The cost-effectiveness and budget impact of stepwise addition of bolus insulin in the treatment of type 2 diabetes: evaluation of the FullSTEP trial

Abstract

Background and aims:

Intensification of basal insulin-only therapy in type 2 diabetes is often achieved through addition of bolus insulin 3-times daily. The FullSTEP trial demonstrated that stepwise addition (SWA) of bolus insulin aspart was non-inferior to full basal-bolus (FBB) therapy and reduced the rate of hypoglycemia. Here the cost-effectiveness and budget impact of SWA is evaluated.

Methods:

Cost-effectiveness and budget impact models were developed to assess the cost and quality-of-life (QoL) implications of intensification using SWA compared with FBB in the US setting. At assessment, SWA patients added one bolus dose to their current regimen if the HbA1c target was not met. SWA patients reaching three bolus doses used FBB event rates. Outcomes were evaluated at trial end and projected annually up to 5 years. Models captured hypoglycemic events, the proportion meeting HbA1c target, and self-measured blood glucose. Event rates and QoL utilities were taken from trial data and published literature. Costs were evaluated from a healthcare-payer perspective, reported in 2013 USD, and discounted (like clinical outcomes) at 3.5% annually. This analysis applies to patients with HbA1c 7.0–9.0% and body mass index <40 kg/m².

Results:

SWA was associated with improved QoL and reduced costs compared with FBB. Improvement in QoL and cost reduction were driven by lower rates of hypoglycemia. Sensitivity analyses showed that outcomes were most influenced by the cost of bolus insulin and QoL impact of symptomatic hypoglycemia. Budget impact analysis estimated that, by moving from FBB to SWA, a health plan with 77,000 patients with type 2 diabetes, of whom 7.8% annually intensified to basal-bolus therapy, would save USD 1304 per intensifying patient over the trial period.

Conclusions:

SWA of bolus insulin should be considered a beneficial and cost-saving alternative to FBB therapy for the intensification of treatment in type 2 diabetes.

Keywords::

• Type 2 diabetes
• Insulin
• Cost-effectiveness
• Budget impact

Introduction

Type 2 diabetes is a major and growing burden on healthcare systems globally. Recent data from the Centers for Disease Control and Prevention indicate that 29.1 million people in the US (9.3% of the population) have diabetes, with a further 86 million adults estimated to have pre-diabetes1. In high income countries, up to 95% of diabetes cases are attributed to type 2 diabetes and it is the main contributor to the substantial resource and economic burden of diabetes due to its greater prevalence2. For the US setting, it was estimated that diagnosed diabetes accounted for USD 176 billion in direct medical costs in 2012, ∼20% of US healthcare spending3. Although treatment of diabetes requires constant therapy, the largest cost contributor is hospitalization (43–55%)3,4, whereas anti-diabetic agents and diabetes supplies make up 18% of diabetes spending3. The aim of diabetes treatment is, thus, to normalize blood glucose and prevent complications and hospitalizations. The UK Prospective Diabetes Study demonstrated that a 1% reduction in glycated hemoglobin (HbA1c) resulted in a 21% reduction in type-2-diabetes-related complications (95% confidence interval [CI] = 17–24%, p < 0.0001)5.

The European Association for the Study of Diabetes (EASD)/American Diabetes Association (ADA) guidelines recommend that the percentage of patients achieving an HbA1c <7.0% acts as a care quality indicator6. Patients with type 2 diabetes often initiate therapy using diet and lifestyle modification, but progressive treatment intensification is common due to the continual deterioration of β-cell function that results in failure to meet glycemic targets7. A 1-year, open-label, uncontrolled, single-arm interventional study conducted by Viana et al.8 found that following EASD/ADA 2009 guidelines the use of oral anti-diabetic drugs (OADs) increased significantly during the trial period. Likewise, 13% of patients initiated insulin therapy and the mean insulin dosage increased significantly in patients who had been on insulin at baseline8. In spite of treatment intensification to maintain glycemic control, the percentage of patients achieving an HbA1c <7% decreased from 53.3% at base-line to 48.9% at study end8.

