Cost-effectiveness of insulin detemir compared with neutral protamine Hagedorn insulin in patients with type 1 diabetes using a basal-bolus regimen in five European countries

Abstract

Objectives: The aim of this analysis was to evaluate the long-term clinical and economic outcomes associated with insulin detemir and neutral protamine Hagedorn (NPH) insulin in combination with mealtime insulin aspart in patients with type 1 diabetes in Belgian, French, German, Italian and Spanish settings.

Methods: The published and validated IMS CORE Diabetes Model was used to make long-term projections of life expectancy, quality-adjusted life expectancy and direct medical costs. The analysis was based on patient characteristics and treatment effects from a 2-year randomised controlled trial. Events were projected for a time horizon of 50 years. Potential uncertainty using a modelling approach was addressed.

Results: Basal-bolus therapy with insulin detemir was projected to improve quality-adjusted life expectancy by 0.45 years versus NPH in the German setting, with similar improvements in the other countries. Insulin detemir was associated with cost savings in Belgium, Germany and Spain. In France and Italy, lifetime costs were slightly higher in the detemir arm, leading to incremental cost-effectiveness ratios of €519 per QALY gained and €3,256 per QALY gained, respectively.

Conclusions: Compared to NPH, insulin detemir is likely to be a dominant treatment strategy in Belgium, Germany and Spain and highly cost-effective in France and Italy in patients with type 1 diabetes.

Keywords::

• cost-effectiveness
• detemir
• Europe
• insulin
• modelling
• NPH
• type 1 diabetes

Introduction

Type 1 diabetes mellitus (T1DM) is associated with a wide range of complications such as cardiovascular disease, nephropathy, retinopathy, neuropathy and foot ulceration. These micro- and macrovascular complications are a key driver of costs. In a review, Liebl reported that diabetes is responsible for over 14% of the total direct medical cost burden in Germany, costing in excess of €60 million annually¹. A cost-of illness study in Spain by Oliva et al found that 6.3–7.4% of the total National Health System expenditures went towards to diabetes care, accounting for €2.4–2.7 billion². Even if patients with T1DM only make an estimated 10% of the patients diagnosed with diabetes, the financial burden of diabetes is still substantial.

Various studies like the Diabetes Control and Complications Trial (DCCT)³ have shown that intensive glycaemic control can delay or reduce the risk of long-term complications in T1DM. However, intensive therapy with conventional insulins is associated with a greater risk of major hypoglycaemia and increased weight gain. Insulin detemir is a long-acting basal soluble insulin with a protracted action profile designed to avoid the variation in absorption and release typically seen with standard long-acting insulins like neutral protamine Hagedorn (NPH). In clinical trials, insulin detemir has demonstrated comparable HbA_1c reduction, lower within-subject fasting blood glucose variability, less hypoglycaemia and less weight gain compared to basal-bolus regimens using NPH insulin^4–10. Many of these trials were, however, of too limited duration to evaluate long-term effects. In a recently published 2-year, multi-national, open-label, randomised, controlled trial (RCT), Bartley et al studied the long-term efficacy and safety of insulin detemir versus NPH in a total of 497 patients with T1DM. Basal insulin, either detemir or NPH, was individually titrated aiming for pre-breakfast and pre-dinner targets of ≤ 6.0 mmol/l. Insulin aspart was injected immediately before each meal, and was titrated to achieve a postprandial plasma glucose level of ≤ 9.0 mmol/l. After 24 months, insulin detemir was associated with significant improvements in glycaemic control (HbA_1c 7.36 vs. 7.58%, mean difference -0.22%, p = 0.022) and major hypoglycaemic events (69% risk reduction, p = 0.001) versus NPH. Patients treated with detemir gained less weight (1.7 vs. 2.7 kg, p = 0.024)¹¹.

