Glycemic Status, Cardiac Structure, and Function by Magnetic Resonance Imaging: The Jackson Heart Study

Abstract

Background

Type 2 diabetes is associated with a high risk of heart failure. Hence, precise cardiac imaging can help early detection of subclinical myocardial dysfunction. We investigated the associations of glycemic status and markers with left ventricular (LV) remodeling assessed by cardiac magnetic resonance imaging among Black adults.

Methods

Peak systolic circumferential strain (Eulerian circumferential systolic strain), LV mass, end‐diastolic volume, end‐systolic volume, LV ejection fraction, and LV remodeling index were measured by cardiac magnetic resonance imaging in JHS (Jackson Heart Study) participants between 2008 and 2012. We evaluated the associations of glycemic markers (fasting plasma glucose and hemoglobin A_1C) and glycemic status (diabetes, prediabetes, and normoglycemia) with cardiac magnetic resonance imaging measures of LV structure and function.

Results

A total of 1609 Black participants (mean age, 59 years; 63% women; 27.4% with diabetes and 46.8% with prediabetes) were included. Participants with diabetes had higher LV mass (adjusted β, 4.94 [95% CI, 0.51–9.37]), higher LV remodeling index (β, 0.05 [95% CI: 0.01– 0.09]), impaired global circumferential systolic strain (β, 0.44 [95% CI, 0.04–0.84]), and a higher odds of abnormal strain compared with those with normoglycemia (adjusted odds ratio, 1.43 [95% CI, 1.01–2.04]). Each SD increase in hemoglobin A_1C was associated with higher LV mass (β, 2.48 [95% CI, 0.76–4.21]), LV remodeling index (β, 0.02 [95% CI, 0.01–0.04]), and impaired global circumferential systolic strain (β, 0.21 [95% CI, 0.06–0.37]).

Conclusions

In a large community‐based cohort of Black individuals, diabetes was associated with differences in LV structure and myocardial strain measured by cardiac magnetic resonance imaging.

Nonstandard Abbreviations and Acronyms

ECC

Eulerian circumferential systolic strain

EDV

end‐diastolic volume

FPG

fasting plasma glucose

Clinical Perspective.

What Is New?

• Subclinical myocardial alterations predict incident heart failure, especially in the setting of diabetes; however, investigations of the association between diabetes and subclinical myocardial alterations assessed by cardiac magnetic resonance imaging have seldom or not been conducted in Black individuals.

• Compared with individuals without diabetes, those with diabetes had higher left ventricular mass, higher left ventricular remodeling index, impaired global circumferential systolic strain, and a higher odds of abnormal strain.

What Are the Clinical Implications?

• Early detection of subtle but potentially significant subclinical myocardial structural and functional abnormalities can be detected early in diabetes using advanced cardiac resonance imaging, which can possibly help prevent heart failure.

In the United States, Black adults experience a higher prevalence of heart failure (HF) and worse HF‐related outcomes than other racial or ethnic groups. ¹ , ² The rates of HF are partially driven by diabetes, a condition that tends to be more frequent among Black individuals than in other racial and ethnic groups. ³ Diabetes confers a 2‐ to 4‐fold higher risk of HF, ⁴ and it is 1 of the most common causes of HF. ⁵

Accruing evidence suggests that subclinical cardiac dysfunction is a potent precursor of HF, ⁶ especially in the context of diabetes. ⁷ Indeed, subclinical diabetes‐related cardiomyopathy may precede the onset of clinically evident HF. ⁵ The American Diabetes Association currently recommends screening for HF among individuals with diabetes using cardiac biomarkers, and then performing follow‐up imaging. ⁸ Prior echocardiographic studies suggest that diabetes is associated with left ventricular (LV) alterations, including systolic dysfunction, diastolic dysfunction, and alterations of the LV geometry. ⁹ , ¹⁰ Traditional 2‐dimensional echocardiography may, however, lack sensitivity for detection of preclinical HF, particularly systolic dysfunction. ¹¹ , ¹² However, cardiac magnetic resonance imaging (CMR) offers superior visualization and quantification of the LV morphology and function compared with echocardiography, ¹³ and thus can provide additional insights into the diabetes‐related myocardial alterations. CMR exhibits superior consistency, enhanced agreement between different observers and within the same observer, and improved image resolution, and it relies on fewer assumptions about cardiac geometry. ¹³ , ¹⁴ This enhanced image quality becomes especially significant in individuals with diabetes, as echocardiography may pose challenges because of accompanying obesity. Tagged CMR, currently the preferred CMR technique for assessing myocardial deformation, enables the measurement of strain and offers valuable information about discrete myocardial impairment. ¹⁵ In brief, CMR is the reference standard for assessment of volumes, ejection fraction, and cardiac mass. ¹⁶

The associations of glycemic status with CMR‐based measures of LV structure and function have not been extensively explored in community‐dwelling Black individuals. Hence, in a community‐based cohort of Black adults (with and without diabetes) recruited to the JHS (Jackson Heart Study), ¹⁷ we investigated the associations of glycemic status (normoglycemia, prediabetes, and diabetes) and glycemic markers (fasting plasma glucose and glycated hemoglobin) with CMR‐based measures of LV function and structure.

