The cost–effectiveness of oral semaglutide versus empagliflozin in Type 2 diabetes in Denmark

Since there is no head-to-head long-term (or short term) clinical trial evaluating hard end points of the two interventions, the IQVIA Core Diabetes Model (CDM) version 9.5 (using the UKPDS 82 cardiovascular risk equations) was used to simulate the clinical and economic results.

The CDM is a non product-specific microsimulation tool that models the effect of glucose monitoring, diabetes therapies and treatment strategies on disease progression and outcomes. Disease progression is based on a series of interdependent Markov submodels that simulate progression of disease-related complications (angina, myocardial infarction, congestive heart failure, stroke, peripheral vascular disease, diabetic retinopathy, macular edema, cataract, hypoglycemia, ketoacidosis, nephropathy and end stage renal disease, neuropathy, foot ulcer, amputation) and non specific mortality. Each submodel uses time-, state- and diabetes type-dependent probabilities derived from published sources. The use of tracker variables bypasses the memoryless properties of standard Markov models. The model facilitates interconnectivity and interaction between the modeled complications, representing the complex and varied sequelae of the disease. The structure of the CDM and the most recent validation are described in detail elsewhere [14,15]. More information on CDM version 9.5 is available online (www.core-diabetes.com/).

The CDM estimated the cost and quality-adjusted life-years (QALYs) with respect to oral semaglutide plus metformin versus empagliflozin plus metformin in persons with T2D uncontrolled on metformin alone. The primary outcome of the model was the incremental cost–effectiveness ratio (ICER). A Danish health sector perspective was used for a long-term (50 years) time horizon for the base case analysis. Future costs, LYs and QALYs were discounted with a rate of 4% as recommended by the Danish guidelines for health economic evaluation of pharmaceuticals [16]. All prices are stated in Danish Krone (DKK) price-level 2020 excluding value-added tax (ex. VAT). Main results are also presented in EUROs (DKK 744 = €100).

The long-term projection of surrogate end points showed that in the base case analysis over a lifetime horizon (50 years), CDM projected 13.21 and 13.21 life-years, 8.78 and 8.75 QALY and DKK447, 633 and DKK387, 786 total costs, respectively, generating an ICER of DKK 1,930,548 for oral semaglutide plus metformin versus empagliflozin plus metformin (Table 2). Patients switched to next-line therapy after 3 years in the oral semaglutide arm and after 2 years in the empagliflozin arm.

Assuming a willingness-to-pay threshold (WTP) of DKK 357,100 per QALY gained (one-time Gross Domestic Product per capita, 2020), ICER for oral semaglutide plus metformin was above the WTP threshold.

The results of the probabilistic sensitivity analysis illustrated that in 16% of the simulations, oral semaglutide plus metformin was a cost-effective therapy at this WTP threshold (Figures 1 & 2).

Over a 5-year time horizon an ICER of DKK612,931 per QALY (€82,383 per QALY) was generated, also above the WTP threshold and as such not cost effective (Table 2).

In general, the performed scenario-analyses confirmed that findings were robust (Supplementary Table 11). However, in the hypothetical three-line scenario analysis where oral semaglutide and empagliflozin were discontinued in all patients after they reached an HbA1c level of 7.5% again, oral semaglutide reached cost–effectiveness.

The cost–effectiveness result was driven by a major difference in treatment costs (Supplementary Table 12), reflecting the large unit cost difference of oral semaglutide versus empagliflozin (Figure 3).

Overall, this study finds that oral semaglutide is not a cost-effective treatment option compared with empagliflozin in the management of T2D patients uncontrolled on metformin alone, both in a lifelong (50 years) and 5-year time horizon. Results showed only a marginal QALY gain with oral semaglutide, but considerable cost savings associated with the use of empagliflozin.

Three cost–effectiveness analyses of oral semaglutide versus empagliflozin in patients uncontrolled on metformin have been published recently and with conflicting results.

In an analysis from the independent research done by the Institute for Clinical and Economic Review, cost–effectiveness in the US setting was assessed based on network meta-analysis of available clinical trials including PIONEER 2 and cardiovascular outcomes trials. Institute for Clinical and Economic Review concluded that oral semaglutide was not cost effective versus empagliflozin in the US setting based on estimated net prices [23].

