Measures of Left Ventricular Diastolic Function and Cardiorespiratory Fitness According to Glucose Metabolism Status: The Maastricht Study

Abstract

Background

This cross‐sectional study evaluated associations between structural and functional measures of left ventricular diastolic function and cardiorespiratory fitness (CRF) in a well‐characterized population‐based cohort stratified according to glucose metabolism status.

Methods and Results

Six hundred seventy‐two participants from The Maastricht Study (mean±SD age, 61±9 years; 17.4% prediabetes and 25.4% type 2 diabetes mellitus) underwent both echocardiography to determine left atrial volume index, left ventricular mass index, maximum tricuspid flow regurgitation, average e′ and E/e′ ratio; and submaximal cycle ergometer test to determine CRF as maximum power output per kilogram body mass. Associations were examined with linear regression adjusted for cardiovascular risk and lifestyle factors, and interaction terms. After adjustment, in normal glucose metabolism but not (pre)diabetes, higher left atrial volume index (per 1 mL/m²), left ventricular mass index (per 1 g/m^2.7), maximum tricuspid regurgitation flow (per 1 m/s) were associated with higher CRF (maximum power output per kilogram body mass; β in normal glucose metabolism 0.015 [0.008–0.023], P _interaction (pre)diabetes <0.10; 0.007 [−0.001 to 0.015], P _interaction type 2 diabetes mellitus <0.10; 0.129 [0.011–0.246], P _interaction >0.10; for left atrial volume index, left ventricular mass index, maximum tricuspid regurgitation flow, respectively). Furthermore, after adjustment, in all individuals, higher average E/e′ ratio (per unit), but not average e′, was associated with lower CRF (normal glucose metabolism −0.044 [−0.071 to −0.016]), P _interaction >0.10).

Conclusions

In this population‐based study, structural and functional measures of left ventricular diastolic function were independently differentially associated with CRF over the strata of glucose metabolism status. This suggests that deteriorating left ventricular diastolic function, although of small effect, may contribute to the pathophysiological process of impaired CRF in the general population. Moreover, the differential effects in these structural measures may be the consequence of cardiac structural adaptation to effectively increase CRF in normal glucose metabolism, which is absent in (pre)diabetes.

Nonstandard Abbreviations and Acronyms

CRF

cardiorespiratory fitness

GMS

glucose metabolism status

LAVI

left atrial volume index

LVMI

left ventricular mass index

NGM

normal glucose metabolism

T2D

type 2 diabetes mellitus

W_max

estimated maximum power output

Clinical Perspective

What Is New?

• In this population‐based study, structural and functional measures of left ventricular diastolic function were independently differentially associated with cardiorespiratory fitness over a strata of glucose metabolism status that suggests that deteriorating left ventricular diastolic function may contribute to the pathophysiological process of impaired cardiorespiratory fitness in the general population.

• In addition, structural adaptations of the left atrium and left ventricle may be different in (pre)diabetes as compared with normal glucose metabolism, indicating an absence of physiological cardiac structural adaptation in (pre)diabetes possibly explained by a diabetic cardiomyopathy.

What Are the Clinical Implications?

• Our findings emphasize the need for future prospective studies that examine the usefulness of left ventricular diastolic function as potential targets in preventive and therapeutic strategies for impaired cardiorespiratory fitness in the general population and in (pre)diabetes, and the exact cardiovascular mechanisms to better understand the cardiac adaptation process in (pre)diabetes (eg, mechanisms of diabetic cardiomyopathy) in the early development of diastolic dysfunction.

Left ventricular (LV) diastolic function is an important determinant of cardiorespiratory fitness (CRF) in individuals with established cardiovascular disease (CVD) or CVD risk factors.¹, ² The latter may be especially so for individuals with type 2 diabetes mellitus (T2D) because of the existence of a hyperglycemia‐driven diabetic cardiomyopathy.³ LV diastolic function may be particularly negatively influenced by T2D and thus its association with CRF, because of synergy between diabetic cardiomyopathy and other CVD risk factors.⁴ However, the concept of diabetic cardiomyopathy is under debate,⁵ and it is currently unclear whether this concept also is present in individuals with prediabetes.⁶

To what extent the link between LV diastolic function and CRF expands to the general population at large, including individuals with a is less clear.⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³ Population‐based data focusing solely on the association between LV diastolic function and CRF in individuals with (pre)diabetes do not exist. Population‐based studies on the association between LV diastolic function and CRF⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³ in the general population are heterogeneous in their results. Moreover, these studies were relatively small,⁷, ⁸, ¹⁰, ¹¹, ¹², ¹³ included only a limited number of individuals with prediabetes,⁹, ¹², ¹³ and/or did not adequately adjust for (multiple) confounders (Table S1).⁷, ⁸, ¹¹, ¹² Furthermore, the pathophysiology of LV diastolic function is complex and may give inhomogeneous responses because of the interplay between different compensatory mechanisms.¹⁴ For instance, unfavorable alterations in LV myocardial relaxation may be compensated for by an increased atrial contribution to the LV filling (at higher filling pressures) without any pathological changes in LV or left atrial volume.¹⁵, ¹⁶, ¹⁷

In view of these considerations, the aim of the present study was to evaluate (1) the associations between structural and functional measures of LV diastolic function¹⁴ and CRF, measured as estimated maximum power output adjusted for body mass (W_max per kilogram), in a well‐characterized population‐based cohort stratified according to glucose metabolism status (GMS); and (2) whether these associations differed between individuals with normal glucose metabolism (NGM) and (pre)diabetes.

