This study employed a case-control design. The sample size was calculated using the following parameters: an odds ratio (OR) of 1, a significance level ($\alpha$) of 0.05, a power (1 - $\beta$) of 0.9, and minor allele frequencies for the rs2000813 T allele of 0.22 (PMAF1) and 0.30 (PMAF2), sourced from the National Center for Biotechnology Information (NCBI) database. Using the two-proportion formula in PASS software (version 21.0.3, Kaysville, UT, USA), we estimated a required sample size of 1260 participants. To account for a 20% attrition rate, an additional 252 participants were included, yielding a total sample size of 1512. This was evenly divided into 756 participants in the coronary heart disease group and 756 in the control group.

Between January 2019 and December 2021, we enrolled 900 patients diagnosed with coronary artery disease (CAD) who underwent coronary angiography or percutaneous coronary intervention (PCI) at the First Affiliated Hospital of Xinjiang Medical University as the case group. Concurrently, 900 patients who did not meet the diagnostic criteria, as confirmed by coronary angiography or coronary computed tomography angiography (CTA) during the same period, were selected as the control group. Subsequently, 720 participants from each group were randomly allocated in an 8:2 ratio to form the modeling subset, with the remaining 180 participants from each group assigned to the validation subset.

Inclusion criteria: (1) Patients aged $\ge$18 years. (2) Patients hospitalized at the First Affiliated Hospital of Xinjiang Medical University between January 2019 and December 2021 with typical angina symptoms or non-invasive evidence of myocardial ischemia, a prerequisite for admission at this institution. (3) CHD diagnosis confirmed by coronary angiography, demonstrating coronary artery stenosis $>$50% in diameter.

Exclusion criteria: Patients were excluded if they met any of the following conditions: (1) Incomplete clinical data. (2) Presence of severe heart failure, cardiogenic shock, malignant arrhythmias, tumors, autoimmune diseases, or conditions associated with aortic stenosis. (3) Refusal to participate or noncompliance with the study protocol.

Inclusion criteria: (1) Participants aged $\ge$18 years. (2) Patients hospitalized at the First Affiliated Hospital of Xinjiang Medical University between January 2019 and December 2021 with typical angina symptoms or noninvasive evidence of myocardial ischemia, a prerequisite for admission at this institution. (3) Absence of CHD confirmed by coronary angiography or CTA.

Exclusion criteria: Patients were excluded if they met any of the following: (1) Incomplete clinical data. (2) Presence of severe heart failure, cardiogenic shock, malignant arrhythmias, tumors, or autoimmune diseases. (3) Refusal to participate or noncompliance with the study protocol.

(1) CHD: Diagnosis required typical angina symptoms and/or non-invasive evidence of myocardial ischemia, coupled with coronary angiography demonstrating $\ge$50% stenosis in the main trunk, left anterior descending branch (including the diagonal branch), left circumflex artery (including the marginal branch), or right coronary artery (including the posterior descending and posterolateral branch) [11, 12].

(2) Hypertension: Defined as systolic blood pressure $\ge$140 mmHg and/or diastolic blood pressure of $\ge$90 mmHg, measured at rest on at least two separate days, or a documented history of hypertension [13].

(3) Type 2 diabetes mellitus: Diagnosed based on random blood glucose $\ge$11.1 mmol/L (200 mg/dL), fasting glucose levels $\ge$7.0 mmol/L (126 mg/dL), 2-hour post-load glucose $\ge$11.1 mmol/L (200 mg/dL) on an oral glucose tolerance test, or a confirmed history of type 2 diabetes mellitus [14].

(4) Smoking calculator: A pack year is defined as twenty cigarettes smoked everyday for one year [15].

(5) Alcohol consumption: Defined as ethanol intake $\ge$28 g/d for men or $\ge$14 g/d for women, calculated as ethanol (g) = alcohol volume (mL) $\times$ ethanol percentage (%) $\times$ 0.8 [16].

Baseline data, including sex, age, body mass index (BMI), smoking history, alcohol consumption, hypertension, and type 2 diabetes mellitus, were collected from all participants.

