Background
The effects of light to moderate alcohol consumption on cardiac mechanics remain poorly understood. The aim of this study was to investigate the dose-response relationship between alcohol consumption and left ventricular (LV) and left atrial (LA) function using myocardial deformation.
Methods
In total 3,946 asymptomatic participants (mean age, 49.7 ± 10.7 years; 65% men) were consecutively studied using comprehensive echocardiography and two-dimensional speckle-tracking in a cross-sectional, retrospective manner. Global LV longitudinal and circumferential strain and LA strain were assessed and related to habitual alcohol consumption pattern (fewer than one, one to six, or more than six drinks per week) before and after propensity matching.
Results
With increasing weekly alcohol consumption, participants displayed greater LV eccentric remodeling, impaired diastolic function, and more attenuated global longitudinal strain, LA strain (adjusted coefficients, −1.07 [95% CI, −1.95 to −0.19] and −3.73 [95% CI, −5.36 to −2.11]), and early diastolic strain rates (adjusted coefficients, 0.07 [95% CI, 0.03–0.11] and 0.33 [95% CI, 0.24–0.42]) for one to six and more than six drinks per week, respectively ( P < .05 for all) in a dose-response manner. Participants with recent alcohol abstinence displayed cardiac mechanics intermediate between those of nondrinkers and current drinkers. After propensity matching ( n = 1,140), participants currently consuming more than one drink per week continued to have significantly attenuated global longitudinal strain and all LA mechanics compared with those consuming fewer than one drink per week ( P < .05 for all).
Conclusions
Habitual alcohol consumption, even at light to moderate doses, is associated with both reduced LV and LA mechanics in a dose-dependent manner. Whether such observations are reversible or related to future atrial fibrillation deserves further study.
Chronic excessive alcohol consumption is known to be associated with detrimental effects on cardiac structure and function. The extreme manifestation of the toxic effects of alcohol on the heart has been termed “alcoholic cardiomyopathy,” in which left ventricular (LV) systolic function is severely impaired. However, disruptions in myofibrillary architecture, fibrosis, and subclinical reductions in myocardial contractility may occur with alcohol consumption, even before overt LV systolic dysfunction. Furthermore, moderate doses of alcohol consumption have recently been shown to be associated with incident atrial fibrillation (AF). Thus, preclinical LV and left atrial (LA) functional alterations may be detected with lower doses of alcohol before the development of overt heart failure or AF and may carry important implications for targeted preventive management.
Accordingly, we aimed to study the association of light to moderate habitual alcohol consumption with changes in cardiac structure and function in clinically asymptomatic participants. We hypothesized that myocardial deformation would enable the detection of subclinical cardiac functional changes and that these changes would be detectable in a dose-response fashion. Further recognizing that alcohol consumption may be associated with lifestyle factors, metabolic changes, and blood pressure effects that can affect cardiac structure and function, we used propensity matching to account for these covariates and assessed the independent effect of alcohol on cardiac mechanics.
Methods
Study Setting and Population
Our study population consisted of consecutive participants in an ongoing cardiovascular health screening program from June 2009 to December 2012 at a tertiary medical center in Northern Taipei, Taiwan. All participants underwent detailed physical examination, baseline anthropometric measurements, biochemical studies, and comprehensive echocardiography. The primary goal of this program was to test the hypothesis that several clinical key demographic characteristics, biochemical data, or lifestyle factors (such as smoking and alcohol consumption) are related to cardiac remodeling or subclinical cardiac dysfunction in terms of worsened myocardial deformation measures in an ethnic Asian population. The study design in our present work relating the dose of alcohol use and deformational functional changes was cross-sectional and conducted in a retrospective manner. The clinical significance of strain measures is further addressed in the section “Speckle-Tracking Analysis Protocol.” A 260-item structured questionnaire was used to obtain baseline clinical information, symptoms and signs, medical history, and lifestyle factors. The exclusion criteria were entry into this screening program before the feasibility of myocardial deformation analysis, inadequate images for myocardial deformation analysis, specific clinical conditions and diseases, and missing data ( Supplemental Figure 1 ). This study was approved by the local ethics board committee (14MMHIS202).