In the UK analysis based on PIONEER 2, it was concluded that oral semaglutide was cost effective versus empagliflozin on top of metformin in a UK setting [19]. This 50-year analysis included the cost of oral semaglutide and empagliflozin for a period of only 3 and 2 years respectively, followed by discontinuation of the two medications under study and initiation of insulin for 47 and 48 years respectively. It also applies the more controversial ‘trial product estimand’ (per protocol) approach rather than the intention-to-treat principle. Under these circumstances the authors conclude that oral semaglutide is cost effective [19].

In the UK analysis (which was adapted to a Danish setting in our study), comparing empagliflozin with oral semaglutide based on intention-to-treat results from PIONEER 2 and keeping the medications under study throughout the 50-year life cycle. Empagliflozin was cost effective versus oral semaglutide, and dominant (better health outcomes with lower costs) when effects on heart failure from a real world study in low-risk patients was taken into account [17]. In a three-line sensitivity analysis, this study also evaluated the effect of discontinuing oral semaglutide or empagliflozin and metformin and going to higher doses of long-acting insulin alone as third-line therapy when patients reached the prespecified HbA1c level of 7.5% again (similar to our three-line sensitivity analysis). This happened after 6 and 5 years respectively and empagliflozin was also cost effective compared with oral semaglutide in this three-line scenario.

It is important to acknowledge that this study has some limitations. First of all, the long-term cost–effectiveness result was based on a 52-week open label head-to-head comparison of surrogate end points for oral semaglutide plus metformin compared with empagliflozin plus metformin. No clinical evidence exists to assess the relative long term benefits of the two interventions, in particular, no evidence exists to compare their relative effects on cardiovascular and renal outcomes. In the PIONEER 2 study, oral semaglutide had a larger glucose reducing effect compared with empagliflozin. The study; however, had weaknesses in terms of a relatively large drop out in the semaglutide arm compared with the empagliflozin arm and the study was unblinded [13]. Since the PIONEER 2 trial is currently the only head-to-head evidence available, it is used as the basis for this cost–effectiveness analysis. Also, in the PIONEER 2 trial two types of analyses were performed. One based on the treatment policy estimand, comparable to intention to treat and as such preferred in health economic analyses, and one called trial product estimand, which is a per protocol analyses. Whereas all the necessary inputs for the trial product estimand were provided in the cost–effectiveness analysis published by Bain et al. [19], unfortunately not all the inputs on the treatment policy estimand were reported in Rodbard et al. [13] what could be considered a limitation of this analyses. Nevertheless, the most important was that HbA1c and BMI change to baseline were reported and as such have been used in the base case analysis. Analyses were run using both estimands anyhow. In our study, the estimated ICER using the trial product estimand (per protocol approach) from the PIONEER 2 study reduced from DKK 1,930,548 to DKK 1,124,537 (Supplementary Table 11). This is still considerably higher than the assumed WTP threshold of DKK357,100; hence, oral semaglutide could not be considered cost effective from a per protocol approach either in our study.

In the model, it was assumed that patients switched treatment when they reached a HbA1c level of 7.5% and as such discontinuation due to for example adverse events is not taken into account. It should be noted that the use of rescue medication and other diabetes medications was not significantly different between the two arms in the PIONEER 2 study. Discontinuation in the oral semaglutide arm was; however, significantly higher than in the empagliflozin arm (10.7 vs 4.4%) [13]; and therefore, adoption of a per protocol analysis (‘trial product estimand’) will run the risk of over-estimating the effects of oral semaglutide. On the other hand, drug cost may be overestimated as not all patients have the full length of therapy in both arms, but more in the oral semaglutide arm and in both estimands analyses. Note, however, that the annual cost of oral semaglutide is more than DKK5,000 higher than the cost of empagliflozin (DKK9,326 vs DKK4,165). Even reducing these costs with 10.7 and 4.4% respectively, it will not be sufficient to bridge the cost difference between the two arms.

The HbA1c progression equation applied was based on the UKPDS study, and as such is not specific for the progression under therapy with oral semaglutide and empagliflozin. Long-term studies that are not allowing rescue therapy options are not available so the real natural progression under those drugs is not known and whether one risk equation is better than the other is also unknown. The UKPDS progression equation is very often used in health economic models in Type 2 diabetes and was also applied by Bain et al. [19] in their analysis. If the progression goes slower, it will take longer to reach the 7.5% and as such switch is made later. This could affect results, mainly in the 3L approach as therapy with oral semaglutide or empagliflozin is stopped in third line but this third line will be reached later, which will result in higher drug costs, and maybe some unknown differences in outcomes. A scenario analysis with a higher switch threshold revealed that patients go to next line 2 years later compared with the base case, with a very similar ICER.

