Cost-effectiveness of ceritinib in previously untreated anaplastic lymphoma kinase-positive metastatic non-small cell lung cancer in the United States

Abstract

Aims: To assess the cost-effectiveness of first-line ceritinib vs crizotinib and platinum doublet chemotherapy for anaplastic lymphoma kinase (ALK)-positive metastatic non-small cell lung cancer (NSCLC) from a US third-party payer’s perspective.

Materials and methods: A partitioned survival model with three health states (stable disease, progressive disease, death) was developed over a 20-year time horizon. Ceritinib’s efficacy inputs (progression-free and overall survival) were estimated from ASCEND-4; parametric survival models extrapolated data beyond the trial period. The relative efficacy of ceritinib vs chemotherapy was obtained from ASCEND-4, the relative efficacy of ceritinib vs crizotinib was estimated using a matching-adjusted indirect comparison based on ASCEND-4 and PROFILE 1014. Drug acquisition, treatment administration, adverse event management, and medical costs were obtained from publicly available databases and the literature, and inflated to 2016 US dollars. Treatment-specific stable-state utilities were derived from trials and progressive-state utility from the literature. Incremental costs per quality-adjusted life year (QALY) were estimated for ceritinib vs each comparator. Cost-effectiveness was assessed based on US willingness-to-pay thresholds. Deterministic and probabilistic sensitivity analyses were performed to test model robustness.

Results: In the base case, first-line ceritinib was associated with total direct costs of $299,777 and 3.28 QALYs (from 4.61 life years gained [LYG]) over 20 years. First-line crizotinib and chemotherapy were associated with 2.73 and 2.41 QALYs, 3.92 and 3.53 LYG, and $263,172 and $228,184 total direct costs, respectively. The incremental cost per QALY gained was $66,064 for ceritinib vs crizotinib and $81,645 for ceritinib vs chemotherapy. In the first 2 years following treatment initiation, ceritinib dominated crizotinib by conferring greater health benefits at reduced total costs. Results were robust to deterministic and probabilistic sensitivity analyses.

Limitations: In the absence of head-to-head trials, an indirect comparison method was used.

Conclusions: Ceritinib is cost-effective compared to crizotinib and chemotherapy in the treatment of previously untreated ALK-positive metastatic NCSLC in the US.

Keywords:

• Anaplastic lymphoma kinase-positive
• non-small cell lung cancer
• ceritinib
• crizotinib
• cost-effectiveness

Introduction

Over the past decade, several genomic mutations have been found to underlie the growth and progression of non-small cell lung cancer (NSCLC)¹, prompting the development of targeted therapies for patients harboring specific mutations². The use of targeted therapies has proved effective in improving both progression-free and overall survival rates². Crizotinib was the first targeted therapy approved in the US for the treatment of metastatic NSCLC with rearrangements in the anaplastic lymphoma kinase (ALK) gene, which are estimated to occur in 3–7% of patients with NSCLC³. Since its approval, crizotinib has been widely used in the first-line setting. However, resistance to crizotinib often arises within 1–2 years after treatment initiation, mostly due to compensatory changes in the ALK gene or downstream signaling pathways^4–7. In addition, crizotinib appears to have sub-optimal penetration across the blood–brain barrier, leaving brain metastases largely untreated⁸.

Ceritinib is a next-generation ALK inhibitor that was first approved in the US in 2014 for the treatment of ALK-positive metastatic NSCLC patients who had been previously treated with chemotherapy. In May 2017, ceritinib was also approved in the US for the treatment of previously untreated ALK-positive metastatic NSCLC patients. The approval of ceritinib in first line was based on the results of the Phase 3 trial ASCEND-4⁹, which directly compared ceritinib and platinum doublet chemotherapy (pemetrexed combined with cisplatin or carboplatin, followed by pemetrexed maintenance therapy). In the ASCEND-4 trial, ceritinib provided a statistically significant and clinically meaningful difference in progression-free survival (PFS) compared to platinum doublet (median = 16.6 vs 8.1 months), with a PFS of over 2 years in patients with no baseline brain metastases. Furthermore, the adverse events (AEs) associated with ceritinib are typically manageable, with the most common being diarrhea, nausea, and vomiting¹⁰. Results from the recently concluded ASCEND-8 trial indicate that such gastrointestinal toxicities can be considerably reduced by modifying the dose of ceritinib and by administering ceritinib with food¹⁰.

