University of Groningen Effectiveness and Safety of Medicines Used in COPD Patients Wang, Yuanyuan

Home , Tetracycline antibiotics, Uricosuric

University of Groningen

Effectiveness and safety of medicines used in COPD patients Wang, Yuanyuan

DOI: 10.33612/diss.123921981

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2020

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Wang, Y. (2020). Effectiveness and safety of medicines used in COPD patients: pharmacoepidemiological studies. University of Groningen. https://doi.org/10.33612/diss.123921981

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license. More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne- amendment.

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 09-10-2021 Efectiveness and safety of medicines used in COPD patients

Pharmacoepidemiological studies

Yuanyuan Wang Paranymphs: Jurjen van der Schans Yanni Li

Efectiveness and safety of medicines used in COPD patients Pharmacoepidemiological studies

ISBN: 978-94-034-2555-9 (printed version) ISBN: 978-94-034-2554-2 (electric version)

Author: Yuanyuan Wang Cover-design: IRINA SHI (photo) & Of Page (content) Printing: Of Page (www.ofpage.nl)

The studies presented in this thesis were funded by University of Groningen and the China Scholarship Council (CSC) Scholarship. This thesis was conducted within the Groningen University Institute for Drug Exploration (GUIDE). Printing of this thesis was fnancially supported by the University of Groningen and the Graduate School of Science and Engineering (GSSE).

Copyright ©Yuanyuan Wang, 2020, Groningen, The Netherlands. All rights reserved. No part of this book may be reproduced or transmitted in any form by any means, electronically or mechanically by photocopying, recording, or otherwise, without the written permission of the author. The copy right of previously published chapters of this thesis remains with the publisher or journal. Effectiveness and safety of medicines used in COPD patients

Pharmacoepidemiological studies

PhD thesis

to obtain the degree of PhD at the University of Groningen on the authority of the Rector Magnificus Prof. C. Wijmenga and in accordance with the decision by the College of Deans.

This thesis will be defended in public on

Friday 8 May 2020 at 14.30 hours

Yuanyuan Wang

born on 15 February 1988 in Henan, China Supervisors Prof. . Ha Prof. H.M. oeen

Assessment Committee Prof. T..M. erheij Prof. . van der Palen Prof. . tienstra Supervisors TABLE OF CONTENTS Prof. . Ha Prof. H.M. oeen Chapter 1 General Introduction 7

Part I Efects of antibiotic use for COPD exacerbations and potential DDIs during COPD exacerbation management Assessment Committee Prof. T..M. erheij Chapter 2 Efects of Prophylactic Antibiotics on Patients with Stable COPD: 21 Prof. . van der Palen A Systematic Review and Meta-Analysis of Prof. . tienstra Randomized Controlled Trials

Chapter 3 The infuence of age on real-life efects of doxycycline for 43 acute exacerbations among COPD outpatients: a population-based cohort study

Chapter 4 Real-world short- and long-term efects of antibiotic therapy on 61 acute exacerbations of COPD in outpatients: a cohort study under the PharmLines Initiative

Chapter 5 Improving antibacterial prescribing safety in 81 the management of COPD exacerbations: systematic review of observational and clinical studies on potential drug interactions associated with frequently prescribed antibacterials among COPD Patients

Part II Neuropsychiatric safety of varenicline use for smoking cessation and the application of prescription sequence symmetry analysis in drug safety evaluation

Chapter 6 Neuropsychiatric safety of varenicline in the general and 119 COPD population with and without psychiatric disorders: a retrospective inception cohort study in a real-world setting

Chapter 7 Risk of neuropsychiatric adverse events associated with 145 varenicline treatment for smoking cessation: a prescription sequence symmetry analysis

Chapter 8 Efect estimate comparison between the prescription 165 sequence symmetry analysis and parallel group study designs: a systematic review

Chapter 9 General discussion 183

Chapter 10 Summary 197 Samenvatting 201 Acknowledgement 204 List of publications 208 About the author 211 CHAPTER 1 General Introduction

General Introduction

GENERAL INTRODUCTION 1 Chronic obstructive pulmonary disease According to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) report 2020, Chronic Obstructive Pulmonary Disease (COPD) is a common, preventable and treatable disease that is characterized by persistent respiratory symptoms and airfow limitation. Such symptoms are caused by airway and/or alveolar abnormalities usually triggered by signifcant exposure to noxious particles or gases.1 Cigarette smoking is the main independent causal risk factor for COPD with indoor and outdoor air pollution and occupational exposure to dust and noxious particles also being the risk factors for COPD.2 Moreover, host factors that may contribute to the development of COPD include age, genetics, airway hyper responsiveness and abnormal lung development.3-5

The prevalence of COPD varies across countries as well as regions within countries. According to the fndings of a global meta-analysis, the number of COPD cases increased to 384 million in 2010, with a global prevalence of 12% (ranging between 8% and 15%).6 COPD is commonly diagnosed in individuals aged 40 years or older based on the presence of associated symptoms and risk factors. However, a defnitive COPD diagnosis requires the performance of spirometry. The presence of a post- bronchodilator FEV1/FVC < 0.7 confrms the presence of airfow limitation and results in a COPD diagnosis. COPD patients are currently categorized into four GOLD stages of severity of airfow limitations based on the predicted value of FEV1: mild (stage I, FEV1

≥ 80% predicted), moderate (stage II, 50% ≤ FEV1 < 80% predicted), severe (stage III, 1 30% ≤ FEV1 < 80% predicted) and very severe (stage IV, FEV1 < 30% predicted). A large prevalence study estimated that the rate of COPD for GOLD stage II and higher is around 10% in the general population, and a little higher in men than women (11.8% and 8.5%, respectively).7

COPD is one of the leading causes of morbidity and mortality worldwide.8 Its burden is predicted to increase in the coming decades as a result of continuous exposure to risk factors in developing countries and aging of the population worldwide, particularly in high-income countries.9 Smoking cessation interventions, increased physical activity, and early diagnosis and treatment of related comorbidities are considered key measures for reducing the health-economic burden of COPD. 10

Up to now, the key goals of COPD treatment have been to improve patients’ prognosis and prevent the disease from worsening. The main treatment used in the daily management of mild and moderate COPD is pharmacotherapy. Bronchodilators, including short- and long-acting β2-agonists (SABA and LABA) and short- and long-acting anticholinergics (SAMA and LAMA) are essential for managing and preventing symptoms, and combined treatment (SABA/SAMA, LABA/LAMA or LABA/ICS) may be used as appropriate. Non-

9 Chapter 1 pharmacological treatment comprises pulmonary rehabilitation (e.g., exercise training, education, and behavioral change).11 Oxygen therapy is necessary for patients with very severe COPD and lung surgery may also be necessary.

Exacerbations of COPD and antibiotic use An exacerbation of COPD is defned as an acute worsening of respiratory symptoms that necessitates additional therapy.1 COPD patients can periodically experience acute exacerbations that may accelerate the decline in lung function, reduce the quality of life, and increase mortality and health-care costs.12,13 14 Infections, especially bacterial infections, and infammation are thought to be an important trigger for exacerbations of COPD. Previous studies have found that bacteria are responsible for around 40% to 50% of exacerbations.15,16 S. pneumoniae, H. infuenza, P. aeruginosa, M. catarrhalis, A. baumannii, and S. aureus were the most frequently reported bacteria that cause exacerbations of COPD.16-18 According to the GOLD guideline, the main goals in treatment of COPD exacerbations are minimizing the negative impact of the current exacerbation and preventing subsequent exacerbations.1 Because almost 40% of exacerbations are bacteria-caused respiratory tract infections,16 the use of antibiotics has become a common component in the management of acute exacerbations among COPD patients, both in terms of treatment and prevention.1,19

Notably, recommendations in prophylactic use of antibiotics in the management of COPD exacerbations are conditional and unspecifc. Only long-term macrolides are currently mentioned as frst-line therapy.1,19 Moreover, in terms of current evidence, an optimal regimen of prophylactic antibiotics for exacerbations has not been well established, and related recommendations regarding an appropriate schedule (continuous vs. intermittent) and the duration of a specifc antibiotic intervention (below or equal to 6 months vs. above 6 months) are still lacking. Besides, the efects of even the most extensively researched antibiotics macrolides—let alone other potentially suitable antibiotics—on the time to the frst exacerbation, changes in lung function, the bacterial load, and airway infammation have not been adequately evaluated.20 Knowledge of these outcomes is also vital for elucidating possible mechanisms behind the reduction of exacerbations through the prophylactic use of antibiotics, and for weighing benefts and risks.21

The benefcial efects of oral corticosteroids as an efective treatment for acute exacerbations of COPD (AECOPD) in improving COPD symptoms and lung function are well established.22 However, although antibiotics have been recommended for the treatment of AECOPOD when signs of bacterial infection are present,1 there is still uncertainty regarding the benefcial efects of antibiotic treatment used in the combination with oral glucocorticoids for AECOPD, particularly in the case of outpatients in real-world settings. In 2012, a Cochrane review that pooled the results

10 General Introduction of fve RCTs conducted among outpatients did not reveal a signifcantly reduced risk of treatment failure associated with antibiotics currently prescribed for outpatients.23 1 However, an updated version of this Cochrane review conducted in 2018 presented statistically signifcant benefcial efects of current antibiotics prescribed for outpatients.24 Two new RCTs were included in this later study in relation to the earlier review conducted in 2012.25,26 Notably, one of the two RCTs did not support the benefcial efects of antibiotics treatment on AECOPD, although it did contribute to almost 25% of the sample size of the updated pooled results.25

Hence, most of the available scientifc evidence on the efects of antibiotics for AECOPD is basically derived from RCTs. It is widely accepted that RCTs provide solid evidence with high internal validity, however, their generalizability in the real world, especially in outpatient settings is low. COPD is a chronic disease and is mostly managed on an outpatient basis within a population that is more heterogeneous compared with populations from RCTs. Moreover, the use of antibiotics for AECOPD treatment is not always appropriate and in line with related guidelines.27,28 Therefore, the efect of antibiotic treatment for AECOPD in real-world settings may difer from those obtained in clinical trials and require further investigation.

Comorbidities of COPD and potential drug-drug interactions COPD is a chronic disease and its prevalence increases with age; around 15% of the general population over 65 years is afected by COPD.29 Hence, age-related comorbidities frequently co-exist with COPD.30 The most common concomitant chronic conditions associated with COPD include cardiovascular disease (e.g., heart failure, ischemic heart disease, and arrhythmias), metabolic disease (e.g., diabetes), osteoporosis, depression and anxiety, lung cancer and gastroesophageal refux (GERD).1,30

COPD itself is a complex disease that entails the need for a variety of medications to improve lung function and treat exacerbations.1 Multiple comorbidities further complicate the medical management of COPD, resulting in polypharmacy among a large section of COPD patients. Polypharmacy poses a potential risk of drug-drug interactions (DDIs) that may induce adverse events and treatment failures. Moreover, COPD is an age-related disease that generally manifests at an older age. Therefore, these older patients are more susceptible to DDIs due to gradual physiologic negative changes that may infuence their pharmacokinetics and the pharmacodynamics of the drugs used.31

As most evidence about drug efects is from clinical trials, more attention should be paid to issues related to polypharmacy and to potential DDIs in the management of COPD in real-world settings.32 This is especially the case for antibiotic therapy as it includes diferent drug classes that vary in their mechanisms relating to absorption and metabolism, making their interaction with other medications more likely. Comprehensive information for clinicians to avoid potential DDIs, however, is lacking.

11 Chapter 1

Smoking cessation drug therapy and neuropsychiatric safety Tobacco smoking is the main risk factor for COPD and other physical and mental disorders.33-35 This preventable behavior poses huge threats to global public health.36,37 Although in recent years, strict tobacco control policies have prompted a global decline in smoking,36 the actual numbers of smokers and smoking-related disease burden continues to increase because of the growing population worldwide.38 More than eight million people continue to die annually as a result of tobacco consumption.39

Therefore, smoking cessation strategies to prevent smoking-related diseases are imperative.40 Varenicline, which was the frst non-nicotine, pharmacotherapeutic, smoking cessation product, has been found to be more efcacious than other therapies, such as single-dose bupropion and nicotine replacement therapy (NRT).41 However, following varenicline’s approval by the FDA in 2006, safety concerns were raised relating to its neuropsychiatric adverse events, which include suicidal thoughts, aggressive behavior, depression, anxiety, and sleep disorders.42 Numerous RCTs were subsequently conducted with varenicline to generate evidence on its safety.43 In light of their fndings, the FDA warning was removed in 2016. However, concerns remain, given the strict inclusion and exclusion criteria applied in RCTs that result in the participation of relatively healthy individuals and the lack of consistent real-world evidence. Notably, special risk populations demonstrating increased smoking prevalence, such as COPD patients, have generally been excluded from RCTs.44

As previously noted, most COPD patients are elderly and have multi-morbidities, making them more susceptible to adverse drug events (ADEs). Similarly, there is evidence that individuals with psychiatric disorders experience relapses of psychiatric symptoms more frequently than those without these disorders.45,46 The safety of varenicline use for these specifc populations has not been established. Although a few studies were conducted among patients with COPD or psychiatric disorders,47,48 the results were inconsistent. Consequently, more observational studies are still needed to generate the real-world evidence relating to the safety of varenicline use.

PSSA and observational study designs in drug safety evaluation Most evidence regarding the efects of drugs is derived from strictly regulated clinical trials. However, the results from RCTs may not refect the real-world situations, given that the participants are relatively healthy and because of the limited scope for detecting rare events with clinical trials. Therefore, real-world evidence derived from traditional, non-randomized, observational study designs is valuable for exploring such drug efects or toxicities within the feld of pharmacoepidemiology. However, the evidence from observational studies is often inconsistent, and such designs have been criticized for their potential of bias (e.g., selection or information bias) and confounding (e.g. unmeasured confounding).49

12 General Introduction

Prescription sequence symmetry analysis (PSSA) is increasingly being used to detect adverse efects or events associated with medications. PSSA is a self-controlled study 1 design in which genetic and other time-invariant confounding can be well controlled, it does not entail the abovementioned bias.50,51 It compares the symmetry in the sequence of exposure medication and marker (outcome) medications as proxy for ADRs within a specifc time window based on prescriptions or claims databases.52 The sequence ratio (SR) refects the association between exposure and outcome. However, PSSA is still sensitive to time-varying variables, notably if the follow-up time is long. The overall validity of PSSA study designs has not been fully evaluated by comparing its result with those from conventional observational parallel group study designs, and such comparisons are urgently required.

AIM OF THIS THESIS

In this thesis, we aim to develop a comprehensive profle on the efectiveness of antibiotic use for acute exacerbations of COPD both prescribed prophylactically and therapeutically, and to provide real-world data on neuropsychiatric safety of varenicline use for smoking cessation, particularly among high risk populations with COPD or psychiatric diseases.

OUTLINE OF THIS THESIS

In part I of this thesis, we present several studies on the role of antibiotics in acute exacerbations of COPD (AECOPD).

In Chapter 2, we report the results from a meta-analysis of RCTs focusing on the benefcial efects and side efects of prophylactic antibiotic therapy in COPD patients.

In Chapter 3, we demonstrate the real-world efects of doxycycline treatment on acute exacerbations among COPD outpatients based on data extracted from the University of Groningen’s prescription database (IADB.nl) and explored the possible infuence of age on the clinical outcomes.

In Chapter 4, we further explored the real-world efects of several antibiotic drugs used for acute exacerbations of COPD patients based on a linked database between the Lifelines Cohort biobank with extensive clinical information and the University of Groningen’s prescription database (IADB.nl).

In Chapter 5, we present a systematic review of drug-drug interactions associated with frequently prescribed antibiotics among COPD patients based on causal evidence obtained from observational cohort studies, case-control studies and clinical studies, aimed at improving the safety of antibacterial prescriptions.

13 Chapter 1

In part II of this thesis, we present studies on the role of varenicline for smoking cessation using diferent designs.

In Chapter 6, we present the results of a retrospective inception cohort study aimed at assessing the risk of neuropsychiatric adverse events (NPAEs) in starters with varenicline versus starters with nicotine replacement therapy (NRT) among both the general and COPD populations, with and without psychiatric disorders. This study was conducted using data extracted from the University of Groningen’s prescription database (IADB.nl).

In Chapter 7, we further examine the association between varenicline use and the onset of NPAEs in a real-world setting using a prescription sequence symmetry analysis (PSSA) study design.

Furthermore, in Chapter 8 we systematically compared efect estimates derived from the PSSA study with efect estimates from conventional observational parallel group study designs, to assess the validity and constraints of the PSSA study design within epidemiological research.

Last, in Chapter 9, we summarized the main fndings of this thesis, discussed these fndings in detail and provided suggestions for future research.

14 General Introduction

REFERENCES 1 1. Global Initiative for Chronic Obstructive 10. Rabe KF, Watz H. Chronic obstructive pulmonary Lung Disease (GOLD). Global Strategy for disease. Lancet. 2017;389(10082):1931-1940. the Diagnosis, Management and Prevention of 11. Spruit MA, Singh SJ, Garvey C, et al. An Chronic Obstructive Pulmonary Disease: 2020 ofcial American Thoracic Society/European Report. https://goldcopd.org/gold-reports/. Respiratory Society statement: key concepts Date last accessed: December 17, 2019. and advances in pulmonary rehabilitation. 2. Postma DS, Bush A, van den Berge Am J Respir Crit Care Med. 2013;188(8):e13-64. M. Risk factors and early origins of 12. Wedzicha JA, Seemungal TAR. COPD chronic obstructive pulmonary disease. exacerbations: defning their cause and Lancet. 2015;385(9971):899-909. prevention. Lancet. 2007;370(9589):786-796. 3. Lange P, Celli B, Agusti A, et al. Lung- 13. Soler-Cataluna JJ, Martinez-Garcia MA, Function Trajectories Leading to Chronic Roman Sanchez P, Salcedo E, Navarro M, Obstructive Pulmonary Disease. N Engl J Ochando R. Severe acute exacerbations Med. 2015;373(2):111-122. and mortality in patients with chronic 4. Stern DA, Morgan WJ, Wright AL, Guerra S, obstructive pulmonary disease. Martinez FD. Poor airway function in early Thorax. 2005;60(11):925-931. infancy and lung function by age 22 years: 14. O’Reilly JF, Williams AE, Rice L. Health status a non-selective longitudinal cohort study. impairment and costs associated with COPD Lancet. 2007;370(9589):758-764. exacerbation managed in hospital. Int J Clin 5. Tashkin DP, Altose MD, Bleecker ER, et al. Pract. 2007;61(7):1112-1120. The lung health study: airway responsiveness 15. Sethi S, Murphy TF. Infection in to inhaled methacholine in smokers with the pathogenesis and course of chronic mild to moderate airfow limitation. The Lung obstructive pulmonary disease. N Engl J Health Study Research Group. Am Rev Respir Med. 2008;359(22):2355-2365. Dis. 1992;145(2 Pt 1):301-310. 16. Moghoofei M, Azimzadeh Jamalkandi S, 6. Adeloye D, Chua S, Lee C, et al. Global and Moein M, Salimian J, Ahmadi A. Bacterial regional estimates of COPD prevalence: infections in acute exacerbation of chronic Systematic review and meta-analysis. J Glob obstructive pulmonary disease: a systematic Health. 2015;5(2):020415. review and meta-analysis. Infection. 2019. 7. Buist AS, McBurnie MA, Vollmer 17. Wilkinson TMA, Aris E, Bourne SC, et WM, et al. International variation in al. Drivers of year-to-year variation in the prevalence of COPD (the BOLD Study): exacerbation frequency of COPD: analysis a population-based prevalence study. of the AERIS cohort. ERJ Open Res. 2019;5(1). Lancet. 2007;370(9589):741-750. 18. Monso E, Garcia-Aymerich J, Soler N, et al. 8. Lozano R, Naghavi M, Foreman K, et al. Bacterial infection in exacerbated COPD Global and regional mortality from 235 with changes in sputum characteristics. causes of death for 20 age groups in Epidemiol Infect. 2003;131(1):799-804. 1990 and 2010: a systematic analysis for 19. Wedzicha JA, Calverley PMA, Albert RK, et al. the Global Burden of Disease Study 2010. Prevention of COPD exacerbations: a European Lancet. 2012;380(9859):2095-2128. Respiratory Society/American Thoracic 9. Lopez AD, Shibuya K, Rao C, et al. Chronic Society guideline. Eur Respir J. 2017;50(3). obstructive pulmonary disease: current 20. Ni WT, Shao XD, Cai XJ, et al. Prophylactic Use burden and future projections. Eur of Macrolide Antibiotics for the Prevention Respir J. 2006;27(2):397-412. of Chronic Obstructive Pulmonary Disease

15 Chapter 1

Exacerbation: A Meta-Analysis. Plos a large cross-sectional study in primary One. 2015;10(3). care. Br J Gen Pract. 2017;67(658):e321-e328. 21. Cameron EJ, McSharry C, Chaudhuri R, Farrow S, 31. Mangoni AA, Jackson SH. Age-related changes Thomson NC. Long-term macrolide treatment in pharmacokinetics and pharmacodynamics: of chronic infammatory airway diseases: risks, basic principles and practical applications. Br J benefts and future developments. Clin Exp Clin Pharmacol. 2004;57(1):6-14. Allergy. 2012;42(9):1302-1312. 32. Hanlon P, Nicholl BI, Jani BD, et al. Examining 22. Walters JA, Tan DJ, White CJ, Gibson PG, Wood- patterns of multimorbidity, polypharmacy Baker R, Walters EH. Systemic corticosteroids and risk of adverse drug reactions in for acute exacerbations of chronic chronic obstructive pulmonary disease: obstructive pulmonary disease. Cochrane a cross-sectional UK Biobank study. BMJ Database Syst Rev. 2014(9):CD001288. Open. 2018;8(1):e018404. 23. Vollenweider DJ, Jarrett H, Steurer-Stey CA, 33. Banks E, Joshy G, Korda RJ, et al. Tobacco Garcia-Aymerich J, Puhan MA. Antibiotics smoking and risk of 36 cardiovascular for exacerbations of chronic obstructive disease subtypes: fatal and non-fatal pulmonary disease. Cochrane Db Syst outcomes in a large prospective Australian Rev. 2012(12). study. BMC Med. 2019;17(1):128. 24. Vollenweider DJ, Frei A, Steurer-Stey CA, 34. Gaudet MM, Carter BD, Brinton LA, et al. Garcia-Aymerich J, Puhan MA. Antibiotics Pooled analysis of active cigarette smoking for exacerbations of chronic obstructive and invasive breast cancer risk in 14 cohort pulmonary disease. Cochrane Db Syst studies. Int J Epidemiol. 2017;46(3):881-893. Rev. 2018(10). 35. Tjora T, Hetland J, Aaro LE, Wold B, Wiium 25. van Velzen P, Ter Riet G, Bresser P, et al. N, Overland S. The association between Doxycycline for outpatient-treated acute smoking and depression from adolescence to exacerbations of COPD: a randomised adulthood. Addiction. 2014;109(6):1022-1030. double-blind placebo-controlled trial. 36. Collaborators GBDT. Smoking prevalence and Lancet Respir Med. 2017;5(6):492-499. attributable disease burden in 195 countries 26. Hassan WA, Shalan I, Elsobhy M. Impact of and territories, 1990-2015: a systematic analysis antibiotics on acute exacerbations of COPD. from the Global Burden of Disease Study 2015. Egypt J Chest Dis Tu. 2015;64(3):579-585. Lancet. 2017;389(10082):1885-1906. 27. Bathoorn E, Groenhof F, Hendrix R, et al. 37. Thun MJ, Carter BD, Feskanich D, et Real-life data on antibiotic prescription al. 50-Year Trends in Smoking-Related and sputum culture diagnostics in acute Mortality in the United States. New Engl J exacerbations of COPD in primary care. Int J Med. 2013;368(4):351-364. Chron Obstruct Pulmon Dis. 2017;12:285-290. 38. The L. Progress towards a tobacco-free 28. Roede BM, Bindels PJ, Brouwer HJ, Bresser world. Lancet. 2018;392(10141):1. P, de Borgie CA, Prins JM. Antibiotics and 39. World Health Organisation. WHO report on steroids for exacerbations of COPD in primary the global tobacco epidemic 2019: ofer care: compliance with Dutch guidelines. Br J help to quit tobacco use. Geneva: World Gen Pract. 2006;56(530):662-665. Health Organisation, 2019: 17-21. 29. Halbert RJ, Natoli JL, Gano A, Badamgarav 40. Mehrotra R, Yadav A, Sinha DN, et al. E, Buist AS, Mannino DM. Global burden of Smokeless tobacco control in 180 countries COPD: systematic review and meta-analysis. across the globe: call to action for full Eur Respir J. 2006;28(3):523-532. implementation of WHO FCTC measures. 30. Chetty U, McLean G, Morrison D, Agur K, Lancet Oncol. 2019;20(4):e208-e217. Guthrie B, Mercer SW. Chronic obstructive 41. Anthenelli RM, Benowitz NL, West R, et al. pulmonary disease and comorbidities: Neuropsychiatric safety and efcacy of

16 General Introduction

varenicline, bupropion, and nicotine patch 47. Kotz D, Viechtbauer W, Simpson CR, in smokers with and without psychiatric van Schayck OCP, West R, Sheikh A. 1 disorders (EAGLES): a double-blind, Cardiovascular and neuropsychiatric risks of randomised, placebo-controlled clinical varenicline and bupropion in smokers with trial. Lancet. 2016;387(10037):2507-2520. chronic obstructive pulmonary disease. 42. Moore TJ, Furberg CD, Glenmullen J, Thorax. 2017;72(10):905-911. Maltsberger JT, Singh S. Suicidal behavior 48. Evins AE, Benowitz NL, West R, et al. and depression in smoking cessation Neuropsychiatric Safety and Efcacy of treatments. Plos One. 2011;6(11):e27016. Varenicline, Bupropion, and Nicotine Patch 43. Thomas KH, Martin RM, Knipe DW, Higgins in Smokers With Psychotic, Anxiety, and JP, Gunnell D. Risk of neuropsychiatric Mood Disorders in the EAGLES Trial. J Clin adverse events associated with varenicline: Psychopharmacol. 2019;39(2):108-116. systematic review and meta-analysis. 49. Boyko EJ. Observational research-- BMJ. 2015;350:h1109. opportunities and limitations. J Diabetes 44. Lawrence D, Mitrou F, Zubrick SR. Smoking Complications. 2013;27(6):642-648. and mental illness: results from population 50. Wahab IA, Pratt NL, Wiese MD, Kalisch LM, surveys in Australia and the United States. Roughead EE. The validity of sequence Bmc Public Health. 2009;9. symmetry analysis (SSA) for adverse drug 45. Garza D, Murphy M, Tseng LJ, Riordan HJ, reaction signal detection. Pharmacoepidemiol Chatterjee A. A double-blind randomized Drug Saf. 2013;22(5):496-502. placebo-controlled pilot study of 51. Lai ECC, Pratt N, Hsieh CY, et al. Sequence neuropsychiatric adverse events in abstinent symmetry analysis in pharmacovigilance and smokers treated with varenicline or placebo. pharmacoepidemiologic studies. European Biol Psychiatry. 2011;69(11):1075-1082. Journal of Epidemiology. 2017;32(7):567-582. 46. Tonstad S, Davies S, Flammer M, Russ C, Hughes 52. Hallas J. Evidence of depression J. Psychiatric adverse events in randomized, provoked by cardiovascular medication: double-blind, placebo-controlled clinical a prescription sequence symmetry analysis. trials of varenicline: a pooled analysis. Drug Epidemiology. 1996;7(5):478-484. Saf. 2010;33(4):289-301.

17 PART I Efects of antibiotic use for COPD exacerbations and potential DDIs during COPD exacerbation management CHAPTER 2 Efects of Prophylactic Antibiotics on Patients with Stable COPD: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Yuanyuan Wang Tanja R. Zijp Muh. Akbar Bahar Janwillem W.H. Kocks Bob Wilfert Eelko Hak

Published as: Wang Y, Zijp TR, Bahar MA, Kocks JWH, Wilfert B, Hak E. Efects of prophylactic antibiotics on patients with stable COPD: a systematic review and meta-analysis of randomized controlled trials. J Antimicrob Chemother. 2018;73(12):3231–3243. Chapter 2

ABSTRACT Background As bacterial infections provoke exacerbations, COPD patients may beneft from prophylactic antibiotics. However, evidence regarding their overall beneft-risk is conficting.

Objectives To update previous evidence and systematically evaluate the benefcial and side efects of prophylactic antibiotics on stable COPD patients.

Methods Several databases were searched up to April 26, 2017 for randomized controlled trials (RCTs) on prophylactic antibiotics in stable COPD patients. The Primary outcomes were exacerbations and quality of life. Duration and schedule of antibiotics were considered in sub-group analyses.

Results Twelve RCTs involving 3,683 patients were included. Prophylactic antibiotics signifcantly reduced the frequency of exacerbations (risk ratio [RR] 0.74, 95% CI 0.60-0.92) and the number of patients with one or more exacerbations (RR 0.82, 95% CI 0.74-0.90). Erythromycin and azithromycin appeared the most efective with the number needed to treat ranging from four to seven. Quality of life was also signifcantly improved by prophylactic antibiotics (mean diference -1.55, 95% CI -2.59 to -0.51). Time to f rst exacerbation was prolonged in six studies with one conficting result. Neither the rate of hospitalization nor the rate of adverse events was signifcantly changed. Furthermore, no signifcant changes were observed in lung function, bacterial load and airway infammation. However, antibiotic resistant isolates were signifcantly increased (OR 4.49, 95% CI 2.48-8.12).

Conclusions Prophylactic antibiotics were efective in preventing COPD exacerbations and improving quality of life among stable patients with moderate to severe COPD. The choice of prophylactic antibiotics should be analysed and considered case by case, especially for long and continuous use.

22 Efects of prophylactic antibiotics on COPD

INTRODUCTION

COPD is an infammatory disease that is characterized by persistent respiratory symptoms and airfow limitation.1 At present, COPD is one of the leading causes 2 of chronic morbidity and mortality worldwide, its burden is predicted to increase in the coming decades due to continuous exposure to risk factors and aging of population globally.2 In the course of COPD, exacerbation as an acute worsening of respiratory symptoms has a profound negative impact on health oucomes.3 A vicious circle of infection and infammation is thought as a key to trigger exacerbations of COPD, about 40-50% of exacerbations are caused by bacteria.4

The use of prophylactic antibiotic has been suggested to prevent exacerbations in COPD patients for a long time. However, a Cochrane review in 2003 concluded that antibiotics only contribute to a small 9% reduction of exacerbations and should not be part of routine treatment considering the risk of antibiotic resistance and adverse efects.5 Ten years later in 2013, the review by Herath et al. concluded a clinically signifcant beneft in reducing COPD exacerbations from continuous use of prophylactic antibiotics, but not from intermittent way due to only one randomized controlled trial (RCT) included in this subgroup.6 Infuence of diferent duration of antibiotic intervention were not explored in this study. The most recent review by Ni et al. in 2015 focused on macrolides only and did not evaluate meaningful outcomes including the time to frst exacerbation, change of lung function, bacterial load and airway infammation.7 The latter outcome is important to support the hypothetical mechanism behind the reduction of exacerbations by antibiotics.8

Current recommendation from guidelines about prophylactic antibiotic use in the management of COPD exacerbations is conditional and unspecifc.1,9 At present, the optimal regimen of prophylactic antibiotics for exacerbations has not been well established, and there are no advices for an appropriate schedule and duration of specifc antibiotic intervention. To further enhance information on the public health beneft-risk associated with this intervention, we here aimed to provide a comprehensive overview of the positive and negative efects of prophylactic antibiotics on COPD patients. METHODS Search strategy We performed an update of the previous review by Herath et al. in 20136 according to the PRISMA guidelines. Cochrane Central Register of Controlled Trials (CENTRAL), Medline, EMBASE, Web of Science, CINAHL, AMED and PsycINFO databases were systematically searched for relevant RCTs published from 29 August 2013 (when the review by Herath et al. ended) until 26 April 2017 using key elements of “COPD”,

23 Chapter 2

“RCT” and “antibiotics” (details are presented in Table S1). References from identifed studies and relevant review articles were also checked manually. No language restrictions were applied. For the fnal analysis, we included both the new studies from this searching strategy and previous studies from the review by Herath et al.

Selection criteria Studies included in this review met the following criteria: (1) focus on the efects of prophylactic antibiotics in COPD patients; (2) study designs must be RCTs with placebo group; (3) COPD patients should be aged over 18 years and with a well-defned diagnosis of COPD and confrmed evidence of persistent airfow limitation (the presence of a post- bronchodilator FEV1/FVC < 0.7); (4) prophylactic antibiotics must be given for a minimum period of 12 weeks; (5) patients must be clinically stable without exacerbation for at least three weeks before enrolment. Studies that focused on combined antibiotics (≥ 2) and studies of patients with other respiratory disease (e.g. bronchiectasis, asthma) or related genetic diseases such as cystic fbrosis and primary ciliary dyskinesia were excluded.

Outcomes and data analysis The Primary outcomes were: number of patients with exacerbations; frequency of exacerbation; health-related quality of life assessed by the St Georges Respiratory Questionnaire (SGRQ).10 The Secondary outcomes were: the median time to frst exacerbation; frequency of hospitalization; all-cause mortality; adverse events; antibiotic resistance; change in lung functions, bacteria load and airway infammation. The infuence of diferent schedules and durations of prophylactic antibiotic use on exacerbations and quality of life in COPD patients were explored. For the missing of standard deviation of SGRQ score change in two studies,11,12 we calculated it according to Cochrane guideline (see Supplement data). All analyses were done in accordance with the intention-to-treat principle using Review Manager Version 5.3. Risk ratio (RR) or OR was calculated for binary outcomes, while mean diference (MD) was for continuous outcomes. Generic inverse variance (GIV) methods were used for non-standard types of both dichotomous and continuous data. Summary measures were pooled using random- efects models. If data could not be combined, we performed a descriptive analysis. Statistical heterogeneity among studies was assessed using conventional chi-squared (X2, or Chi2) test and I2 statistic of inconsistency. Sensitivity analysis was performed by removing studies with a high risk for bias or deviation. A funnel plot was used to assess publication bias.

24 Efects of prophylactic antibiotics on COPD

Figure 1. Flow diagram of literature search and study selection. Figure 1. Flow diagram of literature search and study selection.

data). All analyses were done in accordance with the intention‐to‐treat principle using Review RESULTSManager Version 5.3. Risk ratio (RR) or OR was calculated for binary outcomes, while mean Searchdifference results (MD) was for continuous outcomes. Generic inverse variance (GIV) methods were used for non‐standard types of both dichotomous and continuous data. Summary measures From the 667 records generated by new search strategy, fve new RCT studies were eligiblewere pooled and includedusing random (Figure‐effects 1). models. Together If data with could the not previous be combined, seven we studiesperformed from a thedescriptive review by analysis. Herath Statistical et al.,6 heterogeneitya total of twelve among RCTs studies were was included assessed forusing this conventional systematic 13 review.chi‐squared However, (X2, orof allChi twelve2) test andstudies I2 statistic , one was of inconsistency.a conference Sensitivityabstract, analysis one was was not blinded,14 one did not report efect measures.15 In total, nine studies were qualifed for the meta-analysis.

25 Chapter 2

Table 1. Characteristics of included studies.

Age Duration of

Studies Study Patients (year) FEV1/FVC ratio (%) Prophylactic Antibiotics treatment & follow up (1st author, year) design Country (T/P) (T/P) (T/P) (dose) (months) Maintenance medication

Previous included studies

Albert, 2011 RCT US 570:572 65:66 42:43 Azithromycin, 250 mg daily 12 / 12 ICS, LABA, LAMA He, 2010 RCT UK 18:18 68.8:69.3 46.9:48.6 Erythromycin, 125 mg, 3 times a day; 6 / 6 ICS, Mygind, 2010 RCT Denmark 287:288 71 (Median) NA Azithromycin, 500 mg daily, 3 days a month 36 / 36 ICS, Theophylline, inhaled anticholinergic, inhaled β-adrenergic Sethi, 2010 RCT US 569:580 66.1:66.6 45.0:46.3 Moxifoxacin, 400 mg daily, 5 days every 8 12 / 18 LABA, LAMA, SABA, SAMA, weeks ICS, theophylline; Seemungal, 2008 RCT UK 53:56 66.6:67.8 48.9:50.9 Erythromycin, 250 mg twice daily 12 / 12 LABA, LAMA, theophylline Banerjee, 2005 RCT UK 31:36 65.1:68.1 43.8:45.5 Clarithromycin, 500 mg once daily 3 / 3 ICS Suzuki, 2001 RCT Japan 55:54 69.1:71.7 NA Erythromycin, 200-400 mg daily 12 / 12 Inhaled anticholinergic, theophylline

New included studies

* Brill, 2015 RCT UK (25:25:25):24 (70.9:70.4:67.9): 68.7 (51:51:45):51 T1: Moxifoxacin, 400 mg, 5 times every 4 weeks; 3.25 /3.25 ICS

T2: Doxycycline, 100 mg daily;

T3: Azithromycin, 250 mg, 3 times a week; †Shafuddin, 2015 RCT New Zealand 97:94 67.6:66.7 41.5:43.7 Roxithromycin, 300 mg daily 3 / 12 Not available Simpson, 2014 RCT Australia 15:15 71.7:69.9 52.3:51.3 Azithromycin, 250 mg daily 3 / 6 ICS Uzun, 2014 RCT Netherlands 47:45 64.7:64.9 38.0:40.3 Azithromycin, 500 mg, 3 times a week 12 / 12 LABA, LAMA, SABA, ICS, Prednisolone Berkhof, 2013 RCT Netherlands 42:42 67:68 42.2:43.2 Azithromycin, 250 mg, 3 times a week 3 / 4.5 LABA, LAMA, ICS

T/P: Treatment group versus Placebo group; ICS: inhaled corticosteroid; LABA: long-acting beta-2 agonists; LAMA: long- 3 diferent treatment arms with one common placebo arm; †The study included 2 treatment arms, according to preset acting muscarinic antagonist; SABA: short-acting beta-2 agonists; SAMA: short-acting muscarinic antagonist; NA: data were criteria, we only include the arm about single antibiotic use, the other arm in this study about combined antibiotic treatment not available; *This study designed is excluded;

Characteristics of included studies The characteristics of twelve included studies are shown in Table 1, other specifc baseline characteristics about COPD severity and exacerbation history were summarized in Table S2. All these studies were conducted over the last seventeen years involving 3,683 stable COPD patients, with 2932 patients involved in the meta- analysis. All included studies focused on one antibiotic arm with one placebo arm except the study by Brill et al.,16 which compared three antibiotics with one common placebo and we treated this study as three independent RCTs (Trial 1-3: 11,13,16-19 T1, T2, T3). In all, six antibiotics were investigated in this review: azithromycin, erythromycin,12,14,20 moxifoxacin,16,21 clarithromycin,15 roxithromycin22 and doxycycline.16 The duration of treatment ranged from 3 to 36 months with study size ranging from 30 to 1,149 patients.

26 Efects of prophylactic antibiotics on COPD

Table 1. Characteristics of included studies.

Age Duration of

Studies Study Patients (year) FEV1/FVC ratio (%) Prophylactic Antibiotics treatment & follow up (1st author, year) design Country (T/P) (T/P) (T/P) (dose) (months) Maintenance medication 2

Previous included studies

New included studies

* Brill, 2015 RCT UK (25:25:25):24 (70.9:70.4:67.9): 68.7 (51:51:45):51 T1: Moxifoxacin, 400 mg, 5 times every 4 weeks; 3.25 /3.25 ICS

T2: Doxycycline, 100 mg daily;

Quality assessment The review authors’ judgment about each risk of bias item in each study can be seen in Figure S1.1. The risk of bias items presented as percentage across all included studies were presented in Figure S1.2. There was no reporting bias in all included studies; only 2 studies14,16 have potential high risk in the blinding process. For the remaining of bias items, only a small proportion of unclear bias exists. Overall, low risk of bias dominates in all domains of bias.

Primary outcomes Seven studies involving 2,642 participants11,12,17-21 reported the number of patients with exacerbations (Figure 2), which was signifcantly reduced (RR 0.82, 95% CI 0.74- 0.90) by prophylactic antibiotics and there was no diference between continuous

27 significance (p = 0.07) suggested that using antibiotics 6 months may achieve better

Chaptertreatment 2 effects (RR 0.59, 95% CI 0.40‐0.86) than longer time (RR 0.84, 95% CI 0.77‐0.93), which requires further confirmation. The risk difference (RD) between antibiotic and placebo

FigureFigure 2. 2. Forest Forest plot plot of ofrisk risk ratio ratio (antibiotics (antibiotics versus versus placebo) placebo) for totalfor total number num ofber patients of patients with withone or more exacerbations stratifed by (a) schedule of prophylactic antibiotics and (b) duration of prophylactic one or more exacerbations stratified by (a) schedule of prophylactic antibiotics and (b) antibiotics. M-H: Mantel-Haenszel; *Studies reviewed by Herath et al. in 2013. duration of prophylactic antibiotics. M‐H: Mantel‐Haenszel; *Studies reviewed by Herath et al. in 2013. and intermittent subgroups. However, the diference between other subgroups with a distinct trend toward signifcance (p = 0.07) suggested that using antibiotics ≤ 6 months may achieve better treatment efects (RR 0.59, 95% CI 0.40-0.86) than longer time (RR 0.84, 95% CI 0.77-0.93), which requires further confrmation. The risk diference

(RD) between antibiotic and placebo groups is presented in Figure 3, for erythromycin, the RD is substantial (RD -0.24, 95% CI -0.39 to -0.08), the corresponding number needed to treat (NNT) was 4; for azithromycin, the RD was moderate (RD -0.14, 95% CI -0.20 to -0.08), the NNT was 7; no statistically signifcant efect for moxifoxacin intervention.

28 Efects of prophylactic antibiotics on COPD

FigureFigure 3. 3.Forest Forest plot plot of riskof risk dif erencedifference (antibiotics (antibiotics versus versus placebo) placebo) for total for number total number of patients of patients with one or more exacerbations stratifed by types of antibiotics. M-H: Mantel-Haenszel test; *Studies reviewed with one or more exacerbations stratified by types of antibiotics. M‐H: Mantel‐Haenszel test; by Herath et al. in 2013. *Studies reviewed by Herath et al. in 2013.

Usegroups of prophylacticis presented in antibioticFigure 3, for was erythromycin, also associated the RD is withsubstantial a signi (RDf ‐cant0.24, reduction95% CI ‐0.39 in theto ‐frequency0.08), the correspondingof exacerbations number (RR 0.74,needed 95% to CItreat 0.60-0.92, (NNT) was Figure 4; for 4). azithromycin, As the study the by RD Brill et al.16 has a potential risk of bias in the blinding process, a sensitivity analysis was done was moderate (RD ‐0.14, 95% CI ‐0.20 to ‐0.08), the NNT was 7; no statistically significant effect with the other 6 studies, which resulted in a 31% RR reduction of exacerbations among patientsfor moxifloxacin taking prophylactic intervention. antibiotics (RR 0.69, 95% CI 0.58-0.82). In subgroup analysis showedUse of prophylactic in Figure 4, antibioticmacrolides was (azithromycin, also associated erythromycinwith a significant and reduction roxithromycin) in the frequency showed benefcial efects on frequency reduction of exacerbations, the benefts from both of exacerbations (RR 0.74, 95% CI 0.60‐0.92, Figure 4). As the study by Brill et al.16 has a azithromycin and erythromycin were of clinical signifcance. However, this benefcial potential risk of bias in the blinding process, a sensitivity analysis was done with the other 6 efect was not seen in the use of moxifoxacin and doxycycline. These subgroup distudies,ferences which for frequency resulted in of a exacerbations 31% RR reduction were of of exacerbationsstatistical signi amongfcance patients (p = 0.02). taking prophylactic antibiotics (RR 0.69, 95% CI 0.58‐0.82). In subgroup analysis showed in Figure 4, Health-related quality of life using SGRQ was measured in seven studies.11,12,16-19,21 When macrolides (azithromycin, erythromycin and roxithromycin) showed beneficial effects on we performed a sensitivity analysis by removing the study by Berkhof et al.,18 which wasfrequency very di reductionferent from of exacerbations, the other data, the thebenefits heterogeneity from both azithromycin reduced sharply and erythromycin (I2 changed fromwere 92% of clinicalto 0%). significance. Hence, only However, the remaining this beneficial 6 studies effect were was included not seen for thein the fnal use meta- of analysis. The pooled result indicated that prophylactic antibiotics led to a signifcant improvement in the total SGRQ score (MD -1.55, 95% CI -2.59 to -0.51, Figure 5). In subgroup analysis, the improvement of SGRQ score was not seen in both continuous and intermittent antibiotics. However, another subgroup result indicated that the total

29 moxifloxacin and doxycycline. These subgroup differences for frequency of exacerbations were of statistical significance (p = 0.02).

Health‐related quality of life using SGRQ was measured in seven studies.11,12,16‐19,21 When we performed a sensitivity analysis by removing the study by Berkhof et al.,18 which was very different from the other data, the heterogeneity reduced sharply (I2 changed from 92% to 0%). Hence, only the remaining 6 studies were included for the final meta‐analysis. The pooled result indicated that prophylactic antibiotics led to a significant improvement in the total Chapter 2 SGRQ score (MD ‐1.55, 95% CI ‐2.59 to ‐0.51, Figure 5). In subgroup analysis, the improvement of SGRQ score was not seen in both continuous and intermittent antibiotics. However, another

FigureFigure 4. Forest4. Forest plot plot of risk of ratiorisk ratio(antibiotics (antibiotics versus versusplacebo) placebo) for frequency for frequency of exacerbations of exacerbations stratifed by types of antibiotics. SE: standard error; IV: inverse variance; *Studies reviewed by Herath et al. in 2013; stratified by types of antibiotics. SE: standard error; IV: inverse variance; *Studies reviewed by T1-3: three independent RCTs in study by Brill et al. Herath et al. in 2013; T1‐3: three independent RCTs in study by Brill et al.

SGRQ score signifcantly changed by long-term intervention (MD -1.70, 95% CI% -2.81 to -0.60), although it was not changed by short-term (≤ 6) intervention (MD -0.34, 95% CI -3.43 to 2.75).

Four studies12,17,19,21 also reported the component scores of SGRQ (Figure S2). Both the symptom (MD -3.89, 95% CI -5.48 to -2.31) and impact (MD -1.32, 95% CI -2.61 to -0.03) scores were improved with prophylactic antibiotics. However, the activity score did not show any signifcant improvement. Of note, none of these improvements mentioned above in SGRQ score reached the hypothesized clinically benefcial level (> 4-unit reduction).10

30 Efects of prophylactic antibiotics on COPD

FigureFigure 5. Forest5. Forest plot plot of mean of mean diference difference (antibiotics (antibiotics versus versus placebo) placebo) of quality of quality of life by of SGRQ life by strati SGRQfed by stratified(a) schedule by of prophylactic(a) schedule antibiotics of prophylactic and (b) duration antibiotics of prophylactic and (b) antibiotics.duration of SGRQ: prophylactic St Georges Respiratory Questionnaire; IV: inverse variance; SE: standard error; *Studies reviewed by Herath et al. in antibiotics. SGRQ: St Georges Respiratory Questionnaire; IV: inverse variance; SE: standard 2013; T1-3: three independent RCTs in study by Brill et al. * error; Studies reviewed by Herath et al. in 2013; T1‐3: three independent RCTs in study by Brill Secondaryet al. outcomes

Sevensubgroup studies result involving indicated 2,803 that patients the total reported SGRQ score the mediansignificantly time changed to frst byexacerbation long‐term (Table S3). Four studies indicated that using prophylactic antibiotics lengthened intervention (MD ‐1.70, 95% CI% ‐2.81 to ‐0.60), although it was not changed by short‐term ( the median time to frst exacerbation signfcantly.12,17,19,20 Two other studies found a 6)similar intervention trend, (MD but ‐0.34, without 95% CI ‐ statistical3.43 to 2.75). signi fcance.18,21 Only one study showed the opposite result in antibiotic and placebo arms.22

31 Four studies12,17,19,21 also reported the component scores of SGRQ (Figure S2). Both the symptom (MD ‐3.89, 95% CI ‐5.48 to ‐2.31) and impact (MD ‐1.32, 95% CI ‐2.61 to ‐0.03) scores were improved with prophylactic antibiotics. However, the activity score did not show any significant improvement. Of note, none of these improvements mentioned above in SGRQ score reached the hypothesized clinically beneficial level (> 4‐unit reduction).10

Secondary outcomes

Seven studies involving 2,803 patients reported the median time to first exacerbation (Table S3). Four studies indicated that using prophylactic antibiotics lengthened the median time to first exacerbation signficantly.12,17,19,20 Two other studies found a similar trend, but without statistical significance.18,21 Only one study showed the opposite result in antibiotic and placebo arms.22

The frequency of hospitalization related to COPD was pooled from five studies with 2,576 participants,17‐21 no significant difference was observed between antibiotic and placebo groups (RR 0.94, 95% CI 0.83‐1.06, Figure 6). Also, no difference in the rate of all‐cause mortality were found between the two arms (Figure S3).

Eight studies involving 2,833 participants reported adverse events related to antibiotic use.11,12,17‐22 Overall, there was no significant difference between two comparison arms in the rate of adverse events (RR 1.09, 95% CI 0.84‐1.42, Figure 7). As there was a lack of uniform definition about adverse events, the heterogeneity was substantial (I2 = 73%). Considering that Chapter 2 the result by Shafuddin et al.22 was deviant from the other seven studies, a sensitivity analysis was, therefore, performed after removal of this study. The homogeneous result also did not

FigureFigure 6. 6.Forest Forest plot plot of ofrisk risk ratio ratio (antibiotics (antibiotics versus versus placebo) placebo) for forfrequency frequency of hospitalization.of hospitalization. M-H: Mantel-Haenszel test; *Studies reviewed by Herath et al. in 2013. M‐H: Mantel‐Haenszel test; *Studies reviewed by Herath et al. in 2013.

FigureFigure 7. 7.Forest Forest plot plot of riskof risk ratio ratio (antibiotics (antibiotics versus versus placebo) placebo) for adverse for adverse events. events. M-H: Mantel-HaenszelM‐H: Mantel‐ * test;Haenszel Studies test; reviewed *Studies by Herathreviewed et al by. in Herath 2013. et al. in 2013.

2 Theshow frequency significant of differencehospitalization between related two armsto COPD (RR 0.93, was 95%pooled CI 0.83from‐1.05, fve I studies = 2%). withIn 2,576subgroups, participants, gastrointestinal17-21 no signi disordersfcant were diference more frequentwas observed in the interventionbetween antibiotic group than and placebocontrol groupgroups (RR (RR 1.87, 0.94, 95 CI95% 0.98 CI‐3.59, 0.83-1.06, Figure S4)Figure with 6).a boundary Also, no statistical diference significance in the rate (p = of all-cause mortality were found between the two arms (Figure S3). 0.06). However, no statistical significant difference for respiratory and cardiovascular Eightdisorders studies were involving found. 2,833 participants reported adverse events related to antibiotic use.11,12,17-22 Overall, there was no signifcant diference between two comparison Eight studies had the bacteriological assessments (see Table S4),12,15‐21 however, only three arms in the rate of adverse events (RR 1.09, 95% CI 0.84-1.42, Figure 7). As there was 16,17,19 a studieslack of with uniform five RCTs def reportednition about the quantitative adverse events, results forthe antibiotic heterogeneity resistance. was substantial Due to (I2the = 73%). different Considering definition about that bacterialthe result resistant by Shafuddin outcome etin alstudy.22 was by Uzun deviant et al ,from only the the other other sevenhomogeneous studies, studiesa sensitivity involving analysis four RCTs was, were therefore, included for performed pooled results after (OR removal 4.49, 95% of thisCI study. The homogeneous result also did not show signifcant diference between two 2.48‐8.12, Figure 8), long‐term (versus short‐term) and continuous (versus intermittent) arms (RR 0.93, 95% CI 0.83-1.05, I2 = 2%). In subgroups, gastrointestinal disorders were antibiotic intervention seems to cause more antibiotic resistance, although these subgroup more frequent in the intervention group than control group (RR 1.87, 95 CI 0.98-3.59, difference did not reach statistical significant level. Antibiotic resistance appeared in all types of antibiotics involved (Figure 9), although the result from moxifloxacin did not reach 32 statistical significance. No subgroup differences about this outcome were seen among azithromycin, moxifloxacin and doxycycline (p = 0.63).

Eight studies11,13,16‐18,20‐22 provided the data on changes of lung function (Table S5). However, no study found significant increase by antibiotic intervention compared with placebo. Mygind et al. did not compare the lung function change directly, but measured and compared the lung function in both groups at enrolment and endpoint separately, they also did not find any significant difference.13

Three studies reported the change of bacterial load.11,15,16 Although both Brill et al. and Simpson et al. have found the more reduction of bacterial load by prophylactic antibiotic compared with placebo, the results did not reach the level of statistical significance, even both quantitative culture and 16S qPCR methods were used by Brill et al. Benerjee et al. also did not find a significant difference between pre‐and post‐ sputumEfects of cfu prophylactic numbers/bacterial antibiotics (PPM)on COPD isolates in two arms.

FigureFigure 8. Forest8. Forest plot plotof odds of ratioodds (antibiotics ratio (antibiotics versus placebo) versus forplacebo) antibiotic for resistance antibiotic strati resistancefed by (a) schedulestratified of prophylacticby (a) schedule antibiotics of prophylactic and (b) duration antibiotics of prophylactic and (b) antibiotics. duration IV:of inverse prophylactic variance; *Studies reviewed by Herath et al. in 2013; T : three independent RCTs in study by Brill et al. * 1-3 antibiotics. IV: inverse variance; Studies reviewed by Herath et al. in 2013; T1‐3: three independent RCTs in study by Brill et al. Figure S4) with a boundary statistical signifcance (p = 0.06). However, no statistical signifcant diference for respiratory and cardiovascular disorders were found.

