Remnant Cholesterol Is an Independent Predictor of Type 2 Diabetes: A Nationwide Population-Based Cohort Study

Type 2 diabetes (T2D) is a serious global, public-health concern, and has a high morbidity and mortality rate, secondary to cardiovascular disease (CVD) (1). A recent systematic review revealed that several countries, including Korea, have shown decreasing, or increasing yet stable, incidences of T2D (2). However, approximately a quarter and half of the patients with diabetes in the U.S. and Asia, respectively, remain undiagnosed (3). Therefore, screening and early detection of T2D is an important concern.

Multiple risk factors for T2D, including age, family history, prediabetes, obesity, dyslipidemia, and lack of physical activity, have been used as variables in prediction or screening scores/models to assess the incidence of T2D worldwide (4). Among these, the link between T2D and dyslipidemia is relatively complex since most patients with T2D have typical dyslipidemia, which is characterized by hypertriglyceridemia, low levels of HDL cholesterol (HDL-C), and high concentrations of small dense LDLs; these atherogenic lipid changes may be consequences of T2D (5). However, recent studies showed that these changes may not only cause disturbances in glucose metabolism, (6) but may also be associated with the progression of T2D (7,8).

Remnant cholesterol (remnant-C), characterized by the cholesterol content of triglyceride (TG)-rich lipoproteins, consisting of VLDL and intermediate-density lipoprotein in the fasting state, as well as the chylomicron remnants in the non-fasting state (9), have been linked to an increased risk of all-cause mortality and cardiovascular outcomes (10,11). Studies have shown that the level of serum remnant-C is significantly elevated in patients with T2D (12), and that higher levels of remnant-C not only increase the risk of microvascular complications, but can also lead to macrovascular complications in diabetes (12,13). Recent studies demonstrated that remnant-C predicts newly developed T2D beyond traditional lipid parameters (14–16). Notably, these previous studies were conducted on a small sample size (14,15) or a specific group (16). Therefore, we aimed to evaluate the association between remnant-C and the risk of newly developed T2D in the general population using the National Health Insurance Service (NHIS) health checkup data.

We evaluated the data from the NHIS–Health Screening Cohort (NHIS-HEALS), which was provided by the NHIS from 2002 to 2019 in the Republic of Korea. In Korea, 97% of the population is required to enroll in this program (17). The NHIS database has been described in detail in a previous study (17) and is open to any researchers whose study protocols are approved by the official review committee.

Adults who underwent a health examination in 2009 (n = 10,601,274) were selected; of whom, 2,115,735 adults with the following criteria were excluded: age <20 years (n = 15,431), TG levels ≥3.4 mmol/L (300 mg/dL) (n = 563,442), had diabetes prior to the index year (n = 877,018), and were missing data on fasting blood glucose levels (n = 1,951), cholesterol levels (n = 69,145), or other parameters, including blood pressure and creatinine levels (n = 2,532), body composition indices (n = 4,212), lifestyle factors (n = 317,482), and income levels (n = 214,416). Finally, 8,485,539 participants were included in this analysis and were categorized according to the remnant-C quartiles (Supplemental Fig. 1). These study participants were followed up until the date of the incident diabetes or until 31 December 2019. This study was approved by the Hallym University Sacred Heart Hospital Institutional Review Board (IRB No. HALLYM 2021-08-002), and permission was granted to use the NHIS health checkup data (NHIS-REQ000042107-001). The requirement for written informed consent was waived in the current study because the data in the NHIS database are anonymized in adherence to strict confidentiality guidelines.

Detailed information on the demographics and lifestyles of the participants was obtained through standardized self-reporting questionnaires. The low-income level was defined as being in the lower one‐fifth of the entire population. Smoking status was classified as a nonsmoker, former smoker, or current smoker. Alcoholic drinking was categorized into 0 (none), 30 (mild), or ≥30 g/day (heavy). Regular exercise was defined as a vigorous-intensity or moderate-intensity physical activity performed at least three times or five times per week, respectively (18).

All blood samples were drawn after an overnight fast. Blood samples for the measurement of total cholesterol, HDL-C, and TG levels were obtained at the health examination after the participant had fasted for at least 8 h. Lipid profiles, including the levels of total cholesterol, LDL‐C, HDL‐C, and TG, were measured using an enzymatic method. The LDL-C levels were calculated based on total cholesterol, HDL-C, and TG levels according to the Friedewald formula, unless TG was significantly elevated (>4.5 mmol/L or 400 mg/dL) (19). Quality control of laboratory tests was conducted by the procedures of the Korean Association of Laboratory Quality Control (20). Remnant-C was estimated as total cholesterol minus LDL-C, minus HDL-C. Non–HDL-C was calculated as total cholesterol minus HDL-C (11).

