Results of conventional (frequentist) and Bayesian meta-analyses of the association between impaired CYP2D6 function and disease-free survival among patients with breast cancer treated with tamoxifen, under a variety of study selection criteria
| Selection criterion | Meta-analysis type* | Summary RR (95% CI) |
| All studies | Conventional | 1.53 (1.25 to 1.87) |
| Bayesian, vague | 1.51 (1.23 to 1.90) | |
| Bayesian, informative | 1.49 (1.25 to 1.80) | |
| Caucasian/*4 | Conventional | 1.25 (1.06 to 1.49) |
| Bayesian, vague | 1.25 (1.05 to 1.50) | |
| Bayesian, informative | 1.28 (1.10 to 1.51) | |
| Asian/*10 | Conventional | 2.44 (1.48 to 4.03) |
| Bayesian, vague | 2.15 (1.25 to 3.69) | |
| Bayesian, informative | 1.67 (1.22 to 2.24) | |
| RR ≤2 | Conventional | 1.22 (1.05 to 1.41) |
| Bayesian, vague | 1.21 (1.01 to 1.42) | |
| Bayesian, informative | 1.25 (1.07 to 1.44) | |
| Tumour DNA | Conventional | 1.19 (0.94 to 1.51) |
| Bayesian, vague | 1.18 (0.88 to 1.55) | |
| Bayesian, informative | 1.26 (1.02 to 1.56) | |
| Non-neoplastic DNA | Conventional | 1.85 (1.40 to 2.46) |
| Bayesian, vague | 1.81 (1.33 to 2.51) | |
| Bayesian, informative | 1.65 (1.31 to 2.08) |
Random-effects models were used for the conventional analyses. Bayesian models used either vague or informative priors as described in the methods section.
RR, relative risk.