Eight studies had the bacteriological assessments (see Table S4),12,15-21 however, only three studies with fve RCTs reported the quantitative results for antibiotic resistance.16,17,19 Due to the diferent defnition about bacterial resistant outcome in study by Uzun et al, only the other homogeneous studies involving four RCTs were included for pooled results (OR 4.49, 95% CI 2.48-8.12, Figure 8), long-term (versus short-term) and continuous (versus intermittent) antibiotic intervention seems to cause more antibiotic resistance, although these subgroup diference did not reach statistical signifcant level. Antibiotic resistance appeared in all types of antibiotics involved (Figure 9), although the result

33 Chapter 2

FigureFigure 9. 9:Forest Forest plot plot of odds of oddsratio (antibioticsratio (antibiotics versus placebo)versus placebo) for antibiotic for antibiotic resistance resistancestratifed by types of antibiotics. SE: standard error; IV: inverse variance; *Studies reviewed by Herath et al. in 2013; stratified by types of antibiotics. SE: standard error; IV: inverse variance; *Studies reviewed by T1-3: three independent RCTs in study by Brill et al. Herath et al. in 2013; T1‐3: three independent RCTs in study by Brill et al.

The change of airway inflammation was only reported in two studies.11,16 The study by Brill et fromal. showed moxif oxacinthat no did significant not reach changes statistical were signiseen fincance. cytokines No subgroupIL‐6, IL‐8 and di fILerences‐1 in any about of this outcome were seen among azithromycin, moxifoxacin and doxycycline (p = 0.63). three antibiotic arms compared with placebo.16 Similarly, Simpson et al. also did not report a Eightsignificant studies reduction11,13,16-18,20-22 in sputum provided neutrophil the data proportion on changes level of of IL lung‐8 in functionthose who (Table received S5). However,azithromycin no study compared found to signiplacebofcant group. increase11 by antibiotic intervention compared with placebo. Mygind et al. did not compare the lung function change directly, but measured and compared the lung function in both groups at enrolment and endpoint separately, they also did not fnd any signifcant diference.13

Three studies reported the change of bacterial load.11,15,16 Although both Brill et al. and Simpson et al. have found the more reduction of bacterial load by prophylactic antibiotic compared with placebo, the results did not reach the level of statistical signifcance, even both quantitative culture and 16S qPCR methods were used by Brill et al. Benerjee et al. also did not fnd a signifcant diference between pre-and post- sputum cfu numbers/bacterial (PPM) isolates in two arms.

The change of airway infammation was only reported in two studies.11,16 The study by Brill et al. showed that no signifcant changes were seen in cytokines IL-6, IL-8 and IL-1β in any of three antibiotic arms compared with placebo.16 Similarly, Simpson et al. also did not report a signifcant reduction in sputum neutrophil proportion level of IL-8 in those who received azithromycin compared to placebo group.11

34 Efects of prophylactic antibiotics on COPD

DISCUSSION

This update of previous systematic reviews demonstrates that prophylactic antibiotic use could signifcantly lower the risk of exacerbations by 26% and prevent stable COPD 2 patients from getting exacerbations by 18%, which is consistent with the result by Herath et al.,6 but the diference is that our review with more RCTs suggest that intermittent antibiotics may also be efective in preventing exacerbations, although the result is of boundary signifcance. Moreover, in contrast with the result by Ni et al.,7 we found both short-term (≤ 6 months) and long-term (> 6 months) treatments can prevent patients from exacerbations signifcantly. A short-term treatment even had better prevention efects than long-term treatment. Considering all included patients are clinically stable without exacerbation before enrolment, the above beneft from short therapy is likely due to the benefts of less resistance and adverse events or shorter follow-up time to detect related exacerbations compared with long therapy.

Besides duration and schedule of antibiotics, the types of antibiotics also have a profound infuence on preventing exacerbations of COPD. In our pre-specifed subgroup analysis, we did not fnd signifcant efect from moxifoxacin and doxycycline intervention on preventing exacerbations, although a previous study showed moxifoxacin is equivalent and bacteriologically superior to other antibiotic regimens routinely used.23 However, our results confrmed the superiority of macrolides (azithromycin, erythromycin) in preventing exacerbations of COPD. This beneft of macrolides has also been confrmed previously in patients with cystic fbrosis and non-cystic fbrosis bronchiectasis.24,25

Although the optimal treatment using macrolide for preventing exacerbation was already conditional recommended by related guidelines,1,9 the mechanisms behind are not totally clear.8 Many studies have confrmed that macrolides with 14 and 15- membered macrocyclic lactone ring have properties such as anti-infammatory, antiviral and potential immune-modulation,26 which were proved to be benefcial for COPD patients.27 Therefore, some researchers hypothesized that prevention of exacerbation by macrolides may due to its antimicrobial efects or anti-infammatory efects or both. However, neither of the above mechanisms could be supported by evidence in our review,11,15,16,20 More studies are needed in future to explore the answers to this question.

Regarding the health-related quality of life, our review showed a signifcant reduction in the total score of SGRQ with no heterogeneity. This is consistent with the association study by Martin et al.28 From our study, duration longer than 6 months of antibiotic intervention can signifcantly improve the total score of SGRQ. As the health-related quality of life is infuenced largely by the frequency of exacerbations in COPD patients,29 it will be an ideal therapy if both the exacerbation and quality of life change towards the same positive direction. Our subgroup analysis in both exacerbation and quality of life showed the positive results in longer duration (above 6 months) of prophylactic

35 Chapter 2 antibiotics. However, as the improvements of total SGRQ score did not reach a clinical signifcant level, further research were still in need to explore the infuence of prophylactic antibiotic use on quality of life in the real world.

The benefts achieved by prophylactic antibiotics always came at the expense of a variety of adverse events according to the earlier reviewers.30 However, we did not fnd signifcant diferences in the overall rate of adverse events between antibiotic and placebo arms. It is worth noting that the heterogeneity was substantial due to the variety of defnition and measurement methods, thus much consistent defnition is needed for future study. Furthermore, much attention should be given to the gastrointestinal disorder by antibiotic use as this disorder was also observed in patients with cystic fbrosis.24 Although not established as an endpoint in our review, hearing loss caused by azithromycin also should draw much attention.19

Along the use of prophylactic antibiotics, another growing concern regarding the development of antibiotic resistance also appears. In this review, the increased resistant isolates were seen during the intervention of prophylactic antibiotics, which involved macrolides (azithromycin), tetracycline (doxycycline), and quinolones (moxifoxacin).16,19 At the same time, as lots of conficting reports existed with heterogeneous defnitions,15,17,18,20,21 much related evidence from studies of uniform criteria is needed for further exploration. Before that, clinicians should pay much attention especially to long and continuous use of antibiotics considering potential risk of bacterial resistance for future treatment of infections.4 Furthermore, although the use of macrolides in preventing exacerbations was considered as a cost-efective strategy31, the rather quick bacterial resistance induced by macrolides should not be ignored.32 Its use should at best be limited to high-risk populations based on consideration of age, exacerbation frequency in previous year, COPD severity and comorbidity conditions. The choice of antibiotics should be based on the community resistant pattern and their benefts and potential risks must be weighted by analysing the specifc situation case by case.

Although the obvious benefts of antibiotics in prevention of exacerbations, we did not fnd any reduction in the rate of hospital admission by antibiotic intervention, which are in contrast with the result by Donath et al.33 Moreover, as hospitalization for exacerbation is always associated with poor prognosis and increased mortality in COPD patients,34 there was also no diference in the rate of all-cause mortality between antibiotic and placebo groups. Besides, the lung function was not improved in any of the included studies after the antibiotic intervention. There is still no conclusive clinical trial evidence up to now that any existing medication for COPD could modify the long-term decline in lung function.1

36 Efects of prophylactic antibiotics on COPD

Study limitations and future perspectives There were several limitations in this review. Firstly, there were some notable heterogeneous results between studies. On the one hand, due to limited information available, we could not totally analysis and exclude the infuence of potential diference 2 in distribution of baseline characteristics especially like COPD severity, exacerbation history and bacterial colonization on outcomes, although all included patients were relative stable with similar COPD severity (GOLD 2-4). On the other hand, heterogeneity also existed between antibiotic therapies, as regimen, dosages, durations and follow-up time of antibiotic intervention were diferent. Secondly, the included patients may concomitantly take other therapies such as infuenza vaccines, bronchodilators or inhaled corticosteroid, which could also have a potential impact on related outcomes if these factors are not comparable between antibiotic and placebo groups. For example, LABA/LAMA combination as a maintenance therapy of COPD could reduce the rate of exacerbation.35,36 Thirdly, the defnitions and measurements of some outcomes were diferent, like the varying defnitions of adverse events and varying methods for identifying antibiotic resistance. Finally, due to limited studies included, we could not evaluate the efects of the diferent doses of a specifc antibiotic on COPD patients.

In the future, more RCTs of high quality are needed to explore a more personalized therapy by studying the optimal dose, duration and schedule of specifc antibiotic use, preferably macrolides, with therapeutic drug monitoring on more homogenous COPD patients. Besides, uniform standards for evaluating the efects of antibiotic use should be made. Considering the safety of antibiotics, how to avoid or reduce the side efects such as gastrointestinal events and bacterial resistance during long-term use of antibiotic is still a problem that needs to be tackled. CONCLUSIONS

This updated systematic review confrms the beneft of prophylactic antibiotics in preventing exacerbations in stable patients with moderate to severe COPD, this beneft existed in all subgroups ignoring the diferent duration and schedules of antibiotic intervention. The overall quality of life was also signifcantly increased by prophylactic antibiotics. However, this beneft was only observed in long-term (above 6 months) subgroup of antibiotics. At the same time, considering the possible risk of bacterial resistance, long-term and continuous prophylactic antibiotics are at best limited to high risk of population with severe COPD and history of frequent exacerbations and the choice of antibiotic should be based on local bacterial resistance pattern. Furthermore, much attention should be paid to some adverse efects like gastrointestinal disorders and hearing loss.

37 Chapter 2

SUPPLEMENTARY MATERIALS

Table S1-S5 and Figures S1-S4 are available as Supplementary data at JAC Online (https://doi.org/10.1093/jac/dky326)

38 Efects of prophylactic antibiotics on COPD

REFERENCES

1. Global Strategy for the Diagnosis, stable neutrophilic COPD: a double blind Management and Prevention of COPD, randomised, placebo controlled trial. Plos Global Initiative for Chronic Obstructive One. 2014;9(8):e105609. 2 Lung Disease (GOLD) 2017. Available from: 12. He ZY, Ou LM, Zhang JQ, et al. Efect of 6 months http://goldcopd.org. of erythromycin treatment on infammatory 2. Lopez AD, Shibuya K, Rao C, et al. Chronic cells in induced sputum and exacerbations obstructive pulmonary disease: current in chronic obstructive pulmonary disease. burden and future projections. Eur Respiration. 2010;80(6):445-452. Respir J. 2006;27(2):397-412. 13. Mygind LH, Pedersen C, Vestbo J et al. 3. Wedzicha JA, Seemungal TAR. COPD A randomized, placebo-controlled 3 years exacerbations: defning their cause and study of prophylactic azithromycin in 575 prevention. Lancet. 2007;370(9589):786-796. patients with chronic obstructive pulmonary 4. Sethi S, Murphy TF. Infection in disease (COPD). Abstract 36 Suppl 54: 1018s. the pathogenesis and course of chronic In: Abstracts of the European Respiratory obstructive pulmonary disease. N Engl J Society Annual Congress, Barcelona, Spain, Med. 2008;359(22):2355-2365. 2010. European Respiratory Society. 5. Staykova T, Black PN, Chacko EE, Poole P. 14. Suzuki T, Yanai M, Yamaya M, et al. Prophylactic antibiotic therapy for chronic Erythromycin and common cold in COPD. bronchitis. Cochrane Database Syst Rev. Chest. 2001;120(3):730-733. 2003(1):CD004105. 15. Banerjee D, Khair OA, Honeybourne D. 6. Herath SC, Poole P. Prophylactic antibiotic The efect of oral clarithromycin on health therapy for chronic obstructive pulmonary status and sputum bacteriology in stable disease (COPD). Cochrane Database Syst COPD. Respir Med. 2005;99(2):208-215. Rev. 2013(11):CD009764. 16. Brill SE, Law M, El-Emir E, et al. Efects of 7. Ni W, Shao X, Cai X, et al. Prophylactic use diferent antibiotic classes on airway bacteria of macrolide antibiotics for the prevention in stable COPD using culture and molecular of chronic obstructive pulmonary disease techniques: a randomised controlled trial. exacerbation: a meta-analysis. Plos Thorax. 2015;70(10):930-938. One. 2015;10(3):e0121257. 17. Uzun S, Djamin RS, Kluytmans JA, et al. 8. Cameron EJ, McSharry C, Chaudhuri Azithromycin maintenance treatment in R, Farrow S, Thomson NC. Long-term patients with frequent exacerbations of macrolide treatment of chronic chronic obstructive pulmonary disease infammatory airway diseases: risks, (COLUMBUS): a randomised, double-blind, benefts and future developments. Clin Exp placebo-controlled trial. Lancet Respir Allergy. 2012;42(9):1302-1312. Med. 2014;2(5):361-368. 9. Wedzicha JA, Calverley PMA, Albert RK, 18. Berkhof FF, Doornewaard-ten Hertog NE, et al. Prevention of COPD exacerbations: Uil SM, Kerstjens HA, van den Berg JW. a European Respiratory Society/ Azithromycin and cough-specifc health status American Thoracic Society guideline. Eur in patients with chronic obstructive pulmonary Respir J. 2017;50(3):1602265. disease and chronic cough: a randomised 10. Jones P. St George’s Respiratory controlled trial. Respir Res. 2013;14:125. Questionnaire Manual. Version 2.3. 2009: 1-6. 19. Albert RK, Connett J, Bailey WC, et al. 11. Simpson JL, Powell H, Baines KJ, et al. Azithromycin for prevention of exacerbations The efect of azithromycin in adults with of COPD. N Engl J Med. 2011;365(8):689-698.

39 Chapter 2

20. Seemungal TA, Wilkinson TM, Hurst JR, Perera COPD: a systematic literature review and WR, Sapsford RJ, Wedzicha JA. Long-term regression analysis. Respir Res. 2016;17:40. erythromycin therapy is associated with 29. Miravitlles M, Ferrer M, Pont A, et al. decreased chronic obstructive pulmonary Efect of exacerbations on quality of disease exacerbations. Am J Respir Crit Care life in patients with chronic obstructive Med. 2008;178(11):1139-1147. pulmonary disease: a 2 year follow up study. 21. Sethi S, Jones PW, Theron MS, et al. Thorax. 2004;59(5):387-395. Pulsed moxifoxacin for the prevention 30. Yamaya M, Azuma A, Takizawa H, Kadota J, of exacerbations of chronic obstructive Tamaoki J, Kudoh S. Macrolide efects on pulmonary disease: a randomized the prevention of COPD exacerbations. Eur controlled trial. Respir Res. 2010;11:10. Respir J. 2012;40(2):485-494. 22. Shafuddin E, Mills GD, Holmes MD, 31. Simoens S, Laekeman G, Decramer M. Poole PJ, Mullins PR, Black PN. A double- Preventing COPD exacerbations with blind, randomised, placebo-controlled macrolides: a review and budget impact study of roxithromycin and doxycycline analysis. Respir Med. 2013;107(5):637-648. combination, roxithromycin alone, or matching placebo for 12 weeks in adults 32. Serisier DJ. Risks of population antimicrobial with frequent exacerbations of chronic resistance associated with chronic macrolide obstructive pulmonary disease. J Negat use for infammatory airway diseases. Lancet Results Biomed. 2015;14:15. Resp Med. 2013;1(3):262-274. 23. Liu KX, Xu B, Wang J, et al. Efcacy and safety 33. Donath E, Chaudhry A, Hernandez-Aya LF, Lit of moxifoxacin in acute exacerbations of L. A meta-analysis on the prophylactic use chronic bronchitis and COPD: a systematic of macrolide antibiotics for the prevention review and meta-analysis. J Thorac of disease exacerbations in patients with Dis. 2014;6(3):221-229. Chronic Obstructive Pulmonary Disease. 24. Florescu DF, Murphy PJ, Kalil AC. Efects of Respir Med. 2013;107(9):1385-1392. prolonged use of azithromycin in patients 34. Soler-Cataluna JJ, Martinez-Garcia MA, with cystic fbrosis: a meta-analysis. Pulm Roman Sanchez P, Salcedo E, Navarro M, Pharmacol Ther. 2009;22(6):467-472. Ochando R. Severe acute exacerbations 25. Figueiredo Bde C, Ibiapina Cda C. and mortality in patients with chronic The role of macrolides in noncystic fbrosis obstructive pulmonary disease. bronchiectasis. Pulm Med. 2011;2011:751982. Thorax. 2005;60(11):925-931. 26. Rubin BK. Immunomodulatory properties of 35. Wedzicha JA, Decramer M, Ficker JH, et al. macrolides: overview and historical perspective. Analysis of chronic obstructive pulmonary Am J Med. 2004;117 Suppl 9A:2S-4S. disease exacerbations with the dual 27. Martinez FJ, Curtis JL, Albert R. Role of bronchodilator QVA149 compared with macrolide therapy in chronic obstructive glycopyrronium and tiotropium (SPARK): pulmonary disease. Int J Chron Obstruct a randomised, double-blind, parallel-group Pulmon Dis. 2008;3(3):331-350. study. Lancet Respir Med. 2013;1(3):199-209. 28. Martin AL, Marvel J, Fahrbach K, Cadarette 36. Wedzicha JA, Banerji D, Chapman KR, et SM, Wilcox TK, Donohue JF. The association al. Indacaterol-Glycopyrronium versus of lung function and St. George’s respiratory Salmeterol-Fluticasone for COPD. N Engl J questionnaire with exacerbations in Med. 2016;374(23):2222-2234.

CHAPTER 3 The infuence of age on real-life efects of doxycycline for acute exacerbations among COPD outpatients: a population-based cohort study

Yuanyuan Wang Jens H. Bos H. Marike Boezen Jan-Willem C. Alfenaar Job F.M. van Boven Catharina C.M. Schuiling-Veninga Bob Wilfert Eelko Hak

Publisher as: Wang Y, Bos JH, Boezen HM, et al. Infuence of age on real-life efects of doxycycline for acute exacerbations among COPD outpatients: a population-based cohort study. BMJ Open Respiratory Research 2020;7:e000535. doi: 10.1136/bmjresp-2019-000535. Chapter 3

ABSTRACT Introduction Although bacteria contribute signifcantly to acute exacerbations of COPD (AECOPD), the added value of antibiotics remains controversial, especially in outpatient settings. Age may afect antibiotic efectiveness, but real-world evidence is lacking. We aimed to assess the infuence of age on the efectiveness of doxycycline for AECOPD.

Methods A retrospective cohort study among outpatients with the frst recorded AECOPD treated with oral corticosteroids was conducted using a large pharmacy dispensing database. The primary outcome was treatment failure within 15 to 31 days after treatment start. Secondary outcome was time to second exacerbation. All analyses were stratifed by age groups.

Results We identifed 6300 outpatients with the frst AECOPD. 2261 (36%) received doxycycline and 4039 (64%) did not receive any antibiotic (reference group). Overall, there was no diference in treatment failure (adjusted OR 0.97, 95% CI 0.84 to 1.12) between two groups. Similarly, no diference in treatment failure was observed in younger groups. However, in patients with advanced age (≥ 75 years), treatment failure was signifcantly reduced by doxycycline compared with reference (16% vs 20%, adjusted OR 0.77, 95% CI 0.62 to 0.97). Overall, median time to second exacerbation was 169 days (95% CI 158 to 182 days) in doxycycline group compared with 180 days (95% CI 169 to 191 days) in reference group (adjusted HR 1.06, 95% CI 0.99 to 1.12), Although in older patients there was a trend within 3 months towards longer time of next exacerbation by doxycycline, it did not achieve statistical signifcance.

Conclusions Our fndings showed short-term treatment beneft of doxycycline added to oral corticosteroids for COPD patients with advanced age. This value remains unclear for persons aged under 75 years in current primary care. Long-term preventive benefts of doxycycline for the next exacerbation were not observed, irrespective of age.

44 Age Infuence on Efects of Doxycycline for AECOPD

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a chronic, progressive, infammatory disease and a leading cause of death worldwide.1 Acute exacerbations of COPD (AECOPD) characterized by the sudden worsening of respiratory symptoms may accelerate the progress of COPD and contribute signifcantly to worsened patients’ 3 health status, mortality and medical costs.2,3 As about 50% of AECOPD are triggered by bacterial infections,4 the use of antibiotics has become a common component in the therapeutic management of AECOPD.5,6

The evidence on the benefts of oral corticosteroids for AECOPD is of high quality.6,7 However, the efects of antibiotics in addition to corticosteroids are still uncertain, especially in an outpatient setting. A Cochrane review in 2012 did not show a signifcant reduced risk of treatment failure by antibiotics.8 Although treatment guidelines in 2017 conditionally recommended antibiotics for AECOPD among outpatients,6 this recommendation was based on synthesized evidence from only two earlier RCTs.9,10 In the same year, a new large RCT concluded that antibiotics for AECOPD in an outpatient setting are not efective.11 Later in 2018, an updated Cochrane review included two more RCTs than in 201211,12 and showed statistically signifcant benefcial efects of antibiotics.13 Of note, while most RCTs focused on the short-term efect of antibiotics, the long-term efect in outpatient settings also remains unclear due to conficting results.11,14,15

The majority of AECOPD is treated in primary care and establishing a bacterial infection diagnosis with sputum cultures is not always feasible in routine practice due to technical reasons.5,16 Therefore, accurate prescribing of antibiotics according to guidelines is still low.17-19 Notably, many studies indicate that the susceptibility to infections increases with age.20,21 According to a large population-based observational study, the protective efect of antibiotics against pneumonia is more pronounced in older patients.22 23 Thus, we hypothesized that older patients may beneft more from empirical antibiotic treatment for AECOPD than younger patients.

In addition to prednisone or prednisolone, doxycycline is one of the frst-choice oral antibiotics for AECOPD if antibiotic treatment is indicated.24 5,16 Since only one RCT studied doxycycline, we conducted a cohort study to evaluate if doxycycline has meaningful value added to oral corticosteroids on AECOPD in both the short- and longer-term for outpatients, and examined the potential efect modifcation across age groups.

45 Chapter 3

METHODS Study design and data source We applied a retrospective inception cohort study (Figure 1) using the University of Groningen’s prescription database IADB.nl that contains over 1.2 million dispensings from about 600,000 patients in 60 community pharmacies in the Netherlands since 1994.25,26 IADB.nl provided complete information including date of birth, gender, prescribed drug name, ATC codes, dispensing date, quantity dispensed, and dose regimen.27 Over-the-counter (OTC) drugs and drugs dispensed during hospitalization are not available in the database. As Dutch patients practically always register at a single community pharmacy, the patient’s drug prescription history is usually complete.28 Data from January 1994 to December 2015 were used for this study, which was conducted and reported according to checklists of STROBE guidelines (Supplementary material).

Studycorticosteroids population on AECOPD in both the short‐ and longer‐term for outpatients, and examined the COPDpotential outpatients effect modification with frst across recorded age groups. AECOPD were included in this study. We selected eligible patients according to the following inclusion criteria: (1) Presence of COPD identiMethodsfed based on at least two COPD-related drug prescriptions (see Table S1) within oneStudy year design before and index data source date.5,24 The date of frst recorded AECOPD during the study period wasWe setapplied as index a retrospective date. (2) The inception experience cohort of study the (Figurefrst recorded 1) using AECOPD,the University which of Groningen’s was defned by the prescription of high dose prednisone (ATC-code H02AB07) or prednisolone prescription database IADB.nl that contains over 1.2 million dispensings from about 600,000 (H02AB06) short courses (a daily dose of 40 mg for 5 days or a daily dose of 30 mg for 25,26 7patients days with in 60maximum community extension pharmacies of in14 thedays) Netherlands according since to treatment1994. IADB.nl guidelines. provided14,24,29 (3)complete Registration information in the includingIADB.nl fordate at leastof birth, two gender, years before prescribed and onedrug yearname, after ATC the codes, index date.dispensing (4) Receipt date, quantity of doxycycline dispensed, or and either dose received regimen. 27any Over antibiotics‐the‐counter 3 days(OTC) before drugs andtill 7 daysdrugs after dispensed the index during date. hospitalization Furthermore, are notwe availableexcluded in patients the database. who metAs Dutch the followingpatients exclusion criteria: (1) Receipt of another antibiotic treatment than doxycycline 3 days practically always register at a single community pharmacy, the patient's drug prescription history before till 7 days after the index date. (2) Age under 55 years, to reduce the chance is usually complete.28 Data from January 1994 to December 2015 were used for this study, which of including possible asthma patients.30,31 Age was calculated using the diference betweenwas conducted index and date reported and birth according date. (3)to checklists Presence of of STROBE potential guidelines immunocompromised (Supplementary material).

FigureFigure 1. 1. Retrospective Retrospective cohort cohort study study design. design.

46Study population COPD outpatients with first recorded AECOPD were included in this study. We selected eligible patients according to the following inclusion criteria: (1) Presence of COPD identified based on at least two COPD‐related drug prescriptions (see Table S1) within one year before index date.5,24 The date of first recorded AECOPD during the study period was set as index date. (2) The Age Infuence on Efects of Doxycycline for AECOPD diseases, which were defned by the prescription of antiviral drugs for HIV infection, immunosuppressant drugs or antineoplastic agents within one year before index date and one month after index date.

Exposure and outcomes Among patients with a frst identifed AECOPD, during their treatment period of oral 3 prednisone or prednisolone (3 days before till 7 days after the index date), those who were also prescribed doxycycline and no other antibiotics were classifed as treatment exposure group. Those who did not receive doxycycline (or any other antibiotic) were classifed as reference group. The primary outcome was treatment failure defned as a new prescription of prednisone or prednisolone or an antibiotic treatment within a period of 15 to 31 days after index date according to Dutch NHG guidelines for COPD management. Secondary outcome was time to the second exacerbation within a follow-up period of 12 months. As the frst exacerbation may last for a longer time, and to avoid counting it as second exacerbation, we limited the minimum time from frst exacerbation to second one to 21 days.32 A few patients could be included in the treatment failure outcome and the second exacerbation outcome if the drugs appeared within 21 and 31 days after index date.

Covariates The following covariates were included as potential confounders: age; gender; frequently used maintenance drugs for COPD treatment within 365 days before index date including SABA, SAMA, LABA, LAMA, SABA/SAMA, LABA/LAMA, LABA/ICS and theophylline. Comorbidities in COPD patients were defned on the basis of at least two prescriptions of related drugs within 365 days prior to index date: diabetes (A10), heart failure (C01AA05 or C03C), ischemic heart disease (C01DA), other cardiovascular disease (C02 or C03 or C07 or C08 or C09, but not for C01AA05, C03C and C01DA), dyslipidaemia (C10), osteoporosis (M05B), anxiety (N05B, N05C), dementia (N06D), depression (N06A), rheumatic arthritis (M01 or M02), hypothyroid disease (H03).30

Statistical methods The diferences in distribution of baseline characteristics of COPD outpatients between two exposure groups were compared using t-test and Chi-square test for continuous and categorical variables, respectively. We applied logistic regression to estimate the odds ratio (OR) with 95% confdence interval (CI) for treatment failure and adjusted for possible covariates. The time to second exacerbation was compared by Kaplan- Meier survival analysis. Cox proportional hazards regression was applied to estimate the hazard ratio (HR) and 95% CI for risk of second exacerbation. For all tests, p-values were 2-sided. A p-value < 0.05 was considered statistically signifcant. All analyses were performed using IBM SPSS statistics 22 (IBM Corp., Armonk, NY, USA).

47 Chapter 3

Sensitivity analysis To further assess the robustness of our results, we performed several sensitivity analyses. Treatment failure was defned by the use of prednisone, prednisolone or antibiotics according to Dutch guidelines.24 However, considering that not all antibiotics are used for acute exacerbations, we narrowed the outcome defnition by including only frequently prescribed antibiotics among COPD patients in Netherlands (see Table S2) based on frequencies in the IADB database and previous published paper.14,27 In addition, we further narrowed the defnition of treatment failure by including prednisone or prednisolone only to see if there is any diference with defnition by including antibiotics only. Thirdly, considering the COPD treatment may change during the long period of study time, we did a sensitivity analysis by limiting our study period to the last 10 years and compared the result with those from previous decade.

RESULTS Study participants In total, 8,889 COPD patients with a frst recorded AECOPD were identifed, all received prednisone or prednisolone. Of those, we excluded 2,589 patients who were prescribed another antibiotic than doxycycline, i.e. our exposure of interest. Of the remaining 6,300 patients, 2,261 patients who received doxycycline were included as treatment group, and the remaining 4,039 patients who did not receive any antibiotic were included as reference group (see Figure 2).

The baseline characteristics of both comparison groups are summarized in Table 1. The two groups were balanced for most characteristics. However, the mean age in the doxycycline group was slightly higher than the reference group. A little higher prevalence of LABA/ICS and doxycycline prescriptions and lower prevalence of prescriptions of SABA were seen in the doxycycline group compared with reference.

Primary outcome Between 15 and 31 days after the frst exacerbation, 354 (15.7 %) patients in the doxycycline group versus 640 (15.8 %) patients in the reference group had treatment failure (crude OR 0.99 [95% CI: 0.89 to 1.14], Table 2). After adjustment for potential confounders, there still was no statistical diference between the two groups regarding the rate of treatment failure, the adjusted OR (aOR) of treatment failure was 0.97 [95% CI: 0.84 to 1.12].

In the analysis stratifed by age groups, there was no signifcant diference in the rate of treatment failure between the two treatments for age groups below 75 years old. However, for COPD outpatients aged 75 years and older, less patients in the doxycycline group experienced treatment failure than in the reference group (16.1% versus 19.9%,

48 Study participants In total, 8,889 COPD patients with a first recorded AECOPD were identified, all received prednisone or prednisolone. Of those, we excluded 2,589 patients who were prescribed another antibiotic than doxycycline, i.e. our exposure of interest. Of the remaining 6,300 patients, 2,261

patients who received doxycycline were included Ageas treatment Infuence on gr Eoup,fects and of Doxycycline the remaining for AECOPD 4,039 patients who did not receive any antibiotic were included as reference group (see Figure 2).

FigureFigure 2. 2. Flow Flow chart chart of of participation participation selection. selection.

OR 0.78 [95% CI: 0.62, 0.97]). After adjustments for possible confounders, the value of OR did not change much, and results were compatible with a 23% relative risk reduction of treatment failure observed by doxycycline treatment compared with reference group (aOR 0.77 [95% CI: 0.62 to 0.97]).

Secondary outcome After a follow-up of 12 months, 71.4% and 67.9% COPD outpatients experienced the next exacerbation in doxycycline and reference groups, respectively. The median time to next exacerbation was 169 days [95% CI 156-182] in the doxycycline group compared with 180 days [95% CI 169-191] in the reference group (p=0.07, Figure 3). However, if we included only those patients who experienced a second exacerbation within 12 months follow up, the median time was longer in the doxycycline group than in the reference

49 Chapter 3

Table 1. Baseline characteristics of COPD outpatients with frst exacerbation in treatment groups.

Doxycycline Reference (n=2261) (n= 4039) P-value

Gender, no. (%) Men 1085 (48.0) 1999 (49.5) 0.252 Female 1176 (52.0) 2040 (50.5) Age, years, no. (%) Mean age (SD) 71.08 (9.6) 70.30 (9.4) 0.002# 55-64 667 (29.5) 1285 (31.8) 0.018* 65-74 733 (32.4) 1357 (33.6) ≥75 861 (38.1) 1397 (34.6) Year of index date (%) 1996-2004 893 (39.5) 1676 (41.5) 0.121 2005-2015 1368 (60.5) 2363 (58.5) Prescriber GP 2147 (95.0) 3424 (84.8) <0.001 Specialist 114 (5.0) 615 (15.2) Maintenance medicines, no. (%) SABA 775 (34.3) 1579 (39.1) <0.001 LABA 494 (21.8) 847 (21.0) 0.414 SAMA 689 (30.5) 1216 (30.1) 0.761 LAMA 555 (24.5) 1020 (25.3) 0.534 SABA/SAMA 80 (3.5) 173 (4.3) 0.149 LABA/LAMA 0 (0) 1 (0) 0.454 LABA/ICS 1093 (48.3) 1846 (45.7) 0.044 Theophylline 124 (5.5) 159 (3.9) 0.004 Comorbidity, no. (%) Diabetes mellitus 301 (13.3) 504 (12.5) 0.341 Disorders of lipid metabolism 629 (27.8) 1093 (27.1) 0.517 Heart failure 363 (16.1) 676 (16.7) 0.484 Ischemic heart disease 206 (9.1) 336 (8.3) 0.282 Other cardiovascular disorders 843 (37.3) 1493 (37.0) 0.801 Thyroid disease 115 (5.1) 192 (4.8) 0.556 Rheumatic arthritis 355 (15.7) 660 (16.3) 0.508 Osteoporosis 117 (5.2) 232 (5.7) 0.343 Anxiety 392 (17.3) 649 (16.1) 0.193 Depression 274 (12.1) 438 (10.8) 0.125 Dementias 9 (0.4) 10 (0.2) 0.296

#Student’s t-test; SD: standard deviation; *Pearson Chi-Square test;

50 Age Infuence on Efects of Doxycycline for AECOPD

Table 2. Odds ratio for treatment failure of frst exacerbation among COPD outpatients in diferent age groups.

Doxycycline Reference Crude OR Adjusted OR* (n=2261) (n=4039) (95% CI) (95% CI) Treatment failure (n, %) 3 Overall 354 (15.7) 640 (15.8) 0.99 [0.86, 1.14] 0.97 [0.84, 1.12] Subgroups 55-65 99 (14.8) 166 (12.9) 1.18 [0.90, 1.54] 1.17 [0.89, 1.53] 65-75 116 (15.8) 196 (14.4) 1.11 [0.87, 1.43] 1.11 [0.86, 1.42] ≥75 139 (16.1) 278 (19.9) 0.78 [0.62, 0.97] 0.77 [0.62, 0.97]

OR = odds ratio; CI = confdence interval; *Adjusted for age, SABA, LABA/ICS, theophylline. group, though it was not statistically signifcant (97 days [95% CI 91-103] versus 91 days [95% CI 86-96], p=0.128).

When the results were stratifed according to diferent age groups, we did not fnd signifcant diferences, although older people (aged 65-74 and ≥75) on doxycycline experienced a lower risk of next exacerbation than the reference group at early time points (within 3 months) of the follow up (Figure 3). However, we found that in both treatment groups, the median time to second exacerbation was shorter (p<0.01) in older age groups compared with younger age groups (Table S3 and Figure S1).

Overall, around 30%, 50% and 70% patients in both treatment groups experienced a new exacerbation in the 3, 6 and 12 months follow-up, respectively (Table S4). From the univariate Cox regression model, the risk for the next exacerbation was similar between two treatment groups, the HR (doxycycline versus reference) was 1.00 [95% CI 0.9-1.09], 1.03 [95% CI 0.96-1.11] and 1.07 [95% CI 1.00-1.14] in 3, 6 and 12 months follow up. Similar results were observed after adjusting for potential confounding factors, the HR was 0.98 [95% CI 0.89-1.07], 1.02 [95% CI 0.95-1.09] and 1.06 [95% CI 0.99-1.12], respectively.

Sensitivity analysis When we further defned the primary outcome of treatment failure including only frequently used antibiotics, it showed consistent results in that doxycycline did not reduce treatment failure for the overall cohort (aOR 0.99 [0.85, 1.14]), but that doxycycline treatment showed benefts in patients 75 years or older with 137 patients (15.9 %) and 268 patients (19.2 %) that experienced treatment failure in the doxycycline group and the reference group, respectively (aOR 0.80 [0.63, 1.00]). (Table S5) When we further narrow our treatment failure defnition to a new prescription of prednisone or prednisolone, we also observed reduced treatment failure by doxycycline in the older

51 Chapter 3

Figure 3. Kaplan-Meier curves showing the proportion of patients free of 2nd exacerbation in COPD Figure 3. Kaplan‐Meier curves showing the proportion of patients free of 2nd exacerbation in outpatients up to 12 months’ follow up: a) all-age group patients (p = 0.07); b) patients aged 55-64 COPD(p=0.252); outpatients c) patients up aged to 12 65-74 months' (p=0.564); follow d) up:patients a) all aged‐age ≥group75 (p=0.421). patients (p = 0.07); b) patients aged 55‐64 (p=0.252); c) patients aged 65‐74 (p=0.564); d) patients aged 75 (p=0.421). age group compared with reference (aOR 0.72 [0.55, 0.95], Table S6), while no signifcant Sensitivity analysis diference was observed between groups for the narrow defnition of treatment failure Whenby a new we furtheprescriptionr defined of the antibiotics. primary outcome When we of limittreatment the study failure period including to theonly last frequently decade (2005-2015) and the previous decade (1994-2004) separately, the treatment failure was used antibiotics, it showed consistent results in that doxycycline did not reduce treatment failure also less among patients with advanced age in doxycycline group than reference group. for the overall cohort (aOR 0.99 [0.85, 1.14]), but that doxycycline treatment showed benefits in (aOR 0.75 [0.55, 1.01] and aOR 0.84 [0.60, 1.18], separately, Table S7). patients 75 years or older with 137 patients (15.9 %) and 268 patients (19.2 %) that experienced DISCUSSION Main fndings In a real-world population of primary care patients with AECOPD of any age, doxycycline did not appreciably reduce the failure rate, nor prolong time to next exacerbation. However, when stratifed by age, we found a statistically signifcant 23% relative

52 Age Infuence on Efects of Doxycycline for AECOPD reduction in treatment failure by doxycycline for AECOPD in outpatients aged 75 years and older. These benefts were not seen in younger age groups. In the long-term, we observed that the protective efect of doxycycline for the subsequent exacerbations was only present in the frst 3 months among older patients. After that, the protective efect wanes over time. 3 The observed short-term efect regarding reduction rate in treatment failure for older patients (≥75 years) is compatible with a previous RCT which found that short-term treatment non-response rates are signifcantly lower in the doxycycline group compared to placebo (OR 0.77, 95% CI [0.63, 0.94]).11 Our sub-group result is also consistent with a recent Cochrane review that showed that the current available antibiotics reduce the risk for treatment failure between seven days and one month after treatment initiation (OR 0.72, 95% CI [0.56, 0.94]).13

The beneft of doxycycline in older patients may be primarily due to their increased susceptibility to infection.20 With increasing age, not only the lung function changes, the natural defense mechanisms of the lungs also decrease gradually.33 Intercellular communications become less efective which could contribute to immune-senescence.34 Additionally, mucocilliary clearance is also compromised with age.35 All these changes with age contribute to the greater possibility of bacterial infection and infammation in elderly.20 Therefore, the elderly seem to beneft more from antibiotic treatment than younger patients.

The average age of study patients (about 70 years) was comparable with previous studies.36 We did not fnd a short-term beneft of doxycycline in the younger age group (< 75), which may be explained by the fact that the overall rate of appropriate antibiotic use in practice is rather low.17-19 According to GOLD, general practitioners should only consider antibiotics for patients when signs of bacterial infection are present.5 However, in reality, guidelines regarding the prescription of antibiotics are poorly followed,17,19 on average in only 25% of AECOPD antibiotics were prescribed properly according to the GOLD criteria.18 For AECOPD with other etiology like viral infection and environmental pollution, antibiotics may not have been efective. Of note, a complicating factor in the outpatient setting is that sputum cultures are not feasible as they take at least two days and frequently do not give reliable results.5 Identifcation of bacterial exacerbation still relies on clinical assessment rather than laboratory biomarkers.37 As infection is the most treatable cause of breathlessness, it is not surprising that many patients continue to receive antimicrobials in the absence of clinical, pathological or radiological evidence of infection.38 Therefore, if the proportion of patients who were prescribed doxycycline but in fact should not be given antibiotics is large, it will be difcult to fnd signifcant benefcial efects of doxycycline treatment.

53 Chapter 3

Observed long-term efects from this study for all patients independent of age were also consistent with the fndings of two previous RCTs that antibiotics did not prolong time to next exacerbation.11,39 However, two observational studies showed diferent results in that the time to next exacerbation was signifcantly extended if the exacerbation was treated with antibiotics.14,15 Similarly, one RCT also showed a pronged time to next exacerbation by antibiotic treatment.9 In this study, the prolonged time to next exacerbation by doxycycline was only seen in older outpatients within 3 months. Of note, as diferent defnitions for subsequent exacerbation and diferent types of antibiotics were used in these studies when evaluating the long-term efect of antibiotics, these may led to the inconsistent results.

Besides the efects of ageing on bacterial susceptibility,22 we should also realize that COPD itself is an age-related chronic infammatory disorder. After the lungs reach their maximum function around the age of 25 years, its function progressively declines as a sequence of structural and physiological changes to the lung.33 With ageing, severity and comorbidities of COPD usually also increase. These factors could further infuence the frequency of exacerbations in primary care patients with COPD.40 A higher frequency also means a shorter time to experience the next exacerbation. In this study we have found that the time to next exacerbation was shorter in older than younger patients, and it was consistent in both doxycycline and reference groups.

Strengths and limitations This study has several strengths. One strength is that this study was based on a large real-life prescription database which enabled us to evaluate the efects of doxycycline in a large COPD population. Another strength is that both short-term and long-term efects of additional doxycycline were evaluated, which may ofer more comprehensive support for decision making in clinical practice. Additionally, we chose the frst recorded exacerbation as investigated event for all COPD outpatients, which could exclude the infuence of historical exacerbation frequency as a risk factor on targeted outcomes to a large extent. In addition, as oral steroids and antibiotics cannot be bought over the counter in the Netherlands, the study population from the IADB database represents a generalizable population for AECOPD outpatients treated with doxycycline.

Limitations to observational studies also need to be discussed. First, due to the characteristics of the prescription database, there was no diagnostic information available. Therefore, the defnition of COPD, comorbidities and outcomes were defned using related drugs as proxies, which may result in potential misclassifcation bias. Secondly, although the relevant measured baseline characteristics of the two groups were similar in this study, other clinical information like lung function, GOLD stages (I-IV) of COPD and severity of exacerbations were lacking, which may infuence our outcome to some extent if these unknown characteristics were not balanced between the two

54 Age Infuence on Efects of Doxycycline for AECOPD study groups. In clinical practice, antibiotics may be prescribed to those who in fact did not have enough indication of infection due to limitation of outpatients setting or to those with more severe COPD,5 which may have led to underestimation of the efcacy of additional doxycycline treatment in all age groups compared to corticosteroids only. Thirdly, there were overlap for a few patients within 21 days and 31 days between the short- and long- term outcome defnitions by a new prescription of corticosteroids 3 due to lack of clinical information to distinguish and classify the outcomes. Finally, although we set the age limitation of 55 years older to exclude potential asthma, asthma-COPD overlap patients may still existed as we did not exclude the patients who use asthma drugs at the stage of study design. However, these patients were very few and unlikely to infuence the overall results based on the fact that no patients were prescribed leukotriene receptor antagonists and only 11 patients were prescribed cromoglycates within one year before index date among all the AECOPD patients in our study.

Implications for future research and clinical practice The tendency towards better efects of antibiotics in the elderly COPD patients may ofer clues for clinicians and researchers for more targeted management of AECOPD. In particular, decision making about empirical antibiotic therapy for AECOPD should take the age of patients into consideration. However, before that, more prospective, well-designed studies with more accurate diagnostic information are needed to further confrm the fnding from this study.

Although related guideline and GOLD report about antibiotic use for AECOPD were basically based on secondary care RCT evidence,5,6 decision making in daily practice is infuenced by many factors making AECOPD treatment more challenging in outpatient settings.5 Therefore, identifying high risk populations for infection may improve management and clinical decisions about antibiotic use in COPD outpatients. The high risk of infection and benefcial efects from antibiotics for AECOPD in elderly outpatients should warrant a personalized approach towards antibiotic treatment.

CONCLUSION

Doxycycline in addition to oral corticosteroid treatment was associated with a reduced risk of treatment failure for AECOPD in patients 75 years or older, but not in younger patients. Long-term efects of doxycycline treatment on subsequent exacerbations was not observed, though among older persons there was a non-statistically signifcant benefcial trend within 3 months after doxycycline treatment. Clinicians should take the age of patients into consideration in empirical antibiotic therapy for AECOPD. More real-world studies with high quality, preferably prospective clinical data collections, should be recommended to confrm the infuence of age on efects of antibiotics and

55 Chapter 3 to further explore which patient groups could beneft most from antibiotic treatment for AECOPD. SUPPLEMENTARY DATA

Table S1 to S7 and Figure S1 are available as Supplementary data at BMJ Open Respiratory Research online (https://bmjopenrespres.bmj.com/content/7/1/e000535)

56 Age Infuence on Efects of Doxycycline for AECOPD

REFERENCES

1. Lozano R, Naghavi M, Foreman K, et al. 10. Anthonisen NR, Manfreda J, Warren CP, Global and regional mortality from 235 Hershfeld ES, Harding GK, Nelson NA. causes of death for 20 age groups in Antibiotic therapy in exacerbations of 1990 and 2010: a systematic analysis for chronic obstructive pulmonary disease. Ann the Global Burden of Disease Study 2010. Intern Med. 1987;106(2):196-204. 3 Lancet. 2012;380(9859):2095-2128. 11. van Velzen P, Ter Riet G, Bresser P, et al. 2. Soler-Cataluna JJ, Martinez-Garcia MA, Roman Doxycycline for outpatient-treated acute Sanchez P, Salcedo E, Navarro M, Ochando R. exacerbations of COPD: a randomised Severe acute exacerbations and mortality in double-blind placebo-controlled trial. patients with chronic obstructive pulmonary Lancet Respir Med. 2017;5(6):492-499. disease. Thorax. 2005;60(11):925-931. 12. Hassan WA, Shalan I, Elsobhy M. Impact of 3. O’Reilly JF, Williams AE, Rice L. Health antibiotics on acute exacerbations of COPD. status impairment and costs associated Egypt J Chest Dis Tu. 2015;64(3):579-585. with COPD exacerbation managed in 13. Vollenweider DJ, Frei A, Steurer-Stey CA, hospital. International journal of clinical Garcia-Aymerich J, Puhan MA. Antibiotics practice. 2007;61(7):1112-1120. for exacerbations of chronic obstructive pulmonary disease. Cochrane Database Syst 4. Sethi S, Murphy TF. Infection in Rev. 2018;10:CD010257. the pathogenesis and course of chronic obstructive pulmonary disease. N Engl J 14. Roede BM, Bresser P, Bindels PJ, et al. Med. 2008;359(22):2355-2365. Antibiotic treatment is associated with reduced risk of a subsequent exacerbation 5. Global Initiative for Chronic Obstructive in obstructive lung disease: an historical Pulmonary Disease (GOLD). Global Strategy population based cohort study. for the Diagnosis, Management and Thorax. 2008;63(11):968-973. Prevention of COPD. 2019. Available from: 15. Roede BM, Bresser P, Prins JM, Schellevis F, Verheij http://goldcopd.org. TJ, Bindels PJ. Reduced risk of next exacerbation 6. Wedzicha JAEC-C, Miravitlles M, Hurst JR, et al. and mortality associated with antibiotic use in Management of COPD exacerbations: a European COPD. Eur Respir J. 2009;33(2):282-288. Respiratory Society/American Thoracic Society 16. Team NGU. National Institute for Health and guideline. Eur Respir J. 2017;49(3). Care Excellence: Clinical Guidelines. Chronic 7. Walters JA, Tan DJ, White CJ, Gibson PG, Wood- obstructive pulmonary disease in over 16s: Baker R, Walters EH. Systemic corticosteroids diagnosis and management. London: National for acute exacerbations of chronic Institute for Health and Care Excellence (UK). obstructive pulmonary disease. Cochrane Copyright (c) NICE 2018.; 2018. Database Syst Rev. 2014(9):CD001288. 17. Bathoorn E, Groenhof F, Hendrix R, et al. 8. Vollenweider DJ, Jarrett H, Steurer-Stey CA, Real-life data on antibiotic prescription Garcia-Aymerich J, Puhan MA. Antibiotics and sputum culture diagnostics in acute for exacerbations of chronic obstructive exacerbations of COPD in primary care. Int J pulmonary disease. Cochrane Database Syst Chron Obstruct Pulmon Dis. 2017;12:285-290. Rev. 2012;12:CD010257. 18. Dietrich E, Klinker KP, Li J, Nguyen CT, Quillen 9. Llor C, Moragas A, Hernandez S, Bayona C, D, Davis KA. Antibiotic Stewardship for Miravitlles M. Efcacy of antibiotic therapy Acute Exacerbation of Chronic Obstructive for acute exacerbations of mild to moderate Pulmonary Disease. Am J Ther. 2018. chronic obstructive pulmonary disease. Am 19. Roede BM, Bindels PJ, Brouwer HJ, Bresser J Respir Crit Care Med. 2012;186(8):716-723. P, de Borgie CA, Prins JM. Antibiotics and

57 Chapter 3

steroids for exacerbations of COPD in primary studies: an easy to obtain and reliable tool. care: compliance with Dutch guidelines. Br J Pharmacoepidemiol Drug Saf. 2002;11(5):379-384. Gen Pract. 2006;56(530):662-665. 29. Woods JA, Wheeler JS, Finch CK, Pinner 20. Brandsma CA, de Vries M, Costa R, Woldhuis NA. Corticosteroids in the treatment of RR, Konigshof M, Timens W. Lung ageing and acute exacerbations of chronic obstructive COPD: is there a role for ageing in abnormal pulmonary disease. Int J Chron Obstruct tissue repair? Eur Respir Rev. 2017;26(146). Pulmon Dis. 2014;9:421-430. 21. Gardner ID. The Efect of Aging on 30. van Boven JF, van Raaij JJ, van der Galien R, Susceptibility to Infection. Rev Infect et al. Impact of multiple-dose versus single- Dis. 1980;2(5):801-810. dose inhaler devices on COPD patients’ 22. Petersen I, Johnson AM, Islam A, Duckworth persistence with long-acting beta(2)- G, Livermore DM, Hayward AC. Protective agonists: a dispensing database analysis. efect of antibiotics against serious NPJ Prim Care Respir Med. 2014;24:14069. complications of common respiratory tract 31. Penning-van Beest F, van Herk-Sukel M, infections: retrospective cohort study with Gale R, Lammers JW, Herings R. Three- the UK General Practice Research Database. year dispensing patterns with long-acting BMJ. 2007;335(7627):982. inhaled drugs in COPD: a database analysis. 23. Stone RA, Lowe D, Potter JM, Buckingham Respir Med. 2011;105(2):259-265. RJ, Roberts CM, Pursey NJ. Managing 32. Seemungal TA, Donaldson GC, Bhowmik A, patients with COPD exacerbation: does age Jefries DJ, Wedzicha JA. Time course and matter? Age Ageing. 2012;41(4):461-468. recovery of exacerbations in patients with 24. Snoeck-Stroband JB, Schermer TRJ, Van Schayck chronic obstructive pulmonary disease. Am CP, et al. NHG-Werkgroep Astma bij volwassenen J Respir Crit Care Med. 2000;161(5):1608-1613. en COPD. NHG-Standaard COPD (derde 33. Rojas M, Meiners S, Le Saux CJ. Molecular herziening). Huisarts Wet 2015; 58(4):198-211. aspects of aging : understanding lung 25. Bahar MA, Wang Y, Bos JHJ, Wilfert B, Hak aging. Hoboken, New Jersey: John Wiley & E. Discontinuation and dose adjustment Sons, Inc.; 2014. of metoprolol after metoprolol- 34. Lopez-Otin C, Blasco MA, Partridge L, paroxetine/fuoxetine co-prescription in Serrano M, Kroemer G. The hallmarks of Dutch elderly. Pharmacoepidemiol Drug aging. Cell. 2013;153(6):1194-1217. Saf. 2018;27(6):621-629. 35. Bailey KL, Bonasera SJ, Wilderdyke M, et al. 26. Mulder B, Pouwels KB, Schuiling-Veninga CC, Aging causes a slowing in ciliary beat frequency, et al. Antibiotic use during pregnancy and mediated by PKCepsilon. Am J Physiol Lung Cell asthma in preschool children: the infuence of Mol Physiol. 2014;306(6):L584-589. confounding. Clinical and experimental allergy : journal of the British Society for Allergy and 36. Balcells E, Anto JM, Gea J, et al. Clinical Immunology. 2016;46(9):1214-1226. Characteristics of patients admitted for the frst time for COPD exacerbation. Respir 27. Visser ST, Schuiling-Veninga CC, Bos JH, de Jong-van den Berg LT, Postma Med. 2009;103(9):1293-1302. MJ. The population-based prescription 37. Jacobs DM, Pandit U, Sethi S. Acute database IADB.nl: its development, exacerbations in chronic obstructive usefulness in outcomes research and pulmonary disease: should we use antibiotics challenges. Expert Rev Pharmacoecon and if so, which ones? Curr Opin Infect Dis. 2019. Outcomes Res. 2013;13(3):285-292. 38. Taverner J, Ross L, Bartlett C, et al. 28. Monster TB, Janssen WM, de Jong PE, de Antimicrobial prescription in patients dying Jong-van den Berg LT, REnal PSGPo, Vascular from chronic obstructive pulmonary disease. ENTSD. Pharmacy data in epidemiological Internal medicine journal. 2019;49(1):66-73.

58 Age Infuence on Efects of Doxycycline for AECOPD

39. Wilson R, Jones P, Schaberg T, et al. Antibiotic treatment and factors infuencing short and long term outcomes of acute exacerbations of chronic bronchitis. Thorax. 2006;61(4):337-342. 40. Westerik JA, Metting EI, van Boven JF, Tiersma W, Kocks JW, Schermer TR. Associations 3 between chronic comorbidity and exacerbation risk in primary care patients with COPD. Respir Res. 2017;18(1):31.

59 CHAPTER 4 Real-world short- and long-term efects of antibiotic therapy on acute exacerbations of COPD in outpatients: a cohort study under the PharmLines Initiative

Yuanyuan Wang Victor Pera H. Marike Boezen Jan-Willem C. Alfenaar Bob Wilfert Rolf H.H. Groenwold Eelko Hak

Submitted for publication. Chapter 4

ABSTRACT Introduction Although antibiotic treatment is recommended for acute exacerbations of COPD (AECOPD) under specifc circumstances, the treatment’s value in an outpatient setting is unclear. We aim to evaluate the real-world short- and long-term efects of antibiotic treatment on AECOPD outpatients.

Methods This retrospective inception cohort study was conducted under the PharmLines Initiative that linked the Lifelines cohort study database with IADB.nl, the University of Groningen’s medication prescription database during the period 2005-2017. We included participants with a frst recorded diagnosis of COPD who underwent systemic glucocorticoids treatment for an AECOPD episode. The exposed and reference group was defned as patients who received and not received any antibiotic prescriptions during AECOPD treatment. The short-term outcome was AECOPD treatment failure within 14-30 days after the index date. The long-term outcome was the time to the next exacerbation within a one-year follow-up period. Binary logistic regression analysis combined with propensity scores (PS) were used to estimate the association between antibiotic use and treatment failure. The risk of another exacerbation was assessed using Kaplan-Meier survival and Cox regression analyses. Several subgroup and sensitivity analyses were also performed.

