Predicting neonatal infection in preterm premature rupture of membranes with vaginal microbiology and metagenomics: a prospective cohort study.

OBJECTIVE: Early-onset neonatal sepsis due to ascending infection is a potentially preventable complication of preterm premature rupture of membranes. Our objective was to determine whether the analysis of bacteria from vaginal swab samples is predictive of the risk of early-onset neonatal sepsis in preterm premature rupture of membranes.

STUDY DESIGN: In a prospective 3-center observational cohort, patients with preterm premature rupture of membranes were enrolled between 22 and 36 weeks of gestation + 6 days. Vaginal swab samples at delivery were analyzed using 2 different approaches, classical bacterial cultures and shotgun metagenomic sequencing analysis. A metagenomics score was constructed combining the characterization of the vaginal microbiome and the presence of pathogens and the optimal cutoff to predict early-onset neonatal sepsis was tested on a receiver operating curve.

RESULTS: Five hundred sixty-three preterm premature rupture of membranes cases were enrolled, with 646 live-born neonates. Preterm premature rupture of membranes occurred<32 weeks of gestation in 41.9% and deliveries were<34 weeks of gestation in 41.0%. The incidence of early-onset neonatal sepsis was 29/646 (4.5%). When considering all central and peripheral microbiological samples available for 26 neonates, the main pathogens isolated were Escherichia coli in 14 cases (53.8%), other gram-negatives in 5 (19.2%), strict anaerobes in 3 (11.5%); there was a single case (3.8%) each with Group B Streptococcus, Streptococcus anginosus, Staphylococcus aureus, and Ureaplasma urealyticum. We studied the prediction of early-onset neonatal sepsis among 272 mothers and their 310 neonates (20 early-onset neonatal sepsis, 6.4%) with both culture and metagenomic data available. A culture positive for a major or intermediate pathogen in the vaginal sample at delivery had a sensitivity of 80.0% (95% confidence interval=56.3-94.3) and a specificity of 37.9% (95% confidence interval=32.3-43.8), adjusted odds ratio of 1.6 (95% confidence interval [0.5-5.0]) to predict early-onset neonatal sepsis. The presence of E. coli was associated with an early-onset neonatal sepsis risk of 10.6% vs 4.9%, in the absence of E. coli (P=0.07). The metagenomics score was highly associated with early-onset neonatal sepsis, with an area under the receiver operating curve of 0.75 (95% confidence interval, 0.61-0.90). At the optimal cutoff value, sensitivity was 70% (95% confidence interval, 64%% to 95%), specificity was 85% (95% confidence interval, 81%% to 89%). A metagenomics score greater than 40 was associated with a significantly increased risk of early-onset neonatal sepsis with an adjusted odds ratio of 8.9 (95% confidence interval [3.5; 22.3]) in multivariate analysis adjusted for latency period and gestational age, P<0.001.

CONCLUSION: In preterm premature rupture of membranes, conventional microbial culture of maternal vaginal samples was associated with early-onset neonatal sepsis, but its predictive values remain insufficient to guide perinatal care. Metagenomic microbial signatures improved predictive values. This opens the perspective for a rapid point-of-care test.