The cost-effectiveness of exenatide once weekly compared with exenatide twice daily and insulin glargine for the treatment of patients with type two diabetes and body mass index ≥30 kg/m² in Spain

Abstract

Objective:

The objective of this analysis was to estimate the cost-effectiveness of exenatide once weekly (EQW) for the treatment of type two diabetes mellitus (T2DM) in Spain. EQW was compared against exenatide twice daily (EBID) and insulin glargine (IG).

Methods:

The IMS CORE Diabetes Model was used to project clinical and economic outcomes for patients with T2DM treated with EQW, EBID, and IG. Treatment effects and patient baseline characteristics were taken from the DURATION 3 and pooled DURATION 1 and 5 studies, in the comparison against IG and EBID, respectively. Unit costs and health state utility values were derived from published sources. To reflect diabetes progression, patients started on EQW or EBID, switching to insulin glargine after 3 years. The analysis was conducted from the perspective of the Spanish National Health Service over a time horizon of 35 years with costs and outcomes discounted at 3%. The base case included patients with a BMI > 30 kg/m², which is in line with current prescription restrictions in Spain. Uncertainty was addressed through extensive one-way sensitivity analyses around key model parameters and a comprehensive probabilistic sensitivity analysis.

Results:

When compared with EBID, EQW was the dominant strategy, i.e., less costly and more effective. When compared to IG, the incremental cost-effectiveness ratio was estimated at €12,084 per QALY gained. Sensitivity analysis indicated that the model projections were robust to the various scenarios tested.

Limitations:

Primary limitations of the analysis are common to other T2DM analyses and include the extrapolation of short-term clinical data to the 35 year time horizon and uncertainty around optimum treatment durations.

Conclusion:

The analyses indicate that EQW is a cost-effective option for the treatment of T2DM patients in Spain for patients with a BMI > 30 kg/m² considering a willingness-to-pay threshold of €30,000 per QALY gained.

Keywords::

• Diabetes
• Cost-effectiveness
• Spain
• Exenatide
• Insulin glargine
• Type 2 diabetes

Introduction

Diabetes is increasingly prevalent in Western society affecting up to 11.63% of adults at any one time1. The majority of diabetes patients (∼90%) will suffer from type 2 diabetes mellitus (T2DM), a condition that is often under-reported, with the American Diabetes Association estimating that about one third of individuals with diabetes are unaware that they have the disease2,3. T2DM historically occurs in people older than 40 years and the incidence increases with age with T2DM affecting more than 25% of people aged over 65 years4. However, as society is becoming increasingly obese and children are increasingly subject to poor diet and sedentary lifestyles, T2DM often arises at a younger age5–7. A recent study carried out in Spain estimating both known and undiagnosed rates of diabetes reported an important increase in the prevalence of T2DM among people aged 18 years and older. Overall prevalence was estimated at 13.8%, of which about half had unknown diabetes (6%)8. Further, the Eurostat data indicate that, in Spain, the death rate associated with T2DM is now estimated at 11.8 per 100,000 inhabitants9.

The high prevalence and the socio-economic consequences of T2DM make this disease an important public health concern. The life expectancy of patients with diabetes can be reduced by up to 10 years compared to the general population, mainly due to the increased risk of cardiovascular death and stroke10,11. The total annual health cost per patient with diabetes from a societal perspective in Spain was estimated to be €4278 (of which €2504 were direct costs) (1999 values)12. The management of T2DM-related complications is the major underlying cost driver, with half of diabetes-related medical expenditure attributed to hospital inpatient care13–15. Improving glycaemic control and other cardiovascular risk factors can substantially reduce the risk of diabetes-related complications16,17 and associated costs18–22.

Exenatide is a glucagon-like peptide-1 (GLP-1) receptor agonist and is currently approved for twice-daily (EBID) and once weekly (EQW) administration23. Recent clinical studies conducted as part of the DURATION clinical trial programme have shown that EQW improved glycaemic control compared with exenatide BID, with a similar reduction in body weight24,25 and resulted in improvement in glycaemic control compared to sitagliptin, pioglitazone, and insulin glargine (IG)26,27.

The objective of this study was to assess the cost-effectiveness of EQW when compared to EBID and IG for the treatment of T2DM patients with a body mass index (BMI) ≥ 30 kg/m² in Spain. This base case is in line with the current restriction for prescription in Spain.

