Cost-effectiveness of osimertinib in the UK for advanced EGFR-T790M non-small cell lung cancer

Abstract

Aim: This study presents the cost-utility analysis that was developed to inform the NICE health technology assessment of osimertinib vs platinum-based doublet chemotherapy (PDC) in patients with EGFR-T790M mutation-positive non-small cell lung cancer (NSCLC) who have progressed on epidermal growth factor receptor-tyrosine kinase inhibitor (EGFR-TKI) therapy.

Methods and materials: A partitioned survival model with three health states (progression-free, progressed disease, and death) from a UK payer perspective and over lifetime (15 years) was developed. Direct costs included disease management, treatment-related (acquisition, administration, monitoring, adverse events), and T790M testing costs. Efficacy and safety data were taken from clinical trials AURA extension and AURA2 for osimertinib and IMPRESS for PDC. An adjusted indirect treatment comparison was applied to reduce the potential bias in the non-randomized comparison. Parametric functions were utilized to extrapolate survival beyond the observed period. Health state utility values were calculated from EQ-5D data collected in the trials and valued using UK tariffs. Resource use and costs were based on published sources.

Results: Osimertinib was associated with a gain of 1.541 quality-adjusted life-years (QALYs) at an incremental cost of £64,283 vs PDC (incremental cost-effectiveness ratio [ICER]: £41,705/QALY gained). Scenario analyses showed that none of the plausible scenarios produced an ICER above £44,000 per QALY gained, and probabilistic sensitivity analyses demonstrated a 63.4% probability that osimertinib will be cost-effective at a willingness-to-pay threshold of £50,000.

Limitations: The analysis is subject to some level of uncertainty inherent to phase 2 single-arm data and the immaturity of the currently available survival data for osimertinib.

Conclusions: Osimertinib may be considered a cost-effective treatment option compared with PDC in the second-line setting in patients with EGFR-T790M mutation-positive NSCLC from a UK payer perspective. Further data from the ongoing AURA clinical trial program will reduce the inherent uncertainty in the analysis.

Keywords:

• Osimertinib
• NSCLC
• EGFR-T790M mutation
• cost-effective
• cost-utility

Introduction

Lung cancer is the most commonly diagnosed cancer, and the leading cause of cancer-related death worldwide¹. In England, lung cancer is the second most frequently diagnosed cancer for both males and females, with 37,400 new cases registered in 2014². Non-small cell lung cancer (NSCLC) accounts for more than 85% of all lung cancer cases³, and 8–11% of non-Asian patients with NSCLC have tumor-associated epidermal growth factor receptor (EGFR)-sensitizing mutations^4,5. EGFR tyrosine kinase inhibitors (EGFR-TKIs) are recommended as first-line therapy for patients with NSCLC and locally advanced EGFR-sensitizing mutations, on the basis of clinical trials showing improved response rates, progression-free survival, and health-related quality-of-life (HRQoL) compared with platinum-based doublet chemotherapy (PDC)⁶. The most common mechanism of acquired resistance is the secondary EGFR-T790M mutation, which inhibits TKI binding. This is detectable in ∼60% of EGFR-mutated lung cancers that have developed acquired resistance to erlotinib or gefitinib⁷. Prior to osimertinib, the optimal clinical management strategies for patients with advanced or metastatic NSCLC resistant to EGFR-TKI were undefined, with PDC typically used as a treatment option⁸.

Osimertinib is an EGFR-TKI that is selective for both primary EGFR-sensitizing and T790M-resistant mutations; it is indicated for the treatment of patients with locally advanced or metastatic EGFR-T790M-positive advanced or metastatic NSCLC who have progressed on a prior EGFR-TKI.

Osimertinib (80 mg, orally, once daily) received conditional marketing authorization from the European Medicines Agency (EMA) in February 2016 under the accelerated assessment process. Regulatory approval was achieved based on results from two phase 2, single-arm, trials that assessed the efficacy and safety of osimertinib: AURA extension (ClinicalTrials.gov NCT01802632) and AURA2 (ClinicalTrials.gov NCT02094261).

