This study was conducted following the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) standard (22). Based on the United States Department of Health and Human Services (45 CFR §46), this study did not use individual patient information, nor did it include human or animal research, so permission for ethical approval was not required from an institutional review board or ethics committee (23).

According to EMPOWER-Lung 3 trial (17), the inclusion criteria included men and women over the age of 18 (for Japanese patients, the age was 20); histologically or cytologically verified non-squamous or squamous NSCLC; Eastern Cooperative Oncology Group (ECOG) source ≤1; at least three months of expected life expectancy; and adequate organ and bone marrow function.

We constructed a partitioned survival model including three mutually incompatible health states in this study: PFS, progressing disease (PD), and death ( Figure 1 ). The area under the OS curve was assessed to determine the percentage of patients with OS, the area under the PFS curve was assessed to determine the proportion of patients with PFS, and based on the difference between OS and PFS curves, the proportion of patients with PD was determined.

The cycle length should be dictated by the natural history of disease and it should be the minimum interval over which the pathology or symptoms are expected to alter. Considering the progression of aNSCLC, the cycle length of the partitioned survival model was 1 week. The time horizon was 10 years that more than 98% of the cohort died. Throughout each cycle, the patients either maintained their current health status or advanced to the subsequent health level. The primary outcomes of this study were overall costs, incremental cost-effectiveness ratios (ICERs), quality-adjusted life-years (QALYs), life-years (LYs), incremental net health benefits (INHB), and incremental net monetary benefits (INMB). The threshold for willingness to pay (WTP) was set at $150,000/QALY (24). All cost and utility results were discounted annually by 3% (25, 26).

These OS and PFS survival curves are taken from the EMPOWER-Lung 3 study, which included cemiplimab in combination with chemotherapy and chemotherapy alone (17), and algorithm developed by Guyot et al. (27) was used to construct data beyond the follow-up period of the trial. In the EMPOWER-Lung 3 trial, which was conducted between June 17 2019 and September 30 2020, cemiplimab plus platinum-based chemotherapy was compared with patients who received platinum-based chemotherapy alone for the therapy of NSCLC (17).

We obtained Kaplan-Meier survival curves from the EMPOWER-Lung 3 trial by GetData Graph Digitizer version 2.26 (28), which allows time-to-survival data points to be extracted. In order to fit this data, parametric survival models were used: Weibull, log-normal, log-logistic, exponential, generalized gamma, and Gompertz. The optimal survival model was chosen based on Akaike information criteria and Bayesian information criterion with the lowest scores. The survival fit model results of cemiplimab plus chemotherapy and chemotherapy alone were displayed in Table 1 , and Supplementary Table 1 displays the results of fit in detail. The proportions of patients with PFS and OS were calculated by using the selected survival distribution. It was found that the survival events and survival durations of virtual patients were comparable to the actual number of patients at risk, which was indicative of a close replication of the survival curves. Further details of the model fitting are presented in Supplementary Figure 1 .

We evaluated the costs that included direct medical charges. Direct medical costs consist of prescription costs, patient health-related costs, adverse event (AE) management costs, and costs related to terminal care ( Table 1 ). According to medical-care inflation from Tom’s Inflation Calculator (36), all costs have been converted to 2021 United States dollars ( Table 1 ). According to the results of the EMPOWER-Lung 3 study, patients in cemiplimab plus chemotherapy group received 350 mg of cemiplimab every three weeks in addition to chemotherapy. Researchers used paclitaxel plus carboplatin, paclitaxel plus cisplatin, pemetrexed plus carboplatin, or pemetrexed plus cisplatin as chemotherapy treatments. There was a maximum duration of treatment of 108 weeks, or until the disease progressed or toxicity became intolerable. To assess direct drug prices, the Centers for Medicare and Medicaid Services (CMS) was used to offer the 2021 average retail price for drugs (37). We estimated the chemotherapy dosage based on the assumption that a typical patient has a body surface area of 1.86 m² and a weight of 70 kg (23). Monitoring costs for patients in the PFS and PD stages were $465 and $1,075 per cycle, respectively (29). The cost of terminal care is $16441.83 per aNSCLC patient (29). This study calculated the costs associated with addressing grade ≥ 3 AEs based on the literature ( Supplementary Table 2 ) (30–32).