We designed and performed a computer simulation modelling analysis to estimate the long-term clinical and economic outcomes associated with insulin detemir and NPH in combination with mealtime insulin aspart in patients with T1DM, based on these short-time findings. The cost-effectiveness of insulin detemir versus insulin NPH was evaluated in Belgium, France, Germany, Italy and Spain – five large European economies with high expenditures on health care.

Methods

Model

The IMS CORE Diabetes Model (CDM) is an internet-based computer model developed to determine the long-term health outcomes and economic consequences of interventions in diabetes patients¹². Disease progression is based on a series of inter-dependent semi-Markov sub-models that simulate progression of disease-related complications. Each sub-model uses time, state and diabetes type-dependent probabilities derived from published sources. Each sub-model utilises tracker variables to overcome the memory-less properties of standard Markov models and allows interconnectivity and interaction between individual complication sub-models. Clinical and economic outcomes (means and standard deviations) are calculated within the model using a non-parametric bootstrapping approach. This process simulates the lifetime progression of diabetes in cohorts of 1,000 hypothetical patients and repeats the process 1,000 times. This produces 1,000 mean values of clinical effectiveness and lifetime costs which are then used to generate a scatter plot diagram and acceptability curve to express the likelihood of a treatment being cost-effective versus a comparator. The reliability of simulated outcomes has been tested, with results validated against those reported by clinical trials and epidemiological studies¹³.

The model was used to project life expectancy, quality-adjusted life expectancy, complication rates, time to onset of complications and direct medical costs. For the estimation of quality-adjusted life expectancy, utility scores were derived wherever possible from diabetes populations and have been published previously^14–17.

Simulation cohorts

Country-specific simulation cohorts were generated based on patients' characteristics from the Bartley trial¹¹. The mean patient age at baseline was 35 years, with 54.7% male and an average duration of diabetes of 13 years. Mean HbA_1c at baseline was 8.3% and body mass index (BMI) was 24.7 kg/m². Full cohort characteristics including standard deviations used for all country simulations are given in Table 1. The prevalence of pre-existing complications in each country cohort was taken from recently published country-specific data^5,18–31. Patient management practices in terms of the proportion of patients regularly screened for retinopathy and nephropathy were derived from the Bartley trial. Data about the use of concomitant cardiovascular medications such as aspirin, statins and ACE-inhibitors for primary and secondary prevention were taken form the EUROASPIRE II Euro Heart Survey Programme³².

Costs

Costs were accounted from a third party payer perspective. Direct medical costs of complications were derived from published country-specific sources^33–65. Acquisition costs (public prices) of insulin detemir (Levemir), NPH insulin (e.g., Insulatard) and insulin aspart (Novorapid), as well as needles and devices for self-monitoring of blood glucose were obtained from public pharmacies (Belgium, Germany, Spain) or national health authorities and health-care payers (National Health Service, Italy; RIZIV, Belgium). The annual costs of insulin were calculated based on the reported end-of-trial doses of insulin detemir, NPH and aspart. Costs were inflated as required, using country-specific indices, to 2006 values.

Discounting and time horizon

Future costs and clinical benefits were discounted at country-specific rates in line with published guidances^66–70 (Belgium 3% costs, 1.5% benefits; France 3% both; Germany 5% both; Italy 3% both; Spain 6% both). Sensitivity analysis was performed using a range of discount rates between 0% and 10% for costs and clinical outcomes, based on the recommendations by the same authors. A time horizon of 50 years was used in the base-case analysis to capture all relevant long-term complications, their associated costs and impact on life expectancy and quality-adjusted life expectancy. Sensitivity analysis was performed using time horizons of 5, 10, 15, 20, 30 and 40 years.