A confirmatory phase 3 trial (AURA3, ClinicalTrials.gov NCT02151981) of osimertinib compared with PDC (pemetrexed plus carboplatin or cisplatin) in patients whose disease has progressed following first-line treatment with EGFR-TKI therapy has completed recruitment, and recently met its primary end-point, demonstrating superior progression-free survival (PFS) of osimertinib-treated patients⁶.

Healthcare budget holders rely on heath technology assessments (HTAs) to evaluate health benefits and value for money, and to support reimbursement decisions regarding existing and new pharmaceuticals. While single-arm trial data can inform the regulatory process and lead to marketing authorization, the absence of randomized controlled trial (RCT) data poses challenges for the comparative effectiveness assessments required by budget holders.

During the interim between osimertinib receiving marketing authorization and RCT data from AURA3 becoming available, a number of comparative evidence and HTAs for national pricing and reimbursement reviews were undertaken based on the single-arm studies (AURA extension and AURA2).

Notably, osimertinib was submitted to the National Institute for Health and Care Excellence (NICE) technology appraisal process. In the absence of randomized data and in order to estimate comparative effects for osimertinib compared with PDC, a relevant historical control was identified. This study presents the cost-effectiveness analysis of osimertinib compared with PDC that was developed to inform the NICE technology appraisal⁹.

As a result of the appraisal, osimertinib was recommended by NICE as an option for use within the ‘new’ Cancer Drugs Fund (CDF)^9,10. The CDF is intended to promote faster access to the most promising new treatments, while additional evidence on cost-effectiveness is gathered in ongoing studies.

Methods

Overview

A cost-effectiveness model was developed to assess the relative costs and health outcomes of osimertinib compared with PDC in the treatment of patients with EGFR-T790M mutation-positive NSCLC who have progressed on or after EGFR-TKI therapy.

The analysis was conducted from the perspective of the UK National Health Service (NHS) and personal social services. A lifetime horizon using extrapolations of PFS and overall survival (OS) to 15 years (when <1% of patients are still alive) was used for the base case. Costs were expressed in GBP (£) at 2015 UK prices, and costs and outcomes were discounted at 3.5% per annum, in line with NICE guidelines¹¹.

Direct costs include disease management costs, treatment-related costs (acquisition, administration, monitoring, and adverse events), and T790M testing costs. Health outcomes are measured in quality-adjusted life years (QALYs), which is NICE’s preferred indicator of effectiveness. The incremental cost-effectiveness ratio (ICER) is measured as the incremental cost per QALY gained for osimertinib compared with PDC. Cost-effectiveness is assessed by comparing the estimated ICER with a notional threshold of £50,000/QALY gained, the relevant threshold for osimertinib as it was considered to fulfil end-of-life criteria¹².

Scenario analyses and probabilistic sensitivity analyses (Monte Carlo simulation) were conducted to assess the robustness of the results. The model was developed using Microsoft Excel 2010 with Visual Basic for Applications.

Model structure

The cost-effectiveness of osimertinib compared with PDC was evaluated using a partitioned survival model with three mutually-exclusive health states (progression-free, progressed disease, and death). In partitioned survival modeling, the state occupancy of the simulated cohort is estimated by extrapolating the cumulative probability of PFS and OS to a lifetime, and using the curves to estimate the proportion who are alive and have not progressed (% on PFS curve), those who have died (1 – % on OS curve), and those who are alive but have experienced disease progression (% on OS curve – % on PFS curve), at each time point of the simulation.

Within this framework it is assumed that the modeled health states (progression-free, progressed disease, and death) represent the sequence of events that patients may experience over the course of their treatment (progression of disease and death), on the basis that these events are progressive, mutually exclusive, and irreversible.

A cycle length of 1-week was used and a half-cycle correction was applied to improve the accuracy of the results¹³. Each health state was associated with a cost and utility, and the time spent by patients in each state was used to estimate cumulative costs and health outcomes.

