Cost-effectiveness of saxagliptin vs glimepiride as a second-line therapy added to metformin in Type 2 diabetes in China

Abstract

Objective:

This study aims to estimate the long-term cost-effectiveness of saxagliptin + metformin (SAXA + MET) vs glimepiride + metformin (GLI + MET) in patients with Type 2 diabetes mellitus (T2DM) inadequately controlled with MET in China.

Methods:

The Cardiff Model was used to simulate disease progression and estimate the long-term effect of treatments on patients. Systematic literature reviews and hospital surveys were conducted to obtain patients profiles, clinical data, and costs. Health insurance costs (2014¥) were estimated over a 40-year period. One-way and probabilistic sensitivity analyses were performed.

Results:

SAXA + MET had lower predicted incidences of cardiovascular and hypoglycemia events and a decreased total cost compared with GLI + MET (¥241,072,807 vs ¥285,455,177). There were increased numbers of quality-adjusted life-years (QALYs; 1.01/patient) and life-years (Lys; 0.03/patient) gained with SAXA + MET compared with GLI + MET, and the incremental cost of SAXA + MET vs GLI + MET (−¥44,382) resulted in −¥43,883/QALY and −¥1,710,926/LY gained with SAXA + MET. Sensitivity analyses confirmed that the results were robust.

Conclusion:

In patients with T2DM in China, SAXA + MET was more cost-effective and was well tolerated with fewer adverse effects (AEs) compared with GLI + MET. As a second-line therapy for T2DM, SAXA may address some of the unmet medical needs attributable to AEs in the treatment of T2DM.

Keywords::

• Cost-effectiveness
• Saxagliptin
• Sulfonylurea
• Glimepiride
• Type 2 diabetes
• China
• Cardiff model

Introduction

Type 2 diabetes mellitus (T2DM) is a chronic, progressive disease characterized by insulin resistance and relative insulin deficiency, either or both of which may be present at the time diabetes is diagnosed; while its etiology and pathogenesis is still not clear at present1^,2. T2DM may remain undetected for many years, and the diagnosis can occur at any age, which is often made when a complication occurs or a routine urine or blood glucose test is done1. It is a condition that necessitates critical management of blood glucose levels to prevent hyperglycemia, diabetes-related complications, and disability. The world’s prevalence of diabetes is high, with a total of 387 million patients in 2014, and the number is increasing in every country3. The prevalence is also rapidly rising in China4^,5, affecting ∼113.9 million (11.6%) adults aged 18 years or older in 20106; T2DM accounted for at least 90% of these cases1. In addition, blood glucose control remains elusive for many patients with diabetes in China, where only 39.7% of patients have adequate blood glucose control (glycated hemoglobin, [HbA_1c] < 7.0%)6. T2DM imposes a considerable economic burden on patients and their families, healthcare systems, and national economies7^,8. In China, the direct medical costs of T2DM and its complications were estimated to be $26.0 billion in 2007, and by 2030 may be as high as $47.2 billion8.

Given the central importance of long-term blood glucose control in managing risk factors and enhancing health-related quality-of-life (HRQoL) for patients with T2DM9–11, a variety of anti-diabetic agents, such as metformin (MET), α-glucosidase inhibitors, sulfonylureas (SUs), dipeptidyl peptidase-4 (DPP-4) inhibitors, thiazolidinediones (TZDs), glucagon-like peptide-1 (GLP-1) mimetics, and insulin, have been introduced into clinical practice and incorporated into the treatment algorithm2 in China to improve glycemic control in patients with T2DM. Among them, sulfonylureas, a second-line therapy recommended by the Chinese Diabetes Society (CDS) 2013 T2DM Clinical Guidelines2, have been commonly accepted and widely used in China for patients with T2DM who are inadequately controlled with MET alone12.

Glimepiride (GLI) is a third-generation sulfonylureas anti-diabetic drug. It stimulates insulin release through the closure of adenosine triphosphate-sensitive potassium channels on pancreatic beta cells to increase insulin levels and further lower blood glucose in patients with T2DM13^,14. Its clinical efficacy (glucose-lowering properties) as monotherapy or in combination treatments has been well assessed and demonstrated14; however, glimepiride as one of SUs is also associated with significant adverse effects, such as weight gain and an increased risk of hypoglycemic episodes15–18, which may negatively affect patient HRQoL and treatment adherence19–23 and subsequently increase the economic burden imposed on patients with T2DM and their families24–26. Therefore, there is an urgent demand for effective oral hypoglycemic agents with beneficial hypoglycemia and weight profiles.

Saxagliptin (SAXA), a potent, reversible, and competitive DPP-4 inhibitor, acts by preventing the inactivation of the incretin hormones GLP-1 and glucose-dependent insulinotropic polypeptide, and thereby increases GLP-1 levels, stimulates insulin secretion, and decreases postprandial glucagon and glucose levels27. Studies have demonstrated that SAXA as monotherapy and in combination therapy regimens can effectively lower HbA_1c levels and fasting plasma and postprandial glucose concentrations in patients with T2DM, with good safety and tolerability profiles across all drug doses28^,29. In addition, saxagliptin has demonstrated a non-inferior treatment effect with the significant advantage of lowering HbA_1c without increasing hypoglycemic episodes and weight gain as compared with SUs and other insulin secretagogues27^,30–32. Another second-line option for patients with T2DM recommended by the CDS 2013 T2DM Clinical Guidelines, SAXA has been gaining increasingly more attention in clinical practice in China in recent years33.

