Response

Second, the authors excluded patients with advanced breast cancer because of the significant mortality in this group. Third, the authors argued that exdusion of women with specific exposures is unlikely to have introduced bias. However, we are not told how many of the original study group were successfully contacted, what proportion had died, and what proportion of those recontacted, and eligible, agreed to participate in the second study. Did these proportions differ between cases and controls? Fourth, in addition to the possibility of selection bias due to differential recruitment in the second study, the lack of a significant difference between cases and controls may be due to the restriction of the study to early stage cancer. In the study by Kabat et al. (2), among postmenopausal women the ratio of 2-OHE1/16a-OHE1 was strongly and inversely associated with breast cancer. However, this association was driven primarily by a strong association with later stage cancer (stages III and IV). Finally, in the small sample for which the results were reported, adjustment for breast cancer risk factors (induding age at menarche, age at first pregnancy, parity, family history of breast cancer, and ethnicity) was apparently not carried out. This is critical because the matched-pair design of the original study was not maintained in the current study. We look forward to the full report, which will hopefully provide more detailed information on these questions. It should be clearly understood that these results, as they now stand, are in no way inconsistent with our hypothesis that the metabolite ratio is a valid biochemical marker for breast cancer.