Association of Diabetes and Glycemic Control with Left Atrial Function: The Atherosclerosis Risk in Communities (ARIC) Study

Abstract

Background and Aims:

Although glycemic status is associated with impaired cardiac structure and function, less is known on left atrial (LA) function across the glycemic spectrum. We evaluated the association of diabetes and glycemic control with LA function in a community-based cohort of older adults.

Methods and Results:

This cross-sectional analysis included 5075 participants from the Atherosclerosis Risk in Communities Study (mean age 75.5 years, 58% women, and 20% Black adults) with echocardiographic strain data for LA reservoir, conduit, and contractile function. Multivariable linear regression was used to assess associations of diabetes status and glycemic control with LA function. In participants without diabetes, we used ordinal linear regression to evaluate associations of fasting glucose and HbA1c with LA function. Compared to individuals with a normal fasting glucose, prevalent diabetes was associated with 0.68% lower LA conduit function (95% confidence interval (CI): −1.11 to −0.25) and prediabetes a 0.47% reduction (95% CI: −0.85 to −0.09) in fully adjusted analyses. Persons with diabetes and high HbA1c (HgbA1c ≥7% vs <7%) had 1.05% lower LA conduit function (95% CI: −1.63, −0.48). Among individuals without diagnosed diabetes, higher fasting glucose, but not HbA1c, was significantly associated with worse LA conduit function. No significant associations were observed for LA reservoir and contractile function.

Conclusions:

A history of diabetes, prediabetes, and higher fasting glucose levels in persons without diabetes were associated with worse LA conduit function. Corroborative research is needed in prospective cohorts as well as studies that explore underlying mechanisms.

During the initial stages of left ventricular diastolic dysfunction (LVDD), changes in left atrial (LA) structure and function help to preserve cardiac output. However, over time, these changes can become maladaptive and LA dysfunction ensues, resulting in a reduction in LA compliance, rise in LA pressure, and consequent LA failure.¹ The progression to LA dysfunction plays an important role in the transition from left ventricular diastolic dysfunction (LVDD) to heart failure with preserved ejection fraction (HFpEF).^2–4 LA dysfunction, particularly in those with diabetes, has been associated with an increased risk of heart failure (HF), atrial fibrillation (AF), coronary heart disease (CHD), and stroke.^5–7

LA function has 3 phases, serving as a reservoir in systole, as a conduit in early diastole, and as a booster pump in late diastole.⁸ The prevalence of LVDD in individuals with diabetes is significantly higher than in the general population and individuals with diabetes may be predisposed to LA dysfunction.^9–11 Current data suggest that LA function is worse in those with diabetes compared to those without, and that LA function is worse in those with poorly controlled diabetes compared to those with controlled diabetes.^12–17 Worse LA function, as measured by strain imaging, has also been shown to be more prevalent in adolescents and young adults with type 2 diabetes.¹⁸ However, research based on large population-based cohorts that include individuals with normoglycemia, glucose intolerance, and diagnosed diabetes is lacking. Evaluating whether poorer glycemic control is associated with worse LA function in such population-based cohorts could help better understand whether glycemic abnormalities may be etiologically linked to LA dysfunction. To address this question, we conducted a cross-sectional analysis of a community-based cohort to examine whether diabetes status (diagnosed diabetes, prediabetes, or neither) was associated with LA function. We also evaluated differences in LA function according to glycemic control (HbA1c) among participants with a history of diabetes as well as continuous associations with fasting glucose and HbA1c levels in individuals without diabetes.

Materials and Methods

Study participants

The ARIC study¹⁹ is a prospective, community-based cohort study, which began in 1987-1989. The 15,792 ARIC study participants at inception were aged 45 to 64 years and were recruited from 4 US communities (Forsyth County, NC; Jackson, MS; Washington County, MD; suburbs of Minneapolis, MN). Participants provided written informed consent at each visit, and this study was approved by Institutional Review Boards at all participating institutions.