Sensitivity analyses

One-way sensitivity analyses were performed around key inputs in the base-case analysis. Parameters were varied over a range of possible scenarios to assess their impact on health economic outcomes. Time horizon was varied between 0 and 50 years (here we report values at 5, 10, 15, 20, 30 and 40 years). Discount rates for costs and health outcomes were applied according to country-specific recommendations^66–70 (Belgium 0% and 6%; France 0% and 5%; Germany 0%, 3% and 10%; Italy 0% and 8% and Spain 0% and 10%). The influence of changes in HbA_1c levels on long-term clinical and economic benefits was assessed by abolishing and doubling the HbA_1c benefit of insulin detemir over NPH. The influence of hypoglycaemic events was also evaluated: in one sensitivity analysis the same major hypoglycaemic rates were applied to both treatment arms, in another the same minor hypo-glycaemic rates. To investigate the impact of variation in BMI, simulations were run using the same effect on BMI for both the insulin detemir and the NPH treatment. To assess the view of the societal perspective, sensitivity analyses capturing indirect costs using a human capital approach (lost productivity) were performed.

In addition to the one-way sensitivity analyses, second order simulations were performed by entering associated standard deviations and selecting to run the simulation with sampling. In this case the parameter was then selected at random from the associated distribution thus accounting for uncertainty surrounding the input variable of interest.

Statistical methodology

For each analysis in the base case and sensitivity analysis, a simulated cohort of 1,000 patients was run through the model 1,000 times using a non-parametric bootstrapping approach. From which, mean values and standard deviations were generated⁷¹. One thousand mean values (each of 1,000 patients) of incremental costs and incremental effectiveness in terms of quality-adjusted life expectancy were plotted as scatter plots on a cost-effectiveness plane. For interventions that were not dominant (cost saving with benefits in terms of life expectancy or quality-adjusted life expectancy), it was planned to generate an acceptability curve by calculating the proportion of points below a range of willingness-to-pay thresholds.

Results

Main findings

Clinical outcomes

Basal-bolus therapy with insulin detemir was projected to improve both mean life expectancy (LE) and quality-adjusted life expectancy in all five countries analysed (Table 2). The magnitude of improvement in discounted life expectancy varied in the range of 0.07 to 0.15 years across the countries. In Germany, insulin detemir was associated with a benefit in LE of 0.09 years (12.80 vs. 12.71 years). In France and Italy the benefits were slightly higher, with 0.13 years (15.63 vs. 15.50 years) and 0.15 years (16.21 vs. 16.06 years), respectively. The smallest benefit was observed in Spain with 0.07 years (12.04 vs. 11.97 years). In Belgium, projected LE was extended by 0.14 years (14.36 vs. 14.22 years) with detemir treatment.

Improvements in discounted quality-adjusted life expectancy in the detemir group ranged from 0.40 to 0.58 years. In Germany, insulin detemir treatment was projected to improve quality-adjusted life expectancy with 0.45 years (7.04 vs. 6.59 QALYs). The values were higher in France and Italy with 0.55 (8.47 vs. 7.92 QALYs) and 0.58 (8.98 vs. 8.39 QALYs) years, respectively. In Spain and Belgium the projected improvements with detemir were 0.40 (6.59 vs. 6.19 QALYs) and 0.52 years (7.85 vs. 7.33 QALYs), respectively.

The time to onset of any diabetes-related complication in the German setting was delayed by 0.08 years in the detemir arm (1.18 vs. 1.10 years). Similar effects were observed in the other four countries. Table 3 gives an overview on time free of the most prominent complications for both treatments in all countries studied.

Cumulative incidence of diabetic eye and renal disease, neuropathy and amputations were generally decreased for detemir-based therapy, with greatest benefits observed in renal disease. The cumulative incidences of heart failure, angina and stroke were slightly raised in the detemir-based treatment arm, which is likely attributable to longer survival of patients and therefore higher exposure to macrovascular diseases. Figure 1 shows the cumulative incidences of the most prominent complications in the German setting.