METHODS

The data underlying this article are based on the JHS, and are not publicly available, but are available from the corresponding author on reasonable request.

Study Design

The design and methods used in the JHS have been previously documented. ¹⁷ The JHS is a prospective cohort study of cardiovascular risk factors and diseases among Black individuals. Between 2000 and 2004, the JHS recruited 5306 Black adults, aged 21 to 94 years and residing in the metropolitan area of Jackson, Mississippi. ¹⁷

For the present investigation, we included participants to the JHS CMR component that began in 2008 at the end of visit 2 and continued through visit 3, ending in December 2012. The exclusion criteria for CMR were pregnancy, contraindication to CMR (implanted electrical devices, pacemaker, or history of metal around orbit), claustrophobia, or individuals who could not fit in the CMR machine. Of the 1656 JHS participants who completed the CMR, those with missing data on Eulerian circumferential systolic strain (ECC) (n=39) or glycemic status (n=8) were excluded, resulting in the final sample of 1609 participants (n=232 from visit 2 and n=1377 from visit 3).

The study protocol for the JHS received approval from the Institutional Review Boards of Jackson State University, Tougaloo College, and the University of Mississippi Medical Center. All enrolled participants provided written informed consent.

Assessment of Glycemic Status

The glycemic markers and status were assessed from the visit closest to the CMR. Blood samples were obtained after a fasting period of at least 8 hours. ¹⁷ Fasting plasma glucose (FPG) levels were measured using the glucose oxidase method. The glucose assays were run in duplicate; the intra‐assay coefficient of variation was <3%. Hemoglobin A_1C (HbA_1C) was measured using high‐performance liquid chromatography (Tosoh 2.2; Tosoh Corporation, Tokyo, Japan), an assay certified by the National Glycohemoglobin Standardization Program. ¹⁷ , ¹⁸

Diabetes was defined by FPG ≥126 mg/dL, HbA_1C≥6.5% (48 mmol/mol), or documented use of diabetes medications. ¹⁹ Prediabetes was defined as follows: (1) FPG between 100 and 125 mg/dL or (2) HbA_1C between 5.7% and 6.4% (39–46 mmol/mol); and (3) no self‐reported use antidiabetic medications. ¹⁹ Participants with FPG <100 mg/dL and HbA_1C<5.7% (39 mmol/mol) were classified as having normoglycemia.

Cardiac Magnetic Resonance Imaging

CMR was conducted using a 1.5‐T magnet with a 70‐cm bore and high‐gradient capabilities (Siemens Espree with TIM Cardiac Package, Siemens Medical Solutions USA Inc, Malvern, PA). The examination used a multichannel matrix surface coil to capture short‐ and long‐axis ECG‐gated steady‐state free precession CINE images of the heart. This protocol, designed for measuring cardiac structure and function, was aligned with the CMR examination protocols from the National Heart, Lung, and Blood Institute MESA (Multi‐Ethnic Study of Atherosclerosis) examinations 2 and 3, ensuring standardization for future comparative studies with JHS and MESA. Functional assessment sequences were obtained during brief breath‐holds using white blood sequences (Fast Imaging with Steady Precession, TrueFISP, or TRUFI, Siemens sequence variant Tfi2d1_18) with the following parameters: a field of view of 400 mm, a slice thickness of 8 mm, a matrix of 109×192, repetition time of 45.5 ms, echo time of 1.1 ms, and a flip angle between 78 and 82 degrees. Additional single breath‐hold, ECG‐gated short‐axis sequences at the cardiac base, midsection, and apex were acquired using a CINE radiofrequency grid tagging sequence with these parameters: field of view of 400 mm, slice thickness of 8 mm, a 192×256 matrix, repetition time of 60 seconds, echo time of 4 seconds, and a flip angle of 12 degrees (Siemens Sequence: Tl2d1r5). Radiofrequency pulses were applied to “tag” the myocardium during the CMR.

Global Circumferential Strain (ECC), LV Structure and Function

CMR scans were centrally analyzed at Wake Forest University School of Medicine (Winston‐Salem, NC). The measurements of LV mass, end‐diastolic volume (EDV), and end‐systolic volume were performed using Cardiac Image Modeller software (Auckland UniServices Limited, Auckland, New Zealand). From these measurements, stroke volume, LV ejection fraction (stroke volume/EDV), and the LV remodeling index (mass/EDV) were calculated.

Diagnasoft Harmonic Phase software (HARP, Research Triangle Park, Morrisville, NC) was used to quantify ECC strain. This involved tracking tagged tissues in the midmyocardium through systole. HARP software uses k‐space data, focusing on phase information, to create a binary image that helps in the semiautomated detection of tag lines throughout the cardiac cycle. This method has previously been used to identify subclinical LV dysfunction. Global peak ECC was determined by averaging the midwall strain in the base, mid, and apical LV regions from short‐axis images.

