Endometriosis and Early-Onset Cardiovascular Disease Across Levels of Systemic Inflammation: A Cross-Sectional Study of NHANES 1999–2006

This is a cross-sectional study using data from the 1999–2006 cycles of the National Health and Nutrition Examination Survey (NHANES) [ 39 ]. These survey cycles were selected because they included publicly available data on self-reported endometriosis diagnosis ( https://wwwn.cdc.gov/nchs/nhanes/ ). NHANES is administered by the National Center for Health Statistics (NCHS) and collects person-level demographic, health, nutrition and disease status information of the civilian, non-institutionalized U.S. population at two-year intervals, using a complex stratified, clustered, and four-stage sample design[ 39 ]. Data are collected through household interviews, physical examination and laboratory measures at a mobile examination center (MEC) [ 39 ]. The study was approved by the NCHS ethics review board, and all participants provided written informed consents prior to enrollment. Our study comprised of 41,474 participants from the 1999–2006 NHANES survey cycles, of which 21,210 were female. NHANES administers the reproductive health questionnaire to females aged ≥12 years (n = 15,439) at the mobile examination center (MEC) using the computer-assisted personal interviewing system. Of these, 6,508 were aged 20–54 years and 5,575 provided a response to self-report endometriosis (including “don’t know/refused”). After excluding participants who responded “don’t know” on self-report endometriosis diagnosis (n=18) and those with missing CVD or covariate data, the final analytic sample included 5,345 women ( Figure 1 ). The exposure measure was self-report of endometriosis diagnosis by a physician or health professional via the following survey item on the reproductive health questionnaire: “Has a doctor or other health professional ever told that you had endometriosis? ” Any woman who self-reported “Yes” to the question was considered to have endometriosis and those who responded “No” were not. With an overall confirmation proportion of 84% (range: 72% to 95%) there is evidence to suggest that self-report endometriosis diagnosis is reasonably valid [ 40 ]. The outcome measure was early-onset CVD defined as a self-reported physician diagnosis of any of the following conditions: coronary heart disease, angina/angina pectoris, heart attack, congestive heart failure, or stroke before age 55. Early-onset CVD status was determined using response from the survey items on the medical conditions’ questionnaire, which asked participants, “Has a doctor or other health professional ever told you that you had (specific condition - e.g., coronary heart disease )? Participants who responded “yes” to at least one condition were classified as having CVD, and those who responded “no” to all were classified as not having CVD. Self-report CVD has demonstrated good validity in prior work [ 41 , 42 ]. The potential effect modifier of interest was systemic inflammation, assessed using two complementary biomarkers: CRP and SII. Both biomarkers were assessed from blood samples collected during the mobile examination visit using a standardized protocol [ 43 ]. CRP was quantified using high-sensitivity latex-enhanced nephelometry. Due to sparse cell counts of early-onset CVD events, CRP was categorized as low/moderate (≤3 mg/L) and high (>3 mg/L) to improve model stability while maintaining clinically meaningful cut points [ 44 , 45 ]. SII was calculated as platelet count × neutrophil count / lymphocyte count [ 38 ]. Hematology testing is performed on blood specimens collected from participants and evaluated for neutrophil (10 3 cells/μL), segmented neutrophil (10 3 cells/μL), and platelet (10 3 cells/μL) counts. Cell counts are determined using the Coulter MAXMs method (Beckman Coulter, Miami, FL, USA) [ 43 ]. Values outside physiologically plausible ranges for lymphocyte (n=1) and platelet counts (n=2) were treated as missing (lymphocytes >40 ×10 3 cells/μL; platelets <50 or ≥999 ×10 3 cells/μL), affecting fewer than 1% of observations. SII was divided into three survey-weighted tertiles: low (67th percentile). Covariate selection was informed by a directed acyclic graph (DAG) ( Figure 2 ) constructed based on prior literature to identify a minimal sufficient adjustment set. Selected covariates represent factors associated with both endometriosis and CVD risk and included age, race/ethnicity, body mass index (BMI), smoking status, education, insurance status, age at menarche, and physical activity [ 46 – 50 ]. Age (years), age at menarche (years) and BMI (kg/m 2 ), calculated from measured height and weight were treated as continuous variables. Smoking status was categorized into current, former and never smokers. Participants’ educational level was categorized as less