Diabetes Technology & Therapeutics

Advanced hybrid closed-loop (AHCL) automated insulin delivery systems such as the MiniMed™ 780G have been shown to result in substantial improvements in disease management in people living with type 1 diabetes. The aim of the analysis was to assess the cost utility of the MiniMed 780G system compared with intermittently scanned continuous glucose monitoring (is-CGM) and multiple daily insulin injections (MDI) in people living with type 1 diabetes in France, to estimate the incremental cost-utility ratio (ICUR) and inform decision-making.

The analysis was performed using the CORE Diabetes Model (version 9.5) and clinical input data were sourced from a randomized controlled trial, with glycated hemoglobin reductions of 1.54% (16.8 mmol/mol) and 0.2% (2.18 mmol/mol) assumed for the MiniMed 780G arm and is-CGM + MDI arm, respectively. The analysis was conducted from a national payer perspective over a 40-year time horizon; future costs and clinical outcomes were discounted at 2.5% per annum.

In the base case analysis, use of the MiniMed 780G system was associated with a mean gain in quality-adjusted life expectancy of 2.26 quality-adjusted life years (QALYs) compared with is-CGM + MDI (16.33 QALYs vs. 14.07 QALYs), while mean direct lifetime costs were EUR 78,509 higher (EUR 215,037 vs. EUR 136,528), resulting in an ICUR of EUR 34,732 per QALY gained. Findings from sensitivity analyses showed that analyses were robust to changes in assumptions in most input parameters.

In people with type 1 diabetes in France not achieving glycemic target levels at baseline, the use of the MiniMed 780G system was projected to lead to substantial improvements in quality-adjusted life expectancy compared with continued use of is-CGM + MDI, with an ICUR of EUR 34,732 per QALY gained.

Although use of continuous glucose monitoring (CGM) has been linked with improved glucose control, including reductions in hemoglobin A1c and episodes of hypoglycemia, there has been little investigation of its possible role in reducing other serious clinical events.

To estimate the effect of starting CGM in patients with type 2 diabetes (T2D) on mortality.

A cohort study comparing mortality between propensity score-matched CGM users and non-CGM users over 18 months.

Veterans Affairs Health Care System.

Adult patients with T2D receiving insulin who were identified as CGM users or non-CGM users between January 1, 2015, and December 31, 2020.

Primary outcome of all-cause mortality; secondary outcomes of serious all-cause hospitalization, cardiovascular events, and admissions related to hyperglycemia and hypoglycemia.

A total of 12,729 patients with T2D (94% male with mean age 66) who were new CGM users were 1:1 matched with non-CGM users. Total follow-up time was 17,676 and 17,034 person-years for CGM and non-CGM users. Risk for mortality was lower in CGM users (hazard ratio or HR 0.79: 95% confidence interval or CI 0.73–0.86), as were risks for all-cause hospitalization (0.91: 0.86, 0.96), cardiovascular events (0.84: 0.73, 0.96), and admissions for hyperglycemia (0.88: 0.81, 0.95). Lower risk for mortality persisted after accounting for early deaths, COVID-19, recent onset of diabetes, subsequent use of insulin pumps or newer diabetes medications, or when stratifying by frequency of CGM use, frailty index or mortality risk (all HRs: 0.83 or less, range of CI: 0.60–0.94). No differences between CGM and non-CGM users were seen with negative control outcomes.

Unmeasured health factors, behaviors, or other confounders may exist.

In a large national cohort, initiation of CGM was associated with lower mortality in T2D patients using insulin and indicates use of CGM may have benefits that extend beyond glucose lowering.

This study aims to evaluate the accuracy of continuous glucose monitoring (CGM)-derived metrics, particularly those related to glycemic variability, in the presence of missing data. It systematically examines the effects of different missing data patterns and imputation strategies on both standard glycemic metrics and complex variability metrics.

The analysis modeled and compared the effects of three types of missing data patterns—missing completely at random, segmental, and block-wise gaps—with proportions ranging from 5% to 50% on CGM metrics derived from 14-day profiles of individuals with type 1 and type 2 diabetes. Six imputation strategies were assessed: data removal, linear interpolation, mean imputation, piecewise cubic Hermite interpolation, temporal alignment imputation, and random forest-based imputation.