The EASD/ADA guideline suggests treatment intensification in subjects with type 2 diabetes who do not achieve HbA1c target when treated with basal insulin (±OADs)6. This is often achieved through addition of bolus insulin taken 3-times daily or less. Although leading to lower HbA1c when a basal-bolus regimen is performed, many patients still fail meet targets7,9. In a 1-year, multi-national, open-label, treat-to-target study in the US, patients with type 2 diabetes who had received OADs or insulin (±OADs) were randomized to a basal-bolus regimen of bolus insulin aspart with either insulin detemir or insulin glargine as the basal insulin9. The study demonstrated non-inferiority of detemir compared with glargine, but almost 66% of patients failed to achieve HbA1c <7%9. It has been suggested that insulin injections and more frequent hypoglycemic events are barriers to good glycemic management under a basal-bolus approach10–12.

There have been many clinical studies and health economic analyses assessing the relative value of different insulin products in the treatment of type 2 diabetes13–15. It is also reasonable to ask whether the method of insulin delivery, e.g., the injection schedule, can have a significant impact on patient outcomes16,17. The 32-week, randomized, open-label, two-arm, parallel-group, multi-national, treat-to-target, non-inferiority, phase four FullSTEP study (n = 401) demonstrated that stepwise addition (SWA) of bolus insulin (insulin aspart) was non-inferior to a full basal-bolus (FBB, 3 boluses insulin doses per day) approach in terms of achieving HbA1c targets, and resulted in higher patient satisfaction with treatment18. In the trial, patients (mean age = 59.8 years; mean HbA1c = 7.9%) were either transitioned directly to a FBB regimen (insulin before every main meal) or received a single bolus insulin dose with their largest meal followed by SWA of bolus doses at 11 weeks and 22 weeks if the HbA1c target (<7%) was not achieved18. Compared with FBB, SWA resulted in a non-inferior reduction in HbA1c (−0.98% vs −1.12%) and significantly reduced rate of hypoglycemia (p < 0.001), and was associated with greater patient satisfaction with treatment18. In this paper, the cost-effectiveness and budget impact of SWA of bolus insulin aspart is evaluated. The cost-effectiveness analysis captures both cost and quality-of-life data associated with SWA and FBB therapy, and considers the implication of these factors over an extended time horizon. A shorter term analysis is undertaken for the budget impact analysis. In this case, the focus is on the cost to the health plan of providing care as either SWA or FBB.

Methods

The two models described in this manuscript, although related, were independently coded in Microsoft Excel using related but not identical implementations. Two implementations allowed for each question to be evaluated separately and consistency of data to be assessed, as well as indicating potential impact of model structure on outcomes. The methods described in the following sections apply, unless otherwise stated, to both models.

Model design

The cost-effectiveness and budget impact models are both based on a Markov model for SWA, with an individual health state assigned to one, two, and three bolus doses. For FBB, only one health state exists, three bolus doses. Within each cycle, each health state is associated with a probability of an end-point (e.g., symptomatic hypoglycemia) occurring. In both models, and in line with the trial design, patients using SWA initiated bolus insulin with a single daily dose and the probability of transitioning between states is time-dependent. HbA1c values were assessed at weeks 10, 21, and 32 during the trial, and those patients not meeting the target (<7%) added an extra bolus dose (up to a maximum of three bolus doses) to their regimen. Addition of a bolus dose was assumed to take place in the week following assessment, e.g., weeks 11, 22, and 33. Beyond the trial period, HbA1c in the model was evaluated at 42 weeks, 52 weeks, and once yearly thereafter. The costs and outcomes of SWA were compared with those of the FBB approach. Patients receiving FBB therapy initiated with three bolus insulin doses a day and remained on this regimen for the duration of the model. The base case time horizon was 1 year, and outcomes were assessed at trial end (32 weeks) in the budget impact model and projected annually up to 5 years in the cost-effectiveness analysis.

Patient population

The models were developed using the FullSTEP trial and, as such, represent populations of patients aged 18-years or older who have been diagnosed with type 2 diabetes for at least 12 months. To be included in the trial, patients must have been using a basal insulin (±OADs) regimen for at least 6 months pre-trial, have an HbA1c of 7.0–9.0%, and a body mass index (BMI) <40 kg/m² 18. Mean patient characteristics at baseline in the model were: age = 60.0 (SD = 9.1) years, HbA1c = 7.9% (SD = 0.6%), BMI = 31.5 (SD = 4.8) kg/m², and 51.7% were male18.