The primary outcome was the development of T2D between 1 January 2009 and 31 December 2018 for each participant. However, to avoid the confounding by pre-existing disease and minimize the possible effects of “reverse causation,” we excluded participants who developed T2D within 1 year after the baseline measurements (n = 50,106). Newly developed T2D was defined as a fasting plasma glucose level ≥126 mg/dL during health examinations, or at least one claim per year for the prescription of hypoglycemic drugs under ICD-10 codes E11–14 in the outpatient or inpatient setting and the prescription of at least one antidiabetes drug at any time over 1 year to exclude individuals who had prediabetes or did not have diabetes (21). Hypertension was defined as a blood pressure ≥140/90 mmHg, or at least one claim per year for antihypertensive medication prescription under the ICD-10 codes I10–I15. Dyslipidemia was defined as a total cholesterol level ≥240 mg/dL or one or more claims per year for anti-hyperlipidemic medications with ICD‐10 code E78 (22). Chronic kidney disease (CKD) was defined as a glomerular filtration rate of <60 mL/min/1.73 m² and as a combination of ICD-10 codes N18–19, Z49, Z94.0, and Z99.2. Metabolic syndrome was defined based on the modified criteria of the National Cholesterol Education Program Adult Treatment Panel (NCEP-ATP III) (23), while the Asian-specific waist circumference cutoff was adopted for abdominal obesity (24). We defined a statin user, or fibrate user, as people who had been prescribed those medications at baseline. Obesity was defined as BMI ≥25 kg/m² based on the Asia-Pacific criteria of the World Health Organization guidelines (24).

The general characteristics of study participants are expressed as means ± SDs for continuous variables and numbers (percentages) for categorical variables. Non-normally, distributed variables were log-transformed to achieve normality before analyses. Values between groups were compared using ANOVA for continuous variables and the χ² test for categorical variables. The trend between variables was tested using linear regression for continuous variables and Cochran-Armitage for categorical variables. The incidence rates of diabetes were calculated by dividing the number of events by 1,000 person-years. The hazard ratios (HRs) with 95% CIs for risk of diabetes, according to remnant-C and other lipid levels, were analyzed by multivariable Cox proportional hazard models. The Kaplan–Meier method and log-rank tests were used to compare the cumulative incidence of incident diabetes among remnant-C quartile groups.

In the current study, subgroup analyses were also performed using multivariable, Cox proportional, hazard models stratified by the presence or absence of comorbidities (impaired fasting glucose [IFG], obesity, hypertension, CKD, metabolic syndrome, high levels of TG, non–HDL-C, and LDL-C, and low level of HDL-C), and the use of anti-hyperlipidemic agents (statin and fibrate) through a stratified analysis and interaction testing using a likelihood ratio test. P values <0.05 were considered statistically significant. A restricted cubic spline transformation of remnant-C was used to evaluate nonlinear associations between remnant-C and the risk of T2D. Statistical analysis was performed using SAS 9.4 (SAS Institute, Cary, NC) and R 3.1.0 (R Foundation for Statistical Computing, Vienna, Austria) software.

The baseline characteristics of study participants are presented by remnant-C quartiles in Table 1. Participants in the higher remnant-C quartiles were younger than those in lower remnant-C quartiles, and more likely to be men. Moreover, they had a higher proportion of current smokers and heavy alcohol drinkers, and were more likely to exercise less regularly than the participants in the lower remnant-C quartiles. Hypertension, dyslipidemia, including fibrate or statin users, and CKD were more prevalent in patients in the higher remnant-C quartiles. Furthermore, participants in higher remnant-C quartiles had higher BMIs, waist circumference, blood pressure, fasting blood glucose, and lipid levels than those in lowest remnant-C quartiles.

During the median follow-up period of 9.28 (interquartile range 9.08–9.55) years, 584,649 individuals (6.8% of the total population) developed T2D. The T2D incidence over time, according to baseline lipid concentrations, is shown in Table 2. The serum concentrations of TG, LDL-C, and non-HDL-C were associated with a 3.9%, 1.7%, and 3.5% higher risk of T2D per every 10 mg/dL increase, respectively; whereas, remnant-C was associated with a 13.1% increased risk per 10 mg/dL increase. The risk of T2D increased 16%, 25.4%, 5.8%, and 13.2% per 1-SD increase in remnant-C, TG, LDL-C, and non-HDL-C. It decreased about 15% per 1-SD increase in HDL-C.

