This study used data from the China PEACE-MPP (China Patient-centered Evaluative Assessment of Cardiac Events Million Persons Project), a large-scale cardiovascular disease screening program (2016–2023) (17, 24). Community-dwelling residents aged 35–75 years residing in project areas for ≥6 months were recruited via community mobilization. Written informed consent was obtained, and the Fuwai Hospital Ethics Committee approved the protocol. Data included face-to-face interviews using a standardized questionnaire (demographics, lifestyle, medical history), anthropometrics (height, weight for Body Mass Index calculation), and fasting finger-prick blood lipid measurements (TC, TG, LDL-C, HDL-C) using the CardioChek analyzer (a widely used and reliable analyser in epidemiological investigations) (25). This analysis includes 44,304 elderly individuals (65–75 years) from Jiangsu Province communities who completed examinations and interviews.

Analyses used R 4.3.2. Descriptive statistics summarized participant characteristics by dyslipidemia subtype [mean ± SD or median [IQR] for continuous; n (%) for categorical; Chi-square tests for group differences]. To address missing data, multiple imputation by chained equations (MICE) was utilized. Correspondence analysis (CA) explored associations between dyslipidemia subtypes and patterns (22). Binary logistic regression assessed factors associated with overall dyslipidemia (any abnormality). Multinomial logistic regression (MLR) identified factors associated with specific dyslipidemia patterns (reference: no dyslipidemia) (23). Odds ratios (ORs) with 95% Confidence Intervals (CIs) were calculated. MLR P-values were Bonferroni-adjusted. Forest plots visualized associations (protective green, risk red). Results for patterns with n < 100 (“Other patterns”) are in Supplementary Figure and considered exploratory. Gender-stratified subgroup analyses were performed for CA and logistic regressions.

Odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) were calculated to quantify the associations between variables and both overall dyslipidemia and the specific dyslipidemia patterns. To control for the family-wise Type I error rate arising from multiple comparisons, P-values generated from the MLR were adjusted using the Bonferroni correction method. Forest plots were generated for all patterns to visually represent these associations, clearly indicating protective factors in green and risk factors in red. Results for less prevalent patterns (n < 100), which were grouped as “Other patterns”, are presented in the Supplementary Figure. Logistic regression models require a minimum of approximately 10 events per predictor variable to ensure stable parameter estimates and adequate statistical power (23). Due to the limited sample size, these results should be considered exploratory.

We performed gender-based analyses for both correspondence analysis and two types of logistic regressions to examine gender-specific risk factors associated with dyslipidemia patterns. These analyses allowed us to explore the influence of dyslipidemia subtypes, demographic characteristics, and health status within each gender.

Among 44,304 participants, 13,960 (31.5%) had dyslipidemia. Prevalence rates for subtypes were: Hyper_TC 7.2%, Hyper_TG 14.6%, Hyper_LDL 4.1%, and Hypo_HDL 16.5%. Table 1 details participant characteristics stratified by subtype. Females were more prevalent in Hyper_TC, Hyper_TG, and Hyper_LDL groups, while males predominated in the Hypo_HDL group. Significant differences (P < 0.05) between dyslipidemic and non-dyslipidemic groups were observed for residence, income, education, smoking, alcohol use, BMI level, and comorbidity number, but not insurance status. Lipid levels (TC, TG, LDL-C, HDL-C) and BMI also differed significantly across subtypes.

Table 2 shows pattern distributions. The most common overall patterns were “0 0 0 1” (Hypo_HDL, 36.5%) and “0 1 0 0” (Hyper_TG, 24.6%), followed by “0 1 0 1” (Hyper_TG & Hypo_HDL, 13.2%). Major patterns differed by gender: “0001” (Hypo_HDL) was most common in males (54.6%), while “0 1 0 0” (Hyper_TG) led in females (31.9%). Females had proportionally more Hyper_TC patterns (“1 X X X”), whereas males had a substantially higher proportion of the “0 0 0 1” pattern.

Correspondence analysis (Figure 1) revealed pattern structures. Hypo_HDL and Hyper_TG contributed most to simple patterns (Figure 1A). Dimensions 1 and 2 explained >80% of the association (Figure 1C). Hyper_TC and Hyper_LDL clustered together, strongly associated with mixed and complex patterns (Figure 1D). Hyper_LDL, in particular, showed low contribution to simple patterns (<15%) and higher contribution to complex ones, suggesting more complex metabolic dysregulation when LDL-C is high (Figures 1A,B). Gender differences were observed: males showed stronger tendencies towards “0001” and complex patterns involving Hyper_LDL or Hyper_TC compared to females (detailed breakdown in Supplementary Figures S1.1, S1.2).

