Perinatal & Pediatric
Vaginal Microbiome Structure in Pregnancy and Host Factors Predict Preterm Birth: Results from the ECHO Cohorts Kimberly McKee* Kimberly McKee Christine M. Bassis Jonathan Golob Beatrice Palazzolo Ananda Sen James E. Gern Sarah S. Comstock Christian Rosas-Salazar Tom O’Connor Nigel Paneth Anne L. Dunlop Environmental Influences on Child Health Outcomes (ECHO) Program
The vaginal microbiome is a dynamic system, typically shifting over the course of pregnancy toward increased enrichment of Lactobacillus to support pregnancy maintenance. Proliferation of Lactobacillus may be absent among women with preterm births although specific signatures robust across populations have not been identified. Using vaginal data from the first to third trimester from pregnancy cohorts with available data in the Environmental Influences on Child Health Outcomes Program, we meta-analyzed 16S rRNA gene amplicon sequence data to identify robust vaginal microbiota signatures in pregnancy that along with host factors predicted preterm birth (<37 weeks gestation). We classified community state types and employed penalized logistic regression models, applying Firth’s correction, and mixed models for assessing the association between vaginal community state types and preterm birth. To compare the most predictive combination of vaginal taxa and host factors, we generated supervised random forest model plots with receiver operating curves and validated them using a train (20%) and test (80%) approach. Of N=683 births, 12% were preterm, most of which were spontaneous. Vaginal community types were strongly associated with spontaneous preterm birth (adjusted odds ratio= 3.86[95%CI= 1.57-11.3] for Diverse, non-Lactobacillus dominant communities; adjusted odds ratio= 3.03 [95%CI=1.25-8.78] for L. iners-dominant compared to those L.crispatus dominant). Although strongly associated with preterm birth, after adjustment for vaginal community types and level of maternal education, self-identified race was no longer a robust predictor(p=0.58). We calculated the highest area under the curve (0.72) for the most discriminant vaginal taxa (Shuttleworthia satelles, Prevotella amnii, Gardnerella vaginalis) combined with host age, providing insight into potential predictive risk scoring and targets for preterm birth prevention strategies that may be robust across populations.