Discordance between remnant cholesterol and low-density lipoprotein cholesterol predicts incident hypertension: a 9-year prospective cohort study in Chinese adults

Abstract

Background

Hypertension (HTN), a primary driver of cardiovascular disease (CVD) progression, shares etiological links with dyslipidemia. While low-density lipoprotein cholesterol (LDL-C) remains a cornerstone biomarker for CVD risk, emerging evidence implicates remnant cholesterol (RC)—a triglyceride-rich lipoprotein component—in HTN pathogenesis. This study aimed to examine the association between RC (independent of LDL-C) and incident HTN among Chinese adults aged 45 years and older.

Methods

We analyzed 4,508 normotensive participants from the China Health and Retirement Longitudinal Study (CHARLS). Incident HTN served as the primary endpoint. RC was derived by subtracting directly measured LDL-C from non-high-density lipoprotein cholesterol (non-HDL-C). Adjusted Cox proportional hazards models evaluated the relationship between natural log-transformed RC (ln RC) levels and HTN risk. Discordance analyses categorized participants into RC-LDL-C concordant/discordant groups using percentile differences (> 15 units), median splits, and guideline-based LDL-C thresholds. Subgroup analyses validated the robustness of the results.

Results

During a median follow-up of 9 years, 39.88% (n = 1,798) of participants developed incident HTN. RC-LDL-C discordance was observed in 58.23% of the cohort, comprising 37.18% with discordantly low RC and 21.05% with discordantly high RC. After multivariable adjustment, each 1 standard deviation increase in baseline ln RC was associated with a 7% elevated HTN risk (hazard ratio [HR] 1.07, 95% confidence interval [CI] 1.02–1.12). Restricted cubic spline (RCS) analysis revealed a linear dose-response relationship between continuous ln RC levels and HTN risk (P for trend = 0.013; P for nonlinearity = 0.233). Participants in the discordantly high RC group exhibited a 17% higher HTN risk compared with the discordantly low group (HR 1.17, 95% CI 1.04–1.32). Stratification by three LDL-C clinical thresholds and the median value showed that low LDL-C/high RC individuals had higher HTN risk than the low LDL-C/low RC reference group in the fully adjusted model, with statistically significant associations at 113 and 130 mg/dL. Subgroup analyses supported these findings.

Conclusions

In normotensive populations, higher RC concentrations were independently linked to incident HTN, irrespective of LDL-C levels. The pathophysiological pathways underlying this RC-HTN association—distinct from LDL-C-mediated mechanisms—and the clinical utility of RC-targeted interventions in primary prevention strategies warrant further investigation.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12872-026-06008-z.

Introduction

Hypertension (HTN) remains a critical global public health challenge, affecting approximately 1.3 billion individuals worldwide and contributing to nearly 10 million deaths annually due to cardiovascular complications [1–3]. In China, the burden is particularly severe, with recent epidemiological studies reporting a HTN prevalence of 27.5% among adults, translating to over 300 million affected individuals [4]. This escalating prevalence is exacerbated by rapid urbanization, aging populations, and lifestyle shifts, positioning HTN as a leading modifiable risk factor for stroke, chronic kidney disease, and cardiovascular mortality in China [5, 6].

Traditional risk factors for HTN, including age, obesity, sodium intake, and genetic predisposition, have been extensively characterized [7]. However, emerging evidence underscores the pivotal role of dyslipidemia in HTN pathogenesis. Beyond the well-established association of low-density lipoprotein cholesterol (LDL-C) with atherosclerosis, lipid abnormalities—particularly elevated triglycerides (TG) and altered lipoprotein particle composition—promote endothelial dysfunction and arterial stiffening, thereby contributing to elevated blood pressure. Despite advances in lipid-lowering therapies, a substantial proportion of HTN risk remains unexplained by conventional lipid parameters, highlighting the need to explore novel biomarkers that may contribute to HTN pathogenesis alongside other well-established risk factors.

Remnant cholesterol (RC), the cholesterol content of triglyceride-rich lipoprotein remnants (TRLs) — primarily consisting of very-low-density lipoprotein cholesterol (VLDL-C) and intermediate-density lipoprotein cholesterol (IDL-C) in the fasting state and chylomicron remnant cholesterol (CM-C) in the non-fasting state — has recently emerged as a key mediator of cardiovascular disease (CVD). Mechanistically, RC directly infiltrates the arterial wall, triggering oxidative stress, inflammation, and foam cell formation [8–11]. Importantly, discordance between RC and LDL-C—where elevated RC coexists with optimal LDL-C levels—has been implicated in residual cardiovascular risk, independent of traditional lipid targets [12]. For instance, cohort studies demonstrate that RC-LDL-C discordance predicts incident stroke and atherosclerotic CVD (ASCVD) even among individuals achieving LDL-C goals [13, 14]. This phenomenon suggests that RC may drive vascular injury through pathways distinct from LDL-C-dependent mechanisms. Unlike LDL-C, which requires oxidative modification to become atherogenic, RC particles can infiltrate the arterial intima directly due to their smaller size and cholesterol-enriched core, making them potentially more relevant for early vascular damage preceding HTN. Therefore, investigating RC-LDL-C discordance may uncover residual risk not captured by conventional lipid metrics.

Despite these advancements, the contribution of RC-LDL-C discordance to HTN pathogenesis remains understudied, particularly in populations with distinct lipid profiles such as Chinese adults. To address this gap, we analyzed data from the China Health and Retirement Longitudinal Study (CHARLS)—a nationally representative cohort with serial lipid and health measurements—to evaluate the independent associations of RC and its discordance with LDL-C in relation to incident HTN. We hypothesize that RC-LDL-C discordance contributes to residual risk stratification for HTN beyond absolute RC levels, potentially reflecting unaddressed pathways in lipid-mediated vascular dysfunction. This inquiry holds significant implications for refining risk prediction models and tailoring lipid management strategies in HTN prevention.

