Impact of lipid accumulation product on cardiometabolic multimorbidity and the mediating effect of leukocytes: evidence from a prospective cohort study among middle-aged and elderly chinese adults

Abstract

Background

Cardiometabolic multimorbidity (CMM) is a growing public-health burden in ageing populations, yet conventional adiposity measures such as body mass index and waist circumference do not adequately capture the visceral fat and lipid-metabolic disturbances that underlie it. We investigated the association between the lipid accumulation product (LAP) and incident CMM in middle-aged and older Chinese adults, and examined whether systemic inflammation mediates this association.

Methods

We analysed data from five waves (2011–2020) of the China Health and Retirement Longitudinal Study (CHARLS), comprising 7563 participants aged 45 years or older who were free of heart disease, stroke, and diabetes at baseline and had complete LAP and white blood cell (WBC) count data. CMM was defined as the coexistence of two or more cardiometabolic conditions among heart disease, stroke, and diabetes. The association between LAP and incident CMM was assessed using Cox proportional-hazards models with subgroup and restricted cubic spline analyses, and the mediating effects of inflammatory markers (WBC count and C-reactive protein [CRP]) were evaluated by causal mediation analysis.

Results

Over a median follow-up of 9.0 years, 486 participants (6.43%) developed CMM. After multivariable adjustment, LAP was positively associated with CMM: each one-unit increase in the base-10 logarithm of LAP (LgLAP) corresponded to a hazard ratio (HR) of 3.03 (95% CI 2.33–3.93), and risk rose across LAP quartiles (quartile 4 vs. quartile 1 h 3.41, 95% CI 2.49–4.67; P for trend < 0.001). Adding LAP to a base model raised the C-index from 0.624 to 0.674, with corresponding gains in the integrated discrimination improvement (IDI) and net reclassification improvement (NRI); the model including both LAP and WBC had a C-index of 0.677. Restricted cubic spline analysis showed a non-linear, positive dose–response relationship (P for non-linearity = 0.003). The association was stronger in normotensive than in hypertensive participants (HR 4.39 vs. 2.19; P for interaction = 0.020) but remained significant in both. In mediation analysis, WBC count and log-transformed CRP each significantly mediated the association, accounting for 5.1% and 8.4% of the total effect, respectively.

Conclusions

In middle-aged and older Chinese adults, a higher baseline LAP was associated with an increased risk of incident CMM, and systemic inflammation—indexed by WBC count and CRP—partially mediated this association. As an inexpensive and routinely available index, LAP may help identify people at risk of CMM, particularly among those without hypertension.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s42506-026-00235-0.

Introduction

Cardiometabolic multimorbidity (CMM), the co-occurrence of two or more cardiometabolic diseases such as stroke, diabetes, and cardiovascular disease, has become a major public-health challenge as populations age [1]. In China, the number of people with cardiovascular disease has reached 330 million, and related deaths have risen by 48% since 2005 [2]. Relative to a single condition, CMM markedly increases mortality, shortens life expectancy, and is often accompanied by cognitive decline and reduced quality of life [3–5]. Early identification of people at risk is therefore important, but the measures conventionally used for this purpose have clear limitations: body mass index (BMI) does not distinguish fat from muscle, and waist circumference does not separate subcutaneous from visceral fat [6]. Neither fully captures the metabolic processes that underlie CMM.

Dyslipidaemia promotes atherosclerosis and insulin resistance and thereby raises the risk of cardiovascular disease and diabetes [7]. But most studies have examined single lipid markers, such as total or low-density lipoprotein cholesterol, in relation to individual diseases [8, 9] rather than to multimorbidity. The lipid accumulation product (LAP), calculated from waist circumference and fasting triglycerides ([WC − 65] × TG in men and [WC − 58] × TG in women), combines visceral fat accumulation and lipid-metabolic dysfunction in a single index [10]. It outperforms conventional anthropometric measures in identifying metabolic syndrome and predicting cardiovascular disease [11, 12], is closely related to arteriosclerosis, and has been reported to mediate part of the association between frailty and major adverse cardiovascular and cerebrovascular events [13]. Because LAP reflects underlying insulin resistance [14, 15], which impairs endothelial nitric oxide synthesis and promotes vascular dysfunction [16, 17], it is mechanistically connected to cardiometabolic injury. Despite this growing evidence, prospective evidence on the relationship between LAP and incident CMM remains limited.

Inflammation is a central mechanism in the development of cardiometabolic disease and a plausible link between disordered lipid metabolism and multimorbidity. Visceral adipose tissue is not merely an energy store but an active endocrine organ; when it accumulates excessively, it releases free fatty acids and pro-inflammatory cytokines such as tumour necrosis factor-α and interleukin-6, promoting insulin resistance and endothelial dysfunction [18]. This low-grade inflammatory state impairs glucose homeostasis, raises the risk of type 2 diabetes, and accelerates atherosclerosis [19]. Circulating markers of this state, in particular the white blood cell (WBC) count and C-reactive protein (CRP), are inexpensive, easily measured, and reproducible, which makes them well suited to large-scale screening and follow-up. Whether inflammation mediates the association between LAP and CMM, however, has not been examined.

