Remnant cholesterol and the risk of incident type 2 diabetes: A community-based 14-year prospective cohort study

Abstract

BACKGROUND

Remnant cholesterol, a marker of triglyceride-rich lipoproteins, has been implicated in cardiometabolic risk; however, its role in predicting incident type 2 diabetes remains incompletely defined, particularly in comparison with conventional lipid parameters.

AIM

To investigate the association between remnant cholesterol and incident type 2 diabetes and its predictive performance in a community-based cohort.

METHODS

This prospective study included 7702 Korean adults without diabetes at baseline, who were followed for up to 14 years. Incident diabetes was ascertained using repeated oral glucose tolerance tests in addition to fasting plasma glucose, glycosylated hemoglobin, and medical history.

RESULTS

During follow-up, 22.0% (1694/7702) of participants developed incident diabetes. In multivariable Cox proportional hazards models, higher remnant cholesterol levels were independently associated with an increased risk of diabetes after adjustment for established risk factors (hazard ratio per 1-SD increase = 1.25; 95%CI: 1.20-1.30). This association was consistently observed across subgroups defined by age, sex, body mass index, metabolic syndrome status, and baseline glycemic status, and appeared more evident in metabolically healthier individuals. Overall, the predictive performance of remnant cholesterol was comparable to that of triglycerides and superior to that of total and low density lipoprotein cholesterol. Notably, remnant cholesterol provided incremental risk reclassification value over non-high density lipoprotein cholesterol among individuals with triglyceride levels < 400 mg/dL.

CONCLUSION

These findings suggest that remnant cholesterol may serve as a complementary marker for refining diabetes risk stratification, particularly in individuals with relatively low baseline metabolic risk.

Key Words: Remnant cholesterol; Triglycerides; Non-high density lipoprotein cholesterol; Diabetes; Cohort study

Core Tip: Remnant cholesterol, a surrogate of triglyceride-rich lipoprotein burden, has recently gained attention as a contributor to cardiometabolic disease. Our findings underscore its independent predictive value for incident type 2 diabetes, beyond conventional lipid indices and established metabolic risk factors. Given its robust performance across multiple subgroups, remnant cholesterol may serve as a clinically relevant biomarker for early metabolic risk stratification and complement current approaches in predicting diabetes development.

INTRODUCTION

High triglyceride-rich lipoproteins and hypertriglyceridemia are frequently observed lipid disorders that are commonly associated with obesity, metabolic syndrome, type 2 diabetes, and insulin-resistant status. Remnant cholesterol is a cholesterol component in triglyceride-rich lipoproteins mainly composed of very low-density lipoprotein and intermediate-density lipoproteins in fasting state and additionally by chylomicron remnants in non-fasting state. Remnant cholesterol can be measured directly or calculated as follows: Plasma total cholesterol - low density lipoprotein cholesterol (LDL-C) - high density lipoprotein cholesterol (HDL-C)[1-3]. Although remnant cholesterol is metabolically related to triglycerides, it reflects the cholesterol content of triglyceride-rich lipoproteins rather than triglyceride mass itself. Therefore, remnant cholesterol may better capture the atherogenic and metabolic burden associated with these particles beyond circulating triglyceride levels alone.

Recent evidence suggests that remnant cholesterol is a residual cardiovascular risk factor after attainment of LDL-C goal with statins[1-3]. In overweight or obese people at high cardiovascular risk, remnant cholesterol level, but not LDL-C level, was independently associated with the cardiovascular outcomes[4]. In addition, in an analysis of the secondary prevention cohort, lowering remnant cholesterol to 32 mg/dL could reduce the risk of recurrent major cardiovascular events by 20%[5]. Not only cardiovascular events, lipid parameters, including elevated triglycerides, non-HDL-C, apolipoprotein B, and low HDL-C, were also significantly associated with incident diabetes[6-9]. In individuals from the Framingham Offspring Study, high triglyceride and low HDL-C levels were independently associated with the risk of diabetes during a 7-year follow-up interval[10]. In a Korean population study, atherogenic dyslipidemia profiles, as determined by elevated apolipoprotein B and non-HDL-C, were associated with incident type 2 diabetes independent of diabetes risk factors and other conventional lipid measures[11].

