Cindolo L et al. (2005) Comparison of predictive accuracy of four prognostic models for nonmetastatic renal cell carcinoma after nephrectomy. A multicenter European study. Cancer 104: 1362–1371

Most patients with nonmetastatic renal cell carcinoma undergo nephrectomy, but predicting long-term survival can be difficult. Although several prognostic models exist, few have been validated and it was unknown how well they could be generalized to all patients with nonmetastatic renal cancer undergoing this surgery. This retrospective study aimed to compare several prognostic models directly, with overall survival as the primary endpoint; secondary endpoints were cancer-specific and recurrence-free survival.

Of the four models selected, those designed by Yaycioglu et al. and Cindolo et al. assign preoperative prognostic scores; the other two, the University of California at Los Angeles integrated staging system (UISS) and the model developed by Kattan et al., assign postoperative scores. The authors applied each model to data collected from 2,404 patients who had undergone surgery for renal cell carcinoma in six European clinics between 1984 and 2002. Clinical and pathologic information was gathered from databases compiled in each clinic and information about outcome was collected by phone call or by contacting patients' primary-care doctors or relatives. Each model's predictive ability was calculated using a RECEIVER OPERATING CHARACTERISTIC CURVE.

Although all four models discriminated prognosis well, the Kattan model was the most accurate (with a value of 0.706, where 1.0 is ideal), closely followed by the UISS. Having applied the models to a heterogeneous group of patients across three European countries, the authors suggest their results will help clinicians to predict outcome, although they hope that current models will be supplemented by molecular prognostic markers as technology develops.