A total of 933 14-day CGM profiles from 468 individuals with diabetes were analyzed. Across all metrics, the coefficient of determination (R ²) improved as the proportion of missing data decreased, regardless of the missing data pattern. The impact of missing data on the agreement between imputed and reference metrics varied depending on the missing data pattern. To achieve high accuracy (R ² > 0.95) in representing true metrics, at least 70% of the CGM data were required. While no imputation strategy fully compensated for high levels of missing data, simple removal outperformed others in most scenarios.

This study examines the impact of missing data and imputation strategies on CGM-derived metrics. The findings suggest that while missing data may have varying effects depending on the metric and imputation method, removing periods without data is a general acceptable approach.

To examine the risk and protective factors for severe and recurrent diabetic ketoacidosis (DKA) in a large sample of children in the Southwestern United States.

Retrospective chart review of children age 0–18 years with type 1 diabetes (T1D) seen at a large children’s hospital/integrated care delivery system between October 2019 and December 2022. Data from the preceding 2 years were used to predict postdiagnosis DKA in each subsequent year. Logistic regression and recursive feature elimination (RFE) were used to select significant predictors of any DKA, severe DKA, and recurrent DKA. Model performance was evaluated using fivefold cross-validation, with area under the curve in the receiver operating characteristic plot as the performance metric.

Records were obtained for 4649 encounters, representing 1850 patients and 846 prior DKA events. Based on RFE, single prior DKA, recurrent prior DKA, and hemoglobin A1c were significant shared predictors for subsequent DKA, severe DKA, and recurrent DKA, and female sex was positively associated with any DKA and recurrent DKA. The model for recurrent DKA also included age between 10 and 14 years as an unshared risk factor, and Hispanic ethnicity and use of an insulin pump (with or without automated insulin delivery) as unshared protective factors. Incidence of severe DKA was highly correlated (r = 0.95) with number of prior DKA events. Black and female patients were more likely to experience multiple recurrent DKA episodes and repeated episodes of severe DKA.

Severe and recurrent DKA have both shared and unshared risk factors. Severe DKA may be a singular phenomenon in most cases, although a subset of patients (primarily Black and female) experience repeated severe events, placing them at high risk for adverse health outcomes. Recurrent DKA appears to be more of a chronic issue, although a number of variables emerged as protective factors, suggesting ways in which recurrent DKA might be prevented.

Immune therapies such as teplizumab and antithymocyte globulin (ATG) offer promise in delaying type 1 diabetes (T1D). However, growing availability of automated insulin delivery (AID) systems for insulin management may alter the cost-effectiveness of these therapies. Immune therapies may become more cost-effective when paired with AID instead of conventional insulin management. Meanwhile, as immune therapies delay T1D for only a short period, effective AID may reduce the economic value of prevention. This study provides the first cost-effectiveness analysis of the interplay between immune therapies and AID systems.

Using microsimulation modeling, we examined the cost-effectiveness of six alternative prevention-treatment strategies defined by a combination of three preventive immune therapies (teplizumab, ATG, or no therapy) and two insulin management strategies (AID or conventional insulin management). Effectiveness was measured by quality-adjusted life years (QALYs). Costs were estimated from a payer perspective.

Among the six strategies considered, preventive ATG therapy followed by AID was the most cost-effective. It entailed $394,250 in lifetime costs and yielded 19.13 QALYs. These costs were lower and QALY gains higher than those with strategies that did not involve immune therapy or AID. Preventive teplizumab therapy followed by AID generated 0.25 more QALYs than ATG therapy followed by AID, albeit at an additional cost of $153,670, resulting in an incremental cost-effectiveness ratio of $369,890/QALY.

Preventive ATG therapy followed by AID after T1D onset can be a potentially cost-effective approach. In the absence of randomized clinical trials for ATG in the prevention space, findings in this study assume that ATG is at least half as efficacious as teplizumab. The optimal prevention-treatment strategy will ultimately depend on payers’ ability to negotiate prices for teplizumab and further evidence on efficacy of ATG in preventing T1D.

To evaluate longitudinal real-world outcomes in adults with type 1 diabetes initiating hybrid closed loop (HCL).

