Association of Lipids With Incident Heart Failure Among Adults With and Without Diabetes Mellitus: Multiethnic Study of Atherosclerosis (original) (raw)

. Author manuscript; available in PMC: 2014 May 1.

Abstract

Background

Dyslipidemia is a known risk factor for coronary disease, but its role in heart failure (HF) development is less well-defined.

Methods and Results

We included 5688 participants, aged 45 to 84 years, without clinical cardiovascular disease, and not receiving lipid-lowering medications at baseline, from the Multiethnic Study of Atherosclerosis. Cox-proportional hazards models were used to evaluate associations of triglyceride, total cholesterol/high-density lipoprotein–cholesterol (HDL-C) ratio, HDL-C, and non HDL-C with incident HF. We investigated for effect-modification by diabetes mellitus status and sex. During a median follow-up of 8.5 years, there were 152 incident HF cases. There were no interactions by sex. We observed significant interactions between triglyceride and diabetes mellitus (_P_interaction<0.05). We stratified our analyses by diabetes mellitus status. In participants with diabetes, the hazard ratios were 2.03 (0.97–4.27) and 1.68 (1.18–2.38) for high triglyceride and log of triglyceride, respectively, after adjusting for confounders, comorbidities, and diabetes mellitus severity/treatment. The association of high triglyceride with incident HF was attenuated by interim myocardial infarction. The hazard ratios were greatest in participants with diabetes who also had high triglyceride, low HDL-C, or high total cholesterol/HDL-C ratio (3.59 [2.03–6.33], 3.62 [2.06–6.36], and 3.54 [1.87–6.70], respectively). Lipid measures were not associated with incident HF in individuals without diabetes.

Conclusions

The risk of incident HF is greater in individuals with diabetes mellitus who also have high triglyceride, low HDL-C, or high total cholesterol/HDL-C ratio. The association of high triglyceride with incident HF is partly mediated by myocardial infarction.

Keywords: diabetes mellitus, heart failure, lipids

Dyslipidemia is a known risk factor for coronary disease,1 but its role in heart failure (HF) development is less well-defined. Evidence has implicated high levels of free fatty acids (FFAs)2,3 and triglycerides3 in cardiotoxicity, and elevated levels of lipid fractions may be involved in cardiac remodeling,4 which is a known determinant of the clinical course of HF.5 However, reports from observational studies on the associations of triglycerides and total cholesterol/high-density lipoprotein–cholesterol (TC/HDL-C) ratio with incident HF have been inconsistent.4

In a previous study in the Multiethnic Study of Atherosclerosis (MESA), neither high triglyceride nor low HDL-C were significant predictors of incident HF.6 In other studies, low HDL-C (but not high triglyceride) was an independent predictor of HF.7,8 In the Physicians Health Study, neither HDL-C nor TC/HDL-C ratio were independently associated with incident HF.4 In the Framingham Heart Study, elevated levels of non HDL-C and decreased levels of HDL-C,9 and a high TC/HDL-C ratio10 were associated with increased risks of incident HF. Despite the mixed evidence, current guidelines consider hyperlipidemia as a risk factor for HF.4

Diabetes mellitus predisposes to HF,10 and diabetic cardiomyopathy may partially result from altered substrate metabolism attributable to lipid overstorage and lipotoxic injury to cardiomyocytes.11–14 The coexistence of hyperglycemia and increased circulating FFAs accelerates the progression to cellular dysfunction15,16 and may further increase the risk of incident HF. Our objective was to evaluate the associations of lipid fractions with incident HF in participants with and without diabetes.

