Validation of existing prognostic models in patients with early-stage cervical cancer

Models that predict survival and recurrence in patients with early-stage cervical cancer are important tools in patient management. We validated 12 existing prognostic models in an independent population of patients with early-stage cervical cancer.

We searched the literature for prognostic models in patients with surgically treated early-stage cervical cancer. The prognostic performance of these models was assessed in a consecutive group of surgically treated patients with early-stage cervical cancer treated in our hospital between 1982 and 2004. The performance of the models was visually assessed with calibration plots, which display the relation between the predicted and observed survival.

Twelve published prognostic models met the inclusion criteria. The models categorized the patients into two to four risk groups. Prognostic factors most frequently used in these models were depth of invasion, lymph node metastasis, vascu/vascular space involvement, and tumor size. The models were validated in 563 consecutive patients with early-stage cervical cancer. All of the models underestimated the recurrence-free survival or disease-specific survival in our patients. Only two models performed reasonably well in our population. The use of more than three prognostic categories in the models was not meaningful.

In general, the models underestimated the survival. Only 2 of the 12 prognostic models for patients with early-stage cervical cancer were valid for the prediction of the recurrence-free or disease-specific survival in our patient population.