Table 3.
Rosiglitazone meta-analyses for composite outcome
| Method (fixed or random effects), and data sources | Single zero event trials | Zero total event trials | Continuity correction | Effect estimate (95% CI) | P value | No of trials |
|---|---|---|---|---|---|---|
| Peto (fixed) | ||||||
| IPD only | Included | Excluded | None | OR 1.40 (1.16 to 1.69) | 0.000 | 31 |
| IPD+RECORD | Included | Excluded | OR 1.20 (1.06 to 1.36) | 0.004 | 32 | |
| Mantel-Haenszel (fixed) | ||||||
| IPD only | Included | Excluded | Constant continuity correction of 0.5 | OR 1.39 (1.15 to 1.68) | 0.001 | 31 |
| RR 1.37 (1.14 to 1.64) | 0.001 | 31 | ||||
| IPD+RECORD | Included | Excluded | OR 1.20 (1.06 to 1.36) | 0.005 | 32 | |
| RR 1.17 (1.05 to 1.31) | 0.005 | 32 | ||||
| IPD only | Included | Included | OR 1.39 (1.15 to 1.68) | 0.001 | 33 | |
| RR 1.36 (1.14 to 1.63) | 0.001 | 33 | ||||
| IPD+RECORD | Included | Included | OR 1.20 (1.06 to 1.36) | 0.005 | 34 | |
| RR 1.17 (1.05 to 1.31) | 0.005 | 34 | ||||
| Dersimonian and Laird (random) | ||||||
| IPD only | Included | Excluded | Constant continuity correction of 0.5 | OR 1.33 (1.09 to 1.61) | 0.005 | 31 |
| RR 1.30 (1.08 to 1.56) | 0.005 | 31 | ||||
| IPD+RECORD | Included | Excluded | OR 1.17 (1.03 to 1.33) | 0.02 | 32 | |
| RR 1.14 (1.02 to 1.28) | 0.02 | 32 | ||||
| IPD only | Included | Included | OR 1.33 (1.09 to 1.61) | 0.005 | 33 | |
| RR 1.30 (1.08 to 1.56) | 0.005 | 33 | ||||
| IPD+RECORD | Included | Included | OR 1.17 (1.03 to 1.33) | 0.02 | 34 | |
| RR 1.14 (1.02 to 1.28) | 0.02 | 34 | ||||
| Mantel-Haenszel (fixed) | ||||||
| IPD only | Included | Excluded | Treatment arm correction | OR 1.41 (1.16 to 1.70) | 0.001 | 31 |
| RR 1.38 (1.15 to 1.65) | 0.001 | 31 | ||||
| IPD+RECORD | Included | Excluded | OR 1.20 (1.06 to 1.36) | 0.004 | 32 | |
| RR 1.18 (1.05 to 1.32) | 0.004 | 32 | ||||
| IPD only | Included | Included | OR 1.40 (1.16 to 1.70) | 0.001 | 33 | |
| RR 1.38 (1.15 to 1.65) | 0.001 | 33 | ||||
| IPD+RECORD | Included | Included | OR 1.20 (1.06 to 1.36) | 0.004 | 34 | |
| RR 1.18 (1.05 to 1.32) | 0.004 | 34 | ||||
| Dersimonian and Laird (random) | ||||||
| IPD only | Included | Excluded | Treatment arm correction | OR 1.33 (1.09 to 1.61) | 0.005 | 31 |
| RR 1.30 (1.08 to 1.56) | 0.005 | 31 | ||||
| IPD+RECORD | Included | Excluded | OR 1.17 (1.03 to 1.33) | 0.02 | 32 | |
| RR 1.14 (1.02 to 1.28) | 0.20 | 32 | ||||
| IPD only | Included | Included | OR 1.33 (1.09 to 1.61) | 0.005 | 33 | |
| RR 1.30 (1.08 to 1.56) | 0.005 | 33 | ||||
| IPD+RECORD | Included | Included | OR 1.17 (1.03 to 1.33) | 0.02 | 34 | |
| RR 1.14 (1.02 to 1.28) | 0.02 | 34 | ||||
IPD=individual patient level data; OR=odds ratio; RECORD=Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes study; RR=relative risk.
For all analyses, numbers of composite events observed among the total population of patients were as follows: IPD only, rosiglitazone population: 274 events among 11 837 patients; IPD only, control population: 219 events among 9319 patients; RECORD, rosiglitazone population: 333 events among 2226 patients; RECORD, control population: 316 events among 2232 patients.