Long-term cost-consequence analysis of exenatide once weekly vs sitagliptin or pioglitazone for the treatment of type 2 diabetes patients in the United States

Abstract

Objective:

Exenatide once-weekly (ExQW) is a GLP-1 receptor agonist shown to lower glucose and cardiovascular risk factors in patients with type 2 diabetes mellitus (T2DM). The objective of this study was to estimate the clinical benefits and associated economic benefits of treatment with ExQW compared with sitagliptin or pioglitazone in the US.

Methods:

The IMS CORE Diabetes Model, a validated computer simulation model, was used to project lifetime clinical outcomes and complication costs. The costs of glucose-lowering drugs were excluded as not all prices were available. Baseline patient characteristics (mean values: age, 52.5 years; diabetes duration, 6 years; HbA1_c, 8.51%; body mass index, 32.12 kg/m²) and clinical data were derived from a phase 3 clinical trial that compared ExQW with sitagliptin or pioglitazone in T2DM patients. At 6 months, patients treated with ExQW had greater improvements in HbA1_c and body weight than those treated with sitagliptin or pioglitazone. Complication costs were extracted from published sources. Health outcomes and costs were discounted at 3% per year. Sensitivity analyses were performed.

Results:

Over 35 years, and compared with sitagliptin or pioglitazone, ExQW increased life expectancy by, respectively, 0.28 (13.76 ± 0.17 vs 13.48 ± 0.18) and 0.17 years (13.76 ± 0.17 vs 13.59 ± 0.17), and quality-adjusted life years by, respectively, 0.28 (9.56 ± 0.12 vs 9.28 ± 0.12) and 0.24 years (9.56 ± 0.12 vs 9.32 ± 0.12). ExQW was associated with lower lifetime complication costs: compared with sitagliptin or pioglitazone, ExQW saved, respectively US$2215 (US$55,647 ± 2039 vs US$57,862 ± 2159) and US$933 (US$55,647 ± 2039 vs US$56,580 ± 2007) direct cost per patient. Cost-savings resulted mainly from a lower projected cumulative incidence of cardiovascular diseases and neuropathic complications.

Limitations:

Short-term changes in surrogate end-points were used to project lifetime effects on clinical outcomes. Pharmacy costs were excluded from the analyses.

Conclusions:

Over a patient’s lifetime, ExQW was projected to improve health and decrease diabetes-related complication costs compared with sitagliptin or pioglitazone.

Keywords::

• Type 2 diabetes
• Exenatide
• Sitagliptin
• Pioglitazone
• Modeling
• Costs
• United States

Introduction

In clinical practice, type 2 diabetes mellitus (T2DM) is managed over time to maintain glycosylated hemoglobin (HbA1_c) to a level below 7% in order to prevent or to delay diabetes-related complications1–3. Lifestyle interventions alone and then in combination with metformin monotherapy are usually the first steps in diabetes management1,2. However, ∼50% of patients with diabetes require additional medications after 3 years to achieve targeted HbA1_c levels3,4. Since 2005, glucose-lowering therapies based on the glucose-like peptide-1 (GLP-1) receptor activation pathway, including dipeptidyl-peptidase-4 (DPP-4) inhibitors and GLP-1 receptor agonists, have gradually been introduced in clinical practice as a treatment option for patients with T2DM. These therapies function by activating the GLP-1 receptor, either directly via GLP-1 receptor agonists or indirectly via DPP-4 inhibitors. Activation of this receptor is associated with glucose-dependent insulin secretion, glucagon suppression, slowed gastric emptying, and enhanced satiety5.

Exenatide twice-daily (ExBID) was the first GLP-1 receptor agonist approved for clinical use in the US. It is indicated for the improvement of glycemic control in patients with T2DM as monotherapy or as adjunctive therapy in combination with metformin, a sulfonylurea (with or without metformin), a thiazolidinedione (with or without metformin), or insulin glargine6.

A once-weekly exenatide formulation (ExQW) has been approved for use in the European Union and, more recently, in the US. ExQW has the same active ingredient as ExBID. However, in the ExQW formulation, the exenatide is encapsulated in microspheres of a polymer (poly-[d,l-lactide-co-glycolide]), which slowly degrade after subcutaneous injection7. Once reaching steady state after ∼6–7 weeks, weekly injection of ExQW results in continuous exenatide exposure and thus continuous GLP-1 receptor agonism8. The effects of 2-mg ExQW were compared with those of two oral anti-diabetes agents, 100-mg sitagliptin or 45-mg pioglitazone, in a 26-week, randomized, controlled, phase 3 study (n = 491) known as DURATION-2 9. Results from this study showed statistically significant reduction in HbA1_c levels with ExQW compared with either sitagliptin (p < 0.001) or pioglitazone (p = 0.0165). Furthermore, in comparison with ExQW, patients treated with sitagliptin experienced smaller weight loss from baseline and those randomized to pioglitazone gained weight9.

