AI-Predicted Disease Trajectories in Diabetes: A Retrospective Study (AI-TRYDIA)

A Retrospective Observational Study to Use Artificial Intelligence for Prediction of Disease TRajectorY in DIAbetes

The study explores the utilization of artificial intelligence (AI) to predict disease progression trajectories in patients with diabetes. By analyzing historical data from a retrospective cohort, we aim to identify patterns and predictors of disease evolution. The approach seeks to enhance personalized treatment strategies and improve outcomes by foreseeing potential complications and disease milestones. The findings could pave the way for more targeted and effective management of diabetes through AI-driven insights.

Study Overview

• Status: Not yet recruiting
• Conditions: Diabetes Mellitus, Type 2; Diabetes Mellitus, Type 1
• Intervention / Treatment: Other: AI-Analyis

Detailed Description

The proposed study aims to harness the power of artificial intelligence (AI) and machine learning (ML) to address critical clinical needs in the management of Diabetes Mellitus (DM), a chronic and non-remissive disease that significantly impacts patients' lives. Despite the availability of hypoglycemic therapies, the prevention of both microvascular (retinopathy, nephropathy, neuropathy) and macrovascular (cardiovascular, cerebrovascular disease, and peripheral arterial disease) complications remains a challenge, with diabetic patients at higher risk compared to the general population.

The study focuses on two primary objectives: first, to a priori identify patients with varying probabilities of developing DM complications, allowing for a more resource-intensive approach for those at greater risk; second, to pinpoint the most effective therapeutic choices tailored to individual patient profiles. These objectives stem from distinct clinical characteristics and needs in the management of Type 1 DM (T1DM) and Type 2 DM (T2DM). For T1DM, the phenomenon of partial clinical remission post-diagnosis, marked by reduced insulin need and glycemic variability, suggests a window for improved long-term outcomes. Conversely, T2DM management lacks clear guidance for personalized medication regimens following metformin, highlighting a gap in treatment optimization.

Leveraging AI and ML for the analysis of multidimensional and longitudinal health data presents an innovative approach to predicting disease trajectories and therapeutic outcomes in DM. This observational, retrospective study, initially monocentric with potential for broader data integration, will delve into Electronic Health Records (EHR) using the Smart Digital Clinic Software (Meteda). By screening patients based on specific eligibility criteria, including DM type classification and historical health markers, this research aims to generate two distinct patient cohorts for in-depth analysis.

This study not only addresses a pressing clinical necessity by aiming to enhance personalized DM management and prevent complications but also contributes to the nascent field of AI application in healthcare. Through this exploration, the study seeks to offer new insights, validate AI and ML's utility in medical predictions, and establish a foundation for future research and clinical practices that embrace technological advancements for improved patient care.

• Study Type: Observational
• Enrollment (Estimated): 10000

Contacts and Locations

• Study Contact: Name: Lorenzo Piemonti, MD; Phone Number: +390226432706; Email: piemonti.lorenzo@hsr.it

Study Locations

• Italy
  • Lombardy
    • Milan, Lombardy, Italy, 20132
      • Diabetes Research Institute-IRCCS Ospedale San Raffaele

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Probability Sample

Study Population

The study population comprises individuals diagnosed with Diabetes Mellitus, encompassing both Type 1 Diabetes Mellitus (T1DM) and Type 2 Diabetes Mellitus (T2DM). This population is identified through medical records housed within the Electronic Health Record (EHR) system, specifically utilizing data generated by the Smart Digital Clinic Software (Meteda) since its inception in our hospital environment.

Description

Inclusion Criteria:

• Diagnosis: Individuals with a confirmed diagnosis of T1DM or T2DM, as indicated by their EHR labels or a history of glycated hemoglobin levels and medication usage consistent with diabetes management.
• Age: Patients of all ages are considered, with subgroups possibly defined for more detailed analysis (e.g., pediatric, adult, senior).
• Treatment history: Both patients who are newly diagnosed and those with an established history of diabetes treatment, including insulin and oral hypoglycemic agents, are included to capture a broad spectrum of disease trajectories.