The budget impact model considers a health plan with 1 million members in the base case. The prevalence of type 2 diabetes in 2013 is taken to be 7.7% (n = 77,000), given that the prevalence of diabetes in the US was 8.3% in 2011, and that type 2 diabetes accounts for 92.5% (90–95%) of cases19. Insulin plus oral agents is prescribed to 14% of patients with diabetes, while 12% receive insulin only19. Assuming that all patients with type 1 diabetes (7.5%) receive insulin, 18.5% of patients with type 2 diabetes in the US are on insulin therapy. The database analysis by Blak et al.20 found that 17.2% of type 2 diabetes patients in the UK who used insulin were on a regimen that included prandial insulin. Excluding this percentage of patients from intensification leaves 15.3% of patient with type 2 diabetes (n = 11,795) eligible for intensification. The systematic review by Swinnen et al.21 indicated that, on average, 49% of patients with type 2 diabetes achieved target HbA1c of ≤7% using insulin detemir or insulin glargine21. Therefore, 51% of patients (n = 6015) annually require treatment intensification from basal insulin. The costs for treating this set of patients is calculated for intensification from basal insulin-only therapy to SWA and to FBB. The resulting costs form the basis of comparison between the two treatment options in this model.

Event rates and resource utilization

As per the FullSTEP trial, the model captured hypoglycemic (severe, symptomatic, and asymptomatic) events, the proportion of patients meeting HbA1c target and self-measured blood glucose (SMBG). Within-trial event rates were taken from trial data. Where available, event rates were based on the number of daily bolus injections taken, rather than assessed at the cohort level (Table 1). Within the trial period (weeks 1–32), event rates were calculated separately for FBB and three bolus doses under SWA. Beyond the trial, the event rates in weeks 22–32 were assumed to be maintained. After week 32, SWA patients receiving three bolus doses transferred to FBB event rates but remained within the SWA cohort for analysis.

Table 1. Event rates and resource use as applied in the models.

Item	Stepwise 1 bolus dose	Stepwise 2 bolus doses	Stepwise 3 bolus doses	Full basal-bolus
Hypoglycemia, events per exposure year
Severe	0.06	0.06	0.00	0.13
Symptomatic	14.68	28.36	37.08	32.49
Asymptomatic	10.64	20.03	11.39	25.51
Addition of a bolus dose,^a % of dose group
Week 11	75.84%	N/A	N/A	N/A
Week 22	24.73%	65.90%	N/A	N/A
Week 32	10.71%	32.96%	N/A	N/A
After the trial (each 10 weeks)	10.71%	32.96%	N/A	N/A
Achieving target HbA1c,^a,b % of cohort
At end of trial	56.08	56.08	56.08	63.39
Mortality, %
Background mortality	2.00%	2.00%	2.00%	2.00%
After severe hypoglycemia	0.33%	0.33%	0.33%	0.33%
Change in BMI, kg/m^2
At end of trial	0.70	0.70	0.70	0.90
Basal/Bolus insulin, IU per day
Week 11	53.40/23.80	N/A	N/A	51.70/53.60
Week 22	54.60/22.30	54.60/47.40	N/A	50.00/56.50
Week 32	54.60/23.80	54.60/49.60	54.60/61.50	49.70/56.10
After the trial	54.60/23.80	54.60/49.60	49.70/56.10	49.70/56.10
Self-measured blood glucose
Tests per day	2	3	4	4

^aMean of event rates for patients stratified by HbA1c of 7.0–8.0% and >8.0–90%; ^bTarget <7.0%.

BMI, Body mass index; HbA1c, Glycated hemoglobin; IU, International units. Event rates were taken from the clinical trial report (ANA-3786) date February 6, 2013; with data on file with Novo Nordisk.

Within the FullSTEP trial, patients were stratified by HbA1c to either the 7.0–8.0% or the 8.1–9.0% group¹⁸. SWA patients within the higher HbA1c stratum were generally more likely to add an additional bolus dose compared with patients in the lower stratum. At the first assessment period, 67% of patients in the 7.0–8.0% stratum added a second bolus dose, compared with 81% in the 8.1–9.0% stratum (overall mean = 75.84%). Due to these differences, the stratification was retained within the model with the number of patients in each stratum defined by a normal distribution with the mean and standard deviation defined by the cohort description.

