Cost-effectiveness of 3-years of adjuvant imatinib in gastrointestinal stromal tumors (GIST) in the United States

Abstract

Background:

Recent clinical trial data have demonstrated that 3 years vs 1 year of adjuvant imatinib therapy for patients with surgically resected Kit+ Gastrointestinal Stromal Tumors (GIST) leads to a significant improvement in recurrence-free survival and overall survival. This study assesses the cost-effectiveness of treating patients with 3 years vs 1 year of imatinib from a US payer’s perspective.

Methods:

A Markov model was developed to predict GIST recurrence and treatment costs. Patients enter the model after surgery and transition among three health states: free of recurrence, recurrence, and death. Recurrence, mortality, costs, and utilities were derived from clinical trial and published literature. Expected costs and quality-adjusted life years (QALYs) were estimated and discounted at 3%/year. Deterministic and probabilistic sensitivity analyses were conducted.

Results:

Patients receiving 3 years of imatinib had higher QALYs (8.53 vs 7.18) than those receiving 1 year of imatinib. Total lifetime per-patient cost was $302,100 for 3 years vs $217,800 for 1 year of imatinib. Incremental cost effectiveness ratio of 3 years vs 1 year of imatinib was $62,600/QALY. Model results were sensitive to long-term rate of GIST recurrence (beyond 5 years) and cost of imatinib. At a threshold of $100,000/QALY, 3 years vs 1 year of imatinib was cost-effective in 100% of simulations.

Limitations:

The model is a simplified representation of disease natural history and may not account for all possible health states and complications associated with disease. Resource utilization on treatment was estimated using the resource use data from previous trials, therefore calculated medical costs might be over-estimated compared to the real-world setting.

Conclusions:

Model results suggest that treatment with 3 years vs 1 year of imatinib is cost-effective at a $100,000/QALY threshold. Clinical and economic results suggest treating surgically resected Kit+ GIST patients with 3 years of imatinib would result in improved quality-adjusted survival.

Keywords::

• GIST
• Cost-effectiveness
• Imatinib

Introduction

Gastrointestinal stromal tumors (GIST) are tumors arising in the wall of the gastrointestinal tract, occurring most frequently in the stomach and small intestine and less frequently in colon, rectum, or esophagus1. GIST may also occur outside of the gastrointestinal tract, although this is a fairly rare occurrence. Incidence of GIST in the US is estimated at 3000–5000 cases per year2–4, and surgically resectable primary GIST may account for one third to one-half of all GIST cases5.

The standard treatment for primary resectable GIST is surgery. However, ∼ 50% of patients develop recurrence within 5 years of surgery depending on the tumor characteristics6–10. Since the introduction of tyrosine kinase inhibitors (e.g., imatinib), the management of Kit+ GIST has changed significantly. Imatinib (Gleevec® Novartis Pharmaceuticals) is currently approved for treatment of adult patients with Kit+ unresectable, and/or metastatic malignant GIST, and also as adjuvant therapy following surgical resection of Kit+ GIST.

In 2008, imatinib was approved as adjuvant treatment of adult patients following complete gross resection of Kit (CD117) positive Gastrointestinal Stromal Tumors (GIST). This was based on a clinical trial that evaluated use of 1 year of imatinib treatment vs placebo after complete resection of Kit+ GIST. The results showed improved recurrence-free survival (RFS) in patients with GIST ≥ 3 cm11.

