Introduction
Due to the progressive nature of type 2 diabetes mellitus (T2DM), patients usually require treatment intensification to maintain glycemic control, eventually resulting in insulin therapy [
1]. After treatment failure on basal insulin, guidelines recommend intensification with either a glucagon-like peptide-1 receptor agonist (GLP-1RA) or mealtime (bolus) insulin [
1]. Premixed insulin can offer a convenient alternative to basal-bolus therapy, controlling both fasting and postprandial glucose with fewer injections. However, premixed insulin requires resuspension, and insufficient mixing can result in incorrect dosing [
2].
Insulin degludec/insulin aspart (IDegAsp) is the first soluble co-formulation of a basal and a rapid-acting insulin analog in a single injection. The unique properties of insulin degludec (IDeg) enable combination with the rapid-acting insulin, insulin aspart [
3]. IDeg has a long duration of action, with a longer half-life and less within-patient variability versus insulin glargine [
4,
5]. The IDegAsp soluble co-formulation has the additional advantage that it does not require resuspension, eliminating the risk of incomplete mixing, which can lead to hypoglycemia [
6].
The efficacy and safety of IDegAsp was investigated in a phase 3 clinical program (BOOST). A combined analysis of two trials [
7,
8] comparing IDegAsp twice daily (BID) with biphasic insulin aspart (BIAsp 30) BID in insulin-experienced patients with T2DM showed that, at similar levels of glycemic control, IDegAsp resulted in a lower fasting plasma glucose (PG), lower rate of overall and nocturnal hypoglycemia and less weight gain compared with BIAsp 30, all at a lower insulin dose [
9].
To optimize the use of health-care resources, the decision to prescribe a particular product is dependent on clinical and economic evidence. Therefore, it is important that the cost-effectiveness of diabetes interventions is investigated. The cost-effectiveness analyses of diabetes interventions have historically been conducted by estimating the long-term clinical consequences as a function of differences in HbA
1c. However, according to the US Food and Drug Administration guidance, new insulins should be compared with a standard insulin (and not placebo or a non-insulin agent), aiming to achieve similar glycemic control, thus allowing the comparison of safety end points, such as hypoglycemia, body weight, and insulin dose [
10,
11]. This approach is known as ‘treat-to-target’. It follows that there are no differences in long-term risk parameters related to HbA
1c, and a short-term cost-effectiveness model is more appropriate for an economic evaluation of these secondary end points than a long-term model. Such a model has been used to compare treatment with IDeg versus insulin glargine in patients with type 1 diabetes mellitus (T1DM) and T2DM, from Swedish and UK health-care perspectives [
12‐
14].
The objective of this study was to evaluate the cost-effectiveness of IDegAsp BID compared to BIAsp30 BID in patients with T2DM, from a Danish health-care cost perspective, using a short-term cost-effectiveness model.
Results
This analysis shows that IDegAsp is a cost-effective treatment option compared with BIAsp 30 in patients with T2DM. The estimated ICER in the base-case analysis was 81,507.91 DKK (Table
4).
Table 4
Total costs and effects per patient for 5 years for IDegAsp versus BIAsp 30
Costs |
Pharmacy costs |
Insulin | 60,565.74 | 35,703.42 | 24,862.32 |
Needles | 6616.01 | 6616.01 | 0 |
Routine SMBG tests | 8813.48 | 8813.48 | 0 |
Hypoglycemic event costs |
Non-severe daytime events | 1501.38 | 1935.76 | –434.38 |
Non-severe nocturnal events | 157.54 | 399.29 | –241.75 |
Severe events | 1655.93 | 7716.18 | –6060.25 |
Total costs | 79,310.08 | 61,184.14 | 18,125.94 |
Effects |
QALY | 3.5366 | 3.3142 | 0.2224 |
ICER (cost per QALY) | – | – | 81,507.91 |
The total incremental discounted cost per patient over 5 years of IDegAsp compared to BIAsp 30 treatment was 18,125.94 DKK, mainly due to an increased insulin cost (Table
4). A lower incremental cost of treating non-severe and severe hypoglycemia was observed (Table
4), due primarily to the statistically significant reduction in hypoglycemic events in the maintenance period with IDegAsp versus BIAsp 30 in the combined analysis (Table
1).
One-Way Sensitivity Analyses
In all sensitivity analyses, ICERs were below normally accepted thresholds, with estimates ranging from 71,012 DKK to 209,446 DKK per QALY gained (Table S1). Varying the time horizon had a small impact on the ICER, with a lower cost per QALY gained with a 10 year time horizon. Similarly, changing the discount rate did not have a significant impact on the ICER. The rate of hypoglycemia applied to the model had a larger impact on the results, with the higher published rates from Denmark [
35,
36] resulting in a lower ICER (71,012 DKK) compared to the base case.
Using the observed hypoglycemia event rates from the combined analysis, where fewer severe hypoglycemic events were observed than in the UKHSG study, IDegAsp was cost-effective with an ICER of 153,440 DKK per QALY gained. When the cost of hypoglycemia was increased or decreased by 20%, IDegAsp was still cost-effective versus BIAsp 30, with ICERs of 76,049 DKK and 86,968 DKK, respectively. Using the disutilities per hypoglycemic event published by Currie et al. [
37] resulted in an ICER of 209,446 DKK. Using the clinical data from the global trial instead of the combined analysis resulted in an ICER of 100,289 DKK [
7]. Assuming a disutility per SMBG test of 0 did not have a significant impact on the ICER, compared with the base case.
