Alcohol consumption and risks of cardiovascular disease and all-cause mortality in healthy older adults

Excess alcohol consumption is a leading contributor to the global burden of disease and a major risk factor for mortality.¹ Yet, prior studies suggested that moderate alcohol consumption may be associated with a lower risk of cardiovascular disease (CVD) events.^2,3 However, this evidence is based on data from younger individuals, and confirmation in older adults is lacking. Thus, we sought to investigate the risk of CVD events and all-cause mortality associated with alcohol consumption in initially healthy, older individuals.

For these post hoc analyses we used data from the ASPirin in Reducing Events in the Elderly (ASPREE) trial, which investigated the effect of low-dose aspirin on disability-free survival in healthy older people.⁴ ASPREE recruited 19 114 community-dwelling individuals in Australia and the USA aged ≥70years (≥65 for US minorities). Participants entered the study without prior CVD events, diagnosed dementia or physical disability. Incident CVD events and cause of death were adjudicated outcomes. CVD events included coronary heart disease death, non-fatal myocardial infarction, fatal and non-fatal stroke, non-coronary cardiac or vascular death, and hospitalization for heart failure. All participants provided written informed consent. The study was approved by local ethics committees and registered on Clinicaltrials.gov (NCT01038583).

Information on alcohol consumption (days of drinking per week and average standard drinks per day) was assessed by self-report questionnaire at baseline. Based on this information, the alcohol intake was calculated as grams per week. For US participants a standard drink was equivalent to 14 g and 10 g for Australian participants. For the present analyses, we excluded former alcohol consumers (n = 1136), who may have stopped alcohol consumption for various health reasons, possibly introducing bias from reverse causality. Alcohol consumption was categorized as 0 (never drinkers), 1–50, 51–100, 101–150, and >150 g/week. Cox proportional hazards regression analyses evaluated the association of alcohol consumption and the endpoints of interest, adjusting for age, gender, body mass index (BMI), smoking, non-HDL-cholesterol, diabetes, intake of statins, and randomization as per trial intervention. Analyses were also stratified by gender. Furthermore, the association of log-transformed alcohol g/week with endpoints was evaluated using restricted cubic splines, with median consumption as reference. Analyses were performed using R version 4.0.2.

Of 17 978 participants with median age 74 years [interquartile range (IQR) 71.7–77.7], 10 253 (57.0%) were female, 7789 (43.3%) were current or former smokers and mean BMI was 28.1 kg/m² (standard deviation 4.7). A total of 3336 (18.6%) participants reported 0 g of alcohol per week, 6697 (37.3%) reported 1–50 g/week, 3545 (19.7%) reported 51–100 g/week, 2797 (15.6%) reported 101–150 g/week, and 1603 (8.9%) reported >150 g/week. During follow-up (median 4.7 years), 756 CVD events [median time to event 1657 days (IQR 1238–1825)] and 793 deaths [1720 days (IQR 1224–1825)] occurred.

There was a significantly reduced risk of CVD events for individuals consuming alcohol of 51–100, 101–150, and >150 g/week, compared to never consuming alcohol (Figure 1). Stratified by gender the results were similar. Consumption of 51–100 g/week was also associated with a reduced risk of all-cause mortality; an effect observed for the overall population, and in gender-stratified analyses. Restricted cubic splines analysis confirmed this association between alcohol consumption and all-cause mortality, as well as with incident CVD events in the overall population (Figure 1). In agreement with the categorical analyses, from the splines, the lowest risk was observed for a log-transformed value of ~4 units, corresponding to ~55 g/week (or 6 standard drinks/week). When stratified by gender, there was a slightly u-shaped association in males, but not in females.

Figure 1.

Association of alcohol consumption and incident cardiovascular disease and all-cause mortality. (A) The forest plot panels depict hazard ratios and 95% confidence intervals from Cox regression analyses for the association of categorized alcohol consumption with incident cardiovascular disease and all-cause mortality. The regression analyses were adjusted for age, gender, body mass index, smoking, non-HDL-cholesterol, diabetes, intake of statins, and randomization as per study intervention and were also stratified by gender (adjusted for age, body mass index, smoking, non-HDL-cholesterol, diabetes, intake of statins, and randomization as per trial intervention). (B) The other panels depict splines evaluating the adjusted association of alcohol consumption per week and incident cardiovascular disease and all-cause mortality. For the splines, alcohol consumption per week was log-transformed and an alcohol consumption of 4 (on the log scale) corresponds to ~55 g/week, and 5 on the log scale corresponds to ~150 g/week. The blue line indicates the estimated hazard ratio together with shading for the 95% confidence interval. The grey shaded area above the x-axis indicates the density distribution of alcohol consumption. CVD, cardiovascular disease.

Similar to studies in younger populations, we confirmed the reduced risk of CVD associated with light to moderate alcohol consumption.² The splines gave a visual suggestion of different relationships in males and females although this was not apparent from the hazard ratios of the specific categories we used. Gender differences regarding alcohol consumption have been reported before and showed evidence of increased CVD risk of excessive consumption in males, but not in females.³ In contrast to prior data, we also showed a reduced risk for all-cause mortality with moderate alcohol consumption. This finding could be explained by the healthy study population, as prior analyses from the UK Biobank suggested that physical activity may attenuate health risks associated with increased alcohol consumption.⁵

Strengths of our analyses include the well-characterized population with rigorous baseline screening, adjudicated outcome assessment, and little missing data. Limitations include a potential healthy volunteer bias due to the selection process of a clinical trial, perhaps reflected in underrepresentation of individuals consuming >150 g/week of alcohol. In view of the collinearity between alcohol use with socio-economic status and lifestyle factors, unmeasured or underestimated confounding may also affect these analyses.

In conclusion, in a large population of initially healthy, older individuals we show a reduced risk of incident CVD events and all-cause mortality for moderate alcohol consumption. These findings add to the existing body of evidence, as healthy older people have frequently not been included in prior analyses. Further research is warranted to evaluate causal biological effects and behavioural advantages of social drinking and engagement.