Article type
Year
Abstract
Background: Single-centre trials might differ from multicentre trials in characteristics of included patients, implemented interventions, methodological quality and sample size, which may result in different estimated treatment benefits.Objectives: We examined whether estimated treatment benefits differ between single-centre and multicentre osteoarthritis trials and whether this can be explained by components of methodological quality and sample size.
Methods: We performed a meta-epidemiological study of 13 meta-analyses with 154 trials that compared therapeutic interventions with placebo or non-intervention control in patients with hip or knee osteoarthritis. We calculated standardised mean differences (SMDs) from the differences in means of pain intensity between groups at the end of follow-up divided by the pooled SD and compared SMDs between trials with and without multiple centres. We used stratification by components of methodological quality and sample size to derive differences between single and multicenter trials adjusted for these characteristics.
Results: The 80 trials with multiple centres did not differ clearly from the 74 trials without multiple centres in their methodological quality, but randomised more patients (P = 0.027). On average, multicentre trials showed less beneficial effects compared to single-centre trials (difference in SMDs 0.20, 95% CI 0.38 to 0.02, P = 0.028) with small to moderate heterogeneity between meta-analyses (tau2 = 0.05, Pfor heterogeneity 0.002). Differences in treatment benefits were robust when analyses were adjusted for adequate generation of random sequence (-0.17, 95% CI 0.36 to 0.02, P = 0.07), adequate allocation concealment (-0.17, 0.33 to 0.01, P = 0.034) and blinding of patients (-0.19, 0.34 to 0.02, P = 0.029); were attenuated when adjusted for intention-to-treat analyses (-0.13, 0.27 to 0.01, P = 0.08); and disappeared when adjusted for sample size (-0.07, 95% CI 0.27 to 0.14, P = 0.51).
Conclusions: There is an overestimation of treatment benefits in single-centre osteoarthritis trials. This overestimation is not evident after adjustment for sample size.
Methods: We performed a meta-epidemiological study of 13 meta-analyses with 154 trials that compared therapeutic interventions with placebo or non-intervention control in patients with hip or knee osteoarthritis. We calculated standardised mean differences (SMDs) from the differences in means of pain intensity between groups at the end of follow-up divided by the pooled SD and compared SMDs between trials with and without multiple centres. We used stratification by components of methodological quality and sample size to derive differences between single and multicenter trials adjusted for these characteristics.
Results: The 80 trials with multiple centres did not differ clearly from the 74 trials without multiple centres in their methodological quality, but randomised more patients (P = 0.027). On average, multicentre trials showed less beneficial effects compared to single-centre trials (difference in SMDs 0.20, 95% CI 0.38 to 0.02, P = 0.028) with small to moderate heterogeneity between meta-analyses (tau2 = 0.05, Pfor heterogeneity 0.002). Differences in treatment benefits were robust when analyses were adjusted for adequate generation of random sequence (-0.17, 95% CI 0.36 to 0.02, P = 0.07), adequate allocation concealment (-0.17, 0.33 to 0.01, P = 0.034) and blinding of patients (-0.19, 0.34 to 0.02, P = 0.029); were attenuated when adjusted for intention-to-treat analyses (-0.13, 0.27 to 0.01, P = 0.08); and disappeared when adjusted for sample size (-0.07, 95% CI 0.27 to 0.14, P = 0.51).
Conclusions: There is an overestimation of treatment benefits in single-centre osteoarthritis trials. This overestimation is not evident after adjustment for sample size.