Comparison of first trimester dating methods for gestational age estimation and their implication on preterm birth classification in a North Indian cohort

doi link
Published in "medRxiv"
Ramya Vijayram , Nikhita Damaraju , Ashley Xavier , Koundinya Desiraju , Ramachandran Thiruvengadam , Raghunathan Rengasamy , Himanshu Sinha , Shinjini Bhatnagar , others
clinical data , machine learning

Different formulae have been developed globally for estimation of gestational age (GA) by ultrasonography in the first trimester of pregnancy. In this study, we develop an Indian population-specific dating formula and compare its performance with published formulae. Finally, we evaluate the implications of the choice of dating method on preterm birth (PTB) rate. The data for this study was from GARBH-Ini, an ongoing pregnancy cohort of North Indian women to study PTB. Methods: Comparisons between ultrasonography-Hadlock and last menstrual period (LMP) based dating methods were made by studying the distribution of their differences by Bland-Altman analysis. Using data driven approaches, we removed data outliers more efficiently than by applying clinical parameters. We applied advanced machine learning algorithms to identify relevant features for GA estimation and developed an Indian population-specific formula (Garbhini-1) for the first trimester. PTB rates of Garbhini-1 and other formulae were compared by estimating sensitivity and accuracy.

Performance of Garbhini-1 formula, a non-linear function of crown-rump length (CRL), was equivalent to published formulae for estimation of first trimester GA (limits of agreement, -0.46,0.96 weeks). We found that CRL was the most important parameter in estimating GA and no other clinical or socioeconomic covariates contributed to GA estimation. The estimated PTB rate across all the formulae including LMP ranged 11.54-16.50% with Garbhini-1 estimating the least rate with highest sensitivity and accuracy. While LMP-based method overestimated GA by three days compared to USG-Hadlock formula; at an individual level, these methods had less than 50% agreement in classification of PTB. Conclusions: An accurate estimation of GA is crucial for management of PTB. Garbhini-1, the first such formula developed in an Indian setting, estimates PTB rates with higher accuracy especially when compared to commonly used Hadlock formula. Our results reinforce the need to develop population-specific gestational age formulae.