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) Rik Lories (Belgium) Kimme Hyrich (UK) Johan Askling (Sweden) Ingrid Lundberg (Sweden) Contact Details Rik Lories (Belgium) Sang-Cheol Bae (Korea) Gary MacFarlane (UK) Iain McInnes (UK) Xenofon Baraliakos (Germany) Xavier Mariette (France) Editorial Office Thomas Pap (Germany) Anne Barton (UK) Alberto Martini (Italy) Annals of the Rheumatic Diseases David Pisetsky (USA) Maarten Boers (The Netherlands) Marco Mattuci Cerinic (Italy) BMJ Journals, BMA House, Tavistock Square Désirée van der Heijde Maxine Breban (France) Dennis McGonagle (UK) London WCIH 9JR,UK (The Netherlands) Matthew Brown (Australia) Fred Miller (USA) Kazuhiko Yamamoto (Japan) Maya Buch (UK) Peter Nash (Australia) E: [email protected] Loreto Carmona (Spain) Michael Nurmohamed (The Methodological and Statistical Carlo Chizzolini (Switzerland) Netherlands) Production Editor Advisor Bernard Combe (France) Caroline Ospelt (Switzerland) Teresa Jobson Philip Conaghan (UK) Stian Lydersen (Norway) Monika Østensen (Norway) E: [email protected] Maurizio Cutolo (Italy) Constatino Pitzalis (UK) Social Media Advisors Nicola Dalbeth (Australia) Alessia Alunno (Italy) Jane Salmon (USA) EULAR Christian Dejaco (Austria) Georg Schett (Germany) Mary Canavan (Ireland) Oliver Distler (Switzerland) EULAR Executive Secretariat Thomas Dörner (Germany) José da Silva (Portugal) Meghna Jani (UK) Hendrik Schulze-Koops (Germany) Seestrasse 240, 8802 Kilchberg, Switzerland Elena Nikiphorou (UK) Dirk Elewaut (Belgium) Axel Finckh (Switzerland) Nan Shen (China) E: [email protected] Caroline Ospelt (Switzerland) Greg Silverman (USA) www.eular.org Christophe Richez (France) Rebecca Fischer-Betz (Germany) Roy Fleischmann (USA) Alexander So (Switzerland) Paul Studenic (Austria) Mary Goldring (USA) Hiroshi Takayanagi (Japan) Customer support Laure Gossec (France) Tsutomu Takeuchi (Japan) For general queries and support with existing and Guidelines for Authors and Yoshiya Tanaka (Japan) Walter Grassi (Italy) new subscriptions: Reviewers Ahmet Gül (Turkey) Dimitrios Vassilopoulos (Greece) Full instructions are available Frederic Houssiau (Belgium) Douglas Veale (Ireland) W: support.bmj.com online at http://ard.bmj.com/ Tom Huizinga (The Netherlands) Jiri Vencovsky (Czech Republic) T: +44 (0)20 7111 1105 Arthur Kavanaugh (USA) Ronald van Vollenhoven (Sweden) E: [email protected] pages/authors. Articles must Margreet Kloppenburg (The Erwin Wagner (Spain) be submitted electronically at Netherlands) Michael Ward (USA) Self-archiving and permissions http://mc.manuscriptcentral. Robert Landewé (The Netherlands) Kevin Winthrop (USA) W: bmj.com/company/products-services/ Zhan-Gou Li (China) Huji Xu (China) com/ard. Authors retain copyright rights-and-licensing/ but are required to grant ARD Chairman of Advisory E: [email protected] an exclusive licence to publish. Committee (http://ard.bmj.com/pages/authors/). Johannes Bijlsma Advertising (The Netherlands) W: bmj.com/company/for-advertisers- and-sponsor/ Advisory Committee Ferry Breedveld (The Netherlands) Zhan-guo Li (China) Display Advertising ROW Marco Matucci Cerinic (Italy) Peter Lipsky (USA) Sophie Fitzsimmons Michael Doherty (UK) Sir Ravinder Maini (UK) T: +44 (0)20 3655 5612 Maxime Dougados (France) Emilio Martín-Mola (Spain) Paul Emery (UK) Haralampos Moutsopoulos E: [email protected] Daniel Furst (USA) (Greece) Online Advertising ROW Steffen Gay (Switzerland) Karel Pavelka (Czech Republic) Marc Hochberg (USA) Yehuda Shoenfeld (Israel) Marc Clifford Joachim Kalden (Germany) Leo van de Putte (The Netherlands) T: +44 (0) 20 3655 5610 Edward Keystone (Canada) Frank Wollheim (Sweden) E: [email protected] Lars Klareskog (Sweden) Anthony Woolf (UK) Tore Kvien (Norway) 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 2020 Personal Rates 2020 BMJ Reprints Team E: [email protected] Print Print (includes online access at no additional cost) Commercial Reprints ROW £1088 £450 Nadia Gurney-Randall M: +44 07866 262344 Online Online only £192 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 79 Issue 5 | ARD May 2020

ARDThe Eular Journal Impact Factor: 14.299 Editorial Spondyloarthritis

may 2020 Volume 79 Issue 5 551 COVID-19 and rheumatology: first steps 595 Dual neutralisation of interleukin-17A and 79 5 Annals of the oue7 Ise5 Pages 551–676 Volume 79 Issue 5 Rheumatic towards a different future? interleukin-17F with bimekizumab in patients Diseases I B McInnes with active ankylosing spondylitis: results The EULAR Journal from a 48-week phase IIb, randomised,

AnnAls of RheumAtIc DIseAses double-blind, placebo-controlled, dose-ranging Viewpoint study 553 Disruptive innovation in rheumatology: new D van der Heijde, L S Gensler, A Deodhar, networks of global public–private partnerships X Baraliakos, D Poddubnyy, A Kivitz, M K Farmer, D Baeten, N Goldammer, J Coarse, M Oortgiesen, the home for innovative are needed to take advantage of scientific m ay rheumatology research 2020 ard.bmj.com progress M Dougados T WJ Huizinga, V M Holers, J Anolik, M B Brenner, C D Buckley, V Bykerk, S E Connolly, K D Deane, 605 Reactive arthritis and other musculoskeletal Editor J Guo, M Hodge, S Hoffmann, F Nestle, C Pitzalis, symptoms associated with acquisition of Josef S Smolen S Raychaudhuri, K Yamamoto, Z Li, L Klareskog diarrhoeagenic Escherichia coli (DEC) Associate Editors R Tuompo, T Lääveri, T Hannu, S H Pakkanen, Francis Berenbaum J Kirveskari, M Leirisalo-Repo, A Kantele Dimitrios Boumpas Gerd Burmester Rheumatoid arthritis Mary Crow 556 Two-year cost-effectiveness of different Kimme Hyrich Rik Lories COBRA-like intensive remission induction Systemic lupus erythematosus Iain McInnes schemes in early rheumatoid arthritis: a 612 Impact of the new American College of Thomas Pap David Pisetsky piggyback study on the pragmatic randomised Cardiology/American Heart Association Désirée van der Heijde controlled CareRA trial definition of hypertension on atherosclerotic Kazuhiko Yamamoto S Pazmino, A Boonen, V Stouten, D De Cock, J Joly, vascular events in systemic lupus Editorial office K Van der Elst, R Westhovens, P Verschueren erythematosus Annals of the Rheumatic Diseases K Tselios, D D Gladman, J Su, M Urowitz BMJ Publishing Group Ltd 566 Risk of neuroinflammatory events in arthritis BMA House Tavistock Square patients treated with tumour necrosis factor London WCIH 9JR,UK alpha inhibitors: a collaborative population- T: +44 (0)20 3655 5889 Systemic sclerosis based cohort study from Denmark and Sweden E: [email protected] 618 Riociguat in patients with early diffuse Twitter: @ARD_BMJ T I Kopp, B Delcoigne, E V Arkema, R K Jacobsen, ISSN: 0003-4967 (print) M Magyari, E H Ibfelt, H Locht, F Sellebjerg, cutaneous systemic sclerosis (RISE-SSc): ISSN: 1468-2060 (online) R L Cordtz, D V Jensen, J Askling, L Dreyer randomised, double-blind, placebo-controlled

Disclaimer: The Editor of ARD has been granted multicentre trial editorial freedom and ARD is published in 573 Immunosuppression and the risk of D Khanna, Y Allanore, C P Denton, M Kuwana, accordance with editorial guidelines issued by M Matucci-Cerinic, J E Pope, T Atsumi, R Bečvář, the World Association of Medical Editors and the readmission and mortality in patients with L Czirják, E Hachulla, T Ishii, O Ishikawa, Committee on Publication Ethics. ARD is primarily rheumatoid arthritis undergoing hip fracture, intended for healthcare professionals and its S R Johnson, E De Langhe, C Stagnaro, V Riccieri, content is for information only. The Journal is abdominopelvic and cardiac surgery published without any guarantee as to its accuracy E Schiopu, R M Silver, V Smith, V Steen, or completeness and any representations or M D George, J F Baker, K L Winthrop, W Stevens, G Szücs, M-E Truchetet, M Wosnitza, warranties are expressly excluded to the fullest S D Goldstein, E Alemao, L Chen, Q Wu, F Xie, K Laapas, J de Oliveira Pena, Z Yao, F Kramer, extent permitted by law. Readers are advised to J R Curtis independently verify any information on which O Distler they choose to rely. Acceptance of advertising by ARD does not imply endorsement. Neither EULAR nor BMJ Publishing Group Limited shall have any 581 Risk of breast cancer before and after liability for any loss, injury or damage howsoever rheumatoid arthritis, and the impact of arising from ARD (except for liability which cannot be legally excluded). hormonal factors Copyright: © 2020 BMJ Publishing Group Ltd H Wadström, A Pettersson, K E Smedby, J Askling MORE CONTENTS ► and European League Against Rheumatism. All rights reserved; no part of this publication may be reproduced in any form without permission. 587 Anti-carbamylated proteins antibody repertoire This article has been chosen by the Editor to be of special interest­ ARD is published by BMJ Publishing Group Ltd in rheumatoid arthritis: evidence of a new or importance and is freely available online. typeset by Exeter Premedia Services Private Ltd, Chennai, India and printed in the UK on acid-free autoantibody linked to interstitial lung disease This article has been made freely available online under the BMJ paper. R Castellanos-Moreira, S C Rodríguez-García, Journals open access scheme. Annals of the Rheumatic Diseases (ISSN No: ­ M J Gomara, V Ruiz-Esquide, A Cuervo, See http://authors.bmj.com/open-access/ 0003-4967) is distributed in the USA by Air Business Ltd. Periodicals postage paid at I Casafont-Solé, J Ramírez, S Holgado, Jamaica NY 11431. POSTMASTER: send address J A Gómez-Puerta, J D Cañete, I Haro, changes to Annals of the Rheumatic Diseases, Member since 2008 Air Business Ltd, c/o WN Shipping USA, 156-15, R Sanmarti JM00004 146th Avenue, 2nd Floor, Jamaica, NY 11434, This journal is a member of and subscribes to the principles of the USA. Committee on Publication Ethics http://publicationethics.org/ Contents Volume 79 Issue 5 | ARD May 2020

626 Initial combination therapy of ambrisentan and 668 COVID-19 in a patient with systemic sclerosis tadalafil in connective tissue disease-associated treated with tocilizumab for SSc-ILD pulmonary arterial hypertension (CTD-PAH) in C Mihai, R Dobrota, M Schröder, A Garaiman, the modified intention-to-treat population of the S Jordan, Mike Oliver Becker, Britta Maurer, O Distler AMBITION study: post hoc analysis M Kuwana, C Blair, T Takahashi, J Langley, 669 Long-term effectiveness of live herpes zoster J G Coghlan vaccine in patients with rheumatoid arthritis subsequently treated with tofacitinib K L Winthrop, A Wouters, E H Choy, C Chen, Osteoarthritis P Biswas, L Wang, K Soma, E Needle, H Valdez, 635 Metformin limits osteoarthritis development and W FC Rigby progression through activation of AMPK signalling J Li, B Zhang, W-X Liu, K Lu, H Pan, T Wang, 671 CDAI and DAS28 in the management of C Oh, D Yi, J Huang, L Zhao, G Ning, C Xing, rheumatoid arthritis in clinical practice G Xiao, R Liu-Bryan, S Feng, D Chen S Takanashi, Y Kaneko, T Takeuchi

646 Faecal microbiota transplantation from 674 Efficacy of JAK 1/2 inhibition in the treatment metabolically compromised human donors of diffuse non-scarring alopecia due to accelerates osteoarthritis in mice systemic lupus erythematosus ZY Huang, J Chen, BL Li, B Zeng, C-H Chou, K Maeshima, H Shibata X Zheng, J Xie, H Li, Y Hao, G Chen, FX Pei, B Shen, V B Kraus, H Wei, X Zhou, L Cheng 675 Histopathological subgrouping versus renal risk score for the prediction of end-stage renal disease in ANCA-associated vasculitis Crystal arthropathies O Gercik, E Bilgin, D Solmaz, F Cakalagaoglu, 657 Subtype-specific gout susceptibility loci and A Saglam, O Aybi, R C Kardas, Z Soypacaci, enrichment of selection pressure on ABCG2 and G Kabadayi, T Yildirim, I Kurut Aysin, O Karadag, ALDH2 identified by subtype genome-wide meta- S Akar analyses of clinically defined gout patients A Nakayama, M Nakatochi, Y Kawamura, K Yamamoto, H Nakaoka, S Shimizu, T Higashino, Electronic pages T Koyama, A Hishida, K Kuriki, M Watanabe, e49 Inflammatory and non-inflammatory triggers of T Shimizu, K Ooyama, H Ooyama, M Nagase, acute coronary syndromes Y Hidaka, D Matsui, T Tamura, T Nishiyama, M Eyuboglu C Shimanoe, S Katsuura-Kamano, N Takashima, Y Shirai, M Kawaguchi, M Takao, R Sugiyama, e50 Response to: ‘Inflammatory and non- Y Takada, T Nakamura, H Nakashima, M Tsunoda, inflammatory triggers of acute coronary I Danjoh, A Hozawa, K Hosomichi, Y Toyoda, syndromes’ by Eyuboglu Y Kubota, T Takada, H Suzuki, B Stiburkova, H Westerlind, M Holmqvist, L Ljung, T Frisell, T J Major, T R Merriman, N Kuriyama, H Mikami, J Askling T Takezaki, K Matsuo, S Suzuki, T Hosoya, Y Kamatani, M Kubo, K Ichida, K Wakai, I Inoue, Y Okada, N Shinomiya, H Matsuo, on behalf of Japan e51 Assessment of responsiveness of the Gout Genomics Consortium (Japan Gout) musculoskeletal component of SLE-DAS in an independent cohort S-U Hassan, K Mahmoud, E M Vital Letters 666 To consider or not antimalarials as a e52 Response to: ‘Assessment of responsiveness of prophylactic intervention in the SARS-CoV19 the musculoskeletal component of SLE-DAS in (Covid-19) pandemic an independent cohort’, by Hassan et al F R Spinelli, F Ceccarelli, M Di Franco, F Conti D Jesus, M Zen, A Doria, L S Inês

667 Clinical course of COVID-19 in a series of e53 Adipose stromal vascular fraction and patients with chronic arthritis treated with regenerative therapy in SSc: response to the immunosuppressive targeted therapies article by Magalon et al S Monti, S Balduzzi, P Delvino, E Bellis, P Di Benedetto, P Cipriani, P Ruscitti, V Liakouli, V S Quadrelli, C Montecucco R Giacomelli Contents Volume 79 Issue 5 | ARD May 2020 e54 Response to: ‘Adipose stromal vascular fraction e57 MYCYC: unravelling the long road ahead in and regenerative therapy in SSc: response to the ANCA-associated vasculitis article by Magalon et al’ by De Benedetto et al S Jain, A Chattopadhyay, S Naidu, A Sharma J Magalon, M Velier, S Simoncini, F Dignat-George, B Granel, P Paul, F Sabatier e58 Response to: ‘MYCYC, unravelling the long road ahead in ANCA-associated vasculitis’ by e55 Could autologous adipose-derived stromal Jain et al vascular fraction turn out an unwanted source of L Harper, R B Jones profibrotic myofibroblasts in systemic sclerosis? M Manetti e59 Changing landscape of immunosuppression in ANCA-associated vasculitis e56 Response to: ‘Could autologous adipose-derived S V Moiseev, I Smitienko, N Bulanov, E Karovaikina, stromal vascular fraction turn out an unwanted P I Novikov source of profibrotic myofibroblasts in systemic sclerosis?’ by Manetti M Velier, J Magalon, S Simoncini, F Dignat-George, B Granel, P Paul, F Sabatier Editorial

of immune stimulation and host responses COVID-19 and rheumatology: first steps in the evolution of Covid-19. Antiviral agents are urgently required to address towards a different future? patients with Covid-19, but existing agents as yet appear not consistently effective.3–5 Iain B McInnes Remdesivir may offer more benefit though this is being trialled currently. In this context, focus is falling on immune modu- lation to limit tissue damage and espe- As a rheumatology community, we looked hypertension and diabetes.1 Immune cially progression to multiorgan failure. on the new decade with excitement and suppression is similarly flagged as a risk It is apparent that Covid-19 has a phasic ambition yet now instead find ourselves factor. A key question for our commu- component3 with a majority exhibiting amidst the most significant global public nity therefore is the extent to which relatively mild symptomatology with no healthcare challenge in our lifetimes.1 The risk of infection a priori, or progression known sequelae. A smaller but clinically pandemic arising from severe acute respi- thereafter, are altered for people with significant subgroup shows progression to ratory syndrome coronavirus-2 has rapidly RMDs taking immune modifiers, for a more critical illness. A further subgroup changed our personal and professional example, glucocorticoids, biologicals, apparently develops a syndrome with outlook. The rheumatology community targeted synthetic disease-­modifying anti-­ features of a ‘cytokine storm’, reminis- has risen promptly to this challenge and is rheumatic drugs (DMARDs), or indeed cent of, but identical to, haemophagocytic already demonstrating remarkable part- conventional DMARDs. In the ARD, lymphohistiocytosis or that seen following nership in its global and integrated Monti and colleagues2 provide a clinical chimeric antigen receptor T (CAR-­T) cell 6 approach. Rheumatology is a discipline description of Covid-19 in patients with therapeutics. Notably, we should not that has advanced with remarkable pace in inflammatory arthritis receiving immune assume that this is a sequence of events the last decades, driven by revolutionary suppressive therapies. A reassuring in a given individual as outcome could strategic approaches to the management outcome is reported in this small sample; reflect quite discrete host characteristics; of rheumatic and musculoskeletal diseases moreover, on infection, all patients eventually, these states may be identifiable (RMDs), together with increasingly effec- discontinued their immune modifiers on the basis of molecular stratification. tive application of the wealth of possibili- without flaring despite intercurrent viral Interest has fallen accordingly on the use ties contained in modern molecular stimulus. The authors also helpfully point of currently available immune modifiers. medicine. New therapeutics abound based to prior experiences following Middle Early enthusiasm for hydroxychloroquine (HCQ),7 which has theoretical effects on on ever clearer understanding of the East respiratory syndrome and severe virus cell biology via alteration of pH in pathogenesis of our diseases and the acute respiratory syndrome outbreaks in endosomes, lysosomes and Golgi appa- extraordinary capacity in the biotech and which immune suppression, transplan- pharmaceutical industry sector to capi- ratus, together with capacity to block viral tation and cancer did not notably confer talise on the same for therapeutic benefit. replication ex vivo, has been tempered poorer outcomes. At present, therefore, These are precisely the skills and by the lack of high-­quality randomised most recommendations suggest contin- approaches that are now essential as we controlled trial data to support its use in uation of immune modifiers in general combat Covid-19. The immediate public clinic, this despite its approval in some unless Covid-19 is confirmed. This highly health measures that are currently in force, regions. In ARD, Spinelli and colleagues8 commendable report, generated in chal- however effective, will eventually require offer an insightful review of this issue and lenging circumstances, does not, however, a robust underpinning molecular medical remark on the potential for prophylactic definitively address the incidence and response in the form of vaccines and new use. Others have also commented on the implication of Covid-19 in RMDs. To this therapeutics to achieve the long-term­ goal widespread use of HCQ in the absence of end, the EULAR-­COVID-19 Database has 9 10 of normalising quality of life for the controlled evidence, noting also that been launched, comprising a European majority. Rheumatology may therefore be it is not without side effects, especially ideally placed to support our infectious paediatric and adult database designed in the context of potential virus-induced­ diseases and critical care colleagues as we to monitor and report on outcomes of cardiac damage (and prolonged QT enter poorly charted clinical territory. Covid-19 occurring in patients with interval). The data shall decide as formal This month in the Annals of Rheumatic RMDs. In turn, it is designed to work trials are ongoing but at present there Diseases (ARD), we read the first products alongside the Covid-19 Global Rheuma- should perhaps be some restraint in its of the widespread activities concerning tology Alliance that comprises rheuma- widespread application. Meantime there RMDs emerging as we move to combat tologists and epidemiologists and has now is a further direct implication of this high the virus and its various clinical formed a global repository for Covid-19 uptake of HCQ use, namely a shortage manifestations. cases with RMDs. These initiatives will of drug supply with grave consequences Poorer outcomes to Covid-19 seem to provide short-­term clinical advisory infor- for that group of people with RMDs particularly occur in older patients and mation and will serve the purpose of long-­ dependent on it for their maintained well-­ those with comorbidities, for example, term vigilance across the age groups as being. As such, there is an urgent need chronic obstuctive pulmonary disease we anticipate the psychological impacts, for increased manufacture and protected (COPD), coronary heart disease (CHD), immune deregulatory potential and also supply of HCQ to meet the needs of the perhaps even altered comorbidities that RMD population. may ensue post-­Covid-19 in people with The immune stimulation apparent in Correspondence to Professor Iain B McInnes, underlying RMDs. some patients has led many to consider or Institute of Infection, Immunity and Inflammation, University of Glasgow, Glasgow G12 8QQ, UK; A second topical and rapidly evolving apply the use of cytokine inhibiting agents, iain.​ ​mcinnes@glasgow.​ ​ac.​uk discussion concerns the role or otherwise especially tocilizumab and sarilumab, that

McInnes IB. Ann Rheum Dis May 2020 Vol 79 No 5 551 Editorial inhibit interleukin-6 receptor (IL-6R)­ application. Despite the pandemic, we doi:10.1136/annrheumdis-2020-217494 and have found utility in hemophago- should try to adhere as far as possible to cytic lymphohistiocytosis (HLH) and agreed methodologies and rigour. That References CAR-­T cell cytokine storm syndrome. said, all current approaches are hampered 1 Zhou F, Yu T, Du R, et al. Clinical course and risk factors by a lack of data—those will however for mortality of adult inpatients with COVID-19 in An increasing range of immune modi- Wuhan, China: a retrospective cohort study. Lancet fiers are now being considered for inter- emerge rapidly from initiatives such as 2020;395:1054–62. vention including interleukin-6 receptor those described above and behove our 2 Monti S, Balduzzi S, Delvino P, et al. Clinical course of inhibitor (IL-6Ri),­ interleukin-6 inhibitor community to respond quickly and effi- COVID-19 in a series of patients with chronic arthritis (IL-­6i), granulocyte-­macrophage colony-­ ciently to place the evidence base behind treated with immunosuppressive targeted therapies. our recommendations, that in turn must Ann Rheum Dis 2020;79:667–8. stimulating factor (GM-CSF­ i), Janus 3 Chen G, Wu D, Guo W, et al. Clinical and immunologic Kinase inhibitor (JAKi), tumour necrosis be dynamic and agile to offer the best features in severe and moderate coronavirus disease factor inhibitor (TNFi) and others. These outcomes for our patients. These are chal- 2019. J Clin Invest 2020;137244:1–3. agents are not without a significant adverse lenging times indeed but also rich with 4 Wu J, Li W, Shi X, et al. Early antiviral treatment opportunity for a rheumatology commu- contributes to alleviate the severity and improve event profile, and understandable unease the prognosis of patients with novel coronavirus nity response that is creative, collegiate is developing in the critical care commu- disease (COVID-19). J Intern Med 2020. doi:10.1111/ nity as to their unregulated application, and compassionate. joim.13063. [Epub ahead of print: 27 Mar 2020]. although use is compassionate and often 5 Cao B, Wang Y, Wen D, et al. A trial of Lopinavir-­ Handling editor Josef S Smolen Ritonavir in adults hospitalized with severe Covid-19. in extremis. Rheumatologists are expert in N Engl J Med 2020. doi:10.1056/NEJMoa2001282. Funding The authors have not declared a specific the use of these agents; we should be to [Epub ahead of print: 18 Mar 2020]. grant for this research from any funding agency in the 6 Mehta P, McAuley DF, Brown M, et al. Hlh across the fore in advising around their applica- public, commercial or not-­for-­profit sectors. tion noting risks and benefits are not yet speciality collaboration, UK. COVID-19: consider Competing interests None declared. cytokine storm syndromes and immunosuppression. clear and should not be taken for granted Lancet 2020;395:1033–4. in Covid-19. It is preferable, but not Patient and public involvement Patients and/or the public were not involved in the design, or conduct, 7 Gautret P, Lagier J-­C, Parola P, et al. always achievable, that these agents are or reporting, or dissemination plans of this research. Hydroxychloroquine and azithromycin as a treatment explored in the context of clinical trials of COVID-19: results of an open-­label non-­randomized Patient consent for publication Not required. though the compassionate use and obser- clinical trial. Int J Antimicrob Agents 2020;20:105949. Provenance and peer review Commissioned; 8 Spinelli FR, Ceccarelli F, Di Franco M, et al. To consider vational cohorts already published offer internally peer reviewed. or not antimalarials as a prophylactic intervention in useful insights. the SARS-­CoV-2 (Covid-19) pandemic. Ann Rheum Dis © Author(s) (or their employer(s)) 2020. No commercial 2020;79:666–7. Finally, what are we now to do in terms re-­use. See rights and permissions. Published by BMJ. of treatment recommendations in the 9 AHJ K, Sparks JA, Liew JW, et al. COVID-19 Global Rheumatology Alliance . A Rush to Judgment? era of Covid-19? Ideally learned bodies, Rapid Reporting and Dissemination of Results including European League Against Rheu- and Its Consequences Regarding the Use of matism (EULAR), American College of To cite McInnes IB. Ann Rheum Dis 2020;79:551–552. Hydroxychloroquine for COVID-19. Ann Intern Med Rheumatology (ACR) and Asia Pacific 2020. Received 3 April 2020 10 Guastalegname M, Vallone A. Could chloroquine / League Against Rheumatism (APLAR), Revised 3 April 2020 hydroxychloroquine be harmful in coronavirus disease will offer consistent advice with sufficient Accepted 3 April 2020 2019 (COVID-19) treatment? Clin Infect Dis 2020:pii: regional flavour as to allow pragmatic Ann Rheum Dis 2020;79:551–552. ciaa321.

552 McInnes IB. Ann Rheum Dis May 2020 Vol 79 No 5 Viewpoint Disruptive innovation in rheumatology: new networks of global public–private partnerships are needed to take advantage of scientific progress Tom WJ Huizinga ‍ ‍ ,1 V Michael Holers,2 Jennifer Anolik,3 Michael B Brenner,4 Christopher Dominic Buckley,5 Vivian Bykerk,6,7 Sean E Connolly,8 Kevin D Deane,9 Jianping Guo ‍ ‍ ,10 Martin Hodge,11 Steve Hoffmann,12 Frank Nestle,13 Costantino Pitzalis ‍ ‍ ,14 Soumya Raychaudhuri,15,16 Kazuhiko Yamamoto ‍ ‍ ,17 Zhanguo Li,10 Lars Klareskog ‍ ‍ 18

Handling editor Josef S Patterns of innovation can be sustained (contin- partners) funded project designated the ‘AMP for Smolen uous) or rapid, sometimes even ‘disruptive’. A major RA and systemic lupus erythematosus (SLE)’. For numbered affiliations see difference is that in order to support disruptive inno- To bring these groups together, a meeting desig- end of article. vation the support networks and its infrastructure nated the 11th International Forum on Rheuma- often need to be changed dynamically to accommo- toid Arthritis was held from 25–27 September Correspondence to date a rapidly evolving landscape to establish the 2019 in Washington DC. The primary intention Professor Tom WJ Huizinga, disruptive approach. For example, introducing elec- was to facilitate interactions among academic as Rheumatology, Leiden University tric cars disrupts the support network for gasoline well as industry-based­ scientists and clinicians from Medical Center, Leiden, Netherlands; cars (network of gas and service stations), and at Europe, the USA and also East Asia. t.​ ​w.j.​ ​huizinga@​lumc.​nl the same time requires an entirely new system of This collaborative meeting was supported by charging stations. Such disruptions occur in science, NIH and its director Dr. Francis Collins, as well TWH and VMH are joint first and a wonderful example was the creation of mono- as by the pharmaceutical industry stakeholders in authors. clonal antibody technologies. The discovery of the AMP and the IMI-­RTCure project. Received 19 December 2019 principle for production of monoclonal antibodies The RTCure programme is focused on the under- Revised 25 February 2020 by César Milstein and Georg Köhler fueled a rapid standing of the longitudinal course of the disease, Accepted 26 February 2020 adoption of new antibody-­based technologies in all whereas the AMP is focused on better under- Published Online First areas of medicine. This transformation was strongly standing of the contribution of various cell types 5 March 2020 supported by the open workshops that catalogued to the local processes in the inflamed joints. Thus, antibodies from many laboratories into distinct several of the RTCure presentations focused on the ‘clusters of differentiation’, thereby providing a understanding of the natural history of RA, where new ‘support network’ for the global use of well a prolonged RA-related­ autoantibody-­positive validated and standardised monoclonal antibodies. period is present prior to the development of This new support network helped the pharmaceu- arthritis2 3. Subjects who are in this period of time tical industry transition from a major focus on small are operationally defined through the presence of chemical molecules (screened for effects in vitro) to symptoms such as arthralgia as well as predictive a new targeted approach, first with recombinant biomarkers4–7 and one study reported that develop- proteins, later with monoclonal antibodies with the ment of arthritis was delayed after treatment with introduction of anti-tumour­ necrosis factor (TNF) rituximab.8 With great enthusiasm, the ongoing antibodies as an example. prevention trials in the USA and Europe were During the last two decades, in both the USA reported.8 In the USA, the Stop-RA­ trial and in the and Europe, the need to develop public–private UK the Arthritis prevention in the pre-­clinical phase partnership support networks has been recognised of RA (APPRIPRA) trial randomise people with as a means to accelerate innovation and enable anti-­citrullinated peptide antibodies (ACPA) anti- translation of the rapidly expanding cellular and bodies to an intervention (Hydroxychloroquine— molecular understanding of disease pathogenesis USA, Abatacept—UK) or placebo for 1 year to test into the development of new therapeutic agents. the hypothesis that fewer RA will develop in the In Europe, the Innovative Medicine Initiative (IMI) intervention arm. In the Netherlands, patients with was formed to enhance public–private partnerships, clinical suspect arthralgia and a positive MRI are and in the USA the Accelerating Medicines Partner- randomised to either one dose of prednisone and ship (AMP).1 Identifying, validating and testing new 1 year methotrexate (MTX) versus placebo (Treat-­ © Author(s) (or their employer(s)) 2020. No targets based on enhanced understanding are at the Earlier study) and in Sweden a study (EudraCT-­nr commercial re-­use. See rights core of both AMP and IMI. To facilitate continuous 2019-002673-62) with bifosfonates in one arm and and permissions. Published innovation, multi-­institutional collaborations have placebo in the other is performed with the same by BMJ. started on projects such as the European Union hypothesis. To cite: Huizinga TWJ, funded project on tolerance in rheumatoid arthritis The focus of AMP is to take an unbiased approach Holers VM, Anolik J, (RA; Rheuma Tolerance for Cure (RTCure)) and focused on single-­cell analysis to define the cell and et al. Ann Rheum Dis the National Institutes of Health (NIH)/Founda- molecular basis of synovitis.9 Therefore, tissues 2020;79:553–555. tion for the NIH (representing industry and other and cells from patients with RA with synovitis are

Huizinga TWJ, et al. Ann Rheum Dis 2020;79:553–555. doi:10.1136/annrheumdis-2019-216846 553 Viewpoint

that assessments of cardiovascular risk and their management Box 1 The newly identified cell types: on our way to a have changed the treatment landscape and patient expectations new periodic table for celltypes in cardiovascular diseases. This necessitates a new worldwide support network that includes the development and a global 1. Expanded cell populations in rheumatoid arthritis as compared harmonisation of RA preclinical/at-­risk classification criteria, as with osteoarthritis synovium hi well as the establishment of multiple cohorts of individuals at a. THY1(CD90) +HLA-DRA­ sublining fibroblasts (interleukin 6 risk for RA in which trials can be performed. A similar challenge (IL-6) production) remains for efforts aimed at treatments that induce tolerance b. IL1B+proinflammatory monocytes for people with early RA, as exemplified from ongoing12 studies c. ITGAX+TBX21+autoimmune-­associated B cells in RTCure. These studies, in an identical fashion as occurred d. PDCD1 +peripheral helper T (T ) cells 13 PH in prevention approaches for cardiovascular disease, need e. Follicular helper T (T ) cells FH support networks with well-standardised­ immune-­surveillance 2.New cell subsets with functional implications assays and well-defined­ clinical endpoints, with supporting a. CD8 +T cell subsets by differential GZMK+, GZMB+ and GNLY laboratory-­based assays to determine the cellular basis for toler- +expression ance in RA. A joint effort in this area from RTCure and other b. Fibroblasts separated into multiple populations, including international groups contributing cohorts and the development FAPα+THY1+sublining immune effector fibroblasts (mediate of adaptive immunity cell-based­ assays, using novel single-cell­ synovial inflammation) and synovial lining FAPα+THY1− technologies and bioinformatics from the AMP consortium, destructive fibroblasts (mediate bone loss) will form the basis for the establishment of this highly needed support network. It is important that the RTCure pharma-academia­ efforts are captured using ultrasound-­guided biopsies. Using quantitative coordinated with USA based, as well as Asian and other interna- histological and bulk mRNA analyses as well as single-cell­ tech- tional groups. To accomplish that task, criteria must be devel- nologies, new insights were presented on the diversity of stromal oped that are formally evaluated and internationally approved cells (synovial fibroblasts and macrophages) as well as leucocyte by the relevant professional and regulatory organisations. An populations including T cell subpopulations. Other data empha- additional point for harmonisation, as well as agreement by sised the capacity of synovial fibroblasts to interact with other regulatory bodies, is outcome criteria for prevention and early non-­lymphoid cells of the joint, such as endothelial cells. This intervention trials. New intervention studies will benefit from creates a situation where chronicity of synovial inflammation harmonisation between the currently ongoing trials which may may become partly independent of adaptive immunity, allowing benefit from already existing organisations such as Outcome the development of therapies that target these mechanisms. Measures in Rheumatology (OMERACT) that are focused on Presentations from the Asian investigators highlighted the validating outcome measures. opportunity to exploit for discovery genetic differences in as drivers in disease phenotypes in Asian populations. Important updates on each of these topics were presented, ​Taking on big challenges in addition to making incremental summaries of which are beyond the scope of this article. We gains would like to emphasise two simultaneously occurring disruptive Finally, a strongly stated theme with regard to future studies is elements that will need public–private collaborative ventures. to take on ‘big’ problems, that will require collaborative work between different groups and scientific communities, and as ​Findings derived from single-cell studies emphasised in this viewpoint, new support structures for work Single-­cell technologies define subpopulations that were previ- towards the goals of prevention and cure of RA. ously thought to be homogeneous cell populations 9–11 and Although the exact definition of such big issues is subjec- tive, especially if one envisions international efforts beyond the box 1. This realisation creates the opportunity to identify new 14 targets, a finding that is relevant for the pharmaceutical industry highly successful collaborations focused on the genetics of RA, as opportunities for drug development. At the same time, repli- one attractive possibility is to focus on the cellular, humoral cation on tissues from different centres is required. This necessi- and genetic bases for the transformation from systemic autoim- tates a new support network to make fast and reliable progress. munity to autoimmune-associated­ joint inflammation and joint The delivery of the Human Cell Atlas (an international consor- destruction. Tackling this problem will involve efforts to define tium similar to the Human Genome Project) aims to enumerate initial cellular and humoral targets in the joints, including bone, all the individual cell types in each organ of the human body; cartilage and tendons. Also required is a better understanding AMP is providing an atlas of pathological synovial cells and cell of the relative roles in the synovium of local cellular versus states. This paves the road for a cellular basis to drive a new humoral factors and the relative roles of leucocytes and stromal taxonomy of disease. AMP seems to provide the new periodic cells, respectively. Accomplishing this goal will require following table of cell types, where the challenge will be what the funda- at-­risk populations, including temporal sampling and imaging of mental rules for cells to work together are, almost like the rules the synovium through to the development of classifiable disease, that determine how elements of the periodic table can make and then working to reverse synovitis and develop curative approaches. molecules like H2O.

​Progress in studies focused on prevention ​Next steps With at least four large prevention studies in Europe/USA and To shape a support network that is needed for the disruptive with the current exploration of regulatory pathways for preven- changes, we propose the establishment of a working group tive studies prior to the onset of arthritis by pharma and academic inclusive of academics, regional rheumatology societies, industry groups, the road to preventive interventions in RA is paved. This representatives, patients and funding agencies to design and approach has the potential to be disruptive in the same way implement an International RA Collaborative Network. From

554 Huizinga TWJ, et al. Ann Rheum Dis 2020;79:553–555. doi:10.1136/annrheumdis-2019-216846 Viewpoint that working group, suggestions for the design of such a network, Twitter Soumya Raychaudhuri @soumya_boston the priorities for funding and implementation of precompetitive Contributors Written by MVH and TWJH. Intellectual and textual contributions studies, as well as development of a data portal contributed to from all others. by all groups, could emerge in an organic manner. Also partic- Funding The authors have not declared a specific grant for this research from any ipation of regulatory agencies will be needed to enable clinical funding agency in the public, commercial or not-­for-­profit sectors. studies on prevention to result in approved and reimbursed Competing interests None declared. prevention in clinical practice. Future closer transatlantic and Patient and public involvement Patients and/or the public were involved in the global collaborations will capitalise on the legacy of the existing design, or conduct, or reporting, or dissemination plans of this research. Refer to the initiatives, for example, a collaborative IMI-AMP­ project as a Methods section for further details. follow-­up project of current IMI and AMP initiatives. In fact, Patient consent for publication Not required. many industry partners participate in both consortia, and thus Provenance and peer review Not commissioned; externally peer reviewed. the legal framework for such collaborative efforts can follow the current structures which allow for very effective 1 to 1 collabo- ORCID iDs rations on a deep data level, while the overall programme can be Tom WJ Huizinga http://orcid.​ ​org/0000-​ ​0001-​7033-​7520 Jianping Guo http://orcid.​ ​org/0000-​ ​0002-​5031-​3510 constructed to generate shared global data Costantino Pitzalis http://orcid.​ ​org/0000-​ ​0003-​1326-​5051 The primary goals of the International RA Collaborative Kazuhiko Yamamoto http://orcid.​ ​org/0000-​ ​0001-​9037-​3625 Network would be to build on existing public–private part- Lars Klareskog http://orcid.​ ​org/0000-​ ​0001-​9601-​6186 nerships such as RTCure and AMP and shape a new support network for preventive and curative studies, and foster collab- References orations between individuals and communities prepared to 1 Dolgin E. Massive NIH-industry­ project opens portals to target validation. Nat Rev develop and use these networks. Drug Discov 2019. doi:10.1038/d41573-019-00033-8. [Epub ahead of print: 27 Feb 2019]. 2 Malmström V, Catrina AI, Klareskog L. The immunopathogenesis of seropositive Author affiliations rheumatoid arthritis: from triggering to targeting. Nat Rev Immunol 2017;17:60–75. 1Rheumatology, Leiden University Medical Center, Leiden, Netherlands 2 3 Holers VM, Demoruelle MK, Kuhn KA, et al. Rheumatoid arthritis and the Rheumatology, University of Colorado School of Medicine, Aurora, Denver, Colorado, mucosal origins hypothesis: protection turns to destruction. Nat Rev Rheumatol USA 3 2018;14:542–57. Division of Allergy, Immunology, and Rheumatology, Department of Medicine, 4 Gerlag DM, Raza K, van Baarsen LGM, et al. EULAR recommendations for terminology University of Rochester School of Medicine, Rochester, New York, USA 4 and research in individuals at risk of rheumatoid arthritis: report from the study Group Division of Rheumatology, Immunology, and Allergy, Brigham and Women’s for risk factors for rheumatoid arthritis. Ann Rheum Dis 2012;71:638–41. Hospital, Boston, Massachusetts, USA 5 Demoruelle MK, Harrall KK, Ho L, et al. Anti-­citrullinated protein antibodies are 5Department of Rheumatology, University of Birmingham, Birmingham, UK 6 associated with neutrophil extracellular traps in the sputum in relatives of rheumatoid Rheumatology, The Hospital for Special Surgery, New York, New York, USA arthritis patients. Arthritis Rheumatol 2017;69:1165–75. 7Medicine, Mount Sinai Hospital, Toronto, Ontario, Canada 8 6 Rombouts Y, Ewing E, van de Stadt LA, et al. Anti-­citrullinated protein antibodies Department of Immunology and Inflammation, Bristol Myers Squibb Co. Research acquire a pro-­inflammatory Fc glycosylation phenotype prior to the onset of and Development, Princeton, New Jersey, USA 9 rheumatoid arthritis. Ann Rheum Dis 2015;74:234–41. Division of Rheumatology, Department of Medicine, University of Colorado Anschutz 7 Mankia K, Cheng Z, Do T, et al. Prevalence of periodontal disease and Medical Campus, Aurora, Colorado, USA 10 periodontopathic bacteria in anti-­cyclic citrullinated protein antibody-­positive at-­risk Department of Rheumatology and Immunology, Peking University People’s Hospital, adults without arthritis. JAMA Netw Open 2019;2:e195394. Beijing, China 11 8 Gerlag DM, Safy M, Maijer KI, et al. Effects of B-cell­ directed therapy on the preclinical Department of Rheumatology, Pfizer Global Pharmaceuticals, New York, New York, stage of rheumatoid arthritis: the PRAIRI study. Ann Rheum Dis 2019;78: :179–85. USA 9 Zhang F, Wei K, Slowikowski K, et al. Defining inflammatory cell states in rheumatoid 12Foundation for the National Institutes of Health, Bethesda, Maryland, USA 13 arthritis joint synovial tissues by integrating single-cell­ transcriptomics and mass Immunology and Inflammation Therapeutic Research Area Sanofi US, Sanofi cytometry. Nat Immunol 2019;20:928–42. Genzyme, Cambridge, Massachusetts, USA 14 10 Croft AP, Campos J, Jansen K, et al. Distinct fibroblast subsets drive inflammation and Experimental Medicine and Rheumatology, William Harvey Research Institute, damage in arthritis. Nature 2019;570: :246–51. London, UK 15 11 Mizoguchi F, Slowikowski K, Wei K, et al. Functionally distinct disease-­associated Arthritis Research UK Epidemiology Unit, The University of Manchester, Manchester, fibroblast subsets in rheumatoid arthritis. Nat Commun 2018;9:789. UK 16 12 Benham H, Nel HJ, Law SC, et al. Citrullinated peptide dendritic cell immunotherapy Program in Medical and Population Genetics, Broad Institute of MIT and Harvard, in HLA risk genotype-­positive rheumatoid arthritis patients. Sci Transl Med Cambridge, Massachusetts, USA 17 2015;7:290ra87. Department of Allergy and Rheumatology, Graduate School of Medicine, The 13 Jackson R, Lawes CMM, Bennett DA, et al. Treatment with drugs to lower blood University of Tokyo, Tokyo, Japan 18 pressure and blood cholesterol based on an individual’s absolute cardiovascular risk. Rheumatology, Karolinska Institutet, Stockholm, Sweden Lancet 2005;365:434–41. Correction notice This article has been corrected since it published Online First. 14 Al-­Laith M, Jasenecova M, Abraham S, et al. Arthritis prevention in the pre-­clinical The second author’s name has been corrected and the joint first author statement phase of RA with abatacept (the APIPPRA study): a multi-­centre, randomised, double-­ added. blind, parallel-­group, placebo-­controlled clinical trial protocol. Trials 2019;20:429.

Huizinga TWJ, et al. Ann Rheum Dis 2020;79:553–555. doi:10.1136/annrheumdis-2019-216846 555 Rheumatoid arthritis

Epidemiological science Two-­year cost-­effectiveness of different COBRA-­like intensive remission induction schemes in early rheumatoid arthritis: a piggyback study on the pragmatic randomised controlled CareRA trial Sofia Pazmino ‍ ‍ ,1 Annelies Boonen ‍ ‍ ,2,3 Veerle Stouten,1 Diederik De Cock,1 Johan Joly,4 Kristien Van der Elst,4 Rene Westhovens,1,4 Patrick Verschueren1,4

Handling editor Josef S Abstract Key messages Smolen Objectives To evaluate the cost-­effectiveness of treat-­ to-­target strategies among recently diagnosed patients ►► Additional material is What is already known about this subject? published online only. To view with rheumatoid arthritis (RA) using methotrexate (MTX) ►► While treat-­to-­target strategies are cost-­ please visit the journal online and a step-­down glucocorticoid (GC) scheme (COBRA effective in rheumatoid arthritis (RA), initial (http://dx.​ ​doi.org/​ ​10.​1136/​ Slim) compared with (1) this combination with either biological disease-­modifying antirheumatic drug annrheumdis-2019-​ ​216874). sulphasalazine (COBRA Classic) or leflunomide (COBRA (bDMARD) therapy is not cost-­effective and 1 Avant-­Garde) in high-­risk patients and (2) MTX without Department of Development conventional synthetic DMARDs (csDMARDs) GCs (Tight-­Step-­Up, TSU) in low-­risk patients. and Regeneration, Skeletal are to be preferred over bDMARDs or targeted Biology and Engineering Methods The incremental cost-­utility was calculated synthetic DMARDs as first-­line treatment. Research Centre, KU Leuven, from a healthcare perspective in the intention-­to-­treat Leuven, Flanders, Belgium 2 population (n=379) of the 2-­year open-­label pragmatic Department of Internal What does this study add? Medicine, Division of randomised controlled Care in early RA trial. Healthcare ►► In patients with classical factors of poor Rheumatology, Maastricht costs were collected prospectively through electronic prognosis, csDMARD combination therapy with University Medical Centre, trial records. Quality-­adjusted life years (QALYs) were step-­down glucocorticoids (GCs) was not cost-­ Maastricht, The Netherlands estimated using mapping algorithms for EuroQoL-5 3Care and Public Health effective compared with methotrexate (MTX) Dimension. Multiple imputation was used to handle Research Institute (CAPHRI), monotherapy also with step-­down GCs within a missing data and bootstrapping to calculate CIs. Maastricht University, remission induction strategy. Maastricht, The Netherlands Robustness was tested with biological disease-­modifying 4 ► For patients without classical factors of poor Department of Rheumatology, antirheumatic drugs at biosimilar prices. ► University Hospitals Leuven, prognosis, MTX plus a short-­term course of Results In the high-­risk group, Classic (∆k€1.464, Leuven, Belgium GCs was clearly more cost-­effective than the 95% CI −0.198 to 3.127) and Avant-­Garde (∆k€0.636, traditional MTX only. Correspondence to 95% CI −0.987 to 2.258) were more expensive compared with Slim and QALYs were slightly worse for Dr Sofia Pazmino, Department How might this impact on clinical practice or of Development and Classic (∆−0.002, 95% CI −0.086 to 0.082) and Avant-­ future developments? Regeneration, Skeletal Biology Garde (∆−0.009, 95% CI −0.102 to 0.084). This resulted and Engineering Research ► Initiating a step-­down GC bridging combined in the domination of Classic and Avant-­Garde by Slim. ► Center, KU Leuven, Leuven with MTX in newly diagnosed patients with In the low-­risk group, Slim was cheaper (∆k€−0.617, B-3000, Belgium; RA could be key to delay or even avoid use sofia.​ ​pazmino@kuleuven.​ ​be 95% CI −2.799 to 1.566) and QALYs were higher of second line expensive medication such (∆0.141, 95% CI 0.008 to 0.274) compared with TSU, as bDMARDs in both high-­risk and low-­risk Received 23 December 2019 indicating Slim dominated. Results were robust against Revised 17 February 2020 patients. Accepted 1 March 2020 the price of biosimilars. Published Online First Conclusions The combination of MTX with a GC 2 April 2020 bridging scheme is less expensive with comparable health utility than more intensive step-­down combination (csDMARDs) should be started as soon as possible after diagnosis, preferably methotrexate (MTX) strategies or a conventional step-­up approach 2 years 2 after initial treatment. and a short course of glucocorticoids (GCs). The Trial registration number NCT01172639. ideal dosing scheme and the value of classical risk factors as therapeutic prognostics (theragnostic) are still a matter of debate. Over the years, several trials have demonstrated increased efficacy of Introduction initial csDMARD combinations over monotherapy, © Author(s) (or their employer(s)) 2020. No Insufficient control of disease activity in rheu- but evidence remains scarce for the superiority of commercial re-­use. See rights matoid arthritis (RA) can lead to persistent combining csDMARDs within strategies including and permissions. Published pain, joint destruction, functional impairment a step-­down-­bridge GC scheme. It has not been by BMJ. and thus reduced health-­related quality of life elucidated whether effectiveness of such strategies 3–5 To cite: Pazmino S, (QoL). Evidence suggests that controlling disease differs depending on prognostic risk profile. Boonen A, Stouten V, activity depends on early, intensive and to-target­ Care in early RA (CareRA), a randomised treat-­ et al. Ann Rheum Dis medical therapy.1 Therapy with conventional to-­target strategy trial showed no superiority of 2020;79:556–565. synthetic disease-­modifying antirheumatic drugs combinations of csDMARDs and bridging GCs

556 Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 Rheumatoid arthritis

(COBRA Classic or COBRA Avant-Garde)­ over MTX-only­ LEF was added in the non-combination­ arms (COBRA Slim and with a moderate-­dose step-­down GC bridging scheme (COBRA TSU). Further treatment changes could include bDMARD initi- Slim), but COBRA Slim had a more favourable safety profile.6–8 ation according to Belgian reimbursement rules.10 Details on In addition, COBRA Slim showed a better initial response and patient eligibility criteria, randomisation process, study design overall disease control than the traditional Tight-­Step-­Up (TSU) and treatment intensifications have been published.6–9 Patients approach, starting MTX without oral GCs, in so-called­ low-risk­ were assessed at screening, baseline and then followed up at patients.9 week 4, 8, 16, 28, 40, 52, 65, 78, 91 and 104. Optional visits, From a health economical point of view, it is important to if clinically required, could be performed. Physicians filled out investigate to what extent different initial treatment choices and at every patient visit the electronic case record form (eCRF), consecutive treatment steps within a treat-­to-­target strategy lead comprising American College of Rheumatology core measures,11 to differences in longer-term­ costs. Therefore, we conducted a medications and adverse events (AEs). full economic evaluation of the CareRA trial to study the cost per quality-­adjusted life years (QALYs). Outcomes Health utility and QALYs Patients and methods A health utility represents the preference of value attributed to Using the 2-years­ data from the open label CareRA pragmatic a health state. It is expressed on a continuous scale from zero treat-­to-­target randomised controlled trial (RCT) (EudraCT-­ (equalling death) to one (full health). Scores can also be below number: 2008-007225-39), the incremental cost–utility and 0 (states worse than death). By multiplying the utility value with incremental net monetary benefit (iNMB) from a payer’s years of survival, QALYs are calculated. For this study, QALYs perspective were estimated. In the high-risk­ group, both COBRA were determined as the time-weighted­ average of reconstructed Classic and COBRA Avant-­Garde were compared with COBRA EuroQoL-5 Dimension (EQ-5D)­ values for each visit in the total Slim. In the low-risk­ group, COBRA Slim was compared with follow-­up (area under the curve). EQ-5D­ health utilities were TSU. reconstructed based on Health Assessment Questionnaire, age, pain on a Visual Analogue Scale and gender, using the validated UK algorithm of Hernández Alava et al.12–14 For patients expe- CareRA clinical trial riencing AEs that were either severe or lasted for more than 3 In total, 400 DMARD naïve patients with recently diagnosed RA months, the QALY was adjusted by a disutility, accounting for (≤1 year) were assessed for eligibility between January 2009 and the duration of the AE. Based on an extensive literature search, May 2013 and 379 were included. disutilities per AE were constructed (online supplementary table Patients were stratified into a high-­risk or low-­risk group 15–24 7 8 2). If no specific disutility was found, a general moderate-­ based on an algorithm with prognostic factors (erosions, rheu- intensity treatment associated disutility (0.002) was applied.25 matoid factor (RF) or anticitrullinated cyclic peptide (anti-­CCP) positivity and baseline Disease Activity Score in 28 joints with C-reactive­ protein (DAS28CRP)>3.2) and randomised into four Resource utilisation and direct healthcare costs different treat-­to target schemes. Healthcare costs in the economic analysis were rheumatology High-­risk patients were randomised to one of the following visits, RA-related­ medication (cs- and bDMARDs, GCs and initial treatment schemes: all recorded analgesics including paracetamol, non-­steroidals, ►► COBRA Classic: 15 mg MTX weekly, 2 g sulfasalazine daily tramadol and opioids), hospital admissions, laboratory and and a weekly step-­down scheme of oral prednisone starting radiographs occurring during the 2-­year trial. at 60 mg daily and tapering through 40-25-20-15-10-­mg Costs for rheumatology visits included scheduled and addi- tional visits. Cost per visit was retrieved from the National daily to a maintenance dose of 7.5 mg, with further tapering 26 from week 28, before completely stopping at week 34. Institute of Health and Disability Insurance (RIZIV) tariffs. ►► COBRA Avant-Garde:­ 15 mg MTX weekly, 10 mg lefluno- RA-related­ medication costs were calculated from the eCRF mide (LEF) daily and a weekly step-down­ scheme of oral reported medication name, dose, intake duration and frequency, then valued according to the Belgian Centre for Pharmacothera- prednisone starting at 30 mg daily and tapering through 27 20–12.5-10-7.5 mg daily to a maintenance dose of 5 mg, peutic Information. Hospitalisation costs were calculated from with further tapering from week 28, before completely stop- AEs requiring hospitalisation and were price weighted depending ping at week 34. on coded diagnosis (International Classification of Diseases, ►► COBRA Slim: 15 mg MTX weekly and a weekly step-­down Ninth Revision, Clinical Modification), physician registered scheme of oral prednisone starting at 30 mg daily and severity (mild, moderate, severe) and number of inpatient days with INAMI-­RIZIV tariffs (​tct.​fgov.​be).28 Supplemental costs for tapering through 20–12.5-10-7.5 mg daily to a maintenance 29 dose of 5 mg, with further tapering from week 28, before laboratory and radiographs were also incorporated. All prices completely stopping at week 34. were converted to 2018 euros using the general Belgian health Low-­risk patients were randomised to either: index rate (statbel.​ ​fgov.be).​ Total costs per resource were calcu- lated by multiplying the number of resources by the cost unit ►► COBRA Slim. 26–29 ►► TSU: 15 mg MTX weekly, no oral GCs allowed. price extracted from Belgian national websites. Total costs In all treatment arms, low disease activity (DAS28CRP ≤3.2) per patient were obtained by summing costs of all resources. No discounting was considered due to the study’s short follow-up­ was used for steering treatment adaptations. The first trial year 30 had prespecified (per-protocol)­ treatment adaptations. During period (2 years). the second year, adaptations were left at the discretion of the treating rheumatologist. An increase in the weekly MTX dose to Cost-effectiveness analyses 20 mg was the first adjustment in all treatment schemes. Next, This piggyback study, an economic evaluation alongside a clin- the dose of the other csDMARD was increased in the combina- ical trial, follows the superiority design of CareRA and was tion arms (COBRA Classic and COBRA Avant-Garde)­ or 10 mg performed on the intention-­to-­treat (ITT) population.

Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 557 Rheumatoid arthritis

Differences in costs and QALYs between COBRA Classic trees) by chained equations.34 Missingness in clinical variables and COBRA Avant-­Garde compared with COBRA Slim for used to estimate utility, disease activity per time point, and total the high-­risk group and between COBRA Slim and TSU for costs were imputed. Besides the incomplete variables, treatment the low-­risk group were analysed over 2 years. An incremental strategy, centre of recruitment, age, gender, presence of comor- cost-­effectiveness ratio (ICER) was calculated by dividing the bidities, AEs, RF, anti-­CCP, erosions at baseline and trial comple- cost difference by the QALY difference per pair of treatment tion were included as predictors in the matrix. Fifteen imputed schemes. The uncertainty analysis in the estimation of the ICER datasets were created and analysed separately. Results of the 15 was plotted on cost-­effectiveness planes (via non-parametric­ analyses were pooled using Rubin’s rules.35 36 bootstrapping with 25 000 iterations of random sampling with For comparisons in complete cases, non-­parametric Mann-­ replacement). Cost differences were depicted on the y-axis­ and Whitney U or Kruskal-­Wallis and X2 bootstrapped-­corrected QALY differences on the x-axis.­ were used when appropriate. The incremental net monetary benefit (iNMB) of each compar- All analyses were performed with R V.3.6.1. ison was calculated as iNMB=[incremental benefit * willingness to pay (WTP)]-incremental cost. The impact on the iNMB for varying thresholds of WTP (€0–€150 000) for one QALY gain Results was calculated. This reflects absolute economic gain (positive) Patients or loss (negative), given how much society is willing to pay per Of the 379 patients included in the CareRA trial, 289 patients QALY gained (ƛ). The World Health Organization proposes that were stratified to the high-­risk (COBRA Classic n=98, COBRA it is reasonable to pay for an intervention that provides one addi- Avant-­Garde n=93, COBRA Slim n=98) and 90 to the low-risk­ tional year of healthy life per capita.31 An intervention is consid- group (COBRA Slim n=43, TSU n=47). Patient characteristics ered cost-­effective when the cost for one QALY gained falls in each treatment arm are presented in table 1. Good retention below three times the gross domestic product per capita (GDP) rates of up to 89% were observed (online supplementary figure 32 and highly cost-­effective when below the GDP. In Belgium, no 1). Missingness in the clinical variables over 2 years ranged from 33 prespecified WTP exists, but the 2018 GDP was k€40.320. 12% to 39% per different time point and was 15% for total costs For this study, we used a WTP threshold (ƛ) of k€40 per QALY (n=328 for total cost). gained. Considering the increased use of biosimilars, the base case analysis was repeated pricing bDMARDs at the lowest price of Health outcomes a biosimilar (Benepali) in Belgium (€153.15 for 50 mg weekly as QALYs over 2 years were comparable in the high-­risk group of December 2018). (1.551, 1.544, 1.553) between COBRA Classic, COBRA Avant-­ Sustained remission was used as an alternative health outcome Garde and COBRA Slim, respectively (table 2). In the low-­risk to calculate the ICER. Sustained remission was defined as group, 2-­year-QALYs­ were higher in COBRA Slim (1.629) DAS28CRP <2.6 from week 16 to 104 at every visit. compared with TSU (1.488), resulting in an incremental gain of 0.141. Statistical analyses Sustained remission rates (table 2) were also comparable in Missing data were assumed to be missing at random and were the high-­risk group, whereas in the low-risk­ group COBRA Slim imputed with multiple imputation (classification and regression (42%) had better sustained remission rates than TSU (26%).

Table 1 Demographic and clinical characteristics at baseline per treatment scheme High-risk­ Low-risk­ COBRA Classic n=98 COBRA Avant-­Garde n=93 COBRA Slim n=98 COBRA Slim n=43 TSU n=47 Demographic variables Age, years 53 (12) 51 (13) 52 (13) 51 (14) 51 (14) Women, n (%) 64 (65) 64 (69) 63 (64) 33 (77) 38 (81) Smokers, n smoked ever (%) 56 (57) 56 (60) 58 (59) 21 (49) 18 (38) Current work n (%) 44 (45) 48 (52) 52 (53) 22 (51) 27 (57) Clinical variables Body mass index, kg/m² 26 (4) 27 (4) 27 (4) 25 (4) 27 (4) Symptom duration, weeks; median (IQR) 22 (14–44) 25 (15–51) 24 (15–39) 21 (14–35) 19 (13–33) RF positive, n (%) 78 (80) 70 (75) 82 (84) 11 (26) 11 (23) Anti-­CCP positive, n (%) 76 (78) 72 (77) 78 (80) 12 (28) 11 (23) Erosive disease, n (%) 32 (33) 32 (34) 32 (33) 1 (2) 0 (0) DAS28- ­CRP 4.67 (1.13) 4.45 (1.20) 4.55 (1.12) 4.28 (1.61) 4.30 (1.63) Pain, mm (0–100) 59 (24) 57 (24) 57 (22) 48 (31) 52 (23) Fatigue, mm (0–100) 51 (26) 49 (24) 49 (21) 39 (28) 46 (22) HAQ score (0–3) 1.10 (0.77) 0.93 (0.78) 0.92 (0.78) 0.81 (0.85) 0.85 (0.72) Health utility (−0.59 to 1) 0.47 (0.27) 0.51 (0.27) 0.53 (0.26) 0.59 (0.32) 0.55 (0.25) Data are presented as mean and SD unless otherwise specified. Symptom duration=number of weeks between onset of symptoms and start of treatment. Health utility was derived using an EQ-­5D mapping algorithm. Anti-­CCP, anti-­citrullinated cyclic peptide; CRP, C-­reactive protein; DAS28, Disease Activity Score in 28 joints; EQ-5D,­ EuroQoL-5 Dimension; HAQ, Health Assessment Questionnaire; RF, rheumatoid factor; TSU, Tight-­Step-­Up.

558 Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 Rheumatoid arthritis ­ Up. ­ risk CI)

­ Step- ­ weighted as Reference Reference Reference n=47 SU   26 (13 to 38)  1.488 (1.394 to 1.581) Mean (95% T ­ risk Low- ­ adjusted life years, time- adjusted life years, CI)

lim n=43 S RA B Dominant Dominant −0.004 (−1.718 to 1.709) 3.851 (2.688 to 5.014) 3.856 (2.479 to 5.232) 17 (−3 to 37) Dominant 42 (33 to 52) −0.617 (−2.799 to 1.566) 0.141 (0.008 to 0.274) Dominant 1.629 (1.569 to 1.688) 4.007 (2.631 to 5.383) 4.624 (2.685 to 6.562) Mean (95% CO risk patients and COBRA Slim to TSU on low- ­ risk patients and COBRA Slim to

k€ k€ % Y as outcome k€ k€ ∆ Costs vs TSU, ∆ Costs vs ICER (k€ per QALY) ICER (k€ per %SR) ∆ SR vs TSU, ∆ SR vs ICER (k€ per %SR) SR (w16-104), % SR (w16-104), ∆ Costs vs TSU, ∆ Costs vs TSU vs ∆ QALYs ICER (k€ per QALY)

QALYs Total costs, Total Total costs, Total ­ Garde to COBRA Slim in high- bDMARDs priced to biosimilar Sustained remission as outcome Base case: QAL Base case: Y or cost per per cent sustained remission; k€, thousand euros; QALY, quality- QALY, thousand euros; k€, Y or cost per cent sustained remission; CI)

lim n=98 S Reference Reference Reference RA B 4.340 (3.421 to 5.258) 30 (21 to 39) 1.553 (1.494 to 1.612) 4.622 (3.649 to 5.594)    CO Mean (95% n=93 Garde ­ Garde 000 replications from each of the 15 multiply imputed datasets.

Activity Score in 28 joints with C reactive protein <2.6 from week 16 onwards until week 104 and at every timepoint in between; TSU, Tight- TSU, until week 104 and at every timepoint in between; Activity Score in 28 joints with C reactive protein <2.6 from week 16 onwards CI)

­ risk effectiveness ratio, either for cost per QAL effectiveness ratio, ­ vant- High- RA A B

Dominated 29.113 0.511 (−0.904 to 1.926) 4.851 (3.734 to 5.968)

36.191 2 (−12 to 16) 32 (22 to 42) −0.009 (−0.102 to 0.084) Dominated 0.636 (−0.987 to 2.258) 1.544 (1.473 to 1.615) 5.257 (3.909 to 6.605) CO Mean (95% effectiveness analyses comparing COBRA Classic and COBRA Avant- ­ effectiveness analyses comparing COBRA Classic and CI)

RA Classic n=98 B

Dominated Dominated 1.170 (−0.285 to 2.625) 5.510 (4.347 to 6.673)

Dominated −1 (−14 to 12) 29 (20 to 38) −0.002 (−0.086 to 0.082) Dominated 1.464 (−0.198 to 3.127) 1.551 (1.491 to 1.611) 6.086 (4.710 to 7.462) CO Mean (95% ­ case and additional cost- year trial; SR, sustained remission measured with Disease SR, ­ year trial; modifying antirheumatic drugs; ICER, incremental cost- ICER, ­ modifying antirheumatic drugs; Y as outcome k€ k€ Results of base- biological disease-

ICER (k€ per QALY) ICER (k€ per %SR) ∆ Costs vs Slim, k€ ∆ Costs vs Slim, ICER (k€ per %SR) ∆ SR vs Slim, % ∆ SR vs Slim, SR (w16-104), % SR (w16-104), ∆ QALYs vs Slim ∆ QALYs ICER (k€ per QALY) ∆ Costs vs Slim, k€ ∆ Costs vs Slim,

QALYs Total costs, Total Total costs, Total bDMARDs priced to biosimilar Sustained remission as outcome patients (the last strategy being the comparator/reference scheme in every case) area under the curve for 2- Base case: QAL Base case: Table 2 Table Data are expressed as bootstrapped mean and 95% CIs of costs benefits from all 25 bDMARDs,

Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 559 Rheumatoid arthritis

Figure 1 Costs across treatment schemes: mean € per patient and percentage (%) of the total cost in complete cost cases (n=328).

560 Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 Rheumatoid arthritis

Healthcare use and costs cost, making it consistently beneficial across the WTP range Healthcare costs in complete cases, including costs of medica- (figure 4C). tion, were presented in figure 1. The healthcare cost of rheu- matology consultations and use of laboratory and X-rays­ was comparable across all schemes. In both high-risk­ and low-risk­ Discussion groups, differences in average cost per patient between treat- This study showed that for high-­risk early RA patients, ment strategies could be attributed mainly to bDMARD use and csDMARD combination schemes with GCs (COBRA Classic hospitalisations (online supplementary table 1). and COBRA Avant-Garde)­ were not cost-­effective or even domi- Combination arms (COBRA Classic and COBRA Avant-­ nated in the first 2 years when compared with COBRA Slim, Garde) had a higher number of patients that were started on MTX monotherapy together with a moderate-­dose step-­down bDMARDs in the entire 2 years (online supplementary table 1), GC bridging scheme, for initial remission induction within although not significant in all comparisons. a treat-­to-­target strategy. In the low-risk­ group, COBRA Slim Total healthcare costs in the ITT population were for COBRA dominated the traditional MTX monotherapy without GC in Classic (k€6.086), COBRA Avant-Garde­ (k€5.257) and COBRA patients with early RA treated-to-­ ­target. However, selecting the Slim (k€4.622) in the high-risk­ group. In the low-risk­ group, most appropriate first-­line DMARD regimen remains a complex 1 costs were for COBRA Slim (k€4.007) and TSU (k€4.624). clinical decision. We published previously that COBRA Slim treatment resulted Incremental cost per QALY in similar remission rates but less therapy-related­ AEs compared with COBRA Classic or COBRA Avant-­Garde.8 Such adverse drug Given the higher costs and lower QALYs gained, COBRA Classic events have been associated with higher costs of illness37 and also and COBRA Avant-Garde­ were dominated by COBRA Slim in reduced patient’s QoL. This piggyback, trial-based­ economic the base-case­ analysis of the high-risk­ group (table 2). In the evaluation, provides evidence that MTX with step-down­ GCs is low-­risk group, COBRA Slim also dominated TSU in view of an effective initial treatment choice for all patients with RA by the lower costs and higher number of QALYs gained (table 2). balancing the necessary treatment intensity to control the disease The sensitivity analyses considering biosimilar-­prices for all bDMARDs yielded similar results (table 2). with a favourable safety profile, resulting in an adequate QoL. Results from the 25 000 bootstrapped replications of the incre- This strategy could moreover minimise chances of an interrup- mental cost–utility ratios for each comparison are presented in tion or modification of the treatment scheme leading to more figure 2. In the high-­risk group, COBRA Classic was more costly frequent discomfort and higher costs. than COBRA Slim in 96% of the bootstrapped replications and was This study’s results were comparable to other cost-effectiveness­ dominated by Slim in 68% of these replications (figure 2A). The analyses of early RA strategy trials. In the BeSt trial, the COBRA bootstrapped uncertainty of COBRA Avant-­Garde indicated this Classic-­like strategy had a total cost of k€9.2 of which k€5.0 were direct medical costs (calculated from US dollars; exchange strategy to be more costly in 78% of replications compared with 38 COBRA Slim (figure 2B) and in 56% of replications, this strategy rate of 1:0.90), comparable to our cost of k€6.086 in COBRA was dominated. In the low-­risk group, COBRA Slim dominated Classic. In the COBRA-­light trial, using strategies comparable to TSU (figure 2C) in 71% of the bootstrapped replications. COBRA Classic and Slim, the total costs were k€9.7 and k€5.6, respectively, and differences in QALYs comparable to CareRA.39 Incremental net monetary benefit per QALY The robust comparability with previous trials reinforces our Figure 3 represents the iNMB per comparison of schemes at message that COBRA Slim seems a cost-saving­ strategy. different thresholds for WTP from €0 to k€150 per QALY. In the In patients perceived as high-­risk, the main driver for the high-­risk group, when comparing COBRA Classic and COBRA lower cost of COBRA Slim was the lower number of bDMARDs Avant-­Garde versus COBRA Slim, the iNMB estimate (black initiated in the 2 years of the CareRA trial (online supplemen- full line) was negative and remained below zero on the y-­axis tary table 1). Because COBRA Classic and COBRA Avant-­Garde (figure 3A,B) regardless of the WTP. In other words, there was no combined two csDMARDs, patients insufficiently responding economic benefit of choosing COBRA Classic or COBRA Avant-­ to these schemes were after failing to dose escalation of both Garde over COBRA Slim. In the low-risk­ group, COBRA Slim csDMARDs eligible for bDMARDs according to Belgian reim- versus TSU had a positive iNMB. COBRA Slim’s estimate never bursement criteria.10 In contrast, patients on COBRA Slim crossed the zero, indicating COBRA Slim was cost-effective­ even therapy had to first initiate and fail a second csDMARD before at a very low WTP range (figure 3C). being eligible for bDMARDs. In line with good clinical prac- tice, this approach, including if necessary different adaptation Incremental net monetary benefit per sustained remission steps depending on the initial treatment effect, delays the need Figure 4 represents the iNMB per comparison of schemes at for initiating bDMARDs, resulting in cost benefits but also different thresholds for WTP from €0 to k€150 per sustained potentially patient benefits, in terms of risk:benefit ratio.40 The remission (DAS28CRP <2.6 from week 16 to 104). The iNMB long-­term CareRA outcomes from the observational follow-up­ approach represents the monetary benefit (in €) for each extra (3 years) will provide additional insights into cost-­effectiveness percentage (%) of patients that reached sustained remission at and further bDMARD use. different WTP values. In the high-risk­ group, when comparing Despite the fact that several trials and meta-analyses­ have COBRA Classic versus COBRA Slim, the iNMB estimate (black demonstrated that efficacy outcomes improve when using full line) was negative and remained below zero on the y-­axis tumour necrosis factor inhibitor (TNFi) bDMARDs as first-line­ (figure 4A). Regardless of the WTP, there was no economic treatment41 42 they have poor cost-­effectiveness profiles.43 44 benefit. When comparing COBRA Avant-Garde­ versus COBRA One of the challenges with these earlier analyses is that prices Slim, there was an added economic benefit from k€20 onwards. from before the approval of biosimilar TNFis were used for However, the lower CI remained negative at any WTP range cost-­effectiveness calculations. To estimate the hypothetical (figure 4B). In the low-risk­ group, COBRA Slim was dominant impact of using biosimilars in CareRA, an uncertainty analysis to TSU with higher proportions of sustained remission at a lower was performed changing every bDMARD to the lowest priced

Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 561 Rheumatoid arthritis

Figure 2 Cost-­effectiveness planes of the base-­case analysis (€/QALY). ICER, incremental cost-­effectiveness ratio; QALY, quality-­adjusted life-­year; TSU, Tight-­Step-­Up; WTP, willingness to pay.

562 Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 Rheumatoid arthritis

Figure 3 Mean incremental net monetary benefit (iNMB) with 95% Figure 4 Mean incremental net monetary benefit (iNMB) with 95% CIs across different thresholds of willingness to pay (WTP) of the base-­ CIs across different thresholds of willingness to pay (WTP) of the case cost–utility analyses with quality-­adjusted life-­years (QALYs) as secondary analyses with sustained remission (DAS28CRP <2.6 from health outcome. The black line is the estimate and the dotted lines its week 16 to 104) as health outcome. The black line is the estimate and 95%CIs. the dotted lines its 95%CIs. biosimilar in Belgium at the time of this study. This analysis with less stringent inclusion and exclusion criteria, the study demonstrated robustness of the initial results. To further explore population may represent a typical day-­to-­day healthcare popu- the use of earlier bDMARD use, we initiated the CareRA lation with no artificially enhanced compliance, using strategies 2020 trial (EudraCT # 2017-004054-41) examining the cost-­ already in place in clinical practice.46–48 The protocol-­driven effectiveness of accelerated but temporary bDMARD access after treatment adaptations were limited to two logical escalations failing to MTX monotherapy with a GC bridging scheme. of csDMARDs, meaning that when a bDMARD was needed, Since pragmatic effectiveness trials seem best for economic it was left at the discretion of the treating rheumatologist, just studies, the use of data from the pragmatic CareRA study is a as in daily practice. However, this post hoc study of an RCT, major strength for this economic analysis.45 As CareRA is a prag- provided no data on indirect costs nor direct non-­medical costs. matic, treat-to-­ ­target, multicentre investigator-initiated­ RCT There might also be direct medical costs missing when it comes

Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 563 Rheumatoid arthritis to general practitioner appointments and use of paramedical or 5 den Uyl D, ter Wee M, Boers M, et al. A non-­inferiority trial of an attenuated alternative therapies. Intangible costs are complicated to account combination strategy (’COBRA-­light’) compared to the original cobra strategy: clinical results after 26 weeks. Ann Rheum Dis 2014;73:1071–8. for, yet our study provides a glimpse into them by recreating 6 Verschueren P, De Cock D, Corluy L, et al. Methotrexate in combination with the EQ-5D­ values considering pain and physical function among other DMARDs is not superior to methotrexate alone for remission induction with others. Moreover, the estimation of health utility was corrected moderate-­to-­high-­dose glucocorticoid bridging in early rheumatoid arthritis after for disutility produced by AEs. 16 weeks of treatment: the CareRA trial. Ann Rheum Dis 2015;74:27–34. 7 Verschueren P, De Cock D, Corluy L, et al. Effectiveness of methotrexate with step-­ down glucocorticoid remission induction (cobra slim) versus other intensive treatment strategies for early rheumatoid arthritis in a treat-­to-­target approach: 1-­year results Conclusion of CareRA, a randomised pragmatic open-­label superiority trial. Ann Rheum Dis Based on this economic analysis, compared with more inten- 2017;76:511–20. sive step-­down combination strategies or to a conventional 8 stouten V, Westhovens R, Pazmino S, et al. Effectiveness of different combinations of step-­up approach, COBRA Slim, the combination of MTX and DMARDs and glucocorticoid bridging in early rheumatoid arthritis: two-­year results of CareRA. Rheumatology 2019;58:2284–94. a moderate-dose­ GC bridging scheme, was less expensive and 9 Verschueren P, De Cock D, Corluy L, et al. Patients lacking classical poor prognostic lead to comparable or better gain in QALYs. Therefore, we markers might also benefit from a step-­down glucocorticoid bridging scheme in early consider COBRA Slim a good starting strategy for all patients rheumatoid arthritis: week 16 results from the randomized multicenter CareRA trial. with early RA, irrespective of prognostic markers, in a treat-to-­ ­ Arthritis Res Ther 2015;17:97. 10 putrik P, Ramiro S, Kvien TK, et al. Variations in criteria regulating treatment with target setting. reimbursed biologic DMARDs across European countries. are differences related to Twitter Sofia Pazmino @sophie_33pl country’s wealth? Ann Rheum Dis 2014;73:2010–21. 11 aletaha D, Landewe R, Karonitsch T, et al. Reporting disease activity in clinical trials of Acknowledgements We would like to show our gratitude to all participating patients with rheumatoid arthritis: EULAR/ACR collaborative recommendations. Ann patients, as well as to the investigators and medical staff at all sites. We appreciate Rheum Dis 2008;67:1360–4. the time invested. Furthermore, we are thankful for providing statistical advice 12 Hernández Alava M, Wailoo AJ, Ara R. Tails from the peak district: adjusted limited to Robert Obenchain and Anikó Lovik, economical advice to Gonzalo Villa, data dependent variable mixture models of EQ-­5D questionnaire health state utility values. handling/processing advice to Ramiro Pazmino and proofreading to Geovanna Lucio. Value Health 2012;15:550–61. Contributors PV, RW, AB, SP and DDC made substantial contributions to the 13 Hernández Alava M, Wailoo A, Wolfe F, et al. The relationship between EQ-­5D, HAQ conception or design of the study. SP performed the statistical analysis. The and pain in patients with rheumatoid arthritis. Rheumatology 2013;52:944–50. manuscript was written by SP, PV, RW AB and DDC and subsequently revised critically 14 pennington B, Davis S. Mapping from the health assessment questionnaire to the by all the remaining coauthors. All authors were involved in data interpretation and EQ-­5D: the impact of different algorithms on cost-­effectiveness results. Value Health approved the final version to be submitted for publication. 2014;17:762–71. 15 gøthesen Øystein, Slover J, Havelin L, et al. An economic model to evaluate cost-­ Funding The CareRA trial (EudraCT number: 2008-007225-39) was funded by effectiveness of computer assisted knee replacement surgery in Norway. BMC a Flemish governmental grant (Agency for Innovation by Science and Technology Musculoskelet Disord 2013;14:202. (IWT)). Patrick Verschueren holds the Pfizer chair for early rheumatoid arthritis 16 paracha N, Abdulla A, MacGilchrist KS. Systematic review of health state utility values management at the KU Leuven. in metastatic non-­small cell lung cancer with a focus on previously treated patients. Competing interests None declared. Health Qual Life Outcomes 2018;16. 17 lloyd A, Nafees B, Narewska J, et al. Health state utilities for metastatic breast cancer. Patient and public involvement statement The pragmatic CareRA protocol Br J Cancer 2006;95:683–90. was strongly inspired by daily interactions of the investigators with RA patients in 18 groeneveld PW, Lieu TA, Fendrick AM, et al. Quality of life measurement clarifies daily clinical practice. Patients were not formally involved in setting the research the cost-­effectiveness of Helicobacter pylori eradication in peptic ulcer disease and question or the outcome measures, nor were they invited to comment on study uninvestigated dyspepsia. Am J Gastroenterol 2001;96:338–47. design or the interpretation of results of this manuscript. However, results of this 19 liu D, Lehmann HP, Frick KD, et al. Active surveillance versus surgery for low risk research will be disseminated to study participants, all stakeholders and the general prostate cancer: a clinical decision analysis. J Urol 2012;187:1241–6. public in collaboration with patient organisations and the Belgian patient partners 20 Einarson TR, Bereza BG, Nielsen TA, et al. Utilities for asthma and COPD according to programme (trained patients who educate physicians, medicine students and other category of severity: a comprehensive literature review. J Med Econ 2015;18:550–63. healthcare professionals in collaboration with a rheumatologist). 21 a health state utility model estimating the impact of IVOSIDENIB on EHA library. Patient consent for publication Not required. storm M, 2018. Available: https://​library.ehaweb.​ ​org/eha/​ ​2018/stockholm/​ ​215730/​ michael.​storm.​a.health.​ ​state.utility.​ ​model.estimating.​ ​the.impact.​ ​of.ivosidenib.​ ​html Ethics approval The study was approved by the leading Ethics Committee of the [Accessed 18 Oct 2019]. University Hospitals Leuven after consulting the medical ethics committee of each 22 liviu Preda A, Galieta Mincă D. Cost-­Effectiveness analysis of treatment for metastatic participating centre (ref s51411). renal carcinoma in Romania. J Med Life 2018;11:306–11. Provenance and peer review Not commissioned; externally peer reviewed. 23 doyle S, Lloyd A, Davis M. Public perception of atrial fibrillation and treatment-­related adverse events in the UK. British Journal of Cardiology 2011;18:88–93. Data availability statement Data are available on reasonable request. The 24 Van Wilder L, Rammant E, Clays E, et al. A comprehensive Catalogue of EQ-­5D scores authors commit to making the relevant anonymised patient data available for a in chronic disease: results of a systematic review. Qual Life Res 2019;28:3153–61. specified purpose approved by the institution and the principal investigator of the 25 Zulman DM, Vijan S, Omenn GS, et al. The relative merits of population-based­ and CareRA study and with a signed data access agreement. targeted prevention strategies. Milbank Q 2008;86:557–80. ORCID iDs 26 Honoraria, prijzen en tegemoetkomingen van artsen (deel 1) - RIZIV. Available: Sofia Pazmino http://orcid.​ ​org/0000-​ ​0001-​8579-​6914 https://www.riziv.​ ​fgov.​be/​nl/themas/​ ​kost-terugbetaling/​ ​door-ziekenfonds/​ ​individuele-​ Annelies Boonen http://orcid.​ ​org/0000-​ ​0003-​0682-​9533 verzorging/honoraires/​ ​Paginas/arts-​ ​deel1.aspx​ [Accessed 13 Dec 2018]. 27 BCFI | Homepage. Available: http://www.​bcfi.​be/​nl/​start [Accessed 13 Dec 2108]. 28 Financiële feedback per pathologie. Available: https://​tct.fgov.​ ​be/webetct/​ ​etct-web/​ ​ References anonymous?​lang=​nl [Accessed 13 Dec 2018]. 1 Erkan D, Yazici Y, Harrison MJ, et al. Physician treatment preferences in rheumatoid 29 Fees, prices and reimbursements - RIZIV. Available: https://www.​riziv.​fgov.​be/​nl/​ arthritis of differing disease severity and activity: the impact of cost on first-line­ themas/kost-​ ​terugbetaling/door-​ ​ziekenfonds/​individuele-​verzorging/honoraires/​ ​ therapy. Arthritis Rheum 2002;47:285–90. Paginas/​default.​aspx [Accessed 13 Dec 2018]. 2 smolen JS, Landewé R, Bijlsma J, et al. EULAR recommendations for the management 30 guide to the methods of technology appraisal 2013 2013. of rheumatoid arthritis with synthetic and biological disease-­modifying antirheumatic 31 sachs JD. Macroeconomics and health: investing in health for economic development. drugs: 2016 update. Ann Rheum Dis 2017;76:960–77. Revista Panamericana de Salud Pública 2002;12:143–4. 3 de Jong PH, Hazes JM, Han HK, et al. Randomised comparison of initial triple DMARD 32 making choices in health: who guide to cost-­effectiveness analysis. Revista Española therapy with methotrexate monotherapy in combination with low-­dose glucocorticoid de Salud Pública 2004;78:317. bridging therapy; 1-year­ data of the tREACH trial. Ann Rheum Dis 2014;73:1331–9. 33 Eurostat - Data Explorer. Available: https://​appsso.eurostat.​ ​ec.europa.​ ​eu/​nui/​show.​do?​ 4 goekoop-­Ruiterman YPM, de Vries-Bouwstra­ JK, Allaart CF, et al. Comparison of dataset=nama_​ ​10_pc&​ ​lang=en​ [Accessed 4 Nov 2019]. treatment strategies in early rheumatoid arthritis: a randomized trial. Ann Intern Med 34 Burgette LF, Reiter JP. Practice of epidemiology multiple imputation for missing data 2007;146:406–15. via sequential regression trees. Am J Epidemiol 2010;172:1070–6.

564 Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 Rheumatoid arthritis

35 Rubin DB, Schenker N. Multiple imputation for interval estimation from simple random 42 gulácsi L, Zrubka Z, Brodszky V, et al. Long-­Term efficacy of tumor necrosis factor samples with Ignorable nonresponse. J Am Stat Assoc 1986;81:366–74. inhibitors for the treatment of Methotrexate-­Naïve rheumatoid arthritis: systematic 36 schomaker M, Heumann C. Bootstrap inference when using multiple imputation literature review and meta-­analysis. Adv Ther 2019;36:721–45. statistics in medicine. Stat Med 2018;37:2252–66. 43 national Institute for Health and Care Excellence. Adalimumab, etanercept, infliximab, 37 gyllensten H, Rehnberg C, Jönsson AK, et al. Cost of illness of patient-reported­ certolizumab pegol, golimumab, tocilizumab and abatacept for rheumatoid arthritis not previously treated with DMARDs or after conventional DMARDs only have failed adverse drug events: a population-­based cross-­sectional survey. cost of illness of 2015. patient-­reported adverse drug events: a population-­based cross-­sectional survey. 44 stevenson M, Archer R, Tosh J, et al. Adalimumab, etanercept, infliximab, certolizumab Open 2013;3:2574. pegol, golimumab, tocilizumab and abatacept for the treatment of rheumatoid 38 Korthals-­de Bos I, Van Tulder M, Boers M, et al. Indirect and total costs of arthritis not previously treated with disease-­modifying antirheumatic drugs and after early rheumatoid arthritis: a randomized comparison of combined step-­down the failure of conventional disease-­modifying antirheumatic drugs only: systematic prednisolone, methotrexate, and sulfasalazine with sulfasalazine alone. J Rheumatol review and economic evaluation. Health Technol Assess 2016;20:1–610. 2004;31:1709–16. 45 schwartz D, Lellouch J. Explanatory and pragmatic attitudes in therapeutical trials. J 39 Ter Wee MM, Coupé VM, den Uyl D, et al. Cost-Utility­ of COBRA-­light versus cobra Clin Epidemiol 2009;62:499–505. therapy in patients with early rheumatoid arthritis: the COBRA-­light trial. RMD Open 46 Ramsey SD, Willke RJ, Glick H, et al. Cost-­Effectiveness analysis alongside clinical trials II-­An ISPOR good research practices Task force report. Value Health 2015;18:161–72. 2017;3:1–9. 47 Boers M, Verhoeven AC, Markusse HM, et al. Randomised comparison of combined 40 Fautrel B. Economic benefits of optimizing anchor therapy for rheumatoid arthritis. step-­down prednisolone, methotrexate and sulphasalazine with sulphasalazine alone Rheumatology 2012;51 Suppl 4:iv21–6. in early rheumatoid arthritis. The Lancet 1997;350:309–18. 41 cai W, Gu Y, Cui H, et al. The efficacy and safety of mainstream medications for 48 Verschueren P, Esselens G, Westhovens R. Daily practice effectiveness of a step-­ patients with cDMARD-­Naïve rheumatoid arthritis: a network meta-­analysis. Front down treatment in comparison with a tight step-­up for early rheumatoid arthritis. Pharmacol 2018;9:138. Rheumatology 2008;47:59–64.

Pazmino S, et al. Ann Rheum Dis 2020;79:556–565. doi:10.1136/annrheumdis-2019-216874 565 Rheumatoid arthritis

Epidemiological science Risk of neuroinflammatory events in arthritis patients treated with tumour necrosis factor alpha inhibitors: a collaborative population-­based cohort study from Denmark and Sweden Tine Iskov Kopp ‍ ‍ ,1,2 Bénédicte Delcoigne,3 Elizabeth V Arkema ‍ ‍ ,3 Rikke Kart Jacobsen,2 Melinda Magyari,1,4 Else Helene Ibfelt ‍ ‍ ,2,5 Henning Locht,6 Finn Sellebjerg,4 Rene Lindholm Cordtz ‍ ‍ ,7 Dorte V Jensen,7 Johan Askling,3,8 Lene Dreyer9,10

Handling editor Josef S Abstract Key messages Smolen Objectives To investigate whether tumour necrosis factor alpha inhibitors (TNFis) are associated with an ►► Additional material is What is already known about this subject? published online only. To view increased risk of neuroinflammatory diseases among ►► Numerous case reports suggest that multiple please visit the journal online patients with arthritic diseases. sclerosis-­related disorders may be adverse (http://dx.​ ​doi.org/​ ​10.​1136/​ Methods Cohorts of patients with rheumatoid arthritis events following treatment with tumour annrheumdis-2019-​ ​216693). (RA, n=25 796), psoriatic arthritis (PsA, n=8586) and necrosis factor alpha inhibitor (TNFi) for arthritic ankylosing spondylitis (AS, n=9527) who initiated a For numbered affiliations see diseases. end of article. TNFi treatment year 2000–2017 were identified from nationwide clinical rheumatology registers in Sweden What does this study add? Correspondence to and Denmark. Information on demyelinating disease and ► This study included 175 520 arthritic patients of Dr Tine Iskov Kopp, The Danish inflammatory neuropathy diagnoses asw retrieved from ► whom 43 909 were treated with TNFi in routine Multiple Sclerosis Registry, prospective linkage to National Patients Register. A Cox Department of Neurology, care in Denmark and Sweden and followed for proportional hazard model was used to estimate HRs Rigshospitalet Glostrup, 2600 up to 17 years for neuroinflammatory events in and 95% CI comparing TNFi exposed and non-­exposed, Glostrup, Denmark; National Patient Registers. tine​ .​iskov.k​ opp@​regionh.​dk by disease and country. ► An increased but rare risk of neuroinflammatory Results Among 111 455 patients with RA, we identified ► TIK and BD contributed equally. events following TNFi treatment among 270 (Sweden) and 51 (Denmark) events (all types of patients with psoriatic arthritis and ankylosing neuroinflammatory diseases combined), corresponding Received 22 November 2019 spondylitis compared with non-­treated patients Revised 20 February 2020 to crude incidence rates (per 1000 person-­years) of 0.37 was observed in both Sweden and Denmark. Accepted 20 February 2020 (Sweden) and 0.39 (Denmark) in TNFi-­treated patients ► No consistent and significant risk of Published Online First vs 0.39 (Sweden) and 0.28 (Denmark) in unexposed ► 11 March 2020 neuroinflammatory events following TNFi patients, and an age-­sex-calendar­ -­period-adjusted­ HR treatment among patients with rheumatoid (95% CI) of 0.97 (0.72 to 1.33) (Sweden) and 1.45 (0.74 arthritis was found. to 2.81) (Denmark) in TNFi exposed compared with non-­ exposed patients. For a total of 64 065 AS/PsA patients, How might this impact on clinical practice or the corresponding numbers were: 196 and 32 events, future developments? crude incidence rates of 0.59 and 0.87 in TNFi-­treated ► The risk profile following treatment with TNFis patients vs 0.40 and 0.19 in unexposed patients, and ► may be different for inflammatory arthritides, HRs of 1.50 (1.07 to 2.11) and 3.41 (1.30 to 8.96), for which has impact on decision making in clinical Sweden and Denmark, respectively. For multiple sclerosis, practice. the patterns of HRs were similar. Conclusions Use of TNFi in AS/PsA, but not in RA, was associated with increased risk of incident neuroinflammatory disease, though the absolute risk was example, multiple sclerosis (MS), inflammatory 1–11 below one in 1000 patients/year. neuropathies and optic neuritis. Indeed, available evidence suggests an important role of TNF in the pathogenesis of MS. High levels of TNF have been found in cerebral white matter © Author(s) (or their employer(s)) 2020. No Introduction lesions suggestive of MS and cerebrospinal fluid of 12 13 commercial re-­use. See rights Tumour necrosis factor alpha inhibitors (TNFis) patients with MS ; and a genetic variant in the and permissions. Published have successfully established themselves as main- gene encoding TNF-receptor­ 1 confers suscepti- by BMJ. stay therapies for rheumatoid arthritis (RA), psori- bility to MS.14 Patients with MS treated with TNFi To cite: Kopp TI, atic arthritis (PsA) and ankylosing spondylitis (AS). in randomised clinical trials have shown unfa- Delcoigne B, Arkema EV, Several case reports and small case series have vourable results or have even been halted due to et al. Ann Rheum Dis indicated that treatment with TNFi treatment may dose-­dependent disease exacerbation.15 16 The back- 2020;79:566–572. be linked with neuroinflammatory disorders, for ground risk of developing a neuroinflammatory

566 Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 Rheumatoid arthritis event of these types may, however, depend on the type of arthri- Table 1 ICD-10 codes used to define the neuroinflammatory events tides. For example (and in the absence of TNFi treatment), PsA of interest and specific subgroups of diseases for the analyses patients have been reported to be at increased risk for MS,17 while an inverse association has been suggested for RA and MS.18 Neuroinflammatory event DIC code Subgroup To examine whether such neuroinflammatory events are Demyelinating diseases of the G35 with all sub-­ 1: Demyelinating diseases merely coincidental or causally linked to the use of TNFi, central nervous system codes, G36.0, G36.8, of the central nervous and how any association may vary by treatment indication, G36.9, G37.1, G37.3, system and optic neuritis G37.5, G37.8, G37.9 (subsequently named: large observational studies with sufficiently long follow-up­ are ‘demyelinating diseases’) warranted. To date, only a few observational studies exist, mostly Optic neuritis H46 pointing towards increased risk of neuroinflammatory adverse Retrobulbar neuritis in H48.1 diseases classified elsewhere events following TNFi treatment.19–21 We previously reported an association between TNFi treatment and MS among men (but Other encephalitis, myelitis G04.8, G04.9 and encephalomyelitis and not women) with RA and AS compared with the general popula- unspecified 20 tion. However, that study had limited power, nor did it capture Multiple sclerosis (MS) G35 with all subcodes 2: MS other MS-­related conditions, such as other types of demyelin- Inflammatory G61.0, G61.8, G61.9 3: Inflammatory ation or optic neuritis. polyneuropathies polyneuropathies The purpose of the current study was, therefore, to investi- (subsequently named: gate, in a much larger cohort, the occurrence of MS and other ‘polyneuropathies’) types of neuroinflammatory events, and to assess risks specifi- ICD-10, International Classification of Diseases 10 revision. cally by TNFi treatment and by rheumatological condition. To do this, we combined nationwide clinical and health registers population of all RA, PsA or AS). Patients were at least 18 years from Sweden and Denmark. old at start of follow-­up in both countries.

Materials and methods Exposure assessment and start of follow-up Data sources In both countries, follow-up­ for patients exposed to TNFi (orig- We performed a prospective cohort study in Denmark and inator or biosimilar) started at the date of first TNFi treatment. Sweden in parallel (ie, all analyses were run separately in each For patients who were (yet) naïve to TNFi start of follow-­up was country), with inflammatory arthritis patients identified within defined as the first registered visit in DANBIO (Denmark), and national registers. DANBIO is a Danish nationwide quality and for patients in Sweden, start of follow-up­ was at the second visit research register that covers 90%–98% of adults with rheumatic listing (RA or PsA/AS) in the NPR, or 1 January 2000, whichever diseases treated with TNFi and other biologic disease-modifying­ came last, and with the additional condition that at least one of anti-­rheumatic drugs (bDMARDs) in routine care.22 23 The the two required arthritis diagnoses was established in a hospital Swedish Anti-­Rheumatic Treatment in Sweden Register (ARTIS) Department of Rheumatology or Internal Medicine, a validated 25 30 31 is a subset of the nationwide Swedish Rheumatology Quality method to capture rheumatological diagnoses in NPRs. We Register that covers 87%–95% of bDMARD-treated­ patients did not consider use of non-­TNFi biologicals which was used in with rheumatic diseases.24 In both registers, clinical data are less than 10% of patients in both cohorts. collected prospectively. Diagnoses are validated by yearly medical record audits (DANBIO) with a positive predictive Outcome information and end of follow-up value of 96% for RA diagnoses in 201525 and the quality and In addition to MS, we searched for a broader array of diagnostic coverage of the ARTIS register is regularly checked through codes that may correspond to neuroimmune conditions occur- 1–11 reports and scientific publications.26 The nationwide Danish and ring as a result of TNFi treatment using the WHO Interna- Swedish National Patients Registers (NPR) were used to iden- tional Classification of Diseases 10th revision, which are coded tify outcomes (selected neuroinflammatory diagnoses), and for by the discharging physician when a patient is discharged from identifying Swedish arthritis patients with RA, AS or PsA who hospital (table 1). All registered neuroinflammatory events were were unexposed to TNFi. These registers are administrative and combined and categorised in three subgroups: (1) demyelinating funded through public taxation and contain nearly complete diseases of the central nervous system including optic neuritis, information about inpatient and outpatient care.27 28 Finally, subsequently termed ‘demyelinating diseases’; (2) MS and (3) the Civil Registration System29 and the Swedish Population inflammatory polyneuropathies, subsequently termed ‘polyneu- Register28 were used for retrieving information on dates of birth, ropathies’. Patients with one of the above-mentioned­ neuroin- death or emigration, respectively. The personal identification flammatory diagnoses prior to start of follow-­up were excluded number assigned to every Danish and Swedish citizen permitted from the cohorts. The validity of MS in the Danish NPR has 32 register linkage. been found to be 96.3%. The cohorts were followed up from entry date until the date of the first registered neuroinflammatory diagnosis, death, emigra- Cohort identification and follow-up tion or the end of study (DANBIO: 10 May 2017 and Sweden: Inclusion criteria 31 December 2017), whichever came first. TNF-naïve­ indi- The Danish data included patients from DANBIO with a diag- viduals who, during follow-up,­ started a TNFi treatment were nosis of RA, PsA or AS between 1 January 2000 and 20 January censored from the TNFi-­naïve group and contributed person-­ 2017. The Swedish data included patients with a diagnosis of time and any events to the TNFi exposed group thereafter. RA, PsA or AS between 1 January 2000 and 29 December, 2017 identified from the NPR and/or from ARTIS (for all practical Patient and public involvement purposes, this setting could be seen as an identification of all Patients or the public were not involved in the design, or conduct, TNFi exposed patients from the national underlying source or reporting, or dissemination plans of our research.

Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 567 Rheumatoid arthritis

Statistical analyses younger at entry and at time of diagnosis than non-­treated Baseline demographic and clinical characteristics of the cohort patients in both countries, they had a longer disease duration at at entry are presented as numbers with their percentages and entry, and they had higher disease activity based on clinical and medians with IQR. Due to the prospective design, patients could subjective measures. contribute to both the exposed and the unexposed groups. In that case, baseline demographic and clinical characteristics were Rheumatoid arthritis retrieved at the two time points corresponding to the start of For Sweden and Denmark, respectively, we identified the follow-­up of the episode in question. following number of events per person-­years: 160 events/706 A Cox proportional hazards model, with attained age as time 386 person-­years and 33 events/162 003 person-­years for scale, was used to estimate HRs and 95% CIs for the compar- ‘demyelinating diseases’, 49 events/706 830 person-­years and 11 ison of TNFi treated to non-­treated patients with delayed entry. events/162 118 person-­years for MS; and 110 events/706 650 Adjusted models included gender and year of inclusion in the person-­years and 18 events/162 088 person-­years for ‘polyneu- cohorts (categorised as: 2000–2005, 2006–2011 and 2012– ropathies’. Taking all neuroinflammatory events combined in 2017). Exposure to TNFi was included in the model as a time-­ each country, the crude incident rates (per 1000 person-­years) dependent variable. The proportionality assumption of the Cox across all patients exposed to TNFi were 0.37 (Sweden) and proportional hazard model was checked for all variables using 0.39 (Denmark) vs 0.39 (Sweden) and 0.28 (Denmark) in unex- 33 the method by Lin et al. posed patients (table 3). The following sensitivity or subanalyses were performed: (1) We did not find any association between treatment with TNFi An ‘on drug’ analysis, in which follow-up­ was censored at the and risk of a neuroinflammatory event, (HR 0.97, (95% CI 0.72 date of discontinuation of the TNFi treatment (plus a 3-month­ to 1.33)) for the Swedish cohort and an elevated but not signifi- interval), (2) Stratification by gender, (3) Stratification by cantly increased (HR 1.45, (95% CI 0.74 to 2.81)) in the Danish follow-­up time (<1 year, 1–4 years and ≥5 years) and (4) Strat- cohort (table 3). For MS, a borderline statistically significantly ification by age at treatment start or cohort entry (age <50, reduced risk after exposure to TNFi was found in Sweden (HR 50–64 and ≥65 years). In addition, for the DANBIO cohort 0.44 (95% CI 0.18 to 1.05)). In Denmark, the number of MS only, a subanalysis was made with adjustment for smoking status events was very low (four events among the exposed vs seven (entered categorically as ‘previous’, ‘current’ or ‘never’ smokers events among the unexposed), hence, it was not possible to at time of the first visit in DANBIO), and, among RA patients confirm or refute the Swedish results. only, Disease Activity Score using 28 joint count and C reac- Among TNFi exposed patients experiencing an event (of any tive protein (DAS28-­CRP), when this information was recorded type), the median time to event counting from start of TNFi treat- (complete-­case analysis). Finally, for the Swedish cohort only, we ment was 3.6 years (IQR=1.1–6.8) and 3.2 years (IQR=1.4–5.1) performed a sensitivity analysis that employed alternative defini- in the Swedish and Danish cohorts, respectively (online supple- tions of the unexposed comparator: (1) the unexposed patients mentary table 1). identified from ARTIS (ie, using the TNFi naïve patients from the Rheumatology Register instead of the NPR); (2) the unex- posed patients from NPR restricted to patients treated with a Sensitivity and subanalyses Our sensitivity analyses did not demonstrate any signal of TNFi conventional synthetic (cs) DMARD who either switched or esca- treatment being associated with risk of a neuroinflammatory lated to another csDMARD, the date of switching taken as the event (all types combined) (online supplementary table 2). In follow-­up start, that is, a new-user­ active comparator approach. both cohorts, though, a tendency towards increased risk of a The study period for the latter analysis was restricted to 2006– neuroinflammatory events was seen with increasing age. 2017 due to the availability of the Swedish Prescribed Drug Register from which csDMARD information was retrieved34; (3) the unexposed patients from NPR with a start of follow-up­ at AS and PsA the second visit listing (RA or PsA/AS) in the NPR, or 1 January In Sweden and Denmark, we identified the following number of 2006, whichever came last (study period 2006–2017) and (4) events per person-­years: 141/447 906 person-­years and 24/60 same as the latter one with 2011 instead of 2006 (study period 835 person-­years events of ‘demyelinating diseases’, 48/448 324 2011–2017). person-­years and 7/60 894 person-years­ MS events, 55/448 497 The number needed to harm (NNH) was calculated as the person-­years and 8/60 891 person-years­ polyneuropathy events, inverse of the crude excess incidence (ie, crude incidence rateex- respectively. Taking all neuroinflammatory events combined in posed – crude incidence ratenon-­exposed). each country, the crude incidence rates (per 1000 person-years)­ A p<0.05 was considered significant for all tests. Statistical in patients exposed to TNFi were 0.59 (Sweden) and 0.87 analyses were conducted using SAS V.9.4 (SAS). (Denmark) vs 0.40 (Sweden) and 0.19 (Denmark) in unexposed patients. These rates translated into a 50% increased risk of a neuroinflammatory event (all types combined) among patients Results ever exposed to TNFi in the Swedish cohort (HR 1.50 (95% CI Baseline characteristics 1.07 to 2.11)) and a 3.4-­fold increased risk in the Danish cohort We included a total of 175 520 patients with RA, PsA or AS in (HR 3.41 (95% CI 1.30 to 8.96)) compared with patients not the two cohorts of whom 43 909 (25%) patients were exposed exposed to TNFi (table 3). This corresponds to an NNH of 5300 to TNFi any time during follow-up­ (table 2). patients treated with TNFi/year in Sweden and 1500 patients in The overall proportion of TNFi exposed patients varied from Denmark. 23% among patients with RA in the Swedish cohort to 38% Among TNFi exposed patients experiencing an event, the among patients with PsA and AS in DANBIO. Median follow-up­ time to (any type of) neuroinflammatory event was 3.8 years time varied between 2.7 years for non-­exposed PsA and AS (IQR=1.7–5.9) and 3.1 years (IQR=1.4–8.6), and age at patients in DANBIO to 7.4 years among RA patients exposed to neuroinflammatory event among TNFi exposed patients was TNFi in both cohorts. Generally, TNFi exposed patients were 43 years (IQR=34–52) and 46 years (IQR=38–55) in the

568 Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 Rheumatoid arthritis

Table 2 Baseline characteristics of ever and never TNFi exposed patients in the two cohorts divided by arthritic disease

Rheumatoid arthritis Ankylosing spondylitis and psoriatic arthritis

Sweden Denmark Sweden Denmark

TNFi Demographics characteristics TNFi exposed unexposed TNFi exposed TNFi unexposed TNFi exposed TNFi unexposed TNFi exposed TNFi unexposed

Total, n 18 598 78 719 7198 26 006 13 615 48 288 4498 9611 Women, n (%) 14 011 (75) 56 036 (71) 5350 (74) 18 372 (71) 6286 (46) 23 665 (49) 1872 (42) 4747 (49) Age at entry (years), median (IQR) 57 65 57 63 45 51 45 49 (47–66) (54–74) (48–65) (52–72) (35–55) (39–61) (35–54) (39–60) Median years of follow-­up, median (IQR) 7.4 6.1 7.4 3.6 5.3 6.8 6.6 2.7 (3.7–12) (2.7–11) (4.2–10) (1.8–6.1) (2.6–9.1) (3.1–12) (3.7–9.4) (1.1–4.9) Age at rheumatological diagnosis, years, median (IQR) 46 61 48 55 31 47 37 43 (missing) (35–56) (50–71) (37–57) (44–66) (23–42) (36–58) (28–47) (33–54) (325) (0) (490) (3,743) (130) (0) (622) (2,670) Calendar year of inclusion, n (%) 2000–2005 5350 (29) 39 954 (51) 1806 (25) 627 (2) 1636 (12) 17 951 (37) 611 (14) 106 (1) 2006–2011 6760 (36) 21 917 (28) 3376 (47) 11 014 (42) 4773 (35) 16 253 (34) 2185 (49) 2899 (30) 2012–2017 6488 (35) 16 848 (21) 2016 (28) 14 365 (55) 7206 (53) 14 084 (29) 1702 (38) 6606 (69) Smoking status, n (%) Current 801 (4) – 1680 (23) 4801 (18) 735 (5) – 1301 (29) 1770 (18) Previous 2157 (12) – 1869 (26) 5852 (23) 1837 (13) – 895 (20) 1696 (18) Never 2065 (11) – 2690 (37) 9314 (36) 2502 (18) – 1860 (41) 3638 (38) Missing 13 575 (73) 78 719 (100) 959 (13) 6038 (23) 8541 (63) 48 288 (100) 442 (10) 2507 (26) Disease related characteristics Disease duration, years, median (IQR) (missing) 6.9 (2.5–15) 0.5 (0.2–3.0) 6.0 (2.2–14) 1.4 (0.1–8.9) 8.6 (3.3–17) 0.5 (0.2–2.1) 3.4 (0.9–9.9) 1.1 (0.1–6.5) (321) (0) (490) (3743) (126) (0) (622) (2 670) Disease duration <2 years, n 3693 (20) 56 120 (71) 1542 (23) 11 856 (53) 2211(16) 35 929 1544 3970 (%) (321) (0) (490) (3 743) (126) (74) (44) (57) (missing) (0) (622) (2 670) Tender joint count (0–28 scale), median (IQR) (missing) 6 8 3 2 6 2 (3-11) (8-13) (0–7) (0–6) (2-11) (0–6) (3188) (78 719) (1182) (1701) (4269) (48 288) (699)* (893)* Swollen joint count (0–28 scale), median (IQR) (missing) 6 5 2 1 2 0 (3-11) (2-9) (0–5) (0–4) (0–5) (0–2) (3167) (78 719) (1179) (1643) (4257) (48 288) (709)* (930)* DAS28-­CRP, median (IQR) (missing) 4.8 4.8 3.4 3.6 4.3 3.1 (3.9–5.6) (3.7–5.6) (2.1–4.6) (2.7–4.6) (3.2–5.2) (1.9–4.2) (4296) (78 719) (1513) (5627) (5429) (48 288) (861)* (2152)* BASDAI (0–10 scale), median (IQR) (missing) 60 46 (46-73) (26-64) (13 615) (48 288) (1788) (2867) CRP, mg/L, median (IQR) (missing) 11 12 7 7 9 5 (5–28) (5–27) (3–16) (3–18) (3–20) (2–11) (3214) (78 719) (1100) (2373) (3538) (48 288) (1098) (1918) HAQ (0–3 scale), median (IQR) (missing) 1.1 1.3 0.8 (0.6–1.6) (0.8–1.8) (0.3–1.3) (4076) (78 719) (1304) (3901) Seropositive RA, n (%) (missing) 2437 5412 16 882 (39) (75) (72) (12 348) (78 719) (463) (2452) Concomitant medication, n (%) Methotrexate 11 533 (62) – 4832 (67) 15 041 (58) 4607 (34) – 1726 (38) 3591 (37) Oral steroid 8226 (44) – 1911 (27) 3130 (12) 2131 (16) – 225 (5) 235 (2) TNFi, n (%) Adalimumab (Humira) 3264 (18) 1776 (25) 3620 (27) 1600 (36) Infliximab (Remicade, 6353 (34) 2555 (36) 3947 (29) 1308 (29) Remsima, Inflectra) Etanercept (Enbrel, Benepali) 6959 (37) 1773 (25) 4389 (32) 876 (20) Golimumab (Simponi) 906 (5) 199 (3) 1168 (9) 430 (10) Certolozumab pegol (Cimzia) 1116 (6) 895 (12) 491 (4) 284 (6) Data are expressed as numbers (%) or median values (IQR). *Only among psoriatic arthritis patients. BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; CRP, C reactive protein; DAS28, Disease Activity Score using 28 joint count; HAQ, Health Assessment Questionnaire; RA, rheumatoid arthritis; TNFi, tumour necrosis factor-α inhibitor.

Swedish and Danish cohort, respectively (online supplementary despite the difference in HRs between males and females in table 1). the Danish cohort, gender did not appreciably modify the HR for the association between TNFi and the outcome(s) under Sensitivity and subanalyses study (ie, no interaction was found). The risk for a neuroin- The ‘on-­drug’ analyses’ estimates were in agreement with the flammatory event was increased with increasing follow-­up results of the main analysis although the risk estimates were time in the Swedish cohort, whereas this was less evident lower and less precise in both countries (online supplementary in the Danish cohort due to low number of events in each table 2). The HR for male patients was twice the HR for female stratum. No apparent association with age at cohort entry or patients in the Danish cohort (5.81 vs 2.50), while in the treatments start was found. When restricting the unexposed Swedish cohort, the HRs were of the same order of magnitude patients in Sweden to csDMARD switchers or when the unex- in both sexes (1.37 for males vs 1.67 for females). However, posed patients were retrieved from the Swedish Rheumatology

Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 569 Rheumatoid arthritis CI)†

sus unexposed Fi exposed N T HR (95% ver 1.45 (0.74 to 2.81) 1.48 (0.65 to 3.34) 1.02 (0.23 to 4.46) 1.43 (0.45 to 4.52) 3.41 (1.30 to 8.96) – – 3.00 (1.04 to 8.64) CI)*

sus unexposed Fi exposed N T HR (95% ver 1.35 (0.77 to 2.37) 1.49 (0.75 to 2.98) 1.00 (0.29 to 3.43) 1.10 (0.41 to 2.97) 4.03 (1.66 to 9.81) – 3.35 (1.25 to 8.98) s Fi N IR/1000 pyr unexp among T 0.28 0.17 0.06 0.11 0.19 0.03 0.032 – 0.16 s Fi N exp IR/1000 pyr among T s Fi unexp events/pyr N N among T s ­ exposed patients stratified by country and arthritic disease Fi exp events/pyr enmark 4/53 360 7/108 758 0.08 6/53 346 12/108 742 0.11 6/29 9457/29 943 1/30 949 0.20 1/30,948 0.23 N D N among T 21/53 249 30/108 685 0.39 15/53 279 18/108 724 0.28 26/29 869 6/30 933 0.87 19/29 900 5/30 935 0.64 CI)†

sus Fi exposed N T ver unexposed HR (95% 0.97 (0.72 to 1.33) 1.14 (0.79 to 1.64) 0.44 (0.18 to 1.05) 1.50 (1.07 to 2.11) 0.7 (0.39 to 1.25) 1.65 (1.12 to 2.43) 1.65 (0.86 to 3.17) 1.13 (0.53 to 2.37) CI)*

sus Fi exposed N T ver unexposed HR (95% (0.70 to 1.29) (0.79 to 1.64) (0.18 to 1.03) (0.99 to 1.92) (0.37 to 1.16) (1.01 to 2.11) (0.77 to 2.68) (0.55 to 2.35) s Fi TNFi exposed patients compared with non- TNFi N unexp IR/1000 pyr among T s Fi exp N IR/1000 pyr among T s Fi unexp events/pyr N N among T s among Fi exp 9/83 082 46/365 415 0.11 0.13 1.14 events/ 6/145 717 43/561 113 0.04 0.08 0.44 40/82 913 101/364 993 0.48 0.28 1.46 14/83 006 34/365 318 0.17 0.09 1.43 weden N S N pyr T 40/145 570 120/560 816 0.27 0.21 1.14 14/145 723 96/560 927 0.10 0.17 0.66 inhibitor; unexp, unexposed. unexp, α inhibitor; tumour necrosis factor- TNFi, incidence rate; IR, exposed; , polyneuropathies polyneuropathies Risk of neuroinflammatory events (all diagnoses and subgroups) for Risk of neuroinflammatory events (all diagnoses and subgroups) years; Exp ­ years; sclerosis sclerosis

Demyelinating diseases of the central nervous system and optic neuritis Multiple Inflammatory Demyelinating diseases of the central nervous system and optic neuritis Inflammatory Multiple All neuroinflammatory diagnoses 54/145 501 216/560 320 0.37 0.39 0.95 All neuroinflammatory diagnoses 49/82 878 147/364 715 0.59 0.40 1.38 *Unadjusted HRs from Cox proportional hazards regression with age as underlying time scale. †Models additionally adjusted for sex and calendar time. person- Pyrs: Table 3 Table Rheumatoid arthritis Ankylosing spondylitis and psoriatic arthritis

570 Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 Rheumatoid arthritis

Register instead of NPR, a doubled risk of neuroinflammatory There are limitations to the study. First, confounding by indi- event was seen for TNFi exposed patients compared with non-­ cation, or channelling, may explain some of our associations. exposed patients (online supplementary table 2). Crude inci- This risk has, however, been mitigated by excluding patients dence rates were similar in AS and PsA patients in the Danish who registered with any neuroinflammatory outcome prior cohort. In the Swedish cohort, the incidence rates were higher to cohort entry. TNFi exposed RA patients had a more active in patients with AS than in PsA, but the ratios between exposed disease based on clinical and objective disease activity measures. and unexposed were of the same magnitude in both diseases However, we do not know whether disease activity is associated (online supplementary table 3). with the risk of having a demyelinating disease or inflammatory neuropathy—and the risk estimates did not change considerably when adjusting for DAS28-­CRP in subanalysis. Although valida- Discussion tion studies have found high positive predictive values for RA, In this large population-based,­ prospective cohort study, we AS and PsA (79%–96%), there may be differences in the validity observed an increased risk of neuroinflammatory events sugges- of treated vs untreated disease, though we do not believe these tive of demyelination following treatment with TNFi among differences are large enough to explain our findings. Further, we patients with PsA or AS compared with non-­exposed PsA or AS cannot exclude residual or unknown confounding. patients although the risk was very low. We found no significantly The strength of the study is the use of high-quality­ population-­ increased risk of neuroinflammatory events among RA patients based rheumatic disease registers with long follow-­up times (up to treated with TNFi compared with non-­treated RA patients, 17 years) supplemented with near-­complete nationwide patient suggesting that the observed safety profile of TNFi with respect registers. The use of data from two countries not only increased to MS and similar conditions varies with the treated diagnosis. statistical power, but also provided a built-in­ means to compare Two other studies have assessed the risk of different neuroin- patterns in the two countries, and thus the generalisability of the flammatory events following TNFi treatment in arthritis results. We included patients with the same underlying disease as patients. Both studies used US health claim databases and a comparison group which reduced the risk of surveillance bias.38 nested case–control design. The study by Bernatsky et al found Another feature of the current study is the agreement between a non-­significant 31% increased risk of demyelinating events the results from the two countries: despite slightly different defi- following TNFi exposure among RA patients compared with 19 nitions of the TNFi-naïve­ comparator cohorts in the two coun- non-­exposed RA patients, whereas Etminan et al found a more tries, the results point to joined conclusions. Since defining an than twofold increased risk of peripheral neuropathy among unexposed group that is fully comparable to the TNFi exposed arthritis patients (the study combined RA, PsA and AS into one group is a real challenge in arthritides patients, a series of alter- group) with a past use of TNFi compared with arthritis patients 21 native comparator groups were defined for sensitivity analyses. not exposed to TNFi. Both studies had relatively short mean All results pointed to similar conclusions, which strengthens the follow-­up times (1.9 and 2.3 person-­years, respectively) and the findings. claims-­based database used where unrepresentative, based of In conclusion, this large prospective cohort study including unemployed and older individuals. However, this might suggest 175 520 patients followed for 1 376 149 person-­years showed an increased risk in the elderly patients as corresponds with our that the use of TNFi for the treatment of PsA or AS, but not for findings of a trend of increased risk with age among RA patients, RA, may be associated with increased risk of having a demyelin- although not significant. ating disease or inflammatory neuropathy. This information will Our large population-­based study had a longer follow-­up time be important for risk communication and evaluation in clinical than these studies and demonstrate that the pattern of risks for practice, even though the absolute risk is low. The underlying developing a neuroinflammatory event with TNFi treatment biological mechanism needs to be further explored to charac- differs between RA and PsA/AS patients. Since chronic inflam- terise the mechanism of action and to enable identification of matory diseases have distinct genetic profiles35 36 and different susceptible patients. distribution of age and sex, these diseases should be separated in the risk analyses. Author affiliations The underlying mechanisms of demyelinating disorders and 1The Danish Multiple Sclerosis Registry, Department of Neurology, Rigshospitalet inflammatory polyneuropathies following TNFi treatment that Glostrup, Glostrup, Denmark 2 have been reported in the literature (mostly in case reports and Center for Clinical Research and Prevention, Frederiksberg Hospital, Frederiksberg, small case series), are not completely understood, but several Denmark 3Clinical Epidemiology Division, Department of Medicine Solna, Karolinska Institutet, mechanisms have been proposed including the lack of entry Stockholm, Sweden through the blood–brain barrier theory, aggravation of CNS 4Danish Multiple Sclerosis Center, Department of Neurology, Rigshospitalet, demyelination by decreasing TNF receptors and unfavourable Kobenhavn, Denmark 37 5Clinical Epidemiology, Steno Diabetes Center Copenhagen, Gentofte, Denmark altering of cytokine responses. 6 The risk estimates obtained from the ‘on-­drug’ analysis among Department of Rheumatology, Frederiksberg Hospital, Frederiksberg, Denmark 7Center for Rheumatology and Spine Diseases - Gentofte, Rigshospitalet, Hellerup, PsA and AS patients were lower than in the main analysis which Denmark may suggest that demyelinating events develop after longer 8Rheumatology, Karolinska University Hospital, Stockholm, Sweden exposure time, and even after the treatment has been discon- 9Department of Rheumatology, Aalborg University Hospital, Aalborg, Denmark 10 tinued. However, in our study, and due to data availability, the Department of Clinical Medicine, Aalborg University, Aalborg, Denmark time to event varied across cohorts, thus making it difficult to Acknowledgements We thank the DANBIO and ARTIS registers for allowing confirm the nature of any temporal association between expo- us the use of the clinical data from their patients. Hospital Departments of Rheumatology and the private rheumatologic clinics in Denmark and Sweden are sure and outcome. However, our results are in keeping with a acknowledged for their contribution to DANBIO and ARTIS. previous study from Denmark that observed six patients with Contributors LD and JA conceived the idea for the study and all authors planned RA, PsA or AS with a demyelinating neuroinflammatory disorder the study collaboratively. TIK, BD, EVA, RKJ, JA and LD developed the analysis plan. following TNFi treatment, observed from 5 months to 4 years TIK, BD, EVA and RKJ performed management of data and statistical analyses. after treatment start.5 MM and FS provided specific input on neuroinflammatory diseases. HL, JA and

Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 571 Rheumatoid arthritis

LD provided specific input on rheumatic diseases. RLC and DVJ provided input on 13 Selmaj K, Raine CS, Cross AH. Anti-­Tumor necrosis factor therapy abrogates DANBIO data. JA and LD provided funding. TIK and BD wrote the first draft of the autoimmune demyelination. Ann Neurol 1991;30:694–700. manuscript. All authors reviewed the results and critically reviewed the manuscript 14 Gregory AP, Dendrou CA, Attfield KE, et al. Tnf receptor 1 genetic risk mirrors outcome for intellectual content. All authors approved the final version of the manuscript. of anti-­TNF therapy in multiple sclerosis. Nature 2012;488:508–11. 15 van Oosten BW, Barkhof F, Truyen L, et al. Increased MRI activity and immune Funding This study was funded by Program for Clinical Research Infrastructure activation in two multiple sclerosis patients treated with the monoclonal anti-tumor­ (PROCRIN) established by the Novo Nordisk and Lundbeck Foundations, and necrosis factor antibody Ca2. Neurology 1996;47:1531–4. received funding from the Swedish Research Council, The independent Research 16 Tnf neutralization in MS: results of a randomized, placebo-­controlled multicenter Fund Denmark, and from the Karolinska Institutet Region Stockholm funds (ALF). study. The Lenercept multiple sclerosis Study Group and the University of British Competing interests MM has received grants from Novartis and Biogen, and Columbia MS/MRI analysis group. Neurology 1999;53:457–65. has received personal fees from Novartis, Biogen, Merck, SanofiG enzyme and 17 Islam MM, Poly TN, Yang H-­C, et al. Increase risk of multiple sclerosis in patients Teva. FS has received grants and personal fees from Biogen, Merck, Novartis, Sanofi with psoriasis disease: an evidence of observational studies. Neuroepidemiology Genzyme and Roche. JA has received grants from Abbvie, BMS, Eli Lilly, MSD, Pfizer, 2019;52:152–60. Roche, Samsung Bioepis, Sanofi and UCB. LD has received grants from BMS and has 18 Somers EC, Thomas SL, Smeeth L, et al. Are individuals with an autoimmune disease received personal fees from Eli Lilly and Galderma. TIK, BD, EVA, RKJ, EHI, RLC, HL at higher risk of a second autoimmune disorder? Am J Epidemiol 2009;169:749–55. and DVJ have nothing to disclose. 19 Bernatsky S, Renoux C, Suissa S. Demyelinating events in rheumatoid arthritis after drug exposures. Ann Rheum Dis 2010;69:1691–3. Patient and public involvement Patients and/or the public were not involved in 20 Dreyer L, Magyari M, Laursen B, et al. Risk of multiple sclerosis during tumour necrosis the design, or conduct, or reporting, or dissemination plans of this research. factor inhibitor treatment for arthritis: a population-­based study from DANBIO and Patient consent for publication Not required. the Danish multiple sclerosis registry. Ann Rheum Dis 2016;75:785–6. Ethics approval According to Danish and Swedish legislation, the registration 21 Etminan M, Sodhi M, Samii A, et al. Tumor necrosis factor inhibitors and risk of and publication of data from national registers do not require patient consent or peripheral neuropathy in patients with rheumatic diseases. Semin Arthritis Rheum approval by Ethics Committees. Approval was given by the Danish Data Protection 2019;48:1083-1086. Agency (​j.​no: 04422, with I-S­ uite no: CSU-­FCFS-2016-001). 22 Ibfelt EH, Jensen DV, Hetland ML. The Danish nationwide clinical register for patients with rheumatoid arthritis: DANBIO. Clin Epidemiol 2016;8:737–42. Provenance and peer review Not commissioned; externally peer reviewed. 23 DANBIO, 2017. Available: https://​danbio-online​ .​dk/formidling/​ ​copy2_​of_​ Data availability statement No data are available. DANBIOrsrapport_2017.​ ​pdf 24 Eriksson JK, Askling J, Arkema EV. The Swedish rheumatology quality register: ORCID iDs optimisation of rheumatic disease assessments using register-­enriched data. Clin Exp Tine Iskov Kopp http://orcid.​ ​org/0000-​ ​0003-​3729-​4420 Rheumatol 2014;32:S147–9. Elizabeth V Arkema http://orcid.​ ​org/0000-​ ​0002-​3677-​9736 25 Ibfelt EH, Sørensen J, Jensen DV, et al. Validity and completeness of rheumatoid Else Helene Ibfelt http://orcid.​ ​org/0000-​ ​0002-​5253-​8465 arthritis diagnoses in the nationwide DANBIO clinical register and the Danish national Rene Lindholm Cordtz http://orcid.​ ​org/0000-​ ​0002-​5271-​2574 patient registry. Clin Epidemiol 2017;9:627–32. 26 Wadström H, Eriksson JK, Neovius M, et al. How good is the coverage and how accurate are exposure data in the Swedish biologics register (ARTIS)? Scand J References Rheumatol 2015;44:22–8. S1 olomon AJ, Spain RI, Kruer MC, et al. Inflammatory neurological disease in patients 27 Schmidt M, Schmidt SAJ, Sandegaard JL, et al. The Danish national patient treated with tumor necrosis factor alpha inhibitors. Mult Scler 2011;17:1472–87. registry: a review of content, data quality, and research potential. Clin Epidemiol 2 Tristano AG. Neurological adverse events associated with anti-­tumor necrosis factor α 2015;7:449–90. treatment. J Neurol 2010;257:1421–31. 28 Ludvigsson JF, Andersson E, Ekbom A, et al. External review and validation of the 3 Lozeron P, Denier C, Lacroix C, et al. Long-Term­ course of demyelinating neuropathies Swedish national inpatient register. BMC Public Health 2011;11:450. occurring during tumor necrosis factor-­alpha-­blocker therapy. Arch Neurol 29 Schmidt M, Pedersen L, Sørensen HT. The Danish civil registration system as a tool in 2009;66:490–7. epidemiology. Eur J Epidemiol 2014;29:541–9. 4 Tanno M, Nakamura I, Kobayashi S, et al. New-O­ nset demyelination induced 30 Waldenlind K, Eriksson JK, Grewin B, et al. Validation of the rheumatoid arthritis by infliximab therapy in two rheumatoid arthritis patients. Clin Rheumatol diagnosis in the Swedish national patient register: a cohort study from Stockholm 2006;25:929–33. County. BMC Musculoskelet Disord 2014;15:432. 5 Theibich A, Dreyer L, Magyari M, et al. Demyelinizing neurological disease after 31 Lindström U, Exarchou S, Sigurdardottir V, et al. Validity of ankylosing spondylitis and treatment with tumor necrosis factor alpha-­inhibiting agents in a rheumatological undifferentiated spondyloarthritis diagnoses in the Swedish national patient register. outpatient clinic: description of six cases. Clin Rheumatol 2014;33:719–23. Scand J Rheumatol 2015;44:369–76. 6 Mahil SK, Andrews TC, Brierley C, et al. Demyelination during tumour necrosis 32 Mason K, Thygesen LC, Stenager E, et al. Evaluating the use and limitations of the factor antagonist therapy for psoriasis: a case report and review of the literature. J Danish national patient register in register-­based research using an example of Dermatolog Treat 2013;24:38–49. multiple sclerosis. Acta Neurol Scand 2012;125:213–7. 7 Li SY, Birnbaum AD, Goldstein DA. Optic neuritis associated with adalimumab in the 33 Lin DY, Wei LJ, Ying Z. Checking the COX model with cumulative sums of martingale-­ treatment of uveitis. Ocul Immunol Inflamm 2010;18:475–81. based residuals. Biometrika 1993;80:557–72. 8 Barreras P, Mealy MA, Pardo CA. Tnf-­Alpha inhibitor associated myelopathies: a 34 Wettermark B, Hammar N, Fored CM, et al. The new Swedish Prescribed Drug neurological complication in patients with rheumatologic disorders. J Neurol Sci Register--opportunities for pharmacoepidemiological research and experience from 2017;373:303–6. the first six months. Pharmacoepidemiol Drug Saf 2007;16:726–35. 9 Matsumoto T, Nakamura I, Miura A, et al. New-O­ nset multiple sclerosis associated 35 Sirota M, Schaub MA, Batzoglou S, et al. Autoimmune disease classification by inverse with adalimumab treatment in rheumatoid arthritis: a case report and literature association with SNP alleles. PLoS Genet 2009;5:e1000792. review. Clin Rheumatol 2013;32:271–5. 36 Zhang Q, C-­YY L, Dong Q, et al. Relationship between HLA-­DRB1 polymorphism and 10 Ibrahim WH, Hammoudah M, Akhtar N, et al. Central nervous system demyelination susceptibility or resistance to multiple sclerosis in Caucasians: a meta-analysis­ of non-­ associated with etanercept in a 51 years old woman. Libyan J Med 2007;2:99–102. family-based­ studies. Autoimmun Rev 2011. 11 Sicotte NL, Voskuhl RR. Onset of multiple sclerosis associated with anti-­TNF therapy. 37 Kemanetzoglou E, Andreadou E. Cns demyelination with TNF-α blockers. Curr Neurol Neurology 2001;57:1885–8 http://www.ncbi​ .​nlm.nih.​ ​gov/pubmed/​ ​11723281 Neurosci Rep 2017;17:36. 12 Liu J, Marino MW, Wong G, et al. Tnf is a potent anti-­inflammatory cytokine in 38 Haut ER, Pronovost PJ. Surveillance bias in outcomes reporting. JAMA autoimmune-mediated­ demyelination. Nat Med 1998;4:78–83. 2011;305:2462.

572 Kopp TI, et al. Ann Rheum Dis 2020;79:566–572. doi:10.1136/annrheumdis-2019-216693 Rheumatoid arthritis

Clinical science Immunosuppression and the risk of readmission and mortality in patients with rheumatoid arthritis undergoing hip fracture, abdominopelvic and cardiac surgery Michael D George ‍ ‍ ,1 Joshua F Baker,1,2 Kevin L Winthrop ‍ ‍ ,3 Seth D Goldstein,4,5 E Alemao,6 Lang Chen,7 Qufei Wu,1 Fenglong Xie,7 Jeffrey R Curtis7

Handling editor Josef S Abstract Key messages Smolen Objectives The impact of immunosuppression on postoperative outcomes has primarily been studied in ►► Additional material is What is already known about this subject? published online only. To view patients undergoing joint replacement surgery. We aimed ►► Studies of the risk of immunosuppression in please visit the journal online to evaluate the impact of biologics and glucocorticoids patients undergoing surgery most commonly (http://dx.​ ​doi.org/​ ​10.​1136/​ on outcomes after other major surgeries. involve patients undergoing elective annrheumdis-2019-​ ​216802). Methods This retrospective cohort study used arthroplasty, but few studies have evaluated 1 Medicare data 2006–2015 to identified adults with Rheumatology and other major surgery. Epidemiology, University of rheumatoid arthritis undergoing hip fracture repair, Pennsylvania, Philadelphia, abdominopelvic surgery (cholecystectomy, hysterectomy, What does this study add? Pennsylvania, USA hernia, appendectomy, colectomy) or cardiac surgery 2Rheumatology, Philadelphia ► In this study, patients with rheumatoid arthritis (coronary artery bypass graft, mitral/aortic valve). ► VAMC, Philadelphia, receiving biologics did not have a greater risk Logistic regression with propensity-­score-­based inverse Pennsylvania, USA of postoperative infection after hip fracture, 3Infectious Diseases, Oregon probability weighting compared 90-­day mortality and abdominopelvic or cardiac surgery compared Health & Science University, 30-­day readmission in patients receiving methotrexate with patients receiving methotrexate without a Portland, Oregon, USA (without a biologic or targeted synthetic disease-­ 4Surgery, Northwestern biologic. modifying antirheumatic drug (tsDMARD)), a tumour University, Evanston, Illinois, ►► Glucocorticoids were associated with a dose-­ USA necrosis factor inhibitor (TNFi) or a non-­TNFi biologic/ 5 dependent increase in the risk of postoperative Surgery, Ann and Robert H tsDMARD <8 weeks before surgery. Similar analyses infection. Lurie Children’s Hospital of evaluated associations between glucocorticoids and Chicago, Chicago, Illinois, USA 6 outcomes. Bristol-­Myers Squibb, Princeton, How might this impact clinical practice or Results We identified 10 777 eligible surgeries: New Jersey, USA future developments? 7Immunology and 3585 hip fracture, 5025 abdominopelvic and 2167 ► Prolonged interruptions in biologics before Rheumatology, University cardiac surgeries. Compared with patients receiving ► major surgery is likely not required, but of Alabama at Birmingham, methotrexate, there was no increase in the risk of 90-­ Birmingham, Alabama, USA minimising glucocorticoids before surgery day mortality or 30-­day readmission among patients should be a focus of perioperative management. Correspondence to receiving a TNFi (mortality adjusted OR (aOR) 0.83 Dr Michael D George, University (0.67 to 1.02), readmission aOR 0.86 (0.75 to 0.993)) of Pennsylvania, Philadelphia, or non-­TNFi biologic/tsDMARD (mortality aOR 0.78 PA 19104, USA; (0.49 to 1.22), readmission aOR 1.02 (0.78 to 1.33)). patients undergoing major surgery, in which post- Michael.​ ​George@uphs.​ ​upenn.​ Analyses stratified by surgery category were similar. Risk edu operative complications such as pneumonia and of mortality and readmission was higher with 5–10 mg/ wound infections can have substantial morbidity. Received 9 December 2019 day of glucocorticoids (mortality aOR 1.41 (1.08 to Perioperative immunosuppression management Revised 4 March 2020 1.82), readmission aOR 1.26 (1.05 to 1.52)) or >10 mg/ has been studied primarily in patients with RA Accepted 5 March 2020 day (mortality aOR 1.64 (1.02 to 2.64), readmission aOR Published Online First undergoing arthroplasty. Patients with RA are at 24 March 2020 1.60 (1.15 to 2.24)) versus no glucocorticoids, although higher risk of postoperative infections after arthro- results varied when stratifying by surgery category. plasty, with disease activity, comorbidities and Conclusions Recent biologic or tsDMARD use was immunosuppression all potentially contributing.1–5 not associated with a greater risk of mortality or Conventional disease-modifying­ antirheumatic readmission after hip fracture, abdominopelvic or cardiac drugs (csDMARDs) such as methotrexate have not surgery compared with methotrexate. Higher dose been associated with increased postoperative risk.6 glucocorticoids were associated with greater risk. © Author(s) (or their The risk with biological therapies remains uncer- employer(s)) 2020. No commercial re-­use. See rights tain, although studies have suggested that stop- and permissions. Published ping a biologic before arthroplasty does not reduce by BMJ. Introduction infection risk and have shown no differences in 7–9 To cite: George MD, Patients with rheumatoid arthritis (RA) and other risk across different biologics. Several studies, Baker JF, Winthrop KL, autoimmune conditions are frequently treated with however, have demonstrated increased infection et al. Ann Rheum Dis immunosuppression, which can increase the risk of risk and readmission after arthroplasty with even 2020;79:573–580. infection. Infections are of particular concern in modest glucocorticoid doses.1 4 10–12

George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 573 Rheumatoid arthritis

Few studies, however, have evaluated the contribution of receive other csDMARDs (hydroxychloroquine, sulfasalazine or specific immunosuppressive therapies to outcomes after other leflunomide). types of surgery.13 14 These surgeries have different postopera- We separately evaluated associations between glucocorticoids tive risks than arthroplasty and are often non-elective,­ with no and outcomes. Glucocorticoid dose in the 90 days prior to opportunity for holding medications before surgery. surgery was based on oral prescriptions for prednisone, pred- The goal of this study was to evaluate associations between nisolone and methylprednisolone, using prescribed dose in biologic use and adverse outcomes in patients with RA under- prednisone equivalents and days supply to determine each daily going common major surgical operations other than arthroplasty, dose and truncating prescriptions if a new prescription was filled including hip fracture, abdominopelvic and cardiac surgeries. A before the prescription end date. To account for differences secondary goal was to evaluate associations between glucocorti- between prescription instructions and patient behaviour (eg, a coids and adverse outcomes. 30-­day prescription taken over 90 days), we averaged the daily dose over 90 days, categorising the average daily dose as ≤5 mg, >5 to 10 mg or >10 mg/day). Methods In secondary analyses, rather than compare biologic users to Using Medicare claims data 2006–2015, we identified adults≥18 non-­users, we directly evaluated the timing of biologics before years old with RA based on ≥2 International Classifications of surgery similar to previous studies.7 8 Here, we evaluated only Diseases (ICD-9) provider diagnoses ≥7 days apart, use of a patients receiving the two most common infusions (infliximab DMARD, and absence of psoriatic arthritis, ankylosing spondy- or intravenous abatacept), since the timing of treatment for infu- litis and inflammatory bowel disease diagnoses in the year before sions can be precisely determined in this data source. We evalu- surgery. Included patients underwent one of the following ated whether the time between the last infusion and surgery (<4 surgeries between 1 January 2007 and 31 August 2015 based weeks, ≥4 to <8 weeks or ≥8 weeks to 6 months) was associ- on primary ICD-9 procedure codes from inpatient hospitalisa- ated with different risk of postoperative outcomes. tions using established definitions: hip fracture repair, abdom- inopelvic surgeries (cholecystectomy, hysterectomy, hernia surgery, appendectomy, colectomy for diverticular disease) and Outcomes cardiac surgeries (coronary artery bypass graft (CABG), mitral The primary outcomes of interest were (1) mortality within or aortic valve surgery). We required that surgery occur by 90 days after surgery (including the index hospitalisation) hospital day 3 for hip fracture, hysterectomy, hernia surgery or (2) readmission to an acute care hospital within 30 days and appendectomy and hospital day 7 for the other surgeries to after discharge. Readmission was identified among patients avoid surgeries in which an outcome may have occurred prior to discharged to home, home healthcare, acute rehabilitation or surgery or in which the surgical procedure might be the results a skilled nursing facility, considering rehospitalisations within of a complicated hospitalisation, rather than the main indica- 1 day of discharge to be continuations of the initial hospitalisa- tion for hospitalisation. More than 90% of surgeries fell within tion and not counting hospitalisations with a primary diagnosis 15 these windows. We also applied other restrictions to ensure that indicating rehabilitation. Reasons for readmission were based surgeries were performed for specific indications and were not on primary discharge diagnosis codes classified using Chronic 16 17 part of a larger procedure (online supplementary table S1). We Conditions Warehouse software. further subdivided certain applicable surgeries (cholecystectomy, Secondary outcomes included postoperative pneumonia 18 hernia, colectomy, CABG and valve surgery) into elective and and wound complications based on diagnosis codes from the non-­elective surgery, as postoperative risks are different after index hospitalisation or any acute care hospitalisations within elective and non-­elective surgery. All elective surgeries were 30 days after surgery (online supplementary table S1) because required to occur by hospital day 2 with no admission through of specific concerns about these outcomes with biologics and 19–21 the emergency department, transfer from another acute care glucocorticoids. hospital or admission status of ‘emergent’. Hip fracture repair and appendectomy were all considered non-­elective, and all Covariates included hysterectomies were elective. Covariates measured at baseline included demographics, zip We required 1 year of continuous enrolment in Medicare code-­based median household income from the American A, B and D prior to surgery (baseline; online supplementary Community Survey 2009–2013,22 disability status, skilled nursing figure S1). We excluded patients with a diagnosis of malignancy facility residence, surgery type, admission type (elective, urgent, (except non-­melanoma skin cancer) or HIV in the past year, emergent) and hospital day of surgery. We identified comorbid patients receiving multiple different biologics/targeted synthetic conditions, an adaption of the Charlson Comorbidity Index,23 DMARDs (tsDMARDs) within 6 months before surgery and and measures of healthcare utilisation (outpatient visits, emer- patients with another major surgery within 6 months. Patients gency department visits, hospitalisations and hospitalised infec- could contribute multiple surgeries if >6 months apart. tion) during the baseline period. Medication use based on filled prescriptions≤90 before surgery days included non-­steroidal Exposures anti-­inflammatories, opioids, non-­methotrexate csDMARDs and To evaluate associations between immunosuppression use and antibiotics. Number of previous biologics was determined using postoperative outcomes, we classified patients in mutually all data prior to the surgery date (mean 4 years). exclusive groups: (1) methotrexate prescription fill <8 weeks before surgery with no biologic/tsDMARD <6 months before Statistical analysis surgery (to avoid recent biologic discontinuations), (2) TNFi Associations between exposures and outcomes were evaluated prescription/infusion<8 weeks before surgery with or without using logistic regression using propensity score-­derived inverse methotrexate, or (3) non-­TNFi biologic prescription/infusion or probability weights to balance all measured covariates across tsDMARD prescription fill <8 weeks before surgery (16 weeks exposure groups, as described below. Cluster robust standard for rituximab), with or without methotrexate. All groups could errors accounted for patients contributing multiple surgeries.24

574 George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 Rheumatoid arthritis

The same method was used to generate inverse probability treatment weights in glucocorticoid analyses, also including methotrexate use, biologic use and time of last biologic or methotrexate before surgery in propensity score models. In secondary analyses, associations between infliximab timing or intravenous abatacept timing and outcomes were assessed using multivariable logistic regression models with potential confounders included as covariates. Analyses were performed using Stata V.15.1 (StataCorp). Preliminary results were previously presented.30

Results We identified 10 777 surgeries among 10 483 patients meeting inclusion and exclusion criteria (figure 1). Mean age was 72 and 80% of patients were female (table 1, online supplemen- tary table S2). RA treatment before surgery was methotrexate without a biologic/tsDMARD in 6194 (57%), TNFi in 3523 (33%) (709 adalimumab, 88 certolizumab, 972 etanercept, 46 golimumab, 1708 infliximab), and non-­TNFi biologic/ tsDMARD in 1060 (10%) (701 abatacept, 203 rituximab, 127 tocilizumab, 29 tofacitinib). Surgery type was hip frac- ture repair in 33%, abdominopelvic surgery in 47% (most commonly cholecystectomy) and cardiac surgery in 20%. Patients receiving a TNFi or non-­TNFi biologic/tsDMARD were younger, had more outpatient visits, were more likely to have received prior biologics or be disabled and were less likely to have been in a skilled nursing facility. After applying inverse probability weights, patient characteristics were well balanced between methotrexate and TNFi groups with only slight residual imbalance in the non-­TNFi biologic/tsDMARD Figure 1 Cohort selection. *See online supplementary table 1 for group (online supplementary table S2). surgery specific exclusions. †Includes 2208 patients who received Death occurred within 90 days of surgery in 577 (5.4%) infliximab and 827 who received intravenous abatacept <6 months patients and readmission within 30 days of discharge in 1282 before surgery, used in timing analyses. AS, ankylosing spondylitis; IBD, (12.8%) of 9977 surgeries with qualifying discharge. Outcome inflammatory bowel disease; PsA, psoriatic arthritis; RA, rheumatoid rates varied substantially by surgery type and were highest arthritis; tsDMARD, targeted synthetic disease-­modifying antirheumatic after non-­elective colectomy or valve surgery and lowest after drug. hysterectomy (table 2). The most common reasons for readmis- sion were complication of the procedure (18.6%), septicaemia Analyses were first conducted in the overall population and (6.6%) and congestive heart failure (6.0%; online supplemen- then reanalysed (with recalculated propensity scores) separately tary table S3). for each surgery category (hip fracture, abdominopelvic and There was no significant difference in 90-day­ mortality among cardiac). patients receiving a TNFi (aOR 0.83 (0.67 to 1.02)) or non-­TNFi Propensity scores were generated based on the probability biologic/tsDMARD (aOR 0.78 (0.49 to 1.22)) versus patients of being in each treatment group (methotrexate, TNFi, non-­ receiving methotrexate without a biologic/tsDMARD (table 3). TNFi biologic/tsDMARD) using multinomial logistic regression Compared with methotrexate, the risk of 30-day­ readmission models with all covariates of interest including a squared term was somewhat lower in patients receiving a TNFi (aOR 0.86 for age (to account for nonlinearity) and average glucocorticoid (0.75 to 0.993)) and was similar in those receiving a non-­TNFi/ dose before surgery (included variables in online supplementary tsDMARD (aOR 1.02 (0.78 to 1.33)). Results were similar in 10 25 26 table S2). Propensity scores were used to create stabilised sensitivity analyses excluding patients with previous biologic use inverse probability treatment weights truncated at the 1st and (online supplementary table S4). 26–29 99th percentile. Balance of covariates across treatment cate- Unadjusted 90-­day mortality and 30-­day readmission were gories was assessed compared with the reference group (meth- greater for patients receiving higher doses of glucocorticoids otrexate), with standardised mean difference ≤0.1 indicating (table 3). Results from adjusted analyses were attenuated but good balance. Unbalanced covariates were added as covariates still showed significantly higher risk of 90-­day mortality among to weighted models. Propensity scores and stabilised inverse patients receiving an average of 5–10 mg/day (aOR 1.41 (1.08 probability treatment weights were recalculated in analyses with to 1.82)) or >10 mg/day of glucocorticoids (aOR 1.64 (1.02 readmission as the outcome and in analyses of specific surgery to 2.64)), and significantly higher risk of 30-­day readmission categories. Discharge status was considered an intermediate vari- among patients receiving 5–10 mg/day (aOR 1.26 (1.05 to 1.52)) able and was not included in readmission propensity scores. To or >10 mg/day of glucocorticoids (aOR 1.60 (1.15 to 2.24)) address the possibility that methotrexate-­treated patients with versus no glucocorticoids. previous biologic treatment >6 months before surgery were When hip, abdominopelvic and cardiac surgeries were eval- more ill, we conducted a sensitivity analysis excluding patients uated independently, we found no significant differences with previous biologic use. in 90-­day mortality or 30-­day readmission with a TNFi or

George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 575 Rheumatoid arthritis

Table 1 Select cohort characteristics Methotrexate TNF Non-­TNF biologic/tsDMARD (n=6194) (n=3523) SMD pre SMD post (n=1060) SMD pre SMD post Total Age 74.2±10.7 70.0±11.6 −0.37 −0.01 70.3±11.3 −0.36 −0.12 72.5±11.3 Female 4907 (79.2) 2852 (81.0) 0.04 0.01 867 (81.8) 0.07 0.05 8626 (80.0) Race White 5040 (81.4) 2648 (75.2) −0.15 −0.01 831 (78.4) −0.07 −0.06 8519 (79.0) Black 397 (6.4) 252 (7.2) 0.03 0.00 62 (5.8) −0.02 0.06 711 (6.6) Hispanic 371 (6.0) 343 (9.7) 0.14 0.00 68 (6.4) 0.02 −0.02 782 (7.3) Other/unknown 386 (6.2) 280 (7.9) 0.07 0.02 99 (9.3) 0.12 0.05 765 (7.1) Disability Not disabled 4250 (68.6) 1977 (56.1) −0.26 0.00 602 (56.8) −0.25 −0.09 6829 (63.4) Disabled age <65 872 (14.1) 868 (24.6) 0.27 0.00 236 (22.3) 0.21 0.04 1976 (18.3) Disabled age ≥65 1072 (17.3) 678 (19.2) 0.05 0.00 222 (20.9) 0.09 0.07 1972 (18.3) Year 2011–2015* 3307 (53.4) 1849 (52.5) −0.04 0.01 749 (70.7) 0.39 0.22 5905 (54.8) Surgery type Hip fracture 2213 (35.7) 1050 (29.8) −0.13 −0.01 322 (30.4) −0.11 −0.05 3585 (33.3) Cholecystectomy elective 210 (3.4) 151 (4.3) 0.05 −0.01 36 (3.4) 0.00 −0.02 397 (3.7) Cholecystectomy non-­elective 1020 (16.5) 660 (18.7) 0.06 0.02 191 (18.0) 0.04 0.00 1871 (17.4) Hysterectomy 397 (6.4) 349 (9.9) 0.13 0.00 80 (7.5) 0.05 0.02 826 (7.7) Hernia elective 282 (4.6) 151 (4.3) −0.01 0.00 47 (4.4) −0.01 0.03 480 (4.5) Hernia non-­elective 216 (3.5) 148 (4.2) 0.04 0.00 30 (2.8) −0.04 0.03 394 (3.7) Appendectomy 180 (2.9) 129 (3.7) 0.04 0.01 48 (4.5) 0.09 0.01 357 (3.3) Colectomy elective 160 (2.6) 69 (2.0) −0.04 −0.01 25 (2.4) −0.01 0.00 254 (2.4) Colectomy non-­elective 251 (4.1) 135 (3.8) −0.01 0.01 60 (5.7) 0.08 −0.01 446 (4.1) CABG elective 229 (3.7) 155 (4.4) 0.04 0.00 37 (3.5) −0.01 −0.01 421 (3.9) CABG non-­elective 523 (8.4) 252 (7.2) −0.05 0.00 83 (7.8) −0.02 −0.04 858 (8.0) Valve elective 273 (4.4) 133 (3.8) −0.03 0.01 47 (4.4) 0.00 0.03 453 (4.2) Valve non-­elective 240 (3.9) 141 (4.0) 0.01 0.00 54 (5.1) 0.06 0.08 435 (4.0) Average glucocorticoid dose None 3248 (52.4) 1984 (56.3) 0.08 0.00 488 (46.0) −0.13 0.00 5720 (53.1) ≤5 1851 (29.9) 962 (27.3) −0.06 −0.01 324 (30.6) 0.02 −0.04 3137 (29.1) 5–10 837 (13.5) 442 (12.5) −0.03 0.00 180 (17.0) 0.10 0.01 1459 (13.5) >10 258 (4.2) 135 (3.8) −0.02 0.01 68 (6.4) 0.10 0.05 461 (4.3) Previous biologics 0 5486 (88.6) 2949 (83.7) −0.14 −0.02 390 (36.8) −1.27 −0.11 8825 (81.9) 1 543 (8.8) 448 (12.7) 0.13 0.01 461 (43.5) 0.86 0.07 1452 (13.5) >1 165 (2.7) 126 (3.6) 0.05 0.03 209 (19.7) 0.56 0.05 500 (4.6) NSAIDs 1600 (25.8) 957 (27.2) 0.03 0.00 255 (24.1) −0.04 −0.03 2812 (26.1) Opioids 2986 (48.2) 1860 (52.8) 0.09 0.00 612 (57.7) 0.19 0.15 5458 (50.6) Charlson Score 3.0+/-2.8 2.7+/-2.8 −0.09 0.01 3.1+/-2.9 0.04 0.03 2.9+/-2.8 Extra-articular­ RA 127 (2.1) 92 (2.6) 0.04 0.00 42 (4.0) 0.11 0.08 261 (2.4) Diabetes 1541 (24.9) 834 (23.7) −0.03 0.00 288 (27.2) 0.05 0.02 2663 (24.7) Asthma/COPD 1346 (21.7) 717 (20.4) −0.03 0.00 208 (19.6) −0.05 0.03 2271 (21.1) Hospitalisations past year 0 3925 (63.4) 2365 (67.1) 0.08 0.00 698 (65.8) 0.05 −0.04 6988 (64.8) 1–2 1454 (23.5) 758 (21.5) −0.05 0.00 231 (21.8) −0.04 0.02 2443 (22.7) >2 815 (13.2) 400 (11.4) −0.06 0.00 131 (12.4) −0.02 0.02 1346 (12.5) Hospitalised infection past year 1062 (17.1) 500 (14.2) −0.08 0.00 153 (14.4) −0.07 0.03 1715 (15.9) Outpatient visits 14.5±9.5 16.1±9.3 0.17 0.01 19.5±10.2 0.50 0.21 15.5±9.6 Skilled nursing facility stay past year 963 (15.5) 313 (8.9) −0.21 −0.01 74 (7.0) −0.27 −0.05 1350 (12.5) Mean ± SD or N (%) shown. Full cohort characteristics shown in online supplementary table S2. Standardised mean differences (SMD) are shown compared with the methotrexate-treated­ group, either before (pre) or after (post) inverse probability weighting, with SMD >0.1 indicating imbalance. Characteristics with SMD >0.1 after weighting are shown in bold and were added as covariates to weighted models. *Year modelled as continuous and SMD shown indicates standardised difference in means across groups. CABG, coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; NSAID, nonsteroidal anti-inflammatory­ drugs; RA, rheumatoid arthritis; TNF, tumour necrosis factor; tsDMARD, targeted synthetic disease-­modifying antirheumatic drugs. non-­TNFi biologic/tsDMARD versus methotrexate for any of readmission was not significantly different across glucocorti- the surgery categories (figure 2). Associations between gluco- coid doses in these patients. In contrast, patients undergoing hip corticoid use and adverse outcomes varied by surgery category fracture repair had no significant difference in 90-day­ mortality (figure 3). Associations with mortality were most pronounced across glucocorticoid doses, but did have greater risk of 30-day­ among patients undergoing cardiac surgery although 30-day­ readmission with higher glucocorticoid doses.

576 George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 Rheumatoid arthritis

Table 2 Outcomes by surgery type N 90-­day mortality 30-­day readmission* Pneumonia Wound complication Hip fracture surgery 3585 240 (6.7%) 350/3142 (11.1%) 206 (5.7%) 47 (1.3%) Abdominopelvic surgery Appendectomy 357 ≤11 32/346 (9.2%) ≤11 ≤11 Cholecystectomy elective 397 ≤11 38/394 (9.6%) 14 (3.5%) ≤11 Cholecystectomy non-­elective 1871 60 (3.2%) 239/1794 (13.3%) 109 (5.8%) 40 (2.1%) Colectomy elective 254 ≤11 35/253 (13.8%) ≤11 15 (5.9%) Colectomy non-­elective 446 68 (15.2%) 85/374 (22.7%) 65 (14.6%) 63 (14.1%) Hernia elective 480 ≤11 48/480 (10.0%) ≤11 26 (5.4%) Hernia non-­elective 394 21 (5.3%) 42/374 (11.2%) 28 (7.1%) 17 (4.3%) Hysterectomy 826 ≤11 49/823 (6.0%) ≤11 14 (1.7%) Cardiac surgery CABG elective 421 ≤11 62/411 (15.1%) 20 (4.8%) ≤11 CABG non-­elective 858 55 (6.4%) 127/759 (16.7%) 87 (10.1%) 49 (5.7%) Valve elective 453 24 (5.3%) 93/442 (21.0%) 24 (5.3%) 15 (3.3%) Valve non-­elective 435 63 (14.5%) 82/385 (21.3%) 47 (10.8%) 14 (3.2%) Total 10 777 577 (5.4%) 1282/9977 (12.8%) 627 (5.8%) 328 (3.0%) Cell sizes≤11 suppressed. *Readmission within 30 days of discharge reported among patients discharged to home, home health, skilled nursing facility or inpatient rehabilitation. CABG, coronary artery bypass graft

We found no significant difference in the risk of secondary <8 weeks or ≥8 weeks before surgery after adjusting for poten- outcomes of hospitalised pneumonia or hospitalised wound tial confounders (online supplementary table S6). complication within 30 days of surgery in patients receiving a TNFi or non-­TNFi biologic/tsDMARD versus patients receiving Discussion methotrexate (online supplementary table S5). Higher dose In this large observational study of patients with RA under- glucocorticoids were associated with greater risk of wound going hip fracture, abdominopelvic or cardiac surgery, we complications (aOR 1.72 (1.24 to 2.39) for 5–10 mg/day, aOR found no increased risk of 90-day­ mortality, 30-day­ readmis- 1.68 (0.97 to 2.94) for >10 mg/day). sion, wound complications or postoperative pneumonia versus Among 2208 patients who received infliximab and 827 who patients receiving methotrexate without a biologic/tsDMARD. received intravenous abatacept within 6 months of surgery, there Conversely, even low-­dose glucocorticoid use (5–10 mg/day) was was no significant difference in the risk of mortality or readmis- associated with greater risk of adverse outcomes. sion in patients who received their last infusion <4 weeks before Although previous studies of perioperative immunosup- surgery compared with those who received the last infusion 4 to pression in patients undergoing the surgeries we examined are

Table 3 Associations between biologics/targeted synthetic disease-modifying­ antirheumatic drug (tsDMARDs) and glucocorticoids with mortality and readmission 90-­day mortality 30-­day readmission Propensity-­ Propensity-­ Unadjusted OR weighted OR Unadjusted OR weighted OR N Mortality N (%) (95% CI) (95% CI) Readmission N (%)* (95% CI) (95% CI) Methotrexate versus biologics/tsDMARDs Methotrexate 6194 401 (6.5%) Reference Reference 781/5690 (13.7%) Reference Reference TNF 3523 134 (3.8%) 0.57 0.83 367/3295 (11.1%) 0.79 0.86 (0.47 to 0.70) (0.67 to 1.02) (0.69 to 0.90) (0.75 to 0.993) Non- TNF­ biologic/ 1060 42 0.60 0.78 134/992 0.98 1.02 tsDMARD (4.0%) (0.43 to 0.82) (0.49 to 1.22) (13.5%) (0.81 to 1.20) (0.78 to 1.33) Glucocorticoids† None 5720 252 (4.4%) Reference Reference 596/5329 (11.2%) Reference Reference ≤5 mg/day 3137 181 (5.8%) 1.33 1.20 386/2893 (13.3%) 1.22 1.10 (1.09 to 1.62) (0.98 to 1.48) (1.07 to 1.40) (0.96 to 1.27) 5–10 mg/day 1459 106 (7.3%) 1.70 1.41 216/1336 (16.2%) 1.53 1.26 (1.34 to 2.15) (1.08 to 1.82) (1.29 to 1.81) (1.05 to 1.52) >10 mg/day 461 38 1.95 1.64 84/419 1.99 1.60 (8.2%) (1.37 to 2.78) (1.02 to 2.64) (20.0%) (1.54 to 2.57) (1.15 to 2.24) Propensity-­weighted OR from inverse probability-­weighted logistic regression models with separate analyses for methotrexate versus biologics/tsDMARDs and for glucocorticoids. Propensity score models include all covariates shown in online supplementary table S2. Glucocorticoid propensity score models also include use of methotrexate or type of biologic/tsDMARD and time of last prescription/infusion before surgery. Unbalanced covariates are included in weighed models. *Readmission within 30 days of discharge reported among patients discharged to home, home health, skilled nursing facility or inpatient rehabilitation †Glucocorticoid dose is the average daily dose in the 3 months prior to surgery. TNF, tumour necrosis factor inhibitor.

George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 577 Rheumatoid arthritis

Figure 2 Associations of biologic use with mortality and readmission by surgery type. OR from inverse probability-weighted­ logistic regression models. Propensity score models include the same variables as shown in online supplementary table S2 and are recalculated for each surgery type with unbalance covariates added to weighted models. Abd, abdominopelvic surgery; Cardiac, cardiac surgery; Hip, hip fracture surgery; TNF, tumour necrosis factor inhibitor; tsDMARD, targeted synthetic disease-modifying­ antirheumatic drug. sparse,14 our results showing no significant increase in adverse adjusting for potential confounders. Although biologic-treated­ outcomes in patients receiving a biologic are in line with data patients may have more severe RA (in our study biologic-treated­ from studies of joint replacement surgery. These studies showed patients were more likely to be disabled), biologic-­treated no difference in postoperative risk in patients who stopped versus patients also tend to be younger and have fewer comorbidi- continued biologics before surgery.7–9 Interestingly, in our study, ties,31 as we found here. The small differences in risk favouring biologic-­treated patients had lower rates of adverse outcomes biologic-­treated patients after adjusting for confounders likely than patients receiving methotrexate without a biologic before reflect some residual confounding rather than risks associated with methotrexate or protective effects of biologics—reassur- ingly, we found no difference in outcomes in patients receiving a biologic with versus without methotrexate (data not shown). As we previously observed for arthroplasty,7 8 the lack of association between biologic use and adverse outcomes in this study suggests that holding biologics before the surgeries we examined may not be necessary. We required that all biologic-­ treated patients receive an infusion or fill a prescription for their biologic within 8 weeks of surgery (16 weeks of rituximab), excluding patients without evidence of recent biologic exposure. Although some patients may still have stopped their biologic before surgery, the majority of our surgeries were non-elective­ surgeries, in which patients would not have the opportunity to stop treatment. Additionally, we directly evaluated biologic timing among patients receiving infliximab or intravenous abatacept, for whom we could precisely identify timing based on infusion dates, and found no evidence that patients who received infliximab or intravenous abatacept within 4 weeks of surgery had greater risk of mortality or readmission compared with patients with a longer time between infusion and surgery. Nonetheless, our results should be interpreted cautiously since Figure 3 Associations of glucocorticoid use with mortality and we cannot rule out the possibility of small benefits with biologic readmission by surgery type. OR from inverse probability-­weighted interruptions on the outcomes we studied or impacts on the logistic regression models. Propensity score models include the same frequency or severity of other postoperative outcomes. variables as shown in table 2 and are recalculated for each surgery Our results suggest that necessary non-­elective surgeries need type with unbalance covariates added to weighted models. Abd, not be delayed in patients receiving biological therapy, since abdominopelvic surgery; Cardiac, cardiac surgery; Hip, hip fracture delaying these surgeries has its own consequences. For elective surgery. surgery, a common practice has been to hold biologics before

578 George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 Rheumatoid arthritis surgery. While our results show no obvious benefit to treatment Contributors All authors met criteria for authorship, critically revised the interruptions, we also did not find obvious risks with short inter- manuscript, gave approval of the final version to be published and agree to be ruptions. Prolonged interruptions beyond one interval are likely accountable for all aspects of the work related to accuracy or integrity of the results. Specific additional roles included: conception and study design (MG, JFB, KLW, JRC), not warranted. When deciding on shorter interruptions (eg, one data acquisition (MG, LC, QW, FX), analysis (MG, JFB, KLW, JRC), interpretation (MG, dosing interval), clinicians should consider the risks and benefits, JFB, KLW, SDG, EA, LC, FX, JRC), and drafting the original manuscript (MG, JFB, JRC). including the potential need for glucocorticoids if patients have Funding MG is supported by the National Institute of Arthritis and Musculoskeletal a disease flare. and Skin Diseases 1K23AR073931-01. JFB is supported by a VA Clinical Science In analyses with all surgery types combined, we found a Research and Development Merit Award (I01 CX001703). dose-­dependent association between glucocorticoid use and Competing interests MG has received a research grant from Bristol-­Myers Squibb mortality, readmission and wound complications. The difference and the National Institutes of Health and consulting fees from AbbVie. JFB has between unadjusted and adjusted results did suggest substantial received consulting fees from Bristol-­Myers Squibb and Gilead. KLW has received research grants from Pfizer and Bristol-­Myers Squibb, and consulting fees from Pfizer, confounding; we cannot definitively establish causality and the AbbVie, UCB, Lilly, Galapagos, GSK, Roche, and Gilead. EA is an employee of Bristol-­ potential for residual confounding remains. Additionally, results Myers Squibb. JRC has received research grants from the Patient Centered Outcomes were more variable when stratified by surgery type. Whether Research Institute and research grants and consulting fees from Bristol-­Myers these results reflect differences in glucocorticoid risk in different Squibb, Amgen, AbbVie, Corrona, Janssen, Lilly, Myriad, Pfizer, UCB and Regeneron. surgery types or whether these results are due to chance in these Patient and public involvement Patients and/or the public were not involved in smaller subgroups is uncertain. Glucocorticoids have been asso- the design, or conduct, or reporting, or dissemination plans of this research. ciated with poor wound healing32 and previous studies demon- Patient consent for publication Not required. strated a dose-­dependent association between glucocorticoid Provenance and peer review Not commissioned; externally peer reviewed. 33–35 use and adverse outcomes in the non-­operative setting and Data availability statement Data may be obtained from a third party (Centers 7 10 after orthopaedic surgery. We averaged glucocorticoid dose for Medicare & Medicaid Services) and are not publicly available. Statistical code is in the 90 days prior to surgery and could not evaluate dose available from the authors upon request. changes immediately before surgery or doses during the hospi- ORCID iDs talisation, but our results suggest that glucocorticoid use in the Michael D George http://orcid.​ ​org/0000-​ ​0002-​0398-​2308 months before surgery, especially >10 mg/day, should remain Kevin L Winthrop http://orcid.​ ​org/0000-​ ​0002-​3892-​6947 a concern in patients undergoing major surgery and may be a more important contributor to infection than more targeted References immunosuppression. 1 Cordtz RL, Zobbe K, Højgaard P, et al. Predictors of revision, prosthetic joint Several limitations are inherent to this type of observational infection and mortality following total hip or total knee arthroplasty in patients with study. Outcomes from claims data may miss or misclassify rheumatoid arthritis: a nationwide cohort study using Danish healthcare registers. 36 Ann Rheum Dis 2018;77:281–8. specific postoperative complications ; we instead focused on 2 stundner O, Danninger T, Chiu Y-­L, et al. Rheumatoid arthritis vs osteoarthritis in mortality and readmission which can be accurately identified patients receiving total knee arthroplasty: perioperative outcomes. J Arthroplasty in claims data. We evaluated postoperative pneumonia and 2014;29:308–13. wound complications but could not fully assess morbidity or 3 Ravi B, Escott B, Shah PS, et al. A systematic review and meta-­analysis comparing complications following total joint arthroplasty for rheumatoid arthritis versus for minor complications (eg, delayed wound healing) not leading osteoarthritis. Arthritis Rheum 2012;64:3839–49. to hospitalisation. Although typically biologic users would be 4 salt E, Wiggins AT, Rayens MK, et al. Moderating effects of immunosuppressive expected to have more severe disease, we could not measure medications and risk factors for post-­operative joint infection following total joint disease activity, which could be less well controlled in some arthroplasty in patients with rheumatoid arthritis or osteoarthritis. Semin Arthritis non-­biologic treated patients. We also excluded patients with Rheum 2017;46:423–9. 5 schnaser EA, Browne JA, Padgett DE, et al. Perioperative complications in patients recent changes in therapy who might have the highest disease with inflammatory arthropathy undergoing total hip arthroplasty. J Arthroplasty activity. We cannot rule out the possibility that patients receiving 2016;31:2286–90. biologics were selected differently for surgery (especially elective 6 Grennan DM, Gray J, Loudon J, et al. Methotrexate and early postoperative surgery) or were treated or monitored differently after surgery. complications in patients with rheumatoid arthritis undergoing elective orthopaedic surgery. Ann Rheum Dis 2001;60:214–7. Additionally, biologics may be avoided completely in patients 31 7 George MD, Baker JF, Hsu JY, et al. Perioperative timing of infliximab and the risk who are frail or have comorbidities or survivorship bias could of serious infection after elective hip and knee arthroplasty. Arthritis Care Res lead to a healthier biologic group, although we controlled for 2017;69:1845–54. multiple comorbidities and measures of healthcare utilisation 8 George MD, Baker JF, Winthrop K, et al. Timing of abatacept before elective and attained good balance for measured covariates between arthroplasty and risk of postoperative outcomes. Arthritis Care Res 2019;71:1224–33. 9 abou Zahr Z, Spiegelman A, Cantu M, et al. Perioperative use of anti-­rheumatic exposure groups. Glucocorticoid results were more variable agents does not increase early postoperative infection risks: a veteran Affairs’ and there is even greater potential for unmeasured or residual administrative database study. Rheumatol Int 2015;35:265–72. confounding in these analyses. Some contributors to postop- 10 George MD, Baker JF, Winthrop K, et al. Risk of biologics and glucocorticoids in erative outcomes such as surgeon experience, severity of the patients with rheumatoid arthritis undergoing arthroplasty: a cohort study. Ann Intern Med 2019;170:825. presentation and difficulty of the surgery could not be directly 11 Michaud K, Fehringer EV, Garvin K, et al. Rheumatoid arthritis patients are not at measured, although these factors would not be expected to differ increased risk for 30-­day cardiovascular events, infections, or mortality after total joint in biologic users versus non-­users. Finally, although we analysed arthroplasty. Arthritis Res Ther 2013;15:R195. surgeries in groups that would be susceptible to similar types of 12 Gilson M, Gossec L, Mariette X, et al. Risk factors for total joint arthroplasty infection complications, numbers were not sufficient to evaluate specific in patients receiving tumor necrosis factor α-blockers: a case-­control study. Arthritis Res Ther 2010;12:R145. surgeries (eg, appendectomy, cholecystectomy) separately. 13 sims SM, Kao AM, Spaniolas K, et al. Chronic immunosuppressant use in colorectal In conclusion, neither use of biologics nor biologic timing cancer patients worsens postoperative morbidity and mortality through septic before surgery were associated with increased mortality or read- complications in a propensity-­matched analysis. Colorectal Dis 2019;21:156–63. mission after hip fracture, abdominopelvic or cardiac surgery. 14 Waterman M, Xu W, Dinani A, et al. Preoperative biological therapy and short-­term outcomes of abdominal surgery in patients with inflammatory bowel disease. Gut Higher dose glucocorticoids were associated with adverse 2013;62:387–94. outcomes after these surgeries, suggesting that minimising gluco- 15 Centers for Medicare & Medicaid Services (CMS). 2016 Procedure-­Specific corticoids should be a focus of perioperative medicine. measures updates and specifications report hospital-­level 30-­day Risk-­Standardized

George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 579 Rheumatoid arthritis

readmission measures. elective primary total hip arthroplasty (THA) and/or total knee 26 Feng P, Zhou X-­H, Zou Q-­M, et al. Generalized propensity score for estimating the arthroplasty (TKA) – version 5.0, 2016. Available: https://www.​cms.​gov/​Medicare/​ average treatment effect of multiple treatments. Stat Med 2012;31:681–97. Quality-​Initiatives-​Patient-​Assessment-​Instruments/​HospitalQualityInits/​Measure-​ 27 McCaffrey DF, Griffin BA, Almirall D, et al. A tutorial on propensity score Methodology.​html [Accessed 9 Jan 2018]. estimation for multiple treatments using generalized boosted models. Stat Med 16 HCUP clinical classifications software (CCS) for ICD-9-CM.­ Available: https://www.​ 2013;32:3388–414. hcup-us.​ ​ahrq.​gov/​toolssoftware/​ccs/​ccs.jsp#​ ​download [Accessed 17 Jul 2019]. 28 austin PC. The performance of different propensity-­score methods for estimating 17 lawson EH, Hall BL, Louie R, et al. Association between occurrence of a postoperative differences in proportions (risk differences or absolute risk reductions) in observational complication and readmission: implications for quality improvement and cost savings. studies. Stat Med 2010;29:2137–48. Ann Surg 2013;258:10–18. 29 Cole SR, Hernán MA. Constructing inverse probability weights for marginal structural 18 Van Arendonk KJ, Tymitz KM, Gearhart SL, et al. Outcomes and costs of elective surgery for diverticular disease: a comparison with other diseases requiring colectomy. models. Am J Epidemiol 2008;168:656–64. JAMA Surg 2013;148:316–21. 30 George MD, Baker JF, Winthrop K, et al. Risk of serious infection with long-­term use 19 Wolfe F, Caplan L, Michaud K. Treatment for rheumatoid arthritis and the risk of of low-­dose glucocorticoids in patients with rheumatoid arthritis. Arthritis Rheumatol hospitalization for pneumonia: associations with prednisone, disease-­modifying 2019;71. antirheumatic drugs, and anti-­tumor necrosis factor therapy. Arthritis Rheum 31 George MD, Sauer BC, Teng C-­C, et al. Biologic and glucocorticoid use after 2006;54:628–34. methotrexate initiation in patients with rheumatoid arthritis. J Rheumatol 20 Galloway JB, Mercer LK, Moseley A, et al. Risk of skin and soft tissue infections 2019;46:343–50. (including shingles) in patients exposed to anti-­tumour necrosis factor therapy: 32 Wang AS, Armstrong EJ, Armstrong AW. Corticosteroids and wound healing: clinical results from the British Society for rheumatology biologics register. Ann Rheum Dis considerations in the perioperative period. Am J Surg 2013;206:410–7. 2013;72:229–34. 33 Dixon WG, Abrahamowicz M, Beauchamp M-­E, et al. Immediate and delayed 21 aberra FN, Lewis JD, Hass D, et al. Corticosteroids and immunomodulators: impact of oral glucocorticoid therapy on risk of serious infection in older patients postoperative infectious complication risk in inflammatory bowel disease patients. with rheumatoid arthritis: a nested case-­control analysis. Ann Rheum Dis Gastroenterology 2003;125:320–7. 2012;71:1128–33. 22 U.S. census bureau, 2009-2013 5-­year American community survey. Available: https:// 34 Curtis JR, Winthrop K, O’Brien C, O’Brien C, et al. Use of a baseline risk score to www.census.​ ​gov/programs-​ ​surveys/acs​ [Accessed 3 Oct 2015]. identify the risk of serious infectious events in patients with rheumatoid arthritis 23 Gagne JJ, Glynn RJ, Avorn J, et al. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol 2011;64:749–59. during certolizumab pegol treatment. Arthritis Res Ther 2017;19:276. 24 Colin Cameron A, Miller DL. A Practitioner’s Guide to Cluster-­Robust Inference. J Hum 35 Cohen SB, Tanaka Y, Mariette X, et al. Long-­term safety of tofacitinib for the treatment Resour 2015;50:317–72. of rheumatoid arthritis up to 8.5 years: integrated analysis of data from the global 25 spreeuwenberg MD, Bartak A, Croon MA, et al. The multiple propensity score as clinical trials. Ann Rheum Dis 2017;76:1253–62. control for bias in the comparison of more than two treatment arms: an introduction 36 steinberg SM, Popa MR, Michalek JA, et al. Comparison of risk adjustment from a case study in mental health. Med Care 2010;48:166–74. methodologies in surgical quality improvement. Surgery 2008;144:662–9.

580 George MD, et al. Ann Rheum Dis 2020;79:573–580. doi:10.1136/annrheumdis-2019-216802 Rheumatoid arthritis

Epidemiological science Risk of breast cancer before and after rheumatoid arthritis, and the impact of hormonal factors Hjalmar Wadström ‍ ‍ ,1 Andreas Pettersson,1 Karin E Smedby,1,2 Johan Askling1,3

Handling editor Josef S Abstract Key messages Smolen Objectives To examine the risk of incident breast cancer in women with rheumatoid arthritis (RA), and ►► Additional material is What is already known about this subject? published online only. To view the risk of RA in women with a history of breast cancer, ►► The life-­time risk of breast cancer is in the order please visit the journal online taking antihormonal treatment for breast cancer into of 1 woman out of 10. Several studies have (http://dx.​ ​doi.org/​ ​10.​1136/​ account. reported a 15%–20% reduced risk of breast annrheumdis-2019-​ ​216756). Methods Using nationwide Swedish registers, women cancer in women with rheumatoid arthritis (RA), 1 with new-­onset RA diagnosed in 2006–2016 were Clinical Epidemiology Division, but the reasons behind this decrease remains identified and analysed using a cohort and a case– Department of Medicine Solna, unknown and unstudied. Karolinska Institutet, Stockholm, control design. Each patient with RA was matched on ► Previous studies have reported that tamoxifen Sweden age, sex and place of residence to five randomly selected ► 2Patient Area Hematology, and aromatase inhibitor treatment of breast subjects from the general population. Through register Theme Cancer, Karolinska cancer may increase the risk of RA. University Hospital, Stockholm, linkages, we collected information on breast cancer, Sweden breast cancer risk factors (reproductive history and 3 What does this study add? Rheumatology, Theme hormone replacement therapy) and socio-­economy. The ► We demonstrate that the risk of breast cancer Inflammation andI nfection, relative risk of breast cancer after RA was assessed using ► Karolinska University Hospital, after rheumatoid arthritis (RA) is indeed Cox regression, and the relative risk of RA in women Solna, Sweden reduced, but this association cannot readily with a history of breast cancer was assessed using be explained by traditional breast cancer risk conditional logistic regression. Correspondence to factors. By studying the future risk of RA in Dr Hjalmar Wadström, Clinical Results The risk of incident breast cancer in women women with breast cancer, we demonstrate an Epidemiology Division, with RA was reduced and the association was not equally pronounced risk reduction before RA Department of Medicine Solna, attenuated by adjustment for breast cancer risk factors Karolinska Institutet, 171 77 diagnosis, again not explained by traditional (HR=0.80, 95% CI 0.68 to 0.93). The risk of RA in Solna, Sweden; breast cancer risk factors. These findings women with a history of breast cancer was similarly hjalmar.​ ​wadstrom@ki.​ ​se suggest that other factors, independent of RA, reduced (OR=0.87, 95% CI 0.79 to 0.95). Women drive the inverse association between the two Received 3 December 2019 with breast cancer treated with tamoxifen (OR=0.86, Revised 16 February 2020 diseases. 95% CI 0.62 to 1.20) or aromatase inhibitors (OR=0.97, Accepted 25 February 2020 ► Tamoxifen and aromatase inhibitors do not 95% CI 0.69 to 1.37) did not have an increased risk of ► Published Online First seem to increase the risk of future RA. 11 March 2020 RA compared with women with breast cancer treated differently. How might this impact on clinical practice or Conclusions The decreased occurrence of breast future developments? cancer in patients with RA is present already before RA ► With respect to patient counselling, the diagnosis; these reduced risks are not readily explained ► occurrence of a breast cancer in patients with by hormonal risk factors. Adjuvant antihormonal therapy rheumatoid arthritis (RA) is lower than in for breast cancer does not seem to increase RA risk. the general population, and tamoxifen and aromatase inhibitors appear safe in terms of risk of future RA. Background With a lifetime risk of close to 1 in 10, breast cancer is the most common cancer among women. Rheu- matoid arthritis (RA) is the most common rheumatic RA is less clear. For HRT, available evidence does not disease affecting the joints, and has a marked female indicate an association with risk of RA.8 9 For breast predominance. Although the overall risk of malig- feeding, studies to date have reported a protective nancies in RA is increased by 10%–15% compared effect against RA,10–12 while results for parity and with the general population, large cohort studies risk of RA are inconclusive.12–15 Early menopause have consistently reported decreased occurrence of may be associated with subsequent development of breast cancer among women with RA.1–3 RA.10 16 Importantly, however, no study has assessed © Author(s) (or their The characteristics and the reason(s) behind the the extent to which the decrease in risk of breast employer(s)) 2020. Re-­use decreased risk of breast cancer in women with RA cancer in women with RA can be explained by known permitted under CC BY. remain unknown, indeed also virtually unstudied. breast cancer risk factors. Published by BMJ. Hormonal factors such as hormone replacement If the observed reduction in the risk of breast 4 5 To cite: Wadström H, therapy (HRT), early menarche and late menopause, cancer in women with RA was attributable to shared Pettersson A, Smedby KE, no breast feeding6 and nulliparity and increasing age risk factors rather than a causal effect of the RA et al. Ann Rheum Dis at first birth7 are all established risk factors for breast disease or its treatment, then one would expect that 2020;79:581–586. cancer. The role of these factors in the occurrence of the reduced risk of breast cancer would be present

Wadström H, et al. Ann Rheum Dis 2020;79:581–586. doi:10.1136/annrheumdis-2019-216756 581 Rheumatoid arthritis already before the onset of RA. So far, the link between history of and tamoxifen-like­ substances (99.6% tamoxifen, and hereafter breast cancer and future risk of RA has been little investigated.17 referred to as tamoxifen) and AI, respectively, in July 2005 or The occurrence of RA following breast cancer is clinically and later (see Anatomical Therapeutic Chemical (ATC) Classification aetiologically important also for other reasons. Five to 10 years System codes in online supplementary material). of adjuvant antihormonal treatment, for example, tamoxifen or aromatase inhibitors (AI), has become mainstay for oestrogen Covariates receptor positive breast cancer. Arthralgia is a common side We included information on educational level (from the Register effect of AI, and to a lesser extent also of tamoxifen. Some studies of Total Population), country of birth, number of live births have suggested that these therapies not only induce arthralgia, and age at first full-term­ pregnancy (from the Multigeneration but also inflammatory arthritis.18 19 Register), family history in a first-degree­ relative of breast cancer Our study therefore had the following aims: first, to examine or ovarian cancer (via linkage to the Multigeneration and Cancer the risk of breast cancer in women with RA, and to investigate Registers), oral contraceptives and intrauterine devices and HRT to what extent known breast cancer risk factors may explain the (from the Prescribed Drug Register). Age 50 was used as a proxy association. Second, to examine the risk of RA in women with a for menopausal status. Status on all covariates were ascertained history of breast cancer, and whether antihormonal treatment for at index date. For details and for an assessment of the associa- breast cancer modifies this association. To address the first aim, tion between these covariates and the outcome please see online we used a cohort design, and to address the second aim, we used a supplementary material. case–control design, both applied to a nationwide population with SAS V.9.4 was used for all analyses. The study was approved new-­onset RA and to individually matched population referents. by the Stockholm Ethics Review Board.

Materials and methods Setting and data sources Statistical analyses Swedish healthcare is public and tax funded. Patients with RA To assess and characterise the risk of breast cancer in patients are typically diagnosed and treated by rheumatologists. Sweden with RA, counting from RA diagnosis, we used a matched cohort has national and virtually complete registers on demographics design where patients with RA were considered exposed, and and health data which can be linked together by the unique their population comparators unexposed. Start of follow-up­ for personal identity number issued to all Swedish residents. This patients with RA (and for their matched population referents) study was based on linkages between the Swedish Rheumatology was set to the date when all RA-defining­ inclusion criteria were Quality Register (SRQ), the National Patient Register (NPR), the fulfilled. Women with a history of breast cancer at the time of Cancer Register, the Prescribed Drug Register, the Total Popu- start of follow-up­ were excluded. End of follow-up­ was defined lation Register, the Multigeneration Register and the Causes of as 31 December 2016, death, emigration or breast cancer, Death Register, described elsewhere and in the online supple- whichever occurred first. Relative risks for breast cancer were mentary material.20 21 assessed using Cox regression with follow-­up time as the times- cale. We first analysed age-­adjusted models, and then gradu- ally added more variables. In the full model, adjustments were Study population and study design made for age, calendar year, country of birth, educational level, Women with new-onset­ RA between 1 January 2006 and 31 HRT, oral contraceptives, age at first live birth, number of live-­ December 2016 were identified from two partly overlapping born children, family history of breast or ovarian cancer and sources, the SRQ and the NPR, using previously devised algo- previous invasive cancer. The risk of breast cancer was assessed rithms (online supplementary figure 1). Patients were required overall, stratified by time since start of follow-up­ (<1 year, 1 to have either (1) two visits in the NPR with an RA diagnosis to <5 years, >5 years), RA serostatus and age at RA diagnosis (main or contributory) in outpatient specialist care, and at (18–49 years, 50–75 years and >75 years). Relative risks were least one visit in a rheumatology or internal medicine clinic or also assessed separately for premenopausal and postmenopausal (2) an RA diagnosis in SRQ. An index date was defined as the cancer, and tumour, node, metastases (TNM) cancer stage at date of disease debut in the SRQ, RA diagnosis in the SRQ or diagnosis (online supplementary material). first RA diagnosis in the NPR (inpatient or outpatient care, see To assess the relative risk of RA in women with a history of online supplementary material for International Classification of breast cancer, we used a case–control design, and calculated ORs Diseases (ICD) codes), whichever came first. Subjects who had using conditional logistic regression. Besides the matching vari- not fulfilled the criteria (1 or 2) within 18 months of the index ables (year of birth, sex and place of residence), the fully adjusted date were excluded. For each individual with RA, we randomly model accommodated country of birth, educational level, age at selected five population referents from the general population, first live birth, number of live-born­ children and family history matched on year of birth, sex and place of residence at the time of breast or ovarian cancer. ORs were assessed overall, and strat- of index date. For the cohort analysis of risk of breast cancer in ified by age at RA diagnosis (18–49 years, 50–75 years, >75 women with RA, the population referents functioned as general years) and RA serostatus. Also, ORs were assessed according population comparators. For the case–control analysis of risk of to different exposure definitions, that is, time between the RA in women with a history of breast cancer, the population breast cancer and RA (<1 year, 1 to <5 years, 5 to <10 years, referents functioned as incidence-density­ sampled controls. >10 years), premenopausal or postmenopausal breast cancer and TNM cancer stage at diagnosis (see online supplementary Breast cancer and antihormonal treatment after material). breast cancer To assess the risk of RA following antihormonal breast cancer Through linkage to the Swedish Cancer Register, all incident (1958 treatment, we again used a case–control design. Exposure was or later) cases of breast cancer (ICD 170, non-­invasive and invasive) defined as at least two dispensings of the antihormonal drug in in the study population were identified. In the Prescribed Drug question, between breast cancer diagnosis and the index date. Also, Register, we identified all dispensed prescriptions of tamoxifen analyses of cumulative exposure of antihormonal treatment were

582 Wadström H, et al. Ann Rheum Dis 2020;79:581–586. doi:10.1136/annrheumdis-2019-216756 Rheumatoid arthritis

Table 1 Characteristics of study population of Swedish women with Table 2 Relative risk of breast cancer in RA versus general RA 2006–2016, and matched population referents population comparators (cohort study), adjusted for, age, country of RA Population referents birth, educational level, HRT, oral contraceptives, age at first birth, number of children, family history of breast cancer/ovarian cancer, Persons (N) 15 921 79 441 previous invasive cancer and calendar year Birth year (Q1–Q4) 1952 (1940–1963) 1952 (1940–1963) Mean age at entry (SD) 59 (16) 59 (16) Patients Population with RA, comparators, Education (%) Number Number HR (95% CI) 9 years or less 27 25 10–12 years 56 55 Overall 190 1191 0.80 (0.68 to 0.93) More than 12 years 15 19 Exposure variants RA in a first-­degree relative (%) 10 4 RA, 18–49 years 18 155 0.59 (0.36 to 0.97) Breast or ovarian cancer in a first-­ 10 10 RA, 50–75 years 146 841 0.87 (0.73 to 1.03) degree relative (%) RA, >75 years 26 195 0.68 (0.45 to 1.03) Invasive cancer prior to entry (%) 5 6 RA 18–50 and breast 9 53 0.86 (0.42 to 1.75) Country of birth (%) cancer <50 years Sweden 84 83 Seronegative RA 55 346 0.77 (0.58 to 1.02) Rest of Europe 11 11 Seropositive RA 124 772 0.81 (0.67 to 0.98) Rest of the world 5 6 Outcome variants ACPA and RF negative 28 NA Breast cancer, 18–50 9 102 0.47 (0.24 to 0.93) ACPA or RF positive 68 NA years Mean age at first live birth (SD) 25 (5.0) 25 (5.0) Breast cancer, >50 years 181 1089 0.83 (0.71 to 0.97) Children at index date (SD) 2 (1.4) 2 (1.4) Stage 0 12 97 0.63 (0.34 to 1.14) Combined oestrogen and 5 5 Stage 1 85 490 0.87 (0.69 to 1.09) progestin HRT (%) Stage 2 60 375 0.80 (0.61 to 1.05) Unopposed oestrogen HRT (%) 6 6 Stage 3 4 41 0.49 (0.17 to 1.36) Combined oestrogen and 9 9 Stage 4 6 30 0.99 (0.41 to 2.39) progestin contraceptive (%) HRT, hormone replacement therapy; RA, rheumatoid arthritis. Progestin only contraceptive (%) 7 7 ACPA, Anti-­citrullinated protein antibodies; HRT, hormone replacement therapy; RA, rheumatoid arthritis; RF, Rheumatoid factor. 0.58–1.02) and seropositive RA (HR=0.81, 95% CI 0.67–0.98), and for all age groups. We noted reduced risks for all TNM performed (<6 months, 6 to <12 months, 12 to <18 months, 18 stages, and for both premenopausal and postmenopausal breast to <24 months and ≥24 months), in these only one dispensing cancer (table 2). of antihormonal treatment was required. Since the Prescribed Drug Register started in July 2005, we restricted these analyses to Occurrence and relative risk of RA in women with women diagnosed with breast cancer from 2003 to 2016. As this a history of breast cancer would allow some patients receiving antihormonal treatment being At index date, 555 (3.5%) of the 15 921 patients with RA, and misclassified as non-exposed,­ we performed a sensitivity analysis 3193 (4.0%) of the controls had a history of breast cancer, restricting the study period to July 2005 through December 2016. age-­adjusted OR=0.86 (95% CI 0.78 to 0.94), fully adjusted To avoid the inclusion of occult RA diagnosed shortly after breast OR=0.87 (95% CI 0.79 to 0.95, table 3). ORs stratified by cancer diagnosis, and misclassified RA due to acute arthralgia serostatus and age at RA diagnosis yielded similar results. There following initiation of antihormonal treatment, in a sensitivity was no clear trend when examining the risk by menopausal analysis we excluded subjects with <1 year between the breast status, or cancer stage, at breast cancer diagnosis. However, cancer diagnosis and index date. missing information on cancer stage was substantial, especially Results among earlier cases of cancer. Characteristics of the study population During 2006–2016, we identified 15 921 incident patients with Relative risk of breast cancer by time before and RA, who were matched with 79 441 subjects from the general after RA population (table 1). Mean age at index date was 59 years, and There was no clear trend when stratifying by follow-­up time 68% were seropositive (5% unclassified, 6% had conflicting (figure 1). The lowest OR for breast cancer was observed 5 to information). <10 years before RA, OR=0.76 (95% CI 0.63 to 0.93), and the highest 1 to <5 years before RA, OR=0.93 (95% CI 0.58 to Occurrence and relative risk of breast cancer in 1.38). The lowest HR for breast cancer was observed 1 to <5 women with RA years after RA, HR=0.74 (95% CI 0.60 to 0.91), and the highest During a mean follow-up­ of 5.00 years among patients with RA 5 to <10 years after RA, HR=0.90 (95% CI 0.67 to 1.20). and 5.04 years among general population comparators, we iden- tified 190 cases of breast cancer among 15 356 patients with RA, Occurrence and relative risk of RA following anti- and 1191 cases among 75 854 population comparators, resulting hormonal treatment among women with a history in an age-adjusted­ and calendar year-adjusted­ HR=0.80 (95% CI of breast cancer 0.68 to 0.93). This estimate was virtually unchanged by further A history of a first ever breast cancer occurring during 2003– adjustments (fully adjusted HR=0.80, 95% CI 0.68 to 0.93). 2016 was identified among 259 of the 15 921 patients with RA, The risk was similar among seronegative RA (HR=0.77, 95% CI and 1499 of the 79 441 controls (table 4). Among these subjects,

Wadström H, et al. Ann Rheum Dis 2020;79:581–586. doi:10.1136/annrheumdis-2019-216756 583 Rheumatoid arthritis

Table 3 Relative risk of RA in women with a history of breast Table 4 Relative risk of RA in women, with or without a history cancer (case–control study), adjusted for, age, country of birth, of breast cancer, by antihormonal treatment before RA diagnosis, educational level, age at first birth, number of children and family adjusted for age, education, age at first birth, number of children and history of breast cancer/ovarian cancer family history of breast cancer/ovarian cancer Controls Fully adjusted OR RA (n) (n) OR (95% CI) RA (N) Controls (N) (95% CI) Overall 555 3193 0.87 (0.79 to 0.95) No breast cancer 15 366 76 248 1.07 (0.88 to 1.31) Exposure variants Never AI or tamoxifen 117 631 REF Breast cancer, 18–50 years 123 761 0.82 (0.67 to 0.99) Only tamoxifen versus never 62 383 0.86 (0.62 to 1.20) Breast cancer, >50 years 432 2432 0.89 (0.80 to 0.98) Only AI versus never 59 322 0.97 (0.69 to 1.37) Stage 0, 2002–2016 28 130 1.10 (0.73 to 1.65) Both AI and tamoxifen versus 21 163 0.68 (0.41 to 1.12) Stage 1, 2002–2016 110 530 1.04 (0.84 to 1.27) never Stage 2, 2002–2016 61 456 0.67 (0.51 to 0.87) AIs, aromatase inhibitors; RA, rheumatoid arthritis. Stage 3, 2002–2016 9 33 1.35 (0.65 to 2.83) Stage 4, 2002–2016 1 17 0.29 (0.04 to 2.19) risk with longer exposure time of >24 months OR=0.57, 95% Outcome variants CI 0.39 to 0.82). RA, 18–49 years 24 111 1.09 (0.70 to 1.71) Sensitivity analyses restricting the analysis of risk of RA among RA, 50–75 years 357 2132 0.84 (0.75 to 0.94) women with a history of breast cancer to cancer cases occur- RA, >75 years 174 950 0.91 (0.77 to 1.08) ring after the Prescribed Drug Register was started (July 2005 to Breast cancer 18–50 and RA >50 99 650 0.77 (0.62 to 0.96) December 2016), and excluding subjects with <1 year between years the breast cancer diagnosis and index date, provided results Seronegative RA 157 921 0.85 (0.71 to 1.01) similar to the main analysis (see online supplementary material). Seropositive RA 367 2088 0.88 (0.78 to 0.98) RA, rheumatoid arthritis. Discussion In this large population-­based study, we made a series of important observations. Also in this recent modern cohort of 117 (45%) patients with RA and 631 (42%) controls had not patients followed from RA diagnosis, there was a decreased received tamoxifen or AI treatment before RA diagnosis/index risk of breast cancer, corresponding to a lifetime risk of 1 in date (never vs ever, fully adjusted OR=1.23, 95% CI 0.92 to 12.5, instead of 1 in 10 as in the general population.22 This risk 1.64). reduction was independent of RA serostatus, and remained after Among the remaining 1010 patients with RA and controls, adjusting for breast cancer risk factors. Furthermore, the risk of use of tamoxifen was somewhat more frequent than AI (n=629 future RA in women with a history of breast cancer was reduced vs n=565), with some overlap (n=184). The risk of RA was with risk estimates similar to those of breast cancer risk after RA. not significantly associated with any combination of antihor- Finally, treatment of breast cancer with tamoxifen and AI did not monal breast cancer treatments; both tamoxifen and AI versus constitute risk factors for the development of RA. never treated: fully adjusted OR=0.68 (95% CI 0.41 to 1.12); We noted, as has previously been reported,1 a 20% decreased tamoxifen-only­ versus never treated: OR=0.86 (95% CI 0.62 to risk of breast cancer in women with a history of RA, and extend 1.20); AI-­only versus never treated: OR=0.97 (95% CI 0.69 to this finding by demonstrating that adjusting for several important 1.37; table 4). breast cancer risk determinants did not significantly impact this When examining cumulative tamoxifen exposure (years), result. The relative risk of RA in women with a history of breast we noted no trends in risk of RA (table 5). For AI, we noted cancer was similar to the relative risk of breast cancer in women an increased risk of RA among patients treated with AI for <6 with a history of RA. Taken together, this would argue against months (OR=1.58, 95% CI 1.02 to 2.45), but also a decreased the hypothesis that the decreased risk of breast cancer in patients with RA is due to the RA disease or its treatment. Rather, our data suggest that RA and breast cancer share genetic factors or environmental factors acting earlier in life. In this regard, it is interesting that the breast cancer ORs and HRs were similar for seropositive and seronegative RA. This is one of the first studies examining the relationship between antihormonal breast cancer treatment, and the risk of RA.18 19 Contrary to a study by Caprioli et al, who studied the risk of RA in a cohort of 10 493 women with breast cancer, we did not find a markedly higher risk of RA in AI-treated­ compared with tamoxifen-treated­ patients. Although based on relatively small numbers, our upper confidence limits indicate a low prob- ability for clinically significant risk increases overall. However, we cannot rule out an association between RA and tamoxifen/AI Figure 1 Relative risk of breast cancer (invasive or in situ) as a among specific subgroups, for example, as defined by hormonal function of time before (left of vertical line, ORs) and after (right receptor status, which we did not have information on. We did of vertical line, HRs) RA diagnosis. RA versus general population observe a trend towards a higher risk in the beginning of treat- comparators, adjusted for age, country of birth and educational level. ment, but then a reduced risk after 2 years of AI treatment. Capr- RA, rheumatoid arthritis ioli et al did not compare the rate of RA among patients with

584 Wadström H, et al. Ann Rheum Dis 2020;79:581–586. doi:10.1136/annrheumdis-2019-216756 Rheumatoid arthritis

Table 5 Relative risk of RA in patients with a history of breast cancer, by accumulated time of antihormonal treatment before RA diagnosis, adjusted for age, education, age at first birth, number of children and family history of breast cancer/ovarian cancer Tamoxifen AIs Cumulative exposure RA (n) Controls (n) OR (95% CI) RA (n) Controls (n) OR (95% CI) <6 months 27 134 0.98 (0.65 to 1.49) 27 87 1.58 (1.02 to 2.45) 6 to <12 months 14 76 0.95 (0.53 to 1.68) 9 63 0.72 (0.36 to 1.44) 12 to <18 months 8 62 0.62 (0.30 to 1.30) 12 41 1.45 (0.76 to 2.77) 18 to <24 months 11 65 0.83 (0.43 to 1.57) 12 57 1.08 (0.58 to 2.02) >24 months 42 283 0.74 (0.54 to 1.03) 31 267 0.57 (0.39 to 0.82) AIs, aromatase inhibitors; RA, rheumatoid arthritis. breast cancer to that of the general population. The rate that validity and coverage and that we could adjust our models for they observed (4.33 per 1000 person-­years) is somewhat lower several potentially important confounders. than population-­based age and sex standardised incidence rates In conclusion, we found a decreased risk of breast cancer in reported from Italy23 and Sweden.24 Our results also differed patients with RA, and a similar decrease in risk of RA in patients from that of a cross-sectional­ study by Chen et al that included with a history of breast cancer. We did not find evidence to more than 200 000 cases of breast cancer.18 They found a cumu- support that the decreased risk of breast cancer was due to lative dose-dependent­ risk increase of RA with both tamoxifen known risk determinants. Thus we were ultimately unable to and AI treatment, compared with women with breast cancer explain the origins of this association. Antihormonal therapy as who did not receive these treatments. However, they neither used in secondary breast cancer pharmacoprevention does not took follow-­up time nor age into account. We believe that our seem to increase RA risk. nationwide study with a clearly defined population, which could Contributors HW had full access to all the data in the study and takes account for person-­time at risk, age and several other potential responsibility for the integrity of the data and the accuracy of the data analysis. HW confounders, and the use of an appropriate comparator group, conducted the statistical analyses. All authors participated in designing the analyses, provides a more reliable estimate. interpreting the results and writing the manuscript. We adjusted our models for hormonal risk factors, which Funding This work was supported by the Swedish Cancer Society, the Swedish had little impact on our results. Whether this is due to residual Research Council, the Region Stockholm (ALF), the Swedish Heart Lung Foundation, confounding or a true absence of confounding is difficult to ascer- Nordforsk. Funders had no impact on the design or interpretation of the study or its results. tain, but we did observe that these factors were indeed risk factors for breast cancer in our study population (online supplementary Competing interests HW, AP and KES have no competing interests to declare. JA has or has had research agreements with Abbvie, Astra-­Zeneca, BMS, Eli Lilly, material). Studies have found that certain polymorphisms of the MSD, Pfizer, Roche, Samsung Bioepis and UCB, mainly in the context of safety cyclooxygenase-2 gene promoter are associated with an increased monitoring of biologics via ARTIS/Swedish Biologics Register. Karolinska Institutet risk of breast cancer and a decreased risk of RA.25 26 Also, polymor- has received remuneration for JA participating in advisory boards arranged by phisms in the DRB1 gene, the major genetic susceptibility locus Pfizer andL illy. for RA,27 have recently been linked to a decreased risk of breast Patient and public involvement Patients and/or the public were involved in cancer.28 However, because of the complex inheritance, it is diffi- the design, conduct, reporting or dissemination plans of this research. Refer to the cult to estimate the net effect of such singular genetic factors. Methods section for further details. Our study has some limitations. Although algorithms for iden- Patient consent for publication Not required. tifying incident RA diagnoses in the NPR have been validated Ethics approval Regional Ethics Committee, Stockholm, Sweden. and shown to have a positive predictive value close to 90%, Provenance and peer review Not commissioned; externally peer reviewed. misclassification of RA cannot be excluded.29 We lacked data Data availability statement No data are available. The data sets generated and/ on menarche, menopause and breast feeding. Early menarche or analysed during the present study are not publicly available due to the Swedish has been reported as a risk factor, and long-­term breast feeding legislation (the Personal Data Act). 10–12 seems protective, for both RA and breast cancer. Early Open access This is an open access article distributed in accordance with the menopause, which is negatively associated with breast cancer, Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits has been reported as a risk factor for RA.11 16 We also could others to copy, redistribute, remix, transform and build upon this work for any not account for some additional potential confounding factors purpose, provided the original work is properly cited, a link to the licence is given, including body mass index, smoking and alcohol consumption. and indication of whether changes were made. See: https://​creativecommons.​org/​ licenses/by/​ ​4.0/.​ Body mass index is considered a risk factor for both breast cancer and RA and would thus introduce a positive confounding.30 ORCID iD Smoking is a risk factor for RA, but alcohol has been reported Hjalmar Wadström http://orcid.​ ​org/0000-​ ​0002-​9283-​5536 as being protective against RA.31 However, both are only weak risk factors for breast cancer.32 33 We were not able to account References for mammographic screening. However, the negative associ- 1 simon TA, Thompson A, Gandhi KK, et al. Incidence of malignancy in adult patients with rheumatoid arthritis: a meta-­analysis. Arthritis Res Ther 2015;17:1–10. ation between RA and breast cancer was described in studies 34 35 2 gadalla SM, Amr S, Langenberg P, et al. Breast cancer risk in elderly women with conducted before mammographic screening was mainstay. systemic autoimmune rheumatic diseases: a population-­based case-­control study. Br J Moreover, since the assessment of RA risk after breast cancer is, Cancer 2009;100:817–21. by definition, an assessment among breast cancer survivors, we 3 parikh-­Patel A, White RH, Allen M, et al. Risk of cancer among rheumatoid arthritis cannot exclude that breast cancer among would-­be patients with patients in California. Cancer Causes Control 2009;20:1001–10. 36 37 4 collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and RA could have a worse prognosis irrespective of cancer stage. hormonal contraceptives: collaborative reanalysis of individual data on 53 297 The strengths of this study were its large sample size and long women with breast cancer and 100 239 women without breast cancer from 54 follow-­up, the use of nationwide registers with high internal epidemiological studies. Lancet 1996;347:1713–27.