Methods

Study design and population

Participants were recruited from the CHARLS study, a nationwide prospective cohort study that focuses on residents aged 45 years and older in both urban and rural areas of China. The CHARLS study has been granted ethical approval by the Ethics Review Committee of Peking University and all study participants provided informed consent. This study involves a biennial survey, with data collected in 2011, 2013, 2015, 2018, and 2020. The current research integrates data from all five survey iterations. Among the 17,708 participants, 4,508 were selected based on eligibility criteria. Exclusion criteria encompassed: participants aged below 45 years at baseline (n = 777), pre-existing HTN at baseline (n = 6,741), lacked follow-up data on incident HTN (n = 3,494), the absence of pertinent data necessary for RC calculation (n = 1,913), use of lipid-lowering medication (n = 175), and physiologically impossible RC values less than or equal to zero (n = 100), which arose from superimposed measurement errors of TC/HDL-C/direct LDL-C, triglyceride-induced assay interference, and amplified calculation deviations at the lower detection limit [15, 16]. Selection criteria and procedures are detailed in S1. Fig. Baseline characteristics of participants with and without incident HTN during follow-up are compared in Supplemental Table 2.

Data collection and definition

Trained interviewers collected demographic, geographic, medical, and socioeconomic data via standardized questionnaires following established protocols. Demographic variables encompassed age, gender, residence (rural, urban), marital status (married, other), and health-related behaviors including smoking (never, occasional, frequent) and drinking (never, occasional, frequent). Participants rested seated for 5 min before assessment. A trained interviewer recorded three blood pressure measurements at 45-second intervals using a digital sphygmomanometer. Body mass index (BMI) is calculated as weight in kilograms divided by height in meters squared (kg/m²). Fasting blood glucose (FBG) and serum lipids were measured by enzymatic colorimetric assay, while glycated hemoglobin (HbA1c) was assessed via boronate affinity high-performance liquid chromatography. The levels of TG, total cholesterol (TC), LDL-C, high-density lipoprotein cholesterol (HDL-C), and glucose were determined using the enzymatic colorimetric technique with the Olympus Automatic Biochemical Analyzer (model Hitachi 747). The procedures for sample collection, transportation, storage, and analysis have been detailed in prior reports [15] and are available on the CHARLS website (http://charls.pku.edu.cn/).

HTN diagnosis was established according to the following criteria: (1) a physician-confirmed self-reported diagnosis, (2) current antihypertensive medication use, or (3) systolic/diastolic blood pressure (SBP/DBP) ≥ 140/90 mmHg. The initial diagnosis date was recorded as the HTN onset time. For diabetes mellitus (DM), diagnostic thresholds included FBG ≥ 126 mg/dL, HbA1c ≥ 6.5%, clinician-confirmed diagnosis, or the use of glucose-lowering drugs. Dyslipidemia was defined by either a self-reported physician diagnosis, active lipid-lowering therapy, or meeting any of the following laboratory thresholds: TC ≥ 240 mg/dL, TG ≥ 150 mg/dL, HDL-C < 40 mg/dL, or LDL-C ≥ 160 mg/dL. Heart diseases and kidney diseases were assessed using a questionnaire.

RC calculation

Remnant cholesterol was calculated using the explicit formula [14]:

This formula is mathematically and biologically equivalent to RC being the value obtained by subtracting directly measured LDL-C from non-HDL-C. Non-HDL-C was defined as the value obtained by subtracting HDL-C from TC in accordance with international standards and guidelines. As non-HDL-C encompasses cholesterol from all atherogenic lipoproteins, subtracting directly measured LDL-C yields triglyceride-rich lipoprotein remnant cholesterol, consistent with standard cardiovascular epidemiological definitions. LDL-C was measured using a standardized direct homogeneous assay accredited by ISO 15,189 and the College of American Pathologists (CAP), with consistent internal quality control across all samples [15].

Discordance definition

To evaluate the independence of RC-associated HTN risk from LDL-C levels, discordance assessments were conducted through multiple stratification approaches. First, discordance was defined as an absolute percentile difference exceeding 15 units between RC and LDL-C (RC percentile minus LDL-C percentile). Participants were categorized into three groups: (i) discordantly low RC (RC percentile < LDL-C percentile by > 15 units), (ii) concordant RC-LDL-C (difference within ± 15 units), and (iii) discordantly high RC (RC percentile > LDL-C percentile by > 15 units). This percentile-based approach ensures comparability across populations with different absolute lipid values. Sensitivity analyses were conducted using a 20-unit percentile threshold to verify the robustness of the findings.

Additionally, to enhance clinical interpretability, we applied categorical discordance definitions based on established clinical thresholds. Moreover, the median LDL-C (113 mg/dL) and median RC (18 mg/dL) values from this study served as benchmarks for inconsistency analysis. Additionally, two clinically significant LDL-C thresholds, 130 mg/dL and 100 mg/dL, were evaluated in accordance with international guidelines [17]. By aligning the equivalent percentiles within the cohort to these LDL-C criteria, the corresponding RC thresholds of 27 mg/dL and 13 mg/dL were established, thereby extending beyond conventional median-based classification.

