PBPierre-Yves Benhamou

Pierre-Yves Benhamou

*This is an algorithmically generated author record based on publicly available information.

Pierre-Yves Benhamou has a total of 610 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Towards a digital twin for precision diabetes care: a phenotypic mapping and virtual representation of a person living with type 1 diabetes in the SFDT1 study

    by Mariana Oliveira, Gloria Aguayo, Dulce Canha, Pierre Yves Benhamou, Claire Briet, Catherine Campinos Imberti, Emmanuel Disse, Samy Hadjadj, Hélène Hanaire, Nathalie Jeandidier, et al.

  • Open Access

    Impact of DBLG1 Hybrid Closed‐Loop Software Optimization and Meal Declaration Pattern on Glucose Metrics in Type 1 Diabetes: A 12‐Month Prospective Evaluation in 2014 Users

    by Pierre‐Yves Benhamou, Aurélien Vesin, Lucile Reynaud, Thibault Le Roux‐Mallouf, Benjamin Chatel

    Background: The meal period is critical for the outcome of automated insulin delivery. Our aim was to explore the long‐term impact on glucose metrics of algorithmic modifications of the DBLG1 hybrid closed‐loop system, with special emphasis on the meal declaration pattern. Methods: We prospectively analysed CGM data from consenting adult patients with Type 1 diabetes across Europe who initiated the new version of DBLG1 paired with the Kaleido patch pump upon its release. Inclusion occurred over 7 months, with follow‐up extended up to 12 months. Time in range 70–180 mg/dL (TIR) was the primary endpoint. Utility endpoints included the number of declared meals. Results: This study included 2014 users, age 42.5 ± 13.8 years, baseline HbA1c 7.6% ± 1.2%. Median TIR was 72.0%, 72.3%, 72.0% and 71.7% at M3, M6, M9 and M12, respectively. At M3, time < 54 mg/dL (0.2%), < 70 mg/dL (1.2%) and > 250 mg/dL (5.4%) were controlled and remained stable throughout M12, as well as GMI (7.0%–7.1%). The incidence rates of severe hypo‐ or hyperglycaemia remained very low. TIR in patients declaring ≥ 3 meals/day exceeded TIR in those declaring < 1 meal/day (73.3 [67.3, 78.7]% vs. 65.7 [60.7, 72.2]%, p < 0.0001). Conclusion: This study demonstrates the sustained safety and effectiveness of the latest version of the DBLG1 software, with TIR exceeding 70% throughout the 12‐month follow‐up period in a large group of 2014 users. Furthermore, while glycaemic control was negatively affected when meal declaration was reduced, the software maintained its safety and acceptable effectiveness in this latter situation.

  • Top Cited Papers
    Open Access

    Voice changes associated with stress and distress in people living with diabetes: Results from the Colive Voice study

    by Charline Bour, Abir Elbéji, Dulce Canha, Noémie Topalian, Hanin Ayadi, Christelle Godin, Claire Guyon-Gardeux, Pierre-Yves Benhamou, Malvina Billères, Guy Fagherazzi

    Stress and diabetes distress impact glycemic control, self-care, and health outcomes in people living with diabetes, but current diabetes technologies lack easy-to-use, non-invasive methods to detect these states. The objective was to identify vocal features associated with stress and diabetes distress in people living with diabetes. Thus, we analyzed data from 679 adults with diabetes (381 women, 298 men), recruited via the global vocal biomarker screening platform Colive Voice . Participants recorded a standardized 30-second text. Vocal features were extracted using DisVoice across phonation, prosody, articulation, and phonological domains. Associations between vocal features and stress (self-reported, scale: 0–4) and diabetes distress (Problem Areas in Diabetes (PAID) questionnaire; categories: PAID<20, 20 ≤ PAID<40, 40 ≤ PAID<60, PAID≥60) were analyzed separately for men and women. Multivariate ordinal logistic regression models were adjusted for age, language, diabetes type, and HbA1c, with False Discovery Rate correction.In women, results suggest that stress was associated with 25 voice features across all domains, while distress was linked to one articulatory feature, suggesting broader effects of stress. In men, both stress (25 features) and distress (18 features) affected multiple domains, but with distinct acoustic patterns. No voice features were shared between stress and distress. These patterns were consistent across age, diabetes type, and language.To conclude, voice changes reflect stress and diabetes distress in people with diabetes, each characterized by distinct acoustic features and additional sex-specific signatures. These findings highlight the potential of vocal biomarkers to differentiate and monitor these emotional states, supporting more timely and personalized interventions to improve diabetes outcomes.