Results We analyzed linked data for 1,105 AECOPD patients. Antibiotics were prescribed for 518 patients (46.9%) while 587 patients (53.1%) received no antibiotics. The overall antibiotic use was associated with a signifcant relative risk reduction of AECOPD treatment failure by 33%-37% compared with the risk for the reference group (adjusted odds ratio [aOR]: 0.67 [95% CI: 0.42-1.04] and 0.63 [95% CI: 0.40-0.99] by regression and PS analyses, respectively). Similar protective efects were observed for doxycycline, macrolides and co-amoxiclav, but not for amoxicillin, and only the efects of doxycycline were statistically signifcant (aOR 0.53 [95% CI: 0.28-0.99] by PS analysis). There was no diference between the exposure and reference groups regarding the risk of and time to the next exacerbation, irrespective of the follow-up duration.

Conclusion Our fnding that antibiotics treatment supplementing systemic glucocorticoids treatment reduces short-term AECOPD treatment failure in real-world settings concurs with those of clinical trials. Larger studies of high quality are needed to confrm the benefcial efects for specifc classes of antibiotics. Our results after controlling for confounding suggest that in observational studies on AECOPD, unmeasured confounding may induce underestimated benefcial antibiotic treatment efects. 62 Efects of Antibiotic Therapy on AECOPD

INTRODUCTION

Chronic obstructive pulmonary disease (COPD), which is characterized by persistent respiratory symptoms and airfow limitation is one of the leading causes of morbidity and mortality worldwide.1,2 COPD patients frequently experience acute exacerbations of COPD (AECOPD), defned as acute worsening of respiratory symptoms, necessitating additional therapy.3 AECOPD has major impacts on patients’ health status, accelerates the disease progression and increases health-care costs.4,5 Therefore, reducing 4 the symptoms of current exacerbations and preventing further exacerbations are essential aspects of sound pharmaceutical management of AECOPD. Because AECOPD is associated with increased airway infammations, systemic glucocorticoids treatment is recommended to shorten the recovery time, improve lung function and promote oxygenation, given its proven benefcial efects.3,6

The majority of AECOPD episodes are caused by respiratory infections, especially bacterial infections, which account for around 50% of all exacerbations.7,8 The most widely reported bacteria associated with exacerbations are S. pneumoniae, H. infuenza, P. aeruginosa, M. catarrhalis, A. baumannii, and S. aureus.7,9,10 Accordingly, antibiotics have been recommended for the management of AECOPD when signs of bacterial infection are present.3 However, the benefcial efects of antibiotic treatment in addition to oral glucocorticoids for AECOPD are still uncertain among outpatients. The pooled results from fve randomized controlled trials (RCTs) examined in a Cochrane meta-analysis conducted in 2012 did not show a signifcant reduced risk of treatment failure associated with currently prescribed antibiotics among outpatients.11 However, an updated (2018) version of this Cochrane review that included two new RCTs, presented statistically signifcant benefcial efects of current prescribed antibiotics among outpatients.12,13 Especially the RCT conducted by van Velzen et al in 2017 contributed a large proportion (24%) of the updated pooled results.13 This RCT itself was not statistically signifcant. The limited external validity of RCTs prompts questions about the efects of antibiotics treatment for AECOPD in real-world settings.

Primary care of patients with COPD is mostly managed on an outpatient basis. This population from real-world setting is more heterogeneous than those of RCTs.14 Additionally, antibiotic treatment for AECOPD is often not in accordance with current guidelines.15,16 Therefore, the real-world treatment efects of antibiotics for AECOPD may difer from those obtained from clinical trials and merit further investigation. So far, only few observational studies were conducted to evaluate the treatment efects of antibiotics for AECOPD. Two of these studies focused exclusively on the long-term efects of antibiotics used for AECOPD and lacked any adjustments for potential diferences in lung function and smoking history.17,18 Two other cohort studies were conducted among inpatients.19,20 The PharmLines Initiative presented a unique opportunity to retrieve precise information on many previously unmeasured confounders. Using an 63 Chapter 4 inception cohort design, we assessed the short- and long-term efects of antibiotics used in addition to systemic glucocorticoids treatment for AECOPD in outpatients. METHODS Study setting and data sources This retrospective cohort study was conducted as part of the PharmLines Initiative,21 which linked the Lifelines Cohort Study database (https://www.lifelines.nl) and the IADB. nl prescription database (http://www.iadb.nl) afliated to the University of Groningen. Individuals included in these two databases are representative for the population in the northern Netherlands.22,23

Lifelines is a multi-disciplinary prospective population-based cohort study involving 167,729 participants across three generations from 2006 to 2017. A broad range of investigative procedures were used to assess the biomedical, socio-demographic, behavioural, physical and psychological factors contributing to the health and disease status of the general population, with a particular focus on multi-morbidity and complex genetics.24,25 Following baseline assessments, participants underwent physical examinations at the Lifelines location every 5 years and completed extensive questionnaires every 1.5 years. IADB.nl is an evolving drug prescription database since 1996 that currently covers prescription data for 730,000 participants from 72 community pharmacies.23 Each patient is individually tracked throughout the database’s operational period and their prescription records contain information on the date of dispensing, the quantity of medication dispensed, the dose regimen, the number of days for which the prescription is valid, the prescribing physician and the anatomical therapeutic chemical code (ATC code). Each patient, whose date of birth and gender are recorded, is assigned a unique anonymous identifer. Because of the strong patient- pharmacy commitment in the Netherlands, the medication records for each patient are virtually complete, except for over the counter drugs and medication dispensed during hospitalization.

The Lifelines cohort study was approved by the medical ethical committee of the University Medical Center Groningen, and all participants signed informed consent forms confrming their permission for their (anonymized) data and material to be used for scientifc purposes. IADB.nl data are collected in accordance with the national and European guidelines on privacy requirements for handling human data.

Study population Patients with a frst recorded diagnosis of COPD who took systemic glucocorticoids for an acute exacerbation were selected for this study according to the following inclusion criteria: (1) patients were entered in both the Lifelines and IADB.nl databases. (2) Patients had spirometrically-confrmed COPD with a forced expiratory volume in 1

64 Efects of Antibiotic Therapy on AECOPD second/forced volume capacity (FEV1/FVC) < 70% according to lung function test or had general practitioner (GP)-confrmed COPD according to the self-reported questionnaire in the Lifelines Cohort Study. The date of frst recorded COPD diagnosis was set as enrolment date of this study. (3) Patients had the frst recorded acute exacerbation after enrolment date, which was indicated by the prescription of prednisone or prednisolone push treatments (3-4 defned daily doses for 3-14 days) recorded in the IADB.nl database in line with Dutch College of GPs guideline for COPD management.26 The date of the frst prescriptions for acute exacerbation was set as the index date. (4) Patients were 18 years 4 or older on the index date.

Exposure and outcomes During the treatment for frst recorded acute exacerbation with systemic corticosteroid treatment, patients who also received antibiotics (ATC code: J01) within 3 days before and 7 days after the index date were defned as the exposed group. Those patients who were not prescribed any antibiotics during the same period were defned as the reference group. The short-term outcome was treatment failure defned as any new prescription of prednisolone, prednisone or antibiotics between 14 and 30 days after index date. The long-term outcome was the time to the next exacerbation defned as a new prescription of prednisone or prednisolone within a one-year follow-up period. As the frst exacerbation may last for a long time, to avoid counting its following treatment as a second exacerbation, we restricted the minimum time from the frst to the second exacerbation to 21 days.27 The study design for the exposure and outcome measurements is described in Figure 1.

Data collection and covariates Age was calculated as the diference in years between the date of birth and the index date. On the enrolment date, the following information was extracted as covariates from the Lifelines database to describe the characteristics of cohort members with AECOPD: smoking history, the global initiative for chronic obstructive lung disease (GOLD) stages of COPD, lung function parameters and related comorbidities including cardiovascular diseases, diabetes, depression and other disorders. If information concerning the risk status of AECOPD (e.g., smoking history and chronic comorbidities) was not documented on the enrolment date, we used information from the closest follow-up assessment in the Lifelines, if available. Additionally, information on the frequency of AECOPD and maintenance drugs for COPD in the previous year before the index date was retrieved as covariates from the IADB.nl database.

Subgroup and sensitivity analysis Given that diferent antibiotics may have diferent efects on AECOPD, we conducted a subgroup analysis to explore the efects of four most frequently used antibiotics (doxycycline, macrolides, co-amoxiclav and amoxicillin). As COPD patients with an

65 Chapter 4 asthma component may respond diferently to antibiotics, we conducted a sensitivity analysis by excluding these patients to verify the robustness of our study.

Statistical analysis Continuous variables were presented as means with standard deviations (SDs) or median with interquartile ranges (IQRs) and Student’s t-test or Mann-Whitney U Test was performed, as appropriate, to examine their diference between the two patient groups. Categorical variables were presented as percentages with 95% confdence intervals (95% CI) and compared using a Chi-square test or Fisher’s exact test, as appropriate. Binary logistic regression was performed to estimate the odds ratio (OR) with a 95% CI for treatment failure and adjusted for possible covariates. To better control the diferences of characteristics between groups, propensity score (PS) analysis was also conducted by including the PS as a single covariate in the binary logistic regression model. A Kaplan-Meier survival analysis and log-rank test were conducted to compare the times to the next exacerbations between exposure and reference groups. A cox proportional hazards regression was performed to estimate the hazard ratio (HR) and 95% CI for the risk of the next exacerbation. A p-value < 0.05 was considered as statistically signifcant. All analyses were performed using the IBM SPSS statistics version 22 (IBM Corp., Armonk, NY, USA). RESULTS Baseline characteristics The linkage of the IADB.nl and Lifelines database provided 7,760 adults who were prescribed a prednisone or prednisolone treatment (Figure 2). Of these adults, 2,614 (34%) had a diagnosis of COPD. Of these COPD patients, 1,105 with a frst acute exacerbation recorded after their enrolment dates according to pre-set defnitions were eligible for our study. In all, 518 patients were enrolled in the exposed group, receiving both systemic glucocorticoids and an antibiotic. 587 patients were enrolled in the reference group, only receiving systemic glucocorticoids. The baseline characteristics of the study population are summarized in Table 1. Overall, the measured covariates were very similar for both exposed and reference groups. The number of previous exacerbations and antibiotic courses as well as the prevalence of heart failures were higher in the exposure group compared with the reference group.

Short-term outcome Within 14-30 days after treatment of the index exacerbation, 56 (10.8%) patients in the antibiotic exposed group versus 62 (10.6%) patients in the reference group experienced treatment failure (crude OR: 1.03 [95% CI: 0.70-1.50], Table 2). After adjusting for potential confounders through regression and PS analysis, the OR decreased in the direction of a benefcial efect of antibiotics (aOR: 0.67 [95% CI: 0.42-

66 minimum time from the first to the second exacerbation to 21 days.27 The study design for the exposure and outcome measurements is described in Figure 1.

Data collection and covariates

Age was calculated as the difference in years between the date of birth and the index date. On the enrolment date, the following information was extracted as covariates from the Lifelines database to describe the characteristics of cohort members with AECOPD: smoking history, the global initiative for chronic obstructive lung disease (GOLD) stages of COPD, lung function parameters and related comorbidities including cardiovascular diseases, diabetes, depression and other disorders. If information concerning the risk status of AECOPD (e.g., smoking history and chronic comorbidities) was not documented on the enrolment date, we used information from the closest follow‐up assessment in the Lifelines, if available. Additionally, information on the frequency of AECOPD and maintenance drugs for COPD in the Efects of Antibiotic Therapy on AECOPD previous year before the index date was retrieved as covariates from the IADB.nl database.

Figure 1. Retrospective cohort study design. Figure 1. Retrospective cohort study design.

Subgroup1.04] in the and regression sensitivity analysisanalysis), which was statistically signifcant in the PS analysis (aOR 0.63 [95%CI: 0.40-0.99]). Given that different antibiotics may have different effects on AECOPD, we conducted a subgroupLong-term analysis outcome to explore the effects of four most frequently used antibiotics (doxycycline, macrolides,Within a year co‐ amoxiclavof follow-up and afteramoxicillin). the index As COPD date, patients 153 (29.5%) with an patients asthma incomponent the exposure may respondgroup and differently 147 (25.0 to %) antibiotics, patients in we the conducted reference groupa sensitivity experienced analysis a by next excluding exacerbation these (crude HR: 1.19 [95% CI: 0.95-1.49], see Table 3). After adjusting for confounders, patients to verify the robustness of our study. the point estimate of the HR for subsequent exacerbation did not change substantially (adjusted HR 1.14 [95% CI: 0.87-1.49]). There was also no diference between the two comparison groups for the time to the next exacerbation (Figure 3), which applied to the short follow-up period of 3 and 6 months (Table 3).

Subgroup and sensitivity results The fndings of both the logistic regression and PS analyses indicated that the risk of treatment failure was reduced signifcantly by 47% by doxycycline compared to the reference treatments (aOR 0.53 [95% CI: 0.28-1.00] and 0.53 [95% CI: 0.28-0.99] by regression and PS analyses, respectively, Table 1). Although not statistically signifcant, similar benefcial trends were seen for the macrolides exposed group (aOR 0.49 [95% CI: 0.22-1.11] and 0.58 [95% CI: 0.26-1.29] by regression and PS analyses, respectively) and co-amoxicillin exposed group (aOR 0.50 [95% CI: 0.19-1.32] and 0.46 [95% CI: 0.17-1.24] by regression and PS analyses, respectively) compared to the results in the reference group. No statistical diference was observed between the amoxicillin exposed group and the reference group (aOR 1.56 [95% CI: 0.81-3.00] and 1.49 [95% CI: 0.78-2.84] by regression and PS analyses, respectively) and the point estimate of aOR was in the opposite direction.

Even when we excluded self-reported COPD and focused only on spirometrically- confrmed COPD, the protective efect of antibiotics on treatment failure continued (aOR 0.56 [95% CI: 0.32-0.97] and 0.52 [95% CI: 0.29-0.90] by regression and PS analyses,

67 Chapter 4

LifeLines participants from 2007 IADB participants from 1994 to 2017 (n167.000) (n600.000)

Patients 18 years Prednisone or prednisolone (n=152.728) users

Databases overlap (n=7760) Inclusion: 1. Clinical COPD diagnosis (FEV1/FVC<0.70) n=1596 2. Self‐reported COPD diagnosis: n=1635

COPD patients (n=2614)

COPD patients, using corticosteroids (n=1594)

Inclusion: COPD patients with Using 3 or 4 DDDs daily exacerbations corticosteroids for 3‐14 days (n=1105)

Exposed group: Reference group: corticosteroids & corticosteroids antibiotic use use only (n=518) (n=587)

Figure 2. Flow chart of study subject selection. Figure 2. Flow chart of study subject selection.

Statistical analysis respectively; Table 4). Similarly, after excluding COPD patients with asthma, the aOR Continuous variables were presented as means with standard deviations (SDs) or median with (exposure vs reference) for treatment failure was further reduced towards a protective einterquartilefect (aOR 0.58 ranges [95% (IQRs) CI: 0.32-1.01] and Student’s and t0.57‐test [0.32, or Mann 1.02]‐Whitney by regression U Test was and performed, PS analyses) as withappropriate, a boundary to examinestatistical their signi differencefcance. between the two patient groups. Categorical DISCUSSIONvariables were presented as percentages with 95% confidence intervals (95% CI) and compared using a Chi‐square test or Fisher’s exact test, as appropriate. Binary logistic Primary fndings regression was performed to estimate the odds ratio (OR) with a 95% CI for treatment failure In this study of COPD outpatients with mostly mild to moderate GOLD stages, and adjusted for possible covariates. To better control the differences of characteristics antibiotics prescription, notably doxycycline, in addition to systemic prednisone or prednisolonebetween groups, therapy, prope appearednsity score to (PS) reduce analysis the treatmentwas also conducted failure of byAECOPD including substantially. the PS as a Thesingle supplementation covariate in the binary of antibiotic logistic regression treatment model. to systemicA Kaplan ‐Meier glucocorticoids survival analysis did and not prolong the time to the next exacerbation for up to one follow-up year compared with

68 Efects of Antibiotic Therapy on AECOPD

Table 1. Baseline characteristics of COPD patients included in this study (N=1,105).

Exposed (n=518) Reference (n=587) Patient characteristics N (%) N (%) P-value

Age (yr.) Median 55 (18) 54 (18) 0.24 <= 50 185 (35.7) 222 (37.8) 0.77 50-65 189 (36.5) 208 (35.4) >=65 144 (27.8) 157 (26.7) 4 Gender 0.68 Male 211 (40.7) 232 (39.5) Female 307 (59.3) 355 (60.5) BMI (kg/m2) Median 26.85 (6.5) 26.40 (5.9) 0.24 <= 24.9 171 (33.0) 203 (34.6) 0.35 25.0-29.9 195 (37.6) 239 (40.7) >= 30 152 (29.3) 145 (24.7) Lung function FEV1 (L) 2.59 (1.0) 2.68 (1.0) 0.24 FEV1 (% predicted) 83.91 (23.25) 84.28 (22.29) 0.76 FVC (L) 3.86 (1.0) 3.96 (1.0) 0.13 FVC (% predicted) 65.54 (18.72) 66.45 (17.16) 0.72 FEV1 to FVC ratio 0.68 (0.11) 0.67 (0.09) 0.77 GOLD stage 0.43 I: Mild 266 (59.1) 307 (59.4) II: Moderate 160 (35.6) 193 (37.3) III and IV: Severe/very severe 24 (5.4%) 17 (3.3) Smoking status 0.25 Current smoker 169 (35.6) 167 (30.8) Former smoker 190 (40.0) 227 (41.9) Non smoker 116 (24.4) 148 (27.3) No. of AECOPD in previous yr. 0.01 0 472 (91.1) 557 (94.9) 1 17 (3.3) 17 (2.9) 2 or more 29 (5.6) 13 (2.2) No. of antibiotics prescription in previous yr. <0.01 0 20 (3.9) 322 (54.9) 1 228 (44.0) 139 (23.7) 2 or more 270 (52.1) 126 (21.5) Comorbidities A. Cardiovascular diseases Heart failure 22 (4.2) 12 (2.0) 0.03 Heart attack 23 (4.4) 18 (3.1) 0.23 Stroke < 10 11 (1.9) 0.87 Arrhythmia 74 (14.4) 78 (13.3) 0.63 Hypertension 165 (31.9) 180 (30.7) 0.67

69 Chapter 4

Table 1. (continued)

Exposed (n=518) Reference (n=587) Patient characteristics N (%) N (%) P-value

B. other major disorders Asthma 169 (32.6) 200 (34.1) 0.62 Pulmonary fbrosis < 10 < 10 0.99 Diabetes 28 (5.4) 34 (5.8) 0.78 Cancer < 10 < 10 0.79 Osteoporosis 27 (5.2) 21 (3.6) 0.18 Renal impairment 16 (3.1) 22 (3.7) 0.55 Depression 83 (16.0) 87 (14.8) 0.58 Anxiety < 10 < 10 0.26 Anemia 72 (13.9) 78 (13.3) 0.77 C. other minor disorders Ulcerative colitis < 10 < 10 0.03 Stomach ulcer 24 (4.6) 23 (3.9) 0.56 Irritable bowel syndrome 61 (11.8) 62 (10.6) 0.52 Hepatic impairment 13 (2.5) 10 (1.7) 0.35 COPD maintenance medications in previous yr. SABA 169 (32.6) 183 (31.2) 0.61 LABA 22 (4.2) 31 (5.3) 0.42 SAMA 12 (2.3) 14 (2.4) 0.94 LAMA 72 (13.9) 61 (10.4) 0.07 ICS 67 (12.9) 70 (11.9) 0.61 LABA/ICS 190 (36.7) 194 (33) 0.21 Theophylline < 10 < 10 0.67

Note: Data are presented as mean (standard deviation [SD]) or median with interquartile range (IQR) or numbers with percentage. Due to privacy protection of patients according to contract, the number below 10 was not permitted to present. Abbreviations: BMI: body mass index; SABA: short-acting β agonist; SAMA: short-acting muscarinic antagonist; LABA: long-acting β agonist; LAMA: long-acting muscarinic antagonist; ICS: inhaled corticosteroid; GOLD: global initiative for chronic obstructive lung disease; FEV1: forced expiratory volume in 1 second; FVC: forced vital capacity; AECOPD: acute exacerbation of chronic obstructive pulmonary disease; treatment with only glucocorticoids. Our results, after limiting to the COPD patients to those who had spirometrically-confrmed COPD, were robust. Similar robust results were also obtained after excluding COPD patients with self-reported asthma.

The benefcial treatment efects of antibiotic use for AECOPD observed in this study were consistent with the updated pooled results by Vollenweider et al. for outpatients.12 The fnding that additional antibiotic treatment for AECOPD did not produce long-term benefcial efects is also consistent with results of a previous RCT conducted in COPD outpatients.13 Conversely, two previous observational studies reported that antibiotic treatment is associated with a reduced risk of a subsequent exacerbation.17,18 However,

70 Efects of Antibiotic Therapy on AECOPD

Table 2. Odds ratio for treatment failure of index exacerbation among COPD outpatients with adjustment by logistic regression and propensity score weighted analysis.

Treatment groups Treatment failure Crude OR Adjusted OR PS adjusted OR (No.) No. (%) (95% CI) (95% CI)a (95% CI)b

Reference (587) 56 (10.8) 1 1 1 All antibiotics (518) 56 (10.8) 1.03 (0.70-1.50) 0.67 (0.42-1.04) 0.63 (0.40-0.99)* Doxycycline (214) 19 (8.9) 0.83 (0.48-1.42) 0.53 (0.28-1.00)* 0.53 (0.28-0.99)* Macrolides (102) 11 (10.8) 1.02 (0.52-2.02) 0.49 (0.22-1.11) 0.58 (0.26-1.29) 4 Amoxicillin (100) 18 (18.0) 1.86 (1.05-3.30) 1.56 (0.81-3.00) 1.49 (0.78-2.84) Co-amoxicillin (87) < 10 0.74 (0.33-1.68) 0.50 (0.19-1.32) 0.46 (0.17-1.24)

Abbreviation: OR: odds ratio; CI: confdence interval; PS: propensity score weighted analysis; No.: number; aAdjusted result by logistic regression; bAdjusted result by propensity score weighted analysis; *P<0.05;

Tables 3. Hazard ratio for next exacerbation with follow-up of 1 year among COPD outpatients.

Follow-up time Antibiotics group Reference group Crude HR (95% CI) Adjusted HR (95% CI)a

3 months 57 (11.0) 60 (10.2) 1.10 [0.76, 1.58] 1.11 [0.71, 1.71] 6 months 109 (21.0) 119 (20.3) 1.06 [0.81, 1.37] 1.05 [0.77, 1.42] 12 months 153 (29.5) 147 (25.0) 1.19 [0.95, 1.49] 1.14 [0.87, 1.49]

Abbreviation: HR: hazard ratio; CI: confdence interval; aAdjusted baseline characteristics by using cox hazard logistic regression. insufcient information in these studies on, for example, lung function, smoking history and related comorbidities, which are important risk factors associated with exacerbation events, could have accounted for these discrepancies.28

GOLD guidelines recommend amoxicillin with clavulanic acid, macrolide and tetracycline as the frst-line antibiotics treatment for AECOPD.3 The Dutch primary care guidelines recommend amoxicillin or doxycycline as frst-line antibiotics in AECOPD treatment.26 The combined results of seven RCTs examined in the updated Cochrane review showed that the antibiotics were generally efective in treating AECOPD in outpatients.12 However, three of these studies examined combined antibiotics29-31 and only four focused on the specifc antibiotics recommended in the above-mentioned guidelines: one on doxycycline, two on co-amoxiclav and one on amoxicillin.13,32-34 Therefore, no scientifcally sound conclusion could be drawn about the efects of these specifc antibiotics used for AECOPD. Regarding specifc antibiotics in our study, doxycycline had signifcant benefcial treatment efects on AECOPD, we observed similar trends towards benefcial efects for macrolides and amocillin-clavunalate, though non- signifcant, which may be due to the limited power of our study size to detect efects in subgroup analysis, or there were no efects at all, and more qualifed studies are needed

71 Chapter 4

Figure 3. Kaplan-Meier curves showing the proportion of patients free of next exacerbation in COPD Figure 3. Kaplan‐Meier curves showing the proportion of patients free of next exacerbation in outpatients up to follow-up of 1 year. COPD outpatients up to follow‐up of 1 year. to exploreSubgroup these and sensitivity efects results further. Regarding the short-term efects of doxycycline, the Theresistance findings of of both common the logistic pathogens regression forand AECOPDPS analyses like indicated Haemophilus that the risk in fofuenzae treatment and Streptococcus pneumonia to doxycycline is reported to be rare in the Netherlands,35 this failure was reduced significantly by 47% by doxycycline compared to the reference treatments may contribute to its successful short-term treatment efects. (aOR 0.53 [95% CI: 0.28‐1.00] and 0.53 [95% CI: 0.28‐0.99] by regression and PS analyses, Thererespectively, is a general Table consensus 1). Although that not exacerbation statistically significant, frequency similar increases beneficial with trends COPD were severity. seen 36 Becausefor the the macrolides COPD severity exposed groupof patients (aOR 0.49 in our[95% study CI: 0.22 was‐1.11] generally and 0.58 mild [95% (aroundCI: 0.26‐1.29] 60%) compared with that of patients in previous studies (10%-20%),12,13 the rate of next by regression and PS analyses, respectively) and co‐amoxicillin exposed group (aOR 0.50 [95% exacerbations in our study was relatively low. About 30% of patients experienced re- exacerbationCI: 0.19‐1.32] after and index 0.46 [95% exacerbation CI: 0.17‐1.24] within by oneregression year ofand follow-up. PS analyses, After respectively) adjusting for compared possible confounders,to the results wein the observed reference similar group. rates No statistical of next difference exacerbations was observed between antibioticbetween users the amoxicillin and non-antibiotic exposed group users, and the a reference fnding group that (aOR is consistent 1.56 [95% CI: with 0.81 that‐3.00] of 13 a previousand 1.49 RCT [95% report. CI: 0.78 ‐2.84] by regression and PS analyses, respectively) and the point Althoughestimate the of presence aOR was in of the purulent opposite sputum direction. is widely deemed to be the sole determinant 3 of antibioticEven when treatment we excluded of selfAECOPD,‐reported its COPD accuracy and focused and reproducibility only on spirometrically as an indicator‐confirmed of bacterial infection is limited,37 especially for outpatients. Consequently, guidelines on COPD, the protective effect of antibiotics on treatment failure continued (aOR 0.56 [95% CI: antibiotics prescriptions are not stringently adhered to for treating AECOPD.15 Moreover, antibiotics were unusually overprescribed for patients, notably for patients between 18 and 65 years of age in general practice.16,38 Accordingly, we could not exclude

72 Efects of Antibiotic Therapy on AECOPD b PS adjusted OR PS adjusted (95% CI) 0.57 [0.32, 1.02] [0.32, 0.57 1.48] [0.35, 0.71 0.90] [0.29, 0.52 2.06] [0.43, 0.94 4 a Adjustedresult bypropensity score weighted b Adjusted OR (95% CI) 0.58 [0.32, 1.01] [0.32, 0.58 1.67] [0.39, 0.81 0.97] [0.32, 0.56 2.32] [0.46, 1.03 Crude OR (95% CI) 1.03 [0.64, 1.66] [0.64, 1.03 1.95] [0.55, 1.03 1.36] [0.51, 0.83 2.67] [0.76, 1.42 adjusted result by logistic regression; regression; logistic by result adjusted a Reference 39 (10.1) 23 (11.5) 43 (1.1) 19 (9.5) No. (%) No. Treatment failure Antibiotics 36 (10.3)) 20 (11.8) 30 (9.4) 26 (13.1) P<0.05; Sensitivity analyses: odds ratio for treatment failure of index exacerbation among outpatients. COPD * Limited study population Limited (Number) number; analysis;No.: interval;conweightedf dence propensityCI: PS:ratio; score odds Abbreviation:OR: analysis; COPD with asthma (369) (707) f rmed COPD con Spirometrically (398) COPD Self-reported COPD without asthma (736) Table 4. Table

73 Chapter 4 the possibility that antibiotic treatment for some of patients were improperly prescribed. The general implication of our fnding is that contrasting with an ideal situation in which all patients’ prescriptions adhere to the guidelines, efects of antibiotics for a bacterially caused AECOPD were underestimated in our study.

Clinical implications Although improper use of antibiotics may occur within the COPD outpatient population in a real-world setting, our fndings supported the benefcial efect of antibiotics used for AECOPD. Valid antibiotics prescription could further improve the efects of antibiotic treatments on AECOPD. According to the latest GOLD guideline, the sputum colour can be used to avoid unnecessary antibiotic therapy safely with cream, white or clear sputum indicating very low bacterial infections.3,14 If applicable, a procalcitonin-guided algorithm or C-reactive protein (CPR) test can also be considered before making decisions for GP to reduce the unnecessary administration of antibiotics .39,40

Given signifcant variability between GP practices of prescribing antibiotics to COPD patients experiencing exacerbations,41 we recommend doxycycline as the mainstay based on our fndings that are consistent with the Dutch guidelines.26 Though estimates indicate similar benefcial efects for some specifc antibiotics, larger studies of high quality with extensive control for potential confounders are needed to explore their role in AECOPD management. Importantly, the fnal antibiotic choice should also be based on the local bacterial resistance patterns, and sputum cultures of high-risk patients with frequent exacerbations and severe airfow limitations should be performed, given the possible presence of resistant pathogens.3

Strengths and limitations Our study had several strengths. Firstly, the population included in this study was representative of COPD outpatients. Hence our fndings refect the real-world efects of antibiotic treatment for AECOPD. Secondly, properly diagnosed COPD patients and their complete background information , for example, lung function, smoking status and related comorbidities that were lacking in previous observational studies were included in this study. Moreover, the outcomes were adjusted for possible confounders using both logistic regression and PS analyses. Sensitivity analyses by further narrowing study population by excluding diferent sources of uncertainty were also conducted to test the robustness of the results.

There were, however, several limitations. Firstly, an acute exacerbation was defned by the prescriptions of systemic glucocorticoids as a proxy according to Dutch guideline for AECOPD and this may have led to some misclassifcations. In addition, we lacked clinical information on the severity of AECOPD at the time of diagnosis. Moreover, the severity of exacerbations may not have been evenly distributed between the two comparison

74 Efects of Antibiotic Therapy on AECOPD groups. Secondly, antibiotics may have been prescribed improperly in the absence of confrmed bacterial infections, this could lead to an underestimated efect of antibiotics on AECOPD. Thirdly, the low power of subgroup analyses in this study hindered us to make a defnitive conclusion regarding the efects of some specifc antibiotics on AECOPD. Finally, the IADB.nl did not include prescriptions during a hospitalization. However, given the relatively mild outpatient group, we expect only few patients to have such a serious outcome in our study. 4 CONCLUSION

The results of this study support the use of antibiotic therapy, notably doxycycline, for AECOPD in addition to systematic glucocorticoids treatment among outpatients. Further larger qualifed studies with prospective designs and extensive control of confounders are required to explore the efects of other specifc antibiotics in a real-world settings. No further long-term benefcial efects of antibiotics treatment on AECOPD were found for subsequent exacerbations.

75 Chapter 4

REFERENCES

1. Lozano R, Naghavi M, Foreman K, et al. 11. Vollenweider DJ, Jarrett H, Steurer-Stey CA, Global and regional mortality from 235 Garcia-Aymerich J, Puhan MA. Antibiotics causes of death for 20 age groups in for exacerbations of chronic obstructive 1990 and 2010: a systematic analysis for pulmonary disease. Cochrane Database Syst the Global Burden of Disease Study 2010. Rev. 2012;12:CD010257. Lancet. 2012;380(9859):2095-2128. 12. Vollenweider DJ, Frei A, Steurer-Stey CA, 2. Lopez AD, Shibuya K, Rao C, et al. Chronic Garcia-Aymerich J, Puhan MA. Antibiotics obstructive pulmonary disease: current for exacerbations of chronic obstructive burden and future projections. Eur pulmonary disease. Cochrane Database Syst Respir J. 2006;27(2):397-412. Rev. 2018;10:CD010257. 3. Global Initiative for Chronic Obstructive 13. van Velzen P, Ter Riet G, Bresser P, et al. Lung Disease (GOLD). Global Strategy for Doxycycline for outpatient-treated acute the Diagnosis, Management and Prevention of exacerbations of COPD: a randomised Chronic Obstructive Pulmonary Disease: 2020 double-blind placebo-controlled trial. Report. https://goldcopd.org/gold-reports/. Lancet Respir Med. 2017;5(6):492-499. Date last accessed: December 17, 2019. 14. Kurisu K, Yoshiuchi K, Ogino K, Okada Y, 4. Donaldson GC, Seemungal TAR, Bhowmik Oda T. Peak C-reactive protein levels do not A, Wedzicha JA. Relationship between predict 30-day mortality for bacteremia: exacerbation frequency and lung function A retrospective cohort study. J Infect decline in chronic obstructive pulmonary Chemother. 2020;26(1):23-27. disease. Thorax. 2002;57(10):847-852. 15. Bathoorn E, Groenhof F, Hendrix R, et al. 5. Anzueto A. Impact of exacerbations on Real-life data on antibiotic prescription COPD. Eur Respir Rev. 2010;19(116):113-118. and sputum culture diagnostics in acute 6. Walters JAE, Tan DJ, White CJ, Gibson exacerbations of COPD in primary care. Int J PG, Wood-Baker R, Walters EH. Systemic Chron Obstruct Pulmon Dis. 2017;12:285-290. corticosteroids for acute exacerbations of 16. Roede BM, Bindels PJ, Brouwer HJ, Bresser chronic obstructive pulmonary disease. P, de Borgie CA, Prins JM. Antibiotics and Cochrane Db Syst Rev. 2014(9). steroids for exacerbations of COPD in primary 7. Moghoofei M, Azimzadeh Jamalkandi S, care: compliance with Dutch guidelines. Br J Moein M, Salimian J, Ahmadi A. Bacterial Gen Pract. 2006;56(530):662-665. infections in acute exacerbation of chronic 17. Roede BM, Bresser P, Prins JM, Schellevis obstructive pulmonary disease: a systematic F, Verheij TJM, Bindels PJE. Reduced review and meta-analysis. Infection. 2019. risk of next exacerbation and mortality 8. Sethi S, Murphy TF. Infection in associated with antibiotic use in COPD. Eur the pathogenesis and course of chronic Respir J. 2009;33(2):282-288. obstructive pulmonary disease. N Engl J 18. Roede BM, Bresser P, Bindels PJE, et al. Med. 2008;359(22):2355-2365. Antibiotic treatment is associated with 9. Wilkinson TMA, Aris E, Bourne SC, et reduced risk of a subsequent exacerbation al. Drivers of year-to-year variation in in obstructive lung disease: an historical exacerbation frequency of COPD: analysis population based cohort study. of the AERIS cohort. ERJ Open Res. 2019;5(1). Thorax. 2008;63(11):968-973. 10. Monso E, Garcia-Aymerich J, Soler N, et al. 19. Stefan MS, Rothberg MB, Shieh MS, Pekow Bacterial infection in exacerbated COPD PS, Lindenauer PK. Association between with changes in sputum characteristics. antibiotic treatment and outcomes Epidemiol Infect. 2003;131(1):799-804. in patients hospitalized with acute

76 Efects of Antibiotic Therapy on AECOPD

exacerbation of COPD treated with systemic of COPD in a primary care population: steroids. Chest. 2013;143(1):82-90. a retrospective observational cohort study. 20. Petite SE, Murphy JA. Systemic Corticosteroid Bmj Open. 2014;4(12):e006171. and Antibiotic Use in Hospitalized 29. Hassan WA, Shalan I, Elsobhy M. Impact of Patients With Chronic Obstructive antibiotics on acute exacerbations of COPD. Pulmonary Disease Exacerbation. Ann Egypt J Chest Dis Tu. 2015;64(3):579-585. Pharmacother. 2019;53(2):144-150. 30. Sachs APE, Koeter GH, Groenier 21. Sediq R, van der Schans J, Dotinga A, et al. KH, Vanderwaaij D, Schiphuis J, Concordance assessment of self-reported Meyboomdejong B. Changes in Symptoms, 4 medication use in the Netherlands three- Peak Expiratory Flow, and Sputum Flora generation Lifelines Cohort study with during Treatment with Antibiotics of the pharmacy database iaDB.nl: The PharmLines Exacerbations in Patients with Chronic initiative. Clin Epidemiol. 2018;10:981-989. Obstructive Pulmonary-Disease in General- Practice. Thorax. 1995;50(7):758-763. 22. Klijs B, Scholtens S, Mandemakers JJ, Snieder H, Stolk RP, Smidt N. Representativeness of 31. Anthonisen NR, Manfreda J, Warren CP, the LifeLines Cohort Study. Plos One. 2015;10(9). Hershfeld ES, Harding GK, Nelson NA. Antibiotic therapy in exacerbations of 23. Visser ST, Schuiling-Veninga CCM, Bos JHJ, chronic obstructive pulmonary disease. Ann de Jong-van den Berg LTW, Postma MJ. Intern Med. 1987;106(2):196-204. The population-based prescription database IADB.nl: its development, usefulness in 32. Llor C, Moragas A, Hernandez S, Bayona C, Miravitlles M. Efcacy of antibiotic therapy outcomes research and challenges. Expert for acute exacerbations of mild to moderate Rev Pharm Out. 2013;13(3):285-292. chronic obstructive pulmonary disease. Am 24. Scholtens S, Smidt N, Swertz MA, et J Respir Crit Care Med. 2012;186(8):716-723. al. Cohort Profle: LifeLines, a three- 33. Brusse-Keizer M, VanderValk P, Hendrix generation cohort study and biobank. Int J R, Kerstjens H, van der Palen J. Necessity Epidemiol. 2015;44(4):1172-1180. of amoxicillin clavulanic acid in addition 25. Stolk RP, Rosmalen JGM, Postma DS, et to prednisolone in mild-to-moderate al. Universal risk factors for multifactorial COPD exacerbations. Bmj Open Respir diseases - LifeLines: a three-generation Res. 2014;1(1):e000052. population-based study. Eur J 34. Jorgensen AF, Coolidge J, Pedersen PA, Epidemiol. 2008;23(1):67-74. Petersen KP, Waldorf S, Widding E. Amoxicillin 26. Snoeck-Stroband JB ST, Van Schayck CP, in treatment of acute uncomplicated Muris JW, Van der Molen T, In ’t Veen JCCM, exacerbations of chronic bronchitis. Chavannes NH, Broekhuizen BDL, Barnhoorn A double-blind, placebo-controlled MJM, Smeele I, Geijer RMM, Tuut MK. NHG- multicentre study in general practice. Scand Werkgroep Astma bij volwassenen en COPD. J Prim Health Care. 1992;10(1):7-11. NHG-Standaard COPD (derde herziening). 35. Greef SC, Mouton JW, eds. NethMap 2015. Huisarts Wet 2015;58(4):198-211. Consumption of antimicrobial agents and 27. Seemungal TA, Donaldson GC, Bhowmik A, antimicrobial resistance among medically Jefries DJ, Wedzicha JA. Time course and important bacteria in the Netherlands. recovery of exacerbations in patients with Bilthoven: National Institute for Public chronic obstructive pulmonary disease. Am Health and the Environment, 2015. J Respir Crit Care Med. 2000;161(5):1608-1613. 36. Donaldson GC, Wedzicha JA. COPD 28. Mullerova H, Shukla A, Hawkins A, Quint exacerbations 1: Epidemiology. J. Risk factors for acute exacerbations Thorax. 2006;61(2):164-168.

77 Chapter 4

37. Siddiqi A, Sethi S. Optimizing antibiotic pulmonary disease: to use or not to use. selection in treating COPD exacerbations. Int Curr Opin Pulm Med. 2019;25(2):150-157. J Chron Obstruct Pulmon Dis. 2008;3(1):31-44. 40. Butler CC, Gillespie D, White P, et al. C-Reactive 38. Dekker AR, Verheij TJ, van der Velden Protein Testing to Guide Antibiotic AW. Inappropriate antibiotic prescription Prescribing for COPD Exacerbations. N Engl for respiratory tract indications: most J Med. 2019;381(2):111-120. prominent in adult patients. Fam 41. Boggon R, Hubbard R, Smeeth L, et Pract. 2015;32(4):401-407. al. Variability of antibiotic prescribing 39. Chan HP, Lim TK. Procalcitonin and in patients with chronic obstructive antibiotics in moderate-severe acute pulmonary disease exacerbations: a cohort exacerbation of chronic obstructive study. Bmc Pulm Med. 2013;13:32.

CHAPTER 5 Improving antibacterial prescribing safety in the management of COPD exacerbations: systematic review of observational and clinical studies on potential drug interactions associated with frequently prescribed antibacterials among COPD Patients

Yuanyuan Wang Muh. Akbar Bahar Anouk M.E. Jansen Janwillem W.H. Kocks Jan-Willem C. Alfenaar Eelko Hak Bob Wilfert Sander D. Borgsteede

Published as: Wang Y, Bahar MA, Jansen AME, Kocks JWH, Alfenaar JC, Hak E, Wilfert B, Borgsteede SD. Improving antibacterial prescribing safety in the management of COPD exacerbations: systematic review of observational and clinical studies on potential drug interactions associated with frequently prescribed antibacterials among COPD patients. J Antimicrob Chemother. 2019;74(10):2848–2864. Chapter 5

ABSTRACT Background Guidelines advice the use of antibacterials (ABs) in the management of COPD exacerbations. COPD patients often have multiple comorbidities like diabetes mellitus and cardiac diseases leading to polypharmacy. Consequently, drug-drug interactions (DDIs) may frequently occur, cause serious adverse events and treatment failure.

Objective (i) To review DDIs related to frequently prescribed ABs among COPD patients from observational and clinical studies. (ii) To improve AB prescribing safety in clinical practice by structuring DDIs according to comorbidities of COPD.

Methods We conducted a systematic review by searching Pubmed and Embase up to Feb 8, 2018 for clinical trials, cohort and case-control studies reporting DDIs of ABs used for COPD. Study design, subjects, sample size, pharmacological mechanism of DDI, and efect of interaction were extracted. We evaluated level of DDIs and quality of evidence according to established criteria and structured the data by possible comorbidities.

Results In all, 318 articles were eligible for review describing a wide range of drugs used for comorbidities and their potential DDIs with ABs. DDIs between ABs and co-administered drugs could be subdivided into: (1) co-administered drugs alter the pharmacokinetics of ABs; and (2) ABs interfere with the pharmacokinetics of co-administered drugs. The DDIs could lead to therapeutic failures or toxicities.

Conclusion DDIs related to ABs with clinical signifcance may involve a wide range of indicated drugs to treat comorbidities in COPD. The evidence can support (computer supported) decision-making by health practitioners when prescribing ABs during COPD exacerbations in the case of co-medication.

82 DDIs Associated with Antibacterials for COPD

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a complex respiratory disorder characterized by persistent respiratory symptoms and airfow limitation.1 The chronic and progressive course of COPD is frequently aggravated by exacerbation defned as an acute worsening of respiratory symptoms like increased cough, dyspnea and production of sputum.2 Exacerbations of COPD can be triggered by respiratory tract infections, 40% to 60% of exacerbations are caused by bacteria, especially by Haemophilus infuenzae, Streptococcus pneumoniae and Moraxella catarrhalis.3 Evidence from randomized controlled trials (RCTs) indicated that use of antibacterials (ABs) may 5 reduce the frequency and severity of COPD exacerbations.4-6 Therefore, guidelines have recommended involving ABs in the therapeutic and preventive management of COPD exacerbations.1,7

Patients with COPD often sufer from multiple morbidities.8 Hence, polypharmacy is common and contributes to drug-drug interactions (DDIs). Adverse drug reactions (ADRs) or therapeutic failure may be the result of ABs and co-administered drugs interactions. Besides, COPD is an age-related disease and elderly are more susceptible to the efect of DDIs because of gradual physiologic changes afecting pharmacokinetics and pharmacodynamics.9

The objective of this study was to (1) systematically review DDIs related to frequently prescribed ABs among COPD patients from observational and clinical studies and (2) improve AB prescribing safety in clinical practice by structuring DDIs according to comorbidities of COPD. Studies without comparison groups, and therefore have low quality of the causal evidence, like case reports about QT-interval prolonging interactions are not included in this review. Hence, a DDI handbook like Stockley’s Drug Interactions and the ofcial product information can be referred to see the clinical impact of those kinds of interactions. METHODS Searching strategy We conducted a systematic review following PRISMA guideline. PubMed and Embase databases were searched for related articles published in English up to Feb 8, 2018 using key terms of “drug interactions,” “pharmacokinetics”, “pharmacodynamics”, and a list of most frequently used ABs for COPD (see Table 1). The ABs were selected based on two related Cochrane reviews and their prescription frequency by the University of Groningen prescription database IADB.nl (http://www.iadb.nl/) covering drug prescriptions of approximately 700,000 people.4,5 Additionally, we checked the primary sources of signals from Dutch DDI alert systems: G-Standard and Pharmabase.10

83 Chapter 5

Reference lists from eligible studies were also tracked for additional qualifed papers. Searching details are provided in supplementary data.

Study selection criteria Eligible studies met the following criteria: (1) DDIs in humans; (2) involving the targeted ABs; (3) being clinical trials, RCTs, cohort, or case-control study. We excluded case reports or other descriptive studies. We further excluded studies with subjects whose pharmacokinetics and pharmacodynamics were not comparable to the general COPD patients, e.g. newborn babies, pregnant women and patients with severe renal/hepatic impairment. Other exclusion criteria were: (1) unregistered drugs (by FDA or EMA); (2) involving three or more drug interactions; (3) not DDIs (food-drug, gene-drug); (4) not original studies (reviews, letters and editorials). Besides, pharmacodynamic interactions were beyond the scope of this review and then, excluded.

Data extraction and quality assessment All records were exported to Refworks; title and abstracts were screened by Y.W. and A.M.E.J. independently. Full-text papers were obtained for records that were considered of potential relevance by at least one of the reviewers. Final decisions were made by consensus between two reviewers according to the preset criteria. Discrepancies between reviewers were resolved by discussion, a third reviewer (E.H.) was asked if no consensus was reached. Information about name of ABs and related interacting drug, study design, study subjects, sample size, interacting mechanism, efects of interaction, recommendation by study authors were extracted by the same reviewers (Y.W., A.M.E.J.)

Table 1. Antibacterials (ABs) of study that are frequently prescribed among COPD patients.*

Category Sub-category ABs included

Beta-lactam Penicillins Amoxicillin/clavulanic acid (co-amoxiclav), Amoxicillin, Flucloxacillin, Pheneticillin, phenoxymethylpenicillin (penicillin V), Cephalosporins Cefaclor, Cefuroxime, Ceftriaxone, Cephradine, Ceftazidime Macrolides Erythromycin, Clarithromycin, Azithromycin, Roxithromycin, Clindamycin, Tetracycline Tetracycline, Doxycycline, Minocycline, Quinolones Fluoroquinolone Ciprofoxacin, Moxifoxacin, Levofoxacin, Ofoxacin, Norfoxacin, Other quinolone Pipemidic acid Sulfonamides Sulfamethoxazole Others Nitrofurantoin, Methenamine, Trimethoprim

*based on two Cochrane reviews4 and use within the University Groningen prescription database IADB.nl from Netherlands (http://www.iadb.nl/)

84 DDIs Associated with Antibacterials for COPD and checked by another reviewer (M.A.B.). Quality of evidence was evaluated by grade 0 to 4 based on criteria (Table 2) used by previous studies.11,12

The strength of the DDIs were classifed into four levels (1= strong /2 = substantial /3 = moderate /4 = weak/no) according to the preset published criteria (Table 3).12 In case

Table 2. Quality of Evidence for DDIs11,12

Defnition Score

Clinical researches with appropriate control group and relevant pharmacokinetics and/or 4 5 pharmacodynamics parameters. The studies meet all of the criteria below: t The interacting efect of concomitant medication with investigated drugs is reported in the manuscript. t All of potential confounders are mentioned and taken into account (for example smoking behavior or renal function). t The results of interaction are built on the ‘steady-state kinetics’. t - Variation in dose was adjusted. Clinical researches with appropriate control group and relevant pharmacokinetics and/or 3 pharmacodynamics parameters which do not meet one or more pre-defned criteria above. Complete observational studies with clinically relevant results. 2 Incomplete observational studies. (e.g. without controlling confounders or presence of other 1 explanation factors for the adverse reaction), case reports, SmPc. In vitro studies, in vivo animal studies, prediction modelling studies. 0

Table 3. Description for level of DDIs10

Defnition Score*

Involving inhibitor = > 200% ↑AUC; clearance ↓ > 67% 1 Involving inducer = > 90% ↓ AUC; clearance ↑ ≥ 900% 1 For observational studies, RR/OR ≥ 10 1 Involving inhibitor = 75-200% ↑AUC; clearance ↓ ≥ 43% to < 67% 2 Involving inducer = 60-90% ↓ AUC; clearance ↑ ≥ 150% to < 900% 2 For observational studies, RR/OR = 3~9 2 Involving inhibitor = 25-75% ↑AUC; clearance ↓ ≥ 20% to < 43% 3 Involving inducer = 25-60% ↓AUC; clearance ↑ ≥ 33 % to < 150% 3 For observational studies, RR/OR = 1.5~2.9 3 <25% change in AUC; clearance ↓ < 20% or ↑ < 33 % 4 For observational studies, RR/OR < 1.5 4 a. For the Interacting drugs with narrow therapeutic index, the degree of DDIs will be Exception improved to the one higher degree of level. Exception b. If the DDIs level cannot be judged by the above criteria, we assess it by discussion based on available data and evidence.

*defnition: 1 = strong interaction, 2 = substantial interaction, 3 = moderate interaction, and 4 = weak/no interaction

85 Chapter 5 of several studies on the same DDI combination, we categorized the DDI based on the highest level of severity. Considering that narrow therapeutic index (NTI) drugs are more vulnerable to DDIs, the strength of the DDI was upgraded one level.12 RESULTS Publications identifed by literature search Our search yielded 1,412 and 1,734 studies from Pubmed and Embase, respectively (Figure 1). After removing duplicates, 2,560 articles were screened by title and abstracts, of which 630 papers were included for full-text screening, resulting in 282 eligible articles. With 36 studies identifed from other resources, we got 318 studies fnally for assessment in this review.

The interacting drugs, underlying mechanisms, levels and practice recommendations of the DDIs are presented in Table 4. Details on individual studies of DDIs with a potential clinical signifcance (level 1 to 3) were presented in Supplementary Table S1 and S2 and the data of studies with a low level (weak or no) of DDIs were presented in Table S3.

Prescribing AB in COPD: step-by-step approach

1. Check if comorbidity is present (Table 4). 2. A quick overview on AB and its interacting medication, possible interacting mechanism, level of interaction, and practical recommendations is provided in Table 4. 3. Detailed explanation about related interacting mechanism and recommendation to manage related DDIs is provided in main text.

Mechanisms of DDI AB can act as an inhibitor/inducer and/or a substrate producing moderate to strong DDI with other co-administered medication. There are two scenarios: (1) co- administered drug alters the pharmacokinetics parameters of AB; and (2) AB infuences the pharmacokinetics parameters of co-administered medication. The main mechanisms of these DDIs are complex-forming, inhibition/ induction of drug metabolizing enzymes and alteration of drug transporters (Table 4). The ability to inhibit CYP3A4 makes the ABs prone to interact with many diferent drugs as CYP3A4 metabolizes more than 50% of the clinically prescribed drugs.13

Information structured according to drugs for comorbidities The presentation of information on potential clinically signifcant DDIs with moderate to strong level of interaction is according to the most frequent comorbidities that

86 DDIs Associated with Antibacterials for COPD

FigureFigure 1. 1. Flowchart Flowchart of studyof study selection. selection.

Information structured according to drugs for comorbidities have been reported in COPD patients.8,14 Potential mechanisms of DDIs and actionable recommendationThe presentation toof manageinformation the on DDIs potential are provided clinically in significant Table 4. DDIs with moderate to strong level of interaction is according to the most frequent comorbidities that have been 1. Diabetes reported in COPD patients.8,14 Potential mechanisms of DDIs and actionable Patients with COPD have a 50% higher risk to develop diabetes than persons without COPD.recommendation15 Some antidiabetic to manage the drugs DDIs are are provided substrates in Table of enzymes 4. like CYP3A4 (glipizide, tolbutamide),1. Diabetes CYP2C9 (glipizide, glyburide) and CYP2C8 (repaglinide); and substrates of drug transporter like P-gp transporter (glipizide, glyburide).16-26 ABs may inhibit thePatients function with ofCOPD those have metabolic a 50% higher enzymes risk to and develop transporters diabetes such than aspersons clarithromycin without (CYP3A4COPD.15 andSome P-gp antidiabetic inhibitor), drugs trimethoprim–sulfamethoxazole are substrates of enzymes (CYP2C8/2C9like CYP3A4 (glipizide,inhibitor) andtolbutamide), levofoxacin CYP2C9 (P-gp (glipizide, inhibitor). glyburide) These andmedicines CYP2C8 can(repaglinide); potentially and increase substrates the of blooddrug concentration of those antidiabetic agents.16-26 Consequently, patients may develop transporter like P‐gp transporter (glipizide, glyburide).16‐26 ABs may inhibit the function of hypoglycemia. Therefore, it is suggested to avoid these combinations by replacing relatedthose ABmetabolic or adjusting enzymes the doseand oftransporters antidiabetic such agents as clarithromycinas well as monitoring (CYP3A4 the and patients’ P‐gp bloodinhibitor), glucose. trimethoprim–sulfamethoxazole (CYP2C8/2C9 inhibitor) and levofloxacin (P‐gp

87 Chapter 5

2. Heart and circulatory system diseases 2.1. Antihypertensive agents Hypertension is associated with COPD with relative risk of 1.6.15 Antihypertensive calcium channel blocker (CCB) like diltiazem and verapamil are CYP3A4 substrates.27-29 Therefore, macrolides (CYP3A4 inhibitors) can enhance the pharmacologic activity of CCB.30 Avoiding the combination by substitution of macrolides or CCB to another group of drugs or adjusting the dose of CCB while monitoring the blood pressure is recommended. Erythromycin and clarithromycin are the most potent CYP3A4 inhibitors, while azithromycin and roxithromycin are weak inhibitors.30,31 Hence, if prescribing macrolides, choosing macrolides with minimal inhibitory capacity to be co-prescribed with CCB may minimize the risk of DDI.