In this study, we prospectively evaluated the association between remnant cholesterol levels and the development of type 2 diabetes in a community-based Korean cohort. We also compared its predictive performance with that of conventional lipid parameters.

MATERIALS AND METHODS

Study population

The Korean Genome and Epidemiologic Study Ansan and Ansung Study is an ongoing prospective population-based study conducted by the Korean Center for Disease Control and Prevention[12]. All participants were recruited from two communities in South Korea: The Ansung cohort; the rural community and the Ansan cohort; and the urban community. Baseline examinations were performed in 2001-2002, the follow-up examinations conducted every two years, and data collected up to 2015-2016 were used for the analyses. Details regarding the study’s design and baseline characteristics and follow-up rate at every visit have been published in a previous study[13]. From the 10030 subjects aged 40-69 years, we excluded 1449 patients with diabetes at baseline visit. We further excluded: (1) People with cancer (n = 98); (2) Those taking lipid-lowering medication (n = 57); and (3) Those with triglycerides level > 800 mg/dL (n = 34). Finally, the remaining 7702 subjects with at least one oral glucose tolerance test (OGTT) during follow-up were enrolled in this study. The primary objective of this study was to examine the association between remnant cholesterol and incident type 2 diabetes. Assessment of predictive performance was conducted as a secondary, exploratory analysis.

The study protocol was approved by the Ethics Committee of the Korean Center for Disease Control and the Kyung Hee University Hospital at Gangdong Institutional Review Board (Approval No. KHNMC 2023-05-010). Participants signed a written informed consent form and agreed to participate in the trial after fully understanding the study. All methods were carried out in accordance with relevant institutional guidelines and regulations.

Clinical and biochemical parameters

Baseline information on medical history and sociodemographic characteristics was collected using standardized questionnaires along with direct anthropometric assessments and laboratory testing. Participants were categorized according to their smoking status as never, former, or current smokers. Alcohol intake was quantitatively evaluated, with heavy alcohol consumption defined as an average daily consumption of ≥ 30 g for men and ≥ 20 g for women[14]. Engagement in regular physical activity was determined based on participation in moderate- or vigorous-intensity exercise for at least 30 minutes per day. Body mass index (BMI) was calculated as weight divided by the square of height (kg/m²) in light clothing. Waist circumference was measured at the midpoint between the lower limit of the ribcage and the iliac crest. After overnight fasting for 12 hours, blood samples were collected and analyzed at a central laboratory (Seoul Clinical Laboratories, Seoul, South Korea). The plasma concentrations of glucose, total cholesterol, triglycerides, and HDL-C were measured enzymatically using a 747 Chemistry Analyzer (Hitachi, Tokyo, Japan). Glycosylated hemoglobin (HbA1c) levels were measured by high-performance liquid chromatography (VARIANT II; Bio-Rad Laboratories, Hercules, CA, United States). Plasma insulin concentrations were determined using a radioimmunoassay kit (Linco Research, St. Charles, MO, United States). C-reactive protein concentrations were measured by immunoradiometric assay (ADVIA 1650; Bayer Diagnostics, Tarrytown, NY, United States).

Definitions

The definition of diabetes was based on plasma glucose levels during the 75 g OGTT and HbA1c levels, following the criteria established by the American Diabetes Association[15]. The homeostasis model assessment of β-cell function (HOMA-β%) and insulinogenic index (IGI) were calculated to determine the insulin secretory capacity. Insulin sensitivity was estimated using the homeostasis model assessment of insulin resistance (HOMA-IR) and the composite (Matsuda) insulin sensitivity index. To determine pancreatic β-cell function adjusting for insulin sensitivity, the oral disposition index was calculated[16].

Estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease formula[17], and chronic kidney disease was defined as an eGFR < 60 mL/minute/1.73 m². Hypertension was defined as one of the following: (1) Self-reported previous history of hypertension; (2) Systolic or diastolic blood pressure ≥ 140/90 mmHg; or (3) Use of antihypertensive medications[18]. Hyperlipidemia was defined as one of the following: (1) Self-reported previous history of lipid abnormalities; or (2) Total cholesterol ≥ 240 mg/dL[19]. Cardiovascular disease (CVD) was defined as either myocardial infarction, coronary artery disease, congestive heart failure, stroke, or peripheral artery disease, and data on each event were obtained from the participants’ reports. The CVD events reported by participants were corroborated by an in-depth interview during the initial examination and at interviews repeated at biennial follow-up visit. Non-alcoholic fatty liver disease liver fat score[20] was calculated to ascertain the presence of fatty liver disease. The presence of metabolic syndrome was ascertained using NCEP-ATP III criteria[21].

Estimation of remnant cholesterol

The estimation of remnant cholesterol level was derived by subtracting both LDL-C and HDL-C levels from total cholesterol level. However, the Friedewald equation becomes completely unreliable, when triglycerides are more than 400 mg/dL[22]. Furthermore, remnant cholesterol is equivalent to triglyceride divided by five when calculating LDL-C according to the Friedewald equation, and thus remnant cholesterol does not provide any additional information beyond triglycerides[23]. Therefore, the Sampson-NIH2 equation to calculate LDL-C which is more accurate in people with low LDL-C levels and/or hypertriglyceridemia less than 800 mg/dL was used as follows[24]: LDL-C = total cholesterol/0.948 - HDL-C/0.971 - (triglyceride/8.56 + triglyceride × non-HDL-C/2140 - triglyceride²/16100) - 9.44.

The validity of the Sampson-NIH2 equation in estimating LDL-C levels in the Korean population was determined using data from the Korea National Health and Nutrition Examination Survey (KNHANES) 2015. Because the KNHANES data contains the concentrations of total cholesterol, HDL-C, triglyceride, and directly-measured LDL-C, it enables a comparison between the directly-measured and estimated LDL-C level with using the Friedewald equation or the Sampson-NIH2 equation.

Comparison of calculated and measured LDL-C levels

In a total of 5860 subjects aged 21-80 years, we excluded 20 with triglycerides level > 800 mg/dL, because the Sampson-NIH2 equation is valid for triglyceride levels up to 800 mg/dL.

The coefficients of determination (R²) of the Friedewald and Sampson-NIH2 equations with directly-measured LDL-C were 0.917 and 0.944 respectively in all participants (n = 5840) as well as 0.934 and 0.949 respectively in subjects with triglycerides level < 400 mg/dL (n = 5730; Supplementary Figure 1 and Supplementary Table 1).

Statistical analysis

Data are expressed as mean ± SD for continuous variables and as proportions for categorical variables. Group differences were assessed using Student’s t-test for continuous variables and the χ² test for categorical variables. Spearman’s correlation coefficients were used to evaluate associations between remnant cholesterol and indices of insulin secretion, insulin resistance, and low-grade inflammation.

HRs with 95%CIs per 1-SD increment were estimated using Cox proportional hazards models. Covariates were selected a priori based on clinical relevance and previous literature, with additional consideration of the variables that showed significant associations in univariable analyses. When variables reflected overlapping pathophysiological information, those with stronger associations were preferentially included. Multicollinearity was assessed using variance inflation factors, with values < 5 indicating no significant multicollinearity. The final multivariable model included variables representing demographic (age, sex), anthropometric (BMI, waist circumference), lifestyle (smoking status, heavy alcohol consumption, physical activity), metabolic (fasting plasma glucose and insulin), renal (eGFR), and hepatic (alanine aminotransferase) factors associated with diabetes risk.

