Cost-effectiveness of pembrolizumab compared with chemotherapy in the US for women with previously treated deficient mismatch repair or high microsatellite instability unresectable or metastatic endometrial cancer

Abstract

Aims

There is limited published evidence for the cost-effectiveness of treatments for unresectable or metastatic endometrial cancer (mEC). The objective of this analysis was to assess the cost-effectiveness of pembrolizumab versus chemotherapy for previously treated unresectable or mEC, in women whose tumors have deficient mismatch repair (dMMR) or high microsatellite instability (MSI-H). The analysis was carried out from a US healthcare payer perspective.

Materials and methods

A lifetime partitioned survival model comprising three health states (progression-free, progressed disease and death) was constructed. Chemotherapy was represented by single-agent paclitaxel or doxorubicin. Overall survival, progression-free survival and time on treatment data for pembrolizumab were obtained from a Phase II clinical study that included women with previously treated dMMR/MSI-H unresectable or mEC (KEYNOTE-158, NCT02628067). Survival data for chemotherapy were obtained from a published Phase III study for previously treated advanced endometrial cancer. Costs included were drug acquisition and administration, health-state, end-of-life, and adverse event management. Costs were presented in 2019 US$. Outcomes were calculated as quality-adjusted life-years (QALYs), using EQ-5D data from KEYNOTE-158. Model results were tested extensively in deterministic and probabilistic sensitivity analyses.

Results

Results demonstrated that pembrolizumab is a highly cost-effective treatment option when compared with chemotherapy, with estimated deterministic and probabilistic incremental cost-effectiveness ratios (ICERs) of $58,165 and $57,668 per QALY gained, respectively. Pembrolizumab was associated with a large QALY and life-year gain per person versus chemotherapy over the model time horizon (deterministic 4.68 life year gain, 3.80 QALYs), with the majority of QALYs accrued in the progression-free health state.

Limitations

The key limitation of the analysis was the lack of comparative effectiveness data for pembrolizumab versus chemotherapy.

Conclusions

Pembrolizumab is a highly cost-effective treatment option when compared with chemotherapy for women with previously treated dMMR/MSI-H unresectable or mEC. Results were robust to the changes in parameters and assumptions explored.

Keywords:

• Microsatellite instability
• deficient mismatch repair
• endometrial cancer
• economic analysis
• QALY
• cost-effectiveness
• pembrolizumab

JEL CLASSIFICATION CODES:

• I10
• C60
• D60

Introduction

Endometrial cancer forms in the tissue lining of the uterus, are generally adenocarcinomas and occur most frequently after menopause^1,2. In the US, endometrial cancer is the most common cancer of the female reproductive organs and the sixth most common cause of cancer death among women. In 2020, an estimated 65,620 women will be diagnosed with endometrial cancer in the US and an estimated 12,590 women will die from the disease³. Metastatic endometrial cancer frequently involves the ovaries and other reproductive organs, and can often involve the lymphatic system, blood, lungs and other vital organs⁴. The probability of survival is particularly poor for women with metastatic endometrial cancer (mEC, endometrial cancer that has spread to other organs or tissues); in the US, the 5-year survival probability for early stage endometrial cancer is 95% compared with just 17% for mEC³.

There is currently no agreement on the standard treatment for women with mEC; treatment can comprise a combination of surgery, radiotherapy and multi- or single-agent chemotherapy regimens, the choice of which depends on multiple factors⁵. The NCCN Clinical Practice Guidelines In Oncology (NCCN Guidelinesⁱ) for recurrent, metastatic or high-risk endometrial cancer outline a range of combination and single-agent chemotherapy options for consideration, with platinum compounds, anthracyclines, and taxanes most commonly used; however, rates of response are typically low⁶.

The NCCN Guidelines include combination lenvatinib/pembrolizumab, pembrolizumab monotherapy, nivolumab monotherapy, and larotrectinib or entrectinib as biomarker-directed systemic therapy options for second-line treatment of recurrent, metastatic, or high-risk endometrial cancer, in certain circumstances⁶. In European guidelines, single-agent chemotherapies are preferred^7,8. The European Society for Medical Oncology (ESMO) Clinical Practice Guidelines advise that, while several agents have been tested, only paclitaxel has consistently shown a response rate of more than 20%. Additionally, the British Gynaecological Cancer Society (BGCS) Uterine Cancer Guidelines note that paclitaxel, pegylated liposomal doxorubicin and topotecan are options at second line for women who have relapsed within 6 months of their prior chemotherapy^7,10.

In the NCCN Guidelines for mEC, pembrolizumab is presented as a treatment option for tumors with high tumor burden (TMB-H), high microsatellite instability (MSI-H) or deficient mismatch repair (dMMR) and nivolumab is presented as a treatment option for dMMR. dMMR/MSI-H is a form of genomic instability, with estimates in the literature for prevalence in endometrial cancer ranging from 15% to 31% across all stages^6,12–17. Nivolumab is not approved for use in the US for mEC; however, pembrolizumab was granted accelerated approval by the Food and Drug Administration (FDA) for the treatment of unresectable or metastatic solid tumors with MSI-H or dMMR that have progressed following prior treatment in patients who have no satisfactory alternative treatment options, and for the treatment of unresectable or metastatic, MSI-H or dMMR colorectal cancer that has progressed following treatment with a fluoropyrimidine, oxaliplatin, and irinotecan. For solid tumors, approval was supported from the demonstration of anti-tumor activity for patients in KEYNOTE-158 (NCT02628067)^18,19. KEYNOTE-158 is a Phase II, open-label, single-arm trial in which 1,151 people with previously treated unresectable or metastatic solid tumors were enrolled and treated with 200 mg pembrolizumab every 3 weeks. Of those enrolled, 49 participants had MSI-H/dMMR unresectable or mEC¹⁸. For these patients, the objective response rate was 57%, with 16% patients achieving complete response and 41% achieving partial response. Median progression-free survival (PFS) was 26 months, and 73% patients remained alive at 12 months²⁰.

Evidence suggests that dMMR/MSI-H could be a negative prognostic factor for survival in advanced disease. In a study carried out by Arabi et al. patients with high grade endometrial cancer with three negative biomarkers for MSI-H reported markedly poorer survival when compared to microsatellite stable tumors (median survival 3 months vs 71 months, p < .04)¹¹. Pembrolizumab has shown robust clinical benefit; however, the long-term cost-effectiveness of pembrolizumab in dMMR/MSI-H mEC has not been investigated. A search of the economic literature carried out in 2019 identified a single study that evaluated the cost-effectiveness of pembrolizumab in women with mEC; however, this study was limited to the short term, and the outcomes of the analysis were reported in terms of an incremental cost per 2-year survivorship²¹.