than high school, high school, or college and above. For insurance, participants were classified as “Insured” if they reported having any form of health insurance and “Uninsured” if they reported no insurance. Race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and Others). Physical activity was measured using a survey item asking participants to select the description that best reflected their usual daily activity level. Responses were grouped into three categories: low (mostly sitting), moderate (standing or walking frequently), and high (lifting loads, climbing stairs/hills, or performing heavy work). This item has been validated against accelerometry counts and used in prior NHANES studies [ 51 , 52 ]. To account for the complex, multistage sampling design of the NHANES, we conducted all the statistical analyses with the appropriate sampling weights [ 39 ]. Since both the reproductive health questionnaire items and laboratory data were collected during the mobile examination visit, we used the recommended MEC weights consistent with National Center for Health Statistics (NCHS) analytic guidelines [ 53 ]. To account for pooling multiple survey cycles (1999–2006), 4-year MEC weights (WTMEC4YR) for the 1999–2002 cycles were divided by two, and 2-year MEC (WTMEC2YR) weights for 2003–2006 were divided by four to create an 8-year combined MEC weight, following NCHS analytic guidelines [ 53 ]. Descriptive statistics were used to compare characteristics between women with and without self-report endometriosis. We used survey-weighted means and standard errors to summarize continuous variables for approximately normally distributed variables, while non-normal distributions were summarized as medians and interquartile ranges (IQRs). Categorical variables were summarized using weighted frequencies and percentages. Group differences by self-report endometriosis diagnosis was assessed using weighted chi-square test for categorical variables and weighted t-tests or Wilcoxon rank-sum tests for normally distributed variables and for skewed continuous variables respectively. To address our study aims and reduce bias due to measured confounding, stabilized inverse probability of weighting (IPW) was used to balance measured covariates between women with and without endometriosis [ 54 , 55 ]. IPW was chosen because early-onset CVD is a relatively rare outcome and the number of covariates was large relative to the number of events, conditions under which propensity-based methods such as IPW have been shown to yield more stable, less biased, and more robust estimates than conventional regression adjustment [ 56 , 57 ]. In addition, IPW improves covariate balance and allows estimation of a marginal (population-average) effect that aligns with our study aims. Propensity scores were estimated using an unweighted logistic regression model [ 55 ], that included the covariates from the minimally sufficient adjustment set (age, race/ethnicity, BMI, smoking status, education, insurance status, age at menarche, and physical activity). This approach of variable selection into the propensity score model was chosen because prior studies suggest that including confounders and outcome-related covariates reduces bias more effectively than restricting the model to biological risk factors alone or variables that affect an exposure but not the outcome [ 56 , 58 , 59 ]. Stabilized IPWs were constructed using the marginal probability of endometriosis and applied to create a weighted pseudo-population in which measured covariates were balanced across exposure groups. Covariate balance before and after weighting was assessed using standardized mean differences (SMD), with values ≤0.10 indicating adequate balance [ 54 ]. In cases of residual imbalance, the propensity score model was modified by incorporating nonlinear terms in the relevant covariates (e.g., cubic terms) [ 54 ]. To obtain nationally representative marginal estimates, stabilized IPW were multiplied by NHANES survey weights to generate a combined weight that accounted for both confounding adjustments using IPW and the complex survey design (e.g., stratification and clustering) [ 55 , 60 ]. The estimand of interest was the stratum-specific marginal association between endometriosis on early-onset CVD across levels of systemic inflammation on the odds-ratio scale. We fit weighted outcome models using survey-weighted logistic regression with the combined weight to estimate prevalence odds ratios and corresponding 95% confidence intervals. Since the CVD prevalence was low (3.1%) ( Table 1 ), these odds ratios closely approximate prevalence ratios [ 61 , 62 ]. Effect modification by systemic inflammation was assessed using stratified estimates and interaction terms in weighted models. Statistical significance was defined as a two-sided p < 