This economic evaluation examined whether neoadjuvant pembrolizumab followed by adjuvant pembrolizumab in combination with standard of care (NP/AP + SOC) is cost-effective compared with SOC alone for patients with stage III–IVA resectable locally advanced head and neck squamous cell carcinoma (LA HNSCC) from the Swiss healthcare payer perspective. The clinical efficacy inputs derived from the randomized KEYNOTE-689 trial demonstrated improved event-free survival (EFS) for NP/AP + SOC versus SOC, which motivated this cost-effectiveness assessment.
The analysis adapted a previously published US 4 mutually exclusive health-state Markov cohort model to the Swiss context. Patient-level efficacy data from KEYNOTE-689 were used to fit parametric survival models and inform transition probabilities. The time horizon was lifetime, with outcomes discounted at 3.0% per annum for both costs and effectiveness in the base case. Primary outcomes were costs (in 2025 Swiss francs, CHF), quality-adjusted life-years (QALYs), life-years (LYs), and the incremental cost-effectiveness ratio (ICER) expressed as CHF per QALY gained. The evaluation used a hypothetical willingness-to-pay (WTP) threshold of CHF 100,000 per QALY gained.
A 4-state Markov cohort framework was applied, consistent with the adapted US model. Efficacy and survival inputs were derived from patient-level KEYNOTE-689 data via parametric survival analysis to allow extrapolation beyond the median follow-up reported in the trial. Health-state transitions, event rates, and long-term survival projections were driven by these fitted survival curves.
Resource use and unit costs were collected from Swiss national tariffs and informed by Swiss clinical experts where direct Swiss data were not available. All costs are reported in 2025 CHF. Health-state utilities were calculated from EuroQoL-5 dimension (EQ-5D) data collected within KEYNOTE-689; where Swiss-specific utility values were unavailable, Belgian utility values were employed as reported. The model incorporated treatment acquisition and administration costs, downstream healthcare costs associated with disease states, and other relevant medical costs as reflected in the Swiss adaptation.
The base-case analysis produced incremental costs, QALYs, and LYs over the lifetime horizon. Deterministic sensitivity analyses, probabilistic sensitivity analysis (PSA), and scenario analyses were performed to evaluate parameter uncertainty and structural assumptions. Key assumptions included extrapolation of trial follow-up (median 3.2 years) to a lifetime horizon and use of expert opinion for Swiss resource use where real-world Swiss cost data were not available.
In the base-case model, NP/AP + SOC extended mean life expectancy by 1.81 years relative to SOC and generated 1.42 incremental QALYs. Mean total costs were CHF 172,086 for the NP/AP + SOC strategy versus CHF 151,908 for SOC, yielding an incremental cost of CHF 20,178 per patient. The resulting ICERs were CHF 14,227 per QALY gained and CHF 11,130 per LY gained. Under a hypothetical WTP of CHF 100,000 per QALY, the probabilistic analysis indicated a 96% probability that NP/AP + SOC would be considered cost-effective.
The authors conducted deterministic, probabilistic, and scenario analyses to probe the robustness of the base-case findings. These analyses examined the impact of alternative model inputs and structural choices, including survival extrapolation methods, cost inputs, and utility sources. The overall conclusion—that NP/AP + SOC is cost-effective at the assumed CHF 100,000/QALY threshold—remained robust across the reported sensitivity analyses. Specific ranges, drivers of ICER variation, and scenario details were reported in the original article.
The study authors acknowledged several limitations drawn from the source data and adaptation process. The KEYNOTE-689 median follow-up was 3.2 years, necessitating extrapolation to estimate long-term outcomes. Real-world Swiss cost data were not available for all inputs, requiring reliance on Swiss clinical expert opinion. Swiss-specific utility values were unavailable and Belgian utilities were used instead. These gaps introduce uncertainty in absolute cost and utility estimates despite sensitivity testing.
Based on a Markov cohort model populated with patient-level KEYNOTE-689 efficacy and EQ-5D data, and Swiss cost inputs or expert-informed estimates, neoadjuvant pembrolizumab followed by adjuvant pembrolizumab plus SOC for resectable stage III–IVA LA HNSCC was modelled to be cost-effective versus SOC from the Swiss healthcare perspective. At an assumed WTP of CHF 100,000 per QALY, NP/AP + SOC showed a high probability of being cost-effective, supporting consideration of clinical implementation within the stated limitations and uncertainties reported by the authors.