  • Open Access

    How to Demonstrate the Glucose Specificity of a Non-Invasive CGM: A Case Study of the SKAMo-2 Clinical Trial and Neogly™

    by Romain Blanc, Pierre Blandin, Jean-Guillaume Coutard, Kevin Jourde, Helene Marie, Pierre-Yves Benhamou

    Background: Every non-invasive continuous glucose monitoring (NI-CGM) technology introduced into the landscape faces the same skeptical question, from regulators, clinicians, and competing developers alike: is the candidate signal actually specific to glucose, or does an apparently reasonable accuracy figure simply reflect a model fitting to motion, temperature, calibration offset, or trial-duration artifact? Existing evaluation practice does not answer this question directly. NI-CGM performance is instead reported almost exclusively with metrics inherited from minimally invasive, subcutaneous CGM, the Mean Absolute Relative Difference (MARD), Clarke/Parkes error grids, and ISO 15197-style agreement rates, which were designed for sensors whose glucose specificity is already chemically established and which therefore take specificity as a premise rather than treating it as a result to be demonstrated. Methods: We present a methodology for demonstrating NI-CGM technology glucose specificity during the algorithm-development phase, and illustrate it with a case study based on a quantum-cascade-laser (QCL) photoacoustic NI-CGM device (Neogly™) evaluated in the SKAMo-2 free-living clinical trial (eight participants with type 1 diabetes). The methodology combines a white-noise control, a constant-glycemia control, a sensor-ablation control that removes the candidate physical signal while retaining auxiliary covariates, and explicit reporting of the train/test generalization level, so that a reported MARD can be read as evidence of specificity rather than taken on faith. Results: Removing the mid-infrared photoacoustic (PA) signal from the model while retaining all auxiliary sensors (accelerometer, skin temperature, hygrometry, PPG) degraded performance at every generalization level tested, inter-patient MARD rose from 35.0% with the PA signal to 43.1% without it, and intra-experimentation MARD rose from 22.5% to 23.9%, providing direct, internal evidence that the PA channel itself, and not merely the auxiliary covariates, carries glucose-specific information. At the same time, an algorithm trained on pure Gaussian noise produced a MARD of 25% over short test windows, and a trivial constant-glycemia predictor outperformed every machine-learning model tested when generalization was extended from a single recording to an unseen patient (MARD 55% for the naive constant model versus 37% for a deep neural network on inter-patient splits). Reported in isolation, any of these MARD values is uninterpretable; reported against one another, they jointly demonstrate that the signal is specific to glucose while also bounding how much of the headline accuracy figure that specificity currently explains. Conclusions: We propose a specificity-demonstration methodology for NI-CGM technology development, comprising (1) signal quality gating prior to any algorithm benchmarking, (2) a white-noise control to test for genuine information content, (3) a constant-glycemia control to expose trial-duration bias, (4) a sensor-ablation control that isolates the contribution of the candidate physical signal from auxiliary covariates, (5) explicit reporting of the data-splitting generalization level (intra-experimentation, intra-patient, inter-patient). This methodology answers a question that precedes clinical accuracy reporting and that recognized clinical frameworks such as the IFCC Working Group on CGM's Dynamic Glucose Regions guideline are not designed to answer: not how accurate is the device, but is the device measuring glucose at all. We argue that without these controls, MARD and error-grid values for NI-CGM are not comparable across studies and may either overstate clinical readiness or undermine promising technologies. We recommend that this specificity methodology be applied routinely once a candidate NI-CGM sensor reaches algorithm-development stage, alongside and as a deliberate complement to IFCC-style clinical accuracy reporting once the device is mature enough for that evaluation. Keywords: non-invasive continuous glucose monitoring; glucose specificity; algorithm validation; MARD; benchmarking; machine learning; photoacoustic spectroscopy; sensor ablation; Clarke error grid

