The role of serum CA 125 in monitoring the response of epithelial ovarian cancer to treatment has been extensively investigated. The exponential regression curve [1n(CA 125) = i+s (days after initiation of treatment)] has been reported to describe the rate of change of serum CA 125 during treatment. In this model, the y-axis intercept (i) represents the initial CA 125-secreting tumor burden, while the slope (s) is determined by the response to treatment. The exponential regression curve was calculated for 66 patients undergoing salvage chemotherapy with taxol. At a mean follow-up of 121 days, 50 (75%) patients had progressed and 35 (53%) had died. Stratification of the patients by stage, grade, or histology did not reveal any significant differences in the regression rate. When the patients were stratified by response, the mean regression rate was 0.0157 +/- 0.011 for patients with progressive disease (N = 19) vs -0.0250 +/- 0.031 for those with stable disease (N = 25) and -0.0250 +/- 0.015 for those with a partial response (N = 22) (P < 0.0001). The regression rate did not correlate with progression-free interval or survival (P > 0.05). We conclude that changes in serum CA 125 levels follow an exponential regression curve in patients undergoing salvage chemotherapy with taxol for progressive or recurrent ovarian cancer. A positive regression rate may predict which patients will progress prior to the time progression becomes clinically evident. However, a negative rate fails to provide discriminatory utility in predicting progression-free interval or survival.