Relevant to this analysis, 6538 participants attended the visit 5 clinic exam during 2011 to 2013. Overall, the response rate among survivors for visit 5 was 65%. Visit 5 participants at all 4 sites were offered an echocardiogram. Details have been described previously.²⁰ In total, 5784 participants (88%) had left atrial function measures. Individuals missing LA strain data were significantly older with a higher comorbidity burden (Supplemental Table 1). Figure 1 shows the flow of participants included for this study. We excluded 237 of these participants for not having fasting blood measures and 198 for missing covariate data. Those identifying as other than Black or White (n=14) and Black participants from the Minneapolis and Washington County centers (n=21) were additionally excluded due to small numbers. Lastly, 248 participants with undiagnosed diabetes at visit 5 (i.e., no self-reported diagnosis or medication use at any visit) but whose serum glucose levels were ≥126 mg/dL or HbA1c levels ≥6.5% at visit 5 were also excluded from this analysis. This left 5075 participants included in the final analysis

Figure 1:

Flowchart of participants included in the analysis

Measurement of Glycemia

The HbA1c levels were measured in whole blood with the Tosoh G7 automated high-performance liquid chromatography (HPLC) analyzer (Tosoh Bioscience), standardized to the Diabetes Control and Complications Trial assay. Measurements of HbA1c from samples stored for up to 18 years are highly reliable when using state-of-the-art HPLC instruments, but with some bias introduced over time.²¹ The data were re-calibrated to account for this bias. Glucose was measured in serum using the hexokinase method on the Olympus AU400e analyzer (Beckman Coulter Inc). The laboratory intra-assay coefficient of variation based on blind duplicates at visit 5 was 1.3% for HbA1c and 5.7% for FG.

Assessment of Diabetes Status and Definitions

At visit 5, diabetes was present if a participant self-reported a physician diagnosis of diabetes or use of glucose-lowering medication. Self-reported diabetes in ARIC has been shown to be reliable and highly specific.²² Prediabetes was present if individuals without a history of diabetes had fasting glucose levels ≥100 mg/dL and ≤125 mg/dL or HgbA1c ≥5.7% and ≤6.4%. Among persons with a history of diabetes, we dichotomized glycemic control by HbA1c at 7% (53 mmol/mol); a treatment goal recommended by the American Diabetes Association Standards of Medical Care for many adults.²³ Diabetes duration was stratified according to newly diagnosed (within one year), short-term (less than the median duration (9.7 years) among those with diabetes of at least 1 year duration at baseline), and long-term (greater than the median duration among those with diabetes of at least 1 year duration at baseline)

Echocardiographic Measurements

Echocardiograms were performed using dedicated Philips iE33 Ultrasound systems with Vision 2011. Studies were acquired, stored digitally, and transferred to a secure service at the Echocardiography Reading Center (Brigham and Women’s Hospital, Boston, MA) from each field center. LA analysis was performed using a speckle tracking vendor-dependent software with an auto-strain algorithm (QLAB Advanced Quantification Software 13.0, Philips Ultrasound, Andover, MA), which identifies cardiac motion by tracking multiple chamber reference points over time. LA endocardial borders were automatically traced at the end-diastolic frame—defined by the QRS complex or the frame after mitral valve closure—of the apical 4-chamber views, and speckles were tracked during a cardiac cycle frame by frame. LA phasic function was estimated using peak strain during systole to assess reservoir function, early peak strain during diastole to assess conduit function, and late peak strain during diastole to assess contractile function. Strain indices were calculated as the mean of segments obtained, reported as absolute values, and expressed in percent. Intra-reader and inter-reader variability for LA reservoir strain were assessed in a sample of 40 randomly selected participants. The intraclass correlation coefficients for inter-reader and intra-reader variability were 0.91 and 0.98, respectively.^{24, 25} Vendor-specific normal reference ranges for these strain indices are not currently available. In prior ARIC study, reservoir, conduit, and contractile at or below 26.7%, 9.9%, and 13.5% respectively were associated with a higher incidence of study.²⁴ Similar cut-off values were associated with an increased risk of heart failure with preserved ejection fraction.²⁵

Covariates

Covariates included were age (years), sex, race-field center, education level, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), cigarette smoking (current/not current), anti-hypertensive medication use, estimated glomerular filtration rate (eGFR), C-reactive protein (CRP), atrial fibrillation (AF), coronary heart disease (CHD),²⁶ heart failure (HF),^{26, 27} and visit 5 echocardiogram-defined left ventricular mass index (LVMI), left atrial volume index (LAVI), and left ventricular filling pressures (E/e’). A full description can be found in the supplemental methods.

