OBJECTIVE

This study evaluated the societal cost-effectiveness of continuous glucose monitoring (CGM) in patients with type 1 diabetes (T1D) using multiple insulin injections.

RESEARCH DESIGN AND METHODS

In the Multiple Daily Injections and Continuous Glucose Monitoring in Diabetes (DIAMOND) trial, 158 patients with T1D and HbA1c ≥7.5% were randomized in a 2:1 ratio to CGM or control. Participants were surveyed at baseline and 6 months. Within-trial and lifetime cost-effectiveness analyses were conducted. A modified Sheffield T1D policy model was used to simulate T1D complications. The main outcome was cost per quality-adjusted life-year (QALY) gained.

RESULTS

Within the 6-month trial, the CGM group had similar QALYs to the control group (0.462 ± 0.05 vs. 0.455 ± 0.06 years, P = 0.61). The total 6-month costs were $11,032 (CGM) vs.$7,236 (control). The CGM group experienced reductions in HbA1c (0.60 ± 0.74% difference in difference [DiD]), P < 0.01), the daily rate of nonsevere hypoglycemia events (0.07 DiD, P = 0.013), and daily test strip use (0.55 ± 1.5 DiD, P = 0.04) compared with the control group. In the lifetime analysis, CGM was projected to reduce the risk of T1D complications and increase QALYs by 0.54. The incremental cost-effectiveness ratio (ICER) was $98,108 per QALY for the overall population. By extending sensor use from 7 to 10 days in a real-world scenario, the ICER was reduced to$33,459 per QALY.

CONCLUSIONS

#### Non-CGM Medical Care Costs

Health care service utilizations included routine office visits, after-hour clinic visits (urgent care), 911 calls, ambulance use, emergency department visits, and hospitalizations, as well as average daily test strip use and glucose-lowering medications for the prior 6 months. To calculate the 6-month costs of utilization at baseline and 6 months, the median prices of each health care service were multiplied by the number of each service use in the past 6 months.

#### Indirect Costs

Patients were also surveyed at baseline and 6 months on the number of missed workdays due to diabetes and the number of workdays with underperformance (defined as <50% productivity) (13). A workday with underperformance was considered a half-day of missed work. The costs of missed work and time for self-diabetes care were calculated based on median hourly wage per job category of patients through well-established references (provided in Supplementary Table 3).

### Measurements of QoL

The 5-Level EuroQol 5-Dimension (EQ-5D-5L) questionnaire was used to measure health-related QoL. This QoL measure was converted into a utility score, ranging from 0.0 (“death”) to 1.0 (“perfect life”) (17). For the long-term CEA, we used previously published disutilities for microvascular and cardiovascular complications and for severe and nonsevere hypoglycemia. The utilities were then incorporated into a simulation model of long-term outcomes.

Quality-adjusted life years (QALYs), a measure of health outcomes and disease burden, were calculated by the area under the curve method using the calculated utility scores. The incremental cost effectiveness ratio (ICER) was calculated as the ratio of the difference in costs to the difference in QALYs between the two groups.

### Nonsevere Hypoglycemic Events

Nonsevere hypoglycemic events (NSHEs), self-treated hypoglycemia (18), may be experienced by 24–85% of patients with diabetes, particularly among insulin users (1820). Each patient in the DIAMOND trial was required to wear a blinded CGM device to record glucose concentration (invisible to the patients) for 2 weeks before randomization and for a week before the 12- and 24-week visits. NSHEs were defined as the detection of a glucose value <3.0 mmol/L (<54 mg/dL) for at least 20 consecutive minutes, considered to be clinically significant biochemical hypoglycemia according to the International Hypoglycemia Study Group recommendations (21). A daily rate of NSHE was obtained based on the number of NSHEs observed during each period of blinded CGM use. The numbers of NSHEs at 6 months were pooled from 12- to 24-week visits because the NSHE rates did not differ between the two visits (12).

### Within-Trial CEAs

We followed the intent-to-treat principle in our analyses. A Wilcoxon rank sum test was used to compare the two groups in QALYs, utility, and other continuous outcomes. The Fisher exact test was used for each categorical outcome. An ANCOVA was performed to model the change from baseline in utility, HbA1c, daily rate of NSHEs, and daily strip test use, and to compare the two groups, adjusting for their baseline effects. A repeated ANOVA via a linear mixed model (LMM) was also performed to test the effects of treatment, time, and their interaction. Age, sex, and duration of T1D were treated as potential covariates in each model, and site was considered as a random effect. A test of the interaction between the baseline outcome and treatment arm was also conducted through an ANCOVA or LMM to assess homogeneity of the treatment effect. Because ANCOVA and LMM produced similar results, we only present the ANCOVA results. All P values were two-sided, and P values <0.05 were considered significant. Analyses were conducted with SAS version 9.4 software.

To evaluate the cost-effectiveness of CGM, we simulated the natural history of T1D over the projected lifetime of patients by extrapolating the within-trial findings. Among the existing T1D simulation models in the literature, we selected the patient-level Sheffield T1D policy model (22) based on the following features: the model was constructed solely using T1D studies and trials, it includes HbA1c, a risk factor significantly modified by CGM, in most risk equations, it was validated against major T1D trial studies, and it is completely transparent and hence reproducible. The simulation model is characterized by the simultaneous progression of T1D through major microvascular and macrovascular complications as well as short-term complications (hypo- and hyperglycemia) and their associated costs and health utilities. The risk parameters for the microvascular- and macrovascular-related complications and mortalities of the Sheffield model were taken directly from Tables 1 and 2 in Thokala et al. (22).

