Alcohol consumption and cardiorespiratory fitness in five population-based studies Free

Abstract

Introduction

Although alcohol is a major contributor to the burden of disease,¹ observational studies have documented J- or U-shaped relations between alcohol intake and multiple cardiovascular and pulmonary outcomes,^2–4 including coronary artery disease^5,6 stroke,⁷ congestive heart failure,⁸ lung function⁹ and asthma.¹⁰ In essence, these studies have indicated lower cardiovascular risk in individuals with low-to-moderate levels of alcohol consumption as compared to abstainers, and those with heavy alcohol intake.⁵

Few cross-sectional studies have addressed the relationship between alcohol consumption and cardiorespiratory fitness (CRF). CRF is a measure of the capacity of the cardiovascular system to transport oxygen and the capacity of the muscle to use it, and measured by peak exercise oxygen uptake (VO₂).¹¹ CRF is a strong risk factor for cardiovascular morbidity and mortality.¹² Existing cross-sectional population-based studies examining the association between alcohol consumption and CRF have yielded inconsistent results.^13–15 Previous studies have not examined the potential non-linear dose-response relationship of alcohol intake and CRF. In addition, the association between heavy episodic (binge) drinking and CRF has not been investigated, although heavy episodic drinking increases cardiovascular disease risk.⁶

We examined associations of average alcohol consumption, heavy episodic drinking and CRF using data from five independent general population samples: the Study of Health in Pomerania (SHIP),¹⁶ the German Health Interview and Examination Survey for Adults (DEGS1)¹⁷ and three samples of the US National Health and Nutrition Examination Survey (NHANES) (1999/2000, 2001/2002, 2003/2004).¹⁸

Methods

Study designs and populations

SHIP was conducted from 2008–2012 in a North-eastern German region.¹⁶ A general population sample of 10,000 individuals was drawn from local population registries; 4420 individuals aged 20–81 years participated in the study, and 2391 participants without asthma or lung disease provided CRF measurements. DEGS1 was conducted from 2008–2011 among the residential German population aged 18–79 years.¹⁷ The final sample comprises 8151 men and women, including 4192 first-time participants and 3959 persons who already participated in the German National Health Interview and Examination Survey 1998; 2768 participants without asthma and lung disease had CRF measured. NHANES is a repeated cross-sectional examination of representative samples of non-institutionalised US adult civilians aged 20–85 years;¹⁸ sample sizes of NHANES in 1999/2000, 2001/2002 and 2003/2004 were 4880, 5411, and 5041, respectively. CRF was measured in 2314 NHANES participants aged 20–49 years without asthma and lung disease. Study participants provided written informed consent. The study protocols were approved by the Ethics Committees of the University of Greifswald (SHIP) and the Charité-Universitätsmedizin Berlin (DEGS1), and the National Center for Health Statistics Ethics Review Board (NAHNES). In total, 7358 individuals were available for data analysis.

Measures

Assessment of cardiorespiratory fitness and alcohol consumption

A symptom-limited cycle ergometer test (Ergoselect 100, Ergoline, Germany) was performed in SHIP according to a modified Jones protocol: 3 min of rest, 1 min of unloaded cycling, stepwise increases in work load of 16 watts per min, and 5 min of recovery.¹⁹ Gas exchange and ventilatory variables were analysed using a VIASYS HEALTHCARE system (Oxycon Pro, Combitox mask); peak VO₂ was defined as the highest 10-second average of VO₂ during the last minute of exercise.¹⁹ In DEGS1, a submaximal incremental cycle ergometer test (Ergosana Sana Bike 350/450, Ergosana, Bitz, Germany) was performed according to a World Health Organization protocol: starting workload of 25 watts; stepwise increase in workload of 25 watts every 2 min until 85% of the estimated age-specific maximum heart rate (220 minus age) was exceeded, the maximum intensity level of 350 watts was reached or study personnel terminated testing.¹⁷ Only participants aged 18–64 years who were assessed as suitable for CRF testing according to the ‘Physical Activity Readiness Questionnaire’ were included.¹⁷ The submaximal treadmill exercise protocol of NHANES included a two-minute warm-up, two three-minute exercise stages, and a two-minute cool-down period.²⁰ The objective was to provoke a heart rate that was approximately 75% of the age-predicted maximum heart rate (220 minus age). CRF testing was offered to participants aged 20–49 and various exclusion criteria (e.g. physical functioning limitations, cardiovascular limitations) were applied.²⁰ In DEGS1 and NHANES, peak VO₂ was estimated using an American College of Sports Medicine equation:²¹  

$3.5ml/min/kg+12.24×(maximalpower)/(bodyweight)$

The outcome variable for all datasets was peak VO₂ (in ml/kg/min).