Using a nationwide prospective cohort of Chinese adults aged 45 years and older, we therefore aimed to evaluate the association between LAP and incident CMM and to quantify the extent to which inflammatory markers (WBC count and CRP) mediate it. By addressing the limited prospective evidence on LAP and CMM and the unexamined role of inflammation, this study seeks to identify accessible markers for the early warning of CMM and to inform prevention strategies centred on a modifiable target—visceral lipid accumulation—consistent with the goals of healthy ageing.

Methods

Study population and design

This study utilized data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative prospective cohort study targeting middle-aged and older adults aged 45 years and above (https://charls.pku.edu.cn/). The survey employed a multistage stratified probability-proportional-to-size (PPS) sampling technique, covering 150 county-level administrative units across 28 provinces nationwide, ensuring national representativeness and balanced distribution of geographic and socioeconomic characteristics. The baseline survey (Wave 1) completed in 2011 successfully recruited 17,708 eligible participants, achieving a high response rate of 80.5%. Subsequently, four follow-up surveys were conducted (Waves 2–5 in 2013, 2015, 2018, and 2020, respectively) to dynamically monitor the longitudinal evolution of health outcomes and risk factors.

Data collection was performed by interviewers who received standardized training, utilizing face-to-face interviews combined with computer-assisted personal interviewing (CAPI) systems to comprehensively record multidimensional information including participants’ sociodemographic characteristics, lifestyle exposures, medical history, and healthcare utilization.

The original CHARLS study was conducted in accordance with the Declaration of Helsinki and was approved by the Biomedical Ethics Committee of Peking University (approval number: IRB00001052-11015); all participants provided written informed consent. The present study was a secondary analysis of publicly available, de-identified CHARLS data, and no additional ethics approval was obtained for this analysis.

This study analyzed longitudinal data from five survey waves spanning 2011–2020. Wave 1 participants served as the baseline cohort, with continuous follow-up observations conducted across the subsequent four waves. Exclusion criteria included: (1) missing data on any of the following variables: waist circumference, triglycerides, or white blood cell count (6518 participants excluded); (2) diagnosed cardiovascular disease, stroke, diabetes, or fasting blood glucose ≥ 7.0 mmol/L at baseline (3627 participants excluded). After screening, 7563 participants were ultimately included in the analytical cohort (Fig. 1 flowchart).

Fig. 1.

Flowchart of study participants

Variable measurement and definition

Baseline data collection

The survey team obtained multidimensional data through standardized operating procedures, encompassing three major modules: structured questionnaire interviews, physical measurements, and laboratory tests. Questionnaire content covered domains including demographic characteristics, socioeconomic status, health status assessment, and behavioral lifestyle factors. Physical measurements included basic physiological parameters such as height, weight, blood pressure, and waist circumference. Blood pressure was measured using the standard seated position method with 45-second intervals between measurements; left arm blood pressure was measured three times consecutively, recording systolic pressure, diastolic pressure, and pulse rate for each measurement. The arithmetic mean of three measurements was calculated for analysis. Laboratory biomarker testing aimed to evaluate participants’ metabolic functional status and immune system activity.

Covariate definitions

Confounding factors included in this study were: sex, age, residential area, educational attainment, marital status, smoking habits, alcohol consumption, estimated glomerular filtration rate (eGFR), and hypertension. Residential area was dichotomized as urban versus rural; marital status was classified as married versus other; educational level was categorized into four groups (below primary school, primary school, middle school, and high school or above); smoking status was classified as current/former smoker versus never smoker; alcohol consumption status was classified as current/former drinker versus never drinker. Hypertension was defined as systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg, or self-reported physician-diagnosed hypertension. Malignant tumors, chronic pulmonary disease, dyslipidemia, and chronic kidney disease were all determined based on physician-confirmed diagnostic history.

Biospecimen testing

After overnight fasting, fasting venous blood samples were collected by medical personnel professionally trained by the Chinese Center for Disease Control and Prevention following standard operating procedures. Blood samples were uniformly transported to the Beijing central laboratory for centralized analysis, with testing parameters including: fasting blood glucose, glycated hemoglobin, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, serum uric acid, serum creatinine, hematocrit, hemoglobin concentration, and white blood cell count. Estimated glomerular filtration rate (eGFR) was calculated using the 2012 CKD-EPI creatinine–cystatin C equation [20].

Calculation of lipid accumulation product

The Lipid Accumulation Product (LAP) was calculated using the following sex-specific formulas [21]:

In the CHARLS blood-assay dataset, triglycerides are reported in mg/dL; values were therefore converted to mmol/L (by dividing by 88.57) before LAP was computed. To improve model robustness, the common (base-10) logarithm of LAP (LgLAP) was used as a continuous variable in the regression analyses. Participants were grouped into quartiles (Q1-Q4) according to the distribution of LAP in the present study population; the cut-off values were Q1 < 14.57, Q2 14.57 to < 25.72, Q3 25.72 to < 44.84, and Q4 > = 44.84.