Spironolactone, a potassium sparing diuretic, is used to lower blood pressure. Spironolactone and trimethoprim–sulfamethoxazole combination may produce hyperkalemia because both drugs can inhibit renal excretion of potassium.32 Therefore, avoiding combination by selecting an alternative AB or adjusting the dose of spironolactone and closely monitoring potassium plasma levels is strongly recommended.

2.2. Lipid-lowering drugs Lipid metabolism problem is one of the most prevalent comorbidities in COPD patients.14 The main pharmacologic approach to manage blood cholesterol levels is by statin therapy.33 Some ABs increase the plasma concentration of statins by several mechanisms. Statins like simvastatin and atorvastatin are bio-degraded by CYP3A4.34,35 Therefore, potent CYP3A4 inhibitors (erythromycin and clarithromycin) increase the risk for statin related side efects like rhabdomyolysis.34,35 Other statins like rosuvastatin, pravastatin and fuvastatin are not CYP3A4 substrates.36 Yet, the hepatic clearance of these statins are facilitated by anion–transporting polypeptides.37 These infux transporters facilitate the transportation of statins from systemic blood to liver cells to be metabolized or subsequently delivered into the bile for elimination.37 Clarithromycin and erythromycin have been reported to be inhibitors of these transporters.38 Therefore, replacing erythromycin and clarithromycin with other ABs, temporarily stopping statins, or adjusting the dose of statins while monitoring statin related side efects is recommended.

2.3. Oral anticoagulants Both coumarins and direct oral anticoagulants (DOACs) may interact with ABs. Multiple studies reported that DDIs between ABs with coumarins (warfarin, phenprocoumon, acenocoumarol) led to increased risks of hemorrhage.39-58 Several interacting mechanisms were proposed.59,60 One mechanism is by disruption of intestinal fora

88 DDIs Associated with Antibacterials for COPD that synthesizes vitamin K, as many ABs could alter the balance of gut fora.59 Another mechanism is that ABs (e.g. trimethoprim–sulfamethoxazole and macrolide) alter coumarins’ metabolism which mainly involves CYP2C9 and CYP3A4, respectively.60 Therefore, to choose alternative AB or if not possible, to monitor INR values and adjust the dose of coumarins is recommended.

DOACs are regarded as a safe alternative to replace coumarins.61 However, since some DOACs (edoxaban, rivaroxaban, dabigatran) are substrates of CYP3A4 and/or P-gp transporter, their AUC values can be increased by ABs like macrolides.62,63 Therefore, when macrolides and DOACs are required in combination, careful monitoring the signs 5 of bleeding is needed, and adjusting the dose of DOACs should be done if it is necessary.

2.4. Antiarrhythmic agents Some antiarrhythmic agents like digoxin, quinidine, lignocaine, and procainamide potentially interact with ABs.64-75 Quinidine and lignocaine are CYP3A4 substrates, and therefore, macrolides may inhibit their degradation and increase their bioavailabilities.64,65 Meanwhile, the renal clearance of procainamide and digoxin were inhibited by trimetophrim.66,67,72,73 Mechanism of interaction is inhibition of tubular secretion via inhibition of renal organic cation transporter because they are substrates of the transporter.66,67,72,73 Consequently, blood concentrations of these drugs are increased.66,67,72,73 Digoxin is a substrate of P-gp transporter.68-71 Accordingly, AUC of digoxin is elevated by clarithromycin and therefore, may cause toxicities.68-71 Since quinidine, lignocaine, digoxin, and procainamide are drugs with NTI, avoiding ABs that can lead to DDIs with these drugs is recommended.76,77 However, if they are necessary to be co-prescribed, therapeutic drug monitoring (TDM) of these antiarrhythmic agents is strongly recommended.77

3. Respiratory diseases 3.1. Medication for obstructive airways diseases One of the most prevalent comorbidities in COPD is asthma.14 Some anti-asthma drugs such as methylprednisolone, montelukast, loratadine, rofumilast and theophylline were substrates of CYP3A4 and/or P-gp transporter and therefore, evidenced to interact with macrolides.78-87 Hence, one might consider other ABs to be combined with asthma drugs, or closely monitor patients, especially in case of theophylline which is a NTI drug.88 As theophylline is also metabolized by CYP1A2,89 ciprofoxacin (a CYP1A2 potent inhibitor) should be avoided.90-97

3.2. Anti-mycobacterial agents Tuberculosis and COPD diseases share comparable risk factors and therefore, can coincide in individuals, particularly elderly patients.98 Rifampicin and rifabutin (anti- mycobacterial agents) work as potent inducers of hepatic and intestinal CYP enzymes.99

89 Chapter 5

Table 4. DDI of antibacterials (ABs) for COPD exacerbation and other drugs for treating its comorbidities

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

1. Diabetes Antidiabetic Glipizide, glyburide TMP-SMX Inhibition of CYP2C9 Consider alternative, adjusted dose of 2 16-19 medication Glyburide Clarithromycin Inhibition of P-gp substrate or used cautiously by monitoring patients’ blood glucose. Glipizide, glyburide Levofoxacin Inhibition of P-gp Monitor patients’ blood glucose and if 3 16, 20-26 Tolbutamide Clarithromycin Inhibition of CYP3A4 & necessary, adjusted dose of substrate. P-gp TMP-SMX Inhibition of CYP2C9 Glipizide, repaglinide Clarithromycin Inhibition of CYP3A4 Repaglinide, rosiglitazone TMP-SMX Inhibition of CYP2C8 Metformin TMP-SMX Inhibition of OCT2 & MATE1 2. Heart and circulatory system diseases 2.1 Antihypertensive Spironolactone TMP-SMX Inhibition of potassium Avoid combination or adjusted dose of 1 32 agents secretion substrates & closely monitoring potassium plasma levels. Calcium channel blocker Erythromycin, clarithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 27-29 substrate or used cautiously by monitoring side efects. Azithromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 27 dose of substrate. 2.2 Lipid-lowering Simvastatin Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 34 drugs substrates & closely monitoring side efects. Atorvastatin Clarithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 35 substrate or used cautiously by monitoring side efects. Erythromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 36, 204 Rosuvastatin/pravastatin/fuvastatin Clarithromycin Inhibition of OAT the dose of substrate. 2.3 Oral anticoagulants Warfarin, phenprocoumon / TMP-SMX Inhibition of CYP2C9 Avoid combination or closely monitor 1 39-58 acenocoumarol the change of INR routinely and adjusted the dose if needed. Amoxicillin/co-amoxiclav, ceftriaxone Alterations in normal Choose alternative AB or if not possible, 2 gut fora monitor the change of INR routinely. Clarithromycin, azithromycin, Inhibition of CYP3A4 or ciprofoxacin, levofoxacin, ofoxacin, alterations in normal doxycycline gut fora Edoxaban, dabigatran, rivaroxaban Erythromycin, clarithromycin Inhibition of CYP3A4 &/ Consider alternative/adjusted dose of 2 62, 63 or P-gp substrate or monitor the signs of excessive anticoagulant efect. Warfarin Moxifoxacin Inhibition of CYP3A4 or Monitor the change of INR routinely 3 41 alterations in normal gut fora

90 DDIs Associated with Antibacterials for COPD

Table 4. DDI of antibacterials (ABs) for COPD exacerbation and other drugs for treating its comorbidities

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

1. Diabetes Antidiabetic Glipizide, glyburide TMP-SMX Inhibition of CYP2C9 Consider alternative, adjusted dose of 2 16-19 medication Glyburide Clarithromycin Inhibition of P-gp substrate or used cautiously by monitoring patients’ blood glucose. Glipizide, glyburide Levofoxacin Inhibition of P-gp Monitor patients’ blood glucose and if 3 16, 20-26 Tolbutamide Clarithromycin Inhibition of CYP3A4 & necessary, adjusted dose of substrate. P-gp 5 TMP-SMX Inhibition of CYP2C9 Glipizide, repaglinide Clarithromycin Inhibition of CYP3A4 Repaglinide, rosiglitazone TMP-SMX Inhibition of CYP2C8 Metformin TMP-SMX Inhibition of OCT2 & MATE1 2. Heart and circulatory system diseases 2.1 Antihypertensive Spironolactone TMP-SMX Inhibition of potassium Avoid combination or adjusted dose of 1 32 agents secretion substrates & closely monitoring potassium plasma levels. Calcium channel blocker Erythromycin, clarithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 27-29 substrate or used cautiously by monitoring side efects. Azithromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 27 dose of substrate. 2.2 Lipid-lowering Simvastatin Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 34 drugs substrates & closely monitoring side efects. Atorvastatin Clarithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 35 substrate or used cautiously by monitoring side efects. Erythromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 36, 204 Rosuvastatin/pravastatin/fuvastatin Clarithromycin Inhibition of OAT the dose of substrate. 2.3 Oral anticoagulants Warfarin, phenprocoumon / TMP-SMX Inhibition of CYP2C9 Avoid combination or closely monitor 1 39-58 acenocoumarol the change of INR routinely and adjusted the dose if needed. Amoxicillin/co-amoxiclav, ceftriaxone Alterations in normal Choose alternative AB or if not possible, 2 gut fora monitor the change of INR routinely. Clarithromycin, azithromycin, Inhibition of CYP3A4 or ciprofoxacin, levofoxacin, ofoxacin, alterations in normal doxycycline gut fora Edoxaban, dabigatran, rivaroxaban Erythromycin, clarithromycin Inhibition of CYP3A4 &/ Consider alternative/adjusted dose of 2 62, 63 or P-gp substrate or monitor the signs of excessive anticoagulant efect. Warfarin Moxifoxacin Inhibition of CYP3A4 or Monitor the change of INR routinely 3 41 alterations in normal gut fora

91 Chapter 5

Table 4. (continued)

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

2.4 Antiarrhythmic Digoxin Clarithromycin Inhibition of P-gp Avoid combination or perform TDM and if 1 68-71 agent necessary, adjusted dose of substrate. Quinidine, lignocaine Erythromycin Inhibition of CYP3A4 Consider alternative or perform TDM and if 2 64-67 Procainamide TMP Inhibition of tubular necessary, adjusted dose of substrate. secretion Pindolol, digoxin TMP-SMX Inhibition of tubular Perform TDM and if necessary, adjusted dose 3 72-75 secretion of substrate. Procainamide Levofoxacin, ofoxacin Inhibition of OCT 3. Respiratory diseases 3.1. Medication for Methylprednisolone, montelukast Clarithromycin Inhibition of CYP3A4 & Consider alternative/adjusted dose of 2 78-85, 90-97 obstructive airways P-gp substrate or used cautiously by monitoring diseases Theophylline Erythromycin Inhibition of CYP3A4 side efects. For theophylline, perform TDM. Ciprofoxacin Inhibition of CYP1A2 Loratadine Erythromycin, clarithromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 86, 87 Rofumilast Erythromycin Inhibition of CYP3A4 dose of substrate. 3.2. Anti-TB drugs Rifabutin Clarithromycin Inhibition of CYP3A4 Avoid combination 1 101, 110, 111 Rifampicin, rifabutin Clarithromycin Induction of CYP3A4 Consider alternative AB for COPD 2 100, 101 Rifampicin, rifabutin TMP-SMX, doxycycline Induction of CYP3A4/ Consider alternative AB for COPD or monitor 3 102-104, CYP2C9 the efectiveness of AB and if necessary, 106-109 Rifampicin TMP-SMX Inhibition of adjusted dose of AB. mixed oxidases Moxifoxacin Inducing phase II enzymes 4. Neurological disorders 4.1. Antiparkinson Bromocriptine Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 112 Agents substrates & closely monitoring side efects. Cabergoline Clarithromycin Inhibition of CYP3A4 & Consider alternative/adjusted dose of 2 113 P-gp substrate or used cautiously by monitoring side efects. 4.2. Antiepileptic drugs Carbamazepine, phenytoin Doxycycline Induction of CYP3A4 Consider alternative or perform TDM 2 117, 116 Carbamazepine Ciprofoxacin Inhibition of Consider alternative or perform TDM 2 118 CYP3A4/1A2 Phenytoin TMP-SMX Inhibition of CYP2C8 Consider alternative or perform TDM 2 116, 119 Phenobarbital Doxycycline Induction of CYP3A4 Monitor side efects and if necessary, adjusted 3 115 dose of substrate. 5. Depression and psychiatric disorders Antidepressant, Buspirone Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 125 Anxiolytic, & substrates & closely monitoring side efects. Antipsychotic agents Quetiapine Erythromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 122-124, 129 Pimozide, trazodone Clarithromycin Inhibition of CYP3A4 substrate or used cautiously by monitoring Clozapine Ciprofoxacin Inhibition of CYP1A2 side efects. For clozapine, perform TDM. Diazepam Ciprofoxacin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 127 the dose of substrate.

92 DDIs Associated with Antibacterials for COPD

Table 4. (continued)

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

2.4 Antiarrhythmic Digoxin Clarithromycin Inhibition of P-gp Avoid combination or perform TDM and if 1 68-71 agent necessary, adjusted dose of substrate. Quinidine, lignocaine Erythromycin Inhibition of CYP3A4 Consider alternative or perform TDM and if 2 64-67 Procainamide TMP Inhibition of tubular necessary, adjusted dose of substrate. secretion Pindolol, digoxin TMP-SMX Inhibition of tubular Perform TDM and if necessary, adjusted dose 3 72-75 secretion of substrate. 5 Procainamide Levofoxacin, ofoxacin Inhibition of OCT 3. Respiratory diseases 3.1. Medication for Methylprednisolone, montelukast Clarithromycin Inhibition of CYP3A4 & Consider alternative/adjusted dose of 2 78-85, 90-97 obstructive airways P-gp substrate or used cautiously by monitoring diseases Theophylline Erythromycin Inhibition of CYP3A4 side efects. For theophylline, perform TDM. Ciprofoxacin Inhibition of CYP1A2 Loratadine Erythromycin, clarithromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 86, 87 Rofumilast Erythromycin Inhibition of CYP3A4 dose of substrate. 3.2. Anti-TB drugs Rifabutin Clarithromycin Inhibition of CYP3A4 Avoid combination 1 101, 110, 111 Rifampicin, rifabutin Clarithromycin Induction of CYP3A4 Consider alternative AB for COPD 2 100, 101 Rifampicin, rifabutin TMP-SMX, doxycycline Induction of CYP3A4/ Consider alternative AB for COPD or monitor 3 102-104, CYP2C9 the efectiveness of AB and if necessary, 106-109 Rifampicin TMP-SMX Inhibition of adjusted dose of AB. mixed oxidases Moxifoxacin Inducing phase II enzymes 4. Neurological disorders 4.1. Antiparkinson Bromocriptine Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 112 Agents substrates & closely monitoring side efects. Cabergoline Clarithromycin Inhibition of CYP3A4 & Consider alternative/adjusted dose of 2 113 P-gp substrate or used cautiously by monitoring side efects. 4.2. Antiepileptic drugs Carbamazepine, phenytoin Doxycycline Induction of CYP3A4 Consider alternative or perform TDM 2 117, 116 Carbamazepine Ciprofoxacin Inhibition of Consider alternative or perform TDM 2 118 CYP3A4/1A2 Phenytoin TMP-SMX Inhibition of CYP2C8 Consider alternative or perform TDM 2 116, 119 Phenobarbital Doxycycline Induction of CYP3A4 Monitor side efects and if necessary, adjusted 3 115 dose of substrate. 5. Depression and psychiatric disorders Antidepressant, Buspirone Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 125 Anxiolytic, & substrates & closely monitoring side efects. Antipsychotic agents Quetiapine Erythromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 122-124, 129 Pimozide, trazodone Clarithromycin Inhibition of CYP3A4 substrate or used cautiously by monitoring Clozapine Ciprofoxacin Inhibition of CYP1A2 side efects. For clozapine, perform TDM. Diazepam Ciprofoxacin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 127 the dose of substrate.

93 Chapter 5

Table 4. (continued)

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

6. Dyspepsia Antidyspepsia Aluminum hydroxide, sucralfat Quinolone, tetracyclines Complex-forming Avoid combination or administer quinolone at 1 131-142 medications least 2 hours before or 6 hours after co-agents. Lansoprazole Clarithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 147 substrate or used cautiously by monitoring side efects. Calcium carbonate Quinolone, tetracyclines Complex-forming Avoid co-administration or administration 2 131, 139 interval of at least 2 h or more Bismuth subsalicylate Quinolone, tetracyclines Complex-forming Administration interval of at least 2 h or more 3 143, 205 7. HIV Anti-HIV drugs Didanosine Ciprofoxacin Complex-forming Avoid combination or administer quinolone at 1 149, 150 least 2 hours before or 6 hours after the co-agents. Saquinavir Erythromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 151 substrate or used cautiously by monitoring side efects. Lamivudine, didanosine TMP-SMX Inhibition of tubular Monitor side efects and if necessary, adjusted 3 152, 153 secretion dose of substrate. 8. Other Pulmonary arterial Bosentan Clarithromycin Inhibition of CYP3A4 & Avoid combination or adjusted dose of 1 206 hypertension P-gp substrates & closely monitoring side efects. medications Ambrisentan Clarithromycin Inhibition of CYP3A4 & Monitor side efects and if necessary, adjusted 3 207 P-gp the dose of substrate. Insomnia medications Brotizolam, triazolam, zopiclone Erythromycin Inhibition of CYP3A4 Consider an alternative AB or other hypnotic 2 208-210 drugs (not a CYP3A4 substrate) Zolpidem Ciprofoxacin Inhibition of CYP3A4 Monitor side efects and if necessary, choose 3 211 alternative AB or other hypnotic drugs (not a CYP3A4 substrate) Antifungal agents Voriconazole Erythromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 154, 155 Itraconazole Ciprofoxacin Inhibition of CYP3A4 substrate or used cautiously by monitoring side efects. For voriconazole, perform TDM and adjust the dose if needed. Antineoplastic drugs Vinorelbine Clarithromycin Inhibition of CYP3A4 & Avoid combination or adjusted dose of 1 179 P-gp substrates & closely monitoring side efects. Anti-gout drugs Colchicine Clarithromycin Inhibition of CYP3A4 Avoid combination or perform TDM and adjust 1 180 the dose if needed. Azithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of substrate 2 180 or used cautiously by monitoring side efects. Probenecid Ciprofoxacin Inhibition of OAT Monitor side efects and if necessary, adjusted 3 194, 195 dose of substrate. Anesthesia drugs Midazolam Clarithromycin, erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 156-160 substrates & closely monitoring side efects. Ketamine Clarithromycin Inhibition of CYP3A4 Consider alternative or perform TDM and 2 161 adjust the dose if needed.

94 DDIs Associated with Antibacterials for COPD

Table 4. (continued)

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

6. Dyspepsia Antidyspepsia Aluminum hydroxide, sucralfat Quinolone, tetracyclines Complex-forming Avoid combination or administer quinolone at 1 131-142 medications least 2 hours before or 6 hours after co-agents. Lansoprazole Clarithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 147 substrate or used cautiously by monitoring side efects. Calcium carbonate Quinolone, tetracyclines Complex-forming Avoid co-administration or administration 2 131, 139 5 interval of at least 2 h or more Bismuth subsalicylate Quinolone, tetracyclines Complex-forming Administration interval of at least 2 h or more 3 143, 205 7. HIV Anti-HIV drugs Didanosine Ciprofoxacin Complex-forming Avoid combination or administer quinolone at 1 149, 150 least 2 hours before or 6 hours after the co-agents. Saquinavir Erythromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 151 substrate or used cautiously by monitoring side efects. Lamivudine, didanosine TMP-SMX Inhibition of tubular Monitor side efects and if necessary, adjusted 3 152, 153 secretion dose of substrate. 8. Other Pulmonary arterial Bosentan Clarithromycin Inhibition of CYP3A4 & Avoid combination or adjusted dose of 1 206 hypertension P-gp substrates & closely monitoring side efects. medications Ambrisentan Clarithromycin Inhibition of CYP3A4 & Monitor side efects and if necessary, adjusted 3 207 P-gp the dose of substrate. Insomnia medications Brotizolam, triazolam, zopiclone Erythromycin Inhibition of CYP3A4 Consider an alternative AB or other hypnotic 2 208-210 drugs (not a CYP3A4 substrate) Zolpidem Ciprofoxacin Inhibition of CYP3A4 Monitor side efects and if necessary, choose 3 211 alternative AB or other hypnotic drugs (not a CYP3A4 substrate) Antifungal agents Voriconazole Erythromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of 2 154, 155 Itraconazole Ciprofoxacin Inhibition of CYP3A4 substrate or used cautiously by monitoring side efects. For voriconazole, perform TDM and adjust the dose if needed. Antineoplastic drugs Vinorelbine Clarithromycin Inhibition of CYP3A4 & Avoid combination or adjusted dose of 1 179 P-gp substrates & closely monitoring side efects. Anti-gout drugs Colchicine Clarithromycin Inhibition of CYP3A4 Avoid combination or perform TDM and adjust 1 180 the dose if needed. Azithromycin Inhibition of CYP3A4 Consider alternative/adjusted dose of substrate 2 180 or used cautiously by monitoring side efects. Probenecid Ciprofoxacin Inhibition of OAT Monitor side efects and if necessary, adjusted 3 194, 195 dose of substrate. Anesthesia drugs Midazolam Clarithromycin, erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 156-160 substrates & closely monitoring side efects. Ketamine Clarithromycin Inhibition of CYP3A4 Consider alternative or perform TDM and 2 161 adjust the dose if needed.

95 Chapter 5

Table 4. (continued)

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

Alfentanil Erythromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 162-166 Ropivacaine Clarithromycin Inhibition of CYP3A4 the dose of substrate. Ciprofoxacin Inhibition of CYP1A2 Midazolam Roxithromycin Inhibition of CYP3A4 Analgesics Oxycodone Clarithromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 167 substrates & closely monitoring side efects. Immunosuppressant Cyclosporine Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 168, 169, drugs Everolimus Erythromycin Inhibition of CYP3A4 substrates & perform TDM. 181, 182 and/ P-gp Tacrolimus Levofoxacin Inhibition of CYP3A4 or Consider alternative/adjusted dose of substrate 2 170 P-gp or used cautiously by monitoring side efects. Cyclosporine Ciprofoxacin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 171, 172 the dose of substrate. Vasoactive agent Sildenafl Clarithromycin, erythromycin, Inhibition of CYP3A4 Consider alternative/adjusted dose of substrate 2 173, 174 ciprofoxacin or used cautiously by monitoring side efects. Appetite suppressant Sibutramine Clarithromycin Inhibition of CYP3A4 & Avoid combination or adjusted dose of 1 175, 176 P-gp substrates & closely monitoring side efects. Emergency birth Ulipristal acetate Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 178 control substrates & closely monitoring side efects. Antimalarial agent Halofantrine Tetracycline Probably by CYP3A4 Avoid combination or perform TDM and adjust 1 177 inhibition the dose if needed. Muscle relaxant Tizanidine Ciprofoxacin Inhibition of CYP1A2 Avoid combination or perform TDM and adjust 1 183 the dose if needed. Anti-diarrheal Loperamid TMP-SMX Inhibition of CYP2C8 Consider alternative/adjusted dose of 2 186 substrate or used cautiously by monitoring side efects. Anemia medications Iron supplements Quinolone, tetracyclines Complex-forming Avoid co-administration or administration 2 212-220 interval of at least 2 hours or more Other metal cations Zinc sulfate Quinolone, tetracyclines Complex-forming Avoid co-administration or administration 2 144, 188, 189 interval of at least 2 h or more Calcium acetate, calcium carbonate, Quinolone, tetracyclines Complex-forming Administration interval of at least 2 h or more 3 139, 190-193 calcium polycarbophil, patiromer, lanthanum carbonate, sevelamer Other ABs Linezolid Clarithromycin Inhibition of P-gp Consider alternative or perform TDM and 2 196 adjust the dose if needed. Dapson Trimethoprim Inhibition of CYP2C8 Monitor side efects and if necessary, adjusted 3 187 the dose of substrate. Neomycin Penicillin V NA Consider alternative or adjusted the dose of 3 221 penicillin.

Defnition of level of interaction: 1 = strong interaction, 2 = substantial interaction, 3 = moderate interaction, and 4 = weak/ multidrug and toxin extrusion 1; P-gp: P-glycoprotein; TMP-SMX= Trimethoprim and Sulfonamides; TDM = therapeutic drug no interaction; Ref. = reference; h = hour; OCT= organic cation transporter; OAT= Organic anion transporter; MATE= monitoring; NA = not available yet. All detailed supported information about each DDI were available in Table S1 and S2.

96 DDIs Associated with Antibacterials for COPD

Table 4. (continued)

Level of Comorbidity Medication Interacting AB Mechanism Management suggestion interaction Ref.

Alfentanil Erythromycin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 162-166 Ropivacaine Clarithromycin Inhibition of CYP3A4 the dose of substrate. Ciprofoxacin Inhibition of CYP1A2 Midazolam Roxithromycin Inhibition of CYP3A4 Analgesics Oxycodone Clarithromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 167 substrates & closely monitoring side efects. Immunosuppressant Cyclosporine Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 168, 169, 5 drugs Everolimus Erythromycin Inhibition of CYP3A4 substrates & perform TDM. 181, 182 and/ P-gp Tacrolimus Levofoxacin Inhibition of CYP3A4 or Consider alternative/adjusted dose of substrate 2 170 P-gp or used cautiously by monitoring side efects. Cyclosporine Ciprofoxacin Inhibition of CYP3A4 Monitor side efects and if necessary, adjusted 3 171, 172 the dose of substrate. Vasoactive agent Sildenafl Clarithromycin, erythromycin, Inhibition of CYP3A4 Consider alternative/adjusted dose of substrate 2 173, 174 ciprofoxacin or used cautiously by monitoring side efects. Appetite suppressant Sibutramine Clarithromycin Inhibition of CYP3A4 & Avoid combination or adjusted dose of 1 175, 176 P-gp substrates & closely monitoring side efects. Emergency birth Ulipristal acetate Erythromycin Inhibition of CYP3A4 Avoid combination or adjusted dose of 1 178 control substrates & closely monitoring side efects. Antimalarial agent Halofantrine Tetracycline Probably by CYP3A4 Avoid combination or perform TDM and adjust 1 177 inhibition the dose if needed. Muscle relaxant Tizanidine Ciprofoxacin Inhibition of CYP1A2 Avoid combination or perform TDM and adjust 1 183 the dose if needed. Anti-diarrheal Loperamid TMP-SMX Inhibition of CYP2C8 Consider alternative/adjusted dose of 2 186 substrate or used cautiously by monitoring side efects. Anemia medications Iron supplements Quinolone, tetracyclines Complex-forming Avoid co-administration or administration 2 212-220 interval of at least 2 hours or more Other metal cations Zinc sulfate Quinolone, tetracyclines Complex-forming Avoid co-administration or administration 2 144, 188, 189 interval of at least 2 h or more Calcium acetate, calcium carbonate, Quinolone, tetracyclines Complex-forming Administration interval of at least 2 h or more 3 139, 190-193 calcium polycarbophil, patiromer, lanthanum carbonate, sevelamer Other ABs Linezolid Clarithromycin Inhibition of P-gp Consider alternative or perform TDM and 2 196 adjust the dose if needed. Dapson Trimethoprim Inhibition of CYP2C8 Monitor side efects and if necessary, adjusted 3 187 the dose of substrate. Neomycin Penicillin V NA Consider alternative or adjusted the dose of 3 221 penicillin.

97 Chapter 5

They could markedly reduce the ABs activities of clarithromycin, doxycycline, and trimethoprim–sulfamethoxazole by rapidly elimination.100-104 Since, rifampicin also exhibits other ABs activities such as treating methicillin-resistant staphylococcus aureus (MRSA) in combination with other drugs, rationalizing antimicrobial therapy should be considered accordingly.105 Alternative AB for treating COPD is also recommended to reduce the risk of treatment failures.

Moxifoxacin might be an alternative AB as the evidence of moxifoxacin interaction with rifampicin was not consistent with moderate or weak interactions.106-109 Moxifoxacin is not metabolized by CYP450 and its interacting mechanisms with rifampicin might be facilitated by induction of other enzymes like uridine diphosphate- glucuronosyltransferases and sulfotransferases.106-109

Rifabutin and rifampicin are CYP substrates. Rifabutin is a CYP3A4 substrate, and therefore, macrolides may increase its serum concentration and enhance the risk of related ADR.101,110,111 Another study reported rifampicin concentration in blood is moderately elevated by co-trimoxazole.104 It was assumed that the interaction was facilitated by inhibition of the mixed function oxidases, which is responsible for metabolizing rifampicin.104 Thus, considering alternative AB or monitoring the clinical and biochemical parameters for rifampicin related hepatotoxicity is suggested when rifampicin and co-trimoxazole are combined.

What need to be mentioned is that not all the drugs for atypical mycobacterium spp were included in this review due to the selection limitation of ABs that used frequently among COPD patients. For drugs outside the scope of this review, other references (e.g. SPCs) need to be considered.

4. Neurological disorders 4.1. Anti-Parkinson drugs Bromocriptine and cabergoline (dopamine agonists) are substrates of CYP3A4 and/ or P-gp transporter.112,113 Co-prescription of these drugs with clarithromycin and erythromycin may produce major interactions and therefore, might lead to toxicities.112,113 Thus, avoiding combination is recommended. However, if it is not possible, adjusting the dose of those Parkinson medication and closely monitoring side efects are needed.

4.2. Antiepileptic drugs Carbamazepine, phenytoin, and phenobarbital could stimulate the activity of a variety of CYP (CYP1A2/2C9/3A4) and glucuronyl transferase enzymes, which results in multiple DDIs with other substrates for these enzymes.114-116 Carbamazepine and phenytoin were reported to reduce t1/2 of doxycycline by stimulating the hepatic metabolism of

98 DDIs Associated with Antibacterials for COPD doxycycline.117 It is suggested to consider an alternative AB or to adjust the dose of antiepileptic drugs while monitoring the AB activity of doxycycline.

Moreover, carbamazepine and phenytoin are substrates of CYP1A2/3A4 and CYP2C8, respectively. A CYP1A2/3A4 inhibitor (ciprofoxacin) and a CYP2C8 inhibitor (trimethoprim) were reported to increase the bioavailability of carbamazepine and phenytoin, respectively.116-119 Moreover, phenytoin is a NTI drug and therefore, avoiding using trimethoprim concomitantly or performing TDM of phenytoin is recommended when this DDI is not avoidable.120 5 Meanwhile, ciprofoxacin was reported to increase AUC of carbamazepine by more than 50%.118 Although it is not clear whether carbamazepine can be included as a NTI drug, the rise of carbamazepine plasma concentration because of this DDI needs special caution.121 Dose adjustment and TDM of carbamazepine are suggested to diminish potential toxicities.

5. Depression and psychiatric disorders Depression and psychiatric disorders are common among COPD patients.14 Some antidepressant (trazodone), anxiolytic (buspirone), and antipsychotic (quetiapine, and pimozide) drugs are CYP3A4 substrates and therefore, might trigger clinically relevant DDIs with ABs.122-125 Erythromycin and clarithromycin increased AUCs of these drugs substantially.122-125 Considering alternative AB or adjusting the dose of substrates and monitoring related side efects is the way to control potential ADR.

CYP3A4 is also responsible for metabolizing diazepam, in addition to CYP2C19.126 Ciprofoxacin was reported to decrease diazepam clearance moderately by inhibiting CYP3A4 activity.127 Monitoring diazepam-related side efects can therefore be considered when this combination is prescribed.

Ciprofoxacin is also a potent CYP1A2 inhibitor.128 Therefore, metabolism of an atypical antipsychotic clozapine, a CYP1A2 substrate with NTI, can be relevantly altered by ciprofoxacin which produces a signifcant increase of clozapine serum concentration.129,130 Replacing ciprofoxacin or TDM of clozapine is option that can be chosen in managing this DDI.

6. Dyspepsia Drugs containing metal cations (e.g. antacids, sucralfate and bismuth salts) produced chemical interactions with some ABs like oral tetracyclines (e.g. tetracycline, doxycycline) and fuoroquinolones (e.g. ciprofoxacin, moxifoxacin).131-144 Tetracyclines have a high afnity to form chelates due to their structural features with lots of chelation sites.145

99 Chapter 5

Meanwhile, fuoroquinolones have two main sites of metal chelation: 4-keto oxygen and 3-carboxylic acid groups.146

The formation of metal ion chelation complexes decreased absorption of tetracycline and fuoroquinolones, the reduced bioavailability may lead to inefectiveness of these ABs.131-144 Therefore, it is recommended to avoid combination by replacing tetracyclines and fuoroquinolones with another AB, e.g. amoxicillin or amoxicillin/clavulanic acid. It was reported that antacids did not afect the bioavailability of amoxicillin and amoxicillin/clavulanic acid when they were co-administered.136 If replacement of the AB is not possible, substitution of antacids, sucralfate or bismuth salts to PPI is also favored. Another alternative is to separate administration by using quinolone or tetracycline at least 2 hours before and 6 hours after the dyspepsia drugs.

When considering a PPI, lansoprazole may not be the best alternative as it is partly metabolized by CYP3A4 and found to interact with clarithromycin.147

7. HIV HIV-positive patients have about 50% higher risk to develop COPD than HIV-negative patients.148 Then, the risk of co-prescriptions for treating those chronic conditions is also possibly high. A protease inhibitor (saquinavir) and nucleoside reverse transcriptase inhibitors (didanosine and lamivudine) were found to clinically interact with ABs.149-153

Didanosine is very acid sensitive, and therefore, the didanosine formulations are supplemented by bufering mixtures containing magnesium hydroxide, dihydroxyaluminum sodium carbonate, and sodium citrate to prevent hydrolysis by gastric acid.149 These metal ions may form chelation complexes with quinolones and reduce their serum concentration.149,150 Two studies confrmed the didanosine and ciprofoxacin interaction, and recommended that when co-administration cannot be avoided, ciprofoxacin must be given at least 2 hours before didanosine.149,150

Trimethoprim–sulfamethoxazole may inhibit clearances of didanosine and lamivudine by competitively hinder their renal secretion.152,153 Consequently, AUCs of didanosine and lamivudine elevate moderately.152,153 Monitoring of the presumed side efects should be performed.

Saquinafr is metabolized by CYP3A4 and the presence of erythromycin increased its AUC by almost 100%.151 Choosing an alternative AB or adjusting the dose of saquinafr while monitoring toxicities can be considered to manage this DDI.

8. Other potential clinically signifcant DDI Some other drugs that have indications for comorbidities in COPD patients were found to interact with ABs. Some individual drugs of diferent classes (e.g. voriconazole and

100 DDIs Associated with Antibacterials for COPD vinorelbine) are metabolized by CYP3A4.154-182 Therefore, their metabolism is interfered by CYP3A4 inhibitors (macrolides).154-182 Other drugs are CYP1A2 substrates (e.g. ropivacaine and tizanidine) and therefore, CYP1A2 potent inhibitors like quinolones signifcantly alter their metabolism and elevate their bioavailabilities.164,183-185 Others are CYP2C8 substrates (e.g. loperamid for diarrhea) and therefore, trimethoprim (a CYP2C8 potent inhibitor) inhibits their clearance and increases their AUC values.186,187 Some drugs containing metal cations (e.g. Fe, Zn, Ca) should be avoided or administered separately at least 2 hours or more with quinolones and tetracyclines.139,144,188-193 Other interactions were facilitated by drug transporters. An uricosuric agent (probenecid) interacts moderately with ciprofoxacin via competitive inhibition of organic anion 5 transporters in renal tubules.194,195 Meanwhile, linezolid (other AB), which is a substrate of P-gp transporter, can potentially produce clinically signifcant interaction with P-gp inhibitors (macrolides).196

DDI related to NTIs Some ABs may interact with NTI drugs and therefore, can produce serious ADRs. The NTI drugs in this review includes CYP3A4 substrates (theophylline, ketamine, everolimus, tacrolimus, halofantrine, lignocaine, quinidine, voriconazole, carbamazepine, warfarin, cyclosporine, colchicine, phenprocoumon/acenocoumarol); CYP1A2 substrates (theophylline, carbamazepine, clozapine, tizanidine); CYP2C9 substrates and sensitive to alterations in normal gut fora (warfarin, phenprocoumon/acenocoumarol); a CYP2C8 substrate (phenytoin); substrates of P-gp transporter (digoxin, linezolid); and a substrate of organic cation transporter (procainamide).76,77,88,120,197 DISCUSSION Included articles This study outlined the possible DDIs related to frequently prescribed ABs in COPD patients from clinical and observational studies. We only included well-designed studies (2 points or higher) since they provide more valid evidence than studies without a control or comparison group (0 or 1 point). DDIs based on case-reports or hypotheses may lead to unnecessary warnings if these are not confrmed by well-designed studies. One classic example at this point is ABs and oral contraceptive interactions; lots of cases reported unintended pregnancies after ABs were prescribed to women on oral contraceptives, which attracted much attention from health practitioners.198,199 After scientifc evidence from clinical and pharmacokinetic studies has consistently and repeatedly failed to support such interaction, the warning about DDIs between hormonal contraception and non-rifampicin ABs were fnally canceled by related guidelines.199

101 Chapter 5

Mechanisms of DDI The DDI of potential clinical signifcance between AB and co-administered medication may occur in two situations: (1) co-administered drug infuences the absorption, distribution, metabolism, and elimination (ADME) of AB; and (2) AB infuences the ADME of co-administered medication. When AB acted as substrates, some co-administered drugs reduced the blood concentration of AB and led to treatment failure of AB in reducing exacerbations. Other co-administered drugs increased the blood concentration of AB, which could result in the termination of AB use because of an ADR, and therefore acted against the control of infections. As inhibitors, the blood concentrations of co- administered drugs were increased by AB which may also produce an ADR and lead to termination of co-administered drugs, and therefore, may produce a treatment failure of comorbidities. In all, DDIs related to ABs may hinder efective infection control and exacerbation management among COPD patients as well as treatment of comorbidities in COPD.

Comorbidities among COPD patients The impact of comorbidities on quality of life in COPD patients are well reported, however, potential drug interactions between drugs for these comorbidities and ABs used for COPD has received little specifc attention. From this review, we found that many drugs (e.g. those used for heart and circulatory system disease) should not be co- administered with related AB and other actions such as dose adjustment, choosing an alternative drug and monitoring ADRs are necessary. These drug interactions could not only infuence treatment options of clinical practitioner but also infuence treatment efects for both COPD and comorbidities.

Information collected from this review can be used as input to improve the sensitivity and specifcity of drug-drug interaction alert systems. Moreover, this study may also be attractive for researchers in this feld who may take into account the availability of high- quality studies when evaluating the evidence for many potential interactions.

Special warning for NTI drugs We found that some NTI drugs might potentially interact with ABs. Because of the narrow separation between efective and toxic dosing of these drugs, small alteration on their pharmacokinetics parameters can produce fatal consequences.88,120 Therefore, combination of particular AB which have an ability to inhibit their clearance pathways should be avoided if it is possible. However, if the benefts of combination outweigh the potential side efects, dose adjustment and performing TDM of the NTI drugs are strongly recommended.

102 DDIs Associated with Antibacterials for COPD

Limitations Some limitations to this review are worth to be mentioned. First, although we reviewed a signifcant part of the literature, we did not include all sources that might indicate relevant DDIs such as case reports, summary of product characteristics or theoretical hypotheses. As a result, we did not fnd some DDIs that are considered serious and clinically highly relevant, such as QT-interval prolonging interactions for combinations of macrolides with other QT-prolonging drugs or the risk for pseudotumor cerebri in case of combination of doxycycline with vitamin-A analogs.200,201 Such interactions are commonly found as case reports, as it is unethical to design studies to confrm these serious risks in clinical studies. However for some DDIs, it is possible to study 5 the clinical manifestation of a potential DDI in an observational study using a real world drug utilization data.202 Secondly, selection of ABs included in this review was based on their frequent use in COPD and therefore, information for other ABs used for COPD comorbidities such as atypical Mycobacterium spp is limited and therefore, may restrict the application scope of this review. Thirdly, due to limited comparative analyses for several specifc DDIs included in this review, it may be difcult to make recommendations for a specifc situation. Our classifcation of DDIs levels just ofers a general consideration. The specifc impact of a DDI is decided by many variables like diferent doses and formulations, the comorbidities of patients, etc. Therefore, case- by-case analysis is important in clinical practice and a drug interaction handbook like Stockley’s Drug Interactions further expands on these issues.203 CONCLUSION

Clinically signifcant DDIs related to ABs may involve a wide range of indicated drugs to treat comorbidities in COPD. Clinicians should pay attention to these drug interactions when prescribing AB to reduce the frequency and severity of exacerbations in COPD patients and take necessary actions to ensure therapeutic efect and safety of patients. This study may contribute to better prescribing of ABs to COPD patients with comorbidities using potentially interacting combination. Furthermore, the information may be used to point at gaps in scientifc knowledge about potential adverse efects from DDIs. SUPPLEMENTARY DATA

Table S1 to S3 are available as Supplementary data at JAC Online (https://doi. org/10.1093/jac/dkz221)

103 Chapter 5

REFERENCES

1. Vogelmeier CF, Criner GJ, Martinez FJ 11. van Roon EN, Flikweert S, le Comte M et al. et al. Global strategy for the diagnosis, Clinical relevance of drug-drug interactions : management, and prevention of chronic A structured assessment procedure. Drug obstructive lung disease 2017 report. GOLD Saf 2005; 28: 1131-9. executive summary. Am J Respir Crit Care 12. Bahar MA, Setiawan D, Hak E et al. Med 2017; 195: 557-82. Pharmacogenetics of drug-drug interaction 2. Decramer M, Janssens W, Miravitlles M. and drug-drug-gene interaction: A systematic Chronic obstructive pulmonary disease. review on CYP2C9, CYP2C19 and CYP2D6. Lancet 2012; 379: 1341-51. Pharmacogenomics 2017; 18: 701-39. 3. Sethi S, Murphy TF. Infection in 13. Zhou S, Chan E, Li X et al. Clinical outcomes the pathogenesis and course of chronic and management of mechanism-based obstructive pulmonary disease. N Engl J inhibition of cytochrome P450 3A4. Ther Med 2008; 359: 2355-65. Clin Risk Manag 2005; 1: 3-13. 4. Vollenweider DJ, Jarrett H, Steurer-Stey CA 14. Divo M, Cote C, de Torres JP et al. et al. Antibiotics for exacerbations of chronic Comorbidities and risk of mortality obstructive pulmonary disease. Cochrane in patients with chronic obstructive Database Syst Rev 2012; 12: CD010257. pulmonary disease. Am J Respir Crit Care 5. Herath SC, Poole P. Prophylactic antibiotic Med 2012; 186: 155-61. therapy for chronic obstructive pulmonary 15. Mannino DM, Thorn D, Swensen A et al. disease (COPD). Cochrane Database Syst Prevalence and outcomes of diabetes, Rev 2013; CD009764. hypertension and cardiovascular disease in 6. Wang Y, Zijp TR, Bahar MA et al. Efects of COPD. Eur Respir J 2008; 32: 962-9. prophylactic antibiotics on patients with 16. Schelleman H, Bilker WB, Brensinger CM stable COPD: A systematic review and meta- et al. Anti-infectives and the risk of severe analysis of randomized controlled trials. J hypoglycemia in users of glipizide or glyburide. Antimicrob Chemother 2018; 73: 3231-43. Clin Pharmacol Ther 2010; 88: 214-22. 7. Wedzicha JA, Calverley PMA, Albert RK 17. Lilja JJ, Niemi M, Fredrikson H et al. Efects et al. Prevention of COPD exacerbations: of clarithromycin and grapefruit juice on A european respiratory Society/American the pharmacokinetics of glibenclamide. Br J thoracic society guideline. Eur Respir J 2017; 50: Clin Pharmacol 2007; 63: 732-40. 10.1183/13993003.02265,2016. Print 2017 Sep. 18. Tan A, Holmes HM, Kuo YF et al. 8. Chetty U, McLean G, Morrison D et al. Coadministration of co-trimoxazole with Chronic obstructive pulmonary disease and sulfonylureas: Hypoglycemia events and comorbidities: A large cross-sectional study in pattern of use. J Gerontol A Biol Sci Med primary care. Br J Gen Pract 2017; 67: e321-8. Sci 2015; 70: 247-54. 9. Mangoni AA, Jackson SH. Age-related changes 19. Kradjan WA, Witt DM, Opheim KE et al. Lack in pharmacokinetics and pharmacodynamics: Basic principles and practical applications. Br J of interaction between glipizide and co- Clin Pharmacol 2004; 57: 6-14. trimoxazole. J Clin Pharmacol 1994; 34: 997-1002. 10. Bahar MA, Hak E, Bos JH et al. The burden 20. Jayasagar G, Dixit AA, Kishan V et al. Efect and management of cytochrome P450 2D6 of clarithromycin on the pharmacokinetics (CYP2D6)‐mediated drug–drug interaction of tolbutamide. Drug Metabol Drug (DDI): Co‐medication of metoprolol and Interact 2000; 16: 207-15. paroxetine or fuoxetine in the elderly. 21. Wing LM, Miners JO. Cotrimoxazole as an Pharmacoepidemiol Drug Saf 2017; 26: 752-65. inhibitor of oxidative drug metabolism: Efects

104 DDIs Associated with Antibacterials for COPD

of trimethoprim and sulphamethoxazole 31. Yamazaki H, Shimada T. Comparative studies separately and combined on tolbutamide of in vitro inhibition of cytochrome P450 3A4- disposition. Br J Clin Pharmacol 1985; 20: 482-5. dependent testosterone 6beta-hydroxylation 22. Niemi M, Neuvonen PJ, Kivisto KT. by roxithromycin and its metabolites, The cytochrome P4503A4 inhibitor troleandomycin, and erythromycin. Drug Metab Dispos 1998; 26: 1053-7. clarithromycin increases the plasma concentrations and efects of repaglinide. 32. Antoniou T, Hollands S, Macdonald EM et Clin Pharmacol Ther 2001; 70: 58-65. al. Trimethoprim-sulfamethoxazole and risk of sudden death among patients taking 23. Niemi M, Kajosaari LI, Neuvonen M et spironolactone. CMAJ 2015; 187: E138-43. al. The CYP2C8 inhibitor trimethoprim increases the plasma concentrations of 33. Parhofer KG. The treatment of disorders 5 repaglinide in healthy subjects. Br J Clin of lipid metabolism. Dtsch Arztebl Pharmacol 2004; 57: 441-7. Int 2016; 113: 261-8. 24. Hruska MW, Amico JA, Langaee TY et al. 34. Kantola T, Kivisto KT, Neuvonen PJ. Erythromycin and verapamil considerably The efect of trimethoprim on CYP2C8 increase serum simvastatin and simvastatin mediated rosiglitazone metabolism in acid concentrations. Clin Pharmacol human liver microsomes and healthy Ther 1998; 64: 177-82. subjects. Br J Clin Pharmacol 2005; 59: 70-9. 35. Amsden GW, Kuye O, Wei GC. A study of 25. Grun B, Kiessling MK, Burhenne J et al. the interaction potential of azithromycin and Trimethoprim-metformin interaction and clarithromycin with atorvastatin in healthy its genetic modulation by OCT2 and MATE1 volunteers. J Clin Pharmacol 2002; 42: 444-9. transporters. Br J Clin Pharmacol 2013; 76: 787-96. 36. Li DQ, Kim R, McArthur E et al. Risk of adverse 26. Muller F, Pontones CA, Renner B et al. N(1)- events among older adults following co- methylnicotinamide as an endogenous prescription of clarithromycin and statins probe for drug interactions by renal cation not metabolized by cytochrome P450 3A4. transporters: Studies on the metformin- CMAJ 2015; 187: 174-80. trimethoprim interaction. Eur J Clin 37. Kalliokoski A, Niemi M. Impact of OATP Pharmacol 2015; 71: 85-94. transporters on pharmacokinetics. Br J 27. Wright AJ, Gomes T, Mamdani MM et al. The risk Pharmacol 2009; 158: 693-705. of hypotension following co-prescription of 38. Seithel A, Eberl S, Singer K et al. The infuence macrolide antibiotics and calcium-channel of macrolide antibiotics on the uptake blockers. CMAJ 2011; 183: 303-7. of organic anions and drugs mediated 28. Gandhi S, Fleet JL, Bailey DG et al. Calcium- by OATP1B1 and OATP1B3. Drug Metab channel blocker-clarithromycin drug Dispos 2007; 35: 779-86. interactions and acute kidney injury. 39. Glasheen JJ, Fugit RV, Prochazka AV. The risk JAMA 2013; 310: 2544-53. of overanticoagulation with antibiotic use 29. Fraser LA, Sharif SZ, McArthur E et al. in outpatients on stable warfarin regimens. Calcium channel blocker-clarithromycin J Gen Intern Med 2005; 20: 653-6. drug interaction did not increase the risk of 40. Lane MA, Zeringue A, McDonald JR. Serious nonvertebral fracture: A population-based bleeding events due to warfarin and study. Ann Pharmacother 2015; 49: 185-8. antibiotic co-prescription in a cohort of 30. Westphal JF. Macrolide–induced clinically veterans. Am J Med 2014; 127: 657. relevant drug interactions with cytochrome 41. Ghaswalla PK, Harpe SE, Tassone D et al. P‐450A (CYP) 3A4: An update focused on Warfarin-antibiotic interactions in older clarithromycin, azithromycin and dirithromycin. adults of an outpatient anticoagulation clinic. Br J Clin Pharmacol 2000; 50: 285-95. Am J Geriatr Pharmacother 2012; 10: 352-60.

105 Chapter 5

42. Mergenhagen KA, Olbrych PM, Mattappallil potential interaction between levofoxacin A et al. Efect of azithromycin on and warfarin. Pharm World Sci 2009; 31: 224-9. anticoagulation-related outcomes in 53. Liao S, Palmer M, Fowler C et al. Absence geriatric patients receiving warfarin. Clin of an efect of levofoxacin on warfarin Ther 2013; 35: 425-30. pharmacokinetics and anticoagulation in male 43. McCall KL, Anderson HG,Jr., Jones AD. volunteers. J Clin Pharmacol 1996; 36: 1072-7. Determination of the lack of a drug 54. Stroud LF, Mamdami MM, Kopp A et al. interaction between azithromycin and The safety of levofoxacin in elderly patients warfarin. Pharmacotherapy 2004; 24: 188-94. on warfarin. Am J Med 2005; 118: 1417. 44. Fischer HD, Juurlink DN, Mamdani MM et al. 55. Penning-van Beest, van Meegen E, Hemorrhage during warfarin therapy associated Rosendaal FR et al. Drug interactions with cotrimoxazole and other urinary tract anti- as a cause of overanticoagulation on infective agents: A population-based study. Arch phenprocoumon or acenocoumarol Intern Med 2010; 170: 617-21. predominantly concern antibacterial drugs. 45. Washington C, Hou SY, Hughes NC et al. Clin Pharmacol Ther 2001; 69: 451-7. Ciprofoxacin prolonged-release tablets 56. Jobski K, Behr S, Garbe E. Drug interactions do not afect warfarin pharmacokinetics with phenprocoumon and the risk of serious and pharmacodynamics. J Clin haemorrhage: A nested case-control study in Pharmacol 2007; 47: 1320-6. a large population-based german database. 46. Israel DS, Stotka J, Rock W et al. Efect of Eur J Clin Pharmacol 2011; 67: 941-51. ciprofoxacin on the pharmacokinetics and 57. Abbas S, Ihle P, Harder S et al. Risk of bleeding pharmacodynamics of warfarin. Clin Infect and antibiotic use in patients receiving Dis 1996; 22: 251-6. continuous phenprocoumon therapy. 47. Bianco TM, Bussey HI, Farnett LE et al. Potential A case-control study nested in a large warfarin-ciprofoxacin interaction in patients insurance- and population-based german receiving long-term anticoagulation. cohort. Thromb Haemost 2014; 111: 912-22. Pharmacotherapy 1992; 12: 435-9. 58. Penning-van Beest, Koerselman J, Herings 48. Abdel-Aziz MI, Ali MA, Hassan AK et al. RM. Risk of major bleeding during Warfarin-drug interactions: An emphasis concomitant use of antibiotic drugs on infuence of polypharmacy and high and coumarin anticoagulants. J Thromb doses of amoxicillin/clavulanate. J Clin Haemost 2008; 6: 284-90. Pharmacol 2016; 56: 39-46. 59. Juurlink DN. Drug interactions with 49. Schelleman H, Bilker WB, Brensinger CM warfarin: What clinicians need to know. et al. Warfarin with fuoroquinolones, CMAJ 2007; 177: 369-71. sulfonamides, or azole antifungals: 60. Ufer M. Comparative pharmacokinetics Interactions and the risk of hospitalization of vitamin K antagonists - warfarin, for gastrointestinal bleeding. Clin Pharmacol phenprocoumon and acenocoumarol. Clin Ther 2008; 84: 581-8. Pharmacokinet 2005; 44: 1227-46. 50. Baillargeon J, Holmes HM, Lin YL et al. 61. Thachil J. The newer direct oral Concurrent use of warfarin and antibiotics anticoagulants: A practical guide. Clin Med and the risk of bleeding in older adults. Am (Lond) 2014; 14: 165-75. J Med 2012; 125: 183-9. 62. Parasrampuria DA, Mendell J, Shi M et al. 51. O’Reilly RA, Motley CH. Racemic warfarin and Edoxaban drug-drug interactions with trimethoprim-sulfamethoxazole interaction ketoconazole, erythromycin, and cyclosporine. in humans. Ann Intern Med 1979; 91: 34-6. Br J Clin Pharmacol 2016; 82: 1591-600. 52. Mercadal Orfla G, Gracia Garcia B, Leiva 63. Gouin-Thibault I, Delavenne X, Blanchard A et Badosa E et al. Retrospective assessment of al. Interindividual variability in dabigatran

106 DDIs Associated with Antibacterials for COPD

and rivaroxaban exposure: Contribution 74. Martin DE, Shen J, Griener J et al. Efects of ABCB1 genetic polymorphisms and of ofoxacin on the pharmacokinetics and interaction with clarithromycin. J Thromb pharmacodynamics of procainamide. J Clin Haemost 2017; 15: 273-83. Pharmacol 1996; 36: 85-91. 64. Damkier P, Hansen LL, Brosen K. Efect 75. Ujhelyi MR, Bottorf MB, Schur M et al. Aging of diclofenac, disulfram, itraconazole, efects on the organic base transporter grapefruit juice and erythromycin on and stereoselective renal clearance. Clin the pharmacokinetics of quinidine. Br J Clin Pharmacol Ther 1997; 62: 117-28. Pharmacol 1999; 48: 829-38. 76. Raebel MA, Carroll NM, Andrade SE et 65. Isohanni MH, Neuvonen PJ, Olkkola KT. al. Monitoring of drugs with a narrow Efect of erythromycin and itraconazole on therapeutic range in ambulatory care. Am J 5 the pharmacokinetics of oral lignocaine. Manag Care 2006; 12: 268-75. Pharmacol Toxicol 1999; 84: 143-6. 77. Campbell T, Williams K. Therapeutic drug 66. Kosoglou T, Rocci ML,Jr., Vlasses PH. monitoring: Antiarrhythmic drugs. Br J Clin Trimethoprim alters the disposition of Pharmacol 2001; 52: 21-33. procainamide and N-acetylprocainamide. 78. Fost DA, Leung DY, Martin RJ et al. Inhibition Clin Pharmacol Ther 1988; 44: 467-77. of methylprednisolone elimination in 67. Vlasses PH, Kosoglou T, Chase SL et al. the presence of clarithromycin therapy. J Trimethoprim inhibition of the renal clearance Allergy Clin Immunol 1999; 103: 1031-5. of procainamide and N-acetylprocainamide. 79. Hegazy SK, Mabrouk MM, Elsisi AE et al. Arch Intern Med 1989; 149: 1350-3. Efect of clarithromycin and fuconazole on 68. Zapater P, Reus S, Tello A et al. A prospective the pharmacokinetics of montelukast in human study of the clarithromycin-digoxin volunteers. Eur J Clin Pharmacol 2012; 68: 1275-80. interaction in elderly patients. J Antimicrob Chemother 2002; 50: 601-6. 80. Branigan TA, Robbins RA, Cady WJ et al. The efects of erythromycin on 69. Chan AL, Wang MT, Su CY et al. Risk the absorption and disposition of of digoxin intoxication caused by kinetics of theophylline. Eur J Clin clarithromycin-digoxin interactions in heart Pharmacol 1981; 21: 115-20. failure patients: A population-based study. Eur J Clin Pharmacol 2009; 65: 1237-43. 81. Reisz G, Pingleton SK, Melethil S et al. The efect of erythromycin on theophylline 70. Tanaka H, Matsumoto K, Ueno K et pharmacokinetics in chronic bronchitis. Am al. Efect of clarithromycin on steady- Rev Respir Dis 1983; 127: 581-4. state digoxin concentrations. Ann Pharmacother 2003; 37: 178-81. 82. Prince RA, Wing DS, Weinberger MM et al. Efect of erythromycin on theophylline kinetics. J 71. Rengelshausen J, Goggelmann C, Allergy Clin Immunol 1981; 68: 427-31. Burhenne J et al. Contribution of increased oral bioavailability and reduced 83. May DC, Jarboe CH, Ellenburg DT et al. nonglomerular renal clearance of digoxin The efects of erythromycin on theophylline to the digoxin-clarithromycin interaction. Br elimination in normal males. J Clin J Clin Pharmacol 2003; 56: 32-8. Pharmacol 1982; 22: 125-30. 72. Petersen P, Kastrup J, Bartram R et al. 84. Maddux MS, Leeds NH, Organek HW et al. Digoxin-trimethoprim interaction. Acta Med The efect of erythromycin on theophylline Scand 1985; 217: 423-7. pharmacokinetics at steady state. 73. Bauer LA, Black DJ, Lill JS et al. Levofoxacin Chest 1982; 81: 563-5. and ciprofoxacin decrease procainamide 85. Pfeifer HJ, Greenblatt DJ, Friedman P. Efects and N-acetylprocainamide renal clearances. of three antibiotics on theophylline kinetics. Antimicrob Agents Chemother 2005; 49: 1649-51. Clin Pharmacol Ther 1979; 26: 36-40.