The predictive performance of remnant cholesterol was compared with other lipid parameters using the integrated discrimination index and net reclassification improvement (NRI). Interaction effects were examined by including first-order interaction terms in multivariable models. The NRI was calculated using a continuous NRI approach rather than predefined risk categories. No optimism correction or internal validation (e.g., bootstrapping) was performed for the reclassification analyses. All statistical analyses were performed using SPSS software (version 24.0), and a two-sided P value < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics of study participants who developed incident diabetes

During the 14 years of follow-up, 22.0% (1694/7702) of participants developed incident diabetes. People with incident diabetes were older, more likely to be male, more obese, showed unfavorable metabolic phenotypes, including higher blood pressure and glucose levels, atherogenic dyslipidemia profiles, and higher fasting insulin and liver enzymes, compared with those who did not develop diabetes. In particular, baseline remnant cholesterol level in individuals with incident diabetes was significantly higher (32.0 ± 16.2 mg/dL vs 25.7 ± 12.8 mg/dL; P < 0.001) compared with that in individuals without incident diabetes (Table 1).

Table 1 Baseline characteristics of the study participants.

Variable	Total (n = 7702)	Incident diabetes (-; n = 6008)	Incident diabetes (+; n = 1694)	P value
Age (years)	51.7 ± 8.8	51.3 ± 8.8	52.8 ± 8.5	< 0.001
Female	4089 (53.1)	3233 (53.8)	856 (50.5)	0.017
Current smoking	1917 (25.2)	1457 (24.6)	460 (27.4)	0.018
Heavy alcohol drink	732 (9.8)	541 (9.3)	191 (11.6)	0.007
Physical activity	4522 (59.7)	3533 (59.9)	989 (59.3)	0.652
Family history of diabetes	773 (10.0)	546 (9.1)	227 (13.4)	< 0.001
Hypertension	970 (12.6)	619 (10.3)	351 (20.7)	< 0.001
Hyperlipidemia	697 (9.1)	506 (8.4)	191 (11.3)	< 0.001
Cardiovascular disease	163 (2.1)	125 (2.1)	38 (2.2)	0.682
Chronic kidney disease	167 (2.2)	121 (2.0)	46 (2.7)	0.088
Fatty liver	1563 (20.4)	1005 (16.8)	558 (33.2)	< 0.001
eGFR (mL/minute/1.73 m2)	91.5 ± 16.7	91.6 ± 16.5	91.1 ± 17.3	0.235
Body mass index (kg/m2)	24.4 ± 3.1	24.2 ± 3.0	25.3 ± 3.2	< 0.001
Waist circumference (cm)	82.1 ± 8.7	81.4 ± 8.6	84.7 ± 8.6	< 0.001
Systolic blood pressure (mmHg)	120.5 ± 18.0	119.2 ± 17.7	124.9 ± 18.3	< 0.001
Diastolic blood pressure (mmHg)	79.9 ± 11.4	79.2 ± 11.3	82.5 ± 11.2	< 0.001
Fasting plasma glucose (mg/dL)	82.6 ± 8.5	81.5 ± 7.8	86.5 ± 9.7	< 0.001
Post-load glucose 120 minutes (mg/dL)	114.0 ± 30.0	108.4 ± 26.8	133.7 ± 32.6	< 0.001
Fasting serum insulin (μIU/mL)	7.55 ± 4.78	7.44 ± 4.72	7.95 ± 4.96	< 0.001
Glycated hemoglobin	5.54 ± 0.34	5.49 ± 0.32	5.74 ± 0.34	< 0.001
Total cholesterol (mg/dL)	189.3 ± 34.0	187.9 ± 34.0	194.2 ± 33.7	< 0.001
HDL-C (mg/dL)	45.0 ± 10.1	45.5 ± 10.1	43.2 ± 9.7	< 0.001
Triglycerides (mg/dL)	152.7 ± 84.1	144.4 ± 77.5	182.1 ± 98.8	< 0.001
Sampson-NIH2 LDL-C (mg/dL)	117.2 ± 30.8	116.7 ± 30.8	119.0 ± 30.7	0.006
Non-HDL-C (mg/dL)	144.3 ± 33.2	142.4 ± 33.1	151.0 ± 33.0	< 0.001
Remnant cholesterol (mg/dL)	27.1 ± 13.9	25.7 ± 12.8	32.0 ± 16.2	< 0.001
Blood urea nitrogen (mg/dL)	14.3 ± 3.6	14.2 ± 3.6	14.4 ± 3.7	0.022
Creatinine (mg/dL)	0.84 ± 0.19	0.84 ± 0.19	0.85 ± 0.20	0.042
Aspartate aminotransferase (U/L)	29.1 ± 16.8	28.7 ± 17.0	30.5 ± 15.0	< 0.001
Alanine aminotransferase (U/L)	26.9 ± 21.7	25.7 ± 19.6	31.0 ± 27.6	< 0.001
γ-glutamyl transpeptidase (U/L)	31.9 ± 58.0	28.5 ± 42.4	44.0 ± 93.5	< 0.001

