Abstract 4369963: The Role of Sex-Specific Risk Factors in Sex Discordance unveiled by Artificial Intelligence-enhanced Electrocardiography: from the ELSA-Brasil study

Abstract

Introduction: Artificial intelligence–enhanced electrocardiography (AI-ECG) models can accurately predict sex, and sex misclassification is associated with adverse cardiovascular (CV) outcomes and more male-like cardiac (e.g. greater left ventricular mass and chamber volumes) and non-cardiac phenotypes (e.g. higher muscle mass, lower body fat) in women. However, the underlying factors contributing to sex discordance besides traditional CV risk factors—such as sex-specific CV risk factors—remain unexplored. Objective: To evaluate whether elevated AI-ECG sex-discordance scores are associated with sex-specific risk factors in women, while accounting for social determinants of health (SDoH). Methods: In the community-based ELSA-Brasil cohort baseline (2008-2010), we evaluated whether sex-discordance scores in women—measured by an AI-ECG model—were associated with female-specific CV risk factors: early menarche (≤11 years), menopause status, non-spontaneous menopause, multiparity, infertility, polycystic ovary syndrome, hormone replacement therapy >60 years or >10 years post-menopause, menstrual cycle length, history of abortion, hormonal contraceptives, history of eclampsia, and pregnancy weight gain >30 kg. The sex-discordant score (absolute difference between AI-predicted and self-reported sex, with 0=men, 1=women) was treated as a continuous variable. Associations were tested using multivariable robust linear regression with an M-estimator at 95% efficiency, adjusted for age, race, education, and per capita income. Results: Among 13,730 participants (mean age=52 years,SD:9.1; 54% women; 45% Black), higher sex-discordance scores were significantly associated with menopause (β = 0.092; 95%CI: 0.025–0.159), hormone and chemotherapy induced menopause (β = 0.215; 95%CI: 0.007–0.423), and multiparity (≥4 live births) (β = 0.169; 95%CI: 0.086–0.252), history of eclampsia (β = 0.157; 95%CI: 0.052–0.260), pregnancy-related weight gain >30 kg (β = 0.234; 95% CI: 0.117–0.352), early menarche (β = 0.154; 95% CI: 0.073–0.235), and use of hormonal contraceptives (β = 0.104; 95%CI: 0.002–0.207). All associations remained significant after adjustment for SDoH (Table 1). Conclusions: Higher AI-ECG sex-discordance scores are associated with multiple female-specific CV risk factors. These findings suggest that the score may serve as a novel biomarker for identifying women at increased CV risk.

Ähnliche Arbeiten