Financial conflicts of interest and estimated intervention effects in randomized clinical trials: a systematic review of meta-epidemiological studies

Article type
Authors
Hansen C1, Lundh A2, Jørgensen KJ3, Abdi SAIS4, Gelle MHA4, da Costa BR5, Christensen R6, Hróbjartsson A7
1Centre for Evidence-Based Medicine Odense (CEBMO) and Odense Patient data Explorative Network, Odense University Hospital, Department of Clinical Research, University of Southern Denmark, Nordic Cochrane Centre, Rigshospitalet
2Centre for Evidence-Based Medicine Odense (CEBMO) and Odense Patient data Explorative Network, Odense University Hospital; Department of Clinical Research, University of Southern Denmark; Department of Infectious Diseases, Hvidovre Hospital
3Nordic Cochrane Centre, Rigshospitalet
4Centre for Evidence-Based Medicine Odense (CEBMO), Odense University Hospital
5Applied Health Research Centre (AHRC), Li Ka Shing Knowledge Institute of St. Michael’s Hospital; Institute of Health Policy, Management and Evaluation, University of Toronto
6Musculoskeletal Statistics Unit, The Parker Institute, Bispebjerg and Frederiksberg Hospital; Research Unit of Rheumatology, Department of Clinical Research, University of Southern Denmark, Odense University Hospital
7Centre for Evidence-Based Medicine Odense (CEBMO) and Odense Patient data Explorative Network, Odense University Hospital; Department of Clinical Research, University of Southern Denmark
Abstract
Background: randomised trials are regarded as the most reliable method for estimating effects of healthcare interventions. However, conflicts of interest, such as industry funding or author ties, are a potential threat to their trustworthiness, as a Cochrane Review reported an association between industry funding and favourable trial conclusions.

It may be more relevant to investigate the impact of conflicts of interests on estimated intervention effects, as trial results and not conclusions are used in meta-analyses. Due to methodological heterogeneity, the Cochrane Review did not undertake a meta-analysis on estimated intervention effects. Moreover, some of the studies included in the Cochrane Review likely compared trials that differed in other aspects than conflicts of interest (e.g. investigating different drugs for different diseases). Such confounding may be minimized in a meta-epidemiological study, where trials are compared within meta-analyses.

Objectives: to investigate the association between financial conflicts of interest and estimated intervention effects in trials.

Methods: systematic review of meta-epidemiological studies that compare trials with and without financial conflicts of interest. We searched PubMed, Embase, and Cochrane Methodology Register. We screened studies included in a systematic review of meta-epidemiological studies, and studies included and excluded in two versions of a Cochrane Review on industry funding in trials. We searched reference lists of included studies, Web of Science, conference proceedings, and Google Scholar.

Two review authors independently included studies and extracted study characteristics. One review author extracted outcome data, and a second review author verified this.

We analysed the association between financial conflicts of interest and the magnitude of estimated intervention effects using random-effects meta-analysis. We analysed heterogeneity by comparing trials on drug and device interventions with trials on non-drug and non-device interventions. We tested the robustness of our findings by 1) analysing industry funding separately, 2) analysing study author financial conflicts of interest separately, and 3) excluding studies with undeclared financial conflicts of interest.

We assessed reporting bias by comparing protocols with included studies. Further, we drew a random sample of 20 meta-epidemiological studies not investigating financial conflicts of interest according to publications, and we asked the authors if they had extracted information on financial conflicts of interest.

Results: as of April 2019, we have included 10 meta-epidemiological studies (published between 2003 and 2018) with a total of 297 systematic reviews of 2867 studies. We will present meta-analysis results at the Cochrane Colloquium.

Conclusions: will be presented at the Cochrane Colloquium.

Patient involvement: none