.Nutritional deficiencies 4.Neonatal conditions 4.Unintentional 4.Unintentional 4.Malignant 4.Neonatal 5.Unintentional injuries 5.Neurological conditions 5.Unintentional injuries injuries injuries neoplasms conditions 6.Mental and substance 6.Infectious and parasitic diseases 6.Respiratory Infectious 5.Respiratory 5.Malignant 5.Unintentional 5.Unintentional use disorders Infectious neoplasms injuries injuries 7.Respiratory Infectious 7.Unintentional injuries 7.Respiratory diseases 6.Malignant 6.Respiratory 6.Respiratory 6.Mental and neoplasms Infectious Infectious substance use 8.Respiratory diseases 8.Diabetes mellitus 8.Digestive diseases disorders 9.Musculoskeletal 9.Respiratory diseases 9.Intentional injuries 7.Mental and 7.Mental and 7.Mental and 7.Respiratory diseases substance use substance use substance use Infectious 10.Digestive diseases 10.Cardiovascular diseases 10.Congenital anomalies disorders disorders disorders 11.Neurological 11.Genitourinary diseases 11.Diabetes mellitus 8.Respiratory diseases 8.Respiratory 8.Respiratory 8.Respiratory conditions diseases diseases diseases 12.Nutritional deficiencies 12.Skin diseases 12.Genitourinary diseases 9.Nutritional 9.Musculoskeletal 9.Musculoskeletal 9.Musculoskeletal 13.Intentional injuries 13.Oral conditions 13.Neurological conditions deficiencies diseases diseases diseases 14.Diabetes mellitus 14.Digestive diseases 14.Nutritional deficiencies 10.Musculoskeletal 10.Nutritional 10.Digestive diseases 10.Digestive diseases diseases deficiencies 15.Congenital anomalies 15.Neonatal conditions 15.Endocrine, blood, immune disorders 11.Intentional injuries 11.Digestive diseases 11.Nutritional 11.Neurological deficiencies conditions 16.Sense organ diseases 16.Endocrine, blood, immune 16.Maternal conditions disorders 12.Digestive diseases 12.Intentional 12.Neurological 12.Nutritional injuries conditions deficiencies 17.Genitourinary diseases 17.Congenital anomalies 17.Mental and substance use disorders 13.Congenital 13.Neurological 13.Intentional 13.Intentional anomalies conditions injuries injuries 18.Endocrine, blood, 18.Malignant neoplasms 18.Other neoplasms immune disorders 14.Neurological 14.Congenital 14.Congenital 14.Diabetes mellitus conditions anomalies anomalies 19.Skin diseases 19.Respiratory Infectious 19.Musculoskeletal diseases 15.Sense organ 15.Sense organ 15.Diabetes mellitus 15.Congenital diseases diseases anomalies 20.Maternal conditions 20.Intentional injuries 20.Skin diseases 16.Genitourinary 16.Diabetes mellitus 16.Sense organ 16.Sense organ 21.Oral conditions 21.Maternal conditions 21.Sudden infant death diseases diseases diseases syndrome 17.Diabetes mellitus 17.Genitourinary 17.Genitourinary 17.Genitourinary 22.Other neoplasms 22.Other neoplasms 22.Oral conditions diseases diseases diseases 23.Sudden infant death 23.Sudden infant death syndrome 23.Sense organ diseases 18.Endocrine, blood, 18.Endocrine, blood, 18.Endocrine, blood, 18.Endocrine, blood, syndrome immune disorders immune disorders immune disorders immune disorders DALYs, disability-adjusted life years; YLDs, years lived with disability; YLLs, years of life lost. 19.Maternal 19.Maternal 19.Maternal 19.Skin diseases conditions conditions conditions a significant increase in the proportion of MSK-related YLDs 20.Skin diseases 20.Skin diseases 20.Skin diseases 20.Maternal conditions (p=0.0008). Importantly, in 2015 MSK diseases were the second 21.Oral conditions 21.Oral conditions 21.Oral conditions 21.Oral conditions cause of YLDs (>13% of total YLDs) in the world, after psychi- 22.Other neoplasms 22.Other neoplasms 22.Other neoplasms 22.Other neoplasms atric disorders, whereas they remained the 19th cause of YLLs 23.Sudden infant 23.Sudden infant 23.Sudden infant 23.Sudden infant out of 23. This further confirms the overall low fatality of MSK death syndrome death syndrome death syndrome death syndrome diseases, with the largest proportion of MSK DALYs being due DALYs, disability-adjusted life years. to YLDs. Considering aggregated worldwide data, the first cause of DALYs changed from infectious diseases in 2010 to cardiovas- Discussion cular diseases in 2015. This may be explained by the gener- In this study, we analysed the burden due to MSK diseases using ally better prognosis of infectious diseases and the increase in WHO data available for the years 2000, 2005, 2010 and 2015. Our average life expectancy, especially in developing countries.16 main findings are a significant increase of both the total number The burden of MSK diseases as assessed using DALYs is approx- and proportion of DALYs due to MSK diseases over the past 15 imately one-third of that of cardiovascular diseases, which was years, with Europe being the continent where MSK diseases have the first cause of DALYs worldwide in 2015. the highest impact, compared with all other categories of DALYs. Importantly, Europe is the continent with the highest proportion We further confirmed that the burden of MSK diseases is essentially of MSK DALYs and YLDs: 6.66% (5.30–7.88) and 17.6% (16.9– due to disability (YLDs) rather than to premature death (YLLs). 18.75), respectively. In 2015, the median proportions of DALYs Finally, we observed a strong correlation between MSK disease and YLDs due to MSK diseases in Europe were respectively five burden and GDP per capita, suggesting an increased burden of times and twice higher than those of Africa (table 5). Europe is also MSK diseases in high-income countries. the continent with the highest proportion of MSK YLDs. Approx- The overall burden of MSK diseases, as assessed using the imately one-fifth of Europe YLDs are due to MSK diseases (17.6% proportion of MSK DALYs, has been increasing significantly [16.9–18.75]) while at the same time YLLs remain much lower between 2000 and 2015 (p<0.0001), essentially owing to (0.3% [0.1–0.45]). Infectious diseases are still the leading causes of

846 Sebbag E, et al. Ann Rheum Dis 2019;78:844–848. doi:10.1136/annrheumdis-2019-215142 Epidemiology

Table 4 World ranking of countries according to MSK DALYs, for the years 2000 and 2015 (by decreasing proportion) Year 2000 Year 2015 Rank Country DALYs n (%) YLDs n (%) YLLs n (%) Country DALYs n (%) YLDs n (%) YLLs n (%) 1 Australia 415.9 (8.8) 402.5 (19.9) 13.4 (0.5) Australia 526.8 (9.8) 509.8 (19.6) 17 (0.6) 2 Canada 635.6 (8.4) 612.6 (19.9) 23 (0.5) Canada 821.6 (9.7) 795.6 (21.0) 25.9 (0.5) 3 Netherlands 345.7 (7.9) 332.4 (20.7) 13.3 (0.5) United Arab Emirates 142.6 (9.6) 141.3 (16.3) 1.3 (0.2) 4 Switzerland 143.7 (7.6) 138.4 (18.7) 5.3 (0.5) Bahrain 21.4 (9.2) 20.9 (16.4) 0.5 (0.5) 5 New Zealand 72.7 (7.5) 69.2 (18.0) 3.5 (0.6) Switzerland 171.6 (8.9) 165.6 (19.6) 6.0 (0.6) 6 Norway 95.6 (7.5) 91.7 (19.4) 3.9 (0.5) Norway 107.5 (8.8) 104.2 (19.6) 3.3 (0.5) 7 Italy 1198.9 (7.4) 1169.4 (19.4) 29.5 (0.3) New Zealand 93.3 (8.8) 89.5 (19.0) 3.8 (0.6) 8 Iceland 4.4 (7.3) 4.3 (17.1) 0.1 (0.3) Netherlands 380.9 (8.8) 367.8 (20.6) 13.1 (0.5) 9 United Arab Emirates 37.9 (7.1) 37.4 (13.6) 0.5 (0.2) Denmark 130.5 (8.7) 125.5 (21.0) 5.0 (0.6) 10 Malta 6.8 (7.0) 6.5 (17.5) 0.3 (0.5) Iceland 5.6 (8.6) 5.5 (18.1) 0.1 (0.4) …… … … … … … … … 173 Mozambique 110.4 (0.5) 103.3 (5.3) 7.2 (0) Nigeria 1530.0 (1.0) 1427.7 (8.3) 102.3 (0.1) 174 Niger 74.8 (0.5) 72.8 (7.1) 2.0 (0) Central African Republic 43.4 (1.0) 41.4 (7.9) 2.1 (0.1) 175 Zambia 66.1 (0.5) 62.1 (6.2) 3.9 (0) Burundi 70.1 (1.0) 66.1 (7.8) 4.0 (0.1) 176 Angola 109.1 (0.5) 101.4 (6.6) 7.7 (0) Niger 126.8 (0.9) 123.5 (7.6) 3.3 (0) 177 Malawi 66.3 (0.5) 64.1 (5.6) 2.2 (0) South Sudan 81.7 (0.9) 77.9 (6.1) 3.8 (0) 178 Rwanda 44.9 (0.5) 42.1 (4.8) 2.9 (0) Mozambique 176.1 (0.9) 165.3 (6.1) 10.8 (0.1) 179 Mali 67.3 (0.5) 64.3 (5.7) 2.9 (0) Sierra Leone 52.3 (0.8) 48.1 (8.6) 4.2 (0.1) 180 Uganda 126.7 (0.4) 119.2 (5.2) 7.6 (0) Mali 108.5 (0.8) 102.6 (6.6) 5.9 (0) 181 Sierra Leone 31.4 (0.4) 28.5 (6.7) 2.9 (0) Angola 190.9 (0.7) 171.9 (7.3) 19 (0.1) 182 Eritrea 17.1 (0.4) 16.2 (2.8) 0.8 (0) Chad 97.3 (0.7) 91.9 (6.8) 5.4 (0) 183 Somalia 42.0 (0.4) 40.0 (5.1) 2.0 (0) Somalia 61.2 (0.6) 58.1 (5.2) 3.1 (0) DALYs, disability-adjusted life years;MSK, musculoskeletal; YLDs, years lost due to disability; YLLs, years of life lost.

YLLs in Africa, with infectious diseases and respiratory infections diseases, malignant neoplasms and mental disorders were the three being respectively the first and third causes of DALYs in 2015, main causes of disease burden. However, non-communicable when in Europe an epidemiological transition was observed.16 diseases such as cardiovascular diseases are rapidly increasing in Better understanding and treatment of infectious diseases reduced low-income countries owing to a high natality rate and a rapid shift their burden, and a move to diseases with higher disability, like in lifestyle.17 Owing to this transition, MSK diseases could soon MSK diseases, was seen.16 Conversely, infectious diseases were be expected to have a higher impact in those countries. Yet, the the 14th cause of DALYs in Europe in 2015 when cardiovascular

Table 5 Median (IQR 25–75) percentage of total DALYs, YLDs and YLLs for MSK diseases, by continent Continent (Year) DALY YLD YLL Africa (2000) 0.75 (0.51–1.09) 7.50 (5.97–8.30) 0.00 (0.00–0.10) Americas (2000) 3.61 (2.70–4.55) 11.60 (10.30–12.90) 0.40 (0.20–0.40) Asia (2000) 3.11 (2.12–4.71) 11.95 (10.35–13.62) 0.10 (0.00–0.20) Europe (2000) 5.89 (4.38–6.68) 16.60 (15.70–17.80) 0.20 (0.10–0.45) Oceania (2000) 3.34 (2.03–4.87) 12.60 (10.10–14.90) 0.20 (0.10–0.30) Africa (2005) 0.89 (0.68–1.24) 7.80 (6.25–8.67) 0.10 (0.10–0.10) Americas (2005) 4.24 (3.20–4.74) 12.50 (11.20–14.10) 0.20 (0.10–0.40) Asia (2005) 3.79 (2.44–5.18) 12.65 (11.07–14 22) 0.10 (0.10–0.10) Europe (2005) 6.20 (4.54–7.04) 17.10 (16.05–18.15) 0.20 (0.05–0.50) Oceania (2005) 3.65 (2.50–5.14) 13.40 (10.40–15.50) 0.10 (0.00–0.20) Africa (2010) 1.11 (0.87–1.62) 8.10 (6.52–9.00) 0.10 (0.00–0.10) Americas (2010) 4.47 (3.69–5.21) 13.10 (11.90–14.60) 0.50 (0.30–0.65) Figure 1 Association between gross domestic product (GDP) per Asia (2010) 4.22 (2.77–5.69) 13.50 (11.57–14.87) 0.10 (0.07–0.20) capita and proportion of disability-adjusted life years (DALYs) due to Europe (2010) 6.61 (5.07–7.46) 17.50 (16.50–18.50) 0.30 (0.10–0.45) musculoskeletal (MSK) diseases. AO, Angola; AT, Austria; BN, Brunei Oceania (2010) 3.91 (2.25–5.35) 13.70 (11.00–15.60) 0.20 (0.10–0.30) Darussalam; BS: Bahamas; CH, Switzerland; CR, Costa Rica; DK, Africa (2015) 1.33 (1.03–1.92) 8.35 (7.02–9.37) 0.10 (0.10–0.10) Denmark; GQ, equatorial Guinea; GR, Greece; IE, Ireland; IL, Israel; Americas (2015) 4.66 (3.98–5.59) 13.50 (12.30–15.05) 0.50 (0.30–0.70) IR, Iran; IS, Iceland; IT, Italy; JP, Japan; KR, Republic of Korea; KZ, Asia (2015) 4.17 (3.14–6.25) 13.95 (12.07–15.25) 0.10 (0.10–0.22) Kazakhstan; MT, Malta; NL, Netherlands; NO, Norway; NZ, New Zealand; Europe (2015) 6.66 (5.30–7.88) 17.60 (16.90–18.75) 0.30 (0.10–0.45) PT, Portugal; QA, Qatar; SA, Saudi Arabia; SC, Seychelles; SE, Sweden; TR, Oceania (2015) 4.14 (2.65–5.57) 13.70 (11.40–15.06) 0.20 (0.10–0.40) Turkey; TT, Trinidad and Tobago; UAE. United Arab Emirates; UK, United DALY, disability-adjusted life years;MSK, musculoskeletal; YLD, years lost due to Kingdom; USA, United States of America. disability; YLL, years of life lost.

Sebbag E, et al. Ann Rheum Dis 2019;78:844–848. doi:10.1136/annrheumdis-2019-215142 847 Epidemiology proportion of MSK YLDs remains high in Africa in 2015 (median manuscript. All authors reviewed the drafted manuscript for critical content. All proportion of 8.35%), showing that MSK diseases are still strongly authors approved the final version of the manuscript. incapacitating in this region of the world. Funding The authors have not declared a specific grant for this research from any Correlation of MSK DALYs with countries' GDP per capita funding agency in the public, commercial or not-for-profit sectors. disclosed a strong impact of socioeconomic background on the Competing interests None declared. burden of MSK diseases (figure 1). In 2015, the 10 countries with Patient consent for publication Not required. the highest proportion of MSK DALYs all have a GDP per capita Provenance and peer review Not commissioned; externally peer reviewed. over $20 000 per inhabitant. Conversely, the 10 countries with Data sharing statement The data used for the analyses are publicly available. the lowest proportion of MSK DALYs in 2000 and 2015 are in sub-Saharan Africa. Higher-income countries often have more effi- Author note This study is based on publicly available data and solely reflects the opinion of its authors and not that of the World Health Organization. cient national health systems but other key factors could account for this difference between high- and low-income countries. First, References life expectancy is higher in high-income countries than in low-in- 1 murray CJL, Vos T, Lozano R, et al. Disability-adjusted life years (DALYs) for 291 18 come countries. The increased disability due to MSK diseases diseases and injuries in 21 regions, 1990–2010: a systematic analysis for the Global might therefore be a reflection of the ageing of the population.9 18 Burden of Disease Study 2010. Lancet 2012;380:2197–223. Furthermore, the ratio of older to younger people is expected to 2 smith E, Hoy D, Cross M, et al. The global burden of gout: estimates from the Global increase, especially in low-income countries18 in which life expec- Burden of Disease 2010 study. Ann Rheum Dis 2014;73:1470–6. 3 cross M, Smith E, Hoy D, et al. The global burden of hip and knee osteoarthritis: tancy is quickly increasing. Thus, the MSK burden is expected to estimates from the Global Burden of Disease 2010 study. Ann Rheum Dis continue to grow. Another potential explanation is the role of the 2014;73:1323–30. widely observed epidemiological transition. Owing to the overall 4 Hoy D, March L, Brooks P, et al. The global burden of low back pain: estimates from improvement of medical care, the prognosis of several diseases the Global Burden of Disease 2010 study. Ann Rheum Dis 2014;73:968–74. 5 Hoy DG, Smith E, Cross M, et al. The global burden of musculoskeletal conditions for with initially high mortality has improved, hence a relative increase 2010: an overview of methods. Ann Rheum Dis 2014;73:982–9. in disability-inducing diseases, such as mental disorders and MSK 6 smith E, Hoy DG, Cross M, et al. The global burden of other musculoskeletal diseases. This transition is not completed at the same rate in all disorders: estimates from the Global Burden of Disease 2010 study. Ann Rheum Dis countries,19 which may explain the observed differences. 2014;73:1462–9. One important limitation of this study is the potential reporting 7 cross M, Smith E, Hoy D, et al. The global burden of rheumatoid arthritis: estimates from the Global Burden of Disease 2010 study. Ann Rheum Dis 2014;73:1316–22. bias, with a variable uncertainty range on available data, depending 8 murray CJL, Barber RM, Foreman KJ, et al. Global, regional, and national disability- 15 on the reporting country. Also, the burden of MSK diseases, as adjusted life years (DALYs) for 306 diseases and injuries and healthy life expectancy reported in this study, may vary with evolution of the methods (HALE) for 188 countries, 1990–2013: quantifying the epidemiological transition. The used by WHO to calculate MSK DALYs, YLDs and YLLs.15 Finally, Lancet 2015;386:2145–91. 9 Hoy D, Brooks P, Blyth F, et al. The epidemiology of low back pain. Best Pract Res Clin the grouping by continent may not fully reflect the economic Rheumatol 2010;24:769–81. heterogeneity within the continents. 10 Hoy D, Geere J-A, Davatchi F, et al. A time for action: opportunities for preventing the growing burden and disability from musculoskeletal conditions in low- and middle- income countries. Best Pract Res Clin Rheumatol 2014;28:377–93. Conclusion 11 Woolf AD, Akesson K. Understanding the burden of musculoskeletal conditions. The worldwide burden of MSK diseases, as quantified using DALYs, The burden is huge and not reflected in national health priorities. BMJ has significantly increased between 2000 and 2015. MSK diseases 2001;322:1079–80. 12 WHO. Disease burden and mortality estimates. Available: http://www.who​ .int/​ ​ are the ninth cause of DALYs, second cause of YLDs and 19th cause healthinfo/​global_​burden_​disease/​estimates/​en/ of YLLs over the world. The increasing burden of MSK diseases 13 UN Stats. National accounts. Available: https://​unstats.​un.​org/​unsd/​snaama/​Index mostly affects Europe, partly owing to ageing of the population 14 devleesschauwer B, Havelaar AH, Maertens de Noordhout C, et al. Calculating and the global epidemiological transition observed worldwide.18 disability-adjusted life years to quantify burden of disease. Int J Public Health We confirmed that the overall burden of MSK disease is essentially 2014;59:565–9. 15 WHO. Global burden disease database methods. Available: https://www.​who.​int/​ due to YLDs rather than YLLs, suggesting a high disability but low healthinfo/​global_​burden_​disease/​GlobalDALYmethods_​2000_​2015.​pdf lethality of MSK diseases. We observed a strongly significant asso- 16 le Y, Ren J, Shen J, et al. The changing gender differences in life expectancy in Chinese ciation between MSK DALYs and the GDP per capita, underlining cities 2005-2010. PLoS One 2015;10:e0123320. the role of socioeconomic background as a strong determinant of 17 mawaw PM, Yav T, Mukuku O, et al. Prevalence of obesity, diabetes mellitus, hypertension and associated risk factors in a mining workforce, Democratic Republic the MSK disease burden. These results are crucial to rheumatol- of Congo. Pan Afr Med J 2017;28. ogists, and to policy makers and national healthcare systems to 18 Hoy DG, Smith E, Cross M, et al. Reflecting on the global burden of musculoskeletal implement future adjustments to health plans. conditions: lessons learnt from the Global Burden of Disease 2010 study and the next steps forward. Ann Rheum Dis 2015;74:4–7. Acknowledgements The authors which to acknowledge the valuable contribution 19 gBD 2015 Child Mortality Collaborators. Global, regional, national, and selected of Ms Sylvie Thuong in the handling of the manuscript. subnational levels of stillbirths, neonatal, infant, and under-5 mortality, 1980- Contributors LA and ES designed the study. ES, RF, FS, JS, HD, LA analyzed the 2015: a systematic analysis for the Global Burden of Disease Study 2015. data. LA, ES and HD performed the statistical analyses. LA and ES drafted the initial Lancet;2016:1725–74.

848 Sebbag E, et al. Ann Rheum Dis 2019;78:844–848. doi:10.1136/annrheumdis-2019-215142 Letters Comparisons of hepatitis C viral replication in criteria8 for RA and HCV infection were enrolled (table 1). Twenty-three patients received non-TNF-α bDMARDs (tocili- patients with rheumatoid arthritis receiving zumab, n=8; abatacept, n=15), while nine participants tocilizumab, abatacept and tofacitinib therapy received tofacitinib therapy. Serum alanine aminotransferase (ALT) and HCV viral load were measured before and 1 year after treatment with bDMARDs and tsDMARDs. The majority Despite recent advances in direct antiviral agents for hepatitis of abatacept-treated patients with RA were biologics-naïve C virus (HCV), this infectious disease remains prevalent world- (93.3%) compared with their counterparts (tocilizumab, 75%; 2 wide and presents a major therapeutic challenge in patients tofacitinib, 22.2%; p=0.001 by χ test). None of the partic- with rheumatoid arthritis (RA).1 Our previous report demon- ipants received direct antiviral therapy for HCV. There were strated that use of antitumour necrosis factor (TNF)-α agents no significant differences in serum ALT between baseline in patients with RA with HCV infection appears to be safe.2 and 1 year after treatment with bDMARDs and tsDMARDs However, B-cell-targeted therapy with rituximab may lead to (figure 1; see online supplementary table S1). HCV viral HCV viraemia by causing a decline in exosomal microRNAs.3 loads before and after treatments are shown in figure 1. We Therefore, HCV viral replication may respond differently to found that HCV replications were significantly decreased in biologic disease-modifying antirheumatic drugs (bDMARDs) abatacept-treated patients with RA (p=0.015 by Wilcoxon and targeted synthetic DMARDs (tsDMARDs) owing to the signed-rank test). One patient exhibited an increase in ALT different mechanism of action in patients with RA. of Common Terminology Criteria for Adverse Events grade 1 Viral loads in HCV-infected patients with RA were reported severity after tocilizumab therapy. However viral loads before to be unaffected during short-term therapy with tocilizumab, and 1 year after tocilizumab and tofacitinib treatment were a monoclonal antibody targeting interleukin (IL)-6 receptor.4 5 comparable. Anti-IL-6 receptor and JAK inhibitors treatment Abatacept, a fusion protein binding cytotoxic T-lymphocyte– did not appear to affect HCV viral activity. associated antigen 4 (CTLA-4) that blocks the CD80/CD86 This study is the first to demonstrate that T-cell costimu- costimulatory pathways, has also been shown to be safe latory blocker therapy with abatacept might suppress HCV in patients with RA with HCV.6 It remains unclear whether viral activity. Moreover, IL-6 blocker and JAK inhibitor did tofacitinib, a Janus kinase (JAK) inhibitor,7 affects HCV viral not interfere with HCV replication in patients with RA. These activity in patients with RA. To study the impact of non-TNF-α findings support the 2015 ACR guideline which states that bDMARDs and JAK inhibitors on HCV viral replication, the use of bDMARDs in patients with RA with HCV should we conducted a prospective study of patients with RA with not differ from those without HCV.9 In addition, our results concomitant HCV infection treated with tocilizumab, abata- indicate that tofacitinib might not increase HCV viral replica- cept or tofacitinib. tion, which could have important implications for treatment In total, 32 patients who fulfilled the American College of decision-making. Further study is necessary to clarify whether Rheumatology (ACR)/European League Against Rheumatism tsDMARDs with differential JAK family specificity are safe in patients with RA with HCV. It is well known that the levels of transaminases in patients Table 1 Demographic data and concomitant medication of patients with HCV are not consistently correlated with viral replica- with rheumatoid arthritis and hepatitis C receiving tocilizumab, tion. Our study demonstrated that the changes of ALT and abatacept and tofacitinib therapy HCV viral loads were similar before and after treatment with tocilizumab or tofacitinib (figure 1). Interestingly, we Tocilizumab Abatacept Tofacitinib (n=8) (n=15) (n=9) P value revealed significantly decreased HCV viral loads after abata- Age 70.0 (65.0–71.8) 65.0 (54.0–72.0) 64.0 (59.5–70.0) 0.480 cept treatment, consistent with a previous report that showed Gender 0.881 HCV RNA might become undetectable after costimulatory 6 Female 7 (87.5%) 14 (93.3%) 8 (88.9%) blockade. Patients with RA may show different states of T-cell Male 1 (12.5%) 1 (6.7%) 1 (11.1%) exhaustion, which could be restored after abatacept treat- Disease duration, 12.0 (7.5–13.0) 12.0 (5.0–13.0) 12.0 (9.5–22.5) 0.229 ment, further enhancing T-cell functionality and suppressing 10 years viral replication. Given that the number of abatacept-treated RF-positive 8 (100.0%) 13 (86.7%) 8 (88.9%) 0.567 patients is limited, prospective long-term study is needed to ACPA-positive 8 (100.0%) 12 (80.0%) 7 (77.8%) 0.369 recruit a larger number of patients with RA with concomitant Biologics-naive 6 (75.0%) 14 (93.3%) 2 (22.2%) 0.001**† HCV infection. DAS28 5.9 (5.3–6.5) 6.4 (6.0–6.9) 5.6 (4.6–6.7) 0.088 In conclusion, tocilizumab, abatacept and tofacitinib appear ESR (mm/hour) 51.5 (40.3–69.8) 41.0 (31.0–65.0) 34.0 (23.0–84.0) 0.460 to be safe for treatment of patients with RA with HCV infec- CRP (mg/dL) 2.8 (1.1–4.9) 0.5 (0.3–1.4) 1.2 (0.1–2.2) 0.145 tion. Collaboration between rheumatologists and hepatolo- Daily glucocorticoids 5.0 (3.1–9.4) 7.5 (5.0–10.0) 5.0 (1.3–10.0) 0.385 gists is suggested in order to consider the suitability of direct dose (mg) antiviral therapy and to monitor viral loads in HCV-infected Methotrexate 1 (12.5%) 8 (53.3%) 2 (22.2%) 0.097 patients with RA. Salazopyrin 1 (12.5%) 5 (33.3%) 2 (22.2%) 0.533 1,2,3,4 1,3 2,4 2 Hydroxychloroquine 5 (62.5%) 10 (66.7%) 4 (44.4%) 0.550 Yi-Ming Chen, Wen-Nan Huang, Tsai-Ling Liao, Jun-Pen Chen, Sheng-Shun Yang,3,5 Hsin-Hua Chen,1,2,3,4 Tsu-Yi Hsieh,1 Wei-Ting Hung,1 Leflunomide 4 (50.0%) 4 (26.7%) 2 (22.2%) 0.407 Yi-Hsing Chen,1,3 Der-Yuan Chen6,7 Ciclosporin 1 (12.5%) 3 (20.0%) 1 (11.1%) 0.812 1Division of Allergy, Immunology and Rheumatology, Taichung Veterans General Data are expressed as median (IQR). 2 Hospital, Taichung, Taiwan Comparisons by χ test and Kruskal-Wallis test. **p<0.01 2 †Post-hoc analysis, abatacept vs tofacitinib, p=0.002. Department of Medical Research, Taichung Veterans General Hospital, Taichung, ACPA, anticitrullinated protein antibody; CRP, C reactive protein; DAS28, 28-Joint Disease Taiwan 3 Activity Score; ESR, erythrocyte sedimentation rate; RF, rheumatoid factor. Faculty of Medicine, National Yang Ming University, Taipei, Taiwan

Ann Rheum Dis June 2019 Vol 78 No 6 849 Letters

Patient consent for publication Obtained. Provenance and peer review Not commissioned; externally peer reviewed. © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ.

