The progression of Type 2 diabetes is driven by progressive β-cell dysfunction and increased insulin resistance, which results in hypoglycaemia due to difficulty of achieving glycaemic control. Typically, lifestyle modifications such as diet and exercise fail to achieve and give way to the administration of oral hypoglycaemic agents (OHAs) in order to maintain glucose control. In addition to tolerability issues for patients, the inability of OHAs to stem the decline in β-cell function [1] commonly lead to the introduction of exogenous basal insulin to maintain normoglycaemia [2]. Traditionally regarded as a drastic measure in Type 2 diabetes, physicians are increasingly favouring earlier introduction of basal insulin to control hyperglycaemia and minimise the associated micro- and macrovascular complications of diabetes [3, 4]. Whilst undoubtedly clinically effective, use of insulin regimens also carries some problems, namely:

Hypoglycaemia of any severity has a profound effect on patients' quality of life [7] and is regarded as the single greatest obstacle to achieving normoglycaemia [10]. In addition to reduced quality of life, hypoglycaemia results in substantial direct medical cost and lost productivity [11].

Insulin glargine (Lantus™) is an analogue of human insulin with a prolonged duration of action and once-daily dosing. In Type 2 diabetes, the principal emergent benefit is significantly reduced risk of all forms of hypoglycaemia over Neutral Protamine Hagedorn (NPH) [12]. However, initiation of insulin glargine still requires careful dose titration to an appropriate level over a period of time. This is essential for successful treatment of diabetes and the avoidance of hypoglycaemia [13]. However, recent trial evidence has suggested that insulin glargine could be introduced earlier to achieve glycaemic goals [14] and a further study showed that adding insulin glargine to OHA therapy had a positive effect on treatment satisfaction and quality of life (QoL) without complaints related to hypoglycaemia [15]. Insulin glargine is currently not recommended by the National Institute for Health and Clinical Excellence (NICE) for routine use for people with Type 2 diabetes, but can be considered for people with Type 2 diabetes who require assistance from a third party to administer their insulin injections, who have recurrent symptomatic hypoglycaemic episodes or who would otherwise need twice-daily basal insulin injections in combination with oral antidiabetic drugs. Using their own model (based on United Kingdom Prospective Diabetes Study (UKPDS) 68 [16]), NICE concluded that human insulin analogues are the most cost-effective option and glargine was estimated to be cost-effective in Type 2 diabetes patients at increased risk of hypoglycaemia [17]; the NICE guidelines are due to be reviewed and republished in early 2009.

The incretin mimetics are emerging as a significant new class of hypoglycaemic agents that regulate glucose homeostasis similarly to endogenous glucagon-like peptide-1 (GLP-1). Exenatide (Byetta™), the first such compound to be licensed by the European Medicine Agency (EMEA) [18], enhances glucose-dependent insulin secretion, regulates glucagon release, and delays gastric emptying thereby reducing hyperglycaemia in a similar manner to GLP-1. Unlike the short-lived GLP-1, exenatide resists metabolism by dipeptidyl peptidase-IV giving it a pharmacokinetic profile suitable for chronic administration [19]. Exenatide is recommended for use in the treatment of Type 2 diabetes in combination with metformin, and/or sulphonylureas in patients who have not achieved adequate glycaemic control on maximally tolerated doses of these oral therapies and is given twice daily by subcutaneous injection. Exenatide was approved by the Scottish Medicines Consortium (SMC) in June 2007, as an adjunct therapy to patients with Type 2 diabetes who, despite currently taking metformin and/or a sulphonylurea either singly or in combination, fail to achieve adequate glycaemic control [20]. Exenatide, however, is not currently recommended by NICE for routine use for people with Type 2 diabetes [17] and should only be considered for people with Type 2 diabetes, who have a body mass index (BMI) over 35 kg/m², specific problems arising from high weight, inadequate blood glucose control (glycosylated haemoglobin (HbA_1c) >7.5 %), prescribed conventional oral therapy and where other high cost therapies would otherwise be commenced. Using their own model (based on UKPDS 68 [16]), NICE estimated that exenatide was not cost-effective in any scenario [17] and, furthermore, exenatide has been linked with occurrences of acute pancreatitis [21].

