Peter Hanlon, Elaine W. Butterly, Anoop SV Shah, Laurie J. Hannigan, Jim Lewsey, Frances S. Mair, David M. Kent,Bruce Guthrie, Sarah H. Wild, Nicky J. Welton, Sofia Dias,David A. McAllister
Abstract
Background
People with comorbidities are underrepresented in clinical trials. Empirical estimates of treatment effect modification by comorbidity are lacking, leading to uncertainty in treatment recommendations. We aimed to produce estimates of treatment effect modification by comorbidity using individual participant data (IPD).
Introduction
Multimorbidity, the presence of 2 or more long-term conditions, is a global clinical and public health priority [1,2]. Most people with a given long-term condition also have comorbidities (referring to additional long-term conditions in the context of an index condition). There is uncertainty about how individual long-term conditions should be managed in the presence of comorbidities [3]. A major driver of this uncertainty is the underrepresentation of people with multimorbidity in randomised controlled trials (RCTs) [4,5]. Trial populations are typically younger, healthier, and have fewer comorbidities than people treated in routine clinical practice. This has led clinical guideline developers to caution against the application of single-disease recommendations for people with multimorbidity [6]. However, despite the challenges to clinical management posed by this uncertainty, the efficacy of treatments in the context of comorbidity is rarely assessed. It is therefore not clear, for most treatments, whether relative treatment efficacy differs in people with comorbidity.
Materials and method
For trials of 23 index conditions, we identified comorbid long-term conditions using IPD for each trial. We then summarised these as a comorbidity count (in addition to the index condition) for each participant. Further, we identified the 6 commonest comorbidities for each index condition across trials and defined a presence/absence variable for each. We estimated differences in treatment efficacy by fitting regression models to IPD for each trial to obtain trial-level estimates of covariate–treatment interaction effects. We fit models for age and sex alone, for a comorbidity count, and for each of the 6 commonest comorbidities for each index condition. Trial-level estimates were then meta-analysed to obtain drug and index condition-specific estimates of treatment effect modification by comorbidity.
Results
Trial baseline characteristics have been reported previously [4]. For trials with continuous outcomes, there were 20 index conditions and 47 treatment comparisons across a total of 106 trials (n = 88,150 participants). For 9 index conditions, there was only 1 treatment comparison across all trials. Diabetes,
which was the condition for which there were the most trials (22), had the largest number of treatment comparisons (9) (Table 1). Within each model, all trials had a single common outcome except inflammatory bowel disease, where the ulcerative colitis trials used the MAYO score and Crohn’s disease trials used the Crohn’s Disease Activity Index score. For trials with categorical outcomes, there were 3 index conditions (migraine, osteoporosis, and thromboembolism) and 11 treatment comparisons across a total of 17 trials (n = 11,624 participants). For thromboembolism, there were 3 more specific categories of indication—primary prevention (5 trials), secondary prevention (2 trials), and treatment (2 trials).
Discussion
In an IPD meta-analysis of 120 trials, we examined whether the efficacy of drug treatments differed by comorbidity. For 20 index conditions where the outcome variable was continuous (e.g., glycosylated haemoglobin in diabetes trials), efficacy did not differ by the total number of comorbidities or by the presence or absence of specific comorbidities. Similarly, for 3 conditions (17 trials) examining outcomes which were discrete events (e.g., thromboembolism, bleeding, headaches, and fractures), there was no evidence of treatment effect modification by comorbidity count or by specific comorbidities.
Several previous studies have reported findings on treatment effect modification in IPD meta-analyses and meta-analyses of reported subgroup effects. However, these have largely been confined to major cardiovascular disease trials (e.g., for showing similar efficacy of statin in people with and without diabetes [20], differential benefit of blood pressure lowering therapy in people with and without diabetes [21], or showing questionable net benefit of aspirin in primary prevention [22]) or to concordant conditions defined as those closely related to the index condition or target event for the trial (such as hypertension in stroke trials [23]).
Acknowledgments
This study, carried out under YODA Project # 2017–1746, used data obtained from the Yale University Open Data Access Project, which has an agreement with JANSSEN RESEARCH & DEVELOPMENT, L.L.C. The interpretation and reporting of research using this data are solely the responsibility of the authors and does not necessarily represent the official views of the Yale University Open Data Access Project or JANSSEN RESEARCH & DEVELOPMENT, L.L.C. This study was also carried out under ClinicalStudyDataRequest.com project number 1732, used data from the ClinicalStudyDataRequest.com repository, who provided data from Boehringer-Ingelheim, GSK, Lilly, Roche, Takeda, and Sanofi. The interpretation and reporting of research using these data are solely the responsibility of the authors and does not necessarily represent the official views of ClinicalStudyDataRequest.com or Boehringer-Ingelheim, GSK, Lilly, Roche, Takeda or Sanofi.
Citation: Hanlon P, Butterly EW, Shah AS, Hannigan LJ, Lewsey J, Mair FS, et al. (2023) Treatment effect modification due to comorbidity: Individual participant data meta-analyses of 120 randomised controlled trials. PLoS Med 20(6): e1004176. https://doi.org/10.1371/journal.pmed.1004176
Received: January 6, 2023; Accepted: April 12, 2023; Published: June 6, 2023
Copyright: © 2023 Hanlon et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Aggregated data and code required to run these models, along with full model descriptions, are available at https://doi.org/10.5281/zenodo.7713360. Individual participant data is available upon application to the trial sponsors (via https://www.clinicalstudydatarequest.com/ or https://yoda.yale.edu/), subject to a data transfer agreement.
Funding: This work was funded by the Wellcome Trust (grant number 201492/Z/16/Z, grant recipients DMA, SD) and the Medical Research Council (grant number MR/S021949/1, grant recipient PH). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: SD received fees from the Association of the British Pharmaceutical Industry (ABPI) for delivery of a Masterclass (unrelated to this work). The other authors have declared that no competing interests exist.
Abbreviations: ARR, absolute risk reduction; BASFI, Bath Ankylosing Spondylitis Functional Index; BASDI, Bath Ankylosing Spondylitis Disease Activity Index; CI,
credible interval; COMET, Core Outcome Measures in Effectiveness Trials; COPD, chronic obstructive pulmonary disease; CSDR, Clinical Study Data Request; DAPT, dual antiplatelet therapy; eGFR, estimated glomerular filtration rate; IPD, individual participant data; MBP, mid-blood pressure; MCID, minimum clinically important difference; MDRD, Modification of Diet in Renal Disease; MedDRA, Medical Dictionary for Regulatory Activities; RCT, randomised controlled trial; SGLT2, sodium-glucose co-transporter-2; YODA, Yale Open Data Access
https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1004176#ack
