The Effectiveness of an App (Insulia) in Recommending Basal Insulin Doses for French Patients With Type 2 Diabetes Mellitus: Longitudinal Observational Study

**Background:** For patients with type 2 diabetes (T2D), achieving glycemic targets remains challenging, particularly when titrating basal insulin. Insulia is a digital solution combining a smartphone app that uses clinical algorithms to recommend basal insulin doses and a web portal for healthcare professionals to remotely manage treatment. A predecessor device (Diabeo-Basal) was evaluated in the TeleDiab-2 randomized controlled trial, which showed that patients using the app had significantly greater HbA1c reduction (–1.48% vs –0.92% in standard care, p=0.002) and twice as many achieved target fasting blood glucose (FBG), without severe hypoglycemia. Following its availability in France through the ETAPES program, this study aimed to analyze real-world glycemic outcomes among Insulia users.

**Methods:** This was a retrospective observational analysis of data collected through the Insulia device in adult French patients with T2D treated with basal insulin and oral antihyperglycemic agents who used the system for ≥6 months by September 30, 2021. A subpopulation of "regular and compliant users" was defined as those with ≥5 dose calculations per week on average and >80% of injected doses corresponding to recommended doses over the 6-month period. The primary glycemic outcome was the percentage of patients achieving their individualized FBG target (±5%) between 5.5 and 6 months after initiation. Secondary outcomes included the frequency of hypoglycemia (blood glucose <70 mg/dL) resulting in treatment change over 6 months and the evolution of mean HbA1c over the same period (±1.5 months). A logistic regression model was used to identify factors associated with achieving the FBG target among regular and compliant users. A subgroup analysis was conducted for the center with the most patients (Centre Hospitalier Sud Francilien, CHSF) to assess a possible center effect.

**Key Results:** Of 484 enrolled patients, 373 (77.1%) performed at least one dose calculation and were included in the main analysis. Among these, 91 (24.4%) were identified as regular and compliant users over 6 months. Baseline characteristics: mean age 55.8 (SD 11.9) years, 59.2% male, mean BMI 30.6 (SD 5.9) kg/m², mean HbA1c 10.1% (SD 2.0%), mean first calculated basal insulin dose 25.5 (SD 15.8) IU/day. Regular and compliant users had a slightly lower baseline HbA1c (9.6% vs 10.3%, p=0.002) and more stringent FBG targets (39.6% had target 80-120 mg/dL vs 27.0% in other patients). Over a median use duration of 5.0 (95% CI 3.8-5.7) months, patients used the system 5.8 (SD 1.6) times per week on average, and 73.4% of injected doses were consistent with app suggestions. The proportion of patients achieving their FBG target at 6 months was 60% (55/91) among regular and compliant users vs 51.5% (145/282) among other patients (p=0.15). A trend of +1.86% every 2 weeks in target achievement was observed for regular and compliant users, with no clear improvement in other patients. The logistic model (including age, gender, BMI, HbA1c, and insulin dose at initiation) did not identify any significant predictors of target achievement among regular and compliant users. Among 82 patients with paired HbA1c data, mean HbA1c decreased from 9.9% to 7.2% at 6 months (±1.5 months). The incidence of reported hypoglycemia (BG <70 mg/dL) decreased over time (–0.16% per month), though not statistically significant. The subgroup analysis at CHSF (68 patients, 18.3% of the cohort) showed similar baseline characteristics except higher mean HbA1c (10.9% vs 10.0%) and less strict FBG targets; CHSF patients were less often regular and compliant users (13.2% vs 26.9%, p=0.02), but no center effect on FBG target achievement was observed (p=0.77 overall, p=0.75 for regular and compliant users).

**Clinical Implications:** This real-world study demonstrates that the Insulia app can improve glycemic control in patients with T2D, particularly among those who use it regularly and adhere to dose recommendations. The observed HbA1c reduction from 9.9% to 7.2% over 6 months is clinically meaningful, though less pronounced than in the controlled TeleDiab-2 trial, likely due to the lack of sustained human support in routine practice. The high rate of patient disengagement (only 24.4% remained regular and compliant users at 6 months) highlights a critical barrier to effectiveness. Clinicians should consider providing ongoing education and support to maximize adherence when prescribing digital titration tools. The absence of increased hypoglycemia is reassuring. These findings support the use of algorithm-based insulin dose recommendation apps as an adjunct to standard care, but larger, pragmatic studies are needed to confirm long-term benefits and to identify strategies to improve patient engagement.