NEC is one of the most devastating gastrointestinal emergencies in neonatal intensive care units, particularly among very preterm and very-low-birth-weight infants. It affects approximately 2 to 7 percent of preterm infants admitted to NICUs and is associated with substantial mortality, especially in infants requiring surgery. Early recognition remains difficult because the initial clinical manifestations are nonspecific and overlap with other neonatal conditions such as sepsis and spontaneous intestinal perforation. As a result, diagnosis is often established only after radiographic or surgical confirmation, when intestinal injury may already be advanced. Current management is mainly supportive, consisting of bowel rest, antibiotics, and surgery when indicated, while preventive strategies include human milk feeding and, depending on local practice and the applicable evidence base, probiotic supplementation.

Commonly used blood- and urine-based biomarkers such as C-reactive protein and procalcitonin reflect systemic inflammation but lack specificity for intestinal injury and have limited discriminatory value for NEC. This limitation has fueled interest in stool-based biomarkers, which may offer a noninvasive readout of intestinal inflammation, epithelial barrier dysfunction, and mucosal immune activity closer to the site of disease. However, most previous studies focused on single biomarkers, were conducted at single centers, or evaluated limited numbers of analytes in relatively small cohorts. The new analysis addressed these gaps by measuring six fecal biomarkers simultaneously within a multicenter cohort, selecting markers that reflect complementary domains of NEC biology: intestinal inflammation, epithelial barrier dysfunction, antimicrobial defense, and mucosal immunity.

The study was nested within the multicenter ClosNEC cohort (ClinicalTrials.gov identifier NCT02444624), which enrolled preterm infants born before 32 weeks of gestation from 20 NICUs in France over a 16-month period. Infants with NEC were identified by treating neonatologists and classified as confirmed cases when they met modified Bell stage II or higher criteria based on clinical and radiographic findings. The comparison group comprised preterm infants without clinical or radiographic NEC during hospitalization, with control samples selected to approximate the postnatal age distribution of the NEC samples. Infants with spontaneous intestinal perforation were excluded from both groups. Of 142 collected stool samples, 130 had sufficient material for at least one biomarker analysis and were included: 42 from infants with confirmed NEC and 88 from preterm infants without NEC. Twelve samples were excluded because stool material was unavailable or insufficient.

Stool samples were recovered from diapers, transferred into cryotubes containing brain-heart infusion broth supplemented with 15 percent glycerol, and stored at minus 80 degrees Celsius until analysis. All samples underwent the same collection, dilution, freezing, and storage procedures, and the number of freeze-thaw cycles did not differ between groups. Six biomarkers were quantified using commercial enzyme-linked immunosorbent assay kits performed in duplicate: fecal calprotectin (fCal), human beta-defensin-2 (hBD-2), alpha-1-antitrypsin (α1-AT), secretory immunoglobulin A, cingulin (CGN), and eosinophil-derived neurotoxin (EDN). Concentrations were normalized to stool mass and expressed per gram of stool, with values below the lower limit of detection imputed as one-half of that limit. Discriminatory performance was assessed with receiver operating characteristic analysis and logistic regression, with AUC confidence intervals obtained by bootstrap resampling of 2000 iterations. Because this was an exploratory ancillary analysis, p values were not adjusted for multiple comparisons and were interpreted descriptively.

Baseline characteristics were broadly similar between groups. Median gestational age was 28 weeks in both groups, and median birth weight was 983 grams in infants with NEC versus 973 grams in those without. Cesarean delivery occurred in 62 percent of infants with NEC and 59 percent of controls, and median postnatal age at sampling was 24 days versus 28 days. None of the infants received probiotics. The main difference was human milk exposure: infants with NEC were less likely to receive exclusive human milk at postnatal day 3, with exclusive donor human milk feeding more frequent among controls and combined human milk and formula feeding more frequent among NEC infants, while no difference was observed at postnatal day 7.

Biomarker concentrations differed between groups for all markers except cingulin. Compared with controls, infants with NEC had higher concentrations of fCal (median 198 versus 77 micrograms per gram), EDN (597 versus 325 nanograms per gram), and α1-AT (248 versus 105 micrograms per gram), and lower concentrations of hBD-2 (9 versus 27 nanograms per gram) and sIgA (89 versus 2211 micrograms per gram). CGN concentrations did not differ, with a median of 0.46 nanograms per gram in both groups. In ROC analyses, sIgA showed the best individual discrimination (AUC 0.82), followed by fCal (0.74) and α1-AT (0.73); EDN showed modest discrimination (0.67), whereas CGN was not informative (0.51). Because lower hBD-2 and sIgA concentrations were associated with NEC, the ROC analyses for these markers were oriented so that lower values predicted case status.