Statistical analysis

Continuous variables were summarized as mean ± standard deviation or median (interquartile range), while categorical variables as frequencies (percentages). Group comparisons employed one-way ANOVA for normally distributed data, Kruskal-Wallis test for skewed distributions, and chi-square tests for categorical variables. Participants were stratified into three RC-LDL-C concordance categories: discordant low RC, concordant, and discordant high RC. The occurrence of HTN was evaluated using Kaplan-Meier curves and log-rank testing. The independent association between continuous ln RC levels (natural log-transformed for non-normality) and incident HTN was assessed using multivariable Cox regression, with results visualized by restricted cubic splines (RCS). Three progressively refined models were employed in the analysis: Model 1 served as an unadjusted baseline model; Model 2 included adjustments for demographic variables such as age, gender, rural residence, smoking status, and drinking status; and Model 3 further expanded the adjustments to encompass clinical parameters, including BMI, baseline SBP, baseline DBP, TC, FBG, blood urea nitrogen (BUN), creatinine, C-reactive protein (CRP), uric acid (UA), HbA1c, as well as comorbidities including heart disease, kidney disease, and diabetes. Concordance patterns were further examined through percentile differentials, clinical thresholds, and median-based classifications, with continuous discordance defined as RC%-LDL-C% percentile deviations. Subgroup Analysis across demographic, lifestyle, and metabolic baseline characteristics verified RC’s consistent HTN risk profile. To evaluate the incremental predictive value of RC-LDL-C percentile discordance (15% percentile difference-based) beyond traditional cardiovascular risk factors and lipid parameters, we constructed a baseline model including age, gender, rural residence, smoking status, drinking status, BMI, baseline SBP, baseline DBP, FBG, BUN, creatinine, CRP, UA, HbA1c, heart disease, kidney disease, diabetes, TC, HDL-C, LDL-C, and TG. We then built an extended model by adding RC-LDL-C percentile discordance to the baseline model. Harrell’s C-statistic, continuous net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were calculated to compare the predictive performance between models. We compared baseline characteristics between retained and lost-to-follow-up participants to assess potential attrition bias (Supplemental Table 2). To further address informative censoring, we performed an inverse probability of censoring weighting (IPCW) sensitivity analysis. Statistical analysis was conducted using R software, version 4.1.2. Statistical significance was determined using a two-tailed test, with P-values less than 0.05 indicating statistically significant results.

Results

Characteristics of study population

Baseline clinical and demographic characteristics of participants with or without HTN during follow-up are presented in Supplemental Table 1. The HTN group displayed a significantly older age profile, with a higher proportion of individuals aged 60 years or older. While residential distribution (rural vs. non-rural) was comparable between groups, HTN was associated with significantly higher rates of comorbid DM and heart disease. Metabolic assessment revealed a consistently adverse profile in the HTN cohort. This included a higher prevalence of overweight/obesity, markedly elevated SBP/DBP, increased FBG levels, and dysregulated lipid metabolism characterized by higher TG and RC accumulation. This unfavorable metabolic milieu coincided with significantly depressed levels of HDL-C, alongside elevated non-HDL-C and UA levels.

Using the primary 15 percentile-unit discordance threshold, the rate of discordance between RC and LDL-C was observed to be 58.23%, with 37.18% of individuals exhibiting discordantly low RC and 21.05% displaying discordantly high RC. Table 1 presents the baseline characteristics of the study population categorized by concordant/discordant categories between LDL-C and RC. The discordantly high RC group had a younger age profile, higher male predominance, and more smokers/drinkers. They exhibited a pronounced metabolic disturbance: higher overweight/obesity prevalence, elevated FBG, and disrupted lipid metabolism featuring markedly increased TG and RC with depressed HDL-C. Elevated UA levels further characterized this adverse metabolic state, alongside a higher DM rate. Blood pressure, residence, HTN, heart, and kidney disease prevalence showed no significant group differences.

Table 1.

Characteristics of the study population in concordant and discordant groups: RC versus LDL-C

