Table 4.
Risk of AMI in rosiglitazone users compared with metformin users, overall and subgroup analyses
| Study (author, year) | Overall RR (95 % CI) | Subgroup analyses | ||
|---|---|---|---|---|
| Overall sensitivity analysisa | New users | |||
| Sensitivity analysisa | ||||
| Loebstein, 2011 [35] | Not estimable | Not estimable | Not estimable | Not estimable |
| Walker, 2008 [39] | 1.05 (0.67–1.66) | |||
| Tzoulaki, 2009 [38] | 1.10 (0.87–1.39) | Not reported | Not reported | |
| Dormuth, 2009 [28] | 1.14 (0.90–1.44) | |||
| McAfee, 2007 [36] | 1.19 (0.84–1.68) | |||
| Hsiao, 2009 [32] | 2.09 (1.36–3.24) | Not Included | Not included | |
| Brownstein, 2010 [27] | 2.51 (1.98–3.17) | Not Included | Not reported | Not reported |
| Fixed-effects, sRR (95 % CI) | 1.44 (1.28–1.61) | 1.13 (0.98–1.30) | 1.24 (1.05–1.47) | 1.14 (0.95–1.36) |
| Random-effects, sRR (95 % CI) | 1.42 (1.03–1.98) | 1.13 (0.98–1.30) | 1.29 (0.99–1.67) | 1.14 (0.95–1.36) |
| Hetrogeneity statistics |
τ
2 = 0.14 χ 2 = 36.07, df = 5 (P < 0.00001) I 2 = 86 % |
τ
2 = 0.00 χ 2 = 0.21, df = 3 (P = 0.98) I 2 = 0 % |
τ
2 = 0.04 χ 2 = 6.62, df = 3; (P = 0.09) I 2 = 55 % |
τ
2 = 0.00 χ 2 = 0.17, df = 2; (P = 0.92) I 2 = 0 % |
df degrees of freedom, RR relative risk, sRR summary relative risk
a Sensitivity analysis: we excluded those studies with a combined high or unclear risk of bias for more than 30 % of the items in the RTI item bank