Definition of cardiometabolic multimorbidity

Cardiovascular disease was defined as self-reported physician-diagnosed heart disease in the questionnaire survey (encompassing myocardial infarction, coronary atherosclerotic heart disease, angina pectoris, congestive heart failure, and other cardiac conditions). Stroke diagnosis was based on self-reported physician diagnosis in the questionnaire. Diabetes was diagnosed when meeting any of the following criteria: (1) self-reported physician diagnosis; (2) fasting blood glucose ≥ 7.0 mmol/L; (3) glycated hemoglobin ≥ 6.5%; (4) random blood glucose ≥ 11.1 mmol/L. For this study, heart disease was treated as a single composite condition: a participant reporting more than one cardiac diagnosis (for example, both coronary heart disease and heart failure) was counted as having one cardiometabolic condition, not several. Incident CMM was ascertained across the four follow-up waves: a participant was classified as an incident case at the first wave at which two or more conditions had accumulated, whether these arose simultaneously or sequentially during follow-up, and survival time was defined as the interval from baseline to that wave. Hypertension and isolated dyslipidaemia were not counted as components of CMM; hypertension was instead included as a covariate in the adjusted models. Cardiometabolic multimorbidity was thus defined as the concurrent presence of at least two of the three distinct disease categories, namely heart disease, stroke, and diabetes [3, 22].

Statistical analysis

Study participants were divided into four subgroups based on LAP quartiles, and baseline characteristics were described and compared across groups. Given the skewed distribution of data, continuous variables were expressed as median (interquartile range), and between-group comparisons were performed using the Kruskal-Wallis rank-sum test; categorical variables were expressed as frequency (percentage), and between-group differences were evaluated using the chi-square test.

Cox proportional-hazards regression models were used to quantify the association between LAP and incident CMM, estimating hazard ratios (HR) and 95% confidence intervals (CI); the proportional-hazards assumption was verified using scaled Schoenfeld residuals. Three nested models were constructed: Model 1 was unadjusted; Model 2 adjusted for age and sex; and Model 3, the primary model, further adjusted for educational level, marital status, residence, smoking status, alcohol consumption, eGFR, and hypertension. Body mass index (BMI) was deliberately not included in the primary model to avoid over-adjustment, because LAP incorporates waist circumference and is strongly collinear with BMI; the effect of this choice was examined in a stepwise model ladder that additionally added BMI and, separately, waist circumference (Supplementary Table S2). LAP was modelled both as a continuous variable (per one-unit increase in LgLAP) and as quartiles (Q1 as reference), with a test for linear trend. Cumulative incidence curves were plotted using the Kaplan-Meier method and compared using the log-rank test.

The incremental predictive value of LAP, and of WBC, beyond the Model 3 covariates was assessed using Harrell’s C-statistic (differences in C between nested models were tested with the paired comparison in the survival package), the integrated discrimination improvement (IDI), and the continuous net reclassification improvement (NRI) at 4, 7, and 9 years. Because the outcome is time-to-event, discrimination was additionally quantified with time-dependent receiver operating characteristic (ROC) analysis using inverse-probability-of-censoring weighting, and the area under the curve (AUC) at 9 years was reported. Restricted cubic spline (RCS) models with four knots were used to examine the dose-response relationship; when non-linearity was detected, the inflection point was estimated using a two-segment (piecewise-linear) Cox model with a grid search for the maximum-likelihood threshold, and the HRs below and above the threshold were reported.

Effect modification was examined in pre-specified subgroups defined by sex, educational attainment, marital status, residence, hypertension, smoking, and alcohol consumption, with multiplicative interaction tested by likelihood-ratio tests. The mediating roles of two inflammatory biomarkers, WBC count and CRP, were examined using causal mediation analysis within the counterfactual framework (regression-based approach, CMAverse package). Because CRP was highly right-skewed (skewness = 10.4), it was natural-log-transformed before analysis. For each mediator, a linear regression model for the mediator and a Cox model for the outcome were fitted, both adjusted for the identical Model 3 covariates; total, natural direct, and natural indirect effects were expressed as hazard ratios, the proportion mediated was calculated, and 95% CIs and P values were obtained from 1000 non-parametric bootstrap resamples. Because the exposure-by-mediator interaction was non-significant for both mediators, no interaction term was included. As a sensitivity analysis for CRP, the mediation was repeated after excluding participants with CRP > 10 mg/L. To quantify the potential impact of unmeasured confounding, E-values were calculated for the primary association and its confidence limit. The robustness of the LAP definition to its Western-derived constants was further assessed using China-specific waist thresholds (85 cm for men, 80 cm for women), a within-sex standardised LgLAP (per standard deviation), and sex-stratified analysis.

All statistical analyses were performed using R software version 4.3.3, with a two-sided P-value < 0.05 considered statistically significant.