Insulin doses were derived from the FullSTEP trial. For the SWA cohort, bolus doses were available by meal and the following dosing schedule was assumed: one bolus dose (midday), two bolus doses (midday and dinner), and three bolus doses (breakfast, midday, and dinner). Mean insulin doses by dose group are provided in Table 1. After the trial period, SWA patients receiving three bolus doses transferred to FBB insulin dosing. It was assumed that all patients used one needle per insulin dose and one lancet and one test strip per SMBG test. During the trial, patients performed one SMBG test for each daily insulin (basal and bolus) dose taken. Throughout the model time period, patients receiving one, two, or three bolus doses per day performed SMBG 2, 3, or 4-times daily, respectively.

Event rates were the same in both cost effectiveness and budget impact models; however, the budget impact model did not account for age-related mortality and change in BMI.

Healthcare costs

In both models, the healthcare payer perspective was used for event and resource costs (Table 2), reported in USD, with future costs and clinical benefits discounted at 3.5% annually. The majority of event and resource costs were derived from published literature (as outlined in Table 2), with insulin and needle costs provided by Novo Nordisk Inc.

Table 2. Cost data used in the analysis.

Item	Cost unit	Mean cost, 2013 USD	Reference
Basal insulin^a	Per unit	0.18	Novo Nordisk US
Bolus insulin^a	Per unit	0.19	Novo Nordisk US
Needles (basal and bolus)	Per needle	0.31	Novo Nordisk US
Self-measured blood glucose test	Per test strip	0.98	34
Lancet	Per lancet	0.10	Retail price^b
Office visit to the GP	Per visit	96.07	35
Outpatient hospital visit	Per visit	421.19	30
Emergency department visit	Per visit	1483.87	30
Inpatient admission	Per stay	18,793.66	30
Severe hypoglycemia	Per event	2157.23	30
Symptomatic hypoglycemia	Per event	31.16	29,30
Asymptomatic hypoglycemia	Per event	1.08	Assumes one SMBG test

^aPer unit costs of insulin assume that 80% of patients use an insulin pen and 20% use vial and syringe as the delivery mechanism; ^bCost of Therasense Sterile Lancets (Walmart, 2013). Costs not expressed in 2013 USD were converted using the online tool at: http://stats.areppim.com/calc/calc_usdlrxdeflator.php

Quality-of-life

Health state utility data are applied only to the cost-effectiveness analysis in order to quantify the impact of SWA and FBB therapy on patient quality-of-life (QoL). QoL data directly associated with utilities were not captured as part of the FullSTEP trial. Treatment satisfaction, however, was assessed. The Diabetes Medication Satisfaction (DiabMedSat) tool indicated that SWA was associated with significantly better treatment satisfaction at trial end compared with FBB. Transitioning from basal-only therapy to SWA, patients reported a reduction of 3 points at trial end compared with a loss of 6.3 points at trial end if transitioning to FBB. Pollack et al.22 reported that a 1 point increase in the DiabMedSat total score corresponded to an additional 0.001 points on the EQ-5D Index, providing a utility for SWA of −0.94 × 10⁻⁴ per week compared with −1.97 × 10⁻⁴ per week for FBB. Finger pricks are generally associated with reduced quality-of-life and each SMBG was given a utility of −0.000,022,123.

In terms of clinical outcomes, gain in BMI was associated with a utility of −0.0061, while meeting the HbA1c target had a utility of +0.012 24,25. Utilities associated with hypoglycemia were taken from the publication of Evans et al.26, which reported results from a survey of 551 people with type 1 diabetes, and 1603 people with type 2 diabetes covering five countries. The mean utility (assuming daytime events) over all patients with type 2 diabetes and all US patient was utilized in the model: severe hypoglycemia (−0.0575 per event) and symptomatic hypoglycemia (−0.005 per event). Asymptomatic hypoglycemia was not reported and its utility was assumed to be zero. Utilities associated with clinical outcomes were discounted at 3.5% per annum.