The objective of this analysis was to assess the cost-effectiveness of pembrolizumab versus chemotherapy in the treatment of women with unresectable or mEC whose tumors have dMMR/MSI-H and who have been previously treated with chemotherapy. The chemotherapy comparator was represented by the single-agent chemotherapies paclitaxel and doxorubicin, since these treatments are included as second-line treatment options in published guidelines^9,10. The analysis was carried out from the perspective of the US healthcare payer. Cost-effectiveness results are expressed as incremental cost per additional quality-adjusted life year (QALY) and life year (LY) gained, in line with the reference case for economic evaluations published by the Institute for Clinical and Economic Review, and uses a willingness-to-pay threshold of $100,000 per QALY, in line with updated evidence for the US-based threshold^22,23.

Methods

Model overview

A partitioned survival model comprising three health states (progression free, progressed disease and death) was constructed in Microsoft Excelⁱⁱ to assess the cost-effectiveness of pembrolizumab versus chemotherapy for a cohort of women with previously treated dMMR/MSI-H unresectable or mEC (Figure 1). Women with dMMR/MSI-H unresectable or mEC enter the model in the progression-free health state and, each cycle, can remain progression-free, transition to progressed disease, or die. Patients who transition to progressed disease can remain in the progressed disease health state or die. Death is an absorbing health state where once patients enter, they remain in this state for the remainder of the model time horizon. Overall survival (OS) and progression-free survival (PFS) were modelled based on parametric fitting of the OS and PFS KM data, while the proportion of patients in post-progression was estimated as the difference between OS and PFS. A partitioned survival model approach was selected given that models of this type are commonly used to model therapies in advanced or metastatic cancer and it was judged the most appropriate structure to incorporate the comparative trial evidence, given the limitations of the available data²⁴.

Figure 1. Model structure.

The model has a lifetime duration (30 years), in line with the reference case for economic evaluations published by the Institute for Clinical and Economic Review, and has a cycle length of 1 week to ensure that chemotherapy regimens are accurately costed²². Since the cycle length is small when compared with the model time horizon (1 week versus 30 years), a half-cycle correction was judged unnecessary.

The model considers a US healthcare payer perspective and includes all direct healthcare-related costs and effects. Healthcare costs included in the analysis were drug acquisition and administration (including for subsequent therapies), health state-related costs, end-of-life costs, and adverse event costs. Costs are presented in US$ for a cost year of 2019. Outcomes included were life years and QALYs. Costs and QALYs were accrued within each health state and were discounted at an annual rate of 3.0% from Year 1²². Deterministic and probabilistic results were estimated. In addition, one-way sensitivity analysis and a range of scenario analyses were investigated to test the robustness of the model to changes in inputs and structural assumptions, in line with current health economic evaluation best practice²⁵.

The analysis compared pembrolizumab 200 mg as an intravenous (IV) infusion administered every 3 weeks for up to 2 years (or until progression or death, if earlier), with chemotherapy. Chemotherapy was represented by the single agents paclitaxel or doxorubicin because these agents are included as second-line treatment options in published guidelines^9,10. Paclitaxel was assumed to be administered via IV infusion at a dose of 80 mg/m² for 3 out of 4 weeks, until progression or withdrawal from treatment due to unacceptable toxicity. Doxorubicin was assumed to be administered via IV infusion at a dose of 60 mg/m² every 3 weeks for a maximum of eight 3-weekly cycles²⁶.

Model inputs

A summary of all inputs included in the analysis, and the upper and lower values tested in one-way sensitivity analysis, as well as the distributions used for probabilistic sensitivity analysis, is provided in Table 1.

Table 1. Summary of model inputs.

Input	Mean value	Values used in one-way sensitivity analysis		Distribution used in PSA	Source of data
		Lower value	Upper value
Model settings
Time horizon	30 years	20 years	40 years	NA	Assumption
Cycle length	1 week	NA	NA	NA	Assumption
Discount rate (costs)	3%	0%	6%	NA	Institute for Clinical and Economic Review reference case^22
Discount rate (outcomes)	3%	0%	6%	NA
Baseline characteristics
Starting age of women with dMMR/MSI-H mEC	65 years	63 years	68 years	Normal	Baseline characteristics of women with dMMR/MSI-H mEC and included in KEYNOTE-158^18
Weight of women with dMMR/MSI-H mEC	72.9 kg	67.8 kg	78.0 kg	Normal
Height of women with dMMR/MSI-H mEC	160.16 cm	157.7 cm	162.6 cm	Normal
Overall survival
Pembrolizumab overall survival	Two-piece: Kaplan-Meier to 40 weeks and exponential distribution thereafter; log (rate) 6.2148	NA	NA	Multivariate normal	Statistical analysis using data from KEYNOTE-158 – Data on file^18
Chemotherapy overall survival	One-piece: log-normal distribution; intercept 3.8756, scale 0.0669	NA	NA	Multivariate normal	Statistical analysis of McMeekin 2015 study – Data on file^18^,^27
Progression free survival
Pembrolizumab progression-free survival	Two-piece: Kaplan–Meier to 10 weeks and Weibull distribution thereafter; intercept 5.7697, scale 0.4500	NA	NA	Multivariate normal	Statistical analysis using data from KEYNOTE-158 – Data on file^18
Chemotherapy progression-free survival	One-piece: exponential; rate 3.15	NA	NA	Multivariate normal	Statistical analysis of McMeekin 2015 study – Data on file^27
Costs
Cost of treating adverse events (pembrolizumab)	$2,661	$1,618	$3,704	Normal	Calculation based on trial adverse event incidence and costs obtained from AHRQ, Simeone 2020 and Bice 2020^33^,^47^,^48
Cost of treating adverse events (chemotherapy)	$6,721	$3,813	$8,729	Normal
Cost per person testing for dMMR/MSI-H status	$614.23	$373.46	$855.01	Normal	Estimated using data from the CMS Physician Fee Schedule and Laboratory Fee Schedule^35^,^36
Rate of MSI-H expression	14.52%	12.4%	16.8%	Beta	Le et al.^16
Cost of pembrolizumab 100 mg	$4,862.04	$2,956.16	$6,767.92	NA	IBM Micromedex^40
Cost of doxorubicin 200 mg	$64.40	$39.16	$89.64	NA
Cost of paclitaxel 300 mg	$34.15	$20.76	$47.54	NA
Cost of tamoxifen 20 mg	$0.67	$0.41	$0.93	NA
Proportion of women receiving paclitaxel	28%	43.8% doxorubicin	100% doxorubicin	Dirichlet	McMeekin et al.^27
Proportion of women receiving doxorubicin	72%
Administration cost for pembrolizumab	$154.80	$94.12	$215.48	Normal	Calculation based on administration times (pembrolizumab 30 mins, doxorubicin 10 min and paclitaxel 180 mins) using Medicare Physician Fee Schedule 2019 administration costs (HCPCS 96413 and 96417)^35
Administration cost for doxorubicin	$51.60	$31.37	$71.83	Normal
Administration cost for paclitaxel	$430.54	$261.77	$599.31	Normal
Weekly cost of PFS disease management	$83.05	$50.50	$115.61	Normal	Calculation based on monthly outpatient visits and quarterly CT scans using Medicare Physician Fee Schedule 2019 (HCPCS 99213 and 74177)^36
Weekly cost of progressed disease management	$83.05	$50.50	$115.61	Normal
One-off cost of palliative care	$10,384.03	$6,313.56	$14,454.49	Normal	Bekelman et al.^45
Utilities
Progression free survival	0.817	0.797	0.836	Beta	Analysis of EQ-5D data within KEYNOTE-158 for women with dMMR/MSI-H mEC – Data on file – and based on a US algorithm.^18^,^49
Progressed disease	0.779	0.699	0.859	Beta

