Threats to reliability risk erroneous conclusions: a survey of prospective registration and sample sizes of randomised trials in women's health.

Abstract

Prospective registration of clinical trials has been required since 2005. We aimed to assess concordance between registered and reported sample sizes among prospectively registered randomised controlled trials (RCTs) in obstetrics and gynaecology. Analysis of prospectively registered and published data. Women's health. Obstetrics and gynaecology RCTs published in eight journals in 2015. Specialist (Acta Obstet Gynecol Scand, BJOG, Obstet Gynecol and Am J Obstet Gynecol) and general (BMJ, N Engl J Med, JAMA and Lancet) journals were searched from 1 January to 31 December 2015 for main reports of obstetrics and gynaecology RCTs. Their corresponding registries and protocols were sought and data were extracted. Proportions and 95% CI were calculated using exact methods. Prospective registration of RCT and sample size concordance between RCTs and their registries within those where registration took place before patients were recruited. Of the 75 relevant RCTs, 51 (68%, 95% CI 56-78%, P<0.001 using a null hypothesis requiring 100% compliance) were prospectively registered, a feature found to be more common in general journals versus specialist journals (21/22 versus 30/53, 95% versus 57%, P=0.001). Of the 51 prospectively registered RCTs, 31 (61%, 95% CI 46-74%, P=0.003 using a null hypothesis requiring 40% of studies to achieve their stated sample size) did not reach the target sample size. There are gaps in universal adoption of the prospective trial registration rule. Inability to meet target sample size risks deficiencies in statistical power with unreliability in results. Only two-thirds of RCTs in women's health are prospectively registered, and over half fail to achieve target sample sizes.