Wadström H, et al. Ann Rheum Dis 2020;79:581–586. doi:10.1136/annrheumdis-2019-216756 585 Rheumatoid arthritis

5 collaborative Group on Hormonal Factors in Breast Cancer. Menarche, menopause, 20 Wadström H, Eriksson JK, Neovius M, et al. How good is the coverage and how and breast cancer risk: individual participant meta-­analysis, including 118 964 accurate are exposure data in the Swedish biologics register (ARTIS)? Scand J women with breast cancer from 117 epidemiological studies. Lancet Oncol Rheumatol 2015;44:22–8. 2012;13:1141–51. 21 askling J, Fored CM, Geborek P, et al. Swedish registers to examine drug safety and 6 cgoHFiB C. Breast cancer and breastfeeding: collaborative reanalysis of clinical issues in RA. Ann Rheum Dis 2006;65:707–12. individual data from 47 epidemiological studies in 30 countries, including 50 302 22 statistics on cancer incidence 2017: Socialstyrelsen, 2018. Available: https://www.​ women with breast cancer and 96 973 women without the disease. The Lancet socialstyrelsen.se/​ ​globalassets/sharepoint-​ ​dokument/​artikelkatalog/​statistik/2018-​ ​12-​ 2002;360:187–95. 51.pdf​ 7 Rosner B, Colditz GA, Willett WC. Reproductive risk factors in a prospective study of 23 Rossini M, Rossi E, Bernardi D, et al. Prevalence and incidence of rheumatoid arthritis breast cancer: the nurses’ health study. Am J Epidemiol 1994;139:819–35. in Italy. Rheumatol Int 2014;34:659–64. 8 doran MF, Crowson CS, O’Fallon WM, et al. The effect of oral contraceptives and 24 Eriksson JK, Neovius M, Ernestam S, et al. Incidence of rheumatoid arthritis in estrogen replacement therapy on the risk of rheumatoid arthritis: a population based Sweden: a nationwide population-­based assessment of incidence, its determinants, study. J Rheumatol 2004;31:207–13. and treatment penetration. Arthritis Care Res 2013;65:870–8. 9 Walitt B, Pettinger M, Weinstein A, et al. Effects of postmenopausal hormone therapy 25 gao J, Ke Q, Ma H-­X, et al. Functional polymorphisms in the cyclooxygenase 2 (COX- on rheumatoid arthritis: the women’s health Initiative randomized controlled trials. 2) gene and risk of breast cancer in a Chinese population. J Toxicol Environ Health A Arthritis Rheum 2008;59:302–10. 2007;70:908–15. 10 merlino LA, Cerhan JR, Criswell LA, et al. Estrogen and other female reproductive risk 26 lee KH, Kim H-­S, El-­Sohemy A, et al. Cyclooxygenase-2 genotype and rheumatoid factors are not strongly associated with the development of rheumatoid arthritis in arthritis. J Rheumatol 2006;33:1231–4. elderly women. Semin Arthritis Rheum 2003;33:72–82. 27 Holoshitz J. The rheumatoid arthritis HLA-­DRB1 shared epitope. Curr Opin Rheumatol 11 Karlson EW, Mandl LA, Hankinson SE, et al. Do breast-­feeding and other reproductive 2010;22:293–8. factors influence future risk of rheumatoid arthritis? results from the nurses’ health 28 Karnes JH, Bastarache L, Shaffer CM, et al. Phenome-wide­ scanning identifies multiple study. Arthritis Rheum 2004;50:3458–67. diseases and disease severity phenotypes associated with HLA variants. Sci Transl Med 12 pikwer M, Bergström U, Nilsson J-­A, et al. Breast feeding, but not use of oral 2017;9:eaai8708. contraceptives, is associated with a reduced risk of rheumatoid arthritis. Ann Rheum 29 Waldenlind K, Eriksson JK, Grewin B, et al. Validation of the rheumatoid arthritis Dis 2009;68:526–30. diagnosis in the Swedish national patient register: a cohort study from Stockholm 13 Jørgensen KT, Pedersen BV, Jacobsen S, et al. National cohort study of reproductive County. BMC Musculoskelet Disord 2014;15:432. risk factors for rheumatoid arthritis in Denmark: a role for hyperemesis, gestational 30 Zhou Y, Sun M. A meta-­analysis of the relationship between body mass index and risk hypertension and pre-­eclampsia? Ann Rheum Dis 2010;69:358–63. of rheumatoid arthritis. Excli J 2018;17:1079–89. 14 orellana C, Wedrén S, Källberg H, et al. Parity and the risk of developing rheumatoid 31 di Giuseppe D, Alfredsson L, Bottai M, et al. Long term alcohol intake and risk of arthritis: results from the Swedish epidemiological investigation of rheumatoid rheumatoid arthritis in women: a population based cohort study. BMJ : British Medical arthritis study. Ann Rheum Dis 2014;73:752–5. Journal 2012;345:e4230. 15 guthrie KA, Dugowson CE, Voigt LF, et al. Does pregnancy provide vaccine-­like 32 gaudet MM, Gapstur SM, Sun J, et al. Active smoking and breast cancer risk: original protection against rheumatoid arthritis? Arthritis Rheum 2010;62:NA–48. cohort data and meta-­analysis. J Natl Cancer Inst 2013;105:515–25. 16 pikwer M, Bergström U, Nilsson Jan-Åke, et al. Early menopause is an independent 33 smith-­Warner SA, Spiegelman D, Yaun SS, et al. Alcohol and breast cancer in women: predictor of rheumatoid arthritis. Ann Rheum Dis 2012;71:378–81. a pooled analysis of cohort studies. JAMA 1998;279:535–40. 17 Hellgren K, Smedby KE, Feltelius N, et al. Do rheumatoid arthritis and lymphoma share 34 mellemkjaer L, Linet MS, Gridley G, et al. Rheumatoid arthritis and cancer risk. Eur J risk factors?: a comparison of lymphoma and cancer risks before and after diagnosis Cancer 1996;32A:1753–7. of rheumatoid arthritis. Arthritis Rheum 2010;62:1252–8. 35 gridley G, McLaughlin JK, Ekbom A, et al. Incidence of cancer among patients with 18 chen JY, Ballou SP. The effect of antiestrogen agents on risk of autoimmune disorders rheumatoid arthritis. J Natl Cancer Inst 1993;85:307–11. in patients with breast cancer. J Rheumatol 2015;42:55–9. 36 Ji J, Liu X, Sundquist K, et al. Survival of cancer in patients with rheumatoid arthritis: 19 caprioli M, Carrara G, Sakellariou G, et al. Influence of aromatase inhibitors therapy a follow-­up study in Sweden of patients hospitalized with rheumatoid arthritis 1 year on the occurrence of rheumatoid arthritis in women with breast cancer: results from before diagnosis of cancer. Rheumatology 2011;50:1513–8. a large population-­based study of the Italian Society for rheumatology. RMD Open 37 park JK, Yang JA, Ahn EY, et al. Survival rates of cancer patients with and without 2017;3:e000523. rheumatic disease: a retrospective cohort analysis. BMC Cancer 2016;16:381.

586 Wadström H, et al. Ann Rheum Dis 2020;79:581–586. doi:10.1136/annrheumdis-2019-216756 Rheumatoid arthritis

Clinical science Anti-­carbamylated proteins antibody repertoire in rheumatoid arthritis: evidence of a new autoantibody linked to interstitial lung disease Raul Castellanos-­Moreira ‍ ‍ ,1 Sebastian Cruz Rodríguez-­García ‍ ‍ ,1 Maria Jose Gomara,2 Virginia Ruiz-­Esquide,1 Andrea Cuervo,1 Ivette Casafont-­Solé,3 Julio Ramírez,1 Susana Holgado,3 Jose A Gómez-­Puerta,1 Juan D Cañete,1 Isabel Haro,2 Raimon Sanmarti1

Handling editor Josef S Abstract Key messages Smolen Objective To analyse the association between anti-­carbamylated protein antibodies (Anti-­CarP) and ►► Additional material is What is already known about this subject? published online only. To view interstitial lung disease (ILD) in rheumatoid arthritis (RA) ►► Rheumatoid arthritis-­associated interstitial please visit the journal online patients. lung disease (RA-­ILD) entails a high mortality. (http://dx.​ ​doi.org/​ ​10.​1136/​ Methods Cross-­sectional study including RA Therefore, an early detection is crucial in annrheumdis-2019-​ ​216709). patients fulfilling the 2010A CR/EULAR criteria. establishing an individualised treatment 1 The main population comprised two groups: (1) RA Rheumatology Department, strategy. Arthritis Unit, Hospital Clinic de patients diagnosed with RA-­ILD (RA-­ILD group); (2) ► Anti-­carbamylated protein antibodies (Anti-­ Barcelona, Barcelona, Spain RA patients without ILD (non-­ILD RA group). Non-­ILD ► 2 CarP) are associated with poor disease Consejo Superior de RA patients in whom ILD was suspected underwent a Investigaciones Científicas, Unit outcomes in RA patients and have been diagnostic work-­up and, if ILD was diagnosed, were of Synthesis and Biomedical detected in various chronic lung diseases switched to the RA-­ILD group. ILD was diagnosed by Applications of Peptides, CSIC-­ regardless of the RA history. IQAC, Barcelona, Spain high-­resolution computed tomography and confirmed 3Rheumatology Department, by a multidisciplinary committee. An independent What does this study add? Hospital Germans Trias i Pujol, replication sample was also obtained. Three Anti-­CarP Badalona, Spain ► For the first time an association between RA-­ IgG autoantibodies against fetal calf serum (Anti-­FCS), ► ILD and Anti-­CarP has been found, which was fibrinogen (Anti-­Fib) and chimeric fibrine/filagrine Correspondence to independent of smoking, sex, age, RA disease homocitrullinated peptide (Anti-­CFFHP) and one Anti-­ Dr Raimon Sanmarti, Arthritis duration, anticitrullinated protein antibody and Unit, Rheumatology CarP IgA against FCS (Anti-­FCS-­IgA) were determined rheumatoid factor. Department, Hospital Clinic de by home-­made ELISA. Associations between Anti-­CarP Barcelona, Barcelona 08036, and ILD were analysed using multivariable logistic Spain; sanmarti@​ ​clinic.cat​ How might this impact on clinical practice or regression adjusted by smoking, sex, age, RA disease future developments? RC-­M and SCR-­G contributed duration, rheumatoid factor and anticitrullinated protein ► These findings suggest a possible link between equally. antibodies. ► homocitrullination and the development of ILD Results We enrolled 179 patients: 37 (21%) were Received 25 November 2019 in RA patients exists. Revised 29 January 2020 finally diagnosed with AR -­ILD. Anti-­CarP specificities Accepted 17 February 2020 were more frequent in RA-­ILD patients (Anti-­FCS 70% Published Online First vs 43%; Anti-­Fib 73% vs 51%; Anti-­CFFHP 38% vs 10 March 2020 19%; Anti-­CarP-­IgA 51% vs 20%, p<0.05 for all examined.1 RA-associated­ ILD (RA-ILD)­ entails a comparisons). Serum titers of Anti-­CarP were significantly shortened survival with a mortality rate up to 10 higher in RA-­ILD patients. Anti-­CarP specificities showed times higher than those without ILD,2 prompting a a robust effect towards increasing the odds of ILD in great effort to achieve an earlier diagnosis. the multivariate analysis (Anti-­FCS (OR: 3.42; 95% CI: Various risk factors have been identified, including 1.13 to 10.40), Anti-­Fib (OR: 2.85; 95% CI: 0.83 to smoking, male sex, higher disease activity, longer 9.70), Anti-­CFFHP (OR: 3.11; 95% CI: 1.06 to 9.14) and disease duration, older age, positive rheumatoid Anti-­FCS-­IgA (OR: 4.30; 95% CI: 1.41 to 13.04)). Similar factor (RF) and anticitrullinated protein antibodies findings were observed in the replication sample. (ACPAs).1 3 In recent years, several biomarkers have Conclusions Anti-­CarP were strongly associated with been proposed for RA-ILD­ screening,4 5 although ILD. The role of homocitrullination in RA-­ILD merits there is no universally accepted screening test. © Author(s) (or their Anti-­carbamylated protein antibodies (Anti-­ employer(s)) 2020. No further investigation. commercial re-­use. See rights CarP) recognise homocitrullinated peptides, which and permissions. Published are generated by a post-­translational modification by BMJ. (PTM) of lysine residues.6 First described in 2011, Anti-­CarP have been associated with poor disease To cite: Castellanos-­ Introduction Moreira R, Rodríguez-­ Interstitial lung disease (ILD) is a severe extra-­ outcomes in RA patients, including higher disease García SC, Gomara MJ, articular manifestation of rheumatoid arthritis (RA), activity and radiographic progression.7–9 A recent et al. Ann Rheum Dis that affects 4%–50% of RA patients depending study found higher mortality in Anti-­CarP positive 2020;79:587–594. on the screening method and the population RA patients, which was particularly attributed to

Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 587 Rheumatoid arthritis respiratory diseases.10 Interestingly, Anti-­CarP have been found operating characteristic (ROC) curve analysis and regression in patients with various non-­ILD chronic lung diseases irrespec- analysis were conducted using the GraphPad Prism5 programme tive of smoking or RA history.11 12 and the cut-­off values were determined with a specificity of 95% The objective of this study was to analyse the association compared with a healthy population of blood donors (n=179). between Anti-­CarP and ILD in a population of RA patients. A series of successive dilutions of a pool of sera from four posi- tive patients was used as a reference standard in all plates and to Methods convert optical density (OD) values to arbitrary units (AU). Study design A positive cut-off­ value was defined as ≥173.5 AU/mL, ≥166.9 We performed a cross-­sectional study including RA patients diag- AU/mL, ≥146.5 AU/mL and ≥257.0 AU/mL for Anti-­FCS, nosed according to the 2010 ACR/EULAR criteria assessed in a Anti-­Fib, Anti-­CFFHP and Anti-­FCS-IgA,­ respectively. A test rheumatology department outpatient clinic of a tertiary univer- was only considered positive and specific for homocitrulline sity hospital. Individuals fulfilling other inflammatory arthritis when the AU/mL values were higher than the respective cut-­off or connective tissue disease diagnostic criteria were excluded. and the OD value difference between carbamylated (homocitrul- linated) and native (non-­homocitrullinated) antigens was at least 16 Study population 0.1 in agreement with the methodology previously reported. Further details on the ELISA techniques are provided in the The main population comprised two groups; (1) patients diag- online supplementary text. nosed with RA-ILD­ before study inclusion (RA-ILD­ group); (2) RF was measured by nephelometry and ACPA was assessed consecutive patients with RA with no diagnosis of ILD, assessed by anti-CCP3 chemiluminescent immunoassay (QUANTA Flash by a single rheumatologist (RC-M)­ between July 2017 and July CCP3, Inova Diagnostics). The cut-­offs for positive values were 2018 (non-­ILD RA group). 25 international units (IU) and 20 chemiluminescence units All non-­ILD RA patients in whom ILD was suspected, under- (CU), respectively. went an appropriate work-up­ (see RA-­ILD diagnosis section). Subjects resulting in a new diagnosis were switched to the RA-ILD­ group. Criteria for suspicion were based on the presence Statistical analysis of RA-ILD­ risk factors or persistent symptoms such as dyspnoea Between-­group differences were analysed using descriptive or cough. An independent replication sample with similar char- statistics as appropriate. Proportions were compared using the χ² acteristics was obtained from another hospital. or Fisher’s exact test. Continuous variables were analysed using the Student’s t-­test or the Mann–Whitney U test and presented RA-ILD diagnosis as means with SD (±) or medians with IQR. RA-ILD­ was diagnosed using high-resolution­ computed tomog- The association between Anti-CarP­ and ILD was evaluated raphy and classified according to the American Thoracic Society/ using logistic regression including a set of variables known to European Respiratory Society 2013 multidisciplinary classifica- be related to the development of ILD such as smoking, sex, age, tion criteria of idiopathic interstitial pneumonias.13 Diagnoses RA disease duration, ACPA and RF and others taken from the were confirmed by a multidisciplinary committee including univariate analysis (p<0.2). Potential confounders such as treat- pneumologists, radiologists, pathologists, clinical immunolo- ments or disease activity were not included in the models because gists, internists and rheumatologists. of concerns regarding timing, dosing and study design. Different models were built using the all possible equations approach and the final models were chosen according to the Akaike informa- Demographic, clinical, therapeutic and serological data tion criterion and the area under the ROC curve after assessing In the main population, the variables analysed included demo- possible first order interactions. graphics (age, sex and ethnicity), disease duration at study inclu- Exploratory analysis, including the association between Anti-­ sion, smoking status and smoking cumulative dose (number of FCS-IgA­ and RA-ILD,­ and the unadjusted assessment of the packs per day×number of years of smoking). Disease activity overlap between antibodies and the association between Anti-­ and disability were assessed using the Disease Activity Score in CarP and smoking were performed in the main population. 28 joints (DAS28) and the Health Assessment Questionnaire There was only one observation missing for HAQ in the main Disability Index (HAQ-­DI), respectively. Current conventional population which was imputed using the average HAQ values in synthetic and biological disease-modifying­ antirheumatic drugs the respective group. (csDMARDs and bDMARDs, respectively) and glucocorticoid Two-­tailed p values <0.05 were established to be statistical use were evaluated. Hand and foot X-­rays were obtained at significance in all analyses, which were performed using both study entry for analysis of radiological damage. IBM SPSS for Windows V.23.0 and STATA V.15. Antibodies measurement Autoantibody status was measured in sera collected at study Ethics enrolment. Three carbamylated antigens, namely fetal calf serum Signed informed consent was obtained from all patients before (FCS), fibrinogen (Fib) and chimeric fibrin/filaggrin homoc- study enrolment. The Strengthening the Reporting of Observa- itrullinated peptide (CFFHP) were obtained. CFFHP and its tional Studies in Epidemiology guidelines were followed. non-­homocitrullinated version were synthesised by solid-phase­ peptide synthesis following a 9-­fluorenylmethoxycarbonyl/t-­bu Results tyl strategy with subsequent cyclisation in solution by forming a Study population baseline features disulfide bridge.14 15 Four Anti-­CarP autoantibodies were deter- One hundred and seventy-­nine patients were enrolled in the mined by home-­made ELISA tests with either carbamylated main population: 31 in the RA-ILD­ group and 148 in the non-­ proteins or CFFHP as antigens, respectively; three IgG specific- ILD group. The diagnostic work-­up resulted in six new ILD ities namely Anti-­FCS, Anti-­Fib and Anti-CFFHP­ and one IgA diagnoses: these patients were relocated to the RA-ILD­ (see specificity against carbamylated FCS (Anti-FCS-­ IgA).­ Receiver figure 1). Thus, 37 patients (21%) were finally included in the

588 Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 Rheumatoid arthritis

(±1.05), p<0.005), a higher proportion of erosive disease (70% vs 44%, p<0.005) and functional disability (HAQ-­DI ≥1; 27% vs 12%, p value: 0.02). No differences in current (19% vs 16%, p value: 0.69) and ever-smoking­ (57% vs 44%, p value: 0.15) were observed, although patients with RA-ILD­ had a higher smoking cumulative dose (30.7 (±11.1) vs 21.8 (±12.0) packs/year, p<0.005). No between-group­ differences in the proportions of current csDMARDs (89% vs 86%, p value: 0.60), bDMARDs (30% vs 25%, p value: 0.59) or glucocorticoids (70% vs 58%, p value: 0.17) were observed.

Anti-CarP and RA-ILD The proportions of positive Anti-­CarP in the main population were 49%, 56%, 23% and 27% for Anti-­FCS, Anti-­Fib, Anti-­ CFFHP and Anti-FCS-­ IgA,­ respectively. All Anti-CarP­ fine spec- ificities were more frequent in the RA-­ILD group (Anti-­FCS: 70% vs 43%; Anti-Fib:­ 73% vs 51%; Anti-­CFFHP: 38% vs 19%; Anti-­FCS-IgA:­ 51% vs 20%, p<0.05 for all comparisons) Figure 1 Main population flow chart. ILD, interstitial lung disease; RA, (table 2). Similar results were observed in the replication sample rheumatoid arthritis. (Anti-­FCS: 92% vs 48%, p<0.005; Anti-­Fib: 76% vs 58%, p value: 0.12; Anti-­CFFHP: 36% vs 18%, p value: 0.08). RA-ILD­ group and 142 in the non-­ILD RA group. The repli- In addition, the mean titers were higher in the RA-­ILD group, cation sample was composed of 25 patients with RA-ILD­ and ranging from approximately 1.3 for Anti-­Fib to nearly fourfold 50 RA patients without ILD. The baseline features of the main for Anti-­CFFHP; the differences were statistically significant for population and the replication sample are summarised in table 1 Anti-­FCS, Anti-­CFFHP and Anti-­FCS-­IgA (figure 2A–C,F). and online supplementary table S1, respectively. A logistic regression model adjusted for age, RA disease dura- In the main population, RA-ILD­ patients had a mean age at tion, ACPA, RF, sex and smoking cumulative dose showed that ILD diagnosis of 64.2 (±9.7) years and a mean time from RA Anti-­FCS (OR: 3.42; 95% CI: 1.13 to 10.40), Anti-­CFFHP (OR: diagnosis to ILD diagnosis of 8.5 (±7.4) years. Usual inter- 3.12; 95% CI: 1.06 to 9.14) and Anti-FCS-­ IgA­ (OR: 4.30; 95% stitial pneumonia (38%) and non-­specific pneumonia (38%) CI: 1.41 to 13.04) were independently associated with ILD. The were the most frequent ILD patterns, followed by respiratory results of the analysis for Anti-­Fib (OR: 2.85; 95% CI: 0.83 to bronchiolitis-­associated ILD (11%), cryptogenic organised pneu- 9.71) followed a similar trend, although not significant. This monia (8%) and desquamative interstitial pneumonia (5%). tendency was confirmed in the replication sample; Anti-­FCS Subjects within the RA-ILD­ group were older and had a longer (OR: 10. 42; 95% CI: 1.68 to 64.46); Anti-­Fib (OR: 1. 65; 95% disease duration than their counterparts at inclusion (p<0.005) CI: 0.40 to 6.86) and Anti-CFFHP­ (OR: 1. 49; 95% CI: 0.42 to as well as greater disease activity (DAS28 3.71 (±1.35) vs 2.74 5.29) (table 3).

Table 1 Main population demographic, clinical and therapeutic features Main population, n: 179 RA-­ILD, n: 37 Non-­ILD RA, n: 142 P value Female (%) 141 (79) 25 (68) 116 (82) NS Age mean (±SD) 59.7 (13.0) 67.3 (10.1) 57.7 (12.9) <0.005 Mean disease duration (±SD) 6.6 (5.0) 11.6 (7.1) 5.3 (13.3) <0.005 Current smokers (%) 30 (17) 7 (19) 23 (16) NS Ever smokers (%) 83 (46) 21 (57) 62 (44) NS Smoking cumulative dose (±SD) 24.1 (13.0) 30.7 (11.1) 21.8 (12.0) <0.005 Caucasian (%) 151 (84) 31 (84) 120 (85) NS Other EAMs Sicca syndrome (%) 33 (18) 8 (22) 25 (18) NS Rheumatoid nodules (%) 21 (12) 7 (19) 14 (10) NS Serositis (%) 3 (2) 1 (3) 2 (1) NS Treatment Glucocorticoids (%) 108 (60) 26 (70) 82 (58) NS csDMARDs (%) 155 (87) 33 (89) 132 (86) NS MTX (%) 115 (64) 20 (54) 95 (67) NS bDMARDs (%) 47 (26) 11 (30) 36 (25) NS Mean DAS28 (±SD) 2.94 (1.18) 3.71 (1.35) 2.74 (1.05) <0.005 Erosive disease (%) 89 (50) 26 (70) 63 (44) <0.005 Mean HAQ-­DI (95% CI) 0.39 (0.32 to 0.46) 0.69 (0.53 to 0.85) 0.31 (0.24 to 0.38) <0.005 Functional disability HAQ-­DI ≥1 (%) 27 (15) 10 (27) 17 (12) 0.024 bDMARDs, biological disease-­modifying antirheumatic drugs; csDMARDs, conventional synthetic disease-modifying­ antirheumatic drugs; DAS28, Disease Activity Score in 28 joints; EAMs, extra-­articular manifestations; HAQ-­DI, Health Assessment Questionnaire Disability Index; ILD, interstitial lung disease; MTX, methotrexate; NS, not significant; RA, rheumatoid arthritis.

Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 589 Rheumatoid arthritis

Table 2 Main population autoantibody status in patients with and without ILD Main population, n: 179 RA-­ILD, n: 37 Non-­ILD RA, n: 142 P value ACPA positive (%) 128 (72) 29 (78) 99 (70) NS Median titer ACPA (IQR) CU 222 (1341) 674 (2215) 143 (1132) NS RF positive (%) 111 (62) 28 (76) 83 (59) NS Median titer RF (IQR) IU 40 (182) 105 (298) 34 (110) NS Anti-­FCS positive (%) 87 (49) 26 (70) 61 (43) <0.005 Median titer Anti-­FCS (IQR) AU/mL 168 (354) 332 (1700) 136 (304) <0.005 Anti-­Fib positive (%) 100 (56) 27 (73) 73 (51) 0.019 Median titer Anti-­Fib (IQR) AU/mL 186 (305) 281 (357) 175 (277) NS Anti-­CFFHP positive (%) 41 (23) 14 (38) 27 (19) 0.015 Median titer Anti-­CFFHP (IQR) AU/mL 16 (131) 33 (834) 14 (115) 0.030 Anti-FCS-­ ­IgA positive (%) 48 (27) 19 (51) 29 (20) <0.005 Median titer Anti-FCS-­ ­IgA (IQR) AU/mL 102 (285) 258 (528) 79 (222) <0.005 ACPA, anticitrullinated protein antibody; Anti-­CFFHP, IgG antibodies against chimeric fibrine/filagrine homocitrullinated peptide; Anti-FCS,­ IgG antibodies against carbamylated fetal calf serum; Anti-­FCS-­IgA, IgA antibodies against carbamylated fetal calf serum; Anti-Fib,­ IgG antibodies against carbamylated fibrinogen; AU, arbitrary units; CFFHP, chimeric fibrin/filaggrin homocitrullinated peptide; CU, chemiluminescence units; ILD, interstitial lung disease; IU, international units; NS, not significant; RA, rheumatoid arthritis; RF, rheumatoid factor.

Figure 2 Boxplots of autoantibody titers in patients with and without ILD. ACPA, anticitrullinated protein antibody; Anti-CFFHP,­ IgG antibodies against chimeric fibrine/filagrine homocitrullinated peptide; Anti-FCS,­ IgG antibodies against carbamylated fetal calf serum; Anti-Fib,­ IgG antibodies against carbamylated fibrinogen; Anti-­FCS-­IgA, IgA antibodies against carbamylated fetal calf serum immunoglobulin A; AU, arbitrary units; CU, chemiluminescence units; ILD, interstitial lung disease; IU, international units; RA, rheumatoid arthritis; RF, rheumatoid factor.

590 Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 Rheumatoid arthritis

Table 3 Logistic regression models for the association between Anti-CarP­ specificities with RA-­ILD Main population Replication sample OR p>|z| 95% CI OR p>|z| 95% CI Anti-FCS­ 3.421 0.030 1.125 to 10.404 10.419 0.012 1.684 to 64.459 Age 1.068 0.008 1.017 to 1.120 1.045 0.211 0.974 to 1.122 RA disease duration 1.291 0.000 1.161 to 1.436 0.989 0.708 0.933 to 1.047 AC PA 0.525 0.300 0.155 to 1.775 0.708 0.809 0.043 to 11.624 RF 2.016 0.246 0.616 to 6.596 5.080 0.080 0.823 to 31.330 Sex 0.492 0.201 0.166 to 1.459 3.594 0.128 0.692 to 18.670 Smoking cumulative dose 1.023 0.145 0.992 to 1.055 0.962 0.061 0.923 to 1.001 Anti-Fib­ 2.846 0.095 0 .834 to 9.708 1.649 0.491 0.396 to 6.856 Age 1.070 0.006 1.020 to 1.123 1.047 0.142 0.984 to 1.115 RA disease duration 1.276 0.000 1.147 to 1.420 0.983 0.518 0.934 to 1.034 AC PA 0.618 0.427 0.188 to 2.026 1.768 0.641 0.160 to 19.419 RF 1.688 0.431 0.459 to 6.212 5.613 0.056 0.956 to 32.952 Sex 0.475 0.179 0.160 to 1.408 2.406 0.229 0.575 to 10.066 Smoking cumulative dose 1.026 0.089 0.996 to 1.057 0.969 0.083 0.935 to 1.004 Anti-CFFHP­ 3.115 0.039 1.061 to 9.138 1.494 0.534 0.421 to 5.291 Age 1.058 0.016 1.010 to 1.107 1.046 0.147 0.984 to 1.113 RA disease duration 1.285 0.000 1.157 to 1.426 0.985 0.586 0.937 to 1.037 AC PA 0.494 0.268 0.142 to 1.720 1.868 0.605 0.174 to 19.948 RF 2.617 0.104 0.819 to 8.357 6.231 0.037 1.113 to 34.888 Sex 0.382 0.092 0.124 to 1.171 2.030 0.321 0.501 to 8.215 Smoking cumulative dose 1.023 0.143 0.992 to 1.055 0.971 0.126 0.937 to 1.008 ACPA, anticitrullinated protein antibody; Anti-­CarP, anti-­carbamylated protein antibodies; Anti-­CFFHP, IgG antibodies against chimeric fibrine/filagrine homocitrullinated peptide; Anti-­FCS, IgG antibodies against carbamylated fetal calf serum; Anti-Fib,­ IgG antibodies against carbamylated fibrinogen; ILD, interstitial lung disease; RA, rheumatoid arthritis; RF, rheumatoid factor.

ACPA, RF and RA-ILD Anti-CarP and smoking No differences were found in the proportion of patients who Numerically higher proportions of all Anti-­CarP were observed tested positive for ACPA (78% vs 70%, p value: 0.29) or RF in ever and current smokers, although the difference was only (76% vs 59%, p value: 0.06) according to ILD status. Numeri- significant for Anti-FCS­ (58% vs 41% and 67% vs 45%, respec- cally higher titers of both antibodies (median (IQR), 674 (2215) tively), Anti-Fib­ (65% vs 48% and 77% vs 52%, respectively) vs 143 (1132) CU, p value: 0.07) and (105 (298) vs 34 (110) IU, and Anti-­FCS-IgA­ (37% vs 18% and 50% vs 22%), p<0.05 p value: 0.06) ACPA and RF, respectively) were observed in the for all comparisons. No differences in the smoking cumulative RA-ILD­ group (see table 2 and figure 2D,E), although statistical dose according to the Anti-­CarP status were observed. Details in significance was not reached. the unadjusted association between Anti-CarP­ and smoking are provided in online supplementary table S2 and S3. Overlap between Anti-CarP with ACPA and RF The overlap between Anti-CarP­ specificities in ACPA and RF positive patients is shown in table 4. Among those patients nega- Discussion tive to both RF and ACPA, a small subset of positive Anti-FCS­ This is the first study to analyse the association between Anti-­ (9%), Anti-­Fib (14%), Anti-­CFFHP (3%) and Anti-FCS-­ IgA­ CarP and RA-ILD.­ We found higher proportions and titers up to (6%) was observed. ACPA and RF titers were significantly higher nearly fourfold greater of different Anti-CarP­ specificities in the in patients seropositive for Anti-FCS,­ Anti-­Fib, Anti-­CFFHP and RA-ILD­ group when compared with patients without this lung Anti-­FCS-IgA­ (p<0.005 for all comparisons). involvement. All Anti-­CarP showed a consistent effect towards

Table 4 Overlap between Anti-­CarP specificities with ACPA and RF in the main population Anti-­FCS Anti-­Fib Anti-­CFFHP Anti-­FCS-­IgA Positive Negative Positive Negative Positive Negative Positive Negative n: 87 n: 92 n: 100 n: 79 n: 41 n: 138 n: 48 n: 131 ACPA Positive, n: 128 76 (59%) 52 (41%) 85 (66%) 43 (33%) 40 (31%) 88 (69%) 43 (34%) 85 (66%) Negative, n: 51 11 (22%) 40 (78%) 15 (29%) 36 (71%) 1 (2%) 50 (98%) 5 (10%) 46 (90%) Median titer (IQR) CU 760 (2034) 26 (473) 732 (2062) 28 (415) 994 (2278) 75 (929) 154 (313) 12 (167) RF Positive, n: 111 71 (64%) 40 (36%) 87 (78%) 24 (22%) 32 (29%) 79 (71%) 45 (41%) 66 (59%) Negative, n: 68 16 (24%) 52 (76%) 13 (19%) 55 (81%) 9 (13%) 59 (87%) 3 (4%) 65 (96%) Median titer (IQR) IU 102 (260) 18 (48) 115 (261) 12 (24) 105 (190) 32 (117) 184 (336) 0 (93) ACPA, anticitrullinated protein antibody; Anti-­CFFHP, IgG antibodies against chimeric fibrine/filagrine homocitrullinated peptide; Anti-FCS,­ IgG antibodies against carbamylated fetal calf serum; Anti-­FCS-­IgA, IgA antibodies against carbamylated fetal calf serum immunoglobulin A; Anti-Fib,­ IgG antibodies against carbamylated fibrinogen; CU, chemiluminescence units; IU, international units; RF, rheumatoid factor.

Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 591 Rheumatoid arthritis increasing the odds of ILD, even after adjustment for pertinent Although ACPA and Anti-­CarP present important differences confounders using logistic regression modelling. in their genetic40 and environmental41 backgrounds, both anti- Compared with the general population, RA entails a 50% body systems form part of the AMPA family and have great struc- higher risk of mortality.17 Over the past two decades, the tural similarity.36 As in previous studies,8 9 a substantial overlap mortality rate has declined, mostly due to a reduction in deaths between autoantibodies was observed in our cohort. However, related to cardiovascular disease and malignancies,18 19 possibly inhibition studies and the fact that a subset of ACPA negative attributed to an earlier diagnosis and treat-to-­ ­target strate- patients can harbour Anti-CarP­ (up to 29% in our population) gies. Conversely, incident deaths from pulmonary disease have support the hypothesis that the overlap is only partial.7 42 Recent 19 remained stable and may become the leading cause of death evidence indicates that exposure to a purified specific PTM such in RA patients. RA-ILD­ plays a major role in the burden of as homocitrulline, acetaldehyde or citrulline can induce AMPAs pulmonary-­related deaths and constitutes an unmet need, even against antigens containing other PTMs.43 Hence, a single PTM with available treatment options. RA-ILD­ patients usually have can generate a breaking of tolerance against multiple modified greater comorbidity and a higher incidence of infection, reflected antigens. 2 20 21 by a higher mortality and a greater disability. Therefore, Smoking, the leading environmental risk factor for RA and early detection is crucial in establishing an individualised treat- RA-ILD­ is also a known risk factor for breaking tolerance against ment strategy. multiple modified antigens.44 Smoking can induce carbamyla- Various novel candidates have been proposed for RA-ILD­ tion directly by inhalation of thiocyanate or via myeloperoxi- screening, including imaging techniques such as positron emis- dase release by neutrophils at inflammation sites.8 34 36 Like a 22 23 sion tomography and pulmonary ultrasound, or serum previous study,44 we found a higher seroprevalence of Anti-­CarP biomarkers such as the matrix metalloproteinases 7, surfactant 4 5 among smokers. There were no between-­group differences in protein D and Krebsvon del Lungen-6. RA-ILD­ has been asso- ever or current smoking according to the presence of ILD, yet ciated with different antimodified protein antibodies (AMPAs) a higher smoking cumulative dose was observed. No interaction based on citrulline24 25 and more recently, malondialdehyde– 26 between the smoking cumulative dose and Anti-­CarP specifici- acetaldehyde adducts. Screening techniques can increase ties was observed in the regression models, suggesting that the RA-ILD­ diagnostic accuracy, although a model based on features association between Anti-­CarP and ILD is not a simple surrogate used in clinical practice such as age, sex, smoking, RF and ACPA marker for smoking. has a potent screening capacity for symptomatic and subclinical Our study has some limitations. First, a relatively small RA-­ILD.4 Based on these findings, we decided to include these proportion of patients with RA-ILD­ was included, so we had features in addition to RA disease duration as adjusting variables to be conservative while building our models because of the in our models. risk of overfitting, so residual confounding cannot be ruled-­out. The relationship between ACPA and ILD has been addressed in Treatment and disease activity were not included; the former different studies with most showing a higher seroprevalence and due to concerns about the accuracy of timing and dosing and titers in RA-ILD­ cohorts.1 27–29 However, no association between current literature’s inability to validate a true relationship ACPA and RA-­ILD has been confirmed in other cohorts30 31 and between RA-ILD­ and methotrexate45 or bDMARDs.46–48 The up to 17% of patients with RA-­ILD are reported to be ACPA-­ latter, because of the design of the study with only a single negative.20 In a recent meta-analysis,­ ACPA positivity was asso- measurement, which has not shown to be relevant in previous ciated with a higher risk for ILD (OR: 4.7, 95% CI: 2.1 to 10.6) reports,49 50 as opposed to its longitudinal evaluation.51 Likewise, although a single study contributed almost 75% to the weighted average and was not specifically concerned with RA-ILD.­ 29 We we cannot rule-­out subclinical RA-ILD­ in the control group, found that ACPA and RF were numerically higher in patients with mainly due to the lack of a universally accepted evidence-­based ILD, but the differences were not significant, unlike Anti-CarP.­ screening approach. Finally, most of our patients are of Mediter- The association between Anti-­CarP and RA-ILD­ observed in ranean Caucasian origin, which has been associated with a less aggressive RA course and a lower prevalence of extra-­articular the main population and the replication sample opens a debate 52 on the potential role of homocitrullination (carbamylation) and disease. its antibody response on its development. So far, no studies In conclusion, a robust association between different Anti-CarP­ have evaluated homocitrulline in lung tissue of RA patients as and RA-ILD­ was found after adjusting for multiple confounders opposed to citrulline.32 However, carbamylated peptides have including ACPA and RF. These findings pose the debate whether been detected in the synovial of RA patients33 and lung tissue a link between homocitrullination and the development of this from smokers and asthma patients.34 35 Interestingly, Anti-­CarP devastating extra-articular­ manifestation exists. However, our have been found in subjects with bronchiectasis, cystic fibrosis results should be interpreted with caution and further studies and chronic obstructive pulmonary disease with no smoking or validating our findings are needed. RA history.11 12 Correction notice This article has been corrected since it published Online First. Carbamylation can change the structure, function and anti- The second affiliation has been updated. body antigen binding avidity of a wide range of proteins.36 Thus, Twitter Raul Castellanos-­Moreira @raul_cast_morei and Sebastian Cruz Rodríguez-­ Anti-­CarP may target homocitrullinated peptides found in the García @sdlcrodriguez lung and promote a subsequent inflammatory response, which is Acknowledgements The authors wish to thank Loreto Carmona and Miguel essential in the genesis and progression of ILD.37 38 An alterna- Angel Descalzo for their assistance in the methodology and the Interstitial tive explanation is a direct pulmonary insult by the inflammatory Lung Disease Committee staff at the Hospital Clinic of Barcelona for their cascade precipitated by an increased sensitivity of proteinases collaboration. and the presence of free radicals and other inflammatory medi- Contributors RC-­M, SCR-­G, IH and RS contributed to the conception and study ators enhanced by carbamylated free amino acids found in the design. RC-­M, JR, JG-­P, VR-­E, IC-­S, SH and JDC contributed to data collection. 36 RC-­M, SCR-­G and MJG analysed the data. RC-­M, SCR-­G, VR-­E and IH contributed lung. Notably, the association between ILD and Anti-CarP-­ ­IgA to interpretation of the data. RC-­M, SCR-­G and RS wrote the first version of the presented the greatest effect size, fuelling the hypothesis that manuscript and AC, JR, JG-­P, JDC, VR-­E, IC-­S, SH and IH revised it critically. All RA-related­ antibodies originate in the respiratory mucosa.39 authors read and approved the final manuscript.

592 Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 Rheumatoid arthritis

Funding Financial support from the Hospital Clinic of Barcelona, Research, 19 abhishek A, Nakafero G, Kuo C-­F, et al. Rheumatoid arthritis and excess mortality: Innovation and Education Department (Grant # 37 933 to RC-­M and the Spanish down but not out. A primary care cohort study using data from clinical practice Ministry of Economy, Industry and Competitiveness and the European Regional research Datalink. Rheumatology 2018;57:977–81. Development Fund (Grant # RTI2018-094120-­B-­I00 to IH). 20 solomon JJ, Chung JH, Cosgrove GP, et al. Predictors of mortality in rheumatoid arthritis-­associated interstitial lung disease. Eur Respir J 2016;47:588–96. Competing interests None declared. 21 Zamora-­Legoff JA, Krause ML, Crowson CS, et al. Risk of serious infection in patients Patient and public involvement Patients and/or the public were not involved in with rheumatoid arthritis-­associated interstitial lung disease. Clin Rheumatol the design, or conduct, or reporting, or dissemination plans of this research. 2016;35:2585–9. Patient consent for publication Not required. 22 Uehara T, Takeno M, Hama M, et al. Deep-­inspiration breath-­hold 18F-­FDG-P­ ET/CT is useful for assessment of connective tissue disease associated interstitial pneumonia. Ethics approval The study was conducted in accordance with the Declaration of Mod Rheumatol 2016;26:121–7. Helsinki and was approved by the Hospital Clinic of Barcelona Ethics Committee 23 Xie HQ, Zhang WW, Sun DS, et al. A simplified lung ultrasound for the diagnosis of (approval number 2017/0679). interstitial lung disease in connective tissue disease: a meta-­analysis. Arthritis Res Provenance and peer review Not commissioned; externally peer reviewed. Ther 2019;21. 24 Harlow L, Rosas IO, Gochuico BR, et al. Identification of citrullinated Hsp90 isoforms Data availability statement Data are available upon reasonable request. Data as novel autoantigens in rheumatoid arthritis-­associated interstitial lung disease. is available upon reasonable request, all data relevant to the study are included in Arthritis Rheum 2013;65:869–79. the article. 25 alunno A, Bistoni O, Pratesi F, et al. Anti-­citrullinated alpha enolase antibodies, interstitial lung disease and bone erosion in rheumatoid arthritis. Rheumatology ORCID iDs 2018;57:850–5. Raul Castellanos-Moreira­ http://orcid.​ ​org/0000-​ ​0002-​4104-​4101 26 england BR, Duryee MJ, Roul P, et al. Malondialdehyde-­Acetaldehyde adducts and Sebastian Cruz Rodríguez-­García http://orcid.​ ​org/0000-​ ​0002-​7773-​151X antibody responses in rheumatoid arthritis-­associated interstitial lung disease. Arthritis Rheumatol 2019;71:1483–93. References 27 Kelly CA, Saravanan V, Nisar M, et al. Rheumatoid arthritis-­related interstitial 1 Duarte A, Porter J, Leandro M. The lung in a cohort of rheumatoid arthritis patients— lung disease: associations, prognostic factors and physiological and radiological an overview of different types of involvement and treatment. Rheumatology characteristics--a large multicentre UK study. Rheumatology 2014;53:1676–82. 2019;1:2031–8. 28 Giles JT, Danoff SK, Sokolove J, et al. Association of fine specificity and repertoire 2 Hyldgaard C, Hilberg O, Pedersen AB, et al. A population-­based cohort study of expansion of anticitrullinated peptide antibodies with rheumatoid arthritis associated rheumatoid arthritis-­associated interstitial lung disease: comorbidity and mortality. interstitial lung disease. Ann Rheum Dis 2014;73:1487–94. Ann Rheum Dis 2017;76:1700–6. 29 Zhu J, Zhou Y, Chen X, et al. A metaanalysis of the increased risk of rheumatoid 3 spagnolo P, Lee JS, Sverzellati N, et al. The lung in rheumatoid arthritis. Arthritis arthritis-­related pulmonary disease as a result of serum anticitrullinated protein Rheumatol 2018;70:1544–54. antibody positivity. J Rheumatol 2014;41:1282–9. 4 Doyle TJ, Patel AS, Hatabu H, et al. Detection of rheumatoid Arthritis-­Interstitial 30 inui N, Enomoto N, Suda T, et al. Anti-­Cyclic citrullinated peptide antibodies in lung lung disease is enhanced by serum biomarkers. Am J Respir Crit Care Med diseases associated with rheumatoid arthritis. Clin Biochem 2008;41:1074–7. 2015;191:1403–12. 31 skare TL, Nakano I, Escuissiato DL, et al. Pulmonary changes on high-resolution­ 5 Kinoshita F, Hamano H, Harada H, et al. Role of KL-6 in evaluating the disease severity computed tomography of patients with rheumatoid arthritis and their association of rheumatoid lung disease: comparison with HRCT. Respir Med 2004;98:1131–7. with clinical, demographic, serological and therapeutic variables. Rev Bras Reumatol 6 Wang Z, Nicholls SJ, Rodriguez ER, et al. Protein carbamylation links inflammation, 2011;51:325–30. smoking, uremia and atherogenesis. Nat Med 2007;13:1176–84. 32 Ytterberg AJ, Joshua V, Reynisdottir G, et al. Shared immunological targets in the lungs 7 shi J, Knevel R, Suwannalai P, et al. Autoantibodies recognizing carbamylated proteins and joints of patients with rheumatoid arthritis: identification and validation. Ann are present in sera of patients with rheumatoid arthritis and predict joint damage. Rheum Dis 2015;74:1772–7. Proc Natl Acad Sci U S A 2011;108:17372–7. 33 Turunen S, Huhtakangas J, Nousiainen T, et al. Rheumatoid arthritis antigens 8 Truchetet M-­E, Dublanc S, Barnetche T, et al. Association of the presence of homocitrulline and citrulline are generated by local myeloperoxidase and peptidyl Anti-­Carbamylated protein antibodies in early arthritis with a poorer clinical and arginine deiminases 2, 3 and 4 in rheumatoid nodule and synovial tissue. Arthritis Res radiologic outcome: data from the French ESPOIR cohort. Arthritis Rheumatol Ther 2016;18:239. 2017;69:2292–302. 34 lugli EB, Correia RESM, Fischer R, et al. Expression of citrulline and homocitrulline 9 Humphreys JH, Verheul MK, Barton A, et al. Anticarbamylated protein antibodies are residues in the lungs of non-­smokers and smokers: implications for autoimmunity in associated with long-­term disability and increased disease activity in patients with rheumatoid arthritis. Arthritis Res Ther 2015;17:9. early inflammatory arthritis: results from the Norfolk arthritis register. Ann Rheum Dis 35 Wang Z, DiDonato JA, Buffa J, et al. Eosinophil peroxidase catalyzed protein 2016;75:1139–44. carbamylation participates in asthma. J Biol Chem 2016;291:22118–35. 10 Vidal-­Bralo L, Perez-­Pampin E, Regueiro C, et al. Anti-carbamylated­ protein 36 shi J, van Veelen PA, Mahler M, et al. Carbamylation and antibodies against autoantibodies associated with mortality in Spanish rheumatoid arthritis patients. carbamylated proteins in autoimmunity and other pathologies. Autoimmun Rev PLoS One 2017;12:e0180144. 2014;13:225–30. 11 Janssen KMJ, de Smit MJ, Brouwer E, et al. Rheumatoid arthritis–associated 37 Bringardner BD, Baran CP, Eubank TD, et al. The role of inflammation in autoantibodies in non–rheumatoid arthritis patients with mucosal inflammation: a the pathogenesis of idiopathic pulmonary fibrosis. Antioxid Redox Signal case–control study. Arthritis Res Ther 2015;17:174. 2008;10:287–302. 12 Verheul MK, van Erp SJH, van der Woude D, et al. Anti-carbamylated­ protein 38 Racanelli AC, Kikkers SA, Choi AMK, et al. Autophagy and inflammation in chronic antibodies: a specific hallmark for rheumatoid arthritis. Comparison to conditions respiratory disease. Autophagy 2018;14:221–32. known for enhanced carbamylation; renal failure, smoking and chronic inflammation. 39 solomon JJ, Matson S, Kelmenson LB, et al. Iga antibodies directed against Ann Rheum Dis 2016;75:1575–6. citrullinated protein antigens are elevated in patients with idiopathic pulmonary 13 Travis WD, Costabel U, Hansell DM, et al. An officialA merican thoracic Society/ fibrosis. CHEST 2020. European respiratory Society statement: update of the International multidisciplinary 40 Jiang X, Trouw LA, van Wesemael TJ, et al. Anti-­CarP antibodies in two large classification of the idiopathic interstitial pneumonias. Am J Respir Crit Care Med cohorts of patients with rheumatoid arthritis and their relationship to genetic 2013;188:733–48. risk factors, cigarette smoking and other autoantibodies. Ann Rheum Dis 14 Pérez ML, Gómara MJ, Ercilla G, et al. Antibodies to citrullinated human fibrinogen 2014;73:1761–8. synthetic peptides in diagnosing rheumatoid arthritis. J Med Chem 2007;50:3573–84. 41 Janssen KMJ, de Smit MJ, Brouwer E, et al. Rheumatoid arthritis-­associated 15 Malakoutikhah M, Gómara MJ, Gómez-Puerta­ JA, et al. The use of chimeric vimentin autoantibodies in non-­rheumatoid arthritis patients with mucosal inflammation: a citrullinated peptides for the diagnosis of rheumatoid arthritis. J Med Chem case-control­ study. Arthritis Res Ther 2015;17:174. 2011;54:7486–92. 42 shi J, Willemze A, Janssen GMC, et al. Recognition of citrullinated and carbamylated 16 van der Woude D, Rantapää-Dahlqvist S, Ioan-­Facsinay A, et al. Epitope spreading proteins by human antibodies: specificity, cross-­reactivity and the ’AMC-­Senshu’ of the anti-­citrullinated protein antibody response occurs before disease onset method. Ann Rheum Dis 2013;72:148–50. and is associated with the disease course of early arthritis. Ann Rheum Dis 43 Kampstra ASB, Dekkers JS, Volkov M, et al. Different classes of anti-­modified 2010;69:1554–61. protein antibodies are induced on exposure to antigens expressing only one type of 17 Dadoun S, Zeboulon-­Ktorza N, Combescure C, et al. Mortality in rheumatoid arthritis modification. Ann Rheum Dis 2019;78:908–16. over the last fifty years: systematic review and meta-­analysis. Joint Bone Spine 44 van Wesemael TJ, Ajeganova S, Humphreys J, et al. Smoking is associated with the 2013;80:29–33. concurrent presence of multiple autoantibodies in rheumatoid arthritis rather than 18 Zhang Y, Lu N, Peloquin C, et al. Improved survival in rheumatoid arthritis: a general with anti-­citrullinated protein antibodies per se: a multicenter cohort study. Arthritis population-­based cohort study. Ann Rheum Dis 2017;76:408–13. Res Ther 2016;18:285.

Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 593 Rheumatoid arthritis

45 Fragoulis GE, Conway R, Nikiphorou E. Methotrexate and interstitial lung disease: 49 Koduri G, Norton S, Young A, et al. Interstitial lung disease has a poor prognosis controversies and questions. A narrative review of the literature. Rheumatology in rheumatoid arthritis: results from an inception cohort. Rheumatology 2019;58:1900–6. 2010;49:1483–9. 46 Herrinton LJ, Harrold LR, Liu L, et al. Association between anti-­TNF-α therapy and 50 Restrepo JF, del Rincón I, Battafarano DF, et al. Clinical and laboratory factors interstitial lung disease. Pharmacoepidemiol Drug Saf 2013;22:394–402. associated with interstitial lung disease in rheumatoid arthritis. Clin Rheumatol 47 Theander L, Nyhäll-­Wåhlin B-­M, Nilsson Jan-Åke, et al. Severe Extraarticular 2015;34:1529–36. manifestations in a community-­based cohort of patients with rheumatoid arthritis: risk 51 sparks JA, He X, Huang J, et al. Rheumatoid arthritis disease activity predicting factors and incidence in relation to treatment with tumor necrosis factor inhibitors. J incident clinically apparent rheumatoid Arthritis–Associated interstitial lung disease: a Rheumatol 2017;44:981–7. prospective cohort study. Arthritis Rheumatol 2019;71:1472–82. 48 Curtis JR, Sarsour K, Napalkov P, et al. Incidence and complications of interstitial lung 52 Carmona L, González-­Alvaro I, Balsa A, et al. Rheumatoid arthritis in Spain: disease in users of tocilizumab, rituximab, abatacept and anti-­tumor necrosis factor α occurrence of extra-­articular manifestations and estimates of disease severity. Ann agents, a retrospective cohort study. Arthritis Res Ther 2015;17:319. Rheum Dis 2003;62:897–900.

594 Castellanos-­Moreira R, et al. Ann Rheum Dis 2020;79:587–594. doi:10.1136/annrheumdis-2019-216709 Spondyloarthritis

Clinical science Dual neutralisation of interleukin-­17A and interleukin-­17F with bimekizumab in patients with active ankylosing spondylitis: results from a 48-­week phase IIb, randomised, double-­blind, placebo-­ controlled, dose-­ranging study Désirée van der Heijde ‍ ‍ ,1 Lianne S Gensler,2 Atul Deodhar,3 Xenofon Baraliakos ‍ ‍ ,4 Denis Poddubnyy ‍ ‍ ,5 Alan Kivitz,6 Mary Katherine Farmer,7 Dominique Baeten,8 Nadine Goldammer,9 Jason Coarse,7 Marga Oortgiesen,7 Maxime Dougados10,11

Handling editor Josef S Abstract Key messages Smolen Objectives Bimekizumab selectively neutralises both interleukin (IL)-­17A and IL-­17F. We report efficacy and ►► Additional material is What is already known about this subject? published online only. To view, safety in a phase IIb dose-­ranging study in patients with ►► There remains a need for treatment options in please visit the journal online active ankylosing spondylitis (AS). ankylosing spondylitis (AS) that can provide (http://dx.​ ​doi.org/​ ​10.​1136/​ Methods Adults with AS (fulfilling modified New York sustained, long-­term disease control and annrheumdis-2020-​ ​216980). criteria) were randomised 1:1:1:1:1 to bimekizumab 16 improve patient quality of life. For numbered affiliations see mg, 64 mg, 160 mg, 320 mg or placebo every 4 weeks for 12 weeks (double-­blind period). At week 12, patients end of article. What does this study add? receiving bimekizumab 16 mg, 64 mg or placebo were ►► Bimekizumab, a monoclonal antibody that Correspondence to re-­randomised 1:1 to bimekizumab 160 mg or 320 mg neutralises both interleukin (IL)-­17A and IL-­17F, Professor Désirée van der every 4 weeks to week 48; other patients continued on has shown clinically relevant improvements in Heijde, Department of their initial dose (dose-­blind period). The primary end Rheumatology, Leiden University both psoriasis and psoriatic arthritis, leading to point was Assessment of SpondyloArthritis international Medical Center, Albinusdreef 2, its evaluation in other IL-17-mediated­ diseases. 2333 ZA Leiden, The Society (ASAS) 40 response at week 12 (non-­responder ► This is the first study to assess bimekizumab in Netherlands; imputation (NRI) for missing data). ► patients with active AS. mail@​ ​dvanderheijde.nl​ Results 303 patients were randomised: bimekizumab ► A significant dose-­response was observed with 16 mg (n=61), 64 mg (n=61), 160 mg (n=60), 320 mg ► Received 14 January 2020 bimekizumab for ASAS40 at week 12 (p<0.05), (n=61) or placebo (n=60). At week 12, significantly Revised 25 February 2020 with a rapid onset and greater ASAS40 Accepted 25 February 2020 more bimekizumab-­treated patients achieved ASAS40 response rates for all doses of bimekizumab Published Online First vs placebo (NRI: 29.5%–46.7% vs 13.3%; p<0.05 versus placebo, which continued to increase to 5 April 2020 all comparisons; OR vs placebo 2.6–5.5 (95% CI 1.0 week 48. to 12.9)). A significant dose-­response was observed ► A similar pattern was observed across (p<0.001). The primary end point was supported by all ► secondary outcomes, representing secondary efficacy outcomes.A t week 48, 58.6% and improvements in quality of life measures versus 62.3% of patients receiving bimekizumab 160 and 320 placebo and over time. mg throughout the study achieved ASAS40, respectively ► Safety was in line with previous bimekizumab (NRI); similar ASAS40 response rates were observed in ► studies and comparable with the IL-­17A re-­randomised patients. During the double-­blind period, inhibitor class. treatment-­emergent adverse events occurred in 26/60 (43.3%) patients receiving placebo and 92/243 (37.9%) How might this impact on clinical practice or receiving bimekizumab. future developments? Conclusions Bimekizumab provided rapid and ► Results from this study contribute to the sustained improvements in key outcome measures in ► growing body of evidence supporting dual patients with active AS, with no unexpected safety neutralisation of IL-­17A and IL-­17F with findings versus previous studies. © Author(s) (or their bimekizumab as a novel therapeutic option for Trial registration number NCT02963506. employer(s)) 2020. Re-­use the treatment of AS. permitted under CC BY-­NC. No ► commercial re-­use. See rights ► Phase III studies in patients with AS and and permissions. Published non-­radiographic axial spondyloarthritis are by BMJ. Introduction ongoing. To cite: van der Heijde D, Ankylosing spondylitis (AS) is a chronic disease, Gensler LS, Deodhar A, characterised by inflammation of the axial skeleton.1 by other manifestations such as peripheral enthesitis et al. Ann Rheum Dis It is also referred to as radiographic axial spondy- and arthritis, uveitis, inflammatory bowel disease 2020;79:595–604. loarthritis (r-­axSpA). AS can often be accompanied (IBD) and psoriasis.1 2 Expression of human

van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 595 Spondyloarthritis leucocyte antigen (HLA)-B27­ is strongly associated with the Participants disease, and patients often have elevated levels of inflammatory Eligible patients were ≥18 years of age with AS, as determined markers such as C reactive protein (CRP).1 Patients experience by documented radiographic sacroiliitis (X-rays­ were centrally chronic pain and functional impairment, impacting on sleep, read by two readers and a third in case of discrepancy), fulfilling daily activities and overall quality of life,3–5 with some patients the modified New York (mNY) criteria for AS,23 symptom dura- experiencing physical disability due to structural damage of the tion of ≥3 months, age at onset of <45 years, with active disease spine.6 as defined by Bath Ankylosing Spondylitis Disease Activity Index Non-­steroidal anti-­inflammatory drugs (NSAIDs) are a first-­ (BASDAI) score of ≥4 and spinal pain ≥4 on a 0–10 numer- line treatment to provide symptomatic relief to patients with AS.7 ical rating scale (NRS) (from BASDAI question 2). Patients However, response to NSAIDs may be inadequate or they may were required to have at least one of the following: inadequate be contraindicated. Conventional synthetic disease-modifying­ response to NSAIDs (defined as a lack of response for ≥4 weeks antirheumatic drugs, such as methotrexate or sulfasalazine, of continuous NSAID therapy or the lack of response to ≥2 are not efficacious in axial disease, although the latter may be NSAIDs at the maximum tolerated dose for ≥4 weeks), intol- effective for patients with peripheral arthritis.7 Tumour necrosis erance to ≥1 NSAID or contraindication(s) to NSAIDs. Prior factor (TNF) inhibitors are the first-­line biologic in patients with treatment with up to one TNF inhibitor was permitted, which high disease activity, but not all patients achieve adequate disease must have been discontinued because of inadequate response, 8 9 control or tolerate treatment. Interleukin (IL)-17A­ inhibitors intolerance or loss of access. 10 11 are effective second-­line therapies ; however, some patients Patients with active/symptomatic Crohn’s disease or ulcerative may still experience unsatisfactory response and require alterna- colitis (UC) were excluded; previous history of Crohn’s disease tive treatments. or UC was not an exclusion criterion. Other exclusion criteria The IL-17 axis represents an established target in AS treat- were total ankylosis of the spine, a concurrent or history of malig- ment, and inflammation is associated with an increase in IL-17-­ nancy during the past 5 years, a diagnosis of other inflammatory 12 producing innate immune cells. Two members of the IL-17 conditions (eg, rheumatoid arthritis), active infection or infec- cytokine family, IL-17A­ and IL-17F,­ share ~50% structural tion requiring antibiotics within 2 weeks of baseline, a history homology and have similar pro-­inflammatory function, signal- of chronic or recurrent infections or a serious/life-threatening­ 13 14 ling via the same receptor complex. Preclinical evidence infection within 6 months of baseline, presence of significant, from human in vitro assays has shown that IL-17A­ and IL-17F­ uncontrolled neuropsychiatric disorder, active suicidal ideation, cooperate with other mediators of inflammation, such as TNF, or positive suicide behaviour within the past 6 months. to amplify inflammatory responses.15 The contribution of IL-­17F, in addition to IL-­17A, to pathological bone formation has been shown in a human periosteum-­derived stem cell model Interventions of osteogenic differentiation, indicating that neutralisation of Initially, patients were randomised 1:1:1:1:1 to receive subcuta- both cytokines inhibits this process to a greater extent than neous bimekizumab 16 mg, 64 mg, 160 mg, 320 mg or placebo IL-­17A alone.16 17 In addition, levels of IL-17A­ and IL-17F­ have every 4 weeks. At the end of the double-­blind period at week 12, been found to be higher in the serum of patients with AS versus following all assessments, patients were re-­allocated treatment healthy controls, correlating with markers of systemic inflam- for the dose-­blind period as follows: patients initially receiving mation.18 19 placebo, bimekizumab 16 mg or 64 mg were randomised 1:1 to Bimekizumab is a monoclonal antibody developed to selec- receive bimekizumab 160 or 320 mg every 4 weeks through to tively neutralise both IL-­17A and IL-­17F,20 and has been previ- week 48. Patients in the bimekizumab 160 and 320 mg groups ously evaluated in a first-in-­ ­human study in patients with mild continued on the same schedule through to week 48 (figure 1; psoriasis, a proof-of-­ ­concept study in patients with moderate-­ see online supplementary methods for additional information on to-­severe psoriatic arthritis (PsA) and a phase IIb dose-­ranging interventions and randomisation and masking). study in patients with moderate-­to-­severe plaque psoriasis.15 20 21 The overlap in symptoms between psoriasis, PsA and AS,2 22 and involvement of the IL-17 pathway, led to the evaluation of bime- Outcomes kizumab as a potential therapeutic option in AS. Here, we report The primary efficacy end point was the percentage of patients results of the phase IIb BE AGILE study, the first dose-­ranging with an Assessment of SpondyloArthritis international Society clinical study evaluating the efficacy and safety of bimekizumab (ASAS) 40 response at week 12, defined according to the ASAS 24 in patients with active AS. Handbook. Secondary efficacy end points were ASAS20 response, ASAS5/6 response, change from baseline in BASDAI, Bath Methods Ankylosing Spondylitis Functional Index (BASFI) and Anky- Study design losing Spondylitis Disease Activity Score with CRP (ASDAS), This phase IIb, randomised, 48-­week, double-blind,­ placebo-­ all at week 12. Additional efficacy end points included ASAS40, controlled, parallel-group,­ dose-­ranging study (BE AGILE) ASAS20, ASDAS inactive disease (ID; <1.3), ASDAS low disease (online supplementary files 1; 2) was conducted at 74 centres activity (ASDAS-­LDA; 1.3–<2.1),25 ASDAS major improvement across 10 countries in Europe and the USA. (ASDAS-MI;­ reduction ≥2.0 from baseline), ASAS partial remis- sion (PR), change from baseline in the linear version of the Bath Patient and public involvement Ankylosing Spondylitis Metrology Index (BASMI)26 and change Patients with AS were consulted to understand treatment needs from baseline in Maastricht Ankylosing Spondylitis Enthesitis and recommend ways to facilitate study participation while Score (MASES) index. Additional outcomes included measure- minimising the burden of study visits. Study participants were ment of high-­sensitivity CRP (hs-­CRP) levels in blood samples recruited by the study sites and provided written consent to and MRI performed in a subset of patients (20 planned per treat- participate. ment group) to evaluate the effect of bimekizumab on objective

596 van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 Spondyloarthritis

Figure 1 (A) ASAS40 response at week 12 (primary efficacy end point; AS,F NRI). *P value vs placebo calculated from a logistic regression model including fixed effects for treatment, geographic region and prior TNF inhibitor exposure; *p<0.05, **p<0.01, ***p<0.001. (B) ASAS40 responses over 48 weeks; FAS, NRI (weeks 0–12); DBS, NRI (weeks 12–48). (C) ASDAS over time; FAS, MI (weeks 0–12); DBS, MI (weeks 12–48). ASAS40, Assessment of SpondyloArthritis international Society improvement of ≥40%; ASDAS, Ankylosing Spondylitis Disease Activity Score; BKZ, bimekizumab; DBS, dose-­blind set; FAS, full analysis set; MI, multiple imputation; NRI, non-responder­ imputation. van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 597 Spondyloarthritis signs of inflammation (see online supplementary methods for The primary analysis was based on a Mantel-Haenszel­ test30 additional information). and modified ridit scores with the corresponding p value. Non-­ Patient-reported­ outcomes (PROs) included change from responder imputation (NRI) was used to account for missing baseline in Ankylosing Spondylitis Quality of Life (ASQoL), data. Pairwise comparisons of each bimekizumab dose versus 36-­Item Short Form Survey (SF-36) and sleep quality (Medical placebo for ASAS40 at week 12 were based on a logistic regres- Outcomes Study (MOS)-12 item scale). Post hoc analyses sion model with fixed effects for treatment, region and prior included patients achieving a ≥50% improvement in BASDAI TNF inhibitor exposure. Multiplicity was accounted for using response (BASDAI 50) and change from baseline in fatigue a fixed sequence testing procedure with each bimekizumab (BASDAI question 1). dose tested sequentially versus placebo from the highest to Safety monitoring included incidence of adverse events (AEs), lowest dose; each test was only conducted if the previous test serious AEs (SAEs), withdrawal due to AEs and AEs of special reached significance at a two-­sided significance level of α=0.05, interest. AEs identified for additional safety monitoring included: to control the overall type 1 error rate. Further information is serious infections including opportunistic and tuberculosis, cyto- provided in the online supplementary files 1 and supplemen- penia, severe hypersensitivity reactions, IBD, suicide ideation or tary file 2. Other categorical end points (ASAS20 response and behaviour, depression and anxiety, major cardiovascular events ASAS5/6) were analysed as for the primary efficacy end point, (MACE), malignancies and liver function test changes. using NRI to account for missing patient data. For continuous end points (ASDAS, BASDAI and BASFI), multiple imputation Statistical analysis (MI) was used to account for missing patient data. Data for the The sample size was calculated using a two-­sided test for secondary and additional efficacy end points are descriptive detecting a linear trend across proportions at a significance level only. Post hoc analyses and safety data were summarised using of 0.05.27 Assuming 57 patients in each treatment group, the descriptive statistics by each visit. test for detecting the overall dose response based on ASAS40 treatment response was powered at >99%. Sample size calcula- tions were based on ASAS40 response data from phase III studies Results of TNF inhibitors and secukinumab in patients with AS,11 28 29 In total, 601 patients were screened, with 303 randomised and performed using nQuery Advisor 7.0. included in the FAS and SS, and 282 in the PPS. The majority The full analysis set (FAS) consisted of all randomised patients of screening failures were patients not fulfilling radiographic who received ≥1 dose of study drug and had a valid measure- sacroiliitis according to mNY criteria (online supplementary ment of the ASAS components at baseline. All patients starting figure 2). the dose-­blind period and who received ≥1 dose of study drug The majority of patients (98.0%; 297/303) completed the were included in the dose-­blind set. The per-protocol­ set (PPS) double-­blind period (online supplementary figure 2); 89.5% consisted of all patients in the FAS who had no important (265/296) completed the dose-­blind period. protocol deviation affecting the primary efficacy end point. The Patient demographics and baseline disease characteristics were safety set (SS) consisted of all randomised patients who received similar across dose groups (table 1). The study population was at least one dose of study drug. typical of a cohort of patients with highly active, established AS;

Table 1 Patient demographics and baseline disease characteristics (FAS) Placebo BKZ 16 mg BKZ 64 mg BKZ 160 mg BKZ 320 mg All patients Q4W (n=60) Q4W (n=61) Q4W (n=61) Q4W (n=60) Q4W (n=61) (n=303) Age, years, mean (SD) 39.7 (10.3) 43.3 (12.6) 40.4 (10.9) 42.4 (13.1) 45.0 (11.4) 42.2 (11.8) Sex (male), n (%) 49 (81.7) 53 (86.9) 52 (85.2) 52 (86.7) 50 (82.0) 256 (84.5) Caucasian, n (%) 60 (100) 58 (95.1) 60 (98.4) 59 (98.3) 61 (100) 298 (98.3) HLA-­B27 positive, n (%) 57 (95.0) 51 (83.6) 56 (91.8) 52 (86.7) 54 (88.5) 270 (89.1) Time since onset of first symptoms, years, mean (SD) 14.1 (8.4) 16.2 (10.6) 12.4 (8.3) 14.8 (10.3) 15.3 (10.6) 14.6 (9.7) Time since diagnosis, years, mean (SD) 6.6 (7.2) 8.0 (9.4) 7.3 (7.8) 8.8 (9.2) 8.8 (8.8) 7.9 (8.5) ASDAS, mean (SD) 3.8 (0.9) 3.9 (0.7) 4.2 (0.8) 3.9 (0.8) 3.9 (0.7) 3.9 (0.8) hs-­CRP, mg/L, mean (SD) 17.6 (24.6) 15.2 (17.7) 23.5 (21.6) 20.5 (19.3) 18.4 (20.6) 19.0 (20.9) BASDAI, mean (SD) 6.5 (1.4) 6.7 (1.4) 6.7 (1.3) 6.3 (1.3) 6.5 (1.6) 6.5 (1.4) BASFI, mean (SD) 5.6 (2.0) 5.9 (1.7) 6.0 (1.8) 5.6 (2.2) 5.9 (2.0) 5.8 (2.0) BASMI, mean (SD) 4.4 (1.6) 4.8 (1.7) 4.7 (1.7) 4.6 (1.8) 4.8 (1.8) 4.7 (1.7) Spinal pain score, mean (SD) 7.0 (1.7) 7.2 (1.9) 7.4 (1.6) 6.6 (2.0) 7.3 (1.5) 7.1 (1.7) PGADA, mean (SD) 7.0 (1.7) 7.1 (1.5) 7.3 (1.6) 6.5 (1.8) 7.1 (1.9) 7.0 (1.7) Previous TNF inhibitor therapy, n (%) 7 (11.7) 8 (13.1) 7 (11.5) 7 (11.7) 5 (8.2) 34 (11.2) Current NSAID therapy, n (%) 1 51 (85.0) 53 (86.9) 50 (82.0) 54 (90.0) 56 (91.8) 264 (87.1) 2 2 (3.3) 3 (4.9) 1 (1.6) 1 (1.7) 1 (1.6) 8 (2.6) Receiving csDMARDs, n (%) 13 (21.7) 9 (14.8) 18 (29.5) 18 (30.0) 21 (34.4) 79 (26.1) ASDAS, Ankylosing Spondylitis Disease Activity Score; BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index; BASMI, Bath Ankylosing Spondylitis Metrology Index; BKZ, bimekizumab; CRP, C reactive protein; csDMARD, conventional synthetic disease-modifying­ antirheumatic drug; HLA-­B27, human leucocyte antigen-­B27; NSAID, non-­steroidal anti-­inflammatory drug; PGADA, patient global assessment of disease activity; Q4W, every 4 weeks; TNF, tumour necrosis factor.

598 van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 Spondyloarthritis

Table 2 Primary and secondary efficacy end points (week 12; FAS, NRI/MI) Placebo Q4W (n=60) BKZ 16 mg Q4W (n=61) BKZ 64 mg Q4W (n=61) BKZ 160 mg Q4W (n=60) BKZ 320 mg Q4W (n=61) ASAS40*, n (%) Week 12 8 (13.3) 18 (29.5) 26 (42.6) 28 (46.7) 28 (45.9) ASAS20*, n (%) Week 12 17 (28.3) 25 (41.0) 38 (62.3) 35 (58.3) 44 (72.1) ASAS5/6*, n (%) Week 12 4 (6.7) 18 (29.5) 30 (49.2) 32 (53.3) 33 (54.1) BASDAI†, mean (SD) Baseline 6.5 (1.4) 6.7 (1.4) 6.7 (1.3) 6.3 (1.3) 6.5 (1.6) Week 12 5.5 (2.2) 5.0 (2.1) 3.9 (2.1) 3.8 (2.0) 3.7 (2.1) Change from baseline −1.0 (1.7) −1.7 (2.3) data-fill="true"−2.7­ (2.2) −2.5 (1.8) −2.9 (2.2) BASFI, mean (SD), Baseline 5.6 (2.0) 5.9 (1.7) 6.0 (1.8) 5.6 (2.2) 5.9 (2.0) Week 12 5.0 (2.4) 4.6 (2.4) 4.1 (2.3) 3.9 (2.2) 3.7 (2.5) Change from baseline −0.6 (1.9) −1.4 (2.2) −1.9 (2.4) −1.7 (1.8) −2.2 (2.0) ASDAS†, mean (SD), Baseline 3.8 (0.9) 3.9 (0.7) 4.2 (0.8) 3.9 (0.8) 3.9 (0.7) Week 12 3.5 (1.1) 3.0 (0.9) 2.5 (0.9) 2.5 (1.0) 2.4 (0.9) Change from baseline −0.4 (0.7) −0.9 (1.0) −1.7 (1.1) −1.4 (0.9) −1.5 (0.9) *NRI. †MI. ASAS, Assessment of SpondyloArthritis international Society; ASDAS, Ankylosing Spondylitis Disease Activity Score; BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index; BKZ, bimekizumab; FAS, full analysis set; MI, multiple imputation; NRI, non-responder­ imputation; Q4W, every 4 weeks. mean ASDAS was >3.5 (3.8–4.2) in all treatment groups and by 37.9% and 47.5% of patients receiving bimekizumab 160 and mean time since diagnosis was 7.9 (6.6–8.8) years. 320 mg, respectively (table 3). Rapid reductions from baseline in hs-­CRP were observed with bimekizumab treatment compared with placebo (table 3). In Efficacy the 31 patients with MRI follow-­up, reductions from baseline Double-blind period (to week 12) in Spondyloarthritis Research Consortium of Canada Sacro- The primary efficacy end point was reached with significantly iliac (SPARCC SI) joint score were observed at week 12 in all more patients achieving ASAS40 at week 12 in all bimekizumab-­ bimekizumab treatment groups, and in spine Berlin score were treated groups (16 mg: 29.5%, OR vs placebo 2.6 (95% CI: observed in the three highest bimekizumab dose groups at week 1.0 to 6.5); 64 mg: 42.6%, OR 4.5 (95% CI: 1.8 to 10.9); 160 12 (online supplementary figure 4). mg: 46.7%, OR 5.5 (95% CI: 2.3 to 13.5); 320 mg: 45.9%, Improvements in patient global assessment of disease activity OR 5.3 (95% CI: 2.2 to 12.9)) compared with placebo (13.3%; (PGADA), spinal pain, morning stiffness and BASFI were all comparisons p<0.05) (figure 1A, table 2 and online supple- observed compared with baseline scores, and greater reductions mentary table 1). A statistically significant dose response was from baseline in BASMI were seen in all bimekizumab-treated­ observed for ASAS40 at week 12 (p<0.001), with a dose-­ patients compared with placebo at week 12 (table 3). From post response effect observed across bimekizumab 16–160 mg doses hoc analyses, reductions from baseline to week 12 in fatigue and similar ASAS40 response between the 160 and 320 mg were observed in all patients, with the greatest reductions in the groups. Rapid responses with bimekizumab were observed, with bimekizumab 160 and 320 mg groups (table 3). Change from 8.2%–19.7% and 16.4%–50.8% of patients achieving ASAS40 baseline to week 12 in MOS Sleep Disturbance and Sleep Prob- at week 1 and at week 4 vs 0% and 8.3% receiving placebo, lems Index II were greater with bimekizumab in all dose groups respectively (figure 1B and online supplementary figure 3). versus placebo (table 4). Improvements in patient QoL were The primary end point was supported by all secondary effi- observed in all bimekizumab dose groups compared with placebo cacy end points (online table 2). At week 12, a higher proportion (table 4). Consistent with this, patient perception of physical of patients achieved ASAS20 and ASAS5/6 in the bimekizumab-­ health improved, with increases in SF-36 physical component treated groups compared with placebo. In addition, reductions summary (PCS) scores observed in all bimekizumab dose groups from baseline in BASDAI, BASFI and ASDAS were greater in the and higher scores with bimekizumab compared with placebo bimekizumab-treated­ groups compared with placebo (table 2 and (table 4). SF-36 mental component summary (MCS) scores were figure 1C). Other efficacy end points also indicated improve- similar to baseline, with all patients within the normal range. ments from baseline at week 12. ASDAS-­MI was achieved by 18.0%–34.4% of bimekizumab-treated­ patients vs 0% receiving Dose-blind period (weeks 12–48) placebo; ASDAS-­ID was achieved by 3.3%–10.0% vs 0%, The ASAS40 response rates with bimekizumab treatment ASDAS-LDA­ was achieved by 14.8%–32.8% vs 13.3%, while continued to increase after the first 12 weeks and were main- ASAS-­PR was achieved by 8.2%–23.0% vs 3.3%, respectively. tained to week 48 in patients continuing on the same bimeki- Reductions from baseline in the MASES index were observed zumab dose (160 mg: 58.6%, 320 mg: 62.3%; table 3 and figure in patients with baseline enthesitis across bimekizumab dose 1B). Of patients initially randomised to placebo, ≥50% achieved groups (table 3). In post hoc analyses, BASDAI 50 was achieved ASAS40 response at week 48 following re-­randomisation to van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 599 Spondyloarthritis

Table 3 Other efficacy end points up to 48 weeks of bimekizumab treatment (DBS, NRI and MI)

Placebo → BKZ 16 mg → BKZ 64 mg →

BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg n (%) (n=24) (n=36) (n=31) (n=27) (n=34) (n=25) (n=58) (n=61)

ASAS40* Baseline – – – – – – – – Week 12 5 (20.8) 3 (8.3) 12 (38.7) 6 (22.2) 16 (47.1) 10 (40.0) 28 (48.3) 28 (45.9) Week 48 13 (54.2) 18 (50.0) 11 (35.5) 11 (40.7) 19 (55.9) 16 (64.0) 34 (58.6) 38 (62.3) ASAS20* Baseline – – – – – – – – Week 12 9 (37.5) 8 (22.2) 14 (45.2) 11 (40.7) 21 (61.8) 17 (68.0) 35 (60.3) 44 (72.1) Week 48 17 (70.8) 22 (61.1) 17 (54.8) 14 (51.9) 25 (73.5) 20 (80.0) 45 (77.6) 46 (75.4) ASAS5/6* Baseline – – – – – – – – Week 12 2 (8.3) 2 (5.6) 10 (32.3) 8 (29.6) 19 (55.9) 11 (44.0) 32 (55.2) 33 (54.1) Week 48 15 (62.5) 16 (44.4) 13 (41.9) 13 (48.1) 21 (61.8) 20 (80.0) 37 (63.8) 40 (65.6) ASAS-PR*­ Baseline 0 0 0 0 0 0 1.7 0 Week 12 1 (4.2) 1 (2.8) 3 (9.7) 2 (7.4) 5 (14.7) 4 (16.0) 10 (17.2) 14 (23.0) Week 48 8 (33.3) 8 (22.2) 4 (12.9) 8 (29.6) 7 (20.6) 7 (28.0) 17 (29.3) 21 (34.4) ASDAS-MI†­ Baseline – – – – – – – – Week 12 0 0 7 (22.6) 4 (14.8) 13 (38.2) 8 (32.0) 15 (25.9) 19 (31.1) Week 48 8 (33.3) 11 (30.6) 11 (35.5) 9 (33.3) 20 (58.8) 10 (40.0) 24 (41.4) 34 (55.7) ASDAS-ID†­ Baseline 0 0 0 0 0 0 0 0 Week 12 0 0 1 (3.2) 1 (3.7) 3 (8.8) 3 (12.0) 6 (10.3) 6 (9.8) Week 48 4 (16.7) 5 (13.9) 3 (9.7) 6 (22.2) 6 (17.6) 4 (16.0) 15 (25.9) 15 (24.6) ASDAS-LDA†­ Baseline 0 0 0 0 0 0 1.7 0 Week 12 3 (12.5) 5 (13.9) 6 (19.4) 3 (11.1) 10 (29.4) 9 (36.0) 16 (27.6) 20 (32.8) Week 48 9 (37.5) 16 (44.4) 10 (32.3) 7 (25.9) 13 (38.2) 8 (32.0) 17 (29.3) 23 (37.7) Mean (SD) ASDAS† Baseline 3.71 (0.83) 3.88 (0.93) 3.96 (0.72) 3.84 (0.65) 4.12 (0.80) 4.18 (0.74) 3.86 (0.76) 3.94 (0.74) Change from baseline Week 12 −0.31 (0.73) −0.37 (0.62) −1.11 (1.07) −0.85 (1.01) −1.72 (0.99) −1.64 (1.21) −1.40 (0.94) −1.52 (0.94) Week 48 −1.70 (0.89) −1.80 (0.89) −1.63 (1.05) −1.61 (0.93) −2.01 (1.12) −2.04 (0.95) −1.78 (0.95) −2.03 (0.92) BASDAI 50* Baseline – – – – – – – – Week 12 3 (12.5) 4 (11.1) 9 (29.0) 5 (18.5) 14 (41.2) 12 (48.0) 22 (37.9) 29 (47.5) Week 48 11 (45.8) 20 (55.6) 12 (38.7) 10 (37.0) 20 (58.8) 14 (56.0) 30 (51.7) 35 (57.4) Geometric mean (median) Baseline 6.2 (6.1) 9.3 (10.9) 11.4 (10.7) 8.1 (9.9) 14.2 (16.7) 14.7 (17.6) 12.0 (16.0) 10.4 (11.3) hs-CRP†­ Week 12 9.1 (9.4) 8.0 (9.4) 5.6 (5.5) 3.8 (4.3) 2.9 (3.0) 5.3 (6.1) 3.7 (4.3) 3.8 (3.5) Week 48 2.9 (2.2) 2.9 (3.0) 4.0 (5.0) 3.0 (3.8) 3.4 (3.5) 4.1 (3.9) 3.9 (4.7) 3.0 (3.6) BASFI† Baseline 5.8 (1.8) 5.5 (2.2) 5.8 (1.7) 5.9 (1.9) 5.6 (1.8) 6.2 (1.8) 5.5 (2.2) 5.9 (2.0) Change from baseline Week 12 −1.0 (2.1) −0.3 (1.7) −1.7 (2.0) −1.1 (2.5) −1.6 (2.5) −2.1 (2.3) −1.7 (1.8) −2.2 (2.0) Week 48 −2.9 (2.2) −2.4 (2.2) −2.3 (1.5) −2.5 (2.0) −2.8 (2.4) −2.9 (2.4) −2.5 (2.0) −2.9 (2.2) Spinal pain† Baseline 6.9 (1.4) 7.0 (1.9) 6.8 (2.1) 7.7 (1.4) 7.4 (1.5) 7.4 (1.8) 6.6 (2.0) 7.3 (1.5) Change from baseline Week 12 −1.5 (1.6) −0.7 (1.7) −2.2 (2.4) −2.2 (2.4) −3.3 (2.2) −3.2 (2.7) −2.6 (2.2) −3.6 (2.4) Week 48 −3.7 (2.0) −3.7 (2.6) −3.0 (2.9) −3.8 (2.7) −4.2 (2.4) −4.1 (2.2) −3.8 (2.4) −4.7 (2.1) Morning stiffness† Baseline 6.9 (1.7) 6.7 (2.0) 6.7 (2.0) 6.4 (1.9) 6.5 (2.2) 7.2 (1.7) 6.5 (1.8) 6.6 (2.1) Change from baseline Week 12 −1.5 (1.7) −1.1 (1.5) −2.6 (3.0) −1.7 (2.8) −2.9 (2.9) −3.5 (2.5) −2.8 (2.0) −3.4 (2.7) Week 48 −3.9 (2.2) −3.6 (2.4) −3.7 (3.0) −3.2 (2.3) −3.9 (2.8) −4.4 (2.0) −3.9 (2.2) −4.4 (2.4) BASMI† Baseline 4.3 (1.6) 4.5 (1.6) 4.7 (1.7) 4.9 (1.8) 4.5 (1.8) 4.9 (1.5) 4.6 (1.8) 4.8 (1.8) Change from baseline Week 12 −0.3 (0.6) −0.1 (0.8) −0.5 (1.0) −0.4 (0.6) −0.5 (0.8) −0.4 (0.7) −0.3 (0.7) −0.7 (0.7) Week 48 −0.9 (1.0) −0.7 (1.0) −0.7 (1.1) −0.9 (0.7) −0.6 (0.7) −0.9 (0.7) −0.5 (1.0) −1.0 (0.9) Fatigue† Baseline 6.4 (1.7) 6.8 (1.6) 7.1 (1.6) 7.1 (1.6) 6.8 (1.2) 6.7 (1.5) 6.4 (1.7) 6.4 (1.9) Change from baseline Week 12 −0.7 (2.5) −1.0 (1.7) −1.6 (2.2) −1.6 (2.3) −2.4 (2.1) −2.6 (2.4) −2.1 (2.2) −2.1 (2.5) Week 48 −2.7 (2.2) −2.8 (2.4) −2.7 (2.0) −3.2 (2.7) −3.5 (2.3) −3.1 (2.0) −3.1 (2.1) −3.3 (2.4) PGADA† Baseline 7.0 (1.5) 6.9 (1.9) 7.0 (1.7) 7.4 (1.3) 7.4 (1.5) 7.1 (1.8) 6.5 (1.8) 7.1 (1.9) Change from baseline Week 12 −1.5 (1.8) −0.7 (1.6) −1.9 (2.9) −2.0 (2.4) −3.3 (2.3) −3.2 (2.7) −2.2 (2.6) −3.2 (2.2) Week 48 −3.8 (2.3) −3.5 (2.7) −3.3 (3.1) −3.4 (2.6) −4.0 (2.4) −4.0 (2.4) −3.5 (2.3) −4.5 (2.3)

Placebo → BKZ 16 mg → BKZ 64 mg → BKZ 160 mg BKZ 320 mg MASES‡ BKZ 160 mg(n=17) BKZ 320 mg(n=23) BKZ 160 mg(n=21) BKZ 320 mg(n=18) BKZ 160 mg(n=22) BKZ 320 mg(n=11) (n=40) (n=42)

Change from baseline Baseline 4.1 (3.0) 5.0 (3.5) 3.5 (2.4) 5.9 (3.4) 4.6 (3.4) 3.9 (2.5) 3.8 (2.2) 4.6 (3.2) Week 12 −2.5 (2.1) −2.1 (3.7) −1.2 (4.1) −2.0 (3.1) −2.4 (1.9) −3.6 (2.2) −2.5 (2.3) −2.7 (2.4) Week 48 −3.8 (3.1) −4.4 (3.6) −2.5 (3.0) −4.6 (2.8) −3.1 (2,7) −3.2 (2.6) −3.3 (2.2) −3.4 (3.2) *NRI. †MI. ‡In patients with enthesitis at baseline. ASAS, Assessment of SpondyloArthritis international Society; ASAS-PR,­ ASAS partial remission; ASDAS, Ankylosing Spondylitis Disease Activity Score; ASDAS-ID,­ ASDAS inactive disease; ASDAS-­LDA, ASDAS low disease activity; ASDAS-­MI, ASDAS major improvement; BASDAI 50, Bath Ankylosing Spondylitis Disease Activity Index ≥50% improvement; BASFI, Bath Ankylosing Spondylitis Functional Index; BASMI, Bath Ankylosing Spondylitis Metrology Index; BKZ, bimekizumab; DBS, dose-­ blind set; hs-­CRP, high-­sensitivity C reactive protein; MASES, Maastricht Ankylosing Spondylitis Enthesitis Score; MI, multiple imputation; NRI, non-responder­ imputation; NRS, numerical rating scale; PGADA, patient global assessment of disease activity.

