Computational protocol: A multicentre randomised, 1-year comparative effectiveness, parallel-group trial protocol of a physical therapy approach compared to corticosteroid injections

Similar protocols

Protocol publication

[…] A total of 69 participants per group, or a total of 138 participants, are needed for 80% power to detect a significant group × time interaction effect using an α level of 0.05 and assuming a 12% difference between groups in mean post-treatment total WOMAC scores—corresponding to a partial η2 of 0.006 and an effect size f of 0.078. This assumes two groups measured over five points in time, a common SD of 46.8 points, a between-repeated-measures correlation of r=0.681, and a non-sphericity correction factor of (Greenhouse-Geisser) epsilon=0.890—consistent with data collected in previous trials. Assuming an 11% dropout rate, we will need to enrol 156 (approximately 80 from each of the 2 participating centres) participants in order to have 138 participants completing the study. A sensitivity analysis was performed to compute magnitudes of statistical power reduction given varying scenarios with dropout rates greater than the assumed rate of 11%. If the dropout rate should be 15% (132 participants retained), with all other power determinants remaining unchanged, statistical power would be reduced to 78%; with a 20% dropout rate (125 participants retained), statistical power would be 75%; with a 25% dropout rate (117 participants retained), statistical power would be 72%. Sample size estimation was performed with G*Power software, V.3.1.2. Recruitment will not go over a total of 156 participants without prior Institutional Review Board (IRB) approval, but a site may have more than their proportionate share depending on clinical flow and available participant pool. In this case, the other site will under-recruit. [...] Data analysis will be performed with IBM SPSS Statistics software, V.22. The statistician performing the analyses will be blinded to the intervention associated with the group of assignment. Descriptive statistics will be computed to characterise and compare the two treatment groups for assessment of baseline heterogeneity. Distributions of measured variables will be examined visually with frequency histograms and formally assessed with Kolmogorov-Smirnov and Shapiro-Wilk statistics to test the normality assumption. Levene's test will be used to assess for violations of the homogeneity of variance assumption. All statistical tests will all be performed at an α level of 0.05.The primary analysis of relative effectiveness between the two treatments will be tested using a linear mixed-effects model which is flexible in accommodating data assumed to be missing at random. There will be two levels of treatment and five levels of time if randomisation produces reasonably equivalent groups as measured by important prognostic factors (Kellgren-Lawrence scores, duration of symptoms, baseline WOMAC scores, etc). Otherwise, variables for which non-equivalence is detected at baseline will be entered as covariates into the linear mixed-effects model. A similar approach will be used to analyse the GROC data. An α level of 0.05 will be used to establish significance. The primary analysis will include data from patients according to the group they were assigned. If one treatment is shown to be superior to the other, supplemental analyses will be performed by dichotomising groups based on MCIDs of 12% for WOMAC and +3 points for GROC scores. This will allow computation of absolute risk reduction, relative risk reduction and number needed to treat (with associated 95% CIs) using failure to obtain clinically meaningful benefit as the event of interest. Every effort will be made to recruit 156 participants. No interim analysis or alternative analysis is planned. If factors beyond the control of the investigators, such as military reassignment of members of the research team, or other emerging research studies at the same centres on the same population prolong enrolment beyond a reasonable period, enrolment may be terminated and the specified blind analysis performed at that point. Owing to the low risk associated with the interventions, a data monitoring committee will not be used. […]

Pipeline specifications

Software tools G*Power, SPSS
Application Miscellaneous
Organisms Homo sapiens, Daphne virus S
Diseases Osteoarthritis, Osteoarthritis, Knee