When considering remnant-C as a continuous measure, the cubic spline graph showed a positive dose-response, but nonlinear relationship between remnant-C and incidence of T2D (Supplementary Fig. 2). Increasing remnant-C levels were associated with a corresponding rise in HRs of T2D, but the positive trend plateaued when remnant-C levels reached their highest levels. We further examined the association between quartiles of remnant-C and T2D risk (Table 3). The incidence of T2D was particularly high in participants in the upper quartiles of the remnant-C compared with those in the lowest quartiles. Even when adjusting for confounding factors, such as age, sex, current smoking status, regularity of exercise, income (lowest 20%), BMI, fasting blood glucose levels, presence of hypertension and CKD, and medication for dyslipidemia (use of statin and fibrate), the positive relationship between the remnant-C and the risk of T2D persisted (HR [95% CI]: first quartile [Q1], reference; second quartile [Q2], 1.25 [1.24–1.27]; third quartile [Q3], 1.51 [1.50–1.53]; fourth quartile [Q4], 1.95 [1.93–1.97]; P for trend <0.001). An incrementally higher risk of T2D was observed in individuals in higher remnant-C quartiles compared with those in the lowest quartiles across all models. We also analyzed the risk of T2D according to the quartiles of TG, LDL-C, and non-HDL-C. The HRs (95% CIs) of T2D in the highest quartile were 0.65 (0.64–0.65) in HDL-C, 1.17 (1.16–1.18) in LDL-C, 1.98 (1.96–2.00) in TG, and 1.42 (1.41–1.43) in non–HDL-C in the fully adjusted model (Supplementary Table 1).

As shown in Fig. 1, participants in the upper quartiles of the remnant-C had a significantly higher cumulative incidence of T2D compared with those in the lowest quartiles of the remnant-C. The Kaplan-Meier analysis clearly demonstrated a significantly higher probability of developing T2D in study participants within the highest quartile of remnant-C compared with those within the lowest quartile. Furthermore, the increased risk of T2D in the upper quartiles of the remnant-C observed in the early years persisted up to ∼9.3 years of follow-up (P < 0.001 by log-rank test).

We conducted subgroup analyses stratified by sex and the presence or absence of comorbidities (Supplementary Table 2). In each subgroup stratified, the positive association between remnant-C levels and incident T2D was consistent; however, there was an interaction between subgroups. The relative increase in T2D incidence among participants in the highest quartiles of remnant-C was modestly greater in women and in those without comorbidities, such as IFG, obesity, hypertension, CKD, metabolic syndrome, and atherogenic dyslipidemia, compared to those with comorbidities; this difference was significantly observed in an adjusted model (all P values for interaction <0.001). Moreover, the risk of T2D based on the remnant-C quartiles showed a significant interaction with antihyperlipidemic agents use, with higher T2D risks among participants in the upper quartiles of the remnant-C who were not using fibrate or statins than those who were using these agents (all P values for interaction <0.001).

As age itself is an important risk factor for both T2D and dyslipidemia, we further investigated the association between remnant-C and risk of T2D stratified by age groups (20–29 years, 30–39 years, 40–49 years, 50–59 years, 60–69 years, and ≥70 years) using cubic spline graph (Fig. 2). We found a positive dose-response but nonlinear relationship between remnant-C and the incidence of T2D in all age groups. However, the magnitude of the increased risk for T2D in individuals in the higher remnant-C quartiles was greater in younger than in older participants. The adjusted HR (95% CI) of T2D in the highest remnant-C quartile was 3.06 (2.93–3.20) in ages 20–29 years, 3.07 (2.97–3.2) in ages 30–39 years, 2.47 (2.43–2.52) in ages 40–49 years, 1.90 (1.87–1.94) in ages 50–59 years, 1.51 (1.49–1.54) in ages 60–69 years, and 1.20 (1.17–1.23) in ages ≥70 years, after adjusting for sex, current smoking status, regularity of exercise, income (lowest 20%), BMI, fasting blood glucose levels, presence of hypertension and CKD, and medication for dyslipidemia (use of statin and fibrate).