In a recent multi-center, prospective, randomized Phase III study, the efficacy and safety of imatinib in Kit+ GIST patients was evaluated comparing 3 years vs 1 year of treatment12. This study was carried out as a co-operative group study with the Scandinavian Sarcoma Group (SSG), in co-operation with the Sarcoma Group of the AIO (Arbeitsgemeinschaft Internistische Onkologie) [NCT00116935]. The study participants were randomly assigned to either 12 months or 36 months of adjuvant imatinib treatment (following complete resection of Kit+ GIST) administered orally at a daily dosage of 400 mg/day. All randomized patients who completed 1 year of adjuvant imatinib were followed for 5 years. A total of 400 patients were randomized. Recurrence-free survival for patients receiving 3 years of adjuvant imatinib at years 3 and 5 were 88.1% and 67.4% vs 62.1% and 50.3% for patients receiving 1 year of adjuvant imatinib; hazard ratio [HR] = 0.46, 95% CI = 0.31–0.68, p < 0.0001. Similarly, overall survival (OS) at years 3 and 5 were higher for patients on 3 years adjuvant imatinib, 96.3% and 92.0% vs 94.0% and 81.7%, respectively, for patients on 1-year of treatment, HR = 0.45, 95% CI = 0.22–0.89, p = 0.19.

In January 2012, the FDA expanded the label based on the results from the SSG trial. Although clinical benefits of 3-year imatinib treatment are clearly established in GIST patients through this trial, it is important to assess the impact of longer treatment duration from a payer perspective. We used decision analytic techniques to evaluate the lifetime cost-effectiveness of 3-year imatinib vs 1-year imatinib in the US from a third party payer perspective.

Methods

Model overview

We developed a Markov (state-transition) model13 in Microsoft Excel 2007 to compare 3-year vs 1-year adjuvant imatinib treatment outcomes and costs for patients who undergo complete gross resection of primary GIST over a lifetime horizon. The model focused on disease recurrence/progression as predictors of cost, quality-of-life, and survival following surgical resection. The model consisted of mutually exclusive health states (Figure 1) grouping patients into the following three categories: (1) free of recurrent GIST; (2) recurrent GIST; and (3) death (from GIST or other causes). The risk of disease recurrence/progression is modeled as a series of transitions between these states. All patients began in the model free of recurrent GIST. In subsequent cycles of the Markov process, patients could either remain in the initial Markov state, or transition to the recurrent GIST or death states. The transitions occurred at 1-month intervals and probabilities governing these transitions were conditional on treatment received and health state.

Figure 1.  Model structure.

The target population for the model reflected the study population in the SSGXVIII/AIO clinical study (hereafter referred to as the SSG trial)12. Data from the SSG trial12, along with other published secondary sources and expert clinical opinion were used to estimate the model (Tables 1–3). Measures of effectiveness include life-years saved and quality-adjusted life-years (QALYs) gained. Costs and benefits were discounted at 3% per year.

Table 1.  Probabilities of GIST recurrence, discontinuation, mortality, and treatment pathways.

