Modelling Approach and Overview
A cost-effectiveness analysis was performed using version 9.0 of the IQVIA CORE Diabetes Model . The model is a non-product-specific diabetes policy analysis tool, capable of performing real-time simulations taking into account intensive or conventional diabetes therapy, oral hypoglycaemic medications, screening and treatment strategies for microvascular complications, treatment strategies for end-stage complications and multifactorial interventions. The model is based on a series of interdependent sub-models that simulate non-specific mortality and the complications of diabetes. Each sub-model has a semi-Markov structure and uses time, state, time-in-state and diabetes type-dependent probabilities derived from published sources. Model outputs include time to onset and cumulative incidence of complications, life expectancy, quality-adjusted life expectancy (expressed in quality-adjusted life years, QALYs), direct costs and, where required, incremental cost-effectiveness ratios (ICERs), which describe the cost per additional unit of effectiveness gained for the intervention versus the comparator. Projected long-term outcomes have been validated against real-life data in 2004 and more recently in 2014 [8, 9].
The analysis was consistent with previously published cost-effectiveness analyses in the UK setting [10, 11]. A lifetime time horizon was used, as recommended in modelling guidelines for the UK released by NICE, to ensure that all differences in long-term complications (and their impact on costs and quality of life) and mortality with once-weekly semaglutide 1 mg and empagliflozin 25 mg were captured . Projected cost and clinical outcomes were discounted at 3.5% annually, in line with modelling guidelines for the UK . Base case and scenario analyses were performed using a second-order Monte Carlo approach, with baseline cohort characteristics, treatment effects, costs of complications, utilities and transition probabilities relating to myocardial infarction, stroke, congestive heart failure and angina sampled in each model iteration.
Baseline cohort characteristics were based on pooled data from SUSTAIN 2, SUSTAIN 3, SUSTAIN 8 and PIONEER 2, as these studies informed the meta-analysis. The mean (standard deviation, SD) age of the cohort was 56 (10.3) years, with mean duration of diabetes of 7 (5.9) years, mean HbA1c of 8.2 (1.0)% (66  mmol/mol) and mean BMI of 32.8 (6.7) kg/m2. Approximately 4%, 2% and 1% of patients had a history of myocardial infarction, angina and peripheral vascular disease, respectively. Few patients had a history of renal complications, with less than 1% of patients with a history of microalbuminuria. Of the ophthalmic complications, 8% of patients had a history of background diabetic retinopathy, while proliferative diabetes retinopathy, cataract and severe vision loss were rare. Alcohol and tobacco consumption were assumed to be the same as the general UK population, as these were not collected in the clinical trials [13, 14]. Changes in physiological parameters with once-weekly semaglutide 1 mg and empagliflozin 25 mg were based on the outcomes calculated in the meta-regression based on individual patient data (Table 1) . Once-weekly semaglutide 1 mg was associated with significantly greater improvements in HbA1c, total cholesterol, LDL cholesterol, triglycerides and BMI compared with empagliflozin 25 mg.
Long-term parameter progression and treatment intensification was based on a clinically realistic approach. Following application of the treatment effects in the first year of the analysis, HbA1c was assumed to increase based on the UKPDS progression equation. This resulted in HbA1c increasing in both arms of the analysis, with the difference between the treatment arms gradually diminished. When HbA1c exceeded 7.5% (58 mmol/mol) (the NICE threshold for treatment intensification ) patients discontinued once-weekly semaglutide 1 mg or empagliflozin 25 mg and initiated treatment with basal insulin (assumed to be generic insulin glargine [insulin Abasaglar®]). At this stage, a reduction in HbA1c based on an insulin-naïve population derived from the “Core” multivariate equations estimated by Willis et al. was applied . This approach was chosen to reflect common clinical practice where, because of the progressive nature of type 2 diabetes, glycaemic control cannot be maintained indefinitely and the addition of further medications, such as basal insulin, is required . Reductions in BMI were assumed to persist while patients received once-weekly semaglutide 1 mg or empagliflozin 25 mg, with BMI returning to baseline when basal insulin was initiated (thereby abolishing the difference). In both arms, changes in blood pressure and serum lipids over the long-term were based on the natural progression algorithms built into the IQVIA CORE Diabetes Model, based on the UKPDS and Framingham data, respectively.