Methods

Study Population and Design

We used data from The Maastricht Study, an observational prospective population‐based cohort study. The rationale and methodology have been described previously.¹⁸ In brief, the study focuses on the cause, pathophysiology, complications, and comorbidities of T2D and is characterized by an extensive phenotyping approach. Eligible for participation were all individuals aged between 40 and 75 years living in the southern part of the Netherlands. Participants were recruited through mass media campaigns as well as from the municipal registries and the regional Diabetes Patient Registry via mailings. Recruitment was stratified according to known T2D status, with an oversampling of individuals with T2D, for reasons of efficiency. Eight hundred sixty‐six participants, who completed the baseline survey between November 2010 and March 2012, were included. To augment statistical power, another random sample of 218 participants was added who had completed the baseline survey between April 2012 and April 2013 (following the same recruitment strategy). The examinations of each participant were performed within a time window of 3 months. The study has been approved by the institutional medical ethical committee (NL31329.068.10) and the Minister of Health, Welfare and Sports of the Netherlands (Permit 131088‐105234‐PG) and follows the Declaration of Helsinki. All participants gave written informed consent. The present study was reported as per the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement for observational cohort studies. Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to The Maastricht Study Management Team at research.dms@mumc.nl.

From the 1084 individuals in this extended sample of the study population, echocardiography was obtained in 933 individuals. Individuals were excluded on the basis of exclusion criteria (N=7), missing data on CRF (N=154), missing data on covariates (N=62), and missing data on 2‐dimensional echocardiography variables (N=38) or tissue Doppler imaging data (N=71). This resulted in 672 (2‐dimensional echocardiography study population) and 639 (tissue Doppler imaging echocardiography study population) individuals eligible for the current analyses (Figure S1).

Echocardiography: Measures of LV Diastolic Function

Echocardiograms were obtained according to a standardized protocol consisting of 2‐dimensional, M‐mode, color flow Doppler, pulsed and continuous wave Doppler, and tissue Doppler recordings with use of standard echo equipment (Vivid E9 with 2.5–3.5 MHz and 4 V transducer; GE Vingmed). All recordings were digitally stored and analyzed off‐line (EchoPAC PC, version 112; GE Healthcare) by 4 researchers blinded to GMS and other covariates.

LV diastolic function was assessed according to the 2016 guidelines with the use of both structural and functional echocardiographic variables, namely left atrial volume index (LAVI), LV mass index (LVMI), average E/e′ ratio, average e′, and maximum tricuspid regurgitation flow.¹⁴ Higher values of these measures apart from average e′ indicate deteriorating LV diastolic function in patients with heart failure, which already may be seen in the general population.¹, ¹⁴ Biplane end‐systolic left atrial volume was measured and indexed to body surface area. LV mass was calculated with the use of end‐diastolic LV diameter, interventricular septum diameter, and end‐diastolic LV posterior wall thickness, and indexed by height^2.7. Maximal tricuspid valve regurgitation velocity was measured with continuous wave Doppler recordings. Average E/e′ ratio was calculated from the mitral peak flow velocity of the filling wave during early inflow (E) obtained with pulsed‐wave Doppler and the septal and lateral early (e′) diastolic longitudinal velocity obtained with pulsed Doppler tissue echocardiography. Further details on echocardiographic procedures and variables including reproducibility are provided in Data S1.

Submaximal Cycle Ergometer Test: CRF

The submaximal cycle ergometer test to determine CRF in W_max was performed as described previously.¹⁹ As an objective measure of CRF, estimated W_max per kilogram was used.²⁰, ²¹ W_max was estimated from a graded submaximal exercise protocol performed on a cycle ergometer system (CASETM version 6.6 in combination with e‐bike; GE Healthcare, Milwaukee, WI). Exclusion criteria for the submaximal cycle ergometer test were: having suffered from CVD 3 months before the ergometer test, having a resting ECG with previously unknown abnormalities, having severe hypertension (systolic blood pressure ≥180 and/or diastolic blood pressure ≥110), or being in the possession of an implantable cardioverter‐defibrillator/pacemaker. Further details on the protocol and estimation of W_max are provided in Data S1.

Covariates

We assessed glucose metabolism, clinical characteristics of patients, lipid profile, markers of renal function, educational level, and self‐reported physical activity as described previously.¹⁸, ¹⁹, ²², ²³ Further details on the covariates are provided in Data S1.

Statistical Analysis

Descriptive statistics are presented as mean±standard deviation, median (interquartile range), or frequency (percentage), as appropriate; all variables were checked for the assumption of normality). Comparisons of population characteristics between groups were made by use of ANOVA for continuous variables, log‐transformed if necessary, or by χ² test for dichotomous or categorical variables.

Associations between structural and functional measures of LV diastolic function and CRF were investigated with multivariable linear regression analyses in the study population stratified according to GMS. The analyses were adjusted for sex, age, and height (Model 1), and additionally adjusted for prior CVD, smoking status, total cholesterol/high‐density lipoprotein ratio, triglycerides, use of lipid‐modifying medication, office systolic pressure, use of antihypertensive medication, estimated glomerular filtration rate, albuminuria, health status, and alcohol use (Model 2). Because of its role as a possible mediator or ascending proxy, waist was added in a separate model (Model 3), because a model including waist might be at risk of overadjustment.²⁴ In addition, we investigated whether or not these associations differed among individuals with different GMS by adding interaction terms in Model 2.