The current study, which consists of 8.4 million Korean adults in the NHIS health checkup data, demonstrates a robust relationship between remnant-C concentration and the risk of T2D. During a 9-year period, the incidence of T2D increased with increasing remnant-C concentration, showing a twofold increase in participants in the highest quartile compared with those in the lowest quartile. As remnant-C increased by 10 mg/dL, the risk of T2D increased by 13%, and this association was more pronounced in younger adults than in older adults. Moreover, the risk of developing T2D according to an increase in remnant-C concentration was higher in individuals with a low risk of developing T2D, such as women, those without IFG, hypertension, CKD, or metabolic syndrome, those without low HDL-C or high TG concentrations, and those who do not use statin or fibrate. These results suggest that remnant-C concentration is an independent risk factor for the development of T2D in the general population.

For many years, atherogenic dyslipidemia, characterized by elevated levels of TG, TG-rich lipoproteins, and decreased levels of HDL-C, has been widely studied in CVD (10,11,25,26). Currently, this lipid pattern is considered one of the important factors that explain the residual risk of major cardiovascular events (11). In atherogenic dyslipidemia, overproduction and inefficient lipolytic processing of TG-rich lipoproteins leads to increased formation of remnant-C (9), and this cholesterol content of TG-rich lipoproteins (remnant-C) is more likely to accumulate in the arterial wall and cause atherosclerosis and CVD (27). In addition, a growing body of evidence links atherogenic dyslipidemia with diabetes and overall glucose metabolism (7,8,28). Insulin resistance contributes to atherogenic dyslipidemia (5), and alterations in plasma cholesterol levels can contribute to ectopic lipid deposit in pancreatic islets, leading to β-cell dysfunction and loss of insulin secretion (29). Moreover, as HDL-C stimulates insulin secretion from pancreatic β cells and glucose uptake by skeletal muscle (30), a low concentration of HDL-C can contribute to the progression of prediabetes to T2D. Our study also demonstrated that elevated TG and decreased HDL-C were closely associated with increased risk of T2D (Table 2 and Supplementary Table 1). The HRs for T2D in the highest quartiles of remnant-C seem to not be better than those in the highest quartiles of TG and HDL-C. However, beyond the associations between fasting serum TG or HDL-C and diabetes, more recent evidence suggests that cholesterol within TG-rich remnant lipoproteins (remnant-C) more accurately reflects the pathophysiological processes that underlie progression to diabetes (31). The nature of remnant-C on new-onset T2D has been unclear, although it is one of the most prevalent lipid abnormality patterns in insulin-resistant states. However, considering the known harmful effect of remnant-C on vascular lesions and systemic inflammation, we hypothesized that remnant-C may influence the pancreatic β cells and overall glucose metabolism apart from TG or HDL-C concentrations. In this context, the current study investigated the relevance of remnant-C T2D risk and the view that cholesterol, and not the TGs within remnant lipoproteins, is causative for impaired glucose metabolism in the general population.

We found that the incidence of T2D gradually increased as remnant-C concentrations increased. This trend was consistently seen regardless of the presence or absence of abnormalities in other types of lipid levels (Supplementary Table 2). These findings are consistent with the results of previous studies showing remnant-C may be a risk factor for T2D (14–16). A longitudinal cohort study from Brazil reported that remnant-C particle diameter measured by nuclear magnetic resonance spectroscopy added predictive power to established risk factors for T2D in individuals without diabetes (14). In addition, a single-center cohort study from China showed that remnant-C has greater predictive power for new-onset T2D than that of conventional lipid parameters such as HDL-C, TG, and non–HDL-C (15). In renal transplant recipients in the cohort from Northern Netherlands, remnant-C levels were significantly associated with new-onset T2D independent of several recognized risk factors, including HDL-C and LDL-C (16). However, these studies were based on a small sample size (14,15) or a specific disease group (16). Our study, based on a large-scale nationwide population cohort with a relatively long-term, follow-up period, provides novel and robust insights into the association between remnant-C and the risk of T2D in the general population.

The underlying mechanisms of how increased remnant-C concentrations contribute to the development of T2D remain largely unclear. Generally, lipotoxicity has been considered the most reliable hypothesis, which may explain the association between cholesterol and β-cell dysfunction in diabetes (32). Although fatty acids are known to stimulate insulin secretion from the pancreatic β cells under normal physiological conditions (33), excessive fat contents of pancreatic islets eventually lead to loss of glucose-stimulated insulin secretion and decreased expression of GLUT-2 in pancreatic β cells (34). Obese individuals showed altered pancreatic energy metabolism with a substrate shift from glucose to fatty acids, blunt pancreatic blood flow, and consequently, β-cell dysfunction and diabetes (35). In addition, human pancreatic ectopic fat content was associated with unsaturated fatty acid enrichment (36). The current study also showed that participants with high levels of remnant-C have obesity-related metabolic derangements, such as higher BMI, waist circumference, blood pressure, fasting blood glucose, and lipid levels, than those with lower remnant-C levels. Notably, the impact of remnant-C on incident T2D remains significant, even after adjusting for various metabolic covariates that influence T2D, suggesting that remnant-C may play a considerable role in the pathogenesis of T2D, regardless of the traditional T2D risk factors.

