Article type
Year
Abstract
Background:
Randomised trials are a reliable method for estimating effects of health care interventions. However, various trial characteristics have been attributed to influencing trial effect estimates. This includes domains in the Cochrane risk of bias tool, single- or multicentre status, and trial funding source. One method for studying such relationships is a systematic review of meta-epidemiological studies.
In a meta-epidemiological study, the risk of confounding is minimised, because trials with and without a characteristic are contrasted within the same meta-analysis and thereby are similar in other important aspects. Several systematic reviews of meta-epidemiological studies have been undertaken to answer important methodological questions concerning clinical trials. However, one important challenge is that most analyses are primarily based on published data thereby limiting the analytical possibilities.
Objectives:
To describe our experiences in establishing the COMFIT Study Group and COMFIT Database, especially the willingness of authors to share trial-level data and challenges of establishing a combined database, with the intention of investigating the impact from commercial funding on trial effect estimates.
Methods:
Re-analysis of trial-level data from meta-epidemiological studies. We included meta-epidemiological studies with data on commercial funding. We searched for studies in PubMed, Embase, and Cochrane Methodology Register (inception to February 2020), reference lists of included studies, relevant systematic reviews, Web of Science, conference proceedings, and Google Scholar.
We contacted the authors of the included studies and asked them whether they would be interested in joining the COMFIT Study Group and share trial-level data. For each meta-epidemiological dataset, we checked data quality and excluded non-informative meta-analyses (i.e. meta-analyses where all trials had the same type of funding). We then constructed the COMFIT Database by merging data from the included studies and excluding duplicate meta-analyses and trials.
Results:
As of March 2020, we have included 16 meta-epidemiological studies with 671 meta-analyses and 6130 trials. We have obtaining or re-constructed trial-level data from 14 studies (89% of included meta-analyses and trials), and expect to receive data from a 15th study. Eight studies investigated a generic sample of trials, and eight investigated specific sub-samples (e.g. critical care trials). Seven studies investigated the impact from commercial funding, and nine investigated a different trial characteristic and adjusted for or included data on commercial funding. We will present the work on collecting trial-level data and constructing the COMFIT Database.
Conclusions:
This study is a comprehensive re-analysis of trial-level data from meta-epidemiological studies. Our experiences may be useful for reviewers undertaking systematic reviews. Additional conclusions will be presented at the Cochrane Colloquium.
Patient or healthcare consumer involvement:
None
Randomised trials are a reliable method for estimating effects of health care interventions. However, various trial characteristics have been attributed to influencing trial effect estimates. This includes domains in the Cochrane risk of bias tool, single- or multicentre status, and trial funding source. One method for studying such relationships is a systematic review of meta-epidemiological studies.
In a meta-epidemiological study, the risk of confounding is minimised, because trials with and without a characteristic are contrasted within the same meta-analysis and thereby are similar in other important aspects. Several systematic reviews of meta-epidemiological studies have been undertaken to answer important methodological questions concerning clinical trials. However, one important challenge is that most analyses are primarily based on published data thereby limiting the analytical possibilities.
Objectives:
To describe our experiences in establishing the COMFIT Study Group and COMFIT Database, especially the willingness of authors to share trial-level data and challenges of establishing a combined database, with the intention of investigating the impact from commercial funding on trial effect estimates.
Methods:
Re-analysis of trial-level data from meta-epidemiological studies. We included meta-epidemiological studies with data on commercial funding. We searched for studies in PubMed, Embase, and Cochrane Methodology Register (inception to February 2020), reference lists of included studies, relevant systematic reviews, Web of Science, conference proceedings, and Google Scholar.
We contacted the authors of the included studies and asked them whether they would be interested in joining the COMFIT Study Group and share trial-level data. For each meta-epidemiological dataset, we checked data quality and excluded non-informative meta-analyses (i.e. meta-analyses where all trials had the same type of funding). We then constructed the COMFIT Database by merging data from the included studies and excluding duplicate meta-analyses and trials.
Results:
As of March 2020, we have included 16 meta-epidemiological studies with 671 meta-analyses and 6130 trials. We have obtaining or re-constructed trial-level data from 14 studies (89% of included meta-analyses and trials), and expect to receive data from a 15th study. Eight studies investigated a generic sample of trials, and eight investigated specific sub-samples (e.g. critical care trials). Seven studies investigated the impact from commercial funding, and nine investigated a different trial characteristic and adjusted for or included data on commercial funding. We will present the work on collecting trial-level data and constructing the COMFIT Database.
Conclusions:
This study is a comprehensive re-analysis of trial-level data from meta-epidemiological studies. Our experiences may be useful for reviewers undertaking systematic reviews. Additional conclusions will be presented at the Cochrane Colloquium.
Patient or healthcare consumer involvement:
None