Standard Conventional Echocardiographic Protocol
Echocardiography was uniformly performed using the Vivid i system (GE Healthcare, Little Chalfont, United Kingdom) equipped with a 2- to 4-MHz transducer (3S-RS) during the recruitment period of the present analysis. The standard echocardiographic imaging protocol included measurement of LV end-diastolic and end-systolic diameter, wall thickness, LV mass (per American Society of Echocardiography criteria), and LA and LV volumes (using the biplane Simpson method). LV ejection fraction (LVEF) and LA emptying fraction were calculated as 100 × (maximal volume − minimal volume)/maximal volume. LV diastolic function was determined by pulsed-wave Doppler of the early (E) and late diastolic (A) LV filling velocities at the tip of the mitral leaflets from the apical four-chamber view. Doppler tissue imaging (DTI)–based mitral annular contraction (S′) and relaxation velocity (E′) were assessed at the lateral mitral annulus using spectral Doppler techniques, with LV filling pressure estimated using the E/E′ ratio.
Speckle-Tracking Analysis Protocol
Assessment of LV and LA Deformation
We assessed LV deformation using baseline two-dimensional images from three LV apical views for longitudinal strain, three short-axis views for LV circumferential strain, and LV twist analysis (EchoPAC version 10.8; GE Vingmed Ultrasound AS, Horten, Norway), as described in our previous work. Baseline two-dimensional images were analyzed using offline endocardial border manual tracing by the same experienced technician, using proprietary software (EchoPAC version 10.8). The mean frame rate in all studied participants was between 60 and 80 frames/sec. On the basis of automated speckle-tracking algorithms, LV global longitudinal strain (GLS) was then averaged from three individual LV apical views (two-chamber, four-chamber, and three-chamber views). Global circumferential strain (GCS) curves were similarly obtained by averaging values from three individual short-axis levels (mitral valve, papillary muscles, and apical layer), with LV twist derived by subtracting rotation from the LV mitral annulus (negative value) to the LV apical level (positive value).
Speckle-tracking analyses for LA function were also performed, ensuring that image acquisition was carefully optimized in the apical four-chamber and two-chamber views to avoid foreshortening, with offline analyses performed as previously described. LA longitudinal strain (LAS) and strain rate (SR) curves (systolic SR [SRs], early diastolic SR [SRe], and late diastolic LA SR [SRa]) were generated for each atrial segment from LV apical two- and four-chamber views ( Supplemental Figure 2 ). The representative LA deformation indices in each study participant were then derived from the mean of both LV apical two- and four-chamber data. Larger strain or SR measures on the basis of the absolute values indicated better myocardial contractile function. The detailed methodology for the measurement of LV and LA deformation is further addressed in the Supplemental Materials .
Quantification and Categorization of Alcohol Consumption
Details of alcohol consumption (pattern and frequency) were obtained from predefined questionnaires (available for 97.5% of the study population), including information regarding the habitual intake of 6 kinds of local alcoholic beverages (class I, II, and III beer; wine; strong wine; and spirits), the amount of alcoholic beverage consumed (in milliliters) in a single occasion, the frequency of weekly consumption (from none to current users, or abstinence following prior regular use). The amount consumed was reported in an open-ended manner. Weekly alcohol intake (in grams) was then calculated by multiplying the frequency of weekly consumption by the amount of alcoholic beverage consumed (in milliliters) in a single occasion. We assumed that one standard drink contains 12 g alcohol, and the precise conversion of total alcohol consumption into number of standard drinks on a weekly basis was then derived. To further validate the accuracy of self-reported alcohol use by patients in our laboratory, we selected a random sample of 120 subjects with self-report questionnaires completed first, and recall interviews were performed to validate the accuracy of the data fill-in process. The Spearman rank correlation coefficient values for alcohol consumption pattern (on a weekly basis) and amount (in glasses) were 0.90 and 0.88, respectively, which showed good consistency between questionnaires and recall interview data. We further classified study participants on the basis of graded alcohol consumption on a weekly basis : (1) group A, nondrinkers: nonusers or fewer than one drink per week ( n = 3,464); (2) group B, light drinkers: one to six drinks per week ( n = 323); (3) group C, moderate drinkers: more than six drinks per week ( n = 89); and (4) group D, past drinkers: individuals with histories of regular alcohol consumption but abstinence for ≥3 years (mean, 3.8 ± 1.2 years; 84% available) at the time of the study visit ( n = 70). The maximal weekly consumption of alcohol was 20 drinks per week ( n = 5); two participants considered heavy drinkers with higher alcohol consumption were excluded. For comparisons of nondrinkers, current drinkers, and past drinkers, we combined light and moderate drinkers (groups B and C) into a single group of current light to moderate drinkers ( Supplemental Figure 1B ).