Target population

The target population in the analysis is patients with NSCLC in a second-line setting who have progressed after first-line EGFR-TKI therapy. The focus is on patients who have been identified as T790M-positive. T790M mutation status is known only after biopsy or plasma ctDNA testing on progression, and, therefore, the analysis also considers the costs of T790M testing.

Interventions and clinical data sources

Osimertinib is being investigated in the AURA clinical program comprising AURA, AURA2, and AURA3. AURA extension (phase 2 cohort within AURA) and AURA2 (phase 2) were single-arm, open-label studies designed to assess the safety and efficacy of osimertinib (80 mg, orally, once daily) in patients with advanced NSCLC whose disease has progressed after at least one course of treatment with an EGFR-TKI. Patients had confirmation of T790M status from a biopsy sample taken after progression on the most recent treatment regimen.

In a pooled analysis of patients from AURA extension and AURA2 (November 2015 data cut-off) in the second- and further-line settings, osimertinib was associated with a median PFS of 11.0 months, based on a blinded independent central review (227/411 events; 55.2% maturity) (95% CI = 9.6–12.4), and the proportion of patients who were progression-free at 12 months was 47.5% (95% CI = 42.4–52.5)¹⁴. At data cut-off, the OS data were immature and had not yet reached the median. Aggregate data from the individual clinical studies^15,16 show that 23.8% of patients (98/411) had died.

Data for PDC were sourced from the IMPRESS (gefitinib/chemotherapy vs chemotherapy in EGFR mutation-positive NSCLC after progression on first-line gefitinib) study in which patients with chemotherapy-naïve EGFR mutation-positive advanced NSCLC who progressed on first-line gefitinib were randomized to receive either PDC plus gefitinib or PDC plus placebo. The study included EGFR-T790M-positive, EGFR-T790M-negative, or unidentified patients as measured by plasma ctDNA testing. Based on the final data cut-off for IMPRESS, the median PFS (78.7% maturity) for PDC in the T790M mutation-positive group was 5.3 months (95% CI = 4.0–5.6 months) and the median OS (72.1% maturity) was 14.1 months (95% CI = 11.0–20.5 months)¹⁷. Patients in the T790M mutation-positive control group had similar demographic and disease characteristics to those in AURA extension and AURA2¹⁸. Therefore, the T790M placebo arm of IMPRESS provides the best available evidence for efficacy of PDC in patients with second-line EGFR mutation-positive relapsed NSCLC after progression on first-generation EGFR-TKIs.

Model inputs: efficacy

An adjusted indirect treatment comparison of osimertinib relative to PDC based on pooled patient-level data from AURA extension/AURA2 and IMPRESS¹⁸ was used to inform the efficacy inputs in the model. Further, and in order to parameterize the model, inverse probability weighting (IPW) was applied. Patients on osimertinib were weighted by 1/PS, whilst patients on PDC were weighted by 1/(1–PS), where PS represents the individual’s estimated propensity score (PS). IPW was considered the most appropriate adjustment method, as it balances the two treatment cohorts on common observable baseline characteristics, and reduces the potential bias in the non-randomized efficacy comparison, whilst also estimating parameters that can be incorporated into a partitioned survival simulation model.

The adjusted analysis reported the following results (Table 1): the median PFS for osimertinib (51/92 events; 55.4%) was 10.9 months compared with 5.3 months for PDC (41/53 events; 77.4%). Osimertinib was associated with a statistically significant improvement in PFS with a hazard ratio (HR) of 0.242 (95% CI = 0.148–0.394; p < .0001)¹⁸. The median OS was not calculable for osimertinib (17/92 events; 18.5%), whilst for PDC (38/53; 71.7%) the median OS was 14.1 months. Osimertinib was associated with an OS HR of 0.347 (95% CI = 0.179–0.673; p = .0016). An unadjusted naïve treatment comparison reported similar results to those from the adjusted indirect treatment comparison (Table 1).