Abbreviations. dMMR, deficient mismatch repair; mEC, unresectable or metastatic endometrial cancer; MSI-H, high microsatellite instability; PFS, progression free survival; PSA, probabilistic sensitivity analysis.

Efficacy data – overall survival and progression-free survival

PFS and overall survival (OS) data were included in the model for pembrolizumab and chemotherapy. For pembrolizumab, these data were obtained from KEYNOTE-158 and were extrapolated to obtain long-term estimates of survival.

For chemotherapy, PFS and OS data were obtained via a focused literature review. The focused search was designed to find studies of paclitaxel and/or doxorubicin that reported survival data for women with previously treated mEC. All the identified studies were reviewed and are presented in Table 2. None of the identified studies considered patients with dMMR/MSI-H unresectable or mEC. As MSI-H is a negative prognostic factor for disease, it is possible that these identified survival data could overestimate PFS and OS compared with women with MSI-H mEC¹¹.

Table 2. Identified survival data for women with previously treated mEC treated with paclitaxel and/or doxorubicin

Study	Population	Comparator (N)	Median OS (months)	Median PFS (months)	KM plot for OS	KM plot for PFS
McMeekin^27	Locally advanced, recurrent, or metastatic endometrial cancer with at least one failed prior platinum-based chemotherapeutic regimen and up to two prior therapies	Paclitaxel or doxorubicin (same treatment group, undifferentiated) (n = 248)	12.3	4	Yes	NA
Di Legge^29	Advanced or recurrent endometrial cancer failing one previous carboplatin-paclitaxel chemotherapy	Doxorubicin (n = 18)	5.5	[TTP] 2.1	NA	NA
Lincoln^30	Persistent or recurrent endometrial carcinoma who have failed prior chemotherapy	Paclitaxel (n = 44)	10.3	[MRD] 4.2	NA	NA
Makker^31	Recurrent endometrial cancer who received paclitaxel/carboplatin adjuvant chemotherapy followed by second line doxorubicin therapy at time of recurrence	Doxorubicin (n = 17)	5.8	2.1	Yes	Yes
Miller^28	Advanced, recurrent, or metastatic endometrial cancers who had failed prior platinum and taxane therapy	Doxorubicin (n = 255)	10.8	4.7	NA	NA

Abbreviations. KM, Kaplan–Meier; mEC, metastatic endometrial cancer; MRD, median response duration; NA, not available; OS, overall survival; PFS, progression-free survival; TTP, time to progression.

Of the data identified, McMeekin et al. and Miller et al. provided data from the largest cohorts of patients at n = 248 for McMeekin et al. and n = 255 for Miller et al.^27,28 The remaining three studies included data from between 17 and 44 patients^29–31. In addition, only McMeekin et al. included data for women with mEC who received either paclitaxel or doxorubicin; each other study reported data for either paclitaxel or doxorubicin^27–31. Therefore, in the base case analysis, the study published by McMeekin et al. in 2015 was used to estimate PFS and OS over the time horizon of the model²⁷. McMeekin et al. presented results from a Phase III randomized trial of ixabepilone versus paclitaxel or doxorubicin for the second-line treatment of women with advanced endometrial cancer²⁷. In the comparator arm of this study, women received either paclitaxel (28%) or doxorubicin (72%), though survival data were combined, and not presented separately for paclitaxel or doxorubicin.

Women enrolled in this study were of a similar age to women with dMMR/MSI-H unresectable or mEC in KEYNOTE-158 (median 64 versus 65 years of age, respectively). At baseline, patient performance status in the McMeekin et al. study was not directly comparable with performance status from KEYNOTE-158; McMeekin et al. reported Karnofsky status and KEYNOTE-158 reported Eastern Cooperative Oncology Group (ECOG) scores^18,27. However, we note that women were included in the study reported by McMeekin et al. if they had locally advanced, recurrent, or metastatic disease²⁷. This contrasts with KEYNOTE-158, where women were enrolled with unresectable or metastatic disease only.

Data from all studies identified from the focused search were used in scenario analyses to test the robustness of the model; however, we note that McMeekin et al. reported the largest median OS of all the studies identified, and evaluated a similarly large number of patients as Miller et al. (248 patients in total)^27,28.

The modelled and observed OS and PFS for pembrolizumab and chemotherapy are presented in Figure 2; and are described in further detail in the sections below.

Figure 2. Modelled overall survival and progression free survival curves for pembrolizumab and chemotherapy. Abbreviations. KM, Kaplan–Meier; OS, overall survival; PFS, progression-free survival.

Overall survival

OS data for pembrolizumab were obtained from KEYNOTE-158 for women with dMMR/MSI-H unresectable or mEC. The cumulative hazard plot showed a change in hazard at Week 40, i.e. there was a different slope before and after around 40 weeks for pembrolizumab. Consequently, we considered that a piecewise model may be a more appropriate fit to the data and, alongside a range of one-piece parametric extrapolations (exponential, Weibull, log-normal, log-logistic, Gompertz, generalized gamma), we explored the fit of two-piece parametric extrapolations using Kaplan–Meier data to Week 40 and parametric extrapolation thereafter.