Methods

Study Population

MESA is a population-based cohort comprising 6814 men and women of Caucasian (38%), African American (28%), Hispanic (22%), and Chinese origin (12%), aged 45 to 84 years (2000–2002) and without known clinical cardiovascular disease (CVD) at baseline. Participants were recruited from 6 regions in the United States. Details of MESA design and objectives have been published.17 The protocol was approved by the institutional review board of participating sites, and informed consent was obtained from participants. This cohort study is based on baseline data and HF data collected/adjudicated and available in October 2012 from participants who were not taking lipid-lowering medications at baseline. Eight hundred and thirty-seven participants were lost to follow-up. Participants without lipid and glucose measures at baseline and those for whom no follow-up was completed were excluded.

Baseline Measurements

Standardized questionnaires were used to collect information on educational status, cigarette smoking, hypertension, diabetes mellitus, and medications. The MESA typical week physical activity survey was used to record time and frequency spent on intentional exercise, such as walking for exercise, sports/dancing, and conditioning activities.18 The total minutes spent on each activity per week was multiplied by its metabolic equivalent level and summed (metabolic equivalent-minutes/week). Body mass index was calculated as weight divided by the square of height (kg/m2) and used as an indicator of adiposity. Hypertension was defined as systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg or use of antihypertensive medications.

Serum glucose was measured by the glucose-oxidase method.1 Spot urine albumin and creatinine were measured using the nephelometry and Jaffe reaction, respectively.19 Urinary albumin/creatinine ratios were calculated, and participants were classified as normal (<30 mg/g), having macroalbuminuria (>300 mg/g), or microalbuminuria (30–300 mg/g). Interleukin-6 (IL-6) was measured using an ultrasensitive ELISA with a coefficient of variation of 6.3%20 and used as an indicator of inflammation. Insulin resistance (indicated by homeostasis model assessment) was calculated as fasting glucose (mmol/L)×fasting insulin (µU/mL)/22.5.6 Resting 12-lead electrocardiograms obtained from fasting participants were centrally read and coded for the presence of left ventricular hypertrophy using the Minnesota coding system.

Assessment of Diabetic Status and Severity

Participants were categorized as having diabetes mellitus if their fasting blood glucose was ≥126 mg/dL or they were receiving treatment for diabetes mellitus, and not having diabetes mellitus if their fasting blood glucose was <126 mg/dL and they were not receiving treatment for diabetes mellitus. Diabetes mellitus was classified as more severe if participants required insulin for treatment, less severe if participants required oral hypoglycemic agents only, and least severe if participants did not require insulin or oral agents for blood glucose control.

Measurement of Lipid Fractions

Lipid measures were obtained on fasting samples. Triglyceride was measured in plasma using a glycerol-blanked enzymatic method.1 Hypertriglyceridemia was present if plasma triglyceride was ≥150 mg/dL. HDL-C was measured in plasma using a cholesterol-oxidase reaction after precipitation of non HDL-C with magnesium/dextran.1 Low HDL-C was present if plasma HDL-C was <40 mg/dL in men or <50 mg/dL in women. Plasma total cholesterol was measured by the cholesterol-oxidase method.1 TC/HDL-C ratio was calculated by dividing total cholesterol by HDL-C. High TC/HDL-C ratio was present if TC/HDL-C ratio was ≥5.0.21 Non HDL-C was calculated by subtracting HDL-C from total cholesterol.

Follow-Up and Incident HF Definition

The median follow-up was 8.5 years (interquartile range, 0.97) with a total of 43 753 person-years of observation. Each participant or their next of kin was contacted by a telephone interviewer at 6- to 9-month intervals to inquire about interim hospitalizations, outpatient diagnoses, and deaths caused by cardiovascular causes.6 Records were obtained on ≈99% of hospitalized cardiovascular encounters and some information on 97% of outpatient diagnostic encounters. Hospital records were abstracted and reviewed by paired physicians for independent end point classification and assignment of incidence dates.6 In cases of disagreements, the reviewing pair adjudicated differences, but if disagreements persisted, the full morbidity and mortality classification committee made the final decision.