Reproducibility Analysis
We selected a random sample of 50 subjects from our present work and determined coefficients of variance for GLS, GCS, LAS, and LA SRs, SRe, and SRa to be 4.6%, 5.8%, 6.4%, and 5.6%, 4.4%, and 5.8% between raters (interobserver variability) and 3.5%, 5.2%, 4.8%, and 4.4%, 4.0%, and 4.6% for the same rater (intraobserver variability), respectively. The coefficients of variance for LV end-diastolic volume and maximal LA volume between raters (interobserver variability) were 7.6% and 7.1%; these values were 5.9% and 6.2% for the same rater (intraobserver variability). Intraobserver reliability and interobserver reliability using intraclass correlation coefficients (ICCs) are further detailed in the Supplemental Materials .
Statistical Analysis
Continuous data are reported as mean ± SD and categorical data as proportions; data were compared using the χ 2 or Fisher exact test, as appropriate. One-way analysis of variance was used to compare baseline demographics and conventional echocardiography-based analysis (both LA and LV structure or function analysis), with post hoc paired comparisons (group A vs groups B and C and group D). To assess the dose-response relationship between graded alcohol consumption and LA and LV structural and mechanical changes, analyses were conducted using nondrinkers as the reference category. Covariates tested included known confounders, including age, gender, body mass index, systolic blood pressure, medical history of hypertension, diabetes, cardiovascular diseases, and current smoking in our multivariate model. Finally, propensity matching was used in the comparison of deformation parameters between nondrinkers and light to moderate drinkers ( Supplemental Materials ) after matching for clinical covariates using SAS version 9.4e (SAS Institute, Cary, NC).
Except for propensity matching, all analyses were performed using Stata 11.0 (StataCorp LP, College Station, TX). P values were set for two-tailed probability, with P values < .05 considered to indicate statistical significance.
Results
Baseline Demographics
From June 2009 to December 2012, a total of 3,946 asymptomatic participants (mean age, 49.9 years; 35% women) had both sufficient baseline information and deformation data for the present analysis ( Table 1 , Supplemental Figure 1A ). Across the categories of increasing alcohol consumption from group A to C, we observed increases in the number of men, body size (body mass index, waist circumference), blood pressure, fasting glucose, blood γ-glutamyl transpeptidase, uric acid, and triglyceride but decreases in high-density lipoprotein level ( P < .001 for all). Despite similar age, light to moderate drinkers had higher blood pressures and a greater prevalence of smoking (26.5% vs 6.1%) and hypertension (25.8% vs 16.6%) ( P < .001 for both) compared with nondrinkers. Past drinkers tended to be older, with several clinical covariates falling between nondrinkers and light to moderate drinkers. Among current smokers, few smoked more than one pack per day, though there remained a greater portion of distribution differences across the three groups ( P < .001).
