A review of studies in China has found a marked difference in gut bacteria between pregnant women with gestational diabetes and those with normal glucose tolerance. Women with gestational diabetes had lower levels of Bifidobacterium and Lactobacillus, and higher levels of Enterobacteriaceae, Enterococcus and Fusobacterium. The pattern is an association, however, not evidence that the bacteria cause gestational diabetes or appear before it.
The review also found no statistically significant overall difference in Bacteroides. That result came from 13 studies involving 2,491 participants, with high variation between the studies.
A consistent pattern, with important variation
For Bifidobacterium, abundance was lower in the gestational-diabetes group across 31 studies, including 2,964 women with gestational diabetes and 3,193 controls. The pooled standardized mean difference was −2.47, with a 95% confidence interval from −2.86 to −2.09. The review reported substantial between-study heterogeneity, with I² at 97.10%, meaning the study results varied considerably.
Lactobacillus showed a similar direction. Across 29 studies and 5,813 participants, its abundance was lower among women with gestational diabetes. The pooled standardized mean difference was −1.82, with a 95% confidence interval from −2.04 to −1.60. Heterogeneity was again high, with I² at 91.80%.
Among the bacterial groups that were more abundant in gestational diabetes, Enterobacteriaceae had a pooled standardized mean difference of 1.75 across 23 studies. The 95% confidence interval ran from 1.53 to 1.98, while I² was 90.80%. Enterococcus was also higher, with a pooled standardized mean difference of 1.18 across 20 studies and a 95% confidence interval from 1.03 to 1.32. Its I² was 78.70%.
The increase in Fusobacterium was smaller. Across 15 studies involving 3,517 participants, the pooled standardized mean difference was 0.46, with a 95% confidence interval from 0.08 to 0.83. Heterogeneity remained very high, with I² at 96.40%.
What the review actually measured
The analysis combined quantitative, genus-level gut-microbiota data from eligible case-control or cohort studies of Chinese pregnant women. The search covered eight databases from their start dates through January 2026, using terms for gestational diabetes and gut microbiota, followed by manual searches of compilations, dissertations and grey literature.
The final review included 31 case-control studies and 6,157 participants: 2,964 women with gestational diabetes and 3,193 controls. Two reviewers screened the records and extracted information including sample size, gestational age, pre-pregnancy body-mass index, parity, profiling methods and target-taxon measurements.
The review expressed the continuous measurements as standardized mean differences, or SMDs, with 95% confidence intervals. In the results, a negative SMD indicates lower abundance in gestational diabetes and a positive SMD indicates higher abundance. The researchers assessed between-study variation with I², used random-effects pooling when I² was above 50% and P was 0.05 or lower, and ran subgroup, sensitivity and publication-bias analyses.
Why the findings need caution
The evidence was observational and came mainly from case-control studies. That design can identify differences between groups, but it cannot establish whether gut-microbiota changes preceded gestational diabetes, resulted from it, or were influenced by other factors. The review also says that differences in diet, body-mass index, geography, sample processing, detection methods and sequencing platforms may have affected comparisons.
The review reported publication-bias signals for Bifidobacterium, Lactobacillus and Enterobacteriaceae. A trim-and-fill analysis, which estimates the effect of potentially missing studies, imputed two studies for Bifidobacterium, nine for Lactobacillus and none for Enterobacteriaceae. The corrected estimates remained statistically significant and pointed in the same directions as the original findings.
The main pattern also held up in leave-one-out checks. Removing two potential outliers changed the Bifidobacterium estimate to −2.27, with a 95% confidence interval from −2.62 to −1.91. Removing one study changed the Lactobacillus estimate to −1.71, with a 95% confidence interval from −1.90 to −1.52. Neither analysis reversed the direction of the association, and the review says the other major results were similarly robust to sequential omission.
The study’s quality assessment classified all included studies as high quality on the Newcastle–Ottawa Scale, with scores of at least 7. That assessment does not remove the wider limits of the evidence: the analysis was restricted to genus-level data, may not represent all Chinese regions or other Asian populations, and did not test probiotics, prebiotics, dietary changes or any other treatment.
A research lead, not a clinical test
For now, the findings identify candidate bacterial associations for further study. They do not by themselves establish a screening test or a treatment, because the review was based on observational comparisons and did not test interventions. Prospective studies, standardized microbiome methods and research examining species and microbial function will be needed to clarify what these differences mean.
Paper data and sources
Original title: A systematic analysis of gut microbiome characteristics in Chinese patients with gestational diabetes mellitus
Authors: Wang Y, Zheng H, Yue C et al.
Journal/Repository: Frontiers in nutrition
Status: Peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text