Cost‐effectiveness analysis comparing real‐time continuous glucose monitoring with self‐monitoring of blood glucose for patients with type 2 diabetes treated with insulin in Japan

ABSTRACT

Introduction

Real‐time continuous glucose monitoring (rt‐CGM) systems are associated with reductions in glycated hemoglobin (HbA1c) and hypoglycemic events compared with self‐monitoring of blood glucose (SMBG). This analysis aimed to assess the cost‐effectiveness of rt‐CGM compared with SMBG in the management of people with T2D who require insulin treatment in Japan.

Materials and Methods

A lifetime economic evaluation was conducted using the IQVIA Core Diabetes Model, version 10. Effectiveness data were obtained from a randomized controlled trial that showed rt‐CGM reduced HbA1c by 0.9% compared to SMBG. Cohort characteristics and cost data were primarily sourced from Japanese studies. Costs were inflated to 2023 Japanese Yen (JPY) and discounted at 2.0% per annum. The robustness of the base case results was subsequently assessed through sensitivity analyses.

Results

In the base case, there was an incremental gain of 1.501 quality‐adjusted life years (QALYs) favoring rt‐CGM at an incremental cost of JPY 500,192, resulting in an incremental cost‐effectiveness ratio (ICER) of JPY 333,150/QALY gained. This was far below the assumed willingness‐to‐pay (WTP) threshold of JPY 5,000,000/QALY gained. The result was robust in sensitivity analyses, with all scenarios tested also returning ICERs below the WTP threshold. When examining projected clinical outcomes, rt‐CGM was associated with reductions in the incidence of the majority of complications considered.

Conclusions

From a Japanese healthcare perspective, rt‐CGM is highly cost‐effective for the glycemic management of people with T2D who require insulin treatment.

Abbreviations

ACE‐I

angiotensin‐converting enzyme inhibitors

ARB

angiotensin receptor blockers

ARR

absolute risk reduction

BMI

body mass index

CDM

core diabetes model

CGM

continuous glucose monitoring

CI

confidence interval

CV

cardiovascular

DKA

diabetic ketoacidosis

ECG

electrocardiogram

eGFR

estimated glomerular filtration rate

FoH

fear of hypoglycemia

ICER

incremental cost‐effectiveness ratio

PPPM

per patient per month

QALY

quality‐adjusted life year

QoL

quality of life

RR

relative risk

rt‐CGM

real‐time continuous glucose monitoring

SE

standard error

SHE

severe hypoglycemic event

SMBG

self‐monitoring of blood glucose

T2D

type 2 diabetes

UKPDS

United Kingdom Prospective Diabetes Study

WBC

white blood cells

WTP

willingness‐to‐pay

INTRODUCTION

Diabetes presents a large healthcare problem in Japan, with an estimated 10.8 million adults diagnosed with diabetes in the country in 2024 ¹ . Consequently, the costs of managing diabetes pose a significant economic burden for the nation, accounting for JPY1.2 trillion of the JPY 43 trillion national medical care expenditure in 2018 ² . For type 2 diabetes (T2D), an analysis of a large claims database found the excess cost associated with T2D to be USD 123 (JPY 13,029) per patient per month (PPPM) in 2014 ³ . The complications of T2D, in particular, are associated with large economic burdens, with the same study estimating average excess costs of USD 319 (JPY 33,790) PPPM for renal failure, USD 3,677 (JPY 389,481) PPPM for those requiring dialysis, and up to USD 13,280 (JPY 1,406,667) PPPM for ischemic heart disease ³ . Health policies with the potential to reduce the incidence of such complications are therefore of importance ² .

Achieving glycemic control is crucial for avoiding complications associated with T2D, in addition to reducing mortality ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ . Renal complications due to diabetes are a key issue in Japan, with increasing numbers of patients needing dialysis ¹² , ¹³ . Japan currently has the second highest prevalence of dialysis worldwide, and approximately 40% of those who start dialysis have end‐stage renal disease as a complication of diabetes ¹² . Poor glycemic control is a particular contributor to this issue, with a recent study finding that 21% of patients with diabetes in Japan experience rapid declines in renal function, with those with glycated hemoglobin (HbA1c) levels >7.4% at particular risk ¹³ . These issues may be further compounded by limited access to healthcare services for the management of diabetes, with a study in Kyoto finding that only one‐third of patients with diabetes regularly receive HbA1c testing ¹² .

