Predictive Diagnosis of Ulcero-Necrotizing EnteroColitis in Premature Babies Using an Artificial Intelligence Approach Based on Early Analysis of the Fecal Microbiota

NCT ID: NCT06727877

Last Updated: 2025-04-15

Basic Information

• Recruitment Status: RECRUITING
• Clinical Phase: NA
• Total Enrollment: 1000 participants
• Study Classification: INTERVENTIONAL
• Study Start Date: 2025-04-01
• Study Completion Date: 2026-06-01

Brief Summary

Prematurity affects around 7% of births in France. Necrotizing enterocolitis (NEC) is a dreaded digestive complication. It is responsible for a mortality rate ranging from 15 to 40%, a rate that has remained stable in recent years, and for medium- and long-term digestive and neurodevelopmental morbidity.

Its onset is unpredictable and sudden, usually between 10 and 20 days of life, and requires immediate, aggressive management: hemodynamic support, fasting, systemic antibiotic therapy or even surgery.

Prevention is therefore essential, but systematic measures with proven efficacy (breastfeeding, early enteral feeding, multiple probiotics) are few and far between. What's more, these preventive measures cannot be modulated and adapted individually, since it is not possible to finely predict the risk of developing enterocolitis.

Thus, the use of a predictive diagnostic test for NEC would make it possible to identify high-risk premature babies and develop personalized preventive measures.

Changes in the digestive microbiota precede the onset of NEC, but it has not been possible to identify a reproducible and reliable microbial signature. As a result, the limited power of microbiota analysis and interpretation means that it cannot be used in practice to predict ECUN.

Our partner team (MEDiS) has developed a bioinformatics chain (RiboTaxa) to obtain the precise structure of complex microbial communities from direct metagenomic sequencing data. Stool samples from international cohorts (1562 samples, 208 preterm infants) were then mined to train a deep neural network and generate a predictive diagnostic test for NEC. In a local study (10 cases and 10 controls), the predictive diagnostic performance of this test was 90%, with the 1ère stool identified as "at risk" preceding NEC by 8 days (extremes 4 - 17 days), and the 2nde by 2 days (extremes 0-7 days). We would now like to test our predictive diagnostic technique on a larger number of premature babies in the AURA region.

1000 children included, 200 children tested (50 NEC - 150 controls)

Detailed Description

Systematic collection of stool (excluding meconium) from premature infants up to 21 days of age. Systematic analysis of the first two stools at the MEDiS laboratory: analysis of fecal microbiota by direct metagenomic sequencing (RiboTaxa), coupled with artificial intelligence (deep neural network previously trained on literature data). The test gives us a dichotomous response (yes/no) for each stool.

In the event of discordant analysis between the 2 stools (approximately 35% of cases in our preliminary study), a 3ème stool will be analyzed in order to classify the child as being at risk of NEC or not. The person performing these analyses will not be informed of the child's clinical evolution.

The diagnosis of NEC will be made by the clinician in charge of the child, according to the Bell classification.

Follow-up until return home or transfer to a peripheral center. A telephone call will be made to parents at 3 months of age, to ensure that no NECN has occurred after transfer to a peripheral center.

Conditions

Necrotizing Enterocolitis (NEC); Healthy Control

Study Design

• Allocation Method: NON_RANDOMIZED
• Intervention Model: PARALLEL
• Primary Study Purpose: DIAGNOSTIC
• Blinding Strategy: NONE

Study Groups

NEC

diagnosis of NEC according to the Bell classification

Group Type: EXPERIMENTAL

Ability of early digestive microbiota analysis (using artificial intelligence) to predict the occurrence of NEC diagnosed according to the Bell classification.

Intervention Type: DIAGNOSTIC_TEST

The test gives us a dichotomous response (yes/no) for each stool. We will systematically analyze two stools per child, and in the event of a discrepancy, we will analyze a third to classify the child as being at risk of NEC or not.

The analysis model consists of a deep neural network that has been trained and optimized on data from international cohorts. In a local pilot study (N=20), it enabled accurate prediction for 90% of newborns.

control

children without diagnosis of NEC

Group Type: OTHER

Ability of early digestive microbiota analysis (using artificial intelligence) to predict the occurrence of NEC diagnosed according to the Bell classification.

Intervention Type: DIAGNOSTIC_TEST

The test gives us a dichotomous response (yes/no) for each stool. We will systematically analyze two stools per child, and in the event of a discrepancy, we will analyze a third to classify the child as being at risk of NEC or not.

The analysis model consists of a deep neural network that has been trained and optimized on data from international cohorts. In a local pilot study (N=20), it enabled accurate prediction for 90% of newborns.

Interventions

Ability of early digestive microbiota analysis (using artificial intelligence) to predict the occurrence of NEC diagnosed according to the Bell classification.

The test gives us a dichotomous response (yes/no) for each stool. We will systematically analyze two stools per child, and in the event of a discrepancy, we will analyze a third to classify the child as being at risk of NEC or not.

The analysis model consists of a deep neural network that has been trained and optimized on data from international cohorts. In a local pilot study (N=20), it enabled accurate prediction for 90% of newborns.

Intervention Type: DIAGNOSTIC_TEST

Eligibility Criteria

Inclusion Criteria

• Child born prematurely (i.e. before 34 weeks of amenorrhea) in one of participating university hospitals and hospitalized in neonatal intensive care units of the AURA region's university hospitals
• Child born outside CHU and transferred before 24h of life to the neonatal intensive care unit of one of thehospital participating in the study
• Affiliated with a Social Security scheme

Exclusion Criteria

• Child whose guardians are protected by law (guardianship, curatorship, safeguard of justice)
• Children whose parents are under 18 years of age
• Refusal of parental authority to participate

• Maximum Eligible Age: 1 Day
• Eligible Sex: ALL
• Accepts Healthy Volunteers: Yes

Sponsors

University Hospital, Clermont-Ferrand

OTHER

Sponsor Role: lead

Responsible Party

Responsibility Role: SPONSOR

Principal Investigators

Maguelonne Pons

Role: PRINCIPAL_INVESTIGATOR

University Hospital, Clermont-Ferrand

Locations

CHU de Clermont-Ferrand

Clermont-Ferrand, , France

Site Status: RECRUITING

CHU Grenoble

Grenoble, , France

Site Status: NOT_YET_RECRUITING

HFME

Lyon, , France

Site Status: NOT_YET_RECRUITING

Hopital Croix Rousse

Lyon, , France

Site Status: NOT_YET_RECRUITING

CHU Saint Etienne

Saint-Etienne, , France

Site Status: RECRUITING

Countries

France

Central Contacts

Lise Laclautre

Role: CONTACT

promo_interne_drci@chu-clermontferrand.fr

334.73.754.963

Facility Contacts

Lise Laclautre

Role: primary

+33473750573

Other Identifiers

2024-A01840-47

Identifier Type: OTHER

Identifier Source: secondary_id

PHRC I 2023 PONS

Identifier Type: -

Identifier Source: org_study_id