Sensitivity analyses

Sensitivity analyses assessed input-dependent variability in model outcomes. Probabilistic sensitivity analyses were performed around clinical, QoL, and resource use inputs using normal distributions accept for relative risks of events, for which log normal distributions were used. If standard deviations were not provided these were assumed to be 10% of the mean value of the distribution. One-way sensitivity analyses assessed the impact of unit costs, utilities, discount rates, and time horizon on incremental costs measured in 2013 USD and quality-adjusted life expectancy (QALE) measured in quality-adjusted life years (QALYs). Cost and QoL sensitivity were assessed via both percentage change and utilization of alternative sources27–30.

Results

Cost-effectiveness analysis

At the end of year 1, nearly one third (30.2%) of patients in the model maintained glycemic control using one or two bolus doses under SWA therapy. By dose group, there were 14.3% using one bolus dose and 15.8% using two bolus doses. Over the first year, the mean percentage of patients receiving one bolus dose and two bolus doses was 37.0% and 29.8%, respectively. At the end of the fifth year, only 3.3% of patients remained on one or two bolus doses; indicating the progressive nature of the disease in patients with type 2 diabetes.

Over all time horizons, SWA of insulin aspart dominated FBB. At 1 year, SWA was associated with an increase in QALE of 0.05 QALYs (0.69 vs 0.64 QALYs) and an incremental cost saving of USD 1613 (USD 8723 vs 10,336) compared with FBB. If a 5-year time horizon was used, the incremental benefits associated with SWA were 0.08 QALYs and a saving of USD 2542. SWA reduced the occurrence of severe hypoglycemia by 45% in the first year, however, QoL improvement was driven by reductions in the more frequent symptomatic hypoglycemia, the number of SMBG tests, and improved treatment satisfaction. Cost savings were driven by the lower bolus insulin requirement under SWA compared with FBB and were also influenced by reduced SMBG and less frequent severe and symptomatic hypoglycemia.

Probabilistic sensitivity analyses found that SWA dominated FBB at 1 year in 98.3% of cases (Figure 1) and at 5 years in 97.4% of cases. One-way sensitivity analyses tested the influence of changes in unit costs on the incremental cost and utilities on the incremental QoL of SWA compared with FBB. Two types of one-way sensitivity analyses types were undertaken, in the first instance inputs were varied from 25–400% of their base case value, and secondly base case values were replaced by other input values available from the literature. The unit costs of bolus insulin and SMBG tests have the largest impact on the incremental cost of SWA compared with FBB at 1 year (Figure 2). SWA is associated with reduced reliance on bolus insulin, but this is compensated for by an increased mean dose of basal insulin (Table 1). Increasing the cost of bolus insulin, thus, makes SWA more cost-effective, whereas increasing the cost of basal insulin reduces the cost-effectiveness of SWA, although SWA remains cost-effective. In the base case, the cost of insulin was calculated using the assumption that 80% of patients used insulin pens and 20% used vial and syringe as the insulin delivery mechanism. If all patients used pens or all patients used vial and syringe, the cost-effectiveness of SWA remained relatively unaffected. Switching all patients to bolus insulin via vial and syringe had the largest impact on the cost saving, reducing savings by 10.6% (Figure 2). All other switches resulted in changes of less than 3% in the incremental cost of SWA. Symptomatic hypoglycemia influences both the cost and utility of SWA, increasing either the cost or QoL impact of symptomatic hypoglycemia makes SWA more cost-effective (Figure 2) by increasing its associated cost saving or QALE benefit, respectively.

Figure 1. Probabilistic sensitivity analysis demonstrates that stepwise addition of bolus insulin generally dominates full basal bolus therapy at 1 year. Increment is the difference in cost between stepwise addition (SWA) and full basal bolus (FBB), where by the value of FBB is subtracted from that of SWA, e.g., SWA – FBB. PSA, Probabilistic sensitivity analysis; QALY, Quality-adjusted life years; USD, United States Dollars.

Figure 2. Bolus insulin and symptomatic hypoglycemia have the largest impact on the cost and clinical outcomes of stepwise addition. Graphs indicate the percentage change from the base case values: a cost saving of USD −1613 (a and b) and quality-adjusted life expectancy (QALE) benefit of 0.05 quality-adjusted life years (c and d). Variance in cost savings at 1 year given (A) percentage change (from 20–400%) in input parameters and (B) cost data from alternative sources or resource use27,28. Variance in QALE benefit given (C) percentage change (from 20–400%) in input parameters and (D) utility data from alternative sources29,30. Country codes indicate the country in which the alternative study was set. CA, Canada; DE, Germany; ES, Spain; HbA1c, Glycated hemoglobin; SMBG, self-measured blood glucose; SW, Sweden; UK, United Kingdom; USD, United States Dollars.