We modified the Sheffield model with respect to severe and nonsevere hypoglycemic events. The original hypoglycemia module assumed that the risk of severe hypoglycemic events increases as the HbA1c level decreases. This relationship is altered by CGM, as described in the systematic review and meta-analysis conducted on behalf of Agency for Healthcare Research and Quality (11) and as we found in the DIAMOND trial: patients are able to achieve normoglycemia without an increased risk of hypoglycemia. As an alternative to the Sheffield hypoglycemia module, we used the observed hypoglycemia event rates from the DIAMOND trial. In the model, we also incorporated NSHEs and their QoL and costs effects, which were not considered in the original Sheffield model. All base-case model parameters, including clinical inputs, costs, and health utilities, are described in Supplementary Tables 57. A 3% annual discount rate was applied to costs and health utilities (23). We used bootstrapping of simulation samples to calculate 95% CIs for the key outcomes.

#### Projected CGM Effects

The simulation base-case model carried forward the CGM effects on HbA1c reduction and NSHEs rate found in the trial through the lifetime of patients (Supplementary Table 5). Other risk factors, such as systolic blood pressure, HDL cholesterol, and total cholesterol, which were only measured at the baseline in the trial, were assumed not to change over time in the simulations (Supplementary Table 5). Similar modeling assumptions have been used in prior CEAs of new interventions for diabetes, including CGM (13).

#### One-Way Sensitivity and Subgroup Analyses

We performed subgroup analyses per baseline HbA1c with a cutoff of 8.5%. We also conducted one-way sensitivity analyses on selected model parameters. Several scenarios were considered, including changes in CGM effects and costs and reflecting CGM technology advancements such as Dexcom’s G5 CGM. The Dexcom G5 CGM was approved by the U.S. Food and Drug Administration to allow for replacement of finger-stick blood glucose testing (24). In the G5 scenarios, we assumed that patients use as few as 2.8 test strips per day (25). Because smartphones can be used as a G5 receiver (26), we assumed that 50% of the CGM patients use their phones as the receiver from the 2nd year of obtaining CGM onwards (the U.S. Food and Drug Administration requires that a CGM receiver should be included in the CGM package). Because the lifetime of device sensors can be prolonged from the recommended 7 days to 10 days (27) without compromising safety (28), we also conducted sensitivity analyses for the real-world use of G4 and G5 CGMs (see Supplementary Table 4). Details of all subgroup and sensitivity analyses scenarios are provided in Supplementary Table 8.

In the base-case analysis, we assumed that patients continued to use and benefit from CGM over their lifetime. In clinical practice, patients may in fact discontinue use of CGM over time. To account for this phenomenon, we considered alternative assumptions and evaluated scenarios where CGM was used for different periods of 5, 10, 15, 20, and 25 years. In each scenario, patients accrued the benefits and the costs of CGM only during periods of use.

#### Probabilistic Sensitivity Analysis

To examine the combined effects of model parameter uncertainty, we performed a probabilistic sensitivity analysis (PSA) by varying multiple parameters simultaneously. All input parameters in the base-case model, including costs, utilities, and parameters in risk equations, were sampled using distributions described in the Sheffield model (22). A total of 200 PSA scenarios were generated, and the same simulation method as in the base-case analysis was performed for each scenario. The simulation results were summarized in a cost-effectiveness acceptability curve (CEAC).

The 6-month visit for outcome collection was completed by 102 participants (97%) in the CGM group and all 53 participants (100%) in the control group. The CGM group was younger than the control group (mean ± SD: 45.7 ± 13.6 vs. 51.4 ± 10.9, P < 0.01) (Table 1). No other significant differences were found for the other baseline characteristics.

Table 1

Baseline characteristics of the study populations

Control (n = 53)CGM (n = 105)P value
Demographic characteristics
Female sex, n (%) 23 (43) 47 (45)
Race, n (%)   0.5
White 50 (94) 95 (90)
Black 3 (6) 6 (6)
Other 0 (0) 4 (4)
Age (years)   <0.01
Mean ± SD 51.4 ± 10.9 45.7 ± 13.6
Range 25–73 25–72
T1D duration (years)   0.1
Mean ± SD 23.1 ± 14.5 19.6 ± 13.1
Range 4–56 2–56
Clinical outcomes at baseline
HbA1c (%)   0.6
Mean ± SD 8.6 ± 0.6 8.6 ± 0.7
Range 7.5–9.9 7.5–9.9
Daily strip tests (n  0.4
Mean ± SD 4.1 ± 1.6 3.9 ± 1.3
Range 1–9 1–7
Insulin (units)   0.4
Mean ± SD 60.2 ± 32.7 66.0 ± 36.7
Range 12–185 24–155
Daily event rate of NSHEs   0.5
Mean ± SD 0.35 ± 0.33 0.30 ± 0.29
Range 0–1.34 1–1.33
Utility   0.6
Mean ± SD 0.91 ± 0.13 0.93 ± 0.10
Range 0.26–1.0 0.38–1.0
BMI (kg/m2  0.2
Mean ± SD 26.8 ± 4.6 27.9 ± 5.9
Range 18.2–38.4 16.9–62.6
Patients using noninsulin glucose–lowering medication, n (%) 8 (8) 4 (8) 1.0
Patients having severe hypoglycemia in previous 12 months, n (%) 8 (8) 9 (17) 0.1
Patients having severe hyperglycemia in previous 12 months, n (%) 1 (2) 1 (<1) 1.0
Control (n = 53)CGM (n = 105)P value
Demographic characteristics
Female sex, n (%) 23 (43) 47 (45)
Race, n (%)   0.5
White 50 (94) 95 (90)
Black 3 (6) 6 (6)
Other 0 (0) 4 (4)
Age (years)   <0.01
Mean ± SD 51.4 ± 10.9 45.7 ± 13.6
Range 25–73 25–72
T1D duration (years)   0.1
Mean ± SD 23.1 ± 14.5 19.6 ± 13.1
Range 4–56 2–56
Clinical outcomes at baseline
HbA1c (%)   0.6
Mean ± SD 8.6 ± 0.6 8.6 ± 0.7
Range 7.5–9.9 7.5–9.9
Daily strip tests (n  0.4
Mean ± SD 4.1 ± 1.6 3.9 ± 1.3
Range 1–9 1–7
Insulin (units)   0.4
Mean ± SD 60.2 ± 32.7 66.0 ± 36.7
Range 12–185 24–155
Daily event rate of NSHEs   0.5
Mean ± SD 0.35 ± 0.33 0.30 ± 0.29
Range 0–1.34 1–1.33
Utility   0.6
Mean ± SD 0.91 ± 0.13 0.93 ± 0.10
Range 0.26–1.0 0.38–1.0
BMI (kg/m2  0.2
Mean ± SD 26.8 ± 4.6 27.9 ± 5.9
Range 18.2–38.4 16.9–62.6
Patients using noninsulin glucose–lowering medication, n (%) 8 (8) 4 (8) 1.0
Patients having severe hypoglycemia in previous 12 months, n (%) 8 (8) 9 (17) 0.1
Patients having severe hyperglycemia in previous 12 months, n (%) 1 (2) 1 (<1) 1.0