Furthermore, we conducted several additional analyses to test the robustness of our results. First, we repeated the analyses with diastolic function according to 2016 guidelines as categorical determinant,¹⁴ and in the total study population. Second, we replaced LAVI with nonindexed left atrial volume, and LVMI indexed by height with LVMI indexed by body surface area, or by LV mass, to test the influence of the index used. Third, we replaced CRF as W_max per kilogram by CRF as the percentage of the predicted W_max to put the results in a clinical perspective.²⁵ Fourth, we additionally adjusted Model 2 for moderate‐to‐vigorous physical activity.²⁶ Fifth, to restrict the analyses to subclinical disease, we repeated the analyses excluding individuals with prior coronary heart disease, current atrial fibrillation or flutter, wall motion abnormalities, significant valvular dysfunction, or functional mobility limitations. Sixth, the analyses were repeated with the replacement of office systolic pressure in Model 2 by office diastolic pressure,²⁷ their 24‐hour equivalents,²⁸ the presence of hypertension, and with additional adjustment for renin‐angiotensin system inhibitors or β‐blockers. Seventh, we replaced waist in Model 3 by body mass index or weight. Last, we used interaction terms added to Model 2, additionally adjusted for the interaction between measures of LV diastolic function and GMS, to examine whether the investigated associations were modified by sex.

All statistical analyses were performed using IBM SPSS Statistics version 23 (IBM, Armonk, NY). A 2‐sided P<0.05 was considered statistically significant, except for interaction terms, where a P<0.10 was used. Multicollinearity was assessed by collinearity diagnostics (ie, tolerance <0.2 and/or variance inflation factor >10). Because of the observational nature of our study, we made no corrections for multiple comparisons in our analyses.²⁹

Results

Characteristics of the Study Population

The study population with a mean±SD age of 61±9 years consisted of 366 (57.3%), 111 (17.4%), and 162 (25.4%) individuals with NGM, prediabetes, and T2D, (pre)diabetes were older, more often men, had higher body mass index, lower high‐density lipoprotein, higher triglycerides, lower estimated glomerular filtration rate, and less physical activity. In addition, they more often suffered from prior CVD, hypertension, albuminuria, and mobility limitations, and more frequently used antihypertensive and lipid‐modifying medication (Table 1).

Table 1.

General Characteristics of the Tissue Doppler Imaging Echocardiography Study Population According to Glucose Metabolism Status

	Normal Glucose Metabolism, n=366	Prediabetes, n=111	Type 2 Diabetes Mellitus, n=162	P Value
Demographics
Men, n (%)	43	61	70	<0.001
Age, y	60±8	62±7	63±7	<0.001
Educational level, low/middle/high, %	10.7/39.2/50.1	17.1/44.1/38.7	25.9/43.2/30.9	<0.001
Prior cardiovascular disease, %	10	13	19	0.004
Prior coronary heart disease, %	3	6	9	0.005
Current atrial fibrillation or flutter, %*	0.0	0.0	1.3	0.030
Blood pressure
Office systolic pressure, mm Hg	130±16	139±16	145±17	<0.001
Office diastolic pressure, mm Hg	75±10	79±10	78±10	<0.001
24‐hour systolic pressure, mm Hg†	116±11	121±12	122±11	<0.001
24‐hour diastolic pressure, mm Hg†	74±7	74±8	73±7	0.929
Hypertension, %	36	60	79	<0.001
Metabolic variables
BMI, kg/m2	25.3±3.3	27.5±3.4	28.6±3.4	<0.001
Waist, cm	90.4±10.5	98.2±10.6	102.7±10.1	<0.001
Total cholesterol, mmol/L	5.59±0.99	5.46±1.09	4.49±0.92	<0.001
HDL, mmol/L	1.54±0.49	1.38±0.36	1.19±0.32	<0.001
LDL, mmol/L	3.54±0.87	3.39±0.99	2.53±0.78	<0.001
Triglycerides, mmol/L	1.01 [0.75–1.40]	1.27 [0.88–1.78]	1.63 [1.16–2.08]	<0.001
Total‐to‐HDL‐cholesterol ratio	3.95±1.31	4.17±1.18	3.93±0.94	0.891
HbA1C, in %‡	5.5±0.3	5.8±0.4	6.7±0.9	<0.001
Fasting plasma glucose, mmol/L	5.2±0.4	6.0±0.5	7.6±1.7	<0.001
Kidney function
eGFR, mL/min per 1.73 m2	91.0±13.7	85.1±14.6	86.4±16.2	<0.001
Albuminuria, %	3.3	4.5	17.3	<0.001
Lifestyle variables
Smoking status: never/former/current, %	38.9/45.5/15.6	28.8/61.3/9.9	27.8/57.4/14.8	0.101
Alcohol use: no/low/high, %	13.2/56.2/30.7	14.4/52.3/33.3	26.5/48.1/25.3	0.006
Moderate to vigorous physical activity, h/wk§	5.5 [3.0–9.0]	4.5 [2.7–7.1]	3.6 [2.3–10.0]	<0.001
Medication
Antihypertensive medication, %	20	40	63	<0.001
RAS inhibitors, %	13	30	51	<0.001
β‐blockers, %	6	19	33	<0.001
Diuretics, %	6	16	19	<0.001
Calcium antagonists, %	3	6	13	<0.001
Oral antidiabetics and/or insulin use	…	…	74	…
Lipid‐modifying medication, %	14	36	75	<0.001
Cardiorespiratory fitness, Wmax	168.9±48.6	168.8±46.1	158.1±42.7	0.023
Cardiorespiratory fitness adjusted for body mass, Wmax/kg	2.28±0.55	2.08±0.48	1.89±0.49	<0.001
Predicted cardiorespiratory fitness, predicted Wmax	149.2±49.5	153.5±54.8	158.4±47.9	0.145
Cardiorespiratory fitness, % of predicted Wmax	118.8±29.2	120.7±43.1	106.7±31.2	<0.001
Mobility limitation, %||	12	26	30	<0.001

Data are presented as mean±SD, median [interquartile‐range], or frequency (%) as appropriate. Data present the tissue Doppler imaging echocardiography population for regression Models 1 to 3. Linear trend was tested with ANOVA or χ² test as appropriate. BMI indicates body mass index; eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin A1c; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; NGM, normal glucose metabolism; RAS, renin angiotensin system; T2D, type 2 diabetes mellitus; and W_max, estimated maximum power output.