The cost-effectiveness of ceritinib vs other existing therapies in the treatment of ALK-positive metastatic NSCLC has been previously assessed for the second-line setting and mostly outside the US^11–14. In light of the recent FDA approval of ceritinib in patients with previously untreated NSCLC, this study evaluated the cost-effectiveness of ceritinib vs crizotinib and platinum-based chemotherapy in previously untreated patients with ALK-positive metastatic NSCLC from the perspective of a US third-party payer.

Methods

Model overview

A partitioned survival model was developed in Microsoft Excel 2016 to assess the cost-effectiveness of ceritinib vs crizotinib and platinum doublet with maintenance (i.e. platinum doublet [pemetrexed + cisplatin or pemetrexed + carboplatin] followed by pemetrexed maintenance therapy) as a first-line treatment for adult patients with previously untreated ALK-positive metastatic NSCLC. A partitioned survival model is an economic model that applies clinical end-points (e.g. PFS, overall survival [OS]) in trials to estimate the distribution of patients across different health states over time, based on the area under the survival curves. It is a typical approach in modeling advanced cancers and is common in cost-effectiveness analysis (CEA) studies in NSCLC^15–21. Three mutually exclusive health states were modeled: (1) stable disease; (2) progressive disease; and (3) death (Figure 1). In any given cycle, patients were assumed to be in one of the three health states, with death being the absorbing state.

Figure 1. Partitioned survival model structure.

The target population consisted of adult (i.e. aged 18 years or over) patients with ALK-positive metastatic NSCLC who had not received prior treatment with systemic anti-cancer therapy. The model considered ceritinib 750 mg orally once daily or crizotinib 250 mg orally twice daily or platinum doublet (with maintenance). The platinum doublet treatment was based on the ASCEND-4 trial, in which patients were prescribed pemetrexed 500 mg/m² every 3 weeks in combination with cisplatin (75 mg/m² every 3 weeks, up to 4 cycles) or carboplatin (area under the curve [AUC] 5–6 mg/mL/min every 3 weeks, up to 4 cycles) based on the physician’s choice, followed by pemetrexed maintenance therapy 500 mg/m² every 3 weeks. The distribution of patients on cisplatin vs carboplatin (47% vs. 53%) was based on the proportion of patients receiving each treatment in combination with pemetrexed in the ASCEND-4 trial⁹. Patients were modeled with 1-month cycles over a 20-year time horizon, reflecting a lifetime horizon for this disease population. The model was developed from a US third-party payer’s perspective; as such, only direct costs were considered. In the base case, both costs and effectiveness were discounted at 3.0% annually.

Model inputs

Efficacy inputs

Efficacy inputs included those for PFS and OS. For each treatment arm, the model estimated the amount of time spent in the stable disease, progressive disease, and death states, based on the respective areas under the PFS and OS curves during the 20-year model timeframe. Ceritinib efficacy inputs were derived from the ceritinib arm of the ASCEND-4 trial. For the comparators, the model applied the hazard ratio (HR) of ceritinib vs each comparator to parametric models of ceritinib PFS and OS (Figure 2).

Figure 2. Predicted PFS and OS for ceritinib and comparator arms under HR method. a) The predicted PFS curve for ceritinib was based on exponential functions. PFS curves for comparator treatments were derived by applying HRs to the corresponding curves for ceritinib. b) The predicted OS curve for ceritinib was based on exponential functions. OS curves for comparator treatments were derived by applying HRs to the corresponding curves for ceritinib.

For ceritinib, parametric functions were fitted to patient-level time-to-event data from the ASCEND-4 trial to extrapolate data beyond the trial period⁹. Proportional hazards models based on Weibull, exponential, and Gompertz distributions were applied (Supplementary data). The exponential function was selected as the base case parametric model for both PFS and OS, based on the relative goodness-of-fit criteria and expert opinion on the clinical plausibility of predicted long-term outcomes.