A patient-centric treatment strategy for patients with T2DM should be determined by the patient’s characteristics and related complications, and optimized for efficacy, tolerability, safety, and treatment costs15^,34. Because hypoglycemia and weight gain resulting from T2DM treatment may increase the risk of cardiovascular disease2^,4, as well as treatment costs, a therapeutic regimen with beneficial hypoglycemia and weight profiles, which provides long-term blood glucose control, like SAXA, may represent an advancement in treatment and cost-effectiveness compared with SUs35–37. However, to our knowledge, there are no direct head-to-head studies comparing both the long-term benefits and costs of SAXA with GLI as second-line treatment added to MET in a Chinese population. Therefore, the aim of this study was to estimate the long-term outcomes and cost-effectiveness of treatment with the DPP-4 inhibitor SAXA as second-line therapy in combination with MET compared with a sulfonylureas GLI plus MET, after failure of MET as monotherapy in China, from the perspective of payers (health insurance).

Methods

Cost-effectiveness model

This study used a previously published long-term simulation model, the Cardiff Diabetes Model38–41, which is designed specifically to assess therapeutic regimens in diabetes and has been validated against other available T2DM models42. It is capable of running in two key modes: deterministic (mean values) and probabilistic sensitivity analysis (PSA). Underlying each mode is a Monte Carlo simulation that simulates the progress of each patient within a cohort over a maximum of 40 years. It also allows users to perform univariate sensitivity analyses in a succinct fashion by using its inside Tornado model, which provides the ability to set alternative input values for certain parameters and perform multiple simulation scenarios in order to provide results relating to each choice of sensitivity analysis. The Cardiff model utilizes published United Kingdom Prospective Diabetes Study (UKPDS) 68 or UKPDS 82 risk equations to estimate the risk of clinical end-points associated with T2DM43^,44. We adopted UKPDS 68 equations to simulate disease progression and update model parameters in this study, because they have been widely used in diabetes modeling and extensively validated compared with UKPDS 82 equations35–37^,45. The simulation cohort consisted of 1000 patients over a period of 40 years (a lifetime horizon because the baseline age is 48.51 years)46; patients discontinued the simulation only in the event of death or when the time horizon was reached. The outputs for the model include the incidence of microvascular (e.g., amputation, blindness, end-stage renal disease) and macrovascular (e.g., congestive heart failure, myocardial infarction, stroke, and ischemic heart disease) events, hypoglycemia, diabetes-specific mortality, and all-cause mortality. The costs and quality-adjusted life-years (QALYs) associated with each treatment strategy were calculated from a payer’s perspective, using only direct medical costs. The annual discount rates for costs and benefits were set to be 3.0%, based on the World Health Organization (WHO) guideline47. The final results from the model are expressed as incremental cost-effectiveness ratios (ICERs), which is calculated as the incremental cost (cost in treatment group minus costs in control group) divided by the incremental QALYs (QALYs in treatment group minus QALYs in control group).

Based on the UKPDS series studies, the Cardiff Diabetes Model used time-dependent risk factor profiles to determine the risk of diabetes-related complications. The model differentiates between symptomatic and severe hypoglycemic events (severe hypoglycemia is defined as a major symptomatic event requiring medical assistance owing to severe impairment in consciousness and is associated with healthcare costs)48–50. Because hypoglycemia is a transient event that is relatively infrequent in patients with T2DM treated with oral anti-hyperglycemic agents, fear of hypoglycemia can be a crucial driver of disutility, with a strong relevance to severe episodes22^,51. Another characteristic of this model is the use of weight gain associated with diabetes treatment, which may have a negative effect on patients HRQoL and their long-term risk of cardiovascular complications24, and may, furthermore, lead to increased healthcare costs22^,52^,53.

Data required in the model

Patient profiles, the clinical efficacy of each treatment, costs related to treatment strategies and diabetes-related events, and utilities related to events are included in the model. We conducted a literature review and hospital survey to obtain the required data.

Literature review and hospital survey

A systematic literature search was carried out in both English (PubMed and Web of Science) and Chinese (China National Knowledge Infrastructure, Wanfang Data, and Chongqing VIP) language databases to identify relevant randomized controlled trials or observational studies providing direct head-to-head comparisons of the treatment efficacy of SAXA + MET and GLI + MET in Chinese patients with T2DM (18 years or older). The search terms included saxagliptin, Onglyza, glimepiride, Amaryl, type 2 diabetes mellitus, non-insulin-dependent diabetes mellitus, Chinese, and China. The keywords were combined and adapted to search the above-mentioned databases for publications dating from 1995–2014 to ensure a comprehensive search.