Numbers for specific variables (total, NGM/prediabetes/T2D) are:

^*Current atrial fibrillation or flutter 610, 350/103/157.

^† 24‐hour blood pressure measurements 590, 340/101/149.

^‡ HbA1c 637, 365/110/162.

^§ Moderate to vigorous physical activity 554, 322/94/138.

^|| Mobility limitation 636, 364/110/162.

General characteristics of the tissue Doppler imaging echocardiography study population stratified according to GMS and according to tertiles of average E/e′ ratio are given in Tables 1 and 2 and Tables S2 and S3, respectively. Individuals with (pre)diabetes had higher average E/e′ ratio and higher LVMI indicating deteriorating LV diastolic function, more frequent abnormal diastolic function grade, and lower CRF (Tables 1 and 2). The study population in which 2‐dimensional echocardiography was available (NGM N=380 [56.5%], prediabetes N=115 [17.1%], and T2D N=177 [26.3%]) overlapped for 94% with the tissue Doppler imaging study population and was comparable with regard to age, sex, and cardiometabolic risk factors (Figure S1 and Table S4).

Table 2.

Echocardiographic Characteristics of the Tissue Doppler Imaging Echocardiography Study Population According to Glucose Metabolism Status

	Normal Glucose Metabolism, n=366	Prediabetes, n=111	Type 2 Diabetes Mellitus, n=162	P Value
Measures LV diastolic function
Average E/e′ ratio	7.6±1.9	8.6±2.4	8.9±2.3	<0.001
Average e′, cm/s	9.5±2.2	8.2±2.0	8.0±1.8	<0.001
Maximum tricuspid regurgitation flow, m/s*	1.95±0.42	1.80±0.56	1.90±0.51	0.113
Left atrial volume index, mL/m2 †
Total	30.2±6.3	30.0±7.0	29.8±6.8	0.486
Men	31.1±6.6	30.7±8.0	30.2±6.9	0.321
Women	29.6±6.0	28.8±5.0	28.9±6.7	0.376
LV mass index, g/m2.7 ‡
Total	28.7±6.2	31.5±6.8	31.3±6.7	<0.001
Men	29.8±6.7	31.5±6.7	31.3±7.2	0.077
Women	27.9±5.7	30.9±7.1	32.3±5.5	<0.001
LV mass index, g/m2 ‡
Total	65.2±13.6	65.9±14.0	67.4±14.5	0.047
Men	69.9±15.1	71.5±13.6	69.3±15.5	0.800
Women	61.5±11.1	63.7±13.4	62.9±10.8	0.326
LV diastolic function according to 2016 guidelines (normal, indeterminate, abnormal), n (%)	162/162/42 (44.3/44.3/11.5)	25/67/19 (22.5/60.4/17.1)	38/92/32 (23.5/56.8/19.8)	<0.001
Systolic LV function
LV ejection fraction, %§	60.7±2.5	60.2±2.6	59.7±3.5	0.001
S' septal, cm/s	7.5±1.3	7.2±1.3	7.5±1.7	0.931
S' lateral, cm/s	8.8±2.0	8.7±2.1	8.5±1.9	0.031
Other measures LV diastolic function
Early peak velocity, m/s	0.68±0.15	0.66±0.14	0.67±0.14	0.540
Active peak velocity, m/s||	0.66±0.15	0.72±0.16	0.73±0.15	<0.001
E/A ratio||	1.04 [0.87–1.27]	0.90 [0.76–1.08]	0.91 [0.78–1.10]	<0.001
Deceleration time E‐peak, ms	190±34	200±35	201±36	<0.001
Isovolumetric relaxation time, ms¶	94±20	100±23	95±21	0.200
S/D ratio#	1.38±0.31	1.43±0.35	1.47±0.31	0.004
e′ septal, cm/s	8.4±2.1	7.0±1.7	7.1±1.7	<0.001
a' septal, cm/s	9.8±1.8	9.8±1.7	10.1±1.9	0.166
e′ lateral, cm/s	10.7±2.5	9.4±2.6	8.9±2.2	<0.001
a' lateral, cm/s	10.6±2.4	11.2±2.4	11.3±2.3	0.001
Wall motion abnormalities, n yes (%)	0 (0.0)	1 (0.9)	3 (1.9)	0.012
Valvular dysfunction, moderate or severe, n (%)	23 (6.3)	7 (6.3)	8 (4.9)	0.573

Data are presented as mean±SD, median [interquartile range], or frequency (%) as appropriate. Data present the tissue Doppler imaging echocardiography population for regression Models 1 to 3. Linear trend was tested with ANOVA or χ² test as appropriate. E/A indicates peak flow velocity E/peak flow velocity A; LV, left ventricular; NGM, normal glucose metabolism; S/D, systolic/diastolic pulmonary peak inflow velocity; and T2D, type 2 diabetes mellitus.

