Abstract Body (Do not enter title and authors here): Background: Early identification of women at high risk for pregnancy-related cardiovascular (CV) complications has the potential to significantly reduce maternal mortality. Peak VO₂ is the gold-standard metric of cardiorespiratory fitness and has been shown to predict adverse outcomes in pregnant women, but ascertainment is costly and requires specialized equipment and expertise. We previously developed a deep learning model to accurately estimate peak VO₂ from the resting 12-lead electrocardiogram (ECG). We sought to examine the association of deep learning ECG-predicted peak VO₂ and incident pregnancy-related CV complications.

Methods: We ascertained ECG-estimated peak VO₂ among individuals who underwent clinical 12-lead ECG testing between 1 year prior to pregnancy and 13 weeks of gestation in a multi-institutional electronic health record cohort of pregnant women. Age-adjusted logistic regression models were used to examine the association of ECG-estimated peak VO₂ with subsequent pregnancy-related CV complications up to 1 year postpartum (maternal death, severe hypertensive disorders of pregnancy [HDP], and major adverse cardiac events [MACE]).

Results: Among 3650 pregnancies from 3437 women (mean age at delivery 33 ± 6 years), the median ECG-estimated VO₂ was 26.5 mL/kg/min, and 26% experienced a pregnancy-related CV complication. Lower ECG-estimated peak VO₂was associated with greater risk of a pregnancy-related CV complication (odds ratio [OR] 1.18 per 1-unit lower metabolic equivalent (MET = 3.5 kg/m²), 95% CI 1.15-1.23, p<0.001). Women in the lowest quartile of ECG-estimated peak VO₂had nearly twice the odds of developing a pregnancy-related CV complication compared with the highest quartile (OR 2.36, 95% CI 1.91-2.93, p<0.001, Figure 1).

Conclusions: Lower estimated cardiorespiratory fitness from a validated deep learning model based on resting 12-lead ECG is strongly and independently associated with higher risk of pregnancy-related CV complications. Artificial intelligence-enabled analysis of ECGs performed routinely in antepartum care may enable scalable risk assessment for identifying high-risk pregnancies in routine clinical settings.