For the comparator arms, the model applied the corresponding HR of each comparator vs ceritinib to parametric models of ceritinib PFS and OS (Supplementary data). In the absence of head-to-head randomized trials of ceritinib vs crizotinib in untreated ALK-positive metastatic NSCLC, efficacy outcomes between two treatments were compared using a matching-adjusted indirect comparison (MAIC) method. The MAIC method is an extension of propensity score weighting, which has long been used in epidemiological studies for adjusted comparisons of non-randomized treatment groups^22,23. Specifically, the MAIC is a pairwise indirect comparison method with adjustment for cross-trial differences in patient characteristics, using patient-level data from the ceritinib arm in ASCEND-4⁹ and published aggregate data for the crizotinib arm from PROFILE 1014²⁴—a Phase 3 trial comparing the efficacy of first-line crizotinib and platinum-based doublet (pemetrexed combined with cisplatin or carboplatin) without pemetrexed maintenance therapy. Ceritinib-treated patients in ASCEND-4 were reweighted to match the baseline characteristics reported for crizotinib patients in PROFILE 1014; the PFS and OS efficacy outcomes were then compared based on the balanced trial populations. HRs for PFS and OS generated from the MAIC of ceritinib vs crizotinib were applied in the CEA model.

For ceritinib vs platinum doublet, the PFS and OS HRs were obtained from ASCEND-4 (PFS: HR (95% confidence interval [CI]) = 0.55 (0.42–0.73); OS = 0.73 (0.50–1.08))⁹. The model did not adjust for post-progression crossover to ceritinib that occurred in the platinum doublet arm of the ASCEND-4 trial. The OS data were immature at the time of the data cut-off (June 24, 2016)⁹, resulting in limited follow-up time available after patients crossed over from the platinum doublet to ceritinib treatment. In the absence of reliable estimates of crossover-adjusted OS inputs, survival of patients in the platinum doublet arm was modeled without adjustment for crossover to second-line ceritinib following disease progression; the model instead accounted for crossover through its consideration of second-line treatment costs, discussed below. The predicted PFS and OS for all treatments are shown in Figures 2(a and b), respectively.

Several sensitivity analyses were conducted varying efficacy inputs. Specifically, these included alternative parametric functions for ceritinib PFS and OS; variations of HRs for each comparator vs ceritinib; and using individually estimated parametric functions of PFS and OS for each treatment arm.

Utility inputs

Health state utility inputs for the base case and sensitivity analyses were derived using the EuroQoL Five Dimensions Questionnaire (EQ-5D) data from ASCEND-4, PROFILE 1014, and additional evidence from literature²⁵ (and unpublished information). Following a systematic literature review, utility values reported in Felip et al.²⁵, Chouaid et al.²⁶, and Nafees et al.²⁷ were extracted and used in the model as base case or scenario analyses utility inputs.

Stable and progressive disease utilities were considered in the model. Stable disease utilities were treatment-specific, and alternative scenario analyses were conducted applying different estimates based on the ASCEND-4 trial data and applications of disutilities from AEs. For progressive disease utility, as EQ-5D scores were not collected systematically after treatment discontinuation in ASCEND-4 or PROFILE 1014, the utility value for progressive disease (0.64) in the base case and scenario analyses was estimated based on the utility study by Chouaid et al.²⁶.

In the base case, stable disease utility values were obtained from the ASCEND-4 trial⁹ for ceritinib and platinum doublet, and from PROFILE 1014, as reported by Felip et al.²⁵, for crizotinib. Treatment-specific stable disease utility values were assumed to incorporate the disutility impact of AEs. As such, AE-related disutilities were not applied in the base case. In sensitivity analyses, two alternative utility scenarios were tested: (1) stable disease treatment non-specific utilities based on ASCEND-4; post-progression utility based on Chouaid et al.²⁶; (2) stable disease and post-progression utilities based on Chouaid et al.²⁶, with adjustment for response rate based on Nafees et al.²⁷. In both scenarios, patients incurred a one-time disutility associated with AEs.

Cost inputs

The model considered the following costs: drug, drug administration, and monitoring costs for initial treatment; costs associated with the management of AEs (AE costs); monitoring costs; progression-related costs (i.e. medical and post-progression treatment costs); and terminal care costs. As treatment initiation with an ALK inhibitor requires that patients test positive for an ALK rearrangement, ALK testing was assumed to be routinely performed under current practice in the US, and was not included as a cost component in the base case.