Methods

The IMS Centre for Outcomes Research (CORE) Diabetes Model was used to assess the long-term clinical and economic outcomes of the interventions (Figure 1). This model is a robust and well-established simulation model validated against results from both clinical and epidemiological studies28,29. Analyses were conducted for the two comparisons of interest. Patient characteristics (including the prevalence of diabetes complications) and treatment parameters are described in detail below. The analyses were conducted from the perspective of the Spanish National Health Service, as it represents the main audience for economic evaluations of healthcare technologies, which are predominantly financed with public funds. Consequently, the analyses captured all relevant direct medical costs (pharmacy costs plus complication and management costs). The patient population comprised patients with T2DM and reflected the RCT trial populations on which the analyses were based. Spanish-specific unit cost data were used with an emphasis on costs derived from diabetes-specific sources where possible. Unit costs were reported in 2012 values, inflated when necessary, using the Consumer Price Index published by the Spanish National Statistics Institute30. Health state utilities were estimated based on published data31–40 and are described in detail below. Given the chronic nature of T2DM, a lifetime horizon representing 35 years or death, whichever came first, was used in the analysis. Spanish recommendations on the economic evaluation of health technologies were followed throughout model development39. The model output was in the form of estimated incremental cost-effectiveness ratios (ICERs). Costs and benefits were both discounted at 3% annually in accordance with Spanish guidelines41.

Figure 1. Diagrammatic representation of the IMS CORE diabetes model. Figure from Palmer et al.27 (IMS Health is the owner of the CORE Diabetes Model).

Baseline characteristics of patients

Exenatide twice daily comparison

The baseline patient cohort was defined according to data from the DURATION 1 and 5 studies where EQW was compared against EBID42,43. Although both studies showed a superior reduction in glycated haemoglobin (HbA1c) with EQW therapy compared to EBID therapy, the magnitude of the difference in HbA1c between the two studies differed. There was no strong rationale for selecting the results from one study over another and, therefore, the data from the two studies were pooled. Pooling was considered appropriate as the studies were similar in terms of participants (inclusion and exclusion criteria identical), treatments (intensity, dose, duration, mode of administration were all similar), outcomes (HbA1c was re-defined at same time point for pooling), design (parallel RCTs), and conduct (allocation, concealment, and analysis methods).

Insulin Glargine comparison

The baseline cohort was defined according to data from DURATION 3, where EQW was compared with IG44. However, baseline rates of diabetes complications were not reported in DURATION 3, therefore the rates reported in the combined DURATION 1 and 5 data were used. This was considered appropriate as these studies were from a similar patient population (i.e., patients who have failed on up to two oral anti-diabetes drugs, OADs).

Other patient characteristics such as the proportion of smokers or the weekly average alcohol consumption were derived from official European statistics sources1. Baseline characteristics can be found in Table 1. Table 2 reports the modelled rates of baseline diabetes complications.

Table 1. Baseline characteristics of analysis cohort (Mean, SD).

Characteristic	EQW-EBID analysis*	EQW-IG analysis*
Patients (n)	414	271
HbA1c (%)	8.32, 1.06	8.34, 1.06
Age (years)	54.75, 10.33	57, 9.73
Male (proportion)	0.514	0.498
Ethnicity (proportion)
Caucasian	0.73	0.89
African-descent	0.08	0.01
Asian/Pacific Islander	0.01	0.00
Hispanic	0.18	0.02
Native American and others	0.00	0.09
Duration of diabetes (years)	6.53, 4.87	7, 5.65
	99	99
Systolic blood pressure (mmHg)	129.14, 13.69	135.97, 16.92
Baseline total cholesterol (mg/dL)	182.93, 47.48	190.2, 39.82
Baseline HDL-C (mg/dL)	44.6, 10.65	44.76, 11.13
Baseline LDL-C (mg/dL)	102.42, 39.3	106.32, 35.79
Baseline triglycerides (mg/dL)	203.54, 176.18	201.35, 109.3
Body mass index (kg/m^2)	36.25, 4.21	35.47, 4.1
Proportion smoker†	0.27	0.29
Cigarettes per day†	7	7
Alcohol consumption (ml/week)†	238	238

EQW, exenatide once weekly; EBID, exenatide twice daily; IG, insulin glargine.

*The analysis cohorts are sub-sets of the pooled DURATION 1 and 5 ITT data, n = 547 (EQW-EBID) and the DURATION 3 ITT data, n = 456 (EQW-IG), comprising those patients with a baseline BMI ≥ 30 kg/m²; †The proportion of smokers, cigarettes per day, and alcohol consumption were extracted from the European Health for all database1.

Table 2. Diabetes complications at baseline.

Prevalent conditions	Proportion*
Myocardial infarction	0.033
Angina	0.077
Peripheral vascular disease	0.018
Stroke	0.009
Congestive heart failure	0.002
Atrial fibrilation	0.002
Left ventricular hypertrophy	0.002
Microalbuminuria	0.004
Gross proteinuria	0.024
End-stage renal disease	0.000
Background diabetic retinopathy	0.029
Proliferative diabetic retinopathy	0.000
Severe vision loss	0.002
Macular oedema	0.016
Cataract	0.091
Uninfected ulcer	0.002
Infected ulcer	0.000
Healed ulcer	0.002
Amputation	0.000
Neuropathy	0.155

*Proportions based on pooled DURATION 1 and 5 ITT baseline population data (all BMI, n = 547) and applied to both analysis cohorts (EQW-EBID and EQW-IG).