### Within-Trial CEA

During the trial, the two groups had similar utility ratings (mean ± SD: control: 0.91 ± 0.12 vs. CGM: 0.92 ± 0.10, P = 0.78) and QALYs (control: 0.46 ± 0.06 vs. CGM: 0.46 ± 0.05), both close to 0.5, the maximum QALY attainable in a half-year (Table 2). From a societal perspective, the average 6-month total costs were $11,032 for the CGM group and$7,236 for the control group (P < 0.01). The difference in total costs was primarily attributable to CGM device costs of $2,554. We found no other difference between the groups in other major cost categories such as direct personnel costs, non-CGM medical care costs, and all indirect costs associated with work productivity (all P > 0.4). CGM reduced daily strip use (0.55 ± 1.5 difference in difference [DiD], P = 0.04), resulting in a reduction in costs of strip use (CGM:$612; control: $750). CGM patients had no significant difference from control patients in their frequency of interactions with trial personnel or other forms of health care utilization (Supplementary Table 9), even though they were granted one additional visit at week 1 to review CGM use instructions. No within-trial ICER was calculated due to the lack of difference in QALYs. CGM was dominated by control in the within-trial CEA. Table 2 Within-trial cost-effectiveness results Control (n = 53) CGM (n = 103) Mean (SD)Median (range or IQR)Mean (SD)Median (range or IQR) Utility and QALYs P valueb Utility change from baseline 0.0 (0.08) 0 (−0.27, 0.26) −0.01 (0.09) 0 (−0.33, 0.32) 0.78 QALYs 0.46 (0.06) 0.47 (0.13, 0.50) 0.46 (0.05) 0.47 (0.24, 0.50) 0.61 Costs,$     P valueb
Total direct costs 3,118 (3,120) 2,565 (1,928, 3,277) 5,336 (3,070) 5,092 (4,485, 5,726) <0.01
Direct trial personnel 96 (205) 47 (0, 94) 60 (77) 47 (0, 94) 0.41
Medical care 3,022 (3,088) 2,478 (1,880, 3,122) 2,921 (3,065) 2,509 (1,909, 3,095) 0.86
CGM 0 (0) 2,554 (0) 2,554 <0.01
Total indirect costsa 36 (121) 0 (0, 0) 54 (314) 0 (0, 0) 0.85
Missed work 26 (101) 0 (0, 0) 36 (307) 0 (0, 0) 0.65
Poor performance 10 (40) 0 (0, 0) 18 (70) 0 (0, 0) 0.63
Self-management 4,012 (5,529) 2,829 (0, 5,610) 5,473 (10,300) 2,829 (2,259, 5,658) 0.86
Total costs 7,236 (6,097) 5,287 (4,586, 8,223) 11,200 (11,300) 8,178 (6,864, 10,300) <0.01
Total costsa 3,154 (3,122) 2,565 (1,999, 3,513) 5,593 (3,083) 5,105 (4,496, 5,780) <0.01
Clinical outcomes: reduction from baseline     P valueb
HbA1c −0.39 (0.70) −0.30 (−3.20, 0.90) −0.99 (0.77) −1.00 (−3.00, 0.70) <0.01
Daily strip tests 0.1 (1.5) 0 (−4, 3) −0.5 (1.5) 0 (−5, 3) 0.04
Insulin dose 1.0 (11) 1 (−23, 25) −2.3 (22) 0 (−145, 52) 0.31
Daily rate of NSHEs −0.06 (0.27) 0 (−0.93, 0.47) −0.12 (0.29) −0.08 (−1.07, 0.63) 0.02c
BMI 0.27 (1.07) 0.15 (−2.22, 2.80) 0.59 (1.38) 0.56 (−3.42, 5.28) 0.16
Patients having severe hyperglycemic events, n (%) 1 (2)  0 (0)  0.34
Patients having severe hypoglycemic events, n (%) 2 (4)  2 (2)  0.6
Subgroup analyses: reduction from baseline     P valuec
In the subgroup with high baseline HbA1c (≥8.5%)
HbA1c −0.53 (0.60) −0.50 (−1.5, 0.8) −1.29 (0.77) −1.30 (−3, 0.3) 0.02
Daily rate of NSHEs −0.10 (0.29) −0.07 (−0.93, 0.47) −0.08 (0.27) −0.07 (−1.03, 0.63) 0.27
In the subgroup with low baseline HbA1c (<8.5%)
HbA1c −0.22 (0.78) −0.10 (−3.20, 0.90) −0.63 (0.59) −0.60 (−1.80, 0.70) 0.01
Daily rate of NSHEs −0.02 (0.25) 0.01 (−0.86, 0.32) −0.17 (0.32) −0.14 (−1.07, 0.44) 0.03
Control (n = 53)
CGM (n = 103)
Mean (SD)Median (range or IQR)Mean (SD)Median (range or IQR)
Utility and QALYs     P valueb
Utility change from baseline 0.0 (0.08) 0 (−0.27, 0.26) −0.01 (0.09) 0 (−0.33, 0.32) 0.78
QALYs 0.46 (0.06) 0.47 (0.13, 0.50) 0.46 (0.05) 0.47 (0.24, 0.50) 0.61