Numbers for specific variables (total, NGM/prediabetes/T2D) are:

^*Maximum tricuspid regurgitation flow 636, 366/110/160.

^† Left atrial volume index 637, 366/109/162, men 339, 159/66/114, women 298, 207/43/48.

^‡ LV mass index 634, 363/110/161, men 338, 157/68/113, women 296, 206/42/48.

^§ LV ejection fraction 634, 364/109/161.

^|| Active peak velocity and E/A ratio 636,366/110/160.

^¶ Isovolumetric relaxation time 634, 362/110/162.

^# S/D ratio 634, 364/111/159.

Individuals in the middle and highest tertile of the average E/e′ ratio, as compared with the lowest tertile, were older, had higher body mass index, lower high‐density lipoprotein, higher triglycerides, and lower estimated glomerular filtration rate. In addition, they more often had T2D, prior CVD, hypertension, and albuminuria, more frequently used antihypertensive and lipid‐modifying medication, and had a lower CRF (Table S1).

Individuals with missing values were older, more often had T2D, had a worse cardiovascular risk profile, were less physically active, had higher average E/e′ ratio and lower CRF. Within the strata of GMS, similar patterns were seen, although to a lesser extent, and somewhat more pronounced in individuals with T2D (Tables S4 and S5).

Associations Between Measures of LV Diastolic Function and CRF

Table 3 and the Figure show the associations between measures of LV diastolic function and CRF in W_max per kilogram in individuals with NGM, prediabetes, and T2D.

Table 3.

Associations Between Measures of LV Diastolic Function and Cardiorespiratory Fitness

	Model	Normal Glucose Metabolism, N=380/366			Prediabetes, N=115/111			Type 2 Diabetes Mellitus, N=177/162
		B	95% CI	P Value	B	95% CI	P Value	B	95% CI	P Value
Average E/e′ ratio	1	−0.057	−0.084 to −0.030	<0.001	−0.047	−0.084 to −0.011	0.012	−0.040	−0.075 to −0.006	0.021
	2	−0.044	−0.071 to −0.016	0.002	−0.030	−0.072 to 0.012	0.156	−0.037	−0.072 to −0.001	0.043
	3	−0.033	−0.060 to −0.007	0.014	−0.015	−0.055 to 0.025	0.450	−0.028	−0.059 to 0.003	0.077
Average e′, cm/s	1	0.035	0.008 to 0.062	0.012	0.060	0.014 to 0.106	0.011	0.013	−0.032 to 0.057	0.570
	2	0.013	−0.016 to 0.042	0.388	0.040	−0.016 to 0.095	0.158	−0.010	−0.057 to 0.037	0.662
	3	0.009	−0.018 to 0.037	0.501	0.028	−0.023 to 0.079	0.280	−0.014	−0.055 to 0.026	0.487
Maximum tricuspid regurgitation flow, m/s	1	0.166	0.049 to 0.283	0.005	0.032	−0.128 to 0.191	0.694	0.034	−0.121 to 0.190	0.665
	2	0.137	0.023 to 0.252	0.019	−0.054	−0.213 to 0.134	0.653	−0.021	−0.175 to 0.133	0.785
	3	0.114	0.007 to 0.222	0.037	−0.064	−0.218 to 0.090	0.411	−0.021	−0.153 to 0.111	0.754
Left atrial volume index, mL/m2	1	0.017	0.010 to 0.025	<0.001	−0.003	−0.015 to 0.010	0.674	0.003	−0.008 to 0.014	0.516
	2	0.015	0.008 to 0.022	<0.001	−0.007	−0.021 to 0.008	0.357*	0.001	−0.009 to 0.012	0.804*
	3	0.014	0.007 to 0.021	<0.001	−0.006	−0.018 to 0.007	0.358	−0.002	−0.011 to 0.007	0.618
LV mass index, g/m2.7	1	0.005	−0.003 to 0.013	0.224	−0.002	−0.015 to 0.011	0.722	−0.009	−0.020 to 0.002	0.097
	2	0.009	0.001 to 0.017	0.035	0.001	−0.013 to 0.015	0.857	−0.006	−0.017 to 0.005	0.259*
	3	0.015	0.008 to 0.023	<0.001	0.009	−0.004 to 0.022	0.166	0.001	−0.009 to 0.010	0.859
Diastolic function 2016 guidelines
Indeterminate	1	0.009	−0.104 to 0.123	0.872	−0.204	−0.408 to 0.001	0.051	−0.156	−0.339 to 0.027	0.095
	2	0.050	−0.061 to 0.160	0.374	−0.208	−0.422 to 0.005	0.055	−0.115	−0.302 to 0.073	0.229
	3	0.095	−0.009 to 0.199	0.074	−0.144	−0.345 to 0.057	0.157	−0.064	−0.227 to 0.100	0.441
Abnormal	1	0.100	−0.061 to 0.160	0.055	−0.115	−0.386 to 0.157	0.404	−0.110	−0.342 to 0.122	0.350
	2	0.165	−0.004 to 0.335	0.055	−0.053	−0.329 to 0.224	0.706	−0.065	−0.301 to 0.170	0.229
	3	0.178	0.019 to 0.337	0.028	−0.022	−0.280 to 0.235	0.862	−0.056	−0.260 to 0.149	0.590

N=672 or 639 for the 2‐dimensional or tissue Doppler imaging echocardiography study population, respectively. The unstandardized regression coefficients (B) represent the difference in cardiorespiratory fitness in W_max per kilogram per 1‐unit higher level of measure of diastolic function, and for diastolic function according to 2016 guidelines vs normal diastolic function. Model 1: age, sex, height. Model 2: Model 1+prior cardiovascular disease, smoking status, alcohol use, total‐to‐HDL‐cholesterol ratio, triglycerides, use of lipid‐modifying medication, estimated glomerular filtration rate, health status, office systolic pressure, use of antihypertensive medication, albuminuria. Model 3: Model 2+waist. HDL indicates high‐density lipoprotein; LV, left ventricular; and W_max, estimated maximum power output.