Adults with type 1 diabetes, managed with an insulin pump and intermittently scanned continuous glucose monitoring with hemoglobin A1c (HbA1c) ≥8.5% (69 mmol/mol), were started on HCL between August and December 2021 as part of the National Health Service England HCL pilot. We collected outcomes, including change in HbA1c, sensor glucometrics, Gold score (hypoglycemia awareness), diabetes distress score, acute event rates, and user opinion of HCL.

In total, 420 HCL users across 30 diabetes centers in the United Kingdom were included (median age 40 [interquartile range or IQR 29–50] years, 68% female, 85% White British). Over a median follow-up of 12 months (IQR 8–28) (range 6–38 months), mean adjusted HbA1c reduced by 1.4% (95% confidence interval [CI] −1.5, −1.3; P < 0.001) (16 mmol/mol [95% CI −17, −14]; P < 0.001). Time in range (70–180mg/dL) increased from 33.7% to 60.4% (P < 0.001). The proportion of individuals achieving HbA1c ≤7.5% (58 mmol/mol) increased from 0% to 33.1% (P < 0.001). Diabetes distress score reduced (−1.1; 95% CI −1.3, −1.0; P < 0.001) and Gold score reduced (−0.4; 95% CI −0.5, −0.2; P < 0.001). The percentage of individuals with impaired hypoglycemia awareness (Gold score ≥4) decreased (16.6% [baseline] vs. 9.2% [follow-up]; P < 0.001). Almost all participants stated that HCL had a positive impact on quality of life (94.5%; 361/382). The number of hospital admissions was low.

Long-term real-world use of HCL is associated with sustained improvements in glycemic and person-reported outcomes in adults with type 1 diabetes and above-target HbA1c levels.

This study addressed the challenge of postprandial glycemic variability in type 1 diabetes (T1D), even with AID (automated insulin delivery). We evaluated the effectiveness of a non-carbohydrate counting (non-CC) meal bolus strategy in adults with T1D utilizing open-source AID.

A total of 32 adults with T1D, aged 18 to 50 years, participated in a randomized crossover trial utilizing open-source AID. Following a 7-day run-in period, participants were randomly assigned to one of two groups: automatic mode (closed loop) or manual mode (open loop). After 2 weeks, the participants underwent a crossover to the alternate treatment mode. Prandial boluses were administered according to a sliding scale based on preprandial glucose levels, without utilizing either the exact carbohydrate content of meals or meal announcement buttons. The study compared the differences in time in range (TIR) and insulin dosage across the different phases.

Compared with the open-loop phase, the TIR for patients during the closed-loop phase increased significantly during the night (75.45% ± 12.08% vs. 83.05% ± 7.20%, P < 0.001) and 24 h (73.40% ± 9.98% vs. 79.21% ± 4.84%, P = 0.019), with a more pronounced effect observed at night. During the closed-loop phase, the frequency of 24-h hypoglycemic events (＜3.9 mmol/L) was reduced compared with the open-loop phase, with no difference in nocturnal hypoglycemic events. In addition, compared with the open-loop phase, there were no significant differences in average postprandial blood glucose and peak blood glucose levels during the closed-loop phase; however, the time to reach peak postprandial blood glucose was delayed (86.06 ± 20.80 min vs. 99.08 ± 15.05 min, P < 0.001).

A non-CC meal bolus strategy based on preprandial glucose in adults with T1D utilizing open-source AID effectively prevents glycemic excursions and maintains a mean TIR over 70%.

All continuous glucose monitors (CGMs) have an upper detection limit, typically of 22.2 mmol/L. This might bias CGM metrics. We aimed to develop and validate a statistical model for imputing values above this limit.

We analyzed CGM data from 85 inpatients with type 2 diabetes, 705 outpatients with type 1 diabetes, and 27 outpatients with type 2 diabetes. A Bayesian nonparametric latent Gaussian process regression model was applied to the CGM data intentionally right censored for the top 5%, 10%, 20%, and 30% and compared with the uncensored CGM data by the bias, mean squared error (MSE), and coefficient of determination (R ²) of mean glucose, standard deviation (SD), and coefficient of variation (CV).

