PDF
Citation
Print
Highlights
• In early diabetes or with fewer drugs, higher LDL-C is linked to lower HF risk.
• In long-term diabetes or with multiple drugs or insulin, high lipid levels raise HF risk.
• The impact of dyslipidemia on HF risk changes with diabetes progression.
• Lipid targets should be adjusted based on the progression of diabetes.
INTRODUCTION
Interest in heart failure (HF) has increased significantly due to the robust clinical benefits of newer antidiabetic agents demonstrated in recent, large-scale, cardiovascular outcome trials [1]. Diabetes is associated with a 2- to 4-fold increased risk for HF, and its prevalence, associated hospitalization, and costs of care are increasing, in contrast to the decreasing trend of other cardiovascular diseases such as ischemic heart disease (IHD) or stroke [2-5]. Because patients with HF and diabetes experience worse outcomes than those without diabetes, understanding its pathophysiology and risk factors is important to predict, prevent, or mitigate the impact of this devastating disease [4].
Well-known risk factors for HF include old age, long duration of diabetes, IHD, hypertension, kidney dysfunction, obesity, and smoking [4,6]. Although several epidemiological studies have suggested that dyslipidemia is a risk factor for HF, it has demonstrated an inconsistent association with the development of HF [7-12]. Conflicting results have been reported, depending on the lipid parameters studied, whether the participants had IHD or diabetes, and whether lipid-lowering agents were used. Because a significant number of studies have proposed a “lipid paradox,” with an inverse relationship between circulating lipid levels and HF risk or prognosis in patients with established HF [13-16], it is important to delineate this relationship for the appropriate management and prevention of HF.
In this study, we aimed to investigate the relationship between circulating lipid levels and future HF risk among individuals with diabetes, and whether this association was modified by the duration of diabetes or treatment regimens, using data from nearly 2.5 million individuals in a large-scale, nationwide, population-based, cohort study.
METHODS
Data source
The present study used data from the Korean National Health Information Database (NHID), which combines data from the National Health Insurance Service (NHIS) collected for claims and reimbursements of healthcare services and general health examinations. The NHIS is a single insurer managed by the government for the entire South Korean population and provides comprehensive medical coverage. The NHID houses a complete set of health information, which includes a qualification database (e.g., age, sex, type of subscription, income level, and disability), birth and death database, a treatment database (e.g., principal and additional diagnosis encoded by the International Classification of Diseases, 10th Revision–Clinical Modification [ICD-10], hospitalization, treatment details, prescription details, and medical expenses), a health examination database (e.g., questionnaires addressing lifestyle and behavior, past medical history, family history, anthropometry, blood tests, chest radiography, and electrocardiography) and a medical care institution database (types of medical care institutions, location, equipment, and number of physicians). All enrolled individuals ≥20 years of age are mandated to undergo annual or biannual health examination. Details of the variables included in the NHID and health examinations have been described previously [17,18]. This study was approved by the Institutional Review Board of Yeouido St. Mary’s Hospital, College of Medicine, The Catholic University of Korea (No. SC23ZISE0054), and requirements for informed consent were waived due to the retrospective design of the study and the use of anonymized patient data.
Study population
In total, 2,613,026 subjects with new-onset or confirmed type 2 diabetes mellitus, who underwent a national health examination in 2015 to 2016, were identified. Individuals <20 years of age (n=322), those with a history of HF (n=55,729), missing data (n=73,193), and those who developed HF within 1 year of follow-up (n=43,804) were excluded. Ultimately, 2,439,978 subjects were included in the final analysis.