The end point for our study was symptomatic HF. Multiple HF events in the same participant were considered once and the time to first occurrence was used. End point criteria for HF in MESA included the following: (1) physician-diagnosed HF and medical therapy for HF; (2) pulmonary edema/congestion on chest radiography; or (3) dilated ventricle or poor left ventricular function on echocardiography or ventriculography, or evidence of left ventricular diastolic dysfunction.6,22 Participants not meeting any criteria, including those with a physician diagnosis only, without any other evidence were classified as not having HF. Myocardial infarction (MI) is a potent risk factor for HF23,24 and was defined according to standard criteria as consisting of symptoms, electrocardiographic findings, and cardiac biomarker levels.17,22

Statistical Analysis

Data are presented according to HF status using means±SDs or median (interquartile range) for continuous variables and percentage for discrete variables. Logarithmic transformation was performed for triglyceride, IL-6, and homeostasis model assessment of insulin resistance because of skewness. Intentional exercise was non-normally distributed and was divided into quartiles. Comparisons between HF groups were tested using χ2 test (discrete variables), 2-sample t test (normally distributed continuous variables), and Mann–Whitney test (non-normally distributed continuous variables). Kaplan–Meier plots for incident HF are presented according to lipid and diabetes mellitus categories, and tested with the log-rank test. Participants were censored if they were lost to follow-up or failed to experience HF at the end of follow-up.

We assessed the associations between categorical and continuous measures of lipid fractions and incident HF using separate Cox-proportional hazards models. We checked for interactions of lipid fractions with sex and diabetes mellitus at the 0.05 significance level. We observed significant interactions between triglyceride and diabetes mellitus. We performed our analyses separately for participants with and without diabetes. We sequentially adjusted for confounders and comorbidities known to be associated with HF6,23–26 as follows, model 1: unadjusted analysis; model 2: adjusted for confounders, including age, sex, ethnicity, educational status, cigarette smoking, intentional exercise, and center; model 3: model 2, adjusted for comorbidities, including adiposity, hypertension, left ventricular hypertrophy (by ECG), kidney dysfunction (indicated by albuminuria), inflammation, and insulin resistance. In MESA, IL-6 was the inflammatory marker with the strongest prediction for incident HF.6 Forty-five participants with incident HF also experienced incident MI, therefore we also adjusted for interim MI (model 4). In participants with diabetes, we also adjusted for diabetes mellitus treatment/severity (model 5). Adjustment for HDL-C attenuates the relationship between triglycerides and CVD,27,28 and we added HDL-C to models that contained triglyceride. Ethnicity was not defined a priori as a unit of analysis. Because of our modest number of HF events, there was limited power for ethnic-specific analysis.

We calculated hazard ratios (HRs) per SD greater value for continuous variables, and for high triglyceride, low HDL-C, and high TC/HDL-C ratio categories. We calculated unadjusted and multivariable-adjusted HRs of incident HF, according to lipid and diabetes mellitus categories. To maximize statistical power, only participants with missing data on a variable needed for a particular model were excluded from analyses.6 We checked for proportionality of hazards by visually examining the log-log plots. Two-sided P values of <0.05 were considered significant. Statistical analysis was performed using SAS enterprise guide version 4.3 (SAS Institute, Inc, Cary, NC).

Results

Diagrammatic presentations of our sample size are shown in Figure 1. Our total sample size was 5688 and consisted of 616 and 5072 participants with and without diabetes mellitus, respectively. There were 152 cases (48 in diabetic and 104 in participants without diabetes) of incident HF. The overall HF incidence was 3.47 per 1000 person-years, with incidences of 11.0 per 1000, and 2.6 per 1000 person-years in participants with and without diabetes, respectively.

Figure 1.

Figure 1

Diagrammatic presentation of our sample size for participants with and without diabetes. MESA indicates Multiethnic Study of Atherosclerosis.