Table 1. Non-parametric analysis, unadjusted and adjusted.

Outcome	AURA pooled	IMPRESS T790M mutation-positive	Treatment effect, HR (95% CI; p-value)
Indication	Second-line	Second-line	–
Treatment	Osimertinib	PDC	–
Adjusted indirect comparison
PFS adjusted
Number of patients, n	92	53	0.242 (0.148–0.394; p < .0001)
Total events, n (%)	51 (55.4)	41 (77.4)
Months, median	10.9	5.3
OS adjusted
Number of patients, n	92	53	0.347 (0.179–0.673; p = .0016)
Total events, n (%)	17 (18.5)	38 (71.7)
Months, median	Not reached	14.1
Unadjusted indirect comparison
PFS unadjusted
Number of patients, n	127	61	0.234 (0.155–0.356; p < .001)
Total events, n (%)	69 (54.3)	48 (78.7)
Months, median (95% CI)	11.0 (8.3–13.6)	5.3 (4.0–5.6)
OS unadjusted
Number of patients, n	129	61	0.357 (0.209–0.611; p < .001)
Total events, n (%)	25 (19.4)	44 (72.1)
Months, median (95% CI)	Not reached	14.1 (11.0–20.5)

CI, confidence interval; HR, hazard ratio; OS, overall survival; PDC, platinum-based doublet chemotherapy; PFS, progression-free survival.

Parametric analysis

Standard guidance for fitting and selecting survival functions was followed¹⁹. However, due to the immaturity of the data, testing for proportional hazards was judged uninformative. Independent survival models for osimertinib and PDC were fitted to the Kaplan–Meier data for PFS and OS. Standard parametric models (exponential, Weibull, Gompertz, log-logistic, log-normal, and generalized Gamma) adjusted for IPW were estimated in the base case analysis.

Visual inspection and statistical goodness-of-fit were performed for PFS and OS. However, due to the immaturity of the available survival data at the time of analysis, goodness-of-fit statistics were considered uninformative or inconclusive, and so the selection of models for the base case was mostly guided by the plausibility of the extrapolated curves. Given that a common parametric model is preferred across treatments according to the NICE technical support document on partitioned survival analysis¹⁹:

• The generalized gamma, Weibull, and the Gompertz Models provide plausible PFS estimates for both osimertinib and PDC in second-line use. The Gompertz Model and Weibull are the more conservative options (generalized gamma gives a higher incremental PFS gain for osimertinib). The log-normal, log-logistic, and exponential models give unrealistic tails for osimertinib and PDC.

• The Weibull, generalized gamma, and Gompertz Models are the most plausible distributions for extrapolating OS for osimertinib in second-line use. The generalized gamma has an unrealistic extrapolation for PDC and the log-logistic, log-normal, and exponential give unrealistic extrapolations of osimertinib.

Therefore, the Gompertz and Weibull models are the most plausible distributions. Both of the models provide a plausible fit for PFS for both treatments; however, the Gompertz model has a more conservative tail for osimertinib. Due to the Gompertz OS curves crossing at ∼40–45 months, which was judged clinically implausible, the Weibull is the most appropriate model, although it has a long tail, and there is still considerable uncertainty given the maturity of the OS Kaplan–Meier curve for osimertinib (Figure 1).

Figure 1. Overall survival and progression-free survival in the base case (adjusted analysis).

Model inputs: safety

Adverse events (grade ≥3 according to CTCAE²⁰) occurring in ≥10% of the population in either treatment group were selected for inclusion in the analysis. Adverse event incidence rates for osimertinib were sourced from the AURA pooled dataset, and data for PDC were derived from IMPRESS. Because the rates of grade ≥3 events were generally low, especially for osimertinib, it was considered appropriate to use unadjusted adverse event rates for both treatments. The costs and disutilities related to adverse events were applied as one-off events for a once-weekly cycle at the start of the model simulation.