Measurements and definitions
Hospitals that performed health examinations were certified by the NHIS, and laboratory investigations were performed in accordance with the procedures of the Korean Association of Laboratory Quality Control. Body mass index (BMI) was calculated as weight divided by height squared (kg/m2). Obesity was defined as a BMI ≥25 kg/m2 in accordance with the Korean Society for the Study of Obesity criteria [19]. Blood tests for fasting glucose levels, lipid profiles (total cholesterol [TC], triglycerides, high-density lipoprotein cholesterol [HDL-C], and low-density lipoprotein cholesterol [LDL-C] levels), and serum creatinine were performed after an overnight fast. The estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease formula, as follows: 186×(serum creatinine)–1.154×age–0.203×0.742 (for females), and chronic kidney disease (CKD) was defined as an eGFR <60 mL/min/1.73 m2. Information regarding smoking history, alcohol consumption (heavy alcohol consumption, ≥30 g/day), and regular exercise (>30 minutes of moderate-intensity activity at least five times per week or >20 minutes of vigorous-intensity exercise at least three times per week) was obtained using a self-administered questionnaire. Household income levels were dichotomized to <20%. Hypertension was defined as at least 1 claim per year with ICD-10 code I10–I11 and a prescription for antihypertensive medication or systolic/diastolic blood pressure ≥140/90 mm Hg. Dyslipidemia was defined as at least 1 claim per year under ICD-10 code E78, and the prescription of lipid-lowering medication or a TC level ≥240 mg/dL. Type 2 diabetes mellitus was defined as at least 1 claim per year under ICD-10 code E11–E14, and a prescription for antidiabetic medication or a fasting blood glucose level ≥126 mg/dL. The study participants were categorized into four groups according to the duration of diabetes: new-onset (no previous recorded disease code or history of antidiabetic medication prescription but with a fasting glucose level ≥126 mg/dL at health examination), <5, 5–10, or ≥10 years. Subjects were also categorized into four groups according to the diabetes treatment modality: none, 1–2 oral hypoglycemic agents (OHAs), ≥3 OHAs, or insulin±OHA.
Study outcome
The primary outcome was incident HF, defined as recording of ICD-10 code I50 as the primary diagnosis during hospitalization which was retrieved from the claims database. Subjects were followed up until the end of 2020 or the incidence of HF. This was a longitudinal retrospective observational study with a median follow-up of 4.0 years (interquartile range, 3.3 to 4.3).
Statistical analyses
Baseline characteristics according to the duration of diabetes are expressed as mean±standard deviation (SD), median (interquartile range), or number (%). Subjects were classified into subgroups according to prespecified lipid levels: LDL-C <70 (reference group), and 70–99, 100–129, 130–159, and ≥160 mg/dL; TC <200 (reference group), and 200–240, and ≥240 mg/dL; non-HDL-C <100 (reference group) and 100–129, 130–159, 160–189, and ≥190 mg/dL; HDL-C <40 (reference group) and 40–49, 50–59, 60–69, and ≥70 mg/dL; triglycerides <100 (reference group) and 100–149, 150–199, and ≥200 mg/dL. Subjects were also classified according to the quartiles of TC/HDL-C ratio. The incidence rate of HF was calculated by dividing the number of cases by the total follow-up duration (per 1,000 person-years). The Cox proportional hazards model was used to estimate hazard ratio (HR) and corresponding 95% confidence interval (CI) for HF according to lipid level categories. The proportional hazards assumption was assessed using Schoenfeld residuals test with a logarithm of the cumulative hazard functions based on Kaplan-Meier estimates. There was no significant departure from proportionality to hazards over time. Model 1 was unadjusted; model 2 was adjusted for age and sex; model 3 was further adjusted for BMI, smoking, alcohol consumption, income, hypertension, CKD, and fasting glucose level; and model 4 was further adjusted for the use of lipid-lowering medications, insulin treatment, and OHAs. Because a history of cardiovascular disease may affect the risk for the development of HF, a sensitivity analysis was performed excluding subjects with a history of myocardial infarction/IHD (ICD-10 codes I20–I25), and stroke (ICD-10 codes I63, I64). Subgroup analyses were performed based on age, sex, obesity, use of sodium glucose cotransporter 2 (SGLT2) inhibitors, and the use of lipid-lowering medications during follow-up. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA), and differences with P<0.05 were considered to be statistically significant.
RESULTS
Baseline characteristics of the subjects according to diabetes duration are summarized in Table 1. Subjects with a longer duration of diabetes were older, more likely to be female, have a lower body weight and BMI, a favorable lifestyle with a lower frequency of current smoking and alcohol consumption, and a higher likelihood of regular exercise. The prevalence of dyslipidemia and the use of lipid-lowering medications were similar, except in the new-onset diabetes group; however, lipid levels were lower in subjects with a longer duration of diabetes, possibly due to more intensive treatment. Comorbid conditions, such as hypertension and CKD, and use of antidiabetic medications, were more prevalent in subjects with a longer duration of diabetes.
