Statistical reporting in randomized controlled trials from the dermatology literature: a review of 44 dermatology journals

Abstract

Background: The validity of randomized controlled trials (RCTs) is determined by several statistical factors.

Objectives: To determine the level of recent statistical reporting in RCTs from the dermatology literature.

Methods: We searched MEDLINE for all RCTs published between 1 May 2013 and 1 May 2014 in 44 dermatology journals.

Results: Two hundred and ten articles were screened, of which 181 RCTs from 27 journals were reviewed. Primary study outcomes were met in 122 (67.4%) studies. Sample size calculations and beta values were reported in 52 (28.7%) and 48 (26.5%) studies, respectively, and nonsignificant findings were supported in only 31 (17.1%). Alpha values were reported in 131 (72.4%) of studies with 45 (24.9%) having two-sided P-values, although adjustment for multiple statistical tests was performed in only 16 (9.9% of studies with ≥ two statistical tests performed). Sample size calculations were performed based on a single outcome in 44 (86.3%) and multiple outcomes in six (11.8%) studies. However, among studies that were powered for a single primary outcome, 20 (45.5%) made conclusions based on multiple primary outcomes. Twenty-one (41.2%) studies relied on secondary/unspecified outcomes. There were no differences for primary outcome being met (Chi-square, P = 0.29), sample size calculations (P ≥ 0.55), beta values (P = 0.89), alpha values (P = 0.65), correction for multiple statistical testing (P = 0.59), two-sided alpha (P = 0.64), support of nonsignificant findings (Fisher's exact, P = 0.23) based on the journal's impact factor.

Conclusions: Levels of statistical reporting are low in RCTs from the dermatology literature. Future work is needed to improve these levels of reporting.