Cost–effectiveness analysis of empagliflozin compared with glimepiride in patients with Type 2 diabetes in China

The IQVIA Core Diabetes Model (CDM) version 9.0 was used to quantify survival, cardiovascular and renal outcomes treated either with empagliflozin or glimepiride when glycemic control is not achieved with metformin alone in Chinese patients with T2D.

The CDM is a web-based computer microsimulation model that estimates the long-term health outcomes and economic consequences of interventions in patients with Type 1 diabetes and T2D. To project diabetes progression, it uses a set of 17 inter-dependent Markov submodels, and risk factor progression models, together with standard risk equations (REs) and landmark clinical data that estimate the risk of diabetes complications and adverse events [7].

To capture the cardiovascular effects observed in the trial, the CDM was calibrated in a step-wise approach (Table 1). First, the observed annual progression of risk factors (HbA1c, systolic blood pressure [SBP] and BMI) per treatment arm were programmed in the CDM for the follow-up period of the EMPA-REG H2H-SU trial (4 years). These inputs were incorporated into REs to predict the risk of diabetes complications and to define the trajectory within each Markov submodel. Several well-established REs are available in the CDM: the United Kingdom Prospective Diabetes Study (UKPDS) 68 [8], UKPDS 82 [9], Advance [10], Swedish national Diabetes Register [11,12], ARIC study [13], Fremantle [14] and PROCAM [15]. These different REs were tested and the UKPDS 82 equations were found to be the closest fit to predict the 4-year clinical outcomes of the EMPA-REG H2H-SU trial [4,8]. The second step of the calibration was an iterative adjustment process to identify the best set of relative risk (RR) adjustments needed to fine-tune the UKPDS 82 RE predictions, so that predictions would match the observed 4-year clinical outcomes of EMPA-REG H2H-SU. The outcomes used to calibrate the predictions were: primary and subsequent myocardial infarction, primary and subsequent stroke, hospitalization for angina, hospitalization for heart failure, cardiovascular death and noncardiovascular death.

The outcomes of the CDM are life years (LY), quality-adjusted life years (QALY), event rate per 100 patient-years for diabetes-related complications and associated costs as well as incremental costs, incremental LY and QALY, and the incremental cost–effectiveness ratio (ICER). To assess different levels of uncertainty, scenario and sensitivity analyses, as well as a probabilistic sensitivity analysis (PSA) are also to be conducted.

Baseline characteristics of the EMPA-REG H2H-SU trial were programmed in the CDM. The data represented patients with T2D and no established CVD whose glycemia was inadequately controlled with metformin alone. Thirty-three percent of the patients in the EMPA-REG H2H-SU trial were Asian [4]. Patients had a starting age of 56 years, average diabetes duration of 6 years and baseline HbA1c level of 7.92% (Supplementary Table 1). Patients were defined as Asian in the CDM.

This economic evaluation compared empagliflozin with glimepiride as an add on to metformin. When the patients’ HbA1c level exceeded the defined threshold of 8.5%, patients moved to a rescue therapy. This next-line therapy was defined equally in both arms, as a combination of insulin glargine and glimepiride. The efficacy of this treatment combination was informed by a published meta-analysis [16]. In both lines, HbA1c progression evolved over time following the UKPDS 68 progression model. The average daily dose of insulin was not reported per kg and hence was used as such (40.6 IU/day). For glimepiride, the dose was not reported in the meta-analysis and so we assumed 2 mg daily following the defined daily dose recommended by the WHO [17].

The efficacy data of both empagliflozin and glimepiride were extracted from the trial (Table 2) [18]. These effects show the impact of treatments on the evolution of patients’ physiological parameters, which will influence the progression of the disease and the projections on the occurrence of events (i.e., clinical outcomes). After the 4-year time-point, treatment effect on CV and mortality outcomes were maintained until treatment escalation occurred at HbA1c level of 8.5%. After treatment switch, the occurrence of diabetes complications was predicted by the standard UKPDS 82 REs (i.e., without calibration). Hypoglycemia, urinary tract infection (UTI) and genital infection (GI) were considered as adverse events (Table 2). Table 2 also describes the treatment effects of the second-line therapy, insulin glargine and glimepiride [16]. Mortality data were predicted using the UKPDS82 combined mortality approach.

Baseline quality of life utility of Chinese patients with T2D without complications was 0.876 [19] and BMI-related disutility applied per BMI point increase was 0.0061 [20]. For acute events, a disutility was subtracted from the base utility, leading to a final utility score recorded for the patient experiencing the event. In case, a patient suffered different multiple events (e.g., a myocardial infarction [MI] and a stroke), the model followed the minimum approach, recording the utility value of the health state with the lowest individual score, while still applying a disutility related to adverse events. The full list of utilities and disutilities can be retrieved from Supplementary Table 2.

A healthcare system perspective was taken, therefore only direct medical costs were considered. These included medicine acquisition costs, cost of managing complications, adverse events costs and costs related to long-term clinical management of the disease. Treatment acquisition costs were derived from a Chinese big-data service platform for the health industry (Table 3) [21] and from the tendering prices [22], in line with National Development and Reform Commission’s price-difference policy [23]. The daily cost of empagliflozin 25 mg in China was calculated according to the Chinese rules of price difference and based on the tendering price of empagliflozin 10 mg. The information on costs were derived from a recent publication that assessed the burden of T2D in China using the CDM, which provided detailed cost data in the Chinese setting (Supplementary Table 3) [24]. All costs were inflated and reported as 2020 Chinese Yuan (CNY). Additionally, results were also reported in US dollars to allow convenient interpretation by the international health community.

