Trends of Publication of Negative Trials Over Time

Studies with negative results are less likely to be published than others, potentially leading to publication bias. Introduced in 2000, trial registration could have participated in decreasing the proportion of unpublished studies. We assessed the proportion of negative randomized controlled trials (RCT) over the last 20 years. We searched Medline for RCT published in 2000, 2005, 2010, 2015, and 2020 in the British Medical Journal, the Journal of the American Medical Association, the Lancet, and the New England Journal of Medicine. The primary endpoint was the proportion of negative (final comparison on the primary study-endpoint without statistical significance or favoring the control arm) studies published in 2000 and 2020. Factors independently associated with the publication of negative studies were identified using multivariable analysis. A total of 1,542 studies were included. The proportion of negative RCT significantly increased between 2000 and 2020 (from 27.6% to 37.4%; P = 0.01), however, the trend over time was not significant (P = 0.203). In multivariable analysis, the following factors were associated with a higher proportion of published negative studies: superiority (P < 0.001), two-group trials (P < 0.001), number of patients ≥510 (P < 0.001), cardiology trials (P = 0.003), emergency/critical care trials (P < 0.001), obstetrics trials (P = 0.032), surgery trials (P = 0.006), pneumology trials (P = 0.029). Exclusive industry funding was associated with a lower proportion of published negative studies (P < 0.001). The proportion of published negative studies in 2020 was higher only when compared to 2000. During the two decades, no trend was noticeable. There is no clear relationship between trial registration and the publication of negative results over time.