Lifetime costs and cost-effectiveness

Insulin detemir was associated with savings in direct costs in Germany (€74,880 vs. 75,734), Spain (€44,085 vs. 44,661) and Belgium (€122,737 vs. 134,679). In France and Italy, lifetime direct costs were higher in the detemir arm (€63,605 vs. 63,321 and €92,036 vs. 90,139, respectively). Table 4 shows the total direct costs over patient lifetimes. The breakdown of direct medical costs demonstrated that the key drivers were higher treatment costs in the detemir arm, and higher major hypoglycaemic event rates in the NPH insulin arm. Overall costs of complications in all countries studied were higher in the NPH treatment group, except for cardiovascular disease (CVD), were they were marginally lower. The higher direct medical costs in France and Italy led to incremental cost-effectiveness ratios of €519 per QALY gained and 3,256 per QALY gained, respectively. Capturing lost productivity costs using a human capital approach suggested that detemir treatment was likely to be cost saving in France and Italy as well. Due to the lower rate of acute events like hypoglycaemia or ketoacidosis and fewer diabetes-related complications, lifetime indirect costs with detemir treatment were €7,935 lower than with NPH in France, and €10,318 lower in Italy, respectively. Combining both direct and indirect costs, insulin detemir resulted in being a dominant treatment versus NPH in both France and Italy.

Figure 2 shows an incremental cost-effectiveness scatterplot for quality-adjusted life expectancy generated from the mean values of 1,000 × 1,000 nonparametric bootstrap simulations in the German setting. Most of the points lying in the bottom right quadrant of the plane show the dominant nature (increased effectiveness at lower overall costs) of detemir.

At a willingness-to-pay threshold of €50,000, the likelihood of detemir treatment being cost-effective has shown to be 100% in all five countries (Figure 3).

Sensitivity analyses

One-way sensitivity analyses in the German setting revealed that results were most sensitive to differences in major hypoglycaemic event rates (Table 5). Abolishing the difference between the treatments while setting the effect of detemir to the same level as with NPH resulted in a decrease in quality-adjusted life expectancy with detemir treatment of 0.09 QALYs. Costs in the detemir group simultaneously increased by €3,472, resulting in detemir still being a cost-effective but not dominant treatment versus NPH. Assuming no difference in minor hypoglycaemic events between the two treatments resulted in a greater decrease in quality-adjusted life expectancy in the detemir group, however the overall outcome remained unchanged.

Applying the same change in HbA_1c for detemir as applied for NPH resulted in a slight reduction of quality-adjusted life expectancy in the detemir arm (a decrease of 0.09 QALYs from the base case). With an increase in lifetime direct costs of €922, the ICER of detemir treatment versus NPH became €210 per QALY. When the HbA_1c benefit associated with insulin detemir treatment was doubled, quality-adjusted life expectancy increased with costs slightly decreasing while detemir treatment remained dominant.

Variations in time horizon had a noticeable impact on the mean quality-adjusted life expectancy and total direct medical costs. Increases in the simulated time horizon resulted in elevated quality-adjusted life expectancy and direct costs for both treatment arms. Shorter time horizon failed to capture the long-term clinical outcomes and end-stage complications, resulting in small benefits at lower costs.

Discounting had little impact on the relative outcomes and detemir treatment remained dominant through the whole range of discount rates applied.

Sensitivity analyses also revealed that varying the treatment effects in terms of BMI had no impact on quality-adjusted life expectancy and only a marginal raising effect on costs of detemir, so that overall outcomes did not change.

The same patterns were observed in extensive one-way sensitivity analyses performed in the Belgian, French, Italian and Spanish settings (data not shown).

In France, clinical benefits in terms of quality-adjusted life expectancy improved with time horizon, with greater effects in the detemir group. With costs of both treatments being almost equal, the ICER of detemir versus NPH decreased with time horizon with a trend towards cost-neutrality. Doubling the efficacy of detemir with regard to change in HbA_1c resulted in detemir being dominant versus NPH.