Characteristic	Overall N = 4508	Discordantly Low RC N = 1676	Concordant N = 1260	Discordantly High RC N = 1572	P-value
Age group, n (%)					< 0.001
< 60	1520 (33.72%)	606 (36.16%)	442 (35.08%)	472 (30.03%)
≥ 60	2988 (66.28%)	1070 (63.84%)	818 (64.92%)	1100 (69.97%)
Gender, n (%)					< 0.001
Female	2424 (53.77%)	968 (57.76%)	666 (52.86%)	790 (50.25%)
Male	2084 (46.23%)	708 (42.24%)	594 (47.14%)	782 (49.75%)
Education, n (%)					0.194
Primary school or lower	1465 (32.50%)	530 (31.62%)	397 (31.51%)	538 (34.22%)
Middle school or higher	3043 (67.50%)	1146 (68.38%)	863 (68.49%)	1034 (65.78%)
BMI, kg/m2					< 0.001
< 25	3296 (73.11%)	1280 (76.37%)	932 (73.97%)	1084 (68.96%)
≥ 25	1212 (26.89%)	396 (23.63%)	328 (26.03%)	488 (31.04%)
Rural residence, n (%)	3062 (67.92%)	1150 (68.62%)	861 (68.33%)	1051 (66.86%)	0.526
Hypertension, n (%)	1798 (39.88%)	644 (38.42%)	475 (37.70%)	679 (43.19%)	0.004
Diabetes, n (%)	166 (3.68%)	61 (3.64%)	30 (2.38%)	75 (4.77%)	0.004
Heart diseases, n (%)	335 (7.43%)	119 (7.10%)	96 (7.62%)	120 (7.63%)	0.809
Kidney diseases, n (%)	233 (5.17%)	88 (5.25%)	64 (5.08%)	81 (5.15%)	0.978
Drinking status, n (%)	1551 (34.41%)	540 (32.22%)	421 (33.41%)	590 (37.53%)	0.004
Smoking status, n (%)	1382 (30.66%)	447 (26.67%)	397 (31.51%)	538 (34.22%)	< 0.001
SBP, mmHg	117.00 (109.00-126.50)	117.00 (108.50–126.00)	117.50 (109.00-126.50)	117.00 (108.50–127.00)	0.495
DBP, mmHg	70.00 (64.00-76.50)	70.00 (63.50–76.50)	70.00 (64.00-76.50)	70.00 (64.00-76.50)	0.79
BUN, mg/dL	15.13 (12.52–18.18)	15.45 (12.82–18.54)	15.18 (12.55–18.18)	14.71 (12.32–17.62)	< 0.001
FBG, mg/dL	100.98 (93.60-110.34)	100.26 (93.60–108.00)	100.44 (93.42-109.98)	102.60 (93.96–114.30)	< 0.001
Creatinine, mg/dL	0.75 (0.64–0.87)	0.73 (0.63–0.85)	0.75 (0.64–0.87)	0.76 (0.64–0.87)	0.013
TC, mg/dL	188.27 (165.85-213.02)	204.51 (185.18-226.55)	184.02 (161.99-212.34)	172.04 (153.48-192.53)	< 0.001
TG, mg/dL	99.12 (72.57-144.26)	81.42 (63.72-104.43)	97.35 (71.68–135.40)	138.95 (96.46-202.66)	< 0.001
HDL-C, mg/dL	50.64 (40.98–60.70)	55.28 (47.94–64.66)	50.64 (42.53–59.92)	43.69 (35.57–54.51)	< 0.001
LDL-C, mg/dL	113.27 (92.40-135.70)	135.31 (118.69-154.25)	113.85 (94.72-133.38)	90.85 (75.39-105.16)	< 0.001
CRP, mg/dL	0.88 (0.50–1.81)	0.81 (0.48–1.54)	0.90 (0.50–1.92)	0.95 (0.52–2.09)	< 0.001
HbA1c, %	5.10 (4.90–5.40)	5.10 (4.90–5.40)	5.10 (4.90–5.40)	5.10 (4.80–5.40)	0.374
UA, mg/dL	4.15 (3.47–4.97)	4.05 (3.41–4.87)	4.15 (3.48–4.92)	4.25 (3.57–5.15)	< 0.001
Non-HDL-C, mg/dL	136.08 (113.27-160.54)	147.29 (127.96-169.33)	132.99 (106.60-161.60)	124.87 (104.38-147.78)	< 0.001
RC, mg/dL	18.36 (10.82–29.77)	10.82 (6.96–15.85)	18.56 (11.98–28.22)	30.54 (22.04–44.85)	< 0.001

Discordant groups were established based on a difference exceeding 15 percentile units. Continuous variables are presented as median values along with the interquartile range (25th–75th percentile). Data are means ± SD, median (interquartile range), or n (%)

Abbreviations: BMI body mass index, BUN blood urea nitrogen, SBP systolic blood pressure, DBP diastolic blood pressure, FBG fasting blood glucose, HDL-C high density lipoprotein cholesterol, LDL-C low density lipoprotein cholesterol, TC total cholesterol, TG triglycerides, RC remnant cholesterol, CRP C-reactive protein, HbA1c glycosylated hemoglobin, UA uric acid

Association of RC and HTN

Over a nine-year follow-up period, 1,798 (39.88%) of these participants developed HTN for the first time. The Cox regression analysis presented in Table 2 substantiates a significant association between baseline log RC and the incidence of new-onset HTN. After adjusting for potential confounders, Model 3 revealed each 1 standard deviation (SD) increase in baseline ln RC was associated with a 7% increase in the risk of HTN (hazard ratio [HR] 1.07, 95% confidence interval [CI] 1.02–1.12). The RCS analysis presented in Fig. 1 demonstrates a significant linear relationship between RC as a continuous variable and the risk of HTN, after adjusting for multiple covariates (P for trend = 0.013; P for nonlinearity = 0.233).

Table 2.

Cox models analyzed the relationship between ln RC, ln LDL-C and the risk of hypertension

	Model 1			Model 2			Model 3
	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
ln RC, per SD	1.14	1.09, 1.19	< 0.001	1.16	1.11, 1.20	< 0.001	1.07	1.02, 1.12	0.009
ln LDL-C, per SD	1.04	0.99, 1.09	0.139	1.03	0.98, 1.08	0.221	0.97	0.91, 1.04	0.422

HR hazard ratios, CI confidence interval, Ref reference, RC remnant cholesterol, SBP systolic blood pressure, DBP diastolic blood pressure, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, SD standard deviation, BMI body mass index

Model 1: Adjusted for none

Model 2: Adjusted for age, gender, rural residence, education, smoking status and drinking status

Model 3: Adjusted for age, gender, rural residence, smoking status, drinking status, FBG, BUN, Creatinine, CRP, UA, HbA1c, heart disease, kidney disease, diabetes, BMI, SBP, DBP and TC

Fig. 1.