Baseline characteristics of participants are presented according to HF occurrence during follow-up (Table 1). HF cases were more commonly men, older, African American, past or current smokers, and exercised less frequently than noncases. Hypertension, left ventricular hypertrophy, diabetes mellitus, and kidney abnormalities were more prevalent, and body mass index, homeostasis model assessment of insulin resistance, and IL-6 levels were higher in HF cases than non-cases. HDL-C levels were lower, whereas triglyceride and TC/HDL-C ratios were higher in HF cases. Kaplan–Meier plots of incident HF are presented according to lipid and diabetes mellitus categories (Figures 2–4). The HF free probability was significantly less in participants with both diabetes mellitus and lipid abnormalities, intermediate in those with diabetes mellitus alone, but was similar in those without diabetes mellitus, irrespective of their lipid status.

Table 1.

Characteristics of MESA Participants Who Were Not Taking Lipid-Lowering Medications at Baseline (2000–2002) According to Incident Heart Failure Status

Incident Heart Failure
Characteristics Cases (n=152) Noncases (n=5536) P Value
Age, y 69.3±8.5 61.2±10.3 <0.0001
Male sex, % 60.5 46.8 0.0008
Ethnicity 0.057
White, % 38.8 37.5
Chinese American, % 5.9 12.2
Black, % 34.2 27.5
Hispanic, % 21.1 22.8
>High school education, % 58.6 64 0.17
Cigarette smoking 0.046
Never, % 40.8 50.8
Former, % 44.1 35.6
Current, % 15.1 13.7
Total intentional exercise, median (IQR), MET-min/wk 525 (1282) 840 (1980) 0.004
Left ventricular hypertrophy by ECG, % 7.3 0.7 <0.0001
Hypertension, % 72.4 40.5 <0.0001
Diabetes mellitus, % 31.6 10.3 <0.0001
Body mass index, kg/m2 29.6±6.1 28.2±5.5 0.002
HDL-C, mg/dL 47.9±14.4 51.2±15.0 0.007
Triglycerides, median (IQR), mg/dL 117 (102.5) 109 (81.0) 0.09
Total cholesterol/HDL-C ratio, mg/dL 4.31 (1.40) 4.11 (1.27) 0.05
Non HDL-C, mg/dL 143.2±35.8 145.3±36.1 0.47
Interleukin-6, pg/mL* 1.66±1.95 1.22±1.95 <0.0001
HOMA-IR* 2.36±1.91 1.94±1.86 0.0001
Urine albumin/creatinine ratio <0.0001
Normal (<30 mg/dL) 70.0 91.5
Microalbuminuria (30–300 mg/dL) 24.7 7.4
Macroalbuminuria (>300 mg/dL) 5.3 1.1

Figure 2.

Figure 2

Heart failure free probability in Multiethnic Study of Atherosclerosis according to triglyceride and diabetes mellitus category. There were 261, 355, 1366, and 3706 participants in categories 1, 2, 3, and 4, respectively. High triglyceride refers to triglyceride ≥150 mg/dL.

Figure 4.

Figure 4

Heart failure free probability in Multiethnic Study of Atherosclerosis (MESA) according to total cholesterol (TC)/high-density lipoprotein–cholesterol (HDL-C) ratio and diabetes mellitus category. There were 194, 422, 1068, and 4004 participants in categories 1, 2, 3, and 4, respectively. High TC/HDL-C ratio refers to TC/HDL-C ratio ≥5.

There were no differences in the associations of lipid fractions by sex (_P_interaction>0.05). High triglyceride and log of triglyceride were associated with increased risks of incident HF in diabetic, but not in patients without diabetes (_P_interaction of 0.02 and 0.04, respectively). The interaction terms of low HDL-C, HDL-C, high TC/HDL-C ratio, TC/HDL-C ratio, and non HDL-C with diabetes mellitus were not statistically significant (_P_interaction of 0.07, 0.21, 0.12, 0.15, and 0.19, respectively). We presented our results separately for participants with and without diabetes.