In the Italian setting, in all scenarios of the sensitivity analysis, quality-adjusted life expectancy in the detemir treatment group remained higher than in the NPH group. The same applied to lifetime direct costs, resulting in ICERs in the range of €1,929–8,129 per QALY gained for detemir versus NPH treatment.

In Belgium and Spain, detemir treatment was dominant in all sensitivity analyses performed, except when major hypoglycaemic event rates were the same number in the detemir as in the NPH arm. This led to ICERs of €14,797 and €5,942, respectively, per QALY gained for detemir versus NPH treatment.

The results of the second-order simulations did not reveal major differences to the results observed in the base-case setting (Table 6).

Discussion

Long-term projections using the CORE Diabetes Model, based on intervention effects from the Bartley trial, demonstrated that treatment with insulin detemir is associated with improvements in life expectancy and quality-adjusted life expectancy in comparison to NPH insulin in patients with type 1 diabetes in all five countries studied. The higher rate of major hypoglycaemic events in the NPH insulin arm (approximately three-fold higher versus insulin detemir) had a deleterious impact on the quality of life of patients. Variations in life expectancy and quality-adjusted life expectancy between the five countries were mainly due to differences in non-specific mortality (life tables) and different proportions of pre-existing diseases in the baseline cohorts. For instance, the Belgian population with most underlying diseases had the shortest undiscounted life expectancy. Additionally, clinical benefits were discounted at different rates varying between 1.5% and 6% in the five countries.

Interestingly, treatment with insulin detemir reduced the cumulative incidence of most diabetes-related complications and was associated with a delay in time to onset of all complications over patient lifetimes compared to NPH insulin. The cumulative incidences of heart failure, angina and stroke were slightly raised in the detemir-based treatment arm as overall survival was increased, exposing these patients to a longer ongoing risk of these events.

In the base case, insulin detemir was associated with savings in direct medical costs in Germany, Spain and Belgium. In France and Italy, lifetime costs were slightly higher than with NPH, but detemir treatment still remained a very cost-effective treatment. The high direct costs in Belgium were mostly due to a relatively sick baseline population, with high proportions of pre-existing cardiovascular and renal diseases. This population had an early onset of complications (on average already after 0.58 years) and a high cumulative incidence in costly conditions such as end-stage renal disease. Treatment costs of frequently occurring nephropathy were also relatively high compared to other countries. Spanish costs were the lowest of all five countries, due to low costs for treating complications and a relatively high discount rate of 6% in the base case.

Sensitivity analysis showed that results were highly sensitive to hypoglycaemic event rate. Assuming the same number of events with detemir treatment than with NPH, quality-adjusted life expectancy in the German setting decreased by 0.09 QALYs compared to base case in the detemir group. Due to the higher hypoglycaemic event rate lifetime direct costs with detemir increased by €3,472 compared to base case, resulting in an ICER of €7,393 per QALY gained. This analysis shows that by avoiding a higher number of major hypoglycaemic events, the higher acquisition costs of detemir treatment can be compensated. Abolishing the benefit of detemir on HbA_1c decreased QALYs, inflated costs due to an increased number of complications and led to higher ICERs in all countries. Changing the treatment effect on BMI had a small effect on costs only. BMI in the model is not accounted as an independent risk factor and only affects congestive heart failure in women. All other cardiovascular risk formulae rely on systolic blood pressure and serum lipid markers instead^72,73. Several analyses with varying time horizons also showed that insulin detemir is cost-saving compared to NPH already after 5 years. As long-term clinical outcomes and complications are only captured later on, its benefits clearly increase with time horizon.

Using a modelling approach to make long-term projections based on short-term trial data is a potential weakness of this analysis. In the model, we assumed that HbA_1c benefits are maintained and hypoglycaemic event rates remain constant over a long time period, which is both supported by data in the DCCT⁷⁴. Like in all modelling analyses, many assumptions around the progression of the disease and associated costs had to be made, resulting in mean outcome values with large distributions. Additionally, the differences in clinical and economic outcomes were quite small, as we were comparing two potent and highly efficient treatments. Consequently, no claim about statistical significance between the observed results can be made. However, the CDM is one of the only two currently available models with published validations demonstrating the reliability of the outcomes when recreating the observations made in long-term clinical trials.