SHIP and DEGS1 evaluated average alcohol consumption using beverage-specific quantity-frequency questionnaires: number of days with alcohol consumption and the number of alcoholic drinks consumed on a typical drinking day over the last 30 d. Alcohol consumption (ethanol in g per day (g/d)) was calculated by multiplying frequency and amount of alcohol from beer, wine, and spirits, respectively, using standard ethanol contents of 4.8% (by volume) in beer, 11% in wine, and 33% in spirits for conversion.²² SHIP did not elicit heavy episodic drinking and drinking histories among current abstainers. In DEGS1, heavy episodic drinking was assessed using an item of the Alcohol Use Disorders Identification Test (i.e. six or more drinks per occasion),²³ but information regarding drinking history among former drinkers was unavailable. In NHANES, average daily alcohol consumption was determined from the reported number of drinking days and the average number of drinks per day when drinking alcohol over the past 12 months. We used 12.5 g alcohol per standard drink for conversion into g/d.²⁴ Heavy episodic drinking was defined as the consumption of five or more drinks per occasion.²⁵ NHANES asked whether participants had at least 12 drinks during their lifetime. Based on this item, current abstainers were grouped into former drinkers and lifetime abstainers. We used average alcohol consumption (g/d) and heavy episodic drinking as exposure variables.

Covariates

Educational attainment (in years of school education completed) was categorised as <10 years, 10 years, >10 years. Smoking was grouped into never, former or current smoking. SHIP-participants who engaged in leisure-time physical activity during the summer or winter for at least two hours per week were classified as being physically active.²⁶ In DEGS1, physical activity was defined as 2.5 h per week of physical activity with sweating and breathlessness.²⁷ In NHANES, physical activity was assessed based on report of any moderate or vigorous activities lasting ≥10 min in the past 30 d.²⁸

Statistical analyses

First, average alcohol consumption was categorised as current abstainer, >0 to <20 g/d, 20 to <40 g/d, 40 to <60 g/d, ≥60 g/d for bivariate analyses. Selected participant characteristics were compared across groups of average alcohol consumption using age-adjusted means for continuous covariates and age-adjusted percentage values for categorical covariates. Second, we performed multiple linear regression to examine the association of average alcohol consumption (g of alcohol per day as a continuous variable) and peak VO₂ in SHIP (since SHIP was the only sample that directly measured peak VO₂). Fractional polynomials with a ‘spike at zero’ were used to determine the best-fitting dose-response relationship²⁹ of alcohol and peak VO₂. Fractional polynomials differ from regular polynomials in that they allow logarithms, they allow noninteger powers, and they allow powers to be repeated. The best fitting model, i.e. the one minimising the model deviance, was selected when the best fitting model was non-linear.^29,30 We plotted adjusted peak VO₂ levels against average alcohol intake using a marginal effect plot with 95% pointwise confidence intervals.³¹ Third, pooled analysis of the five samples was conducted using a fractional polynomial model. Fourth, we examined the association between heavy episodic drinking and peak VO₂ using a linear regression model based on DEGS1, NHANES 1999/2000, NHANES 2001/2002 and NHANES 2003/2004 (because heavy episodic drinking was not assessed in SHIP), with and without adjustment for average alcohol consumption. We tested for interaction effects between average alcohol consumption and heavy drinking. All models were adjusted for age, sex, education, smoking and physical activity. In addition, we examined associations of average alcohol consumption and heavy episodic drinking in men and women separately. The pooled models additionally included a study indicator variable. Because participants might reduce or quit alcohol consumption due to poor health (the sick-quitter hypothesis),³² we performed sensitivity analysis in NHANES and included a covariate in the multivariable model that separated never and former drinkers, along with polynomial transformations of average alcohol consumption among those with positive alcohol consumption.^22,29 Further, we examined associations of average alcohol consumption and heavy episodic drinking with peak VO₂ after exclusion of current and former smokers. Since participants of the CRF modules were younger and more physically active due to exclusion criteria,^17,33 we estimated a logistic model with lasso penalization³⁴ to estimate probabilities of taking part in the exercise testing subsamples. The regression models for associating alcohol consumption with peak VO₂ were weighted with the inverse of these participation probabilities.³⁴ Statistical analyses were performed using Stata 14.2 (StataCorp., Texas, USA).

Results

Table 1 shows the age-adjusted characteristics of the samples according to average alcohol consumption categories.