Results

Baseline characteristics

Based on the pre-established inclusion and exclusion criteria, a total of 7563 participants were enrolled in this study. All participants were divided into four groups according to the quartiles of LAP, with baseline characteristics presented in the table. It can be observed that with increasing LAP, participants were more likely to be female, younger in age, urban residents, non-drinkers or non-smokers, obese, and to have hypertension and dyslipidemia. Elevated LAP was associated with higher levels of laboratory parameters, including white blood cell count, platelet count, total cholesterol, and C-reactive protein, while high-density lipoprotein cholesterol decreased with increasing LAP Table 1.

Table 1.

Baseline characteristics of study participants stratified by LAP Levels, China, 2011–2020

Characteristic	Overall (n = 7563)	Q1 (n = 1891)	Q2 (n = 1892)	Q3 (n = 1889)	Q4 (n = 1891)	P value
Age, years	58.0 (51.0–65.0)	59.0 (52.0–67.0)	58.0 (51.0–65.0)	57.0 (50.0–64.0)	57.0 (51.0–64.0)	< 0.001
BMI, kg/m²	22.9 (20.7–25.5)	20.2 (18.9–21.8)	22.3 (20.7–23.9)	23.9 (22.1–26.0)	26.0 (23.9–28.3)	< 0.001
Waist circumference, cm	84.0 (77.5–91.0)	75.2 (71.2–79.0)	82.0 (78.0–86.2)	87.2 (82.6–92.0)	93.0 (88.0–99.0)	< 0.001
LAP	25.7 (14.6–44.8)	9.4 (6.5–11.9)	19.6 (16.9–22.7)	33.9 (29.6–39.1)	66.5 (53.6–91.5)	< 0.001
WBC, ×10⁹/L	5.9 (5.0–7.2)	5.8 (4.8–7.0)	5.8 (4.8–7.1)	5.9 (5.0–7.1)	6.2 (5.2–7.5)	< 0.001
HGB, g/dL	14.2 (13.0–15.5)	14.1 (13.0–15.4)	14.1 (12.9–15.4)	14.1 (13.0–15.4)	14.3 (13.2–15.6)	< 0.001
PLT, ×10⁹/L	206 (162–254)	205 (162–247)	205 (161–253)	206 (162–255)	213 (167–263)	< 0.001
BUN, mg/dL	15.1 (12.5–18.2)	16.0 (13.0–19.1)	15.1 (12.6–18.3)	14.7 (12.3–17.6)	14.7 (12.3–17.4)	< 0.001
Crea, mg/dL	0.8 (0.6–0.9)	0.8 (0.7–0.9)	0.8 (0.6–0.9)	0.7 (0.6–0.9)	0.8 (0.6–0.9)	< 0.001
CysC, mg/L	0.9 (0.9–1.1)	0.9 (0.9–1.1)	0.9 (0.9–1.1)	0.9 (0.9–1.0)	0.9 (0.8–1.0)	< 0.001
eGFR, mL/min/1.73 m²	89.3 (77.7–97.8)	88.1 (76.6–97.3)	89.5 (77.2–97.3)	89.3 (78.7–97.7)	90.1 (78.1–99.4)	0.002
UA, mg/dL	4.3 (3.6–5.1)	4.3 (3.5–5.1)	4.1 (3.4–5.0)	4.2 (3.5–5.0)	4.6 (3.8–5.4)	< 0.001
TC, mg/dL	190 (167–215)	177 (158–201)	186 (165–209)	193 (170–216)	203 (177–230)	< 0.001
HDL-C, mg/dL	49.9 (41.0–60.7)	58.8 (49.9–69.2)	53.9 (45.6–63.0)	48.7 (41.0–57.2)	40.2 (34.0–47.9)	< 0.001
LDL-C, mg/dL	114 (92.8–136)	105 (87.0–125)	115 (95.8–135)	120 (99.4–142)	116 (90.8–143)	< 0.001
CRP, mg/L	1.0 (0.5–2.1)	0.8 (0.4–1.7)	0.8 (0.5–1.8)	1.0 (0.6–2.0)	1.4 (0.8–2.6)	< 0.001
HbA1c, %	5.1 (4.9–5.4)	5.1 (4.8–5.3)	5.1 (4.8–5.3)	5.1 (4.9–5.4)	5.2 (4.9–5.5)	< 0.001
Male, n (%)	3532 (46.7)	1310 (69.3)	880 (46.5)	710 (37.6)	632 (33.4)	< 0.001
Urban, n (%)	2506 (33.1)	484 (25.6)	579 (30.6)	680 (36.0)	763 (40.3)	< 0.001
Married, n (%)	6669 (88.2)	1655 (87.5)	1669 (88.2)	1669 (88.4)	1676 (88.6)	0.749
Smoker, n (%)	2970 (39.3)	1089 (57.6)	717 (37.9)	603 (31.9)	561 (29.7)	< 0.001
Drinker, n (%)	3140 (41.5)	998 (52.8)	771 (40.8)	717 (38.0)	654 (34.6)	< 0.001
Hypertension, n (%)	2824 (37.3)	486 (25.7)	600 (31.7)	755 (40.0)	983 (52.0)	< 0.001
Cancer, n (%)	58 (0.8)	16 (0.8)	15 (0.8)	12 (0.6)	15 (0.8)	0.892
Chronic lung disease, n (%)	669 (8.8)	207 (10.9)	171 (9.0)	152 (8.0)	139 (7.4)	< 0.001
Dyslipidemia, n (%)	495 (6.5)	46 (2.4)	85 (4.5)	145 (7.7)	219 (11.6)	< 0.001
Chronic kidney disease, n (%)	354 (4.7)	91 (4.8)	95 (5.0)	86 (4.6)	82 (4.3)	0.768
Cardiometabolic multimorbidity, n (%)	486 (6.4)	55 (2.9)	91 (4.8)	137 (7.3)	203 (10.7)	< 0.001
Education, n (%)						0.205
Below primary school	3716 (49.1)	924 (48.9)	929 (49.1)	927 (49.1)	936 (49.5)
Primary school	1628 (21.5)	440 (23.3)	422 (22.3)	381 (20.2)	385 (20.4)
Middle school	1505 (19.9)	370 (19.6)	356 (18.8)	389 (20.6)	390 (20.6)
High school or above	714 (9.4)	157 (8.3)	185 (9.8)	192 (10.2)	180 (9.5)