600 van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 Spondyloarthritis

Table 4 Patient-­reported outcomes (week 12 (FAS) and week 48 (DBS); MI) Placebo → BKZ 16 mg → BKZ 64 mg → BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg Mean (SD) (n=24) (n=36) (n=31) (n=27) (n=34) (n=25) (n=58) (n=61) MOS Sleep Disturbance Baseline 45.3 (8.3) 46.2 (6.8) 48.4 (8.3) 44.6 (10.0) 47.6 (9.6) 47.8 (7.5) 47.3 (8.1) 48.0 (8.9) Change from baseline Week 12 2.3 (8.4) 1.8 (6.0) 1.9 (8.4) 4.7 (8.7) 5.8 (7.9) 4.7 (5.6) 5.8 (6.2) 6.6 (7.5) Week 48 7.3 (7.8) 6.9 (7.0) 3.8 (7.9) 8.8 (8.1) 7.6 (8.9) 6.7 (7.4) 6.6 (6.3) 6.7 (7.7) MOS Sleep Problems Index II Baseline 45.5 (8.1) 45.3 (7.9) 48.2 (8.7) 42.4 (10.1) 47.6 (9.4) 47.5 (7.1) 46.9 (7.5) 47.2 (9.4) Change from baseline Week 12 2.1 (8.3) 1.8 (6.8) 2.2 (9.4) 5.9 (9.3) 5.9 (8.1) 4.9 (6.1) 5.6 (6.7) 6.8 (7.5) Week 48 7.6 (8.7) 8.0 (9.1) 4.1 (7.5) 10.0 (9.0) 8.1 (8.6) 8.0 (7.2) 6.5 (6.1) 8.0 (7.92) ASQoL Baseline 8.4 (4.7) 9.2 (4.7) 8.4 (4.4) 9.2 (4.1) 7.9 (4.2) 9.6 (4.0) 8.4 (4.3) 8.7 (4.3) Change from baseline Week 12 −1.3 (5.5) −1.3 (3.7) −2.8 (5.2) −1.9 (5.4) −3.5 (3.6) −5.0 (4.2) −3.5 (4.3) −4.6 (4.8) Week 48 −4.2 (5.6) −5.3 (5.6) −3.9 (4.6) −5.0 (4.7) −5.0 (4.1) −6.3 (4.4) −4.9 (4.1) −5.4 (4.8) SF-36 PCS Baseline 32.8 (6.9) 33.0 (8.4) 32.5 (8.7) 30.9 (6.5) 31.9 (7.8) 30.3 (5.6) 33.0 (8.2) 32.4 (7.7) Change from baseline Week 12 5.8 (6.9) 2.0 (6.2) 7.8 (7.4) 7.4 (9.3) 8.9 (8.8) 9.5 (8.4) 8.5 (7.6) 8.2 (7.2) Week 48 12.8 (9.4) 10.9 (8.1) 10.1 (7.4) 12.6 (9.2) 12.9 (10.2) 13.4 (7.8) 12.0 (9.1) 12.0 (8.5) SF-36 MCS Baseline 54.0 (9.1) 53.6 (8.7) 54.6 (9.5) 52.0 (7.8) 55.8 (7.4) 54.4 (7.2) 53.8 (8.1) 54.4 (8.5) Change from baseline Week 12 −0.4 (8.5) 0.3 (7.2) −0.1 (7.9) 3.8 (8.2) 1.7 (6.9) 4.0 (8.1) 1.0 (7.4) 3.4 (6.9) Week 48 1.3 (10.2) 1.5 (7.8) 1.8 (7.8) 6.2 (8.4) 2.2 (6.6) 4.0 (7.0) 1.72 (8.2) 3.0 (7.7) ASQoL, Ankylosing Spondylitis Quality of Life; BKZ, bimekizumab; DBS, dose-blind­ set; FAS, full analysis set; SF-36 MCS, SF-36 mental component summary; MI, multiple imputation; MOS, Medical Outcomes Study; SF-36 PCS, SF-36 physical component summary; SF-36, 36-Item­ Short Form Questionnaire. bimekizumab 160 or 320 mg (54.2% and 50.0%, respec- study (table 4). After initial improvements in SF-36 PCS scores tively). ASAS20, ASAS5/6, ASDAS-MI,­ ASDAS-ID­ and ASAS-­PR over the first 12 weeks, scores were maintained in bimekizumab-­ responses continued to increase from week 12 to week 48 treated patients; scores in SF-36 MCS remained similar to the (table 3). ASDAS levels continued to decrease after week 12 and double-­blind period (table 4). were maintained to week 48 (figure 1C). In patients with base- line enthesitis, further reductions from baseline in MASES index were observed, with similar reductions seen across all bimeki- Safety zumab dose groups (table 3). In post hoc analyses, BASDAI 50 During the double-blind­ period, treatment-­emergent AEs response rates increased to week 48, with patients re-randomised­ (TEAEs) were reported by 37.9% (92/243) of bimekizumab-­ from placebo to bimekizumab 160 or 320 mg achieving similar treated patients compared with 43.3% (26/60) of patients responses to patients receiving 160 and 320 mg throughout the receiving placebo. Overall, up to week 48, 77.6% (235/303) of study (table 3). patients reported TEAEs while receiving bimekizumab (table 5); Following rapid reductions from baseline to week 12, hs-CRP­ the majority were mild or moderate in severity. The most levels were maintained in the bimekizumab 160 and 320 mg frequently reported TEAE (≥10% of patients in any treatment groups; in re-­randomised patients, hs-­CRP levels were similar to group) in bimekizumab-treated­ patients was nasopharyngitis those who remained on the same bimekizumab dose throughout (11.2%, 34/303; table 5). There was no apparent dose relation- the study (table 3). SPARCC SI scores were low but marked ship with type or frequency of TEAEs. reductions from baseline were observed at week 48, with the During the double-blind­ period, SAEs were reported by greatest reductions observed in the group re-­randomised from 1.2% (3/243) of bimekizumab-treated­ patients and 3.3% (2/60) placebo to bimekizumab 160 and 320 mg. MRI data were of those receiving placebo. Thirteen of 303 (4.3%) patients only available for two patients receiving bimekizumab 160 mg receiving bimekizumab experienced SAEs during the 48-­week throughout the study and their baseline SPARCC SI scores were study duration (see online supplementary file). TEAEs leading to low; therefore, reductions in SPARCC SI were not observed in discontinuation were reported by 2.5% (6/243) of bimekizumab-­ this dose group. Reductions from baseline in spine Berlin score treated patients vs 1.7% (1/60) of patients receiving placebo in at week 48 were also observed in bimekizumab-treated­ patients the double-blind­ period. Overall, 6.6% (20/303) of patients (online supplementary figure 4). discontinued due to TEAEs. Improvements in PROs continued to week 48, with reduc- Oral candidiasis was reported by 3 (4.9%) patients in the tions from baseline in PGADA, spinal pain, morning stiffness, bimekizumab 320 mg group during the double-­blind period and BASFI observed in all treatment groups (table 3). Continued (table 5) and 16/303 (5.3%) of bimekizumab-treated­ patients in improvements in mobility were observed, with bimekizumab-­ total. All cases were mild to moderate, resolved with systemic or treated patients experiencing reductions from baseline in topical antifungal treatment and no events led to patients with- BASMI (table 3). At week 48, similar reductions from baseline in drawing from the study. fatigue were observed across bimekizumab dose groups (table 3). Four cases of IBD were reported (further details in online Improvements in Sleep Disturbance and Sleep Problems Index II supplementary file). Two cases were new diagnoses of Crohn’s seen at week 12 were maintained to week 48 in bimekizumab-­ disease and were mild to moderate in intensity. One case (bime- treated patients (table 4). Patients who initially received placebo kizumab 320 mg group) occurred during the dose-­blind period, and were re-­randomised to bimekizumab reported similar was identified as related to treatment and resolved, the other reductions in ASQoL during the dose-­blind period as those (bimekizumab 160 mg group) was considered unrelated to treat- receiving bimekizumab 160 and 320 mg from the start of the ment, with diagnosis confirmed after the patient had withdrawn van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 601 Spondyloarthritis

Table 5 Summary of safety results (SS) Double-­blind period Overall treatment period BKZ 16 mg BKZ 64 mg BKZ 160 mg BKZ 320 mg BKZ 160 mg BKZ 320 mg Placebo Q4W Q4W Q4W Q4W Q4W Q4W Q4W All BKZ† (n=60) (n=61) (n=58) (n=63) (n=61) (n=149) (n=150) (n=303) n* (%) (#) n* (%) (#) n* (%) (#) n* (%) (#) n* (%) (#) n (EAIR) n (EAIR) n (EAIR) Any TEAE 26 (43.3) (41) 26 (42.6) (50) 17 (29.3) (33) 20 (31.7) (33) 29 (47.5) (61) 103 (168.7) 122 (221.1) 235 (186.2) Nasopharyngitis 0 2 (3.3) (2) 1 (1.7) (1) 3 (4.8) (3) 0 13 (12.0) 19 (16.6) 34 (13.7) Pharyngitis 0 0 0 2 (3.2) (2) 1 (1.6) (1) 11 (10.0) 7 (6.1) 18 (7.1) Bronchitis 1 (1.7) (1) 0 2 (3.4) (2) 0 0 4 (3.6) 12 (10.4) 18 (7.1) Upper respiratory tract infection 1 (1.7) (1) 0 1 (1.7) (1) 0 1 (1.6) (1) 5 (4.5) 11 (9.5) 17 (6.7) Oral candidiasis 0 0 0 0 3 (4.9) (3) 8 (7.2) 8 (7.0) 16 (6.3) Serious TEAEs 2 (3.3) (2) 0 2 (3.4) (5) 1 (1.6) (2) 0 5 (4.4) 6 (5.1) 13 (5.1) Permanent withdrawal of study 1 (1.7) (1) 2 (3.3) (2) 1 (1.7) (1) 1 (1.6) (2) 2 (3.3) (2) 7 (6.2) 10 (8.7) 20 (7.9) medication due to TEAEs Drug-­related TEAEs 6 (10.0) (8) 9 (14.8) (12) 6 (10.3) (10) 7 (11.1) (8) 12 (19.7) (17) 48 (52.0) 54 (58.7) 110 (54.0) Death 0 0 0 1 (1.6) (1) 0 1 (0.9) 0 1 (0.4) TEAEs of special interest 5 (8.3) (5) 9 (14.8) (10) 6 (10.3) (9) 3 (4.8) (3) 5 (8.2) (7) Opportunistic infection 0 1 (1.6) (1) 0 0 0 0 0 1 (0.38) Candida infections 0 0 0 0 3 (4.9) (3) 10 (9.1) 9 (7.9) 19 (7.5) Neutropenia 0 0 0 0 0 1 (0.9) 0 1 (0.4) Severe hypersensitivity reactions 0 0 0 0 0 0 0 0 Suicide ideation 0 0 0 0 0 0 0 0 Psychiatric disorder 2 (3.3) (2) 1 (1.6) (1) 0 0 0 0 0 1 (0.3) Major cardiovascular events 0 0 0 1 (1.6) (1) 0 2 (1.8) 0 2 (0.8) Hepatic events 2 (3.3) (2) 6 (9.8) (6) 4 (6.9) (7) 1 (1.6) (1) 1 (1.6) (2) 13 (12.1) 12 (10.4) 33 (13.6) ALT increased 0 1 (1.6) (1) 1 (1.7) (1) 0 1 (1.6) (1) 5 (4.5) 6 (5.1) 13 (5.1) AST increased 0 0 1 (1.7) (1) 0 1 (1.6) (1) 3 (2.7) 5 (4.3) 9 (3.5) GGT increased 1 (1.7) (1) 2 (3.3) (2) 2 (3.4) (3) 0 0 6 (5.4) 4 (3.4) 13 (5.1) Hepatic enzyme increased 1 (1.7) (1) 2 (3.3) (2) 1 (1.7) (1) 0 0 0 3 (2.5) 6 (2.3) Malignancies 0 0 0 0 0 0 0 0 Inflammatory bowel disease 0 1 (1.6) (1) 0 0 0 1 (0.9) 2 (1.7) 4 (1.5) In addition, one patient received doses of 160 and 320 mg in error and was therefore included in both the 160 and 320 mg groups, but only once in the all-BKZ­ group; (#) the number of individual occurrences of the AE. *Number of patients reporting at least one TEAE. †The all-­BKZ group included 5 patients who received BKZ in the double-blind­ period but did not receive BKZ 160 or 320 mg: 2 patients in the 16 mg group and 2 patients in the 64 mg group discontinued before re-­randomisation; 1 patient in the 16 mg group did not start the dose-blind­ period. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BKZ, bimekizumab; EAIR, exposure-adjusted­ incidence rate per 100 patient-­years; GGT, gamma-­ glutamyltransferase; Q4W, every 4 weeks; SS, safety set; TEAE, treatment-emergent­ adverse event. from the study. Two cases were UC, one with a history of UC and No malignancies, severe hypersensitivity reactions or suicidal one new diagnosis; both were considered mild and unrelated to ideation were reported during the 48-­week study (table 5). treatment. The new diagnosis of UC occurred during the double-­ No clinically meaningful changes from baseline in vital sign blind period (bimekizumab 16 mg, week 4). The other case was measurements (systolic and diastolic blood pressure, and pulse an exacerbation of UC during the dose-­blind period in a patient rate), or ECG results were observed in any treatment group with a history of UC who was receiving concomitant mesalazine during the overall study. (bimekizumab 320 mg, week 24). Overall, the exposure adjusted incidence rate (EAIR) per 100 patient-years­ for Crohn’s disease and UC was 0.77 each. Discussion There were four cases of acute anterior uveitis during the This was the first phase IIb study of bimekizumab in patients study; two of these occurred in patients receiving bimekizumab, with active AS. Bimekizumab-treated­ patients had rapid and one during treatment and one after discontinuation (see online significant ASAS40 responses versus placebo at week 12 and supplementary file). Overall, 15.2% of patients in the study these were supported by all secondary efficacy end points, as reported a history of uveitis or iridocyclitis at baseline (patients well as additional QoL and patient-reported­ measures of func- with a recent history of uveitis were not excluded); the incidence tion and disease activity. All outcomes further improved from rate of uveitis in patients receiving bimekizumab was 0.53 (EAIR week 12 and were sustained to week 48. This study offers 0.77). insights into the potential efficacy of bimekizumab for the Two cases adjudicated as MACE were reported: one SAE treatment of patients with AS and provides further support for (described in the online supplementary file), and one myocar- the encouraging findings from clinical and preclinical studies dial infarction (bimekizumab 160 mg), which resolved, was of bimekizumab,15 21 31 that dual neutralisation of IL-­17A and considered unrelated to treatment and did not lead to treatment IL-17F­ may be effective in the treatment of IL-17-mediated­ discontinuation. inflammatory diseases.

602 van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 Spondyloarthritis

TNF and IL-­17A inhibitors are currently recommended for those observed with ixekizumab10 and secukinumab11 in recent the management of patients with AS7 9; the novel biological phase III studies in AS, involving similar numbers of patients. DMARD bimekizumab may provide an alternative therapeutic The 48-­week duration of this study, providing long-­term effi- approach. A significant dose response was observed in ASAS40 cacy data for bimekizumab ahead of completion of the phase at week 12 with increasing bimekizumab dose from 16 to 320 III studies and demonstrating maintenance of responses over mg; responses were generally similar for the three highest doses time, represents a strength of the study design. Study limitations and the greatest response was in the 160 mg group. The parallel include the relatively short placebo control period of 12 weeks, study of bimekizumab in PsA also demonstrated the greatest small patient numbers per treatment group due to the nature of ACR50 response in the 160 mg dose group.31 In patients with the study design, a high number of Caucasian patients and MRI moderate-­to-­severe plaque psoriasis,21 the greatest responses data being available for only a subset of patients. were observed with bimekizumab 320 mg; this differing dose In summary, bimekizumab may provide a promising ther- response may be due to differences between patient populations apeutic option for patients with AS. Bimekizumab treatment and efficacy in axial or joint disease compared with skin mani- resulted in rapid and sustained improvements across multiple festations. Here, for all measures of disease activity, response outcomes of disease activity, QoL and function. These results rates increased and were sustained from week 12 through week support phase III evaluation of bimekizumab in axSpA (BE 48 in patients continuing on bimekizumab 160 and 320 mg. MOBILE 1, NCT03928704; BE MOBILE 2, NCT03928743) Notably, patients re-randomised­ at week 12 responded rapidly to determine the clinical benefits that may be associated with and achieved similar response rates from week 24 to week 48 neutralising IL-17F­ in addition to IL-­17A. to patients receiving bimekizumab 160 or 320 mg continu- ously throughout the study. Although there was a rapid onset Author affiliations 1 of response for the primary and secondary efficacy end points, Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands more stringent outcomes, such as ASDAS-­ID, required observa- 2 University California San Francisco, San Francisco, California, USA tion over a longer time period; by week 48, one-­quarter of this 3Oregon Health & Science University, Portland, Oregon, USA patient population with highly active disease achieved inactive 4Rheumazentrum Ruhrgebiet Herne, Ruhr-­University Bochum, Herne, Germany disease in both the bimekizumab 160 and 320 mg groups. 5Department of Rheumatology, Infectious Diseases and Rheumatology, Charité— Levels of inflammation, as measured by hs-­CRP, decreased Universitätsmedizin Berlin and Epidemiology Unit, German Rheumatism Research Centre, Berlin, Germany rapidly with bimekizumab treatment, with low levels of inflam- 6Altoona Center for Clinical Research, Duncansville, Pennsylvania, USA mation maintained to week 48 in all bimekizumab-treated­ 7UCB Pharma, Raleigh, North Carolina, USA patients. SPARCC SI and spine Berlin scores generally improved 8Department of Clinical Immunology and Rheumatology, UCB Pharma, Slough, UK 9UCB Pharma, Monheim am Rhein, Germany from week 12 through week 48 across bimekizumab groups; 10 however, results should be interpreted with caution due to the Department of Rheumatology, Université de Paris, Hôpital Cochin, Assistance Publique-­Hôpitaux de Paris, Paris, France small number of patients per treatment group (range, 2–12) who 11INSERM (U1153): Epidémiologie Clinique et Biostatistiques, PRES Sorbonne Paris-­ had MRI assessments. Cité, Paris, France AS has a substantial impact on patients’ QoL in many key Acknowledgements The authors would like to acknowledge Alexandra areas such as completion of daily activities, fatigue and sleep, Webster, MSc, of iMed Comms, an Ashfield Company, part of UDG Healthcare, with pain acknowledged as one of the most burdensome symp- for medical writing support that was funded by UCB Pharma in accordance with toms.32 Bimekizumab rapidly improved patient-­reported spinal Good Publications Practice (GPP3) guidelines (http://www.​ismpp.​org/​gpp3). The authors would also like to thank Simone E Auteri, MSc PhD EMS, of UCB Pharma pain and improvements continued over the duration of the for publication coordination and critical review, Natasha de Peyrecave, DPhil, of study. Changes from baseline to week 12 in sleep disturbance UCB Pharma for critical review, and Rajan Bajracharya, MBBS MPH MFPM, of UCB and sleep problems suggested that bimekizumab treatment Pharma for review of the safety reporting in the manuscript. improved sleep outcomes for patients, which were maintained Contributors All authors were involved in designing the study, analysing and to week 48. In addition, reductions in fatigue from baseline to interpreting the study data, and critically reviewing the manuscript. All authors week 12 and week 48 indicated sustained improvements for approved the final version for submission. patients in another key symptom of their disease. Changes from Funding The study (NCT02963506) was funded by UCB Pharma. baseline in SF-36 PCS domain indicated positive improvements Competing interests DvdH reports personal fees from AbbVie, Amgen, Astellas, in the physical aspects of patients’ disease. AstraZeneca, BMS, Boehringer Ingelheim, Celgene, Cyxone, Daiichi, Eisai, Eli Lilly, Overall, bimekizumab was generally well tolerated across Galapagos, Gilead, GlaxoSmithKline, Janssen, Merck, Novartis, Pfizer, Regeneron, the treatment groups. The safety profile in patients with active Roche, Sanofi, Takeda and UCB Pharma, and is Director of Imaging Rheumatology BV. LSG reports grants and personal fees from AbbVie, Amgen, Novartis, Pfizer and UCB AS was as expected given previous studies of bimekizumab in Pharma, and personal fees from Galapagos, Janssen and Eli Lilly. AD reports personal 15 21 patients with psoriasis or PsA. Consistent with the mecha- fees from Amgen, Boehringer Ingelheim, Bristol-Myers­ Squibb, Galapagos and Janssen, nism of action of therapies targeting the IL-17 pathway, 16 cases and grants and personal fees from AbbVie, Eli Lilly, GlaxoSmithKline, Novartis, Pfizer (5%) of oral candidiasis were reported in bimekizumab-treated­ and UCB. DP reports personal fees from UCB Pharma, BMS, Roche and Celgene, and 11 33 grants and personal fees from AbbVie, Eli Lilly, MSD, Novartis and Pfizer. XB reports patients across the 48 weeks of treatment. However, all personal fees from AbbVie, Bristol-Myers­ Squibb, Celgene, Chugai, Janssen, MSD, cases were mild or moderate and did not lead to discontinua- Novartis, Pfizer and UCB Pharma, and grant/research support fromA bbvie, MSD and tion. Four cases of IBD were reported, including one in a patient Novartis. AK reports research support from Altoona Centernical Research, PC, during with a history of UC; incidence of IBD is a known side effect the conduct of the study; Advisory Committee or Review Panel for AbbVie, Janssen, of IL-­17A inhibition but is also a recognised manifestation of UCB Pharma and Boehringer Ingelheim; speaking and teaching for Celgene, Horizon, Merck and Novartis; Advisory Committee or Review Panel, speaking and training for axSpA. The incidence of IBD reported in this study of bimeki- Genzyme; Advisory Committee or Review Panel, speaking/training; consultant, and 34 zumab is comparable to that seen with TNF inhibitors, secuk- stocks for Pfizer and Sanofi;A dvisory Committee or Review Panel, speaking/training inumab35 and ixekizumab in AS.36 To date, IBD has not been and consultant for Regeneron; consultant for SUN Pharma Advanced Research; Speaker reported with bimekizumab in studies involving patients with and Steering Committee for Flexion; Stocks from Amgen, Gilead and GSK and Speaker 21 31 for AbbVie. MO, DB and NG are employees of UCB Pharma and own stocks and stock psoriasis or PsA. options. JC is an employee of UCB Pharma. MKF was an employee of UCB Pharma The ASAS40 responses and secondary outcomes observed with at the time the study was conducted (current employee of Kezar Life Sciences). MD bimekizumab in this phase IIb study compare favourably with reports grants and personal fees from UCB, Eli Lilly, Novartis, Pfizer,A bbVie and Merck. van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 603 Spondyloarthritis

Patient and public involvement Patients and/or the public were involved in 13 Hymowitz SG, Filvaroff EH, Yin JP, et al. IL-­17s adopt a cystine knot fold: structure the design, conduct, reporting or dissemination plans of this research. Refer to the and activity of a novel cytokine, IL-­17F, and implications for receptor binding. EMBO J ’Methods’ section for further details. 2001;20:5332–41. 14 Yang XO, Chang SH, Park H, et al. Regulation of inflammatory responses by IL-­17F. J Patient consent for publication Not required. Exp Med 2008;205:1063–75. Ethics approval The study was conducted in accordance with the principles of the 15 Glatt S, Baeten D, Baker T, et al. Dual IL-­17A and IL-­17F neutralisation by Declaration of Helsinki and the International Conference on Harmonisation Guidance bimekizumab in psoriatic arthritis: evidence from preclinical experiments and a for Good Clinical Practice. Independent institutional review board approvals were randomised placebo-­controlled clinical trial that IL-­17F contributes to human chronic obtained, and all patients provided written informed consent in accordance with tissue inflammation. Ann Rheum Dis 2018;77:523–32. local requirements. 16 ammari M, Presumey J, Ponsolles C, et al. Delivery of miR-­146a to Ly6C(high) monocytes inhibits pathogenic bone erosion in inflammatory arthritis.Theranostics Provenance and peer review Not commissioned; externally peer reviewed. 2018;8:5972–85. Data availability statement Underlying data from this manuscript may be 17 shah M, Maroof A, Al-­Hosni RN, et al. Bimekizumab blocks T cell-­mediated osteogenic requested by qualified researchers six months after product approval in the US differentiation of periosteal stem cells: coupling pathological bone formation to IL-­ and/or Europe, or global development is discontinued, and 18 months after trial 17A and IL-­17F signaling [abstract]. Arthritis Rheumatol 2017;69 (suppl 10):1936. completion. Investigators may request access to anonymized IPD and redacted 18 Gracey E, Yao Y, Green B, et al. Sexual dimorphism in the Th17 signature of ankylosing study documents which may include: raw datasets, analysis-ready­ datasets, study spondylitis. Arthritis Rheumatol 2016;68:679–89. protocol, blank case report form, annotated case report form, statistical analysis 19 Poddubnyy D, Vicari A, Haibel H, et al. Interleukin-17 serum levels are associated plan, dataset specifications, and clinical study report. Prior to use of the data, with markers of systemic inflammation in patients with active ankylosing spondylitis proposals need to be approved by an independent review panel at www.Vivli.​ ​ [abstract]. ACR/ARHP Annual Meeting 2013;2493. org and a signed data sharing agreement will need to be executed. All documents 20 Glatt S, Helmer E, Haier B, et al. First-­In-­Human randomized study of bimekizumab, a are available in English only, for a pre-specified­ time, typically 12 months, on a humanized monoclonal antibody and selective dual inhibitor of IL-­17A and IL-­17F, in password protected portal. mild psoriasis. Br J Clin Pharmacol 2017;83:991–1001. 21 Papp KA, Merola JF, Gottlieb AB, et al. Dual neutralization of both interleukin 17A and Open access This is an open access article distributed in accordance with the interleukin 17F with bimekizumab in patients with psoriasis: results from BE ABLE 1, Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which a 12-­week randomized, double-­blinded, placebo-­controlled phase 2b trial. J Am Acad permits others to distribute, remix, adapt, build upon this work non-­commercially, Dermatol 2018;79:277–86. e10. and license their derivative works on different terms, provided the original work is 22 Kaeley GS, MacCarter DK, Pangan AL, et al. Clinical responses and synovial vascularity properly cited, appropriate credit is given, any changes made indicated, and the use in obese rheumatoid arthritis patients treated with adalimumab and methotrexate. J is non-­commercial. See: http://creativecommons.​ ​org/licenses/​ ​by-​nc/​4.0/.​ Rheumatol 2018;45:1628–35. 23 van der Linden S, Valkenburg HA, Cats A. Evaluation of diagnostic criteria for ORCID iDs ankylosing spondylitis. A proposal for modification of the New York criteria. Arthritis Désirée van der Heijde http://orcid.​ ​org/0000-​ ​0002-​5781-​158X Rheum 1984;27:361–8. Xenofon Baraliakos http://orcid.​ ​org/0000-​ ​0002-​9475-​9362 24 sieper J, Rudwaleit M, Baraliakos X, et al. The Assessment of SpondyloArthritis Denis Poddubnyy http://orcid.​ ​org/0000-​ ​0002-​4537-​6015 international Society (ASAS) handbook: a guide to assess spondyloarthritis. Ann Rheum Dis 2009;68(Suppl 2):ii1–44. References 25 Machado P, Landewé R, Lie E, et al. Ankylosing spondylitis disease activity score 1 sieper J, Poddubnyy D. Axial spondyloarthritis. Lancet 2017;390:73–84. (ASDAS): defining cut-­off values for disease activity states and improvement scores. 2 de Winter JJ, van Mens LJ, van der Heijde D, et al. Prevalence of peripheral and Ann Rheum Dis 2011;70:47–53. extra-­articular disease in ankylosing spondylitis versus non-­radiographic axial 26 Jenkinson TR, Mallorie PA, Whitelock HC, et al. Defining spinal mobility in ankylosing spondyloarthritis: a meta-­analysis. Arthritis Res Ther 2016;18:196. spondylitis (AS). The Bath AS Metrology Index. J Rheumatol 1994;21:1694–8. 3 Halvorsen S, Vøllestad NK, Fongen C, et al. Physical fitness in patients with ankylosing 27 nam JM. A simple approximation for calculating sample sizes for detecting linear spondylitis: comparison with population controls. Phys Ther 2012;92:298–309. trend in proportions. Biometrics 1987;43:701–5. 4 landewé R, Dougados M, Mielants H, et al. Physical function in ankylosing spondylitis 28 landewé R, Braun J, Deodhar A, et al. Efficacy of certolizumab pegol on signs and is independently determined by both disease activity and radiographic damage of the symptoms of axial spondyloarthritis including ankylosing spondylitis: 24-week­ results spine. Ann Rheum Dis 2009;68:863–7. of a double-­blind randomised placebo-­controlled phase 3 study. Ann Rheum Dis 2014;73:39–47. 5 li Y, Zhang S, Zhu J, et al. Sleep disturbances are associated with increased pain, 29 van der Heijde D, Schiff MH, Sieper J, et al. Adalimumab effectiveness for the disease activity, depression, and anxiety in ankylosing spondylitis: a case-control­ treatment of ankylosing spondylitis is maintained for up to 2 years: long-term­ results study. Arthritis Res Ther 2012;14:R215. from the ATLAS trial. Ann Rheum Dis 2009;68:922–9. 6 Machado P, Landewé R, Braun J, et al. A stratified model for health outcomes in 30 Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective ankylosing spondylitis. Ann Rheum Dis 2011;70:1758–64. studies of disease. J Natl Cancer Inst 1959;22:719–48. 7 van der Heijde D, Ramiro S, Landewé R, et al. 2016 update of the ASAS-­EULAR 31 Ritchlin CT, Kavanaugh A, Merola JF, et al. Dual neutralization of IL-­17A and IL-­17F management recommendations for axial spondyloarthritis. Ann Rheum Dis with bimekizumab in patients with active PsA: results from a 48-­week phase 2b, 2017;76:978–91. randomized, double blind, placebo-­controlled, dose-­ranging study [abstract]. Arthritis 8 sepriano A, Regel A, van der Heijde D, et al. Efficacy and safety of biological and Rheumatol 2018;70 (Suppl 10). targeted-­synthetic DMARDs: a systematic literature review Informing the 2016 update 32 Garrido-­Cumbrera M, Hillmann O, Mahapatra R, et al. Improving the management of of the ASAS/EULAR recommendations for the management of axial spondyloarthritis. psoriatic arthritis and axial spondyloarthritis: roundtable discussions with healthcare RMD Open 2017;3:e000396. professionals and patients. Rheumatol Ther 2017;4:219–31. 9 Ward MM, Deodhar A, Gensler LS, et al. 2019 update of the American College of 33 van der Heijde D, Gladman DD, Kishimoto M, et al. Efficacy and safety of ixekizumab Rheumatology/Spondylitis association of America/Spondyloarthritis research and in patients with active psoriatic arthritis: 52-­week results from a phase III study treatment network recommendations for the treatment of ankylosing spondylitis and (SPIRIT-P1).­ J Rheumatol 2018;45:367–77. Nonradiographic axial spondyloarthritis. Arthritis Care Res 2019;71:1285–99. 34 Braun J, Baraliakos X, Listing J, et al. Differences in the incidence of flares or new 10 van der Heijde D, Cheng-­Chung Wei J, Dougados M, et al. Ixekizumab, an interleukin-­ onset of inflammatory bowel diseases in patients with ankylosing spondylitis 17A antagonist in the treatment of ankylosing spondylitis or radiographic axial exposed to therapy with anti-­tumor necrosis factor alpha agents. Arthritis Rheum spondyloarthritis in patients previously untreated with biological disease-­modifying 2007;57:639–47. anti-­rheumatic drugs (COAST-­V): 16 week results of a phase 3 randomised, double-­ 35 schreiber S, Colombel J-­F, Feagan BG, et al. Incidence rates of inflammatory bowel blind, active-­controlled and placebo-­controlled trial. Lancet 2018;392:2441–51. disease in patients with psoriasis, psoriatic arthritis and ankylosing spondylitis treated 11 Baeten D, Sieper J, Braun J, et al. Secukinumab, an interleukin-­17A inhibitor, in with secukinumab: a retrospective analysis of pooled data from 21 clinical trials. Ann ankylosing spondylitis. N Engl J Med 2015;373:2534–48. Rheum Dis 2019;78:473–9. 12 Rezaiemanesh A, Abdolmaleki M, Abdolmohammadi K, et al. Immune cells 36 Marzo-­Ortega H, Mysler E, Tetsuya T, et al. Long-­term safety of ixekizumab in patients involved in the pathogenesis of ankylosing spondylitis. Biomed Pharmacother with radiographic axial spondyloarthritis/ankylosing spondylitis: an integrated analysis 2018;100:198–204. of COAST-­V and COAST-­W [abstract]. Ann Rheum Dis 2019;78:885–6.

604 van der Heijde D, et al. Ann Rheum Dis 2020;79:595–604. doi:10.1136/annrheumdis-2020-216980 Spondyloarthritis

Clinical science Reactive arthritis and other musculoskeletal symptoms associated with acquisition of diarrhoeagenic Escherichia coli (DEC) Riitta Tuompo ‍ ‍ ,1 Tinja Lääveri,2,3 Timo Hannu,4 Sari H Pakkanen,3 Juha Kirveskari,5,6 Marjatta Leirisalo-Repo,­ 1 Anu Kantele2,3,7

Handling editor Josef S Abstract Key messages Smolen Objectives Using a prospective research design, 1 we evaluated the association between acquisition of Inflammation Center, What is already known about this subject? Department of Rheumatology, diarrhoeagenic Escherichia coli (DEC) and development ►► Enteropathogenic bacteria triggering reactive University of Helsinki and of reactive arthritis (ReA) and other reactive arthritis (ReA) are often contracted while Helsinki University Hospital, musculoskeletal (MSK) symptoms among international visiting (sub)tropical countries in particular. Helsinki, Finland travellers. 2Inflammation Center, Although diarrhoeagenic Escherichia coli (DEC) Methods A total of 526 study participants were asked Department of Infectious are the most common pathogens causing to provide pretravel and post-­travel stool samples diseases, University of Helsinki travellers’ diarrhoea (TD), studies are lacking and Helsinki University Hospital, and fill in questionnaires (pretravel, post-­travel and of whether DEC acquisition is associated with Helsinki, Finland 3-­week follow-­up). A multiplex quantitative PCR 3Human Microbiome Research musculoskeletal (MSK) symptoms. assay was deployed to detect five ED C comprising Program, Faculty of Medicine, ► Data are needed on how antibiotic treatment of enteroaggregative E. coli, enteropathogenic E. coli, ► University of Helsinki, Helsinki, TD influences development of MSK symptoms. Finland enterotoxigenic E. coli, enterohaemorrhagic E. coli 4Department of Public Health, and enteroinvasive E. coli and Salmonella, Shigella, What does this study add? University of Helsinki, Helsinki, Campylobacter, Yersinia, and Vibrio cholerae. Finland ►► This is the first prospective study employing 5 Multivariate analysis was employed to identify factors Department of Bacteriology, multivariate analysis to: (1) investigate whether predisposing to MSK symptoms. New post-­travel MSK Helsinki University Hospital reactive MSK symptoms are associated with Laboratory, Helsinki, Finland symptoms reported by participants with DEC were 6 acquisition of DEC and (2) determine incidence Mobidiag Ltd, Espoo, Finland assessed by phone interviews and, if needed, clinically 7Institute of Clinical Medicine, of MSK symptoms and ReA among travellers confirmed. University of Helsinki, Helsinki, with DEC. Results From among the total of 224 volunteers who Finland ► Acquisition of DEC predisposes to development returned all questionnaires and stool specimens, 38 ► of MSK symptoms. Correspondence to (17.0%) reported MSK symptoms. Multivariate analysis ► Development of MSK symptoms does not Dr Riitta Tuompo, Inflammation revealed that acquisition of DEC was associated with ► depend on the severity of TD. Center, Department of MSK symptoms (OR 3.9; 95% CI 1.2 to 13.3). Of the 151 Rheumatology, Helsinki ► Antibiotic use does not prevent development of with only-­DEC, four (2.6%) had ReA, two (1.3%) reactive ► University Hospital, Helsinki MSK symptoms. 00290, Finland; tendinitis and three (2.0%) reactive arthralgia. ReA was riitta.​ ​tuompo@hus.​ ​fi mostly mild, and all patients with ReA were negative for How might this impact on clinical practice or human leucocyte antigen B27. Antibiotic treatment of future developments? Received 29 November 2019 travellers’ diarrhoea did not prevent development of MSK Revised 20 February 2020 ► Acquisition of DEC predisposes to development symptoms. ► Accepted 20 February 2020 of MSK symptoms. Antibiotic use does not Conclusion A total of 17% of volunteers reported Published Online First prevent them. Therefore, concern about 16 March 2020 post-­travel MSK symptoms. DEC acquisition was potential MSK symptoms does not justify associated with an increased risk of developing them, antibiotic treatment for TD. yet the ReA incidence remained low and the clinical ► As no vaccines are available, it appears that, picture mild. Antibiotic treatment did not protect against ► today, the only reasonable approaches to development of MSK symptoms. prevent MSK symptoms are those decreasing the rate of exposure to intestinal pathogens.

Introduction Reactive arthritis (ReA) is a sterile arthritis typically acquisition and musculoskeletal (MSK) symptoms © Author(s) (or their appearing subsequent to a gastrointestinal or geni- has not been explored to date in prospective studies employer(s)) 2020. Re-­use permitted under CC BY-­NC. No tourinary infection. Salmonella, Yersinia, Shigella among travellers recruited before departure. commercial re-­use. See rights and Chlamydia trachomatis are considered the clas- Up to 78% of travellers with TD contract DEC, and permissions. Published sical causative agents, but also Campylobacter and but asymptomatic carriage is not rare either.8–11 by BMJ. Escherichia coli can trigger ReA.1–6 These bacteria There are five pathotypes of DEC: enteropatho- To cite: Tuompo R, are usually contracted while visiting (sub)tropical genic E. coli (EPEC), enteroaggregative E. coli Lääveri T, Hannu T, countries. Although diarrhoeagenic E. coli (DEC) (EAEC), enterotoxigenic E. coli (ETEC), entero- et al. Ann Rheum Dis are the most common pathogens causing travel- invasive E. coli (EIEC) and enterohaemorrhagic E. 2020;79:605–611. lers’ diarrhoea (TD),7–11 association between DEC coli (EHEC).12 EIEC is closely related to Shigella,

Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 605 Spondyloarthritis indistinguishable from it by quantitative PCR (qPCR). While it is Methods well known that Shigella triggers ReA,13 the various DEC require Study design and patient involvement more research. This research was conducted parallel to a larger prospective In ankylosing spondylitis (AS), both microscopic intestinal investigation of the aetiology and clinical features of TD.11 20 21 inflammation14 and increased IgA levels against E. coli15 16 have The current study is the first to present data on its rheumato- been described. AS and ReA both belong to spondylarthritides logical findings. The subjects were recruited at their pretravel and may have similar aetiological and pathogenetic mecha- visits to the Travel Clinic of Aava Medical Centre, Helsinki, nisms.17 In animal experiments, immunisation with E. coli O:14 between March 2009 and February 2010. The volunteers were and its lipopolysaccharide (LPS) has induced chronic arthritis, asked to provide pretravel and post-­travel stool samples and to with LPS and interleukin 1 detected in synovial and pannus complete three questionnaires (pretravel, post-­travel and 3-­week cells in arthritic joints.18 Furthermore, a higher frequency of E. follow-­up). The questionnaires collected information on travel coli-­reactive Th1 cells have been found in synovial fluid among characteristics, symptoms (eg, TD and MSK) and medications patients with AS than those with rheumatoid arthritis, suggesting (antibiotics, antidiarrhoeals and so on). We have published a involvement of mucosal antigens in the pathogenesis of AS.19 detailed description of the questionnaires and data on behaviour Routine diagnostics of TD previously only covered Salmo- and symptoms.26 nella, Shigella, Campylobacter and Yersinia; EHEC 0157 and Stool samples were analysed with a qPCR for nine bacterial Clostridium difficile were analysed on request. Consequently, pathogens: Salmonella, Yersinia, Campylobacter, Vibrio chol- DEC, the most common TD pathogens, often remained unrec- erae, Shigella/EIEC, EHEC, ETEC, EAEC and EPEC.20 The ognised.20 Recently, however, multiplex qPCR methods covering findings were categorised into four groups: (1) no pathogens, a large variety of TD pathogens have revolutionised the diagnos- (2) only-DEC­ (no other pathogens but EPEC, EAEC, ETEC tics.9 11 20 21 EPEC, EAEC and ETEC are the three most common or EHEC), (3) only non-­DEC (no DEC but Shigella/EIEC, TD pathogens worldwide, while Campylobacter ranks as major Salmonella, Campylobacter, Yersinia or V. cholerae), and (4) pathogen only among travellers to Southeast Asia; Salmonella DEC+non-­DEC. To ensure that the only-DEC­ group would and Shigella are less frequent.7 11 21 22 include no other pathogens, we categorised Shigella/EIEC (not Some studies have described ReA triggered by E. coli following distinguishable by qPCR) as non-­DEC, Shigella a known risk urinary tract3 23 24 and gastrointestinal infections.4–6 25 However, factor of ReA.13 We report separately on the MSK symptoms of the studies are few, their participants were not recruited prospec- those with Shigella/EIEC. tively before travel, controls with no pathogen findings were not The inclusion criteria comprised providing all three question- included and the various DEC (particularly EAEC) were mostly naires and both pretravel and post-­travel stool samples. For the not covered. first part, we included all such participants, and for the second The first part of our bipartite traveller study examined the part, those with only DEC in post-­travel specimen and MSK rate of MSK symptoms, seeking to identify travel-related­ factors symptoms in 3-­week follow-up­ with questionnaires (figure 1). predisposing to/preventing them in a prospective study design If no other explanation for MSK symptoms was evident (eg, encompassing all the various DEC. The second part investigated osteoarthritis and trauma), we contacted the volunteers by tele- the incidence and clinical picture of ReA. phone for a comprehensive interview and, if needed, invited

Figure 1 Study flow chart. *Subjects who returned questionnaire and stool sample before and after journey plus follow-up­ questionnaire. **Osteoarthritis, similar symptoms before journey. DEC, diarrhoeagenic Escherichia coli; MSK, musculoskeletal; ReA, reactive arthritis; ReTe, reactive enthesitis and tendinitis; ReAlg, reactive arthralgia

606 Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 Spondyloarthritis them to visit a rheumatologist for clinical assessment. The phone Table 1 Pathogen findings in all 224 post-­travel stool samples* interview covered their history of MSK diseases and a detailed and for 38 participants reporting MSK symptoms in the follow-up­ account of MSK symptoms. Those requested for an appointment questionnaire were subjected to clinical examination and analyses of human leucocyte antigen B27 (HLA-­B27), rheumatoid factor (RF) and Participants reporting C reactive protein (CRP). post-­travel MSK Participants with symptoms among Written informed consent was obtained from all participants. post-­travel stool participants with stool The participants were not asked to assess the burden of the inter- findings N findings indicated,N (%) vention and time required to participate in the research. Total 224 38 (17.0) No pathogens 51 4 (7.8) Diagnostic criteria Any pathogen 173 34 (19.7) ReA was defined as the development of synovitis (swelling Multiple types of pathogens 86 18 (20.9) accompanied by pain and/or painful movement) in a previ- EPEC 99 20 (20.2) ously asymptomatic joint within 2 months after gastrointestinal ETEC 45 8 (17.8) symptoms. Inflammatory back pain (low back pain worsening at EAEC 104 21 (20.2) 27 night) was also regarded as ReA. Reactive tendinitis and enthe- EHEC 20 5 (25.0) sopathy (ReTe) were defined as tendinitis, heel or elbow pain, Shigella/EIEC 3 1 (33.3) and limitation of joint movement without joint swelling, and Salmonella 6 1 (16.7) reactive arthralgia (ReAlg) as pain in a previously asymptom- Campylobacter 13 1 (7.7) atic joint not swollen and with a normal movement range. MSK symptoms among participants with previous rheumatological *Six (2.7%) pretravel stool samples were positive for EPEC, four (1.8%) for EAEC and one (0.5%) for Campylobacter; of these, solely a participant with diagnosis were not defined as travel acquired. Campylobacter in pretravel stools developed MSK. EAEC was found as the sole 11 28 TD was defined according to WHO criteria for diarrhoea. pathogen in the same individual’s post-­travel stools. Mild TD was defined as 1–2, moderate TD as 3–5 and severe TD EPEC enteropathogenic Escherichia coli; EAEC, enteroaggregative Escherichia coli; as 6 or more loose or liquid stools or diarrhoea accompanied by EHEC, enterohaemorrhagic Escherichia coli; EIEC, enteroinvasive Escherichia coli; fever and/or bloody stools or requiring hospitalisation.11 ETEC, enterotoxigenic Escherichia coli; MSK, musculoskeletal.

Statistical analysis Data were analysed with SPSS software, V.22.0. Categorical vari- stools, the DEC were travel acquired (no DEC in pretravel stools). ables were analysed by Pearson’s χ² test, Fisher’s exact test or binary logistic regression analysis when applicable. The following variables were subjected to multivariate analysis: gender, age as Part one categorical variable, pathogen finding in post-­travel stool sample Risk factors of post-travel MSK symptoms (only-­DEC, only non-DEC,­ DEC and non-DEC­ and no patho- Of all 224 participants, 38 (17.0%) reported MSK symptoms gens); severity of TD; and antibiotic use. Statistical significance in the follow-up­ questionnaire (tables 1 and 2). In total, 155 was defined as p<0.05 or 95% CIs ranging only either above or (69.2%) contracted TD during the journey and 24 (15.5%) below 1. treated it with antibiotics. Scrutinising the relation between MSK symptoms and TD, we found that 18.7% (29/155) of those with and 13.0% (9/69) of those without TD reported MSK symp- Results toms (OR 1.5, 95% CI 0.7 to 3.4; p=0.297). A closer look at The first part of this bipartite study concerns risk factors of post-­ antibiotic use among those 109 with TD during travel and only travel MSK symptoms, while the second describes the nature and DEC in stools showed MSK symptoms for 27.3% (3/11) of those incidence of clinical findings among those with DEC as the only taking antibiotics and 19.4% (19/98) of those not taking (OR travel-­acquired pathogens. Our travellers visited sub-Saharan­ 1.6; 95% CI 0.4 to 6.4, p=0.691). Africa (48.2%); Southeast Asia (21.4%); South Asia (10.7%); In multivariate analysis, the findings of only-DEC­ were asso- South and Central America and the Caribbean (10.7%); Europe, ciated with MSK symptoms (aOR 3.9, 95% CI 1.2 to 13.3). By North America or Australia (5.8%); North Africa and Middle contrast, no association was observed for TD, severity of TD, East (2.7%); and East Asia (0.4%). antibiotic use, age or gender (table 2).

Pathogen findings among all 224 travellers Part two Of the 526 subjects initially recruited,21 224 returned all three MSK symptoms among travellers with only-DEC questionnaires and provided both stool samples. In pretravel Of the 151 participants with only-DEC­ in post-­travel stools, 31 specimens, the volunteers had few positive findings: six had (20.5%) reported MSK symptoms; one was not reached. Perusing (2.7%) EPEC, four (1.8%) EAEC and one (0.5%) Campylo- the questionnaires, the symptoms of 12 participants were either bacter. Only the participant with Campylobacter developed considered unrelated to travel (eg, similar joint symptoms before MSK symptoms. As for post-­travel samples, 173 (77.2%) partic- journey) or ascribed to something else (eg, trauma and infec- ipants had at least one pathogen and 86 (38.5%) multiple types tion). Eighteen were interviewed by phone or invited for a clinic of pathogens: DEC was detected for 168 (75.0%); Shigella/EIEC visit. Of these, nine had MSK symptoms related to some other for 3 (1.3%); Salmonella for 6 (2.7%); and Campylobacter for disorder (eg, osteoarthritis). The joint symptoms of nine partici- 13 (5.8%) (table 1). DEC were the only pathogens identified for pants were considered newly acquired. 151 (67.4%), only non-­DEC were detected for 5 (2.2%), DEC Three participants with Shigella/EIEC were not included in the and non-­DEC for 17 (7.6%) and no pathogens for 51 (22.8%). only-­DEC group, as described above. One reported MSK symp- Among all of those with MSK and a DEC finding in post-­travel toms and was diagnosed with ReAlg in clinical examination.

Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 607 Spondyloarthritis

Table 2 Risk factors of musculoskeletal (MSK) symptoms

Total* MSK +† MSK –† Univariate statistics Multivariable statistics n (%) n (%) n (%) P value OR (95% CI) P value aOR (95% CI) Total 224 38 (17.0) 186 (83.0) Gender Male 83 (37.1) 9 (10.8) 74 (89.2) reference 1.0 reference 1.0 Female 141 (62.9) 29 (20.6) 112 (79.4) 0.061 2.1 (1.0 to 4.8) 0.117 2.0 (0.8 to 4.9) Age Mean±SD range 39.5+/-16.7 48.1+/-12.7 37.7+/-16.9 (0–72) (21–67) (0–72) Age group (years) 0–17 20 (8.9) 0 (0) 20 (100) n/a n/a 18–30 57 (25.4) 5 (8.8) 52 (91.2) 0.656 0.7 (0.1 to 3.8) 0.353 0.4 (0.1 to 2.6) 31–50 85 (37.9) 14 (16.5) 71 (83.5) 0.691 1.4 (0.3 to 6.8) 0.897 1.1 (0.2 to 5.9) 51–64 46 (20.5) 17 (37.0) 29 (63.0) 0.083 4.1 (0.8 to 20.3) 0.131 3.6 (0.7 to 19.0) Over 65 16 (7.1) 2 (12.5) 14 (87.5) reference 1.0 reference 1.0 Pathogen finding Negative 51 (22.8) 4 (7.8) 47 (92.2) reference 1.0 reference 1.0 Only- ­DEC 151 (67.4) 31 (20.5) 120 (79.5) 0.047 3.0 (1.0 to 9.1) 0.030 3.9 (1.1 to 13.2) Only non-­DEC 5 (2.2) 0 (0) 5 (100) n/a n/a n/a n/a DEC+non- ­DEC 17 (7.6) 3 (17.6) 14 (82.4) 0.085 2.5 (0.5 to 12.6) 0.111 4.5 (0.7 to 27.9) TD during travel No TD 69 (30.8) 9 (13.0) 60 (87.9) reference 1.0 reference 1.0 Mild TD 53 (23.7) 9 (17.0) 44 (83.0) 0.544 1.4 (0.5 to 3.7) 0.571 1.4 (0.5 to 4.2) Moderate TD 62 (27.7) 14 (22.6) 48 (77.4) 0.156 1.9 (0.8 to 4.9) 0.272 1.8 (0.6 to 5.1) Severe TD 40 (17.9) 6 (15.0) 34 (85.0) 0.775 1.2 (0.4 to 3.6) 0.585 1.4 (0.4 to 5.3) Antibiotic use during travel‡ No AB 195 (87.1) 32 (16.4) 163 (87.6) reference 1.0 reference 1.0 AB for TD 24 (10.7) 5 (20.8) 19 (79.2) 0.586 1.3 (0.5 to 3.9) 0.311 2.0 (0.5 to 7.1) AB other 5 (2.2) 1 (20.0) 4 (80.0) 0.831 1.3 (0.1 to 11.8) 0.460 2.5 (0.2 to 28.4) *Percentages refer to proportion of risk factors. †Percentages refer to proportion of participants with and without MSK symptoms for each risk factor. ‡Antibiotic regimens: 18 fluoroquinolones, 4 macrolides, 2 third-generation­ cephalosporins, 3 unknown and 5 others. AB, antibiotic; aOR, adjusted OR; DEC, diarrhoeagenic Escherichia coli; TD, travellers' diarrhoea.

Clinical evaluation of nine participants with newly acquired The CRP values ranged between 3 mg/L and 6 mg/L. RF was MSK symptoms and DEC as only pathogen negative for all seven. Of the nine participants with reactive MSK symptoms, four met Seven of the nine (77.8%) participants with ReA/ReTe/ReAlg the criteria for ReA, two had ReTe and for the remaining three, had EPEC in their post-travel­ stools (table 4). Two had multiple ReAlg remained as diagnosis. The symptoms had appeared on pathogens, one EPEC and EAEC and the other EPEC and ETEC. average 21.5 (SD 23.5, range 0–56) days after TD. The mean duration of MSK symptoms was 82.8 days (SD 92.9, range 2–300). The mean age was 47.7 years (SD 7.6, range 30–56). Six Table 3 Clinical picture of the four patients with reactive arthritis (66.7%) were women and two had a history of ReA (one had (ReA) ReA and the other ReTe in this study). Eight (88.9%) reported TD during travel. None of the nine participants had contacted Patient 1 Patient 2 Patient 3 Patient 4 healthcare due to TD, nor taken antibiotics. One (11.1%) Age, years 48 47 50 50 reported urogenital symptoms during travel; none reported Gender (M/F) M F F F extra-­articular symptoms like uveitis or erythema nodosum. Pathogen EPEC EPEC EPEC EPEC Of the nine participants, seven (including all four with ReA) EAEC were evaluated at the rheumatology clinic on average 5.9 months Affected joints Knees Toe; finger Feet; knees; Shoulder; (SD 2.0, range 4–8) after the onset of joint symptoms. The clin- elbows elbow; wrist ical picture was mostly mild. The affected joints of the four ReA Joint swelling − + − + patients were knees, shoulder, elbow, wrist, fingers and toes Arthtralgia + + + + (table 3). At the time of clinical examination, only one patient Low back pain + − + − with ReA still had synovitis 8 months after symptom onset and HLA-B27­ antigen − − − − needed an intra-articular­ corticosteroid injection (left MTP I). Duration of ReA 2 10 3 1 Only one patient with ReA reported fever during diarrhoea, and (months) three patients had had mild TD. Only one with ReTe was HLA-­ EAEC, enteroaggregative E.coli; EPEC, enteropathogenic Escherichia coli; F, female; B27 positive, while all four with ReA were HLA-B27­ negative. M, male; ReA, reactive arthritis.

608 Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 Spondyloarthritis

it could prevent postinfectious sequelae such as ReA or irritable Table 4 Distribution of bacterial pathogens among patients with bowel syndrome.30 31 We found no association between devel- reactive musculoskeletal symptoms (n=9) opment of MSK symptoms and antibiotic treatment of TD. This ReA n=4* ReTe n=2 ReAlg n=3† accords with research showing that antibiotic treatment of infec- EPEC 4 1 2 tions caused by Campylobacter, ETEC or non-­typhoidal Salmo- ETEC 0 1 1 nellae does not prevent postinfectious joint symptoms.32–35 To EAEC 1 0 1 our knowledge, the present investigation is the first to scrutinise *One patient with two pathogens (EPEC+EAEC). the relation between antibiotic treatment of TD and MSK symp- †One patient with two pathogens (EPEC+ETEC). toms among travellers. EAEC, enteroaggregative Escherichia coli ; EPEC, enteropathogenic Escherichia The prospective study design allowed the very first evaluation coli; ETEC, enterotoxigenic Escherichia coli ; ReA, reactive arthritis; ReAlg, reactive of the incidence of various MSK symptoms among individuals arthralgia; ReTe, reactive tendinitis and enthesitis. with travel-acquired­ DEC. MSK symptoms were recorded for a considerable proportion (20.5%) of those with DEC as the All four with ReA had EPEC, one also had EAEC. None had sole post-­travel pathogen; the incidence of ReA was 2.6%, ReTe EHEC. 1.3% and ReAlg 2.0%. These figures resemble those reported for volunteers only recruited after acquisition of some pathogen. 5 Incidence of ReA, ReTe and ReAlg For postinfectious joint symptoms, Schiellerup et al report an On the basis of data on the 151 travellers with DEC as the sole incidence of 9.7% among patients with DEC—none had definite 6 pathogen, nine (6.0%) were diagnosed with travel-related­ ReA, and three (1.0%) had probable ReA. Townes et al describe ReA 29 ReTe or ReAlg, yielding for ReA an incidence of 2.6%, for ReTe for 0.25%, and Rees et al report prolonged joint symptoms 4 1.3% and for ReAIg 2.0%. for 4.5% of participants with E. coli O157. Locht and Krogfelt show joint symptoms among 16% of patients with Campylo- Discussion bacter (n=173) and 6% with ETEC (n=177) over a 6-­month This is the first prospective study to have employed multivariate follow-­up. As for incidence of MSK symptoms among patients 36 analysis in investigating whether reactive MSK symptoms are with other pathogens, in their systematic review, Ajene et al associated with acquisition of DEC and determining the inci- found ReA in the samples of 0.9%–1.2% of subjects with Salmo- 36 dence of enterogenic ReA among travellers with DEC. nella, Shigella or Campylobacter, a rate according with that Our main finding was that acquiring DEC is, indeed, associ- among our participants who had DEC. Finnish studies exploring ated with development of MSK symptoms. This study features ReA following infection with Salmonella, Shigella and Campy- three merits not seen in the four largish investigations that have lobacter have reported incidences of 4.5%, 7% and 7%, respec- 2 13 37 previously examined DEC as a trigger of reactive joint symp- tively. In our investigation, the symptoms of ReA mostly toms4–6 29: we collected stool samples prospectively, included remained mild, although one patient had arthritis persisting over control groups, and covered EAEC, one of the most frequently 8 months. All our patients with ReA were negative for HLA-­ detected pathogens among travellers with TD (table 5). Besides B27, a finding according with a previous study, which found no EAEC, we also covered, using the multiplex qPCR assay, EPEC, relation between HLA-B27­ and MSK symptoms following an 5 ETEC, EIEC and EHEC. infection with E. coli. Low prevalence of HLA-­B27 and mild In addition to this broad pathogen coverage, our prospec- clinical course have also been reported among other pathogens tive study design with pathogen-­positive and pathogen-­negative triggering ReA in population-based­ studies.2 6 13 37 travellers allowed multivariate analysis of predisposing factors. In previous studies, the microbe triggering ReA has often Contrary to previous studies, we found no association between remained unidentified. Fendler et al38 report a causative MSK symptoms and age, female sex or severity of TD, all of pathogen in 56% of ReA cases and, on the other hand, among which have previously been associated with the development of 47% of patients with undifferentiated oligoarthritis without MSK symptoms.2 5 30 31 We also looked at antibiotic treatment preceding symptoms of infection. Our data show a bacterial of TD, as some travel medicine researchers have speculated that pathogen for 87.9% of travellers with MSK symptoms and all

Table 5 Outline of previous studies assessing musculoskeletal symptoms following Escherichia coli infection HLA-­B27 status of those MSK symptoms/ReA Clinical picture/duration of symptoms with MSK symptoms/ Study (ref) Number of patients Type of DEC n (%) (months) ReA Locht and Krogfelt4 177 ETEC ReA: 10 (5.6) Not described / 5.5 Not examined Rees et al29 22 E. coli O157:H7 ReA: 1 (4.5)* Pain in hands and feet Not examined Schiellerup et al5 290 All E. coli 28 (9.7)/ 3 (1.0) Median VAS joint pain 35.5 mm 10% HLA-­B27 positive EPEC (n=17; 5.8%) 1 (5.9) ETEC (n=112; 38.6%) 8 (7.1) A/EEC (n=138; 47.6%) 16 (11.6) VTEC (n=23; 7.9%) 3 (13.0) Townes et al6 395 E. coli O157 35 (8.9)/1 (0.25) Described among other pathogens 12% HLA-­B27 positive† together *Infection with Campylobacter and concomitant diagnosis of ulcerative colitis. †among all participants with ReA tested, comprising findings of Campylobacter, Salmonella, Shigella, E. coli O157 and Yersinia. A/EEC, attaching and effacing E. coli (considered EPEC), E. coli O157; DEC, diarrhoeagenic E. coli; EPEC, enteropathogenic E. coli; ETEC, enterotoxigenic E. coli; HLA-B27,­ human leucocyte antigen B27; MSK, musculoskeletal; ReA, reactive arthritis; VAS, visual analogue scale 0-100mm;­ VTEC, verocytotoxin (shigatoxin) producing E. coli.

Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 609 Spondyloarthritis of those with ReA. Interestingly, the development of MSK symp- Contributors Study concept and design: AK and JK; acquisition of data: RT, SHP, JK toms was not related to the severity of TD but, by contrast, those and AK; analysis and interpretation of results: RT, TL and AK; drafting of manuscript: with MSK symptoms or ReA reported mostly mild or moderate RT, TL and AK; statistical analysis: TL; critical comments: TH, SHP and ML-­R; final approval of version published: all authors. TD. The pathogens were identified only because of participation in this study; travellers mostly do not seek medical care for mild Funding The work was supported by the Finnish Governmental Subsidy for 26 Health Science Research, the SSAC Foundation, the Paulo Foundation, the Sigrid TD symptoms. Furthermore, with the exception of EHEC, Jusélius Foundation, the Finnish Cultural Foundation and the Finnish Society for traditional stool diagnostics fails to detect species of DEC, and Rheumatology. therefore, the pathogens would have remained unidentified Disclaimer The funding sources had no involvement in study design, data in laboratories not using qPCR assays with a broad pathogen collection, analysis, interpretation of data, writing of report, and decision to submit coverage.9 20 25 the article for publication. In our data, EPEC was the pathogen most frequently detected Competing interests AK has received honorary for lectures (Valneva and among those with new MSK symptoms (77.8%) and ReA Immuron) and investigator-­initiated grants (Pfizer and alneva),V none of these (100%). Its role in adult TD has remained disputable,9 11 21 yet relevant to the current manuscript. JK is an employee of Mobidiag developing for this study population, we have previously reported EPEC diagnostics test for infectious diseases. No commercial tests, however, are used in the study. more frequently among those with than those without TD.11 Despite the considerable prevalence of EAEC among travel- Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. lers with TD, it has not been covered in the previous studies reporting association between various DEC and ReA or other Patient consent for publication Not required. postinfectious joint symptoms (table 5). We show EAEC for Ethics approval The study protocol was approved by the Ethics Committee of 55.3% (21/38) of the participants with MSK symptoms versus Helsinki University Hospital. 44.6% (83/186) without them (table 1), and for 25.0% (1/4) Provenance and peer review Not commissioned; externally peer reviewed. with ReA versus 20.0% (1/5) with ReTe and ReAlg (table 4). Data availability statement All data relevant to the study are included in the ETEC deserves special attention, since it is often considered to article or uploaded as supplementary information. Any other data are available from cause more severe TD than the other DEC.7 It was detected in the corresponding author. the specimens of 21.1% of the participants with MSK symptoms Open access This is an open access article distributed in accordance with the and none of those with ReA. 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 Limitations and strengths is non-­commercial. See: http://creativecommons.​ ​org/licenses/​ ​by-​nc/​4.0/.​ First, as we could not collect stool samples while abroad, the primary TD pathogen may in some cases have already disap- ORCID iD peared before sampling. Second, our data may slightly overesti- Riitta Tuompo http://orcid.​ ​org/0000-​ ​0003-​2559-​5904 mate the incidence of MSK symptoms, since those with symptoms after travel may have been more eager to return the follow-up­ References 1 Hannu T. Reactive arthritis. Best Pract Res Clin Rheumatol 2011;25:347–57. questionnaire. Third, the time period between primary pathogen 2 Hannu T, Mattila L, Rautelin H, et al. Campylobacter-­triggered reactive arthritis: a acquisition and clinical examination was rather long. Naturally, population-based­ study. Rheumatology 2002;41:312–8. our lengthy follow-up­ also enabled us to evaluate the patients’ 3 laasila K, Leirisalo-­Repo M. Recurrent reactive arthritis associated with urinary tract recovery. However, we may have missed some ReA cases where infection by Escherichia coli. J Rheumatol 1999;26:2277–9. the joint symptoms appeared later than 3 weeks after travel. 4 locht H, Krogfelt KA. Comparison of rheumatological and gastrointestinal symptoms after infection with Campylobacter jejuni/coli and enterotoxigenic Escherichia coli. The main merits of our investigation are its prospective design Ann Rheum Dis 2002;61:448–52. and multiplex qPCR method with a wide coverage of pathogens. 5 schiellerup P, Krogfelt KA, Locht H. A comparison of self-­reported joint symptoms They allowed us to compare symptomatic versus asymptomatic following infection with different enteric pathogens: effect of HLA-­B27. J Rheumatol participants with and without DEC, to look for an association 2008;35:480–7. between MSK symptoms and travel-acquired­ DEC and estimate 6 Townes JM, Deodhar AA, Laine ES, et al. Reactive arthritis following culture-­confirmed infections with bacterial enteric pathogens in Minnesota and Oregon: a population-­ the incidence of MSK symptoms. Furthermore, data on antibi- based study. Ann Rheum Dis 2008;67:1689–96. otic use afforded a rare opportunity to estimate whether antibi- 7 steffen R, Hill DR, DuPont HL. Traveler’s diarrhea: a clinical review. JAMA otic treatment of TD could prevent MSK symptoms. 2015;313:71–80. 8 Jiang Z-­D, Dupont HL, Brown EL, et al. Microbial etiology of travelers’ diarrhea in Mexico, Guatemala, and India: importance of enterotoxigenic Bacteroides fragilis and Conclusion Arcobacter species. J Clin Microbiol 2010;48:1417–9. 9 lertsethtakarn P, Silapong S, Sakpaisal P, et al. Travelers’ diarrhea in Thailand: a This was the first prospective study to explore whether travel-­ quantitative analysis using TaqMan® array card. Clin Infect Dis 2018;67:120–7. acquired DEC are associated with reactive MSK symptoms 10 Paschke C, Apelt N, Fleischmann E, et al. Controlled study on enteropathogens in and to provide incidences for such development. Multivariate travellers returning from the tropics with and without diarrhoea. Clin Microbiol Infect analysis showed acquisition of DEC to be connected with risk 2011;17:1194–200. of developing MSK symptoms yet the risk did not depend on 11 lääveri T, Antikainen J, Pakkanen SH, et al. Prospective study of pathogens in asymptomatic travellers and those with diarrhoea: aetiological agents revisited. Clin severity of TD symptoms. The incidence of ReA proved fairly Microbiol Infect 2016;22:535–41. low and the course mild; no association was seen with HLA-­ 12 Croxen MA, Law RJ, Scholz R, et al. Recent advances in understanding enteric B27. Further investigations are needed to confirm whether pathogenic Escherichia coli. Clin Microbiol Rev 2013;26:822–80. ReA triggered by DEC remains milder than that caused by the 13 Hannu T, Mattila L, Siitonen A, et al. Reactive arthritis attributable to Shigella conventionally known bacteria. Antibiotic treatment of TD did infection: a clinical and epidemiological nationwide study. Ann Rheum Dis 2005;64:594–8. not prevent post-­travel MSK symptoms. 14 Mielants H, Veys EM, Cuvelier C, et al. Ileocolonoscopic findings in seronegative Acknowledgements We wish to express our debt of gratitude to the late Dr spondylarthropathies. Br J Rheumatol 1988;27 Suppl 2:95–105. Jukka Riutta for recruiting the volunteers. We would also like to thank the nurses 15 nissilä M, Lahesmaa R, Leirisalo-­Repo M, et al. Antibodies to Klebsiella pneumoniae, at the Travel Clinic of Aava Medical Centre for help, and the personnel of Helsinki Escherichia coli, and Proteus mirabilis in ankylosing spondylitis: effect of sulfasalazine University Hospital Laboratory (HUSLAB) for assistance with the stool specimens. treatment. J Rheumatol 1994;21:2082–7.

610 Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 Spondyloarthritis

16 Mäki-­Ikola O, Nissilä M, Lehtinen K, et al. Antibodies to Klebsiella pneumoniae, 27 selmi C, Gershwin ME. Diagnosis and classification of reactive arthritis. Autoimmun Escherichia coli and Proteus mirabilis in the sera of patients with axial and peripheral Rev 2014;13:546–9. form of ankylosing spondylitis. Br J Rheumatol 1995;34:413–7. 28 World Health Organization. Health topics: diarrhoea. Available: http://www.​who.​int/​ 17 stolwijk C, Boonen A, van Tubergen A, et al. Epidemiology of spondyloarthritis. Rheum topics/diarrhoea/​ ​en/ Dis Clin North Am 2012;38:441–76. 29 Rees JR, Pannier MA, McNees A, et al. Persistent diarrhea, arthritis, and other 18 noyori K, Okamoto R, Takagi T, et al. Experimental induction of arthritis in complications of enteric infections: a pilot survey based on California FoodNet rats immunized with Escherichia coli 0:14 lipopolysaccharide. J Rheumatol surveillance, 1998-1999. Clin Infect Dis 2004;38 Suppl 3:S311–7. 1994;21:484–8. 30 Riddle MS, Connor BA, Beeching NJ, et al. Guidelines for the prevention and 19 syrbe U, Scheer R, Wu P, et al. Differential synovial Th1 cell reactivity towards treatment of travelers’ diarrhea: a graded expert panel report. J Travel Med Escherichia coli antigens in patients with ankylosing spondylitis and rheumatoid 2017;1:S57–74. arthritis. Ann Rheum Dis 2012;71:1573–6. 31 Yates JA, Stetz LC. Reiter’s syndrome (reactive arthritis) and travelers’ diarrhea. J Travel 20 antikainen J, Kantele A, Pakkanen SH, et al. A quantitative polymerase chain reaction Med 2006;13:54–6. assay for rapid detection of 9 pathogens directly from stools of travelers with 32 locht H, Mølbak K, Krogfelt KA. High frequency of reactive joint symptoms after an diarrhea. Clin Gastroenterol Hepatol 2013;11:1300–7. outbreak of Salmonella enteritidis. J Rheumatol 2002;29:767–71. 21 lääveri T, Vilkman K, Pakkanen SH, et al. A prospective study of travellers’ diarrhoea: 33 schmitt SK. Reactive arthritis. Infect Dis Clin North Am 2017;31:265–77. analysis of pathogen findings by destination in various (sub)tropical regions. Clin 34 schönberg-­Norio D, Mattila L, Lauhio A, et al. Patient-­Reported complications Microbiol Infect 2018;24:908.e9–908.e16. associated with Campylobacter jejuni infection. Epidemiol Infect 22 Connor BA, Riddle MS. Post-­Infectious sequelae of travelers’ diarrhea. J Travel Med 2010;138:1004–11. 2013;20:303–12. 35 esan OB, Pearce M, van Hecke O, et al. Factors associated with sequelae of 23 Thomas DG, Roberton DM. Reiter’s syndrome in an adolescent girl. Acta Paediatr Campylobacter and Non-­typhoidal Salmonella infections: a systematic review. 1994;83:339–40. EBioMedicine 2017;15:100–11. 24 singh Sangha M, Wright ML, Ciurtin C. Strongly positive anti-­CCP antibodies in 36 ajene AN, Fischer Walker CL, Black RE. Enteric pathogens and reactive arthritis: a patients with sacroiliitis or reactive arthritis post-E. coli infection: A mini case-­series systematic review of Campylobacter, Salmonella and Shigella-­associated reactive based review. Int J Rheum Dis 2018;21:315–21. arthritis. J Health Popul Nutr 2013;31:299–307. 25 Houtman PM, Spoorenberg A, Schuurs TA, et al. Reactive arthritis: does screening 37 Tuompo R, Hannu T, Mattila L, et al. Reactive arthritis following Salmonella infection: stools by polymerase chain reaction for Shiga toxin-­producing Escherichia coli (STEC) a population-­based study. Scand J Rheumatol 2013;42:196–202. make sense? Scand J Rheumatol 2012;41:237–9. 38 Fendler C, Laitko S, Sörensen H, et al. Frequency of triggering bacteria in patients with 26 Vilkman K, Pakkanen SH, Lääveri T, et al. Travelers’ health problems and behavior: reactive arthritis and undifferentiated oligoarthritis and the relative importance of the prospective study with post-­travel follow-­up. BMC Infect Dis 2016;16:328. tests used for diagnosis. Ann Rheum Dis 2001;60:337–43.

Tuompo R, et al. Ann Rheum Dis 2020;79:605–611. doi:10.1136/annrheumdis-2019-216736 611 Systemic lupus erythematosus

Epidemiological science Impact of the new American College of Cardiology/ American Heart Association definition of hypertension on atherosclerotic vascular events in systemic lupus erythematosus Konstantinos Tselios,1 Dafna D Gladman ‍ ‍ ,2 Jiandong Su,3 Murray Urowitz ‍ ‍ 4

Handling editor Josef S Abstract Key messages Smolen Background The 2017 American College of Cardiology/American Heart Association guidelines ►► Additional material is What is already known about this subject? published online only. To view defined hypertension at ≥130/80 mm Hg. Studies on ►► The new guidelines for arterial hypertension please visit the journal online patients with connective tissue diseases were not from the American College of Cardiology/ (http://dx.​ ​doi.org/​ ​10.​1136/​ considered. Our aim was to assess the impact of this American Heart Association define hypertension annrheumdis-2019-​ ​216764). definition on atherosclerotic ascularv events (AVEs) in at a level of 130/80 mm Hg. However, studies 1 systemic lupus erythematosus. Centre for Prognosis Studies in patients with systemic autoimmune diseases Patients methods Individuals from the Toronto Lupus in the Rheumatic Diseases, were not considered and recommendations University of Toronto Lupus Clinic with at least 2 years of follow-­up and no prior AVE were not made for such patients. Clinic, Toronto, Ontario, Canada were divided in three groups according to their mean 2Centre for Prognosis Studies blood pressure (BP) over that period (≥140/90 mm Hg, in the Rheumatic Diseases, What does this study add? University of Toronto, Toronto, 130-139/80-89 mm Hg and <130/80 mm Hg). They were ►► This observational longitudinal study of 1532 Ontario, Canada followed until the first occurrence of anA VE (fatal or 3 patients with systemic lupus erythematosus, Lupus Clinic, Centre for non-­fatal coronary artery disease, cerebrovascular event followed by 10.8 years on average, showed Prognosis Studies in the and peripheral vascular disease) or last visit. Groups were Rheumatic Diseases, University that an adjusted mean blood pressure (BP) compared as per the baseline atherosclerotic risk factors. Health Network, Toronto, level of 130–139/80–89 mm Hg over the first Ontario, Canada A multivariable time-­dependent analysis was performed 4 2 years of observation is associated with a Center for Prognosis Studies in to adjust for the presence of other risk factors. 2.5-­fold increase in the risk of atherosclerotic the Rheumatic Diseases, Toronto Results Of 1532 patients satisfying the inclusion Western Hospital, University of cardiovascular events (AVEs) compared with criteria, 155 (10.1%) had a BP ≥140/90 mm Hg, 316 Toronto, Lupus Clinic, Toronto, normotensive (<130/80 mm Hg) patients (20.6%) 130–139/80–89 mm Hg and 1061 (69.3%) Ontario, Canada with lupus. This level of BP conferred a 73% were normotensives. After a mean follow-­up of 10.8 increased risk for AVEs after adjustment for all years, 124 AVEs were documented. The incidence rates Correspondence to traditional and disease-­related atherosclerotic Dr Murray Urowitz, Center for were 18.9, 11.5 and 4.5 per 1000 patient-­years for risk factors. Prognosis Studies in the the three groups, respectively (p=0.0007 between the Rheumatic Diseases, Toronto 130–139/80–89 mm Hg group and the normotensives). Western Hospital, University of How might this impact on clinical practice or A mean BP of 130–139/80–89 mm Hg over the Toronto, Lupus Clinic, Toronto, future developments? M5T 2S8, Canada; first 2years was independently associated with the ► These findings support that the target level of m.​ ​urowitz@utoronto.​ ​ca occurrence of AVEs (HR 1.73, 95% CI 1.13 to 2.65, ► BP should be <130/80 mm Hg for all patients p=0.011). Received 4 December 2019 with lupus. Revised 14 February 2020 Conclusion Patients with lupus with a sustained Accepted 17 February 2020 mean BP of 130–139/80–89 mm Hg over 2 years had Published Online First a significantly higher incidence ofA VEs compared with 10 March 2020 normotensive individuals. This BP level should be the According to the same guidelines, pharmacologic target for antihypertensive therapy to minimise their treatment should be offered in all individuals with cardiovascular risk. stage 1 hypertension and a 10-­year cardiovascular (CV) risk ≥10% (as calculated by the Atheroscle- rotic Cardiovascular Disease (ASCVD) risk esti- mator), regardless of age.1 Despite the fact that Introduction an entire section was devoted on special patient The recently published guidelines for the manage- groups and certain comorbidities, no recommen- ment of arterial hypertension in adults by the dations were made for patients with connective © Author(s) (or their employer(s)) 2020. No American College of Cardiology/American Heart tissue diseases such as systemic lupus erythematosus commercial re-­use. See rights Association (ACC/AHA) recommended a signifi- (SLE). and permissions. Published cantly lower threshold of systolic and diastolic Hypertension is detected in up to 74% of patients by BMJ. blood pressure (SBP, DBP) for diagnosis.1 The levels with lupus and is one of the strongest predictors To cite: Tselios K, of 130–139 mm Hg for SBP and 80–89 mm Hg for for accelerated atherosclerosis and atherosclerotic Gladman DD, Su J, DBP are now considered as stage 1 hypertension, in vascular events (AVEs).3 4 Patients with lupus have et al. Ann Rheum Dis contrast to the previous recommendations in which a fivefold increased lifetime risk for AVEs, while 2020;79:612–617. hypertension was defined as BP ≥140/90 mm Hg.2 the relative risk is even higher in premenopausal

612 Tselios K, et al. Ann Rheum Dis 2020;79:612–617. doi:10.1136/annrheumdis-2019-216764 Systemic lupus erythematosus women.5 6 Therefore, proper management of hypertension is 1. Group 1: patients with an adjusted mean BP <130/80 mm of paramount importance in SLE. However, on the basis of the Hg (normotensives). recent guidelines, the patient with typical lupus (young female 2. Group 2: patients with an adjusted mean BP 130-139/80-89 with no traditional atherosclerotic risk factors) would be consid- mm Hg (stage 1 hypertension). ered as a low-risk­ individual7 and not offered treatment for a 3. Group 3: patients with an adjusted mean BP ≥140/90 mm BP of 130–139/80–89 mm Hg. Moreover, CV risk calculation Hg (stage 2 hypertension). is problematic in SLE, primarily because the available risk scores In cases where the SBP and DBP would classify a given patient do not take into account the disease itself and most are designed in a different group, the individual was categorised in the group for patients older than 40 years. Although there are no studies with the higher BP (eg, a patient with a BP of 125/85 mm Hg was that used the ASCVD risk estimator7 in patients with lupus, classified in group 2). Hypertension was defined regardless of other risk scores, such as the modified 10-­year Framingham Risk use of antihypertensive therapy. The baseline was defined as the Score indicate that the classic risk scores significantly underesti- 2-­year time point since enrolment to the UTLC. All individuals mate CV risk.8 were followed until the occurrence of an AVE or the last visit. As such, the management of hypertension in lupus may be AVEs were defined as new-onset­ angina, myocardial infarction delayed, particularly at the levels of 130–139/80–89 mm Hg. (MI), revascularisation procedures (percutaneous transluminal The aim of the present study was to assess whether the new coronary angioplasty or coronary artery by-­pass graft), conges- definition of hypertension is predictive for AVEs in patients tive heart failure (CHF) of ischaemic origin, peripheral vascular with lupus and, specifically investigate if the BP threshold of disease (PVD) requiring angioplasty, transient ischaemic attack 130–139/80–89 mm Hg carries an excessive CV risk in SLE. (TIA), stroke and CV deaths (fatal MI or stroke). The three groups of patients were compared at baseline as per the variables that have been reported to increase the risk for Patients and methods AVEs according to a recent systematic review.4 These included The University of Toronto Lupus Clinic (UTLC) has currently demographic (age, sex, race/ethnicity), clinical (disease duration, enrolled 2008 patients since its establishment in 1970. All patients disease activity, accrued damage, current renal function), immu- fulfilled the revised American College of Rheumatology criteria nological (antiphospholipid antibodies) and therapeutic vari- for SLE classification9 or had three criteria and a supportive ables (antimalarials and glucocorticosteroids). The prevalence of biopsy. Patients are followed regularly at 2–6 months intervals traditional atherosclerotic risk factors (diabetes, dyslipidaemia, according to a standardised research protocol, which is regularly smoking) was compared as well as the frequency of the antihy- updated. This protocol captures demographic, clinical, immuno- pertensive, hypolipidaemic and antiplatelet/anticoagulant drugs logical and therapeutic variables as well as most comorbidities. used. Incidence rates of AVEs were calculated as first occurrence Concerning hypertension, the protocol captures the actual BP of an event by 1000 patient-­years for each group, while the (SBP/DBP) during the clinic visit (mean of two measurements effect of baseline BP groups was modelled using time-dependent­ in the sitting position with an aneroid sphygmomanometer Cox regression. after a 5 min rest) as well as the current pharmacological treat- ment, including all major classes of the antihypertensive drugs Statistical analysis (diuretics, ACE inhibitors, angiotensin receptor blockers (ARBs), The adjusted mean BP was used as a categorical variable to clas- beta-­blockers, calcium channel blockers, centrally acting agents sify patients into the three groups (normotensives, stage 1 hyper- and others). All individuals have provided written informed tension, stage 2 hypertension). Patients’ demographics, clinical consent for studies being conducted at the UTLC. and immunological variables and treatments, as well as the tradi- For the purpose of the present study, included patients should tional atherosclerotic risk factors are presented as mean±SD or have had at least 2 years of follow-­up after enrolment in the counts (percentages) and were compared using analysis of vari- UTLC and at least one additional clinic visit after that 2-­year ance for continuous variables and X2 tests for categorical vari- period. Patients with less than 2 years of follow-­up or with an ables. Univariate and multivariable Cox proportional hazards AVE at any time prior to the 2 years time point (baseline) were regressions were performed using the demographic variables as excluded. Given that the levels of BP may fluctuate significantly time fixed and the other features as time-dependent­ predictors over time in SLE and that BP measurements at one point in time 10 (being updated at each visit). All potential confounders along are unreliable for the prediction of long-­term outcomes, the with the main variable of interest (groups based on the adjusted adjusted mean BP was calculated for all eligible patients. This mean BP) were included in the analysis. Proportionality assump- considered all BP measurements (regardless of treatment) that tion tests for Cox models were done using Schoenfeld residuals a patient had had in the first 2 years of follow-­up and the time whereas the linearity of continuous variables in Cox models spent with each BP (time from visit to visit), assuming that this was tested using Martingale residuals.10 Statistical analysis was would remain practically unaltered between clinic visits. The performed with SAS V.9.4; p<0.05 was considered significant. mathematical equation used was n x +x 1 i i− 2 ti i=2 Results ∑( n ) ti Of 2008 patients, 1532 (88.1% females) satisfied the inclusion i=2 criteria. Of the excluded patients, 455 had a follow-up­ shorter ∑ where xi is the level of the variable (SBP or DBP) at visit i, and than 2 years (median 1 year) and 21 had an AVE prior to the ti is the time interval between visit i and i-1. By incorporating baseline. The mean age of the included patients at baseline was the time interval between measurements in its calculation, the 36.2±14.3 years and mean disease duration 6.1±6.3 years. One adjusted mean BP considers the length of time that SBP and DBP hundred and fifty-­five patients (10.1%) had an adjusted mean BP are presumed to have remained at a particular level. ≥140/90 mm Hg and 316 (20.6%) a mean BP of 130–139/80–89 Patients were then categorised in three different groups mm Hg over the first 2 years; the rest 1061 (69.3%) were normo- according to their adjusted mean BP, as follows: tensives. The calculation of the adjusted mean BP was based on

Tselios K, et al. Ann Rheum Dis 2020;79:612–617. doi:10.1136/annrheumdis-2019-216764 613 Systemic lupus erythematosus

Table 1 Traditional and lupus-­related atherosclerotic risk factors at baseline ≥140/90 mm Hg 130-139/80-89 mm Hg <130/80 mm Hg Variable (n=155) (n=316) (n=1061) P value Age (year) 46.8±14.6 39.7±13.2 33.6±13.7 <0.001 Females (n, %) 126 (81.3) 266 (84.2) 957 (90.2) <0.001 Caucasians (n, %) 119 (76.8) 226 (71.5) 647 (61) <0.001 Blacks (n, %) 23 (14.8) 40 (12.7) 154 (14.5) Others (n, %) 13 (8.4) 50 (15.8) 260 (24.5) Adjusted mean BP (mm Hg) 149/91 135/86 112/71 <0.001 Diabetes (n, %) 17 (11.0) 18 (5.7) 23 (2.2) <0.001 Dyslipidaemia* (n, %) 94 (60.6) 121 (38.3) 195 (18.4) <0.001 Total cholesterol (mmol/L) 5.9±1.8 5.1±1.3 4.5±1.1 <0.001 Smoking (current, n, %) 36 (23.2) 57 (18.0) 159 (15.0) 0.004 Disease duration (y) 5.5±5.8 6.9±7.0 6.0±6.0 0.026 SLEDAI-2K­ 5.2±5.0 5.2±5.6 4.3±4.4 <0.001 SDI 1.0±1.3 0.7±1.2 0.5±0.9 <0.001 eGFR (mL/min/1.73 m2) 69.9±31.6 86.0±32.4 102.0±29.4 <0.001 aPL† (n, %) 53 (34.2) 75 (23.7) 216 (20.4) 0.003 Antimalarials (n, %) 49 (31.6) 176 (55.7) 721 (68.0) <0.001 Glucocorticosteroids (GCS) (n, %) 125 (80.6) 230 (72.8) 690 (65.0) <0.001 Cumulative GCS dose (g)†‡ 16.9±17.9 18.9±22.6 15.5±18.7 <0.001 Antihypertensives (n, %) 59 (38.1) 95 (30.1) 168 (15.8) <0.001 Statins (n, %) 0 (0) 1 (0.3) 0 (0) 0.146 Antiplatelets/anticoagulants 8 (5.2) 19 (6) 94 (8.9) 0.106 (n, %) P values: trend from ANOVA for continuous variables, X2 test for categorical variables. *Abnormal total cholesterol or low-­density lipoprotein for two consecutive visits. †Lupus anticoagulant and/or anticardiolipin antibodies (IgM, IgG). ‡From enrolment up to the baseline. ANOVA, analysis of variance; aPL, antiphospholipid antibodies; BP, blood pressure; eGFR, estimated glomerular filtration rate; SDI, Systemic Damage Index; SLEDAI, SLE Disease Activity Index 2000.

6.3 BP measurements (average) over 2 years. The traditional and revascularisation procedures (n=12), CHF of ischaemic origin lupus-­related atherosclerotic risk factors as well as the treatment (n=9), cerebrovascular events (n=18) and PVD (n=2). The variables at baseline are shown in table 1. overall prevalence of AVEs was 32 (20.6%), 41 (13%) and 51 After a mean follow-up­ of 10.8 years (16 601 patient-years),­ (4.8%) in the ≥140/90 mm Hg, the 130–139/80–89 mm Hg and 124 AVEs were documented, including 20 CV deaths. The non-­ the <130/80 mm Hg groups, respectively. The incidence rates fatal AVEs were new-­onset angina (n=40), acute MI (n=23), were 18.9, 11.5 and 4.5 per 1000 patient-­years, respectively. The Kaplan-Meier­ curve for the occurrence of AVEs over time (censored at 15 years) in the three groups is shown in figure 1. The differences were statistically significant among all groups (p=0.0007 between normotensives and stage 1 hyperten- sion, p=0.008 between stage 1 and stage 2 hypertension and p<0.0001 between normotensives and stage 2 hypertension). Similar trends were observed in the subcategories of the AVEs, such as hard AVEs (excluding angina and TIAs), coronary artery disease (CAD) events, cerebrovascular events and CV deaths alone, figure 2. In a time-dependent­ multivariable analysis for the identifica- tion of predictors for AVEs, an adjusted BP of 130–139/80–89 mm Hg over the first 2 years since enrolment in the UTLC was independently associated with the occurrence of AVEs (HR 1.73, 95% CI 1.13 to 2.69, p=0.011). Other associated factors were smoking, use of glucocorticosteroids, use of antiplatelet/anti- coagulant agents and disease activity (as expressed by the SLE Figure 1 Kaplan-­Meier curve for the occurrence of AVEs over Disease Activity Index 2000), details in table 2. time (censored at 15 years) in the three patient groups (red line: A sensitivity analysis excluding 30 patients with end-­ ≥140/90 mm Hg, yellow line: 130–139/80–89 mm Hg, green line stage renal disease (7 in the normotensive group, 13 in the <130/80 mm Hg). The differences were statistically significant among all 130–139/80–89 mm Hg group and 10 in the ≥140/90 mm Hg groups. The baseline (T0) is 2 years after enrolment in the clinic. AVEs, group) yielded similar results (online supplementary tables 1 atherosclerotic vascular events; BP, blood pressure. and 2).

614 Tselios K, et al. Ann Rheum Dis 2020;79:612–617. doi:10.1136/annrheumdis-2019-216764 Systemic lupus erythematosus

occasions.1 This was based on previous studies that showed that BP fluctuates considerably in patients with lupus, primarily because of variations in disease activity (and, consequently, increased use of glucocorticosteroids and/or non-steroidal­ anti-­ inflammatory drugs) and renal involvement.3 This means that a given patient may have high BP readings intermittently and for a short period of time; the BP will normalise after successful control of disease activity. It is unknown if these individuals carry a higher CV risk compared with other patients. None- theless, Nikpour et al showed that time-­adjusted BP, reflecting the cumulative exposure to hypertension over time, was more reliable than single BP readings for the prediction of AVEs in patients with lupus.11 The new ACC/AHA definition inevitably increases the preva- lence of hypertension and, consequently, the number of patients that should be considered for management. In a recent study of 38 276 American adults from the National Health and Nutrition Figure 2 Cumulative incidence of AVEs in the three bP groups. Similar Examination Survey, Dorans et al reported an age-­standardised trends were observed with REGARDS to the hard AVEs (excluding prevalence of 45.4% and an increase of more than 20% in the angina and transient ischaemic attacks), CAD events and CV deaths. total number of individuals with hypertension in the USA.12 The differences in the incidence rates of cerebrovascular events were In our cohort, the prevalence of hypertension (based on the not statistically significant. *P<0.001, **P=0.008. AVEs, atherosclerotic adjusted mean BP over the first 2 years) was 10.1% according vascular events; BP, blood pressure; CAD, coronary artery disease; CV, to the previous definition of ≥140/90 mm Hg and increased cardiovascular. dramatically to 30.7% with the new threshold. In older studies in patients with lupus that used the previous definition of hyper- Discussion tension, the prevalence reached 74%3; this should be expected In the present study, we showed that patients with SLE with an to rise with the new definition. adjusted mean BP of 130–139/80–89 mm Hg (regardless of anti- Several studies in SLE showed that hypertension is inde- hypertensive therapy) over the first 2 years of observation had a pendently associated with increased rates of AVEs with a HR significantly higher risk for the development of AVEs compared ranging from 1.05 to 3.5.6 13–18 Furthermore, hypertension with normotensive patients (adjusted mean BP <130/80 mm Hg). has been related to every surrogate end-point­ for atheroscle- These patients had a higher prevalence of other traditional and rosis, including impaired endothelial function,19 20 arterial stiff- lupus-­related risk factors for atherosclerosis as well. However, a ness,21–23 increased carotid intima–media thickness and plaque time-­dependent multivariable analysis with multiple corrections formation,24–26 coronary artery calcification27 and angiograph- for all these factors revealed that this level of BP independently ically proven CAD.28 It was also an independent risk factor for increased the risk for AVEs by 73% over the entire follow-­up myocardial perfusion defects.29 In all these studies, the threshold period (10.8 years on average). for the definition of hypertension was ≥140/90 mm Hg. It seems A time-adjusted­ measure of BP over the first 2 years of possible that the application of a lower threshold, such as the observation was selected instead of the traditional ACC/AHA 130–139/80–89 mm Hg, may further increase the impact of definition of the average BP of ≥2 readings in two separate hypertension on the CV risk of such patients.

Table 2 Univariable and multivariable time dependent COX regression analysis for predictors of AVEs Univariable analysis Multivariable analysis Variable HR 95% CI P value HR 95% CI P value Stage 1 hypertension versus normotensives 1.77 1.17 to 2.68 0.007 1.73 1.13 to 2.65 0.0112 Stage 2 hypertension versus normotensives 1.96 1.23 to 3.12 0.0047 1.65 1.01 to 2.69 0.0456 Age 1.06 0.94 to 1.19 0.3311 2.01 0.96 to 4.22 0.064 Age squared 1.004 0.99 to 1.01 0.4396 0.99 0.99 to 1.00 0.0675 Female 0.51 0.31 to 0.82 0.0059 0.70 0.42 to 1.74 0.1787 Caucasian 1.16 0.75 to 1.80 0.5036 1.18 0.74 to 1.88 0.4771 Diabetes 1.60 0.93 to 2.86 0.0891 1.31 0.73 to 2.35 0.3686 Dyslipidaemia* 2.55 0.58 to 11.24 0.2169 1.71 0.33 to 8.83 0.5202 Smoking 1.61 1.03 to 2.50 0.0352 1.86 1.17 to 2.96 0.0084 SLEDAI-2K­ 1.11 1.07 to 1.15 <0.0001 1.10 1.06 to 1.15 <0.0001 eGFR 0.99 0.99 to 1.00 0.0593 1.00 0.99 to 1.00 0.2388 Antiphospholipid antibodies† 1.70 1.15 to 2.53 0.00081 1.38 0.92 to 2.07 0.1236 Glucocorticoids 2.02 1.40 to 2.92 0.0002 1.76 1.19 to 2.60 0.0043 Anticoagulants/antiplatelets 3.05 2.09 to 4.46 <0.0001 3.34 2.25 to 4.94 <0.0001 Bold indicates variables associated with increased risk for atherosclerotic vascular events in the multivariable analysis. *Abnormal total cholesterol or low-­density lipoprotein for two consecutive visits. †Lupus anticoagulant and/or anticardiolipin antibodies (IgM, IgG). AVE, atherosclerotic vascular event; eGFR, estimated glomerular filtration rate; SLEDAI-2K,­ Systemic Lupus Erythematosus Disease Activity Index 2000.

Tselios K, et al. Ann Rheum Dis 2020;79:612–617. doi:10.1136/annrheumdis-2019-216764 615 Systemic lupus erythematosus

The findings of the present study support that the target BP over more than 10 years of follow-up­ on average compared with should be less than 130/80 mm Hg in all patients with lupus normotensive individuals. In a multivariable time-dependent­ in order to minimise their CV risk. This seems to be indepen- analysis, this level of BP conferred a 73% higher risk for AVEs dent of age or the absence of other traditional risk factors that after adjusting for other traditional and lupus-related­ risk factors. could erroneously lead to a low predicted 10-year­ CV risk. As The management of hypertension in patients with lupus should mentioned above, most CV risk calculators are not designed to commence early and aim at levels below 130/80 mm Hg. estimate the CV risk in patients younger than 40 years. Hence, Contributors All authors were involved in drafting the article or revising it the CV risk cannot be reliably predicted for the majority of critically for important intellectual content, and all authors approved the final patients with lupus. The average age of the patients with an version to be published. MU had full access to all of the data in the study and adjusted mean BP of 130–139/80–89 mm Hg in our study was takes responsibility for the integrity of the data and the accuracy of the data analysis. 40 years, which is exactly the threshold used in the ASCVD risk estimator.7 Therefore, it seems reasonable that clinicians should Funding The University of Toronto Lupus Clinic is supported by the University Health Network, Lou & Marissa Rocca, Mark & Diana Bozzo and the Lupus not rely on CV risk calculators in SLE and commence treatment Foundation of Ontario. as soon as possible in cases of sustained BP elevation above the Disclaimer The sponsors had no role in study design, collection and interpretation threshold of 130/80 mm Hg. of the data or drafting the report. The optimal level of BP to maximise CV risk reduction is Competing interests None declared. not known. In the Systolic Blood Pressure Intervention Trial, it was shown that intensive treatment of high-risk,­ non-­diabetic Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. patients aiming at a SBP <120 mm Hg conferred a 25% reduc- tion in the rate of AVEs compared with individuals with an Patient consent for publication Not required. SBP target of less than 140 mm Hg.30 The average SBP of the Ethics approval The study is approved by the University Health Network Research intensive treatment patients was 121.4 mm Hg compared with Ethics Board. the 136.2 mm Hg of the standard treatment individuals. Would Provenance and peer review Not commissioned; externally peer reviewed. lower BP levels further reduce the incidence rates of AVEs? In Data availability statement All data relevant to the study are included in the a population-­wide study from Sweden in 187 106 patients with article or uploaded as supplementary information. All authors were involved in the type 2 diabetes and no prior AVE, it was shown that there was a study conception and design, acquisition of data, analysis and interpretation of data. linear reduction of the non-­fatal AVEs for gradually decreasing ORCID iDs 31 BP levels. The minimum CV risk (for non-­fatal AVEs) was Dafna D Gladman http://orcid.​ ​org/0000-​ ​0002-​9074-​0592 observed at the levels of 110–119 mm Hg (for SBP), while there Murray Urowitz http://orcid.​ ​org/0000-​ ​0001-​7506-​9166 was a significant increase at the levels of 120–129 mm Hg and a further significant increase at 130–139 mm Hg. The difference References in the HR for AVEs between the lowest SBP threshold that was 1 Whelton PK, Carey RM, Aronow WS, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/ assessed in that study (110–119 mm Hg) to the 130–139 mm Hg APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College level was 12% for all AVEs. However, these levels (110–119 mm of Cardiology/American heart association Task force on clinical practice guidelines. Hg) were also associated with a 20% higher risk for heart failure Circulation 2018;138:e484–594. and 28% higher risk for all-­cause mortality compared with the 2 James PA, Oparil S, Carter BL, et al. Evidence-­Based guideline for the management of 130–139 mm Hg levels.31 Based on these data, targeting lower high blood pressure in adults: report from the panel members appointed to the eighth joint National Committee. JAMA 2014;2014:507–20. levels of BP might be unsafe in certain patients with lupus (eg, 3 Tselios K, Koumaras C, Urowitz MB, et al. Do current arterial hypertension treatment with prior heart disease or heart failure). guidelines apply to systemic lupus erythematosus patients? A critical appraisal. Semin The purpose of the current study was not the assessment of Arthritis Rheum 2014;43:521–5. the effect of treatment on CV outcomes in patients with lupus. 4 Tselios K, Sheane BJ, Gladman DD, et al. Optimal monitoring for coronary heart Nevertheless, it was recently shown that the effective manage- disease risk in patients with systemic lupus erythematosus: a systematic review. J Rheumatol 2016;43:54–65. ment of the traditional atherosclerotic risk factors (including 5 Zöller B, Li X, Sundquist J, et al. Risk of subsequent coronary heart disease in patients hypertension) led to a decrease of 60% in the incidence of AVEs hospitalized for immune-­mediated diseases: a nationwide follow-­up study from from the 1980s to the 2010s.32 As such, it should be expected Sweden. PLoS One 2012;7:e33442. that the effective treatment of hypertension to levels lower than 6 manzi S, Meilahn EN, Rairie JE, et al. Age-­Specific incidence rates of myocardial infarction and angina in women with systemic lupus erythematosus: comparison with 130/80 mm Hg will further reduce the incidence of AVEs in SLE. the Framingham study. Am J Epidemiol 1997;145:408–15. The present report is an observational longitudinal study and 7 goff DC, Lloyd-­Jones DM, Bennett G, et al. 2013 ACC/AHA guideline on the not a randomised controlled trial. Therefore, its findings should assessment of cardiovascular risk: a report of the American College of Cardiology/ be interpreted with caution, particularly since the patients with American heart association Task force on practice guidelines. J Am Coll Cardiol a BP of 130–139/80–89 mm Hg also had a higher prevalence 2014;63:2935–59. 8 Urowitz MB, Ibañez D, Su J, et al. Modified Framingham risk factor score for systemic of other atherosclerotic risk factors. Moreover, certain risk lupus erythematosus. J Rheumatol 2016;43:875–9. factors (obesity and family history of premature CAD) were not 9 Hochberg MC. Updating the American College of rheumatology revised criteria for the included in the analysis due to the large proportion of missing classification of systemic lupus erythematosus. Arthritis Rheum 1997;40:40. data in our cohort. Nevertheless, CV risk reduction is multidi- 10 Therneau TM, Grambsch PM. Modeling survival data: extending the COX model (ISBN mensional and should not be restricted to the tight control of 9781441931610). New York: Springer, 2000. 11 nikpour M, Urowitz MB, Ibanez D, et al. Importance of cumulative exposure to hypertension alone but include all modifiable risk factors, such elevated cholesterol and blood pressure in development of atherosclerotic coronary as dyslipidaemia, diabetes. artery disease in systemic lupus erythematosus: a prospective proof-­of-­concept cohort study. Arthritis Res Ther 2011;13:R156. 12 dorans KS, Mills KT, Liu Y, et al. Trends in prevalence and control of hypertension Conclusion according to the 2017 American College of Cardiology/American heart association In conclusion, patients with lupus with an adjusted mean BP of (ACC/AHA) guideline. J Am Heart Assoc 2018;7:e008888. 13 Bengtsson C, Ohman M-­L, Nived O, et al. Cardiovascular event in systemic lupus 130–139/80–89 mm Hg over the first 2 years from enrolment erythematosus in northern Sweden: incidence and predictors in a 7-­year follow-­up to the clinic had significantly higher incidence rates of AVEs study. Lupus 2012;21:452–9.

