Abstract: A decision model was developed to compare the cost-effectiveness of afatinib versus erlotinib in the first-line treatment of patients with metastatic non–small cell lung cancer (NSCLC) with epidermal growth factor receptor (EGFR) exon 19 deletion (Del19) mutations. The framework of the model was that of a partitioned survival model. Patients entered the model in the progression-free disease state initiating first-line treatment. In each cycle of the model, patients either remained in the progression-free disease state or advanced to the progressive disease or death state based on progression-free and overall survival curves. The model was populated with data from published literature, publicly available sources, and clinical trials. The results showed that patients taking afatinib accrued more life-years (3.09 vs 2.46) and quality-adjusted life-years (QALYs) (2.16 vs 1.72) than patients taking erlotinib. Total per-patient costs were $246,801 with afatinib and $212,466 with erlotinib. The incremental cost per life-year gained for afatinib versus erlotinib was $52,401. Afatinib was cost-effective at a threshold of $150,000, with an incremental cost per QALY gained of $77,504 versus erlotinib.
Key words: non-small cell lung cancer, EGFR mutations, cost-effectiveness, tyrosine kinase inhibitors, partitioned survival model
Citation: Journal of Clinical Pathways. 2016;2(4):31-39.
Received March 25, 2016; accepted April 14, 2016.
There were approximately 221,200 newly diagnosed cases of lung cancer (non–small cell lung cancer [NSCLC] and small cell lung cancer combined) and 158,040 deaths from lung cancer in the United States in 2015. NSCLC is the most common type of lung cancer, accounting for approximately 85–90% of cases. Most cases of lung cancer are diagnosed in advanced stages (stages IIIb and IV). In patients with advanced disease, modest improvements in median progression-free survival (PFS) are seen with chemotherapy; however, overall survival (OS) remains poor. In patients with stage III disease, the 5-year survival rate is 5–14%; in patients with stage IV disease, the 5-year survival rate is 1%.1
Epidermal growth factor receptor (EGFR) mutation–positive NSCLC is a specific lung cancer subtype characterized by the presence of EGFR mutations and sensitivity to treatment with EGFR tyrosine kinase inhibitors (TKIs). EGFR mutation–positive NSCLC accounts for 50% of all NSCLC cases. Common activating mutations account for 90% of EGFR mutation–positive NSCLC cases, of which exon 19 deletion (Del19) accounts for more than 50%.2 Current treatment guidance provided by the National Comprehensive Cancer Network is limited for patients with EGFR Del19 mutations.3
Timeliness, Thoroughness Lacking in Lung Cancer Care
New Biomarker for Lung Cancer Prognosis Identified
Afatinib, an irreversible EGFR TKI, is the only TKI to demonstrate statistically significant PFS and OS benefits versus standard chemotherapy in patients with metastatic NSCLC with EGFR Del19 mutation, as demonstrated in the LUX-Lung 3 clinical trial.4,5 Erlotinib, a reversible EGFR TKI, has also demonstrated significant improvement in PFS, but not OS, compared with standard chemotherapy in this patient population.6
Given today’s financial constraints, decision makers are faced with the broader challenge of assessing the economic value of these agents for formulary placement. It is important to understand the cost and health impact of prolonged survival with these treatment options. A decision model was developed to evaluate the direct costs, outcomes, and cost-effectiveness of treating with afatinib and erlotinib in patients with metastatic NSCLC with EGFR Del19 mutations.
The framework of the model was that of a partitioned survival model (Figure 1).7 Patients entered in the progression-free disease state initiating first-line treatment with either afatinib or erlotinib. In each cycle of the model, patients either remained in the progression-free disease state or advanced to the progressive disease or death states.
Once patients started first-line TKI treatment in the progression-free disease state, costs and outcomes began to accrue. Costs in the progression-free health state comprised drug acquisition costs, costs associated with managing adverse reactions, and other costs related to additional health care resources utilized during the progression-free period. In the progressive disease health state, patients incurred the cost of continuing care and end-of-life costs.