Regional myocardial circumferential strain was calculated using the Lagrangian formula: [e=(L−Lo)/Lo], where e represents strain, Lo is the baseline length, and L is the instantaneous length at the measurement time. Negative strain values indicate that the segment length is shorter than its original length in the circumferential direction. Therefore, global circumferential systolic strain values closer to 0 (less negative) suggest worse LV function. There is currently no universally accepted reference threshold for abnormal global circumferential strain measured using tagged CMR. ²⁰ Therefore, in the present study, circumferential strain was analyzed primarily as a continuous variable in regression models. For secondary analyses, we defined abnormal strain as an absolute ECC value <15%, consistent with thresholds used in prior studies of myocardial strain. The reproducibility of CMR measurements using HARP was evaluated on 96 scans with blind duplicate examinations, where DICOM headers were relabeled, and the examinations were reinserted into the workflow as new. The interclass correlation coefficient for global ECC analysis was 0.78, whereas the interclass correlation coefficients for EDV, end‐systolic volume, and LV mass were 0.95, 0.88, and 0.96, respectively.

Covariates

The covariates were ascertained from the closest visit preceding the date of acquisition of CMR data, and included demographic information, lifestyle factors (comprising smoking status and alcohol consumption), comorbidities, and medication use. ¹⁷ Body mass index was defined as weight (kg) divided by height² (m²). Blood pressure was calculated as the average of 2 measurements recorded at 5‐minute intervals while seated. Hypertension was defined by a self‐reported diagnosis of hypertension, blood pressure>140/90 mm Hg, or treatment with antihypertensive medications. Current smoking was identified as self‐reported smoking within the past 12 months. The presence of coronary heart disease was determined by participant self‐reports of coronary heart disease history, previous abnormal stress test results, coronary bypass graft surgery, coronary angioplasty, or electrocardiographic evidence of prior myocardial infarction. Plasma total cholesterol, high‐density lipoprotein cholesterol, and triglyceride levels were obtained using standard enzymatic techniques, on a Vitros 950 or 250, Ortho‐Clinical Diagnostics analyzer (Raritan, NJ) in accordance with the College of American Pathologists Proficiency Testing Program. ¹⁸ The Friedewald equation was used to calculate low‐density lipoprotein cholesterol. ²¹ The estimated glomerular filtration rate was based on the Chronic Kidney Disease Epidemiology Collaboration creatinine equation, which was derived from baseline serum concentrations of creatinine and cystatin C. ²² Chronic kidney disease was defined as an estimated glomerular filtration rate ≤60 mL/min per 1.73 m².

Statistical Analysis

We classified participants into clinically relevant glycemic categories (normoglycemia, prediabetes, and diabetes) and compared their characteristics using ANOVA or the Kruskal‐Wallis test (for continuous variables) or the χ² test (for categorical variables).

We used multivariable linear regression models to assess the associations of glycemic markers (FPG and HbA_1C) and glycemic status (normoglycemia, prediabetes, and diabetes) with CMR parameters (including LV mass, EDV, end‐systolic volume, stroke volume, remodeling index, LV ejection fraction, and global circumferential strain). We used multivariable logistic regression (binary or multinomial as appropriate) to evaluate the association of glycemic status (normoglycemia [reference], prediabetes, and diabetes) or glycemic makers (FPG and HbA_1C) with abnormal global circumferential strain. In all regression models, we conducted sequential adjustments. Model 1 was adjusted for age and sex. Model 2 additionally included lifestyle and clinical variables (smoking, alcohol consumption, body mass index, hypertension status, use of blood pressure medications, triglycerides, high‐density lipoprotein cholesterol, history of coronary heart disease, and estimated glomerular filtration rate). Model 3 further included heart rate. This sequential modeling approach allowed assessment of the robustness of associations across progressively adjusted models, including models with potential mediators, such as body mass index and hypertension.

Because there is no universally accepted reference threshold for abnormal global circumferential strain derived from tagged CMR, we conducted a sensitivity analysis using a cohort‐based definition of abnormal strain. Specifically, abnormal strain was defined as values within the lowest quartile of global circumferential strain (ie, the least negative strain values indicating worse myocardial deformation). Logistic regression models identical to those used in the primary analyses were then applied to evaluate the associations of glycemic status and glycemic markers with this alternative definition of abnormal strain.

All statistical analyses were performed using STATA 14.2 (Stata, Inc, College Station, TX). A 2‐sided P<0.05 was considered indicative of statistical significance.

RESULTS

Characteristics of Study Participants

Our study comprised 1609 participants (mean±SD age, 59.2±10.2 years; 63% women; 27.4% with diabetes and 46.8% with prediabetes). Table 1 shows their baseline characteristics by glycemic status. Compared with those with normoglycemia, individuals with diabetes were older and more likely to be smokers and statin users, as well to have obesity, hypertension, and coronary heart disease.

Table 1.

