Objectives After a negative study, substudies may emerge providing a positive spin on the hypothesis, based on subgroup or secondary outcome analyses. This approach suffers from multiplicity problem, imprecision and higher risk of type I error.

To describe the incidence of substudies testing the same hypothesis over a period of 5 years after publication of a primarily negative randomized clinical trial; to access the most frequent subanalysis method, how explicit the exploratory nature of the analysis were and the rate of successful positivation of results.

Method We searched all negative randomized trials published in the New England Journal of Medicine (NEJM) during the year 2014. Then, we made a highly sensible PUBMED search over the following 5 years in order to detect subsequent articles testing the same hypothesis in the same dataset throughout secondary analysis. Methods utilized, lack of recognition as secondary analysis was described and rate of positivation were described.

Results During the year 2014, 46 negative randomized clinical trials were published in the NEJM. Over five years, 14 of those articles had subsequent publication of subanalysis in PUBMED indexed journals, an incidence of 30% (95% CI = 19% - 45%). The most frequent method was secondary endpoints analysis (67% of substudies) and the remaining made use of subgroup analysis. Half of those analysis were not defined a priori (post-hoc analysis). Thirty-two percent of studies did not leave explicit in the conclusion it was an exploratory analysis and the success rate of positivation was 91% (95% CI = 80% - 97%).

Conclusions Subsequent publication of subanalysis from originally negative trials is frequent, commonly not defined a priori, commonly not explicit about the exploratory nature and highly successful in positivating results. This suggests lack of ecosystem scientific integrity.