Ultrasound, blood biomarkers, and the vaginal microbiota in early pregnancy as predictors of pregnancy outcome

Abstract

Background Miscarriage is a common adverse pregnancy outcome with varied causes. It is diagnosed with strict ultrasound criteria. Symptoms of bleeding and pain in the first trimester of pregnancy necessitate clinical assessment which, if criteria for a miscarriage diagnosis are not met, often generates uncertainty about the eventual pregnancy outcome. Even when a miscarriage is diagnosed with certainty, uncertainty remains about the underlying cause. In addition to traditional clinical factors, increasingly there is an emerging body of evidence that blood biomarkers and vaginal microbiota differ between miscarriage and normal pregnancy. These variables have the potential to inform understanding of physiology and the development of predictive tools during early pregnancy, but prior to this study have largely been cross-sectional and have tended to inadequately account for confounding variables. Aims The aim of this work was to investigate associations with first trimester outcome by studying the differences in demographic, clinical, sonographic, blood markers and the vaginal microbiota prior to first trimester outcomes. Methods The studies presented here are from sub-groups of participants recruited into a prospective cohort study, Early Pregnancy events and Outcome Study (EPOS). At a baseline study visit (mean=7.2 weeks; range=5-14 weeks) and then fortnightly until 14 weeks, demographic and clinical data was recorded alongside participants undergoing an ultrasound scan and donating paired blood and vaginal samples. Based upon data availability three EPOS sub-groups were studied: prediction model n=428, microbiota and pregnancy of unknown viability (PUV) n=100, and microbiota in successive pregnancies n=92 women with 192 pregnancies. For model development logistic regression was used. Blood biomarkers were analysed using an automated immune-analyser (Roche). Vaginal microbiota were identified by sequencing of 16S ribosomal ribonucleic acid (rRNA) hypervariable region genes on an Illumina platform. Results A model applicable to patients who present prior to 14 weeks, are certain of menstrual dates, and have an intrauterine pregnancy on ultrasound was derived. Using five variables (maternal age, folic acid use, difference in gestational age by menstrual dates and ultrasound, and levels of progesterone and ß-human chorionic gonadotrophin), receiver-operator characteristic analysis found a highly accurate discrimination of first trimester outcome (area under the curve=0.84). Analysis of vaginal microbiota found Lactobacillus spp. dominance associated with higher serum oestradiol levels in women with PUV (411.3pg/ml vs 299.2pg/ml, P=0.02), and when deplete of Lactobacillus spp., lower serum oestradiol levels were associated with miscarriage (212pg/ml vs 395pg/ml, P=0.003). The successive pregnancy study found that nulliparous women were more likely to have vaginal Lactobacillus crispatus abundance compared to their subsequent pregnancy (62% vs 41%, P=0.04), or compared to women studied after any previous pregnancy (62% vs 26.7%). Conclusion Information available during an early pregnancy assessment can inform the quantification of miscarriage risk in the first trimester, pending further validation in independent cohorts. Biomarkers that associate with miscarriage correlate with the composition of the vaginal microbiota. The relationship between vaginal microbiota and miscarriage appears to be nuanced; this work finds that it is individually specific and influenced by other variables such as pregnancy history and interpregnancy interval.