Parameter	Base-case value	Distribution	Source
Monthly probability of recurrence for patients with recurrence-free GIST
3-years of imatinib 400 mg/day			Joensuu et al.12
Year 0–1	0.003	Normal
Year 1–2	0.003	Normal
Year 2–3	0.003	Normal
Year 3–4	0.005	Normal
Year 4–5	0.014	Normal
Beyond year 5	0.004	Normal
1-year of imatinib 400 mg/day			Joensuu et al.12
Year 0–1	0.003	Normal
Year 1–2	0.020	Normal
Year 2–3	0.012	Normal
Year 3–4	0.002	Normal
Year 4–5	0.004	Normal
Beyond year 5	0.004	Normal
Treatment for recurrence/progression during or after adjuvant imatinib (%)
First recurrence while on active treatment with imatinib 400 mg/day: imatinib 800 mg/day	100	N/A	DeMatteo et al.11
First recurrence after completing 1 year of imatinib 400 mg/day: re-treatment with imatinib 400 mg/day	100	N/A	DeMatteo et al.11
Progression
Sunitinib second-line treatment	80	Beta	Expert opinion
Best supportive care	20	Beta	Expert opinion
Monthly probability of progression while treated with
Imatinib 400 mg/day (first recurrence)	0.0336	Beta	Verweij et al.15 Verweij et al.15
Imatinib 800 mg/day (first recurrence)	0.0285	Beta
Imatinib 800 mg/day (second-line therapy)	0.1259	Beta	Verweij et al.15, Contreras-Hernández et al.17
Sunitinib (second-line therapy)	0.1586	Beta	Demetri et al.16
Monthly probability of discontinuation
Imatinib (due to adverse events) in the first 12 months	0.0062	Beta	DeMatteo et al.11
Imatinib 400 mg/day first-line treatment for recurrence	0.0029	Beta	Verweij et al.15
Imatinib 800 mg/day first-line treatment for recurrence	0.0029	Beta	Verweij et al.15
Monthly probability of death during treatment12
3-years of adjuvant imatinib 400 mg daily			Joensuu et al.12
Year 0–1	0.000	Beta
Year 1–2	0.001	Beta
Year 2–3	0.002	Beta
Year 3–4	0.000	Beta
Year 4–5	0.004	Beta
1-year of adjuvant imatinib 400 mg daily			Joensuu et al.12
Year 0–1	0.000	Beta
Year 1–2	0.002	Beta
Year 2–3	0.003	Beta
Year 3–4	0.005	Beta
Year 4–5	0.007	Beta
Monthly probability of death from GIST without GIST recurrence
Mortality multiplier	1.23	Normal	Rubin et al.5
Monthly probability of death from GIST following GIST recurrence
Imatinib 400 mg/day first-line treatment	0.0135	Beta	Verweij et al.15
Imatinib 800 mg/day second-line treatment	0.0125	Beta	Verweij et al.15
Imatinib 800 mg/day for second recurrence	0.0373	Beta	Demetri et al.16
Sunitinib second-line treatment	0.0373	Beta	Demetri et al.16
Best supportive care	0.0432	Beta	Tran et al.2, Demetri et al.16, Huse et al.19

N/A, Model parameter was not varied in the probabilistic sensitivity analysis.

Table 2.  Utilities.

Parameter	Base-case value	Distribution	Source
Utility multipliers
Recurrence-free GIST	1.00	Normal	Assumption
Recurrent GIST	0.87	Beta	Wilson et al.21 Expert opinion
Second recurrence	0.71	Beta	Expert opinion
Best supportive care	0.58	Beta	Chabot et al.22
Death	0.00	N/A
Utility decrements associated with GIST treatment
Adjuvant imatinib*	−0.02	N/A	Expert opinion

N/A, Model parameter was not varied in the probabilistic sensitivity analysis.

*This utility decrement was applied only in the month where patients were discontinued due to adverse events.

Table 3.  Monthly costs associated with GIST and GIST recurrence.

Parameter	Base-case Default value	Distribution	Source
Resource costs			Physician’s Fee and Coding Guide25
Gastroenterologist visit	$104	N/A
CT scan	$1,167	N/A
General practitioner visit	$75	N/A
Monthly pharmacy costs			Redbook24
Imatinib 400 mg	$5,004	N/A
Imatinib 800 mg	$10,008	N/A
Sunitinib	$6,101	N/A
Monthly monitoring and medical costs by health state			Physician’s Fee and Coding Guide25
Recurrence-free GIST
On treatment
First 3 months	$963	Gamma
4 months to 2 years	$436	Gamma
Years 3–5	$224	Gamma
Off treatment
Years 0–3	$349	N/A
Years 4–5	$224	Gamma
Years 6+	$118	Gamma
Adverse events (cost per event)	$360	Gamma	Physician’s Fee and Coding Guide25
Recurrent GIST
Best supportive care	$2,891	Gamma	Yabroff et al.30
One-time cost of recurrence	$15,242	Gamma	Physician’s Fee and Coding Guide25 DRG Expert26

N/A: Model parameter was not varied in the probabilistic sensitivity analysis.

