Management of Patients with Type 2 Diabetes with Once-Weekly Semaglutide Versus Dulaglutide, Exenatide ER, Liraglutide and Lixisenatide: A Cost-Effectiveness Analysis in the Danish Setting

Analyses were performed using the IQVIA CORE Diabetes Model (version 9.0), a non-product specific diabetes policy analysis tool based on a series of inter-dependent sub-models that simulate short- and long-term diabetes-related complications, with projected outcomes validated against real-life data in 2004 and more recently in 2014 [34, 35]. The model projects outcomes for populations based on user inputs, including baseline cohort characteristics and history of complications; current and future diabetes management and concomitant medications; screening strategies; and changes in physiological parameters over time [36]. Each of the 15 sub-models, which simulate angina, congestive heart failure, myocardial infarction, stroke, peripheral vascular disease, macular edema, cataract, retinopathy, nephropathy, neuropathy, foot ulcer and amputation, hypoglycemia, ketoacidosis, lactic acidosis and non-specific mortality, has a semi-Markov structure and uses time, state, time-in-state and diabetes type-dependent probabilities derived from published sources. Model outputs include life expectancy (measured in years), quality-adjusted life expectancy (measured in quality-adjusted life years [QALYs]), cumulative incidence and time to onset of diabetes-related complications, direct medical costs, and cost-effectiveness scatterplots and acceptability curves. Incremental cost-effectiveness ratios (ICERs), calculated as the incremental cost per unit of effectiveness gained by using the new intervention versus the comparator, can be used as a measure of cost-effectiveness where an intervention is associated with improved clinical outcomes at an increased cost.

All base case analyses were projected over patient lifetimes (up to 50 years) and utilized UK Prospective Diabetes Study (UKPDS) 68 risk equations, with background mortality based on Danish-specific life tables from the World Health Organization (WHO) [37]. This approach was chosen to capture all relevant long-term complications and costs and to assess the impact of each intervention on life expectancy and quality-adjusted life expectancy, in line with previous analyses of GLP-1 receptor agonists and the guidance on assessing the cost-effectiveness of interventions for diabetes [38‐41]. All base case and sensitivity analyses were performed using a first-order Monte Carlo approach, with probabilistic sensitivity analyses performed separately using a second-order Monte Carlo approach. Clinical and cost outcomes were discounted at 3.0% annually, consistent with guidelines for health economic analyses in the Danish setting [42].

Clinical data, including baseline cohort characteristics and treatment effects, were sourced from SUSTAIN 7 for primary comparisons of semaglutide versus dulaglutide (Table 1 and Table S1). Data were obtained from pre-specified end points wherever possible, but to fulfill all of the data requirements for an analysis using the IQVIA CORE Diabetes Model, a number of post hoc analyses of the trial data were required. Pre-specified statistical analyses in SUSTAIN 7 assessed the statistical significance of differences in treatment effects between semaglutide 1 mg and dulaglutide 1.5 mg, but post hoc analyses were required to assess the statistical significance of differences in treatment effects between semaglutide 0.5 mg and dulaglutide 1.5 mg. While dulaglutide 0.75 mg was also assessed in SUSTAIN 7, this is indicated only for monotherapy or in patients aged > 75 years by the European Medicines Agency (EMA), so this dose was not included in the present study [43]. Treatment effects (both statistically and non-statistically significant differences, in line with ISPOR guidance) were applied in the first year of the analysis, while hypoglycemic event rates from SUSTAIN 7 were applied until patients treatment-intensified with the addition of basal insulin [44].

Baseline cohort characteristics and treatment effects for supporting comparison of semaglutide 1 mg with exenatide ER were sourced from SUSTAIN 3 (Table 1 and Table S1). For supporting comparisons of semaglutide with exenatide ER, liraglutide and lixisenatide, treatment effect data were obtained from the NMA, with baseline patient characteristics based on the SUSTAIN 7 cohort (Table 1 and Table S1). Changes from baseline in HbA1c, systolic blood pressure and BMI (the outcomes included in the NMA that were applicable to an analysis using the IQVIA CORE Diabetes Model) versus placebo were applied in all treatment arms, with both statistically significant and non-statistically significant differences included. Where parameters were not included in the NMA, inputs were assumed to be 0 in all arms to ensure that these did not drive cost-effectiveness outcomes.