The model was populated with data from the published literature, publicly available sources, and a network meta-analysis of clinical trial data. The cycle length was 1 month, and an annual discount rate of 3% was applied to all costs and outcomes.8 To ensure that the full costs and benefits of treatment were accounted for, given the low 5-year survival rate in advanced-stage NSCLC (5–14% in patients with stage-III disease; 1% in patients with stage-IV disease), the time horizon assessed was 20 years.1
The model population for both afatinib and erlotinib arms reflected the patient population enrolled in the pivotal LUX-Lung 3 clinical trial. Patients were treatment-naïve with advanced (stage IIIB/IV) EGFR mutation-positive NSCLC and an East Cooperative Oncology Group (ECOG) scale score of 0 or 1. Baseline characteristics for patients enrolled in this study are reported elsewhere.4
The model compared afatinib 40 mg once daily with erlotinib 150 mg once daily. Both drugs were assumed to be given until disease progression or death.
Survival. The survival data used in the model for afatinib were based on parametric models fitted to empirical data from LUX-Lung 3. Both independent (performed by a blinded independent adjudication team) and investigator assessments of disease progression (ie, PFS) were reported in LUX-Lung 3. In the base case, the fitted data were based on the independent assessment of PFS. The following parametric models were considered in order to identify the model that best fit the data: exponential, Weibull, Gompertz, log-logistic, and log-normal. The best fit for PFS was obtained by using a log-logistic parametric model. The best-fitting parametric model for afatinib OS was a Weibull function. These parametric models provided the estimates and predictions for survival during and after the time frame of the empirical data.
For erlotinib, the survival curves were estimated based on hazard ratios that were applied to the hazard functions estimated from each of the afatinib parametric models. The hazard ratios for erlotinib versus afatinib were obtained from a previously published network meta-analysis of trials for the treatment of EGFR mutation–positive NSCLC with similar patient populations and treatment doses.9,10 Figure 2 presents the survival curves for both afatinib and erlotinib used in the base-case analysis.
Adverse Reactions. While on first-line TKI treatment, patients could experience adverse reactions. Only grade 3 and 4 adverse reactions that occurred in at least 5% of afatinib- or erlotinib-treated patients were included in the model. These were assumed to occur in the first month of treatment only, as most adverse reactions would present in the first month.11 After this point, it was assumed that adverse reactions would resolve. The probabilities of adverse reactions were obtained from the pivotal trials for both afatinib and erlotinib (Table 1) and were applied in the first cycle only.4,6
Utilities. The default utility value of the progression-free disease state for both treatment and comparator arms was based on the EQ-5D assessment of patients enrolled in LUX-Lung 3 (data on file). Utility scores were calculated from the five EQ-5D item scores using United Kingdom preference weights and were summarized by randomized treatment and visit. For the model, a mean utility value was used, estimated from the area under the curve to the truncation time. The utility value for the progressive disease state was derived using a utility decrement for progressive disease (0.1798) obtained from Nafees and colleagues,12 subtracted from the utility value for progression-free disease. Similarly, a disutility for each adverse reaction was included to reduce the quality of life in patients experiencing that adverse reaction (Table 2). Because adverse reactions were assumed to occur in the first month of treatment, the disutility associated with adverse reactions was applied only during the first model cycle.
Costs. Costs for the progression-free disease state included drug acquisition costs, the costs to manage adverse reactions, and additional disease management costs (other costs) that occurred in the disease state (Table 3). The monthly drug acquisition costs for afatinib and erlotinib were $7093 and $7351 respectively, based on the wholesale acquisition costs as of February 2016, as reported by Micromedex Red Book Online.13 Dosing was taken from the prescribing information for each TKI.14,15 Afatinib and erlotinib were assumed to be given daily until disease progression or death. For both afatinib and erlotinib arms, 100% adherence to therapy was assumed.