►► Additional material is published online only. To view please visit the journal online (http://​dx.​doi.​org/​10.​1136/​annrheumdis-​2018-​214400).

To cite Chen Y-M, Huang W-N, Liao T-L, et al. Ann Rheum Dis 2019;78:849–850.

Received 5 September 2018 Revised 6 December 2018 Accepted 8 December 2018 Published Online First 3 January 2019 Ann Rheum Dis 2019;78:849–850. doi:10.1136/annrheumdis-2018-214400

References 1. Maillefert JF, Muller G, Falgarone G, et al. Prevalence of hepatitis C virus infection in patients with rheumatoid arthritis. Ann Rheum Dis 2002;61:635–7. 2. Chen YM, Chen HH, Chen YH, et al. A comparison of safety profiles of tumour necrosis factor α inhibitors and rituximab therapy in patients with rheumatoid arthritis and chronic hepatitis C. Ann Rheum Dis 2015;74:626–7. 3. Liao TL, Hsieh SL, Chen YM, et al. Rituximab may cause increased hepatitis C virus viremia in rheumatoid arthritis patients through declining exosomal MicroRNA-155. Arthritis Rheumatol 2018;70:1209–19. 4. Nagashima T, Maruyama A, Kamata Y, et al. Unchanged serum viral load and liver function during tocilizumab treatment in a patient with rheumatoid arthritis and hepatitis C virus infection. Rheumatol Int 2012;32:2231–2. 5. Dragonas C, Ehrenstein B, Fleck M. Tocilizumab treatment in a patient suffering from rheumatoid arthritis and concomitant chronic hepatitis C infection. Rheumatology 2012;51:1520–1. 6. Mahajan TD, Hooker R, Maher L, et al. Abatacept therapy for rheumatoid arthritis in the setting of hepatitis C infection. J Clin Rheumatol 2010;16:332–4. 7. Boor PPC, de Ruiter PE, Asmawidjaja PS, et al. JAK-inhibitor tofacitinib suppresses interferon alfa production by plasmacytoid dendritic cells and inhibits arthrogenic and antiviral effects of interferon alfa. Transl Res 2017;188:67–79. Figure 1 Comparisons of (A) serum alanine aminotransferase (ALT) 8. Aletaha D, Neogi T, Silman AJ, et al. 2010 rheumatoid arthritis classification criteria: and (B) hepatitis C virus (HCV) viral load before and after tocilizumab, an American College of Rheumatology/European League Against Rheumatism collaborative initiative. Ann Rheum Dis 2010;69:1580–8. abatacept and tofacitinib treatment. Data are mean±1 SEM. HCV viral 9. Singh JA, Saag KG, Bridges SL, et al. 2015 American College of Rheumatology loads were expressed as log10 of the detected values. *P=0.015 by guideline for the treatment of rheumatoid arthritis. Arthritis Rheumatol Wilcoxon signed-rank test. 2016;68:1–26. 10. Frenz T, Grabski E, Buschjäger D, et al. CD4(+) T cells in patients with chronic inflammatory rheumatic disorders show distinct levels of exhaustion. J Allergy Clin 4Institute of Biomedical Science and Rong Hsing Research Center for Translational Immunol 2016;138:586–9. Medicine, National Chung Hsing University, Taichung, Taiwan 5Division of Gastroenterology and Hepatology, Taichung Veterans General Hospital, Taichung, Taiwan 6Rheumatology and Immunology Center, China Medical University Hospital, Taichung, Taiwan 7School of Medicine, China Medical University, Taichung, Taiwan Correspondence to Professor Der-Yuan Chen, Rheumatology and Immunology Center, China Medical University Hospital, Taichung, 40447, Taiwan; ​dychen@​vghtc.​gov.​tw and Dr Yi-Hsing Chen, Division of Allergy, Immunology and Rheumatology, Taichung Veterans General Hospital, Taichung 40705, Taiwan; ​ ysanne@​vghtc.​gov.​tw

Handling editor Josef S Smolen Acknowledgements The authors are grateful to the Biostatistics Task Force of Taichung Veterans General Hospital, Taiwan, for assistance with statistical analysis. The authors sincerely appreciate the assistance of the Center for Translational Medicine of Taichung Veterans General Hospital. Contributors Y-MC and W-NH conceived and designed the study, conducted the data analysis, and drafted and revised the manuscript. J-PC performed the statistical analysis and revised the manuscript. T-LL, S-SY, H-HC, T-YH and W-TH acquired clinical data and revised the manuscript. D-YC and Y-HC generated the original hypothesis, conceived and designed the study, acquired clinical data, conducted the data analysis, and drafted and revised the manuscript. Funding This study was supported by a grant from Taichung Veterans General Hospital, Taiwan (TCVGH-1077307C). Competing interests None declared.

850 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Reproductive health outcomes in women with psoriatic arthritis

There are now high-quality data showing rheumatoid arthritis improves in pregnancy and flares post partum, and that active disease in pregnancy may be associated with adverse fetal outcomes such as low birth weights. However, in psoriatic arthritis (PsA), the data are less clear. Previous studies examining the disease course of PsA, in and around pregnancy, have found conflicting results.1–3 Only one of those studies was prospective and many lack disease activity scores. Much of the data also predate the widespread use of biologic disease-modifying antirheumatic drugs (bDMARDs) in pregnancy and so may not reflect current clinical practice. A recent abstract reviewing a Swedish patient registry reported increased odds of preterm birth, induction of labour and caesarean section in PsA pregnancies compared with population comparators.4 Consecutive female patients with PsA planning pregnancy or were pregnant seen by our multidisciplinary rheumatology repro- ductive health service were managed using our evidence-based reproductive care pathway and followed prospectively. Age,

850 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Table 1 Patient characteristics, pregnancy and fetal outcomes in psoriatic arthritis Patients 14 Mean age, years, at referral (range) 34 (25–37)* Patients who became pregnant 12 Pregnancies 20 Patients who failed to become pregnant 2 Patients referred with a further pregnancy wish 4 Medications in pregnancy DMARD therapy (including oral steroids) 14 bDMARD 9† Oral prednisolone 5 Sulfasalazine 2 Hydroxychloroquine 1 NSAIDs 2 Aspirin 2 Pregnancy outcomes Live births 13‡ Vaginal deliveries 7 Caesarean sections 5 (3 emergency) Currently pregnant 2 Postpartum complications 2§ Fetal outcomes Birth weight, kg, mean (range) 3.54 (2.78–4.39)¶ Gestational age, weeks, mean (range) 39.62 (36–41.71)** Figure 1 Disease activity before, during and after pregnancy Spontaneous abortions 6 First trimester 4 Second trimester 2†† trimester (one recommenced in the second trimester) and one Breast feeding at 6 weeks 8 stopped in the second trimester. The bDMARD was continued *National average=32 years. throughout in three pregnancies. †6 discontinued in pregnancy (5 in the first trimester, 1 recommenced in the second Disease activity is shown in figure 1. Overall, the mean disease trimester) and 1 stopped in the second trimester). activity scores decreased during pregnancy and increased post ‡Includes one set of twins. §Grade 4 vaginal tear, patient on infliximab, discontinued in the second trimester, 1 partum. After applying the European League Against Rheuma- caesarean section wound infection. tism-defined response criteria disease activity, almost half of the ¶Nationally=3.49 kg. women had at least a moderate response during pregnancy and **Available for 11 births. more than one-third had at least a moderate flare post partum, ††1 hypercoiled cord, 1 trisomy 7. particularly at 6 weeks. The postpartum flare may be underesti- DMARDs, disease-modifying antirheumatic drug; NSAIDs, non-steroidal anti- mated as medication use was remarkably increased after delivery. inflammatory drugs; bDMARD, biologic disease-modifying antirheumatic drug. At 6 weeks, in five cases the bDMARD therapies had already been restarted. In two cases, the patients missed this appointment. One patient was started on methotrexate at their 6-week review. medications, disease control, and maternal and fetal outcomes In conclusion, we observed that PsA disease control generally were reviewed. There are no validated disease activity criteria for improved during pregnancy and flared post partum, replicating the PsA in pregnancy. Thus, we used the Disease Activity Score-28-C- largest previous study on this topic.3 Awareness of this is important reactive Protein 3 (DAS28-CRP-3) as this has been validated for to ensure prompt reintroduction of disease-modifying antirheu- rheumatoid arthritis in pregnancy.5 matic drug therapy post partum when appropriate. Miscarriage In total, 14 patients were reviewed. Twelve of these patients went rates were higher in our patients than the general population (32% on to become pregnant. There were 20 pregnancies in total. Eight vs 20%), while breastfeeding rates at 6 weeks (62% vs 55%) and patients had one pregnancy. Four patients had multiple pregnan- birth weights were similar.6 cies. Two patients had three pregnancies (one of whom is currently pregnant), one patient was pregnant twice, and one patient had Kieran Murray,‍ ‍ 1 Louise Moore,2 Fionnuala McAuliffe,3,4 four pregnancies and is currently pregnant. Two patients failed Douglas J Veale5,6 to become pregnant. On 11 occasions, patients were reviewed 1Rheumatology, Dublin Academic Medical Centre, Dublin, Ireland antenatally (while pregnancy planning). Pregnancy outcomes are 2Rheumatology Rehabilitation, Our Lady’s Hospice, Dublin, Ireland 3 shown in table 1. Maternal Medicine, National Maternity Hospital, Dublin 2, Ireland 4Perinatal Research Centre, University College Dublin, Dublin, Ireland In five cases, oral prednisolone was used during pregnancy, 5EULAR Centre For Arthritis And Rheumatic Diseases, Dublin Academic Medical three had commenced prior to pregnancy, while in two cases it Centre, Dublin, Ireland was commenced during pregnancy (one patient’s bDMARD 6The Conway Institute, University College Dublin, Dublin, Ireland therapy was ‘on hold’). Sulfasalazine and hydroxychloroquine Correspondence to Dr Kieran Murray, St. Vincent’s University Hospital, Dublin 4, were continued throughout pregnancy and post partum. In nine Ireland; kemurr​ ay@​hotmail.com​ cases, patients had taken bDMARD therapy at some time during pregnancy. On five occasions, this was discontinued in the first Handling editor Josef S Smolen

Ann Rheum Dis June 2019 Vol 78 No 6 851 Letters

Contributors The authors of this paper made the following contributions to the paper: Conception and design: KM, LM, DJV, FM; data collection: KM, LM; revising it critically for important intellectual content: KM, LM, DJV; involved in drafting the manuscript: KM, LM, DJV; have read and given final approval of the version to be published: KM, LM, DJV, FM. Competing interests None declared. Patient consent for publication Not required. Ethics approval Ethical approval was deemed unnecessary for this review, which describes our current routine care in accordance with best practice. No specific interventions were performed. Provenance and peer review Not commissioned; externally peer reviewed. Data sharing statement We are happy to share our data. © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ.

To cite Murray K, Moore L, McAuliffe F, et al. Ann Rheum Dis 2019;78:850–852.

Received 21 November 2018 Revised 22 January 2019 Accepted 4 February 2019 Published Online First 15 February 2019 Ann Rheum Dis 2019;78:850–852. doi:10.1136/annrheumdis-2018-214790

References 1 Mouyis MA, Thornton CC, Williams D, et al. Pregnancy outcomes in patients with psoriatic arthritis. J Rheumatol 2017;44:128–9. 2 Polachek A, Li S, Polachek IS, et al. Psoriatic arthritis disease activity during pregnancy and the first-year postpartum. Seminars in Arthritis and Rheumatism 2017;46:740–5. 3 Ursin KLS, Skomsvoll J, Wallenius M. Disease activity during and after pregnancy in women with psoriatic arthritis. Ann Rheum Dis 2018;77. 4 Remaeus KJ K, Stephansson O, Hellgren K. Pregnancy and birth outcomes in women with psoriatic arthritis, a nation-wide Swedish cohort study, 2007-2014. 10th International Conference on Reproduction, Pregnancy and Rheumatic Diseases, Bern, Switzerland, 2018. 5 Ince-Askan H, Hazes JMW, Dolhain R. Identifying clinical factors associated with low disease activity and remission of rheumatoid arthritis during pregnancy. Arthritis Care Res 2017;69:1297–303. 6 Office CS. Vital statistics. 2018, 2018.

852 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

validated in East Asian (EA) populations.2 In recent years, multiple studies indicated significant genetic differences between different European and EA populations.2 7 To resolve this further, we have examined all reported AS-associated SNPs in non-MHC regions in Chinese patients with AS. A total of 1289 patients with AS and 1536 controls were included in this analysis, which have not been included in prior analyses.2 7 The sociodemographic and clinical characteristics of the cohort are presented in online supplementary table S1. In this study, the candidate SNPs were mainly selected from six GWAS studies published from 2007 to 2016.2–6 8 Overall, 69 SNPs in non-MHC regions were included, 32 of which had not been validated in EA population before. The results of six loci (rs6759298, rs2297518, rs75301646, rs12615545, rs5837881 and rs27044) were consistent with prior reports in the Caucasian populations (table 1). Among them, the SNP rs2297518 in the NOS2 gene (OR/p_value (false discovery rate (FDR))=1.328/8.6E-03) and rs5837881 in an intergenic region (OR/p_value (FDR)=0.80/0.04) were not validated previously in EA populations, whereas four other SNPs were consistent with previous reports in Asian cohorts.2 7 Nine loci that had been previously associated with AS in relatively small cohorts of Chinese population were not validated in the present results (online supplementary table S2). Twenty-four loci that previ- ously showed an association with AS but with p value not at genome-wide significance (>10−8) in both Caucasian and EA populations were not validated in our results (online supplemen- tary table S3). Thirty-one loci previously reported implicated only in Caucasians did not show an association with AS in this study (online supplementary table S4). In this study, all six SNPs showed significant differentiation between AS and controls in both Caucasians and our samples. Among them, rs6759298 located in a ‘gene desert’ at chromo- some 2p15 showed highest significance in both Caucasian and Chinese data (OR/p value=1.31/3.60E-41, 1.22/3.81E-04, respectively). The Genotype-Tissue Expression dataset showed that rs6759298 has expression quantitative trait loci (eQTL) Genetic association of non-MHC region with with B3GNT2 (online supplementary figure 1A). In addition, the database of Mouse Genome Informatics (MGI) showed ankylosing spondylitis in a Chinese population that mice with B3gnt2-knocked-out showed increased levels of inflammatory cytokines such as interleukin (IL)-6, IL-1β and tumour necrosis factor-alpha). An association with another SNP Ankylosing spondylitis (AS) is a chronic inflammatory disease (rs2297518) located in NOS2, which has never been reported mainly affecting the sacroiliac joints and spine.1 In order to in Chinese data of patients with AS, was found. It is a mutation examine overall genetic susceptibility of AS, several genome-wide that causes deleterious alterations in the coding region of NOS2. association studies (GWASs) were performed in Caucasian popu- Rs75301646 is an intron variant on SULT1A2, which has eQTL lations, and a number of genetic polymorphisms in non-major with SULT1A1 expression (online supplementary figure 1B). histocompatibility complex (MHC) regions were found, such as MGI showed Sult1c1-knockout- mice have abnormal immune ERAP1, IL23R as well as many others.2–6 However, only a small responses and increased sacral vertebrae number. Rs12615545 portion of the single nucleotide polymorphisms (SNPs) were is near UBE2E3 and has eQTL with UBE2E3 (ubiquitin enzyme)

Table 1 Genetic associations with AS and controls in the Chinese population Western Asian This study Gene/region SNP Allele OR P values OR P values MAF OR P values P_adj 2p152 rs6759298 G 1.31 3.60E-41 1.28 1.60E-06 0.4418/0.3917 1.22 3.81E-04 8.60E-03 NOS22 rs2297518 A 1.13 6.30E-07 NA NA 0.176/0.1387 1.33 1.83E-04 8.60E-03 SULTIA12 rs75301646 A 1.11 1.40E-07 1.16 0.012 0.292/0.2472 1.26 2.55E-04 8.60E-03 UBE2E32 rs12615545 T 0.90 2.30E-07 0.83 8.50E-04 0.2959/0.3379 0.82 6.66E-04 0.01 Chr2 Indel6 rs5837881 T 0.88 1.26E-13 NA NA 0.149/0.18 0.80 3.14E-03 0.04 ERAP13 10 rs27044 C 0.71 1.00E-06 1.30 9.37E-07 0.4694/0.5046 0.86 4.70E-03 0.05 AS, ankylosing spondylitis; MAF, minor allele frequency.

852 Ann Rheum Dis June 2019 Vol 78 No 6 Letters in oesophageal mucosa, suggesting it might be related to inflam- Open access This is an open access article distributed in accordance with the mation status (online supplementary figure S1C). Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, Other loci were not validated in our samples such as IL12B, 9 and license their derivative works on different terms, provided the original work is ANTXR2 and FCGR2A. We postulated that some different SNPs properly cited, appropriate credit is given, any changes made indicated, and the use located in the same genes or genes in the same pathway can substi- is non-commercial. See: http://​creativecommons.​org/​licenses/​by-​nc/​4.​0/. tute for those loci in the Chinese population as has been observed © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY-NC. No with IL23R.2 Therefore, it is essential to find associated SNPs in commercial re-use. See rights and permissions. Published by BMJ. different cohorts by sequencing or different chip analyses to extend ►► Additional material is published online only. To view please visit the journal our knowledge to the pathogenesis of AS. online (http://​dx.​doi.​org/​10.​1136/​annrheumdis-​2018-​214625). In conclusion, we genotyped 69 previously reported non-MHC JL and WP contributed equally. AS-associated SNPs in a different Chinese cohort and found that six loci showed significant differences between patients with AS and controls in both Caucasian and EA populations. Usually, To cite Liu J, Pu W, Li Y, et al. Ann Rheum Dis 2019;78:852–853. several SNPs may have similar effect to the same gene. There- fore, in the future, we can perform functional study of genes Received 21 October 2018 controlled by several SNPs in the mouse. Revised 8 November 2018 Accepted 12 November 2018 Jing Liu,1,2 Weilin Pu,1,2 Yuan Li,1,2 Yanyun Ma,2,3 Qi Zhu,4 Wei Wan,5 Published Online First 14 December 2018 6 1,2 1,7 8 Chengde Yang, Xiaofeng Wang, Xingdong Chen, Xiaodong Zhou, Ann Rheum Dis 2019;78:852–853. doi:10.1136/annrheumdis-2018-214625 John D Reveille,8 Li Jin,1,2 Hejian Zou,9,10 Jiucun Wang1,11 1State Key Laboratory of Genetic Engineering, Collaborative Innovation Center for References Genetics and Development, School of Life Sciences, Fudan University, Shanghai, 1 Brown MA, Kennedy LG, MacGregor AJ, et al. Susceptibility to ankylosing China spondylitis in twins: the role of genes, HLA, and the environment. Arthritis Rheum 2 Human Phenome Institute, Fudan University, Shanghai, China 1997;40:1823–8. 3 Ministry of Education Key Laboratory of Contemporary Anthropology, Department 2 Wellcome Trust Case Control Consortium, Australo-Anglo-American Spondylitis of Anthropology and Human Genetics, School of Life Sciences, Fudan University, Consortium (TASC), Burton PR, Australo-Anglo-American Spondylitis C, et al. Shanghai, China Association scan of 14,500 nonsynonymous SNPs in four diseases identifies 4 Institute of Arthritis Research, Shanghai Academy of Chinese Medical Sciences, autoimmunity variants. Nat Genet 2007;39:1329–37. Guanghua Integrative Medicine Hospital, Shanghai, China 3 Davidson SI, Wu X, Liu Y, et al. Association of ERAP1, but not IL23R, with ankylosing 5Division of Rheumatology and Immunology, Changhai Hospital, Shanghai, China spondylitis in a Han Chinese population. Arthritis Rheum 2009;60:3263–8. 6Division of Rheumatology, Ruijin Hosptial, Shanghai Jiaotong University Schools of 4 Australo-Anglo-American Spondyloarthritis Consortium (TASC), Reveille JD, Sims AM, Medicine, Shanghai, China et al. Genome-wide association study of ankylosing spondylitis identifies non-MHC 7Fudan-Taizhou Institute of Health Sciences, Taizhou, China susceptibility loci. Nat Genet 2010;42:123–7. 8Division of Rheumatology and Clinical Immunogenetics, The University of Texas- 5 Choi CB, Kim TH, Jun JB, et al. ARTS1 polymorphisms are associated with ankylosing McGovern Medical School, Houston, TX, United States spondylitis in Koreans. Ann Rheum Dis 2010;69:582–4. 9Division of Rheumatology, Huashan Hospital, Fudan University, Shanghai, China 6 Evans DM, Spencer CC, Pointon JJ, et al. Interaction between ERAP1 and HLA-B27 in 10Institute of Rheumatology, Immunology and Allergy, Fudan University, Shanghai, ankylosing spondylitis implicates peptide handling in the mechanism for HLA-B27 in China disease susceptibility. Nat Genet 2011;43:761–7. 11Division of Dermatology, Huashan Hospital, Fudan University, Shanghai, China 7 International Genetics of Ankylosing Spondylitis Consortium (IGAS), Cortes A, Hadler J, et al. Identification of multiple risk variants for ankylosing spondylitis through high- Correspondence to Professor Jiucun Wang, State Key Laboratory of Genetic density genotyping of immune-related loci. Nat Genet 2013;45:730–8. Engineering, School of Life Sciences, Fudan University, Division of Dermatology, 8 Karaderi T, Keidel SM, Pointon JJ, et al. Ankylosing spondylitis is associated with the Huashan Hospital, Shanghai 200438, China; ​jcwang@​fudan.​edu.​cn and Hejian Zou, anthrax toxin receptor 2 gene (ANTXR2). Ann Rheum Dis 2014;73:2054–8. Division of Rheumatology, Huashan Hospital, Fudan University, Shanghai, China; 9 Ellinghaus D, Jostins L, Spain SL, et al. Analysis of five chronic inflammatory diseases ​hjzou@​fudan.​edu.​cn identifies 27 new associations and highlights disease-specific patterns at shared loci. Nat Genet 2016;48:510–8. Handling editor Josef S Smolen 10 Robinson PC, Leo PJ, Pointon JJ, et al. Exome-wide study of ankylosing spondylitis demonstrates additional shared genetic background with inflammatory bowel disease. Contributors JW and HZ designed the project and wrote the manuscript. JL and NPJ Genom Med 2016;1:16008. WP collected data, performed genotyping and analysis and wrote the manuscript. YL and YM managed all samples. QZ, WW and CY provided the ankylosing spondylitis samples. XW and XC provided the control samples. XZ, JDR and LJ revised the manuscript and provided valuable suggestions. Funding The study was supported by research grants from the National Basic Research Program (2014CB541801), National Natural Science Foundation of China (31521003), Shanghai Municipal Science and Technology Major Project (2017SHZDZX01), International S&T Cooperation Program of China (2013DFA30870), 111 Project (B13016) and US NIH NIAID U01 (1U01AI090909) and China Medical Foundation. Computational support was provided by theHigh- End Computing Center located at Fudan University. Competing interests None declared Patient consent for publication Obtained. Ethics approval Fudan University. Provenance and peer review Not commissioned; externally peer reviewed.