The included studies had HbA_1c as the primary outcome indicator, with a follow-up duration (up to 12 weeks) appropriate for each comparison. Other than the target treatments, only diet and/or physical exercise were allowed for the study patients. After removal of duplicated publications, we retrieved 1218 and 56 studies from the Chinese and English language databases, respectively, on the topic of SAXA or GLI in T2DM treatment; however, after title and abstract screening and further examination of the full-text articles, only one Chinese-language clinical study was identified32 that provided a direct head-to-head comparison between SAXA and GLI as add-on therapy to MET in a randomized controlled trial of patients with T2DM. This study selected 75 Chinese patients from a hospital between September 2012 and May 2013. The patients were divided into two groups: treatment group (37 cases) given SAXA and MET, and control group (38 cases) given GLI and MET for 24 weeks. The changes of Fasting Plasma Glucose (FPG), HbA_1c, body mass index (BMI), Insulin, and pro-insulin-to-insulin ratio levels were observed before and after treatment. The adverse events were investigated during the treatment.

For the hospital survey, cost data for diabetes-related events occurring between 2010–2014, were collected from two representative hospitals in eastern China (one secondary hospital and one tertiary hospital). These data were a summation of direct medical costs, such as physician visits, drug and laboratory tests, hospital beds, examinations, drugs, operations, and nursing. Costs collected from both hospitals were synthesized to form a combined cost profile; however, because these costs were not comprehensive owing to the unavailability of outpatient data, we used them only as an alternative cost profile in sensitivity analyses to test and corroborate the data collected from the published literature.

Patient profile and treatment strategies

The patients in this analysis were characterized by the mean cohort values for demographics (e.g., age, percentage female) and risk factors (e.g., HbA_1c level, total cholesterol) reported by Zhu and Song32 in a head-to-head study conducted in China (Table 1). Because only BMI rather than body weight or height was reported in this head-to-head study, the baseline weight was calculated from the baseline BMI by assuming that the average height in the investigated patient populations was 1.63 m54. For other data that were not available in the head-to-head study, such as the duration of diabetes, proportion of smokers, and high-density lipoprotein cholesterol levels, we used data from cross-sectional, multi-center, observational survey studies with nationally representative populations of patients with T2DM in China6^,55.

Table 1. Demographics and risk factors.

Variable*	Value†		Reference
	Mean	SE
Baseline demographics
Baseline age, years	48.51	1.21	Zhu and Song32
Proportion female, value: 0–1	0.48	0.06	Zhu and Song32
Duration of diabetes, years	6.7	0.07	Ji et al.55
Height, m	1.63	0	General Administration of Sport of China54
Proportion of smokers, value: 0–1	0.183	0.0046	Ji et al.55
Modifiable risk factors
HbA1c, %	8.70	0.09	Zhu and Song32
Total-cholesterol, mmol/L	5.13	0.14	Zhu and Song32
HDL cholesterol, mmol/L	1.11	0.0013	Xu et al.6
SBP, mmHg	130.99	2.26	Zhu and Song32
Weight, kg‡	71.34	2.15	Zhu and Song32

BMI, body mass index; HbA_1c, glycated hemoglobin; HDL, high-density lipoprotein; SBP, systolic blood pressure; SE, standard error.

*Most variables are taken from the Zhu and Song32 study. If these were not available in Zhu and Song, they have been obtained from a published observational study.

†Values are mean unless stated otherwise.

‡Weight was calculated from baseline BMI by assuming that the average height was 1.63 m; weight = BMI × (height)2.

Type 2 diabetes mellitus is a progressive disease. Consequently, even in patients receiving anti-diabetic treatment, HbA_1c levels increase over time, resulting in a positive slope of the blood glucose curve. The shape of blood glucose profile determines when the therapy escalation occurs because the treatment effect of T2DM is primarily driven by blood glucose control. The treatment strategy considered in this analysis was that patients inadequately controlled on MET alone started with the combination therapy of SAXA + MET (treatment group) or GLI + MET (control group), and that therapy escalations occurred when the pre-specified HbA_1c threshold of 8.0% was reached.

Clinical data

Clinical results and costs were estimated based on the following treatment strategies in the head-to-head study reported by Zhu and Song32: the SAXA dose was 5 mg/day, GLI was titrated from 1–4 mg/day as needed (the mean dosage was 2.8 mg/day), and MET was used in open-label individual doses.

The primary treatment effect parameters considered in this analysis were changes in HbA_1c level and body weight and the occurrence of hypoglycemia (Table 2). No treatment-associated effect on blood pressure or lipid levels was assumed because it had not been reported in the head-to-head study32. SAXA + MET was associated with weight loss, whereas GLI + MET was associated with weight gain in the head-to-head study. Change in body weight was calculated from change of BMI, by assuming that average height in investigated patient populations was 1.63 m54.

Table 2. Clinical input variables.