Reduced SMBG testing associated with one or two bolus doses under SWA compared with FBB was a key aspect of both the cost and QoL benefit associated with SWA. If no reduction in SMBG use was assumed with SWA, then SWA still dominated FBB. In this scenario, the incremental benefit associated with SWA was a gain of 0.044 QALYs and a cost saving of USD 1205 at 1 year. Sensitivity analysis demonstrated that SWA remained dominant to FBB at 1 year in 94.5% of 2000 test cases.

Budget impact assessment

Budget impact analysis estimated that a health plan (as previously described in the Methods) with one million members and with 6015 members with type 2 diabetes annually intensifying to FBB or SWA would save USD 1304 per intensifying patient with SWA compared with FBB during the FullSTEP trial period. Over these 32 weeks, the per member per month (pmpm) saving was estimated to be USD 1.06 and would save USD 7.8 million.

After 1 year, the health plan would save USD 9.7 million with SWA compared with FBB. The mean saving per intensifying patient is estimated to be USD 1612, a value which corresponds almost exactly to the saving calculated in the cost-effectiveness analysis. The saving at 1 year corresponded to a pmpm saving of USD 0.81 (Figure 3). As expected from the cost-effectiveness analysis, cost savings are driven by reduced reliance on bolus insulin (Figure 4), but are also influenced by reduced use of SMBG and lower incidence of both severe and symptomatic hypoglycemia.

Figure 3. Incremental cost of stepwise addition relative to full basal-bolus therapy. SMBG, self-measured blood glucose; USD, United States Dollars.

Figure 4. Per member per month savings with stepwise addition of bolus insulin. USD, United States dollars.

In the base case scenario, the rate of bolus dose addition was assumed to be equal to that in week 32 of the FullStep trial. Given the progressive nature of type 2 diabetes, this assumption may be conservative. A sensitivity analysis was performed to test the dependence of the result on the rate of bolus dose addition. In this test case, the rate of bolus dose addition was doubled after the 32 weeks of the trial, meaning that SWA patients reached three bolus doses (e.g., FBB) more quickly. In this scenario, SWA was associated with a saving of USD 1595 per intensifying patient at 1 year. The saving for the health plan in this scenario was estimated to be USD 9.6 million, a pmpm saving of USD 0.80.

Discussion

Based on these simple, transparent short-term models, use of SWA is likely to dominate FBB in the US setting and result in cost savings for health plan providers. The results were consistent over all time horizons considered, 1–5 years. Improvements in clinical outcomes were driven by reduced incidence of severe and symptomatic hypoglycemia. Also important in this context was the reduction in SMBG and the improved patient satisfaction with treatment as measured by the DiabMedSat questionnaire. Use of SWA was associated with reduced costs compared with FBB, which was driven by reduced need for bolus insulin and lower costs associated with hypoglycemic events. Sensitivity analyses indicated that the conclusion that SWA dominated FBB was robust to changes in model parameters, with 98.3% of probabilistic sensitivity analysis outcomes resulting in SWA dominating FBB. The highest incremental cost-effectiveness ratio estimated in the 2000 iterations was USD 40,000 per QALY gained, which is still likely to be considered cost-effective.

There are a large number of co-morbidities with important health and cost implications associated with diabetes. For this reason, many cost-effectiveness analyses take a longer-term perspective, assessing cost-effectiveness over decades or patient lifetimes31–33. Long-term analyses are generally driven by changes in HbA1c and its impact on the incidence of diabetes-related complications and comorbidities. There is also the potential advantage of offsetting higher treatment costs in the short-term with lower complication costs in the long-term. The FullSTEP study demonstrated that there was no significant difference in change in HbA1c or achieving target HbA1c between SWA and FBB. No significant differences in long-term outcomes are, therefore, expected and a short-term model is most applicable. This conclusion is supported by the low percentage of patients maintaining good glycemic control using fewer than three bolus doses per day after 5 years in the model. At this time, essentially all patients are on FBB and there should be no substantial differences in cost or clinical outcomes. The models presented here, therefore, focus on the first years after intensification from basal insulin-only therapy. It is expected that the cost savings realized through SWA are generated mainly in the first year. This is demonstrated in the difference in cost savings with SWA between a 1-year and 5-year time horizon (USD 929). After year 1, the mean annual saving over the next 4 years was USD 232.25.