Real‐time continuous glucose monitoring (rt‐CGM) is an important tool to support glycemic control and reduce HbA1c levels among patients with T2D, as has been demonstrated across several studies ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ . Rt‐CGM devices can also provide users with alerts and graphs that can enable better understanding of their glucose levels and help them identify trends ²⁴ , ²⁵ . Despite the clinical benefits, clinicians may face barriers to prescribing continuous glucose monitoring (CGM), with insurance‐related factors and CGM costs being cited as the most frequent barriers ²⁶ . The use of rt‐CGM has been shown to be cost‐effective in other countries for both type 1 and type 2 diabetes ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ . However, the cost‐effectiveness of such devices in Japan is yet to be established. This analysis aimed to evaluate the cost‐effectiveness of rt‐CGM in Japanese people with T2D on insulin therapy.

MATERIALS AND METHODS

Model structure

The IQVIA CORE Diabetes Model (CDM v. 10) ³² , ³³ , ³⁴ was used to assess the cost‐effectiveness of rt‐CGM compared with self‐monitoring of blood glucose (SMBG) for people with T2D on insulin therapy in Japan. The CDM is a generic web‐based simulation model designed to estimate the cost‐effectiveness of diabetes care interventions that can be adapted to different countries and regional care‐specific settings ³³ . Outputs include life expectancy, quality‐adjusted life years (QALYs), direct and indirect costs, cumulative incidence of diabetic complications, time to onset of complications, and incremental cost‐effectiveness ratios (ICERs) per additional life years or QALYs gained ³³ .

The model consists of 17 interdependent Markov sub‐models that interact to predict long‐term health outcomes and costs of diabetes care interventions. It includes equations to predict the progression of specific risk factors over a defined time horizon, such as HbA1c, systolic blood pressure, diastolic blood pressure, total cholesterol, low‐density lipoprotein, high‐density lipoprotein, triglycerides, and estimated glomerular filtration rate (eGFR). These equations are derived from studies such as the Framingham Heart Study, the United Kingdom Prospective Diabetes Study (UKPDS), the Epidemiology of Diabetes Interventions and Complications follow‐up study, among others ³³ , ³⁵ , ³⁶ . The model also includes clinical data to inform the probabilities of the onset of diabetes‐related microvascular complications and event‐specific mortalities (e.g., eye disease, renal disease, ulcer, amputation, neuropathy, cardiovascular [CV] mortality), as well as probabilities of transitioning to different health states and from one complication to a different complication. For T2D, these data are derived from the UKPDS and other sources ³³ .

This study was conducted and reported in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement ³⁷ . A table outlining how the reporting of this study adheres to the CHEERS statement is available in Table S1.

Baseline cohort characteristics

Baseline characteristics were obtained from clinical studies that included Japanese people with T2D ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ . When parameters for additional risk factors were not available from these studies, data from a large retrospective cohort study ¹⁴ and CDM default values were used, which were sourced from the ACCORD trial ⁴⁴ , ⁴⁵ . The mean baseline HbA1c was 8.0% (±0.9%), the mean age was 56.8 (±10.9) years, and the mean duration of diabetes was 8.8 (±6.4) years ³⁸ . The racial distribution was assumed to be 100% Asian. Further details on cohort characteristics are provided in Table S2.

Clinical effectiveness

A HbA1c reduction of 0.90% was assumed for rt‐CGM compared with SMBG based on data from the Steno2Tech randomized controlled trial ²³ . The effect was sustained for 2 years beyond Year 1 in the rt‐CGM arm based on longitudinal studies showing improvements in glycemic outcomes for up to 10 years ¹⁴ , ⁴⁶ , ⁴⁷ . The annual re‐increase of HbA1c in the SMBG arm after Year 1 and in the rt‐CGM arm after Year 3 was 0.15% per year based on the CDM default clinical table.