Costs, $P valueb Total direct costs 3,118 (3,120) 2,565 (1,928, 3,277) 5,336 (3,070) 5,092 (4,485, 5,726) <0.01 Direct trial personnel 96 (205) 47 (0, 94) 60 (77) 47 (0, 94) 0.41 Medical care 3,022 (3,088) 2,478 (1,880, 3,122) 2,921 (3,065) 2,509 (1,909, 3,095) 0.86 CGM 0 (0) 2,554 (0) 2,554 <0.01 Total indirect costsa 36 (121) 0 (0, 0) 54 (314) 0 (0, 0) 0.85 Missed work 26 (101) 0 (0, 0) 36 (307) 0 (0, 0) 0.65 Poor performance 10 (40) 0 (0, 0) 18 (70) 0 (0, 0) 0.63 Self-management 4,012 (5,529) 2,829 (0, 5,610) 5,473 (10,300) 2,829 (2,259, 5,658) 0.86 Total costs 7,236 (6,097) 5,287 (4,586, 8,223) 11,200 (11,300) 8,178 (6,864, 10,300) <0.01 Total costsa 3,154 (3,122) 2,565 (1,999, 3,513) 5,593 (3,083) 5,105 (4,496, 5,780) <0.01 Clinical outcomes: reduction from baseline P valueb HbA1c −0.39 (0.70) −0.30 (−3.20, 0.90) −0.99 (0.77) −1.00 (−3.00, 0.70) <0.01 Daily strip tests 0.1 (1.5) 0 (−4, 3) −0.5 (1.5) 0 (−5, 3) 0.04 Insulin dose 1.0 (11) 1 (−23, 25) −2.3 (22) 0 (−145, 52) 0.31 Daily rate of NSHEs −0.06 (0.27) 0 (−0.93, 0.47) −0.12 (0.29) −0.08 (−1.07, 0.63) 0.02c BMI 0.27 (1.07) 0.15 (−2.22, 2.80) 0.59 (1.38) 0.56 (−3.42, 5.28) 0.16 Patients having severe hyperglycemic events, n (%) 1 (2) 0 (0) 0.34 Patients having severe hypoglycemic events, n (%) 2 (4) 2 (2) 0.6 Subgroup analyses: reduction from baseline P valuec In the subgroup with high baseline HbA1c (≥8.5%) HbA1c −0.53 (0.60) −0.50 (−1.5, 0.8) −1.29 (0.77) −1.30 (−3, 0.3) 0.02 Daily rate of NSHEs −0.10 (0.29) −0.07 (−0.93, 0.47) −0.08 (0.27) −0.07 (−1.03, 0.63) 0.27 In the subgroup with low baseline HbA1c (<8.5%) HbA1c −0.22 (0.78) −0.10 (−3.20, 0.90) −0.63 (0.59) −0.60 (−1.80, 0.70) 0.01 Daily rate of NSHEs −0.02 (0.25) 0.01 (−0.86, 0.32) −0.17 (0.32) −0.14 (−1.07, 0.44) 0.03 All costs data were summarized by interquartile range (IQR) and other continuous outcomes were summarized by range. Bold P values indicate statistical significance (P < 0.05). aBoth total indirect costs and total costs did not include the costs from diabetes self-management due to its 20% missing data and huge variability; that is, ∼20% patients reported unknown daily number of hours of self-management and seven patients from both groups reported ≥12 h/day and two of CGM users reported 24 h/day. bP value was from the Wilcoxon rank sum test to compare the two groups. cP value was from an ANCOVA model adjusting for its baseline outcome and site as a random effect. In addition to the key CEA results, a number of clinical outcomes were evaluated. CGM lowered HbA1c (reduction from baseline: CGM: −1.0 ± 0.8% vs. control: −0.4 ± 0.7%) and the daily rate of NSHEs (CGM: −0.12 ± 0.29 vs. control: −0.06 ± 0.27, P = 0.02). In the subgroup with high baseline HbA1c (≥8.5%), CGM reduced HbA1c (0.76% DiD, P < 0.01), with no change in the daily rate of NSHEs (P = 0.27) compared with the control group (Table 2). In the subgroup with low baseline HbA1c (<8.5%), CGM reduced HbA1c (0.41% DiD, P < 0.01) and the daily rate of NSHEs (0.15 DiD from raw means, 0.125 DiD from adjusted means; P = 0.03). There were no significant differences in insulin dosing, BMI, number of severe hypoglycemic events, and number of patients who changed noninsulin glucose–lowering medicines. With the exception of one CGM patient who stopped using metformin during the trial, all the other patients remained on the same glucose-lowering medicines as at baseline. Age was not significantly associated with any of the clinical outcomes (results not shown). ### Long-Term CEA #### Base-Case Analyses The results of the lifetime analysis indicate that CGM is expected to reduce the average incidence rates of all major T1D complications (Table 3). In particular, the use of CGM may lead to reductions in the lifetime risks of end-stage microvascular complications, including blindness (1.9 → 1.8%), end-stage renal disease (11.7 → 10.1%), and amputation (8.1 → 7.1%). Similarly, CGM reduced the lifetime risk of macrovascular complications, including myocardial infarction (37.8 → 37.0%), stroke (7.2 → 7.0%), and heart failure (11.1 → 10.7%). Further, life expectancy improved by 0.72 years on average under CGM (24.29 → 25.01). The improvement in quality-adjusted life expectancy was 0.54 QALYs. The ICER in the base-case was estimated at$98,108 per QALY, and its 95% CI was $90,298–$105,144 per QALY.