Variable*	SAXA + MET	GLI + MET	Reference
HbA1c reduction, year 1, %	−1.40 (0.15)	−1.20 (0.13)	Zhu and Song32
Weight change, kg	−2.39 (3.06)	0.53 (3.33)	Zhu and Song32
Number symptomatic hypoglycemia	0.0264 (0.0264†)	0.1544 (0.0586†)	Zhu and Song32; Zheng et al.50
Probability severe hypoglycemia	0.0006 (0.0040†)	0.0034 (0.0095†)	Zhu and Song32; Zheng et al.50

GLI, glimepiride; HbA_1c, glycated hemoglobin; MET, metformin; SAXA, saxagliptin; SE, standard error.

*Values are mean (SE; SE for values sampled from a normal distribution in the probabilistic sensitivity analysis). Variables are taken from the Zhu and Song32 study.

†Calculated as √rate (1 − rate)/numbers of subjects.

Hypoglycemia reported by Zhu and Song32 was not clearly differentiated between symptomatic and severe episodes. Because patients are more closely monitored and may be better educated about hypoglycemia in clinical trials compared with the usual care setting, they may report few or no occurrences of severe hypoglycemic events. In the usual care setting, however, patients may be more likely to experience severe hypoglycemia. Therefore, we modified the data from the head-to-head study by referring to an observational survey study50 in which the percentage of severe cases was 2.18% of all hypoglycemic events.

The HbA_1c and body weight profiles changed over time, and the effects of the treatments applied at the relevant thresholds are presented in Figures 1 and 2. Weight increases were estimated to be 0.1 kg/year for the control and treatment cohorts; this estimate was based on UKPDS data after accounting for the initial treatment-related weight change56.

Figure 1. Simulated progression of HbA_1c in the treatment (saxagliptin + metformin) and control (glimepiride + metformin) arms over the modeled time horizon. HbA_1c = glycated hemoglobin.

Figure 2. Simulated progression of weight in the treatment (saxagliptin + metformin) and control (glimepiride + metformin) arms over the modeled time horizon.

Costs

In this analysis, only country-specific direct medical costs were estimated. All costs were converted into 2014 Chinese Yuan (¥) using the Chinese Consumer Price Index (CPI)57. The exchange rate of Chinese Yuan/US dollar was 6.143 in 201458. Costs related to macrovascular and microvascular events were classified as fatal or non-fatal and applied to the year in which the event occurred. Maintenance costs for individuals who survived were applied in all subsequent years until the end of the simulation horizon or until the patient died. The costs were primarily derived from published literature in China59. Because the costs for ulcers were not available in the literature, we estimated these costs by synthesizing data from the hospital survey and published studies (Table 3)60^,61.

Table 3. Annual direct medical costs for diabetes-related complications and severe hypoglycemia (2014¥).

Event	Fatal		Non-fatal		Maintenance		Reference
	Mean	SE	Mean	SE	Mean	SE
Ischemic heart disease	–	0	39,041.39	0	6969.85	0	Gao et al.59
Myocardial infarction	46,547.02	0	46,547.02	0	10,692.45	0	Gao et al.59
Congestive heart failure	15,479.64	0	15,479.64	0	9409.36	0	Gao et al.59
Stroke	14,059.41	0	18,141.13	0	8169.26	0	Gao et al.59
Blind	–	–	12,047.60	0	9297.78	0	Gao et al.59
End-stage renal disease	–	–	114,640.49	0	91,981.79	0	Gao et al.59
Amputation	18,232.95	0	18,232.95	0	14,533.60	0	Gao et al.59
Ulcer*	0	0	13,989.07	443.2	4923.52	0	Survey; Li et al.60; Palmer et al.61
Severe hypoglycemia			3829.96				Zheng et al.50

SE, standard error.

*Ulcer, non-fatal costs, were obtained from hospital survey; maintenance cost obtained from two published studies (Li et al.60 and Palmer et al.61).

Only the costs of severe hypoglycemic events were estimated in this analysis, because mild-to-moderate events, like symptomatic hypoglycemia, do not require medical assistance48^,49. This cost was based on direct medical costs for episodes of hypoglycemia in China50, and the annual cost of severe hypoglycemia was assumed to be ¥3829.96 for each patient (Table 3).

For drug cost acquisition, the highest retail price according to the latest official drug price list of the Price Bureau in eastern China62^,63 was considered. Because MET was used in open-label single doses that could not be averaged in the head-to-head study, we cited the average annual drug cost of MET according to a published study (Table 4)64.

Table 4. Annual treatment costs for different drugs.

Drug (Brand)	Manufacturer	Specification	Highest retail price, ¥	Daily dose, mg/day	Annual treatment cost, ¥	Annual metformin cost, ¥	Total cost, ¥
Saxagliptin (Onglyza)	Bristol-Myers Squibb	5 mg*7 tablets	69.65	5	3631.75*	1577.28‡	5209.03
Glimepiride (Amaryl)	Sanofi	2 mg*15 tablets	77.90	2.8	2653.79†	1577.28‡	4231.07

*Official drug price list of Price Bureau for saxagliptin in eastern China62.

