Data was extracted from the UK Biobank cohort, a prospective population-based study comprised of more than 500,000 people between the ages of 40 to 69. This study only included people over age 60 and excluded anyone with incomplete data on alcohol consumption, never-drinkers, binge and former drinkers, leaving just over 135,000 people included in the final analysis. They were classified by their mean alcohol intake into one of four categories, occasional drinking, low-risk, moderate-risk, and high-risk; for example, low-risk was defined as an alcohol intake between 2.86 to 20g per day for men, or between 2.86 to 10g per day for women, while high-risk was defined as >40g per day for men and >20g per day for women (a standard drink is considered to be 14g of alcohol, as found in 12oz of regular beer or 3-5oz of wine).
Participants in the study were followed for a mean of 12.4 years. Compared with only occasional drinking (<2.86g per day), high-risk drinking was associated with the highest risks for all-cause (hazard ratio of 1.3), cancer (HR 1.39), and cardiovascular (HR 1.21) mortality. Moderate drinking was associated with both higher all-cause and cancer mortality, and even low-risk drinking was linked to higher cancer mortality (HR 1.11).
However, the presence of either health-related or socioeconomic risk factors modified these associations among low-risk and moderate-risk drinkers. The presence of health-related risk factors was based on a calculated score and included 49 “health deficits,” ranging from diabetes and high blood pressure to poor sleep and dental problems. Socioeconomic risk was assessed using the Townsend deprivation index (TDI), an evaluation based on residence only (not individually assessed status).
Among people with no health or socioeconomic risk factors, no excess mortality risk was observed among either moderate or low-risk drinkers. Yet among people with health-related risk factors, cancer mortality was increased with both low-risk (HR 1.15) and moderate drinking (HR 1.19), with moderate drinking also associated with an increase in all-cause mortality (HR 1.1). Similarly, the presence of socioeconomic risk factors increased both all-cause and cancer mortality among both low-risk and moderate drinkers.
Lastly, a preference for wine as well as drinking only with meals were both associated with very small reductions in all-cause mortality, but only among people with either health-related or socioeconomic risk factors. Small reductions in cancer mortality were also associated with drinking only with meals in the presence of either type of risk factor, while among people with socioeconomic risk factors, wine was associated with a small reduction in cancer mortality, and drinking with meals with lower cardiovascular mortality. Adopting both patterns was associated with a lower risk for all-cause, cancer, and cardiovascular mortality among people with risk factors, and mitigated some or all of the excess risk for people without these risk factors.
This study suggests that among people with no health or socioeconomic risk factors, there may not be an excess mortality risk if wine is the preferred choice for alcohol, and it is consumed only with meals. These two patterns of consumption appear to have benefit for people with and without risk factors, and in the former group, may carry a slight reduction in mortality risk.
Like many related studies, there are multiple considerations regarding confounders when interpreting this data. For one, the reference group was not people that never drink, but instead it was people that only drink occasionally. It’s quite possible that in comparison to “never-drinkers,” the risk associated with alcohol would be even more pronounced. But past studies have exposed the dangers of using “never-drinkers” as a comparison group because it includes many former drinkers that now abstain, or those that abstain because of specific health problems (selection bias), potentially skewing the results. To highlight this problem, a previous analysis of the UK Biobank cohort found that when selection bias is eliminated, there is an increase in cardiovascular disease risk even at very low levels of intake (<14 units per week), and the alleged J-shaped curve of alcohol intake is faulty. Indeed, a comprehensive global analysis published in the Lancet concluded that the “level of consumption that minimizes health loss is zero,” while other evidence points to increased cancer risk even at very low levels of intake.
Overall, this recent study points to a possible protective effect (or mitigation of risk) associated with wine intake and drinking only with meals (perhaps due to reduced absorption, antioxidant content, etc.), but a lack of a protective effect of alcohol even at low levels, modulated by health and socioeconomic status.