[Survival prediction for terminal gynecologic cancer patients]

To construct a scoring system of predicting the survival time of terminal gynecologic cancer patients.

The clinical data of 91 patients with terminal gynecologic cancers, aged 56 (13-83), who were not suitable to specific anti-cancer therapy and died in the hospital with a mean survival time of 27 days (1-240 days) were analyzed retrospectively. Initially univariate analysis was performed to evaluate the relationship between 19 clinico-biochemical indices and survival time. Multiple logistic regression was used to analyze 9 out of the 19 clinico-biochemical indices which had significant effects on the survival time, and a regression model with 5 indices was constructed by backward selection procedure. The regression coefficient of any category was divided by the maximum regression coefficient so as to get the score of this category. The cores were added together so as to get the overall score of an individual patient.

Univariate analysis identified 9 factors independently and significantly influencing the survival time: short breath, Karnofsky performance status (KPS), age, high fever, speed of tumor growth, presence or absence of treatment-related complication, blood urea nitrogen (BUN), creatinine, and platelet. A regression equation was composed of 5 factors: short breath, KPS, age, higher fever, and BUN by multiple logistic regression with a correct classification ability of 83.5%. The 91 patients were then divided into 2 groups based on this prognostic score system: Group A (n = 37) with a score < or = 9.5 and the mean survival time of (65 +/- 7) days, and Group B (n = 54) with a score > or = 10 and the mean survival time of (19 +/- 2) days. 31 of the 37 patients in Group A survived > or = 30 days, and 46 of the 54 patients in Group B died within 29 days. The Kaplan-Meier survival curves of these 2 group were significantly different (P < 0.001).

A simple, valid, and useful prognostic score system has been established.