Treatment effects

Treatment effects associated with the three treatments were drawn from the efficacy and safety data from the respective clinical trials24,43,45. Changes in HbA1c, systolic blood pressure (SBP), total cholesterol (TC), high density lipoprotein (HDL)-cholesterol, low density lipoprotein (LDL)-cholesterol, and triglycerides were simulated along with changes in BMI and hypoglycaemic events, as listed in Table 3.

Table 3. Treatment effects from DURATION 1 and 5 (pooled) and DURATION 3 (Mean change from baseline, SD).

	DURATION 1 and 5		DURATION 3
	EQW	EBID	EQW	IG
Patients (n)	213	201	136	135
HbA1c (%)	−1.69, 1.31	−1.2, 1.41	−1.33, 0.82	−1.22, 0.81
Systolic blood pressure (mmHg)	−3.4, 15.76	−1.65, 15.73	−4.54, 16.44	−1.31, 16.27
Total cholesterol (mg/dL)	−12.77, 35.17	−3.35, 35.3	−3.07, 31.72	0.49, 31.60
HDL-cholesterol (mg/dL)	0.16, 8.32	−0.14, 8.36	1.11, 7.58	1.18, 7.55
LDL-cholesterol (mg/dL)	−4.93, 29.77	−0.72, 29.63	−1.24, 26.82	3.96, 26.72
Triglycerides (mg/dL)	−45.49, 158.51	−3.8, 92.55	−15.72, 102.68	−17.86, 123.67
BMI (kg/m^2)	−1.13, 2.19	−0.86, 2.12	−0.97, 1.17	0.58, 1.16
Major hypoglycaemia*	0.00	0.00	0.00	0.00
Minor hypoglycaemia*	21	75.58	22.72	118.07

EQW, exenatide once weekly; EBID, exenatide twice daily; IG, insulin glargine.

*Hypoglycaemia events/100 patient years based on sub-sets of the pooled DUR 1/5 (EQW-EBID) and DUR 3 (EQW-IG) ITT populations, comprising patients with BMI ≥ 30 kg/m².

T2DM is a chronic and progressive disease; glycaemic control typically worsens over time46 and it is likely that patients on EQW or EBID will need to intensify their treatment regime at some point47. Treatment guidelines address this by recommending stepwise intensification of treatment from monotherapy with OADs through to insulin therapy48–50. Therefore, in the model it was assumed that patients do not remain on their initial therapy for the entire duration of the analysis. A common starting regimen for insulin is the initiation of long-acting insulin51. In order to reflect usual clinical practice, after 3 years of initial treatment both exenatide treatment groups were switched to insulin glargine.

In the model the treatment effect was applied in the first year of treatment. In subsequent years, the UK Prospective Diabetes Study (UKPDS) regression approach42 was applied to predict the progression of HbA1c over the course of the simulation. The level of HbA1c was determined for each year according to patient baseline HbA1c, as reported in each of the DURATION clinical trials, and the duration of diabetes as reported in the UKPDS Outcomes Model, which represents the natural progression of HbA1c within the UKPDS cohort. For BMI progression, the respective weight change estimates were applied at the beginning of the simulations, and then it was assumed that patients regained their baseline weight at the initiation of insulin. Note that the switch to insulin glargine results in an initial expected weight gain which is applied across both interventions.

Health state utilities and disutilities

For T2DM and its complications, health state utilities for the model were derived wherever possible from the UKPDS52, supplemented with data from other sources as necessary (Table 4).

Table 4. Health state utilities and disutilities.

Health state		Reference
T2DM without complications	0.814	European Health for all database1
Myocardial infarction, year of event	−0.129	European Health for all database1
Myocardial infarction, starting year 2	0.736	European Health for all database1
Angina	0.682	European Health for all database1
Congestive heart failure	0.633	European Health for all database1
Stroke, year of event	−0.181	European Health for all database1
Stroke, starting year 2	0.545	European Health for all database1
Peripheral vascular disease	0.570	Hogan et al.2
Microalbuminuria	0.814	European Health for all database1
Gross proteinuria	0.814	European Health for all database1
Haemodialysis	0.621	Harris et al.3
Peritoneal dialysis	0.581	Harris et al.3
Renal transplant	0.762	Hogan et al.2
Background diabetic retinopathy	0.790	2011 National Diabetes Fact Sheet4
Background diabetic retinopathy wrongly treated	0.790	2011 National Diabetes Fact Sheet4
Proliferative diabetic retinopathy laser treated	0.790	2011 National Diabetes Fact Sheet4
Proliferative diabetic retinopathy no laser	0.790	2011 National Diabetes Fact Sheet4
Macular oedema	0.790	2011 National Diabetes Fact Sheet4
Severe vision loss	0.702	European Health for all database1
Cataract	0.748	Molnar5
Neuropathy	0.630	Mooy et al.6
Healed ulcer	0.814	European Health for all database1
Active ulcer	0.750	Pinhas-Hamiel and Zeitler7
Amputation event	−0.109	European Health for all database1
Post-amputation	0.680	European Health for all database1
Major hypoglycaemia event	−0.012	Soriguer et al.8
Minor hypoglycaemia event	−0.004	Soriguer et al.8
Excess BMI (each unit above 25 kg/m^2)	−0.0061	Eurostat9