After fasting for $>$12 hours, venous blood was collected from participants in the early morning into anticoagulant tubes. Serum samples were analyzed for the following indices: triglyceride (TG), total cholesterol (TC), high-density protein cholesterol (HDL-C), low-density protein cholesterol (LDL-C), apolipoproteins A and B (ApoA and ApoB), lipoprotein a (Lp(a)), serum uric acid (SUA), serum creatinine (Scr), white blood cell (WBC), neutrophil count (NE), platelet count, creatine kinase-MB (CK-MB), and cardiac troponins I and T (cTnI and cTnT). These measurements were performed by the Laboratory Department of the First Affiliated Hospital of Xinjiang Medical University using an automated biochemical analyzer (Olympus AU1000/2700, SC, USA).

Upon hospital admission, 3 mL of venous blood was collected from the antecubital vein of each participant and anticoagulated with 2% ethylenediaminetetraacetic acid (EDTA) disodium salt. Leukocytes were isolated by centrifugation, and genomic DNA was extracted using the phenol-chloroform method, then stored at –80 °C. The SNPscnTM high-throughput assay was employed to genotype the endothelial lipase gene loci rs2000813 and rs3813082. Based on nucleotide variations, the rs2000813 locus was classified into CC, CT, and TT genotypes, and rs3813082 locus into AA, AC, and CC genotypes (Supplementary Table 1).

Statistical analyses were conducted using SPSS software (version 26.0, IBM Corp., Armonk, NY, USA). Categorical data were reported as counts and percentages (n [%]) and compared using the chi-square ($\chi$²) test. For continuous data, the Kolmogorov-Smirnov test was employed to assess normality. Normally distributed continuous data were presented as means $\pm$ standard deviations and analyzed with independent-sample t-test for two-group comparisons or one-way analysis of variance (ANOVA) followed by the least significant difference (LSD) t-test for multiple-group comparisons. Non-normally distributed continuous data were expressed as medians with the 25th and 75th percentiles (M [P₂₅, P₇₅]) and evaluated using the Mann-Whitney U test for two groups or the Kruskal-Wallis H test for multiple groups. Spearman correlation, univariate, and multivariate logistic regression analyses were performed to identify risk factors of CHD in women. Gene-environment interactions between the endothelial lipase gene and CHD risk factors in women were assessed using multifactor dimensionality reduction (MDR) software (version 3.0.2, Computational Genetics Laboratory, Hanover, NH, USA). Receiver operating characteristic (ROC) curves were generated with GraphPad Prism (version 8.0.2, GraphPad Software, Inc., San Diego, CA, USA) to evaluate the predictive performance of the model for CHD risk in women. Statistically significance was defined as p 0.05.

Baseline characteristics (Table 1) revealed no significant differences between the case and control groups in sex distribution, age, alcohol consumption, TG levels, ApoA, SUA, Scr, or other variables. However, the case group exhibited significantly higher prevalence rates of smoking (44.70% vs. 36.60%, p = 0.001), hypertension (48.60% vs. 43.10%, p = 0.023) and diabetes (26.20% vs. 11.90%, p 0.001) compared to the control group. Additionally, the case group had higher BMI, TC, LDL-C, Lp(a), WBC, NE, platelet (PLT), CK-MB, cTnI and cTnT levels, with specific differences noted for HDL-C (1.00 $\pm$ 0.29 vs. 1.08 $\pm$ 0.32, p 0.001), and ApoB (0.87 [0.73, 1.02] vs. 0.84 [0.68, 0.98], p 0.001). Conversely, HDL-C and ApoB levels were significantly lower in the case group.

Supplementary Table 2 presents the Hardy-Weinberg equilibrium (HWE) test results. Genotype distributions for both the CHD and control groups conformed to HWE, indicating that the study population represents a Mendelian population in genetic equilibrium. This finding supports the genetic representativeness the sample, a critical factor for the validity of genetic association studies. Additionally, haplotype linkage disequilibrium (LD) analysis of the rs2000813 and rs3813082 loci in the endothelial lipase gene was performed using the SHEsis online platform. The results yielded a D’ value of 0.158 and an r² value of 0.001, suggesting no significant LD between these loci. Thus, rs2000813 and rs3813082 are genetically independent, an important consideration for evaluating their individual and combined effects on CHD susceptibility.

Supplementary Table 3 summarizes the distribution of endothelial lipase gene polymorphisms between the CHD and control groups. No significant differences were observed in the distribution of rs2000813 genotypes (CC, CT, TT) and alleles (C, T), or in rs3813082 genotypes (AA, AC, CC) and alleles (A, C) between the groups (all p $>$ 0.05). These findings suggest that, when considered independently, these polymorphisms do not significantly distinguish the CHD group from the control group in this study.