The relationship between baseline LDL-C levels and the risk for incident HF was examined in different groups of patients with progressive diabetes (Table 2). Basically, there was a stepwise increase in the incidence rate of HF according to the duration of diabetes. In subjects with new-onset diabetes or a diabetes duration <5 years, an inverse relationship between lipid levels and HF risk was observed. Using an LDL-C level <70 mg/dL as the reference, all other groups with higher LDL-C levels exhibited a significantly lower risk for HF. In the new-onset diabetes group, the risk was approximately 25% lower in those with LDL-C levels 100–129, 130–159, and ≥160 mg/dL, compared to those with levels <70 mg/dL, with the lowest HR in the 130–159 mg/dL group (HR, 0.71; 95% CI, 0.68 to 0.74). In the group with diabetes duration <5 years, the risk was approximately 10% lower in those with LDL-C levels 100–129, 130–159, and ≥160 mg/dL compared to those with levels <70 mg/dL, with the lowest HR in the 130–159 mg/dL group (HR, 0.86; 95% CI, 0.83 to 0.89). However, a J-shaped association was observed in the long-standing diabetes group. In the group with diabetes duration 5 to 10 years, the risk for HF was approximately 5% lower in those with LDL-C level 70–99, 100– 129, and 130–159 mg/dL, but significantly increased in those with ≥160 mg/dL (HR, 1.07; 95% CI, 1.02 to 1.13), compared to those with LDL-C <70 mg/dL. In the group with diabetes duration ≥10 years, the risk for HF was approximately 5% lower in those with LDL-C level 70–99 and 100–129 mg/dL, but significantly increased in those with 130–159 mg/dL (HR, 1.03; 95% CI, 1.00 to 1.07) and ≥160 mg/dL (HR, 1.16; 95% CI, 1.11 to 1.21), compared to those with LDL-C <70 mg/dL (Fig. 1A). This observation suggests different associations between LDL-C levels and the risk for HF, depending on the duration of diabetes.
The relationship between other lipid parameters, such as non-HDL-C, TC, HDL-C, triglycerides, or TC/HDL-C ratio and the risk for incident HF was examined. Similar patterns were noted in the association between non-HDL-C and the risk for HF, with an inverse association in the new-onset and short-duration groups, and a J-shaped association in the longer diabetes duration groups (Table 3). In addition, individuals with low TC level or TC/HDL-C ratio in the new-onset diabetes group and those with high TC or TC/HDL-C ratio in the longer diabetes duration groups exhibited a higher risk for HF (Supplementary Tables 1 and 2). However, there was a stepwise increase in the risk of HF according to the triglyceride levels independent of the disease duration except new-onset diabetes (Supplementary Table 3). The risk was approximately 10% lower in those with HDL-C levels 40 to 69 mg/dL compared to those with levels <40 mg/dL showing no significant differences in the relationship between HDL-C levels and the risk of HF according to the duration of diabetes (Supplementary Table 4).
Using the number of antidiabetic medications instead of diabetes duration to categorize diabetes status also resulted in similar patterns. In groups using none or 1–2 OHAs, a high LDL-C level was associated with a lower risk for HF. However, a J-shaped association was evident in the groups using ≥3 OHAs or insulin±OHA (Table 4, Fig. 1B).
Sensitivity analysis excluding subjects with a history of cardiovascular disease did not alter the results (Supplementary Table 5). In the subgroup analyses, the patterns of association between LDL-C and HF risk were generally similar. In new-onset diabetes, the inverse relationship was stronger in groups with younger age, male sex, non-obesity, and without dyslipidemia medication (P<0.001). In the group with diabetes duration ≥10 years, a J-shaped association was more prominent in groups with younger age, and with dyslipidemia medication (P for interaction <0.001) (Supplementary Tables 6-9). In subjects without SGLT2 inhibitors use, the result was similar to the original analysis. In subjects with SGLT2 inhibitor use, no statistically significant difference in the risk of HF was noted according to LDL-C levels, possibly due to low event number (Supplementary Table 10).