Table 3

ControlCGM
Diabetic retinopathy, %
Background  33.5 27.3
Proliferative  28.9 24.6
Macular edema, % 8.4 6.4
Blindness, % 1.9 1.8
Macroalbuminuria, % 19.7 17.2
End-stage renal disease, % 11.7 10.1
Neuropathy, % 33.2 27.3
Amputation, % 8.1 7.1
Myocardial infarction, % 37.8 37.0
Stroke, % 7.2 7.0
Angina, % 20.6 20.6
Heart failure, % 11.1 10.7
Expected life–years 24.29 25.01
Difference in expected life-years 0.72
Discounted QALYs, means 12.78 13.32
Difference in QALYs, mean 0.54
Discounted total costs, mean 305,278 360,486
Difference in costs, mean 55,208
ICER, mean (95% CI*98,108 (90,298–105,144)
ControlCGM
Diabetic retinopathy, %
Background  33.5 27.3
Proliferative  28.9 24.6
Macular edema, % 8.4 6.4
Blindness, % 1.9 1.8
Macroalbuminuria, % 19.7 17.2
End-stage renal disease, % 11.7 10.1
Neuropathy, % 33.2 27.3
Amputation, % 8.1 7.1
Myocardial infarction, % 37.8 37.0
Stroke, % 7.2 7.0
Angina, % 20.6 20.6
Heart failure, % 11.1 10.7
Expected life–years 24.29 25.01
Difference in expected life-years 0.72
Discounted QALYs, means 12.78 13.32
Difference in QALYs, mean 0.54
Discounted total costs, mean 305,278 360,486
Difference in costs, mean 55,208
ICER, mean (95% CI*98,108 (90,298–105,144)

*CI of the mean. The CI was calculated by bootstrapping simulation samples (each simulation scenario consists of 2,000,000 simulation samples (1,000,000 for each study arm), which were created by first generating 1,000 sample patients and then simulating their lifetime each 1,000 times per study arm).

Figure 1 shows ICER values for one-way sensitivity and subgroup analyses. The most notable changes in ICER were influenced by the price of CGM. In the real-world scenarios with extended use of CGM components (price justification provided in Supplementary Table 4), the ICERs were $33,459 per QALY for G4 and$41,464 per QALY for G5, whereas a 25% increase in the CGM price increased the ICER to $159,679 per QALY. The second key factor was HbA1c reduction produced by CGM. If the HbA1c reduction observed during the trial was 50% higher, the ICER would be$61,586 per QALY, whereas a 50% reduction of the glycemic benefit would increase the ICER to $177,268 per QALY. Assumptions regarding the QoL effects of NSHEs also had a substantial effect on the ICERs. The major reason behind this observation is that the frequencies of NSHEs are very high (annual rates observed in the trial: CGM: 85.9 vs. control: 131.4 in patients with low baseline HbA1c). As a result, any small change in the per-episode QoL disutility can lead to large changes in the annual NSHE-related QoL disutilities. For example, by multiplying the per-episode disutility of −0.00045 in the base-case model with the annual frequencies of NSHEs, the annual NSHE disutilities are −0.0387 (CGM) and −0.0591 (control), causing a difference of +0.02 in annual utilities. Figure 1 Results of subgroup and one-way sensitivity analyses. Figure 1 Results of subgroup and one-way sensitivity analyses. Close modal #### Sensitivity Analysis on Duration of CGM Use Prolonged use of CGM increased the differences between CGM and control in lifetime costs, life expectancy, and QALYs (Supplementary Table 10). Because costs increased at a higher rate than QALYs, ICERs also increased with longer CGM use. The key CEA outcomes at 25 years of CGM use approach those under the base-case assumption, demonstrating the consistency of our base-case analysis. #### PSA The CEAC (in Supplementary Fig. 1) revealed the consistency of our base-case results. For example, in ∼90% of the PSA scenarios, CGM use was cost-effective at a willingness-to-pay threshold of$100,000 per QALY.

