Objective: To evaluate the completeness of methodological reporting in continuous glucose monitoring (CGM)-based clinical trials, and based on our findings, to propose a standardized reporting framework to improve reproducibility and transparency. Research Design and Methods: We conducted a search to assemble a study sample of CGM-based clinical trials published between 2019 and 2024. Articles were identified via PubMed and screened based on predefined eligibility criteria. We measured the presence or absence of reporting elements across seven core methodological features essential for reproducibility, spanning CGM device characteristics, data platforms, data extraction processes, preprocessing and transformation methods, handling of missing data, CGM metrics, and analytic approaches. Data were manually extracted by trained reviewers, with quality assurance performed on a subset of records. Descriptive statistics were used to summarize reporting completeness. Results: A total of 248 studies met inclusion criteria. On average, studies reported 4.39 (SD 1.22) of the seven core methodological features. Reporting was highest for CGM metrics (100%), analytic methods (100%), and device make/model (96%). In contrast, key elements of the CGM data pipeline were frequently underreported, including data extraction processes (56%), preprocessing methods (43%), handling of missing data (32%), and CGM data platform (30%). Only 6% of studies reported all seven features. These findings demonstrate substantial heterogeneity and consistent gaps in reporting practices across CGM clinical research within the literature sampling frame. Conclusions: Incomplete reporting of CGM data pipelines limits reproducibility, comparability, and the clinical and computational utility of CGM-derived outcomes. We propose a preliminary, candidate CGM-specific reporting checklist to standardize documentation of data acquisition, processing, and analysis. Adoption of a refined, consensus-based version of such a framework may strengthen methodological rigor, support regulatory and clinical interpretation, and enhance the reliability of CGM data for emerging digital health and artificial intelligence applications.
Background: Automated insulin delivery (AID) systems are recommended for individuals with diabetes requiring insulin. Real-world effectiveness may differ from outcomes reported in prospective clinical trials. The twiist™ AID system recently received regulatory clearance, and real-world glycemic outcomes have not been previously described. Methods: This was an ongoing retrospective, observational real-world study with cross-sectional analyses of participant-level summary data. The analysis included individuals with type 1 diabetes (T1D) and type 2 diabetes (T2D) who initiated use of the twiist AID system and consented to data sharing. Deidentified continuous glucose monitoring (CGM) and insulin delivery data were analyzed. Inclusion in the analysis required ≥14 days of device use. Glycemic outcomes were assessed using standard CGM metrics, including percent time in range (%TIR; 70–180 mg/dL), time in tight range (%TITR; 70–140 mg/dL), time below range (%TBR; 54 to <70 and <54 mg/dL), and time above range (%TAR; >180–250 and >250 mg/dL). Subgroup analyses evaluated associations with baseline glycated hemoglobin (HbA1c), age, and bolus frequency. Results: Data from 4000 individuals were included. The duration of device use ranged from a minimum of 14 days to a maximum of 365 days. Mean overall %TIR was 73.5 ± 14.5%, with overnight %TIR of 77.9 ± 16.5%. Hypoglycemia exposure was low, with mean overall %TBR of 54 to <70 mg/dL of 1.9 ± 2.0% and <54 mg/dL of 0.2 ± 0.4%. Glycemic outcomes were generally within recommended targets across age categories except for %TIR among those <18 years of age. Among those with baseline HbA1c values and TIR values ( n = 1131), %TIR was 81.5 ± 10.2% and 71.1 ± 11.4%, and 61.7 ± 16.3% for those HbA1c <7%, 7%–8%, and >8%, respectively. More frequent boluses were associated with improved glycemic outcomes. Conclusions: In this large real-world cohort, use of the twiist AID system was associated with effective overall and overnight glycemic management and minimal hypoglycemia.
Inhaled insulin provides a unique prandial insulin option, offering an alternative, with a more rapid onset of effect, to subcutaneous rapid-acting analog insulin. Clinical studies with inhaled insulin have demonstrated its efficacy and safety, with glycemic outcomes comparable to or in some cases better than those of contemporary insulin delivery approaches. However, the use of inhaled insulin for diabetes therapy in daily life remains limited due to payer restrictions, prior authorization requirements, formulary exclusions, clinician knowledge gaps, clinician time constraints, persistent misconceptions regarding safety, and inadequate patient awareness. A panel of clinical experts in diabetes management met in June 2026 to review the safety and efficacy of Technosphere ® inhaled (TI) insulin, discuss the current obstacles that limit access for people with diabetes (PwD) to TI, identify practical strategies to overcome these obstacles, and issue a call-to-action for payers, clinicians, policymakers, professional organizations, and patient advocacy groups. This article summarizes those discussions and proposes strategies to increase access and choice for PwD.
With the growing availability of over-the-counter continuous glucose monitors, the question arises whether Continuous Glucose Monitoring (CGM) can be used to diagnose prediabetes. Traditionally, prediabetes is diagnosed using the oral Glucose Tolerance Test, a test with poor reproducibility. In a study from Singapore, there was large overlap between Ambulatory Glucose Profiles of people with prediabetes and normoglycemia. The optimal threshold of 5.64% Time above 140 mg/dL, the best discriminator, yielded 88% specificity and 48% sensitivity, indicating insufficient discriminatory value in many cases. In a study from Stanford, administration of a 75-gram oral glucose load while wearing CGM showed poor agreement with venous sampling between the category outcomes normoglycemia, prediabetes and diabetes. Future large cohort studies and AI may improve this situation. But for the moment, when a question of prediabetes arises, the most practical answer is to measure HbA1c, which allows for an elegant classification of normoglycemia, prediabetes and diabetes.