Clinical Characteristics of Participants by Glycemic Status

Characteristic	Total (N=1609)	Normoglycemia (N=415)	Prediabetes (N=753)	Diabetes (N=441)	P value
Age, y	59.2±10.2	56.7±10.5	59.4±10.1	61.3±9.7	<0.001
Women, %	63.0	66.5	60.6	63.9	0.12
Cigarette smoking, %	28.0	25.6	27.3	31.6	0.12
Alcohol intake, %	49.2	54.3	52.7	38.4	<0.001
Body mass index, kg/m2	31.7±6.4	29.8±6.0	31.4±6.2	33.9±6.4	<0.001
Hypertension, %	69.9	55.9	67.7	86.6	<0.001
Use of BP‐lowering drug, %	69.4	57.2	64.6	87.8	<0.001
Systolic BP, mm Hg	126±18	125±17	126±17	128±20	0.023
Diastolic BP, mm Hg	76±11	77±11	76±10	75±11	0.015
Heart rate, beats/min	63±10	62±9	62±9	66±11	<0.001
Fasting plasma glucose, mg/dL	103±31	88±6	97±9	133±52	<0.001
Hemoglobin A1C, %	6.1±1.2	5.3±0.3	5.9±0.3	7.4±1.7	<0.001
Total cholesterol, mg/dL	197.9±40.5	198.7±35.2	201.9±39.4	188.8±46.4	<0.001
LDL cholesterol, mg/dL	121.0±36.5	119.9±31.9	124.9±35.5	114.3±42.0	<0.001
HDL cholesterol, mg/dL	57.8±15.6	61.7±16.9	57.8±15.3	53.6±13.5	<0.001
Triglycerides, mg/dL	84.0 (63.0–114.0)	75.0 (55.5–103.0)	84.5 (65.0–113.0)	93.0 (68.0–124.0)	<0.001
Use of statin, %	36.8	22.6	32.7	55.6	<0.001
eGFR, mL/min per 1.73 m2	89.6±21.1	91.7±20.8	89.2±19.6	88.5±23.7	0.062
Chronic kidney disease, %	2.2	2.9	1.5	2.7	0.18
Coronary heart disease, %	2.4	1.0	1.1	6.1	<0.001

Data are mean±SD, median (quartile 1–quartile 3), or proportion (percentage), as appropriate. BP indicates blood pressure; eGFR, estimated glomerular filtration rate; HDL, high‐density lipoprotein; and LDL, low‐density lipoprotein.

Glycemic Status (Prediabetes and Diabetes) and CMR Parameters

Compared with individuals with normoglycemia (Table 2), those with diabetes had higher LV mass, LV remodeling index, and global circumferential strain (including basal, mid, and apical LV strains).

Table 2.

LV Characteristics by CMR

Characteristic	Total (N=1609)	Normoglycemia (N=415)	Prediabetes (N=753)	Diabetes (N=441)	P value
LV mass, g	134.6±40.4	125.5±34.7	133.1±37.9	145.6±46.5	<0.001
End‐diastolic volume, mL	124.6±34.7	126.2±34.5	123.3±32.8	125.5±37.8	0.32
End‐systolic volume, mL	50.2±21.7	50.9±20.9	49.3±19.6	51.2±25.4	0.27
Stroke volume, mL	74.4±20.9	75.3±19.9	74.0±20.6	74.3±22.2	0.58
Remodeling index, g/mL	1.1±0.3	1.0±0.3	1.1±0.3	1.2±0.3	<0.001
Ejection fraction, %	60.3±9.2	60.3±8.2	60.5±9.1	60.1±10.1	0.78
Global circumferential strain, %
Total strain	−16.1±2.7	−16.7±2.5	−16.2±2.7	−15.3±2.8	<0.001
Basal strain	−15.9±3.2	−16.5±3.0	−16.0±3.2	−15.4±3.2	<0.001
Mid strain	−16.2±3.5	−16.9±3.3	−16.4±3.4	−15.3±3.6	<0.001
Apical strain	−16.1±3.1	−16.9±2.8	−16.2±3.1	−15.3±3.2	<0.001

Data are displayed as mean±SD. CMR indicates cardiac magnetic resonance imaging; and LV, left ventricular.

In multivariable adjusted models, compared with participants with normoglycemia, individuals with diabetes had a higher LV mass (β coefficient [β], 4.94 [95% CI, 0.51–9.37]), higher LV remodeling index (β, 0.05 [95% CI, 0.01–0.09]), and higher global circumferential systolic strain (β, 0.44 [95% CI, 0.04–0.84]; Table 3). Compared with those with normoglycemia, there were nonsignificant positive association of prediabetes with LV mass, LV remodeling index, and global circumferential systolic strain. In secondary analyses using a categorical definition of abnormal strain (ECC <15%), participants with diabetes had higher odds of abnormal global circumferential systolic strain compared with those with normoglycemia (odds ratio [OR], 1.43 [95% CI, 1.01–2.04]), but not for those with prediabetes (OR, 1.32 [95% CI, 0.98, 1.78]) (Figure and Table 4).