107 Chapter 5

86. Carr RA, Edmonds A, Shi H et al. Steady-state toxicity: A population-based study. Eur J Clin pharmacokinetics and electrocardiographic Pharmacol 2011; 67: 521-6. pharmacodynamics of clarithromycin and 97. Batty KT, Davis TM, Ilett KF et al. loratadine after individual or concomitant The efect of ciprofoxacin on theophylline administration. Antimicrob Agents pharmacokinetics in healthy subjects. Br J Chemother 1998; 42: 1176-80. Clin Pharmacol 1995; 39: 305-11. 87. Brannan MD, Reidenberg P, Radwanski E et 98. Inghammar M, Ekbom A, Engström al. Loratadine administered concomitantly G et al. COPD and the risk of with erythromycin: Pharmacokinetic and tuberculosis-a population-based cohort electrocardiographic evaluations. Clin study. PloS One 2010; 5: e10138. Pharmacol Ther 1995; 58: 269-78. 99. Baciewicz AM, Chrisman CR, Finch CK 88. Blix HS, Viktil KK, Moger TA et al. Drugs with et al. Update on rifampin, rifabutin, and narrow therapeutic index as indicators rifapentine drug interactions. Curr Med Res in the risk management of hospitalised Opin 2013; 29: 1-12. patients. Pharm Pract (Granada) 2010; 8: 50-5. 100. Wallace Jr RJ, Brown BA, Grifth DE et al. 89. Ha HR, Chen J, Freiburghaus A et al. Reduced serum levels of clarithromycin Metabolism of theophylline by cDNA‐ in patients treated with multidrug expressed human cytochromes P‐450. Br J regimens including rifampin or rifabutin Clin Pharmacol 1995; 39: 321-6. for mycobacterium avium-M. intracellulare 90. Nix DE, DeVito JM, Whitbread MA et al. infection. J Infect Dis 1995; 171: 747-50. Efect of multiple dose oral ciprofoxacin 101. Hafner R, Bethel J, Power M et al. Tolerance on the pharmacokinetics of theophylline and pharmacokinetic interactions of and indocyanine green. J Antimicrob rifabutin and clarithromycin in human Chemother 1987; 19: 263-9. immunodefciency virus-infected volunteers. 91. Wijnands WJ, Vree TB, van Herwaarden CL. Antimicrob Agents Chemother 1998; 42: 631-9. The infuence of quinolone derivatives 102. Colmenero JD, Fernandez-Gallardo LC, on theophylline clearance. Br J Clin Agundez JA et al. Possible implications Pharmacol 1986; 22: 677-83. of doxycycline-rifampin interaction for 92. Loi CM, Parker BM, Cusack BJ et al. Individual treatment of brucellosis. Antimicrob Agents and combined efects of cimetidine Chemother 1994; 38: 2798-802. and ciprofoxacin on theophylline 103. Winter HR, Trapnell CB, Slattery JT et al. metabolism in male nonsmokers. Br J Clin The efect of clarithromycin, fuconazole, and Pharmacol 1993; 36: 195-200. rifabutin on sulfamethoxazole hydroxylamine 93. Robson RA, Begg EJ, Atkinson HC et al. formation in individuals with human Comparative efects of ciprofoxacin and immunodefciency virus infection (AACTG lomefoxacin on the oxidative metabolism of 283). Clin Pharmacol Ther 2004; 76: 313-22. theophylline. Br J Clin Pharmacol 1990; 29: 491-3. 104. Bhatia RS, Uppal R, Malhi R et al. Drug- 94. Prince RA, Casabar E, Adair CG et al. interaction between rifampicin and Efect of quinolone antimicrobials on cotrimoxazole in patients with tuberculosis. theophylline pharmacokinetics. J Clin Hum Exp Toxicol 1991; 10: 419-21. Pharmacol 1989; 29: 650-4. 105. Garske L, Kidd T, Gan R et al. Rifampicin and 95. Raoof S, Wollschlager C, Khan FA. sodium fusidate reduces the frequency Ciprofoxacin increases serum levels of of methicillin-resistant staphylococcus theophylline. Am J Med 1987; 82: 115-8. aureus (MRSA) isolation in adults with cystic 96. Antoniou T, Gomes T, Mamdani MM et fbrosis and chronic MRSA infection. J Hosp al. Ciprofoxacin-induced theophylline Infect 2004; 56: 208-14.

108 DDIs Associated with Antibacterials for COPD

106. Weiner M, Burman W, Luo CC et al. Efects 117. Penttila O, Neuvonen PJ, Aho K et al. of rifampin and multidrug resistance Interaction between doxycycline and some gene polymorphism on concentrations antiepileptic drugs. Br Med J 1974; 2: 470-2. of moxifoxacin. Antimicrob Agents 118. Shahzadi A, Javed I, Aslam B et al. Chemother 2007; 51: 2861-6. Therapeutic efects of ciprofoxacin on 107. Nijland HM, Ruslami R, Suroto AJ et al. the pharmacokinetics of carbamazepine in Rifampicin reduces plasma concentrations healthy adult male volunteers. Pak J Pharm of moxifoxacin in patients with tuberculosis. Sci 2011; 24: 63-8. Clin Infect Dis 2007; 45: 1001-7. 119. Antoniou T, Gomes T, Mamdani MM et 108. Manika K, Chatzika K, Papaioannou M et al. Trimethoprim/sulfamethoxazole- al. Rifampicin-moxifoxacin interaction in induced phenytoin toxicity in 5 tuberculosis treatment: A real-life study. Int the elderly: A population-based study. Br J J Tuberc Lung Dis 2015; 19: 1383-7. Clin Pharmacol 2011; 71: 544-9. 109. Naidoo A, Chirehwa M, McIlleron H et al. Efect 120. Greenberg RG, Melloni C, Wu H et al. of rifampicin and efavirenz on moxifoxacin Therapeutic index estimation of antiepileptic concentrations when co-administered in drugs: A systematic literature review approach. patients with drug-susceptible TB. Antimicrob Clin Neuropharmacol 2016; 39: 232-40. Chemother 2017; 72: 1441-9. 121. Bialer M, Rene’H L, Perucca E. Does 110. Apselof G, Foulds G, LaBoy-Goral L et carbamazepine have a narrow therapeutic al. Comparison of azithromycin and plasma concentration range? Ther Drug clarithromycin in their interactions Monit 1998; 20: 56-9. with rifabutin in healthy volunteers. Eur Respir J 1998; 38: 830-5. 122. Li KY, Li X, Cheng ZN et al. Efect of erythromycin on metabolism of quetiapine 111. Jordan MK, Polis MA, Kelly G et al. Efects of fuconazole and clarithromycin on rifabutin and in chinese sufering from schizophrenia. Eur 25-O-desacetylrifabutin pharmacokinetics. J J Clin Pharmacol 2005; 60: 791-5. Clin Pharmacol 2000; 44: 2170-2. 123. Desta Z, Kerbusch T, Flockhart DA. Efect 112. Nelson MV, Berchou RC, Kareti D et of clarithromycin on the pharmacokinetics al. Pharmacokinetic evaluation of and pharmacodynamics of pimozide in erythromycin and cafeine administered healthy poor and extensive metabolizers with bromocriptine. Clin Pharmacol of cytochrome P450 2D6 (CYP2D6). Clin Ther 1990; 47: 694-7. Pharmacol Ther 1999; 65: 10-20. 113. Nakatsuka A, Nagai M, Yabe H et al. Efect 124. Farkas D, Volak LP, Harmatz JS et al. Short- of clarithromycin on the pharmacokinetics term clarithromycin administration impairs of cabergoline in healthy controls and clearance and enhances pharmacodynamic in patients with parkinson’s disease. J efects of trazodone but not of zolpidem. Pharmacol Sci 2006; 100: 59-64. Clin Pharmacol Ther 2009; 85: 644-50. 114. Perucca E. Clinically relevant drug 125. Kivisto KT, Lamberg TS, Kantola T et al. interactions with antiepileptic drugs. Br J Plasma buspirone concentrations are greatly Clin Pharmacol 2006; 61: 246-55. increased by erythromycin and itraconazole. 115. Neuvonen PJ, Penttila O. Interaction Clin Pharmacol Ther 1997; 62: 348-54. between doxycycline and barbiturates. Br 126. Dean L. Diazepam therapy and CYP2C19 Med J 1974; 1: 535-6. genotype. In: Pratt V, McLeod H, Rubinstein 116. Hansen JM, Kampmann JP, Siersbaek-Nielsen W et al eds. Medical Genetics Summaries. K et al. The efect of diferent sulfonamides US: National Center for biotechnology on phenytoin metabolism in man. Acta Med information, 2012. Available from: https:// Scand Suppl 1979; 624: 106-10. www.ncbi.nlm.nih.gov/books/NBK379740/.

109 Chapter 5

127. Kamali F, Thomas SH, Edwards C. 138. Stass H, Bottcher MF, Ochmann K. The infuence of steady-state ciprofoxacin Evaluation of the infuence of antacids on the pharmacokinetics and and H2 antagonists on the absorption of pharmacodynamics of a single dose of moxifoxacin after oral administration of diazepam in healthy volunteers. Eur J Clin a 400mg dose to healthy volunteers. Clin Pharmacol 1993; 44: 365-7. Pharmacokinet 2001; 40 Suppl 1: 39-48. 128. Fuhr U, Anders EM, Mahr G et al. Inhibitory 139. Frost RW, Lasseter KC, Noe AJ et al. Efects potency of quinolone antibacterial agents of aluminum hydroxide and calcium against cytochrome P450IA2 activity carbonate antacids on the bioavailability in vivo and in vitro. Antimicrob Agents of ciprofoxacin. Antimicrob Agents Chemother 1992; 36: 942-8. Chemother 1992; 36: 830-2. 129. Raaska K, Neuvonen PJ. Ciprofoxacin increases 140. Motoya T, Shimozono T, Yamaguchi T et al. serum clozapine and N-desmethylclozapine: Efects of milk and aluminum hydroxide on A study in patients with schizophrenia. Eur J the absorption of norfoxacin, ciprofoxacin Clin Pharmacol 2000; 56: 585-9. and tosufoxacin in healthy volunteers. J 130. Meyer JM, Proctor G, Cummings MA et al. Appl Ther 1997; 1: 213-7. Ciprofoxacin and clozapine: A potentially 141. Nguyen VX, Nix DE, Gillikin S et al. fatal but underappreciated interaction. Efect of oral antacid administration on Case Rep Psychiatry 2016; 2016: . the pharmacokinetics of intravenous 131. Nix DE, Wilton JH, Ronald B et al. Inhibition doxycycline. Antimicrob Agents of norfoxacin absorption by antacids. Chemother 1989; 33: 434-6. Antimicrob Agents Chemother 1990; 34: 432-5. 142. Nix DE, Watson WA, Handy L et al. 132. Garty M, Hurwitz A. Efect of cimetidine and The efect of sucralfate pretreatment on antacids on gastrointestinal absorption of the pharmacokinetics of ciprofoxacin. tetracycline. Clin Pharmacol Ther 1980; 28: 203-7. Pharmacotherapy 1989; 9: 377-80. 133. Van Slooten AD, Nix DE, Wilton JH et 143. Albert KS, Welch RD, DeSante KA et al. al. Combined use of ciprofoxacin and Decreased tetracycline bioavailability caused sucralfate. DICP 1991; 25: 578-82. by a bismuth subsalicylate antidiarrheal 134. Parpia SH, Nix DE, Hejmanowski LG et al. mixture. J Pharm Sci 1979; 68: 586-8. Sucralfate reduces the gastrointestinal 144. Mapp RK, McCarthy TJ. The efect of zinc absorption of norfoxacin. Antimicrob sulphate and of bicitropeptide on tetracycline Agents Chemother 1989; 33: 99-102. absorption. S Afr Med J 1976; 50: 1829-30. 135. Lehto P, Kivisto KT. Efect of sucralfate on absorption of norfoxacin and ofoxacin. 145. Chopra I, Roberts M. Tetracycline antibiotics: Antimicrob Agents Chemother 1994; 38: 248-51. Mode of action, applications, molecular biology, and epidemiology of bacterial 136. Deppermann KM, Lode H, Hofken G et resistance. Microbiol Mol Biol Rev 2001; 65: al. Infuence of ranitidine, pirenzepine, 232,60 ; second page, table of contents. and aluminum magnesium hydroxide on the bioavailability of various antibiotics, 146. Kamberi M, Nakashima H, Ogawa K et al. including amoxicillin, cephalexin, doxycycline, The efect of staggered dosing of sucralfate and amoxicillin-clavulanic acid. Antimicrob on oral bioavailability of sparfoxacin. Br J Agents Chemother 1989; 33: 1901-7. Clin Pharmacol 2000; 49: 98-103. 137. Nix DE, Watson WA, Lener ME et al. Efects 147. Miura M, Tada H, Yasui-Furukori N et al. Efect of aluminum and magnesium antacids and of clarithromycin on the enantioselective ranitidine on the absorption of ciprofoxacin. disposition of lansoprazole in relation to Clin Pharmacol Ther 1989; 46: 700-5. CYP2C19 genotypes. Chirality 2005; 17: 338-44.

110 DDIs Associated with Antibacterials for COPD

148. Crothers K, Butt AA, Gibert CL et al. Increased 158. Quinney SK, Haehner BD, Rhoades MB COPD among HIV-positive compared to HIV- et al. Interaction between midazolam negative veterans. Chest 2006; 130: 1326-33. and clarithromycin in the elderly. Br J Clin 149. Sahai J, Gallicano K, Oliveras L et al. Pharmacol 2008; 65: 98-109. Cations in the didanosine tablet reduce 159. Zimmermann T, Yeates RA, Laufen H et al. ciprofoxacin bioavailability. Clin Pharmacol Infuence of the antibiotics erythromycin Ther 1993; 53: 292-7. and azithromycin on the pharmacokinetics 150. Knupp CA, Barbhaiya RH. A multiple-dose and pharmacodynamics of midazolam. pharmacokinetic interaction study between Arzneimittelforschung 1996; 46: 213-7. didanosine (videx) and ciprofoxacin 160. Mattila MJ, Vanakoski J, Idanpaan-Heikkila (cipro) in male subjects seropositive for JJ. Azithromycin does not alter the efects HIV but asymptomatic. Biopharm Drug of oral midazolam on human performance. 5 Dispos 1997; 18: 65-77. Eur J Clin Pharmacol 1994; 47: 49-52. 151. Grub S, Bryson H, Goggin T et al. 161. Hagelberg NM, Peltoniemi MA, Saari TI et The interaction of saquinavir (soft gelatin al. Clarithromycin, a potent inhibitor of capsule) with ketoconazole, erythromycin CYP3A, greatly increases exposure to oral and rifampicin: Comparison of the efect S-ketamine. Eur J Pain 2010; 14: 625-9. in healthy volunteers and in HIV-infected 162. Backman JT, Aranko K, Himberg JJ et al. patients. Eur J Clin Pharmacol 2001; 57: 115-21. A pharmacokinetic interaction between 152. Srinivas NR, Knupp CA, Batteiger B et roxithromycin and midazolam. Eur J Clin al. A pharmacokinetic interaction study Pharmacol 1994; 46: 551-5. of didanosine coadministered with 163. Bartkowski RR, Goldberg ME, Larijani GE trimethoprim and/or sulphamethoxazole et al. Inhibition of alfentanil metabolism in HIV seropositive asymptomatic male by erythromycin. Clin Pharmacol patients. Br J Clin Pharmacol 1996; 41: 207-15. Ther 1989; 46: 99-102. 153. Moore KH, Yuen GJ, Raasch RH et al. 164. Jokinen MJ, Olkkola KT, Ahonen J et al. Efect Pharmacokinetics of lamivudine administered of ciprofoxacin on the pharmacokinetics of alone and with trimethoprim-sulfamethoxazole. ropivacaine. Eur J Clin Pharmacol 2003; 58: 653-7. Clin Pharmacol Ther 1996; 59: 550-8. 165. Jokinen MJ, Ahonen J, Neuvonen PJ et al. 154. Shi HY, Yan J, Zhu WH et al. Efects Efect of clarithromycin and itraconazole of erythromycin on voriconazole on the pharmacokinetics of ropivacaine. pharmacokinetics and association Pharmacol Toxicol 2001; 88: 187-91. with CYP2C19 polymorphism. Eur J Clin 166. Jokinen MJ, Ahonen J, Neuvonen PJ et al. Pharmacol 2010; 66: 1131-6. The efect of erythromycin, fuvoxamine, and 155. Sriwiriyajan S, Samaeng M, Ridtitid their combination on the pharmacokinetics of W et al. Pharmacokinetic interactions ropivacaine. Anesth Analg 2000; 91: 1207-12. between ciprofoxacin and itraconazole in 167. Liukas A, Hagelberg NM, Kuusniemi healthy male volunteers. Biopharm Drug K et al. Inhibition of cytochrome Dispos 2011; 32: 168-74. P450 3A by clarithromycin uniformly 156. Olkkola KT, Aranko K, Luurila H et al. afects the pharmacokinetics and A potentially hazardous interaction pharmacodynamics of oxycodone in between erythromycin and midazolam. Clin young and elderly volunteers. J Clin Pharmacol Ther 1993; 53: 298-305. Psychopharmacol 2011; 31: 302-8. 157. Yeates RA, Laufen H, Zimmermann T. 168. Gupta SK, Bakran A, Johnson RW et al. Interaction between midazolam and Cyclosporin-erythromycin interaction clarithromycin: Comparison with azithromycin. in renal transplant patients. Br J Clin Int J Clin Pharmacol Ther 1996; 34: 400-5. Pharmacol 1989; 27: 475-81.

111 Chapter 5

169. Freeman DJ, Martell R, Carruthers 179. Yano R, Tani D, Watanabe K et al. Evaluation SG et al. Cyclosporin-erythromycin of potential interaction between interaction in normal subjects. Br J Clin vinorelbine and clarithromycin. Ann Pharmacol 1987; 23: 776-8. Pharmacother 2009; 43: 453-8. 170. Federico S, Carrano R, Capone D 180. Terkeltaub RA, Furst DE, Digiacinto JL et al. Pharmacokinetic interaction et al. Novel evidence-based colchicine between levofoxacin and ciclosporin dose-reduction algorithm to predict and or tacrolimus in kidney transplant prevent colchicine toxicity in the presence recipients - ciclosporin, tacrolimus and of cytochrome P450 3A4/P-glycoprotein levofoxacin in renal transplantation. Clin inhibitors. Arthritis Rheum 2011; 63: 2226-37. Pharmacokinet 2006; 45: 169-75. 181. Kovarik JM, Beyer D, Bizot MN et al. 171. Wrishko RE, Levine M, Primmett DR et al. Efect of multiple-dose erythromycin on Investigation of a possible interaction between everolimus pharmacokinetics. Eur J Clin ciprofoxacin and cyclosporine in renal transplant Pharmacol 2005; 61: 35-8. patients. Transplantation 1997; 64: 996-9. 182. Kovarik JM, Hsu CH, McMahon L et al. 172. Tan KK, Trull AK, Shawket S. Co-administration Population pharmacokinetics of everolimus of ciprofoxacin and cyclosporin: Lack of in de novo renal transplant patients: evidence for a pharmacokinetic interaction. Impact of ethnicity and comedications. Clin Br J Clin Pharmacol 1989; 28: 185-7. Pharmacol Ther 2001; 70: 247-54. 173. Hedaya MA, El-Affy DR, El-Maghraby GM. 183. Granfors MT, Backman JT, Neuvonen The efect of ciprofoxacin and clarithromycin on M et al. Ciprofoxacin greatly increases sildenafl oral bioavailability in human volunteers. concentrations and hypotensive efect of Biopharm Drug Dispos 2006; 27: 103-10. tizanidine by inhibiting its cytochrome P450 174. Muirhead GJ, Faulkner S, Harness JA et al. 1A2-mediated presystemic metabolism. The efects of steady-state erythromycin Clin Pharmacol Ther 2004; 76: 598-606. and azithromycin on the pharmacokinetics 184. Tan KK, Allwood MC, Shawket S. Efect of of sildenafl in healthy volunteers. Br J Clin ciprofoxacin on the pharmacokinetics Pharmacol 2002; 53 Suppl 1: 37S-43S. of antipyrine in healthy volunteers. J Clin 175. Shinde DD, Kim HS, Choi JS et al. Diferent Pharm Ther 1990; 15: 151-4. efects of clopidogrel and clarithromycin on 185. Ludwig E, Szekely E, Csiba A et al. The efect the enantioselective pharmacokinetics of of ciprofoxacin on antipyrine metabolism. J sibutramine and its active metabolites in healthy Antimicrob Chemother 1988; 22: 61-7. subjects. J Clin Pharmacol 2013; 53: 550-8. 186. Kamali F, Huang ML. Increased systemic 176. Pan W, Bae SK, Shim EJ et al. Efects of clopidogrel availability of loperamide after oral and clarithromycin on the disposition of administration of loperamide and sibutramine and its active metabolites M1 and loperamide oxide with cotrimoxazole. Br J M2 in relation to CYP2B6*6 polymorphism. Clin Pharmacol 1996; 41: 125-8. Xenobiotica 2013; 43: 211-8. 187. Lee BL, Medina I, Benowitz NL et al. Dapsone, 177. Bassi PU, Onyeji CO, Ukponmwan OE. Efects trimethoprim, and sulfamethoxazole of tetracycline on the pharmacokinetics of plasma levels during treatment of halofantrine in healthy volunteers. Br J Clin pneumocystis pneumonia in patients with Pharmacol 2004; 58: 52-5. the acquired immunodefciency syndrome 178. Pohl O, Osterloh I, Gotteland JP. Efects of (AIDS). evidence of drug interactions. Ann erythromycin at steady-state concentrations Intern Med 1989; 110: 606-11. on the pharmacokinetics of ulipristal 188. Piccolo ML, Toossi Z, Goldman M. Efect acetate. J Clin Pharm Ther 2013; 38: 512-7. of coadministration of a nutritional

112 DDIs Associated with Antibacterials for COPD

supplement on ciprofoxacin absorption. 201. Tabibian JH, Gutierrez MA. Doxycycline- Am J Hosp Pharm 1994; 51: 2697-9. induced pseudotumor cerebri. South 189. Penttila O, Hurme H, Neuvonen PJ. Efect Med J 2009; 102: 310-1. of zinc sulphate on the absorption of 202. Berger FA, Monadian N, Groot N et al. tetracycline and doxycycline in man. Eur J QTc prolongation during ciprofoxacin Clin Pharmacol 1975; 9: 131-4. and fuconazole combination therapy: 190. Kays MB, Overholser BR, Mueller BA et al. Prevalence and associated risk factors. Br J Efects of sevelamer hydrochloride and Clin Pharmacol 2018; 84: 369-78. calcium acetate on the oral bioavailability of 203. Baxter K, Preston C. Stockley’s Drug ciprofoxacin. Am J Kidney Dis 2003; 42: 1253-9. Interactions. London: Pharmaceutical 191. Kato R, Ueno K, Imano H et al. Impairment Press, 2010. 5 of ciprofoxacin absorption by calcium 204. Siedlik PH, Olson SC, Yang BB et al. polycarbophil. J Clin Pharmacol 2002; 42: 806-11. Erythromycin coadministration increases plasma atorvastatin concentrations. J Clin 192. Lesko LJ, Ofman E, Brew CT et al. Evaluation Pharmacol 1999; 39: 501-4. of the potential for drug interactions with patiromer in healthy volunteers. J 205. Ericsson CD, Feldman S, Pickering Cardiovasc Pharmacol Ther 2017; 22: 434-46. LK et al. Infuence of subsalicylate bismuth on absorption of doxycycline. 193. How PP, Fischer JH, Arruda JA et al. Efects JAMA 1982; 247: 2266-7. of lanthanum carbonate on the absorption and oral bioavailability of ciprofoxacin. Clin 206. Markert C, Schweizer Y, Hellwig R et al. J Am Soc Nephrol 2007; 2: 1235-40. Clarithromycin substantially increases steady-state bosentan exposure in healthy 194. Landersdorfer CB, Kirkpatrick CM, Kinzig M et al. volunteers. Br J Clin Pharmacol 2014; 77: 141-8. Competitive inhibition of renal tubular secretion of ciprofoxacin and metabolite by probenecid. 207. Markert C, Hellwig R, Burhenne J et Br J Clin Pharmacol 2010; 69: 167-78. al. Interaction of ambrisentan with clarithromycin and its modulation 195. Jaehde U, Sorgel F, Reiter A et al. Efect by polymorphic SLCO1B1. Eur J Clin of probenecid on the distribution and Pharmacol 2013; 69: 1785-93. elimination of ciprofoxacin in humans. Clin Pharmacol Ther 1995; 58: 532-41. 208. Tokairin T, Fukasawa T, Yasui-Furukori N et al. Inhibition of the metabolism of 196. Bolhuis MS, van Altena R, van Soolingen brotizolam by erythromycin in humans: D et al. Clarithromycin increases linezolid In vivo evidence for the involvement of exposure in multidrug-resistant tuberculosis CYP3A4 in brotizolam metabolism. Br J Clin patients. Eur Respir J 2013; 42: 1614-21. Pharmacol 2005; 60: 172-5. 197. Mitchell PB. Therapeutic drug monitoring 209. Phillips JP, Antal EJ, Smith RB. of psychotropic medications. Br J Clin A pharmacokinetic drug interaction Pharmacol 2000; 49: 303-12. between erythromycin and triazolam. J Clin 198. Dickinson BD, Altman RD, Nielsen NH Psychopharmacol 1986; 6: 297-9. et al. Drug interactions between oral 210. Aranko K, Luurila H, Backman JT et al. The efect contraceptives and antibiotics. Obstet of erythromycin on the pharmacokinetics Gynecol 2001; 98: 853-60. and pharmacodynamics of zopiclone. Br J 199. Taylor J, Pemberton MN. Antibiotics and Clin Pharmacol 1994; 38: 363-7. oral contraceptives: New considerations for 211. Vlase L, Popa A, Neag M et al. dental practice. Br Dent J 2012; 212: 481-3. Pharmacokinetic interaction between 200. Albert RK, Schuller JL, Network CCR. Macrolide zolpidem and ciprofoxacin in antibiotics and the risk of cardiac arrhythmias. healthy volunteers. Eur J Drug Metab Am J Respir Crit 2014; 189: 1173-80. Pharmacokinet 2011; 35: 83-7.

113 Chapter 5

212. Polk RE, Healy DP, Sahai J et al. Efect of 217. Neuvonen PJ, Penttila O. Efect of oral ferrous ferrous sulfate and multivitamins with zinc sulphate on the half-life of doxycycline in on absorption of ciprofoxacin in normal man. Eur J Clin Pharmacol 1974; 7: 361-3. volunteers. Respir Care 1989; 33: 1841-4. 218. Neuvonen PJ, Turakka H. Inhibitory efect 213. Kara M, Hasinof BB, McKay DW et al. Clinical of various iron salts on the absorption and chemical interactions between iron of tetracycline in man. Eur J Clin preparations and ciprofoxacin. Br J Clin Pharmacol 1974; 7: 357-60. Pharmacol 1991; 31: 257-61. 219. Venho VM, Salonen RO, Mattila MJ. 214. Lehto P, Kivisto KT, Neuvonen PJ. The efect Modifcation of the pharmacokinetics of of ferrous sulphate on the absorption of doxycycline in man by ferrous sulphate or norfoxacin, ciprofoxacin and ofoxacin. Br charcoal. Eur J Clin Pharmacol 1978; 14: 277-80. J Clin Pharmacol 1994; 37: 82-5. 220. Stass H, Kubitza D. Efects of iron supplements 215. Gothoni G, Neuvonen PJ, Mattila M et al. on the oral bioavailability of moxifoxacin, Iron-tetracycline interaction: Efect of a novel 8-methoxyfuoroquinolone, in humans. time interval between the drugs. Acta Med Clin Pharmacokinet 2001; 40 Suppl 1: 57-62. Scand 1972; 191: 409-11. 221. Cheng SH, White A. Efect of orally 216. Leyden JJ. Absorption of minocycline administered neomycin on the absorption of hydrochloride and tetracycline penicillin V. N Engl J Med 1962; 267: 1296-7. hydrochloride. efect of food, milk, and iron. J Am Acad Dermatol 1985; 12: 308-12.

114

PART II Neuropsychiatric safety of varenicline use for smoking cessation and the application of prescription sequence symmetry analysis in drug safety evaluation CHAPTER 6 Neuropsychiatric safety of varenicline in the general and COPD population with and without psychiatric disorders: a retrospective inception cohort study in a real-world setting

Yuanyuan Wang Jens H. Bos Catharina C.M. Schuiling-Veninga H. Marike Boezen Job F.M. van Boven Bob Wilfert Eelko Hak

Submitted for publication. Chapter 6

ABSTRACT Background Although varenicline is an efective treatment for smoking cessation, evidence on its real-world neuropsychiatric safety is inconsistent, notably for high-risk populations.

Objectives To evaluate the association between varenicline use and major neuropsychiatric adverse events (NPAEs) in the general and COPD population with and without psychiatric disorders compared with nicotine replacement therapy (NRT) in a real-world setting.

Methods A retrospective inception cohort study was conducted among new users of varenicline or NRT using the University of Groningen pharmacy database IADB.nl. The primary outcome was the incidence of any drug-treated NPAEs including depression, anxiety and insomnia within 24 weeks after treatment initiation. Subgroup and sensitivity analyses were also conducted.

Results In the general population without psychiatric disorders, the incidence of total NPAEs in varenicline and NRT groups was 13.7% and 18.3%, respectively (adjusted OR [aOR] 0.78, 95% confdence interval [CI]: 0.67 to 0.90). In the general population with psychiatric disorders, the incidence of total NPAEs was much higher, 81.3% and 84.3% for varenicline and NRT groups, respectively (aOR 0.81, 95% CI: 0.65 to 0.99). In the COPD population, there were no diferences in the incidence of NPAEs between comparison groups in both the psychiatric cohort (aOR 1.01, 95% CI [0.65, 1.58]) and the non-psychiatric cohort (aOR 0.75, 95% CI [0.53, 1.05]). Results from subgroup or sensitivity analyses did not reveal increased risks of varenicline compared to NRT.

Conclusion Varenicline does not increase the risk for NPAEs in both general and COPD populations compared with NRT, irrespective of the presence of psychiatric disorders. Our results provide reassurance for the patients and physicians and may be of help to enhance the use of varenicline for smoking cessation by weighing its risks and benefts.

120 Neuropsychiatric Safety of Varenicline by Cohort

INTRODUCTION

Tobacco smoking is the leading preventable risk factor for a range of physical and mental illnesses1-4 which poses enormous threats to global public health.5 Although average global smoking rates have declined since 1990 through tobacco control policies,6 the actual number of smokers and disease burden related to smoking continues to increase owing to population growth.7 More than 8 million people are killed by tobacco use each year.8 Therefore, more intensifed eforts are needed to fght this deadly epidemic. Smoking cessation strategies as key interventions to prevent smoking- related diseases are therefore urgently needed.9 Varenicline was the frst non-nicotine pharmacotherapy for smoking cessation and has greater efcacy than single bupropion, nicotine replacement therapy (NRT) or placebo.10,11 However, substantial concerns 6 regarding its neuropsychiatric safety (e.g. suicidal thoughts, aggressive behavior) have been raised since its approval in the United States in 2006.12 Therefore, after the frst safety communication and public health advisory in 2008, the FDA released a black box warning on July 1, 2009.13 Of note, these reports could not establish the causality because of a lack of control or comparator. Afterwards, many randomized controlled trials (RCTs) were conducted to evaluate the possible risk of neuropsychiatric adverse events (NPAEs). Notably, pooled evidence of these RCTs did not indicate an association between varenicline and NPAEs.14 Neither did the Evaluating Adverse Events in a Global Smoking Cessation Study (EAGLES) show a signifcant increase in NPAEs with varenicline relative to NRT or placebo.10

Although the FDA warning was removed in 2016, concerns regarding the external validity of the RCT evidence remained. Indeed, due to the strict inclusion and exclusion criteria of RCTs, trial participants were generally healthy. Special risk populations with increased smoking prevalence, such as those with COPD and psychiatric disorders, were usually excluded.15 Importantly, COPD patients are older and sufer from many comorbidities making these patients more susceptible to drug-drug interactions potentially leading to related adverse drug events (ADEs).16 Similarly, it has been reported that individuals with psychiatric disorders are prone to experience relapse of psychiatric symptoms.17,18 Varenicline safety in these specifc populations is not well explored. Only few studies assessed the neuropsychiatric safety of varenicline in patients with COPD or psychiatric disorders,10,19,20 yet results were inconsistent and related evidence from real-world setting is still lacking.

We therefore conducted this cohort study based on real-life data to assess the risk of NPAEs in starters with varenicline versus NRT starters in both the general and the COPD population with and without psychiatric disorders.

121 Chapter 6

METHODS Study design and setting We conducted a retrospective inception cohort study based on the University of Groningen pharmacy dispensing database IADB.nl (http://www.iadb.nl/) which has been widely used for various drug utilization studies.21 IADB.nl contains information of prescribed medications from 70 community pharmacies covering a representative population of approximately 700,000 persons of the Netherlands, regardless of insurance type. It provides both patient information (e.g. date of birth, gender) and complete prescription records including the date of dispensing, quantity dispensed, dose regimen, the number of days the drug will be used, and the related Anatomical Therapeutic Chemical (ATC) codes.

Study population We included adult patients (>18 years) who started with varenicline or NRT. The individuals only prescribed varenicline (ATC code: N07BA03) or NRT (N07BA01) were included as anti-smoking drug users from the general population. Individuals that were prescribed drugs for obstructive airway disease (R03) at least 3 times within 1 year since 1st prescription after the age of 40 years were defned as COPD patients. COPD patients who were prescribed varenicline or NRT were included as COPD anti-smoking drug users.

For both the general and COPD population with anti-smoking drug use, the frst prescription date of varenicline and NRT was set as entry date (index date) of participants for exposure and control groups, respectively. Those who were prescribed other smoking cessation drugs including bupropion (N06AX12), nortriptyline (N06AA10) and cytisine (N07BA04) rather than the studied drugs (varenicline, NRT) within 180 days before or 180 days after index date were excluded. Those who were registered in IADB. nl less than 24 weeks before or after index date were also excluded. For individuals who were prescribed both varenicline and NRT and met criteria of both groups (see Figure 1), we allocated the study subject to the group with the frst index date according to the intention-to-treat principle.

In both the general and the COPD population, we classifed the individuals into a psychiatric cohort and non-psychiatric cohort according to the presence of psychiatric disorders defned by the prescription of two or more drugs from the neurological ATC group, i.e. N02, N03, N04, N05, N06 within 6 months before index date. In all, our study population covered four separate cohorts (1. General population with psychiatric disorders, 2. General population without psychiatric disorders, 3. COPD population with psychiatric disorders and 4. COPD population without psychiatric disorders) in which the association between exposure and outcomes were assessed (Figure 1).

122 Neuropsychiatric Safety of Varenicline by Cohort

Figure 1. Flow chart of population selection. NRT: nicotine replacement therapies; ATC: anatomical therapeutic chemical.

Exposure and outcomes We defned individuals using varenicline as the exposure group and those using NRT as the reference group. The primary outcome was incidence of any drug-treated neuropsychiatric adverse event (NPAE) including depression (drugs with ATC-codes N06A, N06CA), anxiety (N05B) and insomnia (N05C), defned as one or more prescriptions of the specifed drugs within 24 weeks after the frst prescription of varenicline or NRT.

123 Chapter 6

Covariates The following covariates were included as possible confounders: age, gender, social economic status (SES) based on postal codes, previous psychiatric disorders and other comorbidities including heart failure, ischemic heart disease, hypertension, cancer, diabetes, osteoporosis, peptic ulcer and gastroesophageal refux disease (GERD), rheumatic arthritis, thyroid disorder, anemia, glaucoma, gout, and allergic rhinitis. All the comorbidities were defned by at least two prescriptions of related drugs within 6 months prior to index date (see detailed ATC codes in supplementary Table S1), The ATC codes used to defne the comorbidities in this study were consistent with those in previous published papers.22-26

Statistical methods The continuous and categorical variables are presented as means with standard deviations (SD) and numbers with percentages, respectively. The diferences of characteristics between two groups were compared using student’s t-tests and chi-square tests for continuous and categorical variables, respectively. Binary logistic regression modeling was used to obtain the odds ratio (OR) and corresponding 95% confdence interval (CI) after adjustment for the potential confounders. A two-sided p ≤ 0.05 was considered to be statistically signifcant for all tests. All analyses were performed using IBM SPSS statistics 25 (IBM Corporation, Armonk, NY, USA) for Windows.

Subgroup and sensitivity analyses Considering the possible infuence of age and gender on the association between NPAEs and varenicline use, we further analyzed the primary outcome in the four cohorts (general and COPD populations with or without psychiatric disease, separately) by stratifying the results by age groups and gender. To further test the robustness of the results, we performed several sensitivity analyses. First, because the common treatment duration of varenicline is 12 weeks, we explored the outcome occurring within 12 weeks after treatment initiation. Second, to exclude an active psychiatric status, we further selected subjects who were not prescribed any drugs for psychiatric disorders within 30 days before index date. Finally, considering the infuence of policy changes about reimbursement of smoking cessation treatment on the use of anti- smoking drugs,27 we also performed a sensitivity analysis by excluding the patients whose prescription date of varenicline or NRT may be in the period of Dutch smoking policy changes (i.e. from July 1st 2011 to June 30, 2013). RESULTS Baseline characteristics In total, we included 9077 subjects who initiated varenicline or NRT from the general population, of which 2627 had psychiatric disorders. For the COPD population, we

124 Neuropsychiatric Safety of Varenicline by Cohort included 1598 individuals, of which 649 had psychiatric disorders. In both the general and COPD population, individuals treated with varenicline were younger than those treated with NRT (Table 1). Drug use for heart failure and ischemic heart disease was lower in the varenicline-treated than the NRT-treated group. In patients without psychiatric disorders, drug use for other comorbidities (e.g. diabetes and osteoporosis) was also less in individuals treated with varenicline than in those with NRT.

Primary outcome in the general and COPD population In the general population with psychiatric disorders, the incidence of overall NPAEs within 24 weeks was lower in the varenicline group than the NRT group (81.3% vs 84.3%, OR 0.81 [0.66, 0.99], Table 2). After adjusting for potential confounders, the association did not substantially change (adjusted OR (aOR) 0.81, 95% CI [0.65, 0.99]). All the specifc 6 NPAEs were also less in the varenicline group than the NRT group, although the diference between the two comparison groups for depression and insomnia events did not reach statistical signifcance.

In the general population without psychiatric disorders, the incidence rates of NPAEs were lower than among those with psychiatric disorders. The incidence of overall NPAEs within 24-weeks in the varenicline group was lower than those in the NRT group (13.7% and 18.3%, respectively; aOR 0.78, 95% CI [0.67, 0.90]). No diference was observed between the two treatment groups for depression and anxiety, however, less insomnia was seen for varenicline than NRT (aOR was 0.63, 95% CI [0.49, 0.82]).

In the COPD population, we did not see a statistical signifcant diference for incidence of overall NPAEs between the varenicline and NRT groups for both the psychiatric cohort (OR 1.01, 95% CI [0.65, 1.58], Table 3) and the non-psychiatric cohort (OR 0.75, 95% CI [0.53, 1.05]). There were also no diferences for specifc NPAEs between treatment groups in these two cohorts except for anxiety, which was observed signifcantly less in the varenicline group compared with the NRT group (aOR 0.68, 95% CI [0.49, 0.94] for the psychiatric cohort and aOR 0.69, 95% CI [0.50, 0.96] for the non- psychiatric cohort, respectively).

Subgroup analysis In the general population with psychiatric disorders, the risk of overall NPAEs was even lower among varenicline than NRT users for younger patients (aOR 0.54, 95% CI [0.28, 1.04] for age < 40, aOR 0.78, 95% CI [0.63, 0.98] for age 40-65, Table S2) and females (aOR 0.74, 95% CI [0.57, 0.97]). However, in the general population without psychiatric disorders, a lower risk of overall NPAE by varenicline treatment compared with NRT was seen in older patients (aOR 0.52, 95% CI [0.33, 0.82]) and male subjects (aOR 0.78, 95% CI [0.61, 0.99]).

125 Chapter 6 NRT (N= 341) (9.8) 62.7 -89 42 (54.0) 184 (46.0) 157 (36.4) 124 (63.6) 217 (51.3) 175 (48.7) 166 (7.9) 27 (3.5) 12 (42.5) 145 2 (0.6) (12.6) 43 (4.1) 14 (24.0) 82 (6.2) 21 (n = 949) Non-psychiatric cohort Non-psychiatric Varenicline (N=608) (8.9)* 58.9 - 102 40 (51.0) 310 (49.0) 298 (34.7) 211 (65.3) 397 (47.2) 287 (52.8) 321 (3.3)* 20 9 (1.5)* (38.5) 234 0 (0.0) (7.7)* 47 6 (1.0)* (20.2) 123 40(6.6) COPD population (1598) NRT (N=322) 62.4 (10.2) 62.4 92 41 - (39.1) 126 (60.9) 196 (41.9) 135 (58.1) 187 (53.4) 172 (46.6) 150 42 (13.0) 20 (6.2) (51.9) 167 0 (0.0) 40 (12.4) 20 (6.2) (44.1) 142 41 (12.7) (n = 649) Psychiatric cohort Varenicline (N=327) 59.2 (8.6)* 59.2 87 41 - (33.9) 111 (66.1) 216 (41.3) 135 (58.7) 192 (52.0) 170 (48.0) 157 23 (7.0)* 9 (2.8)* (44.6) 146 0 (0.0) 53 (16.2) 17 (5.2) (41.0) 134 45 (13.8) NRT (N= 1970) (12.5) 53.4 90 22 - (55.8) 1100 (44.2) 870 (45.4) 895 (54.6) 1075 (54.8) 1080 (45.2) 890 53 (2.7) 31 (1.6) (27.8) 547 9 (0.5) (7.4) 146 26 (1.3) (12.8) 253 (6.2) 123 (n = 6450) Non-psychiatric cohort Non-psychiatric Varenicline (N=4480) (9.7)* 52.1 102 32 - (53.7) 2406 (46.3) 2074 (33.4)* 1497 (66.6) 2983 (47.3)* 2117 (52.7) 2363 47 (1.0)* 36 (0.8)* (23.2)* 1039 9 (0.2) (5.8)* 260 25 (0.6)* (10.6)* 474 (4.6)* 204 NRT (N=1200) 55.8 (12.3) 55.8 92 25 - (43.6) 523 (56.4) 677 (41.0) 492 (59.0) 708 (54.6) 655 (45.4) 545 82 (6.8) 39 (3.3) (38.8) 465 0 (0.0) (10.8) 129 32 (2.7) (33.9) 407 (14.3) 172 General population (n = 9077) = (n population General (n = 2627) Psychiatric cohort Varenicline (N=1427) 54.5 (9.6)* 54.5 87 32 - (37.9)* 541 (62.1) 886 (36.4)* 519 (63.6) 908 (50.0)* 714 (50.0) 713 41 (2.9)* 25 (1.8)* (35.6) 508 9 (0.6)* (11.4) 162 39 (2.7) (30.6) 436 (11.9) 170 Baseline characteristics of general population and COPD with and without psychiatric disorders treatment by groups. Mean (SD) Mean Age range Men Female 1994-2010 2011-2017 Low High Heart failure heartIschemic disease Hypertension Cancers mellitus Diabetes Osteoporosis Peptic ulcer and GERD arthritis Rheumatic Characteristics Age (years) Age %) (n, Gender Year of index date (n, %) %) (n, status economic Social Comorbidities (n, %) Table 1. Table

126 Neuropsychiatric Safety of Varenicline by Cohort NRT (N= 341) 5 (1.5) (3.5) 12 7 (2.1) (2.9) 10 (4.4) 15 (n = 949) Non-psychiatric cohort Non-psychiatric Varenicline (N=608) 18 (3.0) 18 4 (0.7)* 3 (0.5)* 3 (0.5)* (5.4) 33 COPD population (1598) NRT (N=322) 13 (4.0) 20 (6.2) 4 (1.2) 3 (0.9) 24 (7.5) 6 (n = 649) Psychiatric cohort Varenicline (N=327) 18 (5.5) 11 (3.4) 4 (1.2) 1 (0.3) 28 (8.6) NRT (N= 1970) 49 (2.5) 38 (1.9) 23 (1.2) 24 (1.2) 34 (1.7) (n = 6450) Non-psychiatric cohort Non-psychiatric Varenicline (N=4480) 96 (2.1) 49 (1.1)* 17 (0.4)* 16 (0.4)* (2.3) 104 NRT (N=1200) 40 (3.3) 52 (4.3) 12 (1.0) 9 (0.8) 47 (3.9) General population (n = 9077) = (n population General (n = 2627) Psychiatric cohort Varenicline (N=1427) 55 (3.9) 49 (3.4) 13 (0.9) 4 (0.3) 76 (5.3) (continued) Thyroid disorder Anemia Glaucoma Gout Allergic rhinitis Characteristics *p<0.05; GERD: Gastroesophageal Re f ux Disease. NRT: Nicotine replacement therapy. Table 1. Table

127 Chapter 6 ** Adjusted OR (95% CI) 0.78 [0.67, 0.90] [0.67, 0.78 1.15 [0.84, 1.57] 1.15] [0.71, 0.90 0.82] [0.49, 0.63 Crude OR (95% CI) 0.71 [0.61, 0.82] [0.61, 0.71 1.56] [0.84, 1.15 1.08] [0.67, 0.85 0.78] [0.47, 0.60 (4480 vs 1970) (4480 vs Varenicline vs NRT Non-psychiatric cohort Non-psychiatric Events (n, %) 612 (13.7): 360 (18.3) (13.7): 612 57 (2.9) (3.3): 148 110 (5..6) (4.8): 215 105 (5.3) (3.3): 147 * Adjusted OR (95% CI) 0.81 [0.65, 0.99] [0.65, 0.81 0.90 [0.77, 1.05] 0.84] [0.61, 0.71 1.06] [0.75, 0.89 Crude OR (95% CI) 0.94 [0.80, 1.09] [0.80, 0.94 0.85] [0.62, 0.72 1.01] [0.72, 0.85 0.81 [0.66, 0.99] [0.66, 0.81 (1427 vs 1200) (1427 vs Psychiatric cohort Varenicline vs NRT Events (n, %) 1160 (81.3): 1012 (84.3) 1012 (81.3): 1160 548 (45.7) (44.1): 629 472 (39.3) (32.0): 456 352 (29.3) (26.1): 372 Incidence of neuropsychiatric adverse events (NPAEs) and association with varenicline compared with NRT in general population with and without Depression Anxiety Insomnia NPAEs NPAEs: neuropsychiatric adverse events; NRT: nicotine replacement therapy; OR: odds ratio; heartCI: condiseasef dence andinterval; cancer;. *Adjusted **Adjusted for forage, age,gender, socioeconomicsocioeconomic status,status, heartheart failure,failure,anemia, glaucomaischemicischemic and gout;heart disease, hypertension, diabetes, osteoporosis, peptic ulcer and GERD, Rheumatic arthritis, Overall Overall Table 2. psychiatric disorders with follow weeks. up of 24

128 Neuropsychiatric Safety of Varenicline by Cohort ** Adjusted for age, ** Adjusted OR (95% CI) 0.75 [0.53,1.05] 0.75 3.03] [0.62, 1.37 0.96] [0.50, 0.69 1.16] [0.59, 0.83 Crude OR (95% CI) 0.68 [0.49, 0.95] [0.49, 0.68 2.66] [0.58, 1.24 1.42] [0.50, 0.84 1.19] [0.36, 0.65 (608 vs 341) (608 vs Varenicline vs NRT

Non-psychiatric cohort Non-psychiatric 6 Adjusted for age, heart failure, ischemic heart disease; * Events (n, %) 104 (17.1): 79 (23.2) (17.1): 104 (2.9) 10 22 (3.6): (7.3) 25 38 (6.3): (6.2) 21 25 (4.1): * Adjusted OR (95% CI) 1.01 [0.65, 1.58] [0.65, 1.01 1.24] [0.65, 0.90 0.94] [0.49, 0.68 1.18] [0.60, 0.84 Crude OR (95% CI) 0.86 [0.55, 1.33] [0.55, 0.86 1.34] [0.72, 0.99 0.95] [0.50, 0.69 1.00] [0.52, 0.72 (327 vs 322) (327 vs Psychiatric cohort Varenicline vs NRT Events (n, %) 276 (84.4): 278 (86.3) (84.4): 276 145 (45.0) (44.6): 146 136 (42.2) (33.6): 110 124 (38.5) (31.2): 102 Incidence of neuropsychiatric adverse events (NPAEs) and association with varenicline compared with NRT in COPD population with and without Depression Anxiety Insomnia NPAEs neuropsychiatric adverse events; NRT: nicotine replacement therapy; NPAEs: OR: odds ratio; CI: f dence con interval; heart failure, ischemic heart disease, diabetes, osteoporosis, anemia, glaucoma and gout; glaucoma anemia, osteoporosis, heart heart ischemic diabetes, failure, disease, Overall Table 3. psychiatric disorders with follow weeks. up of 24

129 Chapter 6

In the COPD population, in both the psychiatric and non-psychiatric cohorts, there was no diference for overall NPAEs between the varenicline and NRT groups in each age group and gender group (aOR 0.99, 95% CI [0.53,1.87]; aOR 0.76, 95% CI [0.44, 1.31]; respectively for men and women, Table S2).

Sensitivity analysis In the general population, considering a follow-up of 12 weeks, the risk of overall NPAEs was less for varenicline compared with NRT in both the psychiatric and non-psychiatric cohorts (aOR 0.78, 95% CI: 0.64 to 0.94; aOR 0.74, 95% CI: 0.62 to 0.89; respectively, Table S3). After limiting the study population to those who were not prescribed drugs for any psychiatric disorder or those who were not prescribed drugs for depression, anxiety and insomnia within 1 month before index date, there was no statistical signifcant diference for overall NPAEs, irrespective of presence of psychiatric disorders in the previous year (Table S4). The result was similar when we limited our cohort to individuals whose study period was not in the period of Dutch smoking policy change; the aOR for varenicline compared with NRT was 0.86, 95% CI [0.69, 1.07] in the psychiatric cohort and 0.86, 95% CI [0.71, 1.03] in non-psychiatric cohort.

