W e read with much interest the report by Lai et al. showing that trastuzumab therapy is not associated with an increased risk of central nervous system (CNS) metastases in patients with breast carcinoma. 1 However, we believe that the authors' findings, as presented, leave one relevant aspect of the hypothetical correlation between HER-2 positivity, treatment with trastuzumab, and the development of CNS metastases unaddressed.Several recent reports have shown that in the era of taxanes, trastuzumab, and many other new anticancer treatments, patients with metastatic breast carcinoma develop CNS metastases more frequently than in the past. 2 One hypothesis is that, although active and resulting in long-term control in extra-CNS sites, the majority of these drugs do not accumulate in the CNS. Consequently, the survival of patients is prolonged, and CNS micrometastases might have time to grow into overt metastases. Under this hypothesis, we are not surprised that in the study by Lai et al. the risk of CNS metastases was high, and was similar in patients receiving and not receiving trastuzumab. However, calculating risks involves including in the denominators all the patients registered in the study cohorts (both those exposed and unexposed to trastuzumab), regardless of the individual time at risk for the development of CNS metastases during the followup period of the study. More simply, if one patient develops CNS metastases after 2 months and another after 20 months from the starting point of the follow-up period, they will contribute equally to the denominator of the risk, together with patients who do not develop CNS metastases during the entire duration of the follow-up period. If HER-2 positivity confers a more aggressive clinical course 3 and, perhaps, CNS tropism, 4 and if trastuzumab does not control CNS disease sites, 5 the relative risk as a measure of effect would surely not account for a possibly different clinical behavior observed in the two cohorts studied by Lai et al. 1 We wonder whether measuring the incidence rate of CNS progression, and reporting the incidence rate ratio as a measure of effect, would have been more appropriate for assessing whether CNS progression occurs more rapidly in patients with HER-2-positive tumors who are being treated with trastuzumab. In fact, calculation of the incidence rate involves tallying the time at risk of developing the event for each patient in the denominators (i.e., "total months at risk"). In the case of a similar number of events in the two cohorts at the end of the follow-up period, faster disease progression would imply a smaller denominator and therefore a higher incidence rate.Providing incidence ratios of CNS metastases, together with data regarding the overall survival from the onset of metastatic disease and the time to CNS progression in the two patient cohorts, would cer-