Cost-effectiveness of dapagliflozin versus DPP-4 inhibitors as an add-on to Metformin in the Treatment of Type 2 Diabetes Mellitus from a UK Healthcare System Perspective

Type 2 diabetes mellitus (T2DM) is a chronic condition characterised by elevated blood glucose levels as a result of resistance to the action of insulin. T2DM can lead to numerous micro- and macro-vascular complications and may cause substantial disability. It is increasingly prevalent, with the T2DM population in the UK expected to rise to 3 million by 2017 [1], and it is currently estimated to account for 7–12 % of the total UK National Health Service (NHS) expenditure [2, 3]. Although drug costs are increasing [1], the greatest component of the economic burden of T2DM is the treatment of diabetic complications [2], which can be reduced with effective management of the disease.

The primary treatment goal of T2DM management is to reduce glycated haemoglobin (HbA1c) levels to below 6.5 % for first line treatment or below 7.5 % for second line treatment. This is recommended in the UK by the National Institute for Health and Care Excellence (NICE) in order to effectively reduce diabetes-related complications [3]. The principles of the NICE guidelines are in line with those outlined in the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) combined position statement, which support a target HbA1c goal for adults with T2DM of around 7 %, depending on individual patient characteristics [4]. However, T2DM represents a major clinical priority, as between 30–40 % of all patients receiving treatment fail to reach the blood glucose targets recommended by NICE and over three-quarters are overweight or obese [4, 5].

Metformin is commonly used as a first-line treatment in diabetes; however, due to the progressive nature of T2DM, many patients at some point will require additional therapy to maintain glycaemic control. The selection of additional treatment options is often complex due to the number of factors that must be considered. Unintended sequelae such as hypoglycaemia, weight changes and side effects are important considerations as they can have a significant impact on patients’ adherence and quality of life [4].

Dapagliflozin was the first in a new class of selective sodium-glucose co-transporter 2 (SGLT2) inhibitors licensed in Europe. Both dapagliflozin and dipeptidyl peptidase-4 inhibitors (DPP-4i) have been recommended by NICE in the UK as second-line therapies (dual therapy, add-on to metformin) in patients with T2DM, when diet and exercise plus metformin fail to achieve glycaemic targets. In order for healthcare decision makers to ensure patients receive the highest standard of care within the available budget, the clinical benefits of each treatment option must be balanced against the economic consequences. This study aimed to assess the long-term cost-effectiveness of dapagliflozin versus DPP-4i, as dual oral therapies in combination with metformin, in patients who were inadequately controlled on metformin alone, from the perspective of the UK NHS. The objective was to present the model here as it was reviewed and accepted by NICE.

In addition to glycaemic control, key factors that may differ across therapies and therefore drive treatment decisions in clinical practice, such as weight and hypoglycaemic risk, were also considered in the analysis. Results of a previously published network meta-analysis (NMA), comparing the major clinical outcomes for dapagliflozin with DPP-4i as an add-on to metformin [6], acted as a key source of clinical inputs for this economic analysis. This reported a non-significant reduction in HbA1c (−0.08 % [95 % CI: −0.25, 0.10]) and a significant reduction in weight (−2.85 kg [95 % CI: −3.39, −2.30]) for dapagliflozin compared with DPP-4i [6]. Assessments of the cost-effectiveness of dapagliflozin versus other antidiabetec agents used as add-ons to metformin [7], for indications other than as an add-on to metformin and in settings other than the UK [8] have been presented elsewhere.

The economic evaluation analysed the cost-effectiveness of dapagliflozin as an add-on to metformin (DAPA + MET) versus DPP-4i as an add-on to metformin (DPP-4i + MET) in adults aged 18 years and older with T2DM who were inadequately controlled on metformin alone. The main assessment metric was the incremental cost-effectiveness ratio for dapagliflozin compared with DPP-4i therapy, with effectiveness measured in quality-adjusted life years (QALYs). QALYs represent a composite measure of estimated post-treatment life years adjusted for the quality of life (or utility) of those life years. The economic evaluation was conducted from the perspective of the UK NHS, and a discount rate of 3.5 % was applied to both costs and health effects as recommended in the NICE Methods Guide [9].