Severe hypoglycemic event (SHE) rates were obtained from the Kaiser retrospective cohort study ¹⁴ . Event rates of 0 and 4 per 100 patient years were used for rt‐CGM and SMBG, respectively. Severe hyperglycemic events, including diabetic ketoacidosis (DKA) events, were also modeled based on data from the Kaiser retrospective study ¹⁴ . The rates of such events were assumed to be 0 and 2.5 per 100 patient years for rt‐CGM and SMBG, respectively.

Cardiovascular risk equation

Due to differences between Japanese and Western populations in CV risk, a risk‐prediction equation designed for a Japanese population was used. That equation was developed by Yatsuya et al. based on data from the Japan Public Health Center‐based Prospective Study Cohort II ⁴⁸ . In sensitivity analyses, an alternative equation relevant to Western populations (UKPDS 82) was also tested.

Costs

Only direct medical costs were included. The costs of complications, concomitant medications, and screening were obtained from published literature and Japanese government public databases. Costs were inflated to JPY 2023, where applicable, using the health component of the Consumer Price Index in Japan ⁴⁹ . Costs for diabetes complications or related events and their sources are listed in Table S3.

Annual treatment costs for rt‐CGM were based on the fee‐for‐service classification defined by the Japanese Ministry of Health, Labour and Welfare. Under the code C150 for self‐monitoring of blood glucose, subcode 7 (intermittent CGM), the fee calculation is based on 1,250 points per month, where 1 point equals JPY 10. The monthly fee was therefore JPY 12,500, corresponding to an annual fee of JPY 150,000.

The treatment costs for SMBG were calculated based on the average number of tests reported on the Steno2Tech study of 1.6 tests/day, corresponding to 48 tests per month, reimbursed under the code C150‐3 (41–60 tests/day). The resulting costs were JPY 5,800 per month and JPY 69,600 per year ⁵⁰ .

Utilities

For patients without complications, the initial base utility value was based on a meta‐analysis published by Beaudet et al ⁵¹ . Diabetes‐related complications and treatment‐related adverse events utilities/disutilities were sourced from the published literature ⁵¹ , ⁵² . A conversion algorithm from two previous cost‐effectiveness studies was applied to convert European health utility values to Japanese values ³⁹ , ⁵³ . This led to a base utility value for patients with T2D without complications of 0.858. Other utility values used in the model are summarized in Table S4.

A utility benefit of 0.03 was assumed to be experienced by patients in the rt‐CGM group due to the avoidance of daily and frequent fingerstick testing. This value was based on the study reported by Matza et al. ⁵⁴ An additional utility was included to represent a reduction in the fear of hypoglycemia (FoH) experienced by patients in the rt‐CGM arm. This was sourced from the Hypoglycemia Fear Survey results from the DIAMOND trial ⁵⁵ , which were converted into a utility of 0.02536, based on a conversion carried out as reported by Currie et al. ⁵⁶

Discounting, time horizon, and willingness‐to‐pay

A time horizon of 50 years was adopted in addition to a discount rate of 2% per year ⁵⁷ , ⁵⁸ . A commonly cited willingness‐to‐pay (WTP) threshold of JPY 5,000,000/QALY gained was used ⁵⁷ , ⁵⁹ .

Sensitivity analyses

A probabilistic sensitivity analysis (PSA) was conducted to assess uncertainties in the estimated cost‐effectiveness and treatment outcomes. In addition, one‐way sensitivity analyses were performed to evaluate the effects of varying different parameters on the ICER. The utility associated with FoH was reduced by 50%, removed, or changed to 0.0155 based on a secondary analysis of the COACH study ⁶⁰ . For HbA1c, the effect size was varied by ±30% of the base case, in addition to being set to 0.6% based on a random‐effects meta‐analysis of the Steno2Tech and MOBILE studies ²² , ²³ , and being set to 0.56% based on results reported by Karter et al. ¹⁴ The rates of DKA and SHE were reduced by 50% or removed completely from the analysis. A combined reduction or exclusion of SHE rates and FoH utility was also tested. A scenario in which the UKPDS risk equation for CV disease was used was also analyzed. Additional variables examined included time horizon (1, 5, 10, 20, and 30 years), mean age of cohort (35, 45, 65, and 75 years), discount rates (0.0%, 4.0%), body mass index (BMI, 1.1 kg/m² reduction), number of tests (31–40, 61–90, 91–120, and >120 tests per month) and biological sex (100% male or 100% female).