†Official drug price list of Price Bureau for glimepiride in eastern China63.

‡Obtained from Hou et al.64. The cost for metformin is ¥366.9 per 12 weeks, and thus the cost of metformin = 366.9 * 4 = 1467.6. Convert to 2014 using CPI from 2013 to 2014, annual cost of metformin = 1467.6*102.6/100*104.75/100 = ¥1577.28.

The BMI-related costs which are relate to increased prescribing costs per BMI unit were calculated and estimated based on a follow-up observation study in China65, and they were assumed to be identical between men and women (Table 5).

Table 5. BMI-related costs.*†.

BMI	Annual cost (2014¥)
20	0
21	0
22	0
23	0
24	1519.5
25	3742.7
26	5965.9
27	8189
28	10,412.2
29	12,635.4
30	14,858.6
31	17,081.7
32	19,304.9
33	21,528.1
34	23,751.2
35	25,974.4
36	28,197.6
37	30,420.8
38	32,643.9
39	34,867.1
40+	37,090.3

BMI, body mass index.

*Obtained from Guo et al.65

†Assumptions: The starting point BMI = 25, BMI-related costs per month = ¥246.8; the slope (cost per month/BMI) = ¥146.6 in 2007. For BMI ≤ 23 the cost was set to zero.

Utilities

Utilities related to complications, hypoglycemia, and weight change are presented in Table 6. Given the absence of utility data in China, we used the utilities from the UKPDS 6266, with the exception of end-stage renal disease (ESRD) and blindness67, hypoglycemia19, and per unit change in BMI53.

Table 6. Utility decrements.

Event disutilities	Utility decrement
	First year	Subsequent year	Reference
Ischemic heart disease	0.090	0.090	Clarke et al.66
Myocardial infarction	0.055	0.055	Clarke et al.66
Congestive heart failure	0.108	0.108	Clarke et al.66
Stroke	0.164	0.164	Clarke et al.66
Blind	0.074	0.074	Currie et al.67
End-stage renal disease	0.263	0.263	Currie et al.67
Amputation	0.280	0.280	Clarke et al.66
Ulcer	0.059	0.059	Clarke et al.66
Hypoglycemia
Symptomatic	0.0142	0.000	Currie et al.19
Severe	0.047	0.000	Currie et al.19
BMI-related changes
Per unit decrease in BMI	0.0171	0.0171	Lane et al.53
Per unit increase in BMI	0.0472	0.0472	Lane et al.53

BMI, body mass index.

Sensitivity analyses

In sensitivity analysis, we first ran the Tornado model for univariate sensitivity analysis in order to investigate key model drivers and parameters among input parameters, including baseline demographics (e.g., HbA_1c, age, total cholesterol), costs and utilities related to diabetes-related events and BMI changes, etc. Then, a series of 1-way sensitivity analyses were conducted to further examine the effects of these key model drivers on ICER. The combined effect of variability around model inputs was assessed using a probabilistic sensitivity analysis (PSA) performed as a second-order Monte-Carlo simulation in the Cardiff diabetes model. Treatment-related HbA_1c effects, weight changes, and symptomatic hypoglycemic events are sampled from a normal distribution, while the probability of severe hypoglycemic events and utility decrements are sampled from a beta distribution, and costs from a gamma distribution. The range of values is expressed by the 95% confidence intervals for each parameter. All of the sensitivity analyses were performed for 1000 patients. The PSA was used to generate a scatter plot of the ICER and a cost-effectiveness acceptability curve (CEAC).

Results

Predicted health events

In the base-case analysis, predicted incidences of macrovascular and microvascular events (except nephropathy) were lower in the SAXA + MET vs GLI + MET group. Similarly, the model also predicted fewer deaths caused by macrovascular events in SAXA + MET. Over a patient’s lifetime, the main differences between the SAXA + MET and GLI + MET groups were the adverse effect profiles. Patients in the SAXA + MET group experienced fewer hypoglycemic symptomatic (11,557 vs 13,981) and severe (411 vs 498) events than those in the GLI + MET group. Furthermore, patients in the SAXA + MET group demonstrated weight loss, whereas those in the GLI + MET group experienced an increase in body weight (Table 7).

Table 7. Results for SAXA + MET vs GLI + MET.

