The PEDro scale had acceptably high convergent validity, construct validity, and interrater reliability in evaluating methodological quality of pharmaceutical trials

摘要

Abstract Background and Objective The Physiotherapy Evidence Database (PEDro) scale has been widely used to investigate methodological quality in physiotherapy randomized controlled trials; however, its validity has not been tested for pharmaceutical trials. The aim of this study was to investigate the validity and interrater reliability of the PEDro scale for pharmaceutical trials. The reliability was also examined for the Cochrane Back and Neck (CBN) Group risk of bias tool. Methods This is a secondary analysis of data from a previous study. We considered randomized placebo controlled trials evaluating any pain medication for chronic spinal pain or osteoarthritis. Convergent validity was evaluated by correlating the PEDro score with the summary score of the CBN risk of bias tool. The construct validity was tested using a linear regression analysis to determine the degree to which the total PEDro score is associated with treatment effect sizes, journal impact factor, and the summary score for the CBN risk of bias tool. The interrater reliability was estimated using the Prevalence and Bias Adjusted Kappa coefficient and 95% confidence interval (CI) for the PEDro scale and CBN risk of bias tool. Results Fifty-three trials were included, with 91 treatment effect sizes included in the analyses. The correlation between PEDro scale and CBN risk of bias tool was 0.83 (95% CI 0.76–0.88) after adjusting for reliability, indicating strong convergence. The PEDro score was inversely associated with effect sizes, significantly associated with the summary score for the CBN risk of bias tool, and not associated with the journal impact factor. The interrater reliability for each item of the PEDro scale and CBN risk of bias tool was at least substantial for most items (0.60). The intraclass correlation coefficient for the PEDro score was 0.80 (95% CI 0.68–0.88), and for the CBN, risk of bias tool was 0.81 (95% CI 0.69–0.88). Conclusion There was evidence for the convergent and construct validity for the PEDro scale when used to evaluate methodological quality of pharmacological trials. Both risk of bias tools have acceptably high interrater reliability.