Table 3.

Association Between Glycemic Status and CMR‐Derived LV Structure and Function

Variable	Model	Normoglycemia	Prediabetes		Diabetes
			β Coefficient	(95% CI)	β Coefficient	(95% CI)
LV mass, g	Model 1	Reference	5.39	(1.48 to 9.30)**	19.91	(15.51 to 24.32)***
	Model 2	Reference	−0.20	(−3.73 to 3.34)	4.49	(0.10 to 8.87)*
	Model 3	Reference	−0.31	(−3.86 to 3.23)	4.94	(0.51 to 9.37)*
End‐diastolic volume, mL	Model 1	Reference	−2.03	(−5.90 to 1.83)	2.46	(−1.89 to 6.81)
	Model 2	Reference	−4.06	(−7.84 to −0.28)*	−3.17	(−7.85 to 1.52)
	Model 3	Reference	−4.29	(−8.01 to −0.56)*	−1.65	(−6.30 to 3.00)
End‐systolic volume, mL	Model 1	Reference	−1.65	(−4.11 to 0.81)	1.27	(−1.50 to 4.04)
	Model 2	Reference	−2.05	(−4.44 to 0.34)	−1.37	(−4.33 to 1.59)
	Model 3	Reference	−2.12	(−4.51 to 0.28)	−1.40	(−4.40 to 1.59)
Stroke volume, mL	Model 1	Reference	−0.39	(−2.79 to 2.02)	1.19	(−1.52 to 3.89)
	Model 2	Reference	−2.01	(−4.41 to 0.38)	−1.80	(−4.77 to 1.17)
	Model 3	Reference	−2.17	(−4.48 to 0.14)	−0.25	(−3.14 to 2.64)
Remodeling index, g/mL	Model 1	Reference	0.06	(0.03 to 0.10)***	0.14	(0.10 to 0.18)***
	Model 2	Reference	0.03	(−0.01 to 0.07)	0.06	(0.01 to 0.10)*
	Model 3	Reference	0.03	(−0.01 to 0.07)	0.05	(0.01 to 0.09)*
Ejection fraction, %	Model 1	Reference	0.38	(−0.70 to 1.46)	−0.14	(−1.36 to 1.08)
	Model 2	Reference	0.05	(−1.06 to 1.16)	−0.10	(−1.47 to 1.27)
	Model 3	Reference	0.02	(−1.07 to 1.11)	0.40	(−0.97 to 1.76)
Global circumferential strain, %
Total strain	Model 1	Reference	0.35	(0.04 to 0.67)*	1.17	(0.82 to 1.53)***
	Model 2	Reference	0.16	(−0.16 to 0.48)	0.53	(0.13 to 0.92)**
	Model 3	Reference	0.17	(−0.15 to 0.48)	0.44	(0.04 to 0.84)*
Basal strain	Model 1	Reference	0.31	(−0.06 to 0.68)	0.86	(0.44 to 1.28)***
	Model 2	Reference	0.14	(−0.23 to 0.52)	0.29	(−0.18 to 0.76)
	Model 3	Reference	0.16	(−0.22 to 0.54)	0.21	(−0.26 to 0.68)
Mid strain	Model 1	Reference	0.29	(−0.12 to 0.69)	1.39	(0.93 to 1.85)***
	Model 2	Reference	0.08	(−0.33 to 0.50)	0.68	(0.16 to 1.19)*
	Model 3	Reference	0.08	(−0.34 to 0.49)	0.56	(0.04 to 1.07)*
Apical strain	Model 1	Reference	0.51	(0.15 to 0.87)**	1.32	(0.91 to 1.73)***
	Model 2	Reference	0.30	(−0.07 to 0.67)	0.66	(0.20 to 1.11)**
	Model 3	Reference	0.31	(−0.06 to 0.68)	0.61	(0.15 to 1.07)**

Model 1 is adjusted for age and sex. Model 2 includes model 1 variables plus smoking, alcohol consumption, body mass index, hypertension status, use of hypertension medications, triglycerides, high‐density lipoprotein cholesterol, history of coronary heart disease, and estimated glomerular filtration rate. Model 3 includes model 2 variables plus heart rate. CMR indicates cardiac magnetic resonance imaging; and LV, left ventricular.

^* P<0.01.

^** P<0.001.

^*** P<0.05.

Figure 1. Odds ratios for abnormal global circumferential strain according to the glycemic status.

Table 4.

ORs for Abnormal Global Circumferential Strain According to Glycemic Markers

Model	Glycemic status			Per 1 SD of HbA1C	Per 1 SD of FPG
	Normoglycemia	Prediabetes	Diabetes
Model 1	Reference	1.51 (1.14–1.99)**	2.37 (1.75–3.21)***	1.33 (1.20–1.47)***	1.21 (1.09–1.35)***
Model 2	Reference	1.29 (0.96–1.73)	1.51 (1.07–2.14)*	1.19 (1.04–1.35)**	1.11 (0.99–1.24)
Model 3	Reference	1.32 (0.98–1.78)	1.43 (1.01–2.04)*	1.13 (0.99–1.29)	1.05 (0.94–1.18)

Values are ORs (95% CIs). Model 1 is adjusted for age and sex. Model 2 includes model 1 variables plus smoking, alcohol consumption, body mass index, hypertension status, use of blood pressure medications, triglycerides, high‐density lipoprotein cholesterol, history of coronary heart disease, and estimated glomerular filtration rate. Model 3 includes model 2 variables plus heart rate.