The only published head-to-head randomised controlled trial (RCT) [22] of exenatide versus glargine, reported similar improvements in the overall glycaemic control in patients with Type 2 diabetes sub-optimally controlled with combination OHAs at maximal doses. For secondary endpoints, patients treated with exenatide experienced significantly fewer nocturnal hypoglycaemic events (0.9 vs. 2.4 events per patient per year; a 63% reduction), but higher incidence of daytime hypoglycaemia (6.6 vs. 3.9 events per patient per year; a 69% increase) and achieved significant weight loss from baseline (Δ -2.3 kg), whereas insulin glargine-treated subjects gained weight of a similar magnitude (Δ +1.8 kg) by the study endpoint. The principal drawback reported in exenatide treated patients was nausea (in 57.1% of subjects) and vomiting (in 17.4%), which led to withdrawal of almost 10% of exenatide-treated patients from the study protocol. This finding is consistent with evidence from other published phase 3 placebo-controlled trials [23]. By comparison, less than 1% of those exposed to insulin glargine withdrew due to adverse events. The most common adverse event reported with insulin glargine was nasopharyngitis (in 9.0%). Within this trial setting, however, the dosage of glargine was very low with an average of 25 international units (IU)/day. This does not reflect clinical practise, where the dosage is typically 42 IU/KG [24]. A higher dose with glargine will lead to better glycaemic control, but also more weight gain and will have an effect on hypoglycaemic events [25, 26]. Regarding patient reported outcomes, Boye and colleagues published the patient reported outcomes measured from the same head-to-head trial [22] and showed no significant difference in improvement of health utility or quality of life [27].

The increasing global financial burden of chronic health conditions is fuelling a rising demand for value for money when introducing new health technologies. In line with many other countries, in the United Kingdom (UK), NICE currently accepts as cost-effective, "those interventions with an incremental cost-effectiveness ratio of less than £20,000 per QALY (quality adjusted life year) and that there should be increasingly strong reasons for accepting as cost effective interventions with an incremental cost-effectiveness ratio of over £30,000 per QALY [28]." A recently published study of Ray et al. [29] showed exenatide to have a cost-effectiveness ratio of £22,420 vs. insulin glargine; however, a major limitation of this study was that it used an estimated UK cost of exenatide based on the US price. The study reported exenatide as dominant (less costly and more effective) when marketed at 20% of the US price in the UK. The study of Ray et al. used cohort details from the Heine trial [22] that were then applied to a previously published cost effectiveness model [30], the CORE Model.

The objective of this analysis was to prospectively evaluate the cost-effectiveness of exenatide from a UK NHS perspective, using insulin glargine as a comparator and the current UK NHS price for exenatide and also applying the cohort profiles and published data from the Heine trial [22]. The model, which has been previously published [31], used to evaluate the cost-effectiveness was a discrete event simulation (DES) model for patients with Type 2 diabetes using UKPDS-derived risk functions [16] and a multivariate regression model for the utility associated with hypoglycaemia [7].

For this analysis, a previously published UK diabetes Type 2 model was used to evaluate the introduction of exenatide in patients with sub-optimally controlled Type 2 diabetes to the UK NHS. As a comparator, insulin glargine, a long-acting once daily human insulin analogue, was chosen. The results of the study reflect the long-term projection of the findings of the Heine trial [22]. Using the baseline profiles and treatment effects outlined here the economic evaluation conducted found that insulin glargine dominated exenatide.

A recently published study of Ray et al. [29] evaluating the cost-effectiveness of exenatide vs. insulin glargine using the same input from the Heine trial [22] showed different results to our findings. Ray et al. demonstrated that exenatide has a cost-effectiveness ratio of £22,420 vs. insulin glargine at an estimated 100% US price. Furthermore, it was estimated that exenatide is dominant (cost and QALY saving), when marketed at 20% of the US price in the UK. A cost-effectiveness ratio below the recommended £20,000/QALY [28] was achieved at 80% of the US price, which approximately equals the current UK market price. Using the same Heine trial data [22] as utilised in this study, Ray et al. [29] estimated that exenatide is a cost-effective treatment option in patients with sub-optimally controlled Type 2 diabetes in the UK. Ray et al. [29] modelled their baseline scenario (100% US price) over a time horizon of 35 years and reported higher direct medical costs for exenatide and insulin glargine £29,401 and £19,489 per patient respectively) and higher QALYs (14.62 and 14.51 per patient respectively) than we derived in our analyses for exenatide and insulin glargine (cost: (£14,568 and £9,280 per patient respectively; QALYs: 7.68 and 7.86 per patient respectively) using a time horizon of 40 years. Both studies showed that lifetime medical costs are higher for patients with exenatide and insulin glargine and Ray et al.'s [29] estimate was much higher over a shorter time period than ours.