RESULTS

In the base case, rt‐CGM was associated with an incremental gain of 1.501 QALYs compared with SMBG. The difference in total direct medical costs was JPY 500,192. This was driven by a JPY 1,453,545 increase in glucose monitoring costs that was largely offset by savings on renal complications (JPY −442,625), severe hyperglycemic events management (JPY −294,397), and CV complications (JPY −98,651). Overall, this resulted in an ICER of JPY 333,150 per QALY gained, falling below the WTP threshold of JPY 5,000,000/QALY gained. The base case results are further summarized in Table 1.

Table 1.

Summary of base case findings

Base case costs and outcomes	rt‐CGM	SMBG	Difference†
Glucose monitoring costs, JPY	2,650,628	1,197,083	1,453,545
Management (preventative screening, medication) costs, JPY	359,426	346,182	13,244
CV complications, JPY	3,615,800	3,714,451	−98,651
Renal complications, JPY	602,660	1,045,285	−442,625
Ulcer/amputation/neuropathy complications, JPY	1,315,573	1,382,742	−67,169
Ophthalmic complications, JPY	214,743	265,165	−50,422
Severe hypoglycemia (req. medical assistance), JPY	0	13,335	−13,335
Hyperglycemia & DKA, JPY	0	294,397	−294,397
Total direct (medical) cost, JPY	8,758,831	8,258,639	500,192
Life years	17.013	16.529	0.484
QALYs	14.729	13.228	1.501
ICER	333,150 JPY/QALY

Bold values were used to indicate cost savings for rt‐CGM relative to SMBG.

^†Negative difference in costs favors RT‐CGM over SMBG and positive difference favors SMBG. CV, cardiovascular; DKA, diabetic ketoacidosis; ICER, incremental cost‐effectiveness ratio; QALY, quality‐adjusted life year; rt‐CGM, real‐time continuous glucose monitoring; SMBG, self‐monitoring of blood glucose.

Sensitivity analyses

In the PSA, 99.9% of the 1,000 model iterations showed rt‐CGM to improve quality‐adjusted life expectancy versus SMBG, while 40% of model iterations showed rt‐CGM to be cost saving versus SMBG (Figure 1). Using the PSA outcomes to generate a cost‐effectiveness acceptability curve showed that there would be a 97.2% likelihood of rt‐CGM being cost‐effective versus SMBG at a WTP threshold of JPY 5,000,000/QALY gained (Figure 2). The cost‐effectiveness of rt‐CGM was found to be robust in one‐way sensitivity analyses (Figure 3), with the intervention remaining cost‐effective in all scenarios and the majority of ICERs not exceeding 10% of the WTP threshold (JPY 5,000,000/QALY gained, Table 2). Key drivers identified in these analyses included time horizon, SHE and DKA incidence, number of SMBG tests per month, rt‐CGM costs, and HbA1c effect. Interestingly, when the number of SMBG tests was increased to 61–90 per month or higher, rt‐CGM became dominant, with cost savings of JPY 15,792 for 61–90 tests, JPY 717,531 for 91–120 tests, and JPY 1,377,990 for >121 tests. The only scenarios resulting in ICERs exceeding 10% of the WTP threshold were those in which no severe hyper‐ or hypoglycemic events were considered to have occurred in the SMBG arm and the FoH utility for the rt‐CGM arm was removed (ICER JPY 860,000), and when the time horizon was reduced to ≤10 years (10 years JPY 708,266, 5 years JPY 865,958, and 1 year JPY 1,028,438). In the case of time horizon, this was largely driven by reductions in the difference in QALYs between the two interventions as the horizon was decreased.

Figure 1.

Cost‐effectiveness scatterplot from the probabilistic base case analysis. JPY, Japanese Yen; QALE, quality‐adjusted life expectancy; QALYs, quality‐adjusted life years.

Figure 2.

Cost‐effectiveness acceptability curve from the probabilistic base case analysis. JPY, Japanese Yen; QALYs, quality‐adjusted life years.

Figure 3.