The sIgA finding required careful handling because human milk exposure differed between groups and maternally derived IgA contributes to infant stool sIgA. On postnatal day 3, median sIgA concentrations were 1592 micrograms per gram among infants receiving exclusive donor human milk, 958 micrograms per gram among those receiving exclusive mother’s own milk, and 320 micrograms per gram among those receiving a combination of human milk and infant formula, an overall difference with a p value of 0.02. At postnatal day 7, sIgA did not differ significantly by milk type. After adjustment for milk type, sIgA concentrations remained significantly lower in infants with NEC at both postnatal days 3 and 7, and stratified analyses showed lower sIgA in NEC infants both among those receiving and not receiving human milk. In an exploratory adjusted analysis in 130 infants, the associations of lower hBD-2 and sIgA and higher EDN and α1-AT with NEC remained statistically significant after adjustment for gestational age, delivery mode, and human milk exposure, although residual confounding related to the dose, duration, and timing of human milk exposure cannot be excluded.

Two multimarker logistic regression models were explored. A four-biomarker model including α1-AT, EDN, hBD-2, and sIgA was fitted in all 130 infants and yielded an AUC of 0.83, numerically higher than most individual biomarkers but only a modest gain over sIgA alone. fCal was excluded because measurements were available for only 108 infants, and CGN was excluded because it showed no individual discriminatory value. A six-biomarker model including all markers was fitted as a complete-case analysis in the 108 infants with measurements for all six and yielded an AUC of 0.86; the remaining 22 infants were excluded only because insufficient stool material remained for the fCal assay. The authors caution that because the six-biomarker model was evaluated in a reduced complete-case sample, its AUC should not be interpreted as a direct improvement over the full-cohort AUC for sIgA. Pairwise correlation analysis in the 108 infants with complete data showed generally weak-to-moderate inter-marker associations, somewhat stronger among NEC infants, for example EDN with α1-AT at r = 0.64 in NEC infants versus r = 0.43 in controls, suggesting partial biological overlap without complete redundancy.

The study has important limitations that frame how the results should be read. The comparison of confirmed NEC cases with preterm infants without NEC does not reproduce real-world diagnostic decision-making in infants with suspected NEC and does not establish early predictive performance; the clinically relevant intended-use population is infants undergoing evaluation for suspected NEC before definitive radiographic confirmation, when clinicians must distinguish NEC from sepsis, feeding intolerance, spontaneous intestinal perforation, and other causes of deterioration. The sample size limited the complexity of multivariable modeling, and early human milk exposure differed between groups. The incremental improvement of the multimarker panel over sIgA alone was modest, and external validation is required before clinical use. In addition, stool availability may be limited in critically ill preterm infants, potentially restricting the feasibility of stool-based testing when stool output decreases, and clinical implementation would require rapid turnaround assays compatible with urgent bedside decision-making. At present, the authors conclude, stool biomarkers should be considered adjunctive tools rather than standalone diagnostics, and prospective validation in infants with suspected NEC and clinically relevant comparator groups is needed.

Subject of Research: Stool biomarkers for discriminating confirmed necrotizing enterocolitis in preterm infants

Article Title: Stool biomarkers discriminate confirmed necrotizing enterocolitis in preterm infants

Article References: Diwan, M. A., Delannoy, J., Ferraris, L., Rozé, J.-C., Kapel, N., Butel, M.-J., Barbut, F., & Aires, J. (2026). Stool biomarkers discriminate confirmed necrotizing enterocolitis in preterm infants. Pediatric Research. https://doi.org/10.1038/s41390-026-05566-4

Image Credits: AI Generated

DOI: 10.1038/s41390-026-05566-4

Keywords: necrotizing enterocolitis, preterm infants, stool biomarkers, secretory IgA, fecal calprotectin, ROC analysis, multicenter cohort, mucosal immunity, intestinal barrier dysfunction, ELISA, neonatal intensive care, diagnostic discrimination