  • Economic Burden of Type 1 Diabetes in France

    by Emmanuel Cosson, Corinne Emery, Elodie Torreton, Bruno Detournay, Bruno Verges, Pierre‐Yves Benhamou, Maxime Brule, Laura Sablone, Guy Fagherazzi, Jean Pierre Riveline

    Aim: To estimate the societal costs of Type 1 diabetes (T1D) in France, including direct and indirect costs. Materials and Methods: We conducted a descriptive retrospective cross‐sectional study on all adult patients (aged 18 and over) with T1D identified in the SFDT1 cohort and compared them to a matched cohort of patients without diabetes considering age, gender, deprivation status and geographical area. An Insurance claim data analysis was conducted for the year 2023 individually and then extrapolated to all of France. Costs were broken down by expense categories and insulin delivery devices. Results: The average individual societal cost of T1D in France in 2023 compared with the absence of diabetes was estimated at €9940 (€9340 when considering direct costs only). Nearly half of the identified costs were attributable to the use of medical devices. An analysis stratifying patients by treatment modality (closed‐loop, pump or multiple daily injections) highlighted the significant costs associated with the use of the most advanced technologies for delivering insulin (average costs were respectively €13 557, €11 062, and €7004, respectively). The total cost of T1D in France was estimated at between €1.59 and €1.79 billion in 2023 excluding undiagnosed cases and premature mortality associated with the disease. Conclusions: Our results suggest that the most advanced insulin delivery devices have a significant short‐term economic impact. Although this paper does not cover the research, model‐based studies do suggest that there are some advantages to using such devices, including potential future savings due to lower complication rates.


  • Diabetes Care in 2050: A Human-Centered Road Map From Time in Range to Time in Happiness

    by Guy Fagherazzi, Emmanuel Cosson, Chantal Mathieu, Renza Scibilia, Kyle Jacques Rose, Kamil Armacky, Korey Hood, Bart Torbeyns, Elise Bismuth, Stéphane Besancon, et al.

    By 2050, diabetes care should transform into a human-centered, technology-augmented ecosystem focused on well-being. In this work, we first present a deliberately idealistic, forward-looking scenario in which advanced treatments, artificial intelligence, and integrated care delivery significantly reduce the burden on individuals with diabetes and enable person-centered health care. In this utopian vision, success in diabetes care is measured not only by glycemic metrics such as time in range but also by time in happiness, i.e., the quality time people spend unencumbered by their condition. Second, while such idealistic progress may be unlikely by 2050 due to socioeconomic and public health constraints, we build on existing gaps to outline a road map of co-designed solutions, health system reforms, and ethical innovations to move closer to it. This work serves as a “North Star” document that articulates the key values and next steps for achieving an ideal diabetes prevention and care ecosystem.


  • Open Access

    What Is a Clinically Meaningful Change in Diabetes Distress? Findings for Diabetes Care and Research From the SFDT1 Cohort

    by Dulce Canha, Jean‐Pierre Riveline, Fatima Kazanlieva, Gloria A. Aguayo, Charline Bour, Bruno Vergès, Pierre‐Yves Benhamou, Michaël Joubert, Fawaz Alzaid, Emmanuel Cosson, et al.

