Table 2. Subgroup and sensitivity analysis of the effect of mediator variables on the outcome of “endpoint symptomatic improvement”.
| Variables | Subjects (studies) | SMDs (95%CI) | I2 (%) | P |
|---|---|---|---|---|
| 1. Chinese studies | 535 (7) | -0.63 (-0.93, -0.33) | 64 | <0.0001 |
| Non-Chinese studies | 55 (2) | -0.92 (-1.99, 0.14) | 0 | 0.09 |
| 2. Double blind/rater-masked | 271 (5) | -0.74 (-1.16, -0.33) | 61 | 0.0005 |
| Non-blinded | 319 (4) | -0.59 (-1.04, -0.15) | 72 | 0.009 |
| 3. Trial duration < 12 weeks | 323 (4) | -0.95 (-1.20, -0.70) | 8 | <0.00001 |
| Trial duration ≥ 12weeks | 267 (5) | -0.46 (-0.82, -0.10) | 51 | 0.01 |
| 4. the number of ECTa: mean <9 sessions | 84 (1) | -0.81 (-1.25, -0.36) | NA | 0.0004 |
| mean ≥9 sessions | 481 (7) | -0.68 (-1.03, -0.33) | 70 | 0.0002 |
| 5. High quality (Jadad score ≥ 3) | 313 (6) | -0.63 (-1.03, -0.23) | 64 | 0.002 |
| Low quality (Jadad score < 3) | 277 (3) | -0.74 (-1.18, -0.31) | 66 | 0.0009 |
| 6. Co-starting with an antipsychotic | 237 (4) | -0.49 (-0.75, -0.23) | 1 | 0.0003 |
| Augmenting with an antipsychotic | 353 (5) | -0.83 (-1.26, -0.41) | 70 | 0.0001 |
aOnly 8 RCTs reported the number of ECT sessions. Bold values are p<0.05
CI: 95% confidence interval; ECT = electroconvulsive therapy; SMDs = standardized mean differences; NA = not applicable