Ann Rheum Dis June 2019 Vol 78 No 6 853 Letters

Revisiting the issue of how to assess pneumococcal vaccine immunogenicity: a post hoc analysis of antipneumococcal antibody responses among adult patients with systemic lupus erythematosus previously immunised with 23-valent pneumococcal polysaccharide vaccine

A judicious analysis and comparison of studies on the immu- nogenicity of pneumococcal vaccine in patients with systemic lupus erythematosus (SLE)1–9 is hampered by wide heteroge- neity in several factors, including differences in the immunoas- says performed, cut-off levels used as serological correlates of protection, number of individual serotypes as well as types of

Ann Rheum Dis June 2019 Vol 78 No 6 853 Letters capsular antigens analysed, response criteria, and SLE-related between 2.0 and 2.9, respectively (p<0.0001 for both rates vs immunosuppressive treatment. In addition, how to interpret threefold responses). We also performed a binary logistic regres- vaccine response when preimmunisation antipneumococcal sion analysis to examine whether any of the seven serotypes antibody titres (anti-PnAbs) are already high (or protective) is tested were more likely to result in seroconversion, and serotype not yet clear, which is of relevance given that non-immunised 14 proved to be the most capable of them all (OR 3.43, 95% CI adults often have high antibody levels for some serotypes as 1.08 to 10.93; p=0.037). a result of prior pneumococcal infections. Moreover, only a Thus, in this exploratory analysis with adult patients with minority of individuals with high prevaccination anti-PnAbs SLE, we showed that high preimmunisation anti-PnAbs are reportedly capable of mounting a ≥4-fold rise in titres on precluded a fourfold response to vaccination. Also, pneumo- immunisation, then a quantitative increase historically associ- coccal vaccine seroconversion was best associated with a fold ated with achievement of serological protection against pneu- increase in antipneumococcal titres of at least three. Of note, the mococcal disease.10 antibody cut-off level that we used for assessing seroconversion Considering the scarcity of data in assessing humoral 10 response to pneumococcal vaccination is in line with experts’ immune response when prevaccination anti-PnAbs are high, 10 we addressed this issue by performing a post hoc analysis on recommendations. Lastly, we suggest that (1) there should be antipneumococcal antibody responses of 54 adult patients with a minimum number of individual serotypes to be tested when SLE previously immunised with 23-valent pneumococcal poly- assessing antipneumococcal antibody responses in order to saccharide vaccine.8 Of note, 42 (78%) participants already counterbalance variations in serotype-specific immunogenicity; had high prevaccination anti-PnAbs (ie, ≥1.3 µg/mL) against and (2) pneumococcal serotypes having a high prevaccination- at least one of seven pneumococcal serotypes tested (table 1). specific antibody level should be disregarded from analysis when Short-term pneumococcal vaccine responses (on a serotype interpreting vaccination response by means of the quantitative basis) are displayed in table 2. According to our results, high increase in anti-PnAbs, therefore allowing immunogenicity preimmunisation anti-PnAbs increased significantly less after studies of pneumococcal vaccine to be more comparable.

vaccination when compared with low preimmunisation anti- ‍ ‍ 1 2 PnAbs (fold increase in geometric mean concentration of 1.87 vs Rodrigo Poubel Rezende, Luís Eduardo Coelho Andrade, Evandro Mendes Klumb3 2.30, respectively; p<0.0001). Interestingly, no titre increased 1 fourfold after vaccination when basal antipneumococcal value Rheumatology, Departamento de Medicina Clínica, Universidade Federal Fluminense, Rio de Janeiro, Brazil was ≥1.3 g/mL. For serotypes having a prevaccination anti- 2 µ Fleury Medicine and Health Laboratories, Disciplina de Reumatologia, Escola PnAb <1.3 µg/mL, attaining a ≥3-fold response in titres led to Paulista de Medicina, Universidade Federal de São Paulo, São Paulo, Brazil a seroconversion rate of 100%, as compared with seroconver- 3Rheumatology, Disciplina de Reumatologia, Universidade do Estado do Rio de sion rates of 11% and 82% for fold responses of less than 2 and Janeiro, Rio de Janeiro, Brazil

Table 1 Characteristics of patients with systemic lupus erythematosus at pneumococcal polysaccharide vaccination* Patients Variable (n=54) Women, n (%) 49 (90.7) Age, median (range), years 38 (19–69) Disease duration, mean±SD, years 10±6.8 Users of antimalarial drug, n (%) 45 (83.3) Users of immunosuppressants, n (%)† 28 (51.8) Users of prednisone, n (%) 47 (87) Dosage (mg/day), median (IQR) 5 (2.5–5) SLEDAI-2K score, mean±SD 2.7±3.5 Number of patients with preimmunisation antipneumococcal titres ≥1.3 µg/mL Against none of the serotypes tested 12 One serotype 10 Two serotypes 11 Three serotypes 7 Four serotypes 7 Five serotypes 4 Six serotypes 3 Against all seven serotypes tested 0 Serotypes with preimmunisation specific antibody titres ≥1.3 µg/mL, n‡ 119 Serotypes with preimmunisation specific antibody titres <1.3 µg/mL, n§ 259 *According to our original work,8 participants could not have received any pneumococcal vaccine in the 5 years preceding study entry, and none of them had been previously immunised with pneumococcal conjugate vaccine. Serum antipneumococcal IgG levels against seven vaccine serotypes (4, 6B, 9V, 14, 18C, 19F and 23F) were determined by ELISA, with blood samples drawn immediately prior to administration of a single dose of 23-valent pneumococcal polysaccharide vaccine and 4–6 weeks thereafter. †Mycophenolate, n=16; azathioprine, n=7; cyclophosphamide, n=5. ‡Calculated by summing the number of tested pneumococcal serotypes, per patient, with preimmunisation specific antibody levels ≥1.3 µg/mL. §Calculated as the number of tested serotypes in the study (=54 patients × 7 pneumococcal serotypes tested per patient=378) minus the number of serotypes with preimmunisation antipneumococcal titres ≥1.3 µg/mL (=119). SLEDAI-2K, Systemic Lupus Erythematosus Disease Activity Index 2000.

854 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Table 2 Pneumococcal vaccine responses stratified by preimmunisation serum antipneumococcal antibody levels. Results are presented on a serotype basis Serotypes with baseline anti-PnAb Serotypes with baseline anti- <1.3 µg/mL (low titres) PnAb ≥1.3 µg/mL (high titres) (n=259) (n=119) P value* Prevaccination GMC (95% CI), µg/mL 0.18 (0.17 to 0.19) 0.54 (0.50 to 0.57) <0.0001 Postvaccination GMC (95% CI), µg/mL 0.42 (0.38 to 0.46) 1.01 (0.95 to 1.08) <0.0001 Fold increase in GMC (95% CI)† 2.30 (2.12 to 2.50) 1.87 (1.74 to 2.02) <0.0001 Fold increase in GMC on immunisation <2 Serotypes, n (%) 118/259 (45.5) 64/119 (53.8) 0.15 Serotypes achieving postvaccination titre ≥1.3 µg/mL (seroconversion)‡, n (%) 13/118 (11) NA – Fold increase in GMC on immunisation ≥2 and <3 Serotypes, n (%) 45/259 (17.4) 36/119 (30.2) <0.01 Serotypes achieving postvaccination titre ≥1.3 µg/mL (seroconversion)‡, n (%) 37/45 (82.2) NA – Fold increase in GMC on immunization ≥3 and <4 Serotypes, n (%) 27/259 (10.4) 19/119 (16) 0.13 Serotypes achieving postvaccination titre ≥1.3 µg/mL (seroconversion)‡, n (%) 27/27 (100) NA – Fold increase in GMC on immunisation ≥4 Serotypes, n (%) 69/259 (26.6) 0/119 (0) <0.0001 Serotypes achieving postvaccination titre ≥1.3 µg/mL (seroconversion)‡, n (%) 69/69 (100) NA – *For comparison between serotypes having high versus low prevaccination antipneumococcal antibody levels. Means were compared using unpaired t-test, and proportions using χ2 or Fisher’s exact test. Since antibodies generally have a log-Gaussian distribution, individual serotype data (measured by ELISA) were converted to logarithms to compare geometric means. †The ratio of the GMC of antipneumococcal antibodies after vaccination to the GMC of antipneumococcal antibodies before vaccination. ‡Seroconversion for each pneumococcal serotype was defined as the quantitative increase in specific antibody titre from a level not considered protective (<1.3 µg/mL) to one that is protective (≥1.3 µg/mL), which is in accordance with the cut-off value set by experts.10 GMC, geometric mean concentration of IgG antibodies to all seven pneumococcal serotypes tested; NA, not applicable; anti-PnAb, antipneumococcal antibody titre.

Correspondence to Dr Rodrigo Poubel Rezende, Rheumatology, Departamento de 4 Lipnick RN, Karsh J, Stahl NI, et al. Pneumococcal immunization in patients with Medicina Clínica, Universidade Federal Fluminense, Rio de Janeiro 24033900, Brazil; systemic lupus erythematosus treated with immunosuppressives. J Rheumatol ​ropoubel@​id.​uff.​br 1985;12:1118–21. 5 Battafarano DF, Battafarano NJ, Larsen L, et al. Antigen-specific antibody responses Handling editor Josef S Smolen in lupus patients following immunization. Arthritis Rheum 1998;41:1828–34. 6 Elkayam O, Paran D, Caspi D, et al. Immunogenicity and safety of pneumococcal Acknowledgements We thank the support from Fleury Laboratory (São Paulo, vaccination in patients with rheumatoid arthritis or systemic lupus erythematosus. Clin Brazil) in performing the antipneumococcal immunoassays. Infect Dis 2002;34:147–53. Contributors RPR and EMK conceived and designed the study, analysed the data 7 Tarján P, Sipka S, Maródi L, et al. No short-term immunological effects of and drafted the manuscript. LECA reviewed the manuscript and contributed to the pneumococcus vaccination in patients with systemic lupus erythematosus. Scand J intellectual content of the article. Rheumatol 2002;31:211–5. 8 Rezende RP, Ribeiro FM, Albuquerque EM, et al. Immunogenicity of pneumococcal Funding The authors have not declared a specific grant for this research from any polysaccharide vaccine in adult systemic lupus erythematosus patients undergoing funding agency in the public, commercial or not-for-profit sectors. immunosuppressive treatment. Lupus 2016;25:1254–9. Competing interests None declared. 9 Grabar S, Groh M, Bahuaud M, et al. Pneumococcal vaccination in patients with systemic lupus erythematosus: a multicenter placebo-controlled randomized double- Patient consent for publication Not required. blind study. Vaccine 2017;35:4877–85. Ethics approval Hospital Universitário Pedro Ernesto. 10 Orange JS, Ballow M, Stiehm ER, et al. Use and interpretation of diagnostic Provenance and peer review Not commissioned; externally peer reviewed. vaccination in primary immunodeficiency: a working group report of the Basic and Clinical Immunology Interest Section of the American Academy of Allergy, Asthma Data sharing statement No additional data available. & Immunology. J Allergy Clin Immunol 2012;130(3 Suppl):S1–S24. © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ.

To cite Rezende RP, Andrade LEC, Klumb EM. Ann Rheum Dis 2019;78:853–855.

Received 9 December 2018 Revised 31 December 2018 Accepted 2 January 2019 Published Online First 1 February 2019 Ann Rheum Dis 2019;78:853–855. doi:10.1136/annrheumdis-2018-214888

References 1 Klippel JH, Karsh J, Stahl NI, et al. A controlled study of pneumococcal polysaccharide vaccine in systemic lupus erythematosus. Arthritis Rheum 1979;22:1321–5. 2 Jarrett MP, Schiffman G, Barland P, et al. Impaired response to pneumococcal vaccine in systemic lupus erythematosus. Arthritis Rheum 1980;23:1287–93. 3 McDonald E, Jarrett MP, Schiffman G, et al. Persistence of pneumococcal antibodies after immunization in patients with systemic lupus erythematosus. J Rheumatol 1984;11:306–8.

Ann Rheum Dis June 2019 Vol 78 No 6 855 Letters

Validity of the Swedish version of the systemic sclerosis quality of life questionnaire (SSCQoL): A novel measure of quality of life for patients with systemic sclerosis

Systemic sclerosis (SSc) is an autoimmune disease character- ised by microvascular injury and excessive fibrosis of the skin and internal organs,1 which influences activities of daily life2 and quality of life (QoL).3 The Medical Outcome Study Short Form-36, the EuroQol-5 Domain (EQ-5D) and the Patient-Re- ported Outcomes Measurement Information System 29-item are generic health-related QoL instruments, valid to be used in SSc.4 However, disease-specific tools could be more sensitive to capture disease-related factors influencing QoL. The Systemic

Ann Rheum Dis June 2019 Vol 78 No 6 855 Letters

Sclerosis Quality of Life Questionnaire (SScQoL), developed by Table 1 Characteristics of the 67 patients with SSc in the study Reay5 using a needs-based QoL model, is an outcome for assessing disease impact on health and well-being. The SScQoL consists of Demographic 29 items, covering the themes: function, emotion, sleep, social Age at completion of survey (years) 61 (49–69) and pain. Each item with ‘yes’‘no’ response options are scored 1 Female , male 60 (90), 7 (10) and 0, respectively. The SScQoL has been sufficiently translated Pension— retirement or agreement pension 27 (40) from English to French, Polish, Spanish and Swedish and worked No work disability, part-time work disability, full 12 (30), 14 (35), 13 (33) well as a five-subscale questionnaire across these countries.4 time disability* We evaluated the construct validity of the Swedish SScQoL Disease characteristics against EQ-5D, the Scleroderma Health Assessment Question- lcSSc, dcSSc 45 (67), 22 (33) naire (SHAQ),6 (including Health Assessment Questionnaire Disease duration at survey (years)† 11 (6–17) Disability Index (HAQ-DI) and the five SSc specific Visual mRss (0–51)† 3 (0–9) Analogue Scales (VAS), VAS fatigue and clinical and demo- VC (mean % predicted value)* 93 (82–104) graphic characteristics. The study was approved by the local Pitting scars 60 (90) ethical committee. The patients answered SScQoL, EQ-5D and Upper GI disease 58 (87) SHAQ when they visited the rheumatology centre. Clinical char- Serology: ACA, ARA, ATA, ANA+, ANA− 15 (22), 9 (13), 11 (16), 23 acteristics within the last 24 months were used in the analyses. (34), 9 (13) The result of EQ-5D is expressed in quality-adjusted life years Patient-reported outcomes (QALYs) with scores ranging from −0.059 to 1.00. To present the SHAQ (0–3) experience of QoL within the SScQoL domains, all items for each HAQ-DI 0.5 (0.3–0.9) domain were summed and divided with the number of items in SSc VAS GI symptoms 0.4 (0.0–1.5) the domain, generating an average of agreement (problem) versus SSc VAS respiratory symptoms 0.3 (0.0–1.3) disagreement (no problem) (table 1). Construct validity was anal- SSc VAS Raynaud 0.7 (0.2–1.7) ysed by using Spearman’s correlation coefficient. We hypothe- SSc VAS digital ulcers 0.3 (0.0–1.2) sised that SScQoL would have a high to moderate correlation SSc VAS global assessment 1.3 (0.9–1.7) with patient reported outcomes as, EQ-5D and SHAQ, and low VAS pain (0–3) 0.9 (0.5–1.5) correlation with clinical manifestations since objective functional VAS fatigue (0–3) 1.3 (0.3–2.1) tests not always capture the patient´s experience of the day-to EQ-5D, QALYs (−0.059 to 1.00) −0.880 (−0.779 – −0.935) day functioning. SScQoL (0–29) 12 (7–20) Sixty-seven patients fulfilling American College of Rheumatol- Average of agreement (problem) for each domain in % 59, 52, 51, 44 and 37 ogy/European League Against Rheumatism criteria for SSc were pain, function, sleep, emotional and social domain included in the study. Demographic, clinical features and patient-re- Data are shown as median (IQR) or number (percentage). ported outcomes are detailed in table 1. There was a moderate Lower QALY indicates worse health. SScQoL: high scores indicates poor QoL. correlation between SScQoL and EQ-5D (rs=−0.66, p=0.001). The SScQoL had strong correlation to HAQ-DI, VAS pain and *One drop out. †Four drop out. VAS global assessment. Moderate correlation to VAS fatigue, and ACA, anticentromeric antibodies; ANA, antinuclear antibody; ANA−, ANA low, but significant correlations was found between SScQoL and negative; ANA+, ANA positive; ARA, antiRNA polymerase III antibodies; ATA, VAS gastrointestinal (GI) symptoms, Raynaud’s phenomenon, VAS antitopoisomerase I antibodies; EQ-5D, EuroQol-5 Domain; GI, gastrointestinal; digital ulcers and VAS respiratory symptoms (online supplementary HAQ-DI, Health Assessment Questionnaire Disability Index; QALY, quality-adjusted table S1). In general, EQ-5D was less associated with the SHAQ life years; QoL, quality of life; SHAQ, Scleroderma Health Assessment Questionnaire; (online supplementary table S1). As expected both vital capacity SSc, systemic sclerosis; SScQoL, Systemic Sclerosis Quality of Life Questionnaire; VAS, Visual Analogue Scale; VC, vital capacity; dcSSc, diffuse cutaneous SSc; lcSSc, and skin score had low, if any correlations with SScQoL (rs = 0.03 limited cutaneous SSc; mRss, modified Rodnan skin score. (CI -0.027 - 0.022), rs = 0.01 (CI -0.024 - 0.025), respectively). Neither were there any differences in SScQoL between diffuse cutaneous SSc and limited cutaneous SSc, between those with or Correspondence to Dr Gunnel Sandqvist, Department of Rheumatology, Skåne without immunosuppression or between those with low (<10) University Hospital, Lund SE-221 85, Sweden; ​gunnel.​sandqvist@​med.​lu.​se or high (>10) skin score. Patients with GI symptoms had lower Handling editor Josef S Smolen SScQoL (p=0.020) and EQ-5D QALYs (p=0.016), and patients with some degree of work ability had better SScQoL (p=0.025) Contributors Both authors contributed equally to this letter. and EQ-5D QALYs (p=0.001). Patients with disease duration Funding The authors have not declared a specific grant for this research from any >5 years had worse SScQoL (p=0.049). Correlations between funding agency in the public, commercial or not-for-profit sectors. SScQoL and long or short disease duration are shown in online Competing interests None declared. supplementary table S2. Patient consent for publication Not required. Our results show that SScQoL to some extent have a better Provenance and peer review Not commissioned; externally peer reviewed. ability to capture the disease-specific factors influencing QoL in © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and SSc than the generic EQ-5D and may therefore be a relevant tool permissions. Published by BMJ. to measure QoL in SSc. However, to assess if SScQoL is useful to evaluate treatment from a patient perspective, future studies should include longitudinal analyses to assess sensitivity to change. To cite Sandqvist G, Hesselstrand R. Ann Rheum Dis 2019;78:855–857. Gunnel Sandqvist, Roger Hesselstrand Received 10 August 2018 Department of Clinical Sciences Lund, Section of Rheumatology, Lund University, Revised 17 December 2018 Skåne University Hospital, Lund, Sweden Accepted 17 December 2018

856 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Published Online First 4 January 2019 Ann Rheum Dis 2019;78:855–857. doi:10.1136/annrheumdis-2018-214260

References 1 Steen VD, Medsger TA. Severe organ involvement in systemic sclerosis with diffuse scleroderma. Arthritis Rheum 2000;43:2437–44. 2 Bassel M, Hudson M, Taillefer SS, et al. Frequency and impact of symptoms experienced by patients with systemic sclerosis: results from a Canadian National Survey. Rheumatology 2011;50:762–7. 3 Johnson SR, Glaman DD, Schentag CT, et al. Quality of life and functional status in systemic sclerosis compared to other rheumatic diseases. J Rheumatol 2006;33:1117–22. 4 Ndosi M, Alcacer-Pitarch B, Allanore Y, et al. Common measure of quality of life for people with systemic sclerosis across seven European countries: a cross-sectional study. Ann Rheum Dis 2018;77:1032–8. 5 Reay N. The quality of life in patients with diffuse and limited systemic sclerosis. Monograph: University of Leeds, 2008. 6 Steen VD, Medsger TA. The value of the Health Assessment Questionnaire and special patient-generated scales to demonstrate change in systemic sclerosis patients over time. Arthritis Rheum 1997;40:1984–91.

Ann Rheum Dis June 2019 Vol 78 No 6 857 Letters

Evaluation of retinal microvascular perfusion in systemic sclerosis: a case–control study Figure 1 Optical coherence tomography angiography (OCTA) generated en face angiograms of the full retina (A) superficial retinal Systemic sclerosis (SSc) is a multisystem connective tissue disease capillary plexus (B) and deep retinal capillary plexus (C). Perfusion of characterised by obliterative vasculopathy that is not limited to the full retina (D), superficial retinal capillary plexus (E) and deep retinal the peripheral microcirculation of the skin, but is also observed capillary plexus (F), macular volume (G), foveal avascular zone (FAZ) in other organs.1 Retinal vascular disease in SSc indicating endo- in the superficial retinal capillary plexus (H) as well as deep retinal thelial cell damage has a prevalence of 34%–55% and might be capillary plexus (I) were compared between patients with systemic apparent long before patients become symptomatic.2 3 The gold sclerosis (SSC) and matched controls. Data are presented as box and standard to evaluate retinal microvascular alterations has been whisker plots with median, lower as well as upper extreme. Significant fluorescence angiography (FA) for the last 60 years.4 However, differences were tested using the Mann–Whitney U test. P<0.05 was the use of FA is quite laborious because it is an invasive proce- considered statistically significant. DRCP, deep retinal capillary plexus; dure that takes time and needs an experienced ophthalmologist FR, full retina; SRCP, superficial retinal capillary plexus. to review the resulting images carefully. Recently, optical coher- ence tomography angiography (OCTA) has been introduced, which generates angiographic images based on motion contrast statistically significant differences between patients with SSc and imaging to high-resolution volumetric blood flow information matched controls. Patients with SSc had a significantly lower in a few seconds.5 OCTA is a non-invasive technique, which retinal perfusion compared with healthy eyes (29% vs 32%, allows the quantification of perfused and non-perfused areas p=0.037) (figure 1D). Substructure analysis revealed these differ- of the retina in primarily ocular pathologies, but it also mirrors ences were mainly located in the SRCP (21% vs 25%, p=0.006) vascular involvement in systemic diseases as most recently shown (figure 1E) rather than in the DRCP (38% vs 40%, p=0.061) for systemic lupus erythematosus.6 The aim of this study was (figure 1F). The macular volume (figure 1G) and foveal avascular to evaluate the retinal microvascular blood flow in patients zone (FAZ) in the SRCP (figure 1H) as well as DRCP (figure 1I) with SSc, without clinical signs of retinal involvement, using did not differ significantly between both groups. In patients with OCTA. This age-matched and sex-matched case–control study SSc, the perfusion in the FR slab correlated significantly with was conducted in accordance with the Declaration of Helsinki. macular volume (ρ=0.420, p=0.041) as well as mean arterial A total of 24 eyes of 12 patients with SSc and 24 eyes of 12 pressure (ρ=0.421, p=0.041). BCVA correlated significantly healthy matched controls were examined by OCTA (HS-100; with perfusion in the SRCP (ρ=−0.477, p=0.018) as well as Canon, Tokyo, Japan), in addition to basic ophthalmological the DRCP (ρ=−0.462, p=0.023) slab. Perfusion values in either and rheumatological examination of the study participants. This slab did not show any significant correlation with age, disease OCTA device uses a modified full-spectrum amplitude decor- duration or modified Rodnan skin score (MRSS). relation algorithm to generate angiograms of the full retina (FR) In conclusion, our small proof-of-concept study suggests that as well as automatically segmented slabs of both the superfi- retinal microcirculatory impairment can be demonstrated in cial retinal capillary plexus (SRCP) and deep retinal capillary ophthalmological asymptomatic patients with SSc using OCTA. plexus (DRCP) (figure 1A–C). Perfusion values of both eyes Hence, OCTA-based analysis of retinal perfusion might provide were averaged for each subject and groups were compared using a new quantitative metric, which is much faster to obtain and the Mann-Whitney U test. OCTA results were correlated with also more independent with regard to classification compared various clinical data by Spearman’s rank correlation coefficient. with nailfold capillaroscopy. However, diagnostic characteristics Results with p<0.05 were considered statistically significant. like sensitivity, specificity as well as the predictive and clinical Demographic and clinical characteristics of enrolled subjects value of this potential biomarker for SSc need to be evalu- are reported in table 1. Important parameters for OCTA image ated in further, multicentre analyses including a larger number quality and validity, like best-corrected visual acuity (BCVA), of patients with greater clinical diversity to corroborate our intraocular pressure (IOP) and axial length (AL), showed no findings.