Drug, drug administration, and monitoring costs for initial treatment. Monthly drug costs were estimated based on unit drug costs, dosing, and relative dose intensity (i.e. proportion of planned dose consumed). Unit drug cost inputs were obtained from ReadyPrice, with price inputs for generic versions of the drugs used when available (Table 1)²⁸. Ceritinib and crizotinib dosing schedules were based on US Food and Drug Administration (FDA) prescribing information^29,30. Platinum doublet dosing schedule was based on the ASCEND-4 treatment arm⁹. Relative dose intensities accounted for the fact that patients may not take the full planned doses due to dose interruption or reduction associated with AEs or non-compliance. Mean relative dose intensities for ceritinib and platinum doublet with maintenance, as well as the dosing schedule for the latter treatment, were based on the ASCEND-4 trial (unpublished information). The relative dose intensity for crizotinib was not reported in PROFILE 1014. As such, an estimate was used from PROFILE 1007, a Phase 3 open-label trial for crizotinib among previously treated patients with ALK-positive NSCLC^24,31.

Table 1. Summary of model inputs.

Model input	Value	Notes/sources
Hazard ratios
PFS: Certinib vs Crizotinib	0.64 (0.47–0.87)	The HRs for ceritinib vs crizotinib was based on the indirect comparison of the ceritinib arm from ASCEND-4^9 vs the crizotinib arm from PROFILE 1014^24, while adjusting for cross-trial differences in patient characteristics. The HRs for ceritinib vs platinum doublet with maintenance were based on the direct comparisons of Blinded Independent Review Committee-assessed (BIRC-assessed) PFS and investigator-assessed OS in the ASCEND-4 trial^9
PFS: Certinib vs Platinum doublet	0.55 (0.42–0.73)
OS: Certinib vs Crizotinib	0.82 (0.54–1.27)
OS: Certinib vs Platinum doublet	0.73 (0.50–1.08)
Unit cost
Ceritinib	$100.53 per pill	The 2016 wholesale acquisition cost (WAC), obtained from ReadyPrice^47, was used to estimate the drug cost. The dosing schedule was obtained from the FDA drug labels
Pemetrexed	$6.41 per mg
Cisplatin	$0.36 per mg
Carboplatin	$0.19 per mg
Crizotinib	$247.43 per pill (based on 200 or 250 mg)
Docetaxel (post-progression treatment)	$9.91 per mg
Drug delivery costs
Intravenous (IV) infusion
Chemo IV push additional drug	$63.16	CMS Physician Fee Schedule^32 (HCPCPS Code: 96411)
Chemo IV infusion 1 hour	$139.61	CMS Physician Fee Schedule^32 (HCPCPS Code: 96413)
Chemo IV infusion additional hour	$28.71	CMS Physician Fee Schedule^32 (HCPCPS Code: 96415)
Intra-muscular injection
Therapeutic, prophylactic, or diagnostic injection	$25.84	CMS Physician Fee Schedule^4^,^32 (HCPCPS Code: 96372)
Monitoring cost (per month)	$315.06	CMS 2017 Physician Fee Schedule^32 and assumption
Medical cost of progression (one-time)	$16,316.89	Fox et al.^38, inflated to 2016 USD
Terminal care cost (one-time)	$17,426.94	Chastek et al.^44, inflated to 2016 USD
Costs associated with AEs (one-time)
Ceritinib	$1,239.83	The list of AEs were extracted from the following sources: ASCEND-4^9 and PROFILE 1014^24; cost inputs for each AE episode were obtained from Elting and Shih^35, HCUP or assumptions
Crizotinib	$2,931.31
Platinum doublet with maintenance	$3,409.69

Monthly drug administration costs consisted of drug delivery costs and pre-medicine costs. Unit costs for intravenous drug delivery were obtained from the 2017 Physician Fee Schedule from the Centers for Medicare & Medicaid Services (CMS) (Table 1)³². No administration costs were assumed for oral drugs. Pemetrexed pre-medicines costs were calculated based on the wholesale acquisition cost package prices, package sizes, and strengths reported by ReadyPrice, while the intravenous dosing schedules were obtained from US FDA prescribing information for pemetrexed and carboplatin^28,33,34.