Current study objectives

The objective of this study was to estimate the long-term clinical and economic benefits of ExQW compared with those of sitagliptin or pioglitazone in the US, based on the results of the DURATION-2 study. Since ExQW did not have a price in the US at the time of the analysis, pharmacy costs were excluded.

Methods

IMS CORE diabetes model

The IMS Health Core Diabetes Model (CDM), a validated computer simulation model that has been described in detail elsewhere10,11, was used to conduct a cost-consequence analysis. The CDM was developed to estimate the long-term health outcomes and economic consequences of interventions in type 1 and type 2 diabetes mellitus. The Core Diabetes Model is a validated and widely used tool for economic evaluations in diabetes. The model has been found to generate reliable results when compared to trial data and to other diabetes models12. CDM inputs include baseline cohort characteristics, history of complications, current and future management of diabetes and concomitant medications, treatment effects, and changes in physiological parameters over time. A series of sub-models within the CDM simulate the major complications of diabetes, including myocardial infarction, congestive heart failure, stroke, peripheral vascular disease, neuropathy, foot ulcer, amputation, retinopathy, macular edema, cataract, and nephropathy. Each of the sub-models uses Markov Monte Carlo simulation. Transition probabilities for the model were obtained from published epidemiological and clinical studies, including the Diabetes Control and Complications Trial (DCCT) and the United Kingdom Prospective Diabetes Study (UKPDS)13,14. The CDM is consistent with American Diabetes Association computer-based modeling guidelines for assessing the long-term costs and clinical outcomes for various diabetes treatments15.

Model outcomes

In this analysis, base-case health outcomes of interest were the cumulative incidence of complications, life years (LYs) gained, quality-adjusted life years (QALYs) gained, and annual cumulative costs per patient. The QALY outcome captures both mortality and impact on the quality-of-life of factors such as diabetes complications, treatment-related adverse reactions, and changes in weight. Key health and economic outcomes are presented separately, as point estimates of difference in LYs, QALYs, and cost per patient (US$/LY). Details of cumulative incidence of complications and individual cost components are also provided.

Simulation cohorts

Data from the DURATION-2 intention-to-treat (ITT) population (e.g. baseline demographics, clinical parameters, and racial characteristics) were used as model inputs (Table 1).

Table 1.  Baseline patient characteristics.

Patient demographics	M	SD
Baseline age, years	52.5	10.3
Duration of diabetes, years	6	5.2
Percentage male	51.7	NA
Baseline risk factors
HbA1c, %	8.5	1.2
Systolic blood pressure, mmHg	126.3	13.8
Total cholesterol, mg/dL	180.3	42.1
High-density lipoprotein cholesterol, mg/dL	42.5	10.0
Low-density lipoprotein cholesterol, mg/dL	107.0	34.2
Triglyceride, mg/dL	179.3	110.2
Body mass index, kg/m^2	32.1	5.3
Racial characteristics, %
White	34.2	NA
Black	10.6	NA
Hispanic	29.1	NA
Native American	0.6	NA
Asian/Pacific Islander	25.5	NA
Baseline CVD complications, %
Myocardial infarction	8.2	NA
Angina	0	NA
Peripheral vascular disease	0	NA
Stroke	4.9	NA
Congestive heart failure	3.7	NA
Atrial fibrillation	0	NA
Left ventricular hypertrophy	0	NA
Baseline renal complications, %
Microalbuminuria	0.7	NA
Gross proteinuria	0	NA
End-stage renal disease	0	NA
Background diabetic retinopathy	17.7	NA
Proliferative diabetic retinopathy	0	NA
Severe vision loss	0	NA
Baseline macular edema, %	0	NA
Baseline cataract, %	0	NA
Baseline neuropathy, %	6.5	NA

NA, not applicable.

Source: Data on file.

Intervention effects

The initial clinical effects of ExQW, sitagliptin, and pioglitazone were derived from the 6-month results of the DURATION-2 trial for the ITT population (Table 2). In the base-case analysis, it was conservatively assumed that glycemic control in each patient group deteriorated over the course of 3 years as a result of diabetes progression. Consequently, patients treated with ExQW, sitagliptin, and pioglitazone were assumed to switch to basal insulin after 3 years.

