eTable 2. Additional ITT analysis for the primary outcome (limitation in social participation) using various imputation methods.
| Imputation method | N (IG/CG) | ΔT0−T1IG | ΔT0−T1CG | p value |
ES [95% CI] |
| Last observation carried forward*1 | 267/261 | 5.77 (18.90) | 2.44 (15.85) | 0.029 | 0.19 [0.02; 0.36] |
| Best/worst scenario*2 Worst/best scenario*3 |
267/261 | 7.91 (18.66) 6.40 (18.71) | 2.93 (15.81) 4.09 (16.03) | 0.001 0.128 | 0.29 [0.12; 0.46] 0.13 [−0.04; 0.30] |
| Multiple imputation*4 | 268/262 | 7.29 (22.86) | 2.98 (18.90) | 0.017 | 0.21 [0.04; 0.38] |
| No imputation (CCA) | 210/219 | 7.33 (21.05) | 2.92 (17.31) | 0.018 | 0.23 [0.04; 0.42] |
*1 Baseline values replace missing values in the 12-month follow-up history.
*2 Best/worst scenario: Missing cases in der IG replaced by mean change in the IG responders who underwent rehabilitation (“best”: improvement of 10 points); missing cases in the CG replaced by mean change in the CG responders who did not undergo rehabilitation (“worst“: improvement of 3 points)
*3 Worst/best scenario: Missing cases in the IG replaced by mean change in the CG responders who did not undergo rehabilitation (“worst“: improvement by 3 points); missing cases in the CG replaced by mean change of the IG responders who underwent rehabilitation (“best“: improvement of 10 points)
*4 Multiple imputation using the Markov chain Monte Carlo method (MCMC), 50 imputations, 10 iterations, in the model all outcomes as well as age, sex, school education, and disease duration
CCA, complete case analysis; ES, effect size; IG, intervention group; ITT analysis, intention to treat analysis; CG, control group; CI, confidence interval; T0, baseline; T1, follow-up; Δ difference