In the COPD population, there were no statistical signifcant diferences for overall and subgroup NPAEs within follow-up of 12 weeks between the two treatments except for anxiety in the psychiatric cohort (aOR 0.64, 95% CI [0.46, 0.90], Table S5). After limiting the study population to those who were not prescribed drugs for any psychiatric disorder or those who were not prescribed drugs for depression, anxiety and insomnia within 1 month before index date, there was no statistical signifcant diference for overall NPAEs, irrespective of presence of psychiatric disorders (Table S6). Similar results were seen when limiting individuals to those whose study period was not in the period of Dutch smoking policy change (OR 0.97, 95% CI: 0.62 to 1.51; OR 1.00, 95% CI: 0.64 to 1.58; respectively for the psychiatric and non-psychiatric cohorts, Table S6)

DISCUSSION Main fndings and interpretation Within 24 weeks following initiation of varenicline treatment, we found no signifcantly increased risk of NPAEs in both the general and COPD population compared with those using NRT, irrespective of the presence of psychiatric disorders. These fndings are consistent with the results of previous RCTs and large observational studies,10,28,29 Considering the fact that the smoking cessation treatment may last for only 12 weeks without further treatment,30 we also explored the NPAEs in this shorter time period and observed no increased risk in overall and specifc NPAEs for varenicline compared with NRT.

130 Neuropsychiatric Safety of Varenicline by Cohort

In contrast to the concerns about a possible increased risk of NPAEs among varenicline users, we found a 19% and 22% relative decrease in NPAEs in varenicline users of the general population with and without psychiatric disorders, respectively, compared with NRT. Regarding the safety of varenicline for specifc NPAEs, we recorded a 29% reduced risk of anxiety by varenicline (vs NRT) in the psychiatric cohort, and a 37% reduced risk of insomnia in the non-psychiatric cohort. Rates of depression events were comparable between the two groups in both psychiatric and non-psychiatric cohorts among the general population. These results were consistent with the pooled results of 39 RCTs in a meta-analysis,14 which indicated that less anxiety (hazard ratio (HR) 0.75, 95% CI: 0.61, 0.93) was also observed in the varenicline group (vs NRT), and depression episodes were also evenly distributed among two treatments (HR 0.96, 95% CI: 0.75, 1.22). Compared with our study, the diference is that in this review an increased risk of 6 insomnia was observed in the varenicline group (HR 1.56, 95% CI: 1.36, 1.78).14 Of note, the aforementioned review did not explore the risk of varenicline on NPAEs separately in those with and without psychiatric disorders, which may contribute to the observed diferences in this review compared to our results. Of note, our result is consistent with another cohort study based data from the general practice that no increased depression were observed to be associated with varenicline (HR 0.88 [0.77-1.00]). 31

COPD patients are considered a high-risk population with high prevalence of smoking and relatively older age, making these persons more susceptible for possible adverse drug reactions (ADEs).16,32 Of note, in our study we did not observe an increased risk of overall NPAEs among COPD patients using varenicline in both the psychiatric and non- psychiatric cohorts. Of note, regarding the occurrence of specifc NPAEs, less anxiety was seen in the varenicline group than in the NRT group in both cohorts. The safety of varenicline was not fully explored among COPD patients in previous studies. To the best of our knowledge, only two studies (one RCT and one cohort study) were previously conducted.19,33 Similar to our results, both of these two studies did not fnd an increased risk of NPAEs for varenicline. Notably, in the cohort study even a reduced risk of depression was observed in varenicline users compared with users of NRT among COPD patients.19 This may be misled by unmeasured confounders, however, after modelling the efects of possible unmeasured confounders, the author concluded that an increased risk of these adverse events was very unlikely.19

It is notable that there is a large heterogeneity in the defnition of NPAEs across studies. In some studies, the investigators focused on moderate to serious adverse events like depression, suicide or mental disorders that require hospitalization or an emergency department visit.19,29,34,35 While other studies included all adverse symptoms (e.g. angry, nervousness) or adverse events such as trafc ofences.10,30 In this study, we used prescriptions to defne neuropsychiatric outcomes for the most commonly reported NPAEs including depression, anxiety and insomnia during the study period.

131 Chapter 6

Despite diferences in clinical defnitions, the observed 24-week event rates of specifc NPAEs (3%-9%) for depression, anxiety or insomnia in our general population without psychiatric disorders were similar to previous studies.14 However, the rate of NPAEs was substantially higher in participants with psychiatric disorders than those without such illness, which was also consistent with fndings from previous studies.35-37 When we further limited our study population to those who did not experience any psychiatric disorder or not experienced any depression, anxiety and insomnia within 1 month of enrollment, we found that both the overall NPAEs and specifc NPAE reduced substantially.

Although the rates of NPAEs are diferent between psychiatric and non-psychiatric cohorts in this study, the presence of psychiatric disorders did not infuence the risk of NPAEs by varenicline compared with NRT which was also consistent with previous studies.10,38 In a prospective longitudinal study among psychiatric patients, there was no exacerbation of psychiatric symptoms detected except gastrointestinal adverse events.39 However, although it is not within the scope of this study, what need to be mentioned is that an increased rate of outpatient visits for schizophrenia was previously reported to be present only in patients with a pre-existing mental health disorder.34 This may be explained by mediation through individual genetic liability.40

In interpreting the absence of increased risks by varenicline (vs NRT) observed in the general population, we can only speculate that the positive efect from varenicline may result from its efect of consistently reducing withdrawal-related symptoms of negative afect and raised levels of positive afect.41 It has been reported that varenicline yields higher abstinence rates than NRT,10,11 irrespective of smoker characteristics.42 The successful quitting of smoking by varenicline may ofer more benefts to the psychiatric status compared with NRT. Moreover, there is evidence that quitting smoking is associated with recovery in stress, anxiety and depression in smokers.43,44 A signifcant and progressive improvement of anxiety and depression was also reported in an observational study, and the protective efect was observed regardless of the presence of psychiatric pathology.38

Many studies demonstrated gender diferences in varenicline efcacy for smoking cessation,45,46 some studies also found a diference in neuropsychiatric events between genders.47 In this study, we found less risk of NPAEs by varenicline for females in the psychiatric cohort compared with NRT. This may be explained by the better therapeutic response to varenicline in women compared with men. 45 However, in the non-psychiatric cohort, we observed less risk of NPAEs in males by varenicline (vs NRT) which we cannot explain and more research needs to be done on gender disparities. Similarly, there are some indications for age group-dependent NPAEs risk by varenicline (vs NRT) in two cohorts of this study. In the psychiatric cohort, there were

132 Neuropsychiatric Safety of Varenicline by Cohort lower event rates in the varenicline group than the NRT group in younger age groups. Contrary, in the non-psychiatric cohort, lower event rates were seen in older users of varenicline than NRT. The age-specifc efectiveness of varenicline relative to NRT patch or gum were also reported that only younger smokers achieved greater likelihood of abstinence than NRTs.48As such, age disparities also need to be studied more closely.

Strengths and limitations A major strength of this study is that we evaluated the safety of varenicline in both the general and COPD population with and without psychiatric disorders based on large real-life population data making the results representative and more applicable to daily clinical practice. Both short- (12 weeks) and long-term (24 weeks) NPAEs after treatment initiation were explored in this study. Besides the infuence of current or 6 previous psychiatric disorders, we also evaluated the infuence of age and gender on the NPAEs between treatment groups. Additionally, to test the robustness of our study results, several sensitivity analyses were conducted.

There are several potential limitations in our study that need to be discussed. First, as no diagnostic information was available in this study, outcome events and comorbidities were defned by prescriptions of related drugs as proxies. Although we used similar ATC codes as in previous studies and it’s reported that pharmacy data can be used to provide reliable prevalence estimates of several chronic conditions,26 it may still have led to some misclassifcation. Second, the prescription of medication may not always lead to intake of the drugs (non-adherence) and such misclassifcation, if random, may have caused associations to be biased towards the null value. Third, some serious behavioral changes like self-harm or suicide could not be evaluated in this study due to the limitation of the prescription database, although we included the most frequently reported NPAEs that are commonly drug treated. Similarly, some minor symptoms like fatigue could not be evaluated. Fourth, although we tried our best to exclude the infuence of baseline diferences between exposure groups by adjusting for potential confounders, possible unmeasured confounding may still exist. From the baseline characteristics, we could see that the prevalence of comorbidity in varenicline users was lower than in NRT users which may have been the result of the reluctance of prescribing varenicline by clinicians considering its possible risk of related adverse events for high-risk populations. Such potential channeling bias may have caused a relatively better profle of varenicline. However, this kind of bias may not be large, as most associations (OR) between varenicline and specifc NPAEs observed in this study were still below 1 and such bias should be large enough to contradict our conclusion. Additionally, the diference in characteristics could also be attributable to a disparity in the cumulative cigarette exposure between treatment groups,49 however, it’s pity that smoking history was not available in this study.

133 Chapter 6

CONCLUSION

From this population-based real-life inception cohort study, we conclude that varenicline is not associated with a signifcant increased risk of NPAEs in both general and COPD patients with or without psychiatric disorders following its initiation compared with NRT. These results provide further support for the safety of varenicline to quit smoking in both the general and COPD populations.

134 Neuropsychiatric Safety of Varenicline by Cohort

REFERENCES

1. Banks E, Joshy G, Korda RJ, et al. Tobacco disorders (EAGLES): a double-blind, smoking and risk of 36 cardiovascular randomised, placebo-controlled clinical disease subtypes: fatal and non-fatal trial. Lancet. 2016;387(10037):2507-2520. outcomes in a large prospective Australian 11. Ebbert JO, Hughes JR, West RJ, et al. Efect of study. BMC Med. 2019;17(1):128. varenicline on smoking cessation through 2. Gaudet MM, Carter BD, Brinton LA, et al. smoking reduction: a randomized clinical Pooled analysis of active cigarette smoking trial. JAMA. 2015;313(7):687-694. and invasive breast cancer risk in 14 cohort 12. Moore TJ, Furberg CD, Glenmullen J, studies. Int J Epidemiol. 2017;46(3):881-893. Maltsberger JT, Singh S. Suicidal behavior 3. Tjora T, Hetland J, Aaro LE, Wold B, Wiium and depression in smoking cessation N, Overland S. The association between treatments. PLoS One. 2011;6(11):e27016. smoking and depression from adolescence 6 13. US Food and Drug Administration. Public to adulthood. Addiction. 2014;109(6):1022- health advisory: FDA requires new boxed 1030. warnings for the smoking cessation drugs 4. Evins AE, Korhonen T, Kinnunen TH, Kaprio Chantix and Zyban. http://wayback. J. Prospective association between tobacco archive-it.org/7993/20170112005513/ smoking and death by suicide: a competing http://www.fda.gov/Drugs/ risks hazard analysis in a large twin cohort DrugSafety/Postmarket Drug with 35-year follow-up. Psychol Med. SafetyInformationforPatientsandProviders/ 2017;47(12):2143-2154. ucm169988.htm. Published July 1, 2009. 5. Thun MJ, Carter BD, Feskanich D, et Accessed June 7, 2019. . al. 50-Year Trends in Smoking-Related 14. Thomas KH, Martin RM, Knipe DW, Higgins Mortality in the United States. New Engl J JP, Gunnell D. Risk of neuropsychiatric Med. 2013;368(4):351-364. adverse events associated with varenicline: 6. Collaborators GBDT. Smoking prevalence systematic review and meta-analysis. BMJ. and attributable disease burden in 195 2015;350:h1109. countries and territories, 1990-2015: 15. Lawrence D, Mitrou F, Zubrick SR. Smoking a systematic analysis from the Global and mental illness: results from population Burden of Disease Study 2015. Lancet. surveys in Australia and the United States. 2017;389(10082):1885-1906. BMC Public Health. 2009;9:285. 7. The L. Progress towards a tobacco-free 16. Wang Y, Bahar MA, Jansen AME, et al. world. Lancet. 2018;392(10141):1. Improving antibacterial prescribing 8. World Health Organisation. WHO report on safety in the management of COPD the global tobacco epidemic 2019: ofer exacerbations: systematic review of help to quit tobacco use. Geneva: World observational and clinical studies on Health Organisation, 2019: 17-21. potential drug interactions associated with 9. Mehrotra R, Yadav A, Sinha DN, et al. frequently prescribed antibacterials among Smokeless tobacco control in 180 countries COPD patients. J Antimicrob Chemother. across the globe: call to action for full 2019;74(10):2848-2864. implementation of WHO FCTC measures. 17. Garza D, Murphy M, Tseng LJ, Riordan HJ, Lancet Oncol. 2019;20(4):e208-e217. Chatterjee A. A double-blind randomized 10. Anthenelli RM, Benowitz NL, West R, et al. placebo-controlled pilot study of Neuropsychiatric safety and efcacy of neuropsychiatric adverse events in abstinent varenicline, bupropion, and nicotine patch smokers treated with varenicline or placebo. in smokers with and without psychiatric Biol Psychiatry. 2011;69(11):1075-1082.

135 Chapter 6

18. Tonstad S, Davies S, Flammer M. Psychiatric data to estimate the prevalence of chronic Adverse Events in Randomized, Double- conditions? a comparison of multiple data Blind, Placebo-Controlled Clinical Trials of sources. Bmc Public Health. 2011;11. Varenicline: A Pooled Analysis (vol 33, pg 27. van Boven JF, Vemer P. Higher Adherence 298, 2010). Drug Safety. 2010;33(6):522-522. During Reimbursement of Pharmacological 19. Kotz D, Viechtbauer W, Simpson CR, Smoking Cessation Treatments. Nicotine Tob van Schayck OCP, West R, Sheikh A. Res. 2016;18(1):56-63. Cardiovascular and neuropsychiatric risks of 28. Kotz D, Viechtbauer W, Simpson C, van varenicline and bupropion in smokers with Schayck OC, West R, Sheikh A. Cardiovascular chronic obstructive pulmonary disease. and neuropsychiatric risks of varenicline: Thorax. 2017;72(10):905-911. a retrospective cohort study. Lancet Respir 20. Evins AE, Benowitz NL, West R, et al. Med. 2015;3(10):761-768. Neuropsychiatric Safety and Efcacy of 29. Thomas KH, Martin RM, Davies NM, Metcalfe Varenicline, Bupropion, and Nicotine Patch C, Windmeijer F, Gunnell D. Smoking in Smokers With Psychotic, Anxiety, and cessation treatment and risk of depression, Mood Disorders in the EAGLES Trial. J Clin suicide, and self harm in the Clinical Practice Psychopharmacol. 2019;39(2):108-116. Research Datalink: prospective cohort 21. Visser ST, Schuiling-Veninga CC, Bos study. BMJ. 2013;347:f5704. JH, de Jong-van den Berg LT, Postma 30. Molero Y, Lichtenstein P, Zetterqvist J, MJ. The population-based prescription Gumpert CH, Fazel S. Varenicline and risk of database IADB.nl: its development, psychiatric conditions, suicidal behaviour, usefulness in outcomes research and criminal ofending, and transport accidents challenges. Expert Rev Pharmacoecon and ofences: population based cohort Outcomes Res. 2013;13(3):285-292. study. BMJ. 2015;350:h2388. 22. Kronig M, Haverkamp C, Schulte A, et al. 31. Gunnell D, Irvine D, Wise L, Davies C, Martin Diabetes and beta-adrenergic blockage are RM. Varenicline and suicidal behaviour: risk factors for metastatic prostate cancer. a cohort study based on data from World J Surg Oncol. 2017;15(1):50. the General Practice Research Database. 23. van Boven JF, van Raaij JJ, van der Galien R, BMJ. 2009;339:b3805. et al. Impact of multiple-dose versus single- 32. Mangoni AA, Jackson SH. Age-related changes dose inhaler devices on COPD patients’ in pharmacokinetics and pharmacodynamics: persistence with long-acting beta(2)- basic principles and practical applications. Br J agonists: a dispensing database analysis. Clin Pharmacol. 2004;57(1):6-14. NPJ Prim Care Respir Med. 2014;24:14069. 33. Tashkin DP, Rennard S, Hays JT, Ma W, 24. Huber CA, Szucs TD, Rapold R, Reich O. Lawrence D, Lee TC. Efects of varenicline on Identifying patients with chronic conditions smoking cessation in patients with mild to using pharmacy data in Switzerland: moderate COPD: a randomized controlled an updated mapping approach to trial. Chest. 2011;139(3):591-599. the classifcation of medications. Bmc Public 34. Cunningham FE, Hur K, Dong D, et Health. 2013;13. al. A comparison of neuropsychiatric 25. Monte S, Macchia A, Romero M, D’Ettorre adverse events during early treatment A, Giuliani R, Tognoni G. Antidepressants with varenicline or a nicotine patch. and cardiovascular outcomes in patients Addiction. 2016;111(7):1283-1292. without known cardiovascular risk. Eur J Clin 35. Pasternak B, Svanstrom H, Hviid A. Pharmacol. 2009;65(11):1131-1138. Use of varenicline versus bupropion 26. Chini F, Pezzotti P, Orzella L, Borgia P, and risk of psychiatric adverse events. Guasticchi G. Can we use the pharmacy Addiction. 2013;108(7):1336-1343.

136 Neuropsychiatric Safety of Varenicline by Cohort

36. Kanner AM, Wuu J, Faught E, et al. therapy in clients using smoking cessation A past psychiatric history may be a risk services. Addiction. 2018;113(2):313-324. factor for topiramate-related psychiatric 43. Zarghami M, Taghizadeh F, Sharifpour A, and cognitive adverse events. Epilepsy Alipour A. Efcacy of Smoking Cessation on Behav. 2003;4(5):548-552. Stress, Anxiety, and Depression in Smokers 37. Croce Nanni R, Troisi A. Maternal with Chronic Obstructive Pulmonary attachment style and psychiatric history as Disease: A Randomized Controlled Clinical independent predictors of mood symptoms Trial. Addict Health. 2018;10(3):137-147. in the immediate postpartum period. J 44. Taylor G, McNeill A, Girling A, Farley A, Afect Disord. 2017;212:73-77. Lindson-Hawley N, Aveyard P. Change in 38. Pacheco VA, Fernandez APGB, Morales AV, mental health after smoking cessation: Crespo EL, Garcia SM, Rubio TM. Anxiety, systematic review and meta-analysis. Depression and Tobacco Abstinence. BMJ. 2014;348:g1151. Adicciones. 2017;29(4):233-244. 45. McKee SA, Smith PH, Kaufman M, Mazure CM, 6 39. Raich A, Ballbe M, Nieva G, et al. Safety Weinberger AH. Sex Diferences in Varenicline of Varenicline for Smoking Cessation in Efcacy for Smoking Cessation: A Meta-Analysis. Psychiatric and Addicts Patients. Subst Use Nicotine Tob Res. 2016;18(5):1002-1011. Misuse. 2016;51(5):649-657. 46. Smith PH, Kasza KA, Hyland A, et al. 40. Agerbo E, Sullivan PF, Vilhjalmsson BJ, et al. Gender diferences in medication use and Polygenic Risk Score, Parental Socioeconomic cigarette smoking cessation: results from Status, Family History of Psychiatric Disorders, the International Tobacco Control Four Country and the Risk for Schizophrenia: A Danish Survey. Nicotine Tob Res. 2015;17(4):463-472. Population-Based Study and Meta-analysis. 47. The Lancet P. Sex and gender in psychiatry. JAMA Psychiatry. 2015;72(7):635-641. Lancet Psychiatry. 2016;3(11):999. 41. Global Strategy for the Diagnosis, 48. Chang PY, Shiu MN, Yuan YT, Chang HC, Management and Prevention of COPD, Su PY, Lan TH. Comparative Efectiveness Global Initiative for Chronic Obstructive of Varenicline and Nicotine Replacement Pulmonary Disease (GOLD) 2019 Report. Therapy for Smoking Cessation in Older and Available from: http://goldcopd.org. Younger Smokers: A Prospective Cohort in Taiwan. Nicotine Tob Res. 2019;21(2):149-155. 42. Walker N, Gainforth H, Kiparoglou V, Robinson H, van Woerden H, West R. Factors 49. Fidler L, Anand A, Stanbrook M. Cardiovascular moderating the relative efectiveness of and neuropsychiatric risks of varenicline. varenicline and nicotine replacement Lancet Respir Med. 2016;4(3):e9-e10.

137 Chapter 6

SUPPLEMENTARY MATERIALS

Table S1. List of ATC codes used for identifcation of related diseases and outcome events.

Variables ATC codes

COPD R03 General psychiatric disorders N02, N03, N04, N05, N06 Depression N06A, N06CA, Anxiety N05B Insomnia N05C Heart failure C01AA, C03C Ischemic heart disease C01DA Hypertension C02, C03 (except C03C), C07, C08, C09 Cancers L01 Diabetes mellitus A10 Osteoporosis M05B Peptic ulcer and GERD A02B Rheumatic arthritis M01, M02 Thyroid disorders H03 Anemia B03 Glaucoma S01E Gout M04 Allergic rhinitis R01AD

ATC: anatomical therapeutic chemical; GERD: gastroesophageal refux disease;

138 Neuropsychiatric Safety of Varenicline by Cohort aOR (95% CI) --- 1.52] [0.57, 0.93 1.11] [0.24, 0.51 1.87] [0.53, 0.99 1.31] [0.44, 0.76 Non-psychiatric cohort Non-psychiatric Crude OR (95% CI) --- 1.46] [0.57, 0.91 0.51 [0.25, 1.05] 1.43] [0.46, 0.81 1.22] [0.44, 0.73

COPD populations 6 aOR (95% CI) --- 0.89 [0.55, 1.43] 2.41] [0.44, 1.03 2.34] [0.66, 1.24 1.36] [0044, 0.78 Psychiatric cohort Crude OR (95% CI) --- 1.38] [0.56, 0.88 2.35] [0.50, 1.08 0.95 [0.54, 1.69] 1.33] [0.46, 0.78 aOR (95% CI) 1.64] [0.54, 0.93 1.07] [0.72, 0.87 0.83] [0.34, 0.53 0.99] [0.61, 0.78 1.14] [0.72, 0.91 Non-psychiatric cohort Non-psychiatric Crude OR (95% CI) 1.66] [0.61, 1.01 1.04] [0.71, 0.86 0.78] [0.33, 0.50 0.91] [0.56, 0.71 1.11] [0.71, 0.89 General populations General aOR (95% CI) 0.54 [0.28, 1.04] [0.28, 0.54 0.98] [0.63, 0.78 1.81] [0.69, 1.12 1.17] [0.68, 0.89 0.97] [0.57, 0.74 Psychiatric cohort Crude OR (95% CI) 0.58 [0.32, 1.05] [0.32, 0.58 1.01] [0.65, 0.81 1.02 [0.66, 1.59] 0.87 [0.67, 1.13] 0.73 [0.56, 0.95] ** Subgroup analysis: incidence of neuropsychiatric adverse andevents association(NPAEs) with varenicline compared with NRT within follow up of 24 ** Age < 40 40-65 >65 Men Female Subgroups NPAEs: neuropsychiatric adverse events; NRTs: nicotine replacement therapy; aOR: adjustedeconomic status, oddscomorbidities. ratio; *adjusted gender, social economic status and comorbidities; **adjusted age, social Age Gender (n, %) (n, Gender Table S2. weeks strati f ed age by and gender.

139 Chapter 6 Adjusted OR# (95% CI) 0.74 [0.62, 0.89] [0.62, 0.74 1.29 [0.84, 1.97] 1.24] [0.65, 0.90 0.62 [0.44, 0.88] (4480 vs 1970) (4480 vs Crude OR (95% CI) 1.27 [0.83, 1.92] [0.83, 1.27 1.17) (0.62, 0.85 0.82] [0.42, 0.58 0.69 [0.58,0.82] 0.69 Varenicline vs NRTs Non-psychiatric cohort Non-psychiatric Events (n, %) 363 (8.1): 223 (11.3) (8.1): 363 (1.5) 30 86 (1.9): 61 (3.1) (2.7): 119 (3.1) 61 82 (1.8): # Adjusted OR (95% CI) 0.89 [0.76, 1.05] [0.76, 0.89 0.74] [0.53, 0.63 1.05] [0.73, 0.88 0.78 [0.64, 0.94] [0.64, 0.78 Crude OR (95% CI) 0.77 [0.64, 0.93] [0.64, 0.77 1.09] 0.93[0.80, 0.75] [0.54, 0.64 0.99] [0.69, 0.83 (1427 vs 1200) (1427 vs Psychiatric cohort Varenicline vs NRTs Events (n, %) 1062 (74.4): 948 (79.0) 948 (74.4): 1062 491 (40.9) (39.2): 560 423 (35.3) (25.7): 367 311 (25.9) (22.5): 321 Sensitivity analysis: incidence of neuropsychiatric adverse andevents association (NPAEs) with varenicline compared with NRT in general population Depression Anxiety Insomnia NPAEs NPAEs neuropsychiatricNPAE: adverse events; OR: odds ratio; NRTs: nicotine replacement therapy; # adjusted for age, gender, social economic status and related comorbidities. Follow-up of 12 weeks Overall Table S3. with and without psychiatric disorders weeks. within follow up of 12

140 Neuropsychiatric Safety of Varenicline by Cohort * indexdate the aOR (95% CI) 0.93 [0.78, 1.12] 1.62] [0.84, 1.16 1.33] [0.78, 1.02 0.80] [0.46, 0.61 1.11] [0.77, 0.92 1.65] [0.86, 1.19 1.26] [0.75, 0.98 0.81] [0.47, 0.61 1.03] [0.71, 0.86 1.82] [0.90, 1.28 1.19] [0.70, 0.91 0.88] [0.49, 0.66 OR (95% CI) 0.91 [0.76, 1.10] [0.76, 0.91 1.64] [0.85, 1.18 1.27] [0.76, 0.98 0.79] [0.46, 0.60 1.07] [0.75, 0.90 1.66] [0.87, 1.20 0.94 [0.73, 1.21] 0.78] [0.46, 0.60 1.02] [0.70, 0.85 1.84] [0.91, 1.30 1.15] [0.69, 0.89 0.87] [0.48, 0.65 6 Non-psychiatric cohort (varenicline vs NRTs) vs cohort (varenicline Non-psychiatric Events 418 (9.6): 192 (10.4) (9.6): 418 50 (2.7) (3.2): 139 85 (4.6) (4.5): 197 91 (4.9) (3.0): 131 203 (10.6) (9.6): 424 52 (2.7) (3.3): 143 92 (4.8) (4.5): 199 94 (4.9) (3.0): 132 205 (12.9) (11.1): 352 45 (2.8) (3.6): 115 95 (6.0) (5.3): 169 83 (5.2) (3.4): 109 adjusted age, gender, social economic status and comorbidities. We set We social economic status and comorbidities. gender, adjusted age, * * General populations General Netherlands, pharmacologic Smoking Cessation Treatments (pSCTs) were reimbursed in 2011. In 2012 the aOR (95% CI) 0.87 [0.63, 1.21] [0.63, 0.87 1.67] [0.84, 1.18 1.01] [0.46, 0.68 1.43] [0.62, 0.94 1.36] [0.82, 1.05 1.66] [0.93, 1.25 1.05] [0.54, 0.75 1.47] [0.72, 1.03 1.07] [0.69, 0.86 0.98 [0.82, 1.18] 0.87] [0.60, 0.72 1.07] [0.71, 0.87 OR (95% CI) 0.89 [0.65, 1.22] [0.65, 0.89 1.68] [0.88, 1.21 0.10] [0.47, 0.69 1.20] [0.74, 0.94 1.43] [0.88, 1.12 1.35 [1.02, 1.79] 1.10] [0.58, 0.80 0.99 [0.70, 1.40] 1.12] [0.74, 0.91 1.25] [0.88, 1.04 0.90] [0.62, 0.75 1.02] [0.69, 0.83 Psychiatric cohort (varenicline vs NRTs) Events 254 (57.6): 157 (60.4) (57.6): 254 82 (31.5) (35.8): 158 62 (23.8) 78 (17.7): 48 (18.5) 75 (17.0): 201 (46.1) (48.9): 314 104 (23.9) (29.8): 191 81 (18.6) 99 (15.4): 63 (14.4) 92 (14.3): 733 (77.7) (76.0): 797 411 (43.6) (44.7): 468 372 (39.4) (32.7): 343 280 (29.7) (26.0): 273 Sensitivity analysis: incidence of neuropsychiatric adverse events (NPAEs) and association with compared with NRT in generalpopulationin NRTassociationfollowcomparedwithin withwithand Sensitivity neuropsychiatricincidenceanalysis:of adverse(NPAEs) events Depression Anxiety Insomina Depression Anxiety Insomnia Depression Anxiety Insomnia reimbursement was discontinued. As of 2013, pSCTs were again reimbursed, provided they are accompanied by behavioral counseling. Limitations therapy; nicotine replacement aOR: adjusted odds ratio; NRTs: events; adverse neuropsychiatric NPAEs: was within period between July 1st, 2011 and June 30, 2013; Policy change: the In Exclude participants prescribed drugs for depression, anxiety and insomnia within 1 month before index date index anxiety before and insomnia within 1 month depression, participants drugs for prescribed Exclude Overall participants whose study period includes policies changes Exclude Overall Exclude participants prescribed drugs for any NSP within 1 month before index date index before NSP within 1 month any participants drugs for prescribed Exclude Overall Table S4. weeks.up of 24

141 Chapter 6 # Adjusted OR (95% CI) 0.84 [0.54, 1.29] [0.54, 0.84 5.91] [0.65, 1.95 1.52] [0.31, 0.68 2.46] [0.51, 1.12 (608 vs 341) (608 vs Varenicline vs NRTs Crude OR (95% CI) 0.76 [0.51, 1.14] [0.51, 0.76 4.15] [0.52, 1.47 1.44] [0.32, 0.68 2.16] [0.48, 1.02 Non-psychiatric cohort Non-psychiatric Events (n, %) 13 (2.1): 5 (1.5) 5 13 (2.1): (3.8) 13 16 (2.6): (3.2) 11 20 (3.3): 63 (10.4): 45 (13.2) 63 (10.4): # Adjusted OR (95% CI) 0.94 [0.67, 1.32] [0.67, 0.94 0.90] [0.45, 0.64 1.03] [0.51, 0.72 0.89 [0.59, 1.36] [0.59, 0.89 (327 vs 322) (327 vs Crude OR OR Crude (95% CI) 1.07 [0.78, 1.46] [0.78, 1.07 0.66 [0.48, 0.92] 0.94] [0.48, 0.67 0.80 [0.54, 1.19] [0.54, 0.80 Psychiatric cohort Varenicline vs NRTs Events (n, %) 134 (41.0): 127 (39.4) (41.0): 134 123 (38.2) 95 (29.1): 114 (35.4) 88 (26.9): 260 (79.5): 267 (82.9) (79.5): 260 Sensitivity analysis: incidence of neuropsychiatric adverse events (NPAEs) and association with varenicline compared with NRT in COPD populationwith COPD in NRT withcomparedvarenicline withassociation and (NPAEs) neuropsychiatriceventsadverseSensitivity of incidence analysis: Depression Anxiety Insomnia Outcomes Follow-up of 12 weeks Overall Table S5. and without psychiatric disorders weeks. within follow up of 12

142 Neuropsychiatric Safety of Varenicline by Cohort * indexdate indexdate the aOR (95% CI) 0.92 [0.60, 1.43] 2.52] [0.50, 1.12 2.01] [0.60, 1.10 1.37] [0.36, 0.71 1.42] [0.61, 0.93 2.99] [0.60, 1.33 1.79] [0.56, 1.00 1.39] [0.37, 0.72 1.58] [0.64, 1.00 4.42] [0.68, 1.73 2.24] [0.66, 1.21 1.45] [0.34, 0.71 OR (95% CI) 0.95 [0.63, 1.43] [0.63, 0.95 2.29] [0.49, 1.06 1.87] [0.60, 1.06 1.56] [0.43, 0.82 1.38] [0.62, 0.92 2.49] [0.54, 1.16 0.99 [0.57, 1.72] 1.50] [0.43, 0.80 1.43] [0.61, 0.93 3.89] [0.66, 1.60 1.91] [0.61, 1.08 0.53] [0.38, 0.77 6 Non-psychiatric cohort (varenicline vs NRTs) vs cohort (varenicline Non-psychiatric Events 72 (12.1): 40 (12.7) 72 (12.1): (3.2) 10 20 (3.4): (6.0) 19 38 (6.4): (5.1) 16 25 (4.2): 43 (13.0) 73 (12.1): (3.0) 10 21 (3.5): (6.4) 21 38 (6.3): (5.2) 17 25 (4.2): 41 (15.1) 63 (14.2): (2.6) 7 18 (4.1): (7.4) 20 35 (7.9): (5.5) 15 19 (4.3): * COPD populations Netherlands, pharmacologic Smoking Cessation Treatments (pSCTs) were reimbursed in 2011. In 2012 the aOR (95% CI) 0.85 [0.41, 1.77] [0.41, 0.85 1.97] [0.39, 0.88. 1.90] [0.40, 0.87 1.49] [0.24, 0.60 2.26] [0.73, 1.29 2.08] [0.56, 1.08 1.91] [0.49, 0.97 1.82] [0.42, 0.87 1.51] [0.62, 0.97 0.96 [0.66, 1.40] 1.05] [0.49, 0.72 1.10] [0.51, 0.75 aOR: adjusted odds ratio; adjusted age, gender, social economic status and comorbidities. We set We social economic status and comorbidities. aOR: adjusted gender, odds adjusted age, ratio; * OR (95% CI) 0.92 [0.48, 1.79] [0.48, 0.92 2.21] [0.55, 1.10 1.42] [0.34, 0.70 1.30] [0.27, 0.59 1.94] [0.69, 1.16 1.21 [0.66, 2.22] 1.62] [0.46, 0.86 0.76 [0.39, 1.49] 1.34] [0.58, 0.88 1.54] [0.77, 1.09 1.05] [0.52, 0.74 0.96] [0.47, 0.67 Psychiatric cohort (varenicline vs NRTs) Events 54 (60.0): 39 (61.9) 54 (60.0): 19 (30.2) 29 (32.2): 20 (31.7) 22 (24.4): 16 (25.4) 15 (16.7): 48 (47.5) 68 (51.1): 23 (22.8) 35 (26.3): 23 (22.8) 27 (20.3): 20 (19.8) 21 (15.8): 211 (80.2) (78.1): 203 110 (41.8) (43.8): 114 109 (41.4) 89 (34.2): 106 (40.3) 81 (31.2): Sensitivity analysis: incidence of neuropsychiatric adverse and association events (NPAEs) with compared with NRT populationin COPD within follow Depression Anxiety Insomina Depression Anxiety Insomnia Depression Anxiety Insomnia reimbursement was discontinued. As of 2013, pSCTs were again reimbursed, provided they are accompanied by behavioural counselling. Limitations therapy; nicotine replacement NRTs: events; adverse neuropsychiatric NPAEs: was within period between July 1st, 2011 and June 30, 2013; Policy change: the In Exclude participants prescribed drugs for depression, anxiety and insomnia within 1 month before index date index anxiety before and insomnia within 1 month depression, participants drugs for prescribed Exclude Overall participants whose study period includes policies changes Exclude Overall Exclude participants prescribed drugs for any NSP within 1 month before index date index before NSP within 1 month any participants drugs for prescribed Exclude Overall Table S6. weeks.up of 24

143 CHAPTER 7 Risk of neuropsychiatric adverse events associated with varenicline treatment for smoking cessation: a prescription sequence symmetry analysis

Yuanyuan Wang Job F.M. van Boven Jens H. Bos Catharina C.M. Schuiling-Veninga H. Marike Boezen Bob Wilfert Eelko Hak

Submitted for publication (under review). Chapter 7

ABSTRACT Background Varenicline is an efective treatment for smoking cessation. While clinical trials among selected patients did not confrm a causal role, spontaneous reports from daily practice have suggested a possible risk of neuropsychiatric adverse events (NPAEs) by varenicline.

Objectives To investigate the risk of NPAEs associated with varenicline initiation among the general population in a real-world setting.

Methods We conducted a prescription sequence symmetry analysis (PSSA) using data from 2007 to 2018 from the University of Groningen IADB.nl prescription database. We selected incident users of both varenicline and marker drugs for NPAEs, including depression, anxiety and sleep disorder within diferent time-intervals (30, 60, 90, 180, 365 days). Adjusted sequence ratios (aSR) were calculated for each time-interval.

Results Within 365-days’ time-interval 1,066 patients were incident users of both varenicline and NPAE marker drugs. In total, 505 patients were prescribed varenicline before NPAE marker drugs and 561 vice versa (crude sequence ratio (cSR) 0.90, 95% CI: 0.80-1.02). After adjustments for trends in prescriptions, overall a null association was found (aSR 1.00, 95% CI: 0.89-1.13). Regarding specifc NPAEs, no increased risks were found for depression nor anxiety within any time-interval. A small transient increased risk was found for sleep disorders, particularly in earlier time-intervals 3 months and 6 months (aSRs 1.52, [1.10, 2.11] and 1.45, [1.15, 1.83], respectively). The results were robust in stratifed analyses by age and gender, and several sensitivity analyses.

Conclusions Findings from this real-world study were generally consistent with the evidence from clinical trials. Varenicline initiation was not associated with an increased risk of taking anti-depressants nor anti-anxiety drugs, yet a small, but statistically signifcant, transient association with drugs for sleep disorders was noticed, possibly associated with withdrawal symptoms caused by smoking cessation.

146 NAPEs Associated with Varenicline by PSSA

INTRODUCTION

Although in many countries the prevalence of tobacco use has been declining in recent years,1 the tobacco epidemic is still one of the largest global public health threats, related to more than 8 million deaths worldwide each year.2 Smoking-related health problems, including cardiovascular and respiratory disease, are associated with a high burden for both family and society.3 To help halt this burden, several pharmacological and non-pharmacological treatments are available. Varenicline, a frst- line pharmacological smoking cessation treatment (PSCT), which was approved by the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) in 2006. It has a unique mechanism of action compared with other PSCTs by acting as a partial agonist/antagonist with afnity and selectivity for α4β2 nicotinic acetylcholine receptors.4 In several randomized clinical trials (RCTs), varenicline was more efective for smoking cessation than bupropion, nicotine replacement therapy (NRT) or placebo.5-7 7 However, subsequent post-marketing reports related to neuropsychiatric adverse events (NPAEs), such as depression, anxiety, sleep disorder but also suicide, among varenicline users raised concerns about the neuropsychiatric safety of varenicline.8 Based on the post-marketing surveillance reports, the FDA placed a black box warning on varenicline about its risk of NPAEs in 2009.9 The spontaneous case-reports and the FDA warning may have confused both smokers and physicians regarding the causal role of varenicline in inducing NPAEs.10 Such misunderstanding undoubtedly has led to the underutilization of varenicline for smoking cessation.11-13

Although case reports are important signals of drug safety, causality could not be established without using an adequate control group.14 Therefore, to identify the causal association between varenicline and risk of NPAEs, several large cohort studies were conducted after the safety warning15-19 as well as RCTs.7,20,21 Notably, the synthesized evidence did not confrm the earlier suggestions from case reports about severe neuropsychiatric risk from varenicline use.22-24 Therefore, the warning about possible suicidal risk from varenicline was removed by FDA in 2016. In subsequent years, doubts remained regarding the decision to lift the FDA warning. In particular, considering the relatively healthy population and limited power to detect rare events in clinical trials, results from RCTs may not refect the situation in the real-world setting. Moreover, evidence from available observational cohort studies is inconsistent, and these studies were criticized for their potential bias (e.g. selection bias) and confounding (e.g. residual confounding).25

To overcome this bias, the prescription sequence symmetry analysis (PSSA) has been proposed to investigate acute adverse efects of medications, with moderate sensitivity, high specifcity and robust performance.26-28 Unlike spontaneous reporting systems, it uses individual prescription or hospitalization data to assess the association between

147 Chapter 7 a medication and an adverse drug reaction (ADR) by examining the symmetry in the sequence of index medication and marker medications as proxy for ADRs within a specifc time window.27,29 Compared with traditional observational studies (i.e. cohort or case-control), PSSA controls genetic and other time-invariant confounding efectively due to its self-controlled study design. Of note, the relationship between varenicline and NPAEs has not been studied before by using PSSA.

The aim of this study was to examine whether there is an association between varenicline use and onset of NPAEs in the real-world setting using PSSA. METHODS Data source and setting This study was conducted using the widely researched University of Groningen’s pharmacy prescription database IADB.nl, a growing database that comprises a population of approximately 730,000 people from 72 community pharmacies in the northern Netherlands since 1994, regardless of type of health insurance.30 The individuals are representative of the Dutch population with respect to drug use. Detailed prescription information includes date of prescription, name of dispensed drug, dosage, duration and related Anatomical Therapeutic Chemical (ATC) code of prescribed drug, but also year of birth and gender. As Dutch patients usually register at one single community pharmacy, the prescription information from pharmacies are relatively complete. Of note, over-the-counter drugs and prescriptions during hospital stay are not included in IADB.nl. Data after registration of varenicline from 2007 to 2018 were used for this study. The IADB.nl has been used in several previous PSSAs.31,32

Study design The applied PSSA design compares the frequency of initiation of a marker drug (as proxy for an ADR) before and after initiation of an index drug within the same individual. The individuals who were prescribed the index drug (i.e. varenicline) before marker drugs for NPAEs were labeled the “causal group”. Conversely, those who were prescribed varenicline after the marker drugs were labeled the “non-causal group”. The crude sequence ratio (cSR) was defned by the number of patients in the causal group divided by the number of patients in the non-causal group which refects the association between exposure and outcome. If there is no association, the distribution of sequence orders is expected to be symmetrical and the SR is close to 1. Of note, the PSSA design is sensitive to changing trends in drug prescriptions over time, which could be caused by factors like reimbursement policy changes and safety warnings. Therefore, a null-efect sequence ratio (nSR) was used to adjust for the temporal prescription trends of varenicline and marker drugs for ADRs.29 This trend can be used to estimate the probability of varenicline

148 NAPEs Associated with Varenicline by PSSA

to be prescribed frst, in the absence of any causal relationship. The adjusted sequence ratio (aSR) was calculated by dividing the cSR by the nSR.

Run-in period As the goal of PSSA is to evaluate the relation between two incident events, we needed to identify the incident users of both index and marker drugs. As our index drug, i.e. varenicline, was authorized for use by EMA on 26 September 2006, no varenicline was prescribed in the database IADB.nl before 2007. Therefore, in theory, patients with the frst recorded prescription of varenicline during the study period were all incident users. Since the marker drugs have long been used for chronic treatment of NPAEs, most of their current users will be captured at the beginning of the study period. To exclude the current users and identify incident users of NPAEs marker drugs, we used the waiting time distribution to determine the run-in period.33 7 Study population and time interval This study included all individuals who were incident users of both varenicline (ATC: N07BA03) and anyany ofmarker the marker drugs as drugs potential for NPAEs treatment was setfor NPAEsas index including date of depression any NPAEs. The first prescription (N06B, N06CA),of anxiety specific (N05B) marker and drugs sleep for disorder depression, (N05C) anxiety from and1 January sleep disorder2007 to were31 set as index date of December 2018 in IADB.nl. The frst prescription of varenicline was the index date of varenicline, thespecific frst prescription NPAE. Those of any who of were the marker prescribed drugs the for index NPAEs and was marker set as index drug on the same day were date of any NPAEs.excluded. The frst prescription of specifc marker drugs for depression, anxiety and sleep disorder were set as index date of specifc NPAE. Those who were prescribed the index and markerWe defined drug ondifferent the same time day‐intervals were excluded. (365 days, 180 days, 90 days, 60 days and 30 days)

We defned difbetweenerent time-intervals the initiation (365 of the days, varenicline 180 days, and 90 days, NPAEs 60 marker days and drugs 30 days)to explore their association. between the initiationTherefore, of incidentthe varenicline users of and both NPAEs varenicline marker and drugs marker to explore drugs, their irrespective of dose and association. Therefore, incident users of both varenicline and marker drugs, irrespective duration, within pre‐set time intervals of each other were included for the PSSA. of dose and duration, within pre-set time intervals of each other were included for the PSSA. Statistical analysis

Statistical analysisThe crude sequence ratio (cSR) was calculated by dividing the number of individuals in the The crude sequence ratio (cSR) was calculated by dividing the number of individuals in causal group with the number of individuals in the non‐causal group. The adjusted SR (aSR) the causal group with the number of individuals in the non-causal group. The adjusted SR (aSR) was calculatedwas calculated by adjusting by adjusting the time the trend time of trend the crudeof the SR. crude A null-e SR.f Aect null SR‐ effect SR (nSR) is the (nSR) is the expectedexpected SR withoutSR without any causalany causal associations. associations. Therefore, Therefore, the aSR th ise calculated aSR is calculated as the ratio of as the ratio of crude SR to null-efect SR (cSR/nSR). The detailed formula were as follows: crude SR to null‐effect SR (cSR/nSR). The detailed formula were as follows: (1) cSR = number of patients in the causal group/number of patients in the non-causal group. (1) cSR = number of patients in the causal group/number of patients in the non‐causal group.

� �� ∑��∑�� (2) (2) nSR= ��/1‐��, �� = � �� ∑��∑�� ∑��

In the above formula, u is the last day of the research period, m and n are the consecutive 149 days of the survey period, d is the time interval between index and marker drugs. Am is the

number of individuals being prescribed the index drug first on the m day. Bn is the number of individuals being prescribed marker drugs first on the n day.

(3) aSR = cSR/nSR

Confidence intervals (95% CI) of cSR and aSR were calculated by using the binomial distribution as follows:

95% CI =��.��,

where standard error

� � (SE) = � � �� any of the marker drugs for NPAEs was set as index date of any NPAEs. The first prescription any of the marker drugs for NPAEs was set as index date of any NPAEs. The first prescription of specific marker drugs for depression, anxiety and sleep disorder were set as index date of of specific marker drugs for depression, anxiety and sleep disorder were set as index date of specific NPAE. Those who were prescribed the index and marker drug on the same day were specific NPAE. Those who were prescribed the index and marker drug on the same day were excluded. excluded. We defined different time‐intervals (365 days, 180 days, 90 days, 60 days and 30 days) We defined different time‐intervals (365 days, 180 days, 90 days, 60 days and 30 days) between the initiation of the varenicline and NPAEs marker drugs to explore their association. between the initiation of the varenicline and NPAEs marker drugs to explore their association. Therefore, incident users of both varenicline and marker drugs, irrespective of dose and Therefore, incident users of both varenicline and marker drugs, irrespective of dose and duration, within pre‐set time intervals of each other were included for the PSSA. duration, within pre‐set time intervals of each other were included for the PSSA. Statistical analysis Statistical analysis The crude sequence ratio (cSR) was calculated by dividing the number of individuals in the The crude sequence ratio (cSR) was calculated by dividing the number of individuals in the causal group with the number of individuals in the non‐causal group. The adjusted SR (aSR) causal group with the number of individuals in the non‐causal group. The adjusted SR (aSR) was calculated by adjusting the time trend of the crude SR. A null‐effect SR (nSR) is the was calculated by adjusting the time trend of the crude SR. A null‐effect SR (nSR) is the expected SR without any causal associations. Therefore, the aSR is calculated as the ratio of expected SR without any causal associations. Therefore, the aSR is calculated as the ratio of crude SR to null‐effect SR (cSR/nSR). The detailed formula were as follows: crude SR to null‐effect SR (cSR/nSR). The detailed formula were as follows: (1) cSR = number of patients in the causal group/number of patients in the non‐causal group. (1) cSR = number of patients in the causal group/number of patients in the non‐causal group. � �� ∑��∑�� (2) nSR= �∑�/1‐�� , ��∑�� = � �� ∑� �� ∑�� ∑�� Chapter 7 (2) nSR= ��/1‐��, �� = � �� ∑��∑�� ∑�� In the above formula, u is the last day of the research period, m and n are the consecutive In the above formula,In the above u is the formula, last day uof is the the research last day period, of the m research and n are period,the consecutive m and n are the consecutive days of the survey period, d is the time interval between index and marker drugs. Am is the days of the surveydays of period, the survey d is the period, time intervald is the betweentime interval index between and marker index drugs. and Amarkerm drugs. Am is the

is the number of individuals beingnumber prescribed of individuals the index being drug prescribed frst on the the m index day. Bdrugn is first on the m day. Bn is the number of number of individuals being prescribed the index drug first on the m day. Bn is the number of the number of individuals beingindividuals prescribed being marker prescribed drugs frst marker on the ndrugs day. first on the n day. individuals being prescribed marker drugs first on the n day. (3) (3)aSR aSR = cSR/nSR= cSR/nSR (3) aSR = cSR/nSR Confdence intervals (95% CI) ofConfidence cSR and aSR intervals were calculated(95% CI) of by cSR using and aSRthe werebinomial calculated by using the binomial distribution Confidence intervals (95% CI) of cSR and aSR were calculated by using the binomial distribution distribution as follows: as follows: as follows: 95% CI =��.��, 95% CI =��.��, where standard error where standardwhere error standard error

� � �(SE) = � � � (SE) = � ��

All statistical analyses were performed using IBM SPSS statistics 25 (IBM Corporation, Armonk, NY, USA) for Windows. We defned p<0.05 as the level of statistical signifcance. All statistical tests were two tailed.

Subgroup and sensitivity analyses Stratifed analysis was conducted according to diferent gender and age groups. We considered that several policy changes occurred during the study period: (1) reimbursement of PSCTs: PSCTs were reimbursed in 2011, non-reimbursed in 2012 and again reimbursed from 2013 onwards. (2) Black box warning: FDA communicated it in 2009 and removed it in 2016; we performed sensitivity analyses by calculating the aSR for each year of the study period, as well as in several year groups. RESULTS Study population In total, there were 6,440 patients who initiated both varenicline and marker drugs for NPAEs between 2007 to 2018 (Figure 1). Of these, 17 patients were excluded because they were prescribed varenicline and marker drugs at the same day. Of the remaining 4,966 patients, 1,457 were excluded because they were in the run-in period of three months. As shown in the waiting-time distribution (Figure 2), there was a steep decrease at the beginning of the study period after which a more or less stable situation was reached after 3 months, i.e. when prevalent users were basically not in the newly captured population. Finally, for our PSSA there were 1,066 patients who were incident users of both varenicline and marker drugs that were prescribed within a 1-year time

150 Subgroup and Sensitivity analyses

Stratified analysis was conducted according to different gender and age groups.We considered that several policy changes occurred during the study period: (1) reimbursement of PSCTs: PSCTs were reimbursed in 2011, non‐reimbursed in 2012 and again reimbursed from 2013 onwards. (2) Black box warning: FDA communicated it in 2009 and removed it in 2016;

we performed sensitivity analyses by calculating the aSRNAPEs for each Associated year of with the Varenicline study period, by PSSA as well as in several year groups.

FigureFigure 1. 1 Flow. Flow chart chart of ofstudy study population population selection. selection.

period of each other. General characteristics on the study population are shown in Table 1. The median age was 47 years (IQR 18) and the median number of medications that were prescribed before enrollment was 1 (IQR 3). There were no statistically signifcant diferences in age, gender, number of medications prescribed one year before enrollment and distribution of specifc diseases identifed by related medications between causal and non-causal groups.

Main outcome In total, there were 505 patients in the causal group and 561 patients in the non-causal group. Over the full 365 days study follow-up, no statistically signifcant diference was observed between the two groups. The aSR between varenicline and any NPAEs was

151 Chapter 7

Table 1. Characteristics of the study population in the prescription sequence symmetry analysis.

Total population Causal groupa Non-casual groupb Characteristics (N = 1066) (n = 505) (n = 561) P-value*

Gender (n, %) 0.677 Male 529 (49.6) 254 (50.3) 275 (49.0) Female 537 (50.4) 251 (49.7) 286 (51.0) Age groups (n, %) Median (IQR) 47 (18) 48 (19) 47 (17) <= 40 323 (30.3) 152 (30.1) 171 (30.5) 0.299 40 - 50 307 (28.8) 135 (26.7) 172 (30.7) 50 - 60 290 (27.2) 150 (29.7) 140 (25.0) > 60 146 (13.7) 68 (13.5) 78 (13.9) Number of total medications 1 year before enrolment# (n, %) Median (IQR) 1 (3) 1 (3) 1 (3) 0 406 (38.1) 189 (37.4) 217 (38.7) 0.924 1 197 (18.5) 95 (18.8) 102 (18.2) 2-3 249 (23.4) 116 (23.0) 133 (23.7) >= 4 214 (20.1) 105 (20.8) 109 (19.4) Number of patients with specifc medication use (n, %) For obstructive airway 198 (18.6) 100 (19.8) 98 (17.5) 0.328 diseases (ATC codes: R03) For cardiac diseases (C01) 27 (2.5) 16 (3.2) 11 (2.0) 0.210 For diabetes (A10) 52 (4.9) 27 (5.3) 25 (4.5) 0.500

Note: ATC: The Anatomical Therapeutic Chemical (ATC); aCausal group: patients prescribed varenicline frst following by marker drugs for NPAEs. bNon-causal group: patients prescribed marker drugs for NPAEs frst, following by varenicline. *Chi-square test; #Enrollment: the original prescribing date of varenicline of patients in causal group or the original prescribing date of marker drugs for NPAEs of patients in non-causal group.

1.00 [95% CI: 0.89, 1.13, Table 2]. Also, no statistical signifcant association was observed between varenicline and the occurrence of depression (aSR 1.09 [95% CI: 0.94, 1.26]) nor anxiety (aSR 0.98 [0.85, 1.14]). There was, however, a small statistically signifcant increased risk of sleep disorders observed associated with varenicline (aSR 1.25 [95% CI: 1.05, 1.48]).

Impact of time intervals When we considered diferent time-intervals between initiation of varenicline and any NPAE marker drugs, we also did not fnd signifcant associations (Table 2) within 30 days (aSR 0.96, [0.66, 1.39]) and 60 days (aSR 1.10, [0.84, 1.44]. However, there was a boundary signifcant increased risk of NPAEs observed with varenicline within 90 and 180 days.

152 NAPEs Associated with Varenicline by PSSA

period.

study

year

st 1

the 7

with year of study period. of year

st 1 the NPAEs

for

drugs

marker

prescriptions

first

the rst prescriptions of marker drugs for NPAEs with NPAEs of markerdrugs for f rst prescriptions

the distribution

time

Waiting Waiting time distribution of

. 2

Figure 2. Figure

153 Chapter 7

Table 2. Results of the prescription sequence symmetry analysis for the association between varenicline use and marker drugs use for NPAEs by diferent time periods.

