The rate of decline of CA 125 in effectively treated epithelial ovarian cancer is described by the exponential regression curve CA 125 = EXP [i - s (days after surgery)]. In this equation i, the y-axis intercept, measures initial tumor burden whereas s, the slope of the regression curve, is determined by the extent of cytoreductive surgery and the subsequent response to chemotherapy. Departure from the regression curve uniformly results in progressive disease. In patients whose cancers had been completely removed, we calculated the mean half-life of CA 125 to be 10.4 days (range 4 to 21). In this case s = 0.0835 and characterizes the ideal regression rate. The model predicts that high-dose cisplatin chemotherapy (s = 0.0671) is more effective than low-dose cisplatin (s = 0.0380) (p less than 0.03) in eliminating residual cancer. Because s can be calculated within 2 to 3 months of treatment and then compared with s for the ideal regression curve and with the values of s reported for standard chemotherapy, evaluation of any new treatment protocol can be facilitated with this method.