^*P_interaction <0.07, represents the P value of the interaction effect between measures of diastolic function and prediabetes as compared with normal glucose metabolism in the association with cardiorespiratory fitness.

Figure 1. Associations between measures of left ventricular (LV) diastolic function and cardiorespiratory fitness.

The standardized regression coefficients represent the standardized difference in cardiorespiratory fitness per standard deviation higher measure of LV diastolic function in Model 2. Higher measures of LV diastolic function, apart from average e′, indicate deteriorating LV diastolic function. Model 2: adjusted for age, sex, height, prior cardiovascular disease, smoking status, alcohol use, lipids, lipid medication, estimated glomerular filtration rate, health status, office systolic pressure, antihypertensive medication, albuminuria. *P _interaction <0.07 represents the P value of the interaction effect between measures of diastolic function and (pre)diabetes as compared with normal glucose metabolism in the association with cardiorespiratory fitness. Circle represents the standardized regression coefficient of E/e'‐ratio; hexagram represents the standardized regression coefficient of Average e'; square represents the standardized regression coefficient of tricuspid flow; triangle represents the standardized regression coefficient of left atrial volume index; diamond represents the standardized regression coefficient of left ventricular mass index. NGM indicates normal glucose metabolism; and T2D, type 2 diabetes mellitus.

Structural Measures of LV Diastolic Function

After adjustment for potential confounders (Model 2), higher LAVI (per 1 mL/m²) was, in NGM, significantly associated with higher CRF (regression coefficient for LAVI beta [95% CI] 0.015; [0.008–0.023]). In prediabetes and T2D, no significant associations were observed between LAVI and CRF (−0.003 [−0.017 to 0.011] and 0.000 [−0.009 to 0.010], respectively). This association did differ significantly in individuals with prediabetes and T2D as compared with NGM (P _interaction between prediabetes and T2D, and LAVI were 0.038 and 0.066, respectively; Table S6).

After adjustment for potential confounders, higher LVMI (per 1 g/m^2.7) was, in NGM, borderline significantly associated with higher CRF (0.007 [−0.001 to 0.015]). In prediabetes and T2D, no significant associations were observed between LVMI and CRF (−0.004 [−0.017 to 0.009]) and −0.006 [−0.016 to 0.004]), respectively). These associations did significantly differ in individuals with T2D, but not prediabetes, as compared with NGM (P _interaction between prediabetes and T2D and LVMI were 0.318 and 0.058, respectively; Table S6).

After further adjustment for waist (Model 3), in NGM, the association between LAVI and CRF was not materially altered, whereas the association between LVMI and CRF was further strengthened.

Functional Measures of LV Diastolic Function

After adjustment for potential confounders (Model 2), higher average E/e′ ratio (per unit) was associated with lower CRF. In NGM (−0.044 [−0.071 to −0.016]) and T2D (−0.033 [−0.064 to −0.001]), the association was statistically significant, but in prediabetes it was not (−0.030 [−0.072 to 0.012]). The associations did not significantly differ in prediabetes and T2D as compared with NGM (P _interaction >0.10; Table S6).

In addition, after adjustment for potential confounders (Model 2), higher average e′ (per centimeter per second) was not significantly associated with CRF in all 3 groups of metabolism status (NGM 0.013 [−0.016 to 0.042], prediabetes 0.040 [−0.016 to 0.095], T2D −0.010 [−0.057 to 0.037]). Analyses after adjustment for age, sex, and height show similar results as with average E/e′ ratio (Table 3). The associations did not significantly differ in prediabetes and T2D as compared with NGM (P _interaction >0.10; Table S6). In contrast, after adjustment for potential confounders, higher maximum tricuspid regurgitation flow (per 1 m/s) was, in NGM, significantly associated with higher CRF (0.129 [0.011–0.246]). In prediabetes and T2D, no significant associations were observed between maximum tricuspid flow and CRF (−0.030 [−0.207 to 0.146] and −0.053 [−0.193 to 0.087], respectively). These associations did not significantly differ in prediabetes and T2D as compared with NGM (P _interaction >0.10; Table S6).

These associations were at most slightly attenuated after further adjustment for waist (Model 3).

Additional Analysis

After adjustment for potential confounders (Model 2), abnormal as compared with normal diastolic function according to 2016 guidelines was, in NGM, borderline significantly associated with higher CRF (0.165 [−0.004;0.335]). In prediabetes and T2D, no significant associations were observed between abnormal as compared with normal diastolic function (−0.053 [−0.329 to 0.224] and −0.065 [−0.301 to 0.170], respectively; Table 3). Indeterminate diastolic function as compared with normal diastolic function was, in NGM, prediabetes, and T2D, not associated with CRF (0.050 [−0.079 to 0.146] and −0.158 [−0.369 to 0.052], and −0.115 [−0.302 to 0.073], respectively; Table 3). These associations did not significantly differ in prediabetes and T2D as compared with NGM (P _interaction >0.10; Table S6).

