In the current race to find a cure for COVID-19, the issue of the integrity of clinical trials is more salient than ever. Skeptics cite several reasons to be concerned.

First, given the astronomical cost of bringing drugs to market and the lure of blockbuster profits, as well as the fact that statistical significance in clinical trials is a key prerequisite for marketing approval, investigators could theoretically face pressure from the pharmaceutical company sponsoring a trial to relax their standards of ethical and scientific integrity.

A second reason for concern is data manipulation by researchers, or so-called "p-hacking," in which several statistical analyses are conducted and those that produce significant results are selectively reported. Previous studies looking at statistical reporting in scientific journals across a number of disciplines have detected a suspiciously high concentration of results immediately above the significance threshold (typically p < 0.05).

Differences in the results reported for phase II trials (which establish a drug's efficacy) and subsequent phase III trials (which are larger and confirm the drug's safety and efficacy) were examined by researchers from Bocconi University in Milan and the University of Zurich. To investigate the integrity of registered clinical trials and look for evidence of p-hacking or other bias, the researchers analyzed the results reported in the largest repository of trials in the world, ClinicalTrials.gov.

The data indicated minimal manipulation of clinical trial results and were therefore largely reassuring. However, the researchers did find that phase III trials sponsored by smaller pharmaceutical companies (those outside the top 10) appeared to produce results that were more positive than should be expected given the corresponding phase II results, raising some alarm bells.

They also observed that results just above the significance threshold were slightly more frequent than expected, but the discrepancies were minor compared to those recorded in other fields such as the social and life sciences.

"Even in the case of phase III trials by small companies, we do not observe the spike of results just above the significance threshold," said Marco Ottaviani, PhD, a co-author of the paper and professor of economics at Bocconi University, in a statement. "We think that the registration process is key for transparency."

In other words, clinical trial registration -- the practice of documenting clinical trials before they are performed in the interest of transparency -- appears to be effective at preventing p-hacking and other biases.

The authors found that statistically significant results were more likely to be found in phase III than phase II trials, which is not surprising given that typically the statistical significance of phase II is confirmed in phase III and trials that fall below the significance threshold in phase II are suspended.

For large pharmaceutical companies, the authors found that the predicted share of trial results above the significance threshold in phase III is 65%, and the actual share was 68%. In the case of small companies, which tend to bring less promising drug candidates into phase III trials, the figures are 57% and 76%, raising questions about why their phase III results were more positive than expected.

Therefore, the results suggest that regulators may want to pay closer attention to smaller industry sponsors, some of whom may not feel the pressure of reputational consequences.