Supplementary Table 4 details the distribution of endothelial lipase genes polymorphisms in male participants, comparing the CHD and control groups. No significant differences were observed in the distribution of rs2000813 genotypes (CC, CT, TT) and alleles (C, T), or rs3813082 genotypes (AA, AC, CC) and alleles (A, C) between the CHD and the control groups among men (all p $>$ 0.05). This suggests that these polymorphisms are similarly distributed in both groups within the male population. Table 2 presents the distribution of endothelial lipase genes polymorphisms in female participants. In women, the rs2000813 CT genotype was significantly more prevalent in the CHD group than in the control group (49.30% vs. 37.80%, p = 0.040), whereas the CC genotype was less frequent (45.00% vs. 55.20%, p 0.05). No significant differences were found in the distribution of the TT genotype or C and T alleles between the groups. For rs3813082, the distribution of genotypes (AA, AC, CC) and alleles (A and C) showed no significant differences between the CHD and control groups in women (all p $>$ 0.05).

Supplementary Table 5 outlines the classification of blood lipid parameters, SUA, and routine hematological indices. These classifications are based on the 2022 Chinese Clinical Blood Lipid Test Guide [17], the 2019 Basic Diagnosis and Treatment Guide for Gout and Hyperuricemia [18], and the ninth edition of the Diagnostic Standard [19]. These authoritative guidelines provide a standardized framework for assigning values and interpreting the diverse biochemical and hematological data in this study, ensuring consistency and reliability.

Spearman correlation analysis revealed that positive associations between CHD in women and multiple factors, including lifestyle factors (smoking and alcohol consumption), clinical conditions (hypertension and type 2 diabetes mellitus), and biochemical markers (BMI, TC, ApoB, Lp(a), WBC, NE, and PLT). The rs2000813 CT genotype also showed a positive correlation with CHD, whereas the rs2000813 CC genotype was inversely associated with CHD in women. Univariate logistic regression analysis identified hypertension, type 2 diabetes mellitus, advanced age, elevated BMI, and increased levels of TC, ApoB, Lp(a), WBC, NE, and PLT as risk factors for CHD in women, each with distinct ORs. Multivariate logistic regression further confirmed the following as significant predictors: hypertension (OR, 2.305; 95% confidence interval (CI), 1.438–3.696; p 0.001), age $\ge$60 years (OR, 2.267; 95% CI, 1.392–3.692; p 0.001), BMI $\ge$28 kg/m² (OR, 8.634; 95% CI, 4.653–16.021; p 0.001), TC $\ge$6.2 mmol/L (OR, 6.437; 95% CI, 1.074–38.577; p = 0.042), ApoB $\ge$1.1 g/L (OR, 2.504; 95% CI, 1.138–5.512; p = 0.023), and the rs2000813 CT genotype (OR, 1.614; 95% CI, 1.010–2.579; p = 0.045). These factors, with their respective ORs and 95% CIs, were significant predictors of CHD in women. Detailed statistical results are provided in Table 3.

This study suggests that the rs2000813 CT genotype may increase the risk of CHD in women, while the rs3813082 locus of the endothelial lipase gene may influence lipid metabolism. To explore these genetic contributions further, we analyzed interactions between these loci and environmental factors—including smoking, alcohol consumption, hypertension, type 2 diabetes mellitus, age, BMI, and specific blood markers—using the MDR method. This approach constructed a multilevel interaction model integrating genetic and environmental variables, demonstrating strong predictive performance (Supplementary Table 6).

MDR analysis focused on interactions between the rs2000813 and rs3813082 loci of the endothelial lipase gene. Supplementary Fig. 1A illustrates risk-factor combinations associated with CHD in women. Dark-colored cells denote high-risk combinations, indicating increased CHD susceptibility, whereas light-colored cells represent low-risk combinations, suggesting reduced disease likelihood. White cells indicate no significant association with CHD. Supplementary Fig. 1B, highlights positive interactions between these loci in red, quantifying their individual main effects. Notably, rs2000813 exhibited a stronger influence on CHD risk than rs3813082.

