Factors predictive of survival after first relapse or progression in advanced epithelial ovarian carcinoma: a prediction tree analysis-derived model with test and validation groups.

OBJECTIVE

To identify factors predictive of overall survival after first relapse or primary progression in patients with advanced epithelial ovarian cancer.

METHODS

"Tree-structured prediction of survival for censored survival data" was used to identify the independent prognostic factors in the test group (n = 352) who were the patients from the previously reported Canadian OV.8 trial. A prognostic model was developed using these factors and subjected to validation in the Canadian OV.4 trial cohort (n = 282).

RESULTS

Based upon three factors, time from diagnosis to first recurrence or progression, tumor grade at diagnosis, and ECOG performance status at original diagnosis, three groups of patients were identified. These were labeled as good, intermediate, and poor prognosis with median survivals post relapse of 18 (12), 6 (5), and 1 (2) months, respectively. The figure in parentheses is the survival in the validation cohort.

CONCLUSIONS

These prognostic groupings enable us to recommend second-line treatment more logically. The patients in the poor prognosis group have such a limited survival that cancer shrinking therapy should not routinely be offered. In addition the use of the individual predictive factors as stratification factors will help to avoid erroneous conclusions about treatment efficacy.