Following application of the treatment effects in the first year of the analysis, HbA1c was modeled to increase by 0.14% per year, based on the metformin arm of the ADOPT study (as no data on long-term HbA1c changes with GLP-1 receptor agonists are available), and patients were assumed to receive semaglutide or comparator treatment until HbA1c reached 7.5% (Fig. 1) [45]. At this stage, treatment was intensified with addition of basal insulin and continuation of GLP-1 receptor agonist therapy (the most common intensification step from GLP-1 receptor agonists in Denmark), with the reduction in HbA1c based on an insulin-naïve population and derived from the “Core” multivariate equations estimated by Willis et al. [46]. HbA1c was subsequently modeled to follow the UKPDS progression equation until reaching 8.0%, after which patients discontinued GLP-1 receptor agonist therapy and intensified to basal-bolus insulin. HbA1c was modeled to drop according to the multivariate “Core” equations for an insulin-experienced population published by Willis et al. and then to progress according to the UKPDS equation, with patients remaining on basal-bolus insulin for the remainder of their lifetimes [46]. Using this approach, patients received semaglutide 0.5 mg, 1 mg and dulaglutide 1.5 mg for 7, 9 and 6 years, respectively, and once-weekly semaglutide 0.5 mg, 1 mg and dulaglutide 1.5 mg plus basal insulin for 3 years, before intensifying to basal-bolus insulin for the remainder of their lifetimes (Fig. 1). This approach was chosen to reflect common clinical practice, where, due to the progressive nature of T2D, glycemic control cannot be maintained indefinitely with one medication [27]. For the majority of patients, and particularly for those with high cardiovascular disease risk, for whom GLP-1 receptor agonists are recommended following metformin failure, treatment intensification with the addition of basal insulin followed by basal-bolus insulin therapy is required to maintain good glycemic control over the long term [18, 27, 29]. Variations in initial treatment duration, including an analysis with GLP-1 receptor agonist therapy maintained for patient lifetimes, were explored in sensitivity analyses.

BMI progression was modeled to correlate with treatment switching patterns, taking into account the weight gain often observed in patients receiving basal insulin therapy (Fig. 2). Initial treatment effects with semaglutide or the comparator were maintained until basal insulin was added to the treatment regimen, where BMI was modeled to increase according to the “Core” equations for an insulin-naïve population provided by Willis et al. [46]. BMI was maintained at this level until patients intensified to basal-bolus insulin therapy, after which BMI was further increased according to the equations for an insulin-experienced population, again published by Willis et al., with the differences between the treatment arms abolished [46].

Systolic blood pressure and serum lipids were assumed to follow the natural progression algorithms built into the IQVIA CORE Diabetes Model from the start of each analysis, based on the UKPDS or Framingham data (as described by Palmer et al.) [36]. Following the first treatment intensification with the addition of basal insulin, non-severe and severe hypoglycemic events were projected to increase to 10.615 and 0.258 events per patient per year, respectively, based on the Danish Hypoglycemia Assessment Tool study (as the multivariate regression results calculated by Willis et al. had a poor fit) [47]. These were applied in all arms and were maintained for the remainder of the analyses.

Costs were estimated from a Danish healthcare payer perspective and expressed in 2017 Danish kroner (DKK). Pharmacy acquisition costs were sourced from the Danish Medicines Agency in March 2018 [48]. Diabetes medication resource use was based on the relevant clinical trials from which data were taken for each comparison.

Following addition of basal insulin to the treatment regimen, patients were assumed to receive the defined daily dose (DDD) of 40 IU, with the cost weighted to the market share of basal insulin in Denmark (to reflect current practice). Once patients intensified to basal-bolus insulin, patients were assumed to receive the DDDs of both basal insulin and bolus insulin (both 40 IU), with the cost again weighted to the market shares of basal and bolus insulin in Denmark. Costs were the same in all arms following the second intensification step. Resource use was used to calculate annual treatment costs for each arm (Table S2, Table S3 and Table S4). Annual costs also captured concomitant medications, needle use and self-monitoring of blood glucose (SMBG) testing with GLP-1 receptor agonist and intensification treatments.

Costs of treating diabetes-related complications were based on a literature review updated in 2018, with prices inflated to 2017 DKK (Table S5). Utilities relating to quality of life were taken from a 2014 review by Beaudet et al., with the exception of those associated with hypoglycemia [49]. A diminishing approach for disutilities relating to non-severe hypoglycemic events was used, based on a 2014 publication by Lauridsen et al., while a disutility applied following severe events was sourced from a 2013 publication by Evans et al., both of which were published after the literature searches by Beaudet et al. had been completed [50, 51].

Projection of long-term outcomes from short-term data is associated with uncertainty. Therefore, to test the robustness of the primary analysis, several one-way deterministic sensitivity analyses and a probabilistic sensitivity analysis were conducted on key parameters and assumptions of the analyses.

This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.

Long-term projections, based on data from SUSTAIN 7, indicated that semaglutide 0.5 mg and 1 mg were associated with improvements in life expectancy of 0.04 and 0.15 years, respectively, and quality-adjusted life expectancy of 0.11 and 0.34 QALYs, respectively, versus dulaglutide (Table 2). Improvements in life expectancy and quality-adjusted life expectancy were a result of the multifactorial risk-reduction effects of semaglutide, delaying the time to treatment intensification and resulting in a delayed time to onset and reduced cumulative incidence of diabetes-relating complications in the long term.