| Alcohol categories | χ 2 / P value for ANOVA: group A to C | |||||
|---|---|---|---|---|---|---|
| Nondrinkers, group A ( n = 3,464) | Light drinkers, group B ( n = 323) | Moderate drinkers, group C ( n = 89) | Light to moderate drinkers, groups B and C ( n = 412) | Past drinkers, group D ( n = 70) | ||
| Baseline characteristics | ||||||
| Age (y) | 49.7 ± 10.8 | 50.4 ± 10 | 50 ± 10.1 | 50.34 ± 10.05 | 54.56 ± 10.31 | .594 |
| Men | 2,160 (62.07%) | 282 (86.77%) | 83 (92.22%) | 365 (87.95%) | 61 (87.14%) | <.001 |
| Height (cm) | 165.2 ± 8.7 | 168.9 ± 7.1 ‡ | 169.2 ± 7.2 ‡ | 168.98 ± 7.09 ∗ | 167.19 ± 7.26 | <.001 |
| Weight (kg) | 66.7 ± 12.4 | 71.6 ± 12.1 ‡ | 74.3 ± 14.6 ‡ | 72.18 ± 12.67 ∗ | 69.54 ± 10.7 | <.001 |
| BMI (kg/m 2 ) | 24.3 ± 3.5 | 24.94 ± 3.8 ‡ | 25.8 ± 3.9 ‡ | 25.13 ± 3.85 ∗ | 24.82 ± 3.11 | <.001 |
| Waist (cm) | 81.4 ± 14.9 | 85.5 ± 12.7 ‡ | 89 ± 9.7 ‡ | 86.23 ± 12.16 ∗ | 84.14 ± 17.15 | <.001 |
| SBP (mm Hg) | 121.3 ± 18 | 124.4 ± 16.7 ‡ | 127 ± 16.3 ‡ | 125 ± 16.63 ∗ | 125.06 ± 15.34 | <.001 |
| DBP (mm Hg) | 75.1 ± 11.4 | 78.54 ± 10.6 ‡ | 80.47 ± 10.8 ‡ | 78.96 ± 10.68 ∗ | 78.22 ± 11.33 | <.001 |
| Heart rate (beats/min) | 65.7 ± 15 | 64.1 ± 9.8 | 64.6 ± 9.4 | 64.23 ± 9.71 | 66.68 ± 11.51 | .313 |
| Biochemical data | ||||||
| Fasting glucose (mg/dL) | 101.1 ± 21.9 | 102.5 ± 21.5 | 115.1 ± 43.3 ‡§ | 105.2 ± 28.01 ∗ | 104.44 ± 22.83 | <.001 |
| HbA 1c (%) | 5.8 ± 0.8 | 5.8 ± 0.8 | 6.2 ± 1.4 ‡§ | 5.88 ± 0.96 | 5.98 ± 1.08 | <.001 |
| Uric acid (mg/dL) | 5.8 ± 1.5 | 6.28 ± 1.4 ‡ | 6.5 ± 1.5 ‡ | 6.33 ± 1.43 ∗ | 6.02 ± 1.48 | <.001 |
| eGFR (ml/min/1.73m 2 ) | 88.7 ± 16.5 | 86.3 ± 16.1 ‡ | 90.6 ± 16.4 | 87.17 ± 16.24 | 88.24 ± 19.22 | .024 |
| Cholesterol (mg/dL) | 203.2 ± 36.4 | 205.4 ± 36 | 207.6 ± 33.6 | 205.88 ± 35.5 | 206.27 ± 37.38 | .259 |
| Triglyceride (mg/dL) | 131.68 ± 91 | 155.95 ± 125.9 ‡ | 198.7 ± 151.9 ‡§ | 164.99 ± 132.81 ∗ | 158.08 ± 156.88 | <.001 |
| LDL (mg/dL) | 131.66 ± 33.7 | 133.2 ± 32.9 | 132.69 ± 3 | 133.09 ± 32.9 | 134.16 ± 31.29 | .698 |
| HDL (mg/dL) | 54.6 ± 15.3 | 52.93 ± 14.1 | 48.7 ± 14.2 ‡ | 52.03 ± 14.23 ∗ | 51.32 ± 12.89 | <.001 |
| γ-GT (mg/dL) | 25.4 ± 21.6 | 36.3 ± 35.4 ‡ | 42.1 ± 28.1 ‡§ | 37.6 ± 34 ∗† | 29.2 ± 25.4 | <.001 |
| Medical history | ||||||
| Hypertension | 576 (16.6%) | 77 (23.7%) | 30 (33.3%) | 107 (25.8%) | 21 (30.0%) | <.001 |