Ann Rheum Dis June 2019 Vol 78 No 6 857 Letters

Competing interests None declared. Table 1 Demographic and clinical characteristics of enrolled subjects Patient consent for publication Obtained. Ethics approval Institutional review board at the University of Lubeck (vote Control group SSc group reference number 17–008). (n=12) (n=12) P value Provenance and peer review Not commissioned; externally peer reviewed. Age (years) 64 (49–77) 64 (49–77) 1.000 Gender (F/M) 9/3 9/3 – © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ. Disease duration (months) N/A 64 (13–228) – MR and FR contributed equally. Classification (lcSSc/dcSSc) N/A 9/3 – VD (CCB, PDE5-I, ERA, PGI2-A) N/A 12/12 – MR and FR are joint first authors. IS (MTX, MMF, AZA, HCQ) N/A 9/12 – MRSS (0–51) N/A 6 (2–24) – Raynaud+ N/A 12/12 – To cite Rothe M, Rommel F, Klapa S, et al. Ann Rheum Dis 2019;78:857–858. Digital ulcers in the past N/A 10/12 – Received 6 October 2018 Arthritis N/A 6/12 – Revised 10 December 2018 Pulmonary fibrosis N/A 5/12 – Accepted 12 December 2018 ESR (mm) N/A 16 (1–60) – Published Online First 3 January 2019 C-reactive protein N/A 5.1 (0.3–16.6) – Ann Rheum Dis 2019;78:857–858. doi:10.1136/annrheumdis-2018-214541 ANA+ N/A 11/12 – Anti-RNA polymerase III+ N/A 1/12 – References 1 Allanore Y, Simms R, Distler O, et al. Systemic sclerosis. Nat Rev Dis Primers Anti-centromere+ N/A 5/12 – 2015;1:15002. Anti-Scl70+ N/A 5/12 – 2 Ushiyama O, Ushiyama K, Yamada T, et al. Retinal findings in systemic sclerosis: a Anti-Ro52+ N/A 6/12 – comparison with nailfold capillaroscopic patterns. Ann Rheum Dis 2003;62:204–7. MAP (mm Hg) 101.7 (87.3–113.3) 95 (85.7–109) 0.440 3 Waszczykowska A, Goś R, Waszczykowska E, et al. Prevalence of ocular manifestations in systemic sclerosis patients. Arch Med Sci 2013;9:1107–13. BCVA (logMAR) 0.0 (0.0–0.1) 0.0 (0.0–0.1) 0.477 4 Grennan DM, Forrester J. Involvement of the eye in SLE and scleroderma. A study using IOP (mm Hg) 15.5 (14–22) 14 (11–25) 0.059 fluorescein angiography in addition to clinical ophthalmic assessment. Ann Rheum Dis AL (mm) 22.8 (21.9–25.1) 22.9 (21.8–25.3) 0.597 1977;36:152–6. 5 Rommel F, Siegfried F, Kurz M, et al. Impact of correct anatomical slab segmentation on MAP was approximated using the measured SBP and DBP values: foveal avascular zone measurements by optical coherence tomography angiography in MAP= DBP+ SBP. Results are reported as median (range). Significant differences ⅔ ⅓ healthy adults. J Curr Ophthalmol 2018;30:156–60. were tested using the Mann–Whitney U test. P<0.05 was considered statistically 6 Conigliaro P, Cesareo M, Chimenti MS, et al. Evaluation of retinal microvascular density significant. in patients affected by systemic lupus erythematosus: an optical coherence tomography AL, axial length; AZA, azathioprine; BCVA, best-corrected visual acuity; CCB, calcium angiography study. Ann Rheum Dis 2019;78:287–9. channel blockers; DBP, diastolic blood pressure; dcSSc, diffuse cutaneous systemic sclerosis; ERA, endothelin receptor antagonist; F, female; HCQ, hydroxychloroquine; lcSSc, limited cutaneous systemic sclerosis; IOP, intraocular pressure; IS, immunosuppressants; M, male; MAP, mean arterial pressure; MMF, mycophenolate mofetil; MRSS, modified Rodnan skin score; MTX, methotrexate; N/A, not applicable; PDE5-I, phosphodiesterase type 5 inhibitor; PGI2-A, prostaglandin I2 (prostacyclin) analogue; SBP, systolic blood pressure; SSc, systemic sclerosis; VD, vasodilators.

Matthias Rothe,1,2 Felix Rommel,1,2 Sebastian Klapa,3 Jens Y Humrich,3 Relana Nieberding,3 Tanja Lange,3 Jan A M Sochurek,2 Pauline Plöttner,2 Salvatore Grisanti,1 Gabriela Riemekasten,3 Mahdy Ranjbar1,2 1Department of Ophthalmology, University of Lübeck, Lübeck, Germany 2Laboratory for Angiogenesis and Ocular Cell Transplantation, University of Lübeck, Lübeck, Germany 3Department of Rheumatology and Clinical Immunology, University of Lübeck, Lübeck, Germany Correspondence to Dr Mahdy Ranjbar, Department of Ophthalmology, University of Lübeck, Lübeck 23538, Germany; ​eye.​research101@​gmail.​com Correction notice This article has been corrected since it published Online First. The footnotes in table 1 have been corrected.

Handling editor Josef S Smolen Acknowledgements Burkhard Zander (Eyetec, Lübeck, Germany) and Ori Zahavi (Canon Europe, Amstelveen, Netherlands) for technical support regarding the OCTA device. Contributors MR and FR coordinated the ophthalmological examinations and data analysis. SK and JYH coordinated the rheumatological examinations and reviewed data analysis as well as revised the manuscript. RN contributed to the clinical data. TL contributed to the clinical data and revised the manuscript. JAMS and PP contributed to the collection of OCTA data. SG revised the manuscript. GR reviewed data analysis as well as revised the manuscript. MRa designed and coordinated the study, reviewed the data and wrote the manuscript.

858 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Comparison of autoantibody specificities tested by a line blot assay and immunoprecipitation- based algorithm in patients with idiopathic inflammatory myopathies

More than 15 autoantibodies have been identified in patients with idiopathic inflammatory myopathies (IIM). Most of them are specific to patients with IIM and therefore called myosi- tis-specific autoantibodies (MSA).1 There is no standardised approach of MSA testing to be used in clinical settings. Immu- noprecipitation (IP) of RNA and/or proteins in cellular lysates has traditionally been considered the golden standard for most autoantibodies but does not differentiate between antibodies targeting proteins with the same molecular weight and is not routinely available. Line blot assays (LB) represent a faster and semi-quantitative option to detect autoantibodies.2 3 Our aim was to compare a LB version with 16 specificities and an IP-based algorithm, as well as to describe clinical associations to autoanti- body specificities in a cohort of patients classified as IIM. The first available sera collected between 2013 and 2017, from 110 patients classified as having IIM4 followed at the rheu- matology clinic at Karolinska University Hospital (Stockholm, Sweden) and sera from 60 healthy controls were included. We used a commercial LB assay (Euroline Myositis Antigen Profile 4, Euroimmun, Lübeck, Germany) with densitometrical evaluation

858 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Table 1 Frequencies of autoantibodies detected by LB and IP assays across EULAR/ACR myositis subtypes DM ADM JDM PM IBM Total Controls 36 (32) 4 (4) 3 (3) 52 (47) 16 (14) 110 (100) 60 No of subjects (%) LB IP LB IP LB IP LB IP LB IP LB IP LB Anti-Jo1 7 4 1 1 1 1 9 6 0 0 18 12 0 Anti-PL7 00000001000 10 Anti-PL12 00000000101 00 Anti-Ej 00000000000 00 Anti-Oj 00000001000 10 Anti-SRP 00000034003 40 Anti-PmScl (75/100kD) 12000023003 50 Anti-Ku 11001121004 31 Anti-SAE1 11000032004 30 Anti-NXP2 20000000002 01 Anti-MDA5 12010000001 30 Anti-TIFγ 7 11 0 0 0 1 1 0 0 0 8 12 0 Anti-Mi2 (α/β) 21000050118 21 Any autoantibody 21 21 1 2 2 3 22 18 2 1 48 45 3 ACR, American College of Rheumatology; ADM, amyopathic dermatomyositis; DM, dermatomyositis; EULAR, European League Against Rheumatism; IBM, inclusion body myositis; IP, immunoprecipitation; JDM, juvenile dermatomyositis; LB, line blot; PM, polymyositis. of staining intensity for 16 IgG autoantibodies at the Depart- by LB 63% and by IP 58% had cancer-associated myositis (Table ment of Clinical Immunology, Uppsala University Hospital. The S1 and S2). Anti-SAE1 autoantibodies were detected in four results were reported as negative or positive (0–10 vs ≥11 densi- patients by LB, and three of these sera immunoprecipitated two tometry units) according to the manufacturer’s instructions. 40 kDa and 90 kDa proteins, consistent with the two subunits of For the IP protocol, samples were analysed by protein-IP and the heterodimer small-ubiquitin-like modifier activating enzyme RNA- IP at Kyoto University.5 6 When these results suggested (SAE1/SAE2) suggesting IP reactivity to SAE in 75% of the antibodies targeting a 140kD protein, an in-house anti-MDA5 LB+ samples.8 Anti-Mi2 autoantibodies were more prevalent ELISA was performed (online Supplementary data). Categorical by LB than by IP. A plausible explanation is that the antigens 2 variables were analysed by χ or Fisher’s test. Agreement was are different between the two methods. The LB applied in our 7 calculated by Cohen’s kappa. study included both alpha and beta proteins of the Mi2 protein, whether this affects the specificity could not be addressed in our study. However, our result is consistent with a previous report Results demonstrating positive anti-Mi2 autoantibodies in PM patients The frequency of any autoantibody was similar by both assays in European populations.9 (LB 48/111=43%; IP 45/111=41%, p=0.79). The most In conclusion, the concordance rate between the LB and IP frequent MSAs were anti-Jo-1 (LB: 16%, IP: 10%) and anti- assays for detection of MSA in patients classified as IIM according TIF1γ (LB: 7%, IP: 11%) followed by anti-Mi-2 α or β (LB: 7% to the European League Against Rheumatism/American College IP: 2%), anti-SAE1 (LB: 4% IP:3%) and anti-SRP (LB: 3% IP: of Rheumatology criteria was moderate for the most prevalent 4%). Patients with inclusion body myositis (IBM) and controls MSA. The low sensitivity for anti-TIF1 gamma antibody by LB is were negative for Jo-1, PL-7, SRP, MDA5 and TIF1γ specifici- of concern, as this is a marker of cancer-associated DM. LB assay ties in all assays. Three controls had either anti-Ku, anti-NXP2 seems to be valid and useful to identify subgroups of IIM with or anti-Mi-2α by LB (mean 20 units), corresponding to a total specific clinical features. diagnostic specificity of 99.7% (table 1). Fabricio Espinosa-Ortega,‍ ‍ 1 Marie Holmqvist,2 Helene Alexanderson,1,3 The overall concordance rate between the two assays was 78% 4 5 1 6 with moderate agreement (κ: 0.54). We found very good agree- Helena Storfors, Tsuneyo Mimori, Ingrid E Lundberg, Johan Rönnelid 1 ment for anti-SRP (k: 0.85), anti-Ku (κ:0.85) and anti-SAE1 Division of Rheumatology, Department of Medicine, Karolinska Institutet, (κ:0.85) and good agreement for anti-Jo-1 (κ:0.69). Agreement Stockholm, Sweden 2Clinical Epidemiology Unit, Department of Medicine Solna, Karolinska Institutet, for anti-PmScl (κ:0.48), anti-MDA5 (κ:0.49), and anti-TIF1γ Stockholm, Sweden (κ:0.56) was moderate (table 2). 3Division of Physiotherapy, Department of Neurobiology Care Sciences and Society, Anti-Jo-1 autoantibodies were detected in 18 sera, 11 were Karolinska Institutet, Stockholm, Sweden LB+/IP+, 7 were LB+/IP- and 1 was LB-/IP+. Sera from five 4Department of Clinical Immunology and Transfusion Medicine, Uppsala University Hospital, Uppsala, Sweden LB+/IP- patients were also evaluated by a commercial anti-Jo1 5 ELISA (Profile, Phadia, Freiburg, Germany): three were positive, Department of Rheumatology and Clinical Immunology, Kyoto University Graduate school of Medicine, Kyoto, Japan one borderline and one negative. Anti-Jo-1 LB+ patients had 6Department Immunology, Genetics and Pathology, Uppsala University, Uppsala, a higher frequency of interstitial lung disease (p<0.001) and Sweden arthritis (p<0.001) compared with anti-Jo-1 negative patients Correspondence to Professor Ingrid E Lundberg, Division of Rheumatology, (Table S1). Nineteen per cent of DM/PM patients were Jo-1+ Department of Medicine, Karolinska University Hospital, SE-171 76 Stockholm, compared with 0% of IBM (p<0.001). Anti-TIF1γ was detected Sweden; ingrid.​ ​lundberg@ki.​ ​se in 14 patients; 6 patients were LB+/IP+ (43%), 2 were LB+/IP- (14%) and 6 were LB-/IP+ (43%). Among anti-TIF1γ+patients, Handling editor Josef S Smolen

Ann Rheum Dis June 2019 Vol 78 No 6 859 Letters

4 Lundberg IE, Tjärnlund A, Bottai M, et al. European League against Rheumatism/ Table 2 Comparisons between immunoprecipitation and line blot American College of rheumatology classification criteria for adult and juvenile assays in 110 patients with inflammatory myopathies according idiopathic inflammatory myopathies and their major subgroups. Ann Rheum Dis EULAR/ACR criteria 2017;2017:2271–82. 5 Nakashima R, Imura Y, Kobayashi S, et al. The RIG-I-like receptor IFIH1/MDA5 is a Concordance Concordance dermatomyositis-specific autoantigen identified by the anti-CADM-140 antibody. LB IP Cohen’s on positive on negative Rheumatology 2010;49:433–40. Antibody n (%) n (%) kappa sera (%)* sera (%)* 6 Forman MS, Nakamura M, Mimori T, et al. Detection of antibodies to small nuclear Anti-Jo-1 18 (16) 12 (11) 0.7 11 (10) 92 (83) ribonucleoproteins and small cytoplasmic ribonucleoproteins using unlabeled cell Anti-PL-12 1 (1) 0 0.0 NA NA extracts. Arthritis Rheum 1985;28:1356–61. 7 Landis JR, Koch GG. The measurement of observer agreement for categorical data. Anti-PL-7 0 0 NA NA NA Biometrics 1977;33:159–74. Anti-Ej 0 0 NA NA NA 8 Betteridge Z, Gunawardena H, North J, et al. Identification of a novel autoantibody Anti-OJ 0 1 (1) 0.0 NA NA directed against small ubiquitin-like modifier activating enzyme in dermatomyositis. Anti-SRP 3 (3) 4 (4) 0.85 3 (3) 107 (96) Arthritis Rheum 2007;56:3132–7. 9 Hengstman GJ, Vree Egberts WT, Seelig HP, et al. Clinical characteristics of patients Anti-PmScl 3 (3) 5 (5.0) 0.48 2 (2) 105 (95) with myositis and autoantibodies to different fragments of the Mi-2 beta antigen. Ann (75/100kD) Rheum Dis 2006;65:242–5. Anti-Ku 4 (4) 3 (3) 0.85 3 (3) 107 (96) Anti-SAE 4 (4) 3 (3)† 0.85 3 (3) 107 (96) Anti-NXP2 2 (2) 0 0.0 NA NA Anti-MDA5 1 (1) 3 (3) 0.49 1 (1) 108 (97) Anti-TIF1γ 8 (7) 12 (11) 0.56 6 (5) 97 (87) Anti-Mi-2(α/β) 8 (7) 2 (2) 0.38 2 (2) 103 (93) *The percentage represent the proportion among the 110 patients. †The IP assay showed 40/90kD bands in ¾ patients, suggesting a positivity for SAE1. ACR, American College of Rheumatology; EULAR, European League Against Rheumatism; IP, immunoprecipitation; LB, line blot.

Contributors Study concept and design: IEL, JR. Acquisition of data: JR, HS, TM, FE. Statistical analysis: FE. Analysis and interpretation of data, drafting manuscript and critical revision of manuscript: all authors. Funding The Swedish Research Council and King Gustav V:th 80-year Foundation. Competing interests IEL has received honoraria from Bristol Myers Squibb and MedImmune and is currentlyreceiving a research grant from Bristol Myers Squibb and from Astra Zeneca; and is a scientificadvisor for Bristol Myers Squibb, aTyr and IDERA. Patient consent for publication Not required. Ethics approval Our study had the permit for antibody analysis from the Ethics Committee in Stockholm, Sweden. Provenance and peer review Not commissioned; externally peer reviewed. © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ.

►► Additional material is published online only. To view please visit the journal online (http://​dx.​doi.​org/​10.​1136/​annrheumdis-​2018-​214690).

To cite Espinosa-Ortega F, Holmqvist M, Alexanderson H, et al. Ann Rheum Dis 2019;78:858–860.

Received 2 November 2018 Accepted 31 January 2019 Published Online First 13 February 2019 Ann Rheum Dis 2019;78:858–860. doi:10.1136/annrheumdis-2018-214690

References 1 Casciola-Rosen L, Mammen AL. Myositis autoantibodies. Curr Opin Rheumatol 2012;24:602–8. 2 Zampieri S, Ghirardello A, Doria A, et al. The use of Tween 20 in immunoblotting assays for the detection of autoantibodies in connective tissue diseases. J Immunol Methods 2000;239:1–11. 3 Ghirardello A, Doria A, Zampieri S, et al. Anti-ribosomal P protein antibodies detected by immunoblotting in patients with connective tissue diseases: their specificity for SLE and association with IgG anticardiolipin antibodies. Ann Rheum Dis 2000;59:975–81.

860 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Addressing immune-related adverse events of cancer immunotherapy: how prepared are rheumatologists?

Cancer immunotherapy blocking immune checkpoints represents a major advance in oncology. Immune checkpoint inhibitors (ICI) targeting cytotoxic T-lymphocyte-associated antigen 4 and programmed cell death 1 (PD-1) or its ligand PD-L1 have become standard of care in many advanced-stage cancers. Yet an important proportion of patients experience inflammatory or autoimmune side effects, also known as immune-related adverse events (irAEs), as a consequence of dysregulated immunity. irAEs can affect almost any organ system.1 Notably, an expanding range of manifestations mimicking our classical rheumatic diseases have been described.2 3 Rheumatic irAEs, including inflammatory arthritis, pseudomyalgia rheumatica, sicca syndrome, myositis and vasculitis, are increas- ingly reported. Since there has been increasing emphasis on this emerging field within the last 3 years, we aimed to evaluate the knowledge of rheumatologists, as well as other specialists, regarding ICI and irAEs through an online survey. Survey questions addressed the demographics (gender, age, practice setting and duration of medical practice), domains of awareness, clinical experience and interest in irAE-specific medical education. The survey was first distributed in the USA (2016) to healthcare provider cohorts at Johns Hopkins University (JHU) and Cleveland Clinic (CC), then in France (2018) to rheumatol- ogists and other specialists via several national expert networks: Société Française de Rhumatologie, Club Rhumatismes et Inflam- mations, Société Nationale Française de Médecine Interne, Groupe d’Etude Thérapeutique des Affections Inflammatoires du Tube Digestif and Société Française d’Endocrinologie. Overall, 153 rheumatologists from the USA (JHU n=39; CC n=114) and 349 French specialists (rheumatologists n=159; internists n=113; endocrinologists n=55; gastroenterologists n=22) participated in this study. As detailed in table 1, rheu- matologists from private practices and academic institutions predominated, with various levels of experience. Half of the other specialists worked in an academic setting. In 2016, 67% of rheumatologists from the USA were unaware or have just heard of irAEs, compared with 28% of French rheumatolo- gists and 22% of other French specialists in 2018. Of note, the majority of French participants reported some basic knowl- edge. This can be explained mainly by the timing of the survey and thus increasing physician exposure to patients with irAEs

860 Ann Rheum Dis June 2019 Vol 78 No 6 Letters in 2018: 86 of 159 rheumatologists (54%) and 128 of 190 gastroenterologists reported more frequently to be familiar other specialists (67%) had exposure to patients with irAEs, with irAEs and their management. Description and recognition compared with 18 of 122 rheumatologists (15%) in 2016. of irAEs were perceived to be the biggest educational need in However, despite this 2-year time interval and the fact that 2016, while in 2018 the type of educational content requested the study was conducted in two different countries hardly by 80% of physicians focused on treatment algorithms, more comparable, the resulting conclusion of both surveys is the relevant for providers already engaged in managing irAEs. same: familiarity with ICI and irAEs had been still reported by This study highlights that rheumatologists, as well as other less than 30% of rheumatologists, and less than 10% declared specialists, had limited experience and lacked confidence in the being very confident in treating such patients. Nevertheless, management of irAEs. The same observation has been reported we did observe some improvement as the majority of French by our Portuguese colleagues during the last European League rheumatologists were somewhat (20%) or moderately (41%) Against Rheumatism (EULAR) meeting, with 64% of rheumatol- confident managing irAEs in 2018. In France, systemic auto- ogists being unfamiliar with these new therapies, according to a immune diseases are managed by both rheumatologists and web-based questionnaire sent in November 2017.4 Since the first internists. We observed very similar results between French dissemination of the survey in 2016, shortly after the approval of rheumatologists and internists, while endocrinologists and ICIs, the number of patients treated with ICIs has considerably

Table 1 Demographics and knowledge of ICI and irAEs among rheumatologists and other specialists, n (%) US rheumatologists (2016; French rheumatologists (2018; French endocrinologists and n=153) n=159) French internists (2018; n=113) gastroenterologists (2018; n=77) Gender Male 93 (61) 81 (51) 72 (64) 31 (40) Female 60 (39) 78 (49) 41 (36) 46 (60) Practice setting (n=150) Academic hospital 43 (29) 54 (34) 57 (50) 36 (47) Non-academic hospital 25 (17) 29 (18) 38 (34) 19 (25) Private practice 68 (45) 41 (26) 9 (8) 10 (13) Mixed (private and hospital) – 32 (20) 6 (5) 11 (14) Residents/in training – 3 (2) 3 (3) 1 (1) Industry 5 (3) – – – Others 9 (6) – – – Years of medical practice <10 Mean (SD): 22 (13) 41 (26) 39 (34.5) 21 (27.5) 10–20 45 (28) 38 (34) 24 (31) 20–30 38 (24) 21 (18.5) 18 (23.5) >30 35 (22) 15 (13) 14 (18) Knowledge of ICI (n=133) None 29 (22) 5 (3) 3 (3) 3 (4) I have heard of ICI 68 (51) 49 (31) 25 (22) 11 (14) Some basic knowledge/unsure 2 (1.5) 71 (44) 62 (55) 39 (51) Familiar 32 (24) 27 (17) 15 (13) 19 (25) Very familiar 2 (1.5) 7 (5) 8 (7) 5 (6) Knowledge of irAEs (n=130) None 54 (42) 14 (9) 8 (7) 5 (6) I have heard of irAEs 33 (25) 30 (19) 21 (19) 7 (9) Some basic knowledge/unsure 3 (2) 72 (45) 57 (50) 32 (42) Familiar 32 (25) 31 (19) 17 (15) 24 (31) Very familiar 8 (6) 12 (8) 10 (9) 9 (12) Management of patients with irAEs (n=122) Never 104 (85) 58 (36) 40 (35.5) 11 (14) Maybe, but I didn’t link symptoms – 15 (10) 7 (6) 4 (5.5) with ICI Rarely 18 (15) 64 (40) 45 (40) 31 (40.5) Several times (>5 patients) 17 (11) 16 (14) 17 (22) Frequently (>10 patients) 5 (3) 5 (4.5) 14 (18) Level of confidence (n=122) I don’t know the topic – 21 (13) 6 (5) 7 (9) Not confident 74 (61) 28 (18) 30 (27) 10 (13) Somewhat confident 30 (25) 32 (20) 35 (31) 16 (21) Moderately confident 14 (11) 65 (41) 34 (30) 27 (35) Very confident 4 (3) 13 (8) 8 (7) 17 (22) ICI, immune checkpoint inhibitors; irAEs, immune-related adverse events.

Ann Rheum Dis June 2019 Vol 78 No 6 861 Letters increased. The increase in ICI use, and the resultant increase in 3 Benfaremo D, Manfredi L, Luchetti MM, et al. Musculoskeletal and rheumatic diseases patients with irAEs, likely explains why more rheumatologists induced by immune checkpoint inhibitors: a review of the literature. Curr Drug Saf 2018;13:150–64. reported managing irAEs and were more familiar with them, as 4 Araújo F, Fonseca JE. Physician awareness of rheumatic immune-related adverse indicated by the second dissemination of the survey in 2018, events in cancer patients treated with immune checkpoint inhibitors. Ann Rheum Dis but we are still a long way from having achieved the objective. 2018;77:A1777. Specific education around rheumatic irAEs has been imple- 5 Haanen J, Carbonnel F, Robert C. Management of toxicities from immunotherapy: ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol mented with (1) local multidisciplinary meetings in several insti- 2018. tutions, (2) dedicated sessions during national and international 6 Puzanov I, Diab A, Abdallah K, et al. Managing toxicities associated with meetings, and (3) guidelines published to help oncologists and immune checkpoint inhibitors: consensus recommendations from the Society for organ specialists in the management of irAEs.5–7 Recommen- immunotherapy of cancer (SITC) toxicity management Working Group. J Immunother dations for the diagnosis and management of musculoskeletal Cancer 2017;5:95. 7 Brahmer JR, Lacchetti C, Schneider BJ, et al. Management of immune- irAEs due to cancer immunotherapy will be proposed by EULAR related adverse events in patients treated with immune checkpoint inhibitor in a near future. All these educational efforts will help rheuma- therapy: american Society of clinical oncology clinical practice guideline. JCO tologists and other specialists to become more confident with 2018;36:1714–68. this evolving clinical field.

Marie Kostine,‍ ‍ 1 Laura C Cappelli,2 Cassandra Calabrese,‍ ‍ 3 Leonard H Calabrese,3 Clifton O Bingham III,2 Christophe Richez,1 Jacques-Eric Gottenberg,4 Olivier Lambotte5 1Rheumatology, Centre Hospitalier Universitaire de Bordeaux, Bordeaux, France 2Rheumatology, Johns Hopkins University, Baltimore, Maryland, USA 3Rheumatology/Immunology, Cleveland Clinic, Cleveland, Ohio, USA 4Rheumatology, Hopitaux Universitaires de Strasbourg, Strasbourg, France 5Internal Medicine and Clinical Immunology, Hopital Bicetre, Le Kremlin-Bicetre, France Correspondence to Dr Marie Kostine, Rheumatology, Centre Hospitalier Universitaire de Bordeaux, Bordeaux 33000, France; ​marie.​kostine@​chu-​bordeaux.​fr

Handling editor Josef S Smolen Acknowledgements We thank all respondents for their active participation. Collaborators Société Française de Rhumatologie (SFR), Club Rhumatismes et Inflammations (CRI), Société Nationale Française de Médecine Interne (SNFMI), Groupe d’Etude Thérapeutique des Affections Inflammatoires du Tube Digestif (GETAID) and Société Française d’Endocrinologie (SFE). Contributors LCC, CC, LHC and COB planned and conducted the study in the USA. OL, J-EG, CR and MK conducted the study in France. MK, LCC and CC drafted the manuscript, and all the authors critically reviewed and approved the final version of the manuscript. Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests MK received speaking fees from BMS. LCC received research grant from BMS and reported consulting for Regeneron. COB has served as consultant for BMS, Regeneron and Genentech/Roche, and received grant support from BMS. CR received speaking fees from BMS, AstraZeneca and Genentech/Roche. J-EG received honoraries from BMS and Pfizer and research grant from BMS. OL received paid expert testimony and consultancy fees from BMS France, MSD and AstraZeneca, consultancy fees from Genzyme, and expert testimony for Janssen. CC and LHC declared no conflict of interest for this work. Patient consent for publication Not required. Provenance and peer review Not commissioned; externally peer reviewed. © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ.