Model inputs: resource use and cost

For all drug costs, the dosing information was sourced from the EMA label for each treatment, and unit costs were sourced from the British National Formulary²¹, NHS Reference Costs 2014–2015²², and Unit Costs of Health and Social Care²³.

T790M testing, disease management, and treatment costs are presented in Table 2. For osimertinib, the costs of testing for T790M were considered to be part of the decision problem; thus, the cost per patient identified was added to the treatment costs. It was assumed that 80% of patients are tested using a sequential strategy of plasma ctDNA testing followed by tissue biopsy in patients identified as T790M negative by ctDNA testing, and that the remaining 20% of patients undergo tissue biopsy alone.

Table 2. Costs of T790M testing, disease management, and treatment.

	Cost, £
	Osimertinib	PDC
T790M testing (per patient identified as eligible for osimertinib)	1,350.80	0
Progression-free (per week)	77.44	77.44
Progressed disease (per week)	139.58	139.58
End-of-life/terminal care (one-off)	3,905.35	3,905.35
Drug acquisition costs (list price per pack)	5,770	160.00
Drug administration costs (per administration)	0.48	First: 245.16 Subsequent: 332.50
Drug monitoring costs (per week)	0	4.61

PDC, platinum-based doublet chemotherapy.

Disease management costs included weekly health state costs for progression-free and progressed disease, and costs of end-of-life/terminal care associated with time spent in hospital, hospice, or at home. Resource use data were sourced from the HTA study by Brown et al.²⁴, and subsequently used by the Assessment Group for the recent NICE multiple technology appraisal of erlotinib and gefitinib for treating patients with NSCLC whose disease has progressed following prior chemotherapy (review of NICE technology appraisals 162 and 175)²⁵.

Drug acquisition costs were calculated based on available formulations; pack sizes, unit costs, and price per milligram for each (combination of) treatment. Vial sizes used for intravenous treatments in the base-case setting were those resulting in the lowest monthly acquisition cost, assuming no wastage.

Patients on PDC were treated until progression or for a maximum of four treatment cycles, in line with NHS protocols for England. Drug administration and monitoring costs (pharmacy dispensing costs, cost of a nurse visit for osimertinib, and the costs of chemotherapy infusion, pre-medication with dexamethasone, liver function test, renal function test, and complete blood count for PDC) were included.

All the analyses were run using osimertinib’s confidential patient access scheme discounted price agreed between AstraZeneca and the Department of Health, England.

In both AURA studies, patients were allowed to continue osimertinib treatment after radiological disease progression. Therefore, costs for osimertinib were based on time-to-discontinuation of treatment (TDT) from AURA Extension/AURA2. The TDT Kaplan–Meier data pooled from AURA Extension/AURA2 were applied in the model up to day 431, after which it was linearly extrapolated to a lifetime horizon. The linear extrapolation offered a good approximation of the Kaplan–Meier data, with a good visual fit and a reasonable extrapolation where all patients had stopped treatment at ∼32 months.

The treatment pathway and distributions of subsequent treatments were elicited from NICE Clinical Guidance CG121²⁶ and clinical expert opinion in the UK. It is anticipated that osimertinib will shift the clinical treatment pathway by delaying treatment with current standard of care. In other words, on progression, patients treated with osimertinib will subsequently be treated first with PDC (80%) or best supportive care (BSC) (20%) and then with mono-chemotherapy (docetaxel, 50%) or BSC (50%). In line with this, on progression, patients on PDC as the primary treatment will move onto mono-chemotherapy (50%) or BSC (50%). The cost of subsequent treatments was calculated following this premise and, in the absence of alternative data, the duration of all subsequent treatments was assumed to be the same as the duration of the primary treatment.

