Development and validation of prognostic models in patients with metastatic renal cell carcinoma (mRCC) treated with sunitinib: A Greek-French collaboration.

e15040 Background: The use of tyrosine kinase inhibitors (TKIs) in mRCC has improved prognosis but the individual outcome remains largely unpredictable. The MSKCC model, used to identify risk groups, was developed in cytokine-treated patients and has not been externally validated in the TKI era. It contains 3 laboratory factors (total 5), making its application to retrospective series somewhat problematic. Subsequently, a more complicated model, using 4 laboratory factors (total 6) has been described. The Hellenic Cooperative Oncology Group recently described a simpler model with only 3 clinical factors. We are describing the application and external validation of this model. Methods: 128 Greek patients with mRCC treated with 1 st line sunitinib were included. All had had nephrectomy. Previous interferon was allowed. Cox regression was used to develop a predictive model for overall survival (OS). Our model was compared to that of MSKCC and Heng’s using ROC curves and Harrell’s Concordance Index. Risk groups were defined by the calculated prognostic index and by clinical factors. External validation was done using a sample of 226 French patients. The Royston and Sauerbrei D statistic was used as a measure of discrimination of the survival model. Results: Time from diagnosis of RCC to start of sunitinib (<12), PS (>1) and number of metastatic sites (>1) were independent adverse prognostic factors in the Greek dataset. The co-efficients for each factor were: 0.51, 0.97, 0.61, respectively. The 3 risk groups were defined by the 25 th and 75 th percentiles of the prognostic index values (Table 1). The model was of equal prognostic value to the MSKCC (p=.272) and Heng’s (p=.075). French had better survival than Greek patients especially in the high risk group (for all models). Validation of our model in the French data showed that it was applicable (R 2 D: 0.14, SE: 0.09), especially for the low/medium risk groups. Conclusions: Our model is the only one externally validated in TKI-treated patients. It may be considered as a simpler alternative to those currently applied. [Table: see text]