Average alcohol consumption showed an inverted U-type dose-response relationship with peak VO₂ in SHIP (overall p-value = 0.0006) and in pooled analysis of the five participating samples (overall p-value <0.0001) (Figure 1, Table 2). Current abstainers had lower adjusted peak VO₂ values than participants consuming 10 g/d in SHIP and in pooled analysis (linear regression coefficient (β) = −0.93 and −1.90, respectively). SHIP participants with an alcohol intake of 20, 30 or 40 g/d had higher peak VO₂ than to those consuming 10 g/d (β = 0.33, 0.35 and 0.28, respectively). In pooled analysis, individuals with an average alcohol consumption of 20, 30 or 40 g/d g/d had higher peak VO₂ (β’s of 0.27, 0.29, 0.17, respectively), and those consuming 50 and 60 g/d had lower peak VO₂ levels than participants drinking 10 g/d (β = −0.06 and −0.31, respectively). The association between average alcohol consumption and peak VO₂ was modified by age (p-interaction in SHIP<0.0001; p-interaction in the pooled data = 0.0267). Sex-specific analyses for the association of average alcohol consumption and peak VO₂ revealed similar inverted U-shaped associations (Table 3, Figure 1, p-interaction = 0.675). In additional analyses of NHANES, we did not find significant differences in peak VO₂ between former drinkers and lifelong abstainers (β = 0.18; 95% confidence interval (CI): −1.49 to 1.86). Regression estimates were similar after exclusion of former and current smokers or when SHIP data were excluded from the pooled analyses (Supplementary Material Tables 1 and 2). Results for associations between average alcohol consumption and peak VO₂ for each sample are available in the Supplementary Material Table 3.

Figure 1.

Marginal effect plots with 95% pointwise confidence intervals for overall and sex-specific associations between average alcohol consumption and peak oxygen uptake (VO₂) in the pooled data.

Note that regression models were adjusted for age, sex, school, smoking status, and physical activity. Pooled data from the Study of Health in Pomerania, German Health Interview and Examination Survey, US National Health and Nutrition Examination Survey (NHANES) 1999–2000, NHANES 2001–2002, NHANES 2003–2004.

Open in new tabDownload slide

We regressed peak VO₂ on heavy episodic drinking and found no association in the total dataset and in sex-specific analyses of the pooled data (Table 4, Supplementary Material Table 4). The effect estimate for heavy episodic drinking was attenuated after adjustment for average alcohol consumption (Table 4-model 2). Associations of heavy episodic drinking and peak VO₂ were not modified by sex (p-interaction=0.831) or age (p-interaction=0.862). Heavy episodic drinking did not modify the average alcohol-peak VO₂ association in the pooled analysis (p-interaction=0.2673).

Discussion

This pooled analysis of five independent population-based studies indicated a curvilinear relation between average alcohol consumption and peak VO₂. An inverted U-shaped association was observed between average alcohol consumption and peak VO₂ in the total sample and in both genders. To our knowledge, only three previous population-based studies addressed the relationship between alcohol consumption and CRF. Similar to our findings, a cross-sectional US community sample of 597 men, aged 16–69 years, showed that high alcohol intake and non-drinking was associated with lower peak VO₂ levels in adult men, while moderate drinkers (defined as 4–15.8 g/d) had higher levels of peak VO₂ .¹⁴ In contrast, a population-based study of 939 middle-aged Finnish men found no association of average alcohol consumption and peak VO₂.¹³ Likewise, a cross-sectional study of 187 Japanese men and women found no association between average alcohol consumption and peak VO₂.¹⁵ Yet, in contrast to the present study, existing studies have not used non-linear modelling strategies to examine non-linear dose-response relationships.

The inverted U-type association observed in our study suggests that moderate drinkers have better CRF than non-drinkers and participants with high alcohol consumption, which is in agreement with findings from a large number of observational studies showing U- or J-shaped relationships with cardiovascular^5,6 and respiratory outcomes.^9,10 The effect size of low-to-moderate alcohol consumption on peak VO₂ (∼1–1.5 ml/kg/min) is relatively small, given that endurance training can augment peak VO₂ by 10–30%.¹¹ Based on biological considerations, moderate drinking could exhibit beneficial effects on CRF by increasing insulin sensitivity, anti-inflammatory effects, enhancing levels of high-density lipoprotein cholesterol and adiponectin, improving pulmonary function, and by lowering of fibrinogen and lower blood pressures.^35–37 Several detrimental actions of alcohol might explain lower CRF among participants with high average alcohol intake, including suppressed fat oxidation leading to insulin resistance, hypertriglyceridaemia, and weight gain and adiposity; inhibition of the pulmonary immune response and airway epithelial permeability, and airway leakage; decrease amount of glucose and amino acids used by the skeletal muscles with decrease in energy supply; decreased muscle capillarity, lower glycogen utilization levels; lower fatty acids; lower liver glucose output; and ethanol-induced hypoglycaemia.^38–40

There is ongoing debate as to whether the protective association of light-to-moderate alcohol consumption with cardiovascular disease is causal.^35,41–43 Confounding is a likely explanation, since socioeconomic groups that drink moderately may exhibit lifestyle habits that translate into lower cardiovascular risk.⁴⁴ Evidence also indicates risks of bias from reverse causation, with those at poor health or taking medication being more likely to reduce or quit alcohol consumption^32,43 Accordingly, we separated life-long abstainers from former drinkers in additional analyses but could not find differences between ex-drinkers and lifetime abstainers. Nevertheless, we cannot rule out the possibility that selection bias, reverse causation or unmeasured confounding explains lower peak VO₂ among abstainers.