0.05. All analyses were performed in R (version 4.3.2). We examined the distribution of estimated propensity scores to verify the overlap assumption. Figure 3 shows the kernel density plots of the propensity score among women with and without endometriosis. The two distributions demonstrated clear common support, with substantial overlap across most of the propensity score range. A small region of non-overlap was observed in the lower tail of the control distribution; however, overall overlap was sufficient to support IPW estimation. Stabilized inverse probability weights showed good overlap, with a mean near one and no evidence of extreme values (99th percentile = 1.97). Therefore, truncation or trimming of weights was not applied. Before weighting, several covariates showed substantial imbalance (unweighted SMDs up to 0.49). After application of IPW, balance improved markedly across all variables. Age showed some residual imbalance (SMD = 0.19) ( Figure 4 ), likely reflecting a nonlinear association with endometriosis ( Figure 5 ). To address this, age was modeled using restricted cubic splines in the propensity score model, which resulted in acceptable balance across all covariates (Supplementary Table 1). Overall missingness across study variables was low (≤5%), with the highest proportion observed for inflammatory markers (CRP and SII) (Supplementary Table 2). Missingness in CRP was not associated with endometriosis (p = 0.298) or CVD status (p = 0.302). Similarly, SII missingness was not associated with endometriosis (p= 0.079) or CVD (p= 0.347). Given the low level of missingness and absence of associations with key study variables, we conducted a complete-case analysis, restricting models to participants with non-missing data on the exposure, outcome, and all covariates. [ 63 ]. To evaluate the robustness of the IPW estimates, we conducted a survey-weighted logistic regression for the overall association between endometriosis and early-onset CVD. As a sensitivity analysis, participants with CRP values greater than 10 mg/L were excluded to reduce the influence of severe acute infection, ensuring findings represent chronic or low-grade inflammation rather than transient acute illness [ 64 , 65 ]. Finally, in addition to SII tertiles in the primary analysis, we used log-transformed SII and survey-weighted median split to test for interaction on the multiplicative scale.

As shown in Table 1 , our study included a total of 5,345 women aged 20–54 years, with a mean age of 37.2 ± 0.2 years. The survey-weighted prevalence of self-report endometriosis diagnosis was 6.9% and that for self-report early-onset CVD was 3.0%. The median (IQR) of CRP was 2.3 mg/L (0.8–5.7 mg/L) and SII was 572 (418.5, 790.6). The majority of the participants were non-Hispanic White women (46.2%) and had health insurance (76.8%). Approximately 53.5% had more than high school education, roughly one-quarter (21.8%) were current smokers and 19.6% reported high physical activity. The mean age at menarche was 12.6 ±0.0 years, while median (IQR) BMI 26.48 kg/m 2 (22.71, 32.21 kg/m 2 ). Women with endometriosis were slightly older and more likely to be Non-Hispanic White and insured. They also had higher levels of CRP and SII and were more likely to report low physical activity, an earlier menarche, current smoking, and early-onset CVD compared with women without endometriosis. BMI was similar across groups. In the unadjusted analysis, women with endometriosis had higher odds of early-onset CVD (POR = 2.35, 95% CI: 1.47–3.76). After IPW adjustment, the association remained significant, but was attenuated (aPOR = 1.91, 95% CI: 1.23–2.97) ( Table 2 ). When stratified by CRP categories, endometriosis was associated with higher odds of early-onset CVD among women with CRP >3 mg/L (aPOR 2.33, 95% CI: 1.49–3.66), whereas no association was observed among those with CRP ≤3 mg/L (aPOR = 0.91, 95% CI: 0.35–2.31). However, the overall interaction test was not statistically significant (p for interaction = 0.086), suggesting that there was no strong statistical evidence of effect modification ( Table 3 ). Table 4 shows the association between endometriosis and early-onset CVD across tertiles of SII. Among women in the lowest SII tertile, (aPOR = 1.25; 95% CI: 0.56–2.80) and highest SII tertile (aPOR = 1.72; 95% CI: 0.63–4.72), there was no evidence of an association between endometriosis and early-onset CVD. In contrast, women in the middle SII tertile had higher odds of early-onset CVD (aPOR = 2.92; 95% CI: 1.47–5.78). However, the global test for interaction was not statistically significant, also suggesting that there was no strong statistical evidence of effect modification (p for interaction = 0.288) in our study. ( Table 4 ). Using the survey-weighted logistic regression, women with endometriosis still had