Costs to manage adverse reactions were applied in the first cycle to those who experienced a reaction. Because only grade 3 and 4 adverse reactions were included, management costs were based on inpatient costs that were obtained from Healthcare Cost and Utilization Project (HCUP)16 using the International Classification of Diseases, Ninth Revision (ICD-9) code associated with each adverse reaction (Table 3). These adverse reaction management costs were applied as one-time costs to the first model cycle.
To model the additional health care resources consumed by patients with metastatic NSCLC during the progression-free disease state for both afatinib-treated patients and erlotinib-treated patients, overall resource utilization during first-line treatment observed in LUX-Lung 3 were used (data on file). Health care resource utilization information was available for outpatient visits (general practitioner, specialist, nurse, physiotherapist); outpatient interventions (computed tomography, magnetic resonance imaging, respiratory surgical procedures, ultrasounds, x-rays, radiotherapy); and unplanned hospitalizations (unscheduled hospitalization stay, intensive care unit visit, emergency department visit). Unit costs were obtained from the essential Resource-Based Relative Value Scale (RBRVS), HCUP, and Medical Expenditure Panel Survey (MEPS) and were applied to the amount utilized for each resource to determine the costs per cycle for each resource.16-19 The costs for each resource were then summed to calculate the total additional disease management costs per cycle for the progression-free disease state (Table 3). Full details regarding the costs for the progression-free health state are provided in Supplemental Table 1.
In the progressive disease health state, management costs of progressive disease were accrued each month (Table 3). These costs were the average monthly expenditure of continuing care (eg, further active therapy, surgery) derived by taking a weighted average of the continuing-care costs for both male and female patients with lung cancer from Yabroff and colleagues.20 Because treatment costs are substantially higher during the last months of life, a one-time end-of-life cost was applied upon transitioning to the death state. This cost was also derived from Yabroff and colleagues20 using the average costs of the last year of life for both male and female patients with lung cancer.
All costs were inflated to 2016 US dollars using the medical care component of the consumer price index.21
The outcomes calculated in the model included time in the progression-free disease state (measured in progression-free life-years), OS (measured in life-years), and quality-adjusted life-years (QALYs). The costs calculated in the model included progression-free disease state costs (drug acquisition costs, adverse reaction management costs, and other disease management costs); progressive disease state costs (continuing care costs and end-of-life costs); and total costs. Two incremental cost-effectiveness ratios were calculated using the model: incremental cost per life-year gained, and incremental cost per QALY gained. Treatment was considered cost-effective if the incremental cost per QALY was below a threshold of $150,000, recommended for use in the United States.22
One-way and probabilistic sensitivity analyses were performed to test the robustness of the model assumptions and the uncertainty around specific parameter estimates. A one-way sensitivity analysis, in which one specific parameter was varied at a time while others were held constant, was run for all parameters and was based on a plausible range of estimates derived from 95% confidence intervals or assumptions.
A probabilistic sensitivity analysis, in which all parameters estimated with uncertainty were varied at the same time, was run for 10,000 iterations. All of the model parameters previously mentioned were varied except for drug acquisition cost and the discount rate of costs and outcomes. We assumed that varied costs followed a gamma distribution, for which the shape and scale parameters were estimated via means and standard error. Utility values and the probabilities of adverse reactions were varied assuming a beta distribution, for which the alpha and betas were derived from the means and standard errors. The hazards ratios were varied assuming a log-normal distribution, for which the location and scale parameters were derived from the means and standard errors.
Alternative scenarios were assessed to evaluate the cost-effectiveness of afatinib when investigator-assessed PFS data for afatinib were used in the model instead of independently assessed PFS data and when other parametric models were used to estimate survival.