We assessed the cost-effectiveness of CGM use compared with usual care with SMBG in adults with T1D with elevated HbA1c levels (≥7.5%). Within the trial, CGM increased costs without immediately improving QoL as measured by the EQ-5D. However, CGM reduced HbA1c (0.6% DiD), daily strip test use (201 strips/year DiD), and NSHEs (25 episodes/year DiD). When these clinical benefits were extrapolated over a lifetime, CGM emerged as a cost-effective intervention with an ICER of $98,108 per QALY, with patients gaining 0.54 QALYs. This base-case result was robust to most of the sensitivity analyses. Most of the incremental cost was attributable to the CGM costs ($5,548 per annum). With reasonable extended use of CGM components, the annual cost of CGM could be reduced to $3,271, and the ICER would decrease to$33,459 per QALY. Similarly, the ICER for G5 declines from $96,376 per QALY to$41,464 per QALY with real-world use of components. The base-case ICER for this device is below acceptable ICER thresholds ($109,000 and$297,000 per QALY), which are estimated from existing coverage decisions for medical services by U.S. insurance plans, reflecting the societal preferences in the U.S. (29).

Our findings are consistent with the results of prior CEA studies of CGM in patients with T1D. In the JDRF trial, the ICER of CGM versus usual care in the trial adults with HbA1c ≥7.0% (80% insulin pump users) was $98,679 per QALY (13). The incremental cost was solely due to CGM and not to the pump. Building on the original JDRF trial results, a subsequent CEA study found that the ICER of CGM use versus SMBG in the broader T1D adult population with a full range of HbA1c was$45,033 per QALY (30). CGM has also been evaluated as a combined intervention with the insulin pump. A CEA study of sensor-augmented pump therapy (CGM + continuous subcutaneous insulin infusion [CSII]) versus CSII in Danish patients with a mean HbA1c of 8.1% found that the ICER was $24,751 per QALY (31). In a study of sensor-augmented pump therapy versus usual care in adults with inadequately controlled glucose, the ICER was$168,104 per QALY (32). In a similar study in the Swedish health care setting, the ICER of CGM+ CSII versus usual care was $41,000 per QALY (23). A meta-analysis study (33) concluded that CGM was likely to be cost-effective in patients with T1D, particularly in those with poor glucose control. From a policy perspective, because the total cost of diabetes care for the U.S. population has been steadily rising in an unsustainable fashion, carefully evaluating the clinical and economic values of new glucose-control technologies is critically important. From 2007 to 2012, the total cost of diabetes care in the U.S. rose from$174 billion to $245 billion, a 41% increase (5). Newer CGM technology has been found to significantly improve the precise and accurate measurement of glucose levels and to help optimize glucose control. As such, CGM can be a useful clinical and lifestyle aid for patients with T1D (9,10). Our study provides an important expanded view on the contemporary economic value of CGM in T1D. The prior JDRF trial and its economic study did not have substantial representation by patients with T1D using multiple daily injections of insulin (13). Our study based on the DIAMOND trial confirms that CGM is equally valuable in this large and important segment of the population with T1D. An important strength of our study is that we selected and modified the Sheffield model (22), one of the most rigorous and thoroughly reported T1D models (34). The Sheffield model is the only model that features all the following properties: ability to account for all major complications, estimation of model parameters solely based on T1D trial data, HbA1c as a predictor in most risk equations, validation with major T1D studies, capable of sensitivity analysis, and full transparency. Based on the trial findings, we modified the Sheffield model by removing the interdependency between rates of severe hypoglycemic events and HbA1c levels. Another strength of this study is that we were able to capture and model NSHEs in our economic model. The cost of a NSHE episode used in the study was$20.32, including $11.09 in direct medical costs and$9.23 in indirect costs (35,36). Lowering NSHEs not only reduces the risk of subsequent severe hypoglycemia (37) but also improves QoL, including psychological well-being, adherence to treatment regimens, work productivity, and quality and quantity of sleep (6,18,19).

Our study has some limitations. The disutility value of a NSHE was difficult to ascertain due to diverse definitions of NSHEs within a limited literature. Prior studies of the QoL effects of NSHE (38,39) were based on life with and without symptomatic hypoglycemia or on the experience of a single symptomatic hypoglycemic event. These patient-reported definitions are distinct from the new international definition of <54 mg/dL for ≥20 successive minutes (21), attainable by CGM device, which has a higher frequency than those based on past definitions and is frequently asymptomatic. The EQ-5D is widely accepted and regularly used for indirect utility assessment in clinical trials but may not be sensitive enough to reflect changes in QoL that are caused by reductions in NSHEs (40). Future work should pursue translating the QoL effects of NSHEs into health state utilities. Although the trial found that CGM reduced the daily rate of NSHEs, most of the CGM patients still experienced NSHEs. This indicates that reducing NSHEs should be an important goal of future interventions. Whether a 6-month treatment effect will be sustained over a person’s lifetime is not clear. Our other study limitation is related to the possible inaccuracy of number of missed workdays, number of workdays with <50% productivity, and daily hours devoted to self-management diabetes care. Because patients were surveyed at baseline and 6 months, a 6-month recall period may be too long to recall this information accurately.