Analysis of the relationship between log RC values and hypertension risk Using RCS methodology. Adjusted for age, gender, rural residence, smoking status, drinking status, FBG, BUN, Creatinine, CRP, UA, HbA1c, heart disease, kidney disease, diabetes, BMI, SBP, DBP and TC. The linear trend was significant (P = 0.013) with no evidence of nonlinearity (P = 0.233). RCS: restricted cubic spline; RC: Remnant cholesterol

Discordance analysis

A linear dose-response relationship existed between RC/LDL-C discordance and incident HTN (P for trend = 0.012, P for nonlinearity = 0.118; Fig. 2). Notably, the discordantly high RC group showed a 17% increased HTN risk versus the discordantly low RC group (HR = 1.17, 95% CI:1.04–1.32; Model 3, Table 3). Kaplan-Meier curves confirmed that the discordantly high RC group exhibited the highest risk (Fig. 3). The log-rank test was statistically significant (P = 0.0024).

Fig. 2.

Dose–response relationship of the discordance between RC and LDL-C with HTN onset. Adjusted for age, gender, rural residence, smoking status, drinking status, FBG, BUN, Creatinine, CRP, UA, HbA1c, heart disease, kidney disease, diabetes, BMI, SBP, DBP and TC. The linear trend was significant (P < 0.05) with no evidence of nonlinearity (P = 0.118). RCS: restricted cubic spline; RC: Remnant cholesterol

Table 3.

Cox models analyzed the relationship between standardized differences in RC and LDL-C percentiles and the risk of hypertension

	Model 1			Model 2			Model 3
	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
Discordance, continuous	1.0015	1.0004, 1.0026	0.008	1.0019	1.0008, 1.0030	< 0.001	1.0016	1.0003, 1.0028	0.013
Discordance groups by 15%
Discordantly low RC	Ref			Ref			Ref
Concordant group	0.98	0.87, 1.10	0.705	0.98	0.87, 1.10	0.699	0.96	0.85, 1.08	0.486
Discordantly high RC	1.17	1.05, 1.30	0.005	1.21	1.09, 1.35	< 0.001	1.17	1.04, 1.32	0.008
P for trend			0.005			< 0.001			0.009

HR hazard ratios, CI confidence interval, Ref reference, RC remnant cholesterol, SBP systolic blood pressure, DBP diastolic blood pressure, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, BMI body mass index

Model 1: Adjusted for none

Model 2: Adjusted for age, gender, rural residence, education, smoking status and drinking status

Model 3: Adjusted for age, gender, rural residence, smoking status, drinking status, FBG, BUN, Creatinine, CRP, UA, HbA1c, heart disease, kidney disease, diabetes, BMI, SBP, DBP and TC

Fig. 3.

Kaplan-Meier analysis of hypertension events across three RC groups

Participants were further classified into three groups for analysis based on LDL-C clinical cutpoints and medians. The baseline characteristics of the participants stratified into four LDL-C/RC concordance/discordance groups are shown in Table 4. Among these groups, the prevalence of HTN was consistently elevated in the discordantly low LDL-C/high RC group compared to the concordantly low LDL-C/low RC reference group across all three LDL-C thresholds. Multi-model Cox regression analyses indicated that participants with low LDL-C and high RC exhibited a significantly higher risk of developing HTN in the fully adjusted model (Model 3), with hazard ratios (HRs) of 1.18 (95% CI: 0.98–1.43, P = 0.078, borderline significant) for the 100 mg/dL LDL-C threshold, 1.21 (95% CI: 1.05–1.39, P = 0.007) for the 113 mg/dL median-based threshold, and 1.19 (95% CI: 1.05–1.34, P = 0.006) for the 130 mg/dL threshold. In contrast, no significant difference in HTN incidence was observed between the high LDL-C/low RC group and the low LDL-C/low RC group across any of the thresholds (all P > 0.05), underscoring the predominant role of RC over LDL-C in driving HTN risk. Concurrently, as the clinical cut-off point for LDL-C decreases, the proportion of individuals with high RC/low LDL-C discordance progressively increases (Fig. 4), highlighting the growing relevance of RC evaluation in populations with lower LDL-C levels.

Table 4.

Hazard ratios (95% CI) for hypertension in LDL-C vs. RC concordance/discordance groups, stratified by LDL-C thresholds (100, 113, 130 mg/dL) and RC percentiles

	Model 1			Model 2			Model 3
	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
Cutpoints: LDL-C 100 mg/dL; RC 13 mg/dL
LDL-C < 100 mg/dL & RC < 13 mg/dL	Ref			Ref			Ref
LDL-C < 100 mg/dL & RC ≥ 13 mg/dL	1.29	1.07, 1.55	0.007	1.32	1.10, 1.58	0.003	1.18	0.98, 1.43	0.078
LDL-C ≥ 100 mg/dL & RC < 13 mg/dL	1.09	0.91, 1.32	0.347	1.09	0.90, 1.32	0.358	1.06	0.86, 1.30	0.579
LDL-C ≥ 100 mg/dL & RC ≥ 13 mg/dL	1.37	1.15, 1.63	< 0.001	1.40	1.18, 1.66	< 0.001	1.17	0.96, 1.42	0.131
Cutpoints: LDL-C 113 mg/dL; RC 18 mg/dL
LDL-C < 113 mg/dL & RC < 18 mg/dL	Ref			Ref			Ref
LDL-C < 113 mg/dL & RC ≥ 18 mg/dL	1.32	1.18, 1.49	< 0.001	1.39	1.23, 1.56	< 0.001	1.21	1.05, 1.39	0.007
LDL-C ≥ 113 mg/dL & RC < 18 mg/dL	1.11	0.99, 1.26	0.085	1.11	0.98, 1.26	0.087	1.07	0.91, 1.26	0.403
LDL-C ≥ 113 mg/dL & RC ≥ 18 mg/dL	1.36	1.15, 1.59	< 0.001	1.37	1.17, 1.62	< 0.001	1.09	0.91, 1.30	0.338
Cutpoints: LDL-C 130 mg/dL; RC 27 mg/dL
LDL-C < 130 mg/dL & RC < 27 mg/dL	Ref			Ref			Ref
LDL-C < 130 mg/dL & RC ≥ 27 mg/dL	1.30	1.13, 1.48	< 0.001	1.37	1.20, 1.57	< 0.001	1.19	1.05, 1.34	0.006
LDL-C ≥ 130 mg/dL & RC < 27 mg/dL	1.13	0.98, 1.30	0.090	1.13	0.98, 1.30	0.081	1.04	0.89, 1.21	0.660
LDL-C ≥ 130 mg/dL & RC ≥ 27 mg/dL	1.34	1.17, 1.53	< 0.001	1.38	1.20, 1.58	< 0.001	1.05	0.85, 1.30	0.619

