Research Letter
Pediatrics
Probiotic-Associated Invasive Infection Among Preterm Infants
Ravi Mangal Patel, Monica M. Bennett, Jennifer Canvasser, et al
JAMA Netw Open 2026;9;(7):e2626530. doi:10.1001/jamanetworkopen.2026.26530
Introduction
Probiotic supplementation administered to preterm infants reduces the risk of necrotizing enterocolitis, death, and late-onset invasive infection.1,2 Despite benefits, valid concerns exist about the risk of sepsis from probiotic organisms. In the largest clinical trial to date (2158 infants),3 there was no reported risk of probiotic-associated infections, although not all trials have reported this.1 However, probiotic-associated sepsis has been reported in case series and case reports, including a systematic review of 16 studies involving 32 preterm infants.4 The isolated organisms were consistent with the most administered probiotic products in preterm infants, including Bifidobacterium and Lactobacillus strains, with genomic analysis confirming the probiotic as the source. In 2023, the US Food and Drug Administration issued a warning based on a case report about the risk of invasive disease in preterm infants given probiotics.5
The absolute risks of probiotic-associated sepsis remain uncertain given the lack of large cohort studies systematically estimating uncommon risks. Therefore, we examined the incidence of invasive infection with Bifidobacterium or Lactobacillus in a large cohort of very preterm infants.
Methods
This retrospective cohort study used the Pediatrix Medical Group Clinical Data Warehouse, a multicenter, WCG Institutional Board Review–approved dataset that is prospectively collected and used by neonatal intensive care units in 33 states and Puerto Rico. The WCG waived informed consent due to the deidentified nature of the data. This study followed the STROBE reporting guideline for cohort studies.
We included very preterm infants (gestational age [GA] 22 weeks 0 days to 32 weeks 6 days) discharged from January 1, 2016, through December 31, 2023, and followed up until discharge, death, or transfer to another hospital. Data were analyzed between November 1, 2024, and April 8, 2026. We defined probiotic supplementation similar to a prior study6 and defined invasive infection as either Lactobacillus or Bifidobacterium identified in a blood culture at age greater than 72 hours. Infections were compared between infants with and without probiotic supplementation (as bacteria similar to those in probiotic products may normally colonize in unsupplemented infants) using risk differences with 95% CIs. We analyzed the overall cohort (unmatched) and performed propensity score matching to address potential confounding by indication using baseline covariables. Propensity scores for probiotic supplementation were calculated using GA, birthweight, inborn status (born in the study hospital), and sex. Data on race and ethnicity were not collected as these were not expected to be important confounders or relevant to the risk of probiotic-associated infection. Matching was performed using the greedy nearest neighbor method and allowed for an up to 3:1 matching ratio to include as many infants receiving probiotic supplementation as possible (eMethods in Supplement 1). Additional outcomes and characteristics were only for description and not intended for causal inference. The analyses were performed using SAS, version 9.4 (SAS Institute Inc). A P < .05 was considered significant.
Results
We evaluated 107 505 very preterm infants (median [IQR] GA, 30 [27-32] weeks; 47% female and 53% male; 27% with birth weight <1000 g; 87% inborn), of whom 19 900 (19%) received probiotics (Table 1). The median GA and birth weight were numerically similar among infants with and without probiotic supplementation.