Statistical analysis

Descriptive statistics were computed for baseline characteristics according to tertiles of LA function measures. Multivariable linear regression was used to assess the associations of (1) diabetes status (diabetes and prediabetes vs. no diabetes), (2) diabetes control (HbA1c <7 vs. ≥7%), and (3) diabetes duration with LA function. Associations were adjusted as follows—Model 1: age, sex, race by clinic site, and educational attainment; Model 2: Model 1 + SBP, BMI, cigarette smoking, HF, CHD, AF, eGFR, CRP, and anti-hypertensive use; Model 3: Model 2 + LAVI, LVMI, and E/e’. We performed stratified analyses for age, race, and sex. Due to the known relationship between impaired LA function and both HF and AF, we also evaluated associations of diabetes status with LA function after excluding individuals with either of these conditions at V5.^2–5

In individuals without diabetes, we modeled fasting glucose and HbA1c using restricted cubic spline terms (5 knots located at 5^th, 25^th, 50^th, 75^th, and 95^th percentiles) to characterize their continuous associations with LA function. Distribution of the exposure variable was presented as the histogram under the spline. Analyses were performed using R version 4.2.0.

Results

The mean age for included participants was 75.5 years, 58% were women, 20% were Black and 32.3% had prevalent diabetes. Compared to participants in the lowest tertile of LA conduit strain, those in higher tertiles were younger, more likely to be female, have more education, less likely to be on anti-hypertensive medications, and have fewer cardiac comorbidities (HF, CHD, and AF) (Table 1). Baseline BMI and hs-CRP were lower while eGFR_cys was higher with higher conduit strain. Findings were similar for LA reservoir strain (Supplemental Table 2) and LA contractile strain (Supplemental Table 3).

Table 1.

Baseline characteristics of study participants according to tertiles of LA conduit strain, Atherosclerosis Risk in Communities Study, 2011-13 (n=5075) ^*

Characteristic	Left Atrial Conduit Strain (%)
	[0.06, 11.6] (n=1692)	(11.6, 16.6] (n=1695)	(16.6, 46.7] (n=1688)
Age, years	77.13 (5.32)	75.39 (4.83)	73.92 (4.70)
Male sex	757 (44.7)	692 (40.8)	658 (39.0)
Black race	348 (20.6)	345 (20.4)	324 (19.2)
Education ≥12 years	1410 (83.3)	1468 (86.6)	1538 (91.1)
BMI, kg/m2	28.82 (5.52)	28.57 (5.33)	28.04 (5.41)
Systolic BP, mmHg	131.29 (18.25)	131.13 (18.20)	128.20 (17.13)
Diastolic BP, mmHg	66.76 (11.27)	66.34 (10.66)	65.12 (9.94)
eGFRcys	65.35 (18.72)	69.12 (18.98)	73.09 (18.29)
hs-CRP, mg/dl	4.27 (8.96)	4.03 (7.64)	3.53 (5.99)
Anti-hypertensive use	1390 (82.2)	1289 (76.0)	1075 (63.7)
Smoking status
Never	90 (5.7)	96 (6.0)	109 (6.8)
Former	858 (54.7)	808 (50.8)	813 (50.6)
Current	620 (39.5)	686 (43.1)	684 (42.6)
Heart failure	324 (19.1)	193 (11.4)	91 (5.4)
Coronary heart disease	368 (21.7)	232 (13.7)	140 (8.3)
Atrial fibrillation	201 (11.9)	107 (6.3)	48 (2.8)

hs-CRP=high sensitivity C-reactive protein, eGFR_cys=estimated glomerular filtration rate_{cystatin C}

^*Continuous variables are expressed as mean (SD). Categorical variables are N (percent).

