PROCC: a predictive score to identify KRAS wild type metastatic colorectal cancer patients who are likely to benefit from panitumumab treatment

BackgroundPractice guidelines recommend using panitumumab in combination with chemotherapy to treat KRAS wild-type (WT) metastatic colorectal cancer (mCRC) patients where it was shown to significantly extend progression-free survival (PFS) and overall survival (OS). Still, a proportion of patients will not achieve this goal. We propose a simplified predictive score to identify patients who are likely to benefit from panitumumab treatment. MethodsNCT00364013 (TRDS) (n=460) was used as training dataset and NCT00339183 (VALDS) (n=479) as validation set. Datasets were obtained from www.projectdatasphere.org and included KRAS WT mCRC patients treated with panitumumab in combination (P/FOL) or not with FOLFOX (FOL) (TRDS) or FOLFIRI (VALDS) as 1st and 2nd line therapy. TRDS was used to generate synthetic representations (SRs) for each patient through the integration of 36 clinical and analytical features collected, respectively, during the screening phase and the first month of inclusion. These SRs were then input into a machine learning (ML) framework to identify subgroups of patients based on their similarities. The resultant subpopulations were correlated with PFS and OS. Differential variables between subgroups were identified through feature contribution analysis and included in a multivariable logistic regression model. Independent predictive factors found to be statistically significant were used to generate a predictive score of panitumumab response at baseline that was validated in VALDS. ResultsML identified two different subpopulations on the TRDS: SPA (n=162) and SPB (n=298). Only SPA patients had a lower risk of death when treated with P/FOL compared to FOL (HR 0.68 95%CI 0.48-0.99; p=.04). Patients in SPB showed no significant differences on OS between P/FOL and FOL (p=.27). Feature contribution analysis identified 15 differential features between both subpopulations. From these, CEA, ALP, LDH, and platelets were selected to create a simplified predictive score for panitumumab response ranging 0-18. When applied to TRDS, this score yielded an area under the curve of 0.81 (95% CI: 0.77 to 0.85). A score [&ge;]8.5 was correlated to a lower risk of progression (HR 0.67 95% CI 0.47-0.97; p=.03) and death (HR 0.65 95%CI 0.43-0.98; p=.04) after P/FOL compared to FOL. No significant differences were observed for PFS and OS between P/FOL and FOL in patients with a score <8.5. The predictive score was then validated in the VALDS set with similar results (score [&ge;]8.5: PFS: HR 0.48 95%CI 0.33-0.70; p=.002; OS: HR 0.60 95%CI 0.42-0.87; p=.007; score <8.5, PFS: p=.2; OS: p=.9). ConclusionsBased on CEA, ALP, LDH and platelet baseline levels, this easily applicable predictive score might be helpful to accurately select KRAS WT mCRC patients who would benefit from addition of panitumumab to chemotherapy treatment in first- or second-line therapy. Further work is required to validate this approach in prospective cohorts of patients.