To cite Kostine M, Cappelli LC, Calabrese C, et al. Ann Rheum Dis 2019;78:860–862.

Received 14 November 2018 Accepted 2 January 2019 Published Online First 18 January 2019 Ann Rheum Dis 2019;78:860–862. doi:10.1136/annrheumdis-2018-214748

References 1 Postow MA, Sidlow R, Hellmann MD. Immune-related adverse events associated with immune checkpoint blockade. N Engl J Med Overseas Ed 2018;378:158–68. 2 Cappelli LC, Gutierrez AK, Bingham CO, et al. Rheumatic and musculoskeletal immune- related adverse events due to immune checkpoint inhibitors: a systematic review of the literature. Arthritis Care Res 2017;69:1751–63.

862 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

3D culture of Erdheim-Chester disease tissues unveils histiocyte metabolism as a new therapeutic target

Erdheim-Chester disease (ECD) is a rare non-Langerhans cell histiocytosis, characterised by tissue infiltration by foamy CD68+ CD1a− histiocytes.1 The disease has pleomorphic clinical mani- festations, including long bones and extraskeletal involvement, and may be life-threatening, particularly when heart and central nervous system are affected.1 ECD histiocytes secrete proinflammatory cytokines2 and carry activating mutations along the RAS-RAF-MEK-ERK protein kinase signalling pathway, most commonly the BRAFV600E onco- genic mutation.3 4 Accordingly, patients with ECD have been treated with cytokine inhibitors, including infliximab,1 5 and, more recently, with the BRAFV600E inhibitor vemurafenib.6 The latter, however, induces sustained but partial clinical responses and recurrences on discontinuation,6 underlining the need for more effective therapeutic strategies. To identify the outcomes downstream constitutive ERK phos- phorylation in ECD histiocytes and their response to small mole- cule-based inhibition, we performed three-dimensional (3D) culture of tissues from three BRAFV600E-mutated ECD patients in the RCCS bioreactor7 (and online supplementary methods) in the presence/absence of vemurafenib or infliximab, used as control. All patient samples maintained production of prototypical cytokines and chemokines2 in bioreactor, thus validating this technology also for ECD; moreover, infliximab, and, to a lesser degree, vemurafenib, significantly decreased cyto-chemokines and soluble tumour necrosis factor (TNF) receptor (sTNF-R) levels (figure 1A), remarkably mirroring the downmodulation reported in treated patients5 (and data not shown). Immunohistochemistry (IHC) on ECD tissues in bioreactor showed preserved viability, histoarchitecture and expression of histiocyte lineage and activation markers (figure 1B) for up to 6 days. Vemurafenib specifically targeted mutated ECD histiocytes, as demonstrated by decreased ERK phosphorylation, without affecting viability or persistence of CD68+ cells (figure 1B and online supplementary figure S1). Conversely, the drug decreased TNF-α production (figure 1B, f, f1) paralleling the significantly reduced cytokine release in supernatants (figure 1A). Cancer metabolism has gained renewed and growing interest, being reprogrammed glucose metabolism from oxidative phos- phorylation to aerobic glycolysis and increased lactate production the best characterised metabolic phenotype.8 Oncogenic BRAF

862 Ann Rheum Dis June 2019 Vol 78 No 6 Letters

Figure 1 Vemurafenib treatment affects glycolytic metabolism in Erdheim-Chester disease (ECD) tissues cultured in bioreactor. (A) Tissues from three patients with ECD were cultured in bioreactor in the absence (black bars) or presence of either the tumour necrosis factor (TNF) inhibitor infliximab (10 µg/mL, grey bars) or the BRAFV600E inhibitor vemurafenib (6 µM, white bars). Concentrations of cytokines, chemokines and soluble TNF receptors (sTNF-R) I and II (the latter bona fide expression of TNF activity5) were determined in day 2 culture supernatants by Bio-Plex Multiplex Cytokine assay and ELISA assay, respectively. Data are means±SD of triplicate values. (B) ECD pleural fragments from patient 3 were cultured in the presence/absence of vemurafenib (vem, 6 µM). Samples were then retrieved at day 2 (a–e, a1–e1) and day 6 (f, f1), fixed and submitted to immunohistochemistry (IHC). Bars represent 500 and 100 µm. (C) Glucose (upper) and lactate (lower) concentrations were determined in the supernatants from ECD tissues in bioreactor in the presence/absence of vemurafenib (vem, 6 µM) and from normal skin samples as a control. Data are means±SD of triplicate values. Statistical analysis was performed using Student’s t-test. *P≤0.05; **P≤0.01; ***P≤0.001. Casp-3, caspase-3; NT, untreated; p-ERK, phospho-ERK; TCM, tissue culture medium. affects glucose metabolism by multiple mechanisms, including expression and lactate production, and its reversal on vemurafenib upregulation of the glucose transporter Glut-1.8 Glut-1 is also treatment, delineate aerobic glycolysis as a novel BRAF-driven upregulated by hypoxia and inflammation in rheumatic and feature of ECD histiocytes. Since Glut-1 is the transporter for the cardiovascular diseases.9 Accordingly, a fraction of CD68+ ECD radiotracer of 18F-fluorodeoxyglucose (FDG-PET), these data may histiocytes expressed Glut-1, as shown by IHC (online supplemen- provide the molecular basis for the well-known efficacy of the tary figure S2A,B) and fluorescence-activated cell sorting analysis technique in monitoring response to therapy in ECD.1 (online supplementary figure S2C), and retained its expression in Although not proliferating,2 ECD histiocytes may experience culture in bioreactor (figure 1B,d). Given the unique opportunity increased energy requirement for the production of immune to assess tissue metabolism through metabolite determination in mediators, as for inflammatory macrophages. In bioreactor, vemu- bioreactor supernatants, we determined lactate levels, which were rafenib decreased their cytokine/chemokine release, conceivably increased, compared with basal levels and normal skin samples because of impaired metabolism, but not their viability, at variance (figure 1C). Notably, vemurafenib downmodulated both Glut-1 with other BRAF-mutated tumours. The outcome of vemurafenib expression (figure 1B,d1) and lactate production (figure 1C). treatment in ECD histiocytes may depend on the activation of Altogether, we here demonstrate, as a proof of concept, that adaptive responses to energy deprivation, including autophagy, the bioreactor technology allows investigating ECD pathophys- or, conversely, on detrimental features (hypoxia, acidosis, limited iology and response to drugs, thus overcoming the lack of cell nutrients availability) of the native microenvironment, possibly lines and suitable animal models. Significantly, upregulated Glut-1 overcome by our culture conditions in bioreactor.7

Ann Rheum Dis June 2019 Vol 78 No 6 863 Letters

Further exploitation of the 3D culture system is needed to Patient consent Obtained. identify and harness metabolic signalling and supporting adaptive © Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and responses in mutated histiocytes, in the perspective of designing permissions. Published by BMJ. new therapeutic strategies for patients with ECD. LD, EF and MF contributed equally.

Antonello Villa,1 Daniela Belloni,2 Barbara Vergani,1 Simone Cenci,3 Giulio Cavalli,4,5 Riccardo Biavasco,5,6 Monica Rodolfo,7 8 5,8 4,5 Maria Giulia Cangi, Claudio Doglioni, Lorenzo Dagna, To cite Villa A, Belloni D, Vergani B, et al. Ann Rheum Dis 2019;78:862–864. Elisabetta Ferrero,2 Marina Ferrarini2 1Consorzio MIA, University of Milano-Bicocca, Milan, Italy Received 12 September 2018 2Division of Experimental Oncology, San Raffaele Scientific Institute, Milan, Italy Revised 14 November 2018 3Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy Accepted 20 November 2018 4Unit of Immunology, Rheumatology, Allergy and Rare Diseases, San Raffaele Published Online First 28 November 2018 Scientific Institute, Milan, Italy Ann Rheum Dis 2019;78:862–864. doi:10.1136/annrheumdis-2018-214432 5Vita-Salute University, Milan, Italy 6San Raffaele Telethon Institute for Gene Therapy (SR-Tiget), San Raffaele Scientific Institute, Milan, Italy References 7Department of Experimental Oncology and Molecular Medicine, IRCCS, Istituto 1 Diamond EL, Dagna L, Hyman DM, et al. Consensus guidelines for the diagnosis and Nazionale per Cura dei Tumori, Milan, Italy clinical management of Erdheim-Chester disease. Blood 2014;124:483–92. 8Pathology Unit, San Raffaele Scientific Institute, Milan, Italy 2 Stoppacciaro A, Ferrarini M, Salmaggi C, et al. Immunohistochemical evidence of a cytokine and chemokine network in three patients with Erdheim-Chester disease: Correspondence to Marina Ferrarini, Division of Experimental Oncology, San implications for pathogenesis. Arthritis Rheum 2006;54:4018–22. Raffaele Scientific Institute, Milan 20132, Italy;​ ferrarini.​marina@​hsr.​it 3 Haroche J, Charlotte F, Arnaud L, et al. High prevalence of BRAF V600E mutations in Erdheim-Chester disease but not in other non-Langerhans cell histiocytoses. Blood Handling editor Josef Smolen 2012;120:2700–3. 4 Cangi MG, Biavasco R, Cavalli G, et al. BRAFV600E-mutation is invariably present and Acknowledgements The authors thank Licia Rivoltini and Giovanna Musco for associated to oncogene-induced senescence in Erdheim-Chester disease. Ann Rheum critical support and advices, and Mauro Motta, Marta Strollo and Silvia Heltai for Dis 2015;74:1596–602. technical assistance. The authors also thank Kathy Brewer and the ECD Global 5 Dagna L, Corti A, Langheim S, et al. Tumor necrosis factor α as a master regulator of Alliance (DeRidder, LA) for their unceasing support. inflammation in Erdheim-Chester disease: rationale for the treatment of patients with Contributors AV, EF and MF designed the study. DB and BV performed the infliximab. J Clin Oncol 2012;30:e286–e290. experiments and collected the data. CD reviewed the histological samples. MR 6 Haroche J, Cohen-Aubart F, Emile JF, et al. Reproducible and sustained efficacy of provided experimental tools. MGC determined the BRAF status. DB conducted the targeted therapy with vemurafenib in patients with BRAF(V600E)-mutated Erdheim- statistical analysis. AV, SC, GC, RB, EF and MF analysed and interpreted the data. GC Chester disease. J Clin Oncol 2015;33:411–8. collected clinical information. LD diagnosed and followed patients with ECD and 7 Ferrarini M, Steimberg N, Ponzoni M, et al. Ex-vivo dynamic 3-D culture of human contributed to discussion. AV, EF and MF wrote the manuscript. All authors critically tissues in the RCCS™ bioreactor allows the study of Multiple Myeloma biology and reviewed and approved the final version of the manuscript. response to therapy. PLoS One 2013;8:e71613. 8 Hay N. Reprogramming glucose metabolism in cancer: can it be exploited for cancer Funding This study was funded by the ECD Global Alliance and by the Italian therapy? Nat Rev Cancer 2016;16:635–49. Ministry of Health (grant number GR-2009-1594586). 9 Biniecka M, Canavan M, McGarry T, et al. Dysregulated bioenergetics: a key regulator Competing interests None declared. of joint inflammation. Ann Rheum Dis 2016;75:2192–200.

864 Ann Rheum Dis June 2019 Vol 78 No 6 Miscellaneous Correction: Autoantibodies to citrullinated proteins induce joint pain independent of inflammation via a chemokine- dependent mechanism

Wigerblad G, Bas DB, Fernandes-Cerqueira C, et al. Autoantibodies to citrullinated proteins induce joint pain independent of inflammation via a chemokine-dependent mechanism. Ann of Rheum Dis 2016;75:730–. doi:10.1136/annrheumdis-2015-208094. The specificity of the human monoclonal antibodies B02 and D10 used in functional exper- iments in this article, originally described as high affinity ACPAs, has been re-evaluated. In accordance with data that were made available to us (Ge et al., Arthritis Rheumatol, 2019; 71:210–221, and others), the two monoclonal antibodies used lack specific binding to citrulli- nated peptides in surface plasmon resonance (SPR) and other assays as described in the retrac- tion note to Journal of Experimental Medicine (Amara et al Retraction J. Exp Med 2019; 216:245). As such the functional results reported for these monoclonal antibodies cannot be attributed to reactivity against citrullinated proteins and/or peptides, but are due to other yet unknown mechanisms. Thus, the pathogenetic implications derived from these experiments cannot be maintained as stated. Specifically, this relates to findings presented in figure 2 A-D and 6 B-E. Note that in: Figure 2A the table ranking reactivity against citrullinated proteins and/or peptides is based on the original ELISA from the retracted Amara et al Retraction J. Exp Med 2019; 216:245; the results from this ELISA were not reproducible in other assays. In light of the lack of specificity of these monoclonal antibodies, the functional results observed in Figures 2C-D (reduction in withdrawal thresholds after injection of B02 and D10 monoclonal antibodies); 6B-C (B02/D10-induced mechanical hypersensitivity described as reduction in mechanical hypersensitivity and algesic index); and 6D-E (B02/D10-induced thermal hypersensitivity described as a change in withdrawal latency), confirming the suggested CXCL1 mediation of hypersensitivity using the CXCL1/XCL2 antagonist reparixin, cannot be attributed to reactivity against citrullinated proteins and/or peptides and must have been due to other, hitherto unknown mechanisms. Since the monoclonal antibodies were used to confirm and expand the data obtained with polyclonal antibody preparations, the remaining conclusions in the paper rely on the data from the polyclonal IgG antibodies purified by affinity chromatography on CCP2-linked Sepharose columns. Although the pronociceptive effects in mice were induced by the CCP2- column eluate IgG (endotoxin free) and not the flow through fractions, also these results have to be interpreted with caution and we conclude that the pathogenetic specificity of these obser- vations and the detailed pronociceptive mechanisms remain to be elucidated. Consequently, we feel that the title on page 730 should be corrected from “Autoantibodies to citrullinated proteins induce joint pain independent of inflammation via a chemokine-dependent mechanism” to: “Autoantibodies to citrullinated proteins may induce joint pain independent of inflammation”. We specifically apologise for the delays from our side in communicating the information in this correction note to the readership of ARD.

Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non- commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.​ ​org/​licenses/​by-​nc/​ 4.0/.​ © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ. Ann Rheum Dis 2019;78:865. doi:10.1136/annrheumdis-2015-208094corr1

Ann Rheum Dis 2019;78:865. doi:10.1136/annrheumdis-2015-208094corr1 865 Miscellaneous Correction: Identification of a novel chemokine-dependent molecular mechanism underlying rheumatoid arthritis- associated autoantibody-mediated bone loss

Krishnamurthy A, Joshua V, Haj Hensvold A, et al. Identification of a novel chemokine-depen- dent molecular mechanism underlying rheumatoid arthritis-associated autoantibody-mediated bone loss. Ann of Rheum Dis 2016;75:721–9. doi:10.1136/annrheumdis-2015-208093. The specificity of the human monoclonal antibodies B02 and D10 used in functional exper- iments in this article, originally described as high affinity ACPAs has been re-evaluated. In accordance with data from others that were made available to us in the past, the two mono- clonal antibodies used lack specific binding to citrullinated peptides in surface plasmon reso- nance (SPR) and other assays as described in the retraction note to Journal of Experimental Medicine (Amara et al Retraction J. Exp Med 2019; 216:245). As such the functional results reported for these monoclonal antibodies cannot be attributed to reactivity against citrul- linated proteins and/or peptides, but are due to other yet unknown mechanisms. Thus, the pathogenetic implications derived from these experiments cannot be upheld as stated. In light of the lack of specificity of these monoclonal antibodies, the functional results observed in Figures 1 B-E (effects of monoclonal antibodies on osteoclast formation and bone loss); Figure 2 A-B (Osteoclast stainings using monoclonal antibodies); and Figure 5 (effect of the monoclonal antibodies on bone density in mice) should not be attributed to reactivity against citrullinated proteins and/or peptides, but must have been due to other, hitherto unknown mechanisms. Since the monoclonal antibodies were used to confirm and expand the data obtained with polyclonal antibody preparations, the remaining conclusions in the paper rely on the data from the polyclonal IgG antibodies purified by affinity chromatography on CCP2-linked Sepharose columns. Although all effects on osteoclast activation were seen for the CCP2-column eluate and not in the flow through fractions and these effects could be blocked by both PAD inhib- itors and blockade of IL-8, also these results have to be interpreted with caution waiting for additional mechanistic studies. Thus, the pathogenetic implications provided in Figure 6 are still hypothetical and rely on data from the polyclonal preparations, in light of the lack of ACPA specificity of the monoclonal antibodies used. The authors would like to correct the conclusion worded “We provide novel insights into the key role of citrullination and PAD enzymes during OC differentiation and ACPA-induced OC activation. Our findings suggest that IL8-dependent OC activation may constitute an early event in the initiation of the joint specific inflammation in ACPA-positive RA.” To be corrected as follows: “While ACPA may induce OC activation, the conclusions concerning the specificity of these observations require additional experiments before detailed mechanisms can be eluci- dated. Further, it is also not yet clear if ACPA are pathogenetically involved in the initiation of the joint specific inflammation in ACPA-positive RA or not.” As a note of clarification, the polyclonal antibody fractions did not contain LPS contaminations according to the limulous amoebocyte lysate (LAL) assay. We specifically apologise for the delays from our side in communicating the information in this correction note to the readership of Annals of the Rheumatic Diseases.

Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non- commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.​ ​org/​licenses/​by-​nc/​ 4.0/.​ © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ. Ann Rheum Dis 2019;78:866. doi:10.1136/annrheumdis-2015-208093corr1

866 Ann Rheum Dis 2019;78:866. doi:10.1136/annrheumdis-2015-208093corr1 Correspondence Unending story of the indirect be published on the electronic pages (references to be added by typesetter). immunofluorescence assay on HEp-2 cells: old Pier Luigi Meroni,1 Edward K Chan,2 Jan Damoiseaux,3 problems and new solutions? Luís Eduardo Coelho Andrade,4 Xavier Bossuyt,5 Karsten Conrad,6 Xavier Mariette,7 Joanna Sheldon,8 Johan Rönnelid,9 Marvin J Fritzler,10 The paper by Pisetsky et al raised some critical points on On behalf of the members of the committees the indirect immunofluorescence assay (IFA) on HEp-2 cells 1Immunorheumatology Research Laboratory, IRCCS Istituto Auxologico Italiano, (HEp-2 IFA): (1) the low antinuclear antibodies (ANA) pretest Milan, Italy and post-test probability, and (2) the variability of the IFA ANA 2Department of Oral Biology, University of Florida, Gainesville, Florida, USA 3 result depending on the method type and reagent source.1 Laboratory Specialist Medical Immunology, Maastricht University Medical Center, Maastricht, The Netherlands One consequence of these points is the heterogeneity in the 4 Departamento de Reumatologia, Fleury Laboratory, Immunology, Universidade classification criteria for systemic lupus erythematosus and/or Federal de São Paulo, Sao Paulo, Brazil the inclusion criteria in clinical trials or in specific treatment 5Department of Laboratory Medicine, University Hospital Leuven, Leuven, Belgium protocols. 6Institut für Immunologie Medizinische, Carl Gustav Carus" der Technischen Unfortunately, the paper by Pisetsky et al conveys no practical Universität Dresden, Dresden, Germany 7Department of Rheumatology, Université Paris-Sud, Paris, France suggestions on how to minimise these issues, and the impact on 8Center for Infection and Immunity Research, St George’s, University of London, the readers could be one of confusion rather than resolution. London, UK Several international committees are joining their efforts in 9Department of Immunology, Genetics and Pathology, Uppsala University, Uppsala, order to avoid misdiagnosis and to develop approaches to the Sweden 10Department of Medicine and Biochemistry and Molecular Biology, University of correct interpretation of the IFA ANA results depending on the Calgary, Calgary, Alberta, Canada technique used for detecting a given autoantibody.2 Screening tests for ANA represent the best example. After the initial recommen- Correspondence to Professor Pier Luigi Meroni, IRCCS Istituto Auxologico Italiano; pierluigi​ .​meroni@unimi​ .​it dation by the American College Rheumatology (ACR) Task Force 3 in 2010, several papers have addressed this issue in the last years. Handling editor Josef S Smolen While IFA ANA offered advantages in comparison with the solid phase assays (SPA) available at the time of the ACR position paper, Collaborators PLM on behalf of the members of the committees: IUIS/AF/CDC, the performance of the newer SPA has recently improved. Advan- ICAP, EASI, WG-IFCC, ECFSG, EULAR Task Force on laboratory diagnostic tests. tages and disadvantages of the two methodological platforms have Contributors All the authors contributed to the letter. been reviewed and discussed, and none of the two immunoassays Competing interests None declared. appears to satisfy completely the required demands.4 However, the Patient consent Not required. combination of HEp-2 IFA and SPA including the most relevant Provenance and peer review Not commissioned; internally peer reviewed. nuclear and cytoplasmic antigens for the diagnosis of systemic rheu- matic autoimmune diseases (SARD) has been reported to display © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2019. All rights reserved. No commercial use is permitted unless otherwise higher specificity and post-test probability than the use of the expressly granted. respective single tests.5 In addition, new tests that employ a panel of autoantigens relevant for a given subset of SARD (eg, lupus- like, systemic sclerosis, myopathies, antiphospholipid syndrome) are now available or are going to be launched soon, increasing To cite Meroni PL, Chan EK, Damoiseaux J, et al. Ann Rheum Dis 2019;78:e46. their specificity/post-test probability in a significant manner. For Received 19 March 2018 example, the combination of HEp-2 IFA and SPA for autoantibody Revised 3 April 2018 screening could decrease ‘false positive or false negative’ results, Accepted 6 April 2018 while the use of one screening assay and the new antigen-specific Published Online First 17 April 2019 multiplex immunoassays in the context of a specific clinical setting might increase the diagnostic power. In parallel, the International Consensus on ANA Patterns initia- tive has defined 30 HEp-2 IFA patterns (www.​ ​ANApatterns.org)​ ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213537 that provide clues for the autoantibodies most likely to be present Ann Rheum Dis 2019;78:e46. doi:10.1136/annrheumdis-2018-213440 in a given sample, thereby adding value to the test and directing 6 the investigation towards specific autoantibody assays. Since the References several HEp-2 IFA patterns have diverse immunological and clinical 1 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of implications, disease classification criteria and inclusion criteria for antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis clinical trials referent to HEp-2 IFA should define which patterns 2018;77:911–3. 2 Meroni PL, Biggioggero M, Pierangeli SS, et al. Standardization of autoantibody testing: are to be included. a paradigm for serology in rheumatic diseases. Nat Rev Rheumatol 2014;10:35–43. In this interim, the strategy for autoantibody testing has been 3 Meroni PL, Schur PH. ANA screening: an old test with new recommendations. Ann under reassessment, and it might be advantageous to consider the Rheum Dis 2010;69:1420–2. combination of the new serological tools for better understanding 4 Mahler M, Meroni PL, Bossuyt X, et al. Current concepts and future directions for the assessment of autoantibodies to cellular antigens referred to as anti-nuclear of the meaning of a given positive (or negative) result. The correct antibodies. J Immunol Res 2014;2014:1–18. use and interpretation of autoantibody testing is mandatory, and 5 Bossuyt X, Fieuws S. Detection of antinuclear antibodies: added value of solid phase a specific European League Against Rheumatism Task Force has assay? Ann Rheum Dis 2014;73:e10. been planned to address the issue in conjunction with the other 6 Chan EK, Damoiseaux J, de Melo Cruvinel W, et al. Report on the second international committees. International Consensus on ANA Pattern (ICAP) workshop in Dresden 2015. Lupus 2016;25:797–804. This paper was already the focus of some correspondence 7 Mahler M. Lack of standardisation of ANA and implications for drug development and 7 addressing the issue of ANA testing and others that will also precision medicine. Ann Rheum Dis 2019;78:e33.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Correspondence response Response to: 'Unending story of the indirect combination of an IFA and SPA; for example, in our study, only one patient showed ANA negativity in all three IFA assays. Stan- immunofluorescence assay on HEp-2 cells: old dardisation is also important to allow comparison of studies. problems and new solutions?' by Meroni et al We agree that confusion currently surrounds the issue of ANA testing in clinical trials and hope that our paper begins the We would like to thank Professor Meroni and colleagues for process to come to a resolution. In view of the pipeline of new their comments1 regarding our paper2 on testing for antinuclear agents that can be explored as novel therapeutics for SLE, such antibodies (ANAs) in patients with systemic lupus erythema- a resolution should be a top priority to advance the testing of tosus (SLE) and the very thoughtful discussion of the approaches new treatments and the addition of effective new agents to the for serological determinations. This letter provides a further armamentarium. perspective on ANA assays that were also discussed in a letter 1 1 2 3 3 David S Pisetsky, Diane M Spencer, Peter E Lipsky, Brad H Rovin to this journal by Dr M Mahler. With respect to approaches 1 to serological testing beyond immunofluorescence assays (IFA), Department of Medicine and Immunology, Duke University Medical Center and Medical Research Service, VA Medical Center, Durham, North Carolina, USA we fully agree that solid phase assays (SPA) can provide a useful 2RILITE Research Institute, Charlottesville, Virginia, USA adjunctive approach and, indeed, we provided data on an ANA 3Division of Nephrology, Ohio State University Wexner Medical Center, Columbus, ELISA in our paper. In our study, the SPA kit that we assessed did Ohio, USA not perform better than the IFA tests (11.7% ANA negative for Correspondence to Dr David S Pisetsky, Department of Medicine and Immunology, the SPA compared with 4.9%–22.3% negative for IFA assays). Duke University Medical Center, Durham, NC 27705, USA; ​david.​pisetsky@​duke.​edu While ANA assays have undergone extensive investigation, Handling editor Josef S Smolen one of the current challenges relates to the setting of clinical Funding The authors have not declared a specific grant for this research from any trials and the high screen failure rate of patients who have an funding agency in the public, commercial or not-for-profit sectors. established diagnosis of SLE but are ANA-negative at screening Competing interests None declared. despite a positive ANA in the past. Screening for trial eligibility Patient consent Not required. involves different considerations from those used for routine care and, indeed, may necessitate different assays. Importantly, Provenance and peer review Commissioned; internally peer reviewed. since products for the treatment of SLE can be approved for © Article author(s) (or their employer(s) unless otherwise stated in the text of the ‘active, autoantibody positive disease’, the assays used for sero- article) 2019. All rights reserved. No commercial use is permitted unless otherwise expressly granted. logical assessment are key and therefore should be rigorously evaluated for use as a ‘theranostic’ or companion diagnostic. Such an evaluation requires different methodology from what is currently used to evaluate a kit or assay to assess the presence of To cite Pisetsky DS, Spencer DM, Lipsky PE, et al. Ann Rheum Dis 2019;78:e47. an ANA in a broad population of patients with a systemic auto- immune rheumatic disease. In his letter, Dr Mahler3 discussed Received 11 April 2018 Revised 18 April 2018 the differences between companion and complementary diag- Accepted 19 April 2018 nostics. The use of the different ANA kits for screening for trial Published Online First 28 April 2018 eligibility has not yet been the subject of such an evaluation, adding uncertainty to the field. We discussed these issues in a previous publication.4