FPG indicates fasting plasma glucose; HbA_1C, hemoglobin A_1C; and OR, odds ratio.

^* P<0.05.

^** P<0.01.

^*** P<0.001.

Glycemic Markers and CMR Parameters

In multivariable adjusted models, each SD increase in HbA_1C was associated with a significantly higher LV mass (β, 2.48 [95% CI, 0.76–4.21]), LV remodeling index (β, 0.02 [95% CI, 0.01–0.04]), and global circumferential systolic strain (β, 0.21 [95% CI, 0.06–0.37]; Table 5). For fasting plasma glucose, after adjustment, each SD increase in FPG was not significantly associated with higher LV mass, LV remodeling index, global circumferential systolic strain, or any other CMR parameter.

Table 5.

Association Between Glycemic Markers and CMR‐Derived LV Structure and Function

Variable	Model 1		Model 2		Model 3
	β Coefficient	(95% CI)	β Coefficient	(95% CI)	β Coefficient	(95% CI)
Glycated hemoglobin, %
LV mass, g	6.56	(4.97 to 8.15)***	2.35	(0.64 to 4.06)**	2.48	(0.76 to 4.21)**
End‐diastolic volume, mL	0.25	(−1.32 to 1.82)	−1.76	(−3.59 to 0.07)	−0.89	(−2.71 to 0.93)
End‐systolic volume, mL	0.93	(−0.08 to 1.93)	−0.01	(−1.17 to 1.15)	0.01	(−1.17 to 1.18)
Stroke volume, mL	−0.68	(−1.66 to 0.30)	−1.75	(−2.91 to −0.59)**	−0.89	(−2.02 to 0.24)
Remodeling index, g/mL	0.05	(0.04 to 0.07)***	0.03	(0.01 to 0.05)***	0.02	(0.01 to 0.04)**
Ejection fraction, %	−0.60	(−1.04 to −0.16)**	−0.66	(−1.20 to −0.12)*	−0.39	(−0.93 to 0.14)
Global circumferential strain, %
Total strain	0.43	(0.30 to 0.56)***	0.27	(0.12 to 0.43)***	0.21	(0.06 to 0.37)**
Basal strain	0.32	(0.17 to 0.47)***	0.2	(0.01 to 0.38)*	0.14	(−0.05 to 0.32)
Mid strain	0.54	(0.37 to 0.70)***	0.38	(0.18 to 0.58)***	0.31	(0.10 to 0.51)**
Apical strain	0.43	(0.29 to 0.58)***	0.23	(0.05 to 0.41)*	0.19	(0.01 to 0.37)*
Fasting plasma glucose, mg/dL
LV mass, g	4.3	(2.65 to 5.95)***	0.9	(−0.60 to 2.40)	1.04	(−0.48 to 2.56)
End‐diastolic volume, mL	−0.37	(−1.95 to 1.21)	−1.49	(−3.09 to 0.11)	−0.54	(−2.14 to 1.05)
End‐systolic volume, mL	0.54	(−0.45 to 1.52)	0.15	(−0.86 to 1.17)	0.21	(−0.82 to 1.23)
Stroke volume, mL	−0.9	(−1.91 to 0.10)	−1.64	(−2.65 to −0.63)**	−0.75	(−1.74 to 0.24)
Remodeling index, g/mL	0.04	(0.02 to 0.05)***	0.02	(0.01 to 0.03)*	0.01	(−0.01 to 0.02)
Ejection fraction, %	−0.66	(−1.11 to −0.20)**	−0.71	(−1.18 to −0.25)**	−0.45	(−0.92 to 0.02)
Global circumferential strain, %
Total strain	0.31	(0.17 to 0.44)***	0.16	(0.03 to 0.30)*	0.10	(−0.04 to 0.23)
Basal strain	0.21	(0.05 to 0.37)**	0.07	(−0.09 to 0.23)	0.01	(−0.15 to 0.17)
Mid strain	0.39	(0.22 to 0.56)***	0.25	(0.07 to 0.42)**	0.17	(−0.01 to 0.35)
Apical strain	0.32	(0.17 to 0.47)***	0.17	(0.02 to 0.33)*	0.13	(−0.03 to 0.29)

Values are reported per 1‐SD increment in the glycemic marker. Model 1 is adjusted for age and sex. Model 2 includes model 1 plus smoking, alcohol consumption, body mass index, hypertension status, use of hypertension medications, triglycerides, high‐density lipoprotein cholesterol, history of coronary heart disease, and estimated glomerular filtration rate. Model 3 includes model 2 plus heart rate. CMR indicates cardiac magnetic resonance imaging; and LV, left ventricular.

