Outcome variables were assessed at baseline and at 3 and 18 months after randomization. Consistent with the protocol design, an intention-to-treat approach was employed. The incomplete cases were handled using the multiple imputation analysis by “mi impute pmm” procedure in Stata IC v12 statistical software. The estimates of the parameters for each imputed data set were combined using Rubin’s rules (26 ). Sensitivity analyses comparing complete cases versus multiple imputation analysis are shown in Supplementary Table 3 .
The change of each outcome during the follow-up was assessed by multilevel mixed-effects linear regression adjusted by age, gender, marital status, educational status, income, and baseline TGUG and MEC-35. Interaction terms between the intervention, gender, and age were also tested and none were statistically significant.
Standardized effect sizes were calculated as the mean difference between IG and CG, divided by the pooled standard deviation of CG and IG change (27 (link)). Estimates were reported with corresponding 95% confidence intervals, and statistical significance was established as a p value less than .05.
The change of each outcome during the follow-up was assessed by multilevel mixed-effects linear regression adjusted by age, gender, marital status, educational status, income, and baseline TGUG and MEC-35. Interaction terms between the intervention, gender, and age were also tested and none were statistically significant.
Standardized effect sizes were calculated as the mean difference between IG and CG, divided by the pooled standard deviation of CG and IG change (27 (link)). Estimates were reported with corresponding 95% confidence intervals, and statistical significance was established as a p value less than .05.