Our HRs for participants with diabetes are presented in Table 2. Low HDL-C was significantly associated with incident HF in participants with diabetes after adjusting for confounders, comorbidities (including interim MI), and diabetes mellitus treatment/severity (models 2–5). The association of high triglyceride with incident HF was attenuated when we included interim MI (models 4–5). The HRs became further attenuated when we added HDL-C to models (model 5) that included high triglyceride and log of triglyceride (1.63 [0.75–3.56] and 1.52 [1.01–2.31], respectively). Our point estimates seemed similar when homeostasis model assessment of insulin resistance was excluded from our models (Table in the online-only Data Supplement). When we substituted C-reactive protein for IL-6, glomerular filtration rate for albuminuria, waist circumference for body mass index, and the composite of systolic and diastolic blood pressure, and antihypertensive medication use for hypertension (Table in the online-only Data Supplement), our point estimates did not change much either.

Table 2.

Associations of Lipid Fractions With Incident Heart Failure in MESA Participants With Diabetes Who Were Not Taking Lipid-Lowering Medications at Baseline (2000–2002)

Model 1 Model 2 Model 3 Model 4 Model 5
Lipid Measure HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
Categorical
High TG 2.11 (1.19–3.77) 2.32 (1.25–4.31) 2.56 (1.30–5.05) 1.95 (0.94–4.04) 2.03 (0.97–4.27)
Low HDL-C 1.75 (0.97–3.15) 1.93 (1.05–3.57) 2.26 (1.15–4.46) 2.40 (1.17–4.93) 2.50 (1.20–5.22)
High TC/HDL-C ratio 1.61 (0.91–2.86) 1.65 (0.90–3.01) 1.66 (0.86–3.20) 1.56 (0.77–3.14) 1.65 (0.80–3.37)
Continuous
Log of TG 1.35 (1.04–1.76) 1.44 (1.09–1.90) 1.58 (1.16–2.14) 1.58 (1.12–2.23) 1.68 (1.18–2.38)
HDL-C 0.74 (0.53–1.02) 0.72 (0.51–1.01) 0.66 (0.45–0.96) 0.67 (0.44–1.00) 0.65 (0.43–0.99)
TC/HDL-C ratio 1.20 (1.00–1.43) 1.25 (1.01–1.54) 1.27 (1.00–1.61) 1.22 (0.90–1.65) 1.26 (0.92–1.72)
Non HDL-C 1.07(0.81–1.40) 1.10 (0.83–1.47) 1.06 (0.78–1.43) 0.90 (0.62–1.30) 0.92 (0.64–1.34)

Our HRs for participants without diabetes are presented in Table 3. In participants without diabetes, there were no significant associations between lipid fractions and incident HF. HRs are also presented according to lipid and diabetes mellitus category in Table 4. The HRs of incident HF were greatest in participants with diabetes, who also had high triglyceride, low HDL-C, or high TC/HDL-C ratio. The percentage of missing variables was <3% in all models. Consequently, sample sizes may have varied slightly between the models.

Table 3.

Associations of Lipid Fractions With Incident Heart Failure in MESA Participants Without Diabetes Who Were Not Taking Lipid-Lowering Medications at Baseline (2000–2002)

Model 1 Model 2 Model 3 Model 4
Lipid Measure HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
Categorical
High triglyceride 0.92 (0.59–1.43) 1.02 (0.65–1.61) 0.96 (0.60–1.55) 0.95 (0.58–1.56)
Low HDL-C 0.89 (0.59–1.34) 1.02 (0.67–1.55) 0.97 (0.62–1.53) 0.81 (0.51–1.29)
High TC/HDL-C ratio 0.90 (0.55–1.46) 0.88 (0.53–1.46) 0.89 (0.53–1.51) 0.97 (0.57–1.65)
Continuous
Log of triglyceride 0.96 (0.79–1.16) 1.03 (0.83–1.27) 1.02 (0.81–1.29) 0.99 (0.79–1.25)
HDL-C 0.91 (0.74–1.11) 0.95 (0.76–1.19) 1.00 (0.78–1.29) 1.11 (0.86–1.43)
TC/HDL-C ratio 0.98 (0.80–1.19) 1.00 (0.81–1.24) 1.02 (0.80–1.28) 0.98 (0.76–1.27)
Non HDL-C 0.86 (0.71–1.05) 0.95 (0.77–1.17) 1.01 (0.81–1.25) 1.01 (0.80–1.28)

Table 4.