If we repeated the analyses between measures of LV diastolic function and CRF in the total population, associations were not materially altered compared with the associations in NGM, but the associations between LAVI and LVMI were attenuated in effect size and significance, because of a significant interaction between those measures and prediabetes in their association with CRF (Table S7).

Associations between measures of LV diastolic function and CRF were not materially altered in the following scenarios (Tables S8 through S12): when we replaced LAVI with nonindexed left atrial volume; when we replaced LVMI indexed by height with LVMI indexed by body surface area or LV mass (Table S8); when we replaced CRF as W_max per kilogram by CRF as the percentage of the predicted W_max (Table S9); when we additionally adjusted for moderate‐to‐vigorous physical activity; when we restricted the analyses to individuals without prior cardiac disease or without functional mobility limitations; when we replaced office systolic pressure with office diastolic pressure, presence of hypertension, or 24‐hour ambulatory systolic or diastolic pressure; when we additionally adjusted for renin‐angiotensin system inhibitors or β‐blockers; or when we replaced waist with body mass index or weight (Tables S10 through S12). In addition, in men as compared with women, associations between measures of LV diastolic function and CRF did not statistically significantly differ (P _interactions >0.10; Table S13).

Discussion

In this population‐based study, structural and functional measures of LV diastolic function were differentially associated with CRF over the strata of GMS, apart from average E/e′ ratio, which was inversely associated with CRF in all individuals. We found positive associations of maximum tricuspid regurgitation flow and structural measures of LV diastolic function with CRF in individuals with NGM, but not in those with (pre)diabetes. These associations were independent of cardiovascular risk factors and lifestyle factors and remained unchanged after excluding individuals with prior cardiac pathology. Taken together, deteriorating LV diastolic function may contribute to the pathophysiological process of impaired CRF in the general population, although the effect was small. In addition, our results suggest that the differential effects over the strata of glucose metabolism are the consequence of cardiac structural adaptation to effectively increase CRF in NGM, but not in (pre)diabetes.

Previous population‐based studies not stratified according to (pre)diabetes have investigated the association between structural and/or functional measures of LV diastolic function and CRF.⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³ However, these studies are difficult to compare because they quantified LV diastolic function heterogeneously. Nevertheless, our study is in line with previous population‐based studies that have reported inverse associations between several functional measures of LV diastolic function and CRF⁷, ⁸, ¹⁰, ¹¹ and a positive association between a structural measure of LV diastolic function and CRF,⁷ but not with others that have reported a negative association between structural measures of LV diastolic function and CRF⁹, ¹² (for details see Table S1). The latter might be explained by the fact that these studies were performed in populations at high risk of CVD only.⁹, ¹² Our study extends previous findings to assessment of the association between LV diastolic function and CRF, with both structural and functional measures of LV diastolic function,¹⁴ in a relatively large population‐based cohort of individuals aged 40 to 75 years stratified according to (pre)diabetes and with adjustment for multiple confounders.

Maladaptive cardiac structural alterations may be of hemodynamic and/or microvascular origin.¹ For instance, patients with heart failure with preserved ejection fraction are known to have lower CRF in whom diastolic filling is delayed, slowed, shortened, or associated with elevated LV pressures, leading to reduced ability to enhance transmitral flow and accelerate diastolic filling during exercise.¹ Conceivably, a similar mechanism may also be operative in individuals with subclinical LV diastolic dysfunction in the general population and might, as shown in previous studies,¹⁴, ¹⁵ be most sensitively detected by the functional measure of average E/e′ ratio.

Inadequate alterations in structural LV diastolic function measures could represent impairment of long‐term cardiac hemodynamic function and perfusion. However, physiological cardiac adaptations also involve an increase of tricuspid flow,³⁰ an increase in LAVI,³¹ and an increase in LVMI,³² questioning their value as markers of early diastolic dysfunction. These adaptations may explain our findings that tricuspid flow and structural measures of LV diastolic function were positively associated with CRF in NGM. The absence of such associations in (pre)diabetes suggests either an absence of these physiological adaptations or early diastolic dysfunction that reverses the original physiological adaptation in (pre)diabetes.

Alternatively or additionally, subclinical LV diastolic dysfunction and CRF may share common risk factors, which over time produce alterations in both entities independently of each other. However, we adjusted extensively for potential confounders such as hypertension, smoking, and other cardiovascular risk factors, which therefore are unlikely explanations for the associations we observed. Still, we cannot exclude that other lifestyle factors (eg, dietary habits), which were not included in the analyses, might play a role in the association between LV diastolic function and CRF.³³

The absence of a cardiac structural adaptation response in (pre)diabetes might be explained by the existence of preclinical hyperglycemia‐driven diabetic cardiomyopathy.³ Hyperglycemia affects the structures of the heart and results in myocardial fibrosis, with accumulation of advanced glycation end products in the myocardium; in increased myocardial content of free radicals and oxidants that decrease nitric oxide levels, worsen endothelial function, and induce myocardial inflammation; in elevation of free fatty acids and their oxidation products that may have direct toxic effects on the myocardium; and in altered intracellular calcium homeostasis that leads to myocardial dysfunction.³ In addition, insulin resistance and hyperinsulinemia may contribute to LV hypertrophy,³⁴ and impaired glucose control and insulin resistance may lead to autonomic dysfunction and consequent myocardial hypertrophy and fibrosis.³⁴ Furthermore, the structure and function of the heart may be indirectly affected via alterations in vascular function (eg, microvascular function).³⁵ Because of synergy among these risk factors,⁴ the presence of diabetic cardiomyopathy could aggravate LV diastolic function, or alternatively, it could inhibit an adequate structural adaptation.