The fit of each distribution was assessed visually against Kaplan–Meier data, statistical fit was assessed using the Akaike information criterion (AIC) and the Bayesian information criterion (BIC), and final selections were validated by two clinicians. The two-piece parametric extrapolations were considered to fit the data well, in particular, given the strong durability of response evident in the data, the two-piece Weibull and exponential produced the most clinically plausible extrapolations of the observed OS from KEYNOTE-158, and remained below the general population survival curve. The two-piece exponential curve was noted to be the most conservative option and was therefore selected for use in the base case. The selected OS curve is presented in Figure 2. This selection was tested in sensitivity analysis. Visual and statistical fits are presented in the Supplementary Appendix.

OS data for chemotherapy were estimated by fitting one-piece parametric distributions to digitized Kaplan–Meier curves from the study by McMeekin et al. As no noticeable change in the rate of cumulative hazards over time was observed for this study, a two-piece approach was not explored. Instead, the log-normal parametric extrapolation was selected as the best fitting curve based on visual and statistical fit and was used in the base case; this choice was verified by two clinicians.

OS data for the general population was also included within the model and estimates of OS for pembrolizumab and chemotherapy were assumed not to exceed these³².

Progression-free survival

Similar to the approach taken for OS, PFS Kaplan–Meier data for pembrolizumab were obtained from KEYNOTE-158 for women with dMMR/MSI-H unresectable or mEC and extrapolated using one- and two-piece approaches. The cumulative hazard plot showed a change in hazard at Week 10 which coincided with the time of the first radiographic assessment in KEYNOTE-158. Thus, the two-piece approaches used Kaplan–Meier data to Week 10 and parametric extrapolation thereafter. As with OS, the fit of each distribution for PFS was assessed visually and statistically, with final selections validated by two clinicians. The two-piece parametric extrapolations were considered to fit the data well and the Weibull distribution was selected for use in the base case and was assumed not to exceed OS. The selected PFS curve is presented in Figure 2. This selection was tested in sensitivity analysis. Visual and statistical fits are presented in the Supplementary Appendix.

PFS data for chemotherapy were obtained from McMeekin et al. using the reported median PFS (4 months) and applying an exponential distribution to estimate long-term survival²⁷. No Kaplan–Meier data for PFS were reported in the study, so it was not possible to digitize these data and test a range of parametric distributions. However, PFS estimates for chemotherapy were tested extensively in scenario analysis using PFS reported and extrapolated from alternative data sources^28–31.

Adverse event management

Adverse events were included in the model as reported in the clinical trials for pembrolizumab and chemotherapy if they were Grade 3 or above and expected to require significant medical intervention or hospitalization. Data for pembrolizumab were obtained from KEYNOTE-158 for women with dMMR/MSI-H unresectable or mEC and, in the base case, adverse event data for chemotherapy were obtained from McMeekin et al. to be consistent with the source of data for PFS and OS^18,27. Adverse event costs were sourced from the Agency for Healthcare Quality and Research or from literature and inflated to 2019 values³³. These costs were multiplied by the estimated adverse event incidence and adjusted using the within trial duration for long-term costs for fulminant type 1 diabetes and acute respiratory distress syndrome, with the resulting total cost of managing adverse events, per patient, was applied in the model for pembrolizumab ($4,041) and chemotherapy ($6,271) as a one-off cost in the first cycle (Table 3). The larger cost associated with chemotherapy is the result of the greater toxicity of chemotherapy compared to pembrolizumab.

Table 3. Adverse event incidence and cost.

Adverse event	Adverse event incidence	Cost per adverse event	Source of cost data^33
Pembrolizumab from KEYNOTE-158
Fulminant Type 1 diabetes mellitus	2.04%	$21,730	Annual cost from Simeone 2015 adjusted for adverse event duration in KEYNOTE-158 (828 days)^47
Hyperglycemia	2.04%	$7,003	Hyperglycemia – unspecified (Ref ICD-10 R73.9)^33
Hypophosphatemia	2.04%	$7,337	Other disorders of phosphorous metabolism (Ref ICD-10 E83.39)^33
Lymphocyte count decreased	4.08%	$6,518	Assumption: decreased white blood cell count, unspecified (Ref ICD-10 D72.819)^33
Neutrophil count decreased	2.04%	$13,300	Neutropenia – unspecified (Ref ICD-10 D70.9)^33
Transaminases increased	2.04%	$7,334	Non-specific elevation of levels of transaminase and lactic acid dehydrogenase (Ref ICD-10 R74.0)^33
White blood cell count decreased	2.04%	$6,518	Decreased white blood cell count, unspecified (Ref ICD-10 D72.819)^33
Acute respiratory distress syndrome	2.04%	$78,778	Hospital costs associated with ARDS, Bice 2020^48
Drug-induced liver injury	2.04%	$9,601	Toxic liver disease, unspecified (Ref ICD-10 K71.9)^33
Enterocolitis	2.04%	$17,877	Indeterminate colitis (Ref ICD-10 K52.3)^33
Primary adrenal insufficiency	2.04%	$9,523	Primary adrenocortical insufficiency (Ref ICD-10 E27.1)^33
Rash generalized	2.04%	$5,971	Rash and other non-specific skin eruption (Ref ICD-10 R21)^33
Total cost of treating pembrolizumab AEs		$4,041
Chemotherapy from McMeekin 2015
Leukopenia	43.0%	$6,518	Decreased white blood cell count, unspecified (Ref ICD-10 Code D72.819)^33
Neutropenia	13.0%	$13,300	Neutropenia, unspecified (Ref ICD-10 Code D70.9)^33
Cardiovascular – ejection fraction	2.0%	$13,500	Combined systolic and diastolic heart failure (Ref ICD10 Code I50.4)^33
Febrile neutropenia	5.0%	$13,802	Other neutropenia (Ref ICD-10 Code D70.8)^33
Neurotoxicity	8.0%	$9,733	Adverse effect of antineoplastic and immunosuppressive drugs, initial encounter (Ref ICD-10 Code T45.1X5A)^33
Total cost of treating chemotherapy AEs		$6,271

Abbreviation. AE, adverse event.

