This analysis examines whether funding sources impact the effect size of clinical trials–differences between treatment and control groups. I examine differences between clinical trials that are solely funded by industry and those funded by other institutions such as the National Institute of Health (NIH), other U.S. federal organizations, research universities, etc. The topic of this study is smoking cessation and examines all clinical trials from clinicaltrials.gov that are interventional, as opposed to observational, and measure abstinence in smoking by number (percentage) of people abstaining, as opposed to cotinine levels or number of cigarettes.
I looked at 159 clinical trials that fit the above criteria and coded the percentage of people who abstained from smoking in a treatment scenario and those who abstained from smoking in a control, or placebo, scenario. I first examine whether the treatment scenario across all trials is significantly different than the control scenario by running a simple t-test–looking for differences in means. Figure 1, below, shows that the mean percentage of those abstaining from smoking for the treatment group is significantly different than the mean percentage of those abstaining from smoking for the control group. The magnitude of the difference is 6.66% and is significant at the 1% level, which means there is only a 1% chance the observed difference is due to chance.
|Figure 1. Differences in Means Between Those Abstaining from Smoking in a Treatment Scenario and Control Scenario|
|Treatment Mean||Control Mean||Difference||T-Statistic||Significant?|
|Percent abstaining from smoking||27.29%||20.63%||6.66%||6.3362||Yes|
Next, I code which clinical trials were exclusively funded by “industry.” Those trials funded solely by industry predominantly come from Pfizer. Arena Pharmaceuticals, GlaxoSmithKline, and Swedish Match AB are also industry funders of clinical trials. The average effect size–treatment minus control–of industry-funded clinical trials is 20.39%. In other words, on average, the intervention in industry-funded clinical trials produces a 20.39% increase in those abstaining from smoking. For comparison, the average effect size for those with “other” funding sources, or those not solely funded by industry, is 7.20%. (Note: I replaced effect sizes that were negative with 0. The rationale is that the treatment cannot have a negative impact on smoking abstinence. In other words, the treatment should not make people smoke more.)
In Figure 2, below, I run another t-test on whether industry-funded effect sizes are greater than other-funded effect sizes. The difference in effect sizes between industry-funded and other-funded is 13.19%. This difference in means is statistically significant at the 1% level. The t-test not only shows that the means between these two groups are different, it also shows that industry-funded clinical trials have a larger effect size than other-funded trials. This finding is significant at the 1% level as well.
|Figure 2. Industry-Funded Effect Sizes are Larger than Other-Funded Effect Sizes|
|Industry Mean||Other Mean||Difference||T-Statistic||Significant?|
|Effect size–treatment minus control||20.39%||7.20%||13.19%||3.9501||Yes|
What is peculiar is that when clinical trials are funded by a collaboration between industry and the NIH, or with the federal government, the effect size drops significantly. Figure 3, below, shows that the difference between clinical trials that are solely funded by industry and clinical trials funded by a collaboration between industry and others is significantly different at the 1% level. Clinical trials solely funded by industry have much larger effect sizes.
|Figure 3. Solely Industry-Funded Effect Sizes are Larger than Industry Collaboration Effect Sizes|
|Soley Industry Funded Mean||Industry Collaboration Mean||Difference||T-Statistic||Significant?|
|Effect size–treatment minus control||20.39%||5.94%||14.45%||3.33||Yes|
The takeaway is a question worth further investigating. Why does a clinical trial’s funding source impact its effect size? Perhaps different funding sources support different types of clinical trial interventions and that is why we see such great disparities. This is a point for further research.