higher odds of early-onset CVD compared to women before (aPOR =2.35, 95% CI: 1.47–3.76) and after adjustment (aPOR =1.96, 95% CI: 1.22–3.15) ( Table 2 ). Overall, the magnitude and direction of associations were consistent with those in the primary analysis. To assess the robustness of our findings to the influence of acute inflammation, we conducted a sensitivity analysis excluding participants with CRP values >10 mg/L as shown in Table 5 . In this restricted sample (N = 4,286), endometriosis was not associated with early-onset CVD among women with CRP ≤3 mg/L (aPOR = 0.91, 95% CI: 0.35–2.31). Among women with CRP >3–10 mg/L, endometriosis remained associated with higher odds of early-onset CVD (aPOR = 2.25, 95% CI: 1.18–4.29) and the global interaction remained non-significant (p = 0.136). In sensitivity analyses, there was no significant interaction using log-transformed SII (p=0.618) and median split SII (p=0.566). In this nationally representative study of U.S. women aged 20–54 years, we examined the association between endometriosis and early-onset CVD and examined heterogeneity by systemic inflammation measured by CRP and SII. First, we found that endometriosis was associated with higher odds of early-onset CVD. Second, the formal tests for interaction were not statistically significant, but the magnitude of the association between endometriosis and early-onset CVD differed across the strata of the inflammatory markers. Specifically, we found that the association was strongest among women with high CRP levels and those in the medium SII tertiles, whereas no association was observed among women with low/moderate CRP levels or in the low and high SII tertiles. Our primary finding was that endometriosis was associated with higher odds of early-onset CVD. This observation is consistent with a growing body of evidence linking endometriosis to increased risk of overall CVD risk. For instance, prospective cohort studies and meta-analyses have reported modestly elevated risks of CVD events among women with endometriosis, including ischemic heart disease, stroke and composite CVD outcomes [ 19 , 32 , 33 ]. The biological plausibility linking endometriosis to CVD is well supported across multiple mechanisms. Genome-wide association studies have identified shared genetic loci between endometriosis and cardiovascular disease, suggesting common genetic susceptibility [ 66 – 68 ]. Beyond genetic overlap, endometriosis is associated with heightened oxidative stress and an atherogenic lipid profile, both of which are established contributors to atherosclerosis [ 69 ]. Additionally, endometriosis have been associated with elevated levels of high-density lipoprotein cholesterol (HDL-C) triglycerides, and pro-inflammatory cytokines including interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α) [ 70 , 71 ]. All of which promote endothelial dysfunction, a key early step in atherosclerotic plaque formation [ 69 ]. Importantly, because endometriosis is typically diagnosed during the reproductive years, prolonged subclinical exposure to these factors may contribute to the development of early-onset CVD at younger ages than would be otherwise expected as noted in our study. Our finding that the observed association between endometriosis and early-onset CVD was strongest among women with elevated CRP levels even after sensitivity analysis, may provide evidence to support the current hypothesis that the association between endometriosis and early-onset CVD may in part be driven by systemic inflammation [ 72 ]. Systemic inflammation creates an environment conducive to the growth and persistence of endometriosis lesions. In turn the ectopic endometriotic lesions become biologically active and capable of producing their own estrogen locally, which drives the growth of more lesions and fuels disease progression [ 73 , 74 ]. This repetitive cycle leads to more inflammation which is an established risk factor for CVD. It is important to note that the formal test for interaction was not statistically significant. This may be due to lack of statistical power or true absence of statistical interaction. Therefore, our findings may only point to CRP as a potential marker associated with a higher prevalence of early-onset CVD in women with endometriosis but requires more studies to confirm this. Finally, we found that the association between endometriosis and early-onset CVD was strongest among women in the medium SII tertile. SII been recognized as an important biomarker for CVD [ 75 ], and offers a more comprehensive assessment of systemic immune and inflammatory processes associated with endometriosis than CRP alone [ 76 ]. The strongest association between endometriosis and early-onset CVD was among women in moderate SII tertiles, but the formal interaction was not statistically significant. We