Hazard Ratios of Incident Heart Failure According to Lipid and Diabetes Mellitus Category

Model 1 Model 2 Model 3 Model 4
Category No. of Sample, % HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
Triglyceride and diabetes mellitus 5688 (100)
High triglyceride and diabetes mellitus 261 (4.6) 5.85 (3.82–8.96) 5.17 (3.31–8.07) 4.49 (2.60–7.74) 3.59 (2.03–6.33)
Normal triglyceride and diabetes mellitus 355 (6.2) 2.77 (1.68–4.57) 2.16 (1.30–3.61) 1.77 (0.99–3.18) 2.18 (1.21–3.90)
High triglyceride, no diabetes mellitus 1366 (24.0) 0.92 (0.59–1.43) 1.01 (0.64–1.58) 0.97 (0.60–1.55) 1.02 (0.63–1.65)
Normal triglyceride, no diabetes mellitus (reference) 3706 (65.2) 1.00 1.00 1.00 1.00
HDL-C and diabetes mellitus 5688 (100)
Low HDL-C and diabetes mellitus 318 (5.60) 5.08 (3.33–7.75) 4.43 (2.85–6.87) 3.89 (2.26–6.69) 3.62 (2.06–6.36)
Normal HDL-C and diabetes mellitus 298 (5.2) 2.91 (1.71–4.94) 2.30 (1.34–3.94) 1.83 (0.99–3.35) 1.66 (0.89–3.08)
Low HDL-C, no diabetes mellitus 1738 (30.6) 0.90 (0.60–1.36) 1.00 (0.66–1.52) 0.94 (0.61–1.47) 0.80 (0.51–1.27)
Normal HDL-C, no diabetes mellitus (reference) 3334 (58.6) 1.00 1.00 1.00 1.00
TC/HDL-C ratio and diabetes mellitus 5688 (100)
High TC/HDL-C ratio and diabetes mellitus 194 (3.4) 5.53 (3.40–9.01) 4.46 (2.67–7.43) 3.82 (2.02–7.05) 3.54 (1.87–6.70)
Normal TC/HDL-C ratio and diabetes mellitus 422 (7.4) 3.42 (2.23–5.25) 2.72 (1.75–4.23) 2.34 (1.41–3.89) 2.51 (1.51–4.15)
High TC/HDL-C ratio, no diabetes mellitus 1068 (18.8) 0.90 (0.55–1.46) 0.88 (0.53–1.44) 0.92 (0.55–1.54) 1.03 (0.61–1.72)
Normal TC/HDL-C ratio, no diabetes mellitus (reference) 4004 (70.4) 1.00 1.00 1.00 1.00

Discussion

Our study has shown greater risks of incident HF in individuals with diabetes, who also have high triglyceride, low HDL-C or high TC/HDL-C ratio. This evidence is supported by the concept of glucolipotoxicity15,16 and elucidates a pathway to HF in diabetic individuals with lipid abnormalities. Participants with diabetes usually have higher triglyceride and lower HDL-C levels.11,29 Our findings were independent of diabetes mellitus treatment/severity, and we speculated that this association could be a reflection of more severe dysglycemia, but may partly result from a synergistic relationship between lipid and glucose abnormalities.