Interestingly, the current study demonstrated that the association between remnant-C concentrations and the risk of T2D may vary depending on the individual’s characteristics. In subgroup analysis, we found that elevated remnant-C was associated with a greater T2D risk across all subgroups, and that the association between remnant-C levels and T2D risk appeared to be notable even in statin or fibrate-treated patients. However, the impact of remnant-C levels on the risk of T2D was more prominent in subgroups that are at relatively lower risk of T2D, such as younger age, women, and absence of comorbidities (e.g., IFG, hypertension, CKD, metabolic syndrome, and atherogenic dyslipidemia). In particular, we observed that the increased risk of T2D in the upper quartiles of remnant-C was remarkably attenuated in older adults, suggesting that the impact of remnant-C on the incident T2D may be reduced among those with conventional risk factors. This pattern was also observed in a previous study that investigated the association between particle size of triacylglycerol-enriched remnant lipoproteins and risk of T2D. The results showed that the association was attenuated in individuals with prediabetes and insulin resistance compared with those with normal glucose tolerance (14). These results imply that individuals with lower metabolic risk factors and increased levels of TG-enriched remnant lipoproteins are susceptible to exposure to diabetogenic particles or may have polymorphisms that increase their risk of T2D (37). These findings indicate that more attention should be paid to individuals who have high concentrations of remnant-C, even those with a lower risk of metabolic diseases.

This study has some limitations. First, remnant-C was estimated as non-fasting total cholesterol minus HDL-C, minus LDL-C in this study, as previously applied (10,11,15,16,25–27). In addition, methods of cholesterol measurement may vary by private health care institutions (38), which may overesimate or underestimate the value of remnant-C compared with direct measurement. However, calculated and measured remnant-C are closely correlated (39), and the European Atherosclerosis Society recently suggested using the measured one for assessing remnant-C levels in individuals (9). Moreover, the calculation of remnant-C is an accessible and affordable method in clinical practice.

Second, the NHIS-HEALS does not include hemoglobin A_1c and data from oral glucose tolerance tests. We could not collect all data related to T2D, such as educational level, dietary factors, and family history of diabetes. Thus, the potential for residual or unmeasured confounding factors remained, although we adjusted for multiple covariates that may influence T2D. Furthermore, due to lack of data on insulin resistance, we could not clarify the association between insulin resistance and remnant-C, which may be able to explain the mechanism about our findings.

Third, we obtained information on incident cases of diabetes and other confounders, including lifestyle factors, based on self-reporting or medical health records. Therefore, discrepancies between the ascertainment made by physicians in clinical practice and those recorded in claims data may have influence on the final results.

Fourth, our findings from the Korean population cannot be extrapolated to other populations. Moreover, calculated remnant-C used in this study was not fully validated in the Korean population.

Finally, the causal relationship between remnant-C and T2D could not be defined due to the intrinsic limitations of the observational study design. Despite these limitations, this study used a large-scale, well-validated, longitudinal, nationwide database that represents the entire Korean population.

In conclusion, we have demonstrated that elevated remnant-C is associated with an increased risk of T2D among 8.4 million Korean adults without T2D during a 9-year follow-up period. This result indicates that remnant-C profiles provide additional information in predicting the future progression of T2D, independent of TG or HDL-C. Furthermore, we demonstrated that an increase in T2D risk with an increase in remnant-C levels was more pronounced in individuals at low risk of T2D. These results suggest that the measurement of remnant-C in people without diabetes may be useful for T2D risk assessment in clinical practice. Further study is needed to explore the pathological role of remnant-C in the development of T2D.

Acknowledgments. The authors thank the staff at the Big Data Steering Department of the National Health Insurance Service and the subjects in this study.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions J.H.H., E.R., S.J.L., S.-H.I., K.-D.H., and J.G.K. contributed to the acquisition, analysis, and interpretation of the data. J.H.H., E.R., G.N.L., K.-D.H., and J.G.K. conceived and designed the study and performed the analyses. J.H.H., E.R., and J.G.K. wrote the first draft of the manuscript. K.-D.H. conducted the statistical analysis. All authors interpreted the data, contributed to the writing of the manuscript, and read and approved the final version. K.-D.H. and J.G.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.