HR hazard ratios, CI confidence interval, Ref reference, RC remnant cholesterol, SBP systolic blood pressure, DBP diastolic blood pressure, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, BMI body mass index

Model 1: Adjusted for none

Model 2: Adjusted for age, gender, rural residence, education, smoking status and drinking status

Model 3: Adjusted for age, gender, rural residence, smoking status, drinking status, FBG, BUN, Creatinine, CRP, UA, HbA1c, heart disease, kidney disease, diabetes, BMI, SBP, DBP and TC

Fig. 4.

Discordance proportions in LDL-C below clinical cutpoints. As the clinical cut-off point for LDL-C decreases, the proportion of individuals with high RC/low LDL-C discordance increases

Subgroup analysis

This study assessed the robustness of the association of RC with HTN using subgroup analysis (S2. Fig). Subgroup analyses revealed that the association between RC and incident HTN was consistent across all subgroups, with no significant interactions observed for any baseline characteristic (all P for interaction > 0.05).

Sensitivity analysis

IPCW-weighted Cox regression analysis yielded nearly identical results to the primary analysis (Supplemental Table 3). Each 1 SD increase in ln RC was associated with a 7% elevated risk of incident HTN (HR 1.07, 95% CI 1.02–1.13, P = 0.013), and the discordantly high RC group had an HR of 1.17 (95% CI 1.04–1.32, P = 0.010). These consistent results confirm that our core conclusions are not materially affected by informative censoring due to loss to follow-up.

Sensitivity analysis using a 20-unit percentile threshold to define RC-LDL-C discordance confirmed the robustness of our findings (Supplemental Table 4).

Incremental predictive value analysis

Adding RC-LDL-C discordance to the full clinical model (including all traditional cardiovascular risk factors and lipid parameters) resulted in a statistically significant but small improvement in HTN risk prediction (Supplemental Table 5). The Harrell’s C-statistic increased minimally from 0.676 to 0.677 (P < 0.001), with a continuous NRI of 0.0868 (P = 0.004) and an IDI of 0.0017 (P = 0.009), confirming modest incremental predictive utility of RC-LDL-C discordance above conventional markers.

Discussion

In this large-scale prospective cohort of middle-aged and older Chinese adults, we identified three principal findings. First, elevated baseline RC levels were independently associated with an increased risk of incident HTN over nine years of follow-up, with a linear dose-response relationship. Second, discordance analysis revealed that individuals with high RC despite low LDL-C (RC-LDL-C discordance) exhibited a significantly higher HTN risk compared to those with concordantly low levels. Third, this discordant risk pattern persisted across multiple LDL-C thresholds, including guideline-recommended cutoffs, suggesting that RC captures atherogenic risk not reflected by conventional LDL-C measurement. These findings underscore the potential utility of RC as a complementary biomarker for HTN risk stratification, particularly in individuals with controlled LDL-C levels.

HTN development involves synergistic interactions between non-modifiable (e.g., aging) and modifiable risk factors (e.g., dyslipidemia, sedentary behavior) [18]. While conventional lipid markers like LDL-C and HDL-C have been mechanistically linked to blood pressure dysregulation, their epidemiological correlations remain inconsistent across studies—a limitation attributed to heterogeneous population characteristics and methodological variability [19, 20]. In contrast, accumulating evidence positions RC, a TRLs derivative, as a potent mediator of vascular injury through inflammation-independent pathways [21].

CVD, the leading global mortality cause, exhibit persistent residual risk despite LDL-C-targeted therapies—a phenomenon partly driven by RC-mediated atherogenesis [22]. RC contributes to this residual risk via three distinct mechanisms: (1) Unlike LDL-C, RC directly infiltrates arterial intima without oxidative modification, accumulating due to impaired efflux [23]; (2) RC particles carry 4-fold more cholesterol per unit than LDL, accounting for ≥ 30% of total arterial cholesterol deposition [24]; (3) TRLs lipolysis releases cytotoxic byproducts (e.g., oxidized lipids) that amplify vascular inflammation via cytokine and adhesion molecule activation [25].

Beyond CVD, RC dysregulation correlates with metabolic disorders. Cohort studies identify RC as an independent predictor of new-onset DM, potentially through cholesterol-induced pancreatic β-cell toxicity and insulin resistance [26, 27]. Notably, insulin resistance and endothelial dysfunction—central to HTN pathogenesis—are mechanistically intertwined with RC-mediated inflammation and lipid toxicity. Thus, elucidating RC’s role in HTN may refine risk stratification and therapeutic targeting in cardiometabolic diseases.