Plasma triglyceride is positively associated with myocardial triglyceride (MTG) content.11 MTG content is increased in uncomplicated type 2 diabetes mellitus and is independently associated with impaired left ventricular diastolic function.11 Individuals with uncomplicated type 2 diabetes mellitus and high MTG content, and without ischemia, have shown greater impairments of biventricular myocardial strain and strain rate.13 Because the association of high triglyceride with incident HF among participants with diabetes was attenuated by interim MI, we contemplated that this association may partly be mediated by MI, but could also involve changes in MTG content. MTG measurements were unavailable in MESA and are not included in our analyses. There are inconsistent data on the association of triglyceride with CVD in individuals with diabetes when HDL-C is taken into account.27,30,31 The association of triglyceride with incident HF among participants with diabetes in our study was attenuated by HDL-C.

Plasma HDL-C is inversely correlated with MTG content.11 HDL-C may prevent myocardial lipid accumulation through reverse cholesterol transport, which enhances elimination of free cholesterol.11 The effects of HDL-C may also be related to its anti-inflammatory effects9 because inflammation has been implicated in HF development.6 Through its anti-inflammatory effects, HDL-C improves endothelial function and promotes repair thereby protecting against cardiac and vascular remodeling.9 HDL-C is strongly associated with MI.32 Because the association of HDL-C with incident HF was independent of interim MI in participants with diabetes, we speculated that alternative pathways possibly occurring before MI may play a greater role in this association. Insulin resistance is an independent predictor of HF,25,33 but also has a role in the development of high triglyceride and low HDL-C levels.33,34 The mechanisms by which high triglyceride and low HDL-C resulted in HF among participants with diabetes in our study were independent of insulin resistance because the associations seemed similar in models that included and excluded insulin resistance.

TC/HDL-C ratio may be a marker of metabolic abnormalities that predict ischemic heart disease.35 The association of TC/HDL-C ratio became nonsignificant when we adjusted for metabolic abnormalities, insulin resistance, and interim MI, and we contemplated that its effects may partly be accountable to these conditions. Non HDL-C is a potent predictor of CVD among various individuals,29,36 but was not associated with incident HF in our study. Our findings for non HDL-C are different from those of Velagaleti et al9 in the Framingham Heart Study. However, there may have been residual confounding in the study of Velagaleti et al because their analyses accounted for fewer risk factors.

Intracellular lipid accumulation occurs when high plasma FFAs and triglyceride levels lead to increased FFA uptake in nonadipose tissues3 or because of defects in lipid oxidation.37 Excess FFAs in cells are esterified and stored in lipid droplets as triglycerides,3 which is not toxic,13,37 and initially acts as a buffer by diverting FFAs from toxic pathways.11 Cells of nonadipose tissues, such as cardiomyocytes, have limited lipid storage capacity,3 and excess FFAs are shunted into nonoxidative pathways, which generate toxic byproducts that lead to apoptosis,11–13,15 contractile and metabolic dysfunction.11,13,15 The order, progression, and role of cellular changes in lipotoxicity is not clearly defined, and may depend on lipid composition, and differ between cell types.3,15,38

Among individuals without diabetes, we failed to demonstrate independent associations of lipid fractions with incident HF. Our findings in nondiabetic participants agree with those of Voulgari et al7 and Wang et al8 for high triglyceride, but not for low HDL-C. This inconsistency may be attributable to racial differences arising from the populations being studied, although we would expect such differences to apply to both lipid parameters.

Our study has strengths. MESA is a prospective study that was conducted in 6 locations in the United States and involved a large number of participants of different ages, ethnicities, and sex. Data collection and HF ascertainment procedures were highly standardized. Our definitions of high triglyceride, low HDL-C, and high TC/HDL-C ratio reflect cutoffs accepted in current practice guidelines.