One‐way sensitivity analyses ICERs. BMI, body mass index; CV, cardiovascular; DKA, diabetic ketoacidosis; FoH, fear of hypoglycemia; ICER, incremental cost‐effectiveness ratio; JPY, Japanese Yen; SHE, severe hypoglycemic event; SMBG, self‐monitoring of blood glucose; UKPDS, United Kingdom Prospective Diabetes Study.

Table 2.

Results of sensitivity analyses comparing rt‐CGM and SMBG

Analysis	Total costs, JPY			Quality‐adjusted life expectancy, QALYs			ICER, JPY per QALY gained
	rt‐CGM	SMBG	Difference	rt‐CGM	SMBG	Difference
Base case	8,758,831	8,258,639	500,192	14.729	13.228	1.501	333,150
Discount rate
4.0%	6,665,818	6,236,663	429,154	11.844	10.720	1.124	381,912
0.0%	12,162,278	11,537,030	625,248	19.118	16.989	2.129	293,668
Mean age of RT‐CGM initiation
Mean age 35 years	12,638,922	12,358,175	280,747	20.615	18.329	2.285	122,849
Mean age 45 years	10,964,005	10,530,890	433,115	18.266	16.286	1.981	218,679
Mean age 65 years	7,177,200	6,722,394	454,806	11.996	10.828	1.168	389,222
Mean age 75 years	5,400,336	5,050,134	350,202	8.698	7.882	0.816	429,274
HbA1c effect
−0.56% HbA1c treatment effect (Karter et al. 14 )	8,881,143	8,258,639	622,504	14.618	13.228	1.390	447,780
−0.60% HbA1c treatment effect (Lind et al. 23 and Martens et al. 22 random‐effect meta‐analysis)	8,881,984	8,258,639	623,345	14.640	13.228	1.413	441,306
−30% HbA1c treatment effect (−0.63%)	8,860,164	8,258,639	601,525	14.631	13.228	1.403	428,619
+30% HbA1c treatment effect (−1.17%)	8,642,119	8,258,639	383,480	14.814	13.228	1.586	241,730
BMI effect
Inclusion of BMI effect (−1.1 kg/m2, Lind et al. 23 )	8,750,578	8,258,639	491,939	14.727	13.228	1.500	328,047
FoH utility
COACH FoH utility	8,758,831	8,258,639	500,192	14.562	13.228	1.334	374,816
−50% of FoH utility	8,758,831	8,258,639	500,192	14.514	13.228	1.287	388,740
No FoH utility	8,758,831	8,258,639	500,192	14.300	13.228	1.072	466,640
DKA and SHE incidence
−50% DKA and SHE in SMBG arm and −50% FoH utility in rt‐CGM arm	8,758,831	8,139,588	619,242	14.514	13.274	1.241	499,067
No DKA or SHE in SMBG arm and no FoH utility for rt‐CGM	8,758,831	7,937,961	820,870	14.300	13.345	0.955	860,000
Number of SMBG tests/month
31–40 tests (C150‐2; JPY 4,650/month)	8,758,831	8,021,287	737,544	14.729	13.228	1.501	491,238
61–90 tests (C150‐4; JPY 8,300/month)	8,758,831	8,774,623	−15,792	14.729	13.228	1.501	Dominant
91–120 tests (C150‐5; JPY 11,700/month)	8,758,831	9,476,362	−717,531	14.729	13.228	1.501	Dominant
>121 tests (C150‐6; JPY 14,900/month)	8,758,831	10,136,821	−1,377,990	14.729	13.228	1.501	Dominant
Time horizon
1 year	375,192	318,217	56,975	0.864	0.809	0.055	1,028,438
5 years	1,762,055	1,523,137	238,918	3.982	3.706	0.276	865,958
10 years	3,331,253	2,949,498	381,755	7.170	6.631	0.539	708,266
20 years	5,898,357	5,439,615	458,742	11.523	10.560	0.963	476,269
30 years	7,620,714	7,175,535	445,179	13.723	12.454	1.269	350,756
Cardiovascular risk
UKPDS 82	9,167,231	8,726,818	440,413	13.150	11.802	1.348	326,644
Biological sex
100% female	8,846,285	8,269,469	576,816	15.608	14.021	1.587	363,417
100% male	8,706,832	8,230,545	476,287	14.368	12.908	1.460	326,291

BMI, body mass index; DKA, diabetic ketoacidosis; FoH, fear of hypoglycemia; HbA1c, glycated hemoglobin; ICER, incremental cost‐utility ratio; QALY, quality‐adjusted life year; rt‐CGM, real‐time continuous glucose monitoring; SHE, severe hypoglycemic events; SMBG, self‐monitoring of blood glucose; UKPDS, United Kingdom Prospective Diabetes Study.

