Improving Power and Sample Size Calculation in Rehabilitation Trial Reports: A Methodological Assessment.

Abstract

To systematically assess the reporting of sample size calculation in randomized controlled trials (RCTs) on rehabilitation interventions for mechanical low back pain. The Cochrane Database of Systematic Reviews was searched through February 2015. We conducted an electronic database search for RCTs published from January 1, 1968 to February 28, 2015 and included in the Cochrane Systematic Reviews. Two investigators independently used an ad hoc 6-item checklist derived from the Consolidated Standards of Reporting Trials (CONSORT) 2010 statement recommendations to extract data on sample size calculation. The primary outcome was the proportion of RCTs that reported sample size calculation; the secondary outcome was the completeness of sample size analysis reporting. We also evaluated improvement in reporting of sample size calculation over time. Sample size calculation was reported in 80 (36.0%) of the 222 eligible RCTs included in 14 Cochrane Systematic Reviews. Only 13 (16.3%) of these RCT reports gave a complete description, and about half reported ≥4 of the 6 elements of sample size calculation (median, 4; interquartile range, 3-5). Completeness of reporting of sample size calculation improved from 1968 to 2013; since 2005, the number of RCTs reporting sample size calculation has increased compared with the number of RCTs not reporting it. Despite improvement, reporting of sample size calculation and power analysis remains inadequate, limiting the reader's ability to assess the quality and accuracy of rehabilitation studies.