Data are shown as mean ± SD or n (%). eGFR: Estimated glomerular filtration rate; HDL-C: High density lipoprotein cholesterol; LDL-C: Low density lipoprotein cholesterol.

The association between remnant cholesterol and incident diabetes

In the univariate Cox hazards model, age, obesity and other traditional risk factors for diabetes were significantly associated with the development of diabetes. Among lipid parameters, total cholesterol, triglycerides, Sampson-NIH2 LDL-C, and non-HDL-C levels were positively associated with incident diabetes and HDL-C level was inversely associated with incident diabetes. Remnant cholesterol level was predictive of incident diabetes (HR per 1-SD = 1.36, 95%CI: 1.31-1.40; P < 0.001; Supplementary Table 2).

To determine whether lipid parameters including remnant cholesterol could independently predict future development of diabetes, a multivariate Cox proportional hazard regression model included variables which were significantly associated with diabetes risk in univariate analyses (Table 2). Since some significant variables in the univariate model are highly correlated and/or provide similar information, we selected the variables for the multivariate model based on its higher HR in univariate analysis and/or clinical availability. In the fully adjusted model 4, triglycerides (HR per 1-SD = 1.25, 95%CI: 1.20-1.30; P < 0.001), non-HDL-C (HR per 1-SD = 1.10, 95%CI: 1.04-1.16; P < 0.001), and remnant cholesterol levels (HR per 1-SD = 1.25, 95%CI: 1.20-1.30; P < 0.001) were significantly associated with future diabetes risk, regardless of age, sex, BMI, waist circumference, systolic blood pressure, current smoking, heavy alcohol consumption, physical activity, family history of diabetes, fasting plasma glucose, fasting serum insulin, eGFR, and alanine aminotransferase. On the other hand, total and Sampson-NIH2 LDL-C levels were not associated with diabetes risk in the fully adjusted model 4 (Table 2).

Table 2 Association of remnant cholesterol with incident type 2 diabetes.

Model	Total cholesterol	Triglycerides	Sampson-NIH2 LDL-C	Non-HDL cholesterol	Remnant cholesterol
1	1.20 (1.14-1.26)	1.35 (1.30-1.39)	1.08 (1.03-1.14)	1.27 (1.22-1.34)	1.36 (1.31-1.40)
2	1.19 (1.13-1.24)	1.34 (1.29-1.38)	1.08 (1.03-1.13)	1.26 (1.20-1.32)	1.35 (1.30-1.40)
3	1.10 (1.05-1.16)	1.25 (1.20-1.30)	1.02 (0.97-1.07)	1.15 (1.10-1.21)	1.26 (1.21-1.31)
4	1.04 (0.99-1.09)	1.25 (1.20-1.30)	0.97 (0.92-1.02)	1.10 (1.04-1.16)	1.25 (1.20-1.30)

