Objective: We sought to systematically ascertain the quality of randomized controlled trials (RCTs) in diabetes.
Research design and methods: We identified the 10 most recently published trials as of 31 October 2003 in each of six general medical, five diabetes, and five metabolism and nutrition journals and further enriched our sample with 10 additional RCTs from each of five journals that published the most eligible RCTs in a year. We explored the association between trial characteristics and reporting quality using univariate analyses and a preplanned multivariate regression model.
Results: After excluding redundant reports of included trials and one trial that measured outcomes on the health system and not on patients, we included 199 RCTs: 119 assessed physiological and other laboratory outcomes, 42 assessed patient-important outcomes (e.g., morbidity and mortality, quality of life), and 38 assessed surrogate outcomes (e.g., disease progression or regression, HbA(1c), cholesterol). Fifty-three percent were of low methodological quality, as were one-third (36-40%) of trials reporting patient-important or surrogate outcomes and two-thirds (64%) of laboratory investigations. Independent predictors of low quality were nonprofit funding source (odds ratio 3.1 [95% CI 1.5-6.2]), measure of physiological and laboratory outcomes (2.3 [1.2-4.4]), and cross-over design (2.3 [1.1-4.8]), all characteristics of laboratory clinical investigations.
Conclusions: There is ample room for improving the quality of diabetes trials. To enhance the practice of evidence-based diabetes care, trialists need to pay closer attention to the rigorous implementation and reporting of important methodological safeguards against bias in randomized trials.