As with all modeling studies, there are limitations with this study that are noted here to allow for informed interpretation of the results. Within the FullSTEP trial, three bolus doses via SWA was associated with a lower rate of severe hypoglycemia than was SWA of one or two bolus doses (Table 1). The FullSTEP trial was a 32-week trial and, due to the trial design there was a maximum of 10-weeks in which patients could receive three bolus doses under the SWA protocol. Due to limited data, patients receiving three bolus doses under SWA were transferred to FBB event rates and resource use after 32 weeks. Furthermore, the FullSTEP trial included only those patients with an HbA1c of 7.0–9.0% and a BMI <40 kg/m²; as such, the results of this analysis should only be taken to represent this patient population. The rate of bolus dose addition within the trial was influenced by the HbA1c stratum of the patient, with those patients in the higher stratum more likely to add a bolus dose at the first assessment. The HbA1c influence further suggests that patients with HbA1c higher than 9.0% may not be represented by the results of this study.

With respect to healthcare costs and health state utilities, the costs of hypoglycemia were derived from published studies with data specific to the US setting34,35; however, the costs for non-severe hypoglycemia were not specific to type 2 diabetes but rather diabetes in general34. Utilities associated with hypoglycemia were means calculated from utilities for hypoglycemia in the US and in patients with type 2 diabetes, with both values taken from the same study26. The utility associated with meeting HbA1c targets was taken from a study of patients with type 2 diabetes in the German setting25.

There are limited data in the published literature regarding the distribution of treatments within a population with type 2 diabetes. The percentage of patients receiving basal insulin-only was estimated from data available from the Centers for Disease Control and Prevention19. Similarly sparse data were available regarding the annual intensification of basal insulin-only therapy in real world clinical practice. The value used in this study (51%) was taken as the mean percentage of patients not meeting an HbA1c target of ≤7% using either basal insulin glargine or insulin detemir, as presented in the systematic review by Swinnen et al.21. Higher values were available, with 72.2% (at 1 year) and 45.4% (at 6 months) of patients reported as not meeting an HbA1c target of 7.0% with basal insulin only ±OADs36,37. Changes to the patient population would influence the total, but not the ‘per intensifying patient’, saving in the budget impact analysis. It is, therefore, reasonable to conclude that SWA is associated with a cost-saving of ∼USD 1600 in the first year compared with FBB. A test case scenario, in which the rate of bolus dose addition was doubled after the 32 week trial, indicated that the savings associated with SWA in the first year were robust to model assumptions.

The benefits associated with SWA as estimated in this study are in line with other publications on SWA of bolus insulin17,18. Both these studies highlight the reduction in hypoglycemia associated with SWA of bolus insulin. The models presented here, however, provide the first quantification of the cost and QoL benefits associated with SWA outside of a clinical trial. Independent development of two models estimating the costs associated with SWA and FBB therapy provides opportunity to evaluate variance not only via probabilistic and one-way sensitivity analyses but also from the perspective of model structure. That the two models closely agree on the incremental cost saving associated with SWA (USD 1613 and USD 1612) at 1 year supports the robustness of the approach taken and the conclusion that SWA has both clinical and economic benefits.

Conclusion

Intensification of treatment is a critical aspect of care in type 2 diabetes. Associated with improved QoL and reduced costs as compared with FBB therapy, SWA should be considered a beneficial and cost-saving alternative to FBB therapy for the intensification of treatment in type 2 diabetes.

Transparency

Declaration of funding

This study was funded by Novo Nordisk Inc.

Declaration of financial/other relationships

BK and JL are employees of Novo Nordisk Inc. RS and WV are employees of Ossian Health Economics and Communications, who received consultancy fees for work associated with this manuscript. Ossian Health Economics and Communications and its employees have worked with many leading pharmaceutical and medical devices companies. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Notes

†Event rates were taken from the clinical trial report (ANA-3786) date February 06, 2013; with data on file with Novo Nordisk.