Statistical significance did not affect time to publication in non-Cochrane systematic reviews: a meta-epidemiological study

OBECTIVES: To investigate time-lag bias based on statistical significance in findings in non-Cochrane systematic reviews (SRs) and explore barriers to publication for review authors STUDY DESIGN AND SETTING: We included SRs of randomized controlled trials with completed analyses for protocols registered in the international prospective register of systematic reviews (PROSPERO) by the end of 2014. To obtain unpublished information, we sent questionnaires regarding barriers to publication of SRs to all authors between May and November 2018 and examined the association between the statistical significance of SR findings and time to acceptance. RESULTS: Among the 241 SR authors contacted, 141 (58%) responded, of whom 103 met the eligibility criteria and agreed to participate. Ninety-three (90%) of the protocols had been published as journal articles and 11 (11%) remained unpublished. Statistical significance was not significantly associated with time to acceptance (adjusted hazard ratio 1.36, 95% confidence interval 0.84 to 2.19). The authors reported lack of time (52%), rejection by journals (42%), duplication of similar topics (42%), and nonsignificant results (29%) as barriers to publication of SRs. CONCLUSIONS: We found no convincing evidence of time-lag bias based on statistical significance in the findings among non-Cochrane SRs registered in the PROSPERO.