hypothesize that this could be due to the following, first, the moderate SII tertiles may reflect chronic low-grade inflammation due to immune activation and inflammation, whereas the high SII tertiles may be as a result of acute infections and reflect a more transient state. Second, we may not have had sufficient power to detect meaningful effect modification. A prior study by Peng et al., [ 21 ] using NHANES data reported a nonlinear, S-shaped association between SII and endometriosis, with little association at lower SII levels and a stronger association emerging only after SII exceeded approximately 1100. In the present study, only 590 (11.0%) of our study population had SII values above 1100. Taken together, our findings show an association between endometriosis and early-onset CVD and suggest that this association may vary by systemic inflammation. However, the non-significant interaction tests limit inferences about statistical interaction and as such our results should be interpreted cautiously and explored in future studies with sufficient power. This study has several notable strengths. First, we used data from NHANES, a nationally representative survey allowing our findings to generalize to U.S. women aged 20–54 years. We also applied inverse probability of weighting to address confounding which provides more reliable effect estimates. By evaluating two complementary markers of systemic inflammation CRP and SII, we were able to extend current literature on the association of systemic inflammation in the endometriosis and prevalent early-onset CVD. To our knowledge, this is the first study to formally examine the potential effect modification by systemic inflammation in the association between endometriosis and early-onset CVD. However, several limitations should be considered. First, endometriosis and CVD were self-reported, which may introduce misclassification. Self-reported endometriosis is likely underdiagnosed, particularly among women with limited healthcare access. Because diagnosis depends on access to gynecologic care, misclassification may be differential with respect to socioeconomic status and healthcare utilization. If women with better access to care are more likely to receive both an endometriosis diagnosis and CVD diagnosis, this could bias associations away from the null. Second, the cross-sectional design eliminates the ability to establish temporality. Although we focused on early-onset CVD, the outcome reflects prevalent disease rather than incident events. As such, we cannot determine whether endometriosis or systemic inflammation preceded by early-onset CVD, resulted from it, or reflects shared underlying risk factors. Reverse causation remains possible. A prospective longitudinal design would be necessary to clarify the temporal sequence, which will be helpful in establishing causal inference. Third, systemic inflammatory markers were measured at a single time point. CRP and SII can fluctuate due to acute illness, circadian variation, medication use, or subclinical conditions. A single measurement may therefore misclassify inflammation levels. This misclassification is likely to be non-differential with respect to early-onset CVD and would generally bias effect estimate toward the null. Importantly, a single CRP value does not capture systemic inflammation. Fourth, although we used inverse probability weighting to address measured confounders, residual confounding due to reproductive history (e.g., parity, hysterectomy, oophorectomy), hormonal therapy use, disease severity, diet, medication use for endometriosis or CVD may influence both systemic inflammation and early-onset CVD which could obscure these associations. Additionally, smoking which is a major risk factor for CVD was categorized based on smoking status which is less reliable as it does not measure cumulative exposure to tobacco as does pack-years. While we used stabilized IPW combined with NHANES survey weights to reduce confounding by measured covariates which strengthens bias and confounding control relative to standard survey-weighted regression. However, combining multiple weighting can increase variance and reduce statistical precision, particularly for rare outcomes such as early-onset CVD and in stratified or interaction analyses. This may contribute to wider confidence intervals and non-significant interaction tests. Finally, the relatively small number of women with early-onset CVD limited statistical power to test for interaction, which may have reduced our ability to detect true effect modification. Although we observed patterns suggestive of heterogeneity by CRP and SII levels, our stratified analyses should be viewed as descriptive rather than causal and should be interpreted cautiously. Larger prospective studies will be needed to clarify these findings.