    Aims: Diabetes distress (DD) is common and evolves heterogeneously over time. We aimed to estimate minimal clinically important differences (MCID) for the Problem Areas in Diabetes (PAID) scale and its sub‐dimensions and to identify predictors of worsening over 1 year in people with type 1 diabetes (PwT1D). Materials and Methods: MCIDs for PAID total and sub‐dimension scores were derived using the standard error of measurement approach. One‐year DD worsening, stability and improvement were defined using the ±1 MCID threshold. We analysed changes in DD using data from the SFDT1 cohort. Logistic regression models identified predictors of worsening for PAID total and sub‐dimensions, adjusting for baseline DD, age, sex and social vulnerability. Results: We analysed data from 2457 adults with type 1 diabetes (51.6% female, 41.3 years old (SD 14.0), 23.7 years diabetes duration (SD 14.0)). The MCID for PAID total was 5.0 points, with sub‐scale MCIDs ranging from 8.5 (emotional distress) to 14.3 points (management distress). Over 1 year, 42% experienced clinically meaningful worsening of total DD, 30% improved and 28% remained stable. Worsening of total PAID was associated with social vulnerability, HbA1c increase and treatment burden. Distress trajectories varied across sub‐dimensions, with emotional distress and burnout most frequently worsening (34% each); initiation of automated insulin delivery was solely associated with lower odds of worsening on the burnout dimension. Conclusions: This study provides MCID values for the PAID scale and its sub‐dimensions. The high temporal variability observed suggests a need for regular monitoring and clinically informed assessment of DD in PwT1D.


  • Safety and glycemic outcomes of DBLG1 coupled to the DANA-i pump in adults and adolescents with type 1 diabetes mellitus. The SP8 trial

    by Sylvia Franc, Y. Reznik, C. Amadou, A. Penfornis, P. Schaepelynck, H. Hanaire, B. Delemer, C. Le Tallec, J. Beltrand, N. Jeandidier, et al.

    Purpose DBLG1 is a hybrid closed-loop (CL) system interoperable with ViCentra Kaleido and Roche Insight pumps. The SP8 trial investigated the safety and glycemic outcomes of DBLG1 coupled to the DANA-i pump (DBLG1-DANAi) in adults and adolescents with type 1 diabetes mellitus (T1D). Methods The safety of DBLG1-DANAi (primary outcome of SP8) was evaluated using two criteria: (i) the incidence of serious adverse device events (SADEs) and (ii) the time spent in hypoglycemia (time below range 70 mg/dL [3.9 mmol/L], TBR70) with DBLG1-DANAi compared to an SAP (sensor-augmented pump) system. Secondary outcomes included non-serious adverse events and standard glycemic metrics. Five months after its start, the trial was halted due to the COVID-19 crisis. One hundred and forty (140) of the 148 planned adults had already been included in the study, as well as all planned adolescents (N = 36). Results The intention-to-treat (ITT) population included 105 adult and 36 adolescent participants in whom no SADE was observed with DBLG1-DANAi. The per-protocol (PP) population consisted of 73 adults and 18 adolescents who spent two weeks with SAP, followed by two weeks with DBLG1-DANAi. No increase in TBR70 was observed during the DBLG1-DANAi period. Furthermore, the time spent in the glucose level range of 70–180 mg/dL [3.9–10.0 mmol/L] (time in range, TIR70-180) increased significantly with DBLG1-DANAi, both in adults (+ 10.5%; p < 0.0001) and adolescents (+ 16.0%; p < 0.0001). Conclusion During the two-week study period, DBLG1 was interoperable with the DANA-i pump. This result requires confirmation in future long-term studies. ClinicalTrials.gov ID NCT04190277