Abbreviations: WBC, white blood cell count; HGB, hemoglobin concentration; PLT, platelet count; BUN, blood urea nitrogen; Crea, serum creatinine; CysC, serum cystatin C; eGFR, estimated glomerular filtration rate; UA, uric acid; TC, total cholesterol; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; CRP, C-reactive protein; HbA1c, glycated hemoglobin

Participants were divided into four groups according to the quartiles of LAP: Quartile 1: LAP < 14.57; Quartile 2: LAP ≥ 14.57 and < 25.72; Quartile 3: LAP ≥ 25.72 and < 44.84; Quartile 4: LAP ≥ 44.84. Continuous variables were compared among the four groups using the Kruskal-Wallis test, and categorical variables were compared using the chi-square test

Association between LAP and cardiometabolic multimorbidity

During the 9.0-year follow-up, 486 participants (6.43%) developed CMM. Of these, 55 were in Quartile 1, 91 in Quartile 2, 137 in Quartile 3, and 203 in Quartile 4.

Table 2 shows the association between LAP and CMM. Analysed continuously, each one-unit increase in LgLAP (a 10-fold increase in LAP) was associated with a substantially higher risk of CMM in the fully adjusted model (HR 3.03, 95% CI 2.33–3.93). Across quartiles in Model 3, and relative to Quartile 1, Quartile 2 was associated with a 64% higher risk (HR 1.64, 95% CI 1.17–2.30), Quartile 3 with a 140% higher risk (HR 2.40, 95% CI 1.74–3.31), and Quartile 4 with a 241% higher risk (HR 3.41, 95% CI 2.49–4.67), with a significant linear trend across quartiles (P for trend < 0.001).

Table 2.

Association of LAP with Cardiometabolic Multimorbidity

	Cases/N (%)	Model 1 h (95% CI)	Model 2 h (95% CI)	Model 3 h (95% CI)
LgLAP, per 1 unit (10-fold increase)	486/7563 (6.43%)	3.48 (2.72–4.45)	3.64 (2.83–4.68)	3.03 (2.33–3.93)
Q1 (< 14.57)	55/1891 (2.91%)	Reference	Reference	Reference
Q2 (14.57–25.72)	91/1892 (4.81%)	1.67 (1.19–2.33)	1.72 (1.23–2.41)	1.64 (1.17–2.30)
Q3 (25.72–44.84)	137/1889 (7.25%)	2.55 (1.86–3.48)	2.66 (1.93–3.65)	2.40 (1.74–3.31)
Q4 (≥ 44.84)	203/1891 (10.74%)	3.85 (2.85–5.18)	4.04 (2.97–5.49)	3.41 (2.49–4.67)
P for trend				< 0.001

Abbreviations: CI, confidence interval; HR, hazard ratio

Model 1: Unadjusted for any covariates

Model 2: Adjusted for age and sex based on Model 1

Model 3: Adjusted for age, sex, education level, marital status, residence, smoking status, alcohol consumption, estimated glomerular filtration rate, and hypertension based on Model 1

Figure 2 depicts the cumulative incidence of CMM by LAP quartile.

Fig. 2.

Cumulative incidence curves of cardiometabolic multimorbidity

Predictive value of LAP for CMM

We evaluated the incremental predictive value of LAP for CMM (Table 3). Adding LAP to the base model improved discrimination, with Harrell’s C-statistic increasing from 0.624 to 0.674, an increase of 0.051 (95% CI 0.034–0.068; P < 0.001) that is markedly larger than the increment observed in the original model that had included BMI (0.008). Because WBC was found to mediate part of the association (Sect.  3.5), we also evaluated a model including both LAP and WBC; its C-statistic was 0.677. Compared with the basic model, the 9-year integrated discrimination improvement (IDI) was 0.010 (95% CI 0.006–0.018; P < 0.001) and the continuous net reclassification improvement (NRI) was 0.159 (95% CI 0.114–0.225; P < 0.001).

Table 3.