Utilities

Utilities were included in the model to estimate QALYs for pembrolizumab and chemotherapy. For PFS and progressed disease, the utilities were derived from EQ-5Dⁱⁱⁱ scores collected from the population of patients with unresectable or mEC and dMMR/MSI-H in KEYNOTE-158, using a published, validated US-specific algorithm¹⁸. The estimated utility values for PFS and progressed disease were 0.817 (95% confidence interval [CI] 0.797, 0.836) and 0.779 (95% CI 0.699, 0.859), respectively. Disutilities related to the loss of utility associated with experiencing an adverse event were considered in the analysis³⁴. A disutility associated with adverse events was not incorporated in the model in the base case based to a small sample of patients reporting EQ-5D values while experiencing adverse events (n = 4) with associated uncertainty. This exclusion was conservative since patients in the chemotherapy arm were more likely to experience toxicities; however, the impact of including adverse event disutility was explored in a scenario analysis.

Resource use and costs

Testing for dMMR/MSI-H status

In the model, to identify women eligible for treatment with pembrolizumab, a cost of testing for dMMR/MSI-H was applied to patients receiving pembrolizumab in the first cycle since women are eligible for pembrolizumab if they are identified as having solid tumors with dMMR/MSI-H which have progressed following prior treatment. This cost was not applied to chemotherapy since eligibility for treatment is not related to being identified as dMMR/MSI-H and would not be carried out in routine clinical practice.

The cost of dMMR testing (immunohistochemistry, $389.76) and MSI-H testing (polymerase chain reaction, $348.56) were included in the base case and obtained from the CMS Physician Fee Schedule and the CMS Laboratory Fee Schedule^35,36. Of those tested, the majority (67%) were tested for both MSI-H and dMMR, with 22% patients were assumed to receive testing for MSI-H only, 11% for dMMR only, which equated to a testing cost per person tested of $614.23. The assumptions reflect testing practice in the US for 2018/2019, based on market research [data on file]³⁷. Assuming a prevalence of MSI-H of 14.52% for unresectable or mEC based on data reported for late stage endometrial cancer from Le et al.¹⁶, seven women would need to be tested to identify one patient eligible for treatment with pembrolizumab. Based on this, the one-off cost of testing for each patient identified as eligible to be treated with pembrolizumab was $4,230. This cost was applied in the first cycle of the model as a one-off cost.

Time on treatment

In the base case, Kaplan–Meier data for pembrolizumab time on treatment (ToT) were obtained from KEYNOTE-158 for women with dMMR/MSI-H unresectable or mEC and applied directly in the model, since these data were mature¹⁸. ToT was censored at 104 weeks to align with the maximum treatment duration for pembrolizumab and, as patients were assumed to discontinue treatment upon progression, ToT was not permitted to exceed PFS. Parametric curves were fitted to the ToT data in sensitivity analysis.

ToT data for paclitaxel and doxorubicin were not reported in McMeekin et al. In the absence of data for ToT, PFS was used as a proxy. Paclitaxel and doxorubicin were administered up to a cumulative dose of 500 mg/m², in line with FDA labelling and consistent with the dosing in the McMeekin study^27,38,39.

This may have resulted in an overestimate of ToT for chemotherapy since a proportion of patients will discontinue treatment before progression due to adverse events or other reasons. This, in turn, may have resulted in an overestimate of cost for chemotherapy in the model. However, in the absence of data, this was considered to be a reasonable simplification and is not expected to impact results since the cost of paclitaxel and doxorubicin are low.

Drug costs

In the base case, the unit costs of pembrolizumab, paclitaxel and doxorubicin were obtained from the IBM Micromedex RED BOOK^iv, a US database providing drug prices for over 300,000 prescription and over-the-counter pharmaceuticals, chemicals used for compounding, and medical devices and supplies based on the wholesale acquisition cost (WAC)⁴⁰.

Drug costs were estimated for each of the regimens modelled (pembrolizumab, paclitaxel, and doxorubicin) and perfect vial sharing was assumed in the base case, i.e. no wastage was costed (Table 4). To estimate drug costs for paclitaxel and doxorubicin, assumptions around patient body surface area (BSA) were required, since both paclitaxel and doxorubicin are administered on a per m² basis. These data were obtained from KEYNOTE-158 for women with dMMR/MSI-H unresectable or mEC; KEYNOTE-158 did not provide BSA data, but mean BSA was calculated from weight and height using the Du Bois formula^18,41. The average BSA was estimated to be 1.76 m². The cost of paclitaxel and doxorubicin applied in the model was weighted based on the proportions observed in McMeekin et al., which were 72% for doxorubicin and 28% for paclitaxel²⁷.

Table 4. Dosing regimens and estimated costs for pembrolizumab, paclitaxel and doxorubicin.

Regimen [ref]	Dose per administration	Dosing schedule	Total mg or capsules needed	Dose per format (smallest cost per mg)	Price per vial^40	Total cost per administration
Pembrolizumab^26	200 mg	Every 3 weeks	200 mg	100 mg	$4,862.04	$9,724.08
Doxorubicin^27	60 mg/m^2	Every 3 weeks for up to 8 cycles	106 mg	200 mg	$64.40	$34.07
Paclitaxel^27	80 mg/m^2	Once weekly 3 weeks out of 4	141 mg	300 mg	$34.15	$16.06

In addition to the cost of pembrolizumab and chemotherapy, a cost associated with subsequent therapies was included in the model and applied at progression. For women with dMMR/MSI-H unresectable or mEC in KEYNOTE-158, only six patients received subsequent treatments (10.2%). The subsequent treatments most commonly received were pembrolizumab (n = 2) and tamoxifen (n = 2). Given that patients receiving pembrolizumab who progress are unlikely to continue treatment with pembrolizumab, and given the paucity of data to reliably inform subsequent treatment, it was assumed that all patients in the model who receive subsequent therapy (10.2%) received tamoxifen as subsequent treatment for 140 days (20 weeks); 140 days was the average treatment duration of subsequent therapies reported for the five dMMR/MSI-H unresectable or mEC patients receiving these therapies in KEYNOTE-158. In addition, pembrolizumab was not available to patients at the time of conduct of the chemotherapy trials. The cost of treatment was applied as a one-off cost at the point of progression and was estimated to be $93.36 based on an oral dose of 20 mg per day, and on acquisition costs reported in the IBM Micromedex RED BOOK^40,42.

Administration costs

The cost of administering IV treatments was estimated to be $309.60 for the first hour of infusion (CPT 96413) and $60.47 for each additional hour (CPT 96417), based on cost data obtained from the CMS Hospital Outpatient Prospective Payment System^36,43. Given assumptions made around duration of infusion for pembrolizumab and chemotherapy, the total cost per administration was estimated to be $154.80 for pembrolizumab (30 min infusion), $430.54 for paclitaxel (180 min infusion) and $51.60 for doxorubicin (10 min infusion).