There are cross-sectional studies based on U.S. populations that have found an association between elevated RC and risk of HTN. A decade-long cohort study of 681 normotensive individuals by Kasahara et al. revealed significantly elevated baseline RC levels in those developing HTN [28]. Similarly, Chen et al. identified RC as the strongest lipid predictor of HTN incidence, surpassing conventional parameters—a finding corroborated by our data [29]. UK Biobank analyses by Guo et al. demonstrated a nonlinear dose-response relationship between RC and HTN risk, particularly pronounced in younger, non-diabetic, and non-obese subgroups [30]. Notably, this current analytical framework was confined to a geriatric-specific cohort (≥ 45 years), necessitating additional longitudinal investigations to validate the generalizability of these pathophysiological mechanisms across younger population strata.

Emerging evidence suggests that lowering RC may mitigate HTN risk. Mechanistically, RC, a component of TRLs, promotes endothelial dysfunction and vascular inflammation, pathways central to HTN pathogenesis. Studies have demonstrated that statins are effective in reducing blood pressure and also decrease LDL-C and RC levels to some extent [31]. Furthermore, the REDUCE-IT trial demonstrated that icosapent ethyl, which lowers RC by 14%, significantly reduced cardiovascular events and incident HTN [32].

Emerging genetic evidence supports the causal role of RC in CVD and HTN. Mendelian randomization studies reveal that genetically elevated RC levels increase CVD risk [33]. Additionally, elevated RC and TG are causally linked to cardiometabolic multimorbidity, including HTN [34]. RC is also identified as a modifier of ASCVD progression [13]. These findings underscore RC’s significance in initiating and progressing CVD, suggesting that targeting RC may benefit HTN management.

Recent studies underscore the pathophysiological significance of divergent RC/LDL-C profiles in cardiometabolic disorders such as ASCVD [13], stroke [14], DM [30], and DM-associated ASCVD [35]. Given the frequent comorbidity of these conditions with HTN, discordant RC-LDL-C status may harbor underexplored prognostic implications for incident HTN. Despite these insights, research on the discordance between RC-LDL-C and HTN incidence remains scarce. To date, only two studies have addressed this association: a US-based investigation explored this relationship using fixed clinical thresholds and median values to categorize discordance, while another UK population-based study examined the correlation without subgroup stratification based on clinical thresholds [36, 37]. The present study advances this paradigm by incorporating a percentile-based stratification approach, thereby enabling a comprehensive evaluation of RC as both a continuous variable and a categorical risk marker. This dual analytical methodology not only confirms the independent predictive capacity of RC but also addresses limitations inherent in previous categorical classification approaches. However, we didn’t include the lowest clinical cutpoint of LDL-C at 70 mg/dL, mainly because too few participants (only dozens) had levels below this value. Future studies need a larger sample size.

Globally, the measurement and characterization of RC face two primary challenges: (1) the dynamic metabolic nature of RC particles, characterized by rapid catabolism and variability in size, quantity, and composition, and (2) the difficulty in distinguishing RC from its precursor TRLs despite differences in particle size and TG content [8]. Clinically, RC is universally defined as the cholesterol content of TRLs (VLDL-C, IDL-C, and CM-C), and is typically calculated as RC = TC - HDL-C - LDL-C [13], where LDL-C is often estimated via the Friedewald equation. However, this approach becomes unreliable at elevated TG levels due to its fixed TG: VLDL-C ratio (1:5 mg/dL), which oversimplifies RC estimation. To address these limitations, Martin et al. proposed an LDL-C estimation method incorporating adjustable TG: VLDL-C ratios based on TC and non-HDL-C concentrations, improving accuracy even at higher TG levels. Alternatively, direct LDL-C measurement enables precise RC determination under hypertriglyceridemia [16]. In this study, RC was calculated using direct LDL-C detection. CHARLS employed LDL-C homogeneous assays standardized via ISO 15,189 and CAP accreditation, delivering reliable results across critical thresholds—especially relevant for East Asian cohorts showing higher baseline TG and lower LDL-C [15, 38]. Despite cost benefits of calculated LDL-C in resource-constrained areas [39], China emphasizes assay precision for complex cases—underestimation may jeopardize ASCVD stratification and drug titration accuracy [40, 41]. Additionally, resultant negative values from calculated RC were excluded in this study to ensure data integrity.

The association between RC and HTN is complex, involving multiple physiological and biochemical pathways. RC may induce endothelial dysfunction, a precursor to HTN, by promoting inflammation and oxidative stress within the vascular endothelium, leading to impaired vasodilation and increased vascular resistance. Additionally, RC demonstrates potent agonistic activity toward epidermal growth factor receptors, inducing excessive vascular smooth muscle cell proliferation and pathological vascular restructuring [42]. RC also exhibits significant pathophysiological crosstalk with insulin signaling pathways, wherein arterial accumulation of remnant lipoprotein particles exacerbates systemic insulin resistance. This metabolic dysregulation culminates in persistent hyperglycemia—an established contributor to HTN pathogenesis [43]. Furthermore, RC can stimulate the release of aldosterone and increase circulating blood volume, both contributing to HTN [44, 45]. RC may also lead to a low-grade inflammatory response, fundamental to the development of HTN [46]. Studies have shown that even when traditional lipid markers such as LDL-C are within normal ranges, elevated RC levels can independently increase the risk of HTN, highlighting its unique atherogenic properties. The interplay between RC and metabolic factors like insulin resistance and obesity further exacerbates its effects on blood pressure.