Predictive performance of LAP for CMM

Model	C-index	IDI (95% CI), P	NRI (95% CI), P
Basic model	0.624	Reference	Reference
Basic + LAP	0.674	4-yr: 0.003 (-0.001 to 0.007), P = 0.130 7-yr: 0.005 (0.001 to 0.010), P < 0.001 9-yr: 0.009 (0.005 to 0.015), P < 0.001	4-yr: 0.205 (-0.013 to 0.403), P = 0.068 7-yr: 0.130 (0.048 to 0.234), P < 0.001 9-yr: 0.172 (0.115 to 0.232), P < 0.001
Basic + WBC	0.633	4-yr: 0.002 (0.000 to 0.004), P = 0.006 7-yr: 0.002 (0.000 to 0.007), P = 0.004 9-yr: 0.001 (0.000 to 0.005), P = 0.064	4-yr: 0.286 (0.080 to 0.469), P = 0.010 7-yr: 0.094 (-0.005 to 0.180), P = 0.058 9-yr: 0.063 (0.000 to 0.119), P = 0.048
Basic + LAP + WBC	0.677	4-yr: 0.004 (0.000 to 0.011), P = 0.032 7-yr: 0.007 (0.002 to 0.014), P = 0.002 9-yr: 0.010 (0.006 to 0.018), P < 0.001	4-yr: 0.314 (0.058 to 0.470), P = 0.016 7-yr: 0.150 (0.063 to 0.241), P = 0.002 9-yr: 0.159 (0.114 to 0.225), P < 0.001

The basic model included age, sex, education level, marital status, place of residence, smoking status, alcohol consumption status, estimated glomerular filtration rate, and hypertension

Time-dependent ROC analysis showed a corresponding increase in the 9-year area under the curve after LAP was added (Fig. 3).

Fig. 3.

Time-dependent ROC curves for predicting 9-year incident CMM. The basic model includes age, gender, education level, marital status, place of residence, smoking status, alcohol consumption, estimated glomerular filtration rate, and hypertension

Multivariable-adjusted restricted cubic spline analysis indicated a non-linear, positive association between LAP and CMM (P for non-linearity = 0.003); a two-segment model located the inflection point at a LAP of approximately 74, with risk rising steeply below this value (Fig. 4).

Fig. 4.

Dose-response relationship between LAP and the risk of incident CMM, adjusted for covariates including age, sex, education level, marital status, place of residence, smoking status, alcohol consumption status, estimated glomerular filtration rate, and hypertension

Subgroup analysis

After adjustment for potential confounders, the association between LgLAP and CMM was consistent across subgroups defined by sex, educational level, marital status, residence, smoking status, and alcohol consumption, with no significant interaction (all P for interaction > 0.05). A significant interaction was observed only for hypertension (P for interaction = 0.020): the association was stronger in normotensive participants (HR 4.39, 95% CI 2.98–6.48) than in hypertensive participants (HR 2.19, 95% CI 1.54–3.09), although it remained statistically significant in both strata. Hypertension thus modified the magnitude, but not the direction, of the association outcome (Figs. 5 and 6).

Fig. 5.

Subgroup analysis for the association between LgLAP and CMM. Models were adjusted for age, sex, educational level, marital status, place of residence, smoking status, alcohol consumption status, estimated glomerular filtration rate, and hypertension, except for the variable used to define each subgroup

Fig. 6.

Mediation analysis of WBC and CRP in the association between LgLAP and CMM. The model was adjusted for age, sex, educational level, marital status, place of residence, smoking status, alcohol consumption status, estimated glomerular filtration rate, and hypertension

Mediation analysis

In the primary Cox model, each one-unit increase in LgLAP (a 10-fold increase in LAP) was associated with an HR of 3.03 (95% CI 2.33–3.93) for CMM. The indirect effect mediated through WBC corresponded to a hazard ratio of 1.04 (95% CI 1.01–1.06), accounting for 5.1% of the total effect (P = 0.016). C-reactive protein was strongly right-skewed at baseline (skewness = 10.4); after natural-log transformation, it also showed a significant mediating effect of magnitude similar to WBC, with an indirect-effect hazard ratio of 1.06 (95% CI 1.02–1.10), accounting for 8.4% of the total effect (P = 0.002). This mediating effect of log-CRP persisted after excluding participants with CRP > 10 mg/L (proportion mediated 13.7%, P < 0.001; Supplementary Table S4), indicating that it reflects chronic low-grade inflammation rather than acute-phase elevations. Full path coefficients and effect decompositions are provided in Supplementary Table S3.

Discussion

In this nationwide prospective cohort of middle-aged and older Chinese adults, a higher baseline LAP was independently associated with the subsequent development of CMM, and part of this association operated through systemic inflammation. This study extends the existing prospective evidence by evaluating baseline LAP in relation to incident CMM and by examining the potential mediating contribution of inflammatory markers. Our main findings were as follows: (1) LAP showed a graded association with CMM, with participants in quartiles 2 to 4 having 64%, 140%, and 241% higher risk, respectively, than those in quartile 1, and each one-unit increase in log-transformed LAP corresponding to a 203% higher risk; (2) adding LAP improved the prediction of CMM, raising the C-statistic of the base model from 0.624 to 0.674, with concordant gains in IDI and NRI; (3) the dose–response relationship was non-linear but consistently positive; (4) the association was broadly consistent across subgroups, with hypertension the only significant effect modifier (P for interaction = 0.020), the relationship being stronger in normotensive participants yet significant in both strata; and (5) WBC count and log-transformed CRP each partially mediated the association, accounting for 5.1% and 8.4% of the total effect, respectively.