Semaglutide 0.5 mg and 1 mg were associated with cost savings of DKK 289 and DKK 13,416 per patient, respectively, versus dulaglutide over the lifetime course of the analysis (Fig. 3). This was primarily due to improved glycemic control with semaglutide, delaying the first intensification step to the costlier treatment combination of a GLP-1 receptor agonist plus basal insulin (Table S2). Further cost savings were achieved through avoidance of diabetes-related complications, most notably severe hypoglycemia (mean cost savings of DKK 2719 per patient with semaglutide 0.5 mg and DKK 6158 per patient with semaglutide 1 mg). With clinical benefits achieved at cost savings, both semaglutide 0.5 mg and 1 mg were considered dominant versus dulaglutide.

Sensitivity analyses, conducted for the primary comparisons of semaglutide 0.5 mg and 1 mg versus dulaglutide, showed that the results of the base case analysis were robust to changes in the input parameters and assumptions used (Table 3 and Table S6). Shortening the time horizon to 10 and 20 years resulted in reduced incremental clinical benefits for semaglutide, but the 1 mg dose remained cost saving and dominant versus dulaglutide, while the 0.5 mg dose was associated with increased costs and ICERs of DKK 12,793 and DKK 24,490 per QALY gained, respectively.

Maintaining GLP-1 therapy for patient lifetimes led to increased clinical benefits for both doses of semaglutide, with the 1 mg dose remaining cost saving and dominant but the 0.5 mg dose associated with slight cost increases and an ICER of DKK 6264 per QALY gained.

Including only statistically significant treatment effects between the treatment arms led to only slightly reduced clinical benefits and cost savings for semaglutide 1 mg, and it remained dominant versus dulaglutide, while semaglutide 0.5 mg was associated with reduced clinical benefits and increased incremental costs, yielding an ICER of DKK 683,944 per QALY gained. This was driven by treatment switching occurring simultaneously, as HbA1c progression was equal in both treatment arms.

Exploration of alternative treatment switching approaches, with a set period of 3 or 5 years for GLP-1 receptor agonist treatment followed by intensification to basal insulin (aligned with a recent cost-effectiveness analysis of semaglutide published in the UK setting), resulted in reduced clinical benefits and increased incremental costs for semaglutide 0.5 mg and 1 mg, yielding ICERs of DKK 68,414 and DKK 30,408 per QALY gained, respectively, when treatment switching occurred after 3 years, and DKK 96,326 and DKK 31,721 per QALY gained, respectively, when treatment switching occurred after 5 years [52]. There were only minor changes to the overall results when altering HbA1c thresholds for treatment switching, changing costs by ± 25% or using alternative BMI and hypoglycemia disutilities versus dulaglutide [51, 53, 54].

Clinical outcomes from the PSA were similar to those from the base case analysis, but semaglutide 0.5 mg and 1 mg were associated with increased incremental costs and reduced cost savings, respectively, with greater variance shown around the mean outcomes (Fig. 4). Semaglutide 0.5 mg and 1 mg were associated with increased quality-adjusted life expectancy of 0.10 and 0.30 QALYs, respectively, versus dulaglutide. Mean costs were DKK 3668 higher with semaglutide 0.5 mg but DKK 6718 lower with semaglutide 1 mg versus dulaglutide over patient lifetimes. Semaglutide 0.5 mg was therefore associated with an ICER of DKK 37,197 per QALY gained, while semaglutide 1 mg was dominant versus dulaglutide in the PSA. At a willingness-to-pay threshold of DKK 250,000 per QALY gained (a representative value based on GBP 20,000 in the UK), the probabilities of once-weekly semaglutide 0.5 mg and 1 mg being considered cost-effective were 58.8% and 79.2%, respectively, versus dulaglutide 1.5 mg (Fig. 5).

In the supporting analysis based on SUSTAIN 3, semaglutide 1 mg was projected to improve life expectancy by 0.24 years, and quality-adjusted life expectancy by 0.47 QALYs, versus exenatide ER (Table 4). Clinical benefits were achieved at a cost saving from a healthcare payer perspective, with semaglutide 1 mg costing DKK 18,088 less over a patient’s lifetime than exenatide ER. Clinical benefits and cost savings were a result of a reduced incidence and delayed time to onset of diabetes-related complications.

In the supporting analyses based on the NMA, semaglutide 0.5 mg and 1 mg were associated with clinical benefits versus all comparators (Table 5). Over patient lifetimes, semaglutide 0.5 mg was associated with higher costs of DKK 3218 versus exenatide ER and DKK 1518 versus liraglutide 1.2 mg, but cost savings of DKK 30,970 versus liraglutide 1.8 mg and DKK 16,961 versus lixisenatide. This yielded ICERs of DKK 17,024 and DKK 7390 per QALY gained for semaglutide 0.5 mg versus exenatide ER and liraglutide 1.2 mg, respectively, while it was considered dominant versus liraglutide 1.8 mg and lixisenatide. Semaglutide 1 mg was associated with cost savings of DKK 11,884, DKK 13,583, DKK 46,071 and DKK 32,063 versus exenatide ER, liraglutide 1.2 mg, liraglutide 1.8 mg and lixisenatide, respectively, and was therefore considered dominant versus all comparators.