^* P<0.05.

^** P<0.01.

^*** P<0.001.

The ORs of abnormal global circumferential systolic strain were 1.13 (95% CI, 0.99–1.29) for each SD increase in HbA_1C and 1.05 (95% CI, 0.94–1.18) for each SD increment in FPG.

In sensitivity analyses using a cohort‐based definition of abnormal strain as the lowest quartile of global circumferential strain, the results were generally consistent with the primary findings. Diabetes and higher HbA_1C levels remained associated with higher odds of abnormal strain, whereas fasting plasma glucose was not significantly associated with abnormal strain after full adjustment (Table S1).

DISCUSSION

In a large community‐based cohort of Black individuals, we evaluated the associations of glycemic markers and glycemic status with CMR‐based measures of LV remodeling. We found that diabetes was associated with modest but statistically significant differences in LV mass, LV remodeling index, and global circumferential systolic strain, even after adjusting for potential confounders. Moreover, Hb_A1C, a robust marker of hyperglycemia, was more strongly associated with alterations in LV mass and global circumferential systolic strain than FPG. Our observations are consistent with the hypothesis that chronic hyperglycemia may be associated with alterations in myocardial structure and function. Furthermore, our findings suggest that not only diabetes, but nondiabetic range hyperglycemia may be associated with subclinical myocardial alterations. The positive association between HbA_1C and CMR markers of LV dysfunction in our study supports the possibility that chronic glycemic exposure is linked with alterations in myocardial structure and function. ²³ The fact that FPG was not significantly associated with these cardiac parameters after adjustment suggests that long‐term chronic glycemic exposure, as reflected by HbA_1C, may be a more important determinant of cardiac changes than short‐term glucose fluctuations. Given the cross‐sectional design of the study, these findings should be interpreted as associations rather than evidence of causality, and longitudinal studies with repeated CMR assessments are needed to determine whether dysglycemia precedes and contributes to the progression of myocardial remodeling.

Although the observed differences in LV mass and circumferential strain were modest in absolute magnitude, similar differences in myocardial structure and deformation have been associated with increased risk of incident HF in population‐based studies. Prior analyses from population‐based cohorts have demonstrated that relatively small differences in myocardial strain and ventricular remodeling indexes may precede the development of overt HF and other adverse cardiovascular outcomes. ²⁴ , ²⁵ Thus, even modest alterations in subclinical myocardial structure and function may have important implications at the population level.

Our observations confirm and extend prior findings, but also provide additional insights into the high burden of subclinical myocardial dysfunction related to impaired glucose regulation states, which confer a high risk of overt HF. ²³ Our study examined the associations of glycemic markers with CMR‐based measures of LV function specifically including strain in a large sample focused on Black individuals. Our findings are consistent with those from a recent analysis of the UK Biobank, which also showed that diabetes was associated with impaired LV global circumferential strain in a large cohort of community volunteers living in the United Kingdom. ²⁶ However, the later analysis did not include HbA_1C assessment and had a limited number of Black individuals. Likewise, our results are in agreement with those from the FHS (Framingham Heart Study) (although it did not evaluate myocardial strain) ²⁷ and the MESA. ²⁸ Our results are also in agreement with other reports of a positive association between glycemic markers and other markers of subclinical cardiovascular disease, such as coronary artery calcification. ²⁹ Our study expands on prior studies, which included a limited number of Black participants, ²⁶ , ²⁷ , ²⁸ by focusing on Black individuals and including a vast array of glycemic markers concomitantly and multiple CMR measures.

Several biological mechanisms have been proposed to explain potential links between diabetes and myocardial remodeling. ²³ In the setting of diabetes, cardiac remodeling may be driven by coronary microvascular endothelial dysfunction, which is possibly attributable to systemic inflammation and deposition of endothelial advanced glycation end products, with uncoupled nitric oxide synthesis with increased production of reactive oxygen species. This results in cardiomyocyte hypertrophy, especially in the presence of hyperinsulinemia. ³⁰ , ³¹ Other drivers of remodeling include lipotoxicity, which leads to cardiomyocyte cell death and extensive replacement of fibrosis because hyperglycemia boosts protein kinase G activity in fibroblasts. ³⁰ , ³¹

The lack of statistically significant associations between prediabetes and most CMR parameters after multivariable adjustment may reflect several factors. First, prediabetes represents a heterogeneous metabolic state with varying degrees of glycemic dysregulation. Second, the magnitude of myocardial alterations associated with early dysglycemia may be relatively small and, therefore, more difficult to detect in cross‐sectional analyses. Finally, it is possible that more pronounced structural and functional myocardial alterations emerge once glycemic abnormalities reach the diabetic range, suggesting a potential threshold effect.