DISCUSSION
Using a nationwide database of almost 2.5 million individuals with diabetes, we demonstrated that the relationship between lipid parameters and the risk for HF varied according to the progression of diabetes. Among individuals with new-onset or short diabetes duration and using fewer OHAs, higher LDL-C or non-HDL-C levels were associated with a lower risk for HF. However, in individuals with long-standing diabetes and using multiple OHAs or insulin, J-shaped associations were noted, with high lipid levels being a risk factor for HF. These findings were similar and independent of the use of lipid-lowering medications.
Although several large-scale cohort studies have examined this issue, the relationship between blood lipid levels and the risk for HF remains controversial. An analysis from the Framingham Heart Study revealed that baseline non-HDL-C and HDL-C concentrations carried hazards for HF of 1.19 (95% CI, 1.11 to 1.27) and 0.82 (95% CI, 0.75 to 0.90) per SD increment, respectively [10]. The Apolipoprotein MOrtality RISk (AMORIS) study explored the influence of lipoprotein components on the development of congestive HF in 84,740 healthy Swedish men and women. Apolipoprotein B, apolipoprotein B/A1 ratio, LDL-C, triglyceride, non-HDL-C, TC, and TC/HDL-C ratio were positively associated with HF risk, while apolipoprotein A1 and HDL-C were negatively associated with HF risk in the general population [11]. Remnant cholesterol, encompassing the cholesterol content of very LDL, chylomicrons, and intermediate-density lipoprotein, was calculated by subtracting the HDL and LDL values from the TC value. An analysis of data from the UK Biobank revealed that remnant cholesterol was significantly associated with HF risk in patients with diabetes, independent of LDL-C levels [20]. In another long-term follow-up study of the Copenhagen General Population Study and the Copenhagen City Heart Study, a stepwise association between higher nonfasting triglyceride concentrations and a higher risk for HF was demonstrated. However, there was no relationship between LDL-C levels and the risk for HF, and even an inverse association in a subgroup of individuals without IHD [7]. From four large cohorts in the United States (Atherosclerosis Risk in Communities Study, Coronary Artery Risk Development in Young Adults Study, Framingham Heart Study Offspring Cohort, and Multi-Ethnic Study of Atherosclerosis [MESA]), it was clearly shown that cumulative exposure to LDL-C, time-weighted average LDL-C, and the LDL-C slope change during young adulthood and middle age were not associated with incident HF [12]. In the MESA cohort, low HDL-C and high triglyceride levels were associated with incident HF in patients with diabetes but not in those without diabetes. LDL-C data were not reported in the MESA cohort [9]. Collectively, these findings are not identical, and the evaluation of LDL-C concentration as a risk factor for HF is scarce, especially in patients with diabetes.
Whether low LDL-C levels or statin treatment are beneficial in patients with established HF is another controversial subject. Several reports have suggested that low cholesterol is associated with worse prognosis in patients with HF, suggesting the existence of a “lipid paradox” [21]. High TC levels independently predicted higher survival in chronic HF and lower in-hospital mortality in patients with acute decompensated HF, after adjusting for multiple prognostic factors [13,15,16]. Low remnant cholesterol levels were also associated with an increased risk for all-cause mortality in patients with HF [14]. In a single-center study, high TC levels predicted worse outcomes in patients with underlying coronary artery disease, but better outcomes in patients without coronary artery disease [22]. Low LDL-C level (<100 mg/dL) was independently associated with increased mortality risk in patients with HF without diabetes, but no association was detected in patients with HF with diabetes [23]. In a study using data from the Korean NHID, a U-shaped association between LDL-C and HF mortality was found, with an optimal range of 130 to 159 mg/dL in statin non-users [24]. Potential mechanisms have been proposed for this counterintuitive phenomenon. Low lipid levels may reflect malnutrition and cachexia or may simply reflect advanced disease status and poor prognosis. In addition, low cholesterol levels may cause or reflect systemic inflammation, which predicts poor prognosis in HF [21]. In contrast, a prospective cohort study involving patients with chronic HF and IHD in Japan revealed that statin use, particularly high-dose statins, was beneficial in preventing all-cause death and HF-related admission [25].
