Mylène Jansen

280 Chapter 14 Statistical analysis This study is a post-hoc analysis on the data of the original RCTs. Potential differences in baseline characteristics between the 3 groups (KJD TKA , KJD HTO , and HTO) were analyzed using 1-way ANOVA with, in case of statistically significant differences, post-hoc Tukey HSD tests. In case of not normally distributed continuous variables, Kruskal-Wallis tests with post- hoc Dunn tests in case of statistically significant differences were used. For categorical variables, chi-square tests were used. Changes between pre- and 2 years post-treatment values for all cartilage thickness and JSW parameters were calculated using paired t -tests. Linear regression was used for comparisons in cartilage thickness and JSWchanges over 2 years between groups, correcting for any significantly different baseline patient characteristics. Consistency was tested by in- and excluding baseline cartilage thickness and JSW as confounder. The influence of baseline characteristics on the change in MAC cartilage thickness was tested using linear regression. As leg axis measurements were only available around half of the KJD TKA patients, this parameter was not used in linear regression models (except when specifically mentioned when testing the effect of leg axis). KJD and HTO patients were divided in subgroups based on the strongest predictor of MAC cartilage thickness change; the same statistical tests for changes over time and differences between groups were applied on these subgroups. Sensitivity analyses were performed by adding the trial in which patients were originally included as potential confounder. Absolute values are presented with mean and standard deviation (SD) while changes over time are presented as mean change and 95% confidence interval (95%CI). A p- value of <0.05 was considered statistically significant. As a measure for effect size of the primary and secondary outcome, Cohen’s d was used when comparing changes between different groups. Values of 0.20, 0.50 and 0.80 indicate small, moderate and large effect sizes, respectively. 35 Results Patients A flowchart of the final patient selection is shown in Figure 1. In the KJD HTO group, 1 patient was excluded before surgery due to inoperability, 2 patients received other surgery and 2 patients had MRI scans of insufficient quality, leaving 18 patients for analysis. In the HTO group, 1 patient was excluded before treatment due to anxiety, 4 patients were lost to follow- up because of comorbidities interfering with follow-up, 2 patients did not undergo MRI at 2 years, 2 patients had MRI scans with different hardware at baseline and 2 years, and 4 patients had MRI scans of insufficient quality, leaving 33 patients for analysis. In the KJD TKA group, 1 patient received a different surgery within 2 years of follow-up and 1 patient refused imaging at 2 years, leaving 18 patients for analysis.