| Diabetes | 227 (6.52%) | 15 (4.62%) | 13 (14.44%) | 28 (6.75%) | 12 (17.14%) | .197 |
| Hyperlipidemia | 261 (7.5%) | 31 (9.5%) | 10 (11.1%) | 41 (9.9%) | 8 (11.4%) | .075 |
| CVD | 199 (5.7%) | 26 (8.1%) | 6 (6.7%) | 32 (7.7%) | 11 (15.7%) | .165 |
| Current smoking | 213 (6.1%) | 70 (21.5%) | 39 (43.3%) | 109 (26.5%) | 22 (31.4%) | <.001 |
| Smoking >1 PPD | 127 (3.7%) | 30 (9.3%) | 18 (20%) | 48 (11.7%) | 13 (24.1%) | <.001 |
| Regular exercise | 423 (12.2%) | 54 (16.6%) | 12 (13.3%) | 66 (15.9%) | 15 (21.4%) | .077 |
Conventional Echocardiographic Findings
Across the categories of increasing alcohol consumption from group A to C, we found increasing LA volume, LV dimension, wall thickness, and LV mass index, as well as more prolonged deceleration time and isovolumic relaxation time and lower E′ ( P for trend < .001 for all) ( Table 2 ). There were no differences in filling pressures (E/E′ ratio) among groups. Interestingly, DTI-based S′ decreased with increasing alcohol consumption ( P for trend < .05) despite the preservation of LVEF. Past drinkers displayed several echocardiographic features resembling light to moderate drinkers.
| Alcohol categories | χ 2 / P for ANOVA: group A to C | |||||
|---|---|---|---|---|---|---|
| Nondrinkers, group A ( n = 3,464) | Light drinkers, group B ( n = 323) | Moderate drinkers, group C, ( n = 89) | Light to moderate drinkers, groups B and C ( n = 412) | Past drinkers, group D ( n = 70) | ||
| Maximal LA volume (mL) | 27.73 ± 10.5 | 30.11 ± 11.2 ‡ | 30.5 ± 10.5 ‡§ | 30.2 ± 11.04 ∗ | 28.15 ± 9.91 | <.001 |
| Minimal LA volume (mL) | 11.9 ± 5.4 | 13.4 ± 6.1 ‡ | 14.1 ± 5.7 ‡§ | 13.53 ± 6 ∗ | 12.6 ± 4.8 | <.001 |
| LVIDd (mm) | 46.5 ± 3.7 | 47.9 ± 3.1 ‡ | 47.9 ± 3.3 ‡§ | 47.87 ± 3.1 ∗ | 47.66 ± 3.66 ∗ | <.001 |
| LVIDs (mm) | 29.24 ± 3.0 | 30.14 ± 2.6 ‡ | 30.28 ± 2.9 ‡§ | 30.17 ± 2.7 ∗ | 30.35 ± 3.12 ∗ | <.001 |
| IVS (mm) | 8.9 ± 1.1 | 9.3 ± 1.0 ‡ | 9.4 ± 1.1 ‡§ | 9.34 ± 1.02 ∗ | 9.32 ± 1.03 ∗ | <.001 |
| PWT (mm) | 0.39 ± 0.05 | 0.39 ± 0.04 ‡ | 0.4 ± 0.05 ‡§ | 0.39 ± 0.04 ∗ | 0.39 ± 0.05 ∗ | <.001 |
| LV mass index (g/m 2 ) | 74.7 ± 15.6 | 79 ± 13.1 ‡ | 79.44 ± 13.9 ‡§ | 79.1 ± 13.3 ∗ | 78.7 ± 17.1 | <.001 |
| M/V ratio | 1.86 ± 0.3 | 1.92 ± 0.3 ‡ | 1.98 ± 0.3 ‡§ | 1.93 ± 0.29 ∗ | 1.91 ± 0.35 | <.001 |
| LVEF (%) | 62.46 ± 5.21 | 62.23 ± 5.34 | 61.74 ± 5.95 | 62.12 ± 5.47 | 61.22 ± 6.05 | .345 |