Total events predicted	GLI + MET (Control)		SAXA + MET (Treatment)		Difference	Total costs,* ¥	GLI + MET (Control)	SAXA + MET (Treatment)
	Non-fatal	Fatal	Non-fatal	Fatal
Macrovascular						Macrovascular
IHD	117.4	0.0	116.3	0.0	−1.0	IHD	8,705,008	8,678,793
MI	108.7	119.1	108.8	116.2	−2.8	MI	14,630,033	14,596,760
CHF	42.8	3.9	42.2	3.8	−0.7	CHF	2,542,758	2,517,399
Stroke	43.6	10.6	43.2	10.2	−0.8	Stroke	2,743,846	2,742,735
Microvascular						Microvascular
Blindness	43.7	0.0	42.5	0.0	−1.2	Blindness	2,983,962	2,939,578
Nephropathy	30.3	2.8	30.5	2.7	0.1	Nephropathy	13,824,176	14,013,694
Amputation	49.3	4.6	46.6	4.3	−3.0	Amputation	3,847,717	3,608,470
Fatal						Hypoglycemia	1,333,739	1,021,374
Macrovascular	–	133.6	–	130.2	−3.4	Treatment	63,754,426	68,357,021
Microvascular	–	7.4	–	7.0	−0.3	BMI costs	171,089,513	122,596,983
Other	–	857.1	–	860.8	3.7	Total	285,455,177	241,072,807
Cost-Effectiveness						Hypoglycemia†
Per Patient
Discounted Cost	285,455.18		241,072.81		−44,382	Symptomatic	13,981	11,557
Discounted QALYs	13.00		14.01		1.01	Severe	498	411
Discounted LY	16.64		16.67		0.03
Cost per QALY					−43,883
Cost per LY					−1,710,926

BMI, body mass index; CHF, congestive heart failure; GLI, glimepiride; IHD, ischemic heart disease; LY, life-year; MET, metformin; MI, myocardial infarction; QALY, quality-adjusted life-year; SAXA, saxagliptin.

*Treatment cost includes insulin therapy and rescue therapy with insulin drug costs. Analysis based on 1000 patients.

†Hypoglycemia in both the treatment and the control group includes hypoglycemia events generated by insulin therapy and rescue therapy.

Predicted costs

Compared with GLI + MET, SAXA + MET was associated with lower total costs. Although the cost for nephropathy was higher with SAXA + MET than GLI + MET, this difference might not significantly affect the total costs of long-term diabetes treatments because the costs for most diabetes-related events (except nephropathy) and hypoglycemia episodes were all lower with SAXA + MET. The major differences between SAXA + MET and GLI + MET were the target drug treatment costs and BMI-related costs, which had a notable effect on the total cost. The drug treatment costs were higher in SAXA + MET (¥68,357,021) than GLI + MET (¥63,754,426); however, this difference was immediately offset by a much bigger difference in BMI-related costs between SAXA + MET (¥122,596,983) and GLI + MET (¥171,089,513) (Table 7).

Incremental cost-effectiveness ratio

The QALY gained with SAXA + MET compared with GLI + MET was 1.01 per patient, and the life-years (LYs) gained was 0.03 per patient. Thus, the main benefits of SAXA + MET was derived from better QALY as well as added LYs. Moreover, there was greater cost savings with SAXA + MET in T2DM treatment than with GLI + MET. The mean incremental cost for SAXA + MET compared with GLI + MET was −¥44,382, and the mean cost savings per QALY gained was ¥43,883 and per LY gained was ¥1,710,926, which indicated that SAXA + MET would lead to better utility and decreased costs for patients. In general, the results showed that SAXA + MET exceeded GLI + MET in cost savings per QALY and per LY (Table 7).

Parameters influencing the incremental cost-effectiveness ratio

In the sensitivity analyses, results from tornado model indicated that the key model drivers were associated to utility, HbA_1c, BMI, and age (Figure 3).

Figure 3. Tornado diagram of the univariate sensitivity analysis. BMI, body mass index; HbA_1c, glycated hemoglobin; HDL, high-density lipoprotein; ICER, incremental cost effectiveness ratio; SBP, systolic blood pressure.

Most of the results of the 1-way sensitivity analyses (except one scenario considering BMI cost to be 0) showed that SAXA + MET was superior to GLI + MET in both cost savings and benefit; the cost savings per QALY gained with SAXA + MET ranged from −¥127,225 to −¥7,923 (Table 8).

Table 8. Sensitivity analyses for SAXA + MET vs GLI + MET (results per patient).

Sensitivity analysis*	Total cost difference, ¥	QALYs gained	Cost/QALY, ¥
Base case	−44,382	1.01	−43,883
1-way sensitivity analysis
Baseline age 60.64 years (48.51·125%)	−33,390	0.79	−42,083
Baseline age 72.77 years (48.51·150%)	−21,972	0.53	−41,657
Baseline HbA1c decreased by 1% (7.7%)	−41,429	0.73	−56,940
GLI + MET HbA1c reduction as SAXA + MET	−40,670	0.74	−54,620
All hypoglycemia assumed to be symptomatic events	−44,380	1.01	−43,882
GLI + MET no. symptomatic hypoglycemia as SAXA + MET	−44,382	1.01	−43,961
GLI + MET prob. severe hypoglycemia as SAXA + MET	−44,372	1.01	−43,878
Discount rate (costs and benefits) 5%	−36,154	0.84	−43,257
Utility decrement of hypoglycemia halved (symptomatic, 0.0071; severe, 0.0235)	−44,382	0.99	−44,667
Utility decrement of hypoglycemia reduced by 75% (symptomatic, 0.00355;  severe, 0.01175)	−44,382	0.98	−45,070
Utility weight 0.014 per unit BMI decrease and −0.014 per unit BMI increased in all years	−44,382	0.38	−117,391
Utility decrement (first and subsequent years) per unit BMI gain halved  (−0.0472·0.5 = −0.0236)	−44,382	0.57	−77,906
Utility decrement (first and subsequent years) per unit BMI gain reduced by 75%  (−0.0472·0.25 = −0.0118)	−44,382	0.35	−127,225
GLI + MET annual therapy costs halved (2115.54)	−42,269	1.01	−41,794
Severe hypoglycemia cost halved (1914.98)	−44,226	1.01	−43,729
BMI costs halved	−20,136	1.01	−19,910
BMI costs reduced by 75%	−8013	1.01	−7923
BMI costs = 0	4110	1.01	4064
Alternative diabetes-related complications costs used	−44,498	1.01	−43,998
Probabilistic sensitivity analysis	−41,265	0.80	−51,588

