The effect of study sponsorship on a systematically evaluated body of evidence of head-to-head trials was modest: secondary analysis of a systematic review

Better data presentation in graphs and tables is possible and needed

“…In this issue again, in connection to earlier published work, [3,4] the phenomenon of sponsorship bias and how to study it evokes discussion. One of the key questions is: what is the most appropriate comparator in clinical trials to make a fair and clinically relevant evaluation of the added value of an intervention.…”

Study designs to detect sponsorship and other biases in systematic reviews

“…In a review of industry-sponsored trials of selective serotonin reuptake inhibitors (SSRIs) for severe depression, Gartlehner et al [1] report that treatments manufactured by the trial sponsor tend to perform better with an average relative risk (RR) of 1.07 (95% confidence interval [CI]: 1.02, 1.12) although this small effect is unlikely to be clinically relevant. Much of the previous literature on ''sponsorship'' or ''industry'' bias, including the very large review by Bekelman et al [2], is hard to interpret because it is based on or dominated by placebo-controlled trials.…”

“…They reported no clear evidence of sponsorship bias. We applied the same method to the data in [1] and found a RR bias of 1.02 in favor of the nonsponsored drug, that is, the mean RR of response on a nonsponsored drug is exaggerated on average by 2% (Bayesian 95% credible interval [CrI] 5 0.94,1.11). This alternative analysis design that could, in principle, incorporate placebo-controlled trials as well makes fewer assumptions about the true treatment effects and leads us to conclude that there is no evidence for sponsorship bias in this network of trials, not even a small one.…”

Pharmaceutical company–sponsored drug trials: the system is broken

“…

Our response to Paul Shekelle's commentary is more an extension of his thoughts rather than a rebuttal, because we whole-heartedly agree with the points he has raised.Our study was set out to determine the magnitude of bias in a systematically evaluated body of evidence of head-tohead trials [1]. By using data from an existing Agency for Healthcare Research and Qualityesponsored comparative effectiveness review, we tried to simulate a real-world situation in which, even with considerable effort, authors were unable to detect more than one unpublished trial.

…”