Disease management costs

A focused literature review was carried out to identify studies reporting costs associated with the management of progression-free and progressed disease for women with dMMR/MSI-H unresectable or mEC in the US; however, no relevant studies were identified. Consequently, a micro-costing approach was undertaken to estimate these costs. It was assumed that patients would require one monthly outpatient visit and one computed tomography (CT) scan every quarter. These frequencies were obtained from a review of the management of ovarian cancer by the Institute for Clinical and Economic Review, with costs obtained from the Medicare Physician Fee Schedule^36,44. The resulting estimated weekly cost of PFS and progressed disease was $83.05, and this cost was applied at each cycle for patients remaining in the PFS or progressed disease health states.

End of life costs were sourced from a study of healthcare utilization and hospital expenditures for patients in the final 30 days of life in the US and inflated to current prices⁴⁵. This cost was estimated to be $10,384 and was applied in the model at the point of death.

Approach to sensitivity analysis

Deterministic and probabilistic sensitivity analyses were carried out to test the robustness of model results. Deterministic analyses present results where model parameter input values are based on their mean estimate. By contrast, probabilistic analysis involves the simultaneous varying of all model input values according to an assigned distribution. In this analysis, model results are recorded for 5,000 iterations and mean costs and QALYs are estimated using results from all iterations. The distributions assigned to each model input are described in Table 1 for each variable included within the probabilistic sensitivity analysis. The choice of distribution was made based on guidance from Briggs et al.⁴⁶ Results of the probabilistic sensitivity analyses were presented on a cost-effectiveness plane and cost-effectiveness acceptability curve.

A series of deterministic sensitivity analyses were carried out to test the robustness of model results to changes in parameters and structural assumptions. A one-way sensitivity analysis tested the impact of varying model inputs independently between their upper and lower bounds; the inputs varied and their upper and lower values are presented in Table 1. Results of the one-way sensitivity analysis were presented on a tornado diagram. A range of scenario analyses were also carried out to test model assumptions; the scenarios considered are summarized in Table 5 and include alternative options for extrapolating OS, PFS and ToT for pembrolizumab and chemotherapy relaxing the assumption of perfect vial sharing, including dis-utilities associated with adverse events.

Table 5. Scenario analyses.

Scenario	Comments
Alternative options for extrapolating overall survival for pembrolizumab	Including one-piece (six standard parametric distributions) and two-piece options (Kaplan-Meier data until Week 40 and the remaining five standard parametric distributions)
Alternative options for extrapolating overall survival for chemotherapy	Including alternative one-piece extrapolations using the remaining five standard parametric distributions for McMeekin and using data from Makker et al. De Legge et al., Lincoln et al. and Miller et al.^28–31
Alternative options for extrapolating progression free survival for pembrolizumab	Including one-piece (six standard parametric distributions) and two-piece options (Kaplan-Meier data until Week 10 and the remaining five standard parametric distributions)
Alternative options for extrapolating progression free survival for chemotherapy	Including one-piece extrapolations using data from Makker et al. De Legge et al., Lincoln et al. and Miller et al.^28–31
Alternative options for extrapolating time on treatment for pembrolizumab	Including one-piece (six standard parametric distributions) options
Alternative time horizons	Using 10, 20 and 35 years as alternative model time horizon
No vial sharing	Investigating the impact of including drug wastage within the analysis
Disutility associated with adverse events	To account for treatment-related adverse events, a QALY decrement of 0.002 is applied to chemotherapy, and a QALY decrement of 0.001 is applied to pembrolizumab.
Dose intensity for pembrolizumab	In the base case it is assumed that 100% of pembrolizumab dose is administered. In this scenario we test a dose intensity of 97.4% based upon results from KEYNOTE-158.
Alternative drug costs	Included drug costs from Medicare perspective using Average Sales Price
dMMR testing costs	Exclude dMMR testing costs from the analysis
End of life costs	Exclude end-of-life costs
Utilities for PFS	Utilities for progression-free survival assumed to apply for progressed disease health state
Utilities for PD	Utilities for progressed disease assimed to apply for progression-free survival health state
CT scan frequency	Apply cost of CT scans monthly

Abbreviation. QALY, quality-adjusted life year.

Results

Base case deterministic results

Base case deterministic results are presented in Table 6. Pembrolizumab was associated with an additional 4.68 life years and an additional 3.80 QALYs versus chemotherapy over the time horizon of the analysis (30 years). The majority of these gains were accrued in the progression-free health state, with an additional 3.33 QALYs for women treated with pembrolizumab versus chemotherapy, reflecting the additional time spent in that health state for women treated with pembrolizumab.

Table 6. Base case deterministic, discounted results.

(a) Incremental cost per LY and QALY gained
	Total			Incremental			ICER (incremental cost per LY)	ICER (incremental cost per QALY)
	Costs	LYs	QALYs	Costs	LYs	QALYs
Pembrolizumab	$246,350	6.21	5.01
Chemotherapy	$25,416	1.53	1.21	$220,934	4.68	3.80	$47,236	$58,165
(b) Disaggregated costs
Breakdown of cost (discounted)	Pembrolizumab	Chemotherapy
Drug acquisition	$199,645	$219
Drug administration	$3,178	$2,361
Testing costs	$4,230	$0
Adverse events	$4,041	$6,271
Progression free survival	$19,608	$1,955
Progressed disease	$7,315	$4,700
Death	$8,333	$9,911
Total	$246,350	$25,416
(c) Disaggregated QALYs
Breakdown of QALYs (discounted)	Pembrolizumab	Chemotherapy
Progression free survival	3.70	0.37
Progressed disease	1.31	0.84
Total	5.01	1.21

Abbreviations. ICER, incremental cost-effectiveness ratio; LY, life years; QALY, quality-adjusted life year.

Abbreviation. QALY, quality-adjusted life years.

Pembrolizumab was associated with an incremental cost of $220,934 versus chemotherapy over the time horizon, with the majority of cost accrued for drug acquisition ($119,426). Combined with the incremental gain in LYs and QALYs, the estimated incremental cost per LY gained was $47,236 and per QALY gained (ICER) was $58,165 for pembrolizumab versus chemotherapy. At a willingness-to-pay of $100,000 per QALY gained, this resulted in a net monetary benefit (i.e. the value of pembrolizumab in monetary terms) of $158,907.

Probabilistic sensitivity analysis

The results of the probabilistic sensitivity analysis are presented in Figure 3. The estimated incremental cost per LY gained was $46,334 and per QALY gained for was $57,668 for pembrolizumab versus chemotherapy for the probabilistic sensitivity analysis. We note that the results of the deterministic and probabilistic analyses are similar. The probability of pembrolizumab being considered cost-effective versus chemotherapy at a willingness-to-pay of $100,000 per QALY gained was 90.1%. The probabilistic analysis indicated the main uncertainty was related to QALY gains, with all iterations indicating a positive incremental QALY gain.