Population* Causal Non-causal NPAEs (n) group (n) group (n) cSR (95% CI) nSR aSR (95% CI)

Within 365 days Overall 1066 505 561 0.90 [0.80, 1.02] 0.900 1.00 [0.89, 1.13] depression 727 364 363 1.00 [0.87, 1.16] 0.924 1.09 [0.94, 1.26] anxiety 716 335 381 0.88 [0.76, 1.02] 0.893 0.98 [0.85, 1.14] sleepdisorder 532 286 246 1.16 [0.98, 1.38] 0.931 1.25 [1.05, 1.48] Within 180 days Overall 603 322 281 1.15 [0.98, 1.34] 0.949 1.21 [1.03, 1.42] depression 389 208 181 1.15 [0.94, 1.40] 0.964 1.19 [0.98, 1.45] anxiety 394 206 188 1.10 [0.90, 1.34] 0.943 1.16 [0.95, 1.42] sleepdisorder 295 172 123 1.40 [1.11, 1.76] 0.965 1.45 [1.15, 1.83] Within 90 days Overall 315 173 142 1.22 [0.98,1.52] 0.975 1.25 [1.00, 1.56] depression 209 110 99 1.11 [0.85, 1.46] 0.984 1.13 [0.86, 1.48] anxiety 189 99 90 1.10 [0.83, 1.46] 0.971 1.13 [0.85, 1.51] sleepdisorder 150 90 60 1.50 [1.08, 2.08] 0.984 1.52 [1.10, 2.11] Within 60 days Overall 212 110 102 1.08 [0.82, 1.41] 0.982 1.10 [0.84, 1.44] depression 138 62 76 0.82 [0.58, 1.14] 0.989 0.83 [0.59, 1.15] anxiety 137 72 65 1.11 [0.79, 1.55] 0.979 1.13 [0.81, 1.58] sleepdisorder 101 56 45 1.24 [0.84, 1.84] 0.990 1.26 [0.85, 1.86] Within 30 days Overall 111 54 57 0.95 [0.65,1.37] 0.988 0.96 [0.66, 1.39] depression 74 31 43 0.72 [0.45, 1.14] 0.993 0.73 [0.46, 1.15] anxiety 73 32 41 0.76 [0.47, 1.21] 0.986 0.77 [0.48, 1.22] sleep disorder 50 25 25 1.00 [0.57, 1.74] 0.993 1.01 [0.58, 1.77]

NPAEs: neuropsychiatric adverse events. cSR: crude sequence ratio; aSR: adjusted sequence ratio; nSR: null-efect sequence ratio; CI: confdence interval;

For the specifc NPAEs, similar to the results observed within 365 days, no signifcant associations were observed between varenicline and specifc NPAEs. Again, sleep disorder was the exception with an aSR of 1.52 [95% CI: 1.10, 2.11] and 1.45 [95% CI: 1.15, 1.83] within 90 and 180 days, respectively. Frequency distributions of patients with any NPAE or specifc NPAE (depression, anxiety and sleep disorder) are shown in Figure 3.

Subgroup and sensitivity analyses In stratifed analyses by gender and age groups (Table S1), a signifcant association between varenicline and sleep disorder was only seen in female and older age groups. Of note, a boundary signifcant risk of depression associated with varenicline was also

154 NAPEs Associated with Varenicline by PSSA

after

initiation before

the months

number

disorder

sleep 7 (d)

anxiety;

(c)

depression;

(b)

NPAEs;

all

(a)

with

year.

patients 1

within

distribution

varenicline

Frequency

. Frequency distribution of patients with (a) all NPAEs; (b) depression; (c) anxiety; (d) sleep disorder by number of months before or after 3

initiation

Figure the Figure 3. 1 year. within of varenicline

155 Chapter 7 observed in older age groups (aSR 1.23, 95% CI: 1.01-1.50]). As shown in Figure S1, there were three main fuctuations in the curve for the number of patients newly prescribed varenicline by years of study period. There was a high rise in newly prescribed varenicline in 2011 and a drop in 2012, followed by a sharp increase in 2013. Also, there was a small increase in newly prescribed varenicline from 2015 to 2016. In the results of sensitivity analysis by year groups or each year for the aSR between varenicline and overall NPAEs, we did not observe any statistical signifcant diference for the order of causal and non- causal-groups, except for year 2011 (aSR 1.72, 95% CI: 1.27, 2.32, Table S2). DISCUSSION Main fndings This is the frst study to assess the risk of NPAEs associated with varenicline by applying a PSSA design. Based on the results of this study, we further confrm that no statistical signifcant increased risk of anti-anxiety and anti-depressant drug prescription was associated with varenicline prescription in all diferent time-intervals. However, within 3, 6 and 12 months, there was a small, but statistically signifcant, increased risk of sleep disorder. We did not observe signifcant risk of sleep disorder associated with varenicline within shorter time-intervals (1 or 2 months).

Interpretation The results of this study are consistent with a large meta-analysis of RCTs published in 2015,22 regarding mood change and sleeping disorders. In this systematic review involving 39 studies, there was no evidence of an increased risk of depression and anxiety among varenicline users, compared with placebo users. Oppositely, a higher risk of sleep problems (e.g. insomnia, abnormal dreams) was observed, which was also seen in our study. Of note, diferent from PSSA as a self-controlled design, most previous observational cohort studies used nicotine replacement therapy (NRTs) as the reference group to explore the risk of NPAEs associated with varenicline. In two large cohort studies, there was also no increased depression risk observed among varenicline users.15,16 However, in a third cohort study, varenicline was found to be associated with a small increase in the risk of anxiety and mood conditions, but this was only observed in people with previous psychiatric disorders.17

Sleep disorder is well recognized as a commonly reported ADR associated with varenicline in clinical trials with an incidence ranging from 14.0% to 37.2%.4 It was also the most frequently reported psychiatric event (1.6%) according to a prescription-event monitoring study based in general practice in England.8 However, it is difcult to identify whether sleeping problems are due to side efects of PSCTs or related to withdrawal from nicotine. Indeed, difculty falling asleep and increased number of awakenings are also common symptoms of nicotine withdrawal.34 Although sleep disorder is not

156 NAPEs Associated with Varenicline by PSSA a serious ADR, it may result in poor adherence to varenicline and therefore potentially afect the possibility of quitting successfully. As such, clinicians should pay particular attention to this kind of side efect among varenicline users.

Of note, there was a traditional self-controlled analysis conducted by Gershon et al. in 2018,35 that compared the relative incidence of hospitalizations and emergency department visits during the period of varenicline use compared to the period without varenicline use. The relative incidence (RI) of NPAEs was signifcantly increased (RI 1.06; 95%CI: 1.00-1.13). Considering that the boundary signifcant result was not robust in sensitivity analyses and subgroup analyses stratifed by age groups, the authors did not come to a frm conclusion about the risk of NPAEs associated with varenicline. Compared with this previous study that focused on inpatients, we focused more on NPAEs that happened among outpatients by using a diferent self-controlled study design (i.e. PSSA). Combining the results from our study and the study by Gershon et al, 7 provides complimentary varenicline safety evidence for the general population among diferent real-world settings.

Considering PSSA is sensitive to fuctuation of medication prescriptions, we did sensitivity analyses by calculating the aSR in each year of the study period. We observed fairly consistent results except for the year 2011, which may be biased by PSCT reimbursement policy changes in the Netherlands since PSCTs were reimbursed in 2011 and the reimbursement was temporarily discontinued in 2012.36 Before the cancellation of reimbursement, more varenicline was prescribed in 2011, the sharp increase of varenicline may have led to more people falling in the causal group (varenicline frst and maker drugs second) and less people falling into the non-causal group, which might have led to the aSR above 1. It’s reported that females and older people are more sensitive to NPAEs.37-39 However, we did not fnd a signifcant risk of any NPAEs associated with varenicline in these sub-groups except for sleep disorders, which is consistent with our original outcomes and showed the robustness of our results.

Strengths and limitations Our study has several strengths. The major strength of PSSA design is that it inherently controls for time-invariant, patient-specifc confounders (e.g., sociodemographic characteristics, genetic and lifestyle-related factors) compared with other observational study designs such as cohort or case-control.26 Second, we used a large prescription database with information about medicines dispensed in community pharmacies in the Netherlands, which is representative for a general, unselected, population. Third, due to our design we went beyond the question whether varenicline-related ADRs occurred, and could also provide in-depth analysis of their timing. Our study also has several potential limitations. First, PSSA is sensitive to time-varying confounding like disease severity, which could possibly afect the prescription of sequence of the index

157 Chapter 7 and marker drugs. To minimize this time-varying bias, we limited the time window between index and marker drugs to a maximum of 12 months. Furthermore, due to absence of diagnostic data, marker drugs were used as proxy for NPAEs. Lastly, some severe NPAEs like suicide, neuropsychiatric hospitalizations and emergency department visits could not be evaluated. Despite these limitations are inherent to PSSA methods and common in real-world data sources, this study provides good supplementary evidence for the risk of NPAEs associated with varenicline use in a real-word setting.

Recommendations for future research, policy and clinical practice Our results re-assure the safety of varenicline and may help further minimizing the doubt regarding potential severe adverse drug reactions related to varenicline and support the removal of FDA’s black box warning. Clinicians and users of varenicline should however remain aware of increased occurrence of sleep disorder, especially in the frst three to six months after varenicline initiation. Proper education on expected timing of this event and personalized coping strategies is particularly required. Eventually, this may result in increased smoking cessation treatment uptake, adherence and, ultimately, cessation rates. Future research should focus on whether this sleep disorder is caused by varenicline itself or more related to withdrawal of nicotine. CONCLUSIONS

Our PSSA results suggest that real-world use of varenicline is not associated with any serious risk of NPAEs. However, consistent with previous evidence, there was a transient increased risk of sleep disorder associated with varenicline initiation, particularly in the frst three to six months. Whether sleep disorder was caused by the adverse efects of varenicline or related to withdrawal symptoms needs further study.

158 NAPEs Associated with Varenicline by PSSA

REFERENCES

1. Collaborators GBDT. Smoking prevalence and 11. Gaballa D, Drowos J, Hennekens CH. attributable disease burden in 195 countries Smoking Cessation: The Urgent Need for and territories, 1990-2015: a systematic analysis Increased Utilization of Varenicline. Am J from the Global Burden of Disease Study 2015. Med. 2017;130(4):389-391. Lancet. 2017;389(10082):1885-1906. 12. Sheals K, Tombor I, McNeill A, Shahab L. 2. World Health Organization. Tobacco fact A mixed-method systematic review and sheet. 2019 [cited 2019 Nov. 3]. Available meta-analysis of mental health professionals’ from: http://www.who.int/mediacentre/fact attitudes toward smoking and smoking sheets/fs339/en. cessation among people with mental 3. Goodchild M, Nargis N, Tursan d’Espaignet E. illnesses. Addiction. 2016;111(9):1536-1553. Global economic cost of smoking-attributable 13. Desai RJ, Good MM, San-Juan-Rodriguez A, diseases. Tob Control. 2018;27(1):58-64. et al. Varenicline and Nicotine Replacement 4. Garrison GD, Dugan SE. Varenicline: a frst- Use Associated With US Food and Drug line treatment option for smoking cessation. Administration Drug Safety Communications. 7 Clin Ther. 2009;31(3):463-491. Jama Network Open. 2019;2(9). 5. Cahill K, Stevens S, Perera R, Lancaster T. 14. Shakir SA, Layton D. Causal Pharmacological interventions for smoking association in pharmacovigilance and cessation: an overview and network pharmacoepidemiology: thoughts on meta-analysis. Cochrane Database Syst the application of the Austin Bradford-Hill Rev. 2013(5):CD009329. criteria. Drug Saf. 2002;25(6):467-471. 6. Ebbert JO, Hughes JR, West RJ, et al. Efect of 15. Kress CM, Obi NU, Prochazka AV. In varenicline on smoking cessation through smokers with COPD, neither varenicline smoking reduction: a randomized clinical nor bupropion was linked to increased CV trial. JAMA. 2015;313(7):687-694. or neuropsychiatric risk vs NRT. Ann Intern 7. Anthenelli RM, Benowitz NL, West R, et al. Med. 2017;167(6):JC31. Neuropsychiatric safety and efcacy of 16. Kotz D, Viechtbauer W, Simpson CR, varenicline, bupropion, and nicotine patch van Schayck OCP, West R, Sheikh A. in smokers with and without psychiatric Cardiovascular and neuropsychiatric risks of disorders (EAGLES): a double-blind, varenicline and bupropion in smokers with randomised, placebo-controlled clinical chronic obstructive pulmonary disease. trial. Lancet. 2016;387(10037):2507-2520. Thorax. 2017;72(10):905-911. 8. Kasliwal R, Wilton LV, Shakir SA. Safety and 17. Molero Y, Lichtenstein P, Zetterqvist J, drug utilization profle of varenicline as used Gumpert CH, Fazel S. Varenicline and risk of in general practice in England: interim results psychiatric conditions, suicidal behaviour, from a prescription-event monitoring study. criminal ofending, and transport accidents Drug Saf. 2009;32(6):499-507. and ofences: population based cohort 9. US Food and Drug Administration. Information study. BMJ. 2015;350:h2388. for healthcare professionals: varenicline 18. Kotz D, Viechtbauer W, Simpson C, van (marketed as Chantix) and bupropion Schayck OC, West R, Sheikh A. Cardiovascular (marketed as Zyban, Wellbutrin and generics). and neuropsychiatric risks of varenicline: FDA Drug Safety Newsletter, 2009. a retrospective cohort study. Lancet Respir 10. Davies NM, Thomas KH. The Food and Med. 2015;3(10):761-768. Drug Administration and varenicline: 19. Thomas KH, Martin RM, Davies NM, Metcalfe should risk communication be improved? C, Windmeijer F, Gunnell D. Smoking Addiction. 2017;112(4):555-558. cessation treatment and risk of depression,

159 Chapter 7

suicide, and self harm in the Clinical Practice a prescription sequence symmetry analysis. Research Datalink: prospective cohort Epidemiology. 1996;7(5):478-484. study. Bmj-Brit Med J. 2013;347. 30. Visser ST, Schuiling-Veninga CC, Bos 20. Garza D, Murphy M, Tseng LJ, Riordan HJ, JH, de Jong-van den Berg LT, Postma Chatterjee A. A double-blind randomized MJ. The population-based prescription placebo-controlled pilot study of database IADB.nl: its development, neuropsychiatric adverse events in abstinent usefulness in outcomes research and smokers treated with varenicline or placebo. challenges. Expert Rev Pharmacoecon Biol Psychiatry. 2011;69(11):1075-1082. Outcomes Res. 2013;13(3):285-292. 21. Smith RC, Amiaz R, Si TM, et al. Varenicline 31. Pouwels KB, Visser ST, Bos HJ, Hak E. Angiotensin- Efects on Smoking, Cognition, and converting enzyme inhibitor treatment and Psychiatric Symptoms in Schizophrenia: the development of urinary tract infections: A Double-Blind Randomized Trial. PLoS a prescription sequence symmetry analysis. One. 2016;11(1):e0143490. Drug Saf. 2013;36(11):1079-1086. 22. Thomas KH, Martin RM, Knipe DW, Higgins 32. van Boven JF, de Jong-van den Berg JP, Gunnell D. Risk of neuropsychiatric LT, Vegter S. Inhaled corticosteroids adverse events associated with varenicline: and the occurrence of oral candidiasis: systematic review and meta-analysis. a prescription sequence symmetry analysis. BMJ. 2015;350:h1109. Drug Saf. 2013;36(4):231-236. 23. Lancaster T, Cahill K. ACP Journal Club. 33. Hallas J, Gaist D, Bjerrum L. The waiting Review: Varenicline does not difer from time distribution as a graphical approach to placebo for adverse neuropsychiatric epidemiologic measures of drug utilization. events. Ann Intern Med. 2015;163(2):JC6. Epidemiology. 1997;8(6):666-670. 24. Gibbons RD, Mann JJ. Varenicline, smoking 34. Ashare RL, Lerman C, Tyndale RF, et al. Sleep cessation, and neuropsychiatric adverse events. Disturbance During Smoking Cessation: Am J Psychiatry. 2013;170(12):1460-1467. Withdrawal or Side Efect of Treatment? J 25. Boyko EJ. Observational research-- Smok Cessat. 2017;12(2):63-70. opportunities and limitations. J Diabetes 35. Gershon AS, Campitelli MA, Hawken S, et al. Complications. 2013;27(6):642-648. Cardiovascular and Neuropsychiatric Events 26. Lai EC, Pratt N, Hsieh CY, et al. Sequence after Varenicline Use for Smoking Cessation. symmetry analysis in pharmacovigilance Am J Respir Crit Care Med. 2018;197(7):913-922. and pharmacoepidemiologic studies. Eur J 36. van Boven JF, Vemer P. Higher Adherence Epidemiol. 2017;32(7):567-582. During Reimbursement of Pharmacological 27. Wahab IA, Pratt NL, Wiese MD, Kalisch LM, Smoking Cessation Treatments. Nicotine Tob Roughead EE. The validity of sequence Res. 2016;18(1):56-63. symmetry analysis (SSA) for adverse drug 37. Li SH, Graham BM. Why are women so reaction signal detection. Pharmacoepidemiol vulnerable to anxiety, trauma-related and Drug Saf. 2013;22(5):496-502. stress-related disorders? The potential role of 28. Park SK, Baek YH, Pratt N, Kalisch Ellett L, sex hormones. Lancet Psychiat. 2017;4(1):73-82. Shin JY. The Uncertainty of the Association 38. Riecher-Rossler A. Sex and gender Between Proton Pump Inhibitor Use and diferences in mental disorders. Lancet the Risk of Dementia: Prescription Sequence Psychiat. 2017;4(1):8-9. Symmetry Analysis Using a Korean 39. Vasiliadis HM, Lamoureux-Lamarche C, Gontijo Healthcare Database Between 2002 and Guerra S. Gender and age group diferences 2013. Drug Saf. 2018;41(6):615-624. in suicide risk associated with co-morbid 29. Hallas J. Evidence of depression physical and psychiatric disorders in older provoked by cardiovascular medication: adults. Int Psychogeriatr. 2017;29(2):249-257.

160 NAPEs Associated with Varenicline by PSSA

SUPPLEMENTARY MATERIALS Supplementary materials

FigureFigure S1. S1.Number Number of patients of patientsnewly prescribed newly vare prescribednicline in each varenicline year of the study in each period year of the study period

161 Chapter 7

Table S1. Prescription sequence symmetry results of the association between varenilcine use and marker drugs for NPAEs within a time window of 1 year, stratifed by year.

Number of Sequence Null-efect Time periods patients* ordera Crude SR SR Adjusted SR

Year groups 2007-2010 430 164/266 0.62 [0.51, 0.75] 0.75 0.82 [0.67, 1.00] 2011-2013 345 199/146 1.36 [1.10, 1.69] 1.10 1.23 [1.00, 1.53] 2014-2018 291 142/149 0.95 [0.76, 1.20] 0.96 0.99 [0.79, 1.25] Year 2007 75 19/56 0.34 [0.20, 0.57] 0.62 0.55 [0.33, 0.93] 2008 122 41/81 0.51 [0.35, 0.74] 0.70 0.72 [0.49, 1.05] 2009 103 42/61 0.69 [0.46, 1.02] 0.97 0.71 [0.48, 1.06] 2010 130 62/68 0.91 [0.65, 1.29] 0.98 0.93 [0.66, 1.32] 2011 173 99/74 1.34 [0.99, 1.81] 0.78 1.72 [1.27, 2.32]# 2012 73 44/29 1.52 [0.95, 2.42] 1.06 1.43 [0.89, 2.28] 2013 99 56/43 1.30 [0.88, 1.94] 1.27 1.02 [0.69, 1.52] 2014 67 30/37 0.81 [0.50, 1.31] 0.98 0.83 [0.51, 1.34] 2015 58 25/33 0.76 [0.45, 1.27] 0.92 0.83 [0.49, 1.39] 2016 65 36/29 1.24 [0.76, 2.02] 0.83 1.50 [0.92, 2.44] 2017 69 35/34 1.03 [0.64, 1.65] 1.08 0.95 [0.59, 1.52] 2018 32 16/16 1.00 [0.50, 2.00] 1.11 0.90 [0.45, 1.80]

*Patients with initial prescription of both index drug varenicline and marker drugs for NPAEs. #p<0.05, with statistical signifcance. a the number of patients who initiated marker drugs for NPAEs after index drug varenicline divided by the number of patients who initiated varenicline after marker drugs for NPAEs. NPAEs: neuropsychiatric adverse events; SR: sequence ratio;

162 NAPEs Associated with Varenicline by PSSA

Table S2. Prescription sequence symmetry results of the association between varenicline use and marker drugs for NPAEs within a time window of 365 days, stratifed by gender and age groups.

Number of Sequence Null-efect Variables patients* ordera Crude SR SR Adjusted SR

Gender Male Any NPAEs 529 254/275 0.92 [0.78, 1.10] 0.91 1.02 [0.86,1.21] Depression 307 149/158 0.94 [0.75, 1.18] 0.92 1.02 [0.82, 1.28] Anxiety 333 151/182 0.83 [0.67, 1.03] 0.90 0.92 [0.74, 1.15] Sleep disorder 254 136/118 1.15 [0.90, 1.47] 0.94 1.22 [0.95, 1.56] Female Any NPAEs 537 251/286 0.88 [0.74, 1.04] 0.90 0.98 [0.83, 1.16] Depression 420 215/205 1.05 [0.87, 1.27] 0.93 1.13 [0.94, 1.37] Anxiety 383 184/199 0.92 [0.76, 1.13] 0.89 1.04 [0.85, 1.27] Sleep disorder 278 150/128 1.17 [0.93, 1.48] 0.92 1.27 [1.00, 1.61]# 7 Age groups < = 45 years Any NPAEs 476 214/262 0.82 [0.68, 0.98] 0.91 0.89 [0.75, 1.07] Depression 326 151/175 0.86 [0.69, 1.07] 0.93 0.93 [0.75, 1.15] Anxiety 323 136/187 0.73 [0.58, 0.91] 0.90 0.81 [0.65, 1.01] Sleep disorder 234 116/118 0.98 [0.76, 1.27] 0.95 1.04 [0.80, 1.34] > 45 years Any NPAEs 590 291/299 0.97 [0.83, 1.14] 0.89 1.10 [0.93, 1.29] Depression 401 213/188 1.13 [0.93, 1.38] 0.92 1.23 [1.01, 1.50] Anxiety 393 199/194 1.03 [0.84, 1.25] 0.89 1.16 [0.95, 1.41] Sleep disorder 298 170/128 1.33 [1.06, 1.67] 0.92 1.44 [1.15, 1.82]#

*Patients with initial prescription of both index drug varenicline and maker drugs for NPAEs. a the number of patients who initiated marker drugs for NPAEs after index drug varenicline divided by the number of patients who initiated varenicline after marker drugs for NPAEs. #P<0.05; NPAEs: neuropsychiatric adverse events; SR: sequence ratio;

163 CHAPTER 8 Efect estimate comparison between the prescription sequence symmetry analysis and parallel group study designs: a systematic review

Demy L. Idema Yuanyuan Wang Michael Biehl Peter L. Horvatovich Eelko Hak

Published as: Idema DL, Wang Y, Biehl M, Horvatovich PL, Hak E. Efect estimate comparison between the prescription sequence symmetry analysis (PSSA) and parallel group study designs: A systematic review. PLoS One. 2018 Dec 6;13(12):e0208389. Chapter 8

ABSTRACT

Prescription sequence symmetry analysis (PSSA), a case-only design introduced in 1996, has been increasingly used to identify unintentional drug efects, and has potential applications as a hypothesis-testing and a hypothesis-generating method, due to its easy application and efective control of time-invariant confounders. The aim of this study is to systematically compare efect estimates from the PSSA to efect estimates from conventional observational parallel group study designs, to assess the validity and constraints of the PSSA study design. We reviewed the MEDLINE®, EMBASE®, and Web of Science® databases until February 2016 to identify studies that compared PSSA to a parallel group design. Data from the eligible articles was extracted and analyzed, including making a Bland-Altman plot and calculating the number of discrepancies between the designs. 63 comparisons (from two studies) were included in the review. There was a signifcant correlation (p < 0.001) between the efect estimates of the PSSA and the parallel group designs, but the bias indicated by the Bland-Altman plot (0.20) and the percentage of discrepancies (70-80%) showed that this correlation was not accompanied by a considerable similarity of the efect estimates. Overall, the efect estimates of the parallel group designs were higher than those of the PSSA, not necessarily further away from 1, and the parallel group designs also generated more signifcant signals. However, these results should be approached with caution, as the efect estimates were only retrieved from two separate studies. This review indicates that, even though PSSA has a lot of potential, the efect estimates generated by the PSSA are usually lower than the efect estimates generated by parallel group designs, and PSSA mostly has a lower power than the conventional study designs, but this is based on limited comparisons, and more comparisons are needed to make a proper conclusion.

166 Comparison between PSSA and Parallel Group Study

INTRODUCTION

Conventional observational parallel group studies, such as the cohort study and the case-control study, are still predominantly used to determine causal efects of risk factors and to assess drug safety 1. An important limitation of these designs is that they use an exposed- and a reference group which are (frequently) not readily comparable. This can lead to biased results 2. Case-only designs, such as the case-crossover study design and the self-controlled case-series are alternatives to parallel group designs, and they aim to decrease the possibility of introducing bias 3,4. These designs are particularly useful to control for time-invariant confounders, even when these confounders are generally not recorded in the databases, such as genetic disposition, diet, and over-the- counter drug use 2,5.

In 1996, Hallas introduced another case-only study design: the prescription sequence symmetry analysis (PSSA) 6. A key advantage of PSSA is that it can be used when there is an extensive amount of prescription data available, but no information is given for diagnoses, co-morbidities, and other possible confounders. In this study design, 8 only patients who flled incident prescriptions for both the index drug (the drug under investigation) and the marker drug (the drug prescribed as a proxy/indicator for the outcome of interest, usually an unintentional efect of the index drug) during a predefned risk period are included in the analysis. The crude sequence ratio (SR) is calculated by dividing the number of patients who flled the prescription for the index drug frst and the prescription for the marker drug second, by the number of patients with their prescriptions in the reverse order 7. Since PSSA can be sensitive to temporal prescribing trends, the null-efect sequence ratio is also calculated. This is the expected SR in absence of a causal relationship between the index- and marker drug. A more detailed explanation of the originally proposed method to calculate the null-efect SR is given by Hallas 6. In a study by Tsiropoulos et al. an adjustment to this calculation method is proposed, that takes into account risk periods that are shorter than the total study period 8. By dividing the crude SR by the null-efect SR, the adjusted sequence ratio (ASR) is determined 9. An ASR (including its confdence interval [CI]) above 1 indicates that the index drug may cause the adverse event for which the marker drug is prescribed, while an ASR (including its CI) below 1 suggests a possible protective efect 10. A schematic representation of the cohort-, case-control, and PSSA study design is shown in Fig 1. Variations on the PSSA, such as (event) sequence symmetry analysis ((E) SSA), are also described in literature, and these variations also look at index- and marker events instead of drugs, such as surgeries or behavioral interventions 11.

PSSA has been used less frequently than other, more conventional, pharmacoepidemiologic study designs, and comparisons of PSSA to these designs are lacking 7-10, 12-26. In this systematic review, we aim to compare PSSA to conventional study designs. In a previous study, the correlation between efect estimates from diferent designs has been

167 below 1 suggests a possible protective effect [10]. A schematic representation of the cohort‐, case‐control, and PSSA study design is shown in Fig 1. Variations on the PSSA, such as (event) sequence symmetry analysis ((E)SSA), are also described in literature, and these variations also

Chapterlook at index 8 ‐ and marker events instead of drugs, such as surgeries or behavioral interventions [11].

FigureFig. 1. Schematic1. Schematic representation representation of of the the cohort cohort-,‐, case case-control-,‐control‐, and and PSSAPSSA study designs. Abbreviations: Y = yes, N = no, I = index drug, and M = marker drug. Abbreviations: Y = yes, N = no, I = index drug, and M = marker drug. measured,PSSA has been but used correlation less frequently is not than a measure other, more of agreementconventional, between pharmacoepidemiologic two efect measures 27,28study. Hence, designs, information and comparisons on ofthe PSSA agreement to these designs and discrepancies are lacking [7‐10, between 12‐26]. In designs this rather thansystematic the correlation review, we aim between to compare them PSSA is needed to conventional to assess study the designs. validity In ofa previous the PSSA. study, We aim tothe systematically correlation between review effect articles estimates that from compared different PSSA designs to ha a sconventional been measured, study but design to assess the efect of a medical intervention, to evaluate the diferences between the study correlation is not a measure of agreement between two effect measures [27,28]. Hence, designs and assess possible limitations of the PSSA method. In this review, we will focus information on the agreement and discrepancies between designs rather than the correlation not only on the correlation between the efect estimates in PSSA and parallel designs, but especially on the agreement and discrepancies between them, and the direction of these discrepancies. Our results indicate that even though there is a strong correlation between the efect estimates from the two study designs, there is limited agreement between them and that there are systematic deviations. METHODS Literature search strategies The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist for this study can be found in the S1 File. We searched the MEDLINE® and EMBASE® databases from inception until February 2016 with the search terms “prescription event analysis” OR “symmetry principle” OR “prescription symmetry” OR

168 Comparison between PSSA and Parallel Group Study

“proximate clinical event ratio*” OR “sequence symmetry” OR “sequence-symmetry” OR “symmetry analys*” OR “sequence rat*” OR “prescription sequence”. We also performed a Web of Science® cited reference search (also from inception until February 2016) for the article in which the PSSA method was introduced: “Hallas J”, “1996”, “Epidemiology”. All search results were limited to studies on humans, articles in English, and articles for which the abstract was available.

Selection criteria All identifed articles were exported to RefWorks (ProQuest, Michigan). Title and abstract screening were performed and the full text of the relevant studies was reviewed for eligibility by two independent reviewers (D.L.I. and Y.W.). Disagreement between the reviewers was solved by consensus. Studies were eligible for inclusion in the review if they met the following criteria: the article compares (P)SSA to a conventional study design, the data for both study designs comes from the same data source and the defnitions for the exposure(s) (index drug/event), outcome(s) (marker drug/event) and risk period(s) are equal for both study designs. Articles were excluded if they were 8 systematic reviews, methodological studies, or studies with simulated data.

Data extraction and analysis For all articles that used (P)SSA as a study design identifed by our search, whether the article was eligible for the review or not, the publication year was extracted to examine trends of application of this study design in time. These articles were split up into articles that used PSSA and articles that used another type of SSA, such as event sequence symmetry analysis. If both PSSA and SSA were used, the article was classifed according to the principal study design, as identifed by the article’s author.

From the eligible articles, we extracted the following data: author(s), year of publication, journal name, type of conventional study design and risk measure, exposure (index drug/event), outcome (marker drug/event), comparator used in the conventional study design, risk period(s), the conventional efect estimate and the PSSA efect estimate. If the study investigated multiple drug pairs, and there was not both a conventional efect estimate and a PSSA efect estimate for all of them, only the data for the drug pairs for which both efect estimates were reported was extracted.

As we compared multiple study designs to each other, rather than using diferent quality assessment tools for each study design, a method of quality assessment that we employed was to assess the reporting of potential confounders in the eligible articles. We based our assessment on the checklist by Pouwels et al., derived from the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) statement 29,30.

169 Chapter 8

The efect estimates were exported to SPSS (IBM, New York, version 23), where they were analyzed using several approaches. First, a scatterplot was made of the efect estimates from the conventional study designs against the efect estimates from the PSSA study design to qualitatively examine potential diferences in efect estimates. Second, the Spearman’s correlation coefcients were determined to evaluate the correlation between the efect estimates. Third, we examined whether the diferent study designs found the same signifcant associations. Fourth, because correlation may not be ideal to measure agreement between two types of study design, a Bland-Altman plot was made to assess this 28. Moreover, the discrepancies between the efect estimates were evaluated, as previously done by Ioannidis et al: results were found to be discrepant if there was an absolute diference of 50% or more between the PSSA efect estimate and the parallel group design efect estimate on the natural logarithmic scale 31. RESULTS Article identifcation The search identifed 183 unique articles. Based on the title and abstract screening, 85 potentially relevant articles were selected for full-text screening. After reviewing for eligibility, two articles were included into the review (Fig 2). The frst article compared PSSA to both cohort- and nested case-control studies and the second article compared types of study design, a Bland‐Altman plot was made to assess this [28]. Moreover, the PSSA solely to a cohort study 27,32. The data extracted from both articles is presented in discrepancies between the effect estimates were evaluated, as previously done by Ioannidis et Table 1. al: results were found to be discrepant if there was an absolute difference of 50% or more between the PSSA effect estimate and the parallel group design effect estimate on the natural logarithmic scale [31].

FigureFig. 2. 2. PRISMA PRISMA flow fow diagram diagram of of the the study study selection selection process. process.

170Results

Article identification

The search identified 183 unique articles. Based on the title and abstract screening, 85 potentially relevant articles were selected for full‐text screening. After reviewing for eligibility, two articles were included into the review (Fig 2). The first article compared PSSA to both cohort‐ and nested case‐control studies and the second article compared PSSA solely to a cohort study [27,32]. The data extracted from both articles is presented in Table 1.

There were 50 articles (S2 File) that used (P)SSA to determine the effect of a medical intervention, and Fig 3 shows the number of these articles published per year. Even though the

Comparison between PSSA and Parallel Group Study Risk period Risk 18 months 18 months 12 Outcome Arrhythmia (incident anti-arrhythmic prescription) Nocturnal muscle cramps (incident prescription) quinine

conventional 8 Comparator in the design No incident use of antibacterials Incident use of anticholinergics -agonists 2 Exposure Incident use of antibacterials Incident use of inhaled long-acting β Conventional design ectand e f estimate Cohort (SIR) and nested case-control (AOR) Cohort (HR) Number of Number of drug pairs 62 1 Journal Pharmacoepidemiol. Saf. Drug Arch. Intern. Med.

Main characteristics of articles included in this review. 27 32 Author, Author, year Abbreviations: SIR, standardized incidence ratio; AOR, adjusted odds HR, ratio, hazard ratio. Corrao, 2005 Garrison, 2012 Table 1. Table

171 Chapter 8

There were 50 articles (S2 File) that used (P)SSA to determine the efect of a medical intervention, and Fig 3 shows the number of these articles published per year. Even though the method was rarely used after its introduction in 1996, there is a clear increasing trend in the number of (published) PSSA studies during the last three to four years. For reference, the total number of articles indexed by MEDLINE® per year has also been added to the fgure 33. The increase in the number of articles using the (P)SSA study design is relatively larger than the increase in total number of articles.

Number of PSSA articles (black bar) and number of SSA articles (grey bar) published per year and the total number of articles indexed in MEDLINE (black line) per year after the introduction of the method in 1996 (frst bar in the graph).

Correlation analysis A scatterplot was made of the conventional efect estimates against the PSSA efect estimates (S3 Fig). This scatterplot showed that there was a visible correlation between the efect estimates, but that for most of the investigated drug-pairs, the efect estimate from the conventional study designs was higher than the efect estimate from the PSSA study design, i.e. most of the data points were above the y = x reference line.

Spearman Rank-Order Correlation tests were performed (Table 2), frst for all results taken into consideration, and followed by tests of the subsets of the results PSSA vs. cohort and PSSA vs. nested case-control designs. Their Spearman’s correlation coefcients were 0.621, 0.553, and 0.676, respectively. All results were highly statistically signifcant, with p ≤ 0.001.

FigureFig. 3. 3. Number Number of of (P)SSA (P)SSA articles articles published published per per year. year.

Spearman Rank‐Order Correlation tests were performed (Table 2), first for all results taken into 172 consideration, and followed by tests of the subsets of the results PSSA vs. cohort and PSSA vs. nested case‐control designs. Their Spearman’s correlation coefficients were 0.621, 0.553, and 0.676, respectively. All results were highly statistically significant, with p 0.001.

Table 2. Summary of the Spearman‐Rank Order Correlation analysis.

Dataset N Spearman’s correlation coefficient p‐value

All results 63 0.621 5.52110‐8

PSSA vs. cohort 35 0.553 0.001

PSSA vs. nested case‐control 28 0.676 7.90010‐5

Comparison between PSSA and Parallel Group Study

Table 2. Summary of the Spearman-Rank Order Correlation analysis.

Dataset N Spearman’s correlation coefcient p-value

All results 63 0.621 5.521×10-8 PSSA vs. cohort 35 0.553 0.001 PSSA vs. nested case-control 28 0.676 7.900×10-5

8 estimate

Conventional effect estimate- Conventionaleffect effect PSSA

Mean of PSSA effect estimate and conventional effect estimate

FigureFig 4. Bland 4. Bland-Altman‐Altman plot plot of the of differencethe diference between between the effectthe e festimatesect estimates plotted plotted against against their their mean.mean .The The black black dots dots represent represent the the di fdifferenceerence against against the the mean mean of the of theefect effect estimate estimate pairs; pairs; the black continuous line is the mean diference between the efect estimates (with 95% CI represented by thethe black black dotted continuous lines); line the isgrey the dashed mean differencelines represent between the upper- the effect and lowerestimates limit (withof agreement 95% CI (with 95%represented CI’s represented by the blackby the dotted grey dotted lines); lines). the grey dashed lines represent the upper‐ and lower limit of agreement (with 95% CI’s represented by the grey dotted lines). Agreement and discrepancy analysis Agreement and discrepancy analysis Besides measuring correlation, a second approach used to compare the two methods wasBesides to measuring assess if thecorrelation, diferent a second study approach designs used found to comparethe same the signi two fmethodscant associations was to betweenassess if the the different index- and study marker designs drugs. found In the the same frst significant article, there associations were two betwe signienfcant the index signals‐ found by all three methods, two signals that were only found with the conventional and marker drugs. In the first article, there were two significant signals found by all three study designs and two additional signals that were identifed only with the cohort methods, two signals that were only found with the conventional study designs and two design. In the second article, all designs found a statistically signifcant association for theadditional investigated signals that drug were pair. identified So, combining only with thedata cohort from design. both Inarticles, the second the article,PSSA methodall designs found a statistically significant association for the investigated drug pair. So, combining data from both articles, the PSSA method identified less significant, potentially causal, 173 associations between the index‐ and the marker drug than the two parallel group designs.

Chapter 8 identifed less signifcant, potentially causal, associations between the index- and the marker drug than the two parallel group designs.

The third approach used to compare the efect estimates from the diferent study designs was to make a Bland-Altman plot to assess the degree of agreement between them (Fig 4). This fgure shows that the mean diference between the conventional study design efect estimates and the PSSA study design efect estimates is 0.20 (95% CI [0.15, 0.26]), with the limits of agreement ranging from the lower limit of -0.24 (95% CI [-0.34, -0.14]) to the upper limit of 0.65 (95% CI [0.55, 0.75]). The Bland-Altman analysis shows that there is a degree of bias because the line of equality (the x-axis, y = 0) is not included in the confdence interval of the mean diference.

A fourth approach to compare the efect estimates is by determining discrepancies between them, as described before. The number and percentage of discrepancies are shown in Table 3. Even though the efect estimates were highly correlated, many results were characterized as discrepant when looking at the diference between them. Around 70-80% of all results were found to be discrepant, irrespective of whether it is a comparison to a cohort design or a comparison to a nested case-control design. Looking at these discrepancies, 92% of the cohort efect estimates and 96% of the nested case-control efect estimates were larger than the corresponding PSSA efect estimates. When assessing for the discrepancies whether the conventional efect estimate or the PSSA efect estimate was further away from 1, it was found that 33% of the cohort efect estimates and 36% of the nested case-control efect estimates were further away from 1 than the PSSA efect estimates.

Table 3. Number of discrepancies between PSSA and parallel group designs.

Number (%) of discrepancies for which Number (%) of discrepancies for Number (%) of the conventional efect which the conventional efect Dataset discrepanciesa estimate is largerb estimate is further away from 1c

All results (n=63) 46 (73) 43 (94) 16 (35) PSSA vs. cohort (n=35) 24 (69) 22 (92) 8 (33) PSSA vs. nested case- 22 (79) 21 (96) 8 (36) control (n=28) aDiscrepancies were characterized by the natural logarithm of the PSSA efect estimate being ≥50% larger or smaller than the natural logarithm of the conventional study design efect estimate. bThe fraction of the total number of discrepancies for which the conventional efect estimate is larger than the PSSA efect estimate. cThe fraction of the total number of discrepancies for which the conventional efect estimate is further away from 1 than the PSSA efect estimate.

174 Comparison between PSSA and Parallel Group Study

DISCUSSION

This study aimed to assess and quantify the correlation, agreement, and discrepancies between efect estimates from the PSSA and two conventional pharmacoepidemiologic study designs, the cohort- and nested case-control study design. We found that there was a signifcant correlation between the efect estimates of the PSSA and efect estimates of the conventional studies, but this strong correlation was not accompanied by similar efect estimates; there were systematic diferences between the efect estimates generated by the two types of design. The Bland-Altman analysis showed signifcant bias between the efect estimates, with the efect estimates from the conventional study designs being, on average, 0.20 higher than the efect estimates from the PSSA.

The diference in efect size between the two types of design could originate from the use of a reference group in the conventional study designs, while the PSSA is a case-only design. Time-invariant confounders, whether registered or unregistered, such as advanced age, female gender, and hypochondriasis, may result in bias in parallel group designs (if they are not adjusted for) but not in the case-only PSSA 8. Most comparisons 8 (all apart from one) used in this review were derived from the study by Corrao et al. 27, and in this study, the comparisons from the cohort design were only adjusted for gender, age and month of observation, and the comparisons from the nested case- control design for gender, age, cumulative number of antibiotic prescriptions, and date of cohort entry. The PSSA method may inherently control for more confounders than this, and better confounder control could account for the diference in efect estimates.

However, this may not be the reason for the discrepancies if important assumptions for the validity of PSSA are not met. The assumptions of PSSA, based on the strengths and limitations of the method, are: there is an appropriate and specifc indicator/ proxy for the outcome, the proxy can be prescribed independently of the sequence of the exposure to the index drug and the occurrence of the outcome (e.g. if the outcome is fatal, the proxy could only be prescribed after incident index drug use), the outcome of interest has no efect on subsequent treatment, the efect of the exposure is transient, and the drug-induced symptom is relatively unknown to the prescribing physician 2,10,34-36. Both articles discussed in this review mostly meet all assumptions, indicating that the diferences between the efect estimates do not originate from invalid use of the PSSA design. The only possible problem is that the use of proxies for the outcomes may miss some cases, such as patients with unrecognized symptoms or patients who are hospitalized because of them, or it may include subjects taking the drug who do not have the outcome of interest.

The included articles used relatively long risk periods; 12- and 18 months. These are quite wide time-windows since exposure to the index drug shortly before the onset of the adverse event is more likely to be causal for the investigated exposures, especially for

175 Chapter 8 the antibacterials exposures from the study by Corrao et al. 27. Here, more accurate efect estimates could be obtained if the risk window would be chosen more appropriately, i.e. would be shorter, based on the expected time that is needed for the manifestation of the adverse event. Also, to reduce the possibility of time-variant confounding, the risk period should be relatively small and should generally not exceed a couple of months to a year maximum. Note that for other drugs that are used for more extensive periods of time, and outcomes that may not be reported right away, longer risk periods may be appropriate.

Additionally, the underestimation of the efect size by the PSSA compared to the conventional study designs could be caused by the use of the relatively long risk periods in the PSSA. Using a longer risk period than necessary, especially in the case of the antibacterials exposure, could have diluted the signal by including more nonspecifc sequences in the calculation of the adjusted sequence ratio. If the use of a longer risk period has more efect on the PSSA study design than on the other two designs, this could account (in part) for the lower efect estimates generated in this study design.

More diferences were observed when assessing the number of adverse event signals (i.e. statistically signifcant results). There were cases where adverse event signals were only measured in the cohort- and nested case-control designs, but not in the PSSA. Since PSSA is only performed on subjects that flled incident prescriptions for both the index- and the marker drug, the sample size of the PSSA is smaller, and therefore the power lower, compared to that of the conventional designs. A possible solution for this was used by Pratt et al. in 2013, when they used a distributed network model to investigate the risk of acute hyperglycemia with antipsychotic use 17. Using data from multiple countries increases the size of the population and the power of the analysis, and PSSA can therefore also detect rarer adverse events.

In general, for all arguments made above, it must be noted that our results are only based on two separate studies, and that it is therefore not possible to generalize these results to all studies that use PSSA and/or conventional study designs. A large part of our results (62 of the 63 comparisons included, 98%) are derived from a single study, and therefore these comparisons cannot be considered independent, and the results of our review are greatly dependent on the methodology and results used in that particular study. It may be the case that some part of the methodology of this study caused a systematic underestimation of the efect size of the PSSA study design compared to a conventional study design, and that application of the PSSA methodology is the cause of the discrepancy rather than the PSSA methodology itself. To be able to draw any generalizable conclusion from comparisons between the PSSA study design and conventional study designs, more independent comparisons between the two are greatly needed.

176 Comparison between PSSA and Parallel Group Study

To assess the reporting of potential confounding in the articles used for this review, we used the checklist by Pouwels et al, derived from the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) statement 29,30. In this checklist, there are eight items for the reporting of confounding. The frst article in our review reported seven out of eight items from this checklist, while the second article reported six out of eight 27,32. This is a high reporting quality compared to other articles, as demonstrated in the review by Pouwels et al, which found a median of four out of eight items reported 29.

One important advantage of PSSA is that it eliminates time-invariant confounders, but PSSA may still be sensitive to other types of confounding. Both articles calculated the adjusted sequence ratio to eliminate confounding by temporal prescribing trends and one article also adjusted for age and seasonal trends in drug prescription. Contra- indication was identifed as a potential confounder, but this confounder is hard to control. A large part of the results of this review originates from a study that aimed to determine the efect of antibacterials on arrhythmia. The study split up the antibacterials into many subgroups and this division could have introduced confounding by contra-indication: 8 physicians might only prescribe certain antibacterials to patients who are known to be at risk of arrhythmia. While this bias would not be the cause of the diference between the study designs, because this same division was made for the conventional study designs, it is an important confounder to take into consideration when using PSSA to compare subgroups of drugs prescribed for a similar indication. This is especially true when prescribing physicians are already aware of the potential adverse event, and may let this infuence their prescribing behavior in choosing a specifc drug for high-risk patients.

Also, when subjects consult their physician when they present with symptoms of an adverse event, some physicians may discontinue the index drug instead of prescribing the marker drug. In this case, these subjects would be missed by the PSSA, resulting in an underestimation of the efect estimate. Furthermore, it would also reduce the sample size, which further decreases the power of PSSA. However, most of the efect estimates from the conventional designs were also based on drug dispensing data, so this would not cause discrepancies between the efect estimates of both types of design studied in this review. Besides, it was demonstrated that even fairly well known adverse events are often treated by prescribing additional drugs, rather than discontinuing the drug that might have caused it 21.

Strengths and limitations This is the frst systematic review that assesses the performance of PSSA compared to conventional study designs by reporting on the agreement and the discrepancies between the efect estimates and the direction of the discrepancies. A challenge that we have faced was the limited data available since empirical comparisons between

177 Chapter 8

PSSA and conventional designs are rare, and therefore it is hard to draw any defnite conclusions from our results. We recognize that this could be due to the fact that researchers may perform both types of study and then publish the study with the most relevant results. This would underestimate the discrepancies between the study designs. Another possibility is that the results for both types of study were so similar that researchers only chose to publish the results of one of the two study designs. This would result in an overestimation of the diference between the study designs. Also, because the study designs use diferent efect measures, they may not be readily comparable, causing inconsistencies that are based on the diferent efect measures used rather than the diferent study designs. Therefore, there is a need for a quantitative statistical test that compares efect estimates from diferent risk measures. CONCLUSIONS

PSSA, due to its simple and quick implementation and its ability to eliminate time- invariant confounding has a lot of potential in assessing drug safety. However, our results indicate that PSSA lacks power in many situations, and its results often signifcantly deviate from efect estimates generated by conventional parallel group study designs. PSSA might, therefore, be more suitable as a hypothesis-generating design, that should be followed by a more conventional parallel group design for hypothesis-testing and confrmation. Our results should be approached with some caution, though, as they are only based on two independent studies. To get a better understanding of the practical diferences between the two types of designs, and to be able to make any generalizations, more comparisons between PSSA and parallel group designs are required. Future studies should also compare PSSA to randomized controlled trials, to assess how PSSA performs against the study design considered to be the golden standard. SUPPLEMENTARY MATERIALS

File S1-S2 are available as Supplementary data at PLOS ONE Online (https://doi. org/10.1371/journal.pone.0208389)

178 Comparison between PSSA and Parallel Group Study

REFERENCES

1. Consiglio GP, Burden AM, Maclure M, analysis of one million prescriptions. Aliment McCarthy L, Cadarette SM. Case-crossover Pharmacol Ther. 2000;14: 1479-1484. study design in pharmacoepidemiology: 11. Cole JA, Farraye FA, Cabral HJ, Zhang Y, systematic review and recommendations. Rothman KJ. Irritable bowel syndrome Pharmacoepidemiol Drug Saf. 2013;22: 1146- and hysterectomy: a sequence symmetry 1153. doi: 10.1002/pds.3508 analysis. Epidemiology. 2007;18: 837-838. 2. Hallas J, Pottegård A. Use of self-controlled doi: 10.1097/EDE.0b013e3181571f1b designs in pharmacoepidemiology. J Intern Med. 12. Caughey GE, Roughead EE, Pratt N, Shakib 2014;275: 581-589. doi: 10.1111/joim.12186 S, Vitry AI, Gilbert AL. Increased risk of hip 3. Maclure M. The case-crossover design: fracture in the elderly associated with a method for studying transient efects prochlorperazine: is a prescribing cascade on the risk of acute events. Am J contributing? Pharmacoepidemiol Drug Saf. Epidemiol. 1991;133: 144-153. 2010;19: 977-982. doi: 10.1002/pds.2009 4. Farrington P, Pugh S, Colville A, Flower 13. Fujimoto M, Higuchi T, Hosomi K, Takada M. A, Nash J, Morgan-Capner P, et al. A new Association of statin use with storage lower method for active surveillance of adverse urinary tract symptoms: data mining of events from diphtheria/tetanus/pertussis claims database. J Pharmacovigilance. 2014;2: 8 and measles/mumps/rubella vaccines. 147. doi: 10.4172/2329-6887.1000147 Lancet. 1995;345: 567-569. 14. Hallas J, Bytzer P. Screening for drug-related 5. Newsom JT, Jones RN, Hofer SM. dyspepsia: an analysis of prescription symmetry. Longitudinal data analysis: a practical guide Eur J Gastroenterol Hepatol. 1998;10: 27-32. for researchers in aging, health, and social 15. Lai EC, Yang YH, Lin SJ, Hsieh CY. Use of sciences. New York: Routledge; 2012. antiepileptic drugs and risk of hypothyroidism. 6. Hallas J. Evidence of depression Pharmacoepidemiol Drug Saf. 2013;22: 1071- provoked by cardiovascular medication: 1079. doi: 10.1002/pds.3498 a prescription sequence symmetry analysis. 16. Pouwels KB, Visser ST, Bos HJ, Hak E. Epidemiology. 1996;7: 478-484. Angiotensin-converting enzyme inhibitor 7. Hersom K, Neary MP, Levaux HP, Klaskala W, treatment and the development of urinary Strauss JS. Isotretinoin and antidepressant tract infections: a prescription sequence pharmacotherapy: a prescription symmetry analysis. Drug Saf. 2013;36: 1079- sequence symmetry analysis. J Am Acad 1086. doi: 10.1007/s40264-013-0085-z Dermatol. 2003;49: 424-432. 17. Pratt N, Andersen M, Bergman U, Choi NK, 8. Tsiropoulos I, Andersen M, Hallas J. Adverse Gerhard T, Huang C, et al. Multi-country rapid events with use of antiepileptic drugs: adverse drug event assessment: the Asian a prescription and event symmetry analysis. Pharmacoepidemiology Network (AsPEN) Pharmacoepidemiol Drug Saf. 2009;18: antipsychotic and acute hyperglycaemia 483-491. doi: 10.1002/pds.1736 study. Pharmacoepidemiol Drug Saf. 2013;22: 9. Lai EC, Hsieh CY, Kao Yang YH, Lin SJ. 915-924. doi: 10.1002/pds.3440 Detecting potential adverse reactions 18. Rasmussen L, Hallas J, Madsen KG, of sulpride in schizophrenic patients by Pottegård A. Cardiovascular drugs and prescription sequence symmetry analysis. erectile dysfunction - a symmetry analysis. PLoS One. 2014;9: e89795. doi: 10.1371/ Br J Clin Pharmacol. 2015;80: 1219-1223. journal.pone.0089795 doi: 10.1111/bcp.12696 10. Bytzer P, Hallas J. Drug-induced symptoms of 19. Roughead EE, Kalisch LM, Pratt NL, Killer G, functional dyspepsia and nausea. A symmetry Barnard A, Gilbert AL. Managing glaucoma

179 Chapter 8

in those with co-morbidity: not as easy as 28. Bland JM, Altman DG. Statistical methods for it seems. Ophthalmic Epidemiol. 2012;19: assessing agreement between two methods of 74-82. doi: 10.3109/09286586.2011.638743 clinical measurement. Lancet. 1986;1: 307-310. 20. Van Boven JF, De Jong-van den Berg LT, Vegter 29. Pouwels KB, Widyakusuma NN, Groenwold S. Inhaled corticosteroids and the occurrence RH, Hak E. Quality of reporting of confounding of oral candidiasis: a prescription sequence remained suboptimal after the STROBE symmetry analysis. Drug Saf. 2013;36: guideline. J Clin Epidemiol. 2016;69: 217-224. 231-236. doi: 10.1007/s40264-013-0029-7 doi: 10.1016/j.jclinepi.2015.08.009 21. Vegter S, De Jong-van den Berg LT. 30. Von Elm E, Altman DG, Egger M, Pocock Misdiagnosis and mistreatment of SJ, Gøtzsche PC, Vandenbroucke JP. a common side-efect – angiotensin- The strengthening the reporting of converting enzyme inhibitor-induced observational studies in epidemiology cough. Br J Clin Pharmacol. 2010;69: 200-203. (STROBE) statement: guidelines for reporting doi: 10.1111/j.1365-2125.2009.03571.x observational studies. PLoS Med. 2007;4: 1623- 1627. doi: 10.1371/journal.pmed.0040296 22. Fujimoto M, Higuchi T, Hosomi K, Takada M. Association between statin use and cancer: 31. Ioannidis JPA, Haidich AB, Pappa M, data mining of a spontaneous reporting Pantazis N, Kokori SI, Tektonidou MG, et al. Comparison of evidence of treatment database and a claims database. Int J Med Sci. efects in randomized and nonrandomized 2015;12: 223-233. doi: 10.7150/ijms.10656 studies. JAMA. 2001;286: 821-830. 23. Kalisch Ellett LM, Pratt NL, Barratt JD, 32. Garrison SR, Dormuth CR, Morrow RL, Rowett D, Roughead EE. Risk of medication- Carney GA, Khan KM. Nocturnal leg cramps associated initiation of oxybutynin in and prescription use that precedes them: elderly men and women. J Am Geriatr Soc. a sequence symmetry analysis. Arch Intern 2014;62: 690-695. doi: 10.1111/jgs.12741 Med. 2012;172: 120-126. doi: 10.1001/ 24. Pouwels KB, Kalkman GA, Schagen D, Visser archinternmed.2011.1029 ST, Hak E. Is combined use of SSRIs and NSAIDs 33. U.S. National Library of Medicine. National associated with an increased risk of starting Institutes of Health. Detailed indexing peptic ulcer treatment? Br J Clin Pharmacol. statistics: 1965-2014. 2015. Available from: 2014;78: 192-193. doi: 10.1111/bcp.12300 https://www.nlm.nih.gov/bsd/index_stats_ 25. Takada M, Fujimoto M, Yamazaki K, comp.html. Accessed March 11 2016. Takamoto M, Hosomi K. Association of statin 34. Wahab IA, Pratt NL, Kalisch LM, Roughead use with sleep disturbances: data mining EE. Comparing time to adverse drug of a spontaneous reporting database and reaction signals in a spontaneous database a prescription database. Drug Saf. 2014;37: and a claims database: a case study of 421-431. doi: 10.1007/s40264-014-0163-x rofecoxib-induced myocardial infarction 26. Takeuchi Y, Kajiyama K, Ishiguro C, and rosiglitazone-induced heart failure Uyama Y. Atypical antipsychotics and signals in Australia. Drug Saf. 2014;37: 53-64. the risk of hyperlipidemia: a sequence doi: 10.1007/s40264-013-0124-9 symmetry analysis. Drug Saf. 2015;38: 35. Kubota K. Symmetry analysis for 641-650. doi: 10.1007/s40264-015-0298 monitoring safety of newly marketed drugs. 27. Corrao G, Botteri E, Bagnardi V, Zambon A, Pharmacoepidemiol Drug Saf. 2016;25: Carobbio A, Falcone C et al. Generating signals 349-351. doi: 10.1002/pds.3886 of drug-adverse efects from prescription 36. Thacker EL, Schneeweiss S. Initiation databases and application to the risk of of acetylcholinesterase inhibitors and arrhythmia associated with antibacterials. complications of chronic airways disorders in Pharmacoepidemiol Drug Saf. 2005;14: 31-40. elderly patients. Drug Saf. 2006; 29:1077-1085.