Data are expressed as HRs with 95%CIs. Model 1: Crude; Model 2: Adjusted for age and sex; Model 3: Adjusted for variables in Model 2 plus body mass index, waist circumference, systolic blood pressure, current smoking, heavy alcohol consumption, physical activity, and family history of diabetes; Model 4: Adjusted for variables in Model 3 plus fasting plasma glucose, fasting serum insulin, estimated glomerular filtration rate, and alanine aminotransferase. HDL-C: High density lipoprotein cholesterol; LDL-C: Low density lipoprotein cholesterol.

Subgroup analysis

In subgroup analyses based on model 4, higher remnant cholesterol levels were significantly associated with incident type 2 diabetes across subgroups defined by age, sex, BMI, baseline glycemic status, and the presence of metabolic syndrome. Formal interaction tests were performed to assess effect modification. Significant interactions were observed for sex, obesity status, baseline glycemic status, and metabolic syndrome status (all P for interaction < 0.05), whereas the interaction with age was not statistically significant. Accordingly, the association between remnant cholesterol and incident diabetes appeared more pronounced in women, non-obese individuals, those with normal glucose tolerance, and those without metabolic syndrome at baseline. When stratified by baseline triglyceride levels, remnant cholesterol was significantly associated with incident diabetes among participants with triglyceride levels < 400 mg/dL (n = 7172), but not among those with triglyceride levels ≥ 400 mg/dL (n = 132; Figure 1).

Figure 1 Association between remnant cholesterol and incident diabetes risk across subgroups. HRs and 95%CIs for incident diabetes per 1-SD increase in remnant cholesterol are shown according to baseline subgroup characteristics. BMI: Body mass index; NGT: Normal glucose tolerance.

Remnant cholesterol for the prediction of incident diabetes

Because total and Sampson-NIH2 estimated LDL-C levels were not independently associated with diabetes risk in the multivariate model, the performance of triglycerides, non-HDL-C, and remnant cholesterol was compared by calculating the NRI and integrated discrimination index (IDI). Table 3 shows the results of the risk re-classification for the prediction based on the fully adjusted model 4. When the NRI and IDI statistics were calculated, remnant cholesterol showed comparable risk prediction to triglycerides and non-HDL-C in the total population. On the other hand, in people with baseline triglycerides less than 400 mg/dL, remnant cholesterol resulted in a significant improvement of 5.0% and 53.4% prediction using IDI and NRI respectively, compared to a model predicted by non-HDL-C (Table 3).

Table 3 Comparison of predictive performance of lipid parameters for incident type 2 diabetes.

	All (n = 7304)	Triglyceride < 400 mg/dL (n = 7172)	Triglyceride ≥ 400 mg/dL (n = 132)
IDI
Remnant cholesterol vs triglycerides	0.001 (-0.005 to 0.006)	-0.001 (-0.007 to 0.004)	0.004 (-0.053 to 0.049)
Remnant cholesterol vs non-HDL-C	0.033 (-0.038 to 0.093)	0.050 (0.015-0.120)	-0.006 (-0.110 to 0.085)
NRI
Remnant cholesterol vs triglycerides	0.061 (-0.433 to 0.567)	0.076 (-0.471 to 0.478)	0.542 (-1.306 to 1.797)
Remnant cholesterol vs non-HDL-C	0.351 (-0.127 to 0.699)	0.534 (0.096-0.882)	-0.470 (-1.758 to 0.990)

Data are expressed as indices with 95%CIs. Comparisons are based on Model 4 for each lipid parameter. IDI: Integrated discrimination index; HDL-C: High density lipoprotein cholesterol; NRI: Net reclassification improvement.

