Inflammatory prognostic markers in clear cell renal cell carcinoma - preoperative C-reactive protein does not improve predictive accuracy

OBJECTIVE To re-evaluate the prognostic and predictive significance of the preoperative white blood cell (WBC) count and C-reactive protein (CRP) that independently predicts patient prognosis and to determine optimal threshold values for CRP. PATIENTS AND METHODS From 1996 to 2005, 327 patients with surgery for clear cell renal cell carcinoma were retrospectively evaluated. Cox proportional hazard models were used, adjusted for the effects of tumour stage, size, Fuhrman grade and Karnofsky index, to evaluate the prognostic significance of WBC count and CRP and to identify threshold values. Identified thresholds were correlated with clinicopathological parameters and used to estimate cancer-specific survival. To prove any additional predictive accuracy of the identified threshold it was compared with a clinicopathological base model using the Harrell concordance index (c-index). RESULTS In univariable analyses WBC count was a significant prognostic marker at a concentration of 9.5/mu L (hazard ratio [HR] 1.83) and 11.0/mu L (HR 2.09) and supported CRP values of 0.25 mg/dL (HR 6.47, P < 0.001) and 0.5 mg/dL (HR 7.15, P < 0.001) as potential threshold values. If adjusted by the multivariable models WBC count showed no clear breakpoint, but a CRP value of 0.25 mg/dL (HR 2.80, P = 0.027) proved to be optimal. Reduced cancer-specific survival was proved for CRP 0.25 mg/dL (median 69.9 vs 92.3 months). Median follow-up was 57.5 months with 72 (22%) tumour-related deaths. The final model built by the addition of CRP 0.25 mg/dL did not improve predictive accuracy (c-index 0.877) compared with the clinicopathological base model (c-index 0.881) which included TNM stage, grading and Karnofsky index. CONCLUSIONS Multivariable analyses revealed that an optimal breakpoint of CRP at a value of 0.25 mg/dL was best to stratify patients at risk of cancer-specific mortality, but CRP 0.25 mg/dL added no additional information in the prediction model. Therefore we cannot recommend measuring CRP as the traditional parameters of TNM stage, grading and Karnofsky index are already of high predictive accuracy.