To move the field forward, we would suggest greater clarity ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213440 in the goal of ANA testing in the trial setting. Is such testing Ann Rheum Dis 2019;78:e47. doi:10.1136/annrheumdis-2018-213537 to confirm a diagnosis of SLE or is it to subset patients on the basis of disease activity, suspected mechanistic underpinning of References the disease and/or the likelihood of treatment response? In this 1 Meroni PL, Chan EK, Damoiseaux J, et al. Unending story of the indirect regard, while anti-DNA has had extensive use to assess disease immunofluorescence assay on HEp-2 cells: old problems and new solutions? Ann activity and is a component of the Systemic Lupus Erythema- Rheum Dis 2019;78:e46. tosus (SLEDAI), for example, testing for either ANA or anti- 2 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis bodies to RNA-binding proteins (ie, Sm, RNP, Ro and La) 2018;77:911–3. 5 has not been considered useful to assess disease activity. We, 3 Mahler M. Lack of standardisation of ANA and implications for drug development and therefore, believe that regulatory agencies in concert with inves- precision medicine. Ann Rheum Dis 2019;78:e33. tigators should directly address the issue of ANA testing to deter- 4 Pisetsky DS, Rovin BH, Lipsky PE. New perspectives in rheumatology: biomarkers as mine trial eligibility and provide guidance on the methodology entry criteria for clinical trials of new therapies for systemic lupus erythematosus: the example of antinuclear antibodies and anti-DNA. Arthritis Rheumatol 2017;69:487–93. that is the most informative and reliable in this specific setting. 5 McCarty GA, Rice JR, Bembe ML, et al. Independent expression of autoantibodies in This methodology could involve more than one IFA kit or the systemic lupus erythematosus. J Rheumatol 1982;9:691–5.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Correspondence Variation in antinuclear antibody detection by In a correspondence to the paper of Pisetsky et al.1 Dr Mahler discussed some relevant points related to ANA detection,2 automated indirect immunofluorescence including the substantial interobserver variability inherent to analysis IFA. Accordingly, Dr Mahler recommended to use automated interpretation systems to reduce variability and subjectivity.2 Pisetsky et al reported on assay variation in the detection of We evaluated variation in ANA detection by automated IFA antinuclear antibodies (ANA) in sera of patients with established systems. Three samples (sample 1: homogeneous 1:320; sample systemic lupus erythematosus (SLE).1 The authors determined 2: fine speckled 1:80; sample 3: negative) were distributed to ANA in sera from 103 patients with established SLE using three 31 Belgian laboratories that use automated IFA systems and 27 different validated and widely used immunofluorescence assays laboratories participated (NOVA View n=12 (Inova Diagnos- (IFA) (from ImmunoConcepts, Inova Diagnostics and Bio-Rad), tics, San Diego, USA); EUROPattern n=6 (Euroimmun, Lübeck, an ELISA and a bead-based multiplex assay (both from Bio-Rad). Germany); G-Sight n=7 (Menarini, Firenze, Italy); Image Navi- With IFA, the frequency of ANA negativity varied from 4.9% to gator n=2 (ImmunoConcepts, Sacramento, California, USA)). 22.3%. Part of the samples (11.7% and 13.6%) were negative Each sample was determined at least four times in different runs with ELISA and multiplex assay, respectively. This study showed (the majority was determined at least eight times) according to differences among IFA ANA kits which might have implications the instructions of the manufacturer. The fluorescence intensity for eligibility for clinical trials.1 units (reported as light intensity units (LIU) or probability index)

Figure 1 Fluorescence intensity (light intensity unit (LIU) or probability index (PI)) of antinuclear antibodies (ANA) by automated immunofluorescence assays (IFA). Three samples (sample 1: homogeneous 1:320; sample 2: fine speckled 1:80; sample 3: negative) were analysed by NOVA View (n=12: L1–L12), Image Navigator (n=2: L13–L14) or G-Sight (n=7: L15–L21). All laboratories used a screening dilution of 1:80 except for L17 that used a screening dilution of 1:40. All samples were diluted by a pipetting robot (QUANTA-Lyser, PhD, Sprinter, Zenit UP, Beeline, Helmed). For G-Sight, HEp-2000 or HEp-2 was used as substrate (as indicated). The cut-off for positivity was 49 LIU for NOVA View, 49 LIU for Image Navigator,<8 PI (negative) and >50 PI (positive) for G-Sight (L15–L17, L19–L21) or 20 PI for G-Sight (L18). All samples were run at least four times in different runs (the majority was determined at least eight times) according to the instructions of the manufacturer. The fluorescence intensity units (reported as LIU or PI) are shown. Statistical analysis of differences in fluorescence intensities between laboratories was evaluated by analysis of variance with Bonferroni correction for multiple comparisons. For NOVA View, there were significant differences between L1 versus L4 (p=0.044), L8 (p=0.001), L9 (p=0.002); L2 versus L8 (p=0.024) and L7 versus L8 (p=0.009), L9 (p=0.022) for sample 1, between L2 versus L4 (p=0.003), L8 (p=0.039), L9 (p=0.022); L7 versus L3 (p=0.034), L4 (p=0.026), L5 (p=0.033), L8 (p=0.003), L9 (p=0.018) and L11 versus L2 (p=0.022), L7 (p=0.007), L10 (p=0.005) for sample 2 and between L1 versus L4 (p=0.005), L8 (p=0.001), L11 (p=0.018) for sample 3. For G-Sight, there were differences between L15 versus L17 (p=0.001) and L16 versus L17 (p<0.001) for sample 1 and between L16 versus L17 (p=0.002) and L19 versus L21 (p=0.013) for sample 3. NOVA View estimates a pattern-specific end point titre based on the 1:80 screening dilution, a feature called single well titre. The single well titre varied between 1:160 and 1:640 for sample 1, between <1:80 and 1:320 for sample 2 and between <1:80 and 1:160 for sample 3.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 3 Correspondence

Figure 2 Intralaboratory performance of antinuclear antibodies (ANA) by automated immunofluorescence assays (IFA) for NOVA View. Light intensity units (LIU) values for an analysis of sample 1 in 2016 and in 2017 (the time difference between the two determination ranged between 7 and 19 months). The results shown are from at least five (2016) or seven (2017) determinations in different runs. For each laboratory (L), the first set of results shown were obtained in 2016 and the second set of results in 2017. are shown in figure 1, except for EUROPattern that does not assignment by automated systems should be verified by an expe- report fluorescence intensity. The results not only show variation rienced technician or immunologist. in the way fluorescence intensity is reported by different compa- Sample 1 had been used to evaluate interinstrument variability nies but also (1) variation in between-run imprecision between for NOVA View in a previous study3 performed 1 year before the automated IFA systems from the same manufacturer and (2) current study. This gave us the possibility to evaluate constancy variation in fluorescence intensity results between instruments of the results over a period of 7–19 months for 11 of the 12 from the same manufacturer. For instance, for NOVA View, the NOVA View users. The results are represented in figure 2 and coefficient of variation for sample 1 varied between 8.6% and show statistically significant differences between LIU values 39.7% and there were statistically significant differences (after obtained at two different time points for 6/11 laboratories. Of correction for multiple comparisons) in LIU values between note, for five laboratories, no statistical significant differences laboratories (figure 1 and its legend). Also for G-Sight, signifi- between the two determinations were observed, which illustrates cant differences between laboratories were found. For sample 2, the potential of automated IFA analysis to provide reproducible six laboratories using NOVA View and three laboratories using results over a longer time period. G-Sight found the sample to be negative in 10%–80% of the Taken together, although we could demonstrate reproducible between-run determinations for NOVA View and in 12.5%– ANA results for some laboratories, our results indicate variation 62.5% of the between run determinations for G-Sight. The other in ANA detection by automated IFA systems. We not only found laboratories found the sample to be positive for all determina- variation between automated IFA analysis using instruments tions, with differences in fluorescence intensities between labo- from different manufacturers but also between instruments ratories. For sample 3 (negative sample), three laboratories using from the same manufacturer. Efforts should be undertaken to NOVA View and four laboratories using G-Sight found positive harmonise automated IFA analysis. This could include the use of results in 10%–62.5% of the determinations for NOVA View standards, calibration of the instruments and monitoring of the and in 10%–100% of the determinations for G-Sight. Some of quality of the slides and reagents. these results were due to artefacts as expert visual review scored Lieve Van Hoovels,‍ ‍ 1 Sofie Schouwers,2 Stefanie Van den Bremt,1 60%–80% of the false positive determinations for NOVA View 3 negative. Xavier Bossuyt 1 For a selection of patterns, NOVA View allows to estimate the Department of Laboratory Medicine, OLV Hospital Aalst, Aalst, Belgium 2Department of Laboratory Medicine, GZA Hospital, Antwerp, Belgium end titre from the analysis of the 1:80 dilution (see legend of 3Department of Laboratory Medicine, University Hospital Leuven, Leuven, Belgium figure 1). EUROPattern (performed by six laboratories) does not provide a measure of fluorescence intensities. It scored sample 1 Correspondence to Lieve Van Hoovels, Department of Laboratory Medicine, OLV Hospital Aalst, 9300 Aalst, Belgium; Lieve.​ ​Van.Hoovels@​ ​olvz-​aalst.​be positive for all determinations (n=61) (with the estimated end titre varying between 1:80 and 1:1280) and sample 2 for 59% of Acknowledgements We thank Sofie Arens (Klinisch laboratorium Declerck Ardooie), Jan Beckers (AZ Sint-Blasius Dendermonde), Mario Berth (AML Antwerpen), the determinations (n=51) (with the estimated end titre varying Juul Boes (AZ Turnhout), XB (UZ Leuven), Francis Corazza and Carole Nagant (CHU between <1:80 and 1:160). Sample 3 was negative in 92% of Brugmann Brussel), Bjorn Ghesquière and Ignace Van Hecke (CRI Zwijnaarde), the determinations (n=60). Sylvie Goletti (UCL Louvain), An Joosten and Patricia Meirlaen (UZ Brussel), Pattern assignment by automated IFA was correct in 76%, Julia Kovaleva (CMA Hasselt), Laurence Lutteri (CHU Liège), Laurence Miller and Patricia Evrard (CHU UCL Namur), An Nijs (AZ Groeninge Kortrijk), Mounzer Reda 95% and 100% of the determinations by NOVA View, EUROPat- (Hôpital de Jolimont Haine-Saint-Paul), SS (GZA Antwerpen), Rita Schroder (Vivalia tern and G-Sight for sample 1. For sample 2, the values were, Arlon), Chistine Tilmant (C.H.I.R.E.C. Braine l’Alleud), Tanja Troch (ASZ Aalst), Wim respectively 32%, 100% and 100%. This indicates that pattern Uyttenbroeck (ZNA Jan Palfijn Antwerpen), LVH (OLVZ Aalst), Hilde Vanhouteghem

2 of 3 Ann Rheum Dis June 2019 Vol 78 No 6 Correspondence

(AZ Maria Middelares Gent), Annemie Van Ruymbeke (AZ Alma Sijsele), Lut Vanneste To cite Van Hoovels L, Schouwers S, Van den Bremt S, et al. Ann Rheum Dis (Labo Bruyland Kortrijk), Martine Vercammen (AZ Sint-Jan Brugge), Ann Verdonck 2019;78:e48. (Anacura Labo Nuytinck Evergem) and Sara Vijgen (Jessa ziekenhuis Hasselt) for participating in the study. We thank Heide De Baere and Nathalie Ruisseau (Menarini Received 8 April 2018 Benelux), Tom Gheskiere (Biognost cvba), Chris Raskin (Biomedical Diagnostics Accepted 9 April 2018 N.V./S.A Benelux) and Nathalie Vandeputte (Werfen N.V./S.A Benelux) for their Published Online First 20 April 2018 practical support. Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests XB has been a consultant for Inova Diagnostics. ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213558 Patient consent Not required. Ann Rheum Dis 2019;78:e48. doi:10.1136/annrheumdis-2018-213543 Ethics approval Local ethical committee OLVZ: amendment for study 2015/100 (B126201525864) received on 07/03/2017. References Provenance and peer review Not commissioned; internally peer reviewed. 1 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of © Article author(s) (or their employer(s) unless otherwise stated in the text of the antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis article) 2019. All rights reserved. No commercial use is permitted unless otherwise 2018. doi: annrheumdis-2017-212599 (Epub ahead of print). expressly granted. 2 Mahler M. Lack of standardisation of ANA and implications for drug development and precision medicine. Ann Rheum Dis 2019;78:e33. 3 Van den Bremt S, Schouwers S, Van Blerk M, et al. ANA IIF automation: moving towards harmonization? Results of a multicenter study. J Immunol Res 2017;2017:1–7.

Ann Rheum Dis June 2019 Vol 78 No 6 3 of 3 Correspondence response

Correspondence to Dr David S Pisetsky, Department of Medicine and Immunology, Response to: ‘Variation in antinuclear antibody Duke University Medical Center, Durham, NC 27705, USA; ​david.​pisetsky@​duke.​edu detection by automated indirect Handling editor Josef S Smolen immunofluorescence analysis’ by van Hoovels Competing interests None declared. et al Patient consent Not required. Provenance and peer review Commissioned; internally peer We thank van Hoovels et al1 for their comments on our article2 reviewed. on the variability of antinuclear antibody (ANA) determinations; © Article author(s) (or their employer(s) unless otherwise stated in the text of the other letters also discussed ANA testing issues.3 4 Meroni et al article) 2019. All rights reserved. No commercial use is permitted unless otherwise noted the value of distinguishing staining patterns by immuno- expressly granted. fluorescence assays (IFAs).3 In his discussion, Dr Mahler recom- mended the use of automated or computer-assisted diagnostic (CAD) systems as a way to reduce the variability and subjectivity To cite Pisetsky DS, Spencer DM, Lipsky PE, et al. Ann Rheum Dis 2019;78:e49. that can occur with visual reading of IFA for ANA determina- tions.4 As the data presented by Van Hoovels and colleagues in Received 26 April 2018 their letter indicate, even automated systems have limitations. A Accepted 26 April 2018 Published Online First 5 May 2018 prior publication on CAD systems highlighted quality assurance approaches to address these problems.5 Clearly, even with well-validated assays that include CAD interpretation, available ANA assays show variability. This vari- ability can affect diagnostic evaluation and decision-making; in ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213543 the context of clinical trials for ‘active, autoantibody positive’ Ann Rheum Dis 2019;78:e49. doi:10.1136/annrheumdis-2018-213558 disease, these assay issues can determine trial entry and lead to screen failures. Furthermore, assay variability could affect the References prescription of agents that have been approved for patients who 1 Van Hoovels L, Schouwers S, Van den Bremt S, et al. Variation in antinuclear antibody are ANA or anti-DNA positive. We, therefore, welcome further detection by automated indirect immunofluorescence analysis. Ann Rheum Dis discussion on ANA testing and the opportunity to develop more 2019;78:e48. 2 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of robust and reproducible assays as well as to achieve greater antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis harmonisation with existing platforms. 2018;77:911–3. 3 Meroni PL, Chan EK, Damoiseaux J, et al. Unending story of the indirect 1 1 2 3 David S Pisetsky, Diane M Spencer, Peter E Lipsky, Brad H Rovin immunofluorescence assay on HEp-2 cells: old problems and new solutions? Ann 1Department of Medicine and Immunology, Duke University Medical Center and Rheum Dis 2019;78:e46. Medical Research Service, Veterans Administration Medical Center, Durham, North 4 Mahler M. Lack of standardisation of ANA and implications for drug development and Carolina, USA precision medicine. Ann Rheum Dis 2019;78:e33. 2RILITE Research Institute, Charlottesville, Virginia, USA 5 Van den Bremt S, Schouwers S, Van Blerk M, et al. ANA IIF automation: 3Division of Nephrology, The Ohio State University, Wexner Medical Center, moving towards harmonization? Results of a multicenter study. J Immunol Res Columbus, Ohio, USA 2017;2017:1–7.

Ann Rheum Dis July 2019 Vol 78 No 6 1 of 1 Correspondence

Correspondence to Professor Pier Luigi Meroni, Immunorheumatology Research Pitfalls of antinuclear antibody detection in Laboratory, IRCCS Istituto Auxologico Italiano, Milan 20095, Italy; systemic lupus erythematosus: the positive pierluigi​ .​meroni@unimi​ .​it experience of a national multicentre study Collaborators Claudia Alpini, Alessia Alunno, Maria Romana Bacarelli, Andrea Doria, Sergio Finazzi, Franco Franceschini, Roberto Gerli, Anna Ghirardello, Luigi 1 Giovannelli, Maria Grazia Giudizi, Milvia Lotzniker, Paola Migliorini, Gabriella The recent paper by Pisetsky et al, which already elicited some Morozzi, Federico Pratesi, Antonella Radice, Valeria Riccieri, Amelia Ruffatti, Paola 2 3 debate, reported data on the use of antinuclear antibodies (ANA) Sabatini, Giandomenico Sebastiani, Renato Alberto Sinico, Angela Tincani, Marta detection in patients with systemic lupus erythematosus (SLE) in the Tonello. real life and raised concerns on the usefulness of the assays because Contributors FP contributed to data acquisition, analysis and interpretation. of the significant percentage of samples tested negative for ANA in FP, MOB and PLM gave substantial contributions to the conception of the work spite of the use of well-validated assays and the inclusion of patients and drafting of the letter, had full access to all of the data in the study and take with an established diagnosis. This finding may have negative impli- responsibility for the integrity of the data and the accuracy of the data analysis. The cations both for the correct classification and the inclusion in clinical FIRMA Study Group provided patients, collected data and critically reviewed the study proposal. All the contributors read and approved the manuscript. trials. We report here a multicentre study carried out by an Italian Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. interdisciplinary group (Forum Interdisciplinare per la Ricerca sulle Malattie Autoimmuni) for the validation of new automated Competing interests None declared. reading systems for ANA detection by indirect immunofluorescence Patient consent Obtained. on HEp-2 cells (HEp-2 IFA). Ninety-one patients with well-estab- Ethics approval Ethics Committee of Istituto Auxologico Italiano, 22-07-2010. 4 lished SLE (18/91 men, mean age 40±11 years) were tested for Provenance and peer review Not commissioned; internally peer reviewed. ANA, anti-extractable nuclear antigens (ENA) and anti-double © Article author(s) (or their employer(s) unless otherwise stated in the text of the stranded (ds)-DNA. Manual HEp-2 IFA (with different HEp-2 article) 2019. All rights reserved. No commercial use is permitted unless otherwise commercial preparations including HEp-2000) was performed at expressly granted. the recruiting centre and then sent to two core labs where three different automated ANA reading systems were used with the manufacturers’ cell substrate (1:80 screening dilution). Moreover, all the samples were tested by two commercial connective tissue To cite Pregnolato F, Borghi MO, Meroni PL, et al. Ann Rheum Dis 2019;78:e50. disease (CTD) screening solid phase arrays (SPA). We found almost a Received 3 April 2018 perfect agreement between manual and automated ANA reading. As Revised 11 April 2018 5 expected, SPA displayed a lower sensitivity (table 1). By testing all Accepted 12 April 2018 the samples with both HEp-2 IFA and SPA we reached 100% sensi- Published Online First 25 April 2018 tivity. One sample only tested positive for DFS70 but the positivity was associated with anti-ENA antibodies, a combination that can be found in systemic autoimmune rheumatic diseases.6 The detec- tion of antibodies against ENA and dsDNA was performed at the ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213582 recruiting centres by using different commercial kits. The percent- Ann Rheum Dis 2019;78:e50. doi:10.1136/annrheumdis-2018-213516 ages of patients positive for anti-ENA and anti-dsDNA were similar to those of the Pisetsky’s series making the two cohorts comparable. References The results of our study show that: (1) ANA is a hallmark of 1 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of established SLE; (2) ANA detection by different commercially avail- antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis able kits is reliable in a multicentre setting and the variability linked 2018;77:911–3. to the single operator does not seem to be a critical issue. The CTD 2 Mahler M. Lack of standardisation of ANA and implications for drug development and screening SPAs display a lower sensitivity than HEp-2 IFA likely due precision medicine. Ann Rheum Dis 2019;78:e33. 3 Pisetsky DS, Spencer DM, Lipsky PE, et al. Response to: ’Lack of standardization of ANA to the limited number of autoantigens, however they can offer an and implications for drug development and precision medicine’ by Mahler. Ann Rheum 5 additional diagnostic value when carried out together with IFA. Dis 2019;78:e34. 4 Petri M, Orbai AM, Alarcón GS, et al. Derivation and validation of the Systemic 1 1,2 1 Francesca Pregnolato, Maria Orietta Borghi, Pier Luigi Meroni, Forum Lupus International Collaborating Clinics classification criteria for systemic lupus Interdisciplinare per la Ricerca sulle Malattie Autoimmuni (FIRMA) Study erythematosus. Arthritis Rheum 2012;64:2677–86. Group 5 Claessens J, Belmondo T, De Langhe E, et al. Solid phase assays versus automated 1Immunorheumatology Research Laboratory, IRCCS Istituto Auxologico Italiano, indirect immunofluorescence for detection of antinuclear antibodies. Autoimmun Rev Milan, Italy 2018. doi: 10.1016/j.autrev.2018.03.002. [Epub ahead of print 8 Mar 2018]. 2Department of Clinical Sciences and Community Health, University of Milan, Milan, 6 Conrad K, Röber N, Andrade LE, et al. The clinical relevance of Anti-DFS70 Italy autoantibodies. Clin Rev Allergy Immunol 2017;52:202–16.