Correlation between remnant cholesterol and the marker of insulin secretion/sensitivity or inflammatory markers

The remnant cholesterol level was significantly associated with the markers of insulin secretion (Supplementary Table 3). However, there was a positive correlation with HOMA-β% (r = 0.154, P < 0.001) and IGI 60 minutes (r = 0.070, P < 0.001), while there was an inverse correlation with disposition index (r = -0.068, P < 0.001). The remnant cholesterol level was also significantly associated with a decreased insulin sensitivity (Supplementary Table 3), as indicated by HOMA-IR (r = 0.173, P < 0.001) and Matsuda index (r = -0.227, P < 0.001). Additionally, there was a notable positive correlation between the remnant cholesterol level and inflammatory markers, including white blood cell counts and CRP (Supplementary Table 3).

DISCUSSION

In this community-based prospective cohort of 7702 Korean adults free of diabetes, remnant cholesterol was independently associated with the development of type 2 diabetes over 14 years of follow-up. Although the effect size was modest, the clinical relevance of remnant cholesterol lies not in replacing established risk factors but in complementing them. In exploratory subgroup analyses, the association appeared to be more evident in metabolically healthier individuals. Overall, the predictive performance of remnant cholesterol was comparable to that of triglyceride and non-HDL-C in the total population, while remnant cholesterol provided superior risk reclassification compared with non-HDL-C among individuals with triglyceride levels < 400 mg/dL. By incorporating long-term follow-up with repeated OGTT and comparative risk reclassification analyses, this study extends existing evidence and suggests that remnant cholesterol may serve as a complementary marker for refining diabetes risk stratification, particularly in individuals with triglyceride levels < 400 mg/dL and relatively low baseline metabolic risk.

Several equations, including the Martin-Hopkins equation[25] and the Sampson-NIH2 equation[24], have been proposed to overcome the limitations of the Friedewald equation for estimating LDL-C. In the present study, remnant cholesterol was estimated using the Sampson-NIH2 equation, which derives the cholesterol content of triglyceride-rich lipoproteins based on total cholesterol, HDL-C, and triglyceride concentrations. While this approach may provide a more accurate estimation of remnant cholesterol compared with traditional formulas[26], we acknowledge that remnant lipoproteins were not directly measured.

Previous studies have consistently reported associations between elevated remnant cholesterol levels and an increased risk of cardiometabolic outcomes, including type 2 diabetes[27-32]. Beyond corroborating these findings, the present study extends prior work in several important ways. First, this study was conducted in a large, community-based prospective cohort with a long follow-up duration of up to 14 years, which is substantially longer than most previous studies. Diabetes was ascertained using repeated OGTT in addition to fasting plasma glucose, HbA1c, and medical history, allowing more accurate classification of diabetes at baseline and during follow-up. Second, we directly compared the predictive performance of remnant cholesterol with multiple conventional lipid parameters using risk reclassification metrics. When incorporated into models with non-HDL-C, remnant cholesterol modestly improved risk discrimination and reclassification, particularly among individuals with triglyceride levels of < 400 mg/dL. Third, the association between remnant cholesterol and incident diabetes was consistently observed across age, sex, BMI, metabolic syndrome status, and baseline glycemic categories, and was more pronounced in metabolically healthier individuals, suggesting a potential role as a complementary marker for early diabetes risk stratification.

In exploratory subgroup analyses, remnant cholesterol appeared to be more strongly associated with incident type 2 diabetes among younger, female, and non-obese individuals without metabolic syndrome or impaired glucose tolerance at baseline. These analyses were not prespecified and should therefore be interpreted as hypothesis-generating rather than confirmatory. Similar patterns have been reported in previous studies, in which the association between remnant cholesterol and diabetes risk was more pronounced among individuals at relatively lower baseline risk and attenuated in those with prediabetes[30], or insulin resistance[31]. Sex-specific differences have also been reported, with a weaker association between remnant cholesterol and metabolic risk observed in men compared with women[33].