Duration of treatment was applied to the monthly drug and drug administration costs to estimate total costs. Treatment duration was based on ASCEND-4 for ceritinib and platinum doublet, and PROFILE 1014 for crizotinib^9,24. In the base case, patients were assumed to continue first-line treatment until discontinuation or progression, whichever occurs first. This accounts for patients in real-world settings who may switch to the next line of therapy immediately after disease progression, and may discontinue before progression if there is unacceptable drug toxicity. Under this rule, monthly drug costs in each treatment arm were adjusted with a time-varying proportion of patients on treatment to account for patients who discontinue treatment prior to progression. Specifically, the ratio between the proportion of patients who are both progression-free and on treatment and the proportion in PFS was estimated for each cycle. Both proportions were estimated using exponential functions fitted to patient-level data from the ASCEND-4 trial. For crizotinib, the ratio was assumed to be equal to that of ceritinib. In each treatment arm, the ratio was applied to the predicted PFS curve (as estimated based on the selected parametric function) to estimate the proportion of patients still on treatment in each month. For platinum doublet, no patients were assumed to remain on treatment with cisplatin or carboplatin beyond the maximum allowed number of treatment cycles (i.e. 4 cycles). It should be noted that, while the cycle for platinum doublet was 3 weeks, this was converted into the number of cycles per month when we calculated the drug costs to match the 1-month cycle length of the model.

AE costs. In addition to drug and drug administration costs, patients in the stable disease state incurred costs associated with the management of all-cause AEs. Grade 3 or 4 AEs were assumed to result in a one-time cost during the first cycle. The model considered Grade 3 or 4 AEs that were reported in at least 5% of patients receiving any treatment of interest. The rate of AEs for each comparator was based on the ASCEND-4 trial (for ceritinib and platinum doublet arms) and the PROFILE 1014 trial (for crizotinib arm) (Table 1)^9,24. laboratory abnormality AEs were assumed to incur no costs. The cost of neutropenia was based on Elting and Shih³⁵. All other AE costs were based on the US Department of Health and Human Services 2014 National Inpatient Sample (NIS) of the Healthcare Cost and Utilization Project (HCUP)³⁶. All costs were inflated to 2016 US dollars (USD)³⁷.

Monitoring costs. The model assumed that patients incurred disease monitoring costs in both the stable and progressive disease states. Monitoring costs included healthcare provider visit costs (i.e. outpatient visits), and laboratory test and procedure costs (i.e. hepatic and renal function panels, complete blood count, computerized tomography [CT] scan, and brain magnetic resonance imagings [MRIs]). Unit costs of services were estimated based on Current Procedural Terminology (CPT) reference costs from the CMS 2017 Physician Fee Schedule³². The frequencies of outpatient visits and laboratory tests were assumed to be once every 3 weeks, corresponding to the shortest treatment cycle for the included treatments. The frequency of chest CT scans and brain MRIs was assumed to be once every 3 months to monitor for disease progression and brain metastases. At the month of progression, no monitoring costs were considered in that month; patients incurred one-time progression-related costs instead.

Progression-related and terminal care costs. For patients who progressed on first-line treatment, a one-time medical cost of progression was assigned ($16,316.89), based on Fox et al.³⁸, a retrospective study conducted among 306 stage IIIB or IV patients with metastatic NSCLC, and inflated to 2016 USD³⁷.

To capture costs associated with disease progression, patients with progressive disease were assumed to incur post-progression anti-neoplastic treatment costs (i.e. subsequent-line treatment costs), which were applied once in each model cycle to patients who newly enter the progressive disease state. In the base case, data from the ASCEND-4 and PROFILE 1014 trials were used to estimate the proportions of patients in each treatment arm receiving different post-progression treatments^9,24. It was assumed that the clinical benefit of post-progression treatment was already represented in the efficacy parameters derived from the same trials. Post-progression treatment options included the main post-progression treatments reported in ASCEND-4 and PROFILE 1014 trials: ceritinib, crizotinib, docetaxel, single-agent pemetrexed, and platinum doublet without pemetrexed maintenance. In addition, 20% of patients who progressed on first-line therapy did not receive any further systemic therapy, due to rapid performance deterioration or death.

Post-progression treatment regimen costs were calculated based on monthly drug and drug administration costs, accounting for relative dose intensity and estimated mean duration of treatment corresponding to second-line therapies. Drug and drug administration costs and associated pre-medicines were based on the same sources as those of the first-line treatments^28,32. Dosing for second-line treatments was based on the respective FDA labels^29,30,39,40. As the duration and relative dose intensity of post-progression treatments were not available from the ASCEND-4 and PROFILE 1014 trials, these parameters were collected from clinical trials conducted in ALK-positive or general NSCLC populations^31,41–43.

All patients who transitioned to death were assumed to incur a one-time terminal care cost ($17,426.94), based on Chastek et al.⁴⁴ with inflation adjustment to 2016 USD³⁷.