616 Tselios K, et al. Ann Rheum Dis 2020;79:612–617. doi:10.1136/annrheumdis-2019-216764 Systemic lupus erythematosus

14 Bertoli AM, Vilá LM, Alarcón GS, et al. Factors associated with arterial vascular events 24 doria A, Shoenfeld Y, Wu R, et al. Risk factors for subclinical atherosclerosis in a in profile: a multiethnic lupus cohort. Lupus 2009;18:958–65. prospective cohort of patients with systemic lupus erythematosus. Ann Rheum Dis 15 Urowitz MB, Ibañez D, Gladman DD. Atherosclerotic vascular events in a single large 2003;62:1071–7. lupus cohort: prevalence and risk factors. J Rheumatol 2007;34:70–5. 25 mcMahon M, Skaggs BJ, Sahakian L, et al. High plasma leptin levels confer increased 16 Urowitz MB, Gladman DD. The SLICC inception cohort for atherosclerosis. Curr risk of atherosclerosis in women with systemic lupus erythematosus, and are Rheumatol Rep 2008;10:281–5. associated with inflammatory oxidised lipids. Ann Rheum Dis 2011;70:1619–24. 17 Urowitz MB, Gladman D, Ibañez D, et al. Atherosclerotic vascular events in a 26 Zhang C-­Y, Lu L-­J, Li F-­H, et al. Evaluation of risk factors that contribute to high multinational inception cohort of systemic lupus erythematosus. Arthritis Care Res prevalence of premature atherosclerosis in Chinese premenopausal systemic lupus 2010;62:881–7. erythematosus patients. J Clin Rheumatol 2009;15:111–6. 18 Wang X-­Y, Tang X-­Q, Huang Y-­J, et al. Frequency of established cardiovascular disease 27 Kiani AN, Post WS, Magder LS, et al. Predictors of progression in atherosclerosis over 2 and its risk factors in Chinese patients with systemic lupus erythematosus. Clin years in systemic lupus erythematosus. Rheumatology 2011;50:2071–9. Rheumatol 2012;31:669–75. 28 sella EMC, Sato EI, Barbieri A. Coronary artery angiography in systemic lupus 19 El-­Magadmi M, Bodill H, Ahmad Y, et al. Systemic lupus erythematosus: an erythematosus patients with abnormal myocardial perfusion scintigraphy. Arthritis independent risk factor for endothelial dysfunction in women. Circulation Rheum 2003;48:3168–75. 2004;110:399–404. 29 Bruce IN, Gladman DD, Ibañez D, et al. Single photon emission computed 20 Kiss E, Soltesz P, Der H, et al. Reduced flow-­mediated vasodilation as a marker for tomography dual isotope myocardial perfusion imaging in women with systemic cardiovascular complications in lupus patients. J Autoimmun 2006;27:211–7. lupus erythematosus. II. predictive factors for perfusion abnormalities. J Rheumatol 21 selzer F, Sutton-­Tyrrell K, Fitzgerald SG, et al. Comparison of risk factors for vascular 2003;30:288–91. disease in the carotid artery and aorta in women with systemic lupus erythematosus. 30 Wright JT, Williamson JD, Whelton PK, et al. A randomized trial of intensive versus Arthritis Rheum 2004;50:151–9. standard blood-­pressure control. N Engl J Med 2015;373:2103–16. 22 cypiene A, Kovaite M, Venalis A, et al. Arterial wall dysfunction in systemic lupus 31 adamsson Eryd S, Gudbjörnsdottir S, Manhem K, et al. Blood pressure and erythematosus. Lupus 2009;18:522–9. complications in individuals with type 2 diabetes and no previous cardiovascular 23 sacre K, Escoubet B, Pasquet B, et al. Increased arterial stiffness in systemic lupus disease: national population based cohort study. BMJ 2016;354:i4070. erythematosus (SLE) patients at low risk for cardiovascular disease: a cross-sectional­ 32 Urowitz MB, Su J, Gladman DD. Atherosclerotic vascular events in systemic lupus controlled study. PLoS One 2014;9:e94511. erythematosus: an evolving story. J Rheumatol 2020;47:66–71.

Tselios K, et al. Ann Rheum Dis 2020;79:612–617. doi:10.1136/annrheumdis-2019-216764 617 Systemic sclerosis

Clinical science Riociguat in patients with early diffuse cutaneous systemic sclerosis (RISE-­SSc): randomised, double-­ blind, placebo-­controlled multicentre trial Dinesh Khanna ‍ ‍ ,1 Yannick Allanore,2 Christopher P Denton ‍ ‍ ,3 Masataka Kuwana ‍ ‍ ,4 Marco Matucci-­Cerinic,5 Janet E Pope ‍ ‍ ,6 Tatsuya Atsumi,7 Radim Bečvář,8 László Czirják,9 Eric Hachulla,10 Tomonori Ishii,11 Osamu Ishikawa,12 Sindhu R Johnson ‍ ‍ ,13 Ellen De Langhe,14 Chiara Stagnaro,15 Valeria Riccieri,16 Elena Schiopu,17 Richard M Silver ‍ ‍ ,18 Vanessa Smith,19 Virginia Steen,20 Wendy Stevens,21 Gabriella Szücs,22 Marie-­Elise Truchetet,23 Melanie Wosnitza,24 Kaisa Laapas,25 Janethe de Oliveira Pena,26 Zhen Yao,27 Frank Kramer,24 Oliver Distler ‍ ‍ 28

Handling editor Josef S Abstract Key messages Smolen Objectives Riociguat is approved for pulmonary arterial hypertension and has antiproliferative, anti-­inflammatory ►► Additional material is What is already known about this subject? published online only. To view and antifibrotic effects in animal models of tissue ►► There is a need for new therapies for patients please visit the journal online fibrosis.We evaluated the efficacy and safety of riociguat with diffuse cutaneous systemic sclerosis (http://dx.​ ​doi.org/​ ​10.​1136/​ in patients with early diffuse cutaneous systemic sclerosis annrheumdis-2019-​ ​216823). (dcSSc). (dcSSc) at high risk of skin fibrosis progression. ►► The soluble guanylate cyclase stimulator For numbered affiliations see Methods In this randomised, double-­blind, placebo-­ riociguat has antiproliferative, anti-­ end of article. controlled, phase IIb trial, adults with dcSSc of inflammatory and antifibrotic effects in vitro <18 months’ duration and a modified Rodnan skin score and in animal models of tissue fibrosis and has Correspondence to (mRSS) 10–22 units received riociguat 0.5 mg to 2.5 mg Dr Dinesh Khanna, Division of been shown to increase digital blood flow in Rheumatology, University of orally three times daily (n=60) or placebo (n=61). The patients with Raynaud’s phenomenon. Michigan, Ann Arbor, MI 48109, primary endpoint was change in mRSS from baseline to USA; khannad@​ ​med.umich.​ ​edu week 52. What does this study add? and Dr Oliver Distler, Results At week 52, change from baseline in mRSS ► Department of Rheumatology, ► The RIociguat Safety and Efficacy in patients University Hospital Zurich, 8091 units was –2.09±5.66 (n=57) with riociguat and with diffuse cutaneous Systemic Sclerosis study Zurich, Switzerland; –0.77±8.24 (n=52) with placebo (difference of least failed to meet its primary endpoint of change oliver.​ ​distler@usz.​ ​ch squares means –2.34 (95% CI –4.99 to 0.30; p=0.08)). in modified Rodnan skin score after 52 weeks In patients with interstitial lung disease, forced vital at p=0.08. However, some secondary and For ’Presented at statement’ see end of article. capacity declined by 2.7% with riociguat and 7.6% exploratory endpoints showed potential efficacy with placebo. At week 14, average Raynaud’s condition signals that should be investigated in further Received 13 December 2019 score had improved ≥50% in 19 (41.3%)/46 patients trials. Riociguat was well tolerated, with no Revised 19 February 2020 with riociguat and 13 (26.0%)/50 patients with placebo. unexpected safety signals. Accepted 9 March 2020 Safety assessments showed no new signals with riociguat and no treatment-­related deaths. How might this impact on clinical practice or Conclusions Riociguat did not significantly benefit future developments? mRSS versus placebo at the predefined p<0.05. ►► Although the primary endpoint was not Secondary and exploratory analyses showed potential significant, this phase IIb study provides efficacy signals that should be tested in further trials. important information that can inform future Riociguat was well tolerated. study design and gave preliminary findings that could be explored in future trials in patients with dcSSc.

Introduction Systemic sclerosis (SSc) is an autoimmune connec- © Author(s) (or their is a significant unmet need, particularly in diffuse 3 employer(s)) 2020. Re-­use tive tissue disease characterised by fibrosis, inflam- cutaneous SSc (dcSSc). 1–3 permitted under CC BY. mation and microvascular injury. Systemic The soluble guanylate cyclase (sGC) stimulator Published by BMJ. organ manifestations include pulmonary arterial riociguat increases intracellular cyclic guanosine 7 To cite: Khanna D, hypertension (PAH), interstitial lung disease (ILD), monophosphate (cGMP). cGMP activates protein Allanore Y, Denton CP, Raynaud’s phenomenon (RP) and digital ulcers kinases G, which are important in the regulation et al. Ann Rheum Dis (DU).3 4 To date, nintedanib is the only approved of vascular tone and remodelling.8 Riociguat was 2020;79:618–625. therapy for the treatment of SSc-ILD.­ 5 6 Thus there approved for treatment of PAH following the phase

618 Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 Systemic sclerosis

III Pulmonary Arterial Hypertension Soluble Guanylate Cyclase-­ Outcomes and follow-up Stimulator Trial 1 (PATENT-1) study, which included a subgroup The primary endpoint was the change in mRSS from base- with PAH-­SSc, in which riociguat prevented the decline in 6 min line to week 52. To prevent interobserver variability, the same walking distance seen with placebo.9 In a single-dose­ pilot study, physician, experienced in skin scoring, scored the same patient riociguat increased digital blood flow in patients with RP.10 throughout the study. Skin fibrosis was also analysed by prespec- Riociguat has demonstrated antiproliferative, anti-inflammatory­ ified exploratory analyses of mRSS progression (increase by >5 and antifibrotic effects mediated by attenuation of transforming units and ≥25% from baseline) and regression (decrease by >5 growth factor beta-1 signalling in animal models and in vitro units and ≥25% from baseline). This definition was based on studies.7 8 11–14 sGC stimulators prevented and treated fibrosis in analyses suggesting that a reduction in mRSS of 3.2–5.3 units models of SSc.12 15 or 15%–25% from baseline is considered a minimally clini- We hypothesised that riociguat may benefit tissue fibrosis in cally important difference.22 23 In addition, descriptive analysis dcSSc. The RIociguat Safety and Efficacy in patients with diffuse in prespecified patient subgroups was performed (see online cutaneous Systemic Sclerosis (RISE-­SSc) trial compared riociguat supplementary file 1, p4). Secondary endpoints were tested with placebo in patients with early dcSSc.16–18 hierarchically in the order: American College of Rheumatology Composite Response Index for Systemic Sclerosis (ACR CRISS) at week 5224 (see online supplementary file 1, p5–6), Health Methods Assessment Questionnaire Disability Index, patient’s global Design overview 19 assessment, physician’s global assessment and change in FVC%. RISE-­SSc (​clinicaltrials.gov​ identifier: NCT02283762 ) was a An independent, blinded Adjudication Committee reviewed clin- randomised, double-­blind, placebo-­controlled, parallel-group,­ ical outcomes potentially representing systemic organ manifes- phase IIb, international, multicentre study, consisting of a tations of dcSSc (see online supplementary file 1, p6), and all screening phase (≤2 weeks), a 52-­week, double-­blind, main treat- causes of death. ment phase and a long-term­ extension (see online supplementary FVC% and DLCO% were assessed overall and (post hoc) in figure S1 and supplementary file 2). All patients provided written patients with ILD according to medical history and restrictive informed consent. Each site’s institutional review board or ethics lung disease (FVC% 50%–75% at baseline). committee approved the protocol. The study was performed in Effects on RP at week 14 versus day 0 and net digital ulcer accordance with the Declaration of Helsinki and Good Clinical burden were prespecified exploratory analyses. For details of Practice. other prespecified exploratory analyses and post hoc assessments see online supplementary file 1, p6. Study participants Investigators enrolled patients ≥18 years old, fulfilling Amer- Statistical analysis ican College of Rheumatology/European League Against Rheu- Assuming a standard deviation (SD) of 8 mRSS units,25 80% matism (ACR/EULAR) classification criteria for SSc,20 with power, a two-­sided significance level of 5% and 1:1 rando- dcSSc according to LeRoy and Medsger.21 Based on European misation, 128 patients would be required to detect a placebo-­ Scleroderma Trials and Research Group (EUSTAR) cohort adjusted difference of 4 units for intent-­to-­treat analysis of observations,16–18 entry criteria specified disease duration ≤18 mRSS. Endpoints were analysed using mixed model repeated months (defined as time from first non-­RP manifestation) and measures, with baseline mRSS as a covariate; treatment arm, modified Rodnan skin score (mRSS) 10–22 units to enrich the region and study visit, the interaction effect between study visit study with patients at risk of skin fibrosis progression. Other and treatment arm as fixed effects and patient-specific­ random inclusion criteria were per cent predicted forced vital capacity effects (see online supplementary file 3). The primary endpoint (FVC%) ≥45% and haemoglobin-­corrected per cent predicted was also analysed by analysis of covariance with baseline mRSS diffusing capacity of the lung for carbon monoxide (DLCO) as a covariate, and treatment arm and region as main effects. ≥40% at screening. Patients receiving concomitant nitrates, Endpoints present or not were estimated using Mantel-­Haenszel nitric oxide donors, phosphodiesterase inhibitors or recent SSc weights. Analyses were performed on all patients randomised therapies were excluded (see online supplementary file 1, p1–3). and treated with study medication using SAS V.9.2 software (SAS Institute Inc, Cary, North Carolina, USA). Since the primary endpoint was not met, all other p values are nominal, are only Randomisation and intervention shown for predefined but not post hoc analyses, cannot be Patients were randomised 1:1 to riociguat or matching placebo, considered statistically significant and are presented for infor- individually adjusted every 2 weeks from 0.5 mg to 2.5 mg orally mation only. three times daily over 10 weeks and continued throughout the treatment phase. From week 26, rescue therapy was permitted at investigator discretion (see online supplementary file 1, p4). Patient involvement Physical examination, disease status and demographics were Patients were not directly involved in the design, recruitment or obtained at day 0. Disease status was re-evaluated­ at weeks 12, conduct of the study. 26 and 52, with additional assessments of mRSS and pulmonary function at week 39. Raynaud’s condition score was assessed by Results a patient diary completed daily for seven consecutive days before Study population the first treatment dose and at week 14. Safety assessments In total, 121 patients were randomised (riociguat, n=60; placebo, included laboratory assessments at screening, on day 0, and at n=61). The study was completed according to the protocol. weeks 2, 4, 6, 8, 10, 26, 39 and 52, and evaluation of vital signs, Five patients in each group received ≥1 new rescue therapy adverse events (AEs) and serious adverse events (SAEs) coded after week 26. Study discontinuation occurred in 18 (30.0%) by Medical Directory for Regulatory Activities preferred terms, riociguat-­treated patients and 15 (24.6%) placebo-treated­ DU net burden and prior and concomitant therapy at every visit. patients (figure 1). At week 52, 34 (80.9%)/42 riociguat-treated­

Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 619 Systemic sclerosis

Figure 1 Patient disposition. patients were receiving riociguat 2 or 2.5 mg three times daily. versus 15.73 (SD 10.48) with placebo: difference of least squares Patients generally had early dcSSc, with mean mRSS 17 and (LS) means –2.34 (standard error (SE) 1.33); 95% confidence mean disease duration 8.6 months. Baseline characteristics were interval (CI) –4.99 to 0.30; relative difference –14%; p=0.0815. generally well balanced across groups (table 1). At week 52, the mean change from baseline in mRSS was –2.09 (SD 5.66) with riociguat and –0.77 (SD 8.24) with placebo Skin fibrosis (figure 2A). Progression of mRSS (increase by >5 units and The primary endpoint was not met at the predefined p<0.05. ≥25% from baseline) was observed in 11 (18.6%)/59 patients At week 52, mean mRSS was 14.63 (SD 6.56) with riociguat with riociguat and 22 (36.7%)/60 patients with placebo (Mantel-­ Haenszel estimate of difference: –17.99% (95% CI –33.57% to –2.40%; nominal p=0.0237); figure 2B). Regression rates Table 1 Baseline characteristics of study participants (decrease by >5 units and ≥25% from baseline) in the riociguat Overall Riociguat Placebo and placebo groups were 27 (45.7%)/59 and 18 (30.0%)/60, Characteristics (n=121) (n=60) (n=61) respectively (Mantel-­Haenszel estimate of difference: 15.29% Mean age (SD), years 50.7 (12.2) 51.9 (11.5) 49.5 (12.9) (95% CI –1.98% to 32.57%; nominal p=0.0827)). Female, n (%) 92 (76.0) 47 (78.3) 45 (73.8) On subgroup analyses, the change in mRSS with riociguat White, n (%) 89 (73.6) 43 (71.7) 46 (75.4) versus placebo showed a nominal p value <0.05 for mRSS Black, n (%) 5 (4.1) 2 (3.3) 3 (4.9) 17–22, anti-­RNA polymerase III positive/SCl-70 negative, base- Asian, n (%) 24 (19.8) 12 (20.0) 12 (19.7) line FVC 50%–75% and high-­sensitivity C-reactive­ protein Native Hawaiian or other Pacific Islander, n (%) 1 (0.8) 1 (1.7) 0 >3.0 mg/L (see online supplementary figure S2). Not reported, n (%) 2 (1.7) 2 (3.3) 0 Mean disease duration (SD), months (from first 9.0 (6.4) 9.5 (7.0) 8.6 (5.8) non-RP­ manifestation) Secondary endpoints Mean mRSS (SD), units 16.8 (3.7) 16.9 (3.4) 16.7 (4.1) ACR CRISS as a measure of improvement did not show signif- Mean % predicted FVC (SD), % 92.8 (17.8) 90.7 (18.5) 94.8 (17.0) icant differences in this trial designed for prevention of wors-

Mean % predicted DLCO (Hb corr.) (SD), % 76.4 (18.5) 76.0 (19.9) 76.8 (17.2) ening. Eighteen per cent of patients in each group had a CRISS Swollen joint count ≥1, n (%) 38 (31.4) 23 (38.3) 15 (24.6) improvement probability score ≥0.60 (estimate of difference: Mean swollen joint count (SD), n 2.0 (4.7) 3.0 (6.1) 1.1 (2.5) 0.20% (95% CI –13.68% to 14.09%; nominal p=0.977)). Tender joint count ≥1, n (%) 51 (42.1) 30 (50.0) 21 (34.4) However, in step 1 of the CRISS analysis, 1 (1.7%) patient in Mean tender joint count (SD), n 3.0 (6.2) 3.9 (7.3) 2.1 (4.8) the riociguat group versus 4 (6.6%) in the placebo group met the Digital ulcer count ≥1, n (%) 15 (12.4) 9 (15.0) 6 (9.8) definition for SSc-­related organ involvement. Other secondary Mean digital ulcer count (SD), n 0.3 (1.1) 0.3 (0.7) 0.4 (1.4) endpoints are shown in table 2. Mean digital ulcer count in patients with ulcers 2.5 (2.3) 1.7 (1.0) 3.7 (3.2) (SD), n Tendon friction rubs ≥1, n (%) 35 (28.9) 15 (25.0) 20 (32.8) Lung function Mean tendon friction rubs (SD), n 3.1 (2.2) 2.4 (1.1) 3.6 (2.7) Overall, the change in FVC% between baseline and week 52 ILD by medical history, n (%) 25 (20.7) 12 (20.0) 13 (21.3) was −2.38% (SD 7.52) with riociguat and −2.95% (SD 9.73) Mean HAQ-­DI (SD), units 0.79 (0.68) 0.89 (0.67) 0.69 (0.69) with placebo (difference of LS means −0.20 (SE 1.61); 95% CI Anti-­RNA polymerase III positive, n (%) 26 (21.5) 10 (16.7) 16 (26.2) −3.40 to 3.00; nominal p=0.901; figure 3A). Two patients in Anti-­SCl-70 (anti-­topoisomerase I) positive, n (%) 49 (40.5) 26 (43.3) 23 (37.7) each group developed new ILD. At baseline, 12 (20.0%) patients Anti-­centromere B positive, n (%) 10 (8.3) 4 (6.7) 6 (9.8) receiving riociguat and 13 (21.3%) patients with placebo had

DLCO, diffusing capacity of the lung for carbon monoxide; DLCO (Hb corr.), diffusing capacity of the SSc-ILD­ by medical history, and 11 (18.3%) and 7 (11.5%), lung for CO, corrected for haemoglobin; FVC, forced vital capacity; HAQ-DI,­ Health Assessment respectively, had baseline FVC% 50%–75%. Baseline characteris- Questionnaire Disability Index; ILD, interstitial lung disease; mRSS, modified Rodnan skin score; RP, Raynaud’s phenomenon. tics by lung fibrosis diagnosis are shown in online supplementary

620 Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 Systemic sclerosis

Figure 2 (A) Change from baseline in mRSS during the study. Mixed model with repeated measurement was applied with baseline value, treatment group, region, visit and treatment by visit interaction as fixed effects, and subject as a random effect. Vertical lines represent 95% CI for change. (B) Proportion of patients with mRSS progression (increase in mRSS by >5 units and ≥25% from baseline: prespecified analysis). Treatment comparison (riociguat −placebo): estimate −17.99%, 95% CI −33.57 to −2.40. Mantel-Haenszel­ estimate of difference: nominal p=0.0237. CI, confidence interval; LS, least squares; mRSS, modified Rodnan skin score. table S1. Depending on the diagnosis, the mean change in FVC% in 2 (3.3%) patients in the riociguat group and 6 (9.8%) in from baseline to week 52 was −7.6 to −8.7% with placebo and the placebo group at week 14, and in 5 (8.3%) patients and +0.7 to −2.7% with riociguat (figure 3B). 12 (19.7%) patients, respectively, at week 52. There were 4

DLCO% decreased by −2.31% (SD 10.08) with riociguat and 26 new DUs with riociguat and placebo, respectively, at and −4.09% (SD 12.19) with placebo (difference of LS means week 14; and 12 and 72 new DUs, respectively, at week 52 2.01 (SE 2.14); 95% CI −2.24 to 6.25; nominal p=0.3502). In (see online supplementary figure S3). Concomitant medica- patients with ILD by medical history the changes in DLCO% were tion with an indication for DU was used by 7 (11.7%) patients –4.55 (SD 8.12) with riociguat (n=11) and –7.63 (SD 13.37) receiving riociguat and 10 (16.4%) patients with placebo. with placebo (n=12). In those with baseline FVC% 50%–75%, Changes from baseline to week 14 in Raynaud’s attack dura-

DLCO% increased by 2.26 (SD 15.16) with riociguat (n=8) and tion, frequency and symptoms favoured riociguat but nomi- fell by –7.32 (SD 17.24) with placebo (n=5). nally did not differ significantly between riociguat and placebo (see online supplementary table S2). The average Raynaud’s Raynaud’s phenomenon and digital ulcers condition score improved by ≥50% in 19 (41.3%)/46 patients At baseline, 9 (15.0%) patients had DUs in the riociguat group with riociguat and in 13 (26.0%)/50 patients with placebo. At versus 6 (9.8%) in the placebo group. New DUs were reported week 52, reductions in net DU burden were –0.09 (SD 0.50)

Table 2 Difference between riociguat group and placebo group in change from baseline to week 52 in secondary endpoints Estimate of difference Endpoint Riociguat (n=60) Placebo (n=61) (95% CI) Nominal p value* ACR CRISS No improvement, n (%) 1 (1.7) 4 (6.6) ≥3 missing criteria, n (%) 6 (10.0) 7 (11.5) 0.20% (–13.68 to 14.09)† 0.977 CRISS probability ≥60%, n (%) 11 (18.3) 11 (18.0) CRISS probability <60%, n (%) 49 (81.7) 50 (82.0) Mean HAQ-­DI (SD), units Baseline 0.89 (0.67) 0.69 (0.69) –0.07 (–0.23 to 0.08)‡ 0.3529 Change at week 52 0.05 (0.38) (n=56) 0.13 (0.42) (n=52) Mean patient global assessment (SD), units Baseline 3.93 (2.50) 3.77 (2.34) 0.79 (–0.12 to 1.69)‡ 0.0887 Change at week 52 0.69 (2.75) (n=45) –0.02 (2.23) (n=46) Mean physician global assessment (SD), units Baseline 4.33 (2.11) 4.02 (2.00) 0.83 (0.11 to 1.54)‡ 0.0241 Change at week 52 –0.07 (2.16) (n=45) –0.75 (2.09) (n=47) Mean % predicted FVC (SD), % Baseline 90.74 (18.52) 94.82 (17.03) –0.20 (–3.40 to 3.00)‡ 0.901 Change at week 52 –2.38 (7.52) (n=55) –2.95 (9.73) (n=51) *Since the primary endpoint was not met, all other p values cannot be considered statistically significant and are presented for information only. †Mantel-Haenszel­ estimate. ‡Mixed model repeated measures applied with baseline value, treatment group, region, visit and treatment by visit interaction as fixed effects, and subject as a random effect. ACR, American College of Rheumatology; CI, confidence interval; CRISS, Composite Response Index for Systemic Sclerosis; FVC, forced vital capacity; HAQ-DI,­ Health Assessment Questionnaire Disability Index.

Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 621 Systemic sclerosis

Figure 3 (A) Change in FVC% from baseline to week 52 in overall population. (B) Change in FVC% from baseline to week 52 in patients with lung fibrosis at baseline by diagnostic subgroups (post hoc). Data points are mean (SE). Numbers close to axes are numbers of patients with data at week 52. CI, confidence interval; FVC, forced vital capacity; ILD, interstitial lung disease; LS, least squares; SE, standard error. and –0.08 (SD 1.47) with riociguat and placebo, respectively Discussion (difference of LS means –0.11 (SE 0.14); 95% CI –0.38 to RISE-­SSc investigated the effects of riociguat on disease progres- 0.17; nominal p=0.4444). No case of critical digital ischaemia sion in patients with early dcSSc. mRSS was selected as the primary occurred in either group. endpoint as it correlates with biopsy measures of skin thickness and reflects disease prognosis and visceral involvement.1 26 mRSS does, however, have challenging and unpredictable changes over Other endpoints the disease course and attempts to enrich trial populations with Findings from prespecified exploratory analyses and post hoc patients likely to progress have not been successful. Neverthe- assessments are provided in online supplementary file 1, p12–19. less, it is a validated surrogate marker of disease progression27 and is accepted by authorities as an endpoint for skin fibrosis.22 Adverse events RISE-­SSc was the first trial in SSc with the EUSTAR inclusion Overall, 58 (96.7%) patients in the riociguat group and 55 criteria designed to enrich the population with patients likely to (90.2%) in the placebo group experienced an AE (see online show progression of skin fibrosis. Between baseline and week 52, 36.7% of placebo-treated­ patients showed skin fibrosis progres- supplementary table S8). Most AEs in the riociguat group 25 28–30 were mild to moderate, and most were gastrointestinal events sion, which is much higher than in similar trials, showing (eg, gastro-­oesophageal reflux disease, diarrhoea or nausea) or that our enrichment strategy was successful. This is consistent nervous system disorders (eg, dizziness, headache). Symptomatic with other evidence that patients with baseline mRSS 15–22 and hypotension was reported in 7 (11.7%) patients with riociguat early disease showed higher progression rates than unselected 17 18 31 and 6 (9.8%) patients with placebo. SAEs were reported in 9 cohorts. (15.0%) patients in the riociguat group and 15 (24.6%) in the There are several potential reasons why the primary placebo group (table 3). Eleven patients in each group had AEs endpoint was not met in this study. First, RISE-SSc­ was resulting in discontinuation of study drug (see online supplemen- designed to detect a placebo-­adjusted change of mRSS between tary table S9). No events of serious haemoptysis were reported. riociguat and placebo with 80% power. For the low baseline One patient in the riociguat group died from myocardial infarc- mRSS expected in this study, a 4-unit­ change would repre- tion 117 days after the last administration of riociguat and one sent a change of 23%. The between-groups­ difference was 2.3, patient in the placebo group died from left ventricular failure which was less than expected. This low treatment effect was 157 days after the last administration of placebo. Neither death probably the main reason why the primary endpoint was not was considered related to study drug. met. In addition, the higher than expected numbers of skin 18 Of those with ILD by medical history, AEs were reported in 10 fibrosis regressors and stable patients reduced the sensitivity (83.3%)/12 patients with riociguat and 12 (92.3%)/13 patients of RISE-­SSc to detect a significant change of mRSS. This is with placebo. AEs reported more frequently with riociguat than consistent with previous trials, in which mRSS improvements 32 33 with placebo were predominantly dizziness and gastrointestinal were observed in the majority of patients receiving placebo. events (see online supplementary table S10). The incidence Other possible explanations include the very large variation in of respiratory, thoracic and mediastinal AEs was similar with mRSS scores during the study. riociguat (4 patients; 33.3%) and placebo (4 patients; 30.8%). As expected, the combined secondary endpoint did not favour SAEs were reported in 1 (8.3%)/12 and 3 (23.1%)/13 patients, riociguat because the ACR CRISS evaluates disease improve- respectively. Safety in patients with baseline FVC% 50%–75% ment, whereas RISE-SSc­ was designed to detect prevention of showed no overall excess of AEs with riociguat (see online progression. ACR CRISS is not expected to be positive in such supplementary table S11). a trial design.24

622 Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 Systemic sclerosis

was, therefore, well tolerated in early dcSSc, particularly when Table 3 Serious adverse events compared with traditional immunosuppressive agents.34 35 Toler- Patients reporting event, n (%) ability was also good in patients with ILD, which is important Event Riociguat (n=60) Placebo (n=61) given the increased rates of death and SAEs with riociguat in a Any SAE 9 (15.0) 15 (24.6) study in patients with pulmonary hypertension associated with 36 Any study drug-­related SAE 0 2 (3.3) idiopathic interstitial pneumonia. Discontinuation of study drug due to SAE 2 (3.3) 7 (11.5) Discontinuation rates (≈30% with riociguat and ≈25% with placebo) were higher in RISE-SSc­ than with active treat- Angina pectoris 1 (1.7) 1 (1.6) ment in recent trials of abatacept (23%)37 or tocilizumab (9%)38 Atrial fibrillation 1 (1.7) 0 in SSc. RISE-SSc­ recruited patients with very early disease Abdominal pain 1 (1.7) 0 (compared with these trials, which recruited patients with Intestinal pseudo-­obstruction 1 (1.7) 0 ≤36 and≤60 months from onset of SSc, respectively). The early Inflammation 1 (1.7) 0 discontinuation may be related to the expectation of worsening Lung infection 1 (1.7) 0 of SSc in early disease (based on natural history), where AEs Pneumonia 1 (1.7) 2 (3.3) may lead the investigator to withdraw the patient (see online RP 1 (1.7) 1 (1.6) supplementary table S9), especially in a placebo-controlled­ trial. Musculoskeletal discomfort 1 (1.7) 0 Indeed, another trial with a comparable very early disease popu- Pain in extremity 1 (1.7) 0 lation showed a discontinuation rate of 40% in the active treat- Dyspnoea 1 (1.7) 0 ment (CAT-192) group.39 Another explanation might be anxiety Intraductal proliferative breast lesion 1 (1.7) 0 associated with early disease in the participants; however, these Pericarditis 0 2 (3.3) are speculations and should be explored in other trials in patients Left ventricular failure 0 1 (1.6) with very early disease. AEs in the riociguat and placebo groups Ventricular tachycardia 0 1 (1.6) contributed substantially to the discontinuations in the current Gastric haemorrhage 0 1 (1.6) study. Gastro-­oesophageal reflux disease 0 1 (1.6) In conclusion, RISE-­SSc failed to meet its primary endpoint and is therefore a negative trial. However, it provides important Nausea 0 1 (1.6) findings for the identification of patients at high risk of skin Vomiting 0 1 (1.6) fibrosis progression that could inform future studies in patients Infected skin ulcer 0 1 (1.6) with dcSSc. In addition, there are potential efficacy signals in Anaemia 0 1 (1.6) early dcSSc and these may be explored further with additional Exposure during pregnancy 0 1 (1.6) randomised controlled trials. Osteolysis 0 1 (1.6) Scleroderma 0 1 (1.6) Author affiliations Acute myeloid leukaemia 0 1 (1.6) 1Division of Rheumatology, University of Michigan, Ann Arbor, Michigan, USA 2 Gastric adenocarcinoma 0 1 (1.6) Rheumatology A department, Cochin Hospital, APHP, Paris Descartes University, Paris, France Ovarian cancer 0 1 (1.6) 3Division of Medicine, Centre for Rheumatology, University College London, London, Cerebellar infarction 0 1 (1.6) UK 4 Syncope 0 1 (1.6) Department of Allergy and Rheumatology, Nippon Medical School Graduate School of Medicine, Tokyo, Japan Scleroderma renal crisis 0 1 (1.6) 5Department of Experimental and Clinical Medicine, University of Florence, Firenze, Acute pulmonary oedema 0 1 (1.6) Italy 6 Skin ulcer 0 1 (1.6) Schulich School of Medicine, Division of Rheumatology, The University of Western Surgical/medical prophylaxis 0 1 (1.6) Ontario, London, Ontario, Canada 7Department of Rheumatology, Endocrinology and Nephrology, Faculty of Medicine MedDRA preferred terms are shown. and Graduate School of Medicine, Hokkaido University, Sapporo, Japan MedDRA, Medical Directory for Regulatory Activities; RP, Raynaud’s phenomenon; 8Institute of Rheumatology, Department of Rheumatology, 1st Faculty of Medicine, SAE, serious adverse event. Charles University, Prague, Czech Republic 9Department of Rheumatology and Immunology, University of Pécs, Pécs, Hungary 10Department of Internal Medicine and Clinical Immunology, Claude Huriez Hospital, Some measures of mRSS, lung function in patients with Lille University School of Medicine, Lille, France 11Clinical Research, Innovation and Education Center, Tohoku University Hospital, evidence for pre-­existing ILD and the prevention of new DU Sendai, Japan and RP symptoms suggest potential signals for efficacy. It is 12Department of Dermatology, Gunma University Postgraduate School of Medicine, important to note that the descriptive analyses of predefined Maebashi, Japan 13 secondary and exploratory endpoints should not be interpreted Division of Rheumatology, Department of Medicine, Toronto Western Hospital, University Health Network, Mount Sinai Hospital, University of Toronto, Toronto as efficacy of riociguat, but as a potential signal that can be inves- Scleroderma Research Program, Toronto, Ontario, Canada tigated in further studies. 14Laboratory of Tissue Homeostasis and Disease, Skeletal Biology and Engineering AEs reported more frequently with riociguat than placebo were Research Center, Department of Development and Regeneration, KU Leuven, Leuven, Belgium mainly gastrointestinal events, dizziness or peripheral oedema. 15 These events are consistent with the effects of riociguat, such as Rheumatology Unit, Department of Clinical and Experimental Medicine, University of Pisa, Pisa, Italy relaxation of smooth muscle cells in the vasculature (often asso- 16Department of Clinical Medicine and Therapy, University of Rome La Sapienza, ciated with blood pressure decrease) or the gastrointestinal tract Rome, Italy and did not increase the incidence of withdrawal due to AEs. 17Division of Rheumatology, Department of Internal Medicine, Michigan Medicine SAEs were less common with riociguat than with placebo, no University Hospitals, Ann Arbor, Michigan, USA 18Division of Rheumatology and Immunology, Medical University of South Carolina, riociguat-­treated patient experienced an SAE considered related Charleston, South Carolina, USA to study treatment, and fewer discontinued study medication 19Department of Rheumatology and Internal Medicine, Ghent University Hospital, because of an SAE with riociguat than with placebo. Riociguat Ghent, Belgium

Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 623 Systemic sclerosis

20Division of Rheumatology, Georgetown University Medical Center, Washington, DC, and personal fees from GlaxoSmithKline, outside the submitted work. MK reports USA grants and personal fees from Actelion, personal fees from Bayer AG, personal 21Department of Rheumatology, St. Vincent’s Hospital Melbourne, Melbourne, fees from Chugai, personal fees from BI, personal fees from Reata, personal fees Victoria, Australia from Corpus, personal fees from CSL Behring, personal fees from GlaxoSmithKline, 22Division of Rheumatology, Department of Internal Medicine, University of personal fees from Mochida, personal fees from Pfizer, personal fees from Nippon Debrecen, Debrecen, Hungary Shinyaku, outside the submitted work. MMC reports grants and personal fees from 23Department of Rheumatology, CHU Bordeaux, Bordeaux, France Actelion, grants from Chemomab, grants and personal fees from BMS, grants from 24Research & Development, Bayer AG, Wuppertal, Germany Pfizer, grants and personal fees from MSD, grants from Sanipedia, personal fees 25StatFinn Oy, Espoo, Finland from Janssen, outside the submitted work. JEP reports personal fees from Actelion, 26Bayer HealthCare Pharmaceuticals Inc, Whippany, New Jersey, USA personal fees from Bayer AG, personal fees from BMS, personal fees from AbbVie, 27Bayer Healthcare, Beijing, China personal fees from Lilly, personal fees from Roche, personal fees from Sanofi, 28Department of Rheumatology, University Hospital, Zurich, Switzerland personal fees from Merck, personal fees from Novartis, personal fees from Pfizer, personal fees from Sandoz, personal fees from UCB, personal fees from Amgen, Presented at outside the submitted work. TA reports grants and personal fees from Astellas, grants Preliminary results from this study have been presented at scientific congresses and personal fees from Takeda, grants and personal fees from Mitsubishi Tanabe, as described below. Khanna D, et al. RISE-­SSc: a double-­blind, randomised study grants and personal fees from Chugai, grants and personal fees from Daiichi-­Sankyo, evaluating the efficacy and safety of riociguat for the treatment of patients with grants from Otuska, grants from Takeda, grants and personal fees from Pfizer, diffuse cutaneous systemic sclerosis. European League Against Rheumatism (EULAR) grants from Alexion, grants from Bayer, grants from Eisai, grants from Bristol-­Myers Congress, Madrid, Spain, 14−17 June 2017. Ann Rheumat Dis 2017;76(Suppl. Squibb, grants from Asahi Kasei, personal fees from Ono, personal fees from Sanofi, 2): Abstract no. AB0627. Distler O, et al. Riociguat in patients with early diffuse personal fees from Eli Lilly, outside the submitted work. LC reports personal fees cutaneous systemic sclerosis: a randomized, double-­blind, placebo-­controlled from Actelion, personal fees from Bayer AG, personal fees from BI, personal fees phase IIb study (RISE-­SSc). Arthritis Rheumatol 2018;70(Suppl. 9): Abstract no. from Roche-­Genentech, personal fees from Lilly, personal fees from Medac, personal 903. Khanna D, et al. The effects of riociguat on Raynaud’s phenomenon and fees from Novartis, personal fees from Pfizer, outside the submitted work.E H reports digital ulcers in patients with diffuse systemic sclerosis: results from the phase IIb grants, personal fees and non-­financial support fromA ctelion, grants, personal fees RISE-­SSc study. Arthritis Rheumatol 2018;70(Suppl. 10): Abstract no. 1879. Khanna and non-­financial support from BayerA G, grants, personal fees and non-­financial D, et al. The effects of riociguat on Raynaud’s phenomenon and digital ulcers in support from GlaxoSmithKline, grants, personal fees and non-­financial support patients with diffuse systemic sclerosis: results from the phase IIb RISE-­SSc study. from MSD, grants and personal fees from Pfizer, outside the submitted work. TI European League Against Rheumatism (EULAR) Congress, Madrid, Spain, 12−15 reports personal fees from Astellas, personal fees from Daiichi-­Sankyo, personal fees June 2019. Ann Rheumat Dis 2019;78(Suppl. 2): Abstract no. FRI0303. Distler O, et from Chugai, personal fees from Sanofi, personal fees from AbbVie, personal fees al. Efficacy and safety of riociguat in patients with early diffuse cutaneous systemic from Bristol-­Myers, personal fees from Mitsubishi Tanabe, personal fees from Eisai, sclerosis and interstitial lung disease (SSc-­ILD): results from the phase IIb RISE-­SSc personal fees from Janssen, personal fees from Asahi Kasei, personal fees from Ono study. European League Against Rheumatism (EULAR) Congress, Madrid, Spain, Pharmaceutical, personal fees from Ayumi Pharmaceutical, personal fees from Pfizer, 12−15 June 2019. Ann Rheumat Dis 2019;78(Suppl. 2):167.1–167. Distler O, et outside the submitted work. Virginia Steen reports participation in advisory boards, al. Efficacy and safety of riociguat in patients with early diffuse cutaneous systemic consultancy for economics of scleroderma management, and clinical trials for Bayer; sclerosis and interstitial lung disease (SSc-­ILD): results from the phase IIb RISE-­SSc investigator-­initiated grant, advisory board, and steering committee (consulting) study. Am J Respir Crit Care Med 2019;199:A4086. Hemmrich M, et al. Efficacy for CSL Behring; advisory board and site primary investigator (PI) of clinical trial for and safety of riociguat in patients with early diffuse cutaneous systemic sclerosis: Roche; site PI of clinical trial for Sanofi; site IP of clinical trial for Immune Tolerance results from the RISE-­SSc study, a randomised, double-­blind, placebo-­controlled Network; and DSMB for open labs phase 2 trial and site PI for the phase 3 trial for phase IIb study. 9th International Conference on cGMP: Generators, Effectors and Corbus. ME-­T reports consulting fees, speaking fees and honoraria from AbbVie, Therapeutic Implications, Mainz, Germany, 14−16 June 2019. J Translational Med BMS, Lilly, Medac, MSD, Pfizer, Roche, UCB, outside the submitted work. MW reports 2019;17:S1–03. employment from Bayer AG, outside the submitted work. KL reports other from Acknowledgements We thank the investigators and patients in the RISE-­SSc trial, StatFinn Oy, outside the submitted work. JdOP reports other from Bayer AG, outside as well as the Data Monitoring Committee, and the Central Adjudication Committee. the submitted work. ZY reports other from Bayer HealthCare, outside the submitted work. FK reports other from Bayer AG, outside the submitted work. Contributors DK and OD attest to the accuracy and completeness of the reported data, had full access to all data, wrote the report and made the final decision to Patient and public involvement Patients and/or the public were not involved in submit the manuscript for publication. DK, OD, YA, CPD, MK, MMC, TA, RB, LC, the design, or conduct, or reporting, or dissemination plans of this research. EH, TI, OI, SRJ, EDL, MM-­C, VR, ES, RMS, VSm, VSt, WS, GS, and MET recruited and Patient consent for publication Not required. monitored patients and collected study data. All authors revised the report and Provenance and peer review Not commissioned; externally peer reviewed. approved the final draft for publication. Data availability statement Availability of the data underlying this publication Funding The study was jointly funded by Bayer AG and Merck Sharp & Dohme will be determined according to Bayer’s commitment to the European Federation Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA. Bayer AG designed of Pharmaceutical Industries and Associations and Pharmaceutical Research and the study in collaboration with the authors. DK and OD have access to the data. KL Manufacturers of America principles for responsible clinical trial data sharing, worked with Bayer AG as an external statistician. MW analysed and interpreted the pertaining to scope, time point and process of data access. Bayer commits to sharing results and led post hoc analysis generation. Medical writing services provided by upon request from qualified scientific and medical researchers patient-­level clinical Richard Murphy PhD, of Adelphi Communications Ltd, Macclesfield, UK were funded trial data, study-­level clinical trial data, and protocols from clinical trials in patients by Bayer AG, Berlin, Germany in accordance with Good Publication Practice (GPP3) for medicines and indications approved in the USA and European Union as necessary guidelines. for doing legitimate research. This commitment applies to data on new medicines Competing interests DK reports personal fees from Actelion, grants and personal and indications that have been approved by the European Union and US regulatory fees from Bayer AG, grants and personal fees from BMS, grants from Pfizer, personal agencies on or after Jan 1, 2014. Interested researchers can use www.​clin​ical​stud​ fees from Arena, personal fees from Eclos Sciences, Inc, personal fees from BI, ydat​arequest.​com to request access to anonymised patient-­level data and supporting personal fees from Arena, personal fees from CSL Behring, personal fees from GSK, documents from clinical studies to do further research that can help advance medical personal fees from Galapagos, personal fees from Genentech/Roche, personal fees science or improve patient care. Information on the Bayer criteria for listing studies from Corbus, personal fees from Cytori, grants from Horizon, outside the submitted and other relevant information is provided in the study sponsors section of the work. OD reports other from Actelion, other from Bayer, other from Boehringer portal. Data access will be granted to anonymised patient-­level data, protocols, and Ingelheim, other from Mitsubishi Tanabe, other from AnaMar, other from ChemonAb, clinical study reports after approval by an independent scientific review panel. Bayer other from espeRare Foundation, other from Genentech/Roche, other from GSK, is not involved in the decisions made by the independent review panel. Bayer will other from Inventiva, other from Italfarmaco, other from iQvia, other from Lilly, other take all necessary measures to ensure that patient privacy is safeguarded. from Medac, other from MedImmune, other from Pharmacyclics, other from Novartis, Open access This is an open access article distributed in accordance with the other from Pfizer, other from Sanofi, other from Serodapharm, other from UCB, other Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits from Amgen, other from AbbVie, other from Mepha, other from MSD, outside the others to copy, redistribute, remix, transform and build upon this work for any submitted work. YA reports personal fees from Actelion, personal fees from Bayer purpose, provided the original work is properly cited, a link to the licence is given, AG, grants and personal fees from BMS, grants from Inventiva, personal fees from and indication of whether changes were made. See: https://​creativecommons.​org/​ BI, grants from Roche, grants from Sanofi, outside the submitted work. CPD reports licenses/by/​ ​4.0/.​ personal fees from Actelion, personal fees from Bayer AG, grants and personal fees from Inventiva, personal fees from BI, personal fees from Roche-­Genentech, ORCID iDs personal fees from SanofiA ventis, grants and personal fees from CSL Behring, grants Dinesh Khanna http://orcid.​ ​org/0000-​ ​0003-​1412-​4453

624 Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 Systemic sclerosis

Christopher P Denton http://​orcid.org/​ ​0000-0003-​ ​3975-8938​ 21 leRoy EC, Medsger TA. Criteria for the classification of early systemic sclerosis. J Masataka Kuwana http://orcid.​ ​org/0000-​ ​0001-​8352-​6136 Rheumatol 2001;28:1573–6. Janet E Pope http://orcid.​ ​org/0000-​ ​0003-​1479-​5302 22 Khanna D, Furst DE, Hays RD, et al. Minimally important difference in diffuse systemic Sindhu R Johnson http://orcid.​ ​org/0000-​ ​0003-​0591-​2976 sclerosis: results from the D-­penicillamine study. Ann Rheum Dis 2006;65:1325–9. Richard M Silver http://orcid.​ ​org/0000-​ ​0002-​2038-​3278 23 Khanna D, Clements PJ, Volkmann ER, et al. Minimal clinically important differences Oliver Distler http://orcid.​ ​org/0000-​ ​0002-​0546-​8310 for the modified Rodnan skin score: results from the scleroderma lung studies (SLS-­I and SLS-­II). Arthritis Res Ther 2019;21:23. References 24 Khanna D, Berrocal VJ, Giannini EH, et al. The American College of Rheumatology provisional composite response index for clinical trials in early diffuse cutaneous 1 Khanna D, Furst DE, Clements PJ, et al. Standardization of the modified Rodnan systemic sclerosis. Arthritis Rheumatol 2016;68:299–311. skin score for use in clinical trials of systemic sclerosis. J Scleroderma Relat Disord 25 Khanna D, Clements PJ, Furst DE, et al. Recombinant human relaxin in the treatment 2017;2:11–18. of systemic sclerosis with diffuse cutaneous involvement: a randomized, double-blind,­ 2 Varga J, Trojanowska M, Kuwana M. Pathogenesis of systemic sclerosis: recent insights placebo-controlled­ trial. Arthritis Rheum 2009;60:1102–11. of molecular and cellular mechanisms and therapeutic opportunities. J Scleroderma 26 Khanna D, Merkel PA. Outcome measures in systemic sclerosis: an update on Relat Disord 2017;2:137–52. instruments and current research. Curr Rheumatol Rep 2007;9:151–7. 3 Denton CP, Khanna D. Systemic sclerosis. Lancet 2017;390:1685–99. 27 Wu W, Jordan S, Graf N, et al. Progressive skin fibrosis is associated with a decline in 4 Jaeger VK, Wirz EG, Allanore Y, et al. Incidences and risk factors of organ lung function and worse survival in patients with diffuse cutaneous systemic sclerosis manifestations in the early course of systemic sclerosis: a longitudinal EUSTAR study. in the European Scleroderma Trials and Research (EUSTAR) cohort. Ann Rheum Dis PLoS One 2016;11:e0163894. 2019;78:648–56. 5 Distler O, Highland KB, Gahlemann M, et al. Nintedanib for systemic sclerosis-­ 28 amjadi S, Maranian P, Furst DE, et al. Course of the modified Rodnan skin thickness associated interstitial lung disease. N Engl J Med 2019;380:2518–28. score in systemic sclerosis clinical trials: analysis of three large multicenter, double-­ 6 US Food and Drug Administration. FDA approves first treatment for patients with blind, randomized controlled trials. Arthritis Rheum 2009;60:2490–8. rare type of lung disease. Available: https://www.​fda.gov/​ ​news-events/​ ​press-​ 29 Clements PJ, Furst DE, Wong WK, et al. High-­dose versus low-­dose D-­penicillamine announcements/fda-​ ​approves-first-​ ​treatment-​patients-​rare-type-​ ​lung-disease​ in early diffuse systemic sclerosis: analysis of a two-­year, double-­blind, randomized, [Accessed 6 Sep 2019]. controlled clinical trial. Arthritis Rheum 1999;42:1194–203. 7 sandner P, Stasch JP. Anti-­fibrotic effects of soluble guanylate cyclase stimulators and 30 Postlethwaite AE, Wong WK, Clements P, et al. A multicenter, randomized, double-­ activators: a review of the preclinical evidence. Respir Med 2017;122:S1–9. blind, placebo-­controlled trial of oral type I collagen treatment in patients with 8 Matei A-­E, Beyer C, GyörfiA -­H, et al. Protein kinases G are essential downstream diffuse cutaneous systemic sclerosis: I. oral type I collagen does not improve mediators of the antifibrotic effects of sGC stimulators. Ann Rheum Dis 2018;77:459. skin in all patients, but may improve skin in late-­phase disease. Arthritis Rheum 9 Ghofrani H-­A, Galiè N, Grimminger F, et al. Riociguat for the treatment of pulmonary 2008;58:1810–22. arterial hypertension. N Engl J Med 2013;369:330–40. 31 Herrick AL, Peytrignet S, Lunt M, et al. Patterns and predictors of skin score change in 10 Huntgeburth M, Kießling J, Weimann G, et al. Riociguat for the treatment of early diffuse systemic sclerosis from the European Scleroderma Observational Study. Raynaud’s phenomenon: a single-­dose, double-­blind, randomized, placebo-­controlled Ann Rheum Dis 2018;77:563–70. cross-­over pilot study (DIGIT). Clin Drug Investig 2018;38:1061–9. 32 Volkmann ER, Tashkin DP, Li N, et al. Mycophenolate mofetil versus placebo for 11 Dees C, Beyer C, Distler A, et al. Stimulators of soluble guanylate cyclase (sGC) inhibit systemic sclerosis-­related interstitial lung disease: an analysis of scleroderma lung experimental skin fibrosis of different aetiologies. Ann Rheum Dis 2015;74:1621–5. studies I and II. Arthritis Rheumatol 2017;69:1451–60. 12 Beyer C, Zenzmaier C, Palumbo-­Zerr K, et al. Stimulation of the soluble guanylate 33 Hsu VM, Denton CP, Domsic RT, et al. Pomalidomide in patients with interstitial lung cyclase (sGC) inhibits fibrosis by blocking non-­canonical TGFβ signalling. Ann Rheum disease due to systemic sclerosis: a phase II, multicenter, randomized, double-blind,­ Dis 2015;74:1408–16. placebo-controlled,­ parallel-group­ study. J Rheumatol 2018;45:405–10. 13 Pan S-­L, Guh J-­H, Peng C-­Y, et al. A potential role of YC-1 on the inhibition of cytokine 34 Ramiro S, Sepriano A, Chatzidionysiou K, et al. Safety of synthetic and biological release in peripheral blood mononuclear leukocytes and endotoxemic mouse models. DMARDs: a systematic literature review informing the 2016 update of the EULAR Thromb Haemost 2005;93:940–8. recommendations for management of rheumatoid arthritis. Ann Rheum Dis 14 ahluwalia A, Foster P, Scotland RS, et al. Antiinflammatory activity of soluble 2017;76:1101–36. guanylate cyclase: cGMP-­dependent down-­regulation of P-­selectin expression and 35 Prado MS, Bendtzen K, Andrade LEC. Biological anti-­TNF drugs: immunogenicity leukocyte recruitment. Proc Natl Acad Sci U S A 2004;101:1386–91. underlying treatment failure and adverse events. Expert Opin Drug Metab Toxicol 15 Beyer C, Reich N, Schindler SC, et al. Stimulation of soluble guanylate cyclase reduces 2017;13:985–95. experimental dermal fibrosis. Ann Rheum Dis 2012;71:1019–26. 36 nathan SD, Behr J, Collard HR, et al. Riociguat for idiopathic interstitial pneumonia-­ 16 allanore Y, Distler O. Systemic sclerosis in 2014: advances in cohort enrichment shape associated pulmonary hypertension (RISE-­IPP): a randomised, placebo-­controlled future of trial design. Nat Rev Rheumatol 2015;11:72–4. phase 2B study. Lancet Respir Med 2019;7:780–90. 17 Maurer B, Graf N, Michel BA, et al. Prediction of worsening of skin fibrosis in patients 37 Khanna D, Spino C, Johnson S, et al. Abatacept in early diffuse cutaneous systemic with diffuse cutaneous systemic sclerosis using the EUSTAR database. Ann Rheum Dis sclerosis: results of a phase II investigator-­initiated, multicenter, double-­blind, 2015;74:1124–31. randomized, placebo-­controlled trial. Arthritis Rheumatol 2020;72:125–36. 18 Dobrota R, Maurer B, Graf N, et al. Prediction of improvement in skin fibrosis 38 Khanna D, CJF L, Kuwana M. Efficacy and safety of tocilizumab for the treatment in diffuse cutaneous systemic sclerosis: a EUSTAR analysis. Ann Rheum Dis of systemic sclerosis: results from a phase 3 randomized controlled trial [abstract]. 2016;75:1743–8. Arthritis Rheumatol 70 (suppl 10), 2018. Available: https://​acrabstracts.org/​ ​abstract/​ 19 Distler O, Pope J, Denton C, et al. RISE-­SSc: riociguat in diffuse cutaneous systemic efficacy-and-​ ​safety-of-​ ​tocilizumab-for-​ ​the-treatment-​ ​of-systemic-​ ​sclerosis-results-​ ​ sclerosis. Respir Med 2017;122:S14–17. from-a-​ ​phase-3-​ ​randomized-​controlled-​trial/ [Accessed 23 Mar 2020]. 20 van den Hoogen F, Khanna D, Fransen J, et al. 2013 classification criteria for 39 Denton CP, Merkel PA, Furst DE, et al. Recombinant human anti-­transforming growth systemic sclerosis: an American College of Rheumatology/European League Against factor beta1 antibody therapy in systemic sclerosis: a multicenter, randomized, Rheumatism collaborative initiative. Ann Rheum Dis 2013;72:1747–55. placebo-­controlled phase I/II trial of CAT-192. Arthritis Rheum 2007;56:323–33.

Khanna D, et al. Ann Rheum Dis 2020;79:618–625. doi:10.1136/annrheumdis-2019-216823 625 Systemic sclerosis

Clinical science Initial combination therapy of ambrisentan and tadalafil in connective tissue disease-­associated pulmonary arterial hypertension (CTD-­PAH) in the modified intention-­to-­treat population of the AMBITION study: post hoc analysis Masataka Kuwana ‍ ‍ ,1 Christiana Blair,2 Tomohiko Takahashi,3 Jonathan Langley,4 John G Coghlan5

Handling editor Josef S Abstract Key messages Smolen Objectives To evaluate initial combination therapy with ambrisentan plus tadalafil (COMB) compared with ►► Additional material is What is already known about this subject? published online only. To view, monotherapy of either agent (MONO), and the utility ►► In the AMBITION study, patients with please visit the journal online of baseline characteristics and risk stratification in connective tissue disease-­associated pulmonary (http://dx.​ ​doi.org/​ ​10.​1136/​ predicting outcomes, in patients with connective tissue arterial hypertension (CTD-­PAH) benefited from annrheumdis-2019-​ ​216274). disease-­associated pulmonary arterial hypertension initial combination therapy with ambrisentan 1 (CTD-­PAH) and the systemic sclerosis (SSc)–pulmonary Allergy and Rheumatology, plus tadalafil compared with either agent alone, arterial hypertension (PAH) subpopulation. Nippon Medical School when patients with risk factors for left heart Graduate School of Medicine, Methods This post hoc analysis of the Ambrisentan and disease were excluded. Tokyo, Japan Tadalafil in Patients with PulmonaryA rterial Hypertension 2Research and Development, (AMBITION) study included patients with CTD-P­ AH from Gilead Sciences, Inc, Foster City, What does this study add? California, USA the modified intention-to-­ ­treat population. Time to clinical ►► In this post hoc analysis of the modified 3Medical Affairs, failure (TtCF) was assessed by baseline characteristics, intention-­to-­treat population of the AMBITION GlaxoSmithKline Plc, Tokyo, treatment assignment and risk group (low, intermediate study, initial combination therapy with Japan and high) at baseline and week 16. TtCF was compared 4Medical Affairs, ambrisentan and tadalafil provided benefit between groups using Kaplan-Meier­ curves and Cox GlaxoSmithKline Plc, Brentford, compared with initial monotherapy with UK proportional hazards regression modelling. 5 either agent alone in patients with CTD-­PAH Cardiology Services, Royal Free Results The analysis included 216 patients (COMB, or systemic sclerosis (SSc)-­pulmonary arterial Hospital, London, UK n=117; MONO, n=99). The risk of clinical failure was hypertension (PAH). The effect was most lower with COMB versus MONO (risk reduction: CTD-­ pronounced in patients with haemodynamic Correspondence to PAH 51.7%, SSc-­PAH 53.7%), particularly in patients Professor Masataka Kuwana, parameters characteristic of typical PAH without with haemodynamic parameters characteristic of typical Allergy and Rheumatology, features of left heart disease and/or restrictive PAH without features of left heart disease and/or Nippon Medical School lung disease at baseline. Graduate School of Medicine, restrictive lung disease at baseline. The risk of clinical ► This analysis also assessed the clinical Tokyo, Japan; failure was lower with COMB versus MONO in the ► kuwanam@​ ​nms.ac.​ ​jp utility of a three-­parameter non-­invasive baseline low-­risk group (HR not calculated due to no risk stratification score. Overall, higher risk events in COMB), baseline intermediate-­risk group (HR Received 5 September 2019 stratification at baseline was correlated Revised 25 February 2020 0.519, 95% CI 0.297 to 0.905) and in the week 16 low-­ with subsequent clinical failure events. A Accepted 25 February 2020 risk group (HR 0.069, 95% CI 0.009 to 0.548). potential trend for lower risk of clinical failure Published Online First Conclusions The benefit of COMB over MONO was 11 March 2020 with combination therapy compared with demonstrated in patients with CTD-P­ AH, particularly in monotherapy in all risk groups was observed those with typical PAH haemodynamic characteristics at in both CTD-­PAH and SSc-­PAH populations. baseline. COMB is appropriate for patients categorised The clinical utility of a simplified risk category as low risk and intermediate risk at baseline and low risk at follow-­up (week 16) was limited, but when at follow-­up. used, the risk of clinical failure was lower in Trial registration number NCT01178073. patients undergoing combination therapy versus monotherapy in the low-­risk group. © Author(s) (or their employer(s)) 2020. Re-­use permitted under CC BY-­NC. No Introduction cases are related to systemic sclerosis (SSc), which commercial re-­use. See rights Pulmonary arterial hypertension (PAH) is a progres- has a poor prognosis compared with other CTDs, and permissions. Published sive, life-threatening­ disease.1 2 Connective tissue with a 3 year survival rate of just 56%.4 6–8 by BMJ. disease-­associated pulmonary arterial hypertension The current treatment strategy in PAH is based on To cite: Kuwana M, (CTD-­PAH) is the second-­most common aetiology; assessment of disease severity at diagnosis, and subse- Blair C, Takahashi T, it has a range of underlying features, the most severe quent treatment escalation based on the patient’s et al. Ann Rheum Dis disease characteristics and the highest mortality of condition.9 10 The 2015 European Society of Cardi- 2020;79:626–634. all PAH subgroups.3–5 Almost 75% of CTD-PAH­ ology (ESC) and European Respiratory Society (ERS)

626 Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 Systemic sclerosis

40 mg) or monotherapy (ambrisentan 10 mg plus placebo or Key messages tadalafil 40 mg plus placebo). Details of the study design have been described previously.12 How might this impact on clinical practice or future developments? ►► Combination therapy is appropriate for patients with typical Patient and public involvement PAH categorised as low and intermediate risk at baseline, Patients were not involved in the design of this study nor in the while high-­risk patients could possibly be considered for post hoc analysis. more advanced therapy. On reassessment of risk at follow-­ up, those remaining in the intermediate-­risk category or Patient population deteriorating to the high-­risk category may potentially benefit The patient population has been described previously.12 13 During from escalation of therapy. the study, a blinded interim review revealed a high prevalence of ►► A simplified risk stratification score at baseline may help risk factors for left ventricular diastolic dysfunction; therefore, inform disease management in patients with CTD-­PAH, but eligibility criteria were amended so that patients with ≥3 of the further studies with larger patient populations are required. following risk factors for left ventricular diastolic dysfunction were excluded: body mass index ≥30 kg/m2, history of essential hypertension, diabetes mellitus and significant coronary artery disease. In addition, inclusion criteria for the diagnosis of PAH guidelines and 2018 World Symposium of Pulmonary Hyperten- were modified: PVR increased from ≥240 to ≥300 dyne·/sec/ 5 5 sion consensus recommend the use of risk stratification to guide cm ; for patients with a PVR 300–499 dyne·sec/cm , a pulmonary treatment decisions, classifying patients as low, intermediate or arterial wedge pressure (PAWP)/left ventricular end diastolic pres- 12 high-­risk.10 However, the clinical utility of risk stratification in sure of ≤12 mm Hg was required (PAS population). This analysis patients with CTD-­PAH or SSc-­PAH has not yet been established.11 included all patients (PAS and ex-PAS)­ who were randomised and The Ambrisentan and Tadalafil in Patients with Pulmonary Arte- received ≥1 dose of study drug (mITT population). rial Hypertension (AMBITION) study demonstrated a reduced risk of clinical failure in treatment-naive­ patients with PAH Assessments receiving initial ambrisentan plus tadalafil compared with initial The AMBITION trial12 primary endpoint was the time from 12 monotherapy of either agent. A subanalysis of AMBITION and randomisation to the first adjudicated clinical failure. Clinical a separate clinical trial showed that combination therapy with failure was defined as death (any cause), hospitalisation for wors- ambrisentan and tadalafil is effective in the treatment of CTD-­ ening PAH, disease progression (defined as a decrease of >15% 8 13 PAH and SSc-­PAH. Additionally, although SSc-PAH­ is typically in 6 min walking distance (6MWD) from baseline combined less responsive to therapy than other forms of PAH, significant with WHO Functional Class (WHO-­FC) III/IV symptoms at improvements in haemodynamic measurements (such as cardiac two consecutive study visits ≥2 weeks apart), or unsatisfactory index, mean pulmonary artery pressure (mPAP) and pulmonary long-­term clinical response (decrease in 6MWD from baseline vascular resistance (PVR)) were observed in response to ambris- (at two consecutive study visits ≥2 weeks apart)), together with entan plus tadalafil compared with monotherapy, suggesting that WHO-­FC III symptoms after 6 months of therapy. Secondary 8 these haemodynamic measurements have prognostic value. endpoints included change from baseline to week 24 in N-ter­ - The AMBITION protocol was amended during the trial to minal pro-­B-type­ natriuretic peptide (NT-­proBNP) and 6MWD. exclude patients with less pronounced haemodynamics and/ Time to clinical failure (TtCF) for the following baseline 12 or multiple risk factors for left heart disease. The primary subgroups was investigated (median values were used for contin- analysis set (PAS) included patients who met these amended criteria; uous variables): age (≤/>63 years); SSc-PAH­ diagnosis (yes/ the modified intention-­to-­treat (mITT) population comprised all no); haemodynamic parameters (mPAP (≤/>43 mm Hg)); PVR randomised patients who received study drug, including those who (≤/>593.5 dyne·sec/cm5); PAWP (≤/>9 mm Hg); cardiac index did not meet the revised eligibility criteria (ex-PAS).­ The mITT (≤/>2.5 L/min/m2); transpulmonary pressure gradient (TPG, population included an additional 29 patients with CTD-­PAH, 19 ≤/>34 mm Hg)); pulmonary function test parameters (forced of whom had SSc-PAH.­ Results for the PAS population12 and a expiratory volume in one second (FEV1, % predicted normal) post hoc analysis of the CTD-­PAH and SSc-­PAH subgroups in the ≤/≥79.025%); total lung capacity (TLC, % predicted normal) 13 PAS have been published, as has a post hoc analysis of the mITT ≤/>86.11%); ≥1 comorbidity or no comorbidities, included in population, which found that initial combination therapy may be the PAS population (yes/no); immunosuppressant use (yes/no); associated with improved long-­term survival compared with initial and abbreviated risk category (low, intermediate or high). These 14 monotherapy. parameters were selected as they have been identified as predic- This post hoc subgroup analysis of the AMBITION mITT-­ tors of poor outcome in prior studies.10 11 15–18 CTD-­PAH population assessed the relationship between base- This study also assessed the clinical utility of an abbreviated line characteristics (including haemodynamic parameters) and version of the risk assessment method proposed in the 2015 outcome, and evaluated the utility of an abbreviated, three-­ ESC/ERS PAH guidelines.10 19 Risk score was determined at parameter non-­invasive risk stratification score in predicting baseline and week 16 based on three parameters graded 1 (low outcomes in patients with CTD-PAH.­ risk), 2 (intermediate risk) or 3 (high risk), according to thresh- olds defined in the 2015 ESC/ERS PAH guidelines10: 6MWD Methods (low, >440 m; intermediate, 165–440 m; and high, <165 m), Study design NT-proBNP­ (low, <300 ng/L; intermediate, 300–1400 ng/L; and This was a post hoc subgroup analysis of the mITT-CTD-­ ­PAH high, >1400 ng/L) and WHO-­FC (low, I or II; intermediate, III; population and SSc-­PAH subpopulation of the AMBITION and high, IV). This grading method was based on a previous study.12 Patients were randomised 2:1:1 to receive one time analysis of the Comparative, Prospective Registry of Newly per day combination therapy (ambrisentan 10 mg plus tadalafil Initiated Therapies for Pulmonary Hypertension (COMPERA)

Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 627 Systemic sclerosis database.20 A mean grade was obtained by dividing the sum of all and age was higher in the SSc-PAH­ group than in the CTD-­PAH grades by the number of available parameters for each patient/ group; other disease characteristics were similar (table 1). time point and rounded to the closest integer. Transition from baseline to week 16 was categorised as improved, maintained or Time to clinical failure exacerbated. Patients in the high-­risk subgroup at baseline and The risk of clinical failure was 51.7% lower in the combination week 16 were included in the exacerbated group. TtCF after therapy group than in the monotherapy group in the CTD-PAH­ week 16 was analysed by week 16 abbreviated risk score cate- population (HR 0.483, 95% CI 0.286% to 0.817%) and 53.7% gory. Week 16 was selected rather than week 24 to minimise the lower in the SSc-­PAH population (HR 0.463, 95% CI 0.240% to number of events excluded. Risk stratification in the CTD-PAH­ 0.895%) (figure 2). population was also assessed according to the French registry non-­invasive method (ie, stratification by number of ESC/ERS low-­risk criteria10 fulfilled for WHO-­FC, NT pro-­BNP and Change in NT-proBNP and 6MWD from baseline to week 24 6MWD) at baseline and week 16.19 In CTD-PAH­ and SSc-PAH­ populations, the decrease in Safety was assessed through incidence of adverse events (AEs). geometric mean NT-proBNP­ levels and the increase in 6MWD Efficacy and safety assessments were performed at screening and from baseline to week 24 were greater in patients receiving randomisation visits, at weeks 4, 8, 16, 24 and every 12 weeks combination therapy than monotherapy (online supplementary thereafter, at the final assessment visit and at the end-­of-­study visit. table S1).

Statistical analysis TtCF by baseline subgroup The Kaplan–Meier product limit method was used to generate In the CTD-PAH­ population, a trend for lower risk of clin- curves for TtCF. HRs and associated 95% CIs were calculated ical failure with combination therapy versus monotherapy using Cox proportional hazards regression. For the subgroup was observed for all baseline subgroups, except PVR≤593.5 analyses, the regression model included a term for subgroup dyne·sec/cm5 and ex-­PAS; however, data should be interpreted by treatment interaction. All analyses are post hoc; therefore, cautiously due to large CIs. The risk of clinical failure was lower p values are not presented, except where a HR cannot be calcu- in patients undergoing combination therapy versus monotherapy lated due to no clinical failure events. For the Cox proportional for the following baseline haemodynamic characteristics: mPAP hazards regression, where there was evidence of non-­proportional >43 mm Hg: HR 0.46, 95% CI 0.24 to 0.89; PVR >593.5 hazards, restricted mean survival time was calculated based on dyne·sec/cm5: HR 0.25, 95% CI 0.13 to 0.51; TPG >34: HR time from randomisation to first clinical failure event or week 0.35, 95% CI 0.18 to 0.68; PAWP ≤9 mm Hg: HR 0.32, 95% 48 (area under the curve for each group). Week 48 was selected CI 0.15 to 0.69; cardiac index ≤2.5 L/min/m2: HR 0.34, 95% as the time point because it provided a sufficient sample size. For CI 0.18 to 0.67; and TLC >86.11%: HR 0.30, 95% CI 0.15 to 6MWD, missing values when calculating change from baseline 0.61 (figure 3). The risk of clinical failure was also lower with to week 24 were imputed using last-observation-­ ­carried-­forward combination therapy versus monotherapy in the PAS population, 12 imputation or worst-­case imputation. patients aged ≤63 years, patients who were not receiving base- line immunosuppressants, patients with ≥1 comorbidity and Results patients with no comorbidities (figure 3). For analyses showing Study population evidence of non-­proportional hazards (PVR >593.5 dyne·sec/ This analysis included 216 patients with CTD-PAH­ (combina- cm5, cardiac index ≤2.5 L/min/m2 and no comorbidities), tion therapy, n=117; monotherapy (pooled), n=99) from the restricted mean survival times are detailed in online supplemen- mITT population (n=605) (figure 1). Most patients were female tary table S2.

Figure 1 Participant disposition and aetiology: (A) population distribution among treatment arms and (B) CTD-PAH­ population aetiologies (post hoc summary). *Five patients did not receive the study drug. CTD, connective tissue disease; ITT, intention to treat; MCTD, mixed connective tissue disease; mITT, modified intention to treat; PAH, pulmonary arterial hypertension; SSc, systemic sclerosis; SLE, systemic lupus erythematosus.

628 Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 Systemic sclerosis . 85.8 (17.38) Monotherapy (pooled) Monotherapy (n=56) 2.6 (0.6) 633 (291) 60.8 (10.3) 43.4 (10.8) 871 (332–2312) 49 (88) 9.2 (3.3) 6 (10.7) 17 (30) 9 (16.1) 39 (70) 19.5 336 (95) 90.2 (21.65) 355 7.6 (4.5) pulmonary arterial hypertension; PAP, pulmonary arterial PAP, pulmonary arterial hypertension; ­ FC, WHO Functional Class FC, SSc ­ type natriuretic peptide; PAH, ­ type natriuretic peptide; ­ B- 87.3 (20.38) PAH- Combination therapy (n=81) 2.6 (0.6) 639 (279) 62.6 (8.9) 42.3 (10.7) 1261 (265–3144) 67 (83) 8.4 (3.0) 6 (7.4) 23 (28) 12 (14.8) 58 (72) 23 319 (91) 88.9 (16.82) 330 7.2 (4.1) ­ terminal pro- ­ proBNP, N- proBNP, n=80; monotherapy, n=56). monotherapy, n=80; n=77; monotherapy, n=55). monotherapy, n=77; n=80; monotherapy, n=54). monotherapy, n=80; apy (pooled) 81.2 (16.32) Monother (n=99) 2.7 (0.7) 627 (277) 58.3 (13.0) 43.7 (10.6) 1047 (337–2334) 88 (89) 9.6 (3.4) 15 (15.2) 30 (30) 21 (21.2) 69 (70) 25 330 (96) 88.1 (19.52) 347 7.7 (4.4) PAH: combination therapy, PAH: ­ PAH: combination therapy, PAH: ­ PAH: combination therapy, PAH: ­ n=55). monotherapy, n=98; SSc- n=98; monotherapy, monotherapy, n=97; SSc- n=97; monotherapy, monotherapy, n=97; SSc- n=97; monotherapy, PAH: monotherapy, ­ PAH: SSc- D ­ CT 85.1 (19.27) Combination therapy (n=117) PAH- 2.6 (0.6) 638 (282) 59.1 (12.1) 42.6 (10.6) 978 (309–2757) 101 (86) 8.5 (3.2) 17 (14.5) 33 (28) 22 (18.8) 84 (72) 23 326 (89) 88.3 (15.83) 334 7.4 (4.2) , forced expiratory volume in one second; 6MWD, 6 min walking distance; NT- distance; 6 min walking 6MWD, forced expiratory volume in one second; , 1 CTD: combination therapy, n=116; combination therapy, ­ CTD: CTD: combination therapy, n=115). combination therapy, ­ CTD: CTD: combination therapy, n=111; combination therapy, ­ CTD: CTD: combination therapy, n=115; combination therapy, ­ CTD: CTD: monotherapy, n=98; monotherapy, ­ CTD: (SD) 1 ­ proBNP (ng/L)§ ), mean (SD) ), 5 ), mean (SD)† ), 2 ailable in some patients (PAH- Baseline demographics and clinical characteristics

FC, n (%) ­ FC, PAP (mm Hg), mean (SD) (mm Hg), PAP Median (Q1–Q3) NT- PAWP (mm Hg), mean (SD)‡ (mm Hg), PAWP

Immunosuppressants

II PVR (dyne s/cm

Oral corticosteroids III Median days from diagnosis to treatment

Mean (SD) Mean % predicted TLC (SD)¶ Mean % predicted

Median Mean % predicted FEV RAP (mm Hg), mean (SD)* RAP (mm Hg),

CI (L/min/m CI, cardiac index; CTD, connective tissue disease; FEV connective tissue disease; CTD, cardiac index; CI, ¶Data not available in some patients (PAH- ¶Data not available §Data not available in some patients (PAH- §Data not available ‡Data not available in some patients (PAH- ‡Data not available *Data not available in some patients (PAH- *Data not available †Data not av Table 1 Table pressure; PAWP, pulmonary arterial wedge pressure; PVR, pulmonary vascular resistance; Q, quartile; RAP, right arterial pressure; SSc, systemic sclerosis; TLC, total lung capacity; WHO- total lung capacity; TLC, systemic sclerosis; SSc, right arterial pressure; RAP, quartile; Q, resistance; pulmonary vascular PVR, pulmonary arterial wedge pressure; PAWP, pressure; Mean age (SD) (years) Female, n (%) Female, Prior medications, n (%) Prior medications, WHO-

6MWD (m) Haemodynamic criteria

Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 629 Systemic sclerosis

observed between treatment groups in the proportion of patients who improved, maintained or worsened from baseline to week 16 (online supplementary table S3). Risk of clinical failure on or after week 16 was lowest in the low-­risk group and highest in the high-­risk group (low vs high: HR 0.198, 95% CI 0.067 to 0.580) (figure 4C). The risk of clinical failure after week 16 was lower with combination therapy versus monotherapy for patients in the week 16 low-risk­ group (HR 0.069, 95% CI 0.009 to 0.548); no significant difference between treatments was observed in the week 16 intermediate-risk­ group (HR 0.705, 95% CI 0.372 to 1.337) (figure 4D). At week 16, patient numbers were insufficient in the high-­risk group to allow a meaningful comparison between treat- ment groups. Findings showed a similar pattern in the SSc-PAH­ population compared with the CTD-PAH­ population (online supplementary figure S1). When assessed via the French registry non-­invasive method, patients fulfilling two or three low-risk­ criteria at baseline or week 16 had a significantly lower risk of subsequent clinical failure than those fulfilling 0 criteria (online supplementary figure S2). The risk of clinical failure at week 16 was lower with combination therapy versus monotherapy in patients fulfilling two or three criteria (online supplementary figure S2).

Safety Overall, there was a similar AE profile across both treatment groups in the CTD-­PAH and SSc-­PAH populations (table 2). The most common AEs in both treatment groups were peripheral oedema and headache. Peripheral oedema was more common with combination therapy than monotherapy (both CTD-PAH­ and SSc-PAH)­ (table 2). Figure 2 Kaplan-­Meier plots of time to first adjudicated clinical failure in the modified intention-­to-­treat population: (A) connective tissue Discussion disease-­associated pulmonary arterial hypertension and (B) systemic In this post hoc analysis of the AMBITION study mITT popu- sclerosis pulmonary arterial hypertension. 95% CIs (using log–log lation, initial combination therapy of ambrisentan and tadalafil transform method) are presented for each treatment group at weeks 4, was more effective than monotherapy in reducing the risk of 8, 16 and 24, and then every 12 weeks up to week 96. clinical failure: 51.7% reduction in patients with CTD-PAH­ and 53.7% reduction in patients with SSc-PAH.­ The risk of clinical Risk stratification failure in the CTD-PAH­ population was lower with combination When using the abbreviated COMPERA risk stratification therapy versus monotherapy for most baseline haemodynamic method, at baseline, 27/216 (12.5%) patients were classified characteristic subgroups. as low risk, 179/216 (82.9%) as intermediate risk and 10/216 The benefit of combination therapy over monotherapy was (4.6%) as high-risk.­ In the CTD-­PAH population, the risk of pronounced in patients with baseline haemodynamic parame- clinical failure was lowest for patients in the low-risk­ group and ters characteristic of typical PAH, such as high PVR, low cardiac highest in the high-­risk group (low vs high: HR 0.192, 95% index, high TPG and low PAWP. By contrast, in patients with an CI 0.047 to 0.777) (figure 4A). Kaplan–Meier curves demon- increased risk of left heart disease (low PVR and high PAWP) and/ strated a potential trend towards lower risk of clinical failure or restrictive lung disease (low TLC), the benefit of combination with combination therapy versus monotherapy in all risk groups therapy was less pronounced. Therefore, in the ex-PAS­ popu- at baseline (figure 4B). The risk of clinical failure was lower with lation (an older subgroup of patients with more cardiovascular combination therapy than monotherapy in the intermediate-risk­ comorbidities),21 less benefit may be gained from combination group (intermediate-­risk: HR 0.519, 95% CI 0.297 to 0.905). therapy.22 Further research to determine the optimal treatment However, in the high-­risk group, although a trend towards a approach in addition to standard PAH therapies is required for lower risk of clinical failure with combination therapy versus this population. monotherapy was observed (HR 0.197, 95% CI 0.022 to 1.803), The overall treatment goal in PAH is to achieve or maintain the sample sizes were small (combination therapy, n=4; mono- low-­risk status10 19; therefore, classification of low-risk­ at base- therapy, n=6). Additionally, in the low-­risk group, no HR could line or transitioning to low-risk­ status at week 16 is expected be calculated due to no events in the combination therapy group; to be associated with improved outcomes.19 Overall, higher risk however, comparison by log-­rank test supported a lower risk of stratification at baseline and week 16 was correlated with higher clinical failure with combination therapy versus monotherapy risk of clinical failure. When stratified by baseline risk group, (figure 4B). Kaplan-­Meier plots suggested a potential trend towards lower At week 16, 19 patients had missing data; 55/197 (28%) risk of clinical failure with combination therapy versus mono- patients were classified as low risk, 135/197 (69%) as intermediate therapy in all risk groups (CTD-PAH­ and SSc-PAH­ populations); risk and 7/197 (4%) as high risk. No significant difference was however, patient numbers in the high-risk­ groups were small.

630 Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 Systemic sclerosis

Figure 3 Forest plot of time to first adjudicated clinical failure: combination therapy versus monotherapy (pooled) per baseline subgroup. *Significant treatment by subgroup interaction; †evidence of non-proportionality­ (see online supplementary table S2); ‡hypertension, diabetes and coronary artery disease; §low-­risk valid HR could not be calculated; ¶indication for immunosuppressant use unknown. FEV1, forced expiratory volume in one second; mPAP, mean pulmonary arterial pressure; NA, not applicable; PAS, primary analysis set; PAWP, pulmonary arterial wedge pressure; PVR, pulmonary vascular resistance; SSc, systemic sclerosis; TLC, total lung capacity; TPG, transpulmonary pressure gradient.

When stratified by risk group at week 16, Kaplan–Meier plots simplified risk stratification at week 16 may have limited utility showed a clear benefit for low-­risk patients receiving combina- in the prediction of future outcomes in this population. tion therapy versus monotherapy. For patients at intermediate Methods of risk score calculation employed in previous studies risk, combination therapy provided a reduced risk of clinical often used a higher number of invasive and/or non-invasive­ failure versus monotherapy from week 16–100, then the lines parameters than the method employed here.23–25 For example, converged and plateaued. This suggests that combination therapy a previous analysis of the AMBITION study used the Registry may delay, but not prevent, progression in patients considered to Evaluate Early and Long-­Term PAH Disease Management intermediate risk at follow-­up; therefore, these patients and (REVEAL)risk score to stratify patients by baseline risk score.25 low-­risk patients on monotherapy may benefit from escalation In the current analysis, the utility of an abbreviated COMPERA of therapy. Additionally, patients considered high risk at base- risk score using three non-­invasive parameters was evaluated. line or deteriorating to the high-­risk group at follow-­up may This methodology ensured a direct comparison between base- benefit from an escalation of therapy, but patient numbers were line and week 16 as these non-invasive­ parameters were assessed insufficient to draw meaningful conclusions. Clinical evaluation at both time points. In addition, stratification using simpli- based on the maintenance or improvement of risk score using fied, minimally invasive parameters has been shown to have a

Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 631 Systemic sclerosis

Figure 4 Kaplan-­Meier curves for time to clinical failure in the CTD-PAH­ population by risk category according to the abbreviated COMPERA method at baseline and week 16: (A) baseline CTD-­PAH: overall, (B) baseline CTD-­PAH: combination versus monotherapy (pooled), (C) week 16 CTD-­ PAH: overall and (D) week 16 CTD-­PAH: combination versus monotherapy (pooled (low risk and intermediate risk only)). *NO HR calculation due to NO event in the combination therapy group (log-­rank test: p=0.048). Week 16 Kaplan-­Meier plots were described based on the period from week 16 to the last assessment visit. Patient numbers were too low in the week 16 high-risk­ group to provide a meaningful comparison. CTD-­PAH, connective tissue disease-­associated pulmonary arterial hypertension. predictive value similar to the full REVEAL risk score;26 27 this observed between the number of low-­risk criteria achieved and technique is likely to be simpler to apply in the clinic compared improved outcomes suggested the validity of a three-­parameter with those involving invasive parameters. non-­invasive risk assessment method in predicting outcomes in The lower number of parameters used in the current analysis patients with PAH.19 may partly explain why risk categorisation did not support the When we determined risk stratification using the French benefit of initial combination therapy compared with mono- registry non-­invasive method, a higher number of low-risk­ therapy across all risk groups at both time points. Additionally, criteria were associated with a lower clinical failure risk at the non-­invasive approach may not adequately discriminate baseline and follow-up;­ this is in line with findings using the low-­risk and intermediate-risk­ patients. Given that our find- abbreviated COMPERA method, supporting its validity as a risk ings suggest the importance of baseline haemodynamic char- assessment tool. Further studies with larger patient populations acteristics in predicting outcomes in patients with CTD-PAH,­ would be valuable in confirming the clinical utility of risk strati- the addition of haemodynamic information may improve the fication in predicting outcomes. sensitivity and validity of risk stratification and may help AEs were consistent with the known safety profile of ambris- inform disease management.11 Another approach to improve entan and tadalafil in the CTD-PAH­ population.12 13 In line risk stratification could be to use the seven-parameter­ strati- with findings in the AMBITION trial,12 peripheral oedema was fication as per the COMPERA analysis24 for risk stratification more common with combination therapy than monotherapy. In at baseline, with the abbreviated four-parameter­ non-­invasive patients receiving multiple therapies, a greater number of side method23 at follow-up.­ However, in a recent post hoc anal- effects compared with patients undergoing single therapy are ysis of the PATENT and CHEST studies,28–31 an association possible, particularly therapies associated with vasodilation.32

632 Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 Systemic sclerosis

Table 2 Summary of AEs in patients with CTD-­PAH and SSc-­PAH (mITT population) CTD-­PAH SSc-­PAH Monotherapy Monotherapy Combination Combination therapy Ambrisentan Tadalafil therapy Ambrisentan Tadalafil Number of patients with an AE, n (%) (n=117) (n=52) (n=47) (n=81) (n=28) (n=28) AEs* Any AE 116 (99) 50 (96) 45 (96) 80 (99) 27 (96) 27 (96) Peripheral oedema 56 (48) 20 (38) 14 (30) 37 (46) 9 (32) 9 (32) Headache 40 (34) 18 (35) 18 (38) 24 (30) 9 (32) 10 (36) Diarrhoea 34 (29) 17 (33) 12 (26) 23 (28) 5 (18) 9 (32) Dizziness 21 (18) 15 (29) 12 (26) 13 (16) 7 (25) 7 (25) Dyspnoea 25 (21) 12 (23) 10 (21) 20 (25) 5 (18) 6 (21) Nausea 21 (18) 12 (23) 11 (23) 13 (16) 6 (21) 8 (29) Arthralgia 20 (17) 7 (13) 12 (26) 10 (12) 3 (11) 4 (14) Cough 19 (16) 5 (10) 11 (23) 13 (16) 0 7 (25) Serious AEs† Any serious AE 53 (45) 20 (38) 24 (51) 36 (44) 12 (43) 17 (61) Pulmonary hypertension 7 (6) 4 (8) 6 (13) 4 (5) 1 (4) 5 (18) Pneumonia 10 (9) 4 (8) 3 (6) 5 (6) 4 (14) 2 (7) Dyspnoea 4 (3) 2 (4) 2 (4) 4 (5) 1 (4) 2 (7) Anaemia 5 (4) 1 (2) 3 (6) 4 (5) 1 (4) 1 (4) *AEs occurring in ≥25% of patients in any group. †Serious AEs occurring in >4% of patients in any group. AE, adverse event; CTD, connective tissue disease; mITT, modified intention to treat; PAH, pulmonary arterial hypertension; SSc, systemic sclerosis.

In patients with CTD-PAH,­ the risk of clinical failure was Patient and public involvement Patients and/or the public were not involved in lower with combination therapy versus monotherapy, partic- the design, or conduct, or reporting, or dissemination plans of this research. ularly in those with baseline haemodynamic parameters char- Patient consent for publication Not required. acteristic of typical PAH, without features of left heart disease Provenance and peer review Not commissioned; externally peer reviewed. and/or restrictive lung disease. A simplified risk stratification Data availability statement Anonymised individual participant data and score at baseline using non-­invasive parameters may be useful study documents can be requested for further research from www.​clin​ical​stud​ydat​ in predicting PAH-­related outcomes in patients with CTD-­PAH; arequest.com.​ however, its clinical utility at week 16 is limited. At baseline, Open access This is an open access article distributed in accordance with the Kaplan-­Meier curves demonstrated a potential trend towards Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which lower risk of clinical failure with combination therapy versus permits others to distribute, remix, adapt, build upon this work non-­commercially, monotherapy in all risk groups. Combination therapy appeared 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 beneficial compared with monotherapy in the week 16 low-­risk is non-­commercial. See: http://creativecommons.​ ​org/licenses/​ ​by-​nc/​4.0/.​ group; other groups may benefit from escalated therapy. Incor- porating additional haemodynamic information may increase ORCID iD risk stratification validity at follow-up;­ further research deter- Masataka Kuwana http://orcid.​ ​org/0000-​ ​0001-​8352-​6136 mining the optimal predictive tools in terms of applicability and accuracy at baseline, and follow-up­ in a large-scale­ cohort is References warranted. 1 Frost A, Badesch D, Gibbs JSR, et al. Diagnosis of pulmonary hypertension. Eur Respir Correction notice This article has been corrected since it published Online First. J 2018:1801904. In table 1, the unit of PVR has been corrected and number of FC III in SSC/Mono 2 Montani D, Günther S, Dorfmüller P, et al. Pulmonary arterial hypertension. Orphanet J amended. Rare Dis 2013;8:97. 3 Prins KW, Thenappan T. World Health organization group I pulmonary hypertension: Acknowledgements Medical writing support was provided by Gillian Wallace, epidemiology and pathophysiology. Cardiol Clin 2016;34:363–74. MSc, of Fishawack Indicia Ltd, UK, and funded by GSK. The results of this study were 4 Thakkar V, Lau EMT. Connective tissue disease-­related pulmonary arterial presented in part as an oral presentation at the annual European League Against hypertension. Best Pract Res Clin Rheumatol 2016;30:22–38. Rheumatism meeting, Madrid, 12–15 June, 2019 (Ann Rheum Dis 2019; 78:105–6) 5 Chung L, Liu J, Parsons L, et al. Characterization of connective tissue disease-­ and in part at the 5th Systemic Sclerosis World Congress, Bordeaux, 15–17 February associated pulmonary arterial hypertension from reveal: identifying systemic sclerosis 2018 (Scleroderma Relat Disord 2018;3: suppl). as a unique phenotype. Chest 2010;138:1383–94. Contributors All listed authors contributed substantially to the planning and 6 Coghlan JG, Denton CP, Grünig E, et al. Evidence-­Based detection of pulmonary implementation of this research. All authors revised the manuscript critically for arterial hypertension in systemic sclerosis: the detect study. Ann Rheum Dis important intellectual content, approved the final version for publication and agreed 2014;73:1340–9. to be listed as authors. 7 Hachulla E, Carpentier P, Gressin V, et al. Risk factors for death and the 3-­year survival of patients with systemic sclerosis: the French ItinérAIR-­Sclérodermie study. Funding This study was funded by GlaxoSmithKline (study number 112565) and Rheumatology 2009;48:304–8. Gilead Sciences, Inc. 8 Hassoun PM, Zamanian RT, Damico R, et al. Ambrisentan and tadalafil up-­front Competing interests MK has received personal fees from GSK, Bayer and combination therapy in Scleroderma-­associated pulmonary arterial hypertension. Am J Nippon Shinyaku, and research grants and personal fees from Actelion and Pfizer. Respir Crit Care Med 2015;192:1102–10. CB is an employee and shareholder at Gilead Sciences. JL is a former employee and 9 Galiè N, Channick RN, Frantz RP, et al. Risk stratification and medical therapy of shareholder of GSK. TT is an employee at GSK. JGC has received research grants pulmonary arterial hypertension. Eur Respir J 2019;53:1801889. and personal fees from Actelion and GSK, and personal fees from Bayer, Endotronix, 10 Galiè N, Humbert M, Vachiery J-­L, et al. 2015 ESC/ERS Guidelines for the diagnosis Pfizer and UnitedTherapeutics. and treatment of pulmonary hypertension. Eur Respir J 2015;46:903–75.

Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 633 Systemic sclerosis

11 Weatherald J, Boucly A, Launay D, et al. Haemodynamics and serial risk assessment 21 Vachiéry J-­L, Tedford RJ, Rosenkranz S, et al. Pulmonary hypertension due to left heart in systemic sclerosis associated pulmonary arterial hypertension. Eur Respir J disease. Eur Respir J 2019;53:1801897. 2018;52:1800678. 22 shirai Y, Kuwana M. Complex pathophysiology of pulmonary hypertension associated 12 Galiè N, Barberà JA, Frost AE, et al. Initial use of ambrisentan plus tadalafil in with systemic sclerosis: potential unfavorable effects of vasodilators. J Scleroderma pulmonary arterial hypertension. N Engl J Med 2015;373:834–44. Relat Disord 2017;2:92–9. 13 Coghlan JG, Galiè N, Barberà JA, et al. Initial combination therapy with ambrisentan 23 Boucly A, Weatherald J, Savale L, et al. Risk assessment, prognosis and guideline and tadalafil in connective tissue disease-­associated pulmonary arterial implementation in pulmonary arterial hypertension. Eur Respir J 2017;50:1700889. hypertension (CTD-­PAH): subgroup analysis from the ambition trial. Ann Rheum Dis 24 Hoeper MM, Kramer T, Pan Z, et al. Mortality in pulmonary arterial hypertension: 2017;76:1219–27. prediction by the 2015 European pulmonary hypertension guidelines risk stratification 14 Hoeper MM, McLaughlin VV, Barberá JA, et al. Initial combination therapy with model. Eur Respir J 2017;50:1700740. ambrisentan and tadalafil and mortality in patients with pulmonary arterial 25 Frost AE, Hoeper MM, Barberá JA, et al. Risk-­Stratified outcomes with initial hypertension: a secondary analysis of the results from the randomised, controlled combination therapy in pulmonary arterial hypertension: application of the reveal risk ambition study. Lancet Respir Med 2016;4:894–901. score. J Heart Lung Transplant 2018;37:1410–7. 15 Mathai SC, Suber T, Khair RM, et al. Health-­Related quality of life and survival in 26 Galiè N, McLaughlin VV, Rubin LJ, et al. An overview of the 6th world Symposium on pulmonary arterial hypertension. Ann Am Thorac Soc 2016;13:31–9. pulmonary hypertension. Eur Respir J 2019;53:1802148. 16 Fisher MR, Mathai SC, Champion HC, et al. Clinical differences between idiopathic 27 Cogswell R, Pritzker M, De Marco T. Performance of the reveal pulmonary arterial and scleroderma-­related pulmonary hypertension. Arthritis Rheum 2006;54:3043–50. hypertension prediction model using non-­invasive and routinely measured parameters. 17 shapiro S, Traiger GL, Turner M, et al. Sex differences in the diagnosis, treatment, and J Heart Lung Transplant 2014;33:382–7. outcome of patients with pulmonary arterial hypertension enrolled in the registry to 28 Ghofrani H-­A, Galiè N, Grimminger F, et al. Riociguat for the treatment of pulmonary evaluate early and long-­term pulmonary arterial hypertension disease management. arterial hypertension. N Engl J Med 2013;369:330–40. Chest 2012;141:363–73. 29 Ghofrani H-­A, D’Armini AM, Grimminger F, et al. Riociguat for the treatment of 18 Weatherald J, Boucly A, Chemla D, et al. Prognostic value of follow-­up hemodynamic chronic thromboembolic pulmonary hypertension. N Engl J Med 2013;369:319–29. variables after initial management in pulmonary arterial hypertension. Circulation 30 Rubin LJ, Galiè N, Grimminger F, et al. Riociguat for the treatment of pulmonary 2018;137:693–704. arterial hypertension: a long-­term extension study (PATENT-2). Eur Respir J 19 Humbert M, Farber HW, Ghofrani H-­A, et al. Risk assessment in pulmonary arterial 2015;45:1303–13. hypertension and chronic thromboembolic pulmonary hypertension. Eur Respir J 31 simonneau G, D’Armini AM, Ghofrani H-­A, et al. Riociguat for the treatment of 2019;53:1802004. chronic thromboembolic pulmonary hypertension: a long-­term extension study 20 Kylhammar D, Kjellström B, Hjalmarsson C, et al. A comprehensive risk stratification at (CHEST-2). Eur Respir J 2015;45:1293–302. early follow-­up determines prognosis in pulmonary arterial hypertension. Eur Heart J 32 Burks M, Stickel S, Galiè N. Pulmonary arterial hypertension: combination therapy in 2018;39:4175–81. practice. Am J Cardiovasc Drugs 2018;18:249–57.