Model parameters and data sources

First-recurrence rates

Monthly rates of a first recurrence for years 0–5 for each treatment arm (3-year imatinib and 1-year imatinib) were estimated using the recurrence-free survival data from the SSG trial (Table 1)12. In the SSG trial, the 1-year RFS rate was slightly lower in the 1-year arm than the 3-year arm. Since all patients (3-year and 1-year arms) received the same treatment in the first year (i.e., imatinib 400 mg daily), in the model, we assumed that the RFS rate to be equal in the first year. Therefore, the monthly rate of recurrence in the 3-year imatinib arm was applied to patients in the 1-year imatinib arm. For recurrence rate beyond 5 years, the monthly recurrence rate observed at Year 5 in the 1-year imatinib arm was applied for both treatment arms as a proxy for expected efficacy once the treatment is completed. The 5-year rate from the 1-year imatinib arm was chosen as a proxy based on the assumption that, by the fifth year, the effect of the drug in the 1-year imatinib arm would have reached a steady state. To estimate the monthly probability of recurrence, data from the RFS curve were used to first estimate the annual probability of recurrence/death. Using the OS data from the trial, the proportion of patients who died was estimated, and this estimate was subtracted from the probability of recurrence/death to estimate the annual probability of recurrence. The annual probability was then converted to monthly rates.

Post-recurrence rates

Base-case post-recurrence rates are listed in Table 1. In accordance with the National Comprehensive Cancer Network (NCCN) guidelines14, all patients in the model who experience recurrence while receiving 400 mg/day imatinib treatment receive an escalated dose of imatinib (800 mg/day). All patients experiencing disease recurrence after completion of 1-year or 3-year imatinib treatment are treated with imatinib 400 mg/day. Estimates of treatment failure while receiving treatment for recurrence (imatinib 400 mg/day or 800 mg/day) due to toxicity or second recurrence/progression were obtained from a randomized trial of imatinib in the treatment of advanced or metastatic GIST15. Patients whose tumors progress while receiving imatinib 400 mg/day can either escalate their dose to 800 mg/day, initiate sunitinib therapy, or move to best supportive care; the proportion of patients that initiate each of these regimens was based on expert opinion.

Data for treatment failure and survival on second line treatment with sunitinib was estimated from the randomized controlled trial of sunitinib in patients with advanced GIST after failure of imatinib16. Failure on second line treatment with imatinib 800 mg/day was estimated from published literature15,17.

Mortality rates

The overall survival data from the SSG trial was applied to derive the monthly mortality rates for both treatment arms for years 0–5 using an approach similar to estimating first recurrence rates (Table 1). For mortality rate beyond 5 years for patients without recurrence, a hazard ratio of 1.235 was applied to the background age-specific mortality18 to account for disease-specific mortality. After Year 5, patients with recurrence were applied treatment-specific risk of mortality from published sources2,15,16,19.

Discontinuation rates

Patients were assumed to discontinue due to adverse events during the first year of adjuvant treatment but not thereafter. In the SSGXVIII/AIO clinical trial20 the discontinuation rate at 12 months in 1-year and 3-year treatment arms were 7.7% and 6.6%, respectively. For the model, an average of the two discontinuation rates was calculated (7.2%) and a monthly probability of 0.006 was applied.

Utility

Utility values were assigned to the various health states in the model to estimate the number of QALYs associated with each treatment. In the model, it was assumed that GIST patients who were surgically resected and without recurrence had the same utility as healthy individuals (Table 2). Similarly, for patients with recurrence, each health-state specific utility values were estimated from published literature21,22 and expert opinion. Utilities associated with disease states are applied as multipliers to the age-specific healthy utility values23 to obtain health-state specific utilities. Clinical experts in the management of GIST were consulted to get an estimate of the expected utility decrement associated with discontinuation due to adverse events. Based on their input, patients on adjuvant imatinib were assigned a utility value 0.02 points higher than patients who discontinued adjuvant imatinib due to adverse events.

Costs

Drug acquisition costs were estimated from national average wholesale prices published in the 2010 Drug Topics® Red Book (Table 3)24. The physician visit and other resource use costs were estimated from the 2010 Physician Fee and Coding Guide and procedure costs were estimated using the 2010 DRG Expert25,26. All other costs from the literature were adjusted to 2010 using the consumer price index (CPI)27.