  • Impact of Algorithmic Modifications Targeting the Meal Period and the Management of Hypoglycemia and Hyperglycemia in Adult Persons With Type 1 Diabetes Using Closed-Loop Insulin Delivery: A Two-Step Observational Report

    by Pierre-Yves Benhamou, Aurélien Vesin, Lucile Reynaud, Benjamin Chatel

    Background: There is room for improvement in the outcome of automated insulin delivery. Our aim was to explore the impact on glucose metrics of algorithmic modifications of a DBLG1 hybrid closed-loop system, targeting the management of meal periods, hypoglycemia and hyperglycemia. Methods: We performed a two-step analysis of CGM data of all consenting adult patients with type 1 diabetes who were equipped in Europe with DBLG1 between November 1, 2023 and January 31, 2025, comparing three successive versions of the algorithm: v1.12 vs. v1.16 (first step, retrospective comparison), then v1.16 vs. v1.17 (second step, ambispective before/after analysis). Time in range of 70 to 180 mg/dL was the primary endpoint. Results: The first step (duration 319 days, 937 patients) compared 269 users of 1.12 version and 668 users of 1.16 version. Median TIR improved from 65.3% [IQR 58.4%-72.1%] to 71.3 [63.3%-78.0%]. Time in Tight Range 70 to 140 mg/dL increased from 37.5% [30.6%-43.1%] to 40.4% [31.7%-48.5%]. Time in Hypoglycemia was stable. Time >250 mg/dL decreased from 8.9% to 5.4%, GMI from 7.3% to 7.1%, CV from 30.4% to 27.4%, and GRI from 38.0 to 30.0. The second step (1212 patients, 120 days) showed a further improvement of TIR from 68.8% [59.6%-76.8%] to 70.8% [63.0%-77.6%] when upgrading from v1.16 to v1.17, with marginal changes in other glucose metrics. The incidence rates of severe hypoglycemia or hyperglycemia remained very low. Conclusion: This large post-market report illustrates the margin of improvement in AID performances through algorithmic refinements that improve the efficacy without deteriorating the safety of closed-loop insulin delivery.


  • A Retrospective Real-Life Study to Compare Glycemic Control Between Simplified Meal Size Estimation and Precise Carbohydrates Counting in Type 1 Diabetes Patients Using DBLG1 Hybrid Closed-Loop System

    by Julie Blervaque, Aurélien Vésin, Pierre-Yves Benhamou, Sandrine Lablanche

    Objective: Qualitative meal size estimation (QMSE) could be an interesting alternative to precise carbohydrates counting (PCC) for patients equipped with hybrid closed-loop systems (HCL). The aim is to compare postprandial glycemic control following meals declared by QMSE of the DBLG1 system with PCC. Methods: We randomly selected a 20% sample of patients from the commercial database of type 1 diabetes patients equipped with the DBLG1 system in Europe. We assumed that when the carbohydrates (CHO) amount was identical to the predefined average meal value (small, medium, or large meal), the patient used the semiquantitative method, and the corresponding meals were assigned to the QMSE group. The others were assigned to the PCC group. The glucose metrics of the meals were computed during the postprandial period, defined as [t meal ; t meal + 4 h], provided that there was no other meal during this 4-h period or during the previous 4 h. Results: A total of 1959 patients from seven Western European countries were included (mean HbA1c 7.6% ± 1.2%; mean age 43.9 ± 14.7 years). Overall, 287,000 meals (47%) were declared with PCC and 327,819 (53%) with QMSE and the mean meal size was 47.2 ± 32.5 g and 48.4 ± 28.6 g, respectively. The postprandial TIR was 62.39% ± 30.86% with QMSE and 63.21% ± 30.62% with PCC. The mean TIR difference of 0.81% was statistically significant but not clinically relevant. Time below range (TBR) was low for both methods of declaration (TBR < 70 mg/dL of 1.4% ± 5.0% with QMSE and 1.4% ± 4.8% with PCC). Conclusion: The semi-quantitative CHO declaration achieves similar glycemic results as CHO counting in this retrospective study. This method could help to reduce the burden of diabetes and offers an alternative to patients reluctant to use CHO counting.