Model inputs: quality-of-life

Health state utilities for progression-free and progressed disease, and adverse event disutilities were applied to estimate QALYs. For progression-free and progressed disease, utility values were derived from EQ-5D patient-level data collected in the AURA2 (5-level questionnaire) and IMPRESS (3-level questionnaire) studies. Utilities were calculated applying the EQ-5D-5L crosswalk index values²⁷ and the EQ-5D-3L UK tariff set²⁸, respectively. An EQ-5D index score was calculated for each subject and visit. The average EQ-5D utility was calculated by averaging the EQ-5D index score for each subject in the full analysis set, and then calculating the average value across all subjects within the progression-free or progressed disease state.

The progression-free values were derived based on utility data collected in the IMPRESS study but adjusted for treatment-specific response rates for osimertinib and PDC in AURA extension, AURA2, and IMPRESS. More specifically, the utility values for patients with a partial response (PR) and a complete response (CR) (PR + CR) and for patients with stable disease (SD) were multiplied by the response rates in each treatment (67.4% for osimertinib and 33.3% for PDC) to derive weighted progression-free values for osimertinib and PDC. The weighted progression-free utility values are 0.805 and 0.778, respectively.

The progressed disease utility was given by the mid-point of the values derived from the AURA2 and IMPRESS studies. The mean progressed disease utility value is 0.715 (SD = 0.029). This mid-point value was chosen because it reflects the sustained tumor response experienced by patients in AURA extension and AURA2, as well as reflecting the significant proportion (77%) of patients who responded to osimertinib and continued treatment following Response Evaluation Criteria in Solid Tumors (RECIST) progression. This approach was also deemed more conservative than applying treatment-specific utility values from AURA2 and IMPRESS.

Utility decrements associated with adverse events were sourced from a study by Nafees et al.²⁹ and previous submissions to NICE for similar indications.

Sensitivity analyses

Uncertainty was assessed using scenario analyses and probabilistic sensitivity analysis (PSA). Scenario analyses were conducted in order to test the impact on the results of using other clinically plausible parametric distributions for OS. Other scenarios included varying the health-state utility values for progression-free and progressed disease including treatment response, assumptions about the costs of osimertinib treatment, and applying a 10-year time horizon. For the PSA, parameter uncertainties were assessed using second-order Monte Carlo simulations. Ten thousand iterations were performed, with the measurement of uncertainty reported in terms of a cost-effectiveness plane and multi-way cost-effectiveness acceptability curves, evaluating the probability of cost-effectiveness by a range of willingness-to-pay (WTP) thresholds. The results of this analysis represent the joint consequences of uncertainty in all key inputs.

Results

Base case analysis

Over a lifetime horizon, treatment with osimertinib is associated with 1.541 QALYs gained at an incremental cost of £64,283 per patient. The incremental cost per QALY gained for osimertinib vs PDC is £41,705. The predicted median times to disease progression are 11.1 months with osimertinib, compared with 4.6 months with PDC. These estimates only slightly over-estimate the adjusted AURA pooled data (median 10.9 months), whilst slightly under-estimating the adjusted IMPRESS (median 5.3 months). The predicted mean and median time to death are 51.2 and 40.2 months with osimertinib, compared with 22.6 and 19.2 months with PDC. The median represents an over-estimate of the expected survival for PDC compared with the observed IMPRESS data (median 14.1 months); it is not possible to validate this for osimertinib without more mature data.

Scenario analyses

Overall, the results of the scenario analyses show that none of the scenarios produced ICERs that exceed £44,000 per QALY gained (Table 3).

Table 3. Results of scenario analyses.