The base case was defined as lifelong analysis. With a baseline age of 56 years, a time horizon of 50 years was considered long enough to have all patients in the death state. An annual discount rate of 5% was applied to both future costs and effects in line with Chinese economic guidelines [26]. To assess the impact of discounting rates on the results, scenarios were run with different discount rates (0, 3 and 8%).

Additional scenarios were run to test the impact of various assumptions on the results. In the first scenario, an HbA1c threshold of 9% was tested in contrast to the 8.5% considered in the base case. Second, the impact of BMI on utility was removed to evaluate the effect of this assumption on the base-case results due to differences in BMI reductions of both arms. Third, the short-term outcomes were assessed by shortening the time horizon to 4 years to align with the duration of the actual clinical trial (EMPA-REG H2H-SU). Fourth, as neutral protamine Hagedorn (NPH) insulin is the most used insulin in China, the costs of glargine in next-line therapy were replaced by NPH insulin costs (daily dosage: 15 IU/day and cost per unit: 0.16 CNY). As NPH insulin is applicable to both arms when first line of treatment fails in maintaining glycemic control, the additional risk of hypoglycemia given by this treatment was ignored to be conservative. Fifth, given that Asians have on average a lower BMI compared with Caucasians, a starting BMI of 24 was considered at baseline instead of 29 in the base case.

A PSA was also conducted to capture parameter uncertainty. This is achieved through Monte Carlo simulations together with nonparametric bootstrapping (using 1000 patients and 1000 iterations), so that the robustness of the cost–effectiveness results could be assessed. The parameters included in the PSA are patient characteristics, treatment efficacy, utility and cost of complications. Log normal distributions and 20% variation were applied to sample the costs of complications. Treatment effects were sampled based on the estimated standard error detailed in Table 2. The utility data were varied according to the variability reported in standard deviations values displayed in Supplementary Table 2. All were sampled following the beta distribution. For sampling patient’ baseline characteristics, truncated Normal distributions were used following the mean and standard error (SE) reported in Supplementary Table 1.

In the long-term, when compared with glimepiride, patients on empagliflozin gained an incremental 0.232 LYs and 0.317 QALYs (Table 5). Furthermore, treatment with empagliflozin costed $1368 (CNY9703) more per patient compared with glimepiride, resulting in an ICER of $4364 (CNY30,648) per QALY gained. The WHO defines the WTP threshold as three-times the national gross domestic product (GDP) per capita [27]. For China, this approximates $30,290 (CNY212,676) [28]. As such, at this threshold empagliflozin is a cost-effective treatment option in patients with T2D, without established CVD and not controlled with metformin alone in China.

Patients receiving empagliflozin stayed on first-line therapy for 9 years, whereas for patients on glimepiride this was 7 years. After that, in both arms, patients received insulin glargine plus glimepiride, however in the glimepiride arm, patients initiated insulin glargine 2 years earlier, which lead to higher hypoglycemia and CVD costs.

Patients receiving empagliflozin showed better survival (24.785 years) than those receiving glimepiride (24.167 years; Table 4). This is driven by the improved cardiovascular and renal event rates of empagliflozin over lifetime compared with glimepiride where the RR were below 1 (0.829–0.961; Table 4). Only the estimated number of stroke events was more favorable for glimepiride compared with empagliflozin (RR = 1.025).

The results of the scenario analyses further anchored the cost–effectiveness profile of empagliflozin observed in the base case ($4364/QALY). Empagliflozin remained cost-effective across all scenarios, with the exception of the shorter-time horizon scenario (Table 5). Applying a higher HbA1c threshold 9% for treatment switch increased the ICER to $8309 (CNY58,346) per QALY gained. When assessing the results without the effect of BMI change on utility, as per Bagust et al., absolute QALYs increased in both arms and the incremental gain declined, leading to an ICER of $5824 (CNY40,890) per QALY gained. Running the analysis over a time horizon of 4 years highlighted the benefit of empagliflozin versus glimepiride as add on to metformin, by reducing the progression of renal disease and the risk of CV events. This scenario showed that, compared with glimepiride plus metformin, patients on empagliflozin plus metformin have lower incidence of congestive heart failure, peripheral vascular disease, angina and MI, as well as of microalbuminuria, gross proteinuria and end-stage renal disease. In this scenario, the ICER was $49,871 (CNY350,161) per QALY gained (Figure 1). By replacing the cost of insulin glargine by NPH insulin, absolute total and treatment costs decreased, nevertheless both incremental cost and ICER slightly increased to $1702 (CNY11,950) and $5376 (CNY37,745), respectively. By replacing the baseline BMI by 24, the average BMI of adult Asians with diabetes, absolute QALYs increased and total costs decreased in both arms. Incremental QALYs and costs decrease, leading to a small rise of the ICER to $5806 (CNY40,767).

Most iterations lie on the first quadrant (North–East) of the scatterplot, showing incremental positive benefits and slightly higher costs (Figure 2). The cost–effectiveness acceptability curve revealed that empagliflozin has a probability of being dominant in 10% of cases in China with a 95% probability of being cost-effective at the defined local WTP threshold of $30,290 (CNY212,676; Figure 3).