Nausea disutility was applied within the model based on nausea rates observed in the DURATION clinical trials. In the EBID analysis, 26.0% of patients treated with EQW and 38.9% of patients treated with EBID experienced transient nausea (data from the DURATION 1 and 5 trials); in the glargine analysis, 12.9% of patients treated with EQW and 1.3% of patients treated with insulin glargine experienced transient nausea (data from the DURATION 3 trial). The disutility for nausea (−0.04) was taken from the Matza et al.53 study. Evidence suggests that nausea incidence decreases over time54, the model reflects this and the disutility was applied to the affected patients only for the first 6 months of the analysis period.

Cost data

The annual therapy cost included the drug costs, based on established or reported dosing regimens55–58 and the blood glucose monitoring cost, detailed in Tables 5 and 6. Needles and lancets were assumed to have zero cost. The self-blood glucose monitoring frequency used in the model follows that recommended by Owens et al.59. For patients using OAD regimens which include sulphonylurea, 3-times weekly monitoring is recommended; for regimens including basal insulin, daily monitoring is recommended. No monitoring is required for patients using OAD regimens that do not include sulphonylurea. During monitoring, it was assumed that patients use a new strip for each blood glucose test.

Table 5. Annual drug costs.

Drug	Dose regimen	Cost/day (EUR)*	Annual cost (EUR)
Exenatide once weekly	2 mg sc QW10	4.92	1797.03
Exenatide twice daily	10 mcg BID11	4.28	1561.97
Insulin glargine (year 1)	31.1 IU daily12	1.53	560.32
Insulin glargine (subsequent years)	40 IU daily13	1.97	717.01

*Drug costs are expressed in public price excluding VAT and were derived from the General Spanish Council of Pharmacists14.

Table 6. Self-blood-glucose monitoring cost.

Regimen	% pts on SU at baseline	Tests/week*	Tests/year	Unit costs (EUR)^$	Annual cost (EUR)
EQW (EBID analysis)†	33.5%	1.01	52.26	0.87	45.47
EQW (IG analysis)^#	29.6%	0.88	46.17	0.87	40.17
EBID†	33.5%	1.01	52.26	0.87	45.47
IG	NA	7	365.25	0.87	317.77

EQW, exenatide once weekly; EBID, exenatide twice daily; IG, insulin glargine.

*Self blood-glucose monitoring (SBGM) required on daily basis for IG but not required for EQW/EBID; frequency of SBGM estimated for the EQW/EBID patient cohort based on % ITT EQW/EBID patients on sulfonylurea (SU) in †DURATION 1 and 5 and #DURATION 3, assuming three tests per week for SU patients; ^$Assumption based on costs reported in Oliva et al.15

Spanish-specific unit cost data were used throughout, as listed in Table 7. Emphasis was placed on the use of costs derived from diabetes-specific sources where possible and in line with published evidence that the cost of treating complications in patients with diabetes may well be higher than in the general population60. Minor hypoglycaemia was assumed to have zero cost.

Table 7. Complication and management-related costs.