Model outputs

Base case analyses

Total costs and total effectiveness were estimated for each therapy included in the model. Costs included drug and drug administration costs (initial treatment and post-progression treatment), treatment-associated AE costs, and medical costs (i.e. stable disease medical costs, post-progression medical costs, and terminal care costs). Effectiveness measures included life years (LYs) and quality-adjusted life years (QALYs).

The incremental cost-effectiveness ratios (ICERs) of ceritinib compared with each comparator treatment regimen (crizotinib or chemotherapy) was evaluated using incremental cost per QALY gained and incremental cost per LY gained (LYG).

Sensitivity analyses

Deterministic sensitivity analyses (DSA) and probabilistic sensitivity analyses (PSA) were used to test the robustness of the model. The DSA was conducted by varying one model input or assumption at a time. The inputs that were varied are illustrated in Figures 3 and 4. The PSA was conducted to estimate the probability of ceritinib being cost-effective compared to comparator arms, based on a willingness-to-pay (WTP) threshold of $150,000 USD⁴⁵. A Monte-Carlo simulation with 1,000 iterations was conducted in which model inputs were randomly drawn from the specified distributions at each iteration. Uncertainty in the ceritinib PFS and OS parameter estimates using exponential functions was assumed to follow normal distributions. Uncertainty in the HRs of PFS and OS for platinum doublet or crizotinib vs ceritinib was modeled using log-normal distributions. Log-normal distributions were also assumed for relative dose intensities. Gamma distributions were assumed for costs, while beta distributions were assumed for utilities of health states.

Figure 3. DSA results for ceritinib vs crizotinib. * In this scenario analysis, PFS and OS survival curves for other comparator arms were modeled based on the HRs vs ceritinib, as in the base case. ^ε Ceritinib was dominant over the comparator arm in this sensitivity analysis.

Results

Base case analyses

In the base case analysis over a 20-year time horizon, life years gained were estimated to be 4.61, 3.92, and 3.53 years (Table 2) for patients treated with ceritinib, crizotinib, and platinum doublet, respectively. After adjusting for utility, total QALYs were estimated to be 3.28 for ceritinib, 2.73 for crizotinib, and 2.41 for platinum doublet. Over the same time horizon, total costs were estimated to be $299,777 for ceritinib, $263,172 for crizotinib, and $228,184 for platinum doublet.

Table 2. Base case results.

	Ceritinib	Crizotinib	Platinum doublet	Ceritinib vs Crizotinib	Ceritinib vs Platinum doublet
Costs (2016 USD)
Drug and drug administration costs, initial treatment	$220,601	$181,541	$73,725	$39,060	$146,876
Drug and drug administration costs, post-progression treatment	$30,677	$33,306	$106,829	–$2,629	–$76,152
Treatment associated AE costs	$1,214	$2,838	$3,285	–$1,624	–$2,071
Medical care costs	$47,285	$45,487	$44,345	$1,798	$2,940
Total costs	$299,777	$263,172	$228,184	$36,605	$71,593
Effectiveness
Total QALYs	3.28	2.73	2.41	0.55	0.88
Total LYs	4.61	3.92	3.53	0.69	1.08
ICER (2016 USD)
Incremental cost per QALY gained				$66,064/QALY	$81,645/QALY
Incremental cost per LYG				$53,207/LYG	$66,441/LYG

In the first 2 years following treatment initiation, ceritinib was estimated to confer greater health benefit (more QALYs) at less cost compared to crizotinib (i.e. ceritinib dominated crizotinib in the first 2 years of treatment). The incremental cost per LYG over a 20-year time horizon was estimated to be $53,207 for ceritinib vs crizotinib and $66,441 for ceritinib vs platinum doublet with maintenance. The incremental cost per QALY gained was estimated to be $66,064 and $81,645, respectively, over the same time horizon (Table 2).

Key model drivers included drug and drug administration costs for initial treatment (74%, 69%, and 32%, of ceritinib, crizotinib, and platinum doublet total costs, respectively) and costs associated with post-progression treatment (10%, 13%, and 47%).