The graded association we observed is consistent with previous longitudinal work. Over the 9.0-year follow-up, CMM incidence rose across LAP quartiles (2.91%, 4.81%, 7.25%, and 10.74%). Zhou et al. first reported that persistently elevated LAP trajectories predicted a higher risk of CMM (HR 2.63) [23], and Yan et al., in a large national sample, found the risk of diabetes to be 2.20-fold higher in the persistently high than in the persistently low LAP group [24]. The larger effect estimate in our study (HR 3.03) probably reflects our endpoint: because CMM represents a clustering of several metabolic diseases, its association with LAP may be tighter than that of any single condition.

Our findings also fit the broader evidence that LAP predicts individual cardiometabolic diseases. For metabolic outcomes, LAP has achieved an area under the curve (AUC) above 0.8 for non-alcoholic fatty liver disease [25, 26] and 0.88 in a meta-analysis for metabolic syndrome [27], exceeding BMI, waist circumference, and the triglyceride–glucose index; in a study of 35,000 Chinese adults its AUC for metabolic syndrome reached 0.901 in men and 0.898 in women [28]. For diabetes and its complications, LAP combined with urinary glucose excretion identified newly diagnosed diabetes with a sensitivity of 93.6% [29], was inversely related to the age of onset of early-onset type 2 diabetes [30], and independently predicted diabetic kidney disease [31]; Wei et al. described a non-linear relationship with diabetic retinopathy, with a turning point at a LAP of 63.4 below which the risk rose [32]. For cardiovascular outcomes, Tan et al. linked higher LAP to stroke and myocardial infarction in the Kailuan cohort [33], the ATTICA study supported LAP as a 10-year cardiovascular risk tool [34], and LAP showed modest predictive value for cardiovascular events in patients with metabolic dysfunction-associated steatotic liver disease, comparable to other lipid-based insulin-resistance markers [35]. These associations appear stronger in women, which may be relevant to perimenopausal metabolic risk [34, 36, 37]. One caveat applies when comparing effect sizes across these reports: they are expressed on different scales—Tan et al., for example, reported risk per unit of natural-log LAP, whereas our hazard ratio refers to a one-unit change in log10 LAP (a ten-fold increase)—so the numerical values are not directly comparable.

LAP also improved risk prediction, which has practical value. Adding LAP to the base model (age, sex, education, marital status, residence, smoking, alcohol consumption, eGFR, and hypertension) raised the C-statistic from 0.624 to 0.674, with concordant improvements in IDI and NRI, and the cumulative-incidence curves showed a clear separation of CMM risk across LAP quartiles. The inflection point of approximately 74 in our study was close to thresholds reported for diabetic retinopathy (63.4) [32] and for prediabetes/diabetes (68.1) [38].

The association was consistent across subgroups defined by sex, education, marital status, residence, smoking, and alcohol consumption, indicating broad applicability. The exception was hypertension (P for interaction = 0.020): the association was stronger in normotensive participants (HR 4.39, 95% CI 2.98–6.48) than in hypertensive ones (HR 2.19, 95% CI 1.54–3.09), though it remained significant in both. This pattern has a plausible physiological basis. Hypertension is a major cardiometabolic risk factor and a component of metabolic syndrome; in people who are already hypertensive, endothelial dysfunction and insulin resistance are established, so the additional information LAP provides about metabolic risk may be smaller. In normotensive people, by contrast, LAP can flag excess visceral fat and lipid dysregulation—and the accompanying insulin resistance, low-grade inflammation, and dyslipidaemia—before blood pressure rises, identifying occult risk that precedes overt disease. This supports LAP screening in apparently healthy populations and a shift in emphasis from treatment toward prevention.

A key contribution of this study is the finding that systemic inflammation partly mediates the LAP–CMM association. The indirect effect through WBC count corresponded to a hazard ratio of 1.04 (95% CI 1.01–1.06), accounting for 5.1% of the total effect (P = 0.016), and the model including both LAP and WBC had a C-statistic of 0.677; compared with the basic model, the 9-year IDI was 0.010 [95% CI 0.006–0.018] and the NRI was 0.159 [95% CI 0.114–0.225] (both P < 0.001). This is biologically plausible. LAP reflects visceral adiposity, and visceral adipose tissue is metabolically active: under excess lipid accumulation, adipocyte hypertrophy and local hypoxia trigger an inflammatory cascade in which adipose-tissue macrophages shift toward a pro-inflammatory phenotype and circulating leukocytes are recruited to the vessel wall, where activated neutrophils and monocytes release mediators that impair endothelial function, oxidise low-density lipoprotein, and interfere with insulin signalling, promoting atherosclerosis and insulin resistance. C-reactive protein mediated the association once its extreme right skew was addressed. In the initial analysis, CRP appeared not to mediate, but this was an artefact: a small number of very high baseline values, reflecting acute-phase responses, distorted the linear mediator model. After natural-log transformation—and after excluding participants with CRP > 10 mg/L in a sensitivity analysis—log-CRP showed a mediating effect similar in size to that of WBC. The agreement between a cellular marker (WBC) and a humoral marker (CRP) strengthens the interpretation that chronic low-grade inflammation is a genuine, though partial, pathway linking visceral lipid accumulation to CMM, whereas isolated acute-phase elevations of CRP do not capture this chronic process.