180

CHAPTER 9 General discussion

General Discussion

GENERAL DISCUSSION

The objective of this thesis was to provide a comprehensive profle about the efectiveness and safety of some drugs commonly used by COPD patients, based on a range of pharmacoepidemiological studies. In the frst part of this thesis, we evaluated the efects of several antibiotics for the prevention and treatment of acute exacerbations of COPD (AECOPD), and the potential for drug-drug-interactions (DDIs) during antibiotic therapeutic management. We conducted a systematic review and meta-analysis to provide a beneft-risk profle of prophylactic antibiotics for AECOPD. The infuence of the drug schedule (continuous vs intermittent) and treatment duration of prophylactic antibiotics (<=6 months vs > 6months) on clinical outcomes were also explored (Chapter 2). The efects of doxycycline in addition to oral corticosteroids in AECOPD and possible infuence of age were explored in Chapter 3. To further reduce the infuence of possible unmeasured confounding bias as potentially present in Chapter 3, a cohort study with complete information on COPD clinical diagnosis and lung function was conducted to explore the efects of therapy with any or a specifc frst-line antibiotic in AECOPD (Chapter 4). Considering the frequent occurrence of polypharmacy in COPD drug management, we presented a systematic DDI review 9 focused on frequently used antibiotics among COPD patients to optimize drug treatment (Chapter 5). In the second part of this thesis, we assessed the neuropsychiatric safety of varenicline for smoking cessation in a real-world setting. Using a traditional cohort study, we especially explored the neuropsychiatric safety in COPD patients and those with previous psychiatric disorders as these are high-risk populations for psychiatric events (Chapter 6). Since conventional cohort studies may be vulnerable to potential for confounding bias, we further explored the neuropsychiatric safety of varenicline by use of a self-controlled study design, called Prescription Sequence Symmetry Analysis (PSSA, Chapter 7). Finally, we performed a systematic review comparing the efect estimates between PSSA and traditional parallel group studies with correlation analysis, agreement and discrepancy analysis (Chapter 8).

Antibiotics in management of COPD exacerbations As respiratory bacterial infection is a major risk factor of COPD exacerbations,1,2 antibiotics can be part of the drug management of COPD according to recent guidelines.3 However, these recommendations are based on a limited body of evidence from randomized controlled trials (RCTs) and lack of consistent results from studies in a real-world setting. Therefore, we conducted four separate scientifc studies to provide more evidence for the optimal use of antibiotics among COPD patients.

Efects of antibiotics for preventive use In Chapter 2, we conducted a systematic review to evaluate all possible benefcial and side efects of prophylactic antibiotics in stable COPD patients. Pooled results from

185 Chapter 9 twelve RCTs showed that the frequency of AECOPD and the number of patients with AECOPD were signifcantly reduced, independent of the drug schedule (continuous vs intermittent) and duration of treatment (<=6 months vs >6 months). However, when we examined specifc antibiotics, we could only confrm the superiority of macrolides in preventing exacerbations of COPD and erythromycin and azithromycin appeared the most efective, a fnding which is in line with clinical recommendations.4,5 Previous research suggested properties of anti-infammation and immune-modulation by macrolides,6,7 however, this was not supported by fndings from our review, where we observed changes in neither bacterial load nor airway infammation. Despite the direct benefcial efects of prophylactic antibiotics regarding the reduction of exacerbations, the patients’ quality of life was only improved by longer use (> 6 months) of prophylactic antibiotics. There were no diferences in the rate of hospitalization, adverse events and the time to next exacerbation between patients with prophylactic antibiotics and those on placebo.

Of note, antibiotic resistance problems always come along with antibiotic use, especially for macrolides, for which the review data showed an increase in resistance.8 Weighing benefts and risks of prophylactic antibiotic therapy, the long and continuous use of such therapy should be advised carefully, and it should be used preferably by the high-risk patient population who are at risk for development of severe infections such as patients who are older, with high-risk comorbidities and with higher frequency of exacerbations in the previous year, and those with more severe AECOPD.

Efects of antibiotics for therapeutic use According to the GOLD guideline, amino-penicillin with clavulanic acid, macrolide, or tetracycline antibiotics are recommended as the initial empirical choice of antibiotic treatment for AECOPD.3 In the Netherlands, according to the Dutch primary care guideline, doxycycline or amoxicillin are recommended as the frst choice in AECOPD treatment.9 These recommendations were basically based on results of RCTs. Real- world evidence from observational studies is valuable to evaluate the applicability of the fndings from RCTs for the real-world setting, and these were presented in chapter 3 and 4.

In chapter 3, we conducted a cohort study including the outpatient population from the University of Groningen’s prescription database IADB.nl, to explore the real-world efects of doxycycline for AECOPD and the infuence of age. We found a 23% reduced risk of treatment failure by doxycycline in addition to systemic corticosteroids among COPD outpatients aged 75 years and older; a fnding consistent with results from few previous RCTs about doxycycline for AECOPD among outpatients.10 However, in younger patients with COPD, we did not fnd any diference between additional antibiotic use and oral corticosteroids use only, which is consistent with our hypothesis that older people may

186 General Discussion beneft more from antibiotics due to their susceptibility to bacterial infections and infammations. The fact that the natural lung function, the natural defense mechanism and mucocilliary clearance are reduced with increasing age is also in line with this hypothesis.11-13

However, due to a lack of detailed clinical diagnostic information, the identifcation of COPD and other comorbidities in the study reported in chapter 3 were based on drug prescriptions as proxies for these diseases, which may have biased the results. Moreover, important baseline information on GOLD stages of severity of airway limitations in COPD and smoking history was also absent, though these are vital risk factors for the prognosis of AECOPD and this may have led to unmeasured confounding.3 Accordingly, in Chapter 4, we further explored the efects of both any antibiotic and some specifc antibiotics in the treatment of AECOPD among outpatients based on the “PharmLines Initiative”, which linked extensive clinical information from both the Lifelines Cohort Study and drug information from the IADB.nl prescription database. Within the Lifelines Cohort information on clinical diagnosis of almost all possible chronic diseases and information on parameters of physical examinations like spirometry of lung function is obtained on a regular basis.14 Largely in line with with fndings in chapter 3, overall the prescription 9 of any antibiotic was associated with a statistically signifcant reduction of treatment failure of AECOPD. Similar trends towards protective efects were also observed for the specifc antibiotics doxycycline, co-amoxiclav, and macrolides, separately, except for amoxicillin which was associated with no efect. The doxycycline treatment was even associated with a statistically signifcant 47% reduced risk of treatment failure of AECOPD, after adjustments in both conventional logistic regression and propensity score analysis. Indeed, in Chapter 4, information about the actual severity of the presented frst AECOPD according to signs and symptoms at diagnosis was absent. However, we believe that antibiotics will most likely be given to more severe AECOPD, which may have biased our result towards a null fnding. Further, severe exacerbations that resulted in hospitalizations were not included, but we believe that in an outpatient setting the chance of such severe cases is low.

Although previously two observational studies indicated long-term benefts from short use of antibiotics for next exacerbations,15,16 we did not observe this in both Chapter 3 and Chapter 4. A long-term efect is also doubtful given the fact that re-infections are not substantially afected by a short-course of antibiotics. Our results about the absence of long-term efects of antibiotic treatment are in line with fndings from a more recent RCT reported by van Velzen et al. in 2017.10

Management of DDIs in AECOPD In chapter 3 and 4, it has been shown that antibiotics play a vital role in the management of patients with AECOPD. However, in COPD patients many comorbidities like

187 Chapter 9 cardiovascular disease, diabetes, and lung cancer coexist.17 These comorbidities may contribute to polypharmacy and result in potential DDIs when antibiotics are prescribed simultaneously.

In chapter 5, based on the causal evidence from clinical trials and observational studies with a control group, we found that many drugs interact with commonly used antibiotics in the treatment of COPD. Some co-administered drugs can alter the pharmacokinetics of antibiotics, while other antibiotics can also interfere with the pharmacokinetics of co-administered drugs. DDIs may result in treatment failures of disease or lead to adverse events. For example, clarithromycin as inhibitor of CYP3A4 can increase the risk of hypoglycemia among diabetes patients by inhibition of the metabolic enzymes of related anti-diabetic drugs (e.g. glipizide, glyburide), which are substrates of CYP3A4.

We presented details of potential clinical signifcant DDIs with moderate to strong levels of interaction in this review according to highly prevalent comorbidities, and such information may be used to improve the sensitivity and specifcity of drug-drug interaction alert systems. Importantly, it may help physicians to improve the prescription of antibiotics to COPD patients with comorbidities.

Varenicline intervention for smoking cessation Pharmaceutical smoking cessation treatment (PSCT) is an important intervention for tobacco smoking. Varenicline as frst-line drug of PSCTs has been proven efective for smoking cessation.18 However, concerns about neuropsychiatric adverse events (NPAEs) were raised since the spontaneous reports about such events and the related warning from the FDA.19 Due to strict selection criteria for participants in RCTs, high-risk populations of smokers like those with COPD and psychiatric disorders were usually excluded which hinders making conclusions for a real-world setting.

Neuropsychiatric safety of varenicline for smoking cessation In chapter 6, the association between varenicline use and major NPAEs was explored among general and COPD patients with or without psychiatric disorder. However, we did not observed a signifcant association between varenicline use and the occurrence of any NPAE in high-risk populations and the general population. Although COPD patients are considered more susceptible for possible adverse drug reactions (ADEs), in our study we did not observe an increased risk of any NPAE among COPD patients using varenicline in the psychiatric and non-psychiatric cohorts. This fnding was consistent with the fndings from two earlier studies.20,21

When we examined specifc NPAEs, we observed a signifcantly reduced risk of anxiety among varenicline users in those with psychiatric disorders compared with NRTs users, which may be due to the combined efects of reduced withdrawal-related symptoms

188 General Discussion and raised level of positive afect on mood due to varenicline treatment for smoking cessation. As varenicline users have higher abstinence rates and successful quitting rates than NRT users, quitting of smoking is associated with recovery in psychiatric status for smokers.22

Of note, although the NPAEs observed in earlier studies were defned in diferent ways, the incidence rates of specifc NPAEs related to depression, anxiety and insomnia defned by the occurrence of prescriptions of related drugs in our study were similar to these reports.23 Much higher rates of NPAEs were observed in the specifc high-risk population with psychiatric disorders in this study, which was also consistent with previous reports.24,25

Role of PSSA in drug safety evaluations To overcome the limitations of traditional cohort studies regarding the control of confounding in chapter 6, we conducted a PSSA study described in chapter 7 using the same IADB.nl prescription database. PSSA has been increasingly used for detecting adverse events of medication. Due to its self-controlled study design, the PSSA design may control genetic and other time-invariant confounding efectively. 9 Consistent with results from chapter 6, results in chapter 7 also showed that varenicline was not associated with increased risks for depression or anxiety. However, it was associated with a small signifcant, but transient, increase in sleep disorders, which was a well-known side efect of varenicline from RCTs and other previous studies.23,26 However, it is difcult to identify whether sleeping problems are due to side efects of PSCTs or related to withdrawal from nicotine as difculty falling asleep and increased number of awakenings are also common symptoms of nicotine withdrawal.27

Of note, most observational studies (e.g. the cohort study described in chapter 6) used NRT as the reference group to explore the risk of NPAEs associated with varenicline to make sure the baseline characteristics of study subjects in the comparison groups are more similar. In the PSSA all patients used varenicline and only the sequence orders of prescriptions of varenicline and marker drugs for NPAEs were compared due to its self-controlled design. The diference in outcomes between study designs may be due to either missing a transient risk in the cohort study or by the absence of a reference group in the PSSA.

Although we found consistent fndings in Chapter 6 and 7, the validity and constraints of the PSSA study design regarding the efect estimate of drug use are never compared with conventional observational parallel group study designs (e.g. cohort, case-control), which has already been proven to be an efective design to obtain causal evidence from real-world data.

189 Chapter 9

Therefore, in chapter 8, we compared the efect estimates from two study designs, by systematically searching for publications that explored the efects of the same drug use by applying both of these study designs. Based on the correlation analysis, agreement and discrepancy analysis, this review indicated that the efect estimates generated by the PSSA are usually lower than the efect estimates generated by parallel group designs, and PSSA usually has a lower power than the conventional study designs. However, these results should be interpreted with caution, as the efect estimates were only retrieved from two separate studies. More comparisons are needed to confrm our conclusion.

FUTURE PERSPECTIVES

Although prophylactic antibiotics, especially macrolides, were shown to be efective for preventing exacerbations of COPD in chapter 2, the optimal regimen of antibiotics regarding dose, duration and schedule has not been well established yet, and such treatment is still far from more personalized therapy. For the long and continuous use of antibiotics, it is still an issue how to balance its advantages towards COPD exacerbations and the development of antibacterial resistance in both the individual and the community. It is wise for clinical practitioners to limit their prescription to high-risk populations in order to reduce unnecessary bacterial resistance. Better understanding of the yet unclear mechanisms behind macrolides preventing exacerbations could help develop targeted treatment for AECOPD in the future.

Regarding the antibiotic treatment for ongoing AECOPD, related guidelines were basically depend on the evidence from RCTs.3,28 However, in reality, it is more complicated to make decisions about antibiotic use considering the heterogeneous characteristics in outpatients and various factors that may infuence the fnal treatment outcome. The tendency towards benefcial efects of antibiotics in the elderly COPD patients shown in chapter 3 may ofer clues for clinicians and researchers to use antibiotics in more targeted populations by considering age. A personalized specifc antibiotic treatment could further improve the therapeutic efects in AECOPD, but such evidence is currently lacking.

Of note, antibiotics were not always properly prescribed in line with the guidelines.29,30 Some antibiotics could have been used in AECOPD without actual bacterial infection. Hence, the benefcial efects as reported in chapter 4 may underestimate the true efects in bacterially confrmed AECOPD. In the future, how to improve the appropriate personalized use of antibiotics is a problem that needs to be studied. From a pragmatic perspective, clinicians may reduce unnecessary antibiotic therapy for AECOPD by using sputum color as a predictor of potentially pathogenic bacterial infection in practice.3 A procalcitonin-guided algorithm or C-reactive protein (CPR) test can also be considered,

190 General Discussion if applicable, as a way to better instruct antibiotic use.31 However, both application and accuracy of bacterial tests come with limitations for the outpatient setting, and more practical tests should be developed.

Consistent with the Dutch primary guideline,9 the fndings in chapter 4 confrmed the benefcial efect of doxycycline treatment for AECOPD among outpatients. However, no defnite conclusion can be drawn for other antibiotics from this study. Considering the variability between GP practices in the prescriptions of antibiotics to patients with AECOPD,32 larger studies of high quality with extensive control for confounding by indication are needed to confrm and support their role in the management of AECOPD. Notably, the fnal antibiotic choice for AECOPD treatment should always consider the local bacterial resistance patterns and possibility of resistant pathogens by performing culture of sputum especially among high risk patients with frequent exacerbations and severe airfow limitations.3

Chapter 5 showed that there is a variety of clinically signifcant DDIs between antibiotics and a wide range of drugs that are used to treat related comorbidities in COPD. Clinicians should pay attention to these drug interactions when prescribing antibiotics by assessing the present comorbidities and polypharmacy of patients to ensure 9 therapeutic efect and reduce the possibility of adverse efects. However, the evidence base for clinical adverse outcomes due to DDIs is still weak, and warrants further study in larger cohorts.

Based on previous evidence and the results from our studies in both cohort and PSSA study design in chapter 6 and 7, varenicline is safe to use for smoking cessation among general and COPD populations. Although there was no increased risk of NPAEs by varenicline compared with NRTs among patients with psychiatric disorders, considering the relatively high rate of NPAEs in smokers with psychiatric disorders, these patients should be instructed carefully. Notably, sleeping disorder is a well-recorded and common adverse event of varenicline, especially in the frst three to six months after varenicline initiation. Although sleep disorders do not belong to the “severe” adverse events, it could infuence the uptake and adherence of varenicline. As a result, it may fnally result in the failure of smoking cessation. However, up to now, it is still not clear whether sleep disorders are caused by varenicline itself or are more related to withdrawal of nicotine. The distinction between these causes is important for pharmacists and clinicians to take appropriate actions to improve compliance and adherence of varenicline use.

Based on the fndings from chapter 8, the lack of a wider variety of comparisons between PSSA and parallel group designs that cover more topics make it difcult to make defnite conclusions about the validity of PSSA in establishing associations between drug use and related events. Thus, more studies that explore the association between drug use and adverse events by using both of these study designs based on

191 Chapter 9 the same database or populations are needed. Ideally, comparisons with high-quality pragmatic trials with similar populations, exposure and outcomes could help to specify the validity of the PSSA design even better. CONCLUSION

In the frst part of this thesis, we confrmed the benefcial efects of antibiotics both in the prevention and treatment of AECOPD. Macrolides should be prescribed as the frst- line antibiotic to prevent recurrence of exacerbations and doxycycline appeared the best choice for preventing treatment failure of a current exacerbation. Those with a higher risk of bacterial infections such as older patients among COPD outpatients beneft the most and personalizing therapy for these patients may possibly reduce the development of antibiotic resistance. Considering the polypharmacy among COPD patients, clinicians should pay attention to related DDIs while prescribing antibiotics to avoid treatment failure or adverse events. In the second part of our thesis, neither traditional cohort nor prescription sequence symmetry analysis showed a potential neuropsychiatric risk of varenicline use for smoking cessation among general and COPD patients. Attention should be given to patients with psychiatric disorders. The PSSA design has shown signifcant promise in detecting related drug adverse events, however, due to the limited available comparisons between PSSA and other traditional studies, it is still necessary to test its validity by comparisons with parallel controlled studies and clinical trials.

192 General Discussion

REFERENCES

1. Sethi S, Murphy TF. Infection in 11. Brandsma CA, de Vries M, Costa R, Woldhuis the pathogenesis and course of chronic RR, Konigshof M, Timens W. Lung ageing and obstructive pulmonary disease. N Engl J COPD: is there a role for ageing in abnormal Med. 2008;359(22):2355-2365. tissue repair? Eur Respir Rev. 2017;26(146). 2. Moghoofei M, Azimzadeh Jamalkandi S, 12. Lopez-Otin C, Blasco MA, Partridge L, Moein M, Salimian J, Ahmadi A. Bacterial Serrano M, Kroemer G. The Hallmarks of infections in acute exacerbation of chronic Aging. Cell. 2013;153(6):1194-1217. obstructive pulmonary disease: a systematic 13. Wilkinson TMA, Aris E, Bourne SC, et review and meta-analysis. Infection. 2019. al. Drivers of year-to-year variation in 3. Global Initiative for Chronic Obstructive exacerbation frequency of COPD: analysis Lung Disease (GOLD). Global Strategy for of the AERIS cohort. ERJ Open Res. 2019;5(1). the Diagnosis, Management and Prevention of 14. Scholtens S, Smidt N, Swertz MA, et Chronic Obstructive Pulmonary Disease: 2020 al. Cohort Profle: LifeLines, a three- Report. https://goldcopd.org/gold-reports/. generation cohort study and biobank. Int J Date last accessed: December 17, 2019. Epidemiol. 2015;44(4):1172-1180. 4. Ni WT, Shao XD, Cai XJ, et al. Prophylactic Use 15. Roede BM, Bresser P, Prins JM, Schellevis of Macrolide Antibiotics for the Prevention F, Verheij TJM, Bindels PJE. Reduced of Chronic Obstructive Pulmonary Disease risk of next exacerbation and mortality Exacerbation: A Meta-Analysis. Plos associated with antibiotic use in COPD. Eur 9 One. 2015;10(3). Respir J. 2009;33(2):282-288. 5. Wedzicha JA, Calverley PMA, Albert RK, et al. 16. Roede BM, Bresser P, Bindels PJE, et al. Antibiotic Prevention of COPD exacerbations: a European treatment is associated with reduced risk of Respiratory Society/American Thoracic Society a subsequent exacerbation in obstructive lung guideline. Eur Respir J. 2017;50(3). disease: an historical population based cohort 6. Rubin BK. Immunomodulatory properties of study. Thorax. 2008;63(11):968-973. macrolides: overview and historical perspective. 17. Chetty U, McLean G, Morrison D, Agur K, Am J Med. 2004;117 Suppl 9A:2S-4S. Guthrie B, Mercer SW. Chronic obstructive 7. Martinez FJ, Curtis JL, Albert R. Role of pulmonary disease and comorbidities: macrolide therapy in chronic obstructive a large cross-sectional study in primary pulmonary disease. Int J Chron Obstruct care. Br J Gen Pract. 2017;67(658):e321-e328. Pulmon Dis. 2008;3(3):331-350. 18. Anthenelli RM, Benowitz NL, West R, et al. 8. Serisier DJ. Risks of population antimicrobial Neuropsychiatric safety and efcacy of resistance associated with chronic macrolide varenicline, bupropion, and nicotine patch use for infammatory airway diseases. Lancet in smokers with and without psychiatric Resp Med. 2013;1(3):262-274. disorders (EAGLES): a double-blind, 9. Snoeck-Stroband JB, Schermer TRJ, Van randomised, placebo-controlled clinical Schayck CP, et al. NHG-Werkgroep Astma trial. Lancet. 2016;387(10037):2507-2520. bij volwassenen en COPD. NHG-Standaard 19. US Food and Drug Administration. Public COPD (derde herziening). Huisarts health advisory: FDA requires new boxed Wet 2015; 58(4):198-211. warnings for the smoking cessation drugs 10. van Velzen P, Ter Riet G, Bresser P, et al. Chantix and Zyban. http://wayback.archive- Doxycycline for outpatient-treated acute it.org/7993/20170112005513/http://www. exacerbations of COPD: a randomised fda.gov/Drugs/DrugSafety/Postmarket Drug double-blind placebo-controlled trial. Safety Information for Patients and Providers/ Lancet Respir Med. 2017;5(6):492-499.

193 Chapter 9

ucm169988.htm. Published July 1, 2009. Used in General Practice in England Interim Accessed June 7, 2019. Results from a Prescription-Event Monitoring 20. Kotz D, Viechtbauer W, Simpson CR, Study. Drug Safety. 2009;32(6):499-507. van Schayck OCP, West R, Sheikh A. 27. Ashare RL, Lerman C, Tyndale RF, et al. Sleep Cardiovascular and neuropsychiatric risks of Disturbance During Smoking Cessation: varenicline and bupropion in smokers with Withdrawal or Side Efect of Treatment? J chronic obstructive pulmonary disease. Smok Cessat. 2017;12(2):63-70. Thorax. 2017;72(10):905-911. 28. Wedzicha JAEC-C, Miravitlles M, Hurst JR, 21. Tashkin DP, Rennard S, Hays JT, Ma W, et al. Management of COPD exacerbations: Lawrence D, Lee TC. Efects of Varenicline on a European Respiratory Society/ Smoking Cessation in Patients With Mild to American Thoracic Society guideline. Eur Moderate COPD A Randomized Controlled Respir J. 2017;49(3). Trial. Chest. 2011;139(3):591-599. 29. Bathoorn E, Groenhof F, Hendrix R, et al. 22. Taylor G, McNeill A, Girling A, Farley A, Lindson- Real-life data on antibiotic prescription Hawley N, Aveyard P. Change in mental health and sputum culture diagnostics in acute after smoking cessation: systematic review exacerbations of COPD in primary care. Int J and meta-analysis. BMJ. 2014;348:g1151. Chron Obstruct Pulmon Dis. 2017;12:285-290. 23. Thomas KH, Martin RM, Knipe DW, Higgins JP, Gunnell D. Risk of neuropsychiatric adverse 30. Roede BM, Bindels PJ, Brouwer HJ, Bresser events associated with varenicline: systematic P, de Borgie CA, Prins JM. Antibiotics and review and meta-analysis. BMJ. 2015;350:h1109. steroids for exacerbations of COPD in primary care: compliance with Dutch guidelines. Br J 24. Kanner AM, Wuu J, Faught E, et al. A past psychiatric history may be a risk factor for topiramate-related Gen Pract. 2006;56(530):662-665. psychiatric and cognitive adverse events. Epilepsy 31. Butler CC, Gillespie D, White P, et al. C-Reactive Behav. 2003;4(5):548-552. Protein Testing to Guide Antibiotic 25. Nanni RC, Troisi A. Maternal attachment Prescribing for COPD Exacerbations. N Engl style and psychiatric history as independent J Med. 2019;381(2):111-120. predictors of mood symptoms in 32. Boggon R, Hubbard R, Smeeth L, et the immediate postpartum period. J Afect al. Variability of antibiotic prescribing Disorders. 2017;212:73-77. in patients with chronic obstructive 26. Kasliwal R, Wilton LV, Shakir SAW. Safety pulmonary disease exacerbations: a cohort and Drug Utilization Profle of Varenicline as study. BMC Pulm Med. 2013;13:32.

194

CHAPTER10 Summary CHAPTER10 Samenvatting Acknowledgement List of publication About the author

Summary

SUMMARY

Chronic obstructive pulmonary disease (COPD) is a very common chronic progressive disease that afects millions of middle-aged and older smokers. COPD patients frequently sufer from exacerbations (worsening of their respiratory symptoms). As bacterial infection is the primary trigger of exacerbations, antibiotics can be given to those patients with potential infections. However, the efects of antibiotics on COPD exacerbations in both treatment and prevention have not been consistently reported and are especially unclear in real-world settings. Smoking is an important risk factor for COPD. Stopping smoking, especially by means of pharmaceutical treatment with varenicline, is a vital intervention to improve the quality of life and reduce smoking- related disease among COPD patients. However, concerns about varenicline’s neuropsychiatric safety for patients still exist due to conficting evidence obtained from randomized controlled trials and case-report systems. It is necessary to evaluate the risk of the neuropsychiatric adverse events (NPAEs) due to varenicline by studies based on a large observational database.

The aim of the frst part of this thesis was to evaluate the efectiveness of antibiotic use to treat COPD exacerbations and explore the potential drug-drug interactions (DDIs) related to antibiotics use among COPD patients. The results from a systematic review described in Chapter 2 show that prophylactic antibiotics are efective in preventing 10 COPD exacerbations and improve the quality of life of stable COPD patients. Concerning the efects of antibiotics treatment for COPD exacerbations, however, the cohort study based on the prescription database described in Chapter 3 shows that the short-term benefts of doxycycline in addition to oral corticosteroids are only observed in COPD patients of advanced age. No long-term benefts were observed. Clinicians should take the patients’ age into consideration when prescribing antibiotics to patients with acute exacerbation of COPD (AECOPD). Another cohort study, described in Chapter 4 and concerning clinically diagnosed COPD, confrmed the short-term protective efects of doxycycline use with AECOPD. However, the treatment efects of other antibiotics in real-world settings are still unclear and need to be explored further in more extensive research. In general practice, physicians may prescribe antibiotics erroneously or improperly, which may underestimate the real-world efects of antibiotic use on AECOPD. In addition, since comorbidities and polypharmacy are common among COPD patients, associated DDIs– of which there are many and which are summarized in a systematic review described in Chapter 5 – may also infuence the efects of antibiotics by directly altering their pharmacokinetics or by indirectly infuencing patient compliance due to adverse drug reactions (ADRs) due to DDIs.

The second part of this thesis focused on the neuropsychiatric safety of varenicline to aid smoking cessation among patients in general and COPD patients in particular.

199 Chapter 10

The fndings from a cohort study described in Chapter 6 do not indicate an increased risk of NPAEs associated with varenicline in either the general or the COPD population, irrespective of the psychiatric status of the patients. Considering the limitation for controlling of time-invariant confounders in traditional observational research, Chapter 7 reports a self-controlled study of a prescription sequence symmetry analysis (PSSA) to explore the varenicline’s risk of NPAEs further. The PSSA results also confrm the neuropsychiatric safety of varenicline use. Remarkably, a transient increased risk of varenicline-induced sleep disorder was observed, which clinicians should take into account to reduce its infuence on the adherence of patients taking varenicline. In the last chapter of the thesis, the efect estimate of the PSSA was compared with parallel group study designs, showing that the efects estimated by the PSSA may be lower than those in the parallel group study designs. However, since only two studies were included, more comparisons are necessary to draw solid conclusions.

In this thesis, we confrm the benefcial efects of prophylactic antibiotic use to prevent COPD exacerbations. However, the antibiotic treatment efects on AECOPD were only seen with doxycycline. Although other antibiotics also showed benefts, these were not statistically signifcant and need be investigated further. We confrm the safety of varenicline use for smoking cessation, varenicline as an efective treatment should be used widely to reduce the burden of smoking-related diseases. However, patients with neuropsychiatric disease in their history must be monitored closely considering their higher risk of NPAEs. Since the PSSA is an efective tool for identifying ADRs, it could be applied more widely to better evaluate the efectiveness and safety of medications. Other study designs and methodologies should be explored to achieve better control of related confounders in observational studies based on real-world data.

200 Samenvatting Samenvatting

SAMENVATTING

Chronic Obstructive Pulmonary Disease, oftewel COPD, is een veelvoorkomende chronische en progressieve ziekte waar miljoenen rokers op middelbare en hogere leeftijd last van krijgen. COPD-patiënten lijden regelmatig aan exacerbaties (verergering van hun ademhalingsklachten). Omdat bacteriële infecties vaak de hoofdoorzaak zijn van exacerbaties, kunnen antibiotica toegediend worden bij patiënten met potentiële infecties. De efecten van antibiotica op COPD exacerbaties bij zowel behandeling als preventie zijn echter niet nauwkeurig geschat en zijn met name onduidelijk in de praktijk. Roken is een belangrijke risicofactor voor COPD. Stoppen met roken, vooral met behulp van farmaceutische behandeling met varenicline, is een essentiële ingreep om de kwaliteit van leven te verbeteren en om ziektes die verband houden met roken te verminderen bij COPD-patiënten. Echter, bezorgdheid om de neuropsychiatrische veiligheid van varenicline voor de patiënten blijft bestaan door tegenstrijdig bewijs dat voortkomt uit randomized controlled trials en case-reports. Het is van belang om het risico op neuropsychiatrische bijwerkingen bij vareniclinegebruik te evalueren met behulp van grootschalige observationele databases.

Het doel van het eerste deel van deze thesis is het evalueren van de efectiviteit van antibiotica voor de behandeling van COPD-exacerbaties, en om de potentiële interacties tussen geneesmiddelen (drug-drug interactions ofwel DDI’s) in relatie 10 tot antibioticagebruik bij COPD-patiënten te onderzoeken. De resultaten van een systematisch onderzoek in hoofdstuk 2 laten zien dat profylactische antibiotica efectief zijn in het voorkomen van COPD-exacerbaties en de kwaliteit van leven bij stabiele COPD-patiënten kan verbeteren. We kijken verder naar van de efecten van antibioticabehandelingen tijdens COPD-exacerbaties in hoofdstuk 3 met behulp van een cohort onderzoek dat gebaseerd is op een prescriptie-database. Dit onderzoek toont aan dat naast het gebruik van orale corticosteroïden de voordelen van doxycycline op korte termijn alleen zichtbaar zijn bij COPD-patiënten op hogere leeftijd. Er werden geen voordelen op de lange termijn gevonden. Clinici zouden daarom de leeftijd van de patiënten in acht moeten nemen als zij antibiotica voorschrijven aan patiënten met acute exacerbatie van COPD (AECOPD). Een vervolgstudie met klinisch gediagnostiseerde COPD, beschreven in hoofdstuk 4, bevestigt dat er kortdurende beschermende efecten van doxycyclinegebruik bij AECOPD zijn. De efecten van andere antibiotica in de praktijk zijn echter nog steeds onduidelijk en moeten nader onderzocht worden in een uitgebreider onderzoek. In het dagelijks leven kan het voorkomen dat artsen antibiotica onterecht of onjuist voorschrijven, wat ervoor kan zorgen dat de efecten van antibiotica op AECOPD in de praktijk onderschat worden. Daar komt nog bij dat comorbiditeiten en polyfarmacie gebruikelijk zijn bij COPD- patiënten. Daardoor kunnen geassocieerde DDI’s – die in groten getale voorkomen en zijn samengevat in een systematisch overzicht in hoofdstuk 5 – ook invloed hebben

201 Chapter 10 op de efecten van antibiotica door directe veranderingen te induceren in hun farmacokinetiek of door het indirect beïnvloeden van de therapietrouw van patiënten door bijwerkingen ten gevolge van de DDI.

Het tweede gedeelte van deze scriptie focust op de neuropsychiatrische veiligheid van varenicline als geneesmiddel om patiënten te helpen stoppen met roken, met name bij COPD-patiënten. De bevindingen van een cohortstudie, zoals beschreven in hoofdstuk 6, tonen niet aan dat er een verhoogd risico van NPAE’s geassocieerd is met varenicline in zowel de algemene als de COPD-populatie, ongeacht wat hun psychiatrische status is. In hoofdstuk 7 houden we rekening met zogenaamde tijd ongebonden verstorende factoren (confounding) die in traditioneel observatieonderzoek kunnen optreden. Hiertoe pasten we de zogenaamde prescription sequence symmetry analysis (PSSA) toe om het risico op neuropsychiatrische bijwerkingen van varenicline verder te onderzoeken. De PSSA-resultaten bevestigen ook de neuropsychiatrische veiligheid van vareniclinegebruik. Opmerkelijk is dat een kortstondig verhoogd risico op slaapstoornis werd gerapporteerd bij vareniciline gebruik, waar clinici rekening mee zouden moeten houden om de invloed van varenicline op de therapietrouw van de patiënten te verminderen. In het laatste hoofdstuk van deze scriptie wordt het geschatte efect van de PSSA vergeleken met een parallelgroep-studie, die aantoont dat de geschatte efecten bij de PSSA lager zouden kunnen zijn dan die in de parallelgroep- studie. Maar omdat dit slechts twee onderzoeken omvat, zijn er meer vergelijkingen nodig om betrouwbare conclusies te trekken.

In deze scriptie bevestigen we de positieve efecten van profylactisch antibioticagebruik om COPD-exacerbaties te voorkomen. Echter, de efecten van de antibioticabehandelingen op AECOPD kwamen alleen voor bij doxycycline. Hoewel andere antibiotica ook voordelen toonden, waren deze resultaten niet statistisch signifcant en zullen daarom verder onderzocht moeten worden. We bevestigen de veiligheid van vareniclinegebruik voor het stoppen met roken, en varenicline als een efectieve behandeling zou veelal ingezet moeten worden om de last van ziektes die met roken verband houden te verminderen. Desalniettemin moeten patiënten met een verleden van neuropsychiatrische aandoeningen nauwlettend in de gaten gehouden worden omdat bij hen een hoger risico van neuropsychiatrische bijwerkingen bestaat. Omdat de PSSA een efectief hulpmiddel is om bijwerkingen te identifceren, zou deze techniek ook wijder verbreid toegepast kunnen worden om de efectiviteit en veiligheid van medicatie beter te kunnen inschatten. Andere methodologieën moeten onderzocht worden om betere beheersing te hebben over de verstorende factoren (confounders) in observationele onderzoeken gebaseerd op data uit de praktijk.

202

Acknowledgement

ACKNOWLEDGEMENT

I am very excited at this moment to defend my PhD degree after four years of study at the University of Groningen. Since 2007, I have been studying in the feld of Public Health for approximately twelve years, when I was so young and had never imagined to go so far towards reaching my dream as being an epidemiologist and an academic researcher. What surprised me most is that I met my Mr. Right and had my adorable son during my PhD study in the Netherlands. I am so grateful for all wonderful things that happened to me and I know that I could not achieve these without the kind help and support from many people, whom I would like to thank sincerely.

First and foremost, I would like to express my deepest gratitude to my supervisor Prof. Eelko Hak. Dear Eelko, thank you for providing me the great opportunity to study in your fantastic group. You always give the continuous support for my studies and life during my PhD trajectory. Although the research did not go smoothly at the beginning of my research, your patience and guidance with professional knowledge helped me to continue the projects in the right direction. For me, you are not only an easy-going supervisor, but also a good friend. Thanks for visiting my family after my son was born and I also enjoyed visiting your family members together with our colleagues in Giethoorn, which is a beautiful memory for me.

In addition, I would like to express my sincere gratitude to my second supervisor Prof. H. Marike Boezen. Dear Marike, thank you very much for your valuable suggestions and discussions on my projects, especially in the vital stage of organizing my thesis. You dedicated lots of time revising my chapters, even when it was close to the Christmas holiday. Through the discussion in our regular meetings, I got lots of insightful comments and encouragement. I benefted a lot from your professional feedback and points of view.

Many thanks also go to my assessment committee members: Prof. T.J.M. Verheij, Prof. J. van der Palen and Prof. Y. Stienstra for taking valuable time in reading and assessing my thesis.

I would like to thank my co-authors Prof. Bob Willfert, Dr. C.C.M.(Nynke) Schuiling- veninga, Mr. Jens. H.J. Bos from our department and Dr. Job J.M.D. van Boven, Dr. Jan-willem C. Alfenaar, Prof. Rolf H.H. Groenwold from other institutes, who gave me enormous support to fnish my projects listed in my thesis. Dear Bob, thank you for lots of support at the beginning of my PhD study and good feedback. As co-author, you always provided detailed revisions to improve my manuscripts. Dear Nynke, I’m very grateful for your help in the stage of protocol writing and study design by providing professional suggestions about drug prescriptions. Dear Jens, thank you for your continuous technical support in working with the prescription database. I have learned

204 Acknowledgement a lot from you about SQL-coding. Dear Job, you always show passion in research and are glad to ofer any help that you could. I am really grateful for the help about calculation of null-efect sequence ratio in PSSA project, which speeded up the completion of that project. Dear Jan-willem, you really ofered many valuable suggestions, based on your research background, in the projects on antibiotic use for COPD exacerbations. Dear Rolf, thank you so much for your statistical support in the PharmLines project. You always ofered detailed explanations for our questions, I learned a lot from you.

Special thanks also to prof. W.J. (Wim) Quax. Dear Wim, thank you for introducing me to the pharmacoepidemiological research at the University of Groningen in November 2014 in Beijing. I still remember that morning when we met each other at the PhD workshop. Without the informal but sweet conversation, my story in Groningen would not have happened.

I would like to show gratitude here to my MS supervisor at the Peking University, Prof. dr. Aiguo Ren. Your previous support made it possible for me to pursue my PhD study abroad. In fact, your words that “Do not label yourself” gave me much courage to explore the possibility of myself in the academic feld. You always show respect and understanding to me, even though I have left our institute for several years. Each contact with you makes me feel warm and full of strength. 10 I also would like to give my gratitude to my colleagues in the department of PharmcoTherapy, -Epidemiology & Economics (PTEE). Dear Jannie, Anja and Felicia, thank you for dealing with all kinds of things for conferences, Dutch translation and more during my PhD study. Dear Bert and Jugo, thank you for solving all the problems related to my computer and software. Dear Sylvi, Akbar and Ivan, as ofcemates, your encouragement and support means a lot to me, I will never forget the joyful moments we shared together in the past four years. Dear Jurjen (My paranymph), Heleen, Linda, Eva, Pieter, Christiaan, Simon and Thea, as typical Dutch, you are so kind and nice to ofer any help during my stay in Groningen. I really enjoyed the time that we spent together. Dear Sofa, Ira, Taichi, Fajri, Abrham, Monik, Tia, Lusi, Affah, yunyu and all other members in PTEE, I have many beautiful memories of all of you. I benefted from our regular meetings and PhD workshop to discuss projects and exchange ideas.

Next, I would like to acknowledge the master students I supervised in the PTEE group: Victor, Tanja, Demy, Anouk and Yingxuan. You are very smart students and I learned a lot by routine discussions with you about the details of our projects. Victor, thank you so much for helping me preparing the protocol forms for the PharmLines project during my maternity leave. Without your help, I could not fnish that project smoothly.

I have met many fantastic Chinese friends during my PhD study in Groningen, they are the precious possessions in my life. Dear Qi, I am so happy for you that you got

205 Acknowledgement a new wonderful job in MSD and found your true love Jia. Without you, I would not have met the wonderful friends that I would like to thank below. Dear Jun, thank you for introducing me to my Mr. Right and helping me know him well. Dear Jiacong, Xiaohong, you are the friends that I could always share with freely and safely, the trip together with you in Gran Canaria is really a wonderful memory. Dear Haoxiao & Bing, Yi & Ting, Yingruo & Qihui, Tiantian & Siqi, Jiaying & Jing, Baojie & Xiaodong, Yifei & Yihui, Changsen & Yingying, Yuanyuan (Shen) & Yanji, Jingjing & Yuanze, Qingkai & Mengmeng, Wangli & Jianjun, Jiapan & Chenyu, Beibei & Jiawen, Ping & Yegang, Rui & Robert and Ying’s family, you are the fantastic couples that I have met and you helped me to understand how to support each other in the marriage that is full of love and understanding. I will not forget the wonderful times that we spent together while enjoying dinners and sharing our stories. Dear Xiaoming, Huatang, Yanan, Zhenchen (& Shilin), Hongyan, Shuai, Cong and Jing (Wu) thank you for the nice meals and especially the kind help ofered in my frst half year in Groningen, I really had a wonderful time with you. Dear Yanni (my paranymph), , Xiaojing, Fangfang, Si, Jing (Li), Yizhou, Keni, Huala, Yehan, Yichen, Yuzhen, Mingming, Kai, Bin (UMCG) & Lin, Bin (Zernike), Haigen, Chao, Hao, Chengtao, Jielin, Surigula, Chan, Jing (Du), Tian (Xie), Qingqing (Cai), Qing (Chen), Lianhong, Cancan & Shixian, Lianmin, Yuntao, Yihui and all other friends that I forgot to mention here, I am so lucky to know you in this wonderful place. It is really amazing that we shared our lives and dreams together and helped each other to overcome difculties here thousands miles away from our home. I could not appreciate it more. Wish you all a bright future.

Besides, I also got lots of support from my friends far away from Groningen for my PhD. Dear Pei Qin in China, We have known each other for twelve years and faced difculties in our life together. I really miss the time we spent together during our bachelor and master study. Wish your daughter Ruoxi grows up happily and will have a wonderful friendship in her future. Dear You Li in UK, thank you for your kind help ofering all the possible positions for me, you are really smart and kind person, wish you have more achievements in the feld of Public Health. Dear Xue Li in UK, thanks for your kind help on the way of my study, especially for sharing me the PhD workshop in Beijing 2014, where I had a chance by surprise with a better choice of my PhD study. Looking forward to seeing your lovely baby. Dear Helen (Ying He) in UK, thank you for your encouragement and support all the time, you are the girl with passion and dreams always. I learned a lot from you.

I would like to thank the China Scholarship Committee (CSC) for ofering the funding for my PhD study, I am lucky to beneft from the fast development of China. China is now experiencing a fght towards the novel coronavirus (2019-nCoV). I wish the 2019-nCoV will be controlled early and the Public Health System in China will be getting better and better.

206 Acknowledgement

Finally, I would like to give my great gratitude to my parents, Guoyin and Fengli. Dear dad and mom, thank you for your love and support all the time. In my last year of my PhD, you even take a long trip from China to the Netherlands to help me get over difculties in my life. You are the best friends and teachers in my life, I love you. Dear brother Yuanpeng, thank you for taking care of our parents when I study abroad, I know it is a little hard for you this year, there are always ups and downs in life, but you will defnitely embrace a bright future in another way. I am grateful for my parents-in-law, thank you for your love and understanding, although we have not yet known each other well. But through your son, I know where the bright sides of him come from. Wish you be healthy and happy in Jinan. Particularly, I would like to express my gratitude to my deeply loved husband, Liqiang. Meeting you is the most unexpected surprise for me in The Netherlands. We have so much diferences in many aspects, but these help me to see the beautiful world in other ways and make my life complete. Thank you for your understanding, support and encouragement all the time. My lovely son, Youran, I love you so much, you are my sunshine, bringing me lots of happiness and helping me to understand love deeper and comprehensively, I really enjoy reading books with you every night. Wish you could grow up happily just like your name.

Finishing the PhD is the end of my education, but it is also a new beginning of my academic career. Knowledge is infnite. I really enjoy exploring and swimming in the sea 10 of knowledge. I may have lots of regret and lost times in my past, but I will hold it tightly in my future to pursue my dreams without any regret.

Yuanyuan Wang February 2020 Groningen

207 List of publications

LIST OF PUBLICATIONS

Yuanyuan Wang, Tanja R. Zijp, Muh. Akbar Bahar, Janwillem W.H. Kocks, Bob Wilfert, Eelko Hak. Efects of prophylactic antibiotics on patients with stable COPD: a systematic review and meta-analysis of randomized controlled trials. Journal of Antimicrobial Chemotherapy. 2018 Dec 1; 73(12): 3231-3243.

Yuanyuan Wang, Muh. Akbar Bahar, Anouk M.E. Jansen, Janwillem W.H. Kocks, Jan-Willem C. Alfenaar, Eelko Hak, Bob Wilfert, Sander D. Borgsteede. Improving antibacterial prescribing safety in the management of COPD exacerbations: systematic review of observational and clinical studies on potential drug interactions associated with frequently prescribed antibacterials among COPD patients. Journal of Antimicrobial Chemotherapy. 2019 Oct 1; 74(10):2848-2864.

Yuanyuan Wang, Jens H. Bos, H. Marike Boezen, Jan-Willem C. Alfenaar, Job F.M. van Boven, Catharina C.M. Schuiling-Veninga, Bob Wilfert, Eelko Hak. The infuence of age on real-life efects of doxycycline for acute exacerbations among COPD outpatients: a population-based cohort study. BMJ Open Respiratory Research. 2020 Feb; 7(1). pii: e000535. doi: 10.1136/bmjresp-2019-000535.

Yuanyuan Wang, Jens H. Bos, Catharina C.M. Schuiling-Veninga, Job F.M. van Boven, Bob Wilfert, Eelko Hak. Neuropsychiatric safety of varenicline in the general and COPD population with and without psychiatric disorders: a retrospective inception cohort study in a real-world setting. European Addiction Research. Submitted on 2020 Jan 27. (Under review)

Yuanyuan Wang, Job F.M. van Boven, Jens H. Bos, Catharina C.M. Schuiling-Veninga, H. Marike Boezen, Eelko Hak. Risk of Neuropsychiatric Adverse Events Associated with Varenicline Treatment for Smoking Cessation in Outpatients: A prescription Sequence Symmetry Analysis. Drug Safety. Submitted on 2020 Feb 4. (Under review)

Yuanyuan Wang, Yingxuan Feng, H. Marike Boezen, Bob Wilfert, Eelko Hak. Pharmaceutical Smoking Cessation Treatment and Risk of Neuropsychiatric Adverse Events: A Systematic Review of Evidence from Observational Studies. (In submission)

Yuanyuan Wang, Victor Pera, Jens H. Bos, H. Marike Boezen, Jan-Willem C. Alfenaar, Bob Wilfert, Rolf H.H. Groenwold, Hans Wouters, Marco Grzegorczyk, Eelko Hak. Real- world short- and long-term efects of antibiotic therapy on acute exacerbations of COPD in outpatients: a cohort study under the PharmLines Initiative jointed database. (In submission)

Muh. Akbar Bahar, Yuanyuan Wang, Jens H. Bos, Bob Wilfert, Eelko Hak. Discontinuation and dose adjustment of metoprolol after metoprolol-paroxetine/fuoxetine co-

208 List of publications prescription in Dutch elderly. Pharmacoepidemiology and Drug Safety. 2018 Jun; 27(6):621-629.

Demy L. Idema, Yuanyuan Wang, Michael Biehl, Peter L. Horvatovich, Eelko Hak. Efect estimate comparison between the prescription sequence symmetry analysis (PSSA) and parallel group study designs: A systematic review. PloS One. 2018 Dec 6; 13(12): e0208389.

Before Groningen Shanshan Lin, Aiguo Ren, Linlin Wang, Yun Huang, Yuanyuan Wang, Caiyun Wang, Nicholas Greene. Oxidative stress and apoptosis in benzo[a]pyrene-induced neural tube defects. Free Radical Biology and Medicine. 2018 Feb 20; 116: 149-158.

Yuanyuan Wang, Lei Jin, Jufen Liu, Yali Zhang, Zhiwen Li, Aiguo Ren. Autopsy fndings of 95 cases of neural tube defects and a comparison with clinical reports. Chinese Journal of Reproductive Health. 2015, 26(3): 207-210. (In Chinese)

International Conference Yuanyuan Wang, Jens H. Bos, Catharina C.M. Schuiling-Veninga, et al. The efects of amoxicillin on acute COPD exacerbations in the outpatient setting: a retrospective cohort study based on real-world data. Poster session presented in 35th International 10 Conference of Pharmacoepidemiology and Therapeutic Risk Management (ICPE), August 24-28, 2019. Philadelphia, Pennsylvania, US.

Yuanyuan Wang, Jens H. Bos, Catharina C.M. Schuiling-Veninga, et al. Real-world data on the efect of doxycycline plus prednisone/prednisolone on COPD exacerbations compared with prednisone/prednisolone alone: a retrospective cohort study among COPD outpatients. Poster session presented in 34th International Conference of Pharmarcoepidemiolgy and Therapeutic Risk Management (ICPE). August 22-26. 2018. Prague, Czech Republic.

Yuanyuan Wang, Tanja Zijp, Muh Akbar Bahar, et al. The clinical efect of prophylactic antibiotics on COPD patients: an updated systematic review and meta-analysis. Poster session presented in 33rd International Conference of Pharmarcoepidemiolgy and Therapeutic Risk Management (ICPE). August 26-30. 2017. Montreal, Canada.

Yuanyuan Wang, Anouk M.E. Jansen, Sander D. Borgsteede, et al. Drug-drug interactions and clinical signifcance of frequently prescribed antibiotics for treating exacerbations in COPD patients with other co-administered agents: a systematic review. Poster session presented in 33rd International Conference of Pharmarcoepidemiolgy and Therapeutic Risk Management (ICPE). August 26-30. 2017. Montreal, Canada.

209 List of publications

Yuanyuan Wang, Anouk J.M. Jansen, Jens Bos, Eelko Hak. Drug interactions as a potential instrumental variable: the example of efectiveness of antibiotic therapy in exacerbated COPD patients. Poster session presented in 32nd International Conference of Pharmarcoepidemiolgy and Therapeutic Risk Management (ICPE). August 25-28. 2016. Dublin, Ireland.

210 About the author ABOUT THE AUTHOR ABOUT THE AUTHOR Yuanyuan Wang wasYuanyuan born on Wang February was born 15 onth, February1988 in 15Henan,th, 1988 China. in Henan, She China. She studied Preventive Medicine at the School of studied Preventive Medicine at the School of Public Heath in Public Heath for fve years in Zhengzhou University, and Zhengzhou University,obtained and obtainedher Bachelor he rof Bachelor Medicine of degree Medicine in July degree 2012. in July 2012. Meanwhile,Meanwhile, she was awardedshe was awarded another another degree degree of Bachelor of Bachelor of Arts of Arts from Zhengzhou University due to her personal from Zhengzhou Universityinterest and due study to her of the personal English interest literature. and After study that, of she the English literature. Aftercontinued that, her she master continued studies in her the masterSchool of studies Public Health in the at the Peking University and achieved the degree of Master School of Public Healthof Public at the Health Peking with University a research anddirection achieved in Reproductive the degree of Master of PublicEpidemiology Health with in a July research 2015. During direction her inmaster Reproductive programme, Epidemiology she established in her July research 2015. interest in the feld of Epidemiology. In October of the same year, she then came to During her master programme, she established her research interest in the field of Epidemiology. In the Netherlands to pursue her PhD studies in the department of PharmacoTherapy, October of the same-Epidemiology year, she then and –Economics came to th (PTEE)e Netherlands at the Groningen to pursue Research her Institute PhD studies of Pharmacy in the department of PharmacoTherapy,(GRIP) of the University -Epidemiology of Groningen and –Economics under the primary (PTEE) supervisionat the Groningen of Prof. Research dr. Eelko Hak. Her doctoral research focused on the efectiveness and safety of medications used Institute of Pharmacyin COPD(GRIP) patients, of the University which as described of Groningen in this under thesis. the primary supervision of Prof. dr. Eelko Hak. Her doctoral research focused on the effectiveness and safety of medications used in 10 COPD patients, which as described in this thesis.

211