Association Between CYP2D6 Polymorphisms and Breast Cancer Outcomes

To the Editor: In their cohort study, Dr Schroth and colleagues¹ reported an association between CYP2D6 polymorphisms and breast cancer outcomes among women treated with tamoxifen. Although the authors described their study as adequately powered, it appears that their power calculations did not account for the substantial overlap between the outcomes reported in this study and the outcomes described in 2 earlier reports.^{2 - 3} This overlapping sample violates the independent trials assumption that underlies all frequentist statistics, including power calculations.

Moreover, the authors did not report the association between CYP2D6 polymorphisms and breast cancer outcomes observed only in the person-time that had not been included in the earlier reports. This association would be helpful in meta-analyses that synthesize the available evidence.⁴ For example, the hazard ratio (HR) for the association between decreased metabolism (compared with extensive metabolism) and recurrence can be estimated as 1.50 (95% confidence interval [CI], 1.15-1.96).

I calculated this HR as an inverse variance-weighted average of the adjusted associations reported in Table 3: heterozygous extensive metabolizers and intermediate metabolizers vs extensive metabolizers (HR, 1.40; 95% CI, 1.04-1.90), averaged with poor metabolizers vs extensive metabolizers (HR, 1.90; 95% CI, 1.10-3.28). This averaged HR should itself be an inverse-variance weighted average of (1) the unreported adjusted association in the new person-time, (2) the comparable adjusted HR reported by Goetz et al² (HR, 1.91; 95% CI, 1.05-3.45), and (3) the comparable adjusted HR reported earlier by Schroth et al³ (HR, 2.24; 95% CI, 1.16-4.33). I estimate that the missing adjusted association in the new person-time is HR, 1.26 (95% CI, 0.91-1.76). It would be far better, however, if the authors would provide the estimate of this association and its CI.