634 Kuwana M, et al. Ann Rheum Dis 2020;79:626–634. doi:10.1136/annrheumdis-2019-216274 Osteoarthritis

Translational science Metformin limits osteoarthritis development and progression through activation of AMPK signalling Jun Li,1 Bin Zhang,2 Wei-­Xiao Liu,2 Ke Lu,3 Haobo Pan,3 Tingyu Wang,4 Chun-­do Oh,1 Dan Yi,1 Jian Huang ‍ ‍ ,1 Lan Zhao,1 Guangzhi Ning,2 Cong Xing,2 Guozhi Xiao ‍ ‍ ,5 Ru Liu-Bryan,­ 6 Shiqing Feng,2 Di Chen ‍ ‍ 1,3

Handling editor Josef S Abstract Key messages Smolen Objectives In this study, we aim to determine the effect of metformin on osteoarthritis (OA) development ►► Additional material is What is already known about this subject? published online only. To view and progression. ►► Metformin is the first-­line medication for the please visit the journal online Methods Destabilisation of the medial meniscus treatment of type 2 diabetes. (http://dx.​ ​doi.org/​ ​10.​1136/​ (DMM) surgery was performed in 10-­week-­old wild type ► Metformin has been shown to activate AMP-­ annrheumdis-2019-​ ​216713). and AMP-­activated protein kinase (AMPK)α1 knockout ► activated protein kinase (AMPK). (KO) mice. Metformin (4 mg/day in drinking water) was For numbered affiliations see ► Osteoarthritis (OA) phenotype induced by given, commencing either 2 weeks before or 2 weeks ► end of article. destabilisation of the medial meniscus was after DMM surgery. Mice were sacrificed 6 and 12 weeks accelerated in AMPKα1 (catalytic alpha subunit Correspondence to after DMM surgery. OA phenotype was analysed by of AMPK) knockout (KO) mice. Dr. Di Chen, Shenzhen Institutes micro-­computerised tomography (μCT), histology of Advanced Technology, and pain-­related behaviour tests. AMPKα1 (catalytic Chinese Academy of Sciences, What does this study add? alpha subunit of AMPK) expression was examined Shenzhen, China; ► Metformin had chondroprotective effect to by immunohistochemistry and immunofluorescence ► di.​ ​chen@siat.​ ​ac.​cn, Dr. Ru Liu-­ decelerate OA development and progression. Bryan, San Diego VA Healthcare analyses. The OA phenotype was also determined by μCT ► Metformin enhanced AMPK expression and System, San Diego, California, and MRI in non-­human primates. ► USA; phosphorylation in articular chondrocytes in Results Metformin upregulated phosphorylated and ruliu@​ ​ucsd.edu​ and Dr. Shiqing mice. Feng, Tianjin Medical University total AMPK expression in articular cartilage tissue. Mild ►► Metformin lost its chondroprotective effect General Hospital, Tianjin, China; and more severe cartilage degeneration was observed on OA development in AMPKα1 KO mice, fengsq@​ ​hotmail.com​ at 6 and 12 weeks after DMM surgery, evidenced by suggesting that the chondroprotective effect of markedly increased Osteoarthritis Research Society JL and BZ contributed equally. metformin is mediated by AMPK signalling. International scores, as well as reduced cartilage ► Metformin had chondroprotective effect on OA areas. The administration of metformin, commencing ► Received 25 November 2019 development in non-­human primates. Revised 13 February 2020 either before or after DMM surgery, caused significant Accepted 24 February 2020 reduction in cartilage degradation. Prominent synovial Published Online First How might this impact on clinical practice or hyperplasia and osteophyte formation were observed 10 March 2020 future developments? at both 6 and 12 weeks after DMM surgery; these were ► This study suggests that metformin may be used significantly inhibited by treatment with metformin either ► in clinics to treat patients with OA. before or after DMM surgery. The protective effects of metformin on OA development were not observed in AMPKα1 KO mice, suggesting that the chondroprotective effect of metformin is mediated by AMPK signalling. diseases.2 3 Metformin could influence metabolic In addition, we demonstrated that treatment with and cellular processes, such as inflammation, oxida- metformin could also protect from OA progression in a tive damage, diminished autophagy, cell senescence partial medial meniscectomy animal model in non-­ and apoptosis.4–7 human primates. Metformin has been shown to activate AMP-­ Conclusions The present study suggests that activated protein kinase (AMPK), a master regulator metformin, administered shortly after joint injury, can of energy balance and metabolism. Dysregulation limit OA development and progression in injury-­induced of AMPK is linked to multiple age-­related diseases OA animal models. including diabetes, atherosclerosis, cardiovascular disease, cancer, neurodegenerative diseases and osteoarthritis (OA).8–12 Reduced AMPK activity, assessed by phosphorylation of a specific threonine © Author(s) (or their employer(s)) 2020. Re-­use Introduction in the catalytic alpha subunit of AMPK (AMPKα1), permitted under CC BY-­NC. No Metformin is the first-­line medication for the is observed in both human and mouse knee OA carti- commercial re-­use. See rights treatment of type 2 diabetes1 and the fourth-­most lage.12–14 Decreased phosphorylation of AMPKα1 and permissions. Published prescribed medication in the USA in 2016 (https:// is also seen in chondrocytes after biomechanical by BMJ. www.​medicinenet.​com/​top_​drugs_​prescribed_​ injury or in response to interleukin-1β (IL-1β) and To cite: Li J, Zhang B, Liu W-­ in_​the_​us/​views.​htm). Evidence suggests that tumour necrosis factor-α (TNF-α).15 The adminis- X, et al. Ann Rheum Dis metformin is generally well tolerated and seems tration of metformin could reverse these catabolic 2020;79:635–645. to also be beneficial for a number of age-related­ responses.16 17 These findings suggest that sustained

Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 635 Osteoarthritis

AMPK activity in chondrocytes may be important for cartilage and 12 weeks after DMM surgery; both were significantly inhib- homeostasis. Indeed, it has been shown that AMPKα1 deficiency ited by treatment with metformin either before or after DMM in chondrocytes accelerates development of both injury-­induced surgery. Results of micro-­computerised tomography (µCT) anal- and age-­related spontaneous OA in mice.12 ysis also showed that significantly increased osteophyte forma- Patients with diabetes are nearly twice as likely to have tion was observed in mice 6 and 12 weeks after DMM surgery arthritis, indicating a diabetes-arthritis­ connection.18 19 Diabetes (figure 1L,M) and this effect was significantly inhibited by treat- can damage joints, a condition called diabetic arthropathy.19 ment with metformin administered commencing either before Joint pain may be a response to an illness, injury or arthritis. or after DMM surgery (figure 1L,M). Considering OA progres- AMPK has been identified as a novel target for pain treat- sion during the 6 to 12 weeks period after DMM surgery, we ment and metformin has been shown to decrease sensitivity to found that in terms of the OARSI score, cartilage area and syno- mechanical and thermal allodynia.20 21 vitis score, the progression rates were similar with or without OA is the most common form of arthritis and the leading metformin treatment in WT mice (see online supplementary cause of chronic disability among elderly people. According to figure S1B–D). However, in terms of osteophyte size and osteo- estimates published in 2015, up to 240 million people around phyte maturity, the progression rates were significantly reduced the world suffer from OA.22 Symptomatic knee OA occurs in after metformin treatment in WT mice (see online supplemen- 10% of men and in 13% of women aged 60 years or older.23 tary figure S1E,F). We also analysed changes in subchondral bone The number of people affected with symptomatic OA is likely mass and found that subchondral bone mass was significantly to increase because of the ageing of the population and the increased 6 and 12 weeks after DMM surgery in WT mice (see obesity epidemic.15 Given the role of AMPK in OA and pain online supplementary figure S1L) and metformin had no signif- and metformin being an AMPK activator,12 20 21 in the present icant effect on DMM-induced­ subchondral bone mass increase studies, we investigated the effects of metformin on prevention (see online supplementary figure S1L). These results suggest that and treatment of OA both structurally and functionally in an OA metformin can limit OA development and delay OA progression mouse model induced by destabilisation of the medial meniscus in the injury-­induced OA mouse model, administered shortly (DMM). We also determined whether metformin limits OA after joint injury. development and progression through activation of AMPK signalling using AMPKα1 knockout (KO) mice. To further deter- Metformin induced AMPK phosphorylation and expression mine the therapeutic effect of metformin in OA pathogenesis in Because metformin is a known AMPK activator,24 we exam- large animals, we performed a pilot experiment testing the effect ined, using immunohistochemistry (IHC) analysis, whether of metformin on OA treatment in a partial medial meniscectomy metformin upregulates the phosphorylation and expression of (PMM) model using a non-­human primate. Our findings demon- AMPKα1 in articular cartilage tissue. We found that metformin, strated that metformin possesses chondroprotective effects on either given before or after DMM surgery, significantly increased OA development and progression in rodents and in non-human­ phosphorylated AMPKα1 expression in articular chondrocytes primates through activation of AMPK signalling. Our studies of knee cartilage at both 6 and 12 weeks after DMM surgery suggest that metformin may be used in clinical interventions of (figure 2A,B). Notably, metformin also increased total AMPKα1 OA disease. expression in articular chondrocytes at the 6-week­ time point when it was given before or after DMM surgery (figure 2A,C). Materials and methods See online supplementary materials and methods. AMPK mediated chondroprotective effect of metformin Next, we performed an in vitro study to determine the Results effect of metformin on TNF-α and IL-1β-induced catabolic Metformin limited OA development and delayed OA response. We found that expression of Mmp3, Mmp13, progression in the mouse DMM model Adamts4 and Adamts5 was significantly upregulated by To examine the effect of metformin on OA development and TNF-α (figure 3A–D) and IL-1β (figure 3E–H) in primary progression, we performed DMM surgery in 10-­week-old­ male murine articular chondrocytes derived from WT mice. The C57 wild-type­ (WT) mice with or without metformin admin- addition of metformin significantly inhibited TNF-α and istration commencing either 2 weeks before (limit or preven- IL-1β-induced Mmp13 expression (figure 3B,F) and IL-1β-in- tion) or 2 weeks after (delay or treatment) DMM surgery (see duced Mmp3 expression (figure 3E). In addition, metformin online supplementary figure S1A). We carried out histological inhibited TNF-α-induced Adamts4 expression (figure 3C) and analyses to assess knee joint damage including articular cartilage TNF-α- and IL-1β-induced Adamts5 expression (figure 3D,H) degradation, synovial tissue hyperplasia and osteophyte forma- in WT chondrocytes. We also examined effects of metformin tion 6 and 12 weeks after DMM surgery. As expected, cartilage on expression of TNF-α- and IL-1β-induced chondrocyte degeneration was mild at 6 weeks and became more severe at catabolic marker genes in articular chondrocytes derived from 12 weeks after DMM surgery, evidenced by markedly increased AMPKα1-/- mice. Metformin had no significant effects on Osteoarthritis Research Society International (OARSI) scores expression of TNF-α and IL-1β-induced chondrocyte marker (figure 1A,B,G), as well as reduced cartilage areas, quantified by genes in articular chondrocytes derived from AMPKα1-/- mice histomorphometrical analysis of the Alcian blue stained articular (figure 3A–H). These findings suggest that metformin is a cartilage areas. The administration of metformin (4 mg/day in potent drug inhibiting catabolic responses caused by inflamma- drinking water, until the animals were sacrificed), commencing tory cytokines. In addition to catabolic genes, we also exam- either 2 weeks before or 2 weeks after DMM surgery, caused ined expression of anabolic genes. We found that metformin partial but significant reduction in cartilage degradation upregulated expression of Col2a1, aggrecan, Sox9 and IGF-1 (figure 1C,H). Prominent synovial hyperplasia (figure 1D,I) and in articular chondrocytes of WT mice (figure 3I–L, N–Q). In osteophyte formation (figure 1E,F,J,K) analysed by synovitis contrast, metformin had no significant effect on Bmp7 expres- score and osteophyte size and maturity were observed at both 6 sion (figure 3M,R). Metformin also reversed the inhibitory

636 Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 Osteoarthritis

Figure 1 Metformin limited osteoarthritis (OA) development and delayed OA progression in the mouse destabilisation of the medial meniscus (DMM) model. (A) Representative histology images of osteoarthritic knee joints, that were collected 6 and 12 weeks after DMM surgery. Yellow arrowheads indicate articular cartilage degradation, green arrowheads indicate osteophytes and red arrowheads indicate synovial hyperplasia. n=7; scale bar: 200 µm. (B, G) The severity of OA-­like phenotype was analysed by grading histological sections in medial femoral condyles (MFCs) and the medial tibial plateau (MTP) using the Osteoarthritis Research Society International (OARSI) score system. (C, H) The articular cartilage areas of MFCs and MTPs were quantified by tracing the Alcian blue positive stained areas using the OsteoMeasure system. (D, I) The degree of synovitis was semi-­quantified by the number of synovial lining layers. (E, F, J, K) Osteophytes were semi-quantified­ by evaluating the osteophyte formation score consisting of two domains, size (E, F) and maturity (J, K). n=7. (L, M) The osteophyte formation was analysed by micro-computerised­ tomography in mice with or without metformin treatment 6 and 12 weeks after DMM surgery. The volume of calcified meniscus and synovial tissue was quantified. n=6; scale bar: 1 mm. Statistical analysis was conducted using two-­way analysis of variance followed by the Tukey-­Kramer test. # p<0.05, ## p<0.01, compared between sham and DMM groups; *p<0.05, **p<0.01, ***p<0.001, compared between groups with or without metformin treatment in mice with DMM surgery. Met, metformin; WT, wild type. effects of TNF-α on expressions of Col2a1, aggrecan, Sox9, supplementary figure 3S). We also examined the effect of IGF-1 and Bmp7 (figure 3I–M). In addition, except upregu- metformin on AMPK phosphorylation in human articular lation of IGF-1 expression (figure 3L,Q), metformin had no chondrocytes and found that metformin significantly upregu- significant effects on other anabolic genes in AMPKα1-/- mice. lated AMPK phosphorylation within 16 hours time period (see These results suggest that metformin exerts its chondroprotec- online supplementary figure S2A). TNF-α and IL-1β inhibited tive effect not only by inhibition of catabolic genes but also by AMPK phosphorylation and these inhibitory effects could upregulation of anabolic genes through an AMPK-dependent­ be reversed by treatment with metformin (see online supple- mechanism. IHC results further demonstrated that expression mentary figure S2B). In addition, IL-1β-induced nitric oxide of MMP13, Adamts5 and Col-­X proteins was upregulated in release could be inhibited by metformin in a dose-dependent­ mice performed with DMM surgery (see online supplemen- manner in articular chondrocytes derived from WT mice; this tary figure 3S). The administration of metformin, commencing effect was abolished in articular chondrocytes derived from either before or after DMM surgery, could significantly inhibit AMPKα1-/- mice (see online supplementary figure S2C). In our MMP13, Adamts5 and Col-­X expression (see online supple- previous studies, we have shown that expressions of phosphor- mentary figure 3S) in WT mice. However, no significant ylated and total AMPKα1 are reduced in both primary chon- attenuation effect of metformin on MMP13, Adamts5 and drocytes isolated from knee joint tissues of OA patients (by Col-­X expression in AMPKα1-/- mice was seen (see online Western blot analysis) and articular cartilage tissues collected

Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 637 Osteoarthritis

Figure 2 Metformin induced AMP-­activated protein kinase (AMPK) phosphorylation and expression. (A) Knee joint sections from sham, destabilisation of the medial meniscus (DMM) surgery (6 or 12 weeks after DMM surgery) or treatment with or without metformin were analysed for AMPK expression using immunohistochemistry methods. Results showed that expression of both phosphorylated AMPKα (pAMPKα) (upper panel) and total AMPKα (catalyticalpha subunit of AMPK) (lower panel) was upregulated by the treatment with metformin. The cells present in the non-­ calcified regions of femoral and tibial cartilage were subjected to further analysis. (B, C) Cells with positive staining for pAMPKα and total AMPKα were quantified (n=3). Statistical analysis was conducted using two-way­ analysis of variance followed by the Tukey-­Kramer test. ## p<0.01, compared between sham and DMM groups; *p<0.05, compared between groups with or without metformin treatment in mice with DMM surgery. Met, metformin; NC, negative control; WT,wild type. from OA patients (by IHC), compared with normal primary score, osteophyte size and osteophyte maturity, had no signif- knee chondrocytes and normal knee cartilage, respectively.15 icant changes in AMPKα1-/- mice (see online supplementary We then went on to test the effects of metformin on OA figure S1G–K). In addition, we found that subchondral bone development in AMPKα1-/- mice. We found that AMPKα1-/- mass was significantly increased 6 and 12 weeks after DMM mice administered metformin still exhibited the OA pheno- surgery in AMPKα1-/- mice (see online supplementary figure type to an extent similar to that of WT mice not administered S1M). Metformin had no significant effect on DMM-­induced metformin. This was demonstrated by analysing histolog- subchondral bone mass increase (see online supplementary ical changes (figure 4A) and changes in OARSI scores figure S1M). These results suggest that the chondroprotective (figure 4B,G), cartilage area (figure 4C,H), synovitis score effect of metformin was mediated by AMPK. (figure 4D,I), osteophyte size (figure 4E,J) and osteophyte maturity (figure 4F,K). Results of µCT analysis showed that significant osteophyte formation was observed 6 and 12 weeks Metformin attenuated OA pain after DMM surgery in AMPKα1-/- mice (figure 4L,M). Treat- We then performed a series of tests to determine if metformin ment with metformin, commencing either before or after lowers OA pain sensitivity. The results of von Frey tests DMM surgery, had no significant effect on inhibition of osteo- showed that significantly reduced paw withdrawal response phyte formation in AMPKα1-/- mice (figure 4L,M). Consid- thresholds were observed after DMM surgery in mice without ering the OA progression rate, all parameters reflecting OA metformin treatment in WT mice (figure 5A,B). The admin- progression, including OARSI scores, cartilage areas, synovitis istration of metformin (4 mg/day in drinking water, until the

638 Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 Osteoarthritis

Figure 3 AMP-­activated protein kinase (AMPK) mediated chondroprotective effect of metformin. (A–H) Murine articular chondrocytes were treated with metformin (2 mg) and then incubated with tumour necrosis factor-α (TNF-α; 100 ng/mL) (A–D) or interleukin-1β (IL-1β; 10 ng/mL) (E–H) for 6 hours. Expression of Mmp3, Mmp13, Adamts4 and Adamts5 mRNAs were analysed by real-time­ polymerase chain reaction (PCR) (n=3). (I) Immunohistochemistry (IHC) assays were performed using histological sections collected from mice with sham operation or destabilisation of the medial meniscus (DMM) surgery with metformin treatment. Expression of MMP13, Adamts5 and Col-X­ was examined (n=3). AMPKα1,catalytic alpha subunit of AMPK; KO, knockout; Met, metformin; mRNA, messenger RNA; WT, wild type. animals were sacrificed), commencing either 2 weeks before or in AMPKα1-/- mice (figure 5G,H,K,L,P,Q). These results are 2 weeks after DMM surgery, could significantly increase paw consistent with those of von Frey tests. withdrawal response threshold, reflecting reduced pain sensi- Because metformin possesses a protective effect against tivity in WT mice (figure 5A,B). In contrast, metformin had DMM-­induced OA pain, we examined whether metformin no significant effect on reversing paw withdrawal response affects phosphorylation and expression of AMPKα1 in dorsal threshold and reducing pain sensitivity in AMPKα1-/- mice root ganglia (DRG) using immunofluorescence (IF) staining. (figure 5C,D). We also tested spontaneous activity of the We found that expressions of pAMPKα1 and total AMPKα1 mice in response to DMM surgery and metformin treatment were significantly reduced 6 and 12 weeks after DMM using the Laboratory Animal Behaviour Observation, Regis- surgery (figure 6A–C). Treatment with metformin upregu- tration and Analysis System (LABORAS). We found that travel lated pAMPKα1 and total AMPKα1 expression in DRG cells distance (figure 5E,F), average walking speed (figure 5I,J), after DMM surgery (figure 6A–C). AMPKα1 expression was rearing frequency (figure 5M,N) and rearing duration (see mainly detected in the cell membrane (figure 6A). We then online supplementary figure 5R,S) were significantly decreased analysed expression of a pain-­related marker downstream of 6 weeks after DMM surgery without metformin treatment. AMPK, transient receptor potential ankyrin 1 (TRPA1), and Treatment with metformin, commencing either before or after found that TRPA1 expression was significantly upregulated 6 DMM surgery, could significantly reverse reduced spontaneous and 12 weeks after DMM surgery (figure 6D). Treatment with activity caused by DMM surgery (figure 5E,F,I,J,M,N,R,S), metformin, commencing either before or after DMM surgery, indicating that in addition to the inhibition of mechanically-­ significantly inhibited TRPA1 expression in DRG tissues induced pain sensitivity, metformin could also reverse reduced (figure 6D). In addition, we also analysed expressions of a spontaneous activity observed in mice with DMM surgery. In series of pain-­related marker genes in articular chondrocytes contrast, treatment with metformin had no significant effect isolated from 4-­day-­old WT and AMPKα1-/- mice. We found on reduced spontaneous activity caused by DMM surgery that TNF-α and IL-1β significantly upregulated pain-related­

Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 639 Osteoarthritis

Figure 4 Metformin had no protective effect on destabilisation of medial meniscus (DMM)-induced­ osteoarthritis (OA) in AMP-­activated protein kinase (AMPK) α1 knockout (KO) mice. (A) DMM surgery or sham operation was performed in 10-week-­ ­old AMPKα1 KO mice. Metformin (4 mg/kg/ day in drinking water, until the animals were sacrificed) was administered to the mice 2 weeks before or 2 weeks after DMM surgery. Representative histology images of osteoarthritic knee joints demonstrated that metformin had no significant effect on OA development in AMPKα1 KO mice. Yellow arrowheads indicate articular cartilage degradation, green arrowheads indicate osteophytes and red arrowheads indicate synovial hyperplasia. n=7; scale bar: 200 µm. (B–K) Quantification or semi-­quantification assessments, including Osteoarthritis Research Society International (OARSI) score (B, G), cartilage area (C, H), synovitis score (D, I), osteophyte size (E, J) and osteophyte maturity (F, K) were performed using histological sections. n=7. (L, M) Osteophyte formation was analysed by micro-­computerised tomography in mice with or without metformin treatment 6 and 12 weeks after DMM surgery. The volume of calcified meniscus and synovial tissue was quantified (n=6). Statistical analysis was conducted using two-way­ analysis of variance followed by the Tukey-­Kramer test. # p<0.05, ## p<0.01, compared between sham and DMM surgery groups. AMPKα1,AMP-activated­ protein kinase α1; Met, metformin.

marker genes, such as MCP-1, CCR2, NGF, TNF-α in both WT macaques with 8.5 to 11.4 years of age which were treated with and AMPKα1-/- mice (see online supplementary figure S3A–J). vehicle (n=4) or metformin (n=5) 1 month after PMM surgery Treatment with metformin significantly inhibited expressions (see online supplementary figure S4). The PMM surgery was of these marker genes induced by TNF-α and IL-1β in WT performed on the left knee joint and a sham operation was chondrocytes but not in AMPKα1-­deficient chondrocytes (see performed on the right knee joint (figure 7A). OA-­like lesions online supplementary figure S3A–J). These results suggest that were observed in rhesus macaques 7 months after PMM surgery metformin may upregulate AMPKα1 expression and phos- at the macroscopic level (figure 7B). The semi-quantification­ phorylation and further inhibit pain-­related signals in DRG score of cartilage damage for femoral and tibial condyles and cartilage tissues. showed that the treatment with metformin significantly allevi- ated the cartilage lesions (figure 7C). μCT analysis showed that Metformin alleviated OA development in non-human subchondral bone mass was significantly increased in the rhesus primates macaques 7 months after PMM surgery (figure 7D). In contrast, To further investigate the role of metformin in large animals, we subchondral bone mass was significantly reduced by treatment determined the effect of metformin on OA development in non-­ with metformin in the rhesus macaques 7 months after PMM human primates. We performed PMM surgery on male rhesus surgery (figure 7D,E). MRI analysis showed that treatment with

640 Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 Osteoarthritis

Figure 5 Metformin attenuated osteoarthritis pain. (A, B) Metformin (4 mg/kg/day in drinking water, until the animals were sacrificed) was administered 2 weeks before (limit group) or after (delay group) destabilisation of the medial meniscus (DMM) surgery. Mechanical sensitivity was measured using von Frey filaments biweekly after DMM surgery. n=10. Statistical analysis was conducted using two-way­ analysis of variance (ANOVA) followed by the Tukey-­Kramer test. *p<0.05, **p<0.01, ***p<0.001, compared between groups with or without metformin treatment in mice with DMM surgery. (C–J) Changes in spontaneous activity, including travel distance (C, D), average walking speed (E, F), rearing frequency (G, H) and rearing duration (I, J) were evaluated using the Laboratory Animal Behaviour Observation Registration and Analysis System (LABORAS, Metris, Netherlands) 2, 6 and 10 weeks after DMM surgery (n=4). Statistical analysis was conducted using two-way­ ANOVA followed by the Tukey-­Kramer test. *p<0.05, compared between groups with or without metformin treatment in mice with DMM surgery. AMPKα1,catalytic alpha subunit of AMPK; Met, metformin; WT, wild type.

metformin protected against cartilage loss, demonstrated by in AMPKα1 KO mice; this suggests that the chondroprotective higher cartilage thickness observed in metformin treated rhesus effect of metformin is mediated by activation of AMPK signal- macaques 3 and 7 months after PMM surgery (figure 7F,G). ling. (3) To determine the role of metformin in large animals, Results of pain-related­ behaviour test also showed that the dura- we performed a pilot experiment using non-­human primates. tion of standing and walking was significantly increased by treat- We found that metformin also possesses a chondroprotective ment with metformin 7 months after PMM surgery (figure 7H). effect in non-­human primates. In addition, we also analysed the Because metformin inhibits OA development in non-­human OA progression rates during 6 to 12 weeks period after DMM primates, these findings suggest that metformin may be used to surgery. We found that, during this period, the progression rates treat patients with OA disease. of some OA features, including OARSI score, cartilage area and synovitis score, were not significantly changed after metformin Discussion administration; in contrast, the progression rates of some other We have made several key observations in the studies presented OA features, including osteophyte size and osteophyte matu- here. (1) We found that metformin, administered shortly after rity were significantly reduced after metformin administration. joint injury, limits OA development and delays OA progres- Our findings suggest that metformin may be used clinically to sion. Metformin also relieves OA-associated­ pain sensitivity treat patients with OA and that selective OA symptoms may in mice. (2) We found that metformin upregulates AMPKα1 be improved after metformin administration shortly after joint expression and that metformin lost its chondroprotective effect injury.

Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 641 Osteoarthritis

Figure 6 Metformin upregulated AMP-­activated protein kinase (AMPK) expression in the dorsal root ganglia (DRG). (A) Immunofluorescence (IF) assays were performed in histological sections of DRG tissues in the mice 6 and 12 weeks after destabilisation of the medial meniscus (DMM) surgery; expression of pAMPKα1 and total AMPKα1 was examined by IF staining. (B, C) Results of quantification analysis demonstrated that AMPKα1 expression (phosphorylated and total AMPK α1) was significantly reduced after DMM surgery. Metformin significantly increased the expression of pAMPKα1 and total AMPK α1 protein levels in DRG tissues (n=3). Statistical analysis was conducted using two-way­ analysis of variance followed by the Tukey-­Kramer test. ##p<0.01, compared between sham and DMM groups; *p<0.05, **p<0.01, compared between groups with or without metformin treatment in mice with DMM surgery. pAMPK,phosphorylated AMPK; t-AMPK,­ total AMPK; TRPA1, transient receptor potentialankyrin 1; WT, wild type.

OA is the most common form of arthritis and the leading thus, a drug, such as metformin, acting on activation of AMPK cause of chronic disability among older people. More than 50% signalling, represents a novel class of drugs to treat OA disease. of people over the age of 65 years have radiological evidence AMPK signalling has been suggested to be involved in OA of OA, with approximately 10% of men and 18% of women development27 and AMPK deficiency in chondrocytes acceler- suffering symptomatic OA.25 In a population-­based cohort study, ates the progression of OA induced by DMM surgery or ageing the lifetime risk of symptomatic knee OA was 45%.26 There in adult mice.12 However, opposite results have also been is currently no medication available to cure OA disease or to reported, showing that no significant difference in OA devel- decelerate OA progression; thus disease-­modifying OA drugs are opment was observed in chondrocyte-­specific AMPKα1 KO urgently needed to not only alleviate pain and improve function, mice.28 It has also been reported that activation of AMPK signal- but also to delay the structural progression of the disease. None ling negatively regulates chondrocyte differentiation.29 In the of drugs currently being tested in Phase II and Phase III clinical present studies, we found that the knee joint phenotype induced trials for OA treatment are aimed at targeting AMPK signalling; by DMM surgery in AMPKα1 KO mice was not significantly

642 Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 Osteoarthritis

Figure 7 Metformin alleviates partial medial meniscectomy (PMM)-induced­ osteoarthritis (OA) development in non-­human primates. (A) The PMM surgery was performed on the left knee joint and the sham operation was performed on the right knee joint in rhesus macaques with 8.5 to 11.4 years of age. (B, C) Cartilage damage was observed in rhesus macaques 7 months after the PMM surgery. The administration of metformin inhibited articular cartilage damage in rhesus macaques with PMM surgery (yellow arrows: articular cartilage damage). (D, E) Two-dimensional­ and three-­ dimensional images of micro-­computerised tomography (μCT) scans were obtained 7 months after PMM surgery (yellow arrow: subchondral sclerosis and red arrow: osteophyte). (F) Knee OA was monitored by MRI 1 month before surgery and 1, 3 and 7 months after surgery (red arrows: osteophytes; yellow arrow: subchondral sclerosis; blue arrow: knee joint medial collateral ligament). (G) The medial articular cartilage thickness including medial femoral condyles and the medial tibial plateau was quantified using MRI. Control group: rhesus macaques received PMM surgery and were treated with vehicle (n=4); experimental group: Rhesus macaques received PMM surgery and were treated with metformin (n=5). (H) Changes in animal behaviour were observed and recorded. *p<0.05, compared between groups with or without metformin treatment. Met,metformin. different from that of WT mice; however, metformin has protec- activation loop of the α subunit by its major upstream AMPK tive effect on OA progression in WT mice, but not in AMPKα1 kinase, liver kinase B1 (LKB1).30 The activity of AMPK and KO mice. These results suggest that normal levels of AMPKα1 its upstream regulator LKB1 is reduced in cartilage from aged may not be required for maintaining joint tissue homeostasis and mice and mice with OA, as well as in bovine chondrocytes after that metformin does play a chondroprotective effect through dynamic compression-­induced biomechanical injury.31 activation of AMPK signalling. The AMPKα1 KO mice used in In our in vitro studies, we found that metformin upregulated our study are AMPKα1 global KO mice, but not chondrocyte-­ pAMPKα1 expression within a short period of time, but not the specific KO mice reported in previous studies.12 28 In addition to total AMPKα1 protein. In contrast, we found that metformin mature chondrocytes, AMPKα1 gene was also deleted in mesen- significantly upregulated total AMPKα1 expression in articular chymal progenitor cells in AMPKα1 global KO mice so both chondrocytes and in DRG cells. This difference may be due to mesenchymal progenitor cells and mature chondrocytes may the duration of treatment. In vivo treatment with metformin contribute to the observations that deletion of AMPKα1 blocks was 6 and 12 weeks. The long-term­ effect of metformin on the the chondroprotective effect of metformin in the current studies. upregulation of total AMPKα1 may also contribute to the effect AMPK exists as a heterotrimeric complex composed of a cata- of metformin on OA treatment. In addition to the treatment lytic α subunit and regulatory β and γ subunits. Binding of AMP duration, the endogenous AMPK levels may also affect the effec- to the γ subunit allosterically activates the complex, making it a tiveness of metformin treatment. OA is a heterogeneous disease. more attractive substrate for phosphorylation on Thr172 in the Metformin may be specifically effective for OA patients with low

Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 643 Osteoarthritis

AMPK levels. In the future, we may further stratify OA patients Author affiliations 1 into two groups with high and low AMPK levels to further deter- Department of Orthopedic Surgery, Rush University Medical Center, Chicago, Illinois, mine the effectiveness of metformin on OA pathogenesis. USA 2Department of Orthopedics, Tianjin Medical University General Hospital, Tianjin, In this study, we also found that metformin is an effective drug China to relieve OA pain. Pain is the most common reason that OA 3Research Center for Human Tissues and Organs Degeneration, Shenzhen patients seek medical attention and is frequently a symptom of Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China underlying joint injury. AMPK activation plays an important role 4 in blocking pain sensitivity.32 33 Pain hypersensitivity resulting Department of Pharmacy, Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China from injury or disease is often paralleled by increased excit- 5School of Medicine, Southern University of Science and Technology, Shenzhen, 15 ability of peripheral sensory neurons in the DRG. The TRPA1 China channel is a widely recognised chemical and thermal sensor that 6Division of Rheumatology, Allergy and Immunology, San Diego VA Healthcare plays vital roles in pain transduction.34 Metformin, as an AMPK System, San Diego, California, USA activator, inhibited TRPA1 activity in DRG neurons by inhib- Acknowledgements We would like to express our gratitude to Ms Lily Yu for her iting membrane-­associated TRPA1 expression and metformin assistance on processing and staining histological samples and Dr Benoit Viollet also inhibited TRPA1-­mediated calcium influx.35 These findings (INSERM, U1016, Paris, France) for providing us with AMPKα1 KO mice. suggest that AMPK is a key regulator of TRPA1 channels and Contributors Conception and Design: DC, RL-­B and SF. Data Acquisition: JL, BZ, that metformin may inhibit pain sensitivity through activation W-­XL, KL, TW, DY, JH, LZ and CX. Analysis and Interpretation of Data: DC, RL-­B, JL, of AMPK signalling in DRGs. In the present studies, we found HP, GN, GX. Manuscript Preparation: DC and JL. Manuscript Revision: DC, JL, RL-­B and SF. Manuscript Final Approval: DC. that treatment with metformin significantly upregulated phos- phorylated and total AMPKα1 expression in DRG tissues and Competing interests None declared. also inhibited the increased pain sensitivity caused by DMM Patient and public involvement Patients and/or the public were not involved in surgery. These findings suggest that metformin may inhibit the design, or conduct, or reporting or dissemination plans of this research. pain-­mediating calcium channel activities through activation of Patient consent for publication Not required. AMPK signalling in DRG tissues. From our limited time-­course Ethics approval Institutional review Board of Rush University Medical Center and studies, it seems that pain relieving effect of metformin may be Tianjin Medical University General Hospital. related to its chondroprotective effect on DMM-induced­ OA Provenance and peer review Not commissioned; externally peer reviewed. pathology. Data availability statement Data are available in a public, open access The current study has several limitations. First, our study repository. The data that support the findings of this study are available within the suggests that metformin could be used clinically for young article and its Supplementary Information files or from the corresponding author patients shortly after their joint injury to limit OA development upon reasonable request. and progression. However, one critical question that we could Open access This is an open access article distributed in accordance with the not address in this study is whether metformin is able to delay Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which OA progression when OA is fully developed. Our long-term­ goal 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 is to perform a longitudinal study to investigate the effect of properly cited, appropriate credit is given, any changes made indicated, and the use metformin on OA progression and OA associated pain when is non-­commercial. See: http://creativecommons.​ ​org/licenses/​ ​by-​nc/​4.0/.​ OA is fully developed; for example, 8 and 12 weeks after DMM surgery. Second, to further determine the role of metformin in ORCID iDs OA treatment, we tested effects of metformin in non-human­ Jian Huang http://orcid.​ ​org/0000-​ ​0002-​6676-​5159 Guozhi Xiao http://orcid.​ ​org/0000-​ ​0002-​4269-​2450 primates. Although a positive effect of metformin on OA treat- Di Chen http://orcid.​ ​org/0000-​ ​0002-​9854-​8685 ment has been obtained using non-human­ primates in the pilot study, one limitation of this study is that animal numbers were References relatively small. The effect of metformin on large animals needs 1 nasri H, Rafieian-­Kopaei M. Metformin: current knowledge. J Res Med Sci to be further investigated. Third, in the presented studies, 2014;19:658–64. we have used young, healthy and male mice and male rhesus 2 Paneni F, Lüscher TF. Cardiovascular protection in the treatment of type 2 diabetes: a macaques for OA studies, to further determine the chondropro- review of clinical trial results across drug classes. Am J Cardiol 2017;120:S17–27. 3 novelle MG, Ali A, Diéguez C, et al. Metformin: a hopeful promise in aging research. tective effects of metformin, we will test the effects of metformin Cold Spring Harb Perspect Med 2016;6:a025932. on female mice, on ageing-associated­ OA models and on obese 4 chen K, Lin Z-­W, He S-­M, et al. Metformin inhibits the proliferation of rheumatoid mice. These studies will provide full spectrum of chondroprotec- arthritis fibroblast-­like synoviocytes through IGF-­IR/PI3K/AKT/m-­TOR pathway. Biomed tive effects of metformin. Pharmacother 2019;115:108875. Our studies suggest that metformin may be used clinically to 5 Teodoro JS, Nunes S, Rolo AP, et al. Therapeutic options targeting oxidative stress, mitochondrial dysfunction and inflammation to hinder the progression of vascular treat young patients shortly after their joint injury. In human complications of diabetes. Front Physiol 2018;9:1–18. clinical studies, metformin application has been demonstrated to 6 li Q, Jia S, Xu L, et al. Metformin‐induced autophagy and irisin improves INS-1 cell have a beneficial effect on long-­term knee joint outcomes in those function and survival in high-­glucose environment via AMPK/SIRT1/PGC-1α signal with knee OA and obesity.36 In contrast, a retrospective cohort pathway. Food Sci Nutr 2019;7:1695–703. study was conducted using information of 3217 patients with 7 chen D, Xia D, Pan Z, et al. Metformin protects against apoptosis and senescence type 2 diabetes in the UK. Patients had recorded type 2 diabetes in nucleus pulposus cells and ameliorates disc degeneration in vivo. Cell Death Dis 2016;7:e2441. and treatment with metformin. The outcome was the record 8 Forouzandeh F, Salazar G, Patrushev N, et al. Metformin beyond diabetes: of OA during follow-­up. The results demonstrated that treat- pleiotropic benefits of metformin in attenuation of atherosclerosis. J Am Heart Assoc ment with metformin had no significant effect on OA progres- 2014;3:e001202. sion.37 This discrepancy could be the result of different dosages 9 Zi F, Zi H, Li Y, et al. Metformin and cancer: an existing drug for cancer prevention and and treatment durations that were administered. Randomised therapy. Oncol Lett 2018;15:683–90. 10 rena G, Lang CC. Repurposing metformin for cardiovascular disease. Circulation controlled clinical trials are needed to determine whether 2018;137:422–4. metformin could be used as a potential disease-­modifying drug 11 rotermund C, Machetanz G, Fitzgerald JC. The therapeutic potential of metformin in for knee OA with or without the obese phenotype. neurodegenerative diseases. Front Endocrinol 2018;9:400.

644 Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 Osteoarthritis

12 Zhou S, Lu W, Chen L, et al. AMPK deficiency in chondrocytes accelerated the 26 Murphy L, Schwartz TA, Helmick CG, et al. Lifetime risk of symptomatic knee progression of instability-­induced and ageing-­associated osteoarthritis in adult mice. osteoarthritis. Arthritis Rheum 2008;59:1207–13. Sci Rep 2017;7:43245. 27 Yui N, Kobayashi H, Terauchi K, et al. The role of ALP-­activated protein kinase (AMPK) 13 stein SC, Woods A, Jones NA, et al. The regulation of AMP-­activated protein kinase by in the chondrocyte energy balance and IL-1β-induced production of MMP-13 in phosphorylation. Biochem J 2000;345:437–43. osteoarthritis(OA). Osteoarthritis Cartilage 2016;24:S154–5. 14 June RK, Liu-­Bryan R, Long F, et al. Emerging role of metabolic signaling in synovial 28 Yang C, Li Z, Lai P, et al. Chondrocyte-­specific ablation ofA MPKα1 does not joint remodeling and osteoarthritis. J Orthop Res 2016;34:2048–58. affect bone development or pathogenesis of osteoarthritis in mice. DNA Cell Biol 15 Terkeltaub R, Yang B, Lotz M, et al. Chondrocyte AMP-­activated protein kinase activity 2016;35:156–62. suppresses matrix degradation responses to proinflammatory cytokines interleukin-1β 29 Bandow K, Kusuyama J, Kakimoto K, et al. AMP-­activated protein kinase (AMPK) and tumor necrosis factor α. Arthritis Rheum 2011;63:1928–37. activity negatively regulates chondrogenic differentiation. Bone 2015;74:125–33. 16 Guma M, Wang Y, Viollet B, et al. AMPK activation by A-769662 controls IL-6 30 shackelford DB, Shaw RJ. The LKB1-­AMPK pathway: metabolism and growth control expression in inflammatory arthritis. PLoS One 2015;10:e0140452. in tumour suppression. Nat Rev Cancer 2009;9:563–75. 17 liu-­Bryan R, Terkeltaub R. Emerging regulators of the inflammatory process in 31 Petursson F, Husa M, June R, et al. Linked decreases in liver kinase B1 and osteoarthritis. Nat Rev Rheumatol 2015;11:35–44. AMP-­activated protein kinase activity modulate matrix catabolic responses to 18 arthritis Foundation,. Rheumatoid vs. osteoarthritis and diabetes. Available: www.​ biomechanical injury in chondrocytes. Arthritis Res Ther 2013;15:R77. arthritis.org/​ ​living-​with-​arthritis/​comorbidities/​diabetes-and-​ ​arthritis/rheumatoid-​ ​ 32 asiedu MN, Han C, Dib-­Hajj SD, et al. The AMPK activator A769662 blocks voltage-­ arthritis-vs-​ ​osteoarthritis-diabetes.​ ​php [Accessed 2 Apr 2019]. gated sodium channels: discovery of a novel pharmacophore with potential utility for 19 Diabetologia. Press release. study links diabetes to elevated risk of osteoporosis and analgesic development. PLoS One 2017;12:e0169882. arthritis. October 2, 2018. Berlin. European Association for the Study of Diabetes (EASD) 33 Tillu DV, Melemedjian OK, Asiedu MN, et al. Resveratrol engages AMPK to attenuate annual meeting. Available: www.​eurekalert.​org/​pub_releases/​ 2018-​ 10/​ ​d-sld093018.​ ​php 20 asiedu MN, Dussor G, Price TJ. Targeting AMPK for the alleviation of pathological ERK and mTOR signaling in sensory neurons and inhibits incision-­induced acute and pain. Exp Suppl 2016;107:257–85. chronic pain. Mol Pain 2012;8:5. 21 Weng W, Yao C, Poonit K, et al. Metformin relieves neuropathic pain after spinal nerve 34 story GM, Peier AM, Reeve AJ, et al. ANKTM1, a TRP-­like channel expressed in ligation via autophagy flux stimulation. J Cell Mol Med 2019;23:1313–24. nociceptive neurons, is activated by cold temperatures. Cell 2003;112:819–29. 22 The Burden of Musculoskeletal Diseases in the United States. Bone and joint decade. 35 Wang S, Kobayashi K, Kogure Y, et al. Negative regulation of TRPA1 by AMPK in 2nd edn, 2011. primary sensory neurons as a potential mechanism of painful diabetic neuropathy. 23 Zhang Y, Jordan JM. Epidemiology of osteoarthritis. Clin Geriatr Med Diabetes 2018;67:98–109. 2010;26:355–69. 36 Wang Y, Hussain SM, Wluka AE, et al. Association between metformin use and disease 24 Meng S, Cao J, He Q, et al. Metformin activates AMP-­activated protein kinase progression in obese people with knee osteoarthritis: data from the osteoarthritis by promoting formation of the αβγ heterotrimeric complex. J Biol Chem Initiative-­a prospective cohort study. Arthritis Res Ther 2019;21:127. 2015;290:3793–802. 37 Barnett LA, Jordan KP, Edwards JJ, et al. Does metformin protect against 25 Woolf AD, Pfleger B. Burden of major musculoskeletal conditions. Bull World Health osteoarthritis? An electronic health record cohort study. Prim Health Care Res Dev Organ 2003;81:646–56. 2017;18:623–8.

Li J, et al. Ann Rheum Dis 2020;79:635–645. doi:10.1136/annrheumdis-2019-216713 645 Osteoarthritis

Translational science Faecal microbiota transplantation from metabolically compromised human donors accelerates osteoarthritis in mice ZeYu Huang ‍ ‍ ,1,2 Jing Chen,1,3 BoLei Li,1,3 Benhua Zeng,4 Ching-­Heng Chou,5 Xin Zheng,1 JingWei Xie,2 Hao Li,1,3 Yu Hao,1,3 Guo Chen,2 FuXing Pei,2 Bin Shen,2 Virginia B Kraus,5,6 Hong Wei,7,8 Xuedong Zhou,1,3 Lei Cheng ‍ ‍ 1,3

Handling editor Josef S Abstract Key messages Smolen Objectives Emerging evidence suggests that the microbiome plays an important role in the pathogenesis ►► Additional material is What is already known about this subject? published online only. To view of osteoarthritis (OA). We aimed to test the two-hit­ model ►► The gut-­microbiome plays an important role in please visit the journal online of OA pathogenesis and potentiation in which one ’hit’ the pathogenesis of many different conditions, (http://dx.​ ​doi.org/​ ​10.​1136/​ is provided by an adverse gut microbiome that activates including osteoarthritis (OA). Germ-­free mice annrheumdis-2019-​ ​216471). innate immunity; the other ’hit’ is underlying joint damage. have reduced susceptibility to OA following Methods Medical history, faecal and blood samples For numbered affiliations see destabilisation of the medial meniscus, although were collected from human healthy controls (OA-­METS-, end of article. the mechanism is not clear. n=4), knee OA without metabolic syndrome (OA+METS-, Correspondence to n=7) and knee OA with metabolic syndrome What does this study add? Professor Lei Cheng, West China (OA+METS+, n=9). Each group of human faecal ► We demonstrated that (1) transplanting School of Stomatology, Sichuan samples, whose microbial composition was identified by ► University, Chengdu 610041, faecal samples from patients with metabolic 16S rRNA sequencing, was pooled and transplanted into China; syndrome (MetS) to germ-­free mice increased germ-­free mice 2 weeks prior to meniscal/ligamentous chengleidentist@​ ​163.com,​ gut permeability, endotoxemia and systemic Professor Xuedong Zhou, injury (MLI) (n≥6 per group). Eight weeks after MLI, mice low-­grade inflammation and lead to Department of Cariology and were evaluated for histological OA severity and synovitis, Endodontics, West China aggravated severity of OA induced by meniscal systemic inflammation and gut permeability. Hospital of Stomatology, ligamentous/injury surgery. (2) The abundances Results Histological OA severity following MLI was SiChuan University, Chengdu of the Fusobacterium, Faecalibacterium 610041, China; minimal in germ-free­ mice. Compared with the other and Ruminococcacea bacterial genera were zhouxd@​ ​scu.edu.​ ​cn and groups, transplantation with the OA+METS+ microbiome consistently associated with both histological Professor Hong Wei, Central was associated with higher mean systemic concentrations Laboratory, Clinical Medicine OA severity and inflammatory biomarkers, of inflammatory biomarkers (interleukin-1β, interleukin-6 Scientific and echnicalT suggesting that they might be some of the and macrophage inflammatory protein-1α), higher gut Innovation Park, Shanghai Tenth bacterial agents linked to the MetS capable of People’s Hospital, Tongji permeability and worse OA severity. A greater abundance aggravating OA severity in germ-­free mice and University, Shanghai 200435, of Fusobacterium and Faecalibaterium and lesser China; potentially their human donors. weihong63528@​ ​163.com​ abundance of Ruminococcaceae in transplanted mice were consistently correlated with OA severity and systemic How might this impact on clinical practice or ZH, JC, BL and BZ contributed biomarkers concentrations. future developments? equally. Conclusion The study clearly establishes a direct gut ► These results support treatments aimed at microbiome-­OA connection that sets the stage for a new ► HW, XZ and LC are joint senior beneficially modifying the intestinal microbiome means of exploring OA pathogenesis and potentially authors. such as a high-­fibre diet, weight loss, exercise, new OA therapeutics. Alterations of Fusobacterium, unabsorbable antibiotic and, as a last resort, For ’Presented at statement’ see Faecalibaterium and Ruminococcaceae suggest a role of microbiota transplantation. end of article. these particular microbes in exacerbating OA. Received 13 October 2019 Revised 17 February 2020 Accepted 3 March 2020 fibre, could restore the gut microbiota of obese OA Published Online First Introduction mice to that of a lean gut microbial profile, suggesting 23 March 2020 The past decade has witnessed a gradual but funda- that systemic inflammation in OA is influenced by the mental shift in our understanding of the mechanisms gut microbiome. A role of the gut microbiome has underlying osteoarthritis (OA) towards the concept been shown for several diseases with inflammatory 8 © Author(s) (or their that it is a complex multitissue pathology in which components including type 2 diabetes, non-­alcoholic 1 9 10 employer(s)) 2020. No low-­grade, chronic inflammation has a central role. steatohepatitis and cardiovascular diseases. A commercial re-­use. See rights A number of risk factors playing important roles in subtype of metabolic OA has been proposed,11 with and permissions. Published OA have been identified, including age,2 obesity,2 some studies proposing that metabolic syndrome in by BMJ. metabolic syndrome (MetS),3 joint trauma2 and OA is largely driven by body mass index (BMI).12 To cite: Huang ZY, Chen J, genetic predisposition.2 Among these factors, MetS However, the mechanisms underlying metabolic Li BL, et al. Ann Rheum Dis has attracted great research interest.4–6 Schott et al7 syndrome are still unclear. Although the gut micro- 2020;79:646–656. reported that oligofructose, a non-­digestible prebiotic biome contributes to activation of low-grade­

646 Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 Osteoarthritis inflammation in MetS conditions,13–15 the link is currently specu- Germ-­free mice transplanted with stool samples from healthy lative in OA.16 controls with MLI surgery (three male and three female); (4) Our group has proposed a two-­hit model of OA pathogen- Non-­MetS with surgery (OA+METS-/MLI+): Germ-free­ mice esis in which, as one hit, factors associated with an adverse gut transplanted with stool samples from patients with OA without microbiome activate the innate immune response. The other hit MetS with MLI surgery (6 male and six female); (5) MetS with results from joint damage that releases damage associated with surgery (OA+METS+/MLI+): Germ-free­ mice transplanted molecular pattern molecules, such as hyaluronan and fibronectin with stool samples from patients with both OA and MetS with fragments that result in a synergistic activation of inflammatory MLI surgery (six male and six female) (figure 1A). Twelve SPF pathways and the inflammasome leading to OA.16 The goal of the mice (six male and six female) underwent MLI at the age of current study was to evaluate our two-­hit hypothesis model of 10 weeks to serve as a SPF/MLI+control group (figure 1A). As OA pathogenesis. We hypothesised that an adverse microbiome germ-­free mice are very sensitive to anaesthesia, one male mouse would exacerbate the histopathological severity of OA induced in the OA+METS+/MLI+ group died postoperatively. We by joint injury in a murine model. To generate a ‘first-hit’,­ we were unable to retrieve faecal samples from the gut of one male transplanted these mice with human faecal samples from healthy mouse from the OA+METS-/MLI+ group and thus were unable controls, or individuals with knee OA without MetS, or knee OA to obtain the 16S rRNA gene sequencing data for this animal. with MetS. To generate a ‘second-hit’,­ we created a meniscal/ In order to provide a reference, we also evaluated the level of ligamentous injury (MLI) in murine knee joints. histological knee OA of three age-­matched naïve germ-­free mice (one male and two female). Methods Study cohort, patient characteristics and sample collection Microbiome transplantation Human blood sample collection and all experimental procedures For microbiota transplantation, frozen human faecal samples were approved by the Institutional Review Board of SiChuan were pooled from donors in the same group, were pulverised University (WCHSIRB-D-2017–111),­ in accordance with the with a mortar and pestle, then resuspended with sterile saline guidelines provided by the Health Sciences Authority of China. and centrifuged to obtain the supernatant. At 8 weeks of age, Informed written consent was obtained from participants for germ-­free mice were orally inoculated (200 µL for each mouse) this study in accordance with Declaration of Helsinki. MetS two times a daily for 2 weeks with the prepared faecal contents was defined according to criteria developed by the International from the different groups of donors. Recipient mice transplanted 17 Diabetes Federation (IDF) (see online supplementary methods). with microbiota were kept in different Trexler-type­ film isola- Primary symptomatic knee OA for this study was defined by the tors, fed with sterile food and water, with bacterial contamina- 18 American College of Rheumatology criteria, together with tion strictly controlled. radiographic severity grade III or IV of the Kellgren/Lawrence scale.19 Based on these criteria, participants (all female) were classified as: (1) Healthy controls without evidence of OA or MLI surgery MetS (OA-METS-,­ n=4); (2) OA without MetS (OA+METS-, MLI was performed at age 10 weeks after successful faecal n=7) and (3) OA with MetS (OA+METS+, n=9). transplantation of mice. MLI was induced by transecting the Stool samples were freshly collected from each participant medial collateral ligament and detaching the anterior horn of within 15 min after excretion, were immediately frozen and stored the meniscus in the right knee, as described previously.20 The in liquid nitrogen. Fasting plasma and serum were collected on the animals were sacrificed 8 weeks after MLI with collection of day of admission and stored at −80℃ for further analysis. knee joints, blood, small intestine, colon, peritoneal fat tissue and stool samples. Experimental animals and study design Germ-­free C57BL/6J mice were bred at the Department of Histologic assessment of OA severity Laboratory Animal Science of the Third Military Medical All right knees were evaluated histologically for cartilage lesions, University in Chongqing, China and housed under a 12 hours cartilage proteoglycan content and synovitis (as described in light-­dark cycle in the gnotobiotic facilities. All mice were fed online supplementary methods) by two evaluators (XJW and with sterile food and water ad libitum, and bacterial contami- CG) blinded to the group assignment (all mice were given a nation was monitored by periodic examination of stools. Age-­ numerical code). matched C57BL/J mice were born, raised and maintained in a specific-­pathogen-­free (SPF) animal facility at the same institu- tion with the same light-­dark cycle, food and water. All animal Multivariable regression analysis of predictors of histological experiments were approved by the SiChuan University Animal outcomes Care and Use Committee and followed the recommendations In order to assess the effects of both sex and body weight on from the Guide for the Care and Use of Laboratory Animals of the histological outcomes, we performed multivariable regres- Chinese Association for Laboratory Animal Science. sion analysis with covariates of group, sex and weight (weight Forty-two­ germ-­free mice (8 weeks old) underwent faecal at start, weight at 18 weeks or ∆weight defined as weight at 18 transplantation over 2 weeks with or without MLI surgery weeks minus weight at start). 2 weeks later; they were divided into five groups as follows: (1) Germ-­free with surgery (SALINE/MLI+): Germ-free­ mice transplanted with sterile normal saline with MLI surgery (three Plasma analysis for inflammatory biomarkers male and three female); (2) Healthy with no surgery (OA-­ EDTA plasma samples were collected from all mice. Details on METS-/MLI-): Germ-­free mice transplanted with stool samples the ELISA assay for inflammatory biomarkers (G-CSF,­ IL-1β, from healthy controls without MLI surgery (three male and IL-6, IL-10, IL-17, IP-10, MCP-1 and MIP-1α) and the LPS three female); (3) Healthy with surgery (OA-­METS-/MLI+): detection assay are provided in online supplementary methods.

Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 647 Osteoarthritis

Figure 1 Gut microbiome profiles of the cohort and recipient mice. (A) Schematic representation of faecal microbiota transplantation. Germ-­ free mice are orally inoculated with prepared faecal contents from three different groups of people (OA-METS-,­ OA+METS and OA+METS+). After the success of gut microbiome colonisation, MLI was induced in mice from the SALINE/MLI+, SPF/MLI+, OA-METS-/MLI+,­ OA+METS-/MLI+ and OA+METS+/MLI+ groups. After 8 weeks, the mice were sacrificed for further analysis. (B) Alpha diversity analysis showed no significant differences between the people from three different groups. (C) Beta diversity of faecal microbiome of people from three different groups. (D) NMDS analysis showed significant differences between Health and OA with MetS groups. (E) Decreased Alpha diversity observed in mice from the OA+METS+/MLI+ group, compared with mice from the SALINE/MLI+, OA-­METS-/MLI-, OA-­METS-/MLI+ and OA+METS-/MLI+ groups. (F) Beta diversity analysis showed that enterotype of mice from the OA-­METS-/MLI- and OA-­METS-/MLI+ groups was clustered together, enterotype of mice from the OA+METS-/MLI+ group was clustered together while enterotype of mice from the OA+METS+/MLI+ group was clustered together. (G) Community bar-plot­ analysis showed relative abundance of faecal microbiota in each group at the genera level. (H) Seven kinds of bacteria showed significant differences in the genera level between mice from the OA-­METS/MLI+, OA+METS-/MLI+ and OA+METS+/MLI+ groups. Data were compared by the Kruskal-Wallis­ test (p values were adjusted by the Bonferroni correction). Each bar represents a group (x-axis)­ with mean and 95% CI (y-­axis). METS, metabolic syndrome; MLI, meniscal/ligamentous injury; NMDS, non-­metric multidimensional scaling; OA, osteoarthritis; SPF, specific-pathogen-­ ­free.

16S rRNA gene sequencing, bioinformatics and statistical Gene expression of tight junction molecules analysis Colonic tissue was harvested from mice for mRNA expression of Total DNA was isolated, amplified and sequenced according to tight junction proteins (ZO-1 and occludin) by RT-­PCR (details standard procedures.21 22 The determination of bacterial species provided in online supplementary methods). is described in online supplementary methods. Statistical analyses FISH analysis for bacterial translocation in the gut All statistical analyses were performed using SPSS 22.5 (SPSS, Evaluation for bacterial translocation via the gut endothelium was Chicago, Illinois, USA) unless otherwise specified. Descriptive evaluated by fluorescence in situ hybridisation (FISH) by modifica- statistics are presented as means±SD for continuous variables. 23 tion of a previous protocol (see online supplementary methods). Normal distribution was tested using the Kolmogorov-­Smirnov test. Normally distributed data from multiple groups were Peritoneal fat gene expression compared by one-way­ analysis of variance and all other data Peritoneal fat from each mouse was harvested and processed for were compared by the Kruskal-­Wallis test. When multiple RNAseq analyses as described in online supplementary methods. comparisons were fewer than 10, p values were adjusted using

648 Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 Osteoarthritis

Seven of the bacterial genera, including Anaerostipes, Butyr- Table 1 Human donor demographics icicoccus, Erysipelatoclostridium, Faecalibacterium, Fusobacte- OA without rium, Lachnoclostridium and Ruminococcaceae_UCG-013, were Healthy MetS OA with MetS significantly different between mice from the OA-METS-/MLI+­ Demographics (n=4) (n=7) (n=9) P value and OA+METS+/MLI+ groups (figure 1H), others are shown Gender (F/M) 4/0 7/0 9/0 1 in online supplementary figure S1. Age (years) 64.3±9.9 61.7±6.2 68.8±8.1 0.244 BMI (kg/m2) 22.19±1.37 21.61±2.50 29.58±2.82 <0.001 Waist circumference 72.2±1.7 73.1±3.2 83.9±3.0 <0.001 Highest histological joint damage observed in the (cm) OA+METS+/MLI+ group LDL (mmol/L) 2.86±0.79 2.89±0.56 3.83±0.67 0.017 For ease of comparison, histological scores (mean, SD) for each HDL (mmol/L) 1.96±0.41 1.86±0.64 1.28±0.42 0.034 group are presented in figure 1A; they are also depicted in bar TG (mmol/L) 1.12±0.32 0.81±0.44 1.85±0.85 0.015 graph form in figure 2B,C and E, with p values presenting the CHOL (mmol/L) 4.57±0.94 4.89±0.81 4.12±1.94 0.588 differences across all main groups (data for naïve germ-­free mice P < 0.05 in bold. provided for reference only). BMI, body mass index; CHOL, cholesterol; F, female; HDL, high-­density lipoprotein; Minimal signs of OA and synovitis were found in the age-­ LDL, low-­density lipoprotein; M, male; MetS, metabolic syndrome; OA, osteoarthritis; matched naïve germ-free­ mice (OARSI score: 2.0±1.0; Safranin TG, triglyceride. O score: 0.7±0.7; Synovitis score: 0.3±0.6) (figure 2B,C and E). Mean OARSI cartilage damage scores were significantly higher in the OA+METS+/MLI+ group (18.7±1.6) in comparison with the Bonferroni correction. Otherwise, p values were adjusted by the other groups. Mice from the OA+METS-/MLI+ (14.3±3.6), false discovery rate (FDR) correction according to the Benjamini-­ OA-METS-/MLI+­ (14.0±2.8) and SPF/MLI+ (12.7±3.0) groups Hochberg procedure. The Shannon index at the genera level was had significantly higher mean OARSI scores than mice from calculated with QIIME (V.1.7.0). Principle component analysis the SALINE/MLI+ (4.8±1.2) and OA-­METS-/MLI- (5.0±2.0) (PCA) was performed using the FactoMineR packages and clus- groups. There were no significant differences between SALINE/ terSim package in R software (V.2.15.3). PLS-DA­ was performed MLI+ and OA-METS-/MLI-­ groups. Also, no significant differ- using SIMCA-P­ software to cluster the sample plots across ences were found among SPF/MLI+, OA-METS-/MLI+­ and groups. For correlation of inflammatory biomarker levels with OA+METS-/MLI+ groups (figure 2A,B). Similar results were OA severity scores, Spearman correlations were used. To test the observed for cartilage safranin O scores indicative of proteo- null hypothesis, we defined the significance level α as 0.05. A glycan content; the mice from the OA+METS+/MLI+ group p<α was considered statistically significant. had a significantly higher mean score (11.6±0.5) compared with those from SALINE/MLI+ (2.3±0.8, p=0.003), SPF/MLI+ Results (8.6±1.2, p=0.004), OA-METS-/MLI-­ (2.5±1.0, p=0.003), Cohort demographics OA-METS-/MLI+­ (9.3±1.0, p=0.034) and OA+METS-/MLI+ For murine transplantation, a total of 20 human female study (9.4±1.9, p=0.03) groups (figure 2A,C). participants provided faecal samples including 4 healthy controls The mean synovitis score of the OA+METS+/MLI+ group without knee OA (OA-METS-),­ 7 with knee OA without MetS (4.1±0.9) was significantly higher than the mean scores of the (OA+METS-) and 9 with knee OA with MetS (OA+METS+). SALINE/MLI+ (0.5±0.5, p=0.005), SPF/MLI+ (1.8±1.0, Baseline characteristics of age and total plasma cholesterol of the p=0.005), OA-METS-/MLI-­ (0.3±0.5, p=0.005), OA-METS-/­ three groups were similar (table 1). However, the OA+METS+ MLI+ (1.7±1.0, p=0.015) and OA+METS-/MLI+ (1.8±1.2, group had significantly higher mean values for BMI, low-density­ p=0.008) groups (figure 2D,E). The mean synovitis scores lipoprotein cholesterol and triglyceride level (table 1) and signifi- of these latter groups, (SALINE/MLI+, OA-METS-/MLI-,­ cantly lower mean high-­density lipoprotein level (table 1). Meta- OA-METS-/MLI+­ and OA+METS-/MLI+), did not differ bolic syndrome criteria for each donor are provided in online significantly. supplementary table S1. Multivariable regression analysis of histological outcomes Gut microbiome profiles of the cohort and recipient mice There were no significant differences in weight at 8 weeks (study By 16S rRNA gene sequencing, there were no significant differ- start), weight at 18 weeks or ∆weight among the six groups (see ences of alpha diversity among the three groups of donor online supplementary table S2 and figure S2). By multivariable microbiomes (figure 1B). By non-­metric multidimensional regression analyses, group was strongly associated with OARSI, scaling analysis, there was a significantly different beta diver- Safranin O and synovitis scores. However, sex, weight at start, sity of the faecal microbiomes of the three groups (p=0.045) weight at 18 weeks and ∆weight were not associated with the (figure 1C,D). The alpha diversity at a genus level was much histological outcomes (see online supplementary table S3). lower in the germ-­free mice transplanted with faecal matter from the OA+METS+/MLI+ group (figure 1E). PCA using Jensen-­ Shannon distance was performed to cluster the 34 samples into Inflammatory biomarkers analysis three distinct enterotypes (figure 1F). At the genus level, a total of OA+METS+/MLI+ had the highest mean plasma G-­CSF, IL-1β, 140 genera were identified among all samples; the five dominant IL-6, IL-10, IL-17, IP-10, MCP-1, MIP-1α. After FDR correc- genera were Bacteroides, Blautia, Parabacteroides, Lachnoclos- tion, compared with the other five groups, the OA+METS+/ tridium and Turicibacter (figure 1G). At the genus level, the rela- MLI+ group had statistically significant higher mean plasma tive abundance of 10 bacteria was significantly different in the IL-1β, IL-6 and MIP-1α (figure 3) (see online supplementary faecal samples collected from mice of the healthy with surgery table S2). Mean IP-10 was significantly lower in the SALINE/ (OA-METS-/MLI+),­ Non-­MetS with surgery (OA+METS-/ MLI+ group compared with the other five groups (figure 3) MLI-) and MetS with surgery (OA+METS+/MLI+) groups. (see online supplementary table S2). There were no significant

Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 649 Osteoarthritis

Figure 2 Histology assessments of osteoarthritis severity and synovitis 8 weeks after the surgery. (A) Representative images of safranin O/fast green section of the mice knee joint from different groups (Scale bar in the left panels, 200 µm; Scale bar in the right panels, 50 µm). (B) Comparison of OARSI scores between the groups. (C) Comparison of Safranin O scores between the groups. (D) Representative images of H&E section showing synovitis in mice from different groups (arrows show synovitis) (Scale bar, 50 µm). (E) Comparison of synovitis scores between the groups. With the exception of the naïve germ-­free mice (provided for reference only as indicated by a vertical dashed line to the right of this group in panels B, C and E), data for all other groups were compared by the Kruskal-Wallis­ test, multiple comparison p values were adjusted by the Benjamini-Hochberg­ procedure. Each bar represents a group (x-­axis) with mean and 95% CI (y-­axis). Green dotted lines are references for mean respective histological scores of naïve germ-­free mice. METS, metabolic syndrome; MLI, meniscal/ligamentous injury; OA, osteoarthritis; OARSI, Osteoarthritis Research Society International.

650 Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 Osteoarthritis

Figure 3 Inflammatory biomarkers analysis. (A) G-­CSF. (B) IL-1β. (C) IL-6. (D) IL-10. (E) IL-17. (F) IP-10. (G) MCP-1. (H) MIP-1α. Data were compared by the Kruskal-­Wallis test, and multiple comparison p values were adjusted by the Benjamini-Hochberg­ procedure. Each bar represents a group (x-­ axis) with mean and 95% CI (y-­axis). G-­CSF, granulocyte-­colony stimulating factor; IL-1β, interleukin one beta; IL-6, interleukin 6; IL-10, interleukin 10; IL-17, interleukin 17; IP-10, interferon gamma-­induced protein 10; METS, metabolic syndrome; MCP-1, monocyte chemoattractant protein 1; MIP-1α, macrophage inflammatory protein α1 ; MLI, meniscal/ligamentous injury; OA, osteoarthritis. differences of mean plasma IL-10 or lipopolysaccharide binding in the SALINE/MLI+ group, p=0.002, Normalised Enrich- protein among the six groups. ment Score (NES)=−2.12, FDR q=0.010), lipids and lipopro- teins metabolism (highly activated in the SALINE/MLI+ group, Gut permeability p<0.001, NES=−1.94, FDR q=0.029), leucocyte transendo- According to the two-­hit hypothesis, MetS can increase gut thelial migration (highly activated in the OA+METS+/MLI+ permeability, causing systematic low-grade­ inflammation. In the group, p<0.001, NES=1.88, FDR q=0.231) and the TGF-β current study, we examined gut permeability by mRNA expres- signalling pathway (highly activated in the OA+METS+/MLI+ sion of intestinal tight junction proteins (ZO-1 and occludin) in group, p=0.011, NES=1.76, FDR q=0.165) (figure 5A–D). colon tissue, plasma lipopolysaccharide (LPS) level and bacte- rial translocation by fluorescence FISH analyses of the gut Correlations between gut bacterial genera, cartilage endothelium. As expected, significantly lower mRNA levels of histology scores and inflammatory biomarkers ZO-1 and occludin were detected in the OA+METS+/MLI+ OARSI scores reflecting overall severity of knee OA were signifi- group compared with the other five groups (figure 4G,H). cantly positively correlated with the abundance of five different Also, OA+METS+/MLI+ group had significantly higher mean gut bacterial genera and significantly negatively correlated with plasma LPS level compared with the other five groups (figure 4F) eight other gut bacterial genera (figure 5E) (see online supplemen- with evidence of bacterial translocation via the gut endothelium tary table S5). Safranin O scores reflecting cartilage proteoglycan confirmed by FISH (figure 4A–E). loss were significantly positively correlated with four different gut bacterial genera and significantly negatively correlated with Differential expression and gene set enrichment analyses of eight other gut bacterial genera (figure 5E). Specifically, Candi- peritoneal fat tissue datus_Stoquefichus, Faecalibacterium, Tyzzerlla_4 and Fusobac- We identified 171 differentially expressed genes (DEGs) in terium were consistently significantly positively correlated with peritoneal fat tissue between the OA+METS+/MLI+ and cartilage histology scores (higher scores worse) while Erysipela- SALINE/MLI+ groups (see online supplementary table S4); toclostridium, Eubacterium, Marvinbryantia, Bacteroides and 101 DEGs were upregulated and 70 DEGs downregulated in Ruminococcaceae_UCG-013 were consistently significantly the OA+METS+/MLI+ group. To obtain a deeper insight into negatively correlated with cartilage histology scores. the function of the DEGs, gene set enrichment analysis (GSEA) Plasma concentrations of IL-1β, IL-6 and MIP-1α were consis- was used. DEGs between the two groups were highly enriched tently significantly positively correlated with the abundance of in pathways related to butanoate metabolism (highly activated Fusobacterium and Faecalibacterium and negatively correlated

Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 651 Osteoarthritis

Figure 4 Germ-­free mice transplanted with faecal samples from patients of OA with metabolic syndrome showed increased gut permeability. (A–E) Representative images of gut barrier visualised by fish (16S rRNA genes of all bacteria (red) and nuclei (green)) (left panels original magnification: ×60; right panels original magnification: ×240). Bacteria invasion into colonic mucosa was captured in mice from the OA+METS+/MLI+ group (E) (marked by asterisk). No signs of bacteria invasion were observed in mice from the SPF/MLI+ (A), OA-METS-/MLI-­ (B), OA-­METS-/MLI+ (C), OA+METS-/MLI+ (D) groups. (F) Significantly increased systemic plasma LPS levels were observed in mice from the OA+METS+/MLI+ group. (G) Significantly decreased mRNA level of tight junction protein ZO-1 was observed in mice from the OA+METS+/MLI+ group, compared with mice from the SALINE/MLI+ (p<0.01), SPF/MLI+ (p<0.01), OA-­METS-/MLI- (p=0.036), OA-­METS-/MLI+ (p=0.036) and OA+METS-/MLI+ (p=0.036) groups. (H) Significantly increased mRNA level of tight junction protein Occludin was preserved in mice from the SALINE/MLI+ (p<0.01), SPF/MLI+ (p<0.01), OA-­ METS-/MLI- (p=0.02), OA-­METS-/MLI+ (p=0.034) and OA+METS-/MLI+ (p=0.048) groups, compared with those from the OA+METS+/MLI+ group. **Stands for p<0.01, *stands for 0.01≤p<0.05. Data were compared by the Kruskal-Wallis­ test, multiple comparison p values were adjusted by the Benjamini-­Hochberg procedure. Each bar represents a group (x-­axis) with mean and 95% CI (y-­axis). METS, metabolic syndrome; MLI, meniscal/ ligamentous injury; OA, osteoarthritis; SPF, specific-­pathogen-­free. with the abundance of Ruminococcaceae_UCG-013 (figure 5F) Our study results support this hypothesis showing a noticeable (online supplementary table S5). Thus, these three gut bacterial impact of the human gut microbiome on injury induced OA genera, Fusobacterium, Faecalibacterium and Ruminococcaceae_ severity in a mouse model system. The MetS+ mice developed UCG-013, showed congruent correlations with both inflamma- increased gut permeability (by colonic epithelial FISH analyses) tory biomarkers and cartilage histology scores. with endotoxemia (elevated LPS) in association with upregula- tion of circulating proinflammatory molecules consistent with Discussion an activated innate immune system. Cani et al24 first showed that We previously hypothesised that the gut microbiome plays a crit- endotoxemia triggers the expression of inflammatory factors ical role in the two-hit­ model of MetS related OA pathogenesis.16 and functions as a stimulus of metabolic syndrome. Several of

652 Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 Osteoarthritis

Figure 5 GSEA of the RNA sequencing data of peritoneal fat tissue between OA+METS+/MLI+ versus SALINE/MLI+ mice (A–D). (A) GSEA for butanoate metabolism. (B) GSEA for metabolism of lipids and lipoproteins. (C) GSEA for leucocyte transendothelial migration. (D) GSEA for TGF-β signalling pathway. The enrichment plot on the left of each panel shows the distribution of genes in the set that is correlated with the SALINE/MLI+ or OA+METS+/MLI+ groups. The heatmap on the right of each panel shows where gene expression is relatively high (red) or low (blue) for each gene in the indicated sample. Correlation between gut bacterial genera, cartilage histology scores and inflammatory biomarkers (E–F). (E) Correlations between abundance of bacterial genera and cartilage histology scores. (F) Correlations between the abundance of bacterial genera and inflammatory biomarkers. **Stands for p<0.01, *stands for 0.01≤p<0.05. Positive correlation: red, negative correlation: green. Spearman correlations were performed. GSEA, gene set enrichment analysis; METS, metabolic syndrome; MLI, meniscal/ligamentous injury; OA, osteoarthritis.

Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 653 Osteoarthritis the plasma cytokines elevated in the MetS+ mice in our study, OA histologic severity and inflammatory biomarkers. The abun- IL-1β, IL-6 and MIP-1α, are considered key mediators of OA dance of Fusobacterium and Faecalibacterium (both increased in in patients.25–27 Emerging evidence suggests that IL-1β,28 29 abundance in OA+METS+/MLI+) was consistently positively IL-630 31 and MIP-1α32 play important roles in lipid metabo- correlated while Ruminococcaceae (decreased in abundance in lism and persistence of low-­grade inflammation. In the context OA+METS+/MLI+) was consistently negatively correlated of joint injury, the activated innate immune system would be with OARSI scores, Safranin O scores, IL-1β, IL-6 and MIP-1α. expected to exacerbate the pathological process of OA. Despite Interestingly, Fusobacterium has been associated with abnormal joint injury, the germ-­free mice (SALINE/MLI+) in our study metabolic status, with enrichment in obese Japanese45 and had signs of mild OA that were reduced compared with the SPF/ Chinese46 populations, in patients with MetS47 and sows (pigs) MLI+ group (62% reduction in OARSI score and 73% reduc- with MetS.48 One of the most famous and notorious members tion in Safranin O score) and the OA-MetS-/MLI+­ group (66% from Fusobacterium genus, Fusobacterium nucleatum, has been reduction in OARSI score and 75% reduction in Safranin O associated with the occurrence of multiple conditions including score). This is consistent with Ulici et al33 who demonstrated inflammatory bowel disease,49 colorectal cancer50 and rheuma- that germ-­free DMM+ mice had reduced OA compared with toid arthritis.51 In contrast, Faecalibacterium, especially Faecali- SPF-­DMM mice (47% reduction in sum articular cartilage struc- bacterium prausnitzii, is decreased in many chronic pathological ture score and 42% reduction in sum Safranin O score). Taken conditions and generally considered a good organism with together, we show for the first time, to our knowledge, a direct versatile metabolic abilities.52 However, we observed a positive gut microbiome-­OA connection whereby the microbiome from correlation of abundance of Faecalibacterium with OA histo- different groups of people could change the pathological process logical severity and inflammatory biomarkers. Our results are of OA induced by surgery in mice. supported by van Dijkhuizen et al who demonstrated signifi- In order to evaluate the metabolic status of our mice, we also cantly greater abundance of Faecalibacterium in patients with performed RNA sequencing of the peritoneal fat tissue. The juvenile idiopathic arthritis compared with controls.53 There are GSEA analysis revealed highly activated butanoate (or butyrate), several potential reasons underlying these differences in results. lipid and lipoprotein metabolism pathways in the germ-­free We only used 16S rRNA sequencing to identify the genus of SALINE/MLI+ group. Butanoate or butyrate is a group of short different bacteria; this might not specifically identify the species. chain fatty acids produced through fermentation of dietary Faecalibacterium prausnitzii is the only species isolated from this fibres in the lower intestinal tract.34 Colonocytes use bacteri- genus until now;54 we still know nothing about the biological ally produced butyrate as their primary energy source; butyrate behaviour of the other members of the Faecalibacterium family. prevents deficits in mitochondrial respiration and autophagy.35 Even among the Faecalibacterium prausnitzii species, there are The expressions of GPR109, a butyrate receptor, and SLC5A8, different subspecies; differing proportions of these subspecies a butyrate transporter, are markedly reduced in the ileum and could result in differences in metabolite production such as colon of germ-­free mice.36 Thus, a germ-­free mouse devoid of a short-­chain fatty acid production. Second, the positive correla- gut microbiome producing butyrate could increase the capacity tion between Faecalibacterium abundance, OA histological of the gut endothelial cells to oxidise n-butyrate,­ as observed in severity and inflammatory biomarkers may be an effect rather this study as a form of rescue or countermeasure to butyrate scar- than a cause of inflammation in the donors. city (supported by Cherbuy et al37). The concurrent elevation of Ruminococcaceae was protective in our model system and lipid and lipoprotein metabolism in germ-­free mice,38 39 provide reduced in abundance in the OA+METS+/MLI+ group. one explanation for the resistance of germ-free­ C57BL/6J mice Decreased abundance of Ruminococcaceae has been found to high-­fat-­diet induced obesity.38 in patients with MetS55 and non-­alcoholic steatohepatitis.56 Moreover, the GSEA analysis revealed highly activate leuco- Ruminococcaceae can mediate the protective effects of dietary cyte transendothelial migration and TGF-β signalling pathways capsaicin against chronic low-grade­ inflammation and associ- in the OA+/METS+/MLI+ group. This finding might reflect ated obesity induced by a high-fat­ diet.57 Its protective effect elevated low-­grade inflammation, which is consistent with against low-grade­ inflammation was attributed to its production our biomarker findings and endotoxemia. TGF-β can regu- of butyrate, serving as a primary energy source for colonic intes- late multiple types of immune cells including T cells, B cells, tinal bacteria and a beneficial factor for gut health by causing macrophage, dendritic cells and natural killer cells.40 Addition- decreased intestinal permeability.58 ally, TGF-β signalling is highly activated during tissue repair Recently, Boer et al59 found in the Rotterdam Study that processes such as OA.41 42 Thus, the highly activated TGF-β Streptococcus species are associated with increased knee pain signalling found in the OA+METS+/MLI+ group might repre- and inflammation. In the current study, compared with the other sent elevated low-­grade inflammation and the highly activated groups, the relative abundance of Streptococcus was increased in tissue repair processes. both the OA+MetS+ humans and OA+MetS+/MLI+ mice (see In the current study, although we did not find significant differ- online supplementary figure S3). However, the differences in our ences of alpha diversity among the three human donor groups, donors were not statistically significantly different (p=0.42) and the mice gut microbiome profile differed significantly in alpha of borderline significance in mice (p=0.053). The gut micro- diversity following transplantation. As shown by Turnbaugh et biome is affected by many factors including diet, ethnicity and al,43 only 77% of the human gut microbiome can colonise germ-­ lifestyle.60 The population of the Rotterdam Study was primarily free mice, likely related to differences in diet, environment and Caucasian of European ancestry; in contrast, our human micro- small-­intestine:colon length ratio. Though humans and mice biome donors were all Han Chinese. This might in part explain share more than 60% of their gut microbiome genera, certain why we did not find similar results. Study power may also have genera, such as Mitsuokella, Megasohera and Dialister, are exclu- contributed to this difference since the Rotterdam study included sive to humans44 and therefore unlikely to be able to colonise n=1427 (with replication in n=867) participants; given the germ-­free mice. trends in our OA+/MetS+ humans and OA+MetS+/ML+ To identify the bacteria responsible for aggravating OA induced mice, with larger sample sizes we may have been able to detect by MLI, we evaluated the correlation of gut bacterial genera, an association of Streptococcal species with OA outcomes.

654 Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 Osteoarthritis

There are several limitations to this study. First, the sample Author affiliations 1 size of the human study was relatively small and all donors were State Key Laboratory of Oral Diseases and National Clinical Research for Oral Chinese. Therefore, our results may only reflect the specific Diseases, West China Hospital of Stomatology, Sichuan University, Chengdu, China 2Department of Orthopaedic Surgery, West China Hospital, West China Medical population tested. Second, we did not include an OA-MetS+­ School, Sichuan University, Chengdu, China group; thus, we could not evaluate the single effect of MetS 3Department of Cariology and Endodontics, West China Hospital of Stomatology, on OA. Third, some studies have shown that sex has effects in Sichuan University, Chengdu, China 4 surgical mouse models.61 Although we used both female and Department of Laboratory Animal Science, College of Basic Medical Sciences, Army Medical University, Chongqing, China male germ-­free mice in the current study, we used only female 5Duke Molecular Physiology Institute, Duke University School of Medicine, Duke human donors; thus, we cannot rule out the effect of sex on our University, Durham, North Carolina, USA study. However, the multivariable regression showed that sex 6Division of Rheumatology, Department of Medicine, Duke University School of effects on the histological outcomes in the current study were Medicine, Duke University, Durham, North Carolina, USA 7 not statistically significant. Fourth, because of the constrained Central Laboratory, Shanghai Tenth People’s Hospital, Tongji University, Shanghai, China spaces of the germ-­free containers during sacrifice, we could 8State Key Laboratory of Agricultural Microbiology, College of Animal Sciences and not collect other tissues, including synovial fluid or intrapatellar Technology, Huazhong Agricultural University, Wuhan, China fat pad to explore their differences between groups. However, 5 Presented at according to Barboza et al, the intrapatellar fat pad is not the This work was previously presented at OARSI 2019 (Oral Presentation Contol No. initiating factor of metabolic induced knee OA nor does it well 230). represent the metabolic status locally in the joint. Fifth, due Acknowledgements ZYH would like to thank Li Li, Fei Chen and Chunjuan Bao to the limited resolution of the 16S rRNA sequencing meth- from Laboratory of Pathology, West China Hospital, SiChuan University for helping odology, we were unable to identify specific species or strains with processing the histological sections and staining. ZYH would also like to thank of Fusobacterium, Faecalibacterium or Ruminococcaceae that Dr Ke Xiao from Department of Orthopaedic Surgery, West China Hospital for helping with drawing the schematic diagram in figure A1 and Dr John Martin from could be driving the correlations between OA histological Department of Orthopaedic Surgery, Duke University for his previous suggestions on severity and inflammatory biomarkers. Sixth, our results support the manuscript revision. the Koch’s second postulate of infectious disease in this ‘non-­ 62 Contributors ZYH, JC, BLL, WH, LC and XDZ conceived of the studies and planned communicable’ disease and cannot unequivocally establish the the experimental design. ZYH, JC, BLL, BHZ, JWX, GC, YH, HL, FXP and BS performed causality between these bacteria and the occurrence of OA. To the experiments and analysed the data. CHC performed the analysis on the gene prove Koch’s first and third postulates, the OA-­gut microbiome expression data. ZYH wrote the manuscript, VBK assisted with data interpretation connections need to be evaluated in more people with OA and and performed critical revisions on the manuscript. ZYH, JC, BLL and LC take responsibility for the integrity of the work as a whole. All authors approved the final the suspect microbes need to be transplanted into germ-free­ mice manuscript. to understand how they are associated with disease in humans. Funding This work was supported by grants from the NSFC to ZYH (NSFC: Seventh, though the metabolic and inflammatory status changed 81702185; NSFC: 81972097), LC (NSFC: 81870759), XDZ (NSFC: 81430011; in the mice after the MetS faecal sample transplantation, we did NSFC: 81991501) and HW (NSFC: 81770434). This research was also supported not observe any differences in body weight (weight at 8 weeks by National Key Research and Development Program of China to HW (No. (study start), weight at 18 weeks or ∆weight) or their association 2018YF2000504), SiChuan Science and Technology Programs to ZYH (No. with histological outcomes. Previously, Ussar et al63 studying the 2018HH0071) and LC (No. 2017JQ0028) and a Claude D. Pepper Older Americans Independece Centers grant (5P30 AG028716 from NIA to VBK). interactions between the gut microbiome, host genetics, diet and obesity, found that all three elements were necessary for weight Competing interests None declared. change. In the current study, we used only a normal diet in all Patient and public involvement Patients and/or the public were not involved in groups and sacrificed mice 8 weeks after confirmed colonisa- the design, or conduct, or reporting, or dissemination plans of this research. tion. Parséus et al64 showed that weight differences caused by Patient consent for publication Not required. different gut microbiome colonisation, even fed with a high fat Provenance and peer review Not commissioned; externally peer reviewed. diet, would take as long as 10 weeks. Finally, the goal of our Data availability statement Data are available on reasonable request. study was to evaluate the effects of human microbiota on trauma induced OA in the mouse. We therefore cannot speculate on ORCID iDs the effects of the mouse microbiome on OA in this study. We ZeYu Huang http://orcid.​ ​org/0000-​ ​0002-​0456-​8379 Lei Cheng http://orcid.​ ​org/0000-​ ​0002-​2762-​4740 also believe these results merit further analysis of the role of key microbiota in human OA. References In summary, we demonstrated that (1) OA severity induced by 1 robinson WH, Lepus CM, Wang Q, et al. Low-grade­ inflammation as a key mediator of MLI surgery was reduced in germ-­free mice; (2) faecal transplan- the pathogenesis of osteoarthritis. Nat Rev Rheumatol 2016;12:580–92. tation from patients with MetS to germ-­free mice could increase 2 Johnson VL, Hunter DJ. The epidemiology of osteoarthritis. Best Pract Res Clin gut permeability causing endotoxemia, systemic low-grade­ Rheumatol 2014;28:5–15. inflammation and aggravate severity of OA induced by MLI 3 Yoshimura N, Muraki S, Oka H, et al. Accumulation of metabolic risk factors such as overweight, hypertension, dyslipidaemia, and impaired glucose tolerance raises the surgery; (3) the abundances of the Fusobacterium, Faecalibac- risk of occurrence and progression of knee osteoarthritis: a 3-­year follow-­up of the terium and Ruminococcacea bacterial genera were consistently road study. Osteoarthritis Cartilage 2012;20:1217–26. associated with both histological OA severity and inflammatory 4 collins KH, Herzog W, MacDonald GZ, et al. Obesity, metabolic syndrome, and biomarkers, suggesting that they might be the bacterial agents musculoskeletal disease: common inflammatory pathways suggest a central role for loss of muscle integrity. Front Physiol 2018;9:112. linked to the MetS that are responsible for aggravating OA 5 Barboza E, Hudson J, Chang W-­P, et al. Profibrotic infrapatellar fat pad remodeling severity in germ-­free mice. These results support our two-hit­ without M1 macrophage polarization precedes knee osteoarthritis in mice with diet-­ hypothesis of OA involving the gut microbiome. Because the gut induced obesity. Arthritis Rheumatol 2017;69:1221–32. microbiome is capable of change through interventions such as 6 Griffin TM, Huebner JL, Kraus VB, et al. Induction of osteoarthritis and metabolic diet and exercise, it could be a promising therapeutic target in inflammation by a very high-­fat diet in mice: effects of short-­term exercise. Arthritis Rheum 2012;64:443–53. OA. Future studies are needed to explore the causal mechanism 7 schott EM, Farnsworth CW, Grier A, et al. Targeting the gut microbiome to treat the and potential translation of the present findings into clinical osteoarthritis of obesity. JCI Insight 2018;3. doi:10.1172/jci.insight.95997. [Epub practice. ahead of print: 19 Apr 2018].

Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 655 Osteoarthritis

8 creely SJ, McTernan PG, Kusminski CM, et al. Lipopolysaccharide activates an innate 37 cherbuy C, Darcy-­Vrillon B, Morel MT, et al. Effect of germfree state on the capacities immune system response in human adipose tissue in obesity and type 2 diabetes. Am of isolated rat colonocytes to metabolize n-­butyrate, glucose, and glutamine. J Physiol Endocrinol Metab 2007;292:E740–7. Gastroenterology 1995;109:1890–9. 9 Brun P, Castagliuolo I, Di Leo V, et al. Increased intestinal permeability in obese mice: 38 rabot S, Membrez M, Bruneau A, et al. Germ-free­ C57BL/6J mice are resistant to new evidence in the pathogenesis of nonalcoholic steatohepatitis. Am J Physiol high-­fat-­diet-­induced insulin resistance and have altered cholesterol metabolism. Gastrointest Liver Physiol 2007;292:G518–25. Faseb J 2010;24:4948–59. 10 ahmadmehrabi S, Tang WHW. Gut microbiome and its role in cardiovascular diseases. 39 Mestdagh R, Dumas M-­E, Rezzi S, et al. Gut microbiota modulate the metabolism of Curr Opin Cardiol 2017;32:761–6. brown adipose tissue in mice. J Proteome Res 2012;11:620–30. 11 Zhuo Q, Yang W, Chen J, et al. Metabolic syndrome meets osteoarthritis. Nat Rev 40 Yoshimura A, Wakabayashi Y, Mori T. Cellular and molecular basis for the regulation of Rheumatol 2012;8:729–37. inflammation by TGF-­beta. J Biochem 2010;147:781–92. 12 niu J, Clancy M, Aliabadi P, et al. Metabolic syndrome, its components, and 41 Varela-­Eirin M, Loureiro J, Fonseca E, et al. Cartilage regeneration and ageing: knee osteoarthritis: the Framingham osteoarthritis study. Arthritis Rheumatol targeting cellular plasticity in osteoarthritis. Ageing Res Rev 2018;42:56–71. 2017;69:1194–203. 42 Grol MW, Lee BH. Gene therapy for repair and regeneration of bone and cartilage. 13 Pascale A, Marchesi N, Marelli C, et al. Microbiota and metabolic diseases. Endocrine Curr Opin Pharmacol 2018;40:59–66. 2018;61:357–71. 43 Turnbaugh PJ, Ridaura VK, Faith JJ, et al. The effect of diet on the human gut 14 Kovatcheva-­Datchary P, Arora T. Nutrition, the gut microbiome and the metabolic microbiome: a metagenomic analysis in humanized gnotobiotic mice. Sci Transl Med syndrome. Best Pract Res Clin Gastroenterol 2013;27:59–72. 2009;1:6ra14. 15 Yoo JY, Kim SS, Probiotics KSS. Probiotics and prebiotics: present status and future 44 nguyen TLA, Vieira-­Silva S, Liston A, et al. How informative is the mouse for human perspectives on metabolic disorders. Nutrients 2016;8:173. gut microbiota research? Dis Model Mech 2015;8:1–16. 16 Huang Z, Kraus VB. Does lipopolysaccharide-­mediated inflammation have a role in 45 andoh A, Nishida A, Takahashi K, et al. Comparison of the gut microbial community OA? Nat Rev Rheumatol 2016;12:123–9. between obese and lean peoples using 16S gene sequencing in a Japanese 17 alberti KGMM, Zimmet P, Shaw J, et al. The metabolic syndrome-­a new worldwide population. J Clin Biochem Nutr 2016;59:65–70. definition. Lancet 2005;366:1059–62. 46 liu R, Hong J, Xu X, et al. Gut microbiome and serum metabolome alterations in 18 altman R, Asch E, Bloch D, et al. Development of criteria for the classification and obesity and after weight-­loss intervention. Nat Med 2017;23:859–68. reporting of osteoarthritis. classification of osteoarthritis of the knee. diagnostic and 47 Haro C, Garcia-­Carpintero S, Alcala-­Diaz JF, et al. The gut microbial community in therapeutic criteria Committee of the American rheumatism association. Arthritis metabolic syndrome patients is modified by diet. J Nutr Biochem 2016;27:27–31. Rheum 1986;29:1039–49. 48 cheng C, Wei H, Yu H, et al. Metabolic syndrome during perinatal period in sows and 19 Kellgren JH, Lawrence JS. Radiological assessment of osteo-­arthrosis. Ann Rheum Dis the link with gut microbiota and metabolites. Front Microbiol 2018;9:9. 1957;16:494–502. 49 strauss J, Kaplan GG, Beck PL, et al. Invasive potential of gut mucosa-­derived 20 Hamada D, Sampson ER, Maynard RD, et al. Surgical induction of posttraumatic Fusobacterium nucleatum positively correlates with IBD status of the host. Inflamm osteoarthritis in the mouse. Methods Mol Biol 2014;1130:61–72. Bowel Dis 2011;17:1971–8. 21 sekiguchi H, Tomioka N, Nakahara T, et al. A single band does not always represent 50 Mima K, Nishihara R, Qian ZR, et al. Fusobacterium nucleatum in colorectal carcinoma single bacterial strains in denaturing gradient gel electrophoresis analysis tissue and patient prognosis. Gut 2016;65:1973–80. 2001;23:1205–8. 51 Ziebolz D, Rupprecht A, Schmickler J, et al. Association of different immunosuppressive 22 Wang A-­H, Li M, Li C-Q,­ et al. Human colorectal mucosal microbiota correlates with its medications with periodontal condition in patients with rheumatoid arthritis: results host niche physiology revealed by endomicroscopy. Sci Rep 2016;6:21952. from a cross-­sectional study. J Periodontol 2018;89:1310–7. 23 Vaishnava S, Yamamoto M, Severson KM, et al. The antibacterial lectin RegIIIgamma 52 lopez-­Siles M, Duncan SH, Garcia-­Gil LJ, et al. Faecalibacterium prausnitzii: from promotes the spatial segregation of microbiota and host in the intestine. Science microbiology to diagnostics and prognostics. ISME J 2017;11:841–52. 2011;334:255–8. 53 van Dijkhuizen EHP, Del Chierico F, Malattia C, et al. Microbiome analytics of the gut 24 cani PD, Amar J, Iglesias MA, et al. Metabolic endotoxemia initiates obesity and microbiota in patients with juvenile idiopathic arthritis: a longitudinal observational insulin resistance. Diabetes 2007;56:1761–72. cohort study. Arthritis Rheumatol 2019;71:1000–10. 25 Mabey T, Honsawek S, Tanavalee A, et al. Plasma and synovial fluid inflammatory 54 Duncan SH, Hold GL, Harmsen HJM, et al. Growth requirements and cytokine profiles in primary knee osteoarthritis. Biomarkers 2016;21:639–44. fermentation products of Fusobacterium prausnitzii, and a proposal to reclassify 26 sohn DH, Sokolove J, Sharpe O, et al. Plasma proteins present in osteoarthritic it as Faecalibacterium prausnitzii gen. nov., comb. nov. Int J Syst Evol Microbiol synovial fluid can stimulate cytokine production via Toll-­like receptor 4. Arthritis Res 2002;52:2141–6. Ther 2012;14:R7. 55 He Y, Wu W, Wu S, et al. Linking gut microbiota, metabolic syndrome and economic 27 Zhao XY, Yang ZB, Zhang ZJ, et al. CCL3 serves as a potential plasma biomarker in status based on a population-­level analysis. Microbiome 2018;6:172. knee degeneration (osteoarthritis). Osteoarthritis Cartilage 2015;23:1405–11. 56 raman M, Ahmed I, Gillevet PM, et al. Fecal microbiome and volatile organic 28 Matsuki T, Horai R, Sudo K, et al. Il-1 plays an important role in lipid metabolism compound metabolome in obese humans with nonalcoholic fatty liver disease. Clin by regulating insulin levels under physiological conditions. J Exp Med Gastroenterol Hepatol 2013;11:868–75. 2003;198:877–88. 57 Kang C, Wang B, Kaliannan K, et al. Gut microbiota mediates the protective effects 29 luheshi GN, Gardner JD, Rushforth DA, et al. Leptin actions on food intake of dietary capsaicin against chronic low-­grade inflammation and associated obesity and body temperature are mediated by IL-1. Proc Natl Acad Sci U S A induced by high-­fat diet. MBio 2017;8. 1999;96:7047–52. 58 Milani C, Duranti S, Bottacini F, et al. The first microbial colonizers of the human gut: 30 Wallenius V, Wallenius K, Ahrén B, et al. Interleukin-6-­deficient mice develop mature-­ composition, activities, and health implications of the infant gut microbiota. Microbiol onset obesity. Nat Med 2002;8:75–9. Mol Biol Rev 2017;81. doi:10.1128/MMBR.00036-17. [Epub ahead of print: 08 Nov 31 Peters M, Schirmacher P, Goldschmitt J, et al. Extramedullary expansion of 2017]. hematopoietic progenitor cells in interleukin (IL)-6-­sIL-­6R double transgenic mice. J 59 Boer CG, Radjabzadeh D, Medina-­Gomez C, et al. Intestinal microbiome composition Exp Med 1997;185:755–66. and its relation to joint pain and inflammation.Nat Commun 2019;10:4881. 32 rosenson RS. Effect of fenofibrate on adiponectin and inflammatory biomarkers in 60 Gupta VK, Paul S, Dutta C. Geography, ethnicity or Subsistence-­Specific variations in metabolic syndrome patients. Obesity 2009;17:504–9. human microbiome composition and diversity. Front Microbiol 2017;8:1162. 33 Ulici V, Kelley KL, Azcarate-­Peril MA, et al. Osteoarthritis induced by destabilization 61 Ma H-­L, Blanchet TJ, Peluso D, et al. Osteoarthritis severity is sex dependent in a of the medial meniscus is reduced in germ-­free mice. Osteoarthritis Cartilage surgical mouse model. Osteoarthritis Cartilage 2007;15:695–700. 2018;26:1098–109. 62 Fedak KM, Bernal A, Capshaw ZA, et al. Applying the Bradford Hill criteria in the 34 liu H, Wang J, He T, et al. Butyrate: a double-­edged sword for health? Adv Nutr 21st century: how data integration has changed causal inference in molecular 2018;9:21–9. epidemiology. Emerg Themes Epidemiol 2015;12:14. 35 Donohoe DR, Garge N, Zhang X, et al. The microbiome and butyrate regulate energy 63 Ussar S, Griffin NW, Bezy O, et al. Interactions between gut microbiota, host genetics metabolism and autophagy in the mammalian colon. Cell Metab 2011;13:517–26. and diet modulate the predisposition to obesity and metabolic syndrome. Cell Metab 36 cresci GA, Thangaraju M, Mellinger JD, et al. Colonic gene expression in conventional 2015;22:516–30. and germ-­free mice with a focus on the butyrate receptor GPR109A and the butyrate 64 Parséus A, Sommer N, Sommer F, et al. Microbiota-­Induced obesity requires farnesoid transporter SLC5A8. J Gastrointest Surg 2010;14:449–61. X receptor. Gut 2017;66:429–37.

