Influence of sponsorship bias on treatment effect size estimates in randomized trials of oral health interventions: a meta-epidemiological study

Article type
Authors
Armijo-Olivo S1, Saltaji H2, Cummings G2, Amin M2, Da Costa B3, Flores-Mir C2
1University of Alberta; Institute of Health Economics
2University of Alberta
3University of Toronto
Abstract
Background: sponsorship bias arises due to potential inappropriate influence of funding on trial findings. Such biased trials can generate misleading information that can cause dental clinicians to make inappropriate clinical decisions, leading to ineffective treatment and poor outcomes. There is much debate in the literature about the impact and magnitude of sponsorship bias on treatment effect size estimates. Bias in industry-sponsored oral health trials can potentially benefit the sponsoring company and might lead to inappropriate treatment decisions

Objective: to quantify the extent of bias associated with sponsorship in oral health randomized controlled trials (RCTs).

Methods: we selected all oral health meta-analyses that included a minimum of five RCTs. We extracted data, in duplicate, related to the influence of the trial sponsor (sponsorship bias). We quantified the extent of bias associated with influence of sponsorship on the magnitude of effect size estimates using a two-level meta-meta-analytic approach with a random-effects model to allow for intra- and inter-meta-analysis heterogeneity.

Results: we identified 540 RCTs, included in 64 meta-analyses, analyzing 137,957 participants. The most common sources of sponsorship were industry (n = 156; 28.9%) followed by government (n = 43; 8.0%) and academic (n = 19; 3.5 %) sources. The influence of the trial sponsor was assessed as being unclear in 72.8% (n = 393) of the trials, while it was assessed as appropriate in 16.7% (n = 90) and as inappropriate in 10.6 (n = 57) of the trials. Treatment effect size estimates were 0.10 larger in trials with inappropriate sponsorship influence than in trials with appropriate sponsorship influence (95% confidence interval: 0.02 to 0.19; P = 0.017). A positive value (> 0) across meta-analyses indicated that a lack of appropriate sponsorship influence inflated the treatment effect size estimate.

Conclusions: we found significant differences in treatment effect size estimates between oral health trials based on lack of appropriate influence of funders. Treatment effect size estimates were 0.10 larger in trials with lack of appropriate influence of funders. Investigators of oral health systematic reviews should perform sensitivity analyses based on appropriateness of influence of sponsorship in included trials.

Patient or healthcare consumer involvement: it is hoped that the results help knowledge synthesis teams to assess and discuss sponsorship biases in the process of knowledge synthesis.