Table 1 ANA detection HEp-2 IFA CTD SPA* Test results Manual Auto 1† Auto 2 Auto 3 a b Anti-ENA Anti-dsDNA Positive % (n/N) 100 (91/91) 100 (91/91) 98 (89/91) 99 (90/91) 88 (80/91) 91 (83/91) 45 (40/89) 59 (54/91) Negative % (n/N) 0 (0/91) 0 (0/0) 2 (2/91) 1 (1/91) 12 (11/91) 9 (8/91) 55 (49/89) 41 (37/91) *CTD SPA, connective tissue diseases screening solid phase assay; (a) Quanta Flash CTD screen plus, INOVA Diagnostics: recombinant Scl-70, Jo-1, SSA/Ro 52, SSA/Ro 60, SSB/La, centromere A and B, RNA Pol III, Mi-2, Ku, Th/To, PCNA, native Sm and RNP, synthetic Pm/Scl and ribosomal-P peptides and synthetic dsDNA; (b) ELiA CTD screen, Thermo Fisher: dsDNA, SSA/Ro 52, SSA/Ro 60, SSB/La, U1-RNP (RNP-70, A, C), Sm, centromere B, Jo-1, Scl-70, Rib-P, fibrillarin, RNA Pol III, PM-Scl, PCNA and Mi-2, all recombinant except native purified dsDNA. †Automated reading systems: (1) AKLIDES, Medipan GMBH; (2) NOVA-View, INOVA Diagnostics; (3) G-Sight, Menarini Diagnostics. ANA, anti-nuclear antibodies; ENA, extractable nuclear antigen; HEp-2 IFA, indirect immunofluorescence on HEp-2 cells.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Correspondence response Response to: ‘Pitfalls of antinuclear antibody Handling editor Josef S Smolen Funding The authors have not declared a specific grant for this research from any detection in systemic lupus erythematosus: the funding agency in the public, commercial or not-for-profit sectors. positive experience of a national multi-center Competing interests None declared. study’ by Pregnalato et al Patient consent Not required. Provenance and peer review Commissioned; internally peer reviewed. We appreciate the comments by Pregnolato et al1 on our 2 © Article author(s) (or their employer(s) unless otherwise stated in the text of the original paper and subsequent correspondence in the article) 2019. All rights reserved. No commercial use is permitted unless otherwise journal.3 4 The data presented are very interesting and suggest expressly granted. that, under certain circumstances, results of antinuclear anti- body (ANA) testing with samples from patients with estab- lished systemic lupus erythematosus can show both a high frequency of positivity and high concordance among testing To cite Pisetsky DS, Spencer DM, Lipsky PE, et al. Ann Rheum Dis 2019;78:e51. laboratories. Such agreement may result from the nature of Received 26 April 2018 the samples themselves, including the titre of the antibodies Accepted 26 April 2018 and the array of specificities present, the kits used for the Published Online First 5 May 2018 assays and the experience and training of the observers. While we would acknowledge that ANA assays can perform very well, our experience as well as that in the literature indicates that variability occurs significantly in a way that complicates ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213516 screening of patients for clinical trials. Notwithstanding the Ann Rheum Dis 2019;78:e51. doi:10.1136/annrheumdis-2018-213582 demonstrations of the kind reported by Pregnolato et al, we think that it is important to try to understand assay variability and improve serological determinations for both routine care References as well as clinical research. 1 Pregnolato F, Borghi MO, Meroni PL. Forum Interdisciplinare per la Ricerca sulle Malattie Autoimmuni (FIRMA) Study Group. Pitfalls of antinuclear antibody detection in systemic lupus erythematosus: the positive experience of a national multicentre study. David S Pisetsky,1 Diane M Spencer,1 Peter E Lipsky,2 Brad H Rovin3 Ann Rheum Dis 2019;78:e50. 1 Department of Medicine and Immunology, Duke University Medical Center and 2 Pisetsky DS, Spencer DM, Lipsky PE, et al. Assay variation in the detection of Medical Research Service, Veterans Administration Medical Center, Durham, North antinuclear antibodies in the sera of patients with established SLE. Ann Rheum Dis Carolina, USA 2018;77:911–3. 2RILITE Research Institute, Charlottesville, Virginia, USA 3 3 Mahler M. Lack of standardisation of ANA and implications for drug development and Division of Nephrology, Department of Internal Medicine, Ohio State University, precision medicine. Ann Rheum Dis 2019;78:e33. Wexner Medical Center, Columbus, Ohio, USA 4 Pisetsky DS, Spencer DM, Lipsky PE, et al. Response to: ’Lack of standardization of ANA Correspondence to Dr David S Pisetsky, Department of Medicine and Immunology, and implications for drug development and precision medicine’ by Mahler. Ann Rheum Dis Duke University Medical Center, Durham, NC 27705, USA; ​david.​pisetsky@​duke.​edu 2019;78:e34.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Correspondence Time to personalize the treatment of anti- often used for ARDS like prone ventilation, neuromuscular blockade and extra-corporeal membrane oxygenation. Our hope MDA-5 associated lung disease is that if we move toward a more personalised approach to the treatment of anti-MDA5-associated lung disease we can improve We read with interest the article by the European League Against outcomes for what appears to be one of the most, if not the Rheumatism/American College of Rheumatology on the classi- most, aggressive forms of rheumatic lung disease. fication criteria for adult and juvenile idiopathic inflammatory myopathies (IIM) and their major subgroups.1 While we believe Michael Lake, Gautam George, Ross Summer these criteria will help with the diagnosis of most of the major Division of Pulmonary and Critical Care, Thomas Jefferson University Hospitals, subgroups of idiopathic inflammatory myopathies, we anxiously Philadelphia, Pennsylvania, USA await society guidelines that direct physicians on how best to Correspondence to Dr Michael Lake, Division of Pulmonary and Critical Care individualise the treatment for specific subgroups of this disease. Medicine, Thomas Jefferson University Hospital, Philadelphia PA 19107, USA; For example, one subgroup we seek immediate direction is michael​ .​lake@jefferson.​ ​edu on the treatment of patients with antibodies directed against the Handling editor Josef S Smolen melanoma differentiation associated protein 5 (anti-MDA5). This antibody is typically identified in patients with clinically Contributors All three authors of this article contributed to the writing of the amyopathic dermatomyositis (CADM), and it is now increas- manuscript. ingly appreciated that some of these individuals develop a highly aggressive form of lung disease that causes severe tissue destruc- Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. tion and often fatal respiratory failure.2 Consistent with this clinical course, we recently diagnosed a patient with anti-MDA5 Competing interests None declared. associated CADM in our hospital who developed rapidly Patient consent Not required. progressive respiratory failure within a span of only a few days Provenance and peer review Not commissioned; externally peer reviewed. and died within just 1 week of his hospitalisation; this was despite © Article author(s) (or their employer(s) unless otherwise stated in the text of the our early initiation of high dose corticosteroids. Further, limited article) 2019. All rights reserved. No commercial use is permitted unless otherwise autopsy of the chest revealed pathological findings consistent expressly granted. with the acute respiratory distress syndrome (ARDS), including diffuse alveolar damage and prominent hyaline membrane depo- sition. Notably, our review of the literature indicates that our To cite Lake M, George G, Summer R. Ann Rheum Dis 2019;78:e52. clinical experience is not unique as nearly 50% of individuals with anti-MDA-associated lung involvement appear to have a Received 21 February 2018 similar clinical course, emphasising the devastating nature of this Revised 23 February 2018 condition.3–5 Accepted 2 April 2018 Published Online First 11 April 2018 The general approach to the treatment of lung disease in patients with idiopathic inflammatory myopathies, including CADM, is to initiate immunosuppressant drugs usually with a regimen containing high dose corticosteroids.6 However, we wonder if it is now time to reconsider this treatment approach ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213519 for anti-MDA-5 associated lung disease given our emerging Ann Rheum Dis 2019;78:e52. doi:10.1136/annrheumdis-2018-213285 understanding of the mechanisms that may underlie its devel- opment. For example, it is known that MDA5 is a viral RNA References binding protein that is important for activating anti-viral immune 1 Lundberg IE, Tjärnlund A, Bottai M, et al. International Myositis Classification Criteria Project consortium, The Euromyositis register and The Juvenile Dermatomyositis Cohort responses. As such, it is believed that antibodies directed against Biomarker Study and Repository (JDRG) (UK and Ireland). 2017 European League MDA-5 act to either activate or inhibit its activity, thereby trig- Against Rheumatism/American College of Rheumatology classification criteria for gering enhanced or blunted immune responses to viral infections, adult and juvenile idiopathic inflammatory myopathies and their major subgroups. Ann respectively. It seems reasonable to assume that the prevention of Rheum Dis 2017;76:1955–64. lung disease might depend on removing anti-MDA5 antibodies 2 Chino H, Sekine A, Baba T, et al. Radiological and pathological correlation in anti-mda5 antibody-positive interstitial lung disease: rapidly progressive perilobular opacities and from the circulation much like the approach used for other anti- diffuse alveolar damage. Intern Med 2016;55:2241–6. body-mediated lung conditions such as Goodpasture’s syndrome 3 Moghadam-Kia S, Oddis CV, Sato S, et al. Anti-Melanoma differentiation-associated or Granulomatosis with Polyangiitis, which often rely on the use gene 5 is associated with rapidly progressive lung disease and poor survival in of plasmapheresis and anti-B lymphocyte therapies. US Patients with amyopathic and myopathic dermatomyositis. Arthritis Care Res In addition to above strategies, we in the pulmonary commu- 2016;68:689–94. 4 Gerami P, Schope JM, McDonald L, et al. A systematic review of adult-onset clinically nity also wonder whether we should begin reconsidering our amyopathic dermatomyositis (dermatomyositis siné myositis): a missing link within approach to the management of respiratory failure once it the spectrum of the idiopathic inflammatory myopathies. J Am Acad Dermatol develops in patients with anti-MDA-5 antibodies. For example, 2006;54:597–613. we postulate whether treatments utilised for other aggressive 5 Labrador-Horrillo M, Martinez MA, Selva-O’Callaghan A, et al. Anti-MDA5 antibodies forms of acute respiratory failure (eg, ARDS) should be consid- in a large Mediterranean population of adults with dermatomyositis. J Immunol Res 2014;2014:1–8. ered so that patients can theoretically be supported for extended 6 Kawasumi H, Gono T, Kawaguchi Y, et al. Recent treatment of interstitial lung disease periods. This would include strategies such as fluid restriction, with idiopathic inflammatory Myopathies. Clin Med Insights Circ Respir Pulm Med lung protective ventilation and the various salvage therapies 2015;9:CCRPM.S23313.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Correspondence response Response to: ‘Time to personalise the treatment 14 patients with a positive anti-MDA5 autoantibody, 13 (93%) are classifiable using the new classification criteria as either of anti-MDA-5 associated lung disease’ by Lake DM (n=9) or ADM (n=4). Only one patient is not classifi- et al able as he was diagnosed while critically ill with RPILD in the intensive care unit and died shortly after diagnosis, not unlike We have with interest read the comments raised by Lake et the case described by Lake et al. In that regard, we believe that al in the e-letter titled ‘Time to personalise the treatment of most IIM classification criteria are performing poorly in an anti-MDA-5 associated lung disease’ to be published in your acute setting such as described above, given the challenge of journal.1 We thank the authors for sharing their insights on the assessing for muscle involvement by manual testing or muscle 2017 European League Against Rheumatism (EULAR)/Amer- biopsy. We believe that in the example presented by Lake et al, ican College of Rheumatology (ACR) classification criteria for the question is not if the 2017 EULAR/ACR IIM classification adult and juvenile idiopathic inflammatory myopathies (IIM) criteria are capturing or not anti-MDA5 cases, as they do. The and their major subgroups.2 Lake and colleagues are concerned question is if the subset assigned is an adequate reflection of that these classification criteria are not including rare specific the clinical phenotype associated with the condition. We must subsets such as dermatomyositis (DM) or amyopathic DM asso- agree with the authors of the letter that in the current version ciated with antimelanoma differentiation-associated protein of the EULAR/ACR IIM classification criteria, it is not possible 5 (MDA5). They fear that this will in turn prevent clinicians to identify subgroups of rare phenotypes subclassified by MSA and researchers from exploring new therapeutic approaches status. Our hope is that with time, international collaboration for this population frequently affected by rapidly progressive and expanding data will permit inclusion of other MSA in a interstitial lung disease (RPILD), a complication associated data-driven validated revision of the EULAR/ACR IIM classi- with fatal outcomes.3 4 Dr Malaviya raised a similar concern fication criteria. when discussing the importance of myositis-specific autoanti- 1,2 1 1 bodies (MSA) in classifying subgroups of IIM, exemplified by Valérie Leclair, Ingrid E Lundberg, Anna Tjärnlund 1 four subgroups of amyopathic DM associated with different Rheumatology Unit, Department of Medicine, Karolinska University Hospital, Karolinska Institutet, Stockholm, Sweden MSA and with different clinical phenotypes and response to 2 Rheumatology Unit, Department of Medicine, Jewish General Hospital, Montreal, treatment, one being anti-MDA5 antibody positive amyopathic Quebec, Canada DM.5 We responded to Dr Malaviya’s concern and addressed Correspondence to Anna Tjärnlund, Rheumatology Unit, Department of Medicine, the ongoing international collaborative effort of systematic Karolinska University Hospital, Karolinska Institutet, Stockholm, Sweden; collection of autoantibody and clinical data for patients with anna.​ ​tjarnlund@ki​ .​se 6 IIM. In this letter, we would like to address the concern raised Handling editor Josef S Smolen by Lake and colleagues. Classification criteria and their influ- Funding The authors have not declared a specific grant for this research from any ence on population selection for clinical trials surely have an funding agency in the public, commercial or not-for-profit sectors. impact on formulating treatment guidelines in conditions such Competing interests None declared. as idiopathic inflammatory myositis (IIM). Classification of a heterogeneous group of diseases with multiple possible organ Patient consent Not required. involvement, considerable overlapping features and different Provenance and peer review Commissioned; internally peer reviewed. underlying pathological molecular mechanisms is a challenge. © Article author(s) (or their employer(s) unless otherwise stated in the text of the The 2017 EULAR/ACR IIM classification criteria were devel- article) 2019. All rights reserved. No commercial use is permitted unless otherwise oped using the expertise of rheumatologists, dermatologists, expressly granted. neurologists and paediatricians in an attempt to reflect as closely as possible the clinical spectrum of IIM. As those criteria are data-driven, certain clinical characteristics such as most of To cite Leclair V, Lundberg IE, Tjärnlund A. Ann Rheum Dis 2019;78:e53. the MSA could not be included in the final variables given their rarity and the fact that some of these MSA were not widely Received 16 April 2018 available for clinical use at time of data collection. However, Accepted 17 April 2018 Published Online First 3 May 2018 the inclusion of anti-histidyl-tRNA synthetase (Jo1) in the current version of the criteria is a step forward in incorporating MSA as important discriminating factors in IIM classification. By no means are we suggesting to clinicians or researchers that MSA status do not have a diagnostic and prognostic utility in ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213285 practice. It seems appropriate to emphasise that classification Ann Rheum Dis 2019;78:e53. doi:10.1136/annrheumdis-2018-213519 criteria should not be used as diagnostic criteria and that clini- cians should consider all available evidence pointing to an IIM References diagnosis to tailor their management.7 1 Lake M, George G, Summer R. Time to personalize the treatment of anti-MDA-5 As emphasised by Lake et al, ILD is a serious and possibly associated lung disease. Ann Rheum Dis 2019;78:e52. fatal extramuscular manifestation of IIM. We agree that revi- 2 Lundberg IE, Tjärnlund A, Bottai M, et al. 2017 European League Against Rheumatism/ American College of Rheumatology classification criteria for adult and juvenile sion of the EULAR/ACR criteria should address this concern, idiopathic inflammatory myopathies and their major subgroups. Ann Rheum Dis as ILD yielded a strong association with having IIM in the 2017;76:1955–64. dataset used to develop the criteria.8 In a previous response 3 Gono T, Kawaguchi Y, Satoh T, et al. Clinical manifestation and prognostic factor in letter, we have discussed the process leading up to the final anti-melanoma differentiation-associated gene 5 antibody-associated interstitial lung disease as a complication of dermatomyositis. Rheumatology 2010;49:1713–9. criteria and the selection process involving ILD and other 9 4 Hall JC, Casciola-Rosen L, Samedy LA, et al. Anti-melanoma differentiation-associated extramuscular manifestations. From our local experience protein 5-associated dermatomyositis: expanding the clinical spectrum. Arthritis Care with the Karolinska University Hospital IIM cohort, out of our Res 2013;65:1307–15.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 2 Correspondence response

5 Malaviya AN. 2017 EULAR/ACR classification criteria for adult and juvenile idiopathic 8 Bottai M, Tjärnlund A, Santoni G, et al. EULAR/ACR classification criteria for adult and inflammatory myopathies and their major subgroups: little emphasis on autoantibodies, juvenile idiopathic inflammatory myopathies and their major subgroups: a methodology why? Ann Rheum Dis 2019;77:e77. report. RMD Open 2017;3:e000507. 6 Lundberg IE, Tjärnlund A. Response to: ’2017 EULAR/ACR classification criteria for 9 Tjärnlund A, Bottai M, Lundberg IE. Response to Comments on the "2017 EULAR/ACR adult and juvenile idiopathic inflammatory myopathies and their major subgroups: little Classification Criteria for Adult and Juvenile Idiopathic Inflammatory Myopathies and emphasis on autoantibodies, why?’ by Malaviya. Ann Rheum Dis 2018;77:e78. Their Major Subgroups". Points of concern. Arthritis Rheumatol 2018. [Epub ahead of 7 Aggarwal R, Ringold S, Khanna D, et al. Distinctions between diagnostic and print 8 Mar 2018]. classification criteria? Arthritis Care Res 2015;67:891–7.

2 of 2 Ann Rheum Dis June 2019 Vol 78 No 6 Correspondence Missing pebble in the mosaic of rheumatic and their interactions vary in different age ranges. This would allow understanding how the balance of powers of disease activity, diseases and mental health: younger does not comorbidities and age specifically influences depressive symptoms/ always mean happier health status in patients with SpA. In conclusion, we acknowledge that the study by Redeker et al provides major insights into this We have read with great interest the article by Redeker et al1 field and points out the need of a prompt intervention to identify investigating the determinants of psychological well-being in and address psychological needs of patients with SpA. However, axial spondyloarthritis (SpA). The authors observed that among looking at the overall mosaic of psychological well-being in SpA, 1736 patients aged 18–79 years, 68% of subjects displayed we might miss a specific pebble relating to the unique experience depressive symptoms, ranging from mild to severe according of young people dealing with the disease in a peculiar phase of to the five-item WHO Well-Being Index. This article clearly their life. We believe that a deeper understanding of the specific points out the burden of depressive symptoms in patients burden of SpA and, in general terms, of RMDs on mental health with SpA and its correlation with disease activity, functional in patients of different age would allow to tailor approaches to be impairment and other parameters including younger age. The eventually implemented in clinical practice. authors emphasise that a potential explanation for the negative ‍ ‍ 1 ‍ ‍ 2 3,4 5,6 association between depressive symptoms and age might be a Alessia Alunno, Paul Studenic, Dieter Wiek, Petra Balážová, Daniel Aletaha2 better capability to cope with the disease of adult compared 1 with young patients. This aspect prompted our reflections, as Rheumatology Unit, Department of Medicine, University of Perugia, Perugia, Italy 2Division of Rheumatology, Department of Internal Medicine 3, Medical University the psychological impact of rheumatic and musculoskeletal Vienna, Vienna, Austria diseases (RMDs) in young people is a major unmet need. 3EULAR PARE, Zurich, Switzerland We previously reported that over 90% of young patients with 4Deutsche Rheuma-Liga, Bonn, Germany 5 RMDs outline an impact of their condition on mental health which, EULAR Young PARE, Zurich, Switzerland 6Slovak League Against Rheumatism, Piestany, Slovakia based on the definition of WHO,2 encompasses depressive symp- toms and anxiety and the fear to not be accepted as full members Correspondence to Dr Alessia Alunno, Rheumatology Unit, Department of of the society due to the disease.3 The spectrum of psychological Medicine, University of Perugia, Perugia I-06129, Italy; ​alessia.​alunno82@​gmail.​com discomfort in young people with RMDs is also related to their Handling editor Francis Berenbaum career stage (eg, if looking for a job or being employed for just a few years), the kind of job qualification (eg, employment in a Contributors All authors provided a substantial contribution to the article. physically demanding job may amplify negative feelings), as well Funding The authors have not declared a specific grant for this research from any as private life matters. With regard to the latter, being in a stable funding agency in the public, commercial or not-for-profit sectors. relationship may help coping with the disease, but some partners, Competing interests None declared. particularly in young age, may not be supportive or willing to adapt to the partner’s needs. Patient consent Not required. Previous studies including adult patients revealed that a consis- Provenance and peer review Not commissioned; internally peer reviewed. tently lower percentage of them, only 30%–60%, report an © Article author(s) (or their employer(s) unless otherwise stated in the text of the impact of their disease on mental health.4 5 These latter percent- article) 2019. All rights reserved. No commercial use is permitted unless otherwise ages fit with the results obtained by Redeker et al in their cohort, expressly granted. which includes the complete age range from 18 years to 79 years. Specific conclusions regarding the actual burden of depression for specific age groups of patients with SpA may not become To cite Alunno A, Studenic P, Wiek D, et al. Ann Rheum Dis 2019;78:e54. apparent from the multivariable logistic model. In this regard, it is interesting to note that the study from Meesters et al6 depicts Received 17 April 2018 Accepted 19 April 2018 higher depression rates in patients with ankylosing spondylitis Published Online First 28 April 2018 aged below 50 years. Considering this, can we be sure that the rates are not substantially different in patients below the age of 35 years, or even more so in young patients who had just made the often challenging transition from adolescence to adulthood? How ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213620 impactful on mental health is being confronted with a chronic disease in early adulthood, not knowing how this may affect the Ann Rheum Dis 2019;78:e54. doi:10.1136/annrheumdis-2018-213611 own future? Another interesting point for discussion is the impact of comorbidities and disease activity on mental health. Among References 1 Redeker I, Hoffmann F, Callhoff J, et al. Determinants of psychological well-being in the wide spectrum of comorbidities in RMDs, some are closely axial spondyloarthritis: an analysis based on linked claims and patient-reported survey depending on the patients’ age, and data from the Consortium of data. Ann Rheum Dis 2018;77:1017–24. Rheumatology Researchers of North America registry reveal that 2 WHO. Mental health: a state of well-being. http://www.​who.​int/​features/​factfiles/​ at least in rheumatoid arthritis, the overall prevalence of comor- mental_​health/​en/ (accessed 15 Mar 2018). bidities is higher in older patients. These subjects are less likely to 3 Alunno A, Studenic P, Nikiphorou E, et al. Person-focused care for young people with 7 rheumatic and musculoskeletal diseases: young rheumatologists’ and EULAR Young achieve disease remission. Furthermore, pain might not neces- PARE perspectives. RMD Open 2017;3:e000514–6. sarily correlate with disease activity in SpA as in other RMDs, 4 Sokolovic S, Dervisevic V, Fisekovic S. Mental health status can reflect disease activity in hence it should be considered as a separate domain that can inde- rheumatoid arthritis. Eur J Rheumatol 2014;1:55–7. pendently affect psychological well-being.8–10 It would therefore 5 Dures E, Almeida C, Caesley J, et al. Patient preferences for psychological support in inflammatory arthritis: a multicentre survey. Ann Rheum Dis 2016;75:142–7. be very interesting to see whether data gathered from Redeker et 6 Meesters JJ, Bremander A, Bergman S, et al. The risk for depression in patients al in SpA would mirror those observed in RA, and how the preva- with ankylosing spondylitis: a population-based cohort study. Arthritis Res Ther lence of depressive symptoms, comorbidities, disease activity, pain 2014;16:418.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 2 Correspondence

7 Ranganath VK, Maranian P, Elashoff DA, et al. Comorbidities are associated with 9 Kvrgic Z, Asiedu GB, Crowson CS, et al. "Like No One Is Listening to Me": A Qualitative poorer outcomes in community patients with rheumatoid arthritis. Rheumatology Study of Patient-Provider Discordance Between Global Assessments of Disease Activity in 2013;52:1809–17. Rheumatoid Arthritis. Arthritis Care Res 2017 (Epub ahead of Print Dec 2017). 8 Studenic P, Smolen JS, Aletaha D. Near misses of ACR/EULAR criteria for remission: 10 Rifbjerg-Madsen S, Christensen AW, Christensen R, et al. Pain and pain mechanisms effects of patient global assessment in Boolean and index-based definitions. Ann in patients with inflammatory arthritis: a Danish nationwide cross-sectional DANBIO Rheum Dis 2012;71:1702–5. registry survey. PLoS One 2017;12:e0180014.

2 of 2 Ann Rheum Dis June 2019 Vol 78 No 6 Correspondence response Response to ‘Missing pebble in the mosaic of thereafter. Disease activity and functional impairment (Bath Ankylosing Spondylitis Disease Activity Index and Bath Anky- rheumatic diseases and mental health: younger losing Spondylitis Functional Index, respectively) were lowest does not always mean happier’ by Alunno et al in the youngest group (18–29 years) and rather comparable in all other age groups. In their letter ‘A missing pebble in the mosaic of rheumatic Data from the Consortium of Rheumatology Researchers diseases and mental health: younger does not always mean of North America registry revealed that among persons with happier’,1 Alunno et al commented on our recent publication2 rheumatoid arthritis (RA), the overall prevalence of comor- and raised the issue of determinants of psychological well- bidities is higher in older patients.3 In our study, we found being in different age groups of persons with axial spondyloar- that among persons with axSpA, the prevalence of cardiovas- thritis (axSpA). We very much appreciate their interest in our cular diseases, neurological disorders, metabolic and endo- study and valuable comments made and would like to address crine disorders, osteoarthritis, spondylosis and disorders of the raised issues here in more details. bone density was also increasing with age (table 2). However, In our study, we investigated determinants of psychological mental disorders (according to the claims data) were more well-being in persons with axSpA using the five-item WHO common in middle-aged persons compared with younger and Well-Being Index (WHO-5), a sensitive and specific screening older persons mirroring the distribution of the prevalence of tool for depression. We found that among 1736 persons with depressive symptoms according to the WHO-5. axSpA aged 18–79 years only 42% had a good well-being, In addition, we conducted a multivariable regression analysis for whereas 28% had mild depressive symptoms and 31% even each age group. In all models, we included sex plus all parameters 2 had moderate-to-severe depressive symptoms. In the linear which were found to be associated with depressive symptoms in regression model, we found a negative association between age the original analysis conducted in the entire group (table 3). Age and the odds of having moderate-to-severe depressive symp- groups 18–29 years and 30–39 years were analysed as one group toms. Interestingly, a closer look at the association between in order to achieve convergence of the procedure that was not age and WHO-5 showed a non-linear trend. In fact, the prev- possible otherwise due to a relatively small number of patients in alence of moderate-to-severe depressive symptoms increased the youngest age group. Higher disease activity and a higher level from 15% for patients aged 18–29 years to almost 40% for of functional impairment were associated with moderate-to-severe patients aged 40–59 years and then decreased to 24% for depressive symptoms in almost all age groups (table 3). At the same patients aged 70–79 years (figure 1). time, suffering from stress and lack of exercise showed the stron- Importantly, on the group level, persons aged 30–49 years gest association with depressive symptoms in the youngest group, had higher levels of stress (self-reported) compared with while lower income played a role only in the young and mid-aged younger and older age groups but also reported the highest groups. level of income (table 1). A self-reported lack of exercise was In conclusion, we found the highest prevalence of both, depres- highest in persons aged between 18 and 49 and decreased sive symptoms derived from the WHO-5 score and mental

Figure 1 Prevalence of depressive symptoms (according to the five-item WHOWell-Being Index (WHO-5)) in different age groups.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 3 Correspondence response

Table 1 Main demographic, disease related, lifestyle and socioeconomic characteristics of patients with axial spondyloarthritis in different age groups Age (years) 18–29, n=80 30–39, n=207 40–49, n=361 50–59, n=376 60–69, n=372 70–79, n=340 Sex, female 41.1 47.9 54.7 50.3 45.0 35.7 BASDAI, 0–10 3.5±0.2 4.0±0.1 4.6±0.1 4.8±0.1 4.5±0.1 4.6±0.1 BASFI, 0–10 2.1±0.2 2.6±0.1 3.6±0.1 4.3±0.1 4.6±0.1 5.0±0.1 Lack of exercise 29.9 32.6 27.6 21.5 19.9 24.9 Suffering from stress 50.2 61.1 59.2 48.6 23.6 16.7 Household income, € <1500 31.5 19.6 14.4 24.8 32.0 32.3 1500–3200 57.6 53.3 52.2 53.3 59.1 60.2 >3200 10.9 27.0 33.3 21.9 9.0 7.6 Values are presented as mean±SE of the mean for continuous variables and as percentages otherwise. BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index.