656 Huang ZY, et al. Ann Rheum Dis 2020;79:646–656. doi:10.1136/annrheumdis-2019-216471 Crystal arthropathies

Epidemiological science Subtype-­specific gout susceptibility loci and enrichment of selection pressure on ABCG2 and ALDH2 identified by subtype genome-­wide meta-­ analyses of clinically defined gout patients Akiyoshi Nakayama,1,2 Masahiro Nakatochi ‍ ‍ ,3 Yusuke Kawamura,1,4 Ken Yamamoto,5 Hirofumi Nakaoka,6 Seiko Shimizu,1 Toshihide Higashino,1,7 Teruhide Koyama,8 Asahi Hishida,9 Kiyonori Kuriki,10 Miki Watanabe,11 Toru Shimizu,12,13 Keiko Ooyama,14 Hiroshi Ooyama,14 Mitsuo Nagase,15 Yuji Hidaka,16 Daisuke Matsui,8 Takashi Tamura,9 Takeshi Nishiyama,11 Chisato Shimanoe,17,18 Sakurako Katsuura-­Kamano,19 Naoyuki Takashima,20,21 Yuya Shirai,22,23 Makoto Kawaguchi,1,24 Mikiya Takao,1,25 Ryo Sugiyama,1 Yuzo Takada,26 Takahiro Nakamura,27 Hiroshi Nakashima,28 Masashi Tsunoda,28 Inaho Danjoh,29 Atsushi Hozawa,30 Kazuyoshi Hosomichi,31 Yu Toyoda,32 Yu Kubota,32 Tappei Takada,32 Hiroshi Suzuki,32 Blanka Stiburkova,33,34 Tanya J. Major,35 Tony R. Merriman,35 Nagato Kuriyama,8 Haruo Mikami,36 Toshiro Takezaki,37 Keitaro Matsuo,38,39 Sadao Suzuki,11 Tatsuo Hosoya,40,41 Yoichiro Kamatani,42,43 Michiaki Kubo,44 Kimiyoshi Ichida,40,45 Kenji Wakai,9 Ituro Inoue,6 Yukinori Okada ‍ ‍ ,22,46 Nariyoshi Shinomiya,1 Hirotaka Matsuo,1 on behalf of Japan Gout Genomics Consortium (Japan Gout) Handling editor Josef S Smolen Abstract Key messages ►► Additional material is Objectives Genome-­wide meta-­analyses of clinically published online only. To view defined gout were performed to identify subtype-­ please visit the journal online What is already known about this subject? (http://dx.​ ​doi.org/​ ​10.​1136/​ specific susceptibility loci. Evaluation using selection ►► Our previous genome-­wide association study annrheumdis-2019-​ ​216644). pressure analysis with these loci was also conducted to (GWAS) was performed on broad subtypes of investigate genetic risks characteristic of the Japanese For numbered affiliations see gout with only 945 gout cases. A recent study end of article. population over the last 2000–3000 years. has revealed genetic adaptive evolution of gout Methods Two genome-­wide association studies in the Japanese population. Correspondence to (GWASs) of 3053 clinically defined gout cases and 4554 Hirotaka Matsuo, Department of controls from Japanese males were performed using the What does this study add? Integrative Physiology and Bio-­ Japonica Array and Illumina Array platforms. About 7.2 Nano Medicine, National ►► This is the first GWAS meta-analyses­ of clinically Defense Medical College, million single-­nucleotide polymorphisms were meta-­ defined gout with more finely differentiated Tokorozawa 359-8513, Japan; analysed after imputation. Patients were then divided subtypes using two GWAS platforms with hmatsuo@​ ​ndmc.ac.​ ​jp into four clinical subtypes (the renal underexcretion type, larger samples (3055 cases and 4554 controls). renal overload type, combined type and normal type), AN, MN, YK, KY and HN We identified multiple subtype-­specific loci contributed equally. and meta-­analyses were conducted in the same manner. including four novel loci such as CD2, which Selection pressure analyses using singleton density score encodes a well-­known surface antigen found on Received 15 November 2019 were also performed on each subtype. all peripheral blood T-­cells. Revised 13 February 2020 Results In addition to the eight loci we reported ►► The present study showed significant Accepted 17 February 2020 Published Online First previously, two novel loci, PIBF1 and ACSM2B, enrichment of selection pressure on two genes, 1 April 2020 were identified at a genome-­wide significance level ABCG2 and ALDH2, for gout susceptibility in the (p<5.0×10–8) from a GWAS meta-­analysis of all gout Japanese population over the last 2000–3000 patients, and other two novel intergenic loci, CD2-­ years. PTGFRN and SLC28A3-NTRK2­ , from normal type gout © Author(s) (or their employer(s)) 2020. Re-­use patients. Subtype-­dependent patterns of Manhattan plots permitted under CC BY. were observed with subtype GWASs of gout patients, Conclusions Our findings on subtype GWAS meta-­ Published by BMJ. indicating that these subtype-­specific loci suggest analyses and selection pressure analysis of gout will differences in pathophysiology along patients’ gout assist elucidation of the subtype-­dependent molecular To cite: Nakayama A, targets and evolutionary involvement among genotype, Nakatochi M, Kawamura Y, subtypes. Selection pressure analysis revealed significant et al. Ann Rheum Dis enrichment of selection pressure on ABCG2 in addition to phenotype and subtype-­specific tailor-­made medicine/ prevention of gout and hyperuricaemia. 2020;79:657–665. ALDH2 loci for all subtypes except for normal type gout.

Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 657 Crystal arthropathies

Methods Key messages Study subjects and patients involvement We performed subtype genome-­wide meta-analyses­ based on How might this impact on clinical practice or future two case–control data sets for gout that included the Japonica developments? Array8 and Illumina Array platforms. Patients with known clin- ►► Our subtype GWASs of gout enabled us to develop subtype-­ ical parameters were recruited from Japanese male outpatients dependent molecular targets that will lead to novel at gout clinics (see online supplementary methods). All 3104 subtype-­specific genome tailor-­made therapies for gout/ cases were clinically diagnosed as having primary gout according hyperuricaemia. to the criteria established by the American College of Rheuma- ►► The present study also elucidates the Japanese genetic tology,9 and their subtypes were also diagnosed along with their evolution of susceptibility to gout/hyperuricaemia and its clinical parameters as described previously3 5 6 (table 1 and online subtypes. supplementary figure S1). As controls, 6081 individuals were assigned from Japanese male participants in the Japan Multi-­ Institutional Collaborative Cohort Study (J-­MICC Study).10 11 Introduction This research was done without patient involvement (see online Gout is a well-­known disease that manifests as acute and severe supplementary methods). non-­infectious arthritis.1 2 According to patients’ clinical param- eters which reflect its causes,2–5 gout can be classified into four Genotyping and imputation for the Japonica Array data set distinct subtypes: the renal underexcretion (RUE) type, renal A total of 1048 male clinically defined gout cases and 1179 male 10 11 overload (ROL) type, combined type and normal type, as controls from the J-­MICC Study were genotyped with the 8 shown in table 1 and online supplementary figure S1. Because use of a Japonica SNP Array. The detail of quality control is these subtypes reflect causes of gout, genome-­wide association described in online supplementary methods. This quality control studies (GWASs) of these subtypes are also likely to indicate filtering resulted in the selection of 1028 case subjects and 1167 control subjects as well as 603 009 single-­nucleotide polymor- its various genetic and pathophysiological backgrounds. While phisms (SNPs). Prephasing and imputation were performed dividing patients into these subtypes is helpful for understanding using SHAPEIT212 and Minimac3,13 respectively. Postimpu- patients’ pathophysiology, GWASs of these subtypes have only tation quality control was also performed as described in the rarely been conducted, partly because clinical data, including online supplementary methods. Ultimately, 1028 case subjects time-­consuming urinary collection, are necessary to categorise and 952 control subjects as well as 7 529 176 SNPs remained for these subtypes. We previously performed a GWAS with clini- the GWAS analysis. cally defined gout patients,6 followed by another with broader subtypes:7 RUE gout and ROL gout (table 1 and online supple- Genotyping and imputation for the Illumina Array data set mentary figure S1), that revealed their specific loci. Although As case data, 2056 male gout cases subjects were genotyped with we were able to show these associations, this process has its the use of HumanOmniExpress or HumanOmniExpressExome limitations, including the use of a custom chip for replication BeadChip Arrays (Illumina, San Diego, CA, USA). The detail studies that did not provide comprehensive genetic association of quality control is described in the online supplementary searching. We use finely differentiated subtypes in daily clinical methods. This quality control filtering resulted in the selection settings but there were not sufficient numbers of patients in the of 2032 case subjects and 4901 control subjects as well as 553 7 previous study to enable a GWAS with these finely differenti- 321 SNPs. Postimputation quality control was also performed ated subtypes. This prompted us to conduct, for the first time, as described in the online supplementary methods. Ultimately, GWASs with four distinct subtypes using meta-analysis­ across 2025 case subjects and 3602 control subjects as well as 7 356 two GWAS platforms with a larger number of patients. We addi- 207 SNPs remained for the GWAS analysis. tionally conducted selection pressure analysis of the Japanese population on gout subtypes with the risk loci identified in the Association analysis for SNPs and gout present study in order to investigate the evolutionary selective The association of SNPs with gout was assessed using logistic pressure on the Japanese population over the last 2000–3000 regression analysis (generalised linear model); the dependent years. variable was gout label (case=1, control=0), and the indepen- dent variables included imputed genotypes of each SNP and covariates. The covariates comprised the first four principal component scores. The effect sizes and standard errors estimated Table 1 Subtypes of gout* used in the present study in logistic regression analysis were used in the subsequent meta-­ Subtype Clinical parameters analysis. The association analysis was performed with the use Differentiated subtype of Efficient and Parallelizable Association Container Toolbox (EPACTS). https://​genome.​sph.​umich.​edu/​wiki/EPACTS).​ RUE type gout FEUA <5.5% and UUE ≤25

ROL type gout FEUA≥5.5% and UUE >25

Combined type gout FEUA <5.5% and UUE >25 Meta-analyses

Normal type gout FEUA≥5.5% and UUE ≤25 The meta-­analyses were performed using a total of 3053 cases Broader subtype and 4554 controls from the two data sets (online supplemen- RUE gout (RUE type gout +combined type gout) FE <5.5% tary table S1–S3). The association results for each SNP across UA 14 ROL gout (ROL type gout +combined type gout) UUE >25 the studies were combined with METAL software using the *Subtypes of hyperuricaemia can be classified in the same manner. fixed-­effects inverse-­variance-­weighted method. Heterogeneity of effect sizes was assessed via the I2 index. The meta-analysis­ FEUA, fractional excretion of uric acid (unit: %); ROL, renal overload; RUE, renal underexcretion; UUE, urinary urate excretion (unit: mg/h/1.73 m2). included 7 206 774 SNPs and the results from both the Japonica

658 Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 Crystal arthropathies

Figure 1 Manhattan plots of GWASs of subtypes of gout. Clinical subtypes and Manhattan plots of GWASs of all gout types, RUE type gout, combined type gout, ROL type gout and normal type gout are shown. The x-axis­ represents chromosomal positions and the y-­axis shows −log10 p –8 values. The dotted lines indicate the genome-­wide significance threshold (p=5.0×10 ). FEUA, fractional excretion of uric acid (%); GWASs, genome-­ wide association studies; ROL, renal overload; RUE, renal underexcretion; UUE, urinary urate excretion (mg/h/1.73 m2). See table 1 and online supplementary figure S1 for a detailed classification of gout/hyperuricaemia. and Illumina Arrays. The genome-­wide significance level α was loci were identified at the genome-wide­ significant level to be set to a p value of <5×10–8. associated with all, the RUE type, ROL type, combined type and normal type gout, respectively. Of these, two loci from all gout Genetic correlation analysis types, rs76499759 of PIBF1 and rs9926388 of ACSM2B, and Genetic correlation analysis using linkage disequilibrium score another two intergenic loci from normal type gout, rs146978188 (LDSC) regression analysis15 was conducted to examine the of CD2-­PTGFRN and rs548944057 of SLC28A3-­NTRK2, were potential genetic overlap between gout subtypes and between detected as novel loci. each subtype and serum uric acid (SUA) levels. For the regres- For all gout cases (table 2), 10 loci showed associa- sion, we used the 1000 genomes phase_3 East Asian LDSC and tion at the genome-wide­ significance level: rs4148155 of –101 summary statistics for high-­quality common SNPs present in the ABCG2 (pmeta=1.81×10 ; ORs=2.23), rs671 of ALDH2 –54 HapMap 3 reference panel for each analysis. (pmeta=3.19×10 ; OR=1.93), rs3775946 of SLC2A9 –41 (pmeta=3.73×10 ; OR=1.63), rs145954970 of SLC22A11 –21 Selection pressure analysis (pmeta=2.25×10 ; OR=12.43), rs3129500 of FAM35A –17 The details of genome-­wide recent natural selection signature (recently renamed as SHLD2, pmeta=4.34×10 ; OR=1.37), 16 –13 using singleton density score (SDS) from high-depth­ whole rs1260326 of GCKR (pmeta=2.07×10 ; OR=1.30), rs1010269 –9 genome sequence data of the Japanese population had been of BCAS3 (pmeta=1.81×10 ; OR=1.24), rs2817188 of 17 17 –8 described in the previous study. Using the same approach, we SLC17A1 (pmeta=1.06×10 ; OR=1.35), rs9926388 of –8 calculated the SDS of gout-­risk variants identified in the present ACSM2B (pmeta=2.30×10 ; OR=1.24) and rs76499759 of –8 study, and evaluated overlaps between enrichment of the natural PIBF1 (pmeta=2.79×10 ; OR=1.27). Among these 10 loci, selection signatures and these variants from each subtype. PIBF1 and ACSM2B (table 2 and figure 2A,B). were identified for the first time as gout-­risk loci at the genome-­wide signif- Results icance level. BCAS3 was identified here for the first time by 18 Subtype GWASs of gout the GWAS approach with Japanese individuals, while Li et al Figure 1 displays Manhattan plots of all and four clinical reported that rs11653176, another SNP of BCAS3, is associated subtypes of gout, and figure 2 shows regional plots of novel loci. with gout based on a GWAS with a Han Chinese population. We Compared with the plotted pattern for all clinically defined gout also replicated its association with gout in a Japanese population patients, each gout subtype (RUE type, ROL type, combined using a candidate gene approach.19 Other loci have been previ- type and normal type) shows a subtype-­specific plotted pattern ously reported to have an association with gout in our previous and indicates the presence of cause-specific­ associated genes such GWASs6 7 and association studies.20 21 As suggestive loci for gout, as SLC2A9 and ABCG2. Table 2 lists the genome-wide­ significant seven loci: PDZK1, TACR1-­EVA1A, LOC100128993, ARID5B, loci from subtype GWASs. In total, 10, five, two, seven and three TOLLIP-­AS1-­BRSK2, SLC38A1 and MLXIP, were detected

Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 659 Crystal arthropathies

Figure 2 Regional association plots of novel gout loci. Two loci were revealed to exceed the genome-wide­ significance level from the meta-­analysis of GWASs from all gout patients, and another two loci from normal type gout patients. The highest association signal in each panel is located on (A) PIBF1, (B) ACSM2B, (C) CD2-PTGFRN­ and (D) SLC28A3-­NTRK2. The region within 1 Mb from the single-nucleotide­ polymorphism (SNP) indicating the lowest p value is shown. (Upper panel) Plots of −log10 p values for the test of SNP association with gout. The SNP showing the lowest p value in the meta-­analysis is depicted as a purple diamond. Other SNPs are colour-coded­ according to the extent of linkage disequilibrium (measured in r2) with the SNP showing the lowest p value. Recombination rates (centimorgans per Mb) estimated from HapMap Phase II data are also plotted. (Lower panel) RefSeq genes. Genomic coordinates are based on NCBI human genome reference sequence build hg19. The r2 data were calculated with 1000 Genomes Project Phase_3 JPT samples.45 GWASs, genome-­wide association studies.

–16 (see online supplementary table S4). Of these, PDZK1, a gene (pmeta=5.80×10 ; OR=1.86), rs9420434 of GLUD1 22 23 –11 encoding a scaffolding protein such as for urate transporters (pmeta=8.62×10 ; OR=1.54) and rs76741582 of SLC22A11 –10 SLC22A12/URAT1 and ABCG2, was reported to have an asso- (pmeta=1.06×10 ; OR=3.93). Since GLUD1 is in LD with ciation with SUA by a GWAS approach24 25 and with gout as a SHLD2/FAM35A, for which we previously showed a significant result of candidate gene approach studies.26–28 association with gout,7 all of these five loci were previously iden- As shown in online supplementary table S2, all the 3053 cases tified as having an association with gout.6 7 We also detected nine were classified into RUE type gout (654 cases), ROL type gout suggestive loci: SMYD3, GCKR-­C2orf16, SMARCC1, FRMD4B-­ (486 cases), combined type gout (905 cases) and normal type MITF, ARL4A-­ETV1, C7orf66-­EIF3IP1, ASB10, PXDNL and gout (92 cases) for GWASs of gout subtypes. LOC100287896-­POLD3, for RUE type gout as shown in online The meta-analysis­ of a GWAS of the RUE type gout (table 2) supplementary table S4. showed significant SNPs in the following five loci: rs3775948 From ROL type gout (table 2), rs4148155 of ABCG2 –22 –46 of SLC2A9 (pmeta=8.01×10 ; OR=1.89), rs4148155 of (pmeta=9.75×10 ; OR=2.79) and rs11066008 of ACAD10 –16 –12 ABCG2 (pmeta=2.54×10 ; OR=1.69), rs4646776 of ALDH2 (ALDH2) (pmeta=4.20×10 ; OR=1.79) were revealed to have 660 Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 Crystal arthropathies HetP 0.418 0.861 0 0.787 0 0.669 00 0.660 0.355 0 0.412 0 0.836 00 0.448 0 0.737 0 0.462 0 0.497 0.839 0 0.559 00 0.748 0.646 0 0.596 0 0.788 0 0 0 0.849 3.8 0.308 4.4 0.307 6.1 0.302 2 10.3 0.291 42.8 0.186 50.5 0.155 65.6 0.088 I 39 0.200 –8 –8 –8 –8 –8 –13 –41 –101 –8 –17 –21 –54 –9 –22 –16 –11 –10 –16 –46 –12 –8 –19 –56 –11 –10 –27 –9 2.91×10 2.79×10 2.30×10 4.93×10 analysis ­ (95% CI) P value R 2.34 (1.73 to 3.16)4.04 (2.47 to 6.62 ) 3.64×10 1.30 (1.21 to 1.39)1.63 (1.52 to 1.75) 2.07×10 2.23 (2.08 to 2.41) 3.73×10 1.35 (1.22 to 1.50) 1.81×10 1.37 (1.28 to 1.48) 1.06×10 4.34×10 1.93 (1.78 to 2.10)1.27 (1.17 to 1.38 ) 3.19×10 1.24 (1.15 to 1.33 ) 1.24 (1.16 to 1.33) 1.81×10 1.89 (1.66 to 2.15)1.69 (1.49 to 1.91) 8.01×10 1.54 (1.35 to 1.75) 2.54×10 3.93 (2.59 to 5.95) 8.62×10 1.86 (1.60 to 2.16) 1.06×10 5.80×10 2.79 (2.42 to 3.22)1.79 (1.52 to 2.11) 9.75×10 4.20×10 1.37 (1.23 to 1.53)1.67 (1.50 to 1.87) 1.05×10 2.46 (2.20 to 2.76) 1.43×10 1.49 (1.32 to 1.68) 1.53×10 1.44 (1.28 to 1.62) 4.30×10 7.66×10 2.38 (2.03 to 2.78)1.37 (1.23 to 1.52) 1.88×10 5.62×10 6.31 (3.26 to 12.24 ) Meta- O 12.43 (7.39 to 20.92) 2.25×10 –3 –2 –3 –2 –2 –3 –14 –33 –4 –4 –8 –20 –4 –9 –6 –5 –5 –6 –12 –6 –2 –7 –22 –2 –6 –15 –4 3.83×10 2.62×10 6.46×10 1.90×10 (95% CI) P value R O AF Case Control 0.160 0.047 4.58 (1.63 to 12.83 ) Japonica Array Japonica R 0.6130.672 0.5650.462 0.555 1.22 (1.07 to 1.39)0.871 0.264 1.67 (1.46 to 1.92)0.423 0.836 3.61×10 2.38 (2.06 to 2.75)0.997 0.371 9.42×10 1.40 (1.16 to 1.69)0.821 0.967 8.13×10 1.30 (1.13 to 1.50)0.212 0.684 5.03×10 25.09 (7.83 to 80.37)0.315 0.183 2.71×10 5.75×10 2.04 (1.75 to 2.37)0.583 0.277 1.20 (1.02 to 1.41 ) 0.528 2.78×10 1.22 (1.06 to 1.42 ) 0.717 1.26 (1.10 to 1.43)0.382 0.5560.342 0.264 7.68×10 2.04 (1.61 to 2.59)0.036 0.243 1.74 (1.39 to 2.19)0.796 0.008 3.82×10 1.65 (1.31 to 2.09) 0.680 1.73×10 4.87 (2.31 to 10.25)0.493 2.37×10 1.82 (1.4 to 2.36) 3.09×10 0.751 0.264 0.600 6.66×10 2.61 (2 to 3.41)0.615 2.12 (1.56 to 2.89)0.672 0.565 2.05×10 0.474 0.556 1.57×10 1.23 (1.03 to 1.48)0.436 0.264 1.63 (1.35 to 1.97)0.375 0.393 2.50×10 2.56 (2.11 to 3.09) 0.282 2.58×10 0.845 1.28 (1.04 to 1.58) 5.87×10 0.581 0.683 1.56 (1.29 to 1.89) 1.94×10 0.499 4.39×10 2.61 (2.06 to 3.31)0.086 1.42 (1.19 to 1.70)0.438 0.021 2.69×10 0.264 1.03×10 5.58 (1.33 to 23.46 ) 2.14 (1.05 to 4.36) 3.67×10 –6 –7 –6 –7 –7 –12 –29 –70 –6 –14 –15 –36 –7 –14 –11 –7 –7 –11 –35 –7 –8 –14 –36 –10 –5 –14 –5 2.13×10 2.56×10 1.05×10 8.35×10 (95% CI) P value R O ­ analyses AF Case Control Illumina Array R ­ wide meta- on- ­ nucleotide polymorphism. risk N isk R T C 0.647 0.545 1.45 (1.27 to 1.66) 4.92×10 G AGC 0.454 A G 0.277 0.423 0.995 2.18 (2.00 to 2.38) 0.358 0.971 1.05×10 1.40 (1.29 to 1.53)C 10.43 (5.83 to 18.66)G 2.36×10 G 2.78×10 AT 0.705G 0.388 C 0.573 CG 0.277 0.028 1.83 (1.57 to 2.14) 0.819 A 1.66 (1.43 to 1.93) 0.009 2.60×10 0.722 2.72×10 0.515 3.57 (2.16 to 5.88) 1.88 (1.57 to 2.26)A 0.277 6.31×10 G 1.73×10 C 2.87 (2.43 to 3.39) A 5.14×10 0.474 0.457 0.276 0.368 2.42 (2.11 to 2.78) 1.60 (1.39 to 1.85) 2.40×10 1.60×10 T A 0.125 0.046 3.89 (2.22 to 6.83 ) AA G G 0.219 0.301 0.180 0.252 1.30 (1.17 to 1.43 ) 1.24 (1.14 to 1.36 ) A G 0.065 0.015 6.53 (3.09 to 13.77 ) T C 0.623 0.545 1.33 (1.22 to 1.44) 8.75×10 G A 0.558 0.503 1.24 (1.14 to 1.34) 6.06×10 G A 0.473 0.277 2.38 (1.71 to 3.33) 3.35×10 G AG 0.677 AG 0.569 0.874 A 1.62 (1.49 to 1.76) 0.825 4.55×10 0.823 1.33 (1.18 to 1.51) 0.725 5.03×10 1.89 (1.71 to 2.08) 8.65×10 C GG 0.690 A 0.573C 0.360 1.70 (1.48 to 1.95) T 0.295 9.54×10 0.558 1.38 (1.19 to 1.60) 0.482 2.44×10 1.34 (1.18 to 1.53) 1.20×10 Alleles

K2

R

T

RN N

B

M2 F1 S 2 - PTGF C28A3 - B D GCKR ABCG2 SHLD2/FAM35A SLC22A11 SLC2A9 ABCG2 GLUD1 (SHLD2 )SLC22A11 CALDH2 TABCG2 0.316ACAD10 (ALDH2 ) A 0.245 G 1.49 (1.27 to 1.74)ABCG2 0.724 6.39×10 SHLD2/FAM35A 0.643 1.67 (1.38 to 2.04) 2.49×10 GCKR BCAS3 ABCG2 SLC2A9 SLC17A1 ALDH2 PI AC SLC2A9 CDC42BPG (SLC22A12 ) NAA25 (ALDH2 )BCAS3 G CC 0.828SL 0.749 2.21 (1.80 to 2.72) 5.86×10 87 174 107 73 568 511 20 558 441 117 383 166 Position (bp)† Gene‡ 1 . ocus Chr L 12q24.12 12 112 241 766 Significant gout loci identified in the present genome-

P* Combined type gout patients rs1260326 2p23.3 2 27 730 940 rs548944057 9q21.33 9 rs146978188 1p13.1 rs4148155rs3129500 4q22.1rs145954970 10q23.2 11q13.1 4rs76499759 89 054 667 10 11 13q22.1 88 915 107 64 273 830 RUE type gout patients 13 rs3775948rs4148155rs9420434 4p16.1rs76741582 4q22.1rs4646776 10q23.2 4 11q13.1ROL type gout patients 4rs4148155 9 995 182 10 12q24.12 89 054 667 11rs11066008 88 843 209 12 64 247 850 4q22.1 112 230 019 12q24.12 12 4rs74904971 112 140 669 89 054 667 rs6586063 4q22.1 10q23.2 4 10 89 050 026 Normal type gout patients 88 949 045 All gout patients rs1260326 2p23.3 2 27 730 940 rs9926388rs1010269 16p12.3 17q23.2 16 17 59 448 945 rs4148155 4q22.1 4 89 054 667 *dbSNP rs number. †SNP positions are based on NCBI human genome reference sequence Build hg19. ‡Novel loci are shown in bold. single- SNP, renal underexcretion; RUE, renal overload; ROL, risk allele frequency; RAF, chromosome; Chr, SN Table 2 Table rs3775946rs2817188 4p16.1 6p22.2 4rs671 9 995 256 6 25 807 603 rs3775948 4p16.1rs11231879rs116873087 4 11q13.1rs9905274 9 995 182 12q24.13 11 17q23.2 64 581 645 12 112 511 913 17 59 450 441

Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 661 Crystal arthropathies an association. We had previously reported both to have a signif- the analysis with all associated SNPs (figure 3 and online supple- icant association with gout.6 7 20 21 Three suggestive loci, CNPY4, mentary table S5). GRID1 and KCNJ2-­CASC17, were also identified from ROL type gout (see online supplementary table S4). Combined type gout displayed the following seven signif- Discussion – We previously performed GWASs of clinically defined gout cases icant loci (table 2): rs74904971 of ABCG2 (pmeta=1.53×10 56 –27 in the Japanese population and found loci including ABCG2, ; OR=2.46), rs116873087 of NAA25 (pmeta=1.88×10 ; –19 SLC2A9, ALDH2 (CUX2), GCKR and SHLD2/FAM35A to be OR=2.38), rs3775948 of SLC2A9 (pmeta=1.43×10 ; –11 associated with gout at a genome-­wide significant level.6 7 While OR=1.67), rs6586063 of SHLD2/FAM35A (pmeta=4.30×10 ; –9 our previous study was performed with broader subtypes, it is OR=1.49), rs9905274 of BCAS3 (pmeta=5.62×10 ; OR=1.37), –10 one of the unique points of this study that the present GWAS rs11231879 of CDC42BPG (pmeta=7.66×10 ; OR=1.44) –8 is the first to be conducted with four differentiated subtypes: and rs1260326 of GCKR (pmeta=1.05×10 ; OR=1.37). There are studies on these which report an association between the RUE type, ROL type, combined type and normal type gout CDC42BPG and hyperuricaemia from the Japanese exome-wide­ (figure 1), which are commonly used in daily clinical settings. association study,29 and GWAS on SUA from a Korean popula- Two platforms for GWASs and meta-­analyses between them were tion.30 The significance around rs11231879 of CDC42BPG was, used to perform a comprehensive genetic association search. however, no longer evident when conditioned on rs11231879 The results allowed four novel loci to be identified from the itself, nor when conditioned on the secondarily significant SNP, present study. The pathophysiological associations of the two rs56093838 of SLC22A12/URAT1, demonstrating that these novel loci from all gout types, PIBF1 and ACSM2B with gout, are signals were from the same locus (see online supplementary totally unknown. PIBF (progesterone-induced­ blocking factor) is figure S2). Because URAT1 is a well-known­ urate transporter induced by progesterone and is a mediator that exerts substan- 34 that markedly affects SUA level, these findings indicate that the tial antiabortive activities, including cytokine secretion. PIBF1 true associated gene for combined type gout on chromosome might be therefore involved in decreased urate production by 11q13.1 locus is not CDC42BPG, but SLC22A12/URAT1. female hormone and/or decreased inflammatory response to rs116873087 of NAA25 is in strong LD with rs671 of ALDH2 gout attack. ACSM2B (acyl-CoA­ synthetase medium chain (r2=0.97 in the 1000 Genomes Project Phase_3: JPT samples). family member 2B) is a predominant transcript in the human All of the seven loci had therefore been previously identified liver and an enzyme catalysing the activation of medium-­chain 35 as having an association with gout.6 7 20 21 Two suggestive loci, length fatty acids. Because ASCM2B is involved in the produc- NCKAP5-­MIR3679 and PRDM8-­FGF5, were also identified tion of ATP, a purine body metabolised to urate, ASCM2B might from combined type gout (see online supplementary table S4). contribute to gout via that mechanism. While the present study Three significant loci were found from normal type showed significance at SNPs of two genes, it is of course possible –8 that these are mere markers and that the true risk SNPs are gout: rs548944057 of SLC28A3-­NTRK2 (pmeta=2.91×10 ; –8 present close by. For example, since UMOD, a causative gene OR=4.04), rs4148155 of ABCG2 (pmeta=3.64×10 ; OR=2.34) and rs146978188 of CD2-­PTGFRN (p =4.93×10–8; of uromodulin-­associated kidney disease (previously known as meta 36 OR=6.31). Of these, two intergenic loci are novel suscepti- familial juvenile hyperuricemic nephropathy) is located 180 kb bility loci for gout (table 2 and figure 2C,D) There were eight downstream from rs9926388 of ACSM2B, there might be a rela- suggestive loci: ZNF639-­MFN1, RUNX2-­CLIC5, DST, HGF, tionship between them. MED27-­NTNG2, LINC00944-­LINC02372, SV2B and GABPA, Another two intergenic loci from normal type gout, for normal type gout as shown in online supplementary table S4. rs146978188 of CD2-­PTGFRN and rs548944057 of SLC28A3-­ The LDSC regression analysis was performed to examine the NTRK2, were detected for the first time to have an association potential genetic overlap between gout subtypes and between with gout. CD2 is well known as a surface antigen found on each subtype and SUA levels. Significant positive genetic correla- all peripheral blood T-­cells, and PTGFRN/CD9P-1 encodes tions were observed among these subtypes as well as between prostaglandin F2 receptor inhibitor. Neither of these was previ- these traits and SUA levels (see online supplementary figure S3). ously known to have an association with gout or uric acid. CD101, which is next to PTGFRN (150 kb downstream from rs146978188), is reported to be expressed on macrophages/ Selection pressure analysis of gout susceptibility monocytes and T-­cells, to confer a modulatory/coregulatory We also performed selection pressure analysis of gout on the basis function, and to be conspicuously downregulated in rheumatoid of a previous report on the recent natural selection signature in arthritis patients.37 Because macrophages are a chief contributor the Japanese population.17 This analysis enables us to elucidate to gouty attack, CD101 might be the true susceptible gene for the genetic risks of gout characterised in the recent evolutionary normal type gout. rs146978188 is also in LD with an SNP of history (2000–3000 years) of the Japanese population. Because SLC22A15/FLIPT1, an orphan transporter gene in the same ALDH2 was reported to be subjected to strong selection pressure family as SLC22A12/URAT1, a well-­known urate transporter in the Japanese population,17 and because ABCG2,31 32 as well as gene that is strongly associated with gout. This transporter gene ALDH2,20 21 33 is a well-known­ strong susceptible gene for gout might have an association with normal type gout. SLC28A3/ in Japanese, selection pressure analysis was initially performed CNT3 is reported to be an Na+-­dependent pyrimidine-­selective outside of these two loci. As a result, only combined type gout and purine-­selective transporter found predominantly in the showed significant enrichment of selection pressure (p=0.026; intestine and kidney,38 39 which are the main urate excretion figure 3 and online supplementary table S5). When the ABCG2 pathways. Further analysis is needed to elucidate the relation- locus was included in the analysis, all of these subtypes except ship between this transporter gene and gout, including normal for normal type gout then showed significant enrichment of type gout. selection pressure. As expected, analysis including ALDH2 and While the present study revealed only suggestive loci from outside of ABCG2 showed significant enrichment of selection other subtype GWASs, some of these loci from subtype gout pressure except for normal type gout. This trend also persisted in also suggest a relationship with RUE, extra-RUE­ and/or

662 Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 Crystal arthropathies

Figure 3 Overlap between natural selection signatures and genetic risk of gout and its subtypes in the Japanese population. For each trait, inflation 2 of the selection χ value is indicated along the x-­axis, and −log10(p) of enrichment is plotted along the y-­axis. The horizontal grey line represents significance threshold (p<0.05). Because ABCG2 and ALDH2 are associated with a well-known­ genetic risk of gout in Japanese individuals, selection pressure analyses without these loci were performed initially (filled circle), and subsequent analyses were conducted with ABCG2 (filled triangle), as well as ABCG2 and ALDH2 (filled square). When calculated with ABCG2 (outside of ALDH2) (filled triangle), all but normal type gout showed significant selective pressure, indicating that ABCG2 is involved in adaptive evolution in Japanese for having higher SUA levels, which can result in gout. Finally, all but normal type gout also showed significant selective pressure with ABCG2 and ALDH2 loci (filled square). ROL, renal overload; RUE, renal underexcretion; SDS, singleton density score; SNPs, single-nucleotide­ polymorphisms.

overproduction of urate, because gout subtypes reflect their in the Japanese population. The results revealed that the Japa- clinical parameters (table 1). GWASs with these subtypes should nese population has evolved to have higher SUA levels, which therefore be useful for estimating the expression and function of can result in gout, due to the ABCG2 locus in addition to the proteins encoded by identified loci. rs557868370 of SLC38A1, already-­known ALDH2 gene. ABCG2 is now a well-known­ a transporter gene, was detected as a suggestive locus (see online susceptible gene for hyperuricaemia and gout, especially in the supplementary table S4). Since there is a study reporting the Japanese population.31 42 Few patients had gout before West- 40 relationship between oxidative stress and SLC38A1/SNAT1, ernisation of Japan about 150 years ago, which brought more it might have a relationship with urate, which also has antiox- purine-­rich foods to Japan. The fact that ABCG2 also has a rela- idative stress effects. Because genetic variants in urate trans- tionship with SUA levels might thus have caused few problems porter genes such as ABCG2, SLC2A9 and SLC22A12 are well to Japanese people until recently. The upside of SUA elevation known to cause SUA variation and gout, it is also possible that in the Japanese population might include resistance to anti- the SLC38A1/SNAT1 transporter is involved in urate or purine oxidative stress, lower cancer risk, neuroprotective effect and transport. 43 44 Very recently, Tin et al41 performed transancestry GWAS meta-­ longevity. Selection pressure analysis with other populations analysis of SUA and gout, including self-­reported gout cases. We will also generate more information on evolutionary association compared these SNPs and found five of the 10 loci detected here with gout to elucidate this hypothesis. (GCKR, SLC2A9, ABCG2, SLC17A1 and BCAS3) to be associ- In summary, we performed GWASs of all gout as well as of ated with gout (see online supplementary table S6), indicating distinct gout subtypes, and identified multiple subtype-specific­ population differences in the genetic basis of gout. loci including four novel loci. Selection pressure analysis Taking into account the evidence of shared genetic back- revealed significant enrichment of selection for the ABCG2 and ground among gout subtypes (see online supplementary figure ALDH2 loci in Japanese gout patients of each subtype. These S3) and the presence of subtype-­specific genetic factors of gout findings will lead to elucidation of the molecular pathophysi- (figure 1, table 2), these results will provide helpful information ology of each gout/hyperuricaemia subtype and the development for the development of novel subtype-­specific genome tailor-­ of novel subtype-­specific genome tailor-made­ medicine/preven- made medicines and/or prevention for gout and hyperuricaemia. tion of gout and hyperuricaemia. Adaptive evolution results from adaptation to environmental changes over generations. Selection pressure analysis using Author affiliations SDS in the present study has elucidated which genes have been 1Department of Integrative Physiology and Bio-­Nano Medicine, National Defense involved in adaptive evolution over the last 2000–3000 years Medical College, Tokorozawa, Japan

Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 663 Crystal arthropathies

2Medical Squadron, Air Base Group, Western Aircraft Control and Warning Wing, Nakajima, K Maehara and M Kirihara (National Defense Medical College) for genetic Japan Air Self-­Defense Force, Kasuga, Japan analysis and to N Hamajima at Nagoya University Graduate School of Medicine for 3Division of Department of Nursing, Nagoya University Graduate School of Medicine, sample collection. We are also indebted to H Fujiwara (Midorigaoka Hospital), M Nagoya, Japan Senda and M Hamamura (Ryougoku East Gate Clinic). 4Department of General Medicine, National Defense Medical College, Tokorozawa, Collaborators Members of the Japan Gout Genomics Consortium (Japan Gout) Japan are: Katsuhisa Inoue (Department of Biopharmaceutics, School of Pharmacy, 5Department of Medical Biochemistry, Kurume University School of Medicine, Tokyo University of Pharmacy and Life Sciences, Tokyo), Tomoya Yasujima, Hiroaki Kurume, Japan Yuasa (Department of Biopharmaceutics, Graduate School of Pharmaceutical 6Division of Human Genetics, Department of Integrated Genetics, National Institute Sciences, Nagoya City University, Nagoya, Aichi), Hiroaki Ikezaki, Masayuki of Genetics, Mishima, Japan Murata (Department of Environmental Medicine and Infectious Disease, Graduate 7Graduate School of Information Science and Technology, Hokkaido University, School of Medical Sciences, Kyushu University), Keito Morimoto (Department Sapporo, Japan of Pharmacy, the University of Tokyo Hospital, Tokyo), Mitsuhiro Yokota 8Department of Epidemiology for Community Health and Medicine, Kyoto Prefectural (Department of Medical Biochemistry, Kurume University School of Medicine, University of Medicine, Kyoto, Japan Kurume, Fukuoka), Sahoko Ichihara (Department of Environmental and Preventive 9Department of Preventive Medicine, Nagoya University Graduate School of Medicine, Jichi Medical University School of Medicine, Shimotsuke, Tochigi), Medicine, Nagoya, Japan Tatsuaki Matsubara (Department of Internal Medicine, School of Dentistry, 10Laboratory of Public Health, School of Food and Nutritional Sciences, University of Aichi Gakuin University, Nagoya, Aichi), Toshimitsu Ito (Department of Internal Shizuoka, Shizuoka, Japan Medicine, Self-­Defense Forces Central Hospital, Tokyo), Miki Ueno, Kimiko Hayano 11Department of Public Health, Nagoya City University Graduate School Medical (Division of Nursing, National Defense Medical College, Tokorozawa, Saitama), Science, Nagoya, Japan Kunio Mizutari, Akihiro Shiotani (Department of Otolaryngology-­Head and 12Midorigaoka Hospital, Takatsuki, Japan Neck Surgery, National Defense Medical College, Tokorozawa, Saitama), Yuka 13Kyoto Industrial Health Association, Kyoto, Japan Miyoshi, Satoko Suzuki, Satoko Iwasawa (Department of Preventive Medicine 14Ryougoku East Gate Clinic, Tokyo, Japan and Public Health, National Defense Medical College, Tokorozawa, Saitama), 15Nagase Clinic, Tokyo, Japan Yuka Aoyagi, Yuka Aoki (Department of Integrative Physiology and Bio-­Nano 16Akasaka Central Clinic, Tokyo, Japan Medicine, National Defense Medical College, Tokorozawa, Saitama), Tatsuya 17Department of Preventive Medicine, Faculty of Medicine, Saga University, Saga, Saitoh (Laboratory of Bioresponse Regulation, Graduate School of Pharmaceutical Japan Sciences, Osaka University, Suita, Osaka). 18Clinical Research Center, Saga University Hospital, Saga, Japan 19Department of Preventive Medicine, Institute of Biomedical Sciences, Tokushima Contributors AN, MNakatochi and HMatsuo conceived and designed this study. University Graduate School, Tokushima, Japan KY, HNakaoka, II, YO and NS assisted with research design. TS, KO, HO, MNagase, 20Department of Health Science, Shiga University of Medical Science, Otsu, Japan YH, MKawaguchi, MTakao, RS, THosoya, KI and HMatsuo collected and analysed 21Department of Public Health, Faculty of Medicine, Kindai University, Osaka-­Sayama, clinical data of cases. TK, AHishida, KK, MW, DM, TTamura, TNishiyama, CS, SKK, Japan NT, NK, HMikami, TTakezaki, KM, SSuzuki and KW collected and analysed clinical 22Department of Statistical Genetics, Osaka University Graduate School of Medicine, data of controls. AN, YKawamura, KY, SShimizu, THigashino, YTakada, ID, YKamatani, Suita, Japan MKubo, NS and HMatsuo performed genetic analysis. AN, MNakatochi, YKawamura, 23Department of Respiratory Medicine and Clinical Immunology, Osaka University HNakaoka, SShimizu, YS, TNakamura, HNakashima and YO performed statistical Graduate School of Medicine, Suita, Japan analysis. HMatsuo organized this collaborative study. KY, HNakaoka, THigashino, 24Department of Urology, National Defense Medical College, Tokorozawa, Japan MTsunoda, ID, AHozawa, KH, YToyoda, YKubota, TTakada, HS, BS, TM, TR, THosoya, 25Department of Surgery, National Defense Medical College, Tokorozawa, Japan YKamatani, MKubo, KI, KW, II, YO and NS provided intellectual input and assisted 26Faculty of Medical Science, Teikyo University of Science, Tokyo, Japan with the preparation of the manuscript. AN, MNakatochi, YKawamura, YO, NS and 27Laboratory for Mathematics, National Defense Medical College, Tokorozawa, Japan HMatsuo wrote the manuscript. 28 Department of Preventive Medicine and Public Health, National Defense Medical Funding This study was supported by grants from the Ministry of Education, College, Tokorozawa, Japan Culture, Sports, Science and Technology (MEXT) of Japan, including JSPS Kakenhi 29Group of Privacy Controls, Tohoku Medical Megabank Organization, Sendai, Japan 30 Grants (Nos. 16H06279, 16H01808, 17H04128, 18KK0247, 19K22786 and Department of Preventive Medicine and Epidemiology, Tohoku Medical Megabank 25293145) and a Grant-­in-­Aid for Scientific Research onI nnovative Areas Organization, Tohoku University, Sendai, Japan 31 (No. 221S0002), the Ministry of Defense, and the Kawano Masanori Memorial Department of Bioinformatics and Genomics, Graduate School of Advanced Foundation for the Promotion of Pediatrics and the Gout Research Foundation of Preventive Medical Sciences, Kanazawa University, Kanazawa, Japan Japan. The J-­MICC Study was supported by Grants-­in-­Aid for Scientific Research 32Department of Pharmacy, The University of Tokyo Hospital, Tokyo, Japan 33 from MEXT, including those for Priority Areas (No. 17015018) and Innovative Areas Department of Pediatrics and Adolescent Medicine, First Faculty of Medicine, (No. 221S0001), as well as by a JSPS Kakenhi grant (No. 16H06277). The BioBank Charles University and General University Hospital, Prague, Czech Republic Japan Project was supported by grants from the Japan Agency for Medical Research 34Institute of Rheumatology, Prague, Czech Republic 35 and Development since April 2015, and the Ministry of Education, Culture, Sports, Department of Biochemisty, University of Otago, Dunedin, New Zealand Science and Technology from April 2003 to March 2015. 36Cancer Prevention Center, Chiba Cancer Center Research Institute, Chiba, Japan 37Department of International Island and Community Medicine, Kagoshima Competing interests TTakada, NS and HMatsuo have a patent pending based on University Graduate School of Medical and Dental Sciences, Kagoshima, Japan the work reported in this paper. 38 Division of Cancer Epidemiology and Prevention, Aichi Cancer Center Research Patient and public involvement Patients and/or the public were not involved in Institute, Nagoya, Japan the design, or conduct, or reporting or dissemination plans of this research. 39Department of Epidemiology, Nagoya University Graduate School of Medicine, Nagoya, Japan Patient consent for publication Not required. 40 Division of Kidney and Hypertension, Department of Internal Medicine, Jikei Ethics approval This study was approved by the institutional ethical University School of Medicine, Tokyo, Japan committees and written consent was obtained from all of its participants. All 41 Department of Pathophysiology and Therapy in Chronic Kidney Disease, Jikei procedures involved were performed in accordance with the Declaration of University School of Medicine, Tokyo, Japan Helsinki. 42Laboratory for Statistical Analysis, RIKEN Center for Integrative Medical Sciences, Yokohama, Japan Provenance and peer review Not commissioned; externally peer reviewed. 43Center for Genomic Medicine, Kyoto University Graduate School of Medicine, Data availability statement No data are available. Kyoto, Japan 44 Open access This is an open access article distributed in accordance with the RIKEN Center for Integrative Medical Sciences, Yokohama, Japan 45 Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits Department of Pathophysiology, Tokyo University of Pharmacy and Life Science, others to copy, redistribute, remix, transform and build upon this work for any Hachioji, Japan 46 purpose, provided the original work is properly cited, a link to the licence is given, Laboratory of Statistical Immunology, Immunology Frontier Research Center (WPI-­ and indication of whether changes were made. See: https://​creativecommons.​org/​ IFReC), Osaka University, Suita, Japan licenses/by/​ ​4.0/.​ Acknowledgements We would like to thank all the participants for their generous involvement in this study. We are grateful to the staff of the institutions participating ORCID iDs in the J-­MICC Study for their outstanding assistance in collecting samples and Masahiro Nakatochi http://orcid.​ ​org/0000-​ ​0002-​1838-​4837 clinical information. We are indebted to M Miyazawa, K Morichika, M Sakiyama, M Yukinori Okada http://orcid.​ ​org/0000-​ ​0002-​0311-​8472

664 Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 Crystal arthropathies

References 25 Kolz M, Johnson T, Sanna S, et al. Meta-analysis­ of 28,141 individuals identifies 1 dalbeth N, Merriman TR, Stamp LK. Gout. Lancet 2016;388:2039–52. common variants within five new loci that influence uric acid concentrations. PLoS 2 dalbeth N, Choi HK, Joosten LAB, et al. Gout. Nat Rev Dis Primers 2019;5:69. Genet 2009;5:e1000504. 3 ichida K, Matsuo H, Takada T, et al. Decreased extra-renal­ urate excretion is a common 26 Higashino T, Matsuo H, Sakiyama M, et al. Common variant of PDZ domain cause of hyperuricemia. Nat Commun 2012;3:764. containing 1 (PDZK1) gene is associated with gout susceptibility: a replication 4 Wortmann RL. Disorders of purine and pyrimidine metabolism. In: Fauci AS, Braunwald study and meta-analysis­ in Japanese population. Drug Metab Pharmacokinet E, Kasper D, et al, eds. Harrison's Principles of Internal Medicine. 17th edn. New York, 2016;31:464–6. N.Y: McGraw-­Hill, 2008: 2444–9. 27 li M, Li Q, Li C-­G, et al. Genetic polymorphisms in the PDZK1 gene and 5 matsuo H, Nakayama A, Sakiyama M, et al. ABCG2 dysfunction causes hyperuricemia susceptibility to gout in male Han Chinese: a case-­control study. Int J Clin Exp Med due to both renal urate underexcretion and renal urate overload. Sci Rep 2015;8:13911–8. 2014;4:3755. 28 Ketharnathan S, Leask M, Boocock J, et al. A non-­coding genetic variant maximally 6 matsuo H, Yamamoto K, Nakaoka H, et al. Genome-­wide association study of associated with serum urate levels is functionally linked to HNF4A-­dependent PDZK1 clinically defined gout identifies multiple risk loci and its association with clinical expression. Hum Mol Genet 2018;27:3964–73. subtypes. Ann Rheum Dis 2016;75:652–9. 29 Yasukochi Y, Sakuma J, Takeuchi I, et al. Identification of CDC42BPG as a novel 7 nakayama A, Nakaoka H, Yamamoto K, et al. GWAS of clinically defined gout and susceptibility locus for hyperuricemia in a Japanese population. Mol Genet Genomics subtypes identifies multiple susceptibility loci that include urate transporter genes. 2018;293:371–9. Ann Rheum Dis 2017;76:869–77. 30 lee J, Lee Y, Park B, et al. Genome-­wide association analysis identifies multiple 8 Kawai Y, Mimori T, Kojima K, et al. Japonica array: improved genotype imputation by loci associated with kidney disease-­related traits in Korean populations. PLoS One designing a population-­specific SNP array with 1070 Japanese individuals. J Hum 2018;13:e0194044. Genet 2015;60:581–7. 31 matsuo H, Takada T, Ichida K, et al. Common defects of ABCG2, a high-­capacity urate 9 Wallace SL, Robinson H, Masi AT, et al. Preliminary criteria for the classification of the exporter, cause gout: a function-­based genetic analysis in a Japanese population. Sci acute arthritis of primary gout. Arthritis Rheum 1977;20:895–900. Transl Med 2009;1:5ra11. 10 Hamajima N, J-­MICC Study Group. The Japan multi-­institutional collaborative cohort 32 matsuo H, Ichida K, Takada T, et al. Common dysfunctional variants in ABCG2 are a study (J-­MICC study) to detect gene-­environment interactions for cancer. Asian Pac J major cause of early-­onset gout. Sci Rep 2013;3:2014. Cancer Prev 2007;8:317–23. 33 Kawamura Y, Nakaoka H, Nakayama A, et al. Genome-­wide association study 11 asai Y, Naito M, Suzuki M, et al. Baseline data of Shizuoka area in the Japan revealed novel loci which aggravate asymptomatic hyperuricaemia into gout. Ann multi-­institutional collaborative cohort study (J-­MICC study). Nagoya J Med Sci Rheum Dis 2019;78:1430–7. 2009;71:137–44. 34 druckmann R, Druckmann M-­A. Progesterone and the immunology of pregnancy. J 12 delaneau O, Zagury J-­F, Marchini J. Improved whole-­chromosome phasing for disease Steroid Biochem Mol Biol 2005;97:389–96. and population genetic studies. Nat Methods 2013;10:5–6. 35 Boomgaarden I, Vock C, Klapper M, et al. Comparative analyses of disease risk genes 13 das S, Forer L, Schönherr S, et al. Next-­generation genotype imputation service and belonging to the acyl-­CoA synthetase medium-­chain (ACSM) family in human liver methods. Nat Genet 2016;48:1284–7. and cell lines. Biochem Genet 2009;47:739–48. 14 Willer CJ, Li Y, Abecasis GR. Metal: fast and efficient meta-­analysis of genomewide 36 Hart TC, Gorry MC, Hart PS, et al. Mutations of the UMOD gene are responsible for association scans. Bioinformatics 2010;26:2190–1. medullary cystic kidney disease 2 and familial juvenile hyperuricaemic nephropathy. J 15 Bulik-­Sullivan BK, Loh P-­R, Finucane HK, et al. LD score regression distinguishes Med Genet 2002;39:882–92. confounding from polygenicity in genome-­wide association studies. Nat Genet 37 Jovanovic DV, Boumsell L, Bensussan A, et al. CD101 expression and function in 2015;47:291–5. normal and rheumatoid arthritis-­affected human T cells and monocytes/macrophages. 16 Field Y, Boyle EA, Telis N, et al. Detection of human adaptation during the past 2000 J Rheumatol 2011;38:419–28. years. Science 2016;354:760–4. 17 okada Y, Momozawa Y, Sakaue S, et al. Deep whole-­genome sequencing reveals 38 Young JD, Yao SYM, Baldwin JM, et al. The human concentrative and recent selection signatures linked to evolution and disease risk of Japanese. Nat equilibrative nucleoside transporter families, SLC28 and SLC29. Mol Aspects Med Commun 2018;9:1631. 2013;34:529–47. 18 li C, Li Z, Liu S, et al. Genome-­wide association analysis identifies three new risk loci 39 Young JD. The SLC28 (CNT) and SLC29 (ENT) nucleoside transporter families: a 30-­ for gout arthritis in Han Chinese. Nat Commun 2015;6:7041. year collaborative odyssey. Biochem Soc Trans 2016;44:869–76. 19 sakiyama M, Matsuo H, Nakaoka H, et al. Common variant of BCAS3 is associated 40 ogura M, Takarada T, Nakamichi N, et al. Exacerbated vulnerability to oxidative stress with gout risk in Japanese population: the first replication study after gout GWAS in in astrocytic C6 glioma cells with stable overexpression of the glutamine transporter Han Chinese. BMC Med Genet 2018;19:96. slc38a1. Neurochem Int 2011;58:504–11. 20 sakiyama M, Matsuo H, Nakaoka H, et al. Identification of rs671, a common ariantv 41 Tin A, Marten J, Halperin Kuhns VL, et al. Target genes, variants, tissues and of ALDH2, as a gout susceptibility locus. Sci Rep 2016;6:25360. transcriptional pathways influencing human serum urate levels. Nat Genet 21 sakiyama M, Matsuo H, Akashi A, et al. Independent effects of ADH1B and ALDH2 2019;51:1459–74. common dysfunctional variants on gout risk. Sci Rep 2017;7:2500. 42 nakayama A, Matsuo H, Nakaoka H, et al. Common dysfunctional variants of ABCG2 22 anzai N, Miyazaki H, Noshiro R, et al. The multivalent PDZ domain-­containing protein have stronger impact on hyperuricemia progression than typical environmental risk PDZK1 regulates transport activity of renal urate-­anion exchanger URAT1 via its C factors. Sci Rep 2014;4:5227. terminus. J Biol Chem 2004;279:45942–50. 43 matsuo H, Tomiyama H, Satake W, et al. ABCG2 variant has opposing effects on onset 23 shimizu T, Sugiura T, Wakayama T, et al. PDZK1 regulates breast cancer resistance ages of Parkinson’s disease and gout. Ann Clin Transl Neurol 2015;2:302–6. protein in small intestine. Drug Metab Dispos 2011;39:2148–54. 44 cutler RG. Urate and ascorbate: their possible roles as antioxidants in determining 24 nakatochi M, Kanai M, Nakayama A, et al. Genome-wide­ meta-­analysis identifies longevity of mammalian species. Arch Gerontol Geriatr 1984;3:321–48. multiple novel loci associated with serum uric acid levels in Japanese individuals. 45 auton A, Brooks LD, Durbin RM, et al. A global reference for human genetic variation. Commun Biol 2019;2:115. Nature 2015;526:68–74.

Nakayama A, et al. Ann Rheum Dis 2020;79:657–665. doi:10.1136/annrheumdis-2019-216644 665 Letters To consider or not antimalarials as a European trials, not yet recruiting, will assess the efficacy of CQ/HCQ in preventing symptomatic Covid-19 in healthcare prophylactic intervention in the SARS-­CoV-2 workers, or other individuals at significant risk (​ClinicalTrials.​ (Covid-19) pandemic gov Identifiers: NCT04303507 and NCT04304053). CQ and HCQ have been used for autoimmune rheumatic diseases since 1940s, being safe and well tolerated in most 5 These days, the entire scientific community is facing the severe patients. Data from the literature, including our own experi- ence, reported a low incidence of side effects, generally mild acute respiratory syndrome coronavirus-2 (SARS-­CoV-2) emer- 5 6 gency, characterised last 11 March by WHO as a pandemic. to moderate. The most serious complication (ie, the retinal Social behaviour modification measures may somehow limit the toxicity) depends on weight-­adjusted daily dose and, most of all, spreading of the infection. However, in the case of an extremely cumulative dose of antimalarials. Similarly, the (rare) cardiotox- icity seems to be related to the cumulative dose, even if mecha- contagious pathogen, the huge number of infected people may 5 be a challenge for the health system. What if there was a prophy- nistic evidence is still lacking. lactic drug? Mass drug administration is an intervention used as malaria-­ In the light of their in vitro effect and early clinical results, control measure delivering safe and inexpensive drugs to prevent antimalarial drugs chloroquine (CQ) and hydroxychloroquine or alleviate symptoms and morbidity while reducing transmis- (HCQ) have been proposed for patients with SARS-CoV2-­ ­ sion and improving global health. Is it ethical to propose CQ or related pneumonia (Covid-19) and are now included in the HCQ for preventing the spreading of Covid-19 without any data Chinese guidelines for the management of Covid-19 (version 7, coming from evidence-­based medicine? Even though ‘primum 3 March 2020). non nocere’: is it permissible to take a controlled risk in the event Antiviral activity of antimalarials has been known for more of a pandemic? In such a case: would it be reasonable to consider than 10 years (see online supplementary text). antimalarials as primary prophylaxis in healthy subjects living in Recently, Wang et al demonstrated that at low micromolar highest risk regions or, at least, to use them in those tested posi- concentration CQ was able to potently block viral replication of tive for Covid-19 but still asymptomatic? The advantage of CQ Covid-19, in vitro; the effective concentration of CQ was that or HCQ is that they are safe and inexpensive to administer for a achievable in patients receiving 500 mg/daily.1 HCQ also showed relatively short time, therefore good candidates for mass admin- an anti-­SARS-­CoV-2 effect, decreasing the viral replication in istration, whenever not contraindicated. Waiting for supportive a time and concentration-dependent­ manner.2 Interestingly, CQ data from clinical trials, the scientific community is moving and HCQ prevent the viral replication also at entry stage (ie, towards pre-­emptive use of antimalarials (see online supplemen- when added in cell culture before the viral challenge).2 To date, tary figure 1). If mass prophylaxis was accepted as an option more than 100 patients have been treated with CQ showing worldwide, this would raise the question of whether there is promising results.3 A very recent study showed that, already enough supply of CQ and HCQ to support this approach. after 6 days, HCQ induced a negativity of viral RNA in nasopha- Francesca Romana Spinelli ‍ ‍ , Fulvia Ceccarelli ‍ ‍ , Manuela Di Franco, ryngeal sample: 70% of patients treated with HCQ alone and Fabrizio Conti 100% of those treated with HCQ in combination with azith- Dipartimento di Scienze Cliniche, Internistiche, Anestesiologiche e Cardiovascolari— romycin determined a viral clearance compared with 12.5% Reumatologia, Sapienza University of Rome, Roma, Lazio, Italy of patients who did not receive HCQ.4 Table 1 summarises the data available to date on CQ and HCQ. Many clinical trials on Correspondence to Dr Francesca Romana Spinelli, Dipartimento di Scienze Cliniche, Internistiche, Anestesiologiche e Cardiovascolari - Reumatologia, Sapienza the use of CQ or HCQ are now recruiting patients. Two more University of Rome, Roma, Lazio 00161, Italy; ​francescaromana.​spinelli@​uniroma1.​it

Handling editor Josef S Smolen Table 1 Preclinical and clinical data on chloroquine (CQ) and Contributors FRS and FC discussed the topic of the letter, did the literature search hydroxychloroquine (HCQ) in coronavirus disease 2019 (Covid-19) and drafted the manuscript. MDF and FC reviewed and approved the manuscript. Type of Funding The authors have not declared a specific grant for this research from any study Main results funding agency in the public, commercial or not-­for-­profit sectors. Wang et al1 In vitro At low micromolar concentration, CQ blocks viral infection Competing interests None declared. at both entry and at post-­entry stages of the 2019-­nCoV Patient and public involvement Patients and/or the public were not involved in infection in Vero E6 cells. the design, or conduct, or reporting, or dissemination plans of this research. Yao et al2 In vitro HCQ is more potent than CQ in inhibiting viral infection Patient consent for publication Not required. at entry and post-­entry stages; EC50 values CQ and HCQ decreased with longer incubation times providing higher Provenance and peer review Not commissioned; internally peer reviewed. intracellular concentrations and a better antiviral effect. © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and Suggested dosing for HCQ: 400 mg/two times a day at day permissions. Published by BMJ. 1, followed by 200 mg/two times a day. 3 ►► Additional material is published online only. To view please visit the journal Gao et al Case CQ phosphate is superior to the control treatment in online (http://​dx.doi.​ ​org/​10.​1136/annrheumdis-​ ​2020-217367).​ series inhibiting the exacerbation of pneumonia, improving lung imaging findings, promoting a virus-­negative conversion FRS and FC contributed equally. and shortening the disease course. No severe adverse events were reported. Gautret et al4 Case HCQ induces viral clearance after 6 days of treatment, To cite Spinelli FR, Ceccarelli F, Di Franco M, et al. Ann Rheum Dis control either alone or in combination with azithromycin 2020;79:666–667. (respectively, 70% and 100% negative nasopharyngeal samples among treated patients compared with 12.5% of Received 18 March 2020 untreated patients). Revised 20 March 2020 EC, effective concentration 50; nCoV, novel coronavirus. Accepted 22 March 2020

666 Ann Rheum Dis May 2020 Vol 79 No 5 Letters

Published Online First 1 April 2020 Ann Rheum Dis 2020;79:666–667. doi:10.1136/annrheumdis-2020-217367

ORCID iDs Francesca Romana Spinelli http://​orcid.​org/​0000-​0003-​1969-​2097 Fulvia Ceccarelli http://​orcid.​org/​0000-​0001-​5026-​8783

References 1 Wang M, Cao R, Zhang L, et al. Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-­nCoV) in vitro. Cell Res 2020;30:269–71. 2 Yao X, Ye F, Zhang M, et al. In vitro antiviral activity and projection of optimized dosing design of hydroxychloroquine for the treatment of severe acute respiratory syndrome coronavirus 2 (SARS-­CoV-2). Clin Infect Dis 2020. doi:10.1093/cid/ciaa237. [Epub ahead of print: 9 Mar 2020]. 3 Gao J, Tian Z, Yang X. Breakthrough: chloroquine phosphate has shown apparent efficacy in treatment of COVID-19 associated pneumonia in clinical studies. Biosci Trends 2020;14:72–3. 4 Gautret P, Lagier J-­C, Parola P, et al. Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an open-­label non-­randomized clinical trial. Int J Antimicrob Agents 2020:105949. 5 Schrezenmeier E, Dörner T. Mechanisms of action of hydroxychloroquine and chloroquine: implications for rheumatology. Nat Rev Rheumatol 2020;16:155–66. 6 Spinelli FR, Moscarelli E, Ceccarelli F, et al. Treating lupus patients with antimalarials: analysis of safety profile in a single-­center cohort. Lupus 2018;27:1616–23.

­

­

Ann Rheum Dis May 2020 Vol 79 No 5 667 Letters

Table 1 Clinical characteristics of the patients with confirmed or suspected COVID-19 Contact with Clinical picture a known Confirmed highly suggestive COVID-19 COVID-19 of COVID-19 patient Number of patients 4 4 5 Age (years) (mean±SD) 58±5 56±8 54±12 Female, n (%) 4 (100) 3 (75) 4 (80) Comorbidities, n (%) Hypertension 1 (25) 2 (50) 1 (20) Diabetes 0 0 0 Cardiovascular disease 0 0 1 (20) Other 4 (100) 4 (100) 3 (60) Smoking, n (%) Active 1 (25) 0 0 Previous 2 (50) 3 (75) 1 (20) Rheumatological diagnosis RA, n (%) 3 (75) 3 (75) 5 (100) SpA/PA,* n (%) 1 (25) 1* (25) 0 Rheumatological treatment, n (%) bDMARD Clinical course of COVID-19 in a series of Adalimumab 0 0 1 (20) patients with chronic arthritis treated with Etanercept 2 (50) 2 (50) 0 Abatacept 1 (25) 1 (25) 0 immunosuppressive targeted therapies Tocilizumab 0 0 1 (20) tsDMARD Tofacitinib 1 (25) 0 1 (20) Different viral agents are associated with an increased risk of Baricitinib 0 1 (25) 2 (40) more severe disease course and respiratory complications in Concomitant csDMARD 1–3 immunocompromised patients. The recent outbreak of Methotrexate 2 (50) 1 (25) 3 (60) severe acute respiratory syndrome coronavirus 2 (SARS-­CoV-2) Leflunomide 1 (25) 0 1 (20) disease 2019 (COVID-19) responsible for a severe acute respi- Sulfasalazine 0 1 (25) 0 ratory syndrome (SARS) represents a source of concern for the Concomitant hydroxychloroquine 1 (25) 2 (50) 2 (40) management of patients with inflammatory rheumatic diseases. Low-­dose glucocorticoids* 2 (50) 2 (50) 2 (40) Lombardy is the region in Northern Italy with the highest inci- Known contact with COVID-19 0 1 (25) 5 (100) dence of COVID-19 cases, with more than 33 000 confirmed Symptoms, n (%) patients and 1250 requiring admission to the intensive care Fever 4 (100) 1 (25) 0 unit within 1 month. Since the first reports of COVID-19 cases Non-­productive cough 3 (75) 2 (50) 0 in Italy, we have circulated a survey with a 2-­week follow-up­ Sputum production 1 (25) 0 0 contact to patients with chronic arthritis treated with biolog- Rhinorrhea 2 (50) 1 (25) 0 ical disease-­modifying antirheumatic drugs (bDMARDs) or Sore throat 0 0 0 targeted synthetic disease-­modifying antirheumatic drugs Fatigue 4 (100) 2 (50) 0 (tsDMARDs) followed up at our biological outpatient clinic in Myalgia 2 (50) 1 (25) 0 Pavia, Lombardy. The survey investigated the patients’ health Arthralgia 1 (25) 1 (25) 0 conditions, the presence of contacts with subjects known to be Anosmia/dysgeusia 3 (75) 3 (75) 0 affected by COVID-19 and management of the DMARDs during Dyspnoea at rest 1 (25) 0 0 the first few weeks of pandemic. All patients had provided their Dyspnoea on exertion 2 (50) 1 (25) 0 informed consent for the use of personal and clinical data for Headache 2 (50) 0 0 scientific purposes, and no patient refused to participate. Diarrhoea 1 (25) 0 0 During the first month, we have collected information on Nausea/vomiting 0 0 0 320 patients (female 68%, mean age 55±14 years) treated with Chest X-­ray performed 4 (100) 0† 0 bDMARDs or tsDMARDs (57% with rheumatoid arthritis, 43% Chest X-­ray pathological findings 0 0 0 with spondyloarthritis, 52% treated with tumour necrosis factor Hospital admission 1 (25) 0 0 inhibitors, 40% with other bDMARDs and 8% with tsDMARDs). *Glucocorticoids≤5 mg/day prednisone equivalent. As shown in table 1, four were confirmed cases of COVID-19 †Subject to home quarantine. identified through rhinopharyngeal swabs. Another four patients bDMARD, biological disease-­modifying antirheumatic drug; COVID-19, coronavirus disease 2019; csDMARD, conventional synthetic disease-­modifying antirheumatic drug; PA, psoriatic reported symptoms which were highly suggestive of COVID-19. arthritis; RA, rheumatoid arthritis; SpA, spondyloarthritis; tsDMARD, targeted synthetic Five additional patients with reported certain contacts remained disease-­modifying antirheumatic drug. asymptomatic at the end of the 2-­week observation period. All patients with confirmed COVID-19 received at least one were on previous stable treatment with hydroxychloroquine. All antibiotic course, and the hospitalised patient also received anti- patients with symptoms of infection temporarily withdrew the viral therapy and hydroxychloroquine. Overall, five patients bDMARD or tsDMARD at the time of symptom onset. To date,

Ann Rheum Dis May 2020 Vol 79 No 5 667 Letters there have been no significant relapses of the rheumatic disease. properly cited, appropriate credit is given, any changes made indicated, and the use None of the patients with a confirmed diagnosis of COVID-19 is non-­commercial. See: http://​creativecommons.​org/​licenses/​by-​nc/​4.​0/. or with a highly suggestive clinical picture developed severe © Author(s) (or their employer(s)) 2020. Re-­use permitted under CC BY-­NC. No respiratory complications or died. Only one patient, aged 65, commercial re-­use. See rights and permissions. Published by BMJ. required admission to hospital and low-flow­ oxygen supplemen- tation for a few days. Our findings do not allow any conclusions on the incidence To cite Monti S, Balduzzi S, Delvino P, et al. Ann Rheum Dis 2020;79:667–668. rate of SARS-­CoV-2 infection in patients with rheumatic diseases, nor on the overall outcome of immunocompromised patients Received 26 March 2020 affected by COVID-19. A high level of vigilance and strict Accepted 27 March 2020 Published Online First 2 April 2020 follow-­up should be maintained on these patients, including the exclusion of superimposed infections. However, our preliminary Ann Rheum Dis 2020;79:667–668. doi:10.1136/annrheumdis-2020-217424 experience shows that patients with chronic arthritis treated with ORCID iD bDMARDs or tsDMARDs do not seem to be at increased risk of Sara Monti http://​orcid.​org/​0000-​0002-​1800-​6772 respiratory or life-­threatening complications from SARS-­CoV-2 compared with the general population. References These findings are not surprising as the severe respiratory 1 Memoli MJ, Athota R, Reed S, et al. The natural history of influenza infection in the complications caused by coronaviruses are thought to be driven severely immunocompromised vs nonimmunocompromised hosts. Clin Infect Dis 2014;58:214–24. by the aberrant inflammatory and cytokine response perpetu- 4 2 Noreña I, Fernández-­Ruiz M, Aguado JM. Viral infections in the biologic therapy era. ated by the host immune system. During different coronavirus Expert Rev Anti Infect Ther 2018;16:781–91. outbreaks, such as SARS and Middle East respiratory syndrome, 3 D’Antiga L. Coronaviruses and immunosuppressed patients. The facts during the third there has been no increased mortality reported in patients under- epidemic. Liver Transpl 2020. doi:10.1002/lt.25756 going immunosuppression for organ transplantation, cancer 4 Shi Y, Wang Y, Shao C, et al. COVID-19 infection: the perspectives on immune 3 5 responses. Cell Death Differ 2020;382. or autoimmune diseases. Accordingly, among 700 patients 5 Hui DS, Azhar EI, Kim Y-­J, et al. Middle East respiratory syndrome coronavirus: risk admitted for severe COVID-19 at our hospital (a referral factors and determinants of primary, household, and nosocomial transmission. Lancet centre for SARS-­CoV-2 infection) during last month, none was Infect Dis 2018;18:e217–27. receiving bDMARDs or tsDMARDs. Although continuous surveillance of patients with rheumatic diseases receiving immunosuppressive drugs is warranted, these data can support rheumatologists for the management and coun- selling of their patients, avoiding the unjustifiable preventive withdrawal of DMARDs, which could lead to an increased risk of relapses and morbidity from the chronic rheumatological condition.

Sara Monti ‍ ‍ ,1 Silvia Balduzzi,2 Paolo Delvino,3 Elisa Bellis,3 Verdiana Serena Quadrelli,3 Carlomaurizio Montecucco3 1Rheumatology Department, IRCCS Fondazione Policlinico San Matteo, Pavia, Italy 2Rheumatology Department, IRCCS Policlinico San Matteo, Pavia, Italy 3Rheumatology Department, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy Correspondence to Dr Sara Monti, Rheumatology, IRCCS Fondazione Policlinico S Matteo, Pavia 27100, Italy; ​sara.​saramonti@​gmail.​com

Handling editor Josef S Smolen Contributors SM and CM contributed to the design of the project interpretation and analysis of the data, and writing of the manuscript. SB, PD, EB and VSQ contributed to the collection of data, interpretation and analysis of the data, and review 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 None declared. Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Patient consent for publication Not required. Provenance and peer review Not commissioned; internally peer reviewed.

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

668 Ann Rheum Dis May 2020 Vol 79 No 5 Letters

­ ​ ​ ​ ​ ​ ​ ​

​ ​ ​ ​ ​ ​

COVID-19 in a patient with systemic sclerosis treated with tocilizumab for SSc-­ILD

During the current global outbreak of coronavirus disease 2019 (COVID-19), risk stratification of patients is of utmost importance. Currently, patients >65 years and those with pre-­existing medical conditions such as cardiovascular disease, chronic respiratory disease or diabetes mellitus are considered at higher risk for severe disease.1 The Swiss Federal Office of Public Health, like similar authorities around the world, has additionally included patients on immunosuppressants in the high-­risk group for developing severe COVID-19.2 However, at this moment we do not have enough evidence either to support or to reject this assumption. We report the case of a 57-­year-old­ woman with systemic scle- rosis (SSc) who developed COVID-19. Comorbidities were insulin-­ dependent type 2 diabetes mellitus and WHO grade I obesity. The anti-­topoisomerase I antibody-­positive patient was diagnosed with SSc in 2017. SSc-­associated interstitial lung disease (SSc-ILD),­ with cough and exertion dyspnoea, was the leading organ manifesta- tion, associated with symmetrical, non-­erosive polyarthritis, and elevated acute phase reactants. Treatment with the anti-interleukin­ (IL) 6 receptor blocker tocilizumab, with 8 mg/kg body weight every 4 weeks intravenously, was started, leading to a good control of both arthritis and SSc-­ILD, with gradual improvement of muscu- loskeletal and respiratory symptoms, lung function and high-­ resolution CT imaging (figure 1). At the last annual assessment in January 2020, her forced vital capacity and carbon monoxide lung diffusion capacity were 92% and 70% of the respective predicted values. Tocilizumab was continued at 5-week­ intervals. On 12 March 2020, 4 weeks after the last tocilizumab infusion, the patient presented to our hospital’s emergency department with cough, headache and general malaise since about 1 week. She reported contact with a patient with COVID-19 2 weeks earlier. As she was in relatively good general health condition, subfebrile

668 Ann Rheum Dis May 2020 Vol 79 No 5 Letters

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 CM had congress support from Actelion and Roche, consultancy fees from Boehringer Ingelheim and Geneva Romfarm; RD had research support from Actelion and Pfizer; BM had grant/research support from AbbVie, Protagen, Novartis, and congress support from Pfizer, Roche, Actelion, and MSD, speakers honorary from Novartis. In addition, BM has a patent mir-29 for the treatment of systemic sclerosis registered (US8247389, EP2331143). OD had speaker fees from Actelion, Bayer, Boehringer Ingelheim, Medscape, Novartis, Roche, Menarini, Mepha, MSD, iQone, Pfizer, consultancy fees from Abbvie, Actelion, Acceleron Pharma, Amgen, AnaMar, Bayer, Baecon Discovery, Blade Therapeutics, Figure 1 High-­resolution CT (HRCT) of the chest, 8 weeks before Boehringer, CSL Behring, ChemomAb, Curzion Pharmaceuticals, Ergonex, Galapagos NV, GSK, Glenmark Pharmaceuticals, Inventiva, Italfarmaco, iQvia, medac, Medscape, onset of coronavirus disease 2019 (COVID-19). Ground glass and Mitsubishi Tanabe Pharma, MSD, Roche, Sanofi, UCB, Amgen, Lilly, Target BioScience, reticulation as signs of systemic sclerosis (SSc)-­associated interstitial Pfizer, project scoring fees from Abbvie, interview fees from Catenion, travel support lung disease are visible, with an overall lung involvement of <20%. from Pfizer, research grants from Actelion, Bayer, Boehringer Ingelheim, Competitive Drug Development International Ltd, Mitsubishi Tanabe, patent issued mir-29 for the treatment of systemic sclerosis (US8247389, EP2331143). (37.6°C), with unchanged bibasal lung crackles and no significant Patient and public involvement Patients and/or the public were not involved in dyspnoea, she was allowed to return home with symptomatic treat- the design, or conduct, or reporting, or dissemination plans of this research. ment only. The nasopharyngeal swab was positive for the Severe Patient consent for publication Obtained. Acute Respiratory Syndrome Coronavirus 2 (SARS-­CoV2) by real-­ Provenance and peer review Not commissioned; internally peer reviewed. time reverse transcription-­PCR. She was quarantined at home and © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and monitored by daily telephone calls, and the upcoming tocilizumab permissions. Published by BMJ. infusion was postponed. The symptoms remained mild, and, 10 days later, she reported to be free of symptoms. A follow-up­ naso- pharyngeal swab for SARS-­Cov2 performed on March 26 turned out negative. She was thus declared cured from the infection and To cite Mihai C, Dobrota R, Schröder M, et al. Ann Rheum Dis 2020;79:668–669. scheduled to receive the next tocilizumab dose 4 days after the Received 27 March 2020 negative test. Accepted 30 March 2020 In this case, a patient with insulin-dependent­ type 2 diabetes Published Online First 2 April 2020 mellitus and SSc-­ILD treated with tocilizumab developed a mild Ann Rheum Dis 2020;79:668–669. doi:10.1136/annrheumdis-2020-217442 form of COVID-19. Her pre-­existing ILD and diabetes are WHO-­ defined risk factors for a more severe course of COVID-19,1 while ORCID iDs Carina Mihai http://​orcid.​org/​0000-​0002-​8627-​8817 immunosuppressive treatment is currently also regarded as a risk Rucsandra Dobrota http://​orcid.​org/​0000-​0001-​9819-​7574 factor.2 In previous coronavirus outbreaks however, immunosup- Britta Maurer http://​orcid.​org/​0000-​0001-​9385-​8097 pression was not a documented risk factor.3 At present, in patients Oliver Distler http://​orcid.​org/​0000-​0002-​0546-​8310 with COVID-19 from Wuhan, China, higher levels of C reactive protein and IL-6 have been associated with increased mortality.4 References Accumulating evidence suggests that a subgroup of patients with 1 WHO. Coronavirus disease 2019 (COVID-19) situation report – 59. Available: https:// 4 www.​who.​int/​docs/​default-​source/​coronaviruse/​situation-​reports/​20200319-​sitrep-​59-​ severe COVID-19 develop a cytokine storm syndrome. Currently, covid-​19.​pdf?​sfvrsn=​c3dcdef9_2 [Accessed 20 Mar 2020]. several ongoing randomised trials study the efficacy and safety of 2 Swiss Federal Office of Public Health. New coronavirus: documents for health anti-­IL-6-­receptor monoclonal antibodies in severe COVID-19.5 6 professionals. Available: https://www.​bag.​admin.​ch/​bag/​de/​home/​krankheiten/​ Although these agents are immunosuppressive and thus formally ausbrueche-​epidemien-​pandemien/​aktuelle-​ausbrueche-​epidemien/​novel-​cov/​ contraindicated in patients with active infections, they may show information-​fuer-​die-​aerzteschaft/​dokumente-​fuer-​gesu​ndhe​itsf​achp​ersonen.​html [Accessed 20 Mar 2020]. benefit in certain subgroups of COVID-19-­associated severe acute 3 D’Antiga L. Coronaviruses and immunosuppressed patients. The facts during the third respiratory distress syndrome. In addition, our case indicates that epidemic. Liver Transpl 2020. IL-6-­blocking treatment given for chronic autoimmune diseases 4 Mehta P, McAuley DF, Brown M, et al. COVID-19: consider cytokine storm syndromes such as rheumatoid arthritis or connective tissue disease may even and immunosuppression. The Lancet 2020;395:1033–4 https://​doi.​org/ 5 ClinicalTrials.​gov. Favipiravir combined with tocilizumab in the treatment of corona prevent the development of severe COVID-19. While we are virus disease, 2019. Available: https://​clinicaltrials.​gov/​ct2/​show/​NCT04310228?​recrs=​ fully aware of the limitations of single-case­ observations, we think ab&​cond=​Coronavirus+​Infections&​draw=​3&​rank=​60 [Accessed 20 Mar 2020]. the presented case report supports this hypothesis. Thus, future 6 ClinicalTrials.​gov. Evaluation of the efficacy and safety of Sarilumab in hospitalized prospective, controlled studies should include an analysis for the patients with COVID-19. Available: https://​clinicaltrials.​gov/​ct2/​show/​NCT04315298?​ potential preventive effects of IL-6 blockade in COVID-19. recrs=​ab&​cond=​Coronavirus+​Infections&​draw=​2&​rank=​86 [Accessed 20 Mar 2020].

Carina Mihai ‍ ‍ , Rucsandra Dobrota ‍ ‍ , Maria Schröder, Alexandru Garaiman, Suzana Jordan, Mike Oliver Becker, Britta Maurer ‍ ‍ , Oliver Distler ‍ ‍ Department of Rheumatology, University Hospital Zurich, Zurich, Switzerland Correspondence to Dr Oliver Distler, Department of Rheumatology, University Hospital Zurich, Zurich 8091, Switzerland; ​oliver.​distler@​usz.​ch

Handling editor Josef S Smolen Acknowledgements The authors thank their patient for the permission to report on her case. Contributors All authors contributed to one or more of the following aspects of the manuscript: conception, acquisition of data, drafting and revising the article.

Ann Rheum Dis May 2020 Vol 79 No 5 669 Letters

Long-­term effectiveness of live herpes zoster vaccine in patients with rheumatoid arthritis subsequently treated with tofacitinib

Herpes zoster (HZ) incidence is higher in patients with rheu- matoid arthritis (RA) compared with the general population,1 and it may be further increased with disease-modifying­ anti- rheumatic drugs (DMARDs).2 Tofacitinib is an oral Janus kinase inhibitor for the treatment of RA. Real-world­ data indicate that

Ann Rheum Dis May 2020 Vol 79 No 5 669 Letters

HZ incidence is approximately twofold higher with tofacitinib Vaccine-related­ AEs in the index study included mild injection-­ versus biologic DMARDs (bDMARDs).3 site pain, swelling, redness, itching and myalgia. Disseminated Current American College of Rheumatology guidelines condi- vaccine-­strain varicella was also reported in a patient with no tionally recommend that patients with RA aged ≥50 years previous exposure to VZV.5 After rollover into ORAL Sequel, 100 receive HZ vaccine prior to tofacitinib or bDMARDs.4 We previ- patients received an average tofacitinib dose of 5 mg (n=46) or ously evaluated the immunogenicity of a live attenuated zoster 10 mg (n=54) two times per day. Mean (range) tofacitinib expo- vaccine (LZV), administered 2–3 weeks prior to tofacitinib or sure was 489 (46–811) days and overall exposure was 139 PY. placebo with background conventional synthetic DMARDs. LZV did not provide adequate protection to all patients. Both groups had similar varicella zoster virus (VZV)-specific­ Five HZ cases (#1–5) occurred in the LTE study 218, 280, immune responses, and overall immune responses were compa- 748, 741 and 544 days post-­vaccination, respectively (IR=3.60 rable with those of healthy volunteers in previous studies.5 We (1.17, 8.39); table 1). Cases #1–4 were monodermatomal and have now followed this patient cohort in an open-­label, long-­ case #5 involved five dermatomes. All HZ events were mild/ term extension (LTE) study of tofacitinib. moderate in severity and resolved with antiviral treatment. Patients enrolled in the index study (A3921237; VZV humoral immunity (immunoglobulin G (IgG) titre) NCT02147587)5 could join ORAL Sequel (LTE study; and VZV cell-­mediated immunity (interferon-γ enzyme-­linked A3921024; NCT00413699) 14 weeks post-­vaccination, where immunosorbent spot (ELISPOT)) in patients receiving tofac- they received open-­label tofacitinib 5 or 10 mg two times per itinib or placebo in the index study5 are shown in table 1. day (online supplementary figure S1); background RA therapy In terms of immunity after LZV in this analysis, cases #1, was also allowed. Patients were followed for 27 months. Post-­ #4 and #5 had undetectable VZV cell-mediated­ immunity, vaccination, adverse events (AEs), including discontinuations at baseline and week 6; cases #2 (patient received tofaci- due to AEs, were recorded during the study within 28 days of the tinib 5 mg two times per day in index and LTE studies) and last dose. Incidence rates (IRs; patients with events/100 patient-­ #3 (patient received placebo and tofacitinib 5 mg two times years (PY)) and 95% CIs for HZ post-­vaccination were calcu- per day in index and LTE studies, respectively) responded lated based on time to first event (patients not reporting an event adequately to vaccination by both IgG and ELISPOT measures were censored at last treatment dose). Short-­term VZV-­specific but had lower than average VZV IgG levels at baseline (case immunity was evaluated at baseline and week 6 post-vaccination­ #2: 36.9 U/mL vs average of 201 U/mL; case #3: 96.6 U/mL during the index study. vs average of 182 U/mL) and week 6 (case #2: 70.9 U/mL

Table 1 Patient profiles of HZ cases Case #1 Case #2 Case #3 Case #4 Case #5 Age, years 65 60 77 74 74 Sex Female Male Female Male Male Race White White White White White Study drug Tofacitinib Tofacitinib Placebo Placebo Placebo (A3921237) 5 mg two times per day 5 mg two times per day Study drug Tofacitinib Tofacitinib Tofacitinib Tofacitinib Tofacitinib (ORAL Sequel) 10 mg two times per day 5 mg two times per day 5 mg two times per day 10 mg two times per day 10 mg two times per day Background RA drugs MTX 15 mg/week MTX 20 mg/week None None MTX 20 mg/week Prednisone 5 mg/day Type of HZ Monodermatomal Monodermatomal Monodermatomal Monodermatomal 5 dermatomes Severity of HZ* Moderate Mild Moderate Mild Mild Duration of HZ, days 49 14 14 16 10 Action to study drug No action taken Stopped temporarily No action taken No action taken Stopped temporarily Outcome of HZ Resolved with acyclovir Resolved with famciclovir Resolved with acyclovir Resolved with Resolved with and azithromycin valacyclovir valacyclovir Occurrence of HZ Time after LZV vaccination, days 218 280 748 741 544 Time after initiation of tofacitinib, days 202 267 702 699 466 VZV humoral immunity (IgG titre), U/mL† Baseline 224.3 36.9 96.6 237.3 208.3 Week 6 444.0 70.9 186.9 231.5 222.5 Change from baseline (fold rise at week 6) 1.98 1.92 1.93 0.98 1.07 VZV cell-­mediated immunity, SFCs/106 PBMCs‡ Baseline 25 41 25 25 25 Week 6 25 76 51 25 25 Change from baseline (fold rise at week 6) 1.00 1.85 2.04 1.00 1.00 *Determined by the investigator. †Assessed by gpELISA (PPD Vaccines and Biologics); mean VZV IgG titres in patients receiving tofacitinib and placebo, respectively, in the index study were 201 and 182 U/mL at baseline and 403 and 323 U/mL at week 6 (fold rise at week 6 was 2.11 with tofacitinib and 1.74 with placebo).5 ‡Assessed by IFNγ ELISPOT (Pfizer Inc Vaccine Research Unit, Pearl River, New York, USA); limit of detection was 25 SFCs/106 PBMCs; values in the table shown as 25 SFCs/106 PBMCs may be below this threshold; mean VZV cell-­mediated immunity in patients receiving tofacitinib and placebo, respectively, in the index study was 48 SFCs/106 PBMCs and 43 SFCs/106 PBMCs at baseline, and 70 SFCs/106 PBMCs and 56 SFCs/106 PBMCs at week 6 (fold rise at week 6 was 1.50 with tofacitinib and 1.29 with placebo).5 ELISPOT, enzyme-­linked immunosorbent spot; gpELISA, glycoprotein-­based enzyme-­linked immunosorbent assay; HZ, herpes zoster; IFNγ, interferon gamma; IgG, immunoglobulin G; LZV, live zoster vaccine; MTX, methotrexate; PBMCs, peripheral blood mononuclear cells; RA, rheumatoid arthritis; SFCs, spot-forming­ cells; VZV, varicella zoster virus.

670 Ann Rheum Dis May 2020 Vol 79 No 5 Letters vs average of 403 U/mL; case #3: 186.9 U/mL vs average of permits others to distribute, remix, adapt, build upon this work non-­commercially, 323 U/mL; table 1). 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 HZ incidence was similar to that in patients receiving is non-­commercial. See: http://​creativecommons.​org/​licenses/​by-​nc/​4.​0/. tofacitinib in phase 1/2/3/LTE studies up to 9.5 years (IR=3.6 6 © Author(s) (or their employer(s)) 2020. Re-­use permitted under CC BY-­NC. No (3.4, 3.9); n=782/7061]), although the present analysis was commercial re-­use. See rights and permissions. Published by BMJ. limited due to the small number of patients, and 95% CIs were ►► Additional material is published online only. To view please visit the journal wide. Cell-­mediated responses in cases #2 and #3 may have online (http://​dx.​doi.​org/​10.​1136/​annrheumdis-​2019-​216566). been short-­lived; however, serial longitudinal data are required These data were presented at the American College of Rheumatology Annual to confirm this. Scientific Meeting in 2017 (Winthrop KL et al. Ann Rheum Dis 2017;76[Suppl 2]) and These results suggest that LZV may not provide adequate are used here with permission from John Wiley & Sons, Inc. long-­term protection, as previously demonstrated in healthy individuals aged ≥60 years 3 years post-­vaccination, in which HZ risk was reduced by 51%.2 While it is possible that LZV booster vaccinations may improve vaccine efficacy, to date there To cite Winthrop KL, Wouters A, Choy EH, et al. Ann Rheum Dis 2020;79:669–671. is a lack of data on the use and timing of booster vaccinations, Received 31 October 2019 and no recommendations on the use of LZV booster vaccina- Revised 3 February 2020 tions currently exist. This highlights the importance of evalu- Accepted 4 February 2020 ating the newly approved subunit non-­live vaccine (Shingrix) in Published Online First 11 March 2020 patients with RA receiving tofacitinib. Ann Rheum Dis 2020;79:669–671. doi:10.1136/annrheumdis-2019-216566

Kevin L Winthrop ‍ ‍ ,1 Ann Wouters,2 Ernest H Choy,3 Connie Chen,2 ORCID iD Pinaki Biswas,2 Lisy Wang,4 Koshika Soma,4 Elie Needle,2 Hernan Valdez,2 Kevin L Winthrop http://​orcid.​org/​0000-​0002-​3892-​6947 William FC Rigby5 1Oregon Health & Science University, Portland, Oregon, USA References 2Pfizer Inc, New York, New York, USA 1 Yun H, Yang S, Chen L, et al. Risk of herpes zoster in autoimmune and inflammatory 3CREATE Centre, Division of Infection and Immunity, Cardiff University School of diseases: implications for vaccination. Arthritis Rheumatol 2016;68:2328–37. Medicine, Cardiff, UK 2 Winthrop KL, Furst DE. Rheumatoid arthritis and herpes zoster: risk and prevention 4Pfizer Inc, Groton, Connecticut, USA in those treated with anti-­tumour necrosis factor therapy. Ann Rheum Dis 5Geisel School of Medicine at Dartmouth, Lebanon, New Hampshire, USA 2010;69:1735–7. 3 Curtis JR, Xie F, Yun H, et al. Real-­world comparative risks of herpes virus infections Correspondence to Kevin L Winthrop, MD, MPH, OHSU-­PSU School of Public in tofacitinib and biologic-­treated patients with rheumatoid arthritis. Ann Rheum Dis Health, 3181 S.W. Sam Jackson Rd, Portland, OR 97239, USA; ​winthrop@​ohsu.​edu 2016;75:1843–7. 4 Singh JA, Saag KG, Bridges Jr SL, et al. 2015 American College of Rheumatology Handling editor Josef S Smolen guideline for the treatment of rheumatoid arthritis. Arthritis Rheumatol 2016;68:1–26. 5 Winthrop KL, Wouters AG, Choy EH, et al. The safety and immunogenicity of live Acknowledgements The authors would like to acknowledge Lisa McNeil for zoster vaccination in patients with rheumatoid arthritis before starting tofacitinib: a managing the ELISPOT assay and for her contributions to the interpretation of the randomized Phase II trial. Arthritis Rheumatol 2017;69:1969–77. ELISPOT and ELISA results. Medical writing support, under the guidance of the 6 Cohen S, Tanaka Y, Mariette X, et al. Long-­term safety of tofacitinib up to 9.5 years: a authors, was provided by Anthony G McCluskey, PhD, CMC Connect, McCann Health comprehensive integrated analysis of the RA clinical development program [abstract]. Medical Communications and was funded by Pfizer Inc, New York, New York, USA Arthritis Rheumatol 2018;70:Abstract 963. in accordance with Good Publication Practice (GPP3) guidelines (Ann Intern Med 2015;163:461–464). Contributors All authors were involved in the analysis and interpretation of data, critically revising the text for important intellectual content. All authors agree to be accountable for all aspects of the work and read and approved the final version to be published. Funding This study was sponsored by Pfizer Inc. Competing interests KLW has acted as a consultant for AbbVie, Bristol-­Myers Squibb, Eli Lilly, Galapagos, Gilead, Pfizer Inc and UCB; and he has received grant/ research support from Bristol-­Myers Squibb. EHC has received grant/research support from BioCancer, Pfizer Inc, Roche and UCB; has acted as a consultant for Amgen, Biogen, Chugai Pharma, Eli Lilly, Janssen, Novartis, Pfizer Inc, Regeneron, Roche, R-­Pharm and Sanofi; and has participated in speakers’ bureaus for Amgen, Boehringer Ingelheim, Bristol-­Myers Squibb, Chugai Pharma, Eli Lilly, Hospira, MSD, Novartis, Pfizer Inc, Regeneron, Roche, Sanofi-­Aventis and UCB. WFCR has received grant/research support from Pfizer Inc and has acted as a consultant for Pfizer Inc and Roche. AW, CC, PB, LW, KS, EN and HV are employees and shareholders of Pfizer Inc. Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. Patient consent for publication Not required. Provenance and peer review Not commissioned; externally peer reviewed.

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

Ann Rheum Dis May 2020 Vol 79 No 5 671 Letters

­ ​ ​ ​ ​ ​ ​ ​

​ ​ ​ ​ ​ ​ ​ ​

​ ​ ​ ​ ​ ​

CDAI and DAS28 in the management of rheumatoid arthritis in clinical practice

The primary therapeutic target for rheumatoid arthritis is remis- sion, assessed using validated composited measures.1 Currently, index-­based remission frequently used in clinical practice are disease activity (CDAI) and Disease Activity Score for 28 joints (DAS28). Generally, it has been reported that CDAI is more stringent than DAS28 in assessing clinical remission.1 2 However, this confirmation was mainly derived from trial results. Hence, this study aimed to investigate the real-world­ performance of CDAI and DAS28-­erythrocyte sedimentation rate (ESR). In total, 1585 consecutive patients with rheumatoid arthritis (mean age: 64 years, sex: 84% were women, mean disease dura- tion: 12.0 years) in Keio University Hospital were reviewed cross-­sectionally. Current treatments were conventional synthetic disease-­modifying antirheumatic drugs (csDMARDs) alone, tumour necrosis factor inhibitors (TNFi), interleukin 6 receptor inhibitors (IL-6i),­ cytotoxic T-­lymphocyte associated antigen (CTLA)-­4Ig and janus kinase inhibitors (JAKi) in 39.2%, 29.0%, 22.8%, 7.1% and 1.8% patients, respectively. Defini- tion of cut-offs­ of each composite scores was following; remis- sion, CDAI ≤2.8, DAS28-­ESR <2.6; low disease activity, CDAI ≤10, DAS28-­ESR <3.2; moderate disease activity; CDAI ≤22,

Ann Rheum Dis May 2020 Vol 79 No 5 671 Letters

Figure 1 (A) Clinical characteristics of patients achieving clinical disease activity (CDAI) remission.a, csDMARDs vs TNFi and IL-6i;­ b, TNFi vs csDMARDs and CTLA-­4Ig; c, CTLA-­4Ig vs TNFi and IL-­6i; d, IL-­6i vs csDMARDs and CTLA-­4Ig;e, csDMARDs vs IL-­6i; f, IL-­6i vs all others; g, CTLA-­4Ig vs TNFi and IL-­6i; h, No significant difference with the Bonferroni adjustment; i, IL-6i­ vs csDMARDs,CTLA-­4Ig and JAKi (B) Disease Activity Score for 28 joints (DAS28)-­erytherocyte sedimentation rate (ESR) in patients with CDAI remission. (C) Venn diagrams of CDAI remission and DAS28-ESR­ remission. CRP, C-­reactive protein; csDMARD, conventional synthetic disease-­modifying antirheumatic drugs; EGA, evaluator global assessment; HAQ, Health Assessment Questionnaire; IL, interleukin; PGA, patient global assessment.