Table 1. Baseline Characteristics
| Characteristic | Unmatched cohort, % | Matched cohort, %a | |||||
|---|---|---|---|---|---|---|---|
| Overall (N = 107 505) | Probiotics supplementation (n = 19 990) | No probiotics supplementation (n = 87 515) | P value | Overall (N = 79 952) | Probiotics supplementation (n = 19 988) | No probiotics supplementation (n = 59 964) | |
| GA, median (IQR), wk | 30 (27-32) | 30 (27-31) | 30 (27-32) | <.001 | 30 (27-31) | 30 (27-31) | 30 (27-31) |
| Birth weight, median (IQR), g | 1340 (960-1680) | 1300 (940-1620) | 1355 (965-1700) | <.001 | 1297 (941-1620) | 1300 (940-1620) | 1296 (942-1620) |
| Birth weight strata, g | |||||||
| <1000 | 29145 (27.1) | 5684 (28.4) | 23461 (26.8) | <.001 | 22749 (28.5) | 5682 (28.4) | 17067 (28.5) |
| 1000-1500 | 37336 (34.7) | 7613 (38.1) | 29723 (34.0) | 30355 (38.0) | 7613 (38.1) | 22742 (37.9) | |
| >1500 | 41024 (38.2) | 6693 (33.5) | 34331 (39.2) | 26848 (33.6) | 6693 (33.5) | 20155 (33.6) | |
| Inborn in study hospital | 93412 (86.9) | 17248 (86.3) | 76164 (87.0) | .005 | 69211 (86.6) | 17246 (86.3) | 51965 (86.7) |
| Sex | |||||||
| Female | 50736 (47.2) | 41292 (47.2) | 9444 (47.2) | .88 | 37759 (47.23) | 9443 (47.2) | 28316 (47.2) |
| Male | 56769 (52.8) | 10546 (52.8) | 46223 (52.8) | 42193 (52.8) | 10545 (52.8) | 31648 (52.8) | |
Eleven infants (<0.01%) developed Bifidobacterium or Lactobacillus invasive infection (8 supplemented, 3 not supplemented) (Table 2). Of the 3 invasive infection cases among unsupplemented infants, 1 was at a center that used probiotics within the same discharge year, and 2 were at centers with no probiotic use. In the overall unmatched and propensity-matched analyses, probiotic supplementation was associated with an absolute increase in Lactobacillus or Bifidobacterium infection of 0.04% (95% Cl, 0.01%-0.06%), equivalent to 1 invasive infection per approximately 2500 supplemented infants (95% CI, 1667-10 000 supplemented infants). Among the 613 deaths in infants receiving supplementation, 1 occurred more than 30 days after positive blood culture in an infant with invasive infection.
Table 2. Outcomes and Characteristics
| Outcome | Infants, No. (%) | Risk difference, % (95% CI) | ||
|---|---|---|---|---|
| Overall | Probiotics supplementation | No probiotics supplementation | ||
| Primary outcome in unmatched cohort | ||||
| No. of infants | 107 505 | 19 990 | 87 515 | NA |
| Invasive infection with Bifidobacterium or Lactobacillus | 11 (0.01) | 8 (0.04) | 3 (0.003) | 0.04 (0.01 to 0.06) |
| Descriptive outcomes and characteristics in unmatched cohort | ||||
| Death | 6151 (5.7) | 613 (3.1) | 5538 (6.3) | −3.3 (−3.6 to −3.0) |
| Any bloodstream infection | 6634 (6.2) | 1532 (7.7) | 5102 (5.8) | 1.8 (1.4 to 2.2) |
| Any bloodstream infection after the start of probiotics in supplemented group | NA | 1163 (5.8) | 5102 (5.8) | −0.01 (−0.4 to 0.4) |
| Primary outcome in propensity matched cohorta | ||||
| No. of infants | 79 952 | 19 988 | 59 964 | NA |
| Invasive infection with Bifidobacterium or Lactobacillus | 11 (0.01) | 8 (0.04) | 3 (0.01) | 0.04 (0.01 to 0.06) |
| Descriptive outcomes and characteristics in propensity matched cohort | ||||
| Death | 4391 (5.5) | 612 (3.1) | 3779 (6.3) | −3.2 (−3.6 to −2.9) |
| Any bloodstream infection | 5203 (6.5) | 1532 (7.7) | 3671 (6.1) | 1.5 (1.1 to 2.0) |
| Any bloodstream infection after the start of probiotics in supplemented group | NA | 1163 (5.8) | 3671 (6.1) | −0.3 (−0.7 to 0.1) |
Discussion
This largest cohort study, to our knowledge, of the incidence of probiotic-associated invasive infection in very preterm infants allowed for quantification of uncommon risks, and consistent findings in propensity-matched results that considered confounding by indication based on baseline factors, including GA and birth weight, were reassuring. We found that infection with commonly supplemented probiotic strains was rare, consistent with a recent large study from Canada.7 As bacterial strains in probiotic supplements may also exist as part of the microbiome in unsupplemented infants, we found 2 cases of Lactobacillus or Bifidobacterium infection in infants in whom there was no known probiotic supplementation at their center.
Limitations included the inability to perform bacterial genomic analysis to determine causality, given the retrospective design. Additionally, we evaluated the 2 most common bacteria contained in probiotics used in preterm infants, which may not capture all bacteria, yeast, or contaminants in every probiotic product.
Given the infrequent incidence of invasive infection, very large cohorts may be needed to quantify such rare risks. Our data may provide additional quantification of the risks of probiotic-associated invasive infection from supplementation in the context of known reported benefits.1-3