There was an associated health value for aNSCLC health status ranging from 0 (death) to 1 (perfect health). The utilities of PFS and PD states for aNSCLC were 0.754 and 0.18, respectively (33, 34). This research assessed the disutility based on grade ≥ 3 AEs and the literature-derived estimate of disutility attributable to treatment-emergent AEs grade ≥3 ( Supplementary Table 2 ) (32, 33, 35).

ICER was represented as the cost per extra QALY gained between cemiplimab plus chemotherapy and chemotherapy alone. When the incremental cost-effectiveness ratio (ICER) is below a certain WTP threshold, cost-effectiveness is recommended (24). The INHB and INMB were presented as the following formulas: INHB ( λ ) = ( μ E Cemiplimab plus chemotherapy   −  μ E C h e m o t h e r a p y ) − ( μ C Cemiplimab plus chemotherapy   −  μ C C h e m o t h e r a p y ) λ =   Δ E −   Δ C λ and INMB ( λ ) = ( μ E Cemiplimab plus chemotherapy   −  μ E C h e m o t h e r a p y )   ×   λ − ( μ C Cemiplimab plus chemotherapy   −  μ C C h e m o t h e r a p y ) =   Δ E   ×   λ   –   Δ C , where μC and μE were the cost and utility of cemiplimab plus chemotherapy or chemotherapy, respectively, and λ was the WTP threshold (38, 39).

To determine the robustness of the model outputs, we conducted one-way sensitivity analysis and probabilistic sensitivity studies in this study. One-way sensitivity analysis were conducted based on various variables, such as costs and utilities. In order to estimate the uncertainty of each variable, we either used the 95% confidence intervals provided by the literature or approximated it by assuming a 25% deviation from the baseline value ( Table 1 ). In order to evaluate the uncertainty of the model, we used Monte Carlo simulation with 10,000 iterations to perform a probabilistic sensitivity analysis. The cost parameters were assigned a gamma distribution, the hazard ratios (HRs) were assigned a log-normal distribution, and the probability, percentage, and utility parameters were assigned a beta distribution. A cost-effectiveness acceptability curve was constructed to determine whether cemiplimab plus chemotherapy or chemotherapy could benefit QALY gains at different WTP levels.

Subgroup analysis was performed to investigate the uncertainty of the model resulting from the various patient characteristics. Variables, such as age, sex, race, histology, PD-L1 expression level, ECOG, area, brain metastases, cancer stage, and smoking history were used while performing subgroup analysis for the various subgroups produced by EMPOWER-Lung 3 (17). In this study, the hesim and heemod packages in R, version 4.0.5, 2021 (R Foundation for Statistical Computing), were used to performed statistical analysis.

Compared to chemotherapy, cemiplimab plus chemotherapy delivered an extra 0.237 QALYs at an additional cost of $50,796, yielding an ICER of $214,256/QALY. An INHB of 0.203 and an INMB of $304,704 were found at a WTP threshold of $150,000/QALY ( Table 2 ).

As a result of one-way sensitivity analysis, the primary driver of the model result is the cost of cemiplimab ( Supplementary Figure 2 ), as it had the greatest impact on ICER. Model results were robust to the uncertainty of other model variables, including costs and disutility associated with AE risk.

These key factors were assessed for their relevance to ICER of cemiplimab plus chemotherapy compared to chemotherapy alone. When the WTP threshold was set at $150,000/QALY and the cost of cemiplimab was less than $15.91 per mg, cemiplimab plus chemotherapy could be considered cost-effective ( Supplementary Figure 3 ).

Based on the cost-effectiveness acceptability curves and ICER scatterplot, cemiplimab plus chemotherapy has a 0.04% chance to be considered as cost-effective when compared with chemotherapy alone when the WTP threshold is set at $150,000 ( Figure 2 ; Supplementary Figure 4 ).

The subgroup analysis, which were performed based on varying the HRs for OS, found that cemiplimab plus chemotherapy was related to a higher than 50% likelihood of being cost-effective at the threshold of $150,000/QALY in the following subgroups ( Table 3 ): female patients, patients with PD-L1 expression ≥50%, patients with an ECOG score of 1, European patients, and never smokers. The INHBs in subgroups ranged from -0.24 for patients with brain metastasis to 0.02 for patients from Europe.

In subgroup analysis conducted by altering the HRs for PFS, cemiplimab with chemotherapy was associated with a likelihood of cost-effectiveness greater than 50% in the following subgroups ( Table 4 ): patients with PD-L1 expression 1-49% and ≥50%, patients with an ECOG score of 0, and patients with locally advanced disease.