	2012 (EUR)	Reference
Management costs
Annual statins	450.72	Gaede et al.16
Annual aspirin	27.70	Gaede et al.16
Annual ACE-I	138.49	Gaede et al.16
Annual eye screening eye	76.80	Gaede et al.16
Annual screening for MA	13.79	Gaede et al.16
Annual screening for GRP	13.79	Gaede et al.16
Annual foot screening programme	13.85	Gaede et al.16
Stopping ACE due to SE	53.28	Assumption based on Gaede et al.16
Non-standard ulcer treatment (e.g., Regranex)	268.56	Assumption based on Oliva et al.15
Direct costs of CVD complications
Myocardial infarction 1st year	24,269.74	Gaede et al.16
Myocardial infarction 2nd+ years	976.88	Gaede et al.16
Angina 1st year	2778.61	Gaede et al.16
Angina 2nd+ years	2778.61	Assumption based on Oliva et al.15
Congestive heart failure 1st year	6254.71	Gaede et al.16
Congestive heart failure 2nd+ years	6254.71	Gaede et al.16
Stroke 1st year	5839.24	Gaede et al.16
Stroke 2nd+ years	2168.00	Gaede et al.16
Stroke death within 30 days	4030.06	Gaede et al.16
Peripheral vascular disease 1st year	1786.09	UKPDS17
Peripheral vascular disease 2nd+ years	1786.09	Assumption based on UKPDS17
Direct costs of renal complications
Haemodialysis 1st year	39,322.35	Gaede et al.16
Haemodialysis 2+ years	39,322.35	Assumption based on Oliva et al.15
Peritoneal Dialysis 1st year	41,176.85	Gaede et al.16
Peritoneal Dialysis 2+ years	41,176.85	Assumption based on Oliva et al.15
Renal transplant costs 1st year	35,717.83	Gaede et al.16
Renal transplant 2+ years	10,495.02	Gaede et al.16
Direct costs of acute events
Major hypoglycaemia (per event)	425.54	Gaede et al.16
Minor hypoglycaemia (per event)	0	Assumed
Direct costs of eye disease
Laser treatment	222.84	Gaede et al.16
Cataract operation	1310.62	Gaede et al.16
Cost following cataract operation	1151.99	Gaede et al.16
Cost blindness—year of onset	0.00	Assumption based on Oliva et al.15
Cost blindness—following years	0.00	Assumption based on Oliva et al.15
Direct costs of neuropathy/foot ulcer/amputation
Neuropathy 1st year	3186.53	Gaede et al.16
Neuropathy 2nd+ years	0.00	Assumption based on Oliva et al.15
Amputation (event based)	18,616.83	Gaede et al.16
Amputation prosthesis (event based)	3122.32	Gaede et al.16
Gangrene treatment (monthly)	7064.25	Gaede et al.16
Healed ulcer	0.00	Assumed
Infected ulcer (monthly)	2308.50	Assumption based on Oliva et al.15
Standard uninfected ulcer (monthly)	1319.30	Assumption based on Oliva et al.15
Healed ulcer history of amputation	0.00	Assumed

ACE-I, Angiotensin-converting enzyme inhibitor; MA, Microalbuminuria; GRP, Gross proteinuria.

Handling uncertainty

A comprehensive series of one-way sensitivity analyses were conducted to assess the robustness of model outcomes to plausible changes in key model parameters. These were conducted both on cost and effectiveness parameters. The effectiveness parameters assessed were: the changes from baseline in HbA1c, weight and SBP (varied by trial-reported upper and lower 95% confidence intervals), utilities (scenarios described below), time-horizon (10 and 20 years), discount rate (0% and 6%), time on treatment (5 and 10 years), and complication and treatment-related costs (±20%).

The effect of treatment-related utilities was explored in a series of scenario analyses. Variation 1 included application of injection site reaction disutility (0.011) which was derived from Boye et al.61, applied for patient lifetimes for affected patients. The percentage of patients affected was assumed to be the percentage of patients in each treatment group with injection site reaction as an adverse event in the respective DURATION trial. Variation 2 included injection site reaction disutility as in Variation 1, together with dose flexibility utility (0.006) applied to both treatments, and dose frequency utility for once-weekly administration (0.023) applied for as long as patients remained on EQW treatment. These were also derived from Boye et al.61. Variation 3 considered a utility change related to body weight derived from Matza et al.62, rather than from Bagust63 as in the base case. Variation 4 considered only health-state utilities related to complications (no treatment-related utilities or disutilities applied).

Probabilistic sensitivity analysis (PSA) was performed using a non-parametric bootstrapping approach in which the progression of diabetes was simulated in a cohort of 25,000 patients, repeated 500-times to calculate the mean and standard deviation (SD) for costs, life expectancy, and quality-adjusted life expectancy. Results from the 500 iterations were used to generate scatter plot diagrams and cost-acceptability curves comparing the two treatment regimens included in each comparison.

Results

EQW vs EBID: base-case analysis

In the base case analysis, treatment with EQW was projected to increase quality-adjusted life expectancy vs treatment with EBID. EQW was the dominant strategy, i.e., more effective and less costly. EQW was associated with a lower cumulative incidence of almost all simulated complications. Stroke (event and death) were the only exception. A breakdown of lifetime costs is provided in Table 8. The model predicts that the difference in acquisition costs of €889 is offset by the reduction in costs associated with complications.

Table 8. Base-case and sensitivity analysis results: EQW vs EBID.