The results showed that patients taking afatinib accrued more life-years (3.09 vs 2.46) and QALYs (2.16 vs 1.72) than patients taking erlotinib (Table 4). Additionally, patients taking afatinib remained progression-free longer than patients on erlotinib (1.66 years vs 1.31 years). As a result, progression-free disease state costs were higher with afatinib compared with erlotinib ($176,048 vs $142,001). Similarly, costs accrued in the progressive disease state were higher with afatinib than with erlotinib ($70,753 vs $70,465), because patients receiving afatinib had longer overall survival. Therefore, total per-patient costs were $246,801 with afatinib and $212,466 with erlotinib. The incremental cost per life-year gained for afatinib versus erlotinib was $54,584. Afatinib was cost-effective at a threshold of $150,000, with an incremental cost per QALY gained of $77,504 versus erlotinib.
Figure 3 presents the impact on incremental cost per QALY gained when changing model parameters in one-way sensitivity analysis. The results were most sensitive to the PFS and OS hazard ratios and drug acquisition costs. Varying the PFS hazard ratio had an effect on the time in the progression-free disease state, but it did not affect OS, for erlotinib. When time in the progression-free disease state was increased for erlotinib (PFS hazard ratio, 1.38), afatinib was more cost-effective because of the higher costs associated with erlotinib; however, the effect on QALYs was minimal. In contrast, when time in the progression-free disease state was reduced for erlotinib (PFS hazard ratio, 0.44) from the level used in the base-case analysis, afatinib was no longer more cost-effective because of the lower costs associated with erlotinib.
Varying the OS hazard ratio had an effect on OS, but did not change the time in the progression-free disease state, for erlotinib. Increasing the OS hazard ratio to the upper bound (1.26) created a scenario in which erlotinib was more cost-effective, as a result of erlotinib-treated patients having a longer life expectancy. Reducing the OS hazard ratio to the lower bound (0.36) did not have an effect on the cost-effectiveness of afatinib (incremental cost per QALY gained, $45,416), despite the increase in erlotinib-treated patients’ life expectancy.
Although varying drug acquisition costs had an effect on the cost-effectiveness of both drugs, afatinib remained more cost-effective than erlotinib.
Results were not sensitive to changes in any of the other parameters.
Using an incremental cost per QALY gained threshold of ≤ $150,000 for cost-effectiveness, the results of the 10,000-iteration probabilistic sensitivity analysis showed that afatinib was cost-effective in 74% of the simulations conducted (Figure 4).
In scenario analysis, we explored the effect of investigator assessment data (parametric survival models and hazard ratios) on cost outcomes. The impact was minimal, with the incremental cost per QALY gained increasing to $81,697 (incremental costs of $36,439 and incremental QALYs of 0.45). We also looked at the impact that use of other parametric survival models had on the results. When using alternative parametric models, afatinib remained cost-effective, with the incremental cost per QALY gained ranging from $46,658 to $97,618. Detailed results of all the alternative parametric model analyses are summarized in Supplemental Table 2.
We developed a decision-analytic model to compare the costs and outcomes associated with afatinib treatment versus erlotinib treatment for patients with metastatic NSCLC with EGFR Del19 mutations. The model used parametric survival models fitted to the empirical trial data for afatinib and estimated survival for erlotinib using hazard ratios from a network meta-analysis.
The model showed that treatment with afatinib in patients with metastatic NSCLC with EGFR Del19 mutations was cost-effective at an incremental-cost-per-QALY-gained threshold of $150,000. Costs accrued in both the progression-free disease state and the progressive disease state were higher for afatinib, but both were attributable to afatinib patients having longer PFS and OS than erlotinib. Sensitivity analysis showed that results were most sensitive to the hazard ratios used to determine survival for erlotinib. Results were insensitive to changes in other parameters. Probabilistic sensitivity analysis showed that afatinib was cost-effective at a willingness-to-pay threshold of $150,000 per QALY in 74% of simulations conducted.