The intervention remained cost‐effective when the rates of DKA, SHE, and the FoH utility were decreased by 50%, the time horizon was reduced to 20 years, or the FoH utility was removed. The use of a CV risk equation for a Western population (UKPDS 82) resulted in an almost identical ICER (JPY 326,644) to that of the base case, in which a risk equation designed for a Japanese population was used. Similarly, ICERs close to that of the base case were observed when the sex distribution was set to all male or all female, or when a BMI effect was included.

Projected clinical outcomes

When comparing the predicted risks of numerous complications between rt‐CGM and SMBG (Table 3), the risks of all were found to be numerically lower with rt‐CGM. The largest reductions observed were for end‐stage renal disease (relative risk [RR] 0.56, 95% confidence interval [95% CI]: 0.52–0.60), gross proteinuria (RR 0.68, 95% CI: 0.64–0.71), proliferative diabetic retinopathy (RR 0.66, 95% CI: 0.62–0.69), and severe vision loss (RR 0.74, 95% CI: 0.70–0.77). Broadly similar results to those of the base case were found when a risk equation for a Western population (UKPDS) was used, except for nephropathy death, for which a reduction was observed when the Japanese risk equation was used, whereas no change was observed with the UKPDS risk equation (Table S5).

Table 3.

Projected diabetes complications for adult patients with T2D in Japan

Complication	Cumulative incidence ±SE		Relative risk (95% CI)	Number needed to treat
	rt‐CGM	SMBG
Ophthalmic
Background diabetic retinopathy	35.54 ± 0.47	45.23 ± 0.47	0.79 (0.76 to 0.81)	10
Proliferative diabetic retinopathy	13.50 ± 0.30	20.56 ± 0.37	0.66 (0.62 to 0.69)	14
Macular edema	28.40 ± 0.41	37.17 ± 0.44	0.76 (0.74 to 0.79)	11
Severe vision loss	19.55 ± 0.35	26.49 ± 0.41	0.74 (0.70 to 0.77)	14
Cataract	13.96 ± 0.17	16.43 ± 0.18	0.85 (0.82 to 0.88)	40
Cardiovascular
Congestive heart failure death	4.35 ± 0.03	4.43 ± 0.04	0.98 (0.96 to 1.00)	1,250
Congestive heart failure event	4.24 ± 0.07	4.68 ± 0.07	0.91 (0.87 to 0.95)	227
Peripheral vascular disease onset	21.27 ± 0.20	24.60 ± 0.19	0.86 (0.84 to 0.89)	30
Angina	6.02 ± 0.12	6.38 ± 0.13	0.94 (0.89 to 1.00)	278
Stroke death	13.16 ± 0.06	13.30 ± 0.07	0.99 (0.98 to 1.00)	714
Stroke event	41.27 ± 0.17	42.60 ± 0.18	0.97 (0.96 to 0.98)	75
Myocardial Infarction death	7.00 ± 0.09	7.32 ± 0.09	0.96 (0.92 to 0.99)	313
Myocardial Infarction event	11.02 ± 0.16	12.03 ± 0.18	0.92 (0.88 to 0.95)	99
Renal
Microalbuminuria	34.47 ± 0.48	45.05 ± 0.51	0.77 (0.74 to 0.79)	9
Gross proteinuria	19.47 ± 0.39	28.79 ± 0.48	0.68 (0.64 to 0.71)	11
End‐stage renal disease	7.43 ± 0.22	13.35 ± 0.33	0.56 (0.52 to 0.60)	17
Nephropathy death	5.70 ± 0.18	10.68 ± 0.28	0.53 (0.49 to 0.58)	20
Extremities
Ulcer	8.20 ± 0.37	9.60 ± 0.41	0.85 (0.76 to 0.96)	71
Recurring foot ulcer	15.86 ± 0.26	16.75 ± 0.29	0.95 (0.90 to 0.99)	112
Amputation from foot ulcer	3.89 ± 0.16	4.26 ± 0.17	0.91 (0.82 to 1.02)	270
Amputation from recurring foot ulcer	2.64 ± 0.10	2.94 ± 0.11	0.90 (0.81 to 1.00)	333
Neuropathy	66.81 ± 0.49	75.17 ± 0.40	0.89 (0.87 to 0.90)	12