Table 2.  Base case clinical data for exenatide once weekly, sitagliptin, and pioglitazone (26-week intent-to-treat).

Change from baseline	Exenatide QW, 2 mg		Sitagliptin, 100 mg		Pioglitazone, 45 mg
	M	SD	M	SD	M	SD
HbA1c, %	−1.55	1.27	−0.92	1.28	−1.23	1.27
Systolic blood pressure, mmHg	−3.60	12.27	0.20	12.24	−1.60	12.20
Total cholesterol, mg/dL	−0.60	31.75	3.10	32.08	6.20	31.60
High-density lipoprotein cholesterol, mg/dL	2.00	7.72	2.00	7.73	6.20	7.58
Low-density lipoprotein cholesterol, mg/dL	−1.00	26.06	1.80	26.4	1.80	26.08
Triglycerides, mg/dL	−11.40	87.69	−7.7	104.61	−39.4	87.92
Body mass index, kg/m^2	−0.85	1.5	−0.28	1.53	1.03	1.46
Minor hypo rate,/100ppy	3	NA	12	NA	1	NA
Major hypo rate,/100ppy	0	NA	0	NA	0	NA

NA, not applicable.

Source: Data on file.

Health state utilities

For T2DM and its complications, health state disutilities were derived wherever possible from the UKPDS, with data gaps filled with the use of utilities from a variety of other relevant sources (Table 3)16–23.

Table 3.  Non-treatment-specific health state utilities and disutilities for complications.

Event/State	Utility/disutility	Reference
Type diabetes 2 no complications	0.814	23
Myocardial infarction, year of event	−0.129	23
Myocardial infarction, years 2+after event	0.736	23
Angina	0.682	23
Congestive heart failure	0.633	23
Stroke, year of event	−0.181	23
Stroke, years 2+ after event	0.545	23
Peripheral vascular disease	0.570	18
Microalbuminuria	0.814^a	–
Gross proteinuria	0.814^a	–
Hemodialysis	0.604	17
Peritoneal dialysis	0.612	17
Kidney transplant	0.782	32
Background diabetic retinopathy	0.790	19
Proliferative diabetic retinopathy	0.790^b	–
Macular edema	0.790	19
Severe vision loss/blindness	0.570	21
Cataract	0.620	22
Neuropathy	0.63	21
Healed diabetic ulcer	0.814^a	–
Active ulcer	0.750	20
Amputation, year of event	−0.109	23
Amputation, years 2+ after event	0.680	23

^aNo state-specific health utility identified—conservatively assumed to be equivalent to complication-free utility.

^bAssumption.

Disutilities were also applied to specific adverse events. Non-treatment-specific disutilities associated with both minor and major hypoglycemic events were applied even though no major events were reported in the trial24. Each disutility captured both the experience of minor/major hypoglycemia and the fear associated with this event24. A loss of utility of 0.0038 was included for each unit of BMI increase above 25 kg/m² (a default setting in the CORE model)25.

Disutilities associated with treatment effects such as nausea/vomiting and injection-site reaction (treatment-specific values) were conservatively applied to the intervention (ExQW) and to the switch treatment (basal insulin), but not to sitagliptin or pioglitazone. Each of these disutilities is an estimate based on the incidence of events and a disutility value derived from the literature9,26,27. Disutility resulting from nausea was applied only to the first year of treatment with ExQW and to the first year of switch therapy.

Costs

The analysis was conducted from the healthcare payer’s perspective to capture direct medical costs (i.e. complication and management costs). Pharmacy costs were excluded as the cost of ExQW in the US was not set at the time of the analysis. For balance, pharmacy costs of sitagliptin and pioglitazone were also excluded.

All complications costs were derived from published sources, except for the annual costs of treating complications with aspirin (81 mg), simvastatin (80 mg), and lisinopril (40 mg) (Table 4). Annual pharmacy costs for these drugs were all estimated based on the 2010 PriceRx Medispan Wholesale Acquisition Costs (WAC)28. Original values from the literature were inflated to reflect 2010 US dollars29. The effect of these three costs on the total cost per patient was marginal.

Table 4.  Direct unit of diabetes complications or events, adjusted to $US 2010.