Figure 3. Results of the probabilistic sensitivity analysis. (a) Cost-effectiveness plane. (b) Cost-effectiveness acceptability curve. Abbreviations. CE, cost-effectiveness; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life-years; WTP, willingness-to-pay.

One-way sensitivity analysis and scenario analyses

Deterministic results were tested in one-way sensitivity analysis and scenario analyses and were found to be robust to all changes explored (Figure 4; Table 7). In one-way sensitivity analysis, the model inputs which had the largest impact on the ICER were the price of pembrolizumab and the discount rates for costs and outcomes, with a maximum estimated ICER of $78,768.

Figure 4. Tornado diagram.

Table 7. Results of the scenario analyses.

Scenario	Incremental cost	Incremental QALYs	ICER
Base Case	$220,934	3.80	$58,165
Time horizon = 35 years	$221,198	3.83	$57,746
Time horizon = 20 years	$218,996	3.51	$62,341
Time horizon = 10 years	$211,783	2.43	$87,168
Include AE related disutility	$220,934	3.86	$57,191
Include drug wastage	$220,720	3.80	$58,109
Pembrolizumab dose intensity at 97.4%	$215,743	3.80	$56,798
Progression free survival for pembrolizumab
One-piece – Exponential	$228,610	3.71	$61,609
One-piece – Gompertz	$221,734	3.84	$57,786
One-piece – Log logistic	$226,207	3.79	$59,701
One-piece – Lognormal	$226,865	3.79	$59,862
One-piece – Weibull	$227,721	3.75	$60,667
One-piece – GenGamma	$221,621	3.83	$57,897
Two-piece – Gompertz	$226,121	3.73	$60,613
Two-piece – Log logistic	$220,245	3.84	$57,425
Two-piece – Log normal	$220,285	3.82	$57,703
Two-piece – Weibull	$219,351	3.82	$57,352
Two-piece – Generalized gamma	$219,818	3.82	$57,534
Progression free survival for chemotherapy
Makker 2013 – Exponential	$221,657	3.80	$58,273
Makker 2013 – GenGamma	$221,637	3.80	$58,331
Makker 2013 – Gompertz	$221,619	3.79	$58,429
Makker 2013 – Loglogistic	$221,881	3.80	$58,318
Makker 2013 – Lognormal	$221,756	3.80	$58,291
Makker 2013 – Weibull	$221,657	3.80	$58,273
Li Legge 2011 – Exponential	$221,929	3.81	$58,315
Lincoln 2003 – Exponential	$220,664	3.80	$58,126
Miller 2018 – Exponential	$220,395	3.79	$58,089
Time on treatment for pembrolizumab
One-piece – Exponential	$195,277	3.80	$51,410
One-piece – GenGamma	$187,465	3.80	$49,353
One-piece – Gompertz	$206,394	3.80	$54,337
One-piece – Loglogistic	$201,823	3.80	$53,133
One-piece – Lognormal	$189,325	3.80	$49,843
One-piece – Weibull	$194,893	3.80	$51,309
Overall survival for pembrolizumab
One-piece – Exponential	$213,882	2.38	$89,962
One-piece – Gompertz	$235,459	6.69	$35,199
One-piece – Loglogistic	$224,955	4.61	$48,779
One-piece – Lognormal	$226,629	4.94	$45,847
One-piece – Weibull	$220,721	3.76	$58,628
One-piece – GenGamma	$231,822	5.97	$38,822
Two-piece – Gompertz	$235,211	6.64	$35,422
Two-piece – Loglogistic	$229,503	5.51	$41,644
Two-piece – Lognormal	$231,542	5.92	$39,143
Two-piece – Weibull	$227,016	5.01	$45,272
Two-piece – GenGamma	$235,248	6.65	$35,388
Overall survival for chemotherapy
McMeekin 2015 – Exponential	$221,590	3.93	$56,444
McMeekin 2015 – GenGamma	$221,183	3.85	$57,497
McMeekin 2015 – Gompertz	$222,375	4.08	$54,535
McMeekin 2015 – Loglogistic	$220,843	3.78	$58,431
McMeekin 2015 – Weibull	$222,425	4.09	$54,417
Makker 2013 – Exponential	$223,156	4.23	$52,773
Makker 2013 – GenGamma	$222,475	4.10	$54,318
Makker 2013 – Gompertz	$221,666	3.93	$56,342
Makker 2013 – Loglogistic	$221,494	3.90	$56,754
Makker 2013 – Lognormal	$222,321	4.07	$54,686
Makker 2013 – Weibull	$223,083	4.21	$52,933
De Legge 2011 – Exponential	$224,464	4.48	$50,089
Lincoln 2003 – Exponential	$222,263	4.06	$54,799
Miller 2018 – Exponential	$222,038	4.01	$55,337
Resource use
Drug costs based on ASP	$227,792	3.80	$59,970
Exclude dMMR testing	$216,704	3.80	$57,051
Exclude end of life costs	$222,512	3.80	$58,580
Exclude AE costs	$223,164	3.80	$58,752
Monthly CT scans	$208,974	3.80	$55,016
Utilities
Assume utilities for PFS	$220,934	3.82	$57,816
Assume utilities for PD	$220,934	3.64	$60,636

Abbreviations. AE, adverse event; ASP, average sales price; CT, computed tomography; ICER, incremental cost-effectiveness ratio; PD, progressed disease; PFS, progression-free survival; QALY, quality-adjusted life year.

In scenario analyses, all scenarios explored reported ICERs below $100,000 per QALY threshold. The three scenarios with the highest ICERs were overall survival using the one-piece exponential approach for pembrolizumab ($89,962) and when the time horizon was reduced to 20 or 10 years ($62,341 and $87,168 respectively). At 20 years, a significant proportion of patients receiving pembrolizumab remain alive in the model (approximately 10%) and therefore these time horizons were considered implausible. Including adverse event disutilities resulted in and ICER of $57,191, which was lower than the base case. The lowest ICERs reported in scenario analysis were pembrolizumab overall survival projections with the lowest ICER $35,199 reported for one-piece Gompertz.

Discussion

For previously treated dMMR/MSI-H unresectable or mEC, results of this analysis have demonstrated that pembrolizumab is a highly cost-effective treatment option when compared with the single chemotherapy agents paclitaxel or doxorubicin, and these results are robust to changes in model parameters or assumptions.