Supplementary Fig. 2 presents the MDR analysis of interactions between the rs2000813 and rs3813082 loci of the endothelial lipase gene and environmental factors in in relation to CHD in women. Bar histograms distinguish the case group (left) from the control group (right), with dark cells indicating high-risk combinations, light cells representing low-risk combinations, and white cells denoting no significant association with CHD. This visualization elucidates the gene-environment interactions influencing CHD in women, offering insights into the complex interplay of genetic and environmental factors.

Supplementary Fig. 3 depicts interaction tree highlighting patterns among risk factors for CHD in women. Traditional cardiovascular risk factors-including neutrophil count, BMI, platelet count, and WBC-cluster on a single main branch, exhibiting a strong negative interaction (shown in dark blue). The introduction of the rs2000813 locus, alongside other risk factors such as alcohol consumption, age, smoking, lipoprotein(a), type 2 diabetes mellitus, hypertension, apolipoprotein B, and total cholesterol, attenuates this negative interaction, transitioning towards a positive interaction (indicated by green and yellow). Ultimately, a weak positive interaction emerges with rs3813082.

In the interaction ring network of the MDR model (Fig. 1), node values represent the information gain from individual attributes (main effects), while values between nodes reflect the information gain from attribute pairs (interaction effects). The main effects of single attributes, ranked in descending order, are as follows: BMI, NE, PLT, WBC, hypertension, age, type 2 diabetes mellitus, TC, ApoB, smoking, alcohol consumption, rs2000813, Lp(a), and rs3813082. Among the interaction effects, the strongest positive interaction occurs between rs2000813 and rs3813082, whereas the most pronounced negative interaction is observed between BMI and NE.

Fig. 1.

Interaction ring network of rs2000813 and rs3813082 loci and gene-environment effects in the multifactor dimensionality reduction (MDR) model. WBC, white blood cell; NE, neutrophil count; PLT, platelet; Lp(a), lipoprotein (a); ApoB, apolipoprotein B; TC, total cholesterol; BMI, body mass index.

Logistic regression analysis, conducted using SPSS software (version 26.0), generated prediction probabilities for four models. Model Y1 was defined by the rs2000813 CT genotype, while Model Y2 by the rs2000813 CC genotype. Model Y3 included traditional cardiovascular risk factors (e.g., smoking, hypertension, BMI), while Model Y4 combined all Model Y3 factors with both rs2000813 genotypes. ROC curve analysis revealed that Model Y3 (risk ratio index [RRI] $>$0.344; sensitivity, 82.5%; specificity, 62.7%) and Model Y4 (RRI $>$0.515; sensitivity, 68.6%; specificity, 77.2%) outperformed Model Y1 and Y2 in predicting CHD risk in women. Model Y4 demonstrated strong predictive performance, achieving area under the curve (AUC) values of 0.804 (95% CI, 0.766–0.843; p 0.001) and 0.793 (95% CI, 0.715–0.871; p 0.001) in the training and validation sets, respectively. Decision curves for the female patient training set show Model Y4 yields high net benefit in predicting female coronary heart disease risk, demonstrating its good clinical applicability. AUC values and sensitivity/specificity metrics are detailed in Table 4 and Fig. 2.

Fig. 2.

Performance evaluation of coronary heart disease risk models in female patients — ROC curves for training (A) and validation (B) sets, and decision curves for training set (C). AUC, area under the curve; ROC, receiver operating characteristic.

In this study, 1800 participants were enrolled and divided into a modeling subset (1440 participants) and a validation subset (360 participants) in an 8:2 ratio. These models were reconstructed and validated in female patients. Of these, 590 were women, with 470 (229 CHD, 241 Controls) in the modeling subset and 120 (51 CHD, 69 Controls) in the validation subset. Statistical analysis confirmed no significant differences in the distribution of factors such as smoking status, hypertension prevalence, or rs2000813 genotypes between the two subsets (p $>$ 0.05), as detailed in Table 5. For predicting CHD risk in women, Model Y4 exhibited the highest predictive accuracy, as detailed in Table 4 and Fig. 3.

Fig. 3.

Calibration diagram of female coronary heart disease risk model. (A) Calibration curve of predicting CHD risk in women in the Model Y1. (B) Calibration curve of predicting CHD risk in women in the Model Y2. (C) Calibration curve of predicting CHD risk in women in the Model Y3. (D) Calibration curve of predicting CHD risk in women in the Model Y4. CHD, coronary heart disease. Calibration curve likely adhered closest to the 45-degree ideal line, indicating minimal deviation between predictions and actual outcomes.