A new prognostic model developed for the checkpoint inhibitor era showed better discrimination of survival outcomes in patients with previously untreated metastatic clear cell renal cell carcinoma (RCC) than the widely used International Metastatic RCC Database Consortium (IMDC) model. The findings, reported by Anke Richters, PhD, and colleagues, were published in The Lancet Oncology.
The IMDC model has been used for more than a decade to classify patients as having favorable, intermediate, or poor prognosis and plays a role in both treatment recommendations and clinical trial design. However, it was developed and validated using patients treated largely with VEGFR inhibitors between 2004 and 2010, before checkpoint inhibitor combinations transformed first-line treatment.
The investigators developed the new Clinical Prognostic Index in the Checkpoint Inhibitor era (CPI2) to provide a more contemporary assessment of prognosis.
Study Details
The researchers used individual patient data from two phase III trials of previously untreated metastatic clear cell RCC. CPI2 was developed using data from 1,096 patients in CheckMate-214 and externally validated using 651 patients from CheckMate-9ER, for a total of 1,747 patients. CheckMate-214 evaluated nivolumab plus ipilimumab vs sunitinib, and CheckMate-9ER evaluated nivolumab plus cabozantinib vs sunitinib.
The new model incorporates 14 routinely available factors: age, Karnofsky performance status, previous nephrectomy, metastases to the liver, lung, bone, and lymph nodes, and seven laboratory measures, including calcium, alkaline phosphatase, albumin, lactate dehydrogenase, and neutrophil, lymphocyte, and white blood cell counts. PD-L1 expression did not improve the model's prognostic accuracy.
Median follow-up was 40 months in CheckMate-214 and 38 months in CheckMate-9ER.
Key Findings
In the external CheckMate-9ER validation population, CPI2 better distinguished among patients with different survival outcomes than the IMDC classification. At 36 months, its C-index—a measure of how well a model distinguishes patients with different outcomes—was 0.72 vs 0.61 with IMDC. The model also closely matched predicted with observed survival; predicted 36-month survival was 50%, compared with an observed rate of 54%.
The two models also placed many patients in different risk groups. In CheckMate-9ER, 38% of patients considered favorable risk by IMDC were classified as intermediate or poor risk by CPI2. Among those considered intermediate risk by IMDC, 41% were classified as favorable risk and 24% as poor risk by CPI2.
The authors noted that this reclassification could have implications beyond estimating an individual patient's prognosis. Pivotal trials that helped shape current first-line treatment recommendations have analyzed treatment effects according to IMDC risk groups. If patients were misclassified by the older model, the investigators suggested, differences in treatment effects among prognostic groups may have been obscured or underestimated.
“Continued reliance on the IMDC model is likely to result in substantial misclassification of patients into prognostic risk groups,” they wrote.
Limitations
The authors also highlighted several limitations. Both datasets came from trials sponsored by the same company, and patients with a Karnofsky performance status below 70%, brain metastases at baseline, or non–clear cell histology were not represented. Attempts to obtain data from the KEYNOTE-426 and CLEAR trials for additional external validation were unsuccessful. The investigators said CPI2 should therefore undergo further validation in other clinical trials as well as populations treated in routine practice.
Although CPI2 requires 14 inputs compared with six risk factors in the IMDC model, the investigators noted that its clinical and laboratory measures are readily available in practice. “There is little justification for using less accurate prognostic models solely for computational simplicity,” they wrote.
Dr. Richters, of the Netherlands Comprehensive Cancer Organisation, is the corresponding author of the study.
DISCLOSURE: The study was partly funded by the Josephine Nefkens Foundation. Dr. Richters reported institutional research grants from Astellas, AstraZeneca, Janssen, and Merck. For full disclosure of all study authors, visit The Lancet Oncology.