Heavy episodic drinking was not related to peak VO₂. This finding is in line with randomised trials indicating that acute excessive drinking does not affect peak VO₂ .^45,46 Previous research suggested that heavy episodic drinking counteracts the putative protective effect of low-to-moderate moderate consumption.⁶ However, the association between average alcohol consumption and peak VO₂ persisted after adjustment for heavy episodic drinking, and the association was not modified by heavy episodic drinking.

Our study has several limitations. Several exclusion criteria were applied for fitness testing and this possibly resulted in younger and more physical active participants in the analytical datasets. Although we performed inverse probability weighting as a strategy to reduce selection bias, we cannot rule out the possibility that older subjects and those with lower CRF are underrepresented in our datasets. SHIP was the only study with a direct measure of peak VO₂ using gas analysis, the gold standard measure for CRF,¹¹ whereas peak VO₂ was estimated in DEGS1 and NHANES. In addition, SHIP and NHANES used cycle ergometers, while NHANES used a treadmill test. It is known that peak VO₂ derived from cycle ergometer is lower than peak VO₂ from treadmill-based testing.⁴⁷ Direct measurement vs estimation of peak VO₂ and exercise modality might explain why we found higher peak VO₂ in DEGS1 and NHANES than in SHIP. However, peak VO₂ divided by body weight was used as an outcome, which minimises differences between cycle and treadmill ergometer.⁴⁷ The studies implemented different instruments for alcohol consumption: SHIP and DEGS applied beverage-specific quantity-frequency questionnaires, while NHANES used standard drinks. Assessment using standard drinks underestimates average volume.⁴⁸ We assume that the measurement error of average alcohol intake is independent of peak VO₂ (i.e. non-differential), diluting the alcohol-peak VO₂ association.⁴⁹ We acknowledge that health status among participants with low CRF may affect the error (e.g. due to recall bias), which could result in a differential error.⁴⁹

In conclusion, our pooled analysis of 7358 participants from five independent samples suggests an inverted U-shaped link of alcohol with CRF. We have not solved the alcohol-CRF puzzle, and we offer several reasons for caution before drawing a cause-effect conclusion. More prospective observational and experimental research is needed, including research on potential biological mechanisms. Alcohol consumption might vary over the life course and is typically lower among older adults. Thus, prospective studies of the relation of temporal variation of alcohol consumption and CRF are needed to provide further evidence as to whether moderate alcohol consumption is associated with better CRF.

Conclusion

This general population study revealed that low-to-moderate drinkers have higher peak VO₂ levels than abstainers and those with high alcohol consumption.

Author contribution

SEB, JDF, KL and MFL contributed to the conception and design of the work. SEB, JDF, SG, MD, MRPM, RE, SBF, HJG, MB, GBMM and HV contributed to the acquisition, analysis or interpretation of data for the work. SEB, JDF, KL and MFL drafted the manuscript. SG, MD, MRPM, RE, SBF, HJG, MB, GBMM and HV critically revised the manuscript. All coauthors gave final approval and agree to be accountable for all aspects of work ensuring integrity and accuracy.

Acknowledgments

The authors wish to thank all study participants and members of the SHIP, DEGS1, NHANES study teams.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship and/or publication of this article: SG reports grants from German Ministry of Education and Research; grants and personal fees from Boehringer Ingelheim, Novartis Pharma, Berlin Chemie, Actelion, Bayer Schering, Roche Pharma. RE received financial support from CareFusion (Wurmlingen, Germany). HJG reports grants from German Research Foundation, German Ministry of Education and Research, DAMP Foundation. No conflicts of interest are declared by SEB, JDF, MD, MRPM, SBF, MB, GBMM, HV, KL, MFL.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from the German Federal Ministry of Education and Research (BMBF, grant 01ZZ96030, 01ZZ0701) for the Study of Health in Pomerania. The German Health Interview and Examination Survey was financed through the German Ministry of Health (BMG). The US National Health and Nutrition Examination Surveys were funded by the National Center for Health Statistics (NCHS), a part of the Centers for Disease Control and Prevention, Department of Health and Human Services. The sponsors had no role in the design of the study, the collection and analysis of the data, or the preparation of the manuscript.

References

Author notes