[Survival tree analysis--a new method for multivariate evaluation of prognostic factors in operative cervix carcinoma].

Objective: We introduce a new multivariate analysis, not yet applied to gynecologic cancer, that includes the creation of a survival tree. The tree structured survival analysis is a statistical method designed to identify meaningful prognostic subsets in a study population, which usually do not emerge from routine proportional hazards analysis. We also applied the scoring systems from other groups to our patients and compared them with our system using a variety of statistical methods.

Material And Methods: After excluding patients with microinvasive and small cell carcinoma, data from remaining 301 patients were analyzed. We performed an univariate and multivariate analysis. Significant single parameters and other variables considered important were chosen for multivariate analysis, including the creation of a survival tree.

Results: Risk factors to define prognosis best were: Depth of invasion, lymph vascular space involvement, age of 40 years and lymph node metastases.

Conclusions: The presented model separates patients with early stage invasive carcinoma of the cervix into three subgroups with a significant different survival and correlates well with other models. Our new model is easy to apply and only requires depth of invasion, lymph vascular space involvement, node status and age of a patient to define the individual risk.