These subgroup findings should be interpreted with caution. The absence of a significant association between remnant cholesterol and incident diabetes in certain subgroups, particularly among individuals with very high triglyceride levels, may reflect both reduced measurement reliability of calculated remnant cholesterol and the predominance of other metabolic abnormalities that overshadow its independent contribution. In contrast, in metabolically healthier individuals, the relative contribution of remnant cholesterol to diabetes risk may be more readily detectable because competing metabolic risk factors are less prominent. These subgroup findings should therefore be interpreted with caution and regarded as hypothesis-generating rather than indicative of a specific biological mechanism.

The biological mechanisms linking elevated remnant cholesterol to diabetes risk remain incompletely understood. Previous studies have suggested that insulin resistance and low-grade inflammation may partially mediate this association[34,35]. Consistent with these observations, we found that remnant cholesterol was significantly correlated with indices reflecting insulin resistance (HOMA-IR and Matsuda index) and systemic inflammation (white blood cell and C reactive protein). In addition, remnant cholesterol was inversely associated with the oral disposition index, a marker of insulin secretory capacity adjusted for insulin sensitivity. Although this observation is novel, it should be interpreted with caution because the underlying mechanisms linking remnant cholesterol to pancreatic β-cell dysfunction have not been established in human studies, and are largely based on experimental evidence. Remnant cholesterol-rich lipoproteins may contribute to β-cell dysfunction indirectly through increased lipid uptake, intracellular cholesterol accumulation, and subsequent impairment of insulin secretory pathways, as suggested by experimental models[36,37]. While these findings provide biological plausibility, they do not establish a causal mechanism, and further studies are warranted to confirm the relationship between remnant cholesterol and β-cell function in humans.

We analyzed a large, community-based prospective cohort without restrictive inclusion criteria and with a long follow-up duration of up to 14 years. In addition, we comprehensively adjusted for potential confounders using both clinical information and laboratory data, which strengthens the robustness of our findings.

However, several limitations should also be acknowledged. First, remnant cholesterol was calculated rather than directly measured, and lipid parameters were assessed only once at baseline, which may introduce measurement uncertainty. Because lipid levels can change over time due to aging, lifestyle modification, or medication use, reliance on a single baseline measurement may have led to regression dilution bias[38], potentially resulting in an underestimation of the true association between remnant cholesterol and incident diabetes. Although the Sampson-NIH2 equation used to estimate LDL-C has been validated in previous studies[24,26] and demonstrated high agreement with directly measured LDL-C in our data[39], calculated remnant cholesterol may still be less reliable at higher triglyceride concentrations. Indeed, previous studies have reported discrepancies between calculated and directly measured remnant cholesterol in predicting cardiovascular outcomes[40], and further studies comparing calculated and directly measured remnant cholesterol in relation to incident diabetes are warranted. In addition, autoantibodies related to type 1 diabetes were not measured; however, given the low incidence of type 1 diabetes in South Korea[41], misclassification is likely minimal. Although our study was conducted in a Korean population, similar associations between remnant cholesterol and incident diabetes have been reported in western[35] and other Asian cohorts[30,32], suggesting that the observed relationship may not be limited to a specific ethnic group.

CONCLUSION

Remnant cholesterol was independently associated with the future development of type 2 diabetes in this Korean population. Although this association was observed across subgroups with diverse baseline characteristics, its predictive value appeared more evident in metabolically healthier individuals, in whom conventional risk factors may underestimate future diabetes risk. In predicting incident diabetes, remnant cholesterol performed better than total and LDL-C and showed comparable performance to triglycerides, while providing complementary information for risk stratification, particularly among individuals with triglyceride levels < 400 mg/dL. Accordingly, remnant cholesterol should be regarded as a complementary marker for diabetes risk stratification rather than a replacement for triglycerides, especially in individuals with relatively low baseline metabolic risk.

ACKNOWLEDGEMENTS

Data in this study were from the Korean Genome and Epidemiology Study (KoGES; 6635-302), National Institute of Health, Korea Disease Control and Prevention Agency, Republic of Korea.