Sensitivity analyses

DSA

The DSA results indicate that the incremental cost per QALY gained for ceritinib vs the comparator arms were robust in most scenarios. In the DSA for ceritinib vs crizotinib (Figure 3), the ICER ranged from dominant (i.e. lower cost and higher QALY for ceritinib) to $171,379 per QALY. The results were most sensitive to ceritinib and crizotinib drug cost per month, drug costs based on treatment until discontinuation, the HR of OS for crizotinib vs ceritinib, and a scenario assuming that the relative dose intensity for ceritinib applied to all treatments. Among all DSA tests, 91.7% of the DSA tests for ceritinib vs crizotinib were below a $150,000 WTP threshold.

For ceritinib vs platinum doublet (Figure 4), the ICER ranged from $18,752 per QALY to $174,489 per QALY. The results were most sensitive to ceritinib and crizotinib drug costs per month, drug costs based on treatment until discontinuation, and use of Gompertz functions to model ceritinib PFS or OS. Among all DSA tests, 96.0% of the DSA tests for ceritinib vs platinum doublet were below a $150,000 WTP threshold.

Figure 4. DSA results for ceritinib vs platinum doublet with maintenance. * In this scenario analysis, PFS and OS survival curves for platinum doublet were modeled based on the HRs vs ceritinib, as in the base case.

PSA

In the PSA for ceritinib vs crizotinib, the average ICER in terms of cost per QALY gained was $66,428, and ceritinib was cost-effective at a WTP threshold of $150,000 in 76.0% of cases including dominant in 6.1% of cases (Figure 5). In the PSA for ceritinib vs platinum doublet with maintenance, the average ICER was $82,284 and ceritinib was cost-effective at a WTP threshold of $150,000 in 84.3% of cases (Figure 6). The incremental cost-effectiveness plane for ceritinib vs each comparator and the cost-effectiveness acceptability curve based on the highest net monetary benefit are shown in Supplementary appendices.

Figure 5. Cost-effectiveness acceptability curve for ceritinib vs crizotinib.

Figure 6. Cost-effectiveness acceptability curve for ceritinib vs platinum doublet.

Discussion

Amidst rapid changes in the treatment landscape of ALK-positive NSCLC and following the recent approval of ceritinib for the treatment of ALK-positive metastatic NSCLC in the first-line setting, it is important for payers and other key healthcare shareholders to better understand the cost-effectiveness of available therapies to maximize outcomes and improve survival. Prior CEA studies have focused on patients with ALK-positive NSCLC who had been previously treated, mostly from outside the US^11–14. Therefore, a major strength of this CEA is the inclusion of US patients with ALK-positive metastatic NSCLC who have not been previously treated with anti-neoplastic therapies, as it allowed the cost-effectiveness of ceritinib vs crizotinib and platinum doublet to be assessed in the first-line setting from a US third-party payer perspective.

In the base case analysis over 20 years, ceritinib was associated with a greater health benefit compared to both crizotinib and platinum doublet, as evidenced by a gain of 0.55 QALYs over crizotinib and 0.88 QALYs over platinum doublet. The incremental cost per QALY gained over 20 years was estimated at $66,064 for ceritinib vs crizotinib and $81,645 for ceritinib vs platinum doublet. Notably, when considering ceritinib vs crizotinib in the first 2 years following treatment initiation, ceritinib dominated crizotinib by conferring greater health benefit at less cost. In sensitivity analyses, the cost-effectiveness was found to be robust in most scenarios.

Overall, these results suggest that ceritinib is a more cost-effective treatment option than crizotinib in first line, both in the short and long term. This is noteworthy, not only because payers are particularly interested in the short-term cost-effectiveness of a drug, but also because both crizotinib and ceritinib are currently recommended as first-line treatments for ALK-positive metastatic NSCLC in the National Comprehensive Cancer Network (NCCN) guidelines⁴⁶. Given the recent approval of ceritinib in first line, guidelines recommendations may change over time as more evidence on the relative benefits of ceritinib and crizotinib as frontline therapies for ALK-positive metastatic NSCLC become available.

To the best of our knowledge, no other study to date has investigated the cost-effectiveness of ceritinib vs currently available therapeutic options in the first-line setting. Nevertheless, the cost-effectiveness of ceritinib has been demonstrated in second line (i.e. after prior treatment with chemotherapy, another ALK inhibitor, or crizotinib) for patients with ALK-positive NSCLC in the United Kingdom¹⁴, Canada¹³, and Mexico¹¹. In the US, one study used a naïve-comparison approach to assess the cost-effectiveness of ceritinib vs alectinib, another second-generation ALK inhibitor, in patients previously treated with crizotinib, and found it to be less cost-effective than alectinib¹². However, heterogeneity across trial populations in the ceritinib and alectinib trials was not accounted for in their study, which may lead to different outcomes in terms of both health benefits and costs.