Monthly monitoring costs were estimated using available data from the American College of Surgeons Oncology Group (ACOSOG) Z9001 trial, the National Comprehensive Cancer Network Practice Guidelines, and other published sources11,28,29. A one-time cost of treating adverse event was applied to all patients who discontinued adjuvant imatinib. The adverse event costs were calculated as a weighted average cost of treating the top 15 adverse events reported in the ACOSOG Z9001 trial11. Patients also incurred a one-time cost for GIST recurrence which was estimated assuming the cost for one visit to the general practitioner (GP) and specialist, one computed tomography (CT) scan, and surgical resection of the tumor. The cost estimate of best supportive care was based on the monthly costs for gastric cancer, reported by Yabroff et al.30 Yarbroff et al. estimated costs for the first year of disease, continuing phase of cancer (defined as the period between their first year after diagnosis and their last year of life) and the last year of life. Using the proportion of patients in the last year of life and those in the continuing phase of cancer, a weighted average cost was calculated.

Analyses

The cost-effectiveness ratio was calculated as the incremental cost divided by life-years saved or QALYs gained given treatment with 3-year of adjuvant imatinib vs 1-year. The model estimated cumulative costs of GIST treatment with 3-year imatinib and 1-year imatinib therapy. The cumulative costs were tabulated by summing all costs associated with each health state to which patients’ transitioned. In addition the model yielded estimates of life years (calculated by summing months spent in all non-fatal health states), and QALYs (calculated by summing months spent in a health state with the utilities associated with each health state in which patients reside, then dividing by 12) for each treatment option. All outcomes were tabulated over the lifetime of the population. The incremental cost-effectiveness ratios (ICER) were estimated by first rank-ordering the treatment regimens by increasing cost and then comparing the more costly to the less costly strategy by dividing the additional cost by the additional benefit.

Sensitivity analyses

Deterministic sensitivity analyses were performed to assess the robustness of the results to changes in key parameters. Parameters were varied by 25% above and below their base-case values, while holding the other parameters fixed and the model was re-run.

To assess uncertainty in the cost-effectiveness analysis, a probabilistic sensitivity analysis (PSA) was performed using a second-order Monte Carlo simulation. The PSA was performed by simultaneously drawing from appropriate distribution functions for key model parameters according to their means and standard errors (Tables 1–3); beta distribution for transition probabilities and utilities and gamma distribution for costs. This process of drawing parameters and running the model was repeated 1000 times and the incremental results are presented graphically in the cost-effectiveness plane.

Results

Base case analyses

Recurrence-free and overall survival curves generated from the model were comparable to the data observed in the SSG trial (Figures 2 and 3). Over a lifetime horizon, the model predicted that patients treated with 3-year adjuvant Imatinib had higher quality adjusted survival than those treated with 1-year imatinib (8.53 QALYs vs 7.18 QALYs). Similarly, treatment with 3-year adjuvant imatinib had more life years than 1-year treatment (10.78 LYs vs 9.30 LYs). The total lifetime costs incurred while achieving this benefit was estimated to be $302,100/patient for those treated with adjuvant 3-year imatinib vs $217,800 for patients treated with 1-year imatinib. The ICER for adjuvant 3-year imatinib vs 1-year imatinib was $62,600/QALY and $57,000/LYS.

Figure 2.  Recurrence-free survival.

Figure 3.  Overall survival.

Sensitivity analyses

When the base-case values for the parameters were varied by ±25%, the results from the one-way sensitivity analyses suggested that the cost effectiveness estimates were most sensitive to changes in the rate of GIST recurrence for Years 6+ (beyond 5 years) in both arms and the cost of adjuvant imatinib (Figure 4).

Figure 4.  Deterministic sensitivity analysis of adjuvant 3-years imatinib vs 1-year imatinib.