Table 2 Comorbidities according to the claims data in patients with axial spondyloarthritis in different age groups Age (years) 18– 30– 40– 50– 60– 70– 29, n=80 39, n=207 49, n=361 59, n=376 69, n=372 79, n=340 Cardiovascular diseases (%) Hypertensive diseases (I10–I15) 2.1 13.6 31.7 48.9 65.4 84.7 Ischaemic heart diseases (I20–I25) – – 1.6 6.4 16.5 33.7 Diseases of arteries (I70–I79) 1.4 1.4 3.6 4.8 15.6 19.7 Diseases of veins (I80–I89) 4.2 7.5 14.4 17.5 22.2 27.2 Mental disorders (%) Depressive disorders (F32, F33) 13.9 17.6 21.6 27.9 20.9 20.5 Anxiety disorders (F40, F41) 6.4 9.0 8.1 12.5 8.9 8.1 Reaction to severe stress and adjustment disorders (F43) 6.0 10.6 11.2 13.5 9.1 4.9 Somatoform disorders (F45) 13.1 16.3 23.3 25.2 17.7 21.8 Neurological disorders (%) Nerve, nerve root and plexus disorders (G50–G59) 3.5 5.7 9.6 14.0 13.2 13.4 Polyneuropathies (G60–G64) – 2.0 3.6 6.4 10.5 17.4 Sleep disorders (G47) – 4.9 6.9 9.6 11.3 15.8 Obstructive sleep apnoea (G47.31) – – 1.4 2.4 4.6 5.0 Musculoskeletal disorders (other than axSpA) (%) Osteoarthritis (M15–M19) 3.9 11.6 19.4 34.8 47.0 57.6 Spondylosis (M47) 17.7 16.4 20.4 24.4 25.8 30.4 Other soft tissue disorders, not elsewhere classified (M79) 20.9 29.5 27.8 28.4 25.7 22.8 Fibromyalgia (M79.7) 2.1 2.5 4.0 5.8 4.8 4.3 Disorders of bone density (M80–M85) 2.1 4.2 5.5 13.6 19.6 18.6 Metabolic and endocrine disorders (%) Disorders of thyroid gland (E00–E07) 9.2 14.9 26.4 29.6 36.4 29.5 Diabetes mellitus (E10–E14) – 4.8 9.0 13.1 21.3 30.6 Type 2 diabetes mellitus (E11) – 1.8 7.7 11.3 18.6 28.0 Overweight (E65–E68) 5.7 10.8 13.0 14.6 15.6 19.2 Respiratory tract diseases (%) Chronic obstructive pulmonary disease (J44) – 1.0 5.2 7.4 13.2 14.2 Asthma bronchiale (J45) 8.1 10.7 11.7 9.8 10.7 7.7 Gastrointestinal diseases (%) Diseases of oesophagus, stomach and duodenum (K20–K31) 19.2 13.7 19.7 25.3 25.8 32.6 disorders according to the claims data, in middle-aged persons to the higher level of depressive symptoms in the mid-aged popu- (40–59 years of age) with axSpA. These persons reported the lation. Therefore, according to our data, a careful evaluation of highest prevalences of stress and lack of exercise which is likely depressive symptoms can be recommended in particular in individ- related to factors such as career-oriented and family demands. The uals with axSpA aged between 40 and 59. In this age group, disease mentioned factors together with the impact of the disease—that specific, socioeconomic and lifestyle factors were associated with was very similar across the age groups—are likely to be responsible the highest risk of a depressive disorder.

2 of 3 Ann Rheum Dis June 2019 Vol 78 No 6 Correspondence response

Table 3 Factors associated with the presence of symptoms suggestive of depression (WHO-5 score of ≤28) in the multivariable regression analysis in different age groups All patients*, n=1736, Age 18–39, Age 40–49, n=361, Age 50–59, n=376, Age 60–69, n=372, Age 70–79, n=340, Reference OR (95% CI) n=287,OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Age Per 10 years 0.98 (0.97 to 0.99) – – – – – Sex, female Male 1.00 (0.77 to 1.29) 1.29 (0.68 to 2.43) 0.70 (0.42 to 1.18) 1.01 (0.59 to 1.72) 0.86 (0.48 to 1.56) 1.57 (0.85 to 2.91) BASDAI Per unit 1.37 (1.27 to 1.49) 1.11 (0.91 to 1.35) 1.40 (1.21 to 1.63) 1.28 (1.06 to 1.54) 1.74 (1.40 to 2.16) 1.38 (1.13 to 1.67) BASFI Per unit 1.25 (1.17 to 1.33) 1.37 (1.13 to 1.66) 1.17 (1.02 to 1.34) 1.35 (1.18 to 1.56) 1.15 (0.99 to 1.34) 1.23 (1.06 to 1.42) Lack of exercise No 1.50 (1.14 to 1.98) 1.91 (1.02 to 3.55) 1.26 (0.73 to 2.15) 1.23 (0.66 to 2.32) 1.84 (0.93 to 3.64) 1.77 (0.90 to 3.49) Suffering from No 2.03 (1.55 to 2.64) 3.39 (1.66 to 6.94) 2.01 (1.19 to 3.37) 2.04 (1.19 to 3.50) 1.20 (0.62 to 2.33) 1.87 (0.87 to 4.01) stress Household > € 3200 1.88 (1.27 to 2.78) 2.42 (0.95 to 6.20) 1.25 (0.55 to 2.81) 2.52 (1.13 to 5.61) 1.65 (0.50 to 5.39) 2.06 (0.45 to 9.4) income, < € 1500 Household > € 3200 1.54 (1.08 to 2.19) 3.12 (1.42 to 6.85) 1.37 (0.78 to 2.42) 1.83 (0.88 to 3.82) 0.76 (0.24 to 2.41) 2.20 (0.49 to 9.92) income, € 1500– 3200 *The original model including age as a continuous covariate is shown for a reference purpose. Odds ratios of variables associated with a WHO-5 score of ≤ 28 are shown in bold. BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index; WHO-5, five-item WHO Well-Being Index.

Imke Redeker,1 Falk Hoffmann,2 Johanna Callhoff,1 Hildrun Haibel,3 Joachim Sieper,3 Angela Zink,1,4 Denis Poddubnyy1,3 1 Epidemiology Unit, German Rheumatism Research Centre, Berlin, Germany To cite Redeker I, Hoffmann F, Callhoff J, et al. Ann Rheum Dis 2019;78:e55. 2Department of Health Services Research, Carl von Ossietzky University, Oldenburg, Germany 3 Received 8 May 2018 Department of Gastroenterology, Infectiology and Rheumatology, Charité – Accepted 14 May 2018 Universitätsmedizin Berlin, Berlin, Germany Published Online First 31 May 2018 4Department of Rheumatology and Clinical Immunology, Charité – Universitätsmedizin Berlin, Berlin, Germany Correspondence to Imke Redeker, Epidemiology Unit, German Rheumatism Research Centre, Berlin 10117, Germany; ​imke.​redeker@​drfz.​de Handling editor Francis Berenbaum ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2018-​213611 Contributors All authors have substantially contributed to conducting the Ann Rheum Dis 2019;78:e55. doi:10.1136/annrheumdis-2018-213620 underlying research and drafting this manuscript. Funding This study was funded by Bundesministerium für Bildung und Forschung References (grant number 01EC1405). 1 Alunno A, Studenic P, Wiek D, et al. Missing pebble in the mosaic of rheumatic Competing interests None declared. diseases and mental health: younger does not always mean happier. Ann Rheum Dis Ethics approval Ethics committee of the Charité – Universitätsmedizin Berlin, 2019;78:e57. Berlin, Germany. 2 Redeker I, Hoffmann F, Callhoff J, et al. Determinants of psychological well-being in axial spondyloarthritis: an analysis based on linked claims and patient-reported survey Provenance and peer review Commissioned; internally peer reviewed. data. Ann Rheum Dis 2018;77:1017–24. © Article author(s) (or their employer(s) unless otherwise stated in the text of the 3 Ranganath VK, Maranian P, Elashoff DA, et al. Comorbidities are associated with article) 2019. All rights reserved. No commercial use is permitted unless otherwise poorer outcomes in community patients with rheumatoid arthritis. Rheumatology expressly granted. 2013;52:1809–17.

Ann Rheum Dis June 2019 Vol 78 No 6 3 of 3 Correspondence Forward to the past: ultrasound might be diversity of affected joints, a standardised ultrasound protocol is unlikely to be suitable for such a study and a more prag- necessary in some patients with matic approach seems prudent. Only when such a study has rheumatoid arthritis been performed, the efficacy on ultrasound-guided therapy for these patients can be adequately evaluated.6 In the January issue of ARD, Caporali and Smolen1 argue Hence, although Caporali and Smolen are probably right that against the use of ultrasound in the management of rheuma- ultrasound surveillance of RA is not necessary on a general basis, toid arthritis (RA), citing the negative findings of the TaSER2 there is a relevant number of patients in whom adequate treat- and ARCTIC3 studies and a lack of standardisation. Even ment might be ultrasound dependent. Categorically, dismissing disregarding that the results were not unequivocal, showing ultrasound in patient management and reserving it for diagnostic a slight, but not statistically significant difference between and differential diagnostic issues would mean that these patients the ultrasound and control groups after a longer time span, I are denied the treatment they need. Rather, we need to identify strongly disagree with the overall conclusion. these patients using all the tools at our disposal and to design There is no doubt that modern disease management has studies that address their situation specifically. led to excellent outcomes for most patients with RA, with the mutilating disease courses of the prebiological era largely a Ottar Gadeholt 4 thing of the past. However, there are still patients, although Correspondence to Dr Ottar Gadeholt, Department of Rheumatology, University few, whose disease defies our expectations and causes consid- Clinic Würzburg, Würzburg 97080, Germany; gadeholt_​ ​o@ukw.​ ​de erable structural damage. Notable among these patients are the Funding This research received no specific grant from any funding agency in the so-called ‘silent progressors’, in whom radiological progres- public, commercial or not-for-profit sectors. sion occurs despite relative clinical well-being and normalised Competing interests None declared. inflammatory markers. This phenomenon can occur in most Patient consent Not required. joints. Overall, this group is small, yet, clinically relevant. By means of clinical examination, these patients cannot be distin- Provenance and peer review Not commissioned; internally peer reviewed. guished from those patients with long-standing RA in remis- © Article author(s) (or their employer(s) unless otherwise stated in the text of the sion, in whom a slight synovial swelling persists. Power Doppler article) 2019. All rights reserved. No commercial use is permitted unless otherwise expressly granted. sonography (PDS), however, can help to differentiate the two groups. It is conceivable that intensified treatment of patients who display PDS activity without any joint pain could lead to an To cite Gadeholt O. Ann Rheum Dis 2019;78:e56. improved outcome; on the other hand, intensifying treatment Received 20 February 2018 for every patient with swollen, but painless joints, would lead Accepted 21 February 2018 to overtreatment of patients with residual swelling without Published Online First 6 March 2018 inflammatory activity. This would have negative effects both Ann Rheum Dis 2019;78:e56. doi:10.1136/annrheumdis-2018-213278 concerning costs and adverse events. Admittedly, I know of no study in which ultrasound-guided therapy has improved outcome in these patients, or even References whether any systemic or intra-articular therapy has a rele- 1 Caporali R, Smolen JS. Back to the future: forget ultrasound and focus on clinical assessment in rheumatoid arthritis management. Ann Rheum Dis vant effect on synovitis in these joints. On the other hand, 2018;77:18–20. this could just as well be due to the lack of designated studies, 2 Dale J, Stirling A, Zhang R, et al. Targeting ultrasound remission in early rheumatoid rather than to a lack of effect in itself. In a large-scale, popu- arthritis: the results of the TaSER study, a randomised clinical trial. Ann Rheum Dis lation-based study, the patient group in question is too small 2016;75:1043–50. 3 Haavardsholm EA, Aga AB, Olsen IC, et al. Ultrasound in management of rheumatoid to have a significant impact on overall results, which is a likely arthritis: ARCTIC randomised controlled strategy trial. BMJ 2016;354:i4205. explanation for the negative results of the TaSER and ARCTIC 4 The Lancet. A platinum age for rheumatology. Lancet 2017;389:2263. trials.2 3 It follows that any study exploring the ideal treatment 5 Sewerin P, Vordenbaeumen S, Hoyer A, et al. Silent progression in patients with of ‘silently progressing RA’ must necessarily be performed rheumatoid arthritis: is DAS28 remission an insufficient goal in RA? Results from the on such patients only, with the therapy and control group German Remission-plus cohort. BMC Musculoskelet Disord 2017;18:163. 6 Nguyen H, Ruyssen-Witrand A, Gandjbakhch F, et al. Prevalence of ultrasound-detected being phenotypically similar; for example, patients with low residual synovitis and risk of relapse and structural progression in rheumatoid arthritis clinical and serological activity, but with positive PDS and/ patients in clinical remission: a systematic review and meta-analysis. Rheumatology or MRI scans indicating subclinical synovitis.5 6 Due to the 2014;53:2110–8.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Correspondence Cardiovascular events in ankylosing spondylitis: In 11 longitudinal studies (n=51 990), 2183 strokes were reported in patients with AS over 14.7 years of follow-up. a 2018 meta-analysis The incidence was 1.9% (95% CI 0.8% to 3.4%), or 0.18/100 pyrs. Seven studies reported 31 871 strokes in controls In an article published in the Annals of the Rheumatic Diseases, 1 (n=1 624 844); thus, the incidence was 2.1% (95% CI 1.2% to Schieir et al reported a trend of increased cardiovascular risk 3.4%). A significant increase in stroke among patients with AS in patients with ankylosing spondylitis (AS). In this letter, we compared with controls was found in a meta-analysis of seven wish to add data to support the conclusions of that article. We studies (RR=1.37; 95% CI 1.08 to 1.73; figure 2). performed an updated meta-analysis to investigate the risks of These results reinforced the conclusions of Schieir et al, who myocardial infarction (MI) and stroke in patients with AS and found that AS was associated with a significant increase in the controls. risks of MI and stroke. The reasons for this increased cardio- We searched PubMed to find reports of interest, published vascular risk are probably multifactorial. Systemic inflamma- up to February 2018. We included all observational or case/ tion and high disease activity play pivotal roles in increased control studies that reported rates of MI or stroke. We calculated cardiovascular risk in rheumatic disease. The role of non-ste- the incidences of MI and stroke in a meta-analysis of propor- roidal anti-inflammatory drugs remains an issue of debate.2 tions (inverse variance method) and expressed them in terms Another possible explanation involves the proatherogenic of 100 patient-years (pyrs) of exposure. We used the Mantel- profile of patients with AS who were smokers and/or hyper- Haenszel procedure to determine the risk ratios (RRs) of MI and stroke. tensive with a poor atherogenic lipid profile. Cardiovascular In the 18 included studies (online supplementary file), 2615 risk factors and systemic inflammation should be managed in MIs were reported in patients with AS (n=51 660) over a mean AS to reduce the high cardiovascular risk. Recently, recom- mendations have been published for improving the cardiovas- follow-up period of 13 years. The incidence was 2.6% (95% 3 CI 1.1% to 4.5%), or 0.18/100 pyrs. Twelve studies revealed cular risk profile. Compared with our previous meta-analysis, the occurrence of 24 472 MIs (mean incidence: 2.0%; 95% CI we found a reduction in the prevalence and risk of MI and 4 1.6% to 2.5%) in control individuals (n=1 735 909). A stroke. This information is encouraging; it might be attrib- meta-analysis of 12 longitudinal studies showed a significant utable to improvements in the control of cardiovascular risk increase in the risk of MI (RR=1.44; 95% CI 1.25 to 1.67) in in AS. However, efforts must continue because management patients with AS compared with controls (figure 1). of cardiovascular comorbidities in rheumatic disease remains insufficient.5

Figure 1 Myocardial infarction risk compared between patients with AS and controls. ACR, American College of Rheumatology; AS, ankylosing spondylitis; EULAR, European League Against Rheumatism; M-H, Mantel-Haenszel.

Figure 2 Stroke risk compared between patients with AS and controls. ACR, American College of Rheumatology; AS, ankylosing spondylitis; M-H, Mantel-Haenszel.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 2 Correspondence

Sylvain Mathieu, Martin Soubrier Rheumatology Department, Clermont University, Gabriel Montpied Teaching Hospital, Clermont-Ferrand, France To cite Mathieu S, Soubrier M. Ann Rheum Dis 2019;78:e57. Correspondence to Dr Sylvain Mathieu, Rheumatology Department, Gabriel Montpied Teaching Hospital, 63003 Clermont-Ferrand, France; Received 27 February 2018 smathieu@​ ​chu-clermontferrand.​ ​fr Accepted 28 February 2018 Published Online First 9 March 2018 Contributors Study concept and design: SM and MS. Acquisition of data: SM. Statistical analysis and interpretation of data: SM. Drafting of the manuscript: SM. Ann Rheum Dis 2019;78:e57. doi:10.1136/annrheumdis-2018-213317 Critical revision of the manuscript for important intellectual content: SM and MS. References 1 Schieir O, Tosevski C, Glazier RH, et al. Incident myocardial infarction associated with Funding This research received no specific grant from any funding agency in the major types of arthritis in the general population: a systematic review and meta- public, commercial or not-for-profit sectors. analysis. Ann Rheum Dis 2017;76:1396–404. Competing interests None declared. 2 Nissen SE, Yeomans ND, Solomon DH, et al. Cardiovascular safety of celecoxib, naproxen, or ibuprofen for arthritis. N Engl J Med 2016;375:2519–29. Patient consent Not required. 3 Agca R, Heslinga SC, Rollefstad S, et al. EULAR recommendations for cardiovascular Provenance and peer review Not commissioned; internally peer disease risk management in patients with rheumatoid arthritis and other forms of reviewed. inflammatory joint disorders: 2015/2016 update. Ann Rheum Dis 2017;76:17–28. 4 Mathieu S, Pereira B, Soubrier M. Cardiovascular events in ankylosing spondylitis: an © Article author(s) (or their employer(s) unless otherwise stated in the text of the updated meta-analysis. Semin Arthritis Rheum 2015;44:551–5. article) 2019. All rights reserved. No commercial use is permitted unless otherwise 5 Tournadre A, Pereira B, Dubost JJ, et al. Management of dyslipidaemia in high-risk expressly granted. patients with recent-onset rheumatoid arthritis: targets still not met despite specific ►► Additional material is published online only. To view please visit the journal recommendations. Results from the ESPOIR cohort during the first five years of follow- online (http://​dx.doi.​ ​org/​10.​1136/annrheumdis-​ ​2018-213317).​ up. Clin Exp Rheumatol 2017;35:296–302.

2 of 2 Ann Rheum Dis June 2019 Vol 78 No 6 Correspondence Checkpoint inhibitors and arthritis to disease. In the case of checkpoint inhibitors, this regulatory check point is lost and cells stimulated by cigarette smoke (nicotine) lead It was with interest that we read the article by Belkhir et al1 in to a heightened production of RF or anti-CCP. Smoking in France the recent Annals of Rheumatic Disease. The article described is much more prevalent than in Australia. It would be interesting the first report of rheumatoid arthritis (RA) and polymyalgia to know what percentage of the patients in the above study were rheumatica developing after the use of immune checkpoint smokers. inhibitors to treat a variety of cancers. We found the article inter- The reporting of musculoskeletal immune adverse effects esting and informative but noted that in all of their 10 reported due to checkpoint inhibitors is in its infancy. We believe it is cases, the patients seroconverted from rheumatoid factor (RF) or important that the experience that groups in different countries cyclic citrullinated peptide antibody (CCP) negative to positive, are finding needs to be reported. supporting their claim that ‘rheumatoid arthritis’ may develop Thank you for this thought-provoking article. after immune checkpoint inhibitor treatment. Nicholas Manolios,1,2 Leslie Schrieber1,3 The report by Belkhir et al is in contrast to our experience 1Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia managing 18 patients with musculoskeletal immune adverse 2Department of Rheumatology, Westmead Hospital, Sydney, New South Wales, events, subsequent to checkpoint inhibitor treatment. None Australia of our patients seroconverted after treatment (manuscript in 3Department of Rheumatology, Royal North Shore Hospital, Sydney, New South preparation) and none had clinical features suggestive of RA. In Wales, Australia addition to the negative serology, MRI of affected joints failed Correspondence to Professor Nicholas Manolios, University of Sydney, Sydney, to identify synovial hypertrophy to suggest the start of pannus NSW 2006, Australia; nicholas​ .​manolios@sydney.​ ​edu.​au formation. In one patient with hypophysitis who did develop Contributors Both authors made substantial contributions to the conception or autoantibodies (cytoplasmic anti-neutrophil cytoplasmic anti- design of the letter and important intellectual content. bodies (cANCA); proteinase 3 (PR3) >100 U), there was no Funding The authors have not declared a specific grant for this research from any clinical evidence of vasculitis. We find that the autoimmune funding agency in the public, commercial or not-for-profit sectors. toxicities differ from the ‘classical’ autoimmune entities known Competing interests None declared. to each specialty, for example, the colitis is not ulcerative colitis Patient consent Not required. or Crohn’s; it is its own entity, with widely varying patholog- Provenance and peer review Not commissioned; internally peer reviewed. ical findings between patients. Similarly with arthritis, there © Article author(s) (or their employer(s) unless otherwise stated in the text of the was no ‘hallmark’ pathological finding even though tenosyno- article) 2019. All rights reserved. No commercial use is permitted unless otherwise vitis featured highly. In accord with Belkhir, all of our patients expressly granted. responded well to non-steroidal anti-inflammatory drugs, low dose steroids and, in certain cases, methotrexate. We cannot explain the different experience we have had with our patient group (France vs Australia). We wonder if there is a To cite Manolios N, Schrieber L. Ann Rheum Dis 2019;78:e58. subgroup of patients that do seroconvert and are curious to find Received 14 March 2018 what the cause may be to account for such dissimilarities. We can Accepted 16 March 2018 only speculate on possibilities such as geography, ethnicity, diet Published Online First 23 March 2018 or smoking. For instance, with regards to smoking, we have data Ann Rheum Dis 2019;78:e58. doi:10.1136/annrheumdis-2018-213415 showing that nicotine can influence T cell receptor function that consequently could influence antibody formation (manuscript in References preparation). Other published modes of action of nicotine include 1 Belkhir R, Burel SL, Dunogeant L, et al. Rheumatoid arthritis and polymyalgia upregulation of peptidylarginine deiminase that enhances the rheumatica occurring after immune checkpoint inhibitor treatment. Ann Rheum Dis 2 2017;76:1747–50. production of CCP. Under normal circumstances, the detection of 2 Makrygiannakis D, Hermansson M, Ulfgren AK, et al. Smoking increases autoimmune clones by the immune system leads to their suppression peptidylarginine deiminase 2 enzyme expression in human lungs and increases by regulatory suppressor immune cells that prevent the progression citrullination in BAL cells. Ann Rheum Dis 2008;67:1488–92.

Ann Rheum Dis June 2019 Vol 78 No 6 1 of 1 Miscellaneous Correction: The development of assessment of SpondyloArthritis international society classification criteria for axial spondyloarthritis (part II): validation and final selection

Rudwaleit M, van de Heijde D, Landewe R, et al. The development of Assessment of Spondy- loArthritis international Society classification criteria for axial spondyloarthritis (part II): vali- dation and final selection. Ann of Rheum Dis 2009;68:777–83. doi: 10.1136/ard.2009.108233 Conflicts of interest (competing interests) for this article were not provided in the paper. This correction now serves to provide conflict of interest statements of individual authors and of ASAS as an organisation with relevance to the years 2008/2009. Conflicts of interest (in the order of authorship): M Rudwaleit: speaking fees and/or consulting fees from Abbott, Centocor, MSD, Pfizer, Roche, Schering-Plough, Wyeth D van der Heijde: research grants and/or consulting fees from Abbott, Amgen, BMS, Centocor, Chugai, Novartis, Pfizer, Schering-Plough, UCB and Wyeth. R Landewé: research grants and/or consulting fees Abbott, Amgen/Immunex, Centocor, MSD (Merck), Shering-Plough, UCB, Wyeth J Listing: no competing interests. N Accoc: speaking fees and/or consultancies from Abbott, Wyeth, UCB, Schering and Novartis. J Brandt: consultancy or speaking fees from Abbott, MSD, Pfizer, Roche. J Braun: nothing declared CT Chou: nothing declared E Collantes-Estevez: research grants or consultation fees from Abbott, Schering Plough and Wyeth M Dougados: research grants, speaking fees and consulting fees from Abbott, Janssen, Lilly, Merck, Novartis, Wyeth, UCB F Huang: no competing interests. J Gu: nothing declared. MA Khan: consultancy and speaker fees from Amgen and Abbott. Y Kirazli: research grants or consultation fees from Allergan, Amgen, Lilly, Novartis, Pfizer and Sandoz. WP Maksymowych: fees for consulting, honoraria, and research grants from Abbott, Amgen-Wyeth, Schering-Plough. H Mielants: no competing interests. IJ Sørensen: lecture fees from Schering-Plough A/S, Wyeth and Abbott; consultancy fees from Bristol-Myers Squibb. S Ozgocmen: no competing interests. E Roussou: no competing interests. R Valle-Oñate: no competing interests. U Weber: no competing interests. J Wei: research grants or consultation fee from Abbott, BMS, Chugai, Eisai, Wyeth. J Sieper: honoraria for consultancies and/or being a member of speaker’s bureau from Abbott, Amgen, Centocor, Merck, Schering-Plough, Wyeth. ASAS organisation: ASAS has received unrestricted grants for corporate membership from Abbott, Centocor, MSD, Pfizer, Schering-Plough, Wyeth.

© Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ. Ann Rheum Dis 2019;78:e59. doi:10.1136/ard.2009.108233corr1

Ann Rheum Dis 2019;78:e59. doi:10.1136/ard.2009.108233corr1 1 of 1 Miscellaneous Correction: Visualisation of interstitial lung disease by molecular imaging of integrin αvβ3 and somatostatin receptor 2

Schniering J, Benešová M, Brunner M, et al. Visualisation of interstitial lung disease by molec- ular imaging of integrin αvβ3 and somatostatin receptor 2. Ann of Rheum Dis ​2018;​78:​218-​ 27.​doi: 10.1136/annrheumdis-2018-214322 The funding statement should read: This work was supported by the Swiss National Science Foundation (Grant: CRSII3_154490), Hartmann-Mueller Foundation and Prof Dr Max Cloetta Foundation.

© Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ. Ann Rheum Dis 2019;78:e60. doi:10.1136/annrheumdis-2018-214322corr1

Ann Rheum Dis 2019;78:e60. doi:10.1136/annrheumdis-2018-214322corr1 1 of 1