Given the high proportion of patients developing resistance to crizotinib within the first year after treatment initiation as well as crizotinib’s modest effect on brain metastases, it is important to determine which therapy should be administered soon after a diagnosis of ALK-positive metastatic NSCLC to improve clinical outcomes and prolong survival while containing costs¹². To this end, in the absence of head-to-head clinical trials directly comparing ceritinib and crizotinib and given a lack of cost-effectiveness data in the first-line setting, the findings of this study provide important insight into the relative clinical and economic value of ceritinib and crizotinib as first-line treatments for ALK-positive metastatic NSCLC, and inform the decision-making among payers and other policy-makers in this indication.

Some limitations should be considered when interpreting the results of this CEA study. The main limitation of the model was the lack of a head-to-head clinical trial comparing crizotinib and ceritinib in previously untreated patients with ALK-positive NSCLC. To overcome this limitation, the MAIC approach was used to indirectly compare efficacy outcomes between crizotinib and ceritinib, while adjusting for cross-trial differences in observed patient characteristics. However, there may have been residual systematic errors resulting from unobserved prognostic variables and effect modifiers. Next, the model applied treatment-specific relative dose intensity inputs based on the respective trials. The ASCEND-4 ceritinib arm had a lower reported relative dose intensity (77.3%) compared with that of the crizotinib and platinum doublet regimen arms (all over 90%)⁹. Sensitivity analyses applying the ceritinib dose intensity to all treatment arms demonstrated that this input has limited impact on the ICERs. Another limitation of the model was the absence of reliable estimates of crossover-adjusted OS data for the platinum doublet regimen. The model instead accounted for crossover through its consideration of post-progression treatment costs, assuming that the clinical benefit of post-progression treatment had already been represented in OS data derived from the trials. Further, EQ-5D scores were not collected systematically after treatment discontinuation in the ASCEND-4 or PROFILE 1014 trials^9,24; therefore, utilities from the literature were applied to the progressive disease health state²⁶. Sensitivity analyses, however, showed that the ICER was not sensitive to the utility values used for the progressive disease health state. Finally, there was a lack of certainty around the time to discontinuation for crizotinib in terms of both trial-based data and clinical practice. In the absence of patient-level time-on-treatment data for crizotinib, the duration of treatment for both agents was estimated based on the truncated median treatment durations reported in ASCEND-4 and PROFILE 1014^9,24. Although this is not ideal, it provides the most balanced comparison based on the available data regarding the treatment duration for crizotinib. Sensitivity analyses investigated the impact of this variable and found that the ICER was sensitive to treatment duration.

Conclusions

Benchmarked on the economic value of novel cancer therapies, the results of this CEA model suggest that ceritinib offers a cost-effective option compared to both crizotinib and chemotherapy for previously untreated patients with ALK-positive metastatic NSCLC from the perspective of a US third-party payer. Notably, in the first 2 years following treatment initiation, ceritinib dominated crizotinib by conferring greater health benefits at a reduced cost. The cost effectiveness of ceritinib was robust in sensitivity analyses. Based on these results, ceritinib has the potential to fulfill a significant unmet need for the treatment of patients with ALK-positive metastatic NSCLC in the first-line setting. Future studies are warranted to understand the cost effectiveness of ceritinib vs other ALK inhibitors among patients with previously untreated ALK-positive NSCLC.

Transparency

Declaration of funding

Funding for this research was provided by Novartis Pharmaceuticals Corporation.

Declaration of financial/other relationships

AM, AD, and KC are employees of Novartis Pharmaceuticals Corporation and may own stock or stock options. Z-YZ, SH, AGB, MLR, and JX are employees of Analysis Group, Inc., which has received consultancy fees from Novartis Pharmaceuticals Corporation for this study. A JME peer reviewer on this manuscript declares that they work with Pfizer Japan Inc., which sells celecoxib. All other JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

A synopsis of the current research was presented in poster format at the Academy of Managed Care Pharmacy (AMCP) 2017 Nexus Meeting, which took place in Dallas, TX in October 2017.

Acknowledgments

Medical writing assistance was provided by Cinzia Metallo, PhD, an employee of Analysis Group, Inc.