A logistic model based on enzyme activities for the prediction of response of breast cancer patients to chemotherapy.

A logistic regression model, utilizing the activities of certain selected glycolytic enzymes and ER status measured on primary or recurrent lesions, has been applied to predict for response to combination chemotherapy regimens administered to women with advanced breast cancer. The clinical outcome of response or no response was evaluated retrospectively using criteria employed by cooperative group protocols. In 93 cases, 58/61 patients classified as nonresponders and 22/32 patients demonstrating objective responses would have been correctly designated, based on the 50% estimated probability as the level for separation of responders from nonresponders. The overall predictive accuracy of this model was 86%, with apparently greater accuracy for prediction of lack of response. Addition of estrogen receptor status to the model imparted no gain in accuracy of prediction. Application of this model to a prospective study is warranted.