Description of complication or event	Annual costs	Reference
Myocardial infarction, year of event	41,695	33
Myocardial infarction, each subsequent year	2200	34
Angina, year of onset	9850	33
Angina, each subsequent year	2039	34
Congestive heart failure, year of onset	30,066	33
Congestive heart failure, each subsequent year	3418	34
Stroke, year of event	14,104	34
Stroke, each subsequent year	14,104	33
Peripheral vascular disease, onset	9800	33
Hemodialysis	25,150	33
Peritoneal dialysis	29,609	33
Kidney transplant, first year	22,465	33
Kidney transplant, each subsequent year	22,465	33
Retinal photocoagulation	879	34
Severe vision loss/blindness, year of onset	3357	33
Severe vision loss/blindness, each subsequent year	3357^a	–
Cataract extraction	1094	33
Neuropathy, onset	3657	33
Uninfected ulcer	11,042^b	–
Infected ulcer	11042	33
Gangrene	25,125	33
Amputation, year of event	7371^c	33
Amputation, prosthesis	1271	34
Annual cost of aspirin	4^d	34
Annual cost of statins	144^d	34
Annual costs of ACE-I	50^d	34
Major hypoglycemic event	454	33
Ketoacidosis event	239	33
Lactic acid event	24,344	33
Edema onset	3116	33
Costs of screening for retinopathy	87	34
Costs of screening for nephropathy	29	34
Costs non-standard ulcer treatment	172	34

^aAssumption that the cost in subsequent years equals to the cost in 1st year.

^bSet equal to the cost of infected ulcer.

^cEstimated as weighted average of various types of amputations.

^dAnnual costs based on Medispan drug unit prices and doses assumptions agreed on with manufacturer.

Discounting, time horizon, and perspective

Health outcomes and costs were discounted at a rate of 3% per annum. A time horizon of 35 years was used to reflect that T2DM is a chronic disease and to capture diabetes-related complications and mortality over time.

Sensitivity analyses

A range of one-way sensitivity analyses was performed with key parameters varied to assess their effect on clinical and cost outcomes. For example, simulation time horizon was set to 5, 10, 15, 20, and 30 years, treatment failure was assumed for each pair of therapies after 4 and 5 years rather than 3 years, and change from baseline in HbA1_c associated with ExQW treatment was varied to both the higher (−1.35%) and lower (−1.74%) ends of the 95% confidence intervals (CI). Regarding the effect of health state utilities on the health benefit, several sensitivity analyses were performed in which the effect of BMI on QALYs was waived, disutility associated with nausea, injection-site reaction, and both nausea and injection-site reaction were one-by-one removed from ExQW treatment, and disutilities associated with minor and major hypoglycemia used in the base-case analysis were replaced with an alternative set of disutilities that did not capture the fear of having a hypoglycemic event26,27.

Results

Risks of complications

The 35-year cumulative incidence of 16 diabetes-related complications for the three treatment groups, and the relative risk of each complication in patients treated with ExQW compared with sitagliptin or pioglitazone were evaluated in the base-case analysis.

Compared with sitagliptin, the relative risks for 14 of 16 complications were projected to be lower for ExQW (Table 5). Relative risks ranged from 0.760 for end stage renal disease (ESRD) to 0.973 for cataract. First stroke and congestive heart failure were the two complications for which exenatide was associated with a higher relative risk (1.001 and 1.005, respectively).

Table 5.  Cumulative incidence and relative risk of diabetes complications for exenatide once weekly vs sitagliptin or pioglitazone.

Complication	Exenatide QW	Sitagliptin		Pioglitazone
	Cumulative incidence (%)	Cumulative incidence (%)	Relative risk	Cumulative incidence (%)	Relative risk
Cardiovascular disease
Congestive heart failure	38.968	38.755	1.005	41.588	0.937
Acute myocardial infarction	25.178	27.846	0.904	24.825	1.014
First stroke	26.573	26.559	1.001	25.764	1.031
Angina	16.292	17.057	0.955	15.535	1.049
Peripheral vascular disease	13.745	15.077	0.912	14.541	0.945
Renal disease
Microalbuminuria	43.154	46.571	0.927	44.742	0.965
Gross proteinuria	9.668	11.412	0.847	10.442	0.926
End-stage renal disease	1.400	1.843	0.760	1.614	0.867
Eye disease
Background retinopathy	27.182	30.040	0.905	28.522	0.953
Proliferative retinopathy	2.148	2.475	0.868	2.238	0.960
Severe vision loss	10.083	11.395	0.885	10.658	0.946
Macular edema	17.356	19.412	0.894	18.214	0.953
Cataract	11.234	11.542	0.973	11.426	0.983
Ulcer, neuropathy
First foot ulcer	23.835	25.241	0.944	24.496	0.973
First amputation	5.924	6.320	0.937	6.154	0.963
Onset of neuropathy	52.607	56.265	0.935	54.248	0.970

Compared with pioglitazone, relative risks for 13 complications were projected to be slightly lower for ExQW (range: 0.867 for ESRD to 0.983 for cataract) (Table 5). Risks for three cardiovascular complications (acute myocardial infarction [AMI], stroke, and angina) were projected to be slightly higher for ExQW (range: 1.014 for AMI to 1.049 for angina).