Scenario	Total cost, £		Total QALYs		Incremental costs, £	Incremental QALYs	ICER, £/QALY gained
	Osimertinib	PDC	Osimertinib	PDC
Base case	87,441	23,159	2.841	1.300	64,283	1.541	41,705
(a) Survival modeling
PFS and OS distribution: Weibull (both arms)	86,975	23,239	2.848	1.300	63,736	1.548	41,173
PFS and OS distribution: log logistic (both arms)	93,784	23,390	3.765	1.356	70,393	2.409	29,224
PFS and OS distribution: exponential (both arms)	91,819	23,587	3.510	1.442	68,231	2.068	32,993
(b) Health State Utility Values
(i) AURA2 EQ-5D-5L Crosswalk Values
No Treatment Response: PF 0.812 (both arms); PD 0.751 (both arms)	87,441	23,159	2.952	1.365	64,283	1.587	40,510
Treatment Response: PF 0.808 (Osimertinib); PF 0.781 (PDC); PD 0.751 (both arms)	87,441	23,159	2.947	1.352	64,283	1.595	40,313
(ii) AURA2 EQ-5D-5L England Valuation Set Values
No Treatment Response: PF 0.874 (both arms); PD 0.821 (both arms)	87,441	23,159	3.214	1.490	64,283	1.724	37,279
Treatment Response: PF 0.870 (Osimertinib); PF 0.848 (PDC); PD 0.821 (both arms)	87,441	23,159	3.210	1.479	64,283	1.731	37,145
(iii) IMPRESS EQ-5D-3L Index Values
No Treatment Response: PF 0.779 (both arms); PD 0.679 (both arms)	87,441	23,159	2.712	1.249	64,283	1.463	43,928

ICER, incremental cost-effectiveness ratio; OS, overall survival; PD, progressed disease; PDC, platinum-based doublet chemotherapy; PF, progression-free; PFS, progression-free survival; QALY, quality-adjusted life-years.

Probabilistic sensitivity analysis

The PSA was run for 10,000 iterations. The probabilistic ICER is £40,581 per QALY gained, which compares with £41,705 in the deterministic analysis (a < 3% difference in the ICER). The results of the PSA suggest that osimertinib is most likely to be cost-effective at WTP thresholds of £42,000 and above (Figure 2). At a cost-effectiveness threshold of £50,000 per QALY gained, the probability of osimertinib being considered cost-effective vs PDC is 63.4%.

Figure 2. Cost-effectiveness acceptability curve including osimertinib and PDC.

Discussion

EGFR-T790M mutation-positive lung cancer is a life-threatening disease. Patients with this condition have limited treatment options and reduced life expectancy, resulting in an urgent need for more therapies³⁰.

Osimertinib received marketing authorization in the EU through the EMA accelerated assessment procedure, the purpose of which is to provide support for the development of medicines of major interest for public health and therapeutic innovation. Conditional approval was granted based on the results from two phase 2 single-arm studies (AURA extension and AURA2), which were noted in the European public assessment report (EPAR) report as showing both high anti-tumor activity of clinical value and response rates expected to translate into clinical benefit. However, the magnitude of such benefit in terms of OS and/or PFS remained unknown³¹.

The lack of comparative long-term efficacy data can pose challenges for health technology appraisals. This study presents the cost-effectiveness analysis of osimertinib compared with PDC as second-line therapy in patients with advanced or metastatic EGFR-T790M mutation-positive NSCLC, based on the two phase 2, single-arm studies and a relevant historical control from the IMPRESS study.

The base case results indicate that, over a lifetime horizon (15 years), treatment with osimertinib is associated with 2.841 QALYs vs 1.300 QALYs for PDC. Thus, osimertinib is expected to result in 1.541 QALYs gained. The costs per patient associated with osimertinib treatment are £87,441 compared with £23,159 for PDC, giving an incremental cost of £64,283 per patient. Thus, the incremental cost per QALY gained for osimertinib vs PDC is £41,705. Osimertinib is most likely to be cost-effective at WTP thresholds of £42,000 and above. There is a 63.4% probability that osimertinib will be cost-effective at a WTP threshold of £50,000.