All costs were accounted from a UK healthcare payer perspective in 2019 pounds sterling (GBP). Direct costs captured included pharmacy costs, costs associated with diabetes-related complications and patient management costs. Unit costs of diabetes medications were taken from the Monthly Index of Medical Specialities (MIMS) database, and were used to calculate annual pharmacy costs with once-weekly semaglutide 1 mg, empagliflozin 25 mg and basal insulin (assumed 40 IU based on the defined daily dose) [18, 19]. Costs of complications and patient management were consistent with previously published cost-effectiveness analyses for the UK setting, with costs taken from published sources and inflated to 2019 values where appropriate (Table 2) [10, 20].
As diabetes progresses, patients develop complications that influence their overall health-related quality of life. It was therefore important to evaluate both mortality and morbidity, and address the utility levels associated with each of the complications modelled. Utilities associated with each diabetes-related complication (Table 3) were taken from a 2014 review by Beaudet et al., with hypoglycaemia disutilities coming from Evans et al. 2013 (published after the literature searches by Beaudet et al. had been completed) [21, 22]. Beaudet et al. reviewed the methods of the identified publications to ensure that they met the criteria of the NICE reference case.
Projection of outcomes over patient lifetimes is associated with uncertainty, and therefore a series of scenario analyses were performed to assess the robustness of the model results. The base case analysis used a 50-year time horizon, and the impact of shortening the time horizon of the analysis was examined by running analyses over 20- and 10-year time horizons. In the base case analysis, discount rates of 3.5% per annum for future clinical and cost outcomes were applied, with a scenario analysis conducted applying 0% discount rates. An analysis was conducted with only the statistically significant differences between the once-weekly semaglutide 1 mg and empagliflozin 25 mg arms applied, whereas the base case analysis applied all treatment effects irrespective of statistical significance.
The base case analysis applied a disutility per BMI unit above 25 kg/m2 of − 0.0061, and a scenario analysis was conducted using an alternative value of − 0.01 to assess the impact of using an alternative source to inform this model input . This larger disutility gives greater impact to weight changes compared with the conservative disutility used in the base case analysis. To assess the impact of the disutilities applied following hypoglycaemic events, a scenario was prepared with alternative disutilities for severe and non-severe hypoglycaemic events applied, as reported by Currie et al. (− 0.0118 per severe hypoglycaemic event and − 0.0035 per non-severe hypoglycaemic event) .
The base case analysis used the UKPDS 68 risk equations, but in February 2014, an update to the IQVIA CORE Diabetes Model was released, incorporating data from the UKPDS 82, with these risk equations applied in a scenario analysis. Whilst a validation study of the revised model has been published, the model proprietors suggest that the update is used in a scenario analysis, with the previous version being used in the base case . The base case analysis assumed that HbA1c increased based on the UKPDS progression equation for the duration of the analysis in both arms. An alternative was explored with HbA1c increasing by 0.14% (1.5 mmol/mol) per year in both arms of the analysis while patients received once-weekly semaglutide 1 mg or empagliflozin 25 mg, based on the metformin arm of the ADOPT study . When patients initiated basal insulin, HbA1c followed the UKPDS progression equation, as in the base case.
In the base case analysis, BMI returned to baseline on treatment intensification, and an alternative was explored with BMI returning to baseline followed by a further increase based on the Willis et al. equations . For simplicity, the base case analysis include only one intensification step, and a scenario analysis was conducted with a second intensification to basal bolus insulin when HbA1c exceeded 7.5% (58 mmol/mol) for the second time. A reduction in HbA1c and an increase in BMI were applied, based on the Willis et al.  equations for “insulin experienced” patients. Instead of the treatment switching approach based on an HbA1c threshold applied in the base case analysis, an alternative was explored with treatment switching after 3 years in both arms. In this analysis, the difference in HbA1c was held constant while patients received once-weekly semaglutide 1 mg or empagliflozin 25 mg, with HbA1c brought to 7% (32 mmol/mol) on treatment intensification to abolish the difference. This approach is consistent with previously published cost-effectiveness analyses of once-weekly semaglutide .
Generic basal insulin glargine was used as the intensification treatment in the base case analysis, as it is the most commonly used generic analogue insulin. Scenario analyses were performed with the costs of insulin Semglee® (the lowest cost analogue insulin available in the UK), and insulin Lantus® (the most commonly used branded insulin analogue) applied to evaluate the impact of intensification with less or more costly basal insulins.
The base case analysis was performed using a second-order Monte Carlo approach, with sampling around inputs to capture both first- and second-order uncertainty. A scenario analysis was performed using a first-order Monte Carlo approach, with no sampling around baseline characteristics, treatment effects, costs of complications utilities or transition probabilities.
Compliance with Ethics Guidelines
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.