A new prognostic model for survival in second line for metastatic renal cell carcinoma: development and external validation

Abstract

Background

In patients with metastatic renal cell carcinoma (mRCC), the oncologic benefit of second-line treatment for high volume tumors or presence of more than five risk factors remain to be defined. Our aim was to develop and externally validate a new model most likely to correctly predict overall survival (OS) categories in second line.

Method

mRCC patients treated within clinical trials at Gustave Roussy Cancer Campus (GRCC) formed the discovery set. Patients from two phase III trials from Pfizer database (PFIZERDB), AXIS (NCT00678392), and INTORSECT (NCT00474786), formed the external validation set. New prognostic factors were analyzed using a multivariable Cox model with a backward selection procedure. Performance of the GRCC model and the prognostic classification scheme derived from it, measuring by R², c-index, and calibration, was evaluated on the validation set and compared to MSKCC and IMDC models.

Results

Two hundred and twenty-one patients were included in the GRCC cohort and 855 patients in the PFIZERDB. Median OS was similar in the discovery and validation cohorts (16.8 [95% CI 12.9–21.7] and 15.3 [13.6–17.2] months, respectively). Backward selection procedure identified time from first to second-line treatment and tumor burden as new independent prognostic factors significantly associated to OS after adjusting for IMDC prognostic factors (HR 1.68 [1.23–2.31] and 1.43 [1.03–1.99], respectively). Dividing patients into four risk groups, based on the number of factors selected in GRCC model, median OS from the start of second line in the validation cohort was not reached (NE) [95% CI 24.9–NE] in the favorable risk group (n = 20), 21.8 months [18.6–28.2] in the intermediate-risk group (n = 367), 12.7 months [11.0–15.8] in the low poor-risk group (n = 347), and 5.5 months [4.7–6.4] in the high poor-risk group (n = 121). Finally, this model and its prognostic classification scheme provided the better fit, with higher R² and higher c-index compared to other possible classification schemes.

Conclusion

A new prognostic model was developed and validated to estimate overall survival of patients with previously treated mRCC. This model is an easy-to-use tool that allows accurate estimation of patient survival to inform decision making and follow-up after first line for mRCC.

http://bit.ly/2E228T8