| DT (msec) | 199.7 ± 39 | 204.4 ± 4 | 217.5 ± 42.5 ‡§ | 207.2 ± 41.6 | 204.5 ± 33.9 | <.001 |
| IVRT (msec) | 89.23 ± 14.35 | 91.89 ± 17.3 ‡ | 93.91 ± 16 ‡§ | 92.33 ± 16.99 ∗ | 92.96 ± 17.89 | <.001 |
| E/A ratio | 1.19 ± 0.43 | 1.23 ± 0.37 | 1.07 ± 0.36 ‡§ | 1.19 ± 0.37 | 1.12 ± 0.48 | .0013 |
| E/E′ ratio | 7.18 ± 2.62 | 7.40 ± 2.77 | 7.31 ± 2.28 | 7.38 ± 2.67 | 8.14 ± 2.44 ∗ | .334 |
| DTI E′ lateral (cm/sec) | 10.47 ± 2.87 | 9.9 ± 2.7 ‡ | 9.19 ± 2.61 ‡ | 9.75 ± 2.69 ∗ | 9.13 ± 2.62 ∗ | <.001 |
| DTI S′ lateral (cm/sec) | 9.25 ± 2.37 | 8.96 ± 1.94 | 8.68 ± 2.04 | 8.81 ± 1.97 ∗ | 8.75 ± 2.34 | .0092 |
LV and LA Deformation Indices
Across the categories of increasing alcohol consumption from group A to C ( Table 3 ), we found a graded reduction of GLS (group A to C, P < .001) and a marginal decrease in GCS but no difference in cardiac twist ( P = NS). Light to moderate drinkers had lower GLS compared with nondrinkers (−19.66% vs −20.36%, P < .001) and past drinkers ( P < .05).
| Alcohol categories | χ 2 / P for ANOVA: group A to C | |||||
|---|---|---|---|---|---|---|
| Nondrinkers, group A ( n = 3,464) | Light drinkers, group B ( n = 323) | Moderate drinkers, group C ( n = 89) | Light to moderate drinkers, groups B and C ( n = 412) | Past drinkers, group D ( n = 70) | ||
| Longitudinal strain 4CH (%) | −20.25 ± 2.03 | −19.6 ± 1.98 ‡ | −19.19 ± 2.11 ‡ | −19.51 ± 2.01 ∗ | −19.95 ± 1.74 | <.001 |
| Longitudinal strain 2CH (%) | −20.5 ± 2.07 | −19.86 ± 1.85 ‡ | −19.33 ± 2.16 ‡ | −19.75 ± 1.93 ∗† | −20.6 ± 1.91 | <.001 |
| Longitudinal strain 3CH (%) | −20.36 ± 2.06 | −19.76 ± 1.93 ‡ | −19.21 ± 2.02 ‡ | −19.64 ± 1.96 ∗† | −20.29 ± 1.91 | <.001 |
| GLS (%) | −20.36 ± 1.94 | −19.78 ± 1.68 ‡ | −19.23 ± 1.95 ‡ | −19.66 ± 1.75 ∗ | −20.16 ± 2.04 | <.001 |
| Circumferential strain MV (%) | −18.32 ± 2.29 | −18.11 ± 2.3 | −17.94 ± 2.39 | −18.07 ± 2.32 | −18.03 ± 2.01 | .113 |
| Circumferential strain PM (%) | −21.7 ± 4.82 | −21.38 ± 4.92 | −20.89 ± 4.83 | −21.28 ± 4.9 | −21.06 ± 4.15 | .176 |
| Circumferential strain AP (%) | −24.13 ± 4.75 | −23.91 ± 4.67 | −23.67 ± 5.55 | −23.86 ± 4.86 | −24.24 ± 5.11 | .511 |
| GCS (%) | −21.39 ± 3.59 | −21.14 ± 3.58 | −20.84 ± 3.92 | −21.07 ± 3.65 | −21.07 ± 3.38 | .211 |
| Cardiac twist (deg) | 13.74 ± 5.42 | 14.11 ± 5.11 | 13.75 ± 4.8 | 14.03 ± 5.04 | 15.07 ± 4.94 | .512 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