In hospitalized patients with diabetes, outpatients with type 1 diabetes, and outpatients with type 2 diabetes for 5% to 30% right censoring, respectively, the bias on mean glucose after imputation ranged from −0.012 to 0.362, −0.018 to 0.485, and −0.008 to 0.130, respectively. Bias on SD ranged from −0.024 to 0.226, −0.033 to 0.381, and −0.016 to 0.138, respectively. Bias on CV ranged from −0.207 to 1.543, −0.316 to 2.609, and −0.222 to 1.721, respectively. Similar results indicating good performance of the imputation model were observed for MSE and R ².

An imputation model for glucose values above the upper detection limit of CGMs was developed and validated in various populations. This enables a more unbiased quantification of CGM metrics for patients with severe hyperglycemia.

Achieving optimal glycemic control in patients with type 1 diabetes mellitus (PwT1DM) is essential to prevent complications. Continuous subcutaneous insulin infusion (CSII) systems combined with continuous glucose monitoring (CGM) have improved outcomes, but the effectiveness of additional technologies, such as mobile apps and hybrid closed-loop systems (HCLSs), remains unclear. This study evaluates glycemic control and quality of life (QoL) in adult PwT1DM transitioning from multiple daily injections (MDI) to a CSII, first with the Mylife™ Dose app and subsequently switching to an HCLS.

This was a 10-month, multicenter, open-label sequential study involving 135 adults with type 1 diabetes (T1D), all of whom were using isCGM and MDI before transitioning to CSII, first with the Mylife Dose app and later to an HCLS. Glycemic control (glycated hemoglobin [HbA1c], time in range [TIR], time below range, time above range), insulin requirements, and QoL/treatment satisfaction/hypoglycemia perception (Diabetes Quality of Life questionnaire, Diabetes Treatment Satisfaction Questionnaire, Clarke’s test) were measured at each of the four study visits.

Transitioning from MDI to CSII showed modest improvements in HbA1c (7.57%–7.42%; P = 0.02) and TIR (56.3%–60.4%; P < 0.01). The introduction of the Mylife Dose app did not provide significant additional improvements in glycemic control or QoL, although it provided an additional tool for diabetes management. However, switching to an HCLS resulted in substantial improvements in HbA1c (6.7%), TIR (73.1%), and QoL, with over 70% of patients achieving an HbA1c <7%. Insulin requirements increased slightly with the HCLS, primarily due to basal insulin. Adherence was high, with 88.1% completing the study.

The Mylife Dose app does not improve glycemic control or QoL significantly but offers convenience for patients with T1D. In contrast, HCLSs provide significant metabolic and QoL benefits, supporting their integration into T1D management with appropriate reimbursement policies.

The root mean squared error (RMSE) is commonly used to evaluate blood glucose prediction algorithms. However, it primarily measures how well predictions align with the most likely future values, rather than supporting optimal and proactive treatment decisions. Since diabetes management data predominantly features blood glucose values within the target range, RMSE tends to favor models that consistently predict target-range values, often at the expense of detecting clinically critical events such as rapid fluctuations, hypoglycemia, or hyperglycemia. This study examines how and why RMSE biases evaluations toward trivial models, highlighting the need for alternative performance criteria that better reflect clinical priorities.

We developed the composite glucose prediction metric (CGPM) to integrate three components: RMSE, temporal gain and geometric mean (glycemic event prediction). A custom loss function was designed to emphasize clinically critical predictions during model training. Pareto frontier analysis was used to assess trade-offs among models with comparable performance.

CGPM was computed for five blood glucose prediction techniques (zero-order hold, naïve linear regression, ridge regression, ridge regression trained with a custom loss function, and a physiology-based model) applied to the OhioT1DM dataset. The data-driven model with the lowest RMSE performed poorly on glycemic event prediction, highlighting RMSE’s bias toward target-range predictions. In contrast, the ridge regressor trained with the custom loss function improved event prediction, showing that clinically weighted optimization mitigates biases.

Blood glucose prediction algorithms require evaluation and optimization criteria beyond accuracy to better support optimal treatment decisions. This study introduced the CGPM as an alternative evaluation framework, along with a loss function designed for model optimization that emphasizes clinically critical but rare events. Further clinical validation is needed to refine these criteria and ensure they align more closely with the needs of diabetes management.