We also acknowledge limitations. We used single, fasting lipid measurements at baseline, which does not reflect intraindividual or postprandial variations during follow-up. We adjusted for interim MI, but the absence of MI does not exclude subclinical myocardial ischemia. Because of limited number of events, we did not explore our associations according to HF subtype. We had inadequate power to explore our associations by ethnicity. Ethnic-specific associations should be pursued in adequately powered studies. We excluded participants who were taking lipid-lowering medications at baseline, but lipid-lowering medications may have been started during follow-up. Lipid-lowering therapy (such as statins) have been shown to decrease HF incidence in clinical trials,9 and their use during follow-up would more likely result in weakening of our associations. Hemoglobin A1C measurements were unavailable at baseline. There may be misclassification of diabetes mellitus treatment/severity because 119 participants were newly diagnosed from their blood glucose measures and were not on treatment at baseline. ECG has limited sensitivity for diagnosing left ventricular hypertrophy.39 Although left ventricular mass measurements (by magnetic resonance imaging) were available in a subset of participants, we had inadequate events for analysis in this subpopulation. Individuals with known clinical CVD were excluded from MESA. However, eligibility for participation was based on self-reported information. It is unlikely but possible that some participants may have experienced HF before enrollment.

Conclusions

Diabetes mellitus increases the risk of HF, and the coexistence of high triglyceride, low HDL-C, or high TC/HDL-C ratio may further aggravate HF risk. Isolated lipid abnormalities may be insufficient to cause overt HF in the absence of diabetes mellitus. Individuals with diabetes may require more aggressive control of triglyceride, HDL-C, and TC/HDL-C ratio for HF prevention. Our findings should be replicated in other cohorts and subsequently confirmed in appropriately designed trials.

Supplementary Material

01

Figure 3.

Figure 3

Heart failure free probability in Multiethnic Study of Atherosclerosis according to high-density lipoprotein–cholesterol (HDL-C) and diabetes mellitus category. There were 318, 298, 1738, and 3334 participants in categories 1, 2, 3, and 4, respectively. Low HDL-C refers to HDL-C <40 mg/dL in men and <50 mg/dL in women.

CLINICAL PERSPECTIVE.

The prevalence and healthcare costs associated with heart failure are increasing. It is important to identify modifiable risk factors that can be targeted for heart failure prevention. Diabetes mellitus has been identified as a risk factor for heart failure. Although results from observational studies have been inconsistent, current clinical guidelines consider dyslipidemia as a risk factor for heart failure. The coexistence of diabetes mellitus and dyslipidemia may further increase the risk of heart failure. The objective of this study was to evaluate the associations of lipid fractions with incident heart failure separately in participants with and without diabetes. The findings from our study have shown that the risk of incident heart failure is greater in individuals with diabetes mellitus who also have high triglyceride, low levels of high-density lipoprotein-cholesterol, or high total cholesterol/high-density lipoprotein–cholesterol ratio. Isolated lipid abnormalities may be insufficient to cause overt heart failure in the absence of diabetes mellitus. Diabetes mellitus increases the risk of incident heart failure, but the coexistence of high triglyceride, low levels of high-density lipoprotein–cholesterol, or high total cholesterol/high-density lipoprotein–cholesterol ratio further aggravates the risk of incident heart failure. This finding has clinical implications because diabetic individuals may require more aggressive control of triglyceride, high-density lipoprotein–cholesterol, and total cholesterol to high-density lipoprotein–cholesterol ratio for heart failure prevention.

Acknowledgments

We thank other investigators, staff, and participants of Multiethnic Study of Atherosclerosis (MESA) for their contributions. The full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org.

Sources of Funding

Multiethnic Study of Atherosclerosis was supported by contracts N01-HC-95159 through N01-HC-95169 from the National Heart, Lung, and Blood Institute (NHLBI). T32 training grant was supported by grant 5 T32 HL087730-03 from NHLBI.

Footnotes

Disclosures

Dr Goff served as a member of Merck’s operations committee and Takeda’s data safety and monitoring board. The other authors have no conflicts to report.

Presented as an oral abstract at American Heart Association 2012 scientific sessions, Los Angeles, CA.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

01