A preoperative prognostic model predicting recurrence-free survival for patients with kidney cancer.

ABSTRACT

OBJECTIVE: To develop a preoperative prognostic model in order to predict recurrence-free survival in patients with nonmetastatic kidney cancer. METHODS: A multi-institutional data base of 1889 patients who underwent surgical resection between 1987 and 2007 for kidney cancer was retrospectively analyzed. Preoperative variables were defined as age, gender, presentation, size, presence of radiological lymph nodes and clinical stage. Univariate and multivariate analyses of the variables were performed using the Cox proportional hazards regression model. A model was developed with preoperative variables as predictors of recurrence after nephrectomy. Internal validation was performed by Harrell's concordance index. RESULTS: The median follow-up was 23.6 months (1-222 months). During the follow-up, 258 patients (13.7%) developed cancer recurrence. The median follow-up for patients who did not develop recurrence was 25 months. The median time from surgery to recurrence was 13 months. The 5-year freedom from recurrence probability was 78.6%. All variables except age were associated with freedom from recurrence in multivariate analyses (P < 0.05). Age was marginally significant in the univariate analysis. All variables were included in the predictive model. The calculated c-index was 0.747. CONCLUSIONS: This preoperative model utilizes easy to obtain clinical variables and predicts the likelihood of development of recurrent disease in patients with kidney tumors.