Results from two sets of meta-analyses using a range of different methods. An OR <1 or an arcsine difference <0 in the X versus Y comparison indicates that treatment X is safer (lower risk of death)
| Method used | LAI-AP versus placebo | LAI-AP versus OAP |
| Inverse-variance OR with 0.5 continuity correction (95% CI) | 0.63 (0.24 to 1.70) | 0.71 (0.38 to 1.34) |
| Inverse-variance OR with treatment-arm continuity correction (95% CI) | 0.70 (0.26 to 1.90) | 0.71 (0.38 to 1.34) |
| Peto (95% CI) | 0.75 (0.24 to 2.36) | 0.67 (0.35 to 1.25) |
| Mantel-Haenszel OR with no continuity correction (95% CI) | 0.77 (0.26 to 2.28) | 0.66 (0.34 to 1.26) |
| Mantel-Haenszel OR with treatment-arm continuity correction (95% CI) | 0.85 (0.36 to 2.00) | 0.71 (0.39 to 1.28) |
| Bayesian meta-analysis, fixed effects OR (95% CrI) | 0.72 (0.22, 2.27) | 0.64 (0.33, 1.23) |
| Bayesian meta-analysis, random effects OR (95% CrI) | 0.71 (0.22, 2.31) | 0.63 (0.32, 1.24) |
| Beta-binomial model, OR (95% CI) | – | 0.63 (0.29 to 1.34) |
| Arcsine difference, fixed effects (95% CI) | −0.007 (−0.034 to 0.021) | −0.019 (−0.041 to 0.003) |
CrI, credible interval; LAI-AP, long-acting injectable antipsychotics; OAP, oral antipsychotics.