Exclusion Criteria:

• Incomplete records: Patients with incomplete medical records that do not provide sufficient information on their diabetes diagnosis, treatment history, or follow-up data are excluded.
• Other significant diseases: Individuals with comorbid conditions that could significantly alter the natural history of diabetes or its management (e.g., end-stage renal disease not related to diabetes, active cancer treatment) might be excluded to ensure the study focuses on the diabetes trajectory.

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
T1DM cohort; A. T1DM label attached in the EHR OR B. patients with at least a record of Glycated Hemoglobin level of >6.5% (48 mmol/mol) AND < 45 years old AND no use of oral antidiabetic drug AND positivity of ≥2 anti-islet antibodies	Other: AI-Analyis; The study will investigate classification (ie logistic regression, decision tree, random forest, support vector machine, K nearest neighbour, naive bayes) ML models and treatment effect estimation ML models (T-learner, X-learner..).
T2DM cohort: A. T2DM label attached in the EHR OR B. patients with at least a record of Glycated Hemoglobin level of >6.5% (48 mmol/mol) AND Medication history of antidiabetic drug comprising insulin or not	Other: AI-Analyis; The study will investigate classification (ie logistic regression, decision tree, random forest, support vector machine, K nearest neighbour, naive bayes) ML models and treatment effect estimation ML models (T-learner, X-learner..).

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Primary Endpoint	Development and validation of a model to predict Partial Clinical Remission (PCR) to identify individuals diagnosed with T1D who are most likely to undergo PCR in the early stages of the natural history of the disease. The definition for PCR, namely glycated hemoglobin adjusted for insulin dose (IDAA1c), will be evaluated at 6 and 12 months after the onset of diabetes. Remitters and nonremitters will be dichotomously divided by IDAA1c ≤9 and IDAA1c >9	0-36 month
Primary Endpoint	Development and validation of a model to predict the development of chronic complications in patients with diabetes	0-36 month
Primary Endpoint	Development and validation of a model to predict the response to different second lines of therapy in addition to metformin in patients with T2D who have failed the first line with metformin alone.	0-36 month

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Exploratory Objectives	Gather experience on the AI workflow in the healthcare setting, from data acquisition to model development and testing.	0-36 month

Collaborators and Investigators

• Sponsor: IRCCS San Raffaele
• Investigators: Principal Investigator: Lorenzo Piemonti, MD, IRCCS Ospedale San Raffaele srl; Study Director: Emanuele Bosi, MD, IRCCS Ospedale San Raffaele srl

Study record dates

Study Major Dates

• Study Start (Estimated): March 1, 2024
• Primary Completion (Estimated): March 1, 2025
• Study Completion (Estimated): March 1, 2026

Study Registration Dates

• First Submitted: February 15, 2024
• First Submitted That Met QC Criteria: February 23, 2024
• First Posted (Estimated): February 28, 2024

Study Record Updates

• Last Update Posted (Estimated): February 28, 2024
• Last Update Submitted That Met QC Criteria: February 23, 2024
• Last Verified: February 1, 2024

More Information

Terms related to this study

• Keywords: Type 1 Diabetes Mellitus (T1DM); Type 2 Diabetes Mellitus (T2DM); Artificial Intelligence (AI); Diabetes Mellitus (DM); Machine Learning (ML); Personalized Treatment; Observational, Retrospective Study; Health Data Analysis
• Additional Relevant MeSH Terms: Glucose Metabolism Disorders; Metabolic Diseases; Immune System Diseases; Autoimmune Diseases; Endocrine System Diseases; Diabetes Mellitus; Diabetes Mellitus, Type 2; Diabetes Mellitus, Type 1
• Other Study ID Numbers: AI-TRYDIA

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No