In this study, we showed that high triglycerides and low HDL-C levels were associated with higher risk of HF confirming previous reports [6,8-10], and this was independent of diabetes progression status. However, the association between LDL-C or non-HDL-C and the risk of HF was modified by diabetes progression status. An explanation for the different associations between LDL-C levels and HF risk according to diabetes progression status remains unclear. The pathophysiology of HF in patients with diabetes is complex and encompasses both macro- or microvasculopathy, and vasculopathy-independent myocardial dysfunction. Inflammation, endothelial dysfunction, oxidative stress, impaired mitochondrial energetics, glycated proteins and lipids, and neurohormonal activation contribute to ischemic and diabetic cardiomyopathies by causing functional and anatomical abnormalities [2,26]. Moreover, the role and function of cholesterol in the myocardium and pathogenesis of HF are largely unknown. Cholesterol catabolism and the accumulation of intracellular bile acids promote inflammation in cardiomyocytes, leading to progression of HF [27]. LDL-C has been shown to bind lipopolysaccharide (LPS) and to protect against the immediate toxic effects of LPS on endothelial cells [28]. Lipoproteins’ protective role in HF may stem from an ability to diminish the LPS-induced elaboration of cytokines detrimental to the heart, such as tumor necrosis factor (TNF). TNF may contribute to HF progression and cardiac injury via its pro-apoptotic and negative inotropic effects [29]. A multidirectional Mendelian randomization study performed in Denmark reported that different patterns of remnant cholesterol and LDL-C cause atherosclerosis and progression to HF. Elevated remnant cholesterol levels are associated with IHD and inflammation. In contrast, elevated LDL-C is causally associated with IHD without inflammation, indicating a lower risk for HF [30]. In patients with long-standing diabetes, high lipid levels were associated with an increased risk of HF. There may be a difference in the causes of HF depending on the diabetes duration or severity of diabetes, although this could not be determined in this study due to the lack of HF subtypes recognition. As diabetes progress, the increased risk of HF seems to be more related to ischemia or coronary artery diseases [31]. Collectively, it is conceivable that different degrees of hyperglycemia, insulin resistance, inflammation, and their interactions with cholesterol at different stages of diabetes led to our findings [26]; however, this requires further clarification.
An important factor modifying the association between lipid levels and HF risk may be the presence or development of coronary heart disease (CHD). A Mendelian randomization study using genome-wide association study data from the UK Biobank provided genetic evidence supporting the relationship between dyslipidemia and a higher risk for HF, which was mainly due to the increased risk for CHD [32]. In patients with CHD or acute coronary syndrome, intensive lipid management with statins reduces the risk for incident HF [33,34]. A meta-analysis of major primary and secondary prevention statin trials demonstrated a modest reduction in the risk for nonfatal HF hospitalizations and a composite of nonfatal HF hospitalizations and HF death, although no significant associations were noted at the individual trial level [35]. In contrast, no such relationship between LDL-C levels and the risk for HF was observed in the general population, and even an inverse association was observed in a subgroup of individuals without IHD [7].
To the best of our knowledge, this was the largest study to examine the association between lipid levels and risk for HF in patients with diabetes. However, the present investigation had some limitations. First, because this was an observational study, causal relationships could not be confirmed. To minimize the possibility of reverse causality, patients with a history of HF or a new event during the first year of follow-up were excluded. Second, data regarding glycosylated hemoglobin levels or postload glucose levels were not available in the database. Therefore, the diagnosis of new-onset diabetes could be underestimated. We also subgrouped subjects based on diabetes duration and number of medications but could not integrate glucose control status. Third, because new-onset diabetes group includes people with relatively high glucose levels due to incidental diagnosis during health examination, lipid levels might have been influenced by hyperglycemia and associated conditions. Fourth, initiation of lipid-lowering therapy during the follow-up period may have affected our results. To avoid this effect, we performed a subgroup analysis based on the use of lipid-lowering agents during follow-up and found similar patterns in both groups. Finally, because Korea has a unique healthcare system, our results may not be applicable to other countries or ethnic populations, and further validation is warranted.
In conclusion, results of our study suggest a differential role of dyslipidemia in the development of HF depending on the progression status of diabetes, as defined by disease duration or antidiabetic medication usage. Higher lipid levels were associated with a lower risk of HF in individuals in the early stages of diabetes. In contrast, high lipid levels are a risk factor for HF in individuals with long-standing diabetes. These findings were generally conserved in the sensitivity and subgroup analyses. Further studies investigating the causality and mechanisms of these associations will guide the development of appropriate strategies for the prevention of HF.
XML Download