CI, Confidence Interval; rt‐CGM, real‐time continuous glucose monitoring; SE, standard error; SMBG, self‐monitoring of blood glucose; T2D, type 2 diabetes.

DISCUSSION

In this analysis, rt‐CGM was found to be highly cost‐effective in the management of patients with T2D receiving treatment with insulin, with ICERs from the base case (JPY 333,150 per QALY gained) and all scenarios tested being well below the WTP threshold (JPY 5,000,000 per QALY gained). The cost‐effectiveness in the base case was driven by an incremental gain of 1.501 QALYs compared with SMBG. Although rt‐CGM was associated with modestly increased costs over SMBG, the overall cost increase was largely offset by savings in areas such as the management of renal complications, hyperglycemic/hypoglycemic events, and CV complications. Furthermore, the use of rt‐CGM was predicted to result in reductions in the incidence of numerous T2D complications when compared with SMBG. rt‐CGM thus likely provides a cost‐effective option to improve the health and quality of life (QoL) of patients with T2D in Japan.

These results were found to be robust in sensitivity analyses across all 28 scenarios tested. When more than 60 SMBG tests were performed each month, rt‐CGM even became cost‐saving. The results were sensitive to the removal of utilities associated with severe hypoglycemic or hyperglycemic events, highlighting the importance of the benefits of avoiding these with rt‐CGM. However, it should perhaps be noted that the inclusion of these utilities was already arguably done in a conservative manner, with only severe events considered. The majority of hypoglycemic and hyperglycemic events experienced by patients tend to be non‐severe but still have large impacts on QoL ⁶¹ , significantly impact function ⁶² , and often result in indirect costs through impacts on productivity ⁶¹ , ⁶³ , ⁶⁴ . A recent study of Japanese patients with diabetes found that after a daytime non‐SHE, 25% of respondents experience negative impacts on daily activities or work, 34% experience impacts on sleep, and 23% experience impacts on their emotional state the following day ⁶¹ . The utility benefit of reductions in FoH was also found to impact cost effectiveness, supported by previous findings that such fear negatively affects the QoL of patients with diabetes ⁶⁵ , ⁶⁶ , ⁶⁷ . However, FoH may also have further impacts on patients' care not considered here through its contribution to clinical inertia and its potential to act as a barrier to effective treatment ⁶⁸ , ⁶⁹ . Finally, cost‐effectiveness was found to be somewhat sensitive to the magnitude of the HbA1c effect, although the ICER remained far below the WTP threshold even when relatively conservative estimates were used ¹⁴ , ²² , ²³ . Interestingly, the ICER was similar regardless of whether the CV risk‐prediction equation used was developed for an Asian ⁴⁸ or a Western population (UKPDS 82). The use of prediction equations specific to the Japanese population is nevertheless important, particularly due to observations that rates of coronary artery disease tend to be lower than those of ischemic stroke in Japan ⁷⁰ , ⁷¹ , in contrast with what is seen in US populations ⁷² .

The use of data from studies of Japanese populations with T2D was a key strength of this analysis ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ . In particular, parameters such as baseline HbA1c, duration of diabetes, and the prevalence of most comorbidities were sourced from data gathered from Japanese patients. In some cases, such as complication or treatment‐related adverse event utilities, data from Western sources were used due to a lack of data availability for the Japanese population. Nevertheless, published conversion algorithms were used to adapt these for use in the model ³⁹ , ⁵³ . The use of parameters specific to Japanese individuals is particularly important, as significant differences tend to be observed between patients with T2D in Japan and patients with T2D in Western populations ⁴⁰ . A study of patients with T2D in J‐Dreams, a medical records database, found that patients in Japan tend to have lower BMIs than those in the United Kingdom or the United States and have lower rates of congestive heart failure but higher rates of retinopathy, neuropathy, and chronic kidney disease ⁴⁰ . These results are consistent with broader studies of East Asian and Western populations, in which lower BMIs and rates of coronary heart disease but higher rates of renal complications and retinopathy have typically been observed ⁷³ . Even the underlying pathophysiology tends to differ, with T2D in East Asian individuals typically involving greater impairments in insulin secretion and a lesser contribution from insulin resistance than in Western individuals ⁷⁴ .