In this nationally representative sample of U.S. women aged 20–54 years, we observed that the cross-sectional association between endometriosis and early-onset CVD was stronger at moderate to high levels of systemic inflammation, as reflected by SII and CRP respectively. Longitudinal studies with detailed clinical data are needed to clarify timing, mechanisms, and the role of inflammation in the progression from endometriosis to early cardiovascular disease.

Cardiovascular disease (CVD) is the leading cause of mortality in women globally and in the United States [ 1 , 2 ], yet younger and middle-aged women are often underrecognized in risk assessment and prevention efforts. This gap has contributed to delayed diagnoses, missed opportunities for intervention, and a rising burden of early-onset CVD (i.e., cardiovascular events occurring before age 65) in women [ 3 ]. Early-onset CVD has substantial implications including higher rates of clinical comorbidities (e.g., metabolic syndrome, diabetes, and obesity), increased cumulative healthcare costs and an elevated lifetime risk of recurrent cardiovascular events [ 3 – 7 ]. Notably, women account for nearly half of incident cases of early-onset CVD despite generally lower prevalence of conventional cardiovascular risk factors [ 8 ]. This pattern may reflect two factors. First, these conventional risk factors (e.g., diabetes, smoking, and hypertension) appear to have stronger harmful effects such as an increased risk of CVD in women, compared to men with similar risk factors [ 9 ]. Specifically, diabetes was associated with a 3–7 times higher risk of CVD among women compared with a 2–3 times higher risk in men [ 10 , 11 ]. Further, smoking increases the relative risk of CVD in women by 25–55% despite lower smoking prevalence [ 12 ]. Second, growing evidence underscores the role of sex-specific, gynecologic conditions such as endometriosis in shaping CVD risk [ 13 – 15 ]. Endometriosis is a chronic, estrogen-dependent inflammatory condition affecting approximately 5–10% of reproductive-aged women [ 13 , 14 , 16 , 17 ]. It is defined by the growth of endometrial-like ectopic tissues and stroma outside the uterine cavity and primarily inside the ovaries and pelvic peritoneum [ 17 , 18 ]. Signs and symptoms include chronic pelvic pain, dysmenorrhea, dyspareunia, dyschezia (pain during defecation) and reduced fertility [ 19 , 20 ]. Shared inflammatory and immune pathways are hypothesized as a leading mechanism that may explain the association between endometriosis and CVD [ 21 – 24 ]. The endometrial ectopic tissues recruit and activate immune cells like macrophages, lymphocytes and neutrophils, which trigger the release of inflammatory markers such as C-reactive protein (CRP). [ 25 – 27 ]. At elevated levels, these inflammatory markers may exacerbate chronic inflammatory processes and affect the mechanism of action of endothelial nitric oxide synthase, which results in endothelial cell dysfunction (i.e., damage to the vascular endothelium), a key process implicated in atherosclerosis and the development of CVD [ 26 , 28 – 30 ]. Since endometriosis is typically diagnosed during the reproductive years [ 31 ], women with endometriosis may experience prolonged exposure to chronic systemic inflammation for decades before cardiovascular risk is typically assessed, potentially contributing to an earlier-onset of CVD. However, most studies have examined overall CVD risk without considering whether endometriosis contributes to an earlier onset of CVD [ 32 , 33 ]. Additionally, although chronic inflammation has been documented in endometriosis [ 22 , 34 ], the extent to which it may modify the association between endometriosis and early-onset CVD has not been formally tested. Therefore, our study aims to examine the association between endometriosis and early-onset CVD and assess whether it differs across levels of systemic inflammation measured by CRP and the systemic immune-inflammation index (SII), using data from the National Health and Nutrition Examination Survey (NHANES). Clarifying this heterogeneity may help identify women with a higher burden of early-onset CVD. We examined both markers because while elevated levels of CRP may signify chronic low-grade inflammation characteristic of chronic condition like endometriosis, [ 35 , 36 ], it may not fully capture the complex immune-inflammatory interactions that occur in endometriosis [ 37 ]. In contrast, SII is a composite measure that integrates the inflammatory, immune activation, and thrombotic mechanisms that are central to the pathogenesis of both endometriosis and CVD [ 37 , 38 ], making it a complementary and biologically meaningful measure.