Pembrolizumab demonstrates a considerable gain in life years (4.68) and QALYs (3.80) versus chemotherapy over the time horizon of the analysis (30 years). These gains are due to improved PFS and OS for pembrolizumab versus chemotherapy, with the majority of gains accrued in the progression-free health state (3.33 of 3.80 incremental QALYs).

The demonstrated substantial lifetime benefits of treatment with pembrolizumab versus chemotherapy are accompanied by an increase in costs, driven largely by the increased treatment cost of pembrolizumab, resulting in an ICER of $58,165 for the base case deterministic analysis. Probabilistic results are supportive, with an estimated ICER of $57,668 and the probability of pembrolizumab being considered cost-effective versus chemotherapy at a willingness-to-pay of $100,000 per QALY gained is 90.1%.

This analysis has several strengths. The model structure was chosen to be consistent with other cost-effectiveness analyses in advanced or metastatic cancer and was constructed with close reference to the Institute for Clinical and Economic Review reference case for economic evaluation²². Furthermore, the analysis was underpinned by data obtained from reliable sources, in particular for the PFS and OS data incorporated in the model for pembrolizumab. These data were obtained from a Phase II clinical trial for pembrolizumab which had completed almost three years of follow-up. Extrapolations beyond the observed data were reviewed and validated by two clinical experts. In addition, EQ-5D data describing utilities for PFS and progressed disease were obtained from patients with dMMR/MSI-H unresectable or mEC within KEYNOTE-158 and valued using a US tariff, and standard US sources for costs were used within the analysis^33,35,40.

We note that the results of the analysis may be conservative. Firstly, adverse event disutility was not included in the model base case due to small patient numbers informing the disutility values; however, pembrolizumab is associated with a good tolerability profile and including these disutilities reduces the deterministic ICER to $57,191. Moreover, there is some evidence to suggest that MSI-H could be a negative prognostic factor in advanced disease; consequently, given that chemotherapy data were obtained from women with unknown MSI-H status, and noting that some patients within McMeekin et al. had locally advanced or recurrent endometrial cancer, it is possible that the survival data obtained from McMeekin et al. overstate PFS and OS in these patients²⁷.

The limitations of this analysis relate to areas where data are unavailable or unclear:

• The OS and PFS data in the model for pembrolizumab were obtained from a single-arm trial, and no matched-adjusted indirect comparison was carried out to assess a relative treatment effect versus paclitaxel or doxorubicin. Such analyses were not possible since individual patient data were not available from McMeekin et al.²⁷ Instead, chemotherapy data were included in the model from a published source without adjustment. This was necessary given the lack of published comparative data within this patient group. However, it is possible that inclusion of these data in this way could overestimate survival for chemotherapy since a major difference in patient characteristics between KEYNOTE-158 and McMeekin et al. was that patients within McMeekin et al. did not have confirmed MSI-H endometrial cancer, and that McMeekin et al. included patients with locally advanced or recurrent disease²⁷ rather than only unresectable or mEC patients, as for KEYNOTE-158.

• Further to this, we note that there is some evidence to suggest that MSI-H could be a negative prognostic factor in advanced disease; consequently, it is possible that the survival data obtained from McMeekin et al. overestimate PFS and OS in these patients^11,27.

• Survival data for pembrolizumab were based on small patient numbers given that the data were obtained from a small subset of patients in the larger KEYNOTE-158 clinical trial who had MSI-H expression and had endometrial cancer (n = 49). However, we note that the data obtained from these patients reflect the patient group of interest, and results are robust to changes in modelled OS and PFS.

• Chemotherapy in the treatment of unresectable or mEC varies, and we have only compared with paclitaxel or doxorubicin in this analysis. Future analyses could consider comparing additional chemotherapies where data allow; however, we consider paclitaxel and doxorubicin to be important comparators for this patient group based upon published guidelines^9,10.

While the analysis has limitations, we have mitigated these factors where possible via clinical validation of the survival analyses. We also extensively tested the model in sensitivity analysis and note that the results were robust to the changes investigated.

Conclusions

The results of this analysis suggest that, from a US healthcare perspective, pembrolizumab is a highly cost-effective treatment option compared with the chemotherapy single agents paclitaxel and doxorubicin for women with previously treated dMMR/MSI-H unresectable or mEC. Results are robust to changes in model parameters and assumptions.

Future analyses should consider incorporating comparative survival data for pembrolizumab versus chemotherapy within a dMMR/MSI-H unresectable or mEC patient group, using this model structure. Finally, it would be beneficial to capture real-world data to inform disease management costs and testing rates.

Transparency

Declaration of funding

This study was supported by funding from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Declaration of financial/other relationships

Elizabeth Thurgar, Suzette Matthijsse, and Mark Gouldson are employees of BresMed Health Solutions, which received consultancy fees from Merck and Co. in connection with this study. Mayur Amonkar, Patricia Marinello, Navneet Upadhyay, Chizoba Nwankwo and Raquel Aguiar-Ibáñez are employees of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

A peer reviewer on this manuscript has disclosed that they have received research grants from Intuitive Surgical and Hope Against Cancer, and an honorarium from GlaxoSmithKline. Another reviewer has disclosed that they have received research funding from Merck. The peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Author contributions

Raquel Aguiar-Ibáñez, Suzette Matthijsse, Mark Gouldson and Elizabeth Thurgar were involved in the development of the economic model, which was reviewed by all named authors. Elizabeth Thurgar developed the text for this manuscript, which was reviewed and edited by all named authors.

Acknowledgements

The authors would like to thank the following individuals for their contributions to the analyses of clinical, safety and quality-of-life data from KEYNOTE-158 to populate the model: Ruifeng Xu, Principal Scientist Outcomes Research; Qian Wang, Principal Scientist Statistical Programming; Amit Bhaniani, Sr. Scientist, Stat. Programming; and Li Ma, Scientist, Statistical Programming, and to Robert Orlowski, Sr. Prin. Scientist, Clinical Research Oncology Clinical Development, for providing clinical expertise regarding the validity of long-term projections used within the model.

Notes

i NCCN Guidelines is a registered trademark of National Cancer Comprehensive Network, Plymouth Meeting, PA, USA.

ii Microsoft Excel is a registered trademark of Microsoft Corporation, Redmond, Washington, USA.

iii EQ-5D is a registered trademark of EuroQol, Rotterdam, The Netherlands.

iv Micromedex RED BOOK is a registered trademark of IBM, Armonk, NY, USA.