Sensitivity analysis	Quality-adjusted life years (QALYs)			Costs (EUR)			ICER (EUR/QALY gained)
	EQW	EBID	Difference	EQW	EBID	Difference
Base case (BMI ≥ 30 kg/m^2)	9.197	8.956	0.241	€50,434	€50,949	−€515	EQW Dominant
Probabilistic sensitivity analysis	8.747	8.526	0.220	€53,089	€53,455	−€366	EQW Dominant
A1c = Upper CI (−1.57%)	9.223	8.592	0.631	€50,221	€45,327	€4894	€7752
A1c = Lower CI (−1.85%)	8.793	8.592	0.201	€45,158	€45,327	−€169	EQW Dominant
Weight = Upper CI (0.84 kg/m^2)	8.849	8.592	0.257	€45,085	€45,327	−€242	EQW Dominant
Weight = Lower CI (−1.24 kg/m^2)	8.818	8.592	0.226	€45,116	€45,327	−€211	EQW Dominant
SBP = Upper CI (−1.54 mmHg)	8.853	8.592	0.261	€44,941	€45,327	−€386	EQW Dominant
SBP = Lower CI (−4.54 mmHg)	8.811	8.592	0.219	€45,097	€45,327	−€230	EQW Dominant
Utility 1	9.185	8.951	0.234	€50,067	€50,919	−€852	EQW Dominant
Utility 2	9.269	8.951	0.317	€50,496	€50,919	−€423	EQW Dominant
Utility 3	9.249	9.011	0.238	€50,496	€50,919	−€423	EQW Dominant
Utility 4	9.198	8.963	0.236	€50,496	€50,919	−€423	EQW Dominant
5 years on treatment	9.202	8.962	0.241	€51,848	€51,875	−€27	EQW Dominant
10 years on treatment	9.220	8.973	0.248	€55,227	€53,767	€1460	€5896
10 years time-horizon	5.483	5.412	0.071	€21,736	€21,635	€101	€3502
20 years time-horizon	8.216	8.044	0.172	€38,731	€39,221	−€490	EQW Dominant
Economics +20%	9.197	8.956	0.241	€55,771	€56,580	−€809	EQW Dominant
Economics −20%	9.197	8.956	0.241	€44,862	€45,084	−€223	EQW Dominant
6% discount rate	7.031	6.875	0.156	€35,411	€35,701	−€290	EQW Dominant
0% discount rate	12.709	12.308	0.401	€77,536	€78,317	−€782	EQW Dominant

EQW, exenatide once weekly; EBID, exenatide twice daily.

Upper CI, upper bound confidence interval from trial data used as model input; Lower CI, lower bound confidence interval from trial data used as model input; Utility 1, injection site reaction disutility added; Utility 2, included injection site reaction disutility and dose flexibility/frequency utility added; Utility 3, BMI utility from Matza et al.61 rather than from Bagust62; Utility 4, utility inputs limited to health-state utilities related to complications (no treatment-related utilities or disutilities applied); Economics, T2D management costs and costs of T2D complications.

Table 9. Base-case results (EQW vs EBID): Breakdown of cost differences.

	EQW (EUR)	EBID (EUR)	Difference (EUR)
Total costs	50,434	50,949	−515
Drug acquisition cost	17,354	16,465	889
Management	5806	5744	62
Cardiovascular complication	21,226	22,380	−1154
Renal complication	983	1183	−200
Ulcer/amputation/neuropathy	2723	2829	−106
Eye complication	2341	2348	−7
Major hypoglycaemia	0	0	0

EQW, exenatide once weekly; EBID, exenatide twice daily.

Sensitivity analyses

For nearly all the scenarios considered, EQW remained the dominant strategy. The exceptions were when a 10-year time horizon was tested and when a 10-year treatment duration was assumed (Table 8). When the lower limits of the confidence intervals were used as model inputs for efficacy, EQW was still expected to be dominant. The only exception was when the 95% upper confidence interval for HbA1c was applied.

Probabilistic sensitivity analyses

It was observed that 72.6% of the cost-effectiveness pairs fell in the south east quadrant of the cost-effectiveness plane, indicating that EQW was expected to be the dominant strategy in the majority of the simulations (Figure 2). The cost-effectiveness acceptability curve (CEAC) shows that EQW is expected to have a 100% probability of being cost-effective for almost all threshold levels considered (Figure 3).

Figure 2. Cost-effectiveness plane (EQW-EBID). EQW, exenatide once weekly; EBID, exenatide twice daily; QALY, Quality adjusted life year.

Figure 3. Cost-effectiveness acceptability curve (EQW-EBID). EQW, exenatide once weekly; EBID, exenatide twice daily.

EQW vs insulin glargine: base-case analysis

Treatment with EQW was projected to increase quality-adjusted life expectancy vs treatment with insulin glargine. The cumulative incidence of diabetes-related complications was also lower in the EQW arm. However, EQW was the more costly strategy, resulting in an estimated incremental cost per quality-adjusted life year of €12,084 (Table 10).

Table 10. Base-case results: EQW vs IG.