Incremental health and cost benefits

Compared with sitagliptin or pioglitazone, ExQW was projected to increase life expectancy by 0.28 (vs sitagliptin) and 0.17 years (vs pioglitazone) as well as QALY by 0.28 (vs sitagliptin) and 0.24 years (vs pioglitazone) (Tables 6 and 7).

Table 6.  Base-case clinical and economic outcomes associated with exenatide once weekly vs sitagliptin.

	Exenatide QW	Sitagliptin	Difference
Life expectancy, mean (SD), years	13.76 (0.17)	13.48 (0.18)	0.276 (0.242)
Quality-adjusted life expectancy, mean (SD)	9.56 (0.12)	9.28 (0.12)	0.284 (0.172)
Lifetime direct medical costs, mean (SD), $US	55,647 (2039)	57,862 (2159)	−2215 (3014)

Table 7.  Base-case clinical and economic outcomes associated with exenatide once weekly vs pioglitazone.

	Exenatide QW	Pioglitazone	Difference
Life expectancy, mean (SD), years	13.76 (0.17)	13.59 (0.17)	0.168 (0.231)
Quality-adjusted life expectancy, mean (SD)	9.56 (0.12)	9.32 (0.12)	0.244 (0.162)
Lifetime direct medical costs, mean (SD), $US	55,647 (2039)	56,580 (2007)	−933 (2786)

ExQW was also associated with lower complication costs: compared with sitagliptin or pioglitazone, ExQW was estimated to save US$2215 (vs sitagliptin) and US$933 (vs pioglitazone) lifetime direct cost per patient (Tables 8 and 9). In both cases, cost savings resulted primarily from lower cumulative incidences of cardiovascular diseases and neuropathic complications (e.g. ulcer, amputation).

Table 8.  Base-case direct medical costs associated with exenatide once weekly vs sitagliptin treatment strategies ($US 2010).

	Exenatide QW	Sitagliptin	Difference
Total costs*	55,647	57,862	−2215
Management	2746	2700	46
CVD	45,158	46,518	−1360
Renal	501	682	−181
Ulcer/amputation/ neuropathy	5149	5558	−409
Eye	2092	2405	−313

* No treatment costs were included as this is a cost-consequence analysis.

Table 9.  Base-case direct medical costs associated with exenatide once weekly vs pioglitazone treatment strategies ($US 2010).

	Exenatide QW	Pioglitazone	Difference
Total Costs*	55,647	56,580	−933
Management	2746	2713	33
CVD	45,158	45,688	−530
Renal	501	590	−89
Ulcer/amputation/ neuropathy	5149	5341	−192
Eye	2092	2249	−157

* No treatment costs were included as this is a cost-consequence analysis.

Sensitivity analyses

Results from sensitivity analyses for ExQW compared with sitagliptin are shown in Table 10. Assuming a shorter time horizon than the 35 years used in the base-case, the number of additional QALY gained was smaller. Varying the time horizon also affected the projected incremental costs associated with ExQW treatment and those of sitagliptin (pharmacy costs excluded).

Table 10.  Sensitivity analyses results for exenatide once weekly vs sitagliptin.

