An exploratory study of nano-titanium dioxide combined with vaginal microbiota for the prediction of preterm birth

Vaginal dysbiosis and microbial metabolites are associated with preterm birth. This study evaluated nano-titanium dioxide (nano-TiO 2 ) enrichment combined with 16 S ribosomal ribonucleic acid (16 S rRNA) gene sequencing and metabolomics and developed an exploratory second-trimester risk model. This retrospective analysis included 208 singleton pregnancies (42 preterm and 166 term) with samples collected at 12–14 and 24–26 weeks of gestation. The second sampling date defined the prediction landmark. A fixed six-predictor logistic model included previous preterm birth, short cervix, Lactobacillus , Gardnerella , lactate, and putrescine. Maximum-likelihood and Firth models were internally validated using the same 1,000 bootstrap samples, without discarding or replacing samples. Likelihood-ratio tests compared the combined model with reduced models. At 24–26 weeks, the preterm group had higher alpha diversity and lower Lactobacillus abundance. Eleven of 14 mass-spectrometry indicators differed after multiple-testing correction. The combined model had an apparent area under the receiver operating characteristic curve (AUC) of 0.824 (95% confidence interval [CI], 0.757–0.891). At the probability threshold of 0.218988, sensitivity, specificity, positive predictive value, and negative predictive value were 73.8%, 80.1%, 48.4%, and 92.4%, respectively. The optimism-corrected AUC, calibration slope, and Brier score were 0.800, 0.873, and 0.135, respectively. The Firth model had an optimism-corrected AUC of 0.801. Adding the Shannon index did not show clear incremental information (likelihood-ratio P = 0.499). The exploratory model showed internally assessed discrimination at 24–26 weeks, with residual overfitting. Improvements in model fit in the development sample do not establish clinical utility. External validation, recalibration when indicated, assay standardization, and assessment of turnaround time, net benefit, and cost are required. Not applicable.