Table 5. Sensitivity analysis of each method of measure of association estimate between medication and placebo for the primary outcome (aggregate measure of mental-health-related symptoms).
| Method | o/k (n) | Estimated measure (95% CI) | SE | p-Value | τ2 | Heterogeneity I2 (%) |
|---|---|---|---|---|---|---|
| Only published | 432/124 (28,196) | −0.58 (−0.64 to −0.52) | 0.03 | <0.001 | 0.061 | 49.64 |
| SD imputation (maximum SD) | 469/135 (30,245) | −0.52 (−0.57 to −0.46) | 0.03 | <0.001 | 0.074 | 55.50 |
| No SD imputation | 425/121 (27,228) | −0.55 (−0.61 to −0.49) | 0.03 | <0.001 | 0.078 | 55.51 |
| Correlation of 0.5 | 469/135 (30,245) | −0.56 (−0.62 to −0.50) | 0.03 | <0.001 | 0.079 | 63.02 |
| Correlation of 0.7 | 469/135 (30,245) | −0.56 (−0.61 to −0.50) | 0.03 | <0.001 | 0.068 | 66.75 |
| Excluding outliers | 462/132 (29,955) | −0.55 (−0.60 to −0.49) | 0.03 | <0.001 | 0.063 | 51.19 |
| Endpoint standardized mean difference | 185/53 (8,256) | −0.43 (−0.50 to −0.36) | 0.04 | <0.001 | 0.047 | 58.20 |
| Only studies at low risk of bias | 179/38 (9,291) | −0.56 (−0.67 to −0.45) | 0.06 | <0.001 | 0.100 | 60.19 |
k, number of studies; n, sample size; o, number of outcomes; SE, standard error.