	Quality-adjusted life years (QALYs)			Costs (EUR)			ICER (EUR/QALY gained)
Sensitivity analysis	EQW	Glargine	Difference	EQW	Glargine	Difference
Basecase (BMI ≥ 30 kg/m^2)	8.712	8.534	0.178	€45,520	€43,366	€2154	€12,084
Probabilistic SA	8.360	8.177	0.183	€53,449	€51,191	€2258	€12,313
A1c = Lower CI (−1.57%)	8.751	8.527	0.224	€45,451	€43,292	€2159	€9628
A1c = Upper CI (−1.37%)	8.675	8.527	0.149	€45,613	€43,292	€2322	€15,631
Weight = Lower CI (−1.07 kg/m^2)	8.739	8.527	0.212	€45,618	€43,292	€2327	€10,978
Weight = Upper CI (−0.80)	8.718	8.527	0.191	€45,607	€43,292	€2315	€12,131
SBP = Lower CI (−5.17 mmHg)	8.748	8.527	0.221	€45,435	€43,292	€2144	€9679
SBP = Upper CI (−0.89 mmHg)	8.695	8.527	0.168	€45,709	€43,292	€2417	€14,357
Utility 1	8.763	8.526	0.237	€45,593	€43,292	€2301	€9702
Utility 2	8.781	8.526	0.255	€45,593	€43,292	€2301	€9033
Utility 3	8.784	8.526	0.258	€45,593	€43,292	€2301	€8925
Utility 4	8.731	8.527	0.204	€45,593	€43,292	€2301	€11,282
5 years on treatment	8.736	8.527	0.209	€46,909	€43,292	€3617	€17,315
10 years on treatment	8.755	8.518	0.237	€49,659	€42,568	€7091	€29,935
10 years time-horizon	5.393	5.322	0.072	€21,089	€18,693	€2396	€33,450
20 years time-horizon	7.912	7.752	0.160	€36,353	€34,291	€2062	€12,876
Economics +20%	8.712	8.534	0.178	€50,140	€48,117	€2022	€11,346
Economics −20%	8.712	8.534	0.178	€40,680	€38,396	€2284	€12,816
6% discount rate	6.736	6.615	0.122	€32,573	€30,466	€2107	€17,294
0% discount rate	11.864	11.583	0.281	€68,412	€66,201	€2211	€7867

EQW, exenatide once weekly; IG, insulin glargine.

Upper CI, upper bound confidence interval from trial data used as model input; Lower CI, lower bound confidence interval from trial data used as model input; Utility 1, injection site reaction disutility added; Utility 2, included injection site reaction disutility and dose flexibility/frequency utility added; Utility 3, BMI utility from Matza et al.61 rather than from Bagust62; Utility 4, utility inputs limited to health-state utilities related to complications (no treatment-related utilities or disutilities applied); Economics, T2D management costs and costs of T2D complications.

Table 11. Base-case results (EQW vs IG): Breakdown of cost differences.

	EQW (EUR)	IG (EUR)	Difference (EUR)
Total costs	45,520	43,366	2154
Drug acquisition cost	16,432	13,653	2779
Management	5441	5410	31
Cardiovascular complication	19,832	20,399	−567
Renal complication	439	465	−26
Ulcer/amputation/neuropathy	2476	2529	−53
Eye complication	901	910	−9
Major hypoglycaemia	0	0	0

EQW, exenatide once weekly; IG, insulin glargine.

Sensitivity analyses

EQW remained more effective and more costly across all the scenarios tested, but resulting ICERs were within acceptable bounds of cost-effectiveness. The results were most sensitive to variations in effectiveness values, the time on initial treatment and the time-horizon of the analysis, with estimated ICERs ranging from €7867–€33,450 per QALY gained. The incremental differences in effectiveness were kept constant when different times on treatment were tested. However, in these scenarios higher ICERs were obtained given the expected increase in the incremental costs. The latter were mainly driven by the higher costs of treatment in the EQW arm. The variation in management costs had little impact on the cost-effectiveness results.

Probabilistic sensitivity analysis

It was observed that nearly 100% of the simulations fell in the north-east quadrant of the cost-effectiveness plane, indicating that EQW was projected to be the more effective but at the same a time more costly treatment than insulin glargine (Figure 4). The cost-effectiveness acceptability curve (CEAC) shows that EQW was expected to have a 99.4% probability of being cost-effective at a willingness to pay threshold of €20,000 per QALY gained (Figure 5).

Figure 4. Cost-effectiveness plane (EQW-IG). EQW, exenatide once weekly; IG, insulin glargine; QALY, Quality-adjusted life year.

Figure 5. Cost-effectiveness acceptability curve (EQW-IG). EQW, exenatide once weekly; IG, insulin glargine.

Discussion

A comprehensive clinical trial programme has been undertaken to demonstrate the relative effectiveness of EQW against available comparators30,3,64–66. The current analysis used a sub-set of the data from the combined DURATION 1 and 5 for EQW vs EBID and DURATION 3 for EQW vs IG comprising those patients with a baseline BMI ≥ 30 kg/m². In all three trials there was a significant reduction in HbA1c from baseline for EQW. In addition to this, for the DURATION 1 and 5 trials, patients in both treatment arms experienced a reduction in body weight and similar rates of hypoglycaemia were seen.

When compared to EBID, treatment with EQW was found to be the dominant strategy, i.e., more effective and less costly. Patients in the EQW arm were associated with an improvement in life-expectancy and quality-of-life at a lower cost. EQW was associated with a lower cumulative incidence of almost all simulated complications. Stroke events were the exception, with higher cumulative incidence predicted in the EQW arm. This can perhaps be explained by the survival paradox whereby patients in the EQW group had a longer life expectancy and were, therefore, exposed to the risk of stroke for a longer period of time—stroke risk is driven largely by age and not by markers of glycaemic control67.