Our study has important implications for public health and clinical practice. Our findings highlight the potential importance of glycemic status in relation to subclinical myocardial alterations detectable by advanced cardiac imaging. These findings suggest that individuals with diabetes may exhibit detectable subclinical myocardial alterations before the onset of overt HF, thus supporting the need to target individuals with diabetes for HF prevention, as suggested in the recent American Diabetes Association consensus statement on HF detection. ⁸ This is critical among Black individuals who experience the highest burden of diabetes and HF in the United States, and thus may need imaging‐based HF screening. From a clinical perspective, the detected individuals may benefit from appropriate therapies, such as glucagon‐like peptide‐1 receptor agonists, to reduce the HF risk. ³² , ³³ Lifestyle interventions, including increasing physical activity and dietary changes, can also be beneficial in such a context. ³⁴ However, further research is needed to explore the cost‐effectiveness of using imaging for risk stratification among individuals with diabetes.

Our study is not without limitations. First, given its observational design, residual or unmeasured confounding is a possibility. The cross‐sectional nature of our investigation also limits our ability to establish temporality and thus make causal inference. Third, the glycemic and CMR assessments were conducted at a single examination. Such single‐occasion measurements are prone to regression dilution bias, and we may have underestimated the true strength of the associations. Fourth, to avoid overfitting of the multivariable models, we were parsimonious in our adjustments, and did not include physical activity or dietary intake. We did not have concurrent insulin‐based measures of glycemic regulation, such as insulin resistance markers. We also did not have detailed information on diabetes duration, treatment modality, or glycemic variability, all of which may influence myocardial remodeling and cardiac function. For example, longer duration of diabetes and more severe metabolic dysregulation have been associated with greater myocardial structural and functional abnormalities in prior studies. Future investigations incorporating these measures may provide additional insights into the relationship between diabetes severity and subclinical myocardial dysfunction. Fifth, longitudinal systolic strain, potentially a more sensitive marker than ECC for detecting changes in myocardial deformation, was not measured in the JHS population. In addition, the CMR protocol did not include myocardial tissue characterization techniques, such as T1 mapping, extracellular volume quantification, or late gadolinium enhancement, which may provide additional insights into diffuse myocardial fibrosis and early diabetic cardiomyopathy. This presents an opportunity for future research. Sixth, there is currently no universally accepted threshold for abnormal global circumferential strain measured using tagged CMR. Hence, we primarily analyzed as a continuous variable, which may better capture subtle myocardial functional differences. Given the lack of standardized thresholds for abnormal circumferential strain measured using tagged CMR, we conducted additional analyses using an ECC threshold <15% and a cohort‐based quartile definition of abnormal strain. The consistency of results across these approaches supports the robustness of the observed associations. In addition, hypertension and obesity may represent intermediate mechanisms linking dysglycemia with myocardial remodeling. Adjustment for these factors may therefore attenuate the observed associations. However, the consistency of results across sequential models suggests that the observed relationships were not entirely explained by these factors. Also, contemporary cardioprotective glucose‐lowering therapies, such as sodium‐glucose cotransporter 2 inhibitors and glucagon‐like peptide‐1 receptor agonists, were not widely used during the time period of the CMR examinations in the JHS. Finally, the generalizability of our data should be interpreted in the context of our study limited to Black adults from Jackson, Mississippi.

Despite these limitations, our study possesses notable strengths, which include a well‐characterized, community‐based sample of Black individuals, the examination of a comprehensive set of CMR parameters reflecting various aspects of the subclinical effects of dysglycemia, the examination of various stages of dysglycemia, including prediabetes and diabetes, and the availability of several glycemic markers (HbA_1C and FPG). Finally, the observed associations were robust to adjustment for obesity, which frequently coexists with diabetes, and was shown to  be associated with lower peak systolic circumferential strain among Black individuals. ³⁵

CONCLUSIONS

In a community‐based sample of Black adults, diabetes and higher levels of HbA_1C were associated with differences in CMR‐derived measures of LV structure and myocardial strain. Longitudinal studies are needed to determine whether these imaging findings predict the development of overt HF and to better understand the mechanisms linking dysglycemia with myocardial alterations. Overall, our findings are of contemporary significance given the increasing burden of glucose intolerance and HF in the United States, especially among Black individuals.

Sources of Funding

The Jackson Heart Study is supported and conducted by HHSN268201800014I, HHSN268201800015I, HHSN268201800010I, HHSN268201800011I and HHSN268201800012I, from the National Heart, Lung, and Blood Institute (NHLBI) and the National Institute on Minority Health and Health Disparities (NIMHD), Dr Justin B. Echouffo‐Tcheugui is funded by NHLBI grant K23 HL153774.

Disclosures

RJM reports receiving grants from American Regent, Amgen, AstraZeneca, Bayer, Cytokinetics, Merck, Novartis, and Zoll; and personal consulting fees from Abbott, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Boston Scientific, Cytokinetics, Medtronic, Merck, Novartis, Pharmacosmos, Respicardia, Roche, Sanofi, Vifor, Windtree, and Zoll. All the other authors have nothing to disclose.

Supporting information

Table S1

STROBE Statement—Checklist