Prognostic significance of pretreatment variables in patients with invasive cervical cancer

Abstract

In 119 consecutive patients with invasive cervical cancer 92 variables of potential significance in predicting recurrence were determined before treatment. In 49 patients recurrent disease was diagnosed during follow-up after therapy. The investigated variables include: clinical data, routinely performed hematological and biochemical tests, histopathological data, oncofetal antigens, placental and pregnancy associated substances, serum levels of acute phase reactants, complement components, immunoglobulins and circulating immune complexes, and assays measuring the general immune competence of the patient. Thirty-three of the variables appeared to give statiatically significant differences between values of future recurrence and non-recurrence patients. Further selection was carried out by stepwise forward discriminant analysis. This selection resulted in a prognosis rule using a subset of 7 variables. Carcino-embryonic antigen and the complement component C3PA appeared to be the most powerful variables. Although the stage of disease was the third most important single variable, it did not add prognostic information, once CEA and C3PA were used. By applying the prognosis rule a probabilistic prediction is obtained for each patients. In about 60% of the future recurrences and 40% of the future non-recurrence patients, these predictions reached a high level of accuracy. It was argued that the prediction of non-recurrence does not warrant to change the treatment protocol at present. Whether the prediction of recurrence is clinically useful remains to be clarified.