Conclusions

The findings of this analysis suggest that, compared to NPH, insulin detemir is likely to improve life expectancy, delay the onset of and reduce the cumulative incidence of most diabetes-related complications. Detemir treatment is considered to be dominant in Belgium, Germany and Spain and highly cost-effective in France and Italy in patients with type 1 diabetes.

Figure 1. Cumulative incidences of complications in Germany. SVL, severe vision loss; GPU, gross proteinuria; ESRD, end-stage renal disease; AMP, amputation ulcer; CHF, congestive heart failure event; MI, myocardial infarction event; ST, stroke event.

Figure 2. ICER scatterplot for insulin detemir versus NPH in the German setting. Scatter plot of 1,000 samples of mean incremental costs plotted against mean incremental effectiveness in terms of quality-adjusted life expectancy (QALE). Costs are expressed in euros.

Figure 3. Acceptability curves for all countries based on QALYs (quality-adjusted life years).

Table 1. Baseline demographic characteristics of all patients.

Characteristic	Value	range	SD
Patient demographics
Age at baseline (years)	35	18–75	12
Duration of diabetes (years)	13.0	1.0–50.4	9.8
Proportion male (%)	54.7	—	—
Baseline risk factors
Mean HbA1c (%)	8.3	5.0–11.6	1.2
Mean SBP (mmHg)	121.5	—	—
Mean total cholesterol (mg/dl)	201.1	100.5–564.6	46.4
Mean HDL (mg/dl)	54.1	19.3–131.5	15.5
Mean LDL (mg/dl)	123.7	23.2–255.2	34.8
Mean triglycerides (mg/dl)	106.3	35.4–2692.6	159.4
Mean BMI (kg/m^2)	24.7	15.4–34.7	3.7

SD, standard deviation; HbA_1c, glycosylated haemoglobin; SBP, systolic blood pressure; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; BMI, body mass index.

Values shown are means.

Table 2. Summary of clinical outcomes in five European countries.

Country	Insulin detemir	NPH	Difference
Belgium
LE (years)	14.36 (0.21)	14.22 (0.21)	+0.14 (0.30)
QALE (QALYs)	7.85 (0.11)	7.33 (0.11)	+0.52 (0.16)
France
LE (years)	15.63 (0.21)	15.50 (0.21)	+0.13 (0.30)
QALE (QALYs)	8.47 (0.12)	7.92 (0.11)	+0.55 (0.16)
Germany
LE (years)	12.80 (0.14)	12.71 (0.14)	+0.09 (0.19)
QALE (QALYs)	7.04 (0.08)	6.59 (0.07)	+0.45 0.11)
Italy
LE (years)	16.21 (0.19)	16.06 (0.20)	+0.15 (0.27)
QALE (QALYs)	8.98 (0.11)	8.39 (0.10)	+0.58 (0.15)
Spain
LE (years)	12.04 (0.10)	11.97 (0.11)	+0.07 (0.14)
QALE (QALYs)	6.59 (0.06)	6.19 (0.06)	+0.40 (0.08)

NPH, neutral protamine Hagedorn; LE, life expectancy; QALE, quality-adjusted life expectancy; QALY, quality-adjusted life year.

Values shown are means with standard deviations in parentheses.

Table 3. Time alive and free of complications (years).