Neither age nor gender was associated with overall dyslipidemia risk in this 65–75 age group (ORs = 1.00, 0.96; P > 0.05). Insurance and income were also not associated with overall dyslipidemia. However, being married, urban residence, higher education (junior high+), frequent smoking (≥3/week), higher BMI (≥24.0 kg/m²), and more comorbidities were significantly associated with higher odds of dyslipidemia (P < 0.05), while frequent alcohol consumption was protective (Supplementary Figure S2). Gender-specific differences emerged: age and frequent alcohol use conferred protection only in males; higher income was associated with increased dyslipidemia odds in males but with reduced odds in females. Moreover, higher education elevated dyslipidemia odds solely in males, and being married raised the odds exclusively in females (Supplementary Figures S3.1, S3.2).

Associations for the eight most prevalent patterns are shown in Figures 2, 3 (less prevalent patterns in Supplementary Figures S4, S5).

Marked gender dimorphism was evident. Females showed higher prevalence of patterns involving elevated TC, TG, and LDL-C, potentially linked to post-menopausal hormonal changes (16, 19, 26), while males predominantly exhibited Hypo_HDL patterns, possibly related to androgens or lifestyle factors like historically higher smoking rates and lower healthcare engagement (27–29). Interestingly, overweight/obesity paradoxically protected against the isolated Hyper_TC (“1 0 0 0”) pattern, an effect primarily seen in females, perhaps reflecting metabolically healthy obesity phenotypes or associated higher HDL_C levels (30).

Lifestyle factors showed complex associations. Frequent alcohol consumption demonstrated a dual association, particularly in males, being associated with lower odds of Hypo_HDL related patterns (“0 0 0 1”, “0 1 0 1”) but associated with higher odds for others involving Hyper_TG or Hyper_TC (“0 1 0 0”, “1 0 0 0”, “1 1 1 0”). No protective association was observed in females; instead, less frequent drinking was associated with the “1 0 1 0” pattern. The two-way link between alcohol and dyslipidemia is well-known (17, 18). Drinking alcohol in normal amounts is often related to higher HDL-C (“good” cholesterol), but drinking too much could lead to higher triglycerides and other lipid problems (21). Literature suggests alcohol elevates HDL-C through pathways like stimulating ApoA-I synthesis, enhancing reverse cholesterol transport, inhibiting CETP, and augmenting lipoprotein lipase activity (31). While these mechanisms explain HDL's functionality and its response to moderate alcohol, the overall benefit of HDL-C remains debated (32). Given alcohol's known health risks, our finding is an epidemiological observation, not a public health recommendation.

This study, conducted in Jiangsu Province –a province in China with a GDP per capita exceeding $20,000 USD –revealed notable associations between socioeconomic status and dyslipidemia in this elderly Chinese population. Socioeconomic status also revealed paradoxes. Urban living and higher education were associated with higher odds of dyslipidemia, especially in males (“0 0 0 1”, “0 1 0 1”), possibly reflecting adverse lifestyle shifts (diet, activity) accompanying urbanization despite potentially better healthcare access (8, 33–35). A notable observation was the elevated odds of low HDL-C (Hypo_HDL, pattern “0 0 0 1”) among men with higher education, likely stemming from increased sedentary behavior. This is supported by researches indicating that highly educated individuals tend to engage more frequently in sedentary activities, such as reading (36, 37). While occupational type was not collected in our dataset, this unmeasured factor could potentially mediate the observed association. Annual family income showed opposing associations by gender: higher income was associated with higher odds for the male-dominant “0 0 0 1” pattern but was associated with lower odds for the “1 0 1 0” pattern in females. These disparities might stem from gender differences in health prioritization and healthcare utilization influenced by economic standing. A research conducted in Guangdong Province, China, (economically strong province) has shown that individuals with higher incomes prioritize their health, especially in accessing chronic disease healthcare (38). And females' greater willingness to seek healthcare, compared to men, might be one contributing factor to this difference. These complicated associations, observed in elderly residents of a high-GDP Chinese province, highlighted the critical need to consider the specific socioeconomic and cultural context when assessing dyslipidemia risk in the elderly.

Key limitations include: First, the cross-sectional design, preventing causal inference, and the focus on a specific elderly (65–75 years) Chinese population, which may limit generalizability to other demographics. Second, POCT devices (the CardioChek analyzer) offer advantages for field screening, their accuracy/precision may differ somewhat from standard venous blood laboratory testing methods (25). Third, our “comorbidity number” variable relied on self-reported disease history, which is susceptible to recall bias and potential under-reporting.

The value of this research lies in revealing novel health paradoxes and association patterns that may emerge in regions at the forefront of economic transition and development, rather than aiming to provide generalizable conclusions for the entirety of East China. Our findings offer important insights into the health characteristics of economically advanced regions in China and lay the groundwork for future inter-regional comparative studies across a broader and more diverse East China.