The treatment group, DAPA + MET, was associated with a mean incremental benefit of 0.032 QALYs (95 % CI: −0.022, 0.140) when compared with the DPP-4i + MET control arm (Table 2). This effect is largely explained by differences in patient weight, which has a significant impact on HRQoL [26]. The mean incremental cost was estimated to be £216 (95 % CI: £-258, £795) and was mainly driven by the higher acquisition cost of dapagliflozin (Table 2). An ICER point estimate of £6,761 per QALY gained was calculated.

The results of the univariate SA are presented as tornado graphs (Fig. 1). These highlight the range of both the incremental costs and incremental QALYs for the parameters that most affect these outcomes. The base case value is represented by the central vertical line and the outcome values are plotted for the maximum and minimum values of each selected parameter. As can be seen from Fig. 1, the point estimate for incremental costs was relatively insensitive to the variation applied to the model parameters. Improving the HbA1c lowering effect of dapagliflozin resulted in an incremental cost increase of £165 compared with the base case, due to the increased treatment duration, which eventually led to higher drug acquisition costs for the dapagliflozin strategy. However, due to the incremental gain in QALYs observed (0.06), the overall ICER decreased to £4,140. The point estimate for incremental QALYs was shown to be most sensitive to variation of the treatment effect on body weight of DPP-4i (Fig. 1). When varying the weight effect of DPP-4i between the outer limits of its 95 % CI, the incremental QALYs ranged from 0.015 to 0.169; changing this parameter for dapagliflozin resulted in a QALY range of −0.002 to 0.066.

The assumption around the HbA1c switching threshold was investigated by increasing the value from 8.05 to 8.5 % as studies have shown that many patients will exceed this initial threshold in practice [27, 28]. The resulting ICER decreased from £6,761 to £5,227 per QALY gained. However, by using alternative, lower estimates for BMI utility effect [29], the ICER increased to £12,763 per QALY gained. In both cases, the ICERs still fell below the lower limit of generally accepted ICER values in the UK for diabetes medicines (£20,000 per QALY).

The distribution of the ICER estimates from the PSA shows that in most instances, DAPA + MET is both more effective and more costly than DPP-4i + MET (Fig. 2, top panel). Analysis of the PSA results demonstrated that DAPA + MET had an 85 % probability of being cost-effective compared with the DPP-4i + MET treatment strategy at a willingness-to-pay threshold of £20,000 per QALY gained (Fig. 2, bottom panel).

The economic analysis of DAPA + MET versus DPP-4i + MET, as accepted by NICE in the UK, has shown dapagliflozin to be a cost-effective use of NHS resources and therefore a valuable treatment option for T2DM patients who are inadequately controlled on metformin monotherapy. Results from a previously published NMA showed dapagliflozin added on to metformin resulted in a superior weight reduction outcome compared with DPP-4i [6]. This outcome has been shown to be important to patients in terms of quality of life and was a key driver in the associated gain in QALYs observed with dapagliflozin in this economic evaluation [30]. Ultimately, the results show that the incremental QALYs come at an acceptable incremental cost when employing the commonly accepted payer willingness-to-pay threshold of £20,000 per additional QALY gained employed in the UK by NICE, indicating the cost-effectiveness of dapagliflozin versus DPP-4i as an add-on to metformin for the treatment of T2DM.

The model has previously undergone peer-review [11, 12] and health technology assessments [31] and in terms of structure is very similar to a previously developed and validated cost-effectiveness model [32, 33]. The same model has also been used to compare the cost-effectiveness of DAPA + MET with sulphonylurea + MET; this analysis produced a similar magnitude of costs and QALYs for DAPA + MET as estimated in the current analysis and also concluded that DAPA + MET was a cost-effective use of NHS resources in the UK setting [7]. In addition, the cost-effectiveness of dapagliflozin as an add-on to insulin has been investigated from a Dutch healthcare perspective, where it was also found to have an ICER within acceptable cost-effectiveness limits [8].