The present study contributes to our understanding of the association between structural and functional measures of LV diastolic dysfunction and CRF in a population‐based cohort stratified according to (pre)diabetes. First, we were able to accurately examine CRF and measures of LV diastolic function through a submaximal ergometer test instead of questionnaires²⁰ and the use of continuous structural and functional measures, in addition to a clinical cutoff definition of LV diastolic function according to recent guidelines.¹⁴ Second, we adjusted for an extensive series of potential confounders including CVD risk factors such as systolic blood pressure and use of antihypertensive medication. Risk of overadjustment bias²⁴ (in Model 3) was small, because associations were at most slightly attenuated by adjustment for waist. Moreover, associations remained after excluding individuals with prior coronary heart disease, atrial fibrillation, wall abnormalities, and significant valvular pathology, indicating a role for LV diastolic function in CRF in individuals without cardiac disease. Third, although reversed causality is biologically not unlikely,³⁶, ³⁷ our results were not materially altered when we additionally adjusted the analyses for physical activity.

Strengths of our study include its population‐based design with oversampling of individuals with T2D, the consideration of prediabetes, and the use of extensive phenotyping, which allowed us to adjust for extensive series of CVD risk factors including 24‐hour ambulatory blood pressure. Importantly, a broad array of additional analyses all gave consistent results.

Our study also has limitations. First, the cross‐sectional design of the study does not allow us to draw strong causal inferences. However, from the association between LV diastolic dysfunction and CRF in patients with heart failure,¹, ² it follows that there is a strong prior likelihood that preclinical LV diastolic dysfunction to a certain extent may contribute to impairment in CRF. Second, the use of W_max per kilogram instead of percentage of predicted maximum may limit the clinical interpretation of the found associations. However, the use of the recommended and often used clinical formula by Jones et al²⁵, ³⁸ may also be questioned, because it is less precise in measuring differences in CRF and has led, in our study population, to an underestimation of the predicted value of W_max and consequently to an overestimation of the percent of the predicted W_max (Table S2). Nevertheless, when we replaced the outcome measure with the percentage of the predicted value of W_max, the results were not altered. Third, although average e′ could be a sensitive marker for early diastolic dysfunction, it was no longer associated with CRF after adjustment for potential confounders in our study. Therefore, the use of average e′ still needs further investigation in relation to CRF on population‐based level as compared with E/e′ ratio.¹, ¹⁴, ¹⁵, ³⁹ Fourth, we may have underestimated associations between measures of LV diastolic function and CRF especially in T2D, because individuals who were excluded because of missing values had an adverse cardiometabolic risk profile, worse LV diastolic function, and lower CRF, the healthy participant effect (ie, sicker potential participants are less likely to participate, which is particularly likely for those with T2D). Fifth, LV strain imaging was not available in our study, which might be a potential alternative, but also an experimental measure for subclinical LV dysfunction to assess in association with CRF according to GMS.⁴⁰ Sixth, the generalizability of our findings to other populations can be questioned, but because of its population‐based design, the results of our study may at least be generalized to middle‐aged and older individuals with a similar cardiometabolic risk profile.

In conclusion, our population‐based study shows that average E/e′ ratio as a measure of LV diastolic function was inversely associated with CRF, independently of cardiovascular risk factors, lifestyle factors, cardiac pathology, and GMS. Structural changes in left atrium and left ventricular muscle mass were positively associated with CRF in NGM but not in (pre)diabetes. Our findings suggest that deteriorating LV diastolic function may contribute to the pathophysiological process of impaired CRF in the general population. They also suggest that maximum tricuspid regurgitation flow and structural measures of LV diastolic function may not be accurate for the definition of early LV diastolic dysfunction in the general population. Moreover, structural adaptations of the left atrium and left ventricle may be different in (pre)diabetes as compared with NGM. Future prospective studies need to unravel the exact cardiovascular mechanisms to better understand the cardiac adaptation process in (pre)diabetes (eg, mechanisms of diabetic cardiomyopathy) in the early development of diastolic dysfunction and the usefulness of LV diastolic function as potential targets in preventive and therapeutic strategies for impaired CRF in the general population and in (pre)diabetes.

Sources of Funding

This work was supported by the European Regional Development Fund via OP‐Zuid, the Province of Limburg, the Dutch Ministry of Economic Affairs (grant 31O.041); Stichting De Weijerhorst (Maastricht, the Netherlands); the Pearl String Initiative Diabetes (Amsterdam, the Netherlands); CARIM School for Cardiovascular Diseases (Maastricht, the Netherlands); Stichting Annadal (Maastricht, the Netherlands); Health Foundation Limburg (Maastricht, the Netherlands); and by unrestricted grants from Janssen‐Cilag B.V. (Tilburg, the Netherlands), Novo Nordisk Farma B.V. (Alphen aan den Rijn, the Netherlands), and Sanofi‐Aventis Netherlands B.V. (Gouda, the Netherlands).

Disclosures

Prof Brunner‐La Rocca reports unrestricted research grants from Roche Diagnostics, Vifor, and Novartis, and serves as a member of the advisory board of Roche Diagnostics, Vifor, Novartis, Boehringer Ingelheim, and AstraZeneca outside the submitted work. The remaining authors have no disclosures to report.

Supporting information

Data S1

Tables S1–S13

Figure S1

References 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59

(J Am Heart Assoc. 2021;10:e020387. DOI: 10.1161/JAHA.120.020387.)