The analysis is subject to some level of uncertainty in light of the accelerated EMA approval based on single-arm data only. To mitigate against this, a robust and clinically appropriate comparator was identified, providing evidence for a sub-group of patients in the IMPRESS study who are T790M mutation-positive. The results from an adjusted indirect treatment comparison to reduce bias between AURA extension and AURA2 for osimertinib and IMPRESS for PDC¹⁸ were used to model long-term efficacy in the cost-effectiveness analysis. Early results from the RCT AURA3 based on a median follow-up of 8.3 months show that the duration of PFS was significantly longer in the osimertinib group than in the PDC group (ITT: median PFS 10.1 months vs 4.4 months, HR 0.30 (95% CI = 0.23–0.4))⁶. The consistency of AURA3 results with those reported in the adjusted indirect treatment comparison suggest that the T790M mutation-positive control group from IMPRESS is an appropriate comparator and unlikely to have biased the results of the cost-effectiveness analysis in favour of osimertinib.

Due to the immaturity of the PFS and OS data for osimertinib currently available from the AURA extension and AURA2 trials, the extrapolated survival estimates are uncertain. Sensitivity analyses were undertaken to show the impact of different fits to the results. For the base case, the Gompertz function was selected for PFS and the Weibull function was chosen for OS. Scenario analyses were conducted using the exponential and log-logistic distributions for PFS and OS. None of the scenarios produced ICERs that exceed the cost-effectiveness ratio threshold for end-of-life treatments in the UK, and the base case represents a rather conservative estimate of the cost-effectiveness compared with other scenarios.

The utility values applied in the analysis, particularly for progression-free patients on osimertinib (0.805) compared with PDC (0.778), reflect the improved quality-of-life associated with osimertinib due to its good tolerability profile, symptom improvement, the high tumor response rate, and the oral route of administration. Furthermore, the utility values derived from the AURA2 study are comparable to utility estimates obtained from previous studies of targeted therapies for locally advanced or metastatic NSCLC. For example, in the PROFILE 1007 trial of crizotinib vs chemotherapy in previously treated patients with ALK-positive advanced NSCLC, the utility value for patients on crizotinib treatment (0.82) was greater than for patients on chemotherapy (0.73)³².

Despite the uncertainties inherent to phase 2, single-arm data, the clinical and cost-effectiveness evidence is compelling, and osimertinib was recommended for treating EGFR-T790M mutation-positive NSCLC within the CDF. In the recently modified appraisal process for cancer drugs, a “recommendation for use within the CDF” can be made when NICE considers there to be plausible potential for a drug to satisfy the criteria for routine commissioning, but where there is significant remaining clinical uncertainty⁹. As a condition for the positive recommendation, a managed access arrangement including data collection arrangements for osimertinib was agreed with NICE and NHS England. The data collection arrangement specifies the longer-term data from the phase 2, single-arm studies (AURA extension and AURA2) and AURA3 studies that need to be collected in order to resolve the key areas of uncertainty.

Until the end of the managed access period, when NICE will re-appraise the drug with a view to decide whether or not it can be recommended for routine use within the NHS, patients can access osimertinib through the CDF. Early data from the RCT (AURA3) have demonstrated that, for patients receiving osimertinib, clinical outcomes are in line with results of the phase 2 trials (AURA extension and AURA2). Similarly, the outcomes for patients on PDC are broadly in line with those in the IMPRESS trial³³. Therefore, there is growing evidence on the superior efficacy of osimertinib to support the positive recommendation.

Conclusions

Osimertinib may be considered a cost-effective treatment compared with PDC as second-line therapy in patients with EGFR-T790M mutation-positive NSCLC from a UK payer perspective. Additional data are expected from AURA extension, AURA2, and from the confirmatory phase 3 trial AURA3, which will increase the maturity of the long-term comparative efficacy estimates and reduce uncertainty in the cost-effectiveness analysis

Transparency

Declaration of funding

This study and manuscript were funded by AstraZeneca PLC.

Declaration of financial/other relationships

MD is an employee of AstraZeneca PLC, the sponsor of this study and manuscript; CB was an employee of AstraZeneca at the time of the study. EB, VD, SL, and AG provided advisory and consultancy services paid for by AstraZeneca. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Acknowledgments

Nicholas Rusbridge of PAREXEL Access provided the authors with editorial support during the preparation of this article.