DAS28-­ESR≤5.1; high disease activity, CDAI >22, DAS28-ESR­ patients in CDAI remission and DAS28-ESR­ non-­remission >5.1. and those in CDAI non-­remission and DAS28-­ESR remission is First, we focused on the patients in CDAI remission (62.7% in shown in table 1. Many differences were noted, but patients in all, 66.4% treated with csDMARDs, 65.8% treated with TNFi, CDAI remission and DAS28-ESR­ non-­remission had higher acute 58.6% treated with IL-6i,­ 51.3% treated with CTLA-4Ig­ and phase reactants and comorbidity rates, and those in CDAI non-­ 31.0% treated with JAKi, p＝0.06). Figure 1A shows the charac- remission and DAS28-­ESR remission had worse patient-reported­ teristics of patients in CDAI remission by treatment. As expected, outcomes. However, actual tender and swollen joint counts were DAS28-­ESR and acute phase reactants were significantly lower a little higher in patients in DAS28-­ESR non-­remission than in the IL-­6i-­treated group. Among patients in CDAI remission, those in CDAI remission. the proportion of DAS28-ESR­ non-­remission was 19.4% in Our data demonstrate that despite the belief that CDAI is those treated with csDMARDs, 18.2% treated with TNFi, 4.2% more stringent than DAS28-­ESR when defining remission, in a treated with IL-6i,­ 27.6% treated with CTLA-4Ig­ and 33.3% real-­world setting where more than half of patients were well treated with JAKi (figure 1B). In contrast, among patients in controlled, quite a few patients who satisfied the CDAI remis- DAS28 remission, the proportion of CDAI non-­remission was sion criteria did not meet the DAS28-­ESR remission criteria, 11.7% in those treated with csDMARDs, 15.4% treated with except for those who were treated with IL-­6i, because of the TNFi, 29.5% treated with IL-6i,­ 16.0% treated with CTLA-4Ig­ direct inhibition of high acute phase reactants. The fact that and 14.3% treated with JAKi. Figure 1C depicts diagrams of many patients fulfilled the CDAI but not DAS28-­ESR remis- CDAI remission and DAS28-­ESR remission, demonstrating that sion could be explained by several reasons including residual more patients satisfied the CDAI remission criteria without satis- synovitis in joints that are not included in the main 28 joints, fying the DAS28-­ESR remission criteria than vice versa, except which could lead to an increase in acute phase reactants and for those treated with IL-­6i. A comparison of variables between elevate only DAS28-ESR,­ extra-articular­ involvement or other

672 Ann Rheum Dis May 2020 Vol 79 No 5 Letters

Table 1 Clinical characteristics of patients who met both or either CDAI and/or DAS28-ESR­ remission ALL csDMARDs TNFi IL-­6i DAS28-­ESR DAS28-­ESR DAS28-­ESR Both REM CDAI REM DAS28-­ESR REM CDAI REM REM CDAI REM REM CDAI REM REM n=831 n=163 n=183 n=80 n=44 n=55 n=45 n=9 n=85 Age, years 61.0±14.2 69.6±12.6 63.4±15.0 70.8±12.7 63.0±14.9 68.3±11.6 61.2±16.3 63.2±16.7 63.1±14.2 Disease duration, years 10.5±7.6 13.0±9.3 12.8±9.7 12.5±9.4 11.5±11.3 14.8±9.1 12.6±7.5 8.0±8.0 13.3±9.7 Female, % 80.9 85.3 89.6 82.5 90.9 85.5 80.0 88.9 94.1 CDAI 0.8±0.8 1.4±0.9 5.0±2.1 1.4±0.9 4.4±1.5 1.5±0.8 4.6±1.7 1.3±0.9 5.6±2.3 DAS28-ESR­ 1.8±0.6 2.9±0.2 2.1±0.4 2.9±0.2 2.2±0.3 2.9±0.2 2.2±0.3 2.8±0.2 1.9±0.5 TJC28 0.0±0.2 0.1±0.4 0.4±0.8 0.1±0.4 0.3±0.6 0.1±0.3 0.3±0.6 0.0±0.0 0.6±0.9 TJC=0, % 96.8 88.3 73.2 90.0 79.5 87.3 75.6 100 67.1 TJC≤1, % 99.5 99.4 88.5 98.8 90.9 100 93.3 100 84.7 SJC28 0.1±0.3 0.1±0.4 0.9±1.3 0.2±0.4 0.4±0.7 0.1±0.3 1.0±1.5 0.1±0.3 1.1±1.4 SJC=0, % 95.2 87.7 56.8 86.3 68.2 90.9 60.0 88.9 48.2 SJC≤1, % 98.9 99.4 75.4 98.8 90.9 100 66.7 100 70.6 EGA, mm 1.1±2.3 2.0±3.1 6.2±6.0 2.0±3.2 4.5±5.3 2.0±2.9 7.1±6.6 2.9±5.8 6.8±6.1 EGA≤10 mm, % 98.8 97.5 76.5 97.5 81.8 98.2 68.9 88.9 76.5 PGA, mm 5.7±6.6 9.5±8.2 31.1±20.0 9.1±8.3 32.0±20.1 10.6±8.5 25.7±17.1 9.4±8.9 32.9±21.1 PGA≤10 mm, % 79.3 62.6 15.8 66.3 15.9 54.5 24.4 66.7 11.8 HAQ-DI­ 0.19±0.38 0.41±0.59 0.82±0.78 0.36±0.58 0.51±0.66 0.46±0.67 0.63±0.71 0.43±0.29 1.03±0.80 HAQ-­DI ≤0.5, % 89.3 75.4 46.4 78.8 63.6 74.5 60.0 66.7 32.9 ESR, mm 13.4±8.9 45.7±16.9 8.0±4.8 45.2±16.2 9.9±4.3 45.1±15.8 10.2±4.6 46.7±17.1 5.6±3.9 CRP, mg/dL 0.09±0.16 0.37±0.59 0.07±0.14 0.48±0.74 0.09±0.14 0.21±0.27 0.11±0.17 0.30±0.38 0.02±0.05 MMP-3, ng/mL 48.4±42.5 58.8±61.2 62.9±67.0 62.9±75.4 60.3±55.2 49.0±36.4 50.8±42.5 71.9±66.9 66.2±73.3 Complications, % 11.3 26.4 18.0 25.0 9.1 18.2 4.4 33.3 27.1 ILD, % 4.6 11.7 7.1 7.5 6.8 12.7 0 22.2 10.6 Chronic infection, % 3.5 11.0 4.9 13.8 0 3.6 0 0 7.1 Sjögren syndrome, % 3.2 3.7 6.0 3.8 2.3 1.8 2.0 11.1 9.4 Data of patients treated with CTLA-­4Ig and JAK inhibitors were omitted because of a small samples size. Figures are a mean±SD unless otherwise stated. CDAI, clinical disease activity; CRP, C-­reactive protein; csDMARDs, conventional synthetic disease-modifying­ antirheumatic drug; DAS28, Disease Activity Score for 28 joints; EGA, evaluator global assessment; ESR, erythrocyte sedimentation rate; HAQ-DI,­ Health Assessment Questionnaire-­disability index; ILD, interstitial lung disease; IL-6i,­ interleukin 6 inhibitors; MMP-3, matrix metalloproteinase-3; PGA, patient global assessment; REM, remission; SJC, swollen joint count; TJC, tender joint count; TNFi, TNF inhibitors. comorbidities that could elevate the C-­reactive protein irrele- Division of Rheumatology, Department of Internal Medicine, Keio University School vant to arthritis. The prevalence of complications was higher of Medicine, Tokyo, Japan in patients in CDAI remission and DAS28-ESR­ non-­remission Correspondence to Professor Tsutomu Takeuchi; tsutake@​ ​z5.keio​ .​jp regardless of rheumatoid-­related (interstitial lung disease or Sjögren syndrome) or non-­related (chronic infection) condi- Handling editor Gerd R Burmester tions. Our data indicate that while CDAI reflects mainly arthritis, Contributors ST, YK and TT designed the study. ST wrote first draft of the DAS28-­ESR reflects not only arthritis (probably to a lesser extent manuscript. All authors edited the manuscript and approved the final version. than CDAI) but also extra-articular­ or extrarheumatoid compli- Funding The authors have not declared a specific grant for this research from any cations. Also, the idea that CDAI is more stringent than DAS28 funding agency in the public, commercial or not-­for-­profit sectors. in assessing remission has been mainly derived from clinical trial Competing interests YK has received grants or speaking fees from AbbVie, data. Clinical trials usually exclude patients with comorbidities, Astellas, Ayumi, Bristol-­Myers Squibb, Chugai, Eisai, Eli Lilly, Hisamitsu, Jansen, Kissei, and that can be one of the causes which lead to the discrepancy Pfizer, Sanofi, Takeda, Tanabe-­Mitsubishi and UCB. TT has received research grants or speaking fees from Astellas Pharma, Bristol–Myers K.K., Chugai Pharmaceutical, in index-based­ remission between trials and clinical practice Daiichi Sankyo, Takeda Pharmaceutical, Teijin Pharma, AbbVie GK, Asahikasei Pharma as we have shown in the present study. However, patients in Corp., Mitsubishi Tanabe Pharma, Astra Zeneca K.K., Eli Lilly Japan K.K., Novartis both CDAI and DAS28-ESR­ remission were apparently in better Pharma K.K., Abbivie GK, Nipponkayaku, Janssen, Pharmaceutical K.K., Taiho condition than those who met either criteria; therefore, in the Pharmaceutical and Pfizer Japan. management of rheumatoid arthritis, assessing patients with Patient consent for publication Not required. two composite measures can yield important opportunities to Provenance and peer review Not commissioned; externally peer reviewed. consider what causes the discrepancy between the measures and © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and adjust treatment appropriately. permissions. Published by BMJ. This was only a cross-­sectional study, which did not evaluate longitudinal radiological and functional progression. A large prospective cohort is needed to establish the optimal usage and To cite Takanashi S, Kaneko Y, Takeuchi T. Ann Rheum Dis 2020;79:671–674. interpret composite measures and their components in clinical practice. Received 8 November 2019 Revised 13 January 2020 Satoshi Takanashi ‍ ‍ , Yuko Kaneko, Tsutomu Takeuchi Accepted 22 January 2020

Ann Rheum Dis May 2020 Vol 79 No 5 673 Letters

Published Online First 29 January 2020 Ann Rheum Dis 2020;79:671–674. doi:10.1136/annrheumdis-2019-216607

ORCID iD Satoshi Takanashi http://​orcid.​org/​0000-​0002-​3607-​2140

References 1 Smolen JS, Aletaha D, Bijlsma JWJ, et al. Treating rheumatoid arthritis to target: recommendations of an international Task force. Ann Rheum Dis 2010;69:631–7. 2 Fleischmann RM, Genovese MC, Enejosa JV, et al. Safety and effectiveness of upadacitinib or adalimumab plus methotrexate in patients with rheumatoid arthritis over 48 weeks with switch to alternate therapy in patients with insufficient response. Ann Rheum Dis 2019;78:1454–62.

674 Ann Rheum Dis May 2020 Vol 79 No 5 Letters

Efficacy of JAK 1/2 inhibition in the treatment of diffuse non-­scarring alopecia due to systemic lupus erythematosus

The Janus kinase (JAK)-­signal transducer and activator of tran- Figure 1 Hair before and after treatment with baricitinib. scription (STAT) pathway is crucial for the autocrine loop of type I interferons (IFNs) and is located upstream of important systemic lupus erythematosus (SLE)-­associated pathogenic 12 weeks, the PSL dose was gradually reduced to 12.5 mg, and proinflammatory cytokines and chemokines. Baricitinib is a JAK no lesions—including alopecia—relapsed. In the present patient, 1/2 inhibitor approved for treating rheumatoid arthritis and is baricitinib treatment was well tolerated and no severe adverse also being tested in other diseases, including SLE. Baricitinib events occurred. improved the proportion of patients with SLE who achieved The mechanisms involved in alopecia in SLE are unknown, arthritis or rash resolution; however, the subset of patients with but the clinical course of the current patient suggests that the 1 skin involvement was not reported in the study. Alopecia is one JAK-­STAT pathway—which cannot be controlled via treatment of the most common cutaneous symptoms in SLE. Patients can including pulse corticosteroids—may be integrally involved. It is 2 exhibit different subtypes of alopecia. Diffuse non-­scarring hair thought that JAK inhibitors eliminate the IFN signature by inhib- loss, which is the most frequently observed type of alopecia in iting the downstream signalling of IFN-gamma­ and gamma cyto- patients with SLE, is not life-­threatening, but substantially affects kine receptors, and thus prevent the development of alopecia quality of life for cosmetic reasons and may reflect latent SLE areata.5 While considerable attention has been focused on type 3 activity. To our knowledge, the efficacy of JAK inhibitors for I IFNs in studies investigating IFN-­associated gene expression treating alopecia in SLE is unknown. Herein, we describe a profiles in SLE, the type II IFN (IFN-gamma)­ pathway is also patient with SLE who experienced substantial improvement in considered to be pathogenic in the cutaneous involvement.6 The diffuse non-­scarring alopecia following baricitinib therapy. pathogenesis of SLE in patients with severe diffuse non-­scarring The patient was a 27-year-­ ­old Japanese woman who was diag- alopecia may be primarily driven by IFN-­gamma, and JAK inhib- nosed with SLE at age 21, requiring corticosteroid and tacrolimus itors may constitute another therapeutic option in patients in treatment. Arthritis, renal damage with proteinuria, leucopenia, which standard treatments are ineffective. Our case and the high-­titre antidouble-­stranded DNA antibody and hypocomple- initial evidence of effectiveness of baricitinib in a recent phase mentaemia were observed at onset. SLE disease activity had been II trial1 indicate cautious optimism, while awaiting larger trials. stable after remission but relapsed at the age of 26 (prednisolone Notably, the overall risk-­benefit assessment of SLE treatment (PSL) 3.0 mg/day and hydroxychloroquine 200 mg/day at the with baricitinib should be discussed further. time). Along with the relapse of serum markers and proteinuria, chilblain lupus erythematosus and diffuse hair loss developed. Keisuke Maeshima ‍ ‍ , Hirotaka Shibata Because the reinitiation of tacrolimus had no beneficial effect, Department of Endocrinology, Metabolism, Rheumatology and Nephrology, Faculty the PSL dose was increased to 17.5 mg/day and mycophenolate of Medicine, Oita University, Yufu, Japan mofetil 1500 mg/day was initiated. Renal involvement and serum Correspondence to Dr Keisuke Maeshima, Rheumatology, Faculty of Medicine markers improved, but skin lesions progressed. Adding belim- Graduate School of Medicine, Oita University, Yufu 879-5593, Japan; umab (400 mg/4 weeks after loading dose) was ineffective. When ​maeshima@​oita-​u.​ac.​jp PSL was increased to 40 mg after steroid pulse therapy, the chil- Handling editor Josef S Smolen blain lupus erythematosus improved but the hair loss progressed (see online supplementary figure S1). She was referred to our Acknowledgements We thank the patient and her family for making this work possible. hospital at the age of 27. Since promising outcomes of treatment with the JAK inhibitors tofacitinib, ruxolitinib and baricitinib for 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. alopecia areata unrelated to SLE have been reported,4 we intro- duced baricitinib (4 mg/day) along with combination therapy Competing interests None declared. with PSL, mycophenolate mofetil and hydroxychloroquine. Patient consent for publication Obtained. Surprisingly, no progression of hair loss was observed after 4 Ethics approval We obtained ethics approval for this study from the Ethical weeks of treatment, and prominent hair regrowth was observed Committee of the Faculty of Medicine, Oita University (approval number: 1722). after 8 weeks (figure 1 and online supplementary figure S2). At Provenance and peer review Not commissioned; externally peer reviewed.

674 Ann Rheum Dis May 2020 Vol 79 No 5 Letters

© Author(s) (or their employer(s)) 2020. 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-​2019-​216571).

To cite Maeshima K, Shibata H. Ann Rheum Dis 2020;79:674–675.

Received 31 October 2019 Revised 11 December 2019 Accepted 12 December 2019 Published Online First 3 January 2020 Ann Rheum Dis 2020;79:674–675. doi:10.1136/annrheumdis-2019-216571

ORCID iD Keisuke Maeshima http://​orcid.​org/​0000-​0002-​9251-​5997

References 1 Wallace DJ, Furie RA, Tanaka Y, et al. Baricitinib for systemic lupus erythematosus: a double-­blind, randomised, placebo-­controlled, phase 2 trial. Lancet 2018;392:222–31. 2 Udompanich S, Chanprapaph K, Suchonwanit P. Hair and scalp changes in cutaneous and systemic lupus erythematosus. Am J Clin Dermatol 2018;19:679–94. 3 Gong Y, Ye Y, Zhao Y, et al. Severe diffuse non-­scarring hair loss in systemic lupus erythematosus - clinical and histopathological analysis of four cases. J Eur Acad Dermatol Venereol 2013;27:651–4. 4 de Oliveira AB, Alpalhão M, Filipe P, et al. The role of Janus kinase inhibitors in the treatment of alopecia areata: a systematic review. Dermatol Ther 2019;32:e13053. 5 Xing L, Dai Z, Jabbari A, et al. Alopecia areata is driven by cytotoxic T lymphocytes and is reversed by JAK inhibition. Nat Med 2014;20:1043–9. 6 Hile GA, Gudjonsson JE, Kahlenberg JM. The influence of interferon on healthy and diseased skin. Cytokine 2018. doi:10.1016/j.cyto.2018.11.022. [Epub ahead of print: 06 Dec 2018].

­

Ann Rheum Dis May 2020 Vol 79 No 5 675 Letters

Patients’ of two centres were treated similarly, starting with cyclophosphamide or rituximab in conjunction with high-­dose glucocorticoids for induction or major relapses, and maintenance treatment was composed of a combination of oral methylpred- nisolone and azathioprine, rituximab or mycophenolate mofetil for at least 24 months after remission had been achieved. Factors predictive of renal survival were evaluated by the Kaplan-­Meier method and the Cox proportional hazard model. In total, 167 patients with AAV (90 with granulomatosis with polyangiitis (54%), 39 with microscopic polyangiitis (23%), 30 with RLV (18%) and 8 with eosinophilic granulomatosis with polyangiitis (5%)) (95 men (57%), median age at diagnosis of 55 (IQR 19) years, median serum creatinine level at diagnosis of 3.74 (IQR 5.32 mg/dL) with renal involvement at presentation were analysed. ANCAs were detected in 87% (141/163) of the patients with indirect immunofluorescence (IIF) and/or ELISA. ESRD developed in 52 patients (34%) over a median follow-­up of 39.6 (IQR 65) months, and the median renal survival was 31 (IQR 65) months. In 72 patients (46%), haemodialysis was performed at presentation. Of the 106 kidney biopsies which were available for the final analysis, 14 (13%) were sclerotic, 41 (39%) were crescentic, 33 (31%) were mixed and 18 (17%) were focal, and among them, ESRD developed in 79%, 51%, 32% and 18%, respectively (p=0.003). Among our patients, 14% were at low risk, 63% were at medium risk and 23% were at high risk group according to ARRS, and ESRD developed in 8%, 42% and 67%, respectively (p=0.005). In univariate anal- ysis, age at diagnosis (HR 1.02, 95% CI 1.00 to 1.04, p=0.045), AAV subgroups (p=0.01), Berden’s classification (p=0.009), ARRS (p=0.01), serum creatinine level (HR 1.14, 95% CI 1.09 Histopathological subgrouping versus renal risk to 1.19, p<0.001) and GFR (HR 0.92, 95% CI 0.89 to 0.95, score for the prediction of end-­stage renal p<0.001) were the baseline characteristics associated with renal survival. In multivariate analysis (table 1), Berden’s classification disease in ANCA-­associated vasculitis predicted renal survival in model 1, but when GFR at diagnosis was included (model 2) in the model, it lost its significance. However, the ARRS was found to be an independent prognostic In patients with antineutrophil cytoplasmic antibody-­associated factor for renal survival (model 3) with the same characteristics vasculitis (AAV) and renal involvement, the development of in the model. end-­stage renal disease (ESRD) remains an undesired issue. To In baseline evaluation, the prediction of the development date, reported predictors of renal outcome are mainly patients’ of ESRD in patients with AAV may be important for treating age, severe renal dysfunction and histopathological findings at physicians. Histopathological findings are proposed to be presentation.1 2 Histopathological classification as defined by helpful; however, it does not consider baseline renal function. Berden et al was proposed to be helpful with the highest renal In the present study, ARRS seems to be more advantageous than survival rates in the focal group and the poorest in the scle- Berden’s classification, and the possible explanation might be the rotic group.3 4 Recently, Brix et al suggested the antineutrophil incorporation of baseline GFR to the histopathological findings. cytoplasmic antibody renal risk score (ARRS) to predict ESRD The main limitations of the present study are being retrospective in patients with AAV.5 Unlike Berden's classification, ARRS and the absence of interobserver reliability. combines histopathological findings (the percentage of normal In conclusion, high ESRD development rates in AAV empha- glomeruli, tubular atrophy and interstitial fibrosis) with baseline sises the importance of identifying patients at risk. At this point, glomerular filtration rate (GFR). Here, we aimed to assess the our results support the usage of both clinical and histopatholog- prognostic factors for renal survival and to evaluate the perfor- ical findings to predict the renal outcome. mances of Berden’s histopathological classification and ARRS 1 2 3 4 for predicting ESRD. Onay Gercik ‍ ‍ , Emre Bilgin ‍ ‍ , Dilek Solmaz ‍ ‍ , Fulya Cakalagaoglu, Arzu Saglam,5 Ozge Aybi,2 Riza Can Kardas,2 Zeki Soypacaci,6 We reviewed the medical records of all patients diagnosed Gokhan Kabadayi,3 Tolga Yildirim,7 Idil Kurut Aysin,3 Omer Karadag,2 with AAV according to biopsy and/or antineutrophil cytoplasmic Servet Akar8 antibody (ANCA) serology. Patients with renal involvement 1Rheumatology, Izmir Katip Celebi University Faculty of Medicine, Izmir, Turkey and categorised according to the 2012 Chapel Hill consensus 2Rheumatology, Hacettepe University Faculty of Medicine, Ankara, Turkey nomenclature were included. Renal-limited­ vasculitis (RLV) was 3Rheumatology, Izmir Katip Celebi University Faculty of Medicine, Izmir, Turkey considered as a separate group. We reviewed renal biopsies in 4Pathology, Izmir Katip Celebi University Faculty of Medicine, Izmir, Turkey 5 order to calculate ARRS and according to Berden’s classification. Pathology, Hacettepe University Faculty of Medicine, Ankara, Turkey 6Nephrology, Izmir Katip Celebi University Faculty of Medicine, Izmir, Turkey Renal survival was defined as the time between diagnosis and the 7Nephrology, Hacettepe University Faculty of Medicine, Ankara, Turkey 2 development of ESRD (GFR<15 mL/min/1.73 m or the start of 8Department of Internal Medicine, Division of Rheumatology, Izmir Katip Celebi a permanent dialysis programme). University Faculty of Medicine, Izmir, Turkey

Ann Rheum Dis May 2020 Vol 79 No 5 675 Letters

Table 1 Multivariate analysis models which predict renal survival Model 1 Model 2 Model 3 Parameters HR 95% CI P value HR 95% CI P value HR 95% CI P value Age at diagnosis 1.01 0.98 to 1.03 0.39 0.98 0.96 to 1.01 0.41 1.01 0.99 to 1.04 0.18 AAV subgroups 0.36 0.21 0.06 GPA versus MPA 2.10 0.87 to 5.08 0.09 2.72 1.07 to 6.92 0.03 4.07 1.45 to 11.39 0.007 EGPA versus MPA 0 0 n/A 0.98 0.02 0 n/A 0.99 0 0 n/A 0.98 RLV versus MPA 2.22 0.86 to 5.75 0.09 2.20 0.80 to 6.05 0.12 2.92 0.99 to 8.60 0.05 Histopathological classification – – 0.04 –– 0.08 – – – Focal versus sclerotic 0.20 0.05 to 0.75 0.01 0.31 0.07 to 1.27 0.10 – – – Mixed versus sclerotic 0.32 0.11 to 0.87 0.02 0.25 0.08 to 0.76 0.01 – – – Crescentic versus sclerotic 0.57 0.23 to 1.41 0.22 0.53 0.21 to 1.29 0.16 – – – GFR at diagnosis –– – 0.90 0.86 to 0.95 <0.001 –– – ANCA renal risk score – – ––– ––– 0.04 Moderate versus low –– ––– – 5.62 0.74 to 42.35 0.09 High versus low –– ––– – 10.48 1.32 to 82.88 0.02 Bold indicates statistically significant values at group level. AAV, antineutrophil cytoplasmic antibody-­associated vasculitis; ANCA, antineutrophil cytoplasmic antibody; EGPA, eosinophilic granulomatosis with polyangiitis; GFR, glomerular filtration rate; GPA, granulomatosis with polyangiitis; MPA, microscopic polyangiitis; N/A, not applicable; RLV, renal-limited­ vasculitis.

Correspondence to Professor Servet Akar, Department of Internal Medicine, Revised 3 February 2020 Division of Rheumatology, Izmir Katip Celebi University Faculty of Medicine, 35620 Accepted 4 February 2020 Çiğli/İzmir, Turkey; servet.​ ​akar@gmail.​ ​com Published Online First 10 February 2020 Ann Rheum Dis 2020;79:675–676. doi:10.1136/annrheumdis-2019-216742 Handling editor Josef S Smolen ORCID iDs Funding The authors have not declared a specific grant for this research from any Onay Gercik http://orcid.​ ​org/0000-​ ​0002-​0994-​8899 funding agency in the public, commercial or not-­for-­profit sectors. Emre Bilgin http://orcid.​ ​org/0000-​ ​0002-​2260-​4660 Competing interests None declared. Dilek Solmaz http://orcid.​ ​org/0000-​ ​0002-​9035-​689X Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. References 1 Flossmann O, Berden A, de Groot K, et al. Long-­Term patient survival in ANCA-­ Patient consent for publication Not required. associated vasculitis. Ann Rheum Dis 2011;70:488–94. Provenance and peer review Not commissioned; externally peer reviewed. 2 de Lind van Wijngaarden RAF, Hauer HA, Wolterbeek R, et al. Clinical and histologic determinants of renal outcome in ANCA-­associated vasculitis: a prospective analysis of © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and 100 patients with severe renal involvement. J Am Soc Nephrol 2006;17:2264–74. permissions. Published by BMJ. 3 Berden AE, Ferrario F, Hagen EC, et al. Histopathologic classification of ANCA-­ associated glomerulonephritis. J Am Soc Nephrol 2010;21:1628–36. 4 Chen Y-­X, Xu J, Pan X-­X, et al. Histopathological classification and renal outcome in patients with antineutrophil cytoplasmic antibodies-­associated renal vasculitis: a study To cite Gercik O, Bilgin E, Solmaz D, et al. Ann Rheum Dis 2020;79:675–676. of 186 patients and metaanalysis. J Rheumatol 2017;44:304–13. 5 Brix SR, Noriega M, Tennstedt P, et al. Development and validation of a renal risk score Received 29 November 2019 in ANCA-­associated glomerulonephritis. Kidney Int 2018;94:1177–88.

676 Ann Rheum Dis May 2020 Vol 79 No 5 Correspondence

Correspondence to Dr Mehmet Eyuboglu, Cardiology, Bergama State Hospital, Inflammatory and non-­inflammatory triggers of Izmir 35000, Turkey; mhmtybgl@​ ​gmail.com​ acute coronary syndromes Handling editor Josef S Smolen Inflammation plays an important role in the pathogenesis of acute coronary syndromes (ACS).1 However, ACS include different Contributors I am the sole author. clinical conditions that can present as an acute cardiac event, Funding The authors have not declared a specific grant for this research from any and it may also be seen in various clinical settings independent funding agency in the public, commercial or not-­for-­profit sectors. of underlying cardiovascular status and traditional risk factors.2 Competing interests None declared. 3 Recently, Westerlind et al reported that siblings of patients with Patient consent for publication Not required. rheumatoid arthritis (RA) are at increased risk of ACS that could Provenance and peer review Not commissioned; internally peer reviewed. not be explained by traditional cardiovascular risk factors. This © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and article provides excellent information regarding importance of permissions. Published by BMJ. inflammation and its hereditary role in the pathophysiology of ACS. Nevertheless, because of many clinical conditions may be the underlying cause of ACS and myocardial infarction (MI), there are some important points to be discussed to clarify the To cite Eyuboglu M. Ann Rheum Dis 2020;79:e49. findings of the present article. Received 24 February 2019 ACS and MI may also occur as an acute cardiac event in the Accepted 25 February 2019 absence of significant coronary artery disease (CAD) and inflam- Published Online First 6 March 2019 mation. This type of MI is known as type 2 MI, and may occur in the context of a mismatch between myocardial oxygen supply and demand due to various diseases.4 Therefore, ACS is not only triggered by atherosclerotic plaque disruption, it may occur due ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215293 to non-­cardiovascular reasons, and this situation should be taken Ann Rheum Dis 2020;79:e49. doi:10.1136/annrheumdis-2019-215277 into consideration while defining ACS. Additionally, inflamma- tion also seems to play a role in the mechanism of ACS secondary ORCID iD to non-­atherosclerotic spontaneous coronary artery dissection.5 Mehmet Eyuboglu http://orcid.​ ​org/0000-​ ​0002-​5754-​9382 Hence, it would be worth to know the CAD severity and the underlying mechanism of ACS of the study groups in the study References by Westerlind et al,3 to comment on the relationship between 1 Mulvihill NT, Foley JB. Inflammation in acute coronary syndromes. Heart index event and patients’ clinical conditions. 2002;87:201–4. 2 Roffi M, atronoP C, Collet JP, et al. ESC guidelines for the management of acute In conclusion, rheumatoid diseases and inflammation have coronary syndromes in patients presenting without persistent ST-­segment elevation: important roles in the pathophysiology of CAD and ACS. Task Force for the management of acute coronary syndromes in patients presenting However, ACS and MI do not occur only due to atherosclerotic without persistent ST-­segment elevation of the European Society of cardiology (ESC). plaque disruption. Non-­cardiovascular and non-­atherosclerotic Eur Heart J 2015;2016:267–315. reasons also may trigger ACS and MI in various clinical condi- 3 Westerlind H, Holmqvist M, Ljung L, et al. Siblings of patients with rheumatoid arthritis are at increased risk of acute coronary syndrome. Ann Rheum Dis 2019;78:683–7. tions. Hence, the underlying mechanism of ACS should be 4 Thygesen K, Alpert JS, Jaffe AS, et al. Fourth universal definition of myocardial infarction clearly stated to claim a relationship between siblings of patients (2018). Eur Heart J 2019;40:237–69. with RA and an increased risk of ACS. 5 Canga Y, Guvenc TS, Calik AN, et al. Systemic inflammatory activation in patients with acute coronary syndrome secondary to nonatherosclerotic spontaneous coronary artery Mehmet Eyuboglu dissection. North Clin Istanb 2018;5:186–94.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 1 Correspondence response Response to: ‘Inflammatory and non-­ Helga Westerlind ‍ ‍ ,1 Marie Holmqvist,1 Lotta Ljung,1,2 Thomas Frisell ‍ ‍ ,1 Johan Askling1,3 inflammatory triggers of acute coronary 1Clinical Epidemiology Division, Department of Medicine, Solna, Karolinska Institutet, Stockholm, Sweden syndromes’ by Eyuboglu 2Department of Public Health and Clinical Medicine/Rheumatology, Umeå University, Umeå, Sweden 1 We thank Dr Eyuboglu for his comments on our article ‘Siblings 3Department of Rheumatology, Theme Infection and Inflammation, Karolinska of rheumatoid arthritis patients are at increased risk for acute University Hospital, Stockholm, Sweden 2 coronary syndrome’. Correspondence to Dr Helga Westerlind, Department of Medicine, Karolinska Familial risks tell us if risk factors for a disease are shared Institutet, S-171 77 Solna, Sweden; ​Helga.​Westerlind@​ki.​se within families. By studying familial risks, we can also investi- Handling editor Prof Josef S Smolen gate shared susceptibility between diseases. In our present study, Funding This study was funded by Stockholm County Council (ALF), The Nordic we have, by using the full siblings of patients with rheumatoid Research Council (Nordforsk), Karoliska Institutet (Strategic Research Area arthritis (RA), created a study population that are enriched Epidemiology), Stiftelsen för Strategisk Forskning, Heart Lung Foundation, Swedish for both the genetic and (familial) environmental risk factors Research Council and The Rheumatology Research Foundation (FOREUM). involved in RA. If any disease has an increased frequency in our Competing interests JA has or has had research agreements with Abbvie, cohort of siblings of patients with RA, a cohort with an overrep- Astra-­Zeneca, BMS, Eli Lilly, MSD, Pfizer, Roche, Samsung Bioepis and UCB, mainly in the context of safety monitoring of biologics via ARTIS/Swedish Biologics Register. resentation of RA risk factors compared with the general popu- Karolinska Institutet has received remuneration for JA participating in advisory lation, it provides evidence for a shared susceptibility between boards arranged by Pfizer and Lilly. that disease and RA. Patient consent for publication Not required. What a potential observed risk increase does not tell us is Provenance and peer review Commissioned; internally peer reviewed. whether it is due to genetic or environmental factors. And, as Dr © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and Eyoboglu points out, it also does not tell us what type of under- permissions. Published by BMJ. lying mechanisms are involved. But it does tell us that there is a relationship between the two diseases, regardless of mechanisms of actions. We share Dr Eyuboglu’s interest in what the underlying mech- To cite Westerlind H, Holmqvist M, Ljung L, et al. Ann Rheum Dis 2020;79:e50. anisms are and completely agree that knowing more about the Received 4 March 2019 pathophysiology underlying the identified acute myocardial Accepted 5 March 2019 infarctions (AMI) among the acute coronary syndromes (ACS) in Published Online First 28 March 2019 our study population would be of great interest. Unfortunately, that information was not within our data and it was not possible for us to shed light on this in our present study. We do know, however, that close to 90% of the AMI registered during 2011 ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215277 3 in Sweden were type 1 AMI, meaning that the majority of the Ann Rheum Dis 2020;79:e50. doi:10.1136/annrheumdis-2019-215293 events among the siblings of the patients with RA, are likely to be type 1 AMI with an underlying atherothrombotic coronary ORCID iDs Helga Westerlind http://orcid.​ ​org/0000-​ ​0003-​3380-​5342 event. Thomas Frisell http://orcid.​ ​org/0000-​ ​0002-​5735-​9626 So in conclusion, the increase in ACS risk among the siblings tells us that there is a family-­shared link between RA and ACS, References but it does not tell us how and why. We thus agree with Dr 1 Eyuboglu M. Inflammatory and non-­inflammatory triggers of acute coronary syndromes. Eyuboglu that looking further into the underlying mechanisms Ann Rheum Dis 2020;79:e49. of the ACS events are of great interest and importance to learn 2 Westerlind H, Holmqvist M, Ljung L, et al. Siblings of patients with rheumatoid arthritis are at increased risk of acute coronary syndrome. Ann Rheum Dis 2019;78:683–7. more about the disease mechanisms, and we hope to study this 3 Baron T, Hambraeus K, Sundström J, et al. Type 2 myocardial infarction in clinical further in coming studies. practice. Heart 2015;101:101–6.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 1 Correspondence Assessment of responsiveness of the without a swollen joint. Ultrasound-only­ synovitis was clinically significant, being associated with worse serology, and patient-­ musculoskeletal component of SLE-­DAS in an reported and physician-­reported outcomes.4 Since the muscu- independent cohort loskeletal component of SLE-DAS­ is based entirely on swollen joint count, it would not be able to detect or monitor disease We read with interest the report by Jesus et al on the Systemic activity in this subgroup. Lupus Erythematosus Disease Activity Score (SLE-­DAS).1 We In summary, we validate the superior responsiveness of the agree that the systemic lupus erythematosus disease activity musculoskeletal component of SLE-DAS­ compared with the index 2000 (SLEDAI-­2K) in its original form has limitations as a SLEDAI-­2K in an independent longitudinal study. However, disease activity tool. Since each organ manifestation can only be our data suggest that a better outcome measure focusing on this graded as present or absent, it is unable to differentiate between specific organ system could be designed. mild, moderate or severe disease activity, nor partial response in Sabih-­Ul Hassan ‍ ‍ ,1,2 Khaled Mahmoud,1,2 Edward M Vital1,2 any feature. 1Department of Rheumatology, University of Leeds, Leeds Institute of Rheumatic and Our group is researching outcome measures in musculoskel- Musculoskeletal Medicine, Leeds, UK etal SLE. We published a previous study of responsiveness of 2Leeds Biomedical Research Centre, Leeds, UK various musculoskeletal outcome measures in patients treated 2 Correspondence to Dr Edward M Vital,University of Leeds Leeds Institute of with glucocorticoids. As expected, the musculoskeletal compo- Rheumatic and Musculoskeletal Medicine, Leeds, Leeds, UK; e​ .​m.j.​ ​vital@​leeds.​ac.uk​ nent of the SLEDAI-­2K was less responsive than the physician Contributors KM and EMV collected the data. SUH analysed the data and wrote visual analog scale (physician VAS), musculoskeletal component the manuscript. All authors reviewed the manuscript and approved the final version. of the British Isles Lupus Assessment Group (BILAG) index or Funding This study was funded by NIHR (grant no: CS-2013-13-032). ultrasound, all of which allow for various grades of severity. Competing interests None declared. The musculoskeletal component of the new SLE-­DAS should, in theory, improve its responsiveness as it includes a term for Patient consent for publication Not required. swollen joint count as a continuous variable. We therefore calcu- Provenance and peer review Not commissioned; internally peer reviewed. lated musculoskeletal component of new SLE-­DAS tool in our Data sharing statement Readers may contact the corresponding author to previously published data set of 20 patients. These patients had discuss access to the data. objective synovitis, were treated with 120 mg of intramuscular © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and depomedrone and assessed at 0 and 4 weeks. We calculated the permissions. Published by BMJ. musculoskeletal component of the SLE-DAS­ using the terms: 3.132 x (Presence of Arthritis Yes/No)+ 0.454 x (swollen joint count in 28 joints)‍ To cite Hassan S-­U, Mahmoud K, Vital EM. Ann Rheum Dis 2020;79:e51. As in our publication, change was assessed using Wilcoxon Received 20 March 2019 signed-­rank test. Effect size was calculated using standardised Accepted 24 March 2019 test statistic, Z, using the formula r=Z/sqrt(n1+n2). Effect sizes Published Online First 25 April 2019 were judged using Cohen’s criteria as large (≥0.5), medium (≥0.3) or small (≥0.1).3 As expected, and like the other variables, the musculoskeletal SLE-­DAS significantly reduced from median 13.4 (IQR 12.6– ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215430 14.7) to median 6.2 (IQR 0.0–12.5, p=0.001). The effect size Ann Rheum Dis 2020;79:e51. doi:10.1136/annrheumdis-2019-215411 was large at −0.548. As predicted by its developers, this was superior to the musculoskeletal component of SLEDAI-­2K which ORCID iD Sabih-Ul Hassan­ http://orcid.​ ​org/0000-​ ​0001-​6635-​8011 had an effect size of −0.418. However, the musculoskeletal SLE-­ DAS was not as responsive as the musculoskeletal component of References the BILAG index (−0.576), which could be graded as improving, 1 Jesus D, Matos A, Henriques C, et al. Derivation and validation of the SLE disease or physician VAS (−0.599), which was a continuous variable. activity score (SLE-­DAS): a new SLE continuous measure with high sensitivity for However, the disadvantage of these two variables is that they are changes in disease activity. Ann Rheum Dis 2019;78:365–71. heavily dependent on the training and expertise of the assessors. 2 Mahmoud K, Zayat AS, Yusof Y, et al. Responsiveness of clinical and ultrasound Measures based on joint counts like SLE-DAS­ may show more outcome measures in musculoskeletal systemic lupus erythematosus. Rheumatology 2019;62. robust inter-­reader reliability. 3 Cohen J. Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum, There is another crucial feature of musculoskeletal lupus in 1988. which the BILAG index and physician VAS are advantageous: 4 Zayat AS, Mahmoud K, Md Yusof MY, et al. Defining inflammatory musculoskeletal subclinical synovitis. We recently showed that in patients with manifestations in SLE. Rheumatology. 2018;58:304–12. 5 Jesus D, Zen M, Doria A, et al. Response to: ’Assessment of responsiveness of the SLE presenting with inflammatory joint pain, 38% had swollen musculoskeletal component of SLE-­DAS in an independent cohort’, by Hassan, et al. joints. However, a further 27% had ultrasound-proven­ synovitis Ann Rheum Dis 2020;79:e52.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 1 Correspondence response

1Rheumatology Department, Centro Hospitalar e Universitario de Coimbra EPE, Response to: ‘Assessment of responsiveness of Coimbra, Portugal 2School of Health Sciences, University of Beira Interior, Covilhã, Portugal the musculoskeletal component of SLE-­DAS in 3 an independent cohort’, by Hassan et al Rheumatology Unit, Department of Medicine, University of Padova, Padova, Italy Correspondence to Dr Diogo Jesus, Serviço de Reumatologia, Centro Hospitalar e Universitário de Coimbra, Coimbra, Portugal; Jesus​ .​p.diogo@​ ​gmail.​com It was with great interest that we read the letter ‘Assessment of responsiveness of the musculoskeletal component of SLE-DAS­ in Handling editor Josef S Smolen an independent cohort’ by Hassan et al.1 Competing interests None declared. We recently published the derivation and validation of the Patient consent for publication Not required. Systemic Lupus Erythematosus Disease Activity Score (SLE-­ Provenance and peer review Commissioned; internally peer reviewed. DAS), a new continuous measure of SLE disease activity with 2 © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and high sensitivity and specificity to change. permissions. Published by BMJ. Hassan et al report in their letter that in an analysis of 20 SLE patients, the responsiveness of the arthritis component of SLE-­DAS (large effect size of −0.548) is superior to that of the arthritis component of SLE Disease Activity Index (DAI)-2K­ To cite Jesus D, Zen M, Doria A, et al. Ann Rheum Dis 2020;79:e52. (medium effect size of −0.418).1 3 These results corroborate our 2 4 Received 2 April 2019 data. Moreover, Hassan et al report that the arthritis compo- Accepted 2 April 2019 nent of SLE-DAS­ presents a similarly large effect size comparing Published Online First 25 April 2019 with the arthritis component of the British Isles Lupus Assess- ment Group (BILAG) (−0.576). As highlighted by Hassan et al, the SLE-DAS­ is expected to present a more robust inter-­reader reliability than physician visual analogue scale and BILAG, and ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215411 to be less dependent on the training and expertise of the asses- Ann Rheum Dis 2020;79:e52. doi:10.1136/annrheumdis-2019-215430 sors than BILAG. We further emphasise that SLE-DAS­ is a global disease activity ORCID iDs index (non-­organ specific), which presents a major advantage Diogo Jesus http://orcid.​ ​org/0000-​ ​0003-​3136-​0722 over BILAG for use in daily clinical practice, as it is much less Andrea Doria http://orcid.​ ​org/0000-​ ​0003-​0548-​4983 time-­consuming and requires a collection time similar to SLEDAI. Hassan et al report that joint ultrasonography (US) is able References 1 Hassan S, Mahmoud K, Vital E. Assessment of responsiveness of the musculoskeletal to detect subclinical synovitis in SLE patients. We agree that in component of SLE-­DAS in an independent cohort. Ann Rheum Dis 2020;79:e51. selected patients joint US can be useful to complement clinical 2 Jesus D, Matos A, Henriques C, et al. Derivation and validation of the SLE disease assessment. However, the requirement for US joint assessment activity score (SLE-­DAS): a new SLE continuous measure with high sensitivity for is not advisable to integrate a disease activity instrument, such changes in disease activity. Ann Rheum Dis 2019;78:365–71. as SLE-DAS,­ as it would constitute a major barrier for its prac- 3 Mahmoud K, Zayat AS, Yusof Y, et al. Responsiveness of clinical and ultrasound outcome measures in musculoskeletal systemic lupus erythematosus. Rheumatology ticality. Importantly, previous studies did not find any added 2019;62. value of systematic joint US assessment for treatment decision in 4 Jesus D, Rodrigues M, Matos A, et al. Performance of SLEDAI-­2K to detect a clinically chronic arthritis.5–7 meaningful change in SLE disease activity: a 36–month prospective cohort study of In summary, this letter by Hassan et al corroborates the 334 patients. Lupus 2019;35. improved responsiveness of the arthritis component of SLE-DAS­ 5 Dale J, Stirling A, Zhang R, et al. Targeting ultrasound remission in early rheumatoid arthritis: the results of the TaSER study, a randomised clinical trial. Ann Rheum Dis in an independent longitudinal cohort. The relative sensitivity 2016;75:1043–50. and specificity of SLE-DAS­ and BILAG will be further tested in 6 Haavardsholm EA, Aga A-­B, Olsen IC, et al. Ultrasound in management of rheumatoid our upcoming study comparing BILAG and SLE-DAS.­ arthritis: Arctic randomised controlled strategy trial. BMJ 2016;354. 7 Aletaha D, Smolen JS. Achieving clinical remission for patients with rheumatoid Diogo Jesus ‍ ‍ ,1,2 Margherita Zen,3 Andrea Doria ‍ ‍ ,3 Luís Sousa Inês1,2 arthritis. JAMA 2019;321:457–8.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 1 Correspondence Adipose stromal vascular fraction and Paola Di Benedetto ‍ ‍ ,1 Paola Cipriani,2 Piero Ruscitti,2 Vasiliki Liakouli,2 Roberto Giacomelli2 regenerative therapy in SSc: response to the 1Clinical Pathology, Department of Biotechnological and Applied Clinical Sciences, University of L’Aquila, L’Aquila, Italy article by Magalon et al 2Rheumatology Unit, Department of Biotechnological and Applied Clinical Sciences, University of L’Aquila, L’Aquila, Italy We read with interest the report by Magalon et al,1 supporting Correspondence to Dr Paola Di Benedetto, Department of Biotechnological and the use of stromal vascular fraction (SVF) derived from adipose Applied Clinical Sciences, Clinical Pathology, University of L’Aquila, L’Aquila 67100, tissue, as an innovative autologous biotherapy for patients Italy; paola.​ ​dibenedetto1@univ​ aq.​it affected by systemic sclerosis (SSc). This disease is a complex multisystem disorder, characterised by microvascular damage, Handling editor Josef S Smolen dysregulation of innate and adaptive immunity and generalised fibrosis in the skin and multiple organs, lacking a specific Contributors PDB and RG wrote, analysed and conceived the study. PC, PR and therapy. SSc is classified into two subsets: limited cutaneous VL analysed and conceived the study. All authors contributed to critical review and revision of the paper and approved the final version. (lcSSc), in which the skin fibrosis is limited to the extremities (hands, arms and face), and diffuse cutaneous (dcSSc), in which 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. the skin fibrosis progresses more rapidly, affecting the trunk and extremities, frequently involving the internal organs.2 Due Competing interests None declared. to this heterogeneity, in terms of extent, severity and rate of Patient consent for publication Not required. progression, it is still a matter of debate if both the subsets Provenance and peer review Not commissioned; internally peer reviewed. share a common pathogenic mechanism. At present, a growing © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and body of evidence has been published concerning mesenchymal permissions. Published by BMJ. stem cells (MSCs) therapy for patients with SSc—the cell popu- lation larger represented in the SVF, in the last few years. In this work, the authors1 showed that the transcriptome profile To cite Di Benedetto P, Cipriani P, Ruscitti P, et al. Ann Rheum Dis 2020;79:e53. of SVF, derived from patients with SSc, exhibited a profibrotic signature, without compromising the vascular repair capacity of Received 22 January 2019 SVF. In this context, we previously published data, confirmed Accepted 24 January 2019 also by other groups, showing a strong profibrotic signature Published Online First 2 February 2019 in the MSCs, derived from patients with dcSSc,2–6 and thus suggesting their involvement in the rapid evolution of fibrosis. Furthermore, the authors enrolled nine patients with dcSSc and 15 patients with lcSSc. Of note, the authors1 did not stratify ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215132 the results between the two subsets and analysing the graphs Ann Rheum Dis 2020;79:e53. doi:10.1136/annrheumdis-2019-215113 of the paper, we may observe a large distribution of the data ORCID iD which partially impairs the results. Furthermore, the majority Paola Di Benedetto http://orcid.​ ​org/0000-​ ​0003-​2927-​8703 of the experiments were performed in very small number of patients (from 4 to 8) and without any explanation about References how they were selected. Given the different profibrotic fate 1 Magalon J, Velier M, Simoncini S, et al. Molecular profile and proangiogenic activity of patients with dcSSc and the significant profibrotic signature of the adipose-­derived stromal vascular fraction used as an autologous innovative of MSCs derived from these patients, it should be important medicinal product in patients with systemic sclerosis. Ann Rheum Dis 2019;78:391–8. to compare the presence of possible different SVF effects, 2 Gabrielli A, Avvedimento EV, Krieg T. Scleroderma. N Engl J Med 2009;360:1989–2003. 3 Hegner B, Schaub T, Catar R, et al. Intrinsic deregulation of vascular smooth muscle and deriving from the two different subsets of the disease. Paradox- myofibroblast differentiation in mesenchymal stromal cells from patients with systemic ically, due to the MSCs profibrotic signature in SSc, it could sclerosis. PLoS One 2016;11:e0153101. be possible that the pathogenic milieu may force autologous 4 Cipriani P, Marrelli A, Benedetto PD, et al. Scleroderma mesenchymal stem cells display cells to transdifferentiate to myofibroblasts, thus increasing a different phenotype from healthy controls; implications for regenerative medicine. 2–6 Angiogenesis 2013;16:595–607. the fibrotic burden. In fact, the transcriptomic analysis, in 5 Di Benedetto P, Liakouli V, Ruscitti P, et al. Blocking CD248 molecules in perivascular 1 the study of Magalon et al confirmed the GDF15 gene over- stromal cells of patients with systemic sclerosis strongly inhibits their differentiation expression which is strongly related to the fibrotic pathway. toward myofibroblasts and proliferation: a new potential target for antifibrotic therapy. Further refinements in the understanding of SSc-­SVF profile, Arthritis Res Ther 2018;20. together with targeted approaches to both SSc endothelial cells 6 Cipriani P, Di Benedetto P, Liakouli V, et al. Mesenchymal stem cells (MscS) from scleroderma patients (SSC) preserve their immunomodulatory properties although and pathological microenvironment, are needed to improve the senescent and normally induce T regulatory cells (Tregs) with a functional phenotype: regenerative approach of SSc. implications for cellular-­based therapy. Clin Exp Immunol 2013;173:195–206.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 1 Correspondence response Response to: ‘Adipose stromal vascular fraction in vitro and in vivo functional assays to address the contribution of each of the cellular subsets and their interactive benefit in the and regenerative therapy in SSc: response to the modulation of fibrosis, angiogenesis and inflammation. These article by Magalon et al’ by De Benedetto et al targeted approaches, together with outputs from the ongoing clin- ical trials, are needed to further define the optimal strategies for We would like to thank Di Benedetto et al1 for their construc- cell-based­ therapies of SSc. tive comments on our recent publication in Annals of Rheumatic 1,2 1,2 2 Disease entitled ‘Molecular profile and proangiogenic activity of Jérémy Magalon ‍ ‍ , Mélanie Velier, Stéphanie Simoncini, Françoise Dignat-­George,2 Brigitte Granel,2,3 Pascale Paul,1,2 the adipose-­derived stromal vascular fraction used as an autologous Florence Sabatier1,2 innovative medicinal product in patients with systemic sclerosis’.2 1Cell Therapy Department, Hôpital de la Conception, Marseille, France Di Benedetto et al mentioned that the profibrotic signature 2Aix Marseille University, Marseille, France of mesenchymal stem cells (MSCs) derived from patients with 3Internal Medicine Department, Hôpital Nord, Marseille, France the diffuse cutaneous form of systemic sclerosis (dc-­SSc) has Correspondence to Professor Florence Sabatier, Cell Therapy Department, Hôpital been established from previous works. In line with this, they de la Conception, Assistance Publique Hôpitaux de Marseille, Marseille 13005, discussed the potential risk of increasing the fibrotic burden France; florence​ .​sabatier@ap-​ ​hm.​fr by using autologous adipose-­derived stromal vascular fraction Handling editor Josef S Smolen (ADVSF) in which MSCs are highly represented. Contributors Drafting of the manuscript:JM, MV, SS. Critical revision of the Although this is a fair question in the context of developing manuscript for important intellectual content: FDG, BG, PP and FS. MSCs-­based therapy for SSc, we would like to emphasise the fact Competing interests None declared. that the biological findings referred to by Di Benedetto et al, may not be fully extrapolated to the ADSVF used in our study. Indeed, Patient consent for publication Not required. the mentioned studies rather relate to MSCs derived in culture Provenance and peer review Commissioned; internally peer reviewed. from the bone marrow using research-­grade reagents that are very © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and different from the mesenchymal cell compartment that we aimed permissions. Published by BMJ. to describe in the non-­cultured clinical-grade­ ADSVF. Indeed, the use of MSCs necessitate an in vitro amplification procedure which can have a major impact on the differentiation and characteristics To cite Magalon J, Velier M, Simoncini S, et al. Ann Rheum Dis 2020;79:e54. of MSCs,3 4 thereby limiting the comparison between native MSCs present within ADSVF. ADSVF is also a complex product whose Received 4 February 2019 biological properties result from multiple interactions between Revised 5 February 2019 various cell subpopulations, and the role of each cell subset remains Accepted 5 February 2019 Published Online First 20 February 2019 to be identified. In addition, it is well established that cultured MSCs derived from bone marrow and adipose tissue are quite different populations5 and to our knowledge, such a profibrotic profile of MSCs derived from the adipose tissue of patients with SSc was not reported, as recently confirmed by the review from ►► https://​doi.org/​ ​10.1136/​ ​annrheumdis-2019-​ ​215113 Rozier et al.6 In 2017, report from the team of the Department of Ann Rheum Dis 2020;79:e54. doi:10.1136/annrheumdis-2019-215132 Biotechnological and Applied Clinical Sciences (l’Aquila University, ORCID iD Italy) indicated that MSCs isolated from adipose tissue of dc-SSc­ Jérémy Magalon http://orcid.​ ​org/0000-​ ​0003-​1494-​7011 may represent a possible therapeutic option as they show similar 7 biological properties compared with MSCs from healthy donors. References From a clinical point of view, we are surprised that Benedetto 1 Di Benedetto P, Cipriani P, Ruscitti P, et al. Adipose stromal vascular fraction and et al omitted to mention that autologous ADSVF, injected in the regenerative therapy in SSc: response to the article by Magalon et al. Ann Rheum Dis fingers of 12 patients with SSc (including five dc-SSc)­ displaying 2020;79:e53. hand disability, was shown to have a good safety profile and a 2 Magalon J, Velier M, Simoncini S, et al. Molecular profile and proangiogenic activity 8 of the adipose-­derived stromal vascular fraction used as an autologous innovative potential efficacy. In particular, results of Rodnan score applied medicinal product in patients with systemic sclerosis. Ann Rheum Dis 2019;78:391–8. to the hands, with a follow-­up until 2 years, did not indicate any 3 Hagmann S, Moradi B, Frank S, et al. Different culture media affect growth worsening of the fibrosis aspect.9 characteristics, surface marker distribution and chondrogenic differentiation of human We would also like to emphasise that the potential strength of our bone marrow-­derived mesenchymal stromal cells. BMC Musculoskelet Disord 2013;14. study was that it was performed on clinical-grade­ ADSVF, meaning 4 Guimarães-Camboa­ N, Cattaneo P, Sun Y, et al. Pericytes of Multiple Organs Do Not Behave as Mesenchymal Stem Cells In Vivo. Cell Stem Cell 2017;20:345–59. that the analysed ADSVF correspond to the products potentially 5 Noël D, Caton D, Roche S, et al. Cell specific differences between human adipose-­ injected in the SCLERADEC II randomised placebo-controlled­ derived and mesenchymal-­stromal cells despite similar differentiation potentials. Exp trial (NCT02558543). From our point of view, evaluation of this Cell Res 2008;314:1575–84. therapeutic ADSVF makes our findings of particular clinical rele- 6 Rozier P, Maria A, Goulabchand R, et al. Mesenchymal stem cells in systemic sclerosis: vance but limits the amount of ADSVF available for biological allogenic or autologous approaches for therapeutic use? Front Immunol 2018;9. 7 Capelli C, Zaccara E, Cipriani P, et al. Phenotypical and functional characteristics of investigations. This is the reason why functional angiogenic assays in Vitro-­Expanded adipose-­derived mesenchymal stromal cells from patients with could not be performed on a larger number of samples prepared systematic sclerosis. Cell Transplant 2017;26:841–54. from various forms of the disease. We fully agree that additional 8 Granel B, Daumas A, Jouve E, et al. Safety, tolerability and potential efficacy of injection studies are needed to consider whether the limited or diffuse cuta- of autologous adipose-­derived stromal vascular fraction in the fingers of patients with neous form of SSc have a similar impact on ADSVF properties. As systemic sclerosis: an open-­label phase I trial. Ann Rheum Dis 2015;74:2175–82. suggested by Benedetto et al, refinements in the characterisation 9 Daumas A, Magalon J, Jouve E, et al. Long-­term follow-­up after autologous adipose-­ of SSc-ADSVF­ profile is still required. Beyond transcriptomic and derived stromal vascular fraction injection into fingers in systemic sclerosis patients. phenotypic signature, further work should implement accurate Curr Res Transl Med 2017;65:40–3.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 1 Correspondence Could autologous adipose-­derived stromal from patients with diffuse cutaneous SSc and reported somehow different results.5 In fact, Virzì et al5 have shown that SSc-­ vascular fraction turn out an unwanted source ADSVF displays an altered cell composition characterised by a of profibrotic myofibroblasts in reduced number of cells with a stem-­like phenotype, as well as a high content of proinflammatory cytokines and a shortage of systemic sclerosis? angiogenic factors. Moreover, at variance with an earlier study showing that isolated SSc-ADSC­ exhibit similar biological prop- With great interest, I read the recent publication by Magalon erties (ie, surface antigenic profile, proliferation and differen- et al1 in the Annals of the Rheumatic Diseases. The results of tiation potentials, immunosuppressive properties and capacity this study are placed in the scenario of a quite complex disease to support endothelial cell tube formation) compared with featuring vasculopathy, autoimmunity and extensive multiorgan HD-­ADSC,6 Virzì et al5 found that SSc-­ADSC may retain high fibrosis whose life-threatening­ nature is well testified by a 5-year­ multipotency but fail to sustain terminally differentiated adipo- mortality of 30%–50% in a subset of patients with diffuse cuta- cyte, osteocyte and chondrocyte progenies. Second, similarly to neous systemic sclerosis (SSc) and internal organ involvement.2 previous findings on SSc bone marrow-­derived mesenchymal Although substantial basic/translational and clinical research stem/stromal cells,7 Lee et al8 have recently reported that early progresses have been achieved over the past decade, current passage SSc-ADSC­ have a profibrotic and antiadipogenic pheno- treatments are mainly organ based and do not result in a cure type characterised by high levels of the myofibroblast marker which highlights the urgent need of developing new poten- α-smooth muscle actin and low expression of both caveolin-1 tially disease-modifying­ therapies and personalised medicine and the adipogenic marker FABP4. Of note, they also demon- approaches.2 strated that a myofibroblast-­like phenotype could be induced In this context, the adipose-derived­ stromal vascular frac- in HD-­ADSC by treatment with transforming growth factor-β8 tion (ADSVF) has recently gained attention as an innovative which suggests that the SSc pathological environment might be biotherapy because of its abundance of mesenchymal-like­ stem/ relevant in determining an unwanted profibrotic fate of ADSC. stromal cells (referred to as adipose-derived­ stem cells; ADSC), Finally, we should not overlook that cell fate mapping studies accessibility and ease of harvest. The same research team has in the bleomycin-­induced mouse model of skin fibrosis clearly previously reported encouraging results from a phase I clinical demonstrated that adiponectin-­positive progenitors that are trial showing a good safety profile and a potential efficacy of normally confined to the intradermal adipose tissue compart- local injection of autologous ADSVF to treat hand disability in ment redistribute into the lesional dermis, where they lose the patients with SSc.3 This evidence was further substantiated by adipocytic markers and acquire a myofibroblast-like­ pheno- an up to 2-­year follow-­up analysis indicating an improvement of type.9 Taken together, it is clear that the functional experimental ischaemic vasculopathy (ie, Raynaud’s phenomenon and digital approach employed by Magalon et al1 could exclusively disclose ulcers), hand pain and global quality of life without any signifi- the vascular repair performance of SSc-ADSVF,­ but much work cant worsening of fibrosis as assessed by modified Rodnan skin is still to be done (eg, xenogeneic human SSc-ADSVF­ injection score applied to the hands.4 The aforementioned observations in mouse models) to definitely rule out that autologous ADSVF represented the obvious groundwork for the herein commented might even behave as an unwanted source of profibrotic myofi- study,1 in which the authors provided a thorough characterisa- broblasts in a therapeutic setting. Based on either the evidence tion of the cellular and molecular profiles of therapeutic-­grade that the specific pathological environment might be crucial in ADSVF harvested from patients with SSc in comparison with affecting the ADSC fate or the lesson we recently learnt from healthy donors (HD). Such an analysis is of extraordinary gene expression profiling studies on the existence of different importance, especially when considering that the composition of molecular subtypes of SSc fibrotic skin disease,8 10 a future ADSVF is extremely heterogeneous as it includes not only ADSC, in-­depth molecular and functional characterisation of ADSVF in that represent its major constituents, but also endothelial cells/ larger SSc cohorts has the great potential to help in predicting progenitors, pericytes, and haematopoietic and immune cells.1 which patients are more likely to benefit from an autologous The authors reported no substantial difference in the distribu- ADSVF-­based therapeutic approach. tion of the various cell populations in SSc-ADSVF­ compared with HD-ADSVF,­ and the two products performed similarly with Mirko Manetti only slight differences in terms of their vasculogenic/angiogenic capacity assessed by a number of in vitro and in vivo assays.1 Correspondence to Dr Mirko Manetti, Department of Experimental and Clinical Medicine, Section of Anatomy and Histology, University of Florence, Florence Nonetheless, global and single-cell­ RNA-sequencing­ approaches I-50134, Italy; mirk​ o.​manetti@unifi​ .​it and analysis of ADSVF-derived­ secreted factors revealed a differential endothelial and stromal cell molecular signature of Handling editor Josef Smolen the SSc-ADSVF­ reflecting endothelial activation, deregulation of angiogenesis and fibrosis, such as an upregulation of growth Contributors MM has completely drafted and finalised the manuscript. differentiation factor-15 that has previously been implicated in Funding The author has not declared a specific grant for this research from any SSc-­related fibrosis.1 funding agency in the public, commercial or not-­for-­profit sectors. I particularly appreciated the balanced conclusions of the Competing interests None declared. authors that not only emphasised the potential of autologous Patient consent for publication Not required. ADSVF to treat SSc-­related vasculopathy on the basis of its Provenance and peer review Not commissioned; internally peer reviewed. preserved vascular repair capacity, but also frankly acknowl- edged that the significance of the molecular profile alterations © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and permissions. Published by BMJ. detected in the SSc-ADSVF­ deserves an in-depth­ investigation, since they may be relevant for the fibrotic process.1 In this regard, I believe that some additional cues are worth consid- ering. First, another study has previously analysed the ADSVF To cite Manetti M. Ann Rheum Dis 2020;79:e55.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 2 Correspondence

Received 25 February 2019 of patients with systemic sclerosis: an open-­label phase I trial. Ann Rheum Dis Accepted 2 March 2019 2015;74:2175–82. Published Online First 13 March 2019 4 Daumas A, Magalon J, Jouve E, et al. Long-­term follow-­up after autologous adipose-­ derived stromal vascular fraction injection into fingers in systemic sclerosis patients. Curr Res Transl Med 2017;65:40–3. 5 Virzì F, Bianca P, Giammona A, et al. Combined platelet-­rich plasma and lipofilling treatment provides great Improvement in facial skin-­induced lesion regeneration for ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-215327​ scleroderma patients. Stem Cell Res Ther 2017;8. 6 Capelli C, Zaccara E, Cipriani P, et al. Phenotypical and functional characteristics of Ann Rheum Dis 2020;79:e55. doi:10.1136/annrheumdis-2019-215288 in vitro-­expanded adipose-­derived mesenchymal stromal cells from patients with systematic sclerosis. Cell Transplant 2017;26:841–54. ORCID iD 7 Hegner B, Schaub T, Catar R, et al. Intrinsic deregulation of vascular smooth muscle Mirko Manetti http://orcid.​ ​org/0000-​ ​0003-​3956-​8480 and myofibroblast differentiation in mesenchymal stromal cells from patients with systemic sclerosis. PLoS One 2016;11:e0153101. References 8 Lee R, Del Papa N, Introna M, et al. Adipose-­derived mesenchymal stromal/stem cells 1 Magalon J, Velier M, Simoncini S, et al. Molecular profile and proangiogenic activity in systemic sclerosis: alterations in function and beneficial effect on lung fibrosis are of the adipose-­derived stromal vascular fraction used as an autologous innovative regulated by caveolin-1. J Scleroderma Relat Disord 2019. medicinal product in patients with systemic sclerosis. Ann Rheum Dis 2019;78:391–8. 9 Marangoni RG, Korman BD, Wei J, et al. Myofibroblasts in murine cutaneous fibrosis 2 Manetti M, Matucci-­Cerinic M. The new frontier in systemic sclerosis: from epigenetics originate from adiponectin-­positive intradermal progenitors. Arthritis Rheumatol to new treatments. Rheumatology 2015;54:1757–8. 2015;67:1062–73. 3 Granel B, Daumas A, Jouve E, et al. Safety, tolerability and potential efficacy of 10 Martyanov V, Whitfield ML. Molecular stratification and precision medicine in systemic injection of autologous adipose-­derived stromal vascular fraction in the fingers sclerosis from genomic and proteomic data. Curr Opin Rheumatol 2016;28:83–8.

2 of 2 Ann Rheum Dis May 2020 Vol 79 No 5 Correspondence response Response to: ‘Could autologous adipose-­derived dysfunction constitutes the primum movens for fibrosis progression. Thus, the SCLERADEC phase I clinical trial stromal vascular fraction turn out an unwanted (NCT :01813279) and the preserved proangiogenic activity source of profibrotic myofibroblasts in systemic of SSc-ADSVF­ described in our study suggest that the pres- sclerosis?’ by Manetti ence of regenerative endothelial progenitors cells in ADSVF is a major advantage in preventing and/or limiting SSc-­ associated vasculopathy.2 13 In addition, transcriptomic and We would like to thank Dr Manetti for his relevant comments1 molecular signatures of SSc are not strictly correlated to the on our recent article in the Annals of Rheumatic Disease in vivo properties of progenitor cells. Indeed, we have found entitled ‘Molecular profile and proangiogenic activity of the several markers of endothelial activation and dysregulation adipose-­derived stromal vascular fraction used as an autolo- in the secretome of ADSVF from patients with SSc, without gous innovative medicinal product in patients with systemic 2 compromising its vascular repair capacity compared with sclerosis’ . We are pleased that this study has raised such a high interest in the scientific community and allows us to provide ADSVF from healthy donors. additional comments to those recently published3 4 in order Nonetheless, we agree with Dr Manetti that implementa- to discuss the potential profibrotic profile of adipose-­derived tion of in vitro and/or in vivo potency assays addressing both stem cells (ADSC) derived from patients with diffuse cutaneous vascular and fibrotic properties of ADSVF are needed to define systemic sclerosis (SSc). We agree that SSc is a rare but poten- the optimal cellular-based­ strategies in SSc. From our point tially lethal autoimmune disease that challenges the urgent of view, such developments can benefit from investigations need to develop novel therapeutic approaches. In this context, conducted using a pharmaceutical grade cell therapy product autologous fat grafting has been successfully used to limit SSc-­ infused in the frame of a controlled clinical trial in order to associated clinical complications such as SSc-related­ perioral integrate not only the patients and disease characteristics but thickening and mouth opening limitation5 facial handicap6 also the possible impact of the manufacturing process. 7 or digital ulcers. These clinical results opened up research Melanie Velier,1,2 Jeremy Magalon ‍ ‍ ,1,2 Stephanie Simoncini,2 perspectives on autologous adipose tissue based therapies for Françoise Dignat-­George,2 Brigitte Granel,2,3 Pascale Paul,1,2 patients suffering from SSc. However, we agree that the ques- Florence Sabatier1,2 tion raised by Dr Manetti : ‘Could autologous adipose-derived­ 1Cell Therapy Department, Hopital de la Conception, Marseille, France stromal vascular fraction turn out an unwanted source of 2Cardiovascular and Nutrition Research Center, INSERM, INRA, Aix-­Marseille profibrotic myofibroblasts in systemic sclerosis?’ is of a major Universite, Marseille, France 3 concern. Internal Medicine Department, Hopital Nord, Marseille, France Indeed, some studies performed on skin biopsies in patients Correspondence to Professor Florence Sabatier, Marseille, France; with SSc suggested that perivascular cells from the mesen- florence​ .​sabatier@ap-​ ​hm.​fr chymal lineage were prone to myofibroblastic differentiation.8 Handling editor Josef S Smolen However, recent investigations regarding the impact of SSc Contributors Drafting of the manuscript: JM, MV, SS. Critical revision of the disease on the disorganisation of the adipose tissue archi- manuscript for important intellectual content: FD-­G, BG, PP and FS. 9 9 tecture or altered differentiation capabilities and reduced Funding The authors have not declared a specific grant for this research from any proliferation rate and metabolic activity of ADSC10 remain funding agency in the public, commercial or not-­for-­profit sectors. controversial. Conversely, Capelli and colleagues showed that Competing interests None declared. ADSC obtained from patients with SSc exhibit phenotypic Patient consent for publication Not required. pattern, proliferation, immunosuppressive properties and Provenance and peer review Commissioned; internally peer reviewed. differentiation potential that are similar to the ones observed in healthy controls, emphasising the safety of using autolo- © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and permissions. Published by BMJ. gous ADSC grafting as a therapeutic option for SSc.11 The discordance between these studies may be explained by (1) the variable sources and expansion protocols that allow to derive ADSC and (2) the high interindividual heterogeneity To cite Velier M, Magalon J, Simoncini S, et al. Ann Rheum Dis 2020;79:e56. between patients with SSc in terms of severity, duration and manifestations of the disease and ongoing treatment. More- Received 18 March 2019 Accepted 18 March 2019 over, these preliminary studies have concerned investigation Published Online First 9 April 2019 at the cellular levels whereas it is now well established that the mechanisms sustaining the function of ADSC mainly rely on the secretion of factors able to regulate endogenous cell activity. Although ADSC is one of the most represented cell ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215288 subtype in adipose-­derived stromal vascular fraction (ADSVF), we believe that conclusions driven from ADSC derived in Ann Rheum Dis 2020;79:e56. doi:10.1136/annrheumdis-2019-215327 culture from patients with SSc cannot be extrapolated to the ORCID iD freshly isolated autologous therapeutic ADSVF characterised Jeremy Magalon http://orcid.​ ​org/0000-​ ​0003-​1494-​7011 in our study. In line with this assumption, recent studies have highlighted differences in the paracrine content of expanded References and non-­expanded ADSC and suggest that ADSVF may secrete 1 Manetti Met al. Could autologous adipose-­derived stromal vascular fraction turn out a larger panel of soluble factors with beneficial properties for an unwanted source of profibrotic myofibroblasts in systemic sclerosis? Ann Rheum 12 Dis 2020;79:e55. cell therapy than ADSC. 2 Magalon J, Velier M, Simoncini S, et al. Molecular profile and proangiogenic activity Furthermore, although pathogenesis of SSc is not of the adipose-­derived stromal vascular fraction used as an autologous innovative fully understood, it is now recognised that endothelial medicinal product in patients with systemic sclerosis. Ann Rheum Dis 2019;78:391–8.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 2 Correspondence response

3 Di Benedetto P, Cipriani P, Ruscitti P, et al. Adipose stromal vascular fraction and 9 Virzì F, Bianca P, Giammona A, et al. Combined platelet-­rich plasma and lipofilling regenerative therapy in SSc: response to the article by Magalon et al. Ann Rheum Dis treatment provides great Improvement in facial skin-­induced lesion regeneration for 2020;79:e53. scleroderma patients. Stem Cell Res Ther 2017;8:236. 4 Magalon J, Velier M, Simoncini S, et al. Response to: ’Adipose stromal vascular 10 Griffin M, Ryan CM, Pathan O, et al. Characteristics of human adipose derived fraction and regenerative therapy in SSc: response to the article by Magalon et al’ by stem cells in scleroderma in comparison to sex and age matched normal controls: De Benedetto et al. Ann Rheum Dis 2020;79:e54. 5 Del Papa N, Caviggioli F, Sambataro D, et al. Autologous fat grafting in the treatment implications for regenerative medicine. Stem Cell Res Ther 2017;8:23 of fibrotic perioral changes in patients with systemic sclerosis. Cell Transplant 11 Capelli C, Zaccara E, Cipriani P, et al. Phenotypical and functional characteristics of 2015;24:63–72. in Vitro-­Expanded adipose-­derived mesenchymal stromal cells from patients with 6 Sautereau N, Daumas A, Truillet R, et al. Efficacy of autologous Microfat graft on facial systematic sclerosis. Cell Transplant 2017;26:841–54. handicap in systemic sclerosis patients. Plast Reconstr Surg Glob Open 2016;4. 12 Hirose Y, Funahashi Y, Matsukawa Y, et al. Comparison of trophic factors secreted from 7 Del Papa N, Di Luca G, Sambataro D, et al. Regional implantation of autologous human adipose-­derived stromal vascular fraction with those from adipose-­derived adipose tissue-­derived cells induces a prompt healing of long-­lasting indolent digital stromal/stem cells in the same individuals. Cytotherapy 2018;20:589–91. ulcers in patients with systemic sclerosis. Cell Transplant 2015;24:2297–305. 13 Granel B, Daumas A, Jouve E, et al. Safety, tolerability and potential efficacy of 8 Benedetto PD, Liakouli V, Ruscitti P, et al. Blocking CD248 molecules in perivascular stromal cells of patients with systemic sclerosis strongly inhibits their differentiation injection of autologous adipose-­derived stromal vascular fraction in the fingers toward myofibroblasts and proliferation: a new potential target for antifibrotic of patients with systemic sclerosis: an open-­label phase I trial. Ann Rheum Dis therapy. Arthritis Research 2018;20. 2015;74:2175–82.

2 of 2 Ann Rheum Dis May 2020 Vol 79 No 5 Correspondence MYCYC: unravelling the long road ahead in Medicale in 2013, the non-­inferiority endpoint was not met.4 The primary remission endpoint occurred in 46/70 (66%) ANCA-associated v­ asculitis MMF vs 48/70 (69%) CYC, with the risk difference being −3% (90% CI −16% to 10%), leading to a p value of 0.06 for We read with great interest the article on ‘Mycophenolate non-­inferiority. mofetil versus cyclophosphamide for remission induction in It would be interesting to know the indication for prerando- ANCA-­associated vasculitis: a randomized, non-­inferiority 1 misation treatment with plasmapheresis or pulse methylpred- trial’ by Jones et al. The trials of mycophenolate mofetil nisolone. With 24% patients in the MMF arm receiving one (MMF) for induction in antineutrophil cytoplasmic antibody dose of pulse CYC, 59% receiving intravenous pulse meth- (ANCA)-­associated vasculitis (AAV) done previously were ylprednisolone and 11% receiving plasma exchange before single-­centre studies mostly restricted to patients with micro- randomisation, the efficacy estimates of MMF, and the time scopic polyangiitis (MPA), predominantly from South East 2 3 taken for remission are prone to overestimation, although the Asia. We commend the authors for including a large popu- prerandomisation treatment did not differ between the two lation of patients with AAV of different age groups, with true arms representation of both granulomatosis with polyangiitis, and The steroid dosage protocol followed in the MYCYC mimics patients with MPA from different countries, thus overcoming the current standards of practice in AAV. Higher remission rates the limitations of previous studies and possibly bringing us at 6 months irrespective of steroid compliance compared with closer to the true estimate of efficacy of MMF as an induction those with steroid compliance, along with comparable remission agent in AAV. However, certain aspects of this study require rates of most induction agents in AAV, make one wonder if the clarification and discussion. efficacy estimates of second induction agents in AAV in MYCYC First, median baseline estimated glomerular filtration rate 2 and other induction trials like CYCLOPS, RAVE, RITUXVAS (eGFR) of 51 mL/min/m of the trial patients did not improve or NORAM have been confounded by the use of background considerably at 6 or 18 months in either of the two arms (online steroids. Although the preliminary results of the PEXIVAS have supplementary figure 1B of the paper). Non-­reversibility of found reduced dose (<60%) glucocorticoids (GC) to be non-­ renal dysfunction with immunosuppressive therapy raises inferior to standard dose GC in terms of efficacy,5 the exact a suspicion of a component of chronicity to the underlying contribution of steroids in the currently used steroid-­based renal disease. Since patients with recent decline (eGFR fall combination induction regimens in AAV should be clearer with >20%) in renal function in previous 2 weeks were excluded, the results of low-­steroid induction trials in AAV like the LoVAS.6 the baseline renal dysfunction at entry could be reflective of With the documented efficacy of MMF in advanced prolifera- previously unrecognised or relapsing/refractory renal disease. tive lupus nephritis and the inclusion of a small subset of patients Thus, although newly diagnosed (within 6 months) patients with AAV with severe renal involvement in the study by Han et were included in the trial, it might be prudent to examine the al,3 comparison of MMF versus CYC in this cohort of patients duration of disease (taken from onset of first AAV- related represents an unmet need that needs to be studied in systemati- symptom) at the time of recruitment. Were the recently diag- cally designed prospective studies. nosed patients with AAV included in the study recruited at first Finally, with a statistically significant increase in relapse rates presentation or at a relapse? in the MMF arm after 18 months of follow-up­ in the current It would be interesting to see the predictors of relapse in the study, it would be interesting to see if the long-term­ outcome study, and whether there was any correlation with the dose of of this cohort of patients parallels the NORAM trial.7 8 With MMF used for induction or the dose of steroids at the time of the still uncertain way of handling steroids during induction and relapse. Was there any incremental benefit of increasing the maintenance of AAV and the expanding therapeutic armamen- MMF dose to 3 g/day in terms of remission and subsequent tarium for AAV, these certainly are challenging although exciting relapse rates? Did the 18% patients who received MMF <2 times in the management of AAV. g/day relapse more? In today’s world where we are gradu- ally attempting a transition towards low or no steroid use in Siddharth Jain, Arghya Chattopadhyay ‍ ‍ , Shankar Naidu, AAV, knowledge of a correlation (or a lack thereof) between Aman Sharma ‍ ‍ steroid dose and risk of relapse would be quintessential. Since Clinical Immunology and Rheumatology Services, Department of Internal Medicine, a sizeable number (26) of patients in each arm were not on Post Graduate Institute of Medical Education and Research, Chandigarh, India maintenance azathioprine by the end of 18 months, did this Correspondence to Dr Aman Sharma, Department of Internal Medicine, Post correlate with an increased risk of relapse? With the post hoc Graduate Institute of Medical Education and Research, Chandigarh 160012, India; subgroup analysis showing that most relapses in the MMF arm amansharma74@​ ​yahoo.com​ occurred in patients with PR3-­ANCA, further clarification Handling editor Josef S Smolen on the subsequent comment by the authors, ‘There was no evidence that the effect of MMF on relapse differed by ANCA Contributors All contributed equally. subtype (p=0.52 for interaction)’ may be helpful. Competing interests None declared. We could not help but notice a few factual errors and text-­ table discrepancies. Two withdrawals or loss to follow-up­ by Patient consent for publication Not required. 6 months in the MMF group (figure 1 of the paper) versus Provenance and peer review Not commissioned; internally peer reviewed. three in the text. The p value for the incidence rate ratio (IRR) © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and for relapse rates in MMF versus cyclophosphamide (CYC) is permissions. Published by BMJ. <0.05 even though the lower limit of 95% CI crosses 1 (IRR 1.97, 95% CI 0.96 to 4.23, p=0.049). Interestingly, as per the results of the primary efficacy analysis in the same data To cite Jain S, Chattopadhyay A, Naidu S, et al. Ann Rheum Dis 2020;79:e57. set presented at the 16th International Vasculitis and ANCA workshop and published in abstract form in the La Presse Received 14 February 2019

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 2 Correspondence

Accepted 17 February 2019 3 Han F, Liu G, Zhang X, et al. Effects of mycophenolate mofetil combined with Published Online First 20 March 2019 corticosteroids for induction therapy of microscopic polyangiitis. Am J Nephrol 2011;33:185–92. 4 Jones R, Harper L, Ballarin J, et al. A randomized trial of mycophenolate mofetil versus cyclophosphamide for remission induction of ANCA-­associated vasculitis: “MYCYC”. On behalf of the European vasculitis study group. La Presse Médicale 2013;42:678–9. ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-215254​ 5 Walsh M, Merkel P, Peh CA, et al. LB02 the effect of reduced-­dose glucocorticoids Ann Rheum Dis 2020;79:e57. doi:10.1136/annrheumdis-2019-215241 during remission induction in severe ANCA associated vasculitis. Nephrology Dialysis Transplantation 2018;33(Suppl 1). ORCID iDs 6 Furuta S, Sugiyama T, Umibe T, et al. Low-­dose glucocorticoids plus rituximab versus Arghya Chattopadhyay http://orcid.​ ​org/0000-​ ​0003-​3917-​1977 high-­dose glucocorticoids plus rituximab for remission induction in ANCA-­associated Aman Sharma http://orcid.​ ​org/0000-​ ​0003-​0813-​1243 vasculitis (LoVAS): protocol for a multicentre, open-­label, randomised controlled trial. BMJ Open 2017;7:e018748. References 7 De Groot K, Rasmussen N, Bacon PA, et al. Randomized trial of cyclophosphamide 1 Jones RB, Hiemstra TF, Ballarin J, et al. Mycophenolate mofetil versus cyclophosphamide versus methotrexate for induction of remission in early systemic antineutrophil for remission induction in ANCA-­associated vasculitis: a randomised, non-­inferiority cytoplasmic antibody-­associated vasculitis. Arthritis Rheum 2005;52:2461–9. trial. Ann Rheum Dis 2019;78:399–405. 8 Faurschou M, Westman K, Rasmussen N, et al. Brief report: long-­term outcome of a 2 Hu W, Liu C, Xie H, et al. Mycophenolate mofetil versus cyclophosphamide for inducing randomized clinical trial comparing methotrexate to cyclophosphamide for remission remission of ANCA vasculitis with moderate renal involvement. Nephrol Dial Transplant induction in early systemic antineutrophil cytoplasmic antibody-­associated vasculitis. 2008;23:1307–12. Arthritis Rheum 2012;64:3472–7.

2 of 2 Ann Rheum Dis May 2020 Vol 79 No 5 Correspondence response Response to: ‘MYCYC, unravelling the long road statistical difference in relapse rates between MMF and CYC in Myeloperoxidase (MPO)-ANCA­ disease, the increase in ahead in ANCA-­associated vasculitis’ by Jain relapse rate was accounted for by patients with PR3-­ANCA et al positivity. However, we have not included this statistical data in the manuscript as the study was underpowered to test this In their letter to the editor, Jain et al raised some issues regarding hypothesis. We are concerned about an overinterpretation of our recent paper. We appreciate their interest in our study and our data if this is stated. We, therefore, tested for an interac- their positive comments.1 tion of MMF and disease type and could show no evidence Jain et al highlight the fact that the presenting glomerular that the effect of MMF on relapse differed by ANCA subtype, filtration rate (GFR) did not improve considerably during this is included in the manuscript. the study and were concerned that this raised the suspicion Jain et al were concerned that the p value for the incidence that there was a component of chronicity associated with the risk ratio (IRR) for relapse rates in MMF versus CYC is <0.05 underlying disease. This is unlikely; at study entry it was an even though the lower limit of 95% confidence intervals (CI) inclusion criteria that all patients had newly diagnosed disease, crosses 1 (IRR 1.97, 95% CI 0.96 to 4.23, p=0.049). This is not relapsing disease and had not received either rituximab not a factual error as Jain et al suggest. The CI and p value are therapy or more than one intravenous pulse of cyclophos- derived from different models, so it is possible that these disagree phamide or more than 2 weeks daily oral cyclophosphamide. particularly in a ‘borderline’ case like this. Relapse rate was a Renal vasculitis can present in a number of different ways; secondary outcome in this study which was not powered to give from microscopic haematuria with a normal renal function to robust evidence on relapse. The abstract published in 2013 was the syndrome of rapidly progressive glomerulonephritis neces- based on preliminary findings and was submitted before final sitating dialysis. The stability of GFR over the course of the adjudication of the data and analysis. The final analysis of the study reflects the facts that GFR at entry was relatively high study is as published in Annals of Rheumatic Diseases, and the (51 mL/min), that patients with a clinical syndrome of rapidly study met the primary endpoint as stated. progressive glomerulonephritis were excluded (defined as a The indication for additional treatments intravenous meth- 20% fall in GFR over a 2 weeks), that patients with a GFR ylprednisolone and plasma exchange or initiation of oral <30 mL/min were excluded and that active renal vascu- steroids, and cyclo in the 2 weeks before study entry were litis was not a requirement for entry. Eighty-­one per cent of not recorded, however, these treatments were permitted prior patients had active renal vasculitis at study entry and the lack to enrolment as these are standard treatments initiated for 3 of change in GFR is very likely a reflection of less severe renal severe disease and reflect usual clinical practice. Exclusion of disease rather than more chronic renal disease. Change in GFR patients receiving these treatments would have restricted the demonstrated includes all patients from the study and is not study population to those with milder disease, thus limiting a subgroup analysis of patients with active renal vasculitis at generalisability of results. Recognising that intravenous MP entry.2 However, we did not collect data on time from diag- or plasma exchange may affect short-­term outcomes, use of nosis to inclusion in the study and we have not analysed renal these additional treatments was a stratification factor in the biopsies in this study to determine chronicity of the lesions randomisation process to ensure groups were balanced in this present in those who had a renal biopsy performed. regard. Given this balance it is expected that any overestima- The indication for additional treatments intravenous meth- tion of time to remission is likely to have affected both treat- ylprednisolone (MP) and plasma exchange or initiation of oral ment groups similarly. steroids, and cyclophosphamide (CYC) in the 2 weeks before We agree that the steroid dosing in MYCYC may have contrib- study entry were not recorded; however, these treatments were uted to remission and relapse outcomes, and for this reason permitted prior to enrolment as these are standard treatments was standardised across treatment groups and compliance with initiated for severe disease and reflect usual clinical practice.3 steroid taper was a component of the primary remission outcome Exclusion of patients receiving these treatments would have definition. We also agree that remission efficacy observed with restricted the study population to those with milder disease, MMF and steroids in this study should not be interpreted as thus limiting generalisability of results. Recognising that MMF efficacy alone or with lower steroid regimen without intravenous MP or plasma exchange may affect short-­term further randomised data. outcomes, use of these additional treatments was a stratifica- We agree with Jain et al that there remain many questions tion factor in the randomisation process to ensure groups were about the management of ANCA associated vasculitis (AAV) with regard to steroids and maintenance therapy but these are balanced in this regard. out with the realms of this publication. Jain et al were interested in predictors of relapse but the study was an induction trial and not designed to investi- Lorraine Harper ‍ ‍ ,1 Rachel B Jones2 gate relapse of disease. The study was not powered for this 1Institute of Clinical Sciences, University of Birmingham, Birmingham, UK endpoint and follow-­up duration was short, with only 18 2Department of Renal Medicine, Addenbrooke’s Hospital, Cambridge, UK months follow-­up. We do not feel it appropriate to report risk Correspondence to Lorraine Harper, University of Birmingham, Birmingham B15 factors of relapse in this manuscript. Jain et al are concerned 2TT, UK; l​ .​harper@bham.​ ​ac.​uk that mycophenolate mofetil (MMF) dosing may have impacted Handling editor Josef S Smolen on treatment efficacy and risk of relapse, however, a number Contributors LH and RBJ both wrote and approved the response. of patients receiving different doses of MMF are too small to investigate this; only 12 patients received <2 g and 4 patients 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. received >2 g. Clarification was requested on relapses relating to patients Competing interests None declared. with Proteinase 3-anti-­ ­neutrophil cytoplasm antibody (PR3-­ Patient consent for publication Not required. ANCA)-­associated disease. We were unable to detect a Provenance and peer review Commissioned; internally peer reviewed.

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 2 Correspondence response

© Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and ►► http://​dx.doi.​ ​org/10.​ ​1136/annrheumdis-​ ​2019-​215241 permissions. Published by BMJ. Ann Rheum Dis 2020;79:e58. doi:10.1136/annrheumdis-2019-215254

ORCID iD Lorraine Harper http://orcid.​ ​org/0000-​ ​0003-​1343-​9234 To cite Harper L, Jones RB. Ann Rheum Dis 2020;79:e58. References Received 27 February 2019 1 Jain S, Chattopadhyay A, Naidu G, et al. MYCYC: unravelling the long road ahead in Accepted 28 February 2019 ANCA-associated­ vasculitis. Annals Rheum Dis 2020;79:e57. Published Online First 20 March 2019 2 Jones RB, Hiemstra TF, Ballarin J, et al. Mycophenolate mofetil versus cyclophosphamide for remission induction in ANCA-­associated vasculitis: a randomised, non-­inferiority trial. Ann Rheum Dis 2019;78:399–405. 3 Yates M, Watts RA, Bajema IM, et al. EULAR/ERA-­EDTA recommendations for the management of ANCA-­associated vasculitis. Ann Rheum Dis 2016;75:1583–94.

2 of 2 Ann Rheum Dis May 2020 Vol 79 No 5 Correspondence

Changing landscape of immunosuppression in Table 1 Possible tailored regimens of remission induction treatment ANCA-associated v­ asculitis in patients with AAV Non-­severe AAV Severe AAV Over last decades, there have been substantial advances in the Proteinase-3 ANCA Methotrexate* or rituximab Rituximab† or immunosuppressive treatment for antineutrophil cytoplasmic cyclophosphamide antibody (ANCA)-­associated vasculitides (AAV), that is, granu- MPO-ANCA­ Methotrexate* or MMF or Cyclophosphamide or lomatosis with polyangiitis (GPA) and microscopic polyangiitis rituximab rituximab† or MMF† (MPA). The recently published MYCYC study added another ANCA-negative­ Methotrexate* or Cyclophosphamide piece of evidence to our knowledge of treatment options in cyclophosphamide 1 patients with AAV. In an open-­label, two-group,­ parallel-­ Immunosuppressive medications should be used in combination with design, randomised trial, 140 newly diagnosed patients with glucocorticoids (at least 40 mg daily of prednisone in non-severe­ AAV and at least 60 mg daily of prednisone in severe AAV). active GPA or MPA were allocated to remission induction with 2 mycophenolate mofetil (MMF) or pulsed cyclophosphamide *Only in patients with estimated glomerular filtration rate >30 mL/min/1.73 m . †Preferably in younger patients as well as in patients with relapsing disease. in combination with the same oral glucocorticoid regimen. AAV, ANCA-­associated vasculitides; ANCA, antineutrophil cytoplasmic antibody; Non-­inferiority was demonstrated for the primary remission MMF, mycophenolate mofetil; MPO, myeloperoxidase. endpoint, which occurred by 6 months in 67% of patients in the MMF group and 61% of patients in the cyclophosphamide would be even less effective and should be avoided. In the group (risk difference 5.7%, 90% CI −7.5% to 19%). The IMPROVE trial, MMF was also inferior to azathioprine for median time to remission was also similar in the MMF and 4 maintaining disease remission in 156 patients with AAV. cyclophosphamide groups. However, treatment with MMF was Currently, rituximab is increasingly used for initial induc- associated with a higher occurrence of relapses compared with tion remission treatment in patients with both de novo and cyclophosphamide (33% vs 19%, p=0.049). As expected, the relapsing AAV. Moreover, in the USA, rituximab label has higher relapse rate following remission induction with MMF been recently updated to include information for mainte- was accounted for by more frequent relapses in patients with nance therapy in adult patients with GPA and MPA who have proteinase-3 ANCA, but not patients with myeloperoxidase achieved remission with induction treatment. The apparent (MPO) ANCA. The authors suggested that MMF induction benefits of rituximab question the future of MMF in patients therapy might be a suitable alternative to cyclophosphamide in with AAV. High cost of rituximab is one of the main factors patients at low risk of relapse, such as those with MPO-ANCA.­ limiting its use in AAV in many countries. The independent Unlike methotrexate, MMF can be used in patients with Evidence Review Group commissioned by the National Insti- moderate or severe renal disease. Moreover, remission induc- tute for Health and Care Excellence (NICE) has recently tion treatment with MMF was similar in efficacy to pulsed 2 concluded that rituximab was likely to represent a cost-­ cyclophosphamide in patients with lupus nephritis. It seems effective addition to the treatment sequence if given after that MPO-­ANCA associated renal vasculitis may be a primary cyclophosphamide treatment.5 In Russia, there is almost a indication for administration of MMF. At entry into MYCYC 7-­fold difference between the acquisition costs for rituximab study, 81% of patients in the MMF and cyclophospha- and MMF. However, the emergence of less expensive (by up mide groups presented with proteinuria, haematuria and/or to 30%) biosimilars of rituximab improved patients’ access to impaired kidney function. Only two patients in both groups modern care. Accumulating evidence suggest that a proportion progressed to end stage renal disease, while average estimated of patients with AAV are overtreated with rituximab and may glomerular filtration rate at 18 months did not differ between benefit from safer and more cost-­effective low-­dose regimens groups. of rituximab administration.6 Unlike cyclophosphamide, MMF does not induce infertility In summary, the results of MYCYC study and other that causes particular concerns in younger patients with AAV. randomised controlled trials pave the way for precision medi- Therefore, adolescents and young adults should also be consid- cine in AAV. We agree with the authors that stratified treat- ered as candidates for MMF induction regimen. MYCYC ment approaches are indicated in order to optimise long-­term study was the first randomised trial in AAV to include paedi- outcomes. The choice of treatment depends on many factors, atric patients (n=8). The response rates were similar in the including age of patient, severity of AAV, ANCA specificity, MMF and cyclophosphamide groups in children. However, renal function, the presence of granulomatous lesions and so the small number of paediatric participants precludes any on (table 1). Emerging treatment options, that is, complement definite conclusions regarding the efficacy or safety of MMF inhibitors (avacopan, IFX-1) may result in implementation in this population. Of note, an advantage of MMF in safety of safer, steroid-free­ remission induction and maintenance over cyclophosphamide should not be overestimated given the regimens. similar rates of adverse events, including serious infections, in the two groups. Over 18 months of follow-up,­ malignan- Sergey V Moiseev ‍ ‍ ,1 Ilya Smitienko,2 Nikolai Bulanov,1 1 1 cies were reported in one patient from MMF group and one Ekaterina Karovaikina, Pavel I Novikov patient from cyclophosphamide group. 1Department of Rheumatology, Tareev Clinic of Internal Diseases, Sechenov First Only patients with newly diagnosed AAV participated in the Moscow State Medical University, Moscow, Russian Federation 2Medical Center K+31, Moscow, Russian Federation MYCYC study. Therefore, the efficacy of MMF as compared with cyclophosphamide in patients with relapsing AAV is not Correspondence to Professor Sergey V Moiseev, Tareev Clinic of Internal Diseases, established. In RAVE trial, conventional immunosuppression Sechenov First Moscow State Medical University, Moscow 119435, Russian Federation; clinpharm@​ ​mtu-net.​ ​ru with cyclophosphamide followed by azathioprine was inferior to rituximab at 6 and 12 months in 101 patients who had Correction notice This article has been corrected since it published Online First. Table 1 has been corrected. relapsing disease at baseline.3 We can speculate that in patients with relapsing AAV remission induction treatment with MMF Handling editor Josef S Smolen

Ann Rheum Dis May 2020 Vol 79 No 5 1 of 2 Correspondence

Contributors All authors participated in the preparation of the manuscript. ORCID iD Sergey V Moiseev http://orcid.​ ​org/0000-​ ​0002-​7232-​4640 Funding Russian Academic Excellence Project 5-100. Competing interests None declared. References Patient consent for publication Not required. 1 Jones RB, Hiemstra TF, Ballarin J, et al. Mycophenolate mofetil versus cyclophosphamide for remission induction in ANCA-­associated vasculitis: a randomised, non-­inferiority Provenance and peer review Not commissioned; internally peer reviewed. trial. Ann Rheum Dis 2019;78:399–405. © Author(s) (or their employer(s)) 2020. No commercial re-­use. See rights and 2 Ginzler EM, Dooley MA, Aranow C, et al. Mycophenolate mofetil or intravenous permissions. Published by BMJ. cyclophosphamide for lupus nephritis. N Engl J Med 2005;353:2219–28. 3 Specks U, Merkel PA, Seo P, et al. Efficacy of remission-­induction regimens for ANCA-­ associated vasculitis. N Engl J Med 2013;369:417–27. 4 Hiemstra TF, Walsh M, Mahr A, et al. Mycophenolate mofetil vs azathioprine for To cite Moiseev SV, Smitienko I, Bulanov N, et al. Ann Rheum Dis 2020;79:e59. remission maintenance in antineutrophil cytoplasmic antibody-­associated vasculitis: a randomized controlled trial. JAMA 2010;304:2381–8. Received 24 January 2019 5 Latimer NR, Carroll C, Wong R, et al. Rituximab in combination with corticosteroids Accepted 26 January 2019 for the treatment of anti-­neutrophil cytoplasmic antibody-­associated vasculitis: a NICE single technology appraisal. Pharmacoeconomics 2014;32:1171–83. Published Online First 12 February 2019 6 Moiseev SV, Bulanov NM, Zykova AS, et al. Rituximab in ANCA-­associated vasculitis: Ann Rheum Dis 2020;79:e59. doi:10.1136/annrheumdis-2019-215123 Fewer infusions or ultra low-­dose maintenance therapy? Ann Rheum Dis 2019;78:e99.

2 of 2 Ann Rheum Dis May 2020 Vol 79 No 5