JNCI | Editorials 159 Biomarker studies with negative results can be just as important to publish as those with positive results if the study had sufficient power to exclude a positive interaction because clinging to a longfavored but incorrect hypothesis in the face of negative evidence impedes scientific and clinical progress. The study by Viale et al. ( 1 ) in this issue of the Journal is a case in point. The investigators studied tumor tissue samples accrued from two large randomized trials that compared endocrine therapy vs cyclophosphamide, methotrexate, and 5-fluorouracil (CMF) chemotherapy followed by endocrine therapy among breast cancer patients with node-negative hormone receptor – positive disease (International Breast Cancer Study Group Trials VIII and IX). In contrast to their expectations, Viale et al. ( 1 ) found that tumor Ki-67 labeling index (a measure of the percentage of tumors cells that express the cell proliferation antigen Ki-67 and are therefore in an active phase of the cell cycle) provided information on prognosis but did not predict which patients would benefit from adding CMF to endocrine therapy. This result is striking because it indicates that patients with aggressive node-negative, hormone receptor – positive breast tumors that have a high growth fraction — the patients most in need of additional therapy — obtain no extra benefit from the addition of CMF to their endocrine regimen when compared with patients with lower risk disease. At the most basic level, this study supports the long-held position of the American Society of Clinical Oncology Tumor Markers Expert Panel that measurements of the cell cycle (ie, S phase fraction or Ki-67 labeling index) should not be used for chemotherapy decision making ( 2 ). When compared with the positive results obtained by Paik et al. ( 3 ) with the 21-gene recurrence score assay, which was able to identify a group of high-risk, nodenegative, hormone receptor – positive patients who differentially ex perienced benefit from CMF, the results reported by Viale et al. ( 1 )