Different from Bain et al. [19] we assumed that after treatment escalation empagliflozin and oral semaglutide were continued, not only on a cost side, but also on effectiveness side. We did this to be in line with the clinical guidelines stating that both drug classes are best continued even if escalation is needed [5–7]. Nevertheless, we also included a 3L scenario where oral semaglutide and empagliflozin were discontinued when again a threshold of 7.5% in HbA1c was reached. This happened after 6 and 5 years, respectively. In this scenario oral semaglutide reached cost–effectiveness (Supplementary Table 11) pointing out that the duration of therapy with oral semaglutide is the key driver of the analyses.

The assumption; however, that patients as third-line therapy would discontinue oral semaglutide or empagliflozin treatment and switch to higher doses of long-acting insulin (instead of adding higher doses of long-acting insulin) would not be in line with current clinical practice in Denmark. Neither would it be in line with recommendations from clinical guidelines which do not recommend discontinuation of previously initiated medications [5–7]. Also, clinical guidelines recommend continuation of treatment with SGLT2 inhibitor or GLP-1 receptor agonist regardless of HbA1c level in persons who have developed CVD, heart failure or renal disease over the cause of this 50-year period [6,7].

Recent guidance from the Danish Health Authorities recommends SGLT2 inhibitors over GLP-1 receptor agonists based on clinical efficacy, safety and price, in patients with T2D in general and in subgroups of patients with either atherosclerotic CVD, heart failure or diabetic nephropathy and/or for whom weight loss is crucial [24]. Effectively, this covers the vast majority of persons with T2D.

In this light, real-life use and costs of the SGLT2 inhibitors and GLP-1 receptor agonists could be an important future research topic.

In Denmark, 75% of 240,000 persons with T2D [1] are estimated to have no CVD [25], equaling 180,000 individuals. Most patients with T2D will progress to second-line treatment. In year 2015, 6343 out of 180,742 persons in Denmark with T2D treated with metformin (3.5%) initiated second-line treatment such as DPP-4 inhibitor, SGLT2 inhibitor, GLP-1 receptor agonist, SU, insulin or other treatment [26]. The potential budget impact of second-line choice among patients with T2D uncontrolled on metformin is; therefore, significant and detailed budget impact analysis incorporating all dynamics of the market is warranted.

This study suggests that oral semaglutide plus metformin is not cost effective in Denmark at current price levels compared with empagliflozin plus metformin in people with T2D uncontrolled on metformin. Key drivers of the results are the drug acquisition cost and the duration of therapy. The better impact of oral semaglutide on HbA1c leads to a modestly increased life expectancy and QALY.

Direct clinical long-term comparisons between SGLT2 inhibitors and GLP-1 receptor agonists could be an important future research topic and real-life data could shed more light on the above predictions.

LH Ehlers, M Lamotte, MC Ramos, S Sandgaard, P Holmgaard, EC Frary and N Ejskjaer conceptualized the study. LH Ehlers, M Lamotte, MC Ramos, S Sandgaard and P Holmgaard performed calculations. LH Ehlers wrote the first draft. M Lamotte, MC Ramos, S Sandgaard, P Holmgaard, EC Frary and N Ejskjaer provided inputs too and reviewed first draft.

LH Ehlers is employed by Aalborg University, Aalborg, Denmark. His contributions to the study were supported by an honorarium funded by Boehringer Ingelheim. M Lamotte and MC Ramos are employees of IQVIA, which received consulting fees from Boehringer Ingelheim for adapting the Core Diabetes Model and running the analyses. S Sandgaard, P Holmgaard and EC Frary are employees of Boehringer Ingelheim Denmark A/S. N Ejskjaer received honorarium from Boehringer Ingelheim for his contributions to this study. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

The authors would like to thank C Kumar and A Singhal from IQVIA, Real World Solutions, India for providing writing and editing assistance, funded by Boehringer Ingelheim.

This work is licensed under the Attribution-NonCommercial-NoDerivatives 4.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/