(1166) First trimester machine learning to predict preeclampsia in normotensive pregnancies by American Heart Association guidelines

Friday, January 31, 2025

4:00 PM – 6:00 PM

Submitting Author and Presenting Author(s)

Rebecca Horgan, MD

Assistant Professor
Macon & Joan Brock Virginia Health Sciences Eastern Virginia Medical School at Old Dominion University
Norfolk, VA, United States

Coauthor(s)

EK

Erkan Kalafat, MD, MSc

Associate Professor
Koc University Hospital
Istanbul, Istanbul, Turkey
Elena Sinkovskaya, MD, PhD

Professor
Macon & Joan Brock Virginia Health Sciences Eastern Virginia Medical School at Old Dominion University
Norfolk, Virginia, United States
Alfred Z. Abuhamad, MD

President, Provost and Dean
Macon & Joan Brock Virginia Health Sciences Eastern Virginia Medical School at Old Dominion University
Norfolk, Virginia, United States
George R. Saade, MD (he/him/his)

Professor and Chair, Associate Dean for Women's Health Obstetrics and Gynecology
Macon & Joan Brock Virginia Health Sciences Eastern Virginia Medical School at Old Dominion University
Norfolk, Virginia, United States

Objective: The threshold to define normal blood pressure (BP) outside pregnancy have been lowered. We assessed if machine learning can identify patients at high risk of hypertensive disorders of pregnancy (HDP) amongst normotensive first-trimester patients per American Heart Association (AHA) guidelines.

Study Design: This was a prospective cohort study which enrolled patients at ≤ 13+6 weeks’ gestation without fetal, umbilical cord or placental abnormalities. For this analysis, patients with chronic hypertension were excluded. Baseline maternal BP was classified according to the AHA guidelines (normal, elevated, stage I hypertension, stage II hypertension). The primary outcome was the incidence of HDP diagnosed per ACOG criteria. First trimester BP values were clustered using unsupervised machine learning models and patients were categorized as; AHA normal BP, low risk clustering; AHA normal BP, high risk clustering and AHA abnormal BP (elevated or above). Association with outcomes were assessed with regression analyses, which were adjusted for race, body-mass index, and risk factors for preeclampsia (PE).

Results: 570 patients were included in this analysis, with 43 developing PE and 73 any HDP. After clustering, 350 patients fell within the AHA normal BP low-risk cluster, 28 AHA normal BP high-risk cluster and 192 patients AHA abnormal BP (elevated or above). The PE rates in these categories were 3.1%, 25.0% and 13.0%, respectively (P< 0.001). The rate of small for gestational age (SGA) were 6.6%, 21.4% and 6.8% respectively (P=0.032). The hazard of PE was higher in patients with AHA normal BP, high risk clustering (adjusted HR: 8.51, 95% CI: 3.30-22.0, P< 0.001) compared to AHA abnormal BP (adjusted HR: 4.33, 95% CI: 2.12-8.84, P< 0.001) (Figure). Similar findings were observed for any HDP (adjusted HR:4.15 vs 5.77, P< 0.001 both, Figure).

Conclusion:

A cohort of patients with AHA normal BP, identified by unsupervised machine learning models, had significantly higher risk of preeclampsia and neonatal SGA than those with AHA abnormal BP or AHA normal BP clustered in the low-risk group.