Sponsorship Effect and Publication Bias

• In Event: Conflicts of Interest and Professional Misconduct in Health Care

Abstract

Introduction/Background:

Drug development is a long and costly process, and there is an extensive literature on the value of pharmaceutical innovation, sources of bias in randomized trials and economic evaluations, and the association between study funding and reported outcomes. The last 20 years have seen significant progress in the development of new cancer drugs. These developments have been accompanied by a decline in cancer-specific mortality rates. Considerable attention has been focused on the high cost of new cancer drugs, prompting calls for a stronger alignment between their prices and their clinical and economic value. The literature describes a link between pharmaceutical industry sponsorship and biased research outcomes; suggested explanatory factors may include publication bias and research design. 

Purpose/Research Question: 

In this paper, I show that industry-funded research leads to significantly biased estimates of the clinical effectiveness and economic value of new cancer drugs.

Methods:

I constructed a novel data set of cancer drug randomized clinical trials (RCTs) and cost-effectiveness analyses (CEAs). The identification strategy leverages observed variation in sponsorship, where the same drugs are assessed in RCTs and CEAs conducted by parties with different financial interests. Multiple sources of bias were controlled with dummy variables and their interactions; drug-level and study-level endogeneity were addressed using instrumental variables.

Results/Findings:

A cancer drug appears considerably more effective when the manufacturer is involved, compared to its evaluation within the same drug set without the influence of sponsorship bias; 28% more effective, 37% more likely to report a statistically significant favorable outcome, and 60% more likely to be the more effective option. There is strong evidence of publication bias for both RCTs and cost-effectiveness models. Conversely, cost-effectiveness analyses were almost universally found to be favorable to the study sponsor, and study design choices explain most of the sponsorship effect as modeling approaches are chosen to maximize the perceived value-for-money of the sponsored drug.

Author

• Matthew Chenoweth, University of California - Los Angeles
  Presenting Author