Statistical analysis
Statistical analysis was performed with R Studio software (Version
1.3.959). Categorical variables were compared using χ2or Fisher’s exact test where appropriate, and continuous variables were
compared using T-test. The Mann Whitney U Test was used to compare
quantitative CGA measurements and their associations with the primary
outcome. Further, the analogous continuous outcome of GA at delivery was
assessed relative to the quantitative CGA measurements using Spearman’s
correlations. The relationship between PTB, CL, and CGA absence was
modeled using multivariate logistic regression via a stepwise backward
elimination approach, controlling for variables identified as
significant in univariate analysis or that were deemed clinically
relevant for inclusion in the model. In constructing the model,
Spearman’s correlation testing was used to make pair-wise comparisons
between predictor variables to ensure absence of multicollinearity. Full
and reduced versions of the final model, which differed only by
inclusion of CGA evaluation with TVUS CL screening, were compared using
the likelihood ratio test. Receiver operating curves (ROCs) were
constructed to evaluate performance of the nested models (with and
without evaluation of CGA at TVUS CL screening) for predicting PTB
<37 weeks, and the areas under the curve (AUC) were
determined. Statistical significance was defined as p<0.05.