	Quality-adjusted life expectancy (QALYs)			Lifetime direct costs ($)
	Exenatide QW	Sitagliptin	Difference	Exenatide QW	Sitagliptin	Difference
Base case	9.56 (0.12)	9.28 (0.12)	0.28 (0.17)	55,647 (2039)	57,862 (2159)	−2215 (3014)
Time horizon and treatment duration:
5 year time horizon	3.25 (0.02)	3.23 (0.02)	0.02 (0.03)	7949 (526)	8420 (547)	−472 (769)
10 year time horizon	5.70 (0.05)	5.63 (0.05)	0.07 (0.07)	16,510 (889)	17,530 (1014)	−1019 (1375)
15 year time horizon	7.43 (0.08)	7.30 (0.07)	0.13 (0.10)	26,150 (1278)	27,948 (1309)	−1798 (1888)
20 year time horizon	8.56 (0.10)	8.37 (0.10)	0.19 (0.13)	36,636 (1544)	39,083 (1589)	−2447 (2178)
30 year time horizon	9.45 (0.12)	9.19 (0.12)	0.26 (0.16)	52,804 (1936)	55,108 (1989)	−2304 (2791)
Assumption of 4 years of ExQW treatment	9.55 (0.13)	9.28 (0.12)	0.27 (0.19)	55,843 (1970)	57,863 (2163)	−2020 (2819)
Assumption of 5 years of ExQW treatment	9.55 (0.12)	9.28 (0.12)	0.26 (0.17)	55,524 (2052)	57,863 (2165)	−2339 (3066)
Clinical values:
ExQW HbA1c–upper bound of 95% CI (−1.35%)	9.50 (0.12)	9.28 (0.12)	0.23 (0.17)	56,274 (2086)	57,863 (2159)	−1589 (3130)
ExQW HbA1c–lower bound of 95% CI (−1.74%)	9.60 (0.12)	9.28 (0.12)	0.32 (0.18)	55,441 (1908)	57,863 (2159)	−2422 (2789)
Health state utilities:
No BMI adjustment for QALY	9.89 (0.13)	9.63 (0.13)	0.26 (0.18)	55,647 (2039)	57,863 (2159)	−2216 (3014)
No injection site reaction disutility	9.56 (0.12)	9.28 (0.12)	0.29 (0.17)	55,647 (2040)	57,863 (2159)	−2216 (3015)
No nausea disutility	9.56 (0.12)	9.28 (0.12)	0.29 (0.17)	55,647 (2040)	57,863 (2159)	−2216 (3015)
No injection reaction and nausea disutilities	9.57 (0.12)	9.28 (0.12)	0.29 (0.17)	55,647 (2040)	57,863 (2159)	−2216 (3015)
CDM default utility of hypoglycemia	9.53 (0.12)	9.25 (0.12)	0.28 (0.17)	55,647 (2039)	57,863 (2159)	−2216 (3014)
Discount rate:
Discount rate = 0%	13.19 (0.20)	12.70 (0.20)	0.48 (0.28)	93,126 (3404)	95,779 (3628)	−2653 (5002)
Discount rate = 5%	7.96 (0.09)	7.75 (0.09)	0.21 (0.13)	41,259 (1558)	43,162 (1636)	−1903 (2311)

Setting the HbA1_c change from baseline for ExQW to the lower bound (−1.74%) of the 95% CI resulted in an additional 0.04 incremental QALY increase, whereas setting it to the upper bound (−1.35%) resulted in a 0.05 incremental QALY decrease. With these changes to HbA1_c, incremental costs were projected to be −$1589 and −$2422, respectively, compared with −$2215 in the base-case analysis.

Assumptions regarding the duration of treatment effect had a minor influence on the incremental QALY and costs associated with ExQW compared with sitagliptin. When the time of switch to basal insulin was changed from 3 to 4 and 5 years, incremental QALYs decreased from 0.28 QALY to 0.27 QALY and 0.26 QALY, respectively, and incremental cost changed from −$2215 to −$2020 and −$2339, respectively.

Incremental QALYs were not sensitive to the utility parameters tests in the sensitivity analyses except for BMI, which had a marginal impact on the results. When removing the disutility associated with BMI, the incremental QALY was slightly lower than it was in the base-case analysis (0.26 vs 0.28 QALY, respectively). Some incremental costs were slightly different from those of base case (−$2216 vs −$2215, respectively); differences resulted from rounding methods.

Assuming a discount rate of 0% rather than the 3% used in the base-case, the number of additional QALY gained was larger (0.48 vs 0.28 QALY), and the projected incremental cost difference associated with ExQW was larger (−$2653 vs −$2215). Assuming a discount rate of 5% rather than 3%, the number of additional QALY gained was slightly smaller (0.21 vs 0.28 QALY), and the projected incremental cost difference associated with ExQW was also slightly smaller (−$1903 vs −$2215).

Results from sensitivity analyses for ExQW compared with pioglitazone are shown in Table 11. Overall, the results of these sensitivity analyses were similar to those for ExQW compared with sitagliptin. With shorter time horizons, incremental QALY was smaller and ExQW was overall projected to be more cost-saving than in the base-case.

Table 11.  Sensitivity analyses results for exenatide once weekly vs pioglitazone.

