To the editor

We thank Borg et al. for their News and Views article1 highlighting our publication in the 14 November 2002 New England Journal of Medicine, in which we correlated the expression of low-molecular-weight isoforms of cyclin E to poor survival in breast cancer2. We would like to draw attention, however, to an error in their review, in which they mistakenly stated that we did not include cell proliferation indices such as Ki-67 in conjunction with our cyclin E analysis. The prognostic value of Ki-67 and ploidy as measures of proliferation and S-phase fraction were clearly shown in Table 1 of our paper. When evaluated by Ki-67 staining, 15% of patients with a negative proliferation index had died of breast cancer within five years, compared with 37% with a high proliferation index. In addition, 19% of patients with diploid tumors died of disease by five years, compared with 33% with aneuploid tumors. In contrast, only 5% of patients with low cyclin E levels died of breast cancer within five years, whereas 78% of patients with high cyclin E levels died of their disease. Although Ki-67 and ploidy were significant in the univariate analysis, they lost significance in the multivariate model. Cyclin E remained the single most significant predictor of survival, at least ten times more powerful than either proliferation or ploidy. The reviewers also questioned the exclusion of tumor grade from our analysis. Data regarding the prognostic value of tumor grade in patients with smaller, node-negative tumors, a substantial proportion of our study cohort, are not consistent3,4,5. In fact, variability in the literature regarding tumor grade and prognosis led the Breast Task Force to exclude this variable from the recent American Joint Committee on Cancer (AJCC) breast cancer staging update6. Because of these issues of reliability and clinical use, we decided not to compare tumor grade with cyclin E. We hope this clarifies the issues raised by Borg et al. regarding our data analysis.