Objective: To evaluate factors influencing survival and compare curren
t classification systems in women treated for malignant gestational tr
ophoblastic tumor. Methods: A consecutive series of 454 women treated
between 1968-1992 was reviewed retrospectively to identify potential c
linical prognostic factors using univariate analysis of life tables. A
ll patients were evaluated using clinical classification, World Health
Organization, and recently modified International Federation of Gynec
ology and Obstetrics (FIGO) staging systems, applied retrospectively.
Multivariate Cox regression analysis was used to model potential indep
endent prognostic factors within subsets of the patient population. Re
sults: Factors identified by univariate analysis as potential prognost
ic influences included age, duration of disease, type of antecedent pr
egnancy, clinicopathologic diagnosis, site of metastases, number of me
tastatic sites and foci, tumor size, and prior therapy. The pre-therap
y hCG level was not significantly associated with survival (P < .04).
Multivariate Cox modeling consistently identified prior therapy, type
of antecedent pregnancy, number of metastatic sites, and duration of d
isease as independent prognostic factors. Clinicopathologic diagnosis
and hCG level were of borderline significance only in some models of t
he total patient population. All classification systems were able to i
dentify low- and high-risk subsets of patients with approximately equa
l efficiency. The addition of FIGO substages enhanced discrimination b
etween prognostic groups in patients with stage III disease. Conclusio
ns: Existing systems for the classification of malignant gestational t
rophoblastic tumor are based in part on factors that are not independe
ntly prognostic, such as hCG level or tumor size. These systems discri
minate between low- and high-risk patients with approximately equal ef
ficiency. The clinical classification system is currently preferred fo
r determining initial therapy in women with malignant gestational trop
hoblastic tumors.