To our knowledge, no prior US-based analysis has examined the cost-effectiveness of afatinib or erlotinib treatment of patients with EGFR Del19 mutations. Ting and colleagues23 analyzed the cost-effectiveness of erlotinib versus afatinib in a broader population of patients with NSCLC that included patients with common EGFR mutations. Their analysis simply used the empirical survival data from the respective clinical trials; an exponential extrapolation of these data to predict survival beyond the duration of the trial was used. In addition, their analysis did not use formal statistical methodology for indirect comparison of the baseline characteristics of the trial populations.
The present study findings must be interpreted in light of the following limitations. A key limitation is the lack of head-to-head trial data. Thus, an indirect comparison of afatinib and erlotinib was performed. This approach is commonly used in decision modeling in the absence of head-to-head data.24 Another possible limitation of the model is the assumption that costs and utility decrements due to adverse reactions occurred only in the first cycle. Because timing and durations of adverse reactions are not available from the clinical trial publications, we assumed the adverse reaction would occur in the first cycle and last only for that cycle. This is consistent with other cost-effectiveness analyses, including a study by Carlson and colleagues11 that assumed that the majority of adverse reactions occurred in the first cycle and lasted 1 month.
Additionally, because the LUX-Lung 3 trial was a multinational trial, the resource-use data incorporated in the model might not be applicable to a US population. Also, the analysis that estimated the utility value relied on UK preference weights. It is possible that health-state preferences may differ in a US population, although results from sensitivity analyses showed that the model was not sensitive to changes in utility values.
We assumed that treatment continued until disease progression for both afatinib and erlotinib. This is the goal when TKIs are given; however, individuals can stop treatment before progression, which would lower drug-acquisition costs and possibly lower PFS and OS. The LUX-Lung 3 results showed that the median time on treatment for the intent-to-treat population (which included patients with all EGFR mutations) was 11.0 months, and the median PFS was 11.1 months.4 Therefore, the model assumption of remaining on therapy until disease progression had very little effect on the drug costs for afatinib or on outcomes. The EURTAC trial results showed that the median time on erlotinib in the intent-to-treat population (common EGFR mutations) was 8.2 months, and the median PFS was 9.7 months.6 Although erlotinib-treated patients may not stay on treatment until disease progression, it is unknown what the mean time on therapy was and how that compared with the mean PFS. Regardless, varying time on treatment for both afatinib and erlotinib had little effect on the model results.
The model does not explicitly analyze subsequent treatment after disease progression. However, the survival associated with subsequent treatment are considered in the overall survival curves used in the model, and the subsequent treatment costs are considered in the progressive disease state costs. The cost of continuing care maybe an underestimate of subsequent treatment costs. However, the sensitivity analyses showed that results were not sensitive to changes to the costs.
Our study suggests that afatinib, as a first-line therapy for metastatic NSCLC with EGFR Del19 mutations, is a cost-effective alternative to erlotinib in the US, assuming a $150,000 cost-effectiveness threshold. Decision makers should consider the specific type of EGFR mutation when selecting treatment for patients with metastatic NSCLC.
1. American Cancer Society. Non-small cell lung cancer survival rates, by stage. American Cancer Society web site. http://www.cancer.org/cancer/lungcancer-non-smallcell/detailedguide/non-small-cell-lung-cancer-survival-rates. Revised February 8, 2016. Accessed January 13, 2016.
2. Sharma SV, Bell DW, Settleman J, Haber DA. Epidermal growth factor receptor mutations in lung cancer. Nat Rev Cancer. 2007;7(3):169–181.
3. National Comprehensive Cancer Network. Non-Small Cell Lung Cancer (Version 4.2016). https://www.nccn.org/professionals/physician_gls/pdf/nscl.pdf. AccessedApril 27, 2016.
4. Sequist LV, Yang JC, Yamamoto N, et al. Phase III study of afatinib or cisplatin plus pemetrexed in patients with metastatic lung adenocarcinoma with EGFR mutations. J Clin Oncol. 2013;31(27):3327–3334.
5. Yang JC, Wu YL, Schuler M et al. Afatinib versus cisplatin-based chemotherapy for EGFR mutation-positive lung adenocarcinoma (LUX-Lung 3 and LUX-Lung 6): analysis of overall survival data from two randomised, phase 3 trials. Lancet Oncol. 2015;16(2):141–151.