Despite higher within-trial costs, for adults with T1D multiple insulin injections and suboptimal glycemic control, CGM is cost-effective at the $100,000 per QALY willingness-to-pay threshold with improved glucose control and reductions in nonsevere hypoglycemia. With real-world use, CGM can be highly cost-effective. W.W. and M.R.S. share equal first authorship. Funding. This study was funded by a grant from Dexcom and the National Institute of Diabetes and Digestive and Kidney Diseases Chicago Center for Diabetes Translation Research (P30-DK-092949). A.W. was supported by the Royster Society of Fellows at the University of North Carolina at Chapel Hill. E.S.H. was supported by a National Institute of Diabetes and Digestive and Kidney Diseases Midcareer Investigator Award in Patient-Oriented Research (K24-DK-105340). Duality of Interest. No potential conflicts of interest relevant to this article were reported. Author Contributions. W.W. and M.R.S. contributed to statistical analysis. W.W., M.R.S., A.M., A.G.N., A.W., P.Z., M.O., and E.S.H. contributed to administrative, technical, or material support. W.W., M.R.S., and E.S.H. drafted the manuscript. All authors contributed to study supervision and to critical revision of the manuscript for important intellectual content. E.S.H. obtained funding. W.W., M.R.S., and E.S.H. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Prior Presentation. Parts of this study were presented in abstract form at the 39th Annual Meeting of the Society for Medical Decision Making, Pittsburgh, PA, 22–25 October 2017. 1. Nathan DM ; DCCT/EDIC Research Group . The Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications study at 30 years: overview . Diabetes Care 2014 ; 37 : 9 16 [PubMed] 2. Kibirige D , Akabwai GP , Kampiire L , Kiggundu DS , Lumu W . Frequency and predictors of suboptimal glycemic control in an African diabetic population . Int J Gen Med 2017 ; 10 : 33 38 [PubMed] 3. American Diabetes Association . Approaches to glycemic treatment . Sec 7. In Standards of Medical Care in Diabetes–2015. Diabetes Care 2015 ; 38 ( Suppl. 1 ): S41 S48 [PubMed] 4. Miller KM , Foster NC , Beck RW , et al.; T1D Exchange Clinic Network . Current state of type 1 diabetes treatment in the U.S.: updated data from the T1D Exchange Clinic registry . Diabetes Care 2015 ; 38 : 971 978 [PubMed] 5. American Diabetes Association . Economic costs of diabetes in the U.S. in 2012 . Diabetes Care 2013 ; 36 : 1033 1046 [PubMed] 6. Fulcher G , Singer J , Castañeda R , et al . The psychosocial and financial impact of non-severe hypoglycemic events on people with diabetes: two international surveys . J Med Econ 2014 ; 17 : 751 761 [PubMed] 7. Heller SR , Frier BM , Hersløv ML , Gundgaard J , Gough SC . Severe hypoglycaemia in adults with insulin-treated diabetes: impact on healthcare resources . Diabet Med 2016 ; 33 : 471 477 [PubMed] 8. Bode BW , Schwartz S , Stubbs HA , Block JE . Glycemic characteristics in continuously monitored patients with type 1 and type 2 diabetes: normative values . Diabetes Care 2005 ; 28 : 2361 2366 [PubMed] 9. Schnell O , Barnard K , Bergenstal R , et al . Role of continuous glucose monitoring in clinical trials: recommendations on reporting . Diabetes Technol Ther 2017 ; 19 : 391 399 [PubMed] 10. Walker MS , Fonda SJ , Salkind S , Vigersky RA . Advantages and disadvantages of realtime continuous glucose monitoring in people with type 2 diabetes . US Endocrinol 2012 ; 8 : 22 26 11. Yeh HC , Brown TT , Maruthur N , et al . Comparative effectiveness and safety of methods of insulin delivery and glucose monitoring for diabetes mellitus: a systematic review and meta-analysis . Ann Intern Med 2012 ; 157 : 336 347 [PubMed] 12. Beck RW , Riddlesworth T , Ruedy K , et al.; DIAMOND Study Group . Effect of continuous glucose monitoring on glycemic control in adults with type 1 diabetes using insulin injections: the DIAMOND randomized clinical trial . JAMA 2017 ; 317 : 371 378 [PubMed] 13. Huang ES , O’Grady M , Basu A , et al.; Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group . The cost-effectiveness of continuous glucose monitoring in type 1 diabetes . Diabetes Care 2010 ; 33 : 1269 1274 [PubMed] 14. Hirsch IB . The first hybrid closed-loop insulin pump: will it meet its potential ? Diabetes Technol Ther 2017 ; 19 : 140 141 [PubMed] 15. Sanders GD , Neumann PJ , Basu A , et al . Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: Second Panel on Cost-Effectiveness in Health and Medicine . JAMA 2016 ; 316 : 1093 1103 [PubMed] 16. Husereau D , Drummond M , Petrou S , et al.; CHEERS Task Force . Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement . Value Health 2013 ; 16 : e1 e5 [PubMed] 17. Devlin NJ , Shah KK , Feng Y , Mulhern B , van Hout B . Valuing health-related quality of life: an EQ-5D-5L value set for England . Health Econ 2018 ; 1 : 7 22 18. Brod M , Højbjerre L , Bushnell DM , Hansen CT . Assessing the impact of non-severe hypoglycemic events and treatment in adults: development of the Treatment-Related Impact Measure-Non-severe Hypoglycemic Events (TRIM-HYPO) . Qual Life Res 2015 ; 24 : 2971 2984 [PubMed] 19. Davis RE , Morrissey M , Peters JR , Wittrup-Jensen K , Kennedy-Martin T , Currie CJ . Impact of hypoglycaemia on quality of life and productivity in type 1 and type 2 diabetes . Curr Med Res Opin 2005 ; 21 : 1477 1483 [PubMed] 20. Cariou B , Fontaine P , Eschwege E , et al . Frequency and predictors of confirmed hypoglycaemia in type 1 and insulin-treated type 2 diabetes mellitus patients in a real-life setting: results from the DIALOG study . Diabetes Metab 2015 ; 41 : 116 125 [PubMed] 21. International Hypoglycaemia Study Group . Erratum to: glucose concentrations of less than 3.0 mmol/l (54 mg/dl) should be reported in clinical trials: a joint position statement of the American Diabetes Association and the European Association for the Study of Diabetes [published correction appears in Diabetologia 2017;60:3–6] . Diabetologia 2017 ; 60 : 377 [PubMed] 22. Thokala P , Kruger J , Brennan A , et al . Assessing the cost-effectiveness of type 1 diabetes interventions: the Sheffield type 1 diabetes policy model . Diabet Med 2014 ; 31 : 477 486 [PubMed] 23. Roze S , Saunders R , Brandt AS , de Portu S , Papo NL , Jendle J . Health-economic analysis of real-time continuous glucose monitoring in people with Type 1 diabetes . Diabet Med 2015 ; 32 : 618 626 [PubMed] 24. U.S. Food and Drug Administration . Dexcom G5 Mobile Continuous Glucose Monitoring System - P120005/S041. 2016 . Available from https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P120005S041. Accessed 21 December 2016 25. Aleppo G , Ruedy KJ , Riddlesworth TD , et al.; REPLACE-BG Study Group . REPLACE-BG: a randomized trial comparing continuous glucose monitoring with and without routine blood glucose monitoring in adults with well-controlled type 1 diabetes . Diabetes Care 2017 ; 40 : 538 545 [PubMed] 26. DEXCOM. Dexcom G5 Mobile - CGM on Your Phone. Receiver Optional. 2015. Available from https://diatribe.org/dexcom-g5-mobile-cgm-your-phone-receiver-optional. Accessed 16 October 2015 27. Vaddiraju S , Burgess DJ , Tomazos I , Jain FC , Papadimitrakopoulos F . Technologies for continuous glucose monitoring: current problems and future promises . J Diabetes Sci Technol 2010 ; 4 : 1540 1562 [PubMed] 28. Garg SK , Voelmle MK , Gottlieb P . Feasibility of 10-day use of a continuous glucose-monitoring system in adults with type 1 diabetes . Diabetes Care 2009 ; 32 : 436 438 [PubMed] 29. Braithwaite RS , Meltzer DO , King JT Jr , Leslie D , Roberts MS . What does the value of modern medicine say about the$50,000 per quality-adjusted life-year decision rule
?
Med Care
2008
;
46
:
349
356
[PubMed]
30.
McQueen
RB
,
Ellis
SL
,
Campbell
JD
,
Nair
KV
,
Sullivan
PW
.
Cost-effectiveness of continuous glucose monitoring and intensive insulin therapy for type 1 diabetes
.
Cost Eff Resour Alloc
2011
;
9
:
13
[PubMed]
31.
Roze
S
,
de Portu
S
,
Smith-Palmer
J
,
Delbaere
A
,
Valentine
W
,
Ridderstråle
M
.
Cost-effectiveness of sensor-augmented pump therapy versus standard insulin pump therapy in patients with type 1 diabetes in Denmark
.
Diabetes Res Clin Pract
2017
;
128
:
6
14
[PubMed]
32.
Kamble
S
,
Schulman
KA
,
Reed
SD
.
Cost-effectiveness of sensor-augmented pump therapy in adults with type 1 diabetes in the United States
.
Value Health
2012
;
15
:
632
638
[PubMed]
33.
Pickup
JC
,
Freeman
SC
,
Sutton
AJ
.
Glycaemic control in type 1 diabetes during real time continuous glucose monitoring compared with self monitoring of blood glucose: meta-analysis of randomised controlled trials using individual patient data
.
BMJ
2011
;
343
:
d3805
[PubMed]
34.
Henriksson
M
,
Jindal
R
,
Sternhufvud
C
,
Bergenheim
K
,
E
,
Willis
M
.
A systematic review of cost-effectiveness models in type 1 diabetes mellitus
.
Pharmacoeconomics
2016
;
34
:
569
585
[PubMed]
35.
Foos
V
,
Varol
N
,
Curtis
BH
, et al
.
Economic impact of severe and non-severe hypoglycemia in patients with Type 1 and Type 2 diabetes in the United States
.
J Med Econ
2015
;
18
:
420
432
[PubMed]
36.
Parekh
WA
,
Ashley
D
,
Chubb
B
,
Gillies
H
,
Evans
M
.
Approach to assessing the economic impact of insulin-related hypoglycaemia using the novel Local Impact of Hypoglycaemia Tool
.
Diabet Med
2015
;
32
:
1156
1166
[PubMed]
37.
Fiallo-Scharer
R
,
Cheng
J
,
Beck
RW
, et al.;
Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group
.
Factors predictive of severe hypoglycemia in type 1 diabetes: analysis from the Juvenile Diabetes Research Foundation continuous glucose monitoring randomized control trial dataset
.
Diabetes Care
2011
;
34
:
586
590
[PubMed]
38.
Harris
S
,
Mamdani
M
,
Galbo-Jørgensen
CB
,
Bøgelund
M
,
Gundgaard
J
,
Groleau
D
.
The effect of hypoglycemia on health-related quality of life: Canadian results from a multinational time trade-off survey
.
Can J Diabetes
2014
;
38
:
45
52
[PubMed]
39.
Currie
CJ
,
Poole
CD
,
Woehl
A
, et al
.
The health-related utility and health-related quality of life of hospital-treated subjects with type 1 or type 2 diabetes with particular reference to differing severity of peripheral neuropathy
.
Diabetologia
2006
;
49
:
2272
2280
[PubMed]
40.
Polonsky
WH
,
Hessler
D
,
Ruedy
KJ
,
Beck
RW
;
DIAMOND Study Group
.
The impact of continuous glucose monitoring on markers of quality of life in adults with type 1 diabetes: further findings from the DIAMOND randomized clinical trial
.
Diabetes Care
2017
;
40
:
736
741
[PubMed]