This study presents several novel and methodologically robust contributions to the field of cardiometabolic research. First, to our knowledge, this is the first investigation to evaluate the association between RC-LDL-C discordance and incident HTN in an Asian population. By employing RC as a continuous variable, we demonstrated its independent predictive capacity for HTN risk, while discordance analyses revealed incremental risk stratification beyond absolute RC levels. Importantly, the consistency of findings across multiple clinical cutpoints (e.g., percentile-based, guideline-recommended thresholds) underscores the robustness of our conclusions. Second, the longitudinal design of this study overcomes limitations inherent to cross-sectional analyses by establishing temporal relationships between RC-LDL-C discordance and HTN development, minimizing reverse causality bias and enabling dynamic risk assessment. Collectively, these strengths position our study as a critical reference for understanding lipid-driven HTN mechanisms and refining risk prediction models in understudied Asian cohorts.

This study has several limitations. First, potential confounding factors such as diet and exercise, which could have influenced the results, were not accounted for. Second, despite the longitudinal design of the CHARLS study, the long intervals between surveys resulted in imprecise time points. Additionally, given the complex and dynamic nature of RC metabolism, non-fasting assays may provide more meaningful insights. Furthermore, considering that LDL-C levels may be lower in East Asian populations, the generalizability of our findings to other ethnic groups requires further investigation [14]. Fifth, the loss to follow-up rate was 43.66%, which is a limitation of this study. However, we have demonstrated that the core exposure variable (RC) was fully balanced between retained and lost-to-follow-up participants, and IPCW sensitivity analysis yielded nearly identical results, indicating that attrition bias is unlikely to have a substantive impact on our conclusions. Sixth, in the most extreme low LDL-C subgroup (LDL-C < 100 mg/dL), the association between high RC and incident HTN reached only borderline statistical significance after full adjustment for the strongest HTN predictors. This is likely due to the relatively small sample size in this subgroup, which limited statistical power. However, the consistent direction of the HR across all three LDL-C thresholds suggests that the association persists even at very low LDL-C levels. Larger cohort studies with sufficient numbers of participants with extremely low LDL-C levels are needed to confirm this finding. Lastly, the results of this study need to be validated using different criteria and in the context of HTN diagnosis.

Conclusions

In summary, this investigation demonstrates that a discrepancy between RC and LDL-C is independently linked to an elevated risk of HTN onset in middle-aged and older Chinese populations. Notably, this association persists even when LDL-C levels are within or below the recommended range. Furthermore, the observed disparity between RC and LDL-C concentrations highlights its potential role as an independent risk factor for HTN in primary prevention strategies, emphasizing the need to address residual lipid-related risks beyond conventional LDL-C management.

Supplementary Information

12872_2026_6008_MOESM1_ESM.docx ^{(593.2KB, docx)}

Additional file 1: Supplemental Table 1. Baseline clinical characteristics of participants stratified by incident hypertension status during follow-up. Supplemental Table 2. Baseline characteristics of excluded participants (lost to follow-up) and retained participants. Supplemental Table 3. IPCW-weighted Cox model for RC and hypertension risk. Supplemental Table 4. Cox models analyzed the relationship between standardized differences in RC and LDL-C percentiles and the risk of hypertension (20 percentile-unit threshold). Supplemental Table 5. Enhancement of Hypertension Discrimination and Risk Reclassification Through the Incorporation of RC-LDL-C discordance (15% percentile difference-based). Supplement Figure 1. Flow chart of participants selection. Supplement Figure 2. Subgroup analysis for the association between RC and incident hypertension.

Abbreviations

ASCVD

Atherosclerotic cardiovascular disease

BMI

Body mass index

BUN

Blood urea nitrogen

CHARLS

China Health and Retirement Longitudinal Study

CI

Confidence interval

CRP

C-reactive protein

CVD

Cardiovascular disease

DBP

Diastolic blood pressure

DM

Diabetes mellitus

FBG

Fasting blood glucose

HbA1c

Glycated hemoglobin

HDL-C

High-density lipoprotein cholesterol

HR

Hazard ratio

HTN

Hypertension

LDL-C

Low-density lipoprotein cholesterol

non-HDL-C

Non-high-density lipoprotein cholesterol

RC

Remnant cholesterol

RCS

Restricted cubic spline

SBP

Systolic blood pressure

TC

Total cholesterol

TG

Triglycerides

TRLs

Triglyceride-rich lipoproteins

UA

Uric acid

CAP

College of American pathologists

VLDL-C

Very-low-density lipoprotein cholesterol

IDL-C

Intermediate-density lipoprotein cholesterol

CM-C

Chylomicron remnant cholesterol

Authors’ contributions

**Conceptualization: ** Xiang-Xiang Li, Bin Yan, Xiao-Fei Wu.**Data curation: ** Xiang-Xiang Li, Chun-Fang Ma, Yi-Chi Zhang, Xiao-Fei Wu.**Formal analysis: ** Xiang-Xiang Li, Yi-Chi Zhang, Xiao-Fei Wu.**Software: ** Xiang-Xiang Li, Xiao-Fei Wu.**Supervision: ** Yi-Chi Zhang, Shan Liu.**Writing – original draft: ** Yi-Chi Zhang, Chun-Fang Ma, Xiao-Fei Wu.**Writing – review & editing: ** Xiang-Xiang Li, Bin Yan.**Funding acquisition: ** Shan Liu, Chun-Fang Ma, Xiao-Fei Wu.

Funding

This study was supported by the Project of Suzhou Ninth People's Hospital (YK202220, YK202425, and YK202517).

Data availability

The current study’s dataset is available for public access on the following website: http://charls.pku.edu.cn/en.

Declarations

Ethics approval and consent to participate

This research was carried out in accordance with the principles outlined in the Declaration of Helsinki, and received approval from the Ethics Committee of Peking University (IRB00001052-11015). Approval for data collection was obtained from the Ethical Review Committee at Peking University. The participants provided their written informed consent to participate in this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.