Results from the PSA suggest that, at a $100,000/QALY threshold, 3-year imatinib therapy is cost-effective in 100% of the simulations compared to 1-year imatinib. A scatter plot of incremental costs and QALYs is presented in Figure 5; all the points in the scatter plot fell in the first quadrant, suggesting that 3-year of adjuvant imatinib was more costly and more effective than 1-year of adjuvant imatinib. Figure 6 presents the cost-effectiveness acceptability curve, at a willingness-to-pay threshold of $100,000/QALY, showing a 100% probability of 3-year imatinib therapy being cost-effective compared to 1-year imatinib. The mean incremental costs and QALYS were $72,502 (95% CI = $72,063–$72,941) and 1.32 (95% CI 1.31–1.33), respectively. The mean cost-effectiveness ratios based on the probabilistic analyses was $54,901 (95% CI = $54,514–$55,288).

Figure 5.  Probabilistic sensitivity analysis of adjuvant 3-year imatinib vs 1-year (scatter plot).

Figure 6.  Probabilistic sensitivity analysis of adjuvant 3-year imatinib vs 1-year (cost-effectiveness acceptability curve).

Discussion

Adjuvant imatinib was approved in 2008 by the FDA based on improved recurrence-free survival from a clinical trial evaluating the use of 1 year imatinib treatment vs placebo after complete resection of Kit+ GIST11. The label was recently expanded (January 2012) based on a recent Phase III study (SSG trial) that confirmed significant recurrence-free survival benefit and survival advantage for longer duration adjuvant therapy in a study of 400 patients (3-year vs 1-year adjuvant imatinib)12.

To evaluate if the clinical benefit observed in the SSG trial translates into long-term economic benefits, a Markov model was developed to assess the cost-effectiveness of 3-year vs 1-year adjuvant imatinib from a US third-payer perspective. In the model, the risk of recurrence and mortality were estimated using the results from the SSG trial. Model results suggest that patients treated with 3-year imatinib can expect to have improved recurrence-free survival and overall survival compared to 1-year imatinib treatment, as demonstrated in the SSG trial. Model results ($62,600/QALY and $57,000/LYS for 3 years vs 1 year of imatinib) compares quite favorably to results from a recent survey by Nadler et al.31, indicating that many US physicians felt that oncology treatments could be considered cost-effective at a threshold below $300,000/QALY. The model results were sensitive to rate of GIST recurrence at Year 6 and the monthly cost of adjuvant imatinib.

The model results should be interpreted in light of numerous modeling assumptions. First, the decision analytic model is a simplified representation of disease natural history and associated outcomes and may not account for all possible health states and complications associated with disease. For instance, in the model, all patients who experience disease recurrence while on treatment receive imatinib 800 mg based on input from expert opinion. Although in the SSG trial, the 1-year RFS rate was slightly lower in the 1-year arm than the 3-year arm, in the model, the 1-year RFS rates were assumed to be similar for both 1-year and 3-year adjuvant imatinib.

Additionally, in the absence of published literature on resource utilization while on adjuvant imatinib, the medical costs were estimated using the resource use data from the ACOSOG Z9001 trial; therefore, costs might be over-estimated compared to in a clinical setting.

The results from this study clearly indicate that the clinical benefits (improved recurrence-free survival and improved overall survival) of treating patients with imatinib for 3 years vs 1 year can be clearly translated into economic benefits, providing patients and physicians with an effective treatment option, and offering payers cost-efficient use of resources.

Transparency

Declaration of funding

Funding for this research was provided by Novartis Pharmaceuticals, Florham Park, NJ, USA. The data and methodology used in the manuscript was not proprietary in nature and the publication of this paper is not contingent on the sponsor’s approval.

Declaration of financial/other relationships

Myrlene Sanon, Douglas C. A. Taylor, and Anju Parthan have disclosed that they have a compensated consultant/advisory role at OptumInsight, a company sponsored by Novartis to conduct this study. John Coombs and Medha Sasane have disclosed that they are employees of Novartis. Marc Paolantonio has disclosed that he was an employee of Novartis at the time this study was conducted.

Acknowledgments

No assistance in the preparation of this article is to be declared.