BMI, body mass index; GLI, glimepiride; HbA_1c, glycated hemoglobin; MET, metformin; QALY, quality-adjusted life-year; SAXA, saxagliptin.

*Analysis for 1000 patients. Everything else is as described for the base-case analysis.

The highest absolute value of the cost per QALY (−¥127,225) for SAXA + MET vs GLI + MET was attained under the extreme assumption that the utility decrement (first and subsequent years) per unit BMI gain was reduced by 75%, which provided a 190% increase in cost savings per QALY gained by SAXA + MET compared with the base case (−¥43,883). In the scenario in which the utility decrement per unit BMI gain was halved, the resulting cost savings per QALY by SAXA + MET remained high at −¥77,906 (an absolute increase of 78%). When adopting an alternative weight utility associated with body weight changes68, in which the absolute value of the weight utility related to the gain or loss per unit BMI was set to be identical (both were 0.014), SAXA + MET again yielded a dominant cost savings per QALY, increasing QALYs and reducing costs (Table 8).

Although a decrease in BMI-related costs would have a negative effect on cost savings per QALY gained by SAXA + MET, the results still favored SAXA + MET vs GLI + MET. When BMI-related costs decreased by half, the magnitude of the cost savings per QALY gained by SAXA + MET decreased by ∼55% (−¥19,910 vs −¥43,883); when reduced by 75%, the magnitude decreased by 82% (−¥7923 vs −¥43,883). Only when BMI-related costs were not taken into consideration, there presented a reverse picture where the cost per QALY gained with SAXA + MET was ¥4064 compared with GLI + MET; however, it was still within the acceptable range of gross domestic product (GDP) per capita of China (Table 8). The GDP per capita in China was ¥43,320 in 201369.

Although hypoglycemia is associated with utility decrement and, for severe hypoglycemia, a related cost50, the adjustment of hypoglycemia events and the utility associated with them in alternative scenarios resulted in only a negligible effect on the magnitude of cost savings by SAXA + MET, indicating that episodes of hypoglycemia might not be an important factor in our analysis. Both HbA_1c and age were identified as important factors by the tornado model; however, after reducing the base-case HbA_1c level by 1%, assuming that the reduction of HbA_1c in GLI + MET was equal to that in SAXA + MET, or increasing the base-case age by 25% or 50%, the conclusion remained unchanged (Table 8).

In the PSA, SAXA + MET generated a mean QALY gained of 0.80 and an incremental cost of −¥41,265 over a patient’s lifetime compared with the QALY gained and incremental cost associated with GLI + MET, which was a little lower than that determined in the base case (1.01 and −¥44,382, respectively). However, the overall cost savings per QALY for SAXA + MET vs GLI + MET was slightly higher in the PSA (−¥51,588 vs −¥43,883; Table 8).

SAXA + MET provides an additional benefit as well as lower costs compared with GLI + MET. Most simulations were located in the southeast quadrant of the ICER scatter plot figure, which means SAXA + MET gained more benefits and took less costs than GLI + MET. SAXA + MET was cost-effective in 74.3% of the simulations by a threshold value of GDP per capita (Figures 4 and 5).

Figure 4. Scatter plot of incremental cost-effectiveness ratios for saxagliptin + metformin vs glimepiride + metformin with a CE threshold value of ¥43,320 (GDP per capita in China in 2013). CE, cost effectiveness; GDP, gross domestic product.

Figure 5. Cost-effectiveness acceptability curve for saxagliptin + metformin vs glimepiride + metformin.

Discussion

This study is the first pharmacoeconomic analysis conducted using the Cardiff Diabetes Model to evaluate long-term outcomes and cost-effectiveness of SAXA vs GLI as add-on therapy to MET in a cohort of patients with T2DM in China. We found that the combination therapy of SAXA + MET resulted in better QALY and LY gained, as well as a lower total cost compared with GLI + MET. In addition, the cost-effectiveness results were robust, remaining stable to changes in various assumptions of input parameters.

