Associations between lipid profiles, body mass index, race/ethnicity, and endometriosis diagnosis in a nationally diverse U.S. cohort

Endometriosis is a chronic inflammatory gynecologic disorder associated with pelvic pain, infertility, and systemic metabolic alterations [1]. Despite growing recognition of metabolic involvement in disease pathophysiology, the combined role of metabolic and demographic factors remains incompletely defined. This study aimed to evaluate complementary associations of body mass index (BMI), lipid profiles, and race/ethnicity with recorded endometriosis diagnosis using the All of Us Research Program database. In this retrospective observational analysis, participants aged 18–45 years with a recorded endometriosis diagnosis were compared with participants without a recorded diagnosis using propensity score–based methods. BMI differences were assessed using Welch t-tests stratified by race, and racial associations were evaluated using chi-square testing and multivariable logistic regression. Lipid analyses — including HDL, LDL, total cholesterol, and triglycerides — utilized matched regression models and inverse probability of treatment weighting (IPTW) sensitivity analyses adjusted for age and BMI. Endometriosis cases had modestly lower BMI than controls in the matched-design analysis (mean difference, − 1.19 kg/m²; 95% CI, − 1.97 to − 0.41; P =0.003), although formal interaction testing indicated heterogeneity across race/ethnicity (P for interaction=0.020). Race/ethnicity was associated with recorded endometriosis diagnosis, with Asian and Hispanic or Latino participants demonstrating lower odds than White participants. In the primary matched lipid analyses, triglycerides were higher among cases in the adjusted model, whereas HDL, LDL, and total cholesterol differences were not statistically significant. IPTW sensitivity analyses yielded directionally similar estimates and statistically significant differences across all four lipid outcomes. Race/ethnicity, BMI, and lipid measures were associated with recorded endometriosis diagnosis in this EHR-based population, although the direction of BMI differences varied across demographic subgroups and lipid findings differed in statistical significance across analytic approaches. Because socioeconomic factors, healthcare access, treatment exposures, fasting status, and temporality were not fully characterized, these associations should not be interpreted as causal. Prospective studies incorporating longitudinal metabolic measurements and more comprehensive clinical and social covariates are warranted.