Complication	Belgium		France		Germany		Italy		Spain
	IDet	NPH	IDet	NPH	IDet	NPH	IDet	NPH	IDet	NPH
Free of all complications	0.58	0.54	1.15	1.06	1.18	1.10	1.07	0.99	1.50	1.39
Gross proteinuria	11.93	11.43	18.49	17.95	17.20	16.62	17.25	16.63	20.25	19.62
End-stage renal disease	19.62	19.33	23.14	22.85	23.71	23.38	23.83	23.51	25.33	25.00
Severe vision loss	17.91	17.61	19.16	18.83	21.33	20.94	20.57	20.20	19.58	19.29
Amputation	20.07	19.80	21.95	21.67	23.07	22.78	23.19	22.87	24.77	24.46
Congestive heart failure	19.50	19.24	22.10	21.82	22.97	22.66	23.01	22.70	24.62	24.34
Myocardial infarction	19.51	19.22	21.44	21.16	23.03	22.70	23.24	22.92	24.81	24.49
Stroke	20.37	20.11	22.83	22.55	23.76	23.44	23.85	23.55	25.64	25.34

IDet, Insulin detemir; NPH, neutral protamine Hagedorn insulin.

Table 4. Total direct costs over patient lifetimes and ICERs.

Country	Direct costs (€)			ICER
	Detemir	NPH	Difference	Δ direct costs/ Δ QALE
Belgium	122,737 (3,301)	134,679 (3,140)	–11,943 (4,803)	DOMINANT
France	63,605 (1,778)	63,321 (1,899)	284 (2,366)	519
Germany	74,880 (2,248)	75,734 (2,097)	–854 (3,034)	DOMINANT
Italy	92,036 (2,193)	90,139 (2,092)	1,897 (3,036)	3,256
Spain	44,085 (1,026)	44,661 (1,029)	–577 (1,463)	DOMINANT

ICER, incremental cost-effectiveness ratio; QALE (quality-adjusted life expectancy); Δ, difference.

Values shown are means in € with standard deviations in parentheses.

Table 5. Sensitivity analysis comparing insulin detemir versus NPH: Germany.

Assumption	Quality-adjusted life expectancy (QALYs)			Lifetime direct costs per patient (€)			ICER (€ per QALY gained)
	Detemir	NPH	Δ	Detemir	NPH	Δ
Base case	7.04 (0.08)	6.59 (0.07)	+0.45 0.11)	74,880 (2,248)	75,734 (2,097)	–854 (3,034)	DOMINANT
One-way sensitivity analysis
5-year time horizon	2.63 (0.02)	2.50 (0.02)	+0.13 (0.02)	16,448 (770)	16,634 (741)	–187 (1,060)	DOMINANT
10-year time horizon	4.37 (0.03)	4.14 (0.03)	+0.23 (0.05)	31,263 (1,298)	31,585 (1,271)	–322 (1,838)	DOMINANT
15-year time horizon	5.50 (0.05)	5.19 (0.05)	+0.31 (0.07)	44,088 (1,733)	44,727 (1,626)	–639 (2,355)	DOMINANT
20-year time horizon	6.21 (0.06)	5.84 (0.06)	+0.37 (0.08)	54,614 (2,025)	55,186 (1,780)	–572 (2,763)	DOMINANT
30-year time horizon	6.86 (0.07)	6.44 (0.07)	+0.42 (0.10)	68,437 (2,078)	69,336 (2,117)	–898 (2,905)	DOMINANT
40-year time horizon	7.02 (0.08)	6.58 (0.07)	+0.44 (0.10)	74,067 (2,205)	74,677 (2,299)	–610 (3,192)	DOMINANT
Same HbA1c change applied for detemir as for NPH (–0.72%)	6.95 (0.08)	6.59 (0.07)	+0.36 (0.11)	75,810 (2,081)	75,734 (2,097)	+76 (3,006)	210
Double HbA1c change applied for detemir (–1.16%)	7.13 (0.08)	6.59 (0.07)	+0.53 (0.11)	74,525 (2,279)	75,734 (2,097)	–1,209 (3,106)	DOMINANT
0% discount rate	13.53 (0.22)	12.57 (0.21)	+0.96 (0.29)	176,372 (5,689)	177,820 (5,612)	–1,448 (7,644)	DOMINANT
3% discount rate	8.91 (0.12)	8.32 (0.11)	+0.59 (0.15)	101,381 (3,058)	102,438 (2,894)	–1,057 (4,100)	DOMINANT
10% discount rate	4.40 (0.04)	4.13 (0.04)	+0.27 (0.05)	41,807 (1,306)	42,336 (1,225)	–529 (1,812)	DOMINANT
No difference in BMI (both +0.95)	7.04 (0.08)	6.59 (0.07)	+0.45 (0.11)	74,981 (2,284)	75,734 (2,097)	–753 (3,039)	DOMINANT
No difference in major hypoglycaemia (both 80/100 pt.y.)	6.95 (0.08)	6.59 (0.07)	+0.35 (0.10)	78,352 (2,105)	75,734 (2,097)	+2,618 (2,747)	7,393
No difference in minor hypoglycaemia (both 2820/100 pt.y.)	6.77 (0.08)	6.59 (0.07)	+0.18 (0.10)	74,880 (2,248)	75,734 (2,097)	–854 (3,034)	DOMINANT
Probabilistic sensitivity analysis
	6.92 (0.76)	6.45 (0.76)	+0.47 (1.10)	74,250 (10,970)	74,778 (11,189)	–528 (15,851)	DOMINANT
Societal perspective
	Lifetime direct costs per patient (EUR)			Lifetime indirect costs per patient (€)			ICER (€ per QALY gained)
	Detemir	NPH	Δ	Detemir	NPH	Δ
	74,880 (2,248)	75,734 (2,097)	–854 (3,034)	125,788 (4,207)	136,333 (4,013)	–10,545 (5,602)	DOMINANT