Although no similar studies have been conducted in China, our findings partially support the results of studies performed in Sweden35, Germany36, and Argentina37, which used a similar Cardiff simulation model and demonstrated that the SAXA + MET treatment strategy can improve both QALYs and LY compared with an SU + MET. Compared with SU + MET, our study has shown that SAXA + MET can reduce total costs; however, in the above-mentioned studies, SAXA + MET was found to potentially increase costs. These discrepancies may be related to the differences in BMI-related costs. Because BMI-related costs were not included in their studies, the large difference in BMI-related costs between SAXA + MET and SU + MET in our study might have contributed to this difference.

A patient-centric treatment strategy for patients with T2DM should be determined based on the patient characteristics, and optimized for efficacy, tolerability, safety, and treatment costs15^,34. The CDS 2013 T2DM Clinical Guidelines2 suggest that, when choosing among different therapies, their advantages and disadvantages should be considered. Our analysis highlights that it is essential to consider not only the effect of a drug on long-term blood glucose, but also its effect on hypoglycemia and body weight in the treatment of T2DM.

Hypoglycemia and fear of hypoglycemia impede the attainment of stringent blood glucose control, either through sub-optimal dosing and/or poor treatment adherence, which increases the risk of macrovascular complications21^,51^,70. Although episodes of hypoglycemia did not affect the outcomes significantly in our analysis, this might be explained by the minor difference in the occurrence of symptomatic and severe episodes between SAXA + MET and GLI + MET treatments, which could result in a similar risk of developing macrovascular complications for both treatments.

Being overweight, or obese, was common among patients with T2DM and was usually associated with hypertension and dyslipidemia, which is significantly associated with an increased risk of developing cardiovascular disease2^,55. SAXA + MET was associated with body weight loss and a low risk of hypoglycemia, whereas GLI + MET was associated with weight gain and a relatively higher risk of hypoglycemia32, which could lead to a major difference in diabetes-related complications as well as the corresponding utility and costs associated with the change in BMI between these two treatments. Therefore, SAXA + MET had a cost-effectiveness advantage compared with GLI + MET. Hence, from the perspective of payers, adding SAXA to MET might be preferable to adding an SU in patients with inadequate blood glucose control on MET alone, especially patients who are overweight or at higher risk of hypoglycemia.

In the context of improving the management of diabetes in China, as well as decreasing the disease burden, the use of treatments like SAXA could provide a benefit by addressing some of the unmet medical needs in T2DM treatment that are attributable to undesirable adverse effects (e.g., hypoglycemia and weight gain) caused by SUs, reduce the long-term disease burden on patients and their families, and lessen health expenditures for diabetes treatments for the healthcare system and government in China.

The present study is limited by the paucity of head-to-head studies; only one short-term head-to-head trial comparing the treatment effect of SAXA + MET vs GLI + MET in China was identified32, causing uncertainty in the input parameters. In addition, as with other Cardiff modeling studies35^,36^,71, our study projects long-term outcomes based on clinical input parameters from short-term trials owing to a shortage of lifetime follow-up data from a well-designed epidemiologic or clinical study. However, the use of UKPDS 68 equations to simulate a range of long-term outcomes provided a disease progression that closely matched that observed in the UKPDS study43, potentially resulting in a more realistic simulation and reducing the bias to a certain extent.

Because this study is from the perspective of payers, only direct medical costs are estimated, although diabetes-related events, weight gain, and hypoglycemia may have a considerable effect on productivity (indirect costs) at the base age of 48.51 years. The costs of diabetes-related complications are mainly drawn from a study59 conducted in China that abstracted cost data from previously published literature providing an economic evaluation of treatments in the Chinese diabetes population. The costs were established in 2010, which might bias and limit the generalizability of the results. Therefore, another set of costs derived from a hospital survey in eastern China in 2014, along with literature review data as a supplement, was investigated in an alternative scenario.

The sensitivity analysis corroborated the robustness of the results obtained in the base case. Although it is more reasonable to apply a country-specific utility for diabetes-related complications, hypoglycemia, and weight change, there are no studies related to utilities covering all events, especially in China. We used utilities from the UKPDS 62 and other published studies of foreign populations in this study, which may have resulted in some bias.

All the concerns above suggest that the results of our analysis should be applied with caution to avoid misleading conclusions rather than to deny their real value for decision-makers and administrators.

Conclusion

The combination of SAXA + MET is a preferred option from the perspective of payers, demonstrating greater effectiveness and lower costs compared with GLI + MET. SAXA is likely to represent a durable, well-tolerated, and cost-effective treatment option for physicians and decision-makers as second-line therapy in combination with MET for T2DM inadequately controlled with MET alone in China, and it may address some of the unmet medical needs in the treatment of T2DM attributable to adverse effects (e.g., hypoglycemia and weight gain) and reduce the economic burden of diabetes.

Transparency

Declaration of funding

This study was sponsored by AstraZeneca.

Declaration of financial/other relationships

SG, JD, LS, YM, and HD have disclosed that they have no significant relationships with or financial interests in any commercial companies related to this study or article.

Acknowledgments

The authors thank Xuemei Zhen and Yuhang Zeng (Zhejiang University School of Public Health, Hangzhou, China) for their help in collecting data.