6. Rosell R, Carcereny E, Gervais R, et al. Erlotinib versus standard chemotherapy as first-line treatment for European patients with advanced EGFR mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial. Lancet Oncol. 2012;13(3):239–246.
7. Amdahl J, Manson SC, Isbell R, et al. Cost-effectiveness of pazopanib in advanced soft tissue sarcoma in the United Kingdom. Sarcoma. 2014;2014:481071.
8. Gold MR, Siegel JE, Russell LB, Weinstein MC (eds). Cost-effectiveness in Health and Medicine. New York: Oxford University Press; 1996.
9. Popat S, Mok T, Yang JC, et al. Afatinib in the treatment of EGFR mutation-positive NSCLC--a network meta-analysis. Lung Cancer. 2014;85(2):230–238.
10. Popat S, Yang JC, Lungershausen J, Griebsch I, Märten A, Wu YL. Treatments for EGFR mutation-positive (M+) NSCLC patients – a network meta-Analysis (NMA) by mutation type. Presented at the 17th Annual European Congress of the International Society for Pharmacoeconomics and Outcomes Research; 2014. Amsterdam, The Netherlands.
11. Carlson JJ, Garrison LP, Ramsey SD, Veenstra DL. The potential clinical and economic outcomes of pharmacogenomic approaches to EGFR-tyrosine kinase inhibitor therapy in non-small-cell lung cancer. Value Health. 2009;12(1):20–27.
12. Nafees B, Stafford M, Gavriel S, Bhalla S, Watkins J. Health state utilities for non small cell lung cancer. Health Qual Life Outcomes. 2008;6:84.
13. Micromedex Red Book Online. Accessed through Micromedex 2.0. Truven Health Analytics Inc; 2016. Accessed April 27, 2016.
14. Gilotrif (afatinib tablets) prescribing information. Ridgefield (CT): Boehringer Ingelheim Pharmaceuticals, Inc; 2015.
15. Tarceva (erlotinib tablets) prescribing information. Seymour (IN): Kremers Urban Pharmaceuticals, Inc; 2015.
16. Agency for Healthcare Research and Quality. Healthcare Cost and Utilization Project. 2013. HCUPnet. http://hcupnet.ahrq.gov. Accessed January 27, 2016.
17. Optum360. The Essential RBRVS: A Comprehensive Listing of RBRVS Values for CPT and HCPCS Codes. Optum360, LLC; 2016.
18. Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey. 2013. http://meps.ahrq.gov. Accessed January 27, 2016.
19. Candrilli S, Mauskopf J. How much does a hospital day cost? Presented at the 11th Annual International Meeting of the International Society for Pharmacoeconomics and Outcomes Research; May 20-24, 2006. Philadelphia, PA.
20. Yabroff KR, Lamont EB, Mariotto A, et al. Cost of care for elderly cancer patients in the United States. J Natl Cancer Inst. 2008;100(9):630–641.
21. Bureau of Labor Statistics. Consumer price index—all urban consumers; medical care. http://data.bls.gov. Accessed January 29, 2016.
22. Neumann PJ, Cohen JT, Weinstein MC. Updating cost-effectiveness—the curious resilience of the $50,000-per-QALY threshold. N Engl J Med. 2014;371(9):796–797.
23. Ting J, Tien Ho P, Xiang P, Sugay A, Abdel-Sattar M, Wilson L. Cost-effectiveness and value of information of erlotinib, afatinib, and cisplatin-pemetrexed for first-line treatment of advanced EGFR mutation-positive non-small-cell lung cancer in the united states. Value Health. 2015;18(6):774–782.
24. Sutton A, Ades AE, Cooper N, Abrams K. Use of indirect and mixed treatment comparisons for technology assessment. Pharmacoeconomics. 2008;26(9):753–767.