QALYs, quality-adjusted life years; ICER, incremental cost-effectiveness ratio; pt.y., patient years.

Table 6. Probabilistic sensitivity analysis for all countries (2nd order with sampling).

Assumption	Quality-adjusted life expectancy (QALYs)			Lifetime direct costs per patient (EUR)			ICER (€ per QALY gained)
	Detemir	NPH	Δ	Detemir	NPH	Δ
Belgium	7.64 (1,22)	7.15 (1,09)	+0.49 (1.66)	118,236 (21,529)	129,055 (21,575)	–10,819 (30,593)	–22,037
France	8.26 (1.19)	7.73 (1.07)	+0.52 (1.58)	62,934 (8,979)	63,338 (8,919)	–404 (12,716)	–774
Germany	6.92 (0.76)	6.45 (0.76)	+0.47 (1.10)	74,250 (10,970)	74,778 (11,189)	–528 (15,851)	–1,133
Italy	8.81 (1.21)	8.24 (1.18)	+0.57 (1.66)	90,768 (12,380)	88,429 (12,157)	2,339 (16,844)	4,087
Spain	6.47 (0.65)	6.08 (0.59)	0.38 (0.88)	43,893 (4,605)	44,681 (4,636)	–788 (6,445)	–2,060

ICER, incremental cost-effectiveness ratio.

Acknowledgements

Declaration of interest: This study was supported by an unrestricted grant from Novo Nordisk, Denmark. Manuela Gschwend and William Valentine are current employees of IMS Health, which has received consulting fees from Novo Nordisk for the use of the IMS CORE Diabetes Model, the performance of the analyses and the writing of this manuscript. Mark Aagren is a current employee of Novo Nordisk.

Manuela Gschwend was responsible for the overall design, model inputs, performance of the analysis, collection and interpretation of data and writing of this manuscript. William Valentine gave scientific advice and input into the design of the analysis and critically reviewed the manuscript. Mark Aagren was involved in choosing model input parameters and has given final approval of the version to be published. All authors read and approved the final manuscript.

Notes

* Levemir, Insulatard and Novorapid are all registered trademarks of Novo Nordisk A/S