The main difference between the outcomes reported by Ray et al. [29] and our own study were the QALY estimates with Ray et al. [29] estimating a QALY gain for patients treated with exenatide, while our analysis showed a QALY gain for patients treated with insulin glargine. Reviewing the input factors for the model of Ray et al. [29] showed that the authors used more evidence from the Heine trial [22] in favour of exenatide than we used in our study. While there is clear evidence for a difference in hypoglycaemic events, weight loss and patients with nausea, there was no clear evidence of any significant difference in other factors favouring exenatide. Ray et al. [29] included lower systolic blood pressure (SBP), lower total cholesterol (TC), lower low-density lipoprotein (LDL) and lower triglycerides in patients treated with exenatide. These risk factors are influential predictors of future diabetes-related complications and improvements reported in the Heine trial [22] were non-significant. We did not see this as evidence for a significant treatment effect and therefore did not include these in our model. As outlined in the methods section, we did not include disutility associated with weight gain, gastrointestinal side effects or insulin initiation. The justification for this approach being the analysis of patient reported outcomes in a head to head comparison of exenatide and glargine. Furthermore, while cross sectional studies have demonstrated a disutility of insulin use compared to oral agents [52], prospective longitudinal studies have shown that insulin therapy does not impact on quality of life [53].

The major drawback of the Heine trial [22] was that the dosing of glargine did not reflect clinical practice. Within the trial design, the dosing of glargine was very low, with an average of 25 IU/day. A higher dose of glargine naturally leads to improved glycaemic control, but also more weight gain and, potentially, hypoglycaemic events. Our sensitivity analyses showed that applying long term data to the exenatide arm and a more clinically relevant dosing scenario for glargine produced results consistent with the base case findings.

As with all modelling simulations, our economic evaluation has a number of strengths and weaknesses. The model can only be as reliable as the predictive algorithms and other data that were used in its construction. The model was based on contemporary data largely drawn from the same source – the United Kingdom Prospective Diabetes Study; however, the dependence of our findings on the UKPDS model is not without drawbacks. In particular, the UKPDS outcomes model does not completely capture the complexities of all diabetes related complications. For example, complications such as peripheral neuropathy, ulceration, gross proteinuria or proliferative retinopathy and macular oedema are handled crudely, primarily due to insufficient patient numbers experiencing these events in the UKPDS study. This is largely due to the inclusion criteria of the UKPDS study in which patients were newly diagnosed with Type 2 diabetes with no pre-existing cardiovascular complications. Type 2 diabetes, however, is primarily characterised by macrovascular complications and the equations published from this study represent a robust approach to describing the relationship between the major complications occurring in Type 2 diabetes and modifiable risk factors. No modelled economic evaluation can ever provide point estimates that are demonstrably precise; nevertheless, the fact that exenatide is significantly more costly than insulin glargine, but provides only marginal benefits with respect to BMI provides some degree of confidence that the conclusions presented here are reliable.

This study strictly applied those statistically significant differences between exenatide and insulin glargine emergent reported by Heine et al [22]. Although not necessarily an accurate prediction of the performance of exenatide in day-to-day practice, it is, nevertheless, consistent with NICE economic modelling guidance [28], and therefore appropriate. The study did not include utility estimates regarding weight changes and adverse events due to treatment. Published utility data related to weight changes are not well reported, but preliminary analyses of HODaR data [40], including longitudinal weight and utility data suggested that utility changes start to become significant if a patient gains/loses a great amount of weight (>5 kg) and it is dependent on the initial weight (the higher the BMI the more likely the patient benefit from weight loss). Certain other factors not reported by the Heine trial [22] could also have had an impact on these outcomes. For example, what was the utility decrement associated with nausea and vomiting in the exenatide group? Did the significant weight loss among exenatide-treated subjects improve utility, in spite of gastro-intestinal symptoms, and was the weight loss a result of nausea, pancreatic and other gastro oesophageal problems? How did lipid profiles change as a result of weight loss experienced in the exenatide arm? At the time of writing, these data are not available in the public domain but, in their absence, it is reasonable to conservatively assume there were no significant differences between the groups otherwise they would presumably have been reported in the trial-related publication.

This study could have a number of implications for policy makers. Given therapeutic equivalence of exenatide and insulin glargine concerning glycaemic control it is likely that a forthcoming NICE appraisal of exenatide possibly would apply the lower £20,000 threshold. This economic evaluation showed that glargine, when used in the treatment of Type 2 diabetes patients, was dominant in comparison to exenatide. Maintaining the current UK price difference between insulin glargine and exenatide, the recommendation of exenatide should be considered cautiously as a cost-effective alternative to insulin glargine for tertiary treatment of Type 2 diabetes given the evidence present for exenatide.