Table 2 shows the associations of diabetes status with LA function. In a model adjusted for demographics, CVD risk factors and prevalence, and echocardiographic measures, individuals with diabetes and prediabetes had 0.68% (95% confidence interval (CI): −1.11 to −0.25) and 0.47% (95% CI: −0.85 to −0.09) lower conduit strain compared to those without diabetes. Associations persisted after excluding individuals with HF or AF (Supplemental Table 4). Diabetes status was not independently associated with LA reservoir and contractile strain. In stratified analyses, however, males with diabetes and prediabetes had significantly worse reservoir and contractile strain compared to women (Supplemental Table 5). The only association that reached significant was for LA reservoir strain in males with diabetes.

Table 2:

Association of diabetes status with left atrial function, Atherosclerosis Risk in Communities Study, 2011-13 (n=5075, 1196 normal fasting glucose, 2239 prediabetes, 1640 diabetes)^*

LA strain	Diabetes status	Model 1†		Model 2‡		Model 3§
		Beta (95% CI)*	P-value	Beta (95% CI)*	P-value	Beta (95% CI)*	P-value
LA reservoir strain, %	Normal	Reference
	Prediabetes	−0.59 (−1.19, 0.01)	0.053	−0.17 (−0.74, 0.40)	0.558	−0.33 (−0.86, 0.20)	0.217
	Diabetes	−2.29 (−2.93, −1.64)	<0.001	−0.41 (−1.05, 0.24)	0.215	−0.43 (−1.03, 0.16)	0.156
LA conduit strain, %	Normal	Reference
	Prediabetes	−0.76 (−1.14, −0.38)	<0.001	−0.48 (−0.87, −0.09)	0.015	−0.47 (−0.85, −0.09)	0.016
	Diabetes	−1.81 (−2.21, −1.40)	<0.001	−0.89 (−1.33, −0.45)	<0.001	−0.68 (−1.11, −0.25)	0.002
LA contractile strain, %	Normal	Reference
	Prediabetes	0.19 (−0.26, 0.64)	0.404	0.33 (−0.11, 0.77)	0.139	0.17 (−0.25, 0.58)	0.429
	Diabetes	−0.42 (−0.91, 0.06)	0.087	0.52 (0.02, 1.01)	0.041	0.30 (−0.16, 0.76)	0.201

^*Beta-coefficient represents difference in LA strain (%) for diabetes and prediabetes groups compared to no diabetes group (referent)

^†Model 1: Age, Sex, Race by Clinic site, and Educational attainment

^‡Model 2: Model 1 + Systolic blood pressure, Body mass index, Smoking, Heart failure, Coronary heart disease, Atrial fibrillation, Estimated glomerular filtration rate cystatin C, high-sensitivity C-reactive protein, and anti-hypertensive use

^§Model 3: Model 2 + Left atrial volume index, Left ventricular mass index, Left ventricular filling pressures (E/e’)

Amongst individuals with diabetes, those with HbA1c ≥7% had a 1.05% worse conduit function compared to those with levels <7% (95% CI: −1.63, −0.48) (Table 3). Those with poorer HbA1c control had 0.73% better contractile strain (95% CI: 0.08, 1.38). No significant interactions were observed in analyses stratified by sex, race, or age (Supplemental Table 6). Findings were similar when HbA1c levels were evaluated continuously with the exception that higher levels were also associated with worse reservoir strain (Supplemental Figure 1). No significant associations were observed for diabetes duration with any of the LA function parameters (Supplemental Table 7).

