Determining Survival Impact and Cost-Effectiveness of Multi-Gene Panel Sequencing in Metastatic Colorectal Cancer With Super Learning Approaches

OBJECTIVES: Mutated forms of the KRAS and NRAS genes confer lack of response to third-line standard-of-care anti-EGFR-targeted therapies in metastatic colorectal cancer. Compared to single-gene KRAS tests, multi-gene panel sequencing expands treatment de-escalation by also detecting NRAS variants. Reimbursement of panels remains unequal across healthcare systems given uncertain clinical and economic impacts. We determined the population-level effectiveness and cost-effectiveness of publicly reimbursed multi-gene sequencing compared to single-gene KRAS testing for metastatic colorectal cancers.

METHODS: Our population-based retrospective study design used patient-level linked administrative health databases capturing all cancer and non-cancer care. We considered adult British Columbia, Canada residents with a metastatic colorectal cancer diagnosis receiving single-gene or multi-gene testing between 2016-2018. To maximize balance on characteristics across groups, we estimated inverse probability of treatment weights using super learning (SL). SL assigns weights to multiple machine learning algorithms, minimizing ten-fold cross-validated negative log-likelihood loss of the logistic-linear combination of regression estimators. We estimated mean three-year survival time and costs (public healthcare perspective; 2021CAD) and calculated the incremental net monetary benefit (INMB) for life-years gained (LYG) at $50,000/LYG using weighted linear regression and nonparametric bootstrapping. Final weighting also accounted for inverse probability of censoring weights. We then estimated overall survival using targeted maximum likelihood estimation (TMLE) with super learning. TMLE is a double-robust statistical method minimizing bias in estimates.

RESULTS: Our study included 892 patients receiving panels and 166 single-gene testing. SL weighted five learners, achieving good balance on all 20 covariates. Incremental costs were -$265 (95%CI: -$14,314, $14,530) and LYG were 0.27 (95%CI: 0.06, 0.49). INMB was $13,546 (95%CI: $2,281, $26,552), with a 97.6% probability of being cost-effective. LYG from TMLE were 0.22 (95%CI: 0.05, 0.39).

CONCLUSIONS: Real-world evidence demonstrates reimbursing multi-gene sequencing for more precise targeting of metastatic colorectal cancer treatments provides value for healthcare systems and clinically important benefits to patients.