Study limitations

This study has several limitations. First, only the total WBC count was available in CHARLS; differential counts (neutrophils, lymphocytes), which might mediate the association more specifically, could not be examined, and the modest proportion mediated by total WBC should be read in this light. Second, LAP, WBC, and CRP were each measured once, at baseline; single measurements are subject to within-person variation, and the resulting non-differential misclassification would tend to bias the mediated effects toward the null, so the true mediation may be somewhat stronger than estimated. Third, data on diet, physical activity, and use of antihypertensive, glucose-lowering, and lipid-lowering drugs were unavailable, so residual confounding cannot be excluded; however, the E-value indicates that an unmeasured confounder would need to be associated with both LAP and CMM by a hazard ratio of at least 4.09 to explain away the association fully, which is implausibly strong for such factors. Fourth, heart disease was based on self-reported physician diagnosis and treated as a single composite, which may introduce misclassification. Finally, the LAP formula and its sex-specific constants were derived in Western populations; although our sensitivity analyses (China-specific waist thresholds, within-sex standardisation, and sex-stratified models) supported the robustness of the association, the absolute cut-points and generalisability to other age groups and ethnicities warrant caution. Large, multi-ethnic cohorts with repeated measurements, leukocyte subpopulations, and multi-omics data are needed to confirm and extend these findings.

Conclusion

In this large prospective cohort, a higher LAP was independently associated with a non-linear, positive increase in the risk of incident CMM and improved the discrimination of CMM prediction models. Systemic inflammation, indexed by both WBC count and CRP, partially mediated this association, offering insight into the mechanisms linking LAP to CMM. These findings support LAP as an accessible tool for risk assessment in cardiometabolic multimorbidity and may inform prevention efforts in chronic disease.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations and Acronyms

AUC

Area under the curve

BMI

Body mass index

BUN

Blood urea nitrogen

CAPI

Computer-assisted personal interviewing

CHARLS

China Health and Retirement Longitudinal Study

CI

Confidence interval

CMM

Cardiometabolic multimorbidity

Crea

Serum creatinine

CRP

C-reactive protein

CVD

Cardiovascular disease

CysC

Serum cystatin C

DBP

Diastolic blood pressure

DKD

Diabetic kidney disease

DR

Diabetic retinopathy

eGFR

Estimated glomerular filtration rate

HbA1c

Glycated hemoglobin

HDL-C

High-density lipoprotein cholesterol

HGB

Hemoglobin concentration

HR

Hazard ratio

IDI

Integrated discrimination improvement

IRS-1

Insulin receptor substrate-1

LAP

Lipid Accumulation Product

LDL-C

Low-density lipoprotein cholesterol

LgLAP

Base-10 logarithm of Lipid Accumulation Product

MACCE

Major cardiovascular and cerebrovascular events

MASLD

Metabolic dysfunction-associated steatotic liver disease

MetS

Metabolic syndrome

MPO

Myeloperoxidase

NAFLD

Non-alcoholic fatty liver disease

NRI

Net reclassification improvement

PI3K/Akt

Phosphoinositide 3-kinase/protein kinase B

PLT

Platelet count

PPS

Probability-proportional-to-size

RCS

Restricted cubic spline

ROC

Receiver operating characteristic

SBP

Systolic blood pressure

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

T2DM

Type 2 diabetes mellitus

TC

Total cholesterol

TG

Triglycerides

UA

Uric acid

UGE

Urinary glucose excretion

VAT

Visceral adipose tissue

WBC

White blood cell count

Author contributions

AM: performed the formal analysis, curated the data, prepared the visualizations, and drafted the manuscript. YL: contributed to study conceptualization and methodology, supervised the study, managed the project, and critically revised the manuscript. DL: contributed to validation and investigation, provided resources, and critically revised the manuscript. All authors read and approved the final manuscript.

Funding

This research did not receive funding.

Data availability

The datasets used and/or analyzed during the current study are publicly available from the China Health and Retirement Longitudinal Study (CHARLS) [39].

Declarations

Human ethics and consent to participate

The present study was a secondary analysis of publicly available, de-identified data from the China Health and Retirement Longitudinal Study (CHARLS); no additional ethics approval was obtained for this secondary analysis. The original CHARLS study was approved by the Biomedical Ethics Committee of Peking University (approval number: IRB00001052-11015), and all participants provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.