Use of 95% confidence intervals in the reporting of between-group differences in randomized controlled trials: analysis of a representative sample of 200 physical therapy trials

Abstract

Objectives: To assess the prevalence of the use of 95% confidence intervals in the reporting of between-group differences in randomized controlled trials of physical therapy interventions and to determine if the prevalence is changing over time. Methods: Observational study, including an analysis of 200 trials from the Physiotherapy Evidence Database: 50 from each of the years 1986, 1996, 2006, and 2016. The primary outcome used was the prevalence of the between-group difference presented with 95% confidence intervals. We also extracted trial characteristics for descriptive purposes (i.e., number of participants, number of sites involved in recruitment, country(ies) of data collection, funding, subdiscipline of physical therapy, publication language and total Physiotherapy Evidence Database score). Results: Most commonly, the trials were published in English (89%) and classified in the musculoskeletal subdiscipline (23%). The overall prevalence of use of confidence intervals was 29% and there was a consistent increase in reporting between 1986 and 2016, with peak usage in the 2016 cohort (42%). Confidence intervals were more likely to be used in trials that had received funding, were conducted in Europe and Oceania, and in trials with a Physiotherapy Evidence Database score of at least 6/10. Conclusions: Most trials of physical therapy interventions do not report confidence intervals around between-group differences. However, use of confidence intervals is increasing steadily, especially among high-quality trials. Physical therapists must understand confidence intervals so that they can understand a growing number of trials in physical therapy. (C) 2018 Associacao Brasileira de Pesquisa e Pos-Graduacao em Fisioterapia. Published by Elsevier Editora Ltda. All rights reserved.