	Quality-adjusted life expectancy (QALYs)			Lifetime direct costs ($)
	Exenatide QW	Pioglitazone	Difference	Exenatide QW	Pioglitazone	Difference
Base case	9.56 (0.12)	9.32 (0.12)	0.24 (0.16)	55,647 (2039)	56,580 (2007)	−933 (2786)
Time horizon and treatment duration:
5 year time horizon	3.25 (0.02)	3.22 (0.02)	0.03 (0.03)	7949 (526)	8093 (555)	−144 (0758)
10 year time horizon	5.70 (0.05)	5.62 (0.05)	0.07 (0.07)	16,510 (889)	16,882 (947)	−371 (1286)
15 year time horizon	7.43 (0.08)	7.32 (0.07)	0.11 (0.11)	26,150 (1278)	27,002 (1262)	−852 (1819)
20 year time horizon	8.56 (0.10)	8.40 (0.09)	0.16 (0.13)	36,636 (1544)	37,769 (1524)	−1133 (2173)
30 year time horizon	9.45 (0.12)	9.24 (0.12)	0.21 (0.16)	52,804 (1936)	53,606 (2018)	−802 (2886)
Assumption of 4 years of ExQW treatment	9.55 (0.13)	9.32 (0.12)	0.23 (0.18)	55,843 (1970)	56,581 (2008)	−738 (2714)
Assumption of 5 years of ExQW treatment	9.55 (0.12)	9.32 (0.12)	0.22 (0.17)	55,524 (2052)	56,581 (2010)	−1057 (2958)
Clinical values:
ExQW HbA1c–upper bound of 95% CI (−1.35%)	9.50 (0.12)	9.32 (0.12)	0.19 (0.17)	56,274 (2086)	56,580 (2007)	−306 (2869)
ExQW HbA1c–lower bound of 95% CI (−1.74%)	9.60 (0.12)	9.32 (0.12)	0.28 (0.17)	55,441 (1908)	56,580 (2007)	−1139 (2663)
Health state utilities:
No BMI adjustment for QALY	9.89 (0.13)	9.74 (0.12)	0.15 (0.17)	55,647 (2039)	56,580 (2007)	−933 (2786)
No injection site reaction disutility	9.56 (0.12)	9.32 (0.12)	0.25 (0.16)	55,646 (2040)	56,580 (2007)	−934 (2787)
No nausea disutility	9.56 (0.12)	9.32 (0.12)	0.25 (0.16)	55,646 (2040)	56,580 (2007)	−934 (2787)
No injection reaction and nausea disutilities	9.57 (0.12)	9.32 (0.12)	0.25 (0.16)	55,646 (2040)	56,580 (2007)	−934 (2787)
CDM default utility of hypoglycemia	9.53 (0.12)	9.29 (0.12)	0.24 (0.16)	55,647 (2039)	56,580 (2007)	−933 (2786)
Discount rate:
Discount rate = 0%	13.19 (0.20)	12.79 (0.19)	0.40 (0.26)	93,126 (3404)	93,926 (3402)	−801 (4659)
Discount rate = 5%	7.96 (0.09)	7.78 (0.09)	0.18 (0.12)	41,259 (1558)	42,134 (1517)	−875 (2127)

Compared with the base-case analysis, changing the HbA1_c improvement for ExQW to the lower bound (−1.74%) of the 95% CI resulted in an additional 0.04 incremental QALY increase, whereas setting it to the upper bound (−1.35%) resulted in a 0.05 incremental QALY decrease. Incremental cost projections with these changes to HbA1_c were −$1139 and −$306, respectively, compared with −$933 in the base-case analysis.

Sensitivity analyses that altered the duration of treatment indicated that this parameter had only small effects on QALY and incremental costs. When the switch to basal insulin was changed from 3 to 4 and 5 years, projected incremental QALYs decreased from 0.24 to 0.23 and 0.22, respectively, and projected incremental cost changed from −$933 to −$738 and −$1057, respectively.

The exclusion of treatment-related disutilities for injection-site reaction and nausea had little effect on incremental QALY in the comparison of ExQW with pioglitazone. The removal of the effect of BMI resulted in a 0.09 decrease in incremental QALY, compared with base-case analysis.

Assuming a discount rate of 0% rather than the 3% used in the base-case, the number of additional QALY gained was larger (0.40 vs 0.24 QALY); the projected incremental cost difference associated with ExQW was slightly smaller (−$801 vs −$993). Assuming a discount rate of 5% rather than 3%, the number of additional QALY gained was slightly smaller (0.18 vs 0.24 QALY), and the projected incremental cost difference associated with ExQW was slightly smaller (−$875 vs −$993).

Discussion

Results from the DURATION-2 trial demonstrated that treatment with ExQW was associated with a significantly greater reduction in mean HbA1_c than treatment with either sitagliptin (p < 0.0001) or pioglitazone (p = 0.0165)9. Additionally, patients treated with ExQW experienced significantly greater weight loss than did those treated with sitagliptin (p = 0.0002); patients treated with pioglitazone gained weight.

