Clinical utility and cost-effectiveness of a molecular biomarker for immune checkpoint inhibitor response monitoring and treatment decisions.

e23338 Background: Clinical validation studies have shown that response using ctDNA may detect treatment response following immune checkpoint inhibitor therapy (ICI) earlier and more accurately than radiologic imaging. However, there is no clinical utility evidence to determine if molecular response monitoring-based treatment decisions can improve clinical outcomes compared to imaging-based standard of care (SOC). Here, we simulate an interventional clinical study to explore the potential clinical utility and cost-effectiveness (CE) of molecular response monitoring-based therapy decisions using realistic assumptions based on real-world (rw) and clinical trial data. Methods: We updated a patient-level microsimulation model of a cohort of patients with advanced solid tumors. We compared molecular-response monitoring-based decision-making (intervention) to scan-based treatment decision-making (control). We evaluated progression-free survival (PFS) and overall survival (OS) over a two-year period. We assume all patients receive ICI as first-line therapy, chemotherapy as second-line therapy, and then were placed on best supportive care if they progress on second-line. We incorporate cancer-specific rates of imaging and adverse events from real-world data. Cancer-specific PFS and OS rates were taken from published clinical trials. In the base case, we assume molecular non-responders (nMRs) have 80% longer PFS and OS on therapy after first-line in the intervention arm. We also assume that nMRs that stay on ICI therapy have 80% shorter PFS and OS while on ICI therapy in the control arm. We conducted a sensitivity analysis, varying this parameter from 40% to 120%, or 2-5 months. Imaging, treatment, and adverse event costs were calculated from Medicare’s 2023 perspective (USD). Results: In the base case model, we observed a median OS of 23.6 months in the molecular-response monitoring driven intervention arm compared to 15.4 months in the SOC arm. This was driven by longer PFS in second-line chemotherapy in the intervention arm (median PFS of 10.3 months in the intervention arm vs. 5.3 months in the SOC arm). This resulted in an average total cost of $101,300 in the intervention arm vs. $132,400 in the SOC arm due to early discontinuation of ineffective ICI therapy. When we varied the clinical benefit of therapy switching in the intervention arm, the clinical benefit and CE of the intervention arm remained. Conclusions: Using data-driven, realistic assumptions, we show that use of molecular response monitoring based treatment decision-making may improve OS and may be cost-effective compared to CT scan-based treatment decision-making. Data-driven simulation studies are a useful tool for evaluating the potential clinical utility and CE of emerging technologies, and can be used as a blueprint for design of clinical utility studies.