Probabilistic Sensitivity Analysis
The PSA input variables are outlined in Table S2 in the supplementary material. At a willingness-to-pay threshold of 250,000 DKK per QALY gained, the probability that IDegAsp was cost-effective relative to BIAsp 30 was 99.5% (Fig. S1 in the supplementary material).
Discussion
Due to limited health-care resources, demonstrating the value of new therapies is an essential part of clinical and economical decision making. This simple and transparent short-term cost-effectiveness model, conducted from a Danish health-care perspective, focuses on the impact of important aspects of insulin therapy, including hypoglycemia and dosing. For patients with T2DM, IDegAsp is a cost-effective treatment compared with BIAsp 30, with an ICER of 81,507.91 DKK, over a 5-year time horizon. Several one-way sensitivity analyses and a PSA also found that IDegAsp was cost-effective versus BIAsp 30, supporting the robustness of the analysis. The ICERs were stable when the rates of hypoglycemia, unit cost of hypoglycemia, and the disutility of a hypoglycemic event were varied, ranging from 71,012 DKK to 209,446 DKK, all cost-effective based on a threshold of 250,000 DKK. In Denmark, there is no official published cost-effectiveness threshold. However, two recent Danish cost-effectiveness studies have applied thresholds varying from 250,000 to 500,000 DKK per QALY, with the one in the field of diabetes applying a threshold of 250,000 DKK [
39,
40].
The two-trial combined analysis was used in the base-case model to increase the sample size and thereby strengthen the certainty of the parameter estimates derived from individual trials. Previously, a pre-specified hypoglycemia meta-analysis using pooled data from seven trials in the IDeg clinical program was conducted [
41]. The combined analysis used here was not pre-specified; however, it was conducted using the same methodology as the IDeg meta-analysis [
41] the hypoglycemia classification described in a second IDeg meta-analysis [
42].
The combined analysis showed that, at similar HbA
1c levels, IDegAsp resulted in lower rates of overall and nocturnal confirmed hypoglycemia, particularly during the maintenance period, and less weight gain compared with BIAsp 30, while using a lower dose [
9]. Although hypoglycemia is not a problem for all insulin-treated patients, its frequency tends to increase with longer disease duration and more intensive insulin regimens [
23]. Additionally, treatment of hypoglycemia is associated with considerable resource use and cost and is a burden to patients [
29,
43,
44]. In this analysis, the difference in the rates of hypoglycemia had a noticeable impact on the ICER. In this model, the mean population rates of hypoglycemia from published studies were used in the base case and sensitivity analysis [
18,
35,
36] to more closely reflect the rates observed in routine clinical practice versus clinical trials (where patients at high risk of severe hypoglycemia are excluded). The UKHSG rates used in the base-case analysis [
18] were a conservative estimate compared with the rates in a Danish population used for a sensitivity analysis, where the rates of non-severe daytime and nocturnal hypoglycemia were higher [
35,
36]. Hypoglycemia disutilities from a global (US, UK, Germany, Sweden) TTO study were applied to the base-case model [
32] and those from a UK setting in a sensitivity analysis [
37]. Therefore, in subpopulations of patients, who experience recurrent hypoglycemia, hypoglycemia unawareness, or nocturnal hypoglycemia, a treatment such as IDegAsp might provide additional value for money as compared to the base case. The lower dose required for equivalent glycemic control for IDegAsp compared to BIAsp 30 in the combined analysis contributed to the cost-effectiveness of IDegAsp. The lower weight gain (measured by BMI) only had a limited impact on the cost-effectiveness results.
All modeling approaches are critically influenced by the quality of the input parameters. This model only used parameter estimates for which a statistically significant difference between the treatment arms was documented and assumed that all other differences were due to random variation. This cost-effectiveness model is limited by a variety of factors. The clinical data are derived from post hoc analyses of two clinical trials. The generalizability of clinical trials with high internal validity but usually low external validity, due to the highly selected population, which may not be representative of a real world clinical practice setting, is a common limitation inherent to economic modeling. The model assumed patients continued treatment with IDegAsp and BIAsp 30 for the 5-year period, without changing to another insulin regimen. As the data are based on treat-to-target clinical trials with no differences in HbA1c, any differences in mortality and morbidity are not expected and so not included in the model.
As with most models, the cost data for hypoglycemia were collected from a variety of publicly available sources, which measure parameters differently and may not accurately reflect the economic burden of hypoglycemia. The resources used following a hypoglycemia event were collected during the clinical trials for a selected population and may not represent the burden seen in clinical practice. The actual costs of hypoglycemia may be higher as the estimations did not take into consideration out-of-pocket expenses or lost work productivity. Finally, the model is based on comparable glycemic control between IDegAsp and BIAsp 30, with a potentially lower insulin dose requirement with IDegAsp. To address some of these limitations, more extensive clinical practice experience with IDegAsp would be useful for further health economic evaluations.
Acknowledgments
The sponsorship for this study was funded by Novo Nordisk A/S, Søborg, Denmark. Medical writing assistance and editorial/submission support were provided by Adele Buss and Beverly La Ferla of Watermeadow Medical, an Ashfield Company, a part of UDG Healthcare plc, funded by Novo Nordisk A/S. All named authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship for this manuscript, take responsibility for the integrity of the work as a whole, and have given final approval for the version to be published. Marc Evans contributed to the study design, interpretation of the results, and to the drafting, revising, and final approval of this manuscript. As corresponding author, Marc Evans agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Brian Bekker Hansen contributed to the design, conduct/data collection, analysis, and interpretation of the results and to the drafting, revising, and final approval of this manuscript. Jens Gundgaard contributed to the design, conduct/data collection, analysis, and interpretation of the results and to the drafting, revising, and final approval of this manuscript.