The objective of the current analysis was to estimate the effect on complications and outcomes when treating T2DM patients with ExQW compared with sitagliptin or pioglitazone. Baseline characteristics of the patients in the DURATION-2 study as well as the clinical results of that study were used to populate a validated economic model. The base-case analyses estimated that treatment with ExQW was associated with lifetime QALY gain of 0.28 year per patient (vs sitagliptin) and of 0.24 year per patients (vs pioglitazone). ExQW was also projected to generate cost savings of $2215 per patient (vs sitagliptin) and $933 per patient (vs pioglitazone).

The projected cost savings were primarily the result of lower cumulative incidences of cardiovascular diseases and neuropathic complications (e.g. ulcer, amputation) associated with ExQW treatment compared with sitagliptin or pioglitazone treatment. For most of the 16 complications captured in the model, the relative risks were projected to be lower with ExQW than with sitagliptin (14 of 16 complications) or with pioglitazone (13 of 16 complications).

Type 2 diabetes is a progressive disease. Therefore, glycemic control typically deteriorates over time30. A stepwise intensification of therapy is recommended in published treatment guidelines, from monotherapy with oral anti-hyperglycemic drugs, to dual and triple therapy and eventually to insulin therapy1,2. Thus, it was assumed in the base-case analysis that patients would switch to basal insulin after 3 years of treatment with ExQW, sitagliptin or pioglitazone. Sensitivity analyses performed with the time of the switch set to 4 or 5 years showed that the duration of treatment had little effect on the projected incremental QALY and costs associated with ExQW compared with sitagliptin or pioglitazone.

Sensitivity analyses conducted on shorter time horizons than in the base-case analysis (i.e. 35 years) generated smaller incremental QALY gained for ExQW compared with either comparator. Changes in BMI have been shown to influence utility scores of injectable anti-hyperglycemic drugs25. The impact of BMI was modeled in the base-case analysis. In contrast with the other disutilities modeled (hypoglycemia, injection-site reaction, and nausea), sensitivity analyses showed that BMI had a slight effect on incremental results, especially in the comparison of ExQW with pioglitazone, where the comparative benefit of ExQW in terms of weight lost was greater.

To our knowledge, this is the first study to assess lifetime outcomes associated with ExQW compared with those of sitagliptin or pioglitazone from a US payer perspective. However, a study that also used the CDM to estimate clinical and economic outcomes for ExQW compared with sitagliptin projected clinical outcomes similar to those in this study31. In addition, it was projected that ExQW would be cost-effective compared with sitagliptin from the UK National Health Service payer perspective, with a cost per QALY gained of £6418.

There are several limitations to the current study. Relatively short-term changes (26 weeks) in surrogate end-points (e.g. HbA1_c, weight, complications) were used to project lifetime effects on clinical outcomes. However, this caveat is inherent to any economic evaluation using trial data to project incremental costs and health benefits over a patient’s lifetime. However, transition probabilities used in the CDM have been validated against clinical and epidemiological data11. Furthermore, the impact of the frequency of treatment administration could have been tested since ExQW patients are treated once per week as opposed to once daily in patients under sitagliptin or pioglitazone. It was assumed this effect would be washed out by that of injection-site reactions in clinical practice (in the clinical trial, the sitagliptin and pioglitazone groups received placebo QW injections; all three groups experienced injection-site reactions). The representativeness of the US T2DM population may be limited in this analysis, since the DURATION-2 trial involved patients from sites located outside of the US (e.g. India and Mexico). Lastly, pharmacy costs were excluded from these analyses.

Conclusion

Over the lifetime of a patient with T2DM, treatment with ExQW was projected to improve health and decrease complication costs compared with treatment with sitagliptin or pioglitazone. The cost-effectiveness of ExQW compared with those of both sitagliptin and pioglitazone needs to be assessed after pricing of ExQW in the US has been established.

Transparency

Declaration of funding

This study was funded by Amylin Pharmaceuticals, Inc. San Diego, US.

Declaration of financial/other relationships

MBDY and JAG are employees and stockholders of Amylin pharmaceuticals. A-LG, YS, and AL are employees of IMS Health. IMS Health has received researched funding from Merck, Takeda, Sanofi Aventis and Amylin Pharmaceuticals, Inc. who manufacture drugs described in this analysis. JB owns stock in Amylin Pharmaceuticals.

Acknowledgments

The authors thank Steven C. Brunell for editorial assistance.