Objectives: To investigate changes in vaginal microbiota during pregnancy, and the association between vaginal dysbiosis and reproductive outcomes.

Methods: A total of 730 (week 24) and 666 (week 36) vaginal samples from 738 unselected pregnant women were studied by microscopy (Nugent score) and characterized by 16S rRNA gene sequencing. A novel continuous vaginal dysbiosis score was developed based on these methods using a supervised partial least squares model.

Results: Among women with bacterial vaginosis in week 24 (n = 53), 47% (n = 25) also had bacterial vaginosis in week 36. In contrast, among women without bacterial vaginosis in week 24, only 3% (n = 18) developed bacterial vaginosis in week 36. Vaginal samples dominated by Lactobacillus crispatus (OR 0.35, 95% CI 0.20–0.60) and Lactobacillus iners (OR 0.40, 95% CI 0.23–0.68) in week 24 were significantly more stable by week 36 when compared with other vaginal community state types. Vaginal dysbiosis score at week 24 was associated with a significant increased risk of emergency, but not elective, caesarean section (OR 1.37, 955 CI 1.15–1.64, p < 0.001), suggesting a 37% increased risk per standard deviation increase in vaginal dysbiosis score.

Conclusions: Changes in vaginal microbiota from week 24 to week 36 of pregnancy correlated with bacterial vaginosis status and vaginal community state type. A novel vaginal dysbiosis score was associated with a significantly increased risk of emergency, but not elective, caesarean section. This was not found for bacterial vaginosis or any vaginal community state type and could point to the importance of investigating vaginal dysbiosis as a nuanced continuum instead of crude clusters.