This study had some limitations. First, there was a reliance on studies of type 1 diabetes for data on long‐term outcomes of the use of rt‐CGM ⁴⁶ , ⁴⁷ , ⁷⁵ due to a lack of such data from studies of patients with T2D. Second, the analysis largely focused on studies of a single rt‐CGM device, the Dexcom G6 system. This may limit the generalizability of the analysis, as this system provides predictive alarms that can potentially reduce hypoglycemia without frequent device interaction ⁷⁶ , ⁷⁷ , which may not be found on other systems. Nevertheless, additional studies have shown that glycemic outcomes are comparable among different Dexcom rt‐CGM systems ¹⁴ , ²¹ , ²² , ²³ , ⁷⁸ . A further complication of this analysis is the lack of a formally defined WTP threshold in Japan. However, the value used here (JPY 5,000,000) is consistent with published estimates ⁵⁹ , ⁷⁹ , ⁸⁰ and has been used in previous economic analyses ³⁹ . Furthermore, the conclusions of this analysis would remain robust even if the most conservative estimates from those publications (JPY 2,000,000–2,600,000) were used ⁵⁹ , ⁷⁹ , ⁸⁰ . A final limitation was that treatment effects had to be sourced from a single‐center study conducted in Denmark ²³ due to a lack of data from Japan. However, the results were nevertheless robust when the HbA1c effect was varied in sensitivity analyses.

Overall, the results of this analysis clearly demonstrate that rt‐CGM is a cost‐effective intervention for the management of patients in Japan with T2D who require treatment with insulin. Although further studies of Japanese patients with T2D may facilitate analyses more relevant to Japanese payers, the base case results and sensitivity analyses presented here suggest that rt‐CGM is highly cost‐effective when compared with SMBG and reduces the incidence of complications associated with T2D.

DISCLOSURES

H.W. has received honoraria for lectures from Dexcom, Bayer Pharma Japan, Teijin Pharma Ltd., MSD, Sanofi‐Aventis K.K., Novo Nordisk, Nippon Boehringer Ingelheim, Eli Lilly, Sumitomo Pharma, Mitsubishi Tanabe Pharma, Daiichi Sankyo Company, Ltd. Abbott, Kowa Co., Ltd., Taisho Pharmaceutical, GlaxoSmithKline, Embecta, Sanwa Kagaku, Kyowa Kirin, and Roche DC Japan; and for research activities for Takeda Pharmaceuticals, Nippon Boehringer Ingelheim, Lifescan Japan, Sumitomo Pharma, Teijin Pharma, Taisho Pharmaceutical, Abbott, Soiken Inc., Sanwa Kagaku, SBI Pharma, and Kowa. S.T. has received consultation fees and outsourcing fees from Boehringer Ingelheim, Satt and the Public Health Research Foundation. ST has received research grants from the Japan Agency for Medical Research and Development, the Japanese Ministry of Health Labor and Welfare, the Japanese Ministry of Education, Science, and Technology, and Novo Nordisk Pharma Ltd. J.Y.M. and G.J.N. are employees of Dexcom and hold stock or stock options in Dexcom. H.A. is a former employee of Dexcom. R.F.P. and M.F. are full‐time employees of Covalence Research Ltd, which received consultancy fees from Dexcom to review the analyses and prepare the manuscript.

Approval of the research protocol: N/A.

Informed Consent: N/A.

Approval date of Registry and the Registration No. of the study/trial: N/A.

Animal Studies: As an economic modeling analysis based on published literature, the present study did not enroll human participants or make use of human biological samples or individually identifiable data, and ethics approval was therefore not required.

Supporting information

Data S1. CHEERS checklist.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.