The results were robust across the majority of changes in parameters tested in the sensitivity analyses, with EQW being the dominant strategy in almost all the scenarios tested. The only exceptions were when an analysis time horizon of 10 years was tested, when 10 years on treatment was assumed (the base case assumption being limited to 3 years), and when the upper limit of the HbA1c effect was used to define efficacy. The impact of longer duration of treatment was higher cost but, even with this, the generated ICER was low (€5896 per QALY gained), suggesting that even at the outside of plausible treatment limits, EQW would still be considered cost-effective. The time horizon result is expected, however, given the chronic nature of the disease, outcomes are better captured in longer time horizons, ideally life-time, so shorter horizons are not applicable. The result of the HbA1c analysis suggests that, even at the outside of the plausible limit of treatment effect, EQW would remain cost-effective with an ICER of €7752 per QALY, although no longer dominant.

In the base case analyses performed against IG, the model predicted that EQW resulted in better outcomes and higher costs than IG, with an estimated ICER of €12,084 per QALY gained. Higher pharmacy costs were partially offset by lower complications costs. This can, in part, be explained by the fact that the model predicted that IG-treated patients would have complications on average 3 months before those treated with EQW.

The model was robust across the majority of assumptions, with the cost per QALY gained being between €7867–€29,935 for the vast majority of simulations, under the assumed €30,000 per QALY cost-effectiveness threshold. The exception to this was the 10-year time horizon simulation.

There are a limited number of studies conducted in the Spanish setting with which to compare the outcome of these analyses. While not directly comparable, a previous analysis conducted by Goodall et al.68 to estimate the cost-effectiveness of EBID vs IG resulted in a similar magnitude of cost-effectiveness ratio. In this analysis, EBID was projected to have an incremental cost-effectiveness ratio of €15,068 per QALY gained vs IG69 compared against our current EQW estimate of €12,084 per QALY gained relative to IG.

PSA results suggest that, despite uncertainty, EQW is expected to be a cost-effective option, with nearly 100% probability of being cost-effective assuming a willingness-to-pay threshold of €30,000 per QALY gained70.

Limitations

As with any cost-effectiveness model, in the absence of life-time data, several assumptions had to be made in order to extrapolate the available short-term clinical trial figures. For example, within the model, short-term intermediate end-points (HbA1c, systolic blood pressure, lipids, and BMI) were used to project long-term outcomes such as death, cardiovascular, renal, and ocular complications. Given the chronic nature of T2DM and the pressing need to develop better interventions to treat the disease, it is not practical to fully evaluate these outcomes by the time of regulatory approval. Computer simulation modelling remains the best option currently available to estimate the clinical and economic consequences of therapeutic interventions in the medium- to long-term. No model can claim to be perfectly accurate, but the IMS CORE Diabetes Model is one of the few models currently available with published validations that demonstrate the reliability of outcomes26,27.

There remain a number of limitations regarding both specific parameters included in these analyses and the general application of DURATION trial data to a country-specific analysis. For the sake of consistency, some non-significant changes in treatment effect parameters were included in the modelling analyses. However, extensive sensitivity analyses were undertaken to identify the role of each treatment effect parameter. A further potential limitation is that, at this stage in the clinical development process, the optimum treatment duration of EQW is unclear. Additional research will best inform this issue, but at the moment a pragmatic evidence-based approach to treatment duration was assumed and the impact was fully explored in sensitivity analyses. The results should, however, be viewed in context of the broader data limitations, in that data from a series of global heterogeneous trial populations were applied in an analysis pertaining to a Spanish population. Full disclosure of the characteristics of patients included in the trials does, however, allow readers to assess the generalizability of the analysis findings to their own patient populations.

Diabetes is a chronic and progressive disease and many T2DM patients will eventually need daily injections of insulin71. A common starting regimen is with long acting insulin such as insulin glargine, but some patients will eventually require short-acting insulin injection before meals72. This element was not considered in the current set of simulations, since data were not available to inform an accurate estimation of long-term expected use of short-acting insulin. Longer-term clinical data for EQW regarding efficacy and safety and duration of treatment effect would improve the current modelling analysis.

Conclusions

This analysis indicated that, in a Spanish setting, EQW would be considered a dominant treatment option vs EBID and a cost-effective treatment option vs insulin glargine in T2D patients with a BMI ≥ 30 kg/m² assuming a threshold of €30,000 per QALY gained.

Transparency

Declaration of funding

This study was funded by Eli Lilly and company Limited.

Declaration of financial/other relationships

TF and GC serve as paid consultants for IMS Health. KN is an employee of and has received sponsorship from Eli Lilly Finland, and owns stock in the company. TD and MA are employees of and have received sponsorship from Eli Lilly Spain, and own stock in the company. JC has no conflicts of interest.