Table 3:

Association of diabetes control with left atrial function, Atherosclerosis Risk in Communities Study, 2011–13 (n=1640 participants with diabetes, 1186 participants with HgbA1c<7%)^*

	Model 1†		Model 2‡		Model 3§
	Beta (95% CI)*	p-value	Beta (95% CI)*	p-value	Beta (95% CI)*	p-value
LA reservoir strain, %	−1.38 (−2.33, −0.42)	0.005	−0.46 (−1.36, 0.44)	0.318	−0.46 (−1.29, 0.37)	0.281
LA conduit strain, %	−1.69 (−2.25, −1.12)	<0.001	−1.31 (−1.90, −0.73)	<0.001	−1.05 (−1.63, −0.48)	<0.001
LA contractile strain, %	0.43 (−0.30, 1.16)	0.247	0.97 (0.27, 1.67)	0.007	0.73 (0.08, 1.38)	0.027

^*Beta-coefficient represents difference in LA strain (%) between HgbA1c<7% (referent) and HgbA1c≥7% in individuals with diabetes

^†Model 1: Age, Sex, Race by Clinic site, and Educational attainment

^‡Model 2: Model 1 + Systolic blood pressure, Body mass index, Smoking, Heart failure, Coronary heart disease, Atrial fibrillation, Estimated glomerular filtration rate cystatin C, high-sensitivity C-reactive protein, and anti-hypertensive use

^§Model 3: Model 2 + Left atrial volume index, Left ventricular mass index, Left ventricular filling pressures (E/e’)

Amongst individuals without diabetes, the only significant association observed for higher fasting glucose was with worse LA conduit function (Figure 2). Higher HbA1c was not associated with any of the three LA function parameters (Figure 3).

Figure 2:

Association of fasting glucose with left atrial function in individuals without diabetes, Atherosclerosis Risk in Communities Study, 2011-13 (n=3435, 2239 prediabetes, 1196 normal fasting glucose)*

*Model adjusted for Age, Sex, Race by Clinic site, Educational attainment, Systolic blood pressure, Body mass index, Smoking, Heart failure, Coronary heart disease, Atrial fibrillation, Estimated glomerular filtration rate cystatin C, high-sensitivity C-reactive protein, Anti-hypertensive use, Left atrial volume index, Left ventricular mass index, and Left ventricular filling pressures (E/e’)

Figure 3:

Association of HgbA1c with left atrial function in individuals without diabetes, Atherosclerosis Risk in Communities Study, 2011-13 (n=3435, 2239 prediabetes, 1196 normal fasting glucose)*

*Model adjusted for Age, Sex, Race by Clinic site, Educational attainment, Systolic blood pressure, Body mass index, Smoking, Heart failure, Coronary heart disease, Atrial fibrillation, Estimated glomerular filtration rate cystatin C, high-sensitivity C-reactive protein, Anti-hypertensive use, Left atrial volume index, Left ventricular mass index, and Left ventricular filling pressures (E/e’)

Limitations

This was a cross-sectional study, and we were not able to determine whether glycemic status was associated with changes in LA function. We cannot infer causality or exclude residual confounding. Participants were older individuals (mean age 75 years) and findings may not be generalizable to younger adults. The study was also limited by survival bias as it included only those healthy enough to attend the clinic visit and undergo echocardiography, which occurred over twenty years after initial recruitment.

Discussion

In an older community-based cohort of Black and White men and women, we observed that individuals with prevalent diabetes and, to a lesser degree, prediabetes were associated with worse LA conduit strain compared to individuals with normal fasting glucose levels. Poorer diabetes control amongst individuals with diabetes, and higher fasting glucose amongst those without diabetes, were also associated with both worse LA conduit strain.

