The functional relationship between in vivo bromodeoxyuridine labeling index and Ki-67 proliferation index in human breast cancer.

Proliferation indices are used, along with other parameters, to estimate the risk of recurrence of breast cancer for individual patients. Because it is unlikely one index will be practical for all patients, it is important to understand the relationship between various indices of proliferation. For this reason, we compared a proliferation index based on in vivo labeling of S-phase tumor cells with the thymidine analog bromodeoxyuridine (BrdUrd), to a proliferation index based on an estimate of the growth fraction with the MIB-1 antibody to the Ki-67 antigen. With informed consent, we gave 145 patients 200 mg/m2 BrdUrd intravenously just prior to surgical removal of breast cancer. On histology sections, we visually counted S-phase cells which had incorporated BrdUrd using the Br-3 antibody which is specific to DNA-incorporated BrdUrd, and we counted cells in the growth fraction using the MIB-1 antibody to the Ki-67 antigen. We found that both indices were positively correlated with tumor size, number of positive nodes, and tumor grade, and both were negatively correlated with age and estrogen-progesterone receptor positivity. Using a linear functional relationship model, we found that the best (i.e. the maximal) fit between the two indices (correlation coefficient 0.79; p < 0.0001) occurred when each index was square root transformed, as is appropriate when counts follow a Poisson distribution. When we used the median as a cutpoint for each index, the classification of 19 percent of data pairs changed depending upon which index was used. We also estimated that the Ki-67 intercept (1.02 +/- 0.25) was significantly greater than zero. We conclude that the BrdUrd index of DNA synthesis in S-phase correlates highly with the MIB-1 index of the growth fraction, and both indices correlate well with other parameters of tumor aggressiveness. Because this correlation is driven by concordance of the extremes of high and low counts, clinical comparison will be necessary to determine which is the better prognostic marker for human breast cancer.