Although the model has been validated against UKPDS datasets and extensive scenario and sensitivity analyses have been performed, several assumptions exist which may have an impact on the robustness of the study outcomes. Firstly, the model does not include less severe health states, such as microalbuminuria and foot ulcers; however, the utility and cost impact of such health states would be expected to be minimal and therefore have a negligible impact on the final ICER. Secondly, the model treats the patients as a cohort with mean baseline values and mean treatment estimates, as data availability did not allow for a more sensitive approach. Heterogeneity in the population was tested in the model through the sensitivity analysis, but future modelling could simulate individual patients with different characteristics and link these to treatment effect if clinical trial data allowed. Thirdly, the assumption that the mean HbA1c at baseline is a valid representative of the switching threshold does not take into account the potentially skewed nature of this parameter. Unfortunately, data was not available to allow a normality test; however, the switching threshold has been tested in one-way sensitivity analysis and was found not to change the conclusion over the cost-effectiveness of DAPA + MET over DPP-4i + MET.

In terms of limitations in the clinical inputs, the main one was the lack of an RCT directly comparing DAPA + MET with DPP-4i + MET. In the absence of such a head-to-head trial, an NMA was conducted using Bayesian methodology, the methods and limitations of which have been previously discussed in the publication of Goring et al. [6] The uncertainty around the efficacy of dapagliflozin was investigated through one-way sensitivity analysis and showed that as the efficacy of dapagliflozin on lowering HbA1c levels was increased, the total costs were also increased due to the longer treatment period estimated. The additional QALYs gained, however, mean that the resultant ICER still fell as would be expected. Additionally, there was a lack of available long-term data regarding the effect of dapagliflozin or DPP-4i as an add-on to metformin on the development of diabetes-related micro- and macro-vascular complications; it was assumed instead that valid lifetime predictions of events can be made using the UKPDS 68 risk equations [13]. These risk equations have been widely used by researchers modelling diabetes treatments [10, 11, 33] and although they are not without limitations [34], there are no other sources for risk prediction that have been based on such a large number of T2DM patients. The UKPDS risk equations are also derived from a study of over 5,000 UK patients, making them highly applicable to the perspective of the current analysis. Since this analysis for NICE was performed, more recent UKPDS risk equations than the UKPDS 68 ones have been made available [35], but their use in health economic models has not yet been validated by health technology assessment agencies. Therefore we decided to maintain the use of the UKPDS 68 equations in the model, as these have been reviewed and accepted by NICE during the appraisal of dapagliflozin [36]. We also acknowledge that an alternative source of utility values has been published [37]; however, we think it is important that the model that was reviewed by NICE is published.

Assumptions to extrapolate HbA1c beyond trial outcomes were designed to best represent the progressive nature of T2DM in clinical practice, and the time paths were shown to be in line with those reported in the UKPDS 68 study [13]. Weight change, which was associated with CV risk and a decreased HRQoL whilst on treatment, was extrapolated beyond the 2-year trial data and was shown to be a key risk factor in determining QALYs during the sensitivity analysis. However, as mentioned previously, a conservative approach was adopted when extrapolating these data, and as such, any beneficial effect associated with weight loss was likely underestimated. Recently published long-term data reporting sustained weight effect over 4 years confirms the conservative nature of this assumption [18].

Key uncertainties within the economic evaluation arose around the BMI utilities, as there is some uncertainty over the precise relationship between change in BMI and disutility in T2DM. The values used to estimate the impact of increasing/decreasing BMI on utilities were seen to vary in the literature, although sensitivity analyses showed that this did not impact on the cost-effectiveness of dapagliflozin. The outcomes from the NMA, which included international studies, can be considered to be generalisable to patients in a UK setting as the populations defined were representative of the treatment indication in the UK and the average baseline demographics were similar to the UK T2DM population recruited for the UKPDS studies [13]. Additionally, cost and resource use was derived from UK sources, and utility data were largely sourced from the UKPDS studies. The results may also be of interest to other countries where DPP-4 inhibitors are commonly used in clinical practice as add-ons to metformin, especially as the treatment effect data are sourced from international trials. However, available treatments and advised strategies may differ between countries, and factors such as the risk equations, utilities and costs may be subject to change. As such, extrapolation of these results to countries outside of the UK should be performed with caution and local adaptation of the economic model would be advised.

Dapagliflozin represents the first-in-class selective SGLT2 inhibitor licensed in Europe, and has shown in clinical trials to have an effect on HbA1c comparable with existing treatments and a superior outcome in terms of weight reduction [38]. This analysis confirmed that DAPA + MET, in comparison to DPP-4i + MET, is cost-effective within acceptable UK thresholds for the treatment of patients with T2DM.

Supporting data used for the development of the model is available on request.