Prior single-center studies in much smaller populations support the results of this study. Amongst 218 Macedonian individuals with HFpEF, global average peak atrial longitudinal strain (PALS) and peak atrial contraction strain were significantly lower for the 108 participants with diabetes compared to the 110 participants without. Similar findings have also been demonstrated in individuals without cardiovascular disease. In a study of 150 Italian individuals with normal LA volumes and normal LV systolic function average global PALS, atrial longitudinal strain during early diastole, and atrial longitudinal strain during late diastole were all significantly lower in individuals with diabetes compared to those without.^{12, 14} Importantly, differences for late diastolic atrial longitudinal strain were of smaller magnitude compared to the other strain parameters. The mean age of participants for these studies was between 60 to 65 years of age. In a separate study of 170 Serbian individuals, LA reservoir and conduit function worsened while booster pump function improved, from healthy controls to individuals with prediabetes, and to individuals with diabetes.¹⁵ These studies stand in some contrast to a smaller investigation of 80 individuals without hypertension or clinical cardiovascular disease, 40 of whom had diabetes. LA total emptying fraction and active emptying fraction were significantly lower in those with diabetes, reflecting depressed LA reservoir and pump functions, whereas no significant differences between groups were found for LA conduit function.¹⁷

Our findings build upon the established literature in many important ways. Our study included a substantially larger population, had a comprehensive array of LA function measures, and adjusted results for multiple possible confounders, thereby allowing for a more definitive assessment of the cross-sectional association between diabetes and LA function. More importantly, we demonstrated that hyperglycemia was associated with progressively worse LA function across the entire spectrum of glycemic control, including those with normal fasting glucose, with glucose intolerance, and with diagnosed diabetes. Additionally, we found that higher glucose levels were associated with worse LA function amongst those without diabetes. Finally, our results suggest that poorer glucose regulation specifically affects LA conduit function compared to other phases of LA function.

A variety of mechanisms beyond atrial enlargement itself have been proposed to explain the relationship between impaired glucose regulation and reduced LA function. Subendocardial fibrosis is often seen in the presence of diabetes, and this can lead to decreased wall elasticity, which is important for regulating atrial reservoir function.²⁸ Higher blood pressure induced by insulin resistance has been shown to adversely impact LA mechanics.^{14, 29} Electrophysiologic remodeling of the LA (such as conduction and refractory disorders) induced by diabetes can also have a direct negative impact on atrial phasic function.³⁰ Autonomic remodeling seen in diabetes and that occurs due to autonomic dysfunction with increased sympathetic activation can also alter electrical properties in the LA and impair function.³¹ Increased oxidative stress and inflammation seen in diabetes can result in deformational changes in the LA.¹¹ Finally, increased activation of the renin-angiotensin-aldosterone system can directly impair LA function via pathways described above such as enhanced atrial fibrosis, higher pressures, and increased oxidative stress.¹¹

It is unclear why associations were significant for LA conduit function but not for reservoir function. Our findings are consistent with prior aforementioned studies.^{12, 14, 15} Conduit function is mainly modulated by LV relaxation and, to a lesser degree, atrial compliance. Conduit and booster function are mainly determined by LA compliance and intrinsic LA contractility, respectively. The BErlin Female RIsk evaluation trial evaluated LA function in preclinical LVDD and found that both LA reservoir and conduit function progressively decreased with worsening LVDD while booster function augmented in grade 1 LVDD before being reduced in patients with grade 2 LVDD.³² Similar findings have been observed with progression of hypertensive heart disease.^33–35 The positive point estimates for LA contractility function observed with diabetes (non-significant) and poorer glycemic control (significant) may be due to the low prevalence of advanced LVDD or HF in our population (HF prevalence, 12%)

In conclusion, prevalent diabetes and prediabetes were both associated with worse LA conduit function compared to individuals with normal fasting glucose levels. Poorer diabetes control amongst individuals with diabetes, and higher fasting glucose amongst those without diabetes, were also associated with worse LA conduit function. Findings suggest impaired glucose may adversely affect LA function. Considering the established relationship between LA dysfunction, particularly in those with diabetes, and an increased risk of HF, AF, and CHD, there is potential benefit in. being able to better recognize these individuals with both LA dysfunction and diabetes well before the onset of such cardiac diseases.^5–7 It may provide an opportunity to more aggressively manage cardiovascular risk factors and institute lifestyle measure to help reduce disease incidence. Additional research is needed in prospective cohorts as well as studies that identify that can better individuals with impaired glucose who may benefit most from targeted screening for left atrial dysfunction before this can be applied clinically.