In a new study published in the BMJ researchers report brain changes associated with chronic alcohol consumption, even at moderate levels. Roundup comments accompanied this analysis.
Title, Date of Publication & Journal
Moderate alcohol consumption as risk factor for adverse brain outcomes and cognitive decline: longitudinal cohort study
Published: June 6th 2017
Study’s main claims – and are they supported by the data
The paper does not support the claim that moderate alcohol consumption is associated with negative effects on brain function, however the paper does support the claim that moderate alcohol consumption is associated with abnormal effects on brain structure.
The paper uses ten tests of brain function, but only one significant result was reported in the paper (decline in lexical fluency, meaning how many words beginning with a specific letter a participant can generate in one minute). We would expect one result to be significant by chance alone.
Furthermore, the results for lexical fluency are extremely counterintuitive – the authors’ model suggests that heavier drinkers have a higher IQ to begin with, before declining to about the same IQ as very light drinkers by age 70. The authors report this as being a ‘faster rate of decline’ in heavy drinkers, which is true but misleading; since heavy drinkers start from a position of higher IQ (probably because of some confounding variable the authors were unable to adjust for), in order for them to finish in the same place as everyone else they obviously have to decline faster. This is a very good illustration of ‘reversion to the mean’ effect, where groups that do particularly better than average in one time period by sheer chance appear to decline faster than everyone else because those chance factors do not persist. There is no dose-response relationship by the end of the study; the lightest drinkers score the best at the end of the test, and the second-lightest drinkers score the worst, which casts doubt on there being a simple relationship between alcohol consumption and brain function.
(A) Grey Matter
There is a significant difference in grey matter density between drinkers and non-drinkers, which is dose dependent (grey matter becomes less and less dense the more alcohol you drink). The paper uses ‘voxel based morphometry’ to reach this conclusion, which is an extremely complicated statistical technique for detecting changes over a time series of MRI scans. It is extremely prone to false positives if misapplied (such as, in other studies, detecting brain activity in a dead salmon) but the dose-response relationship and the fact that the results are consistent with the authors’ theoretical model and other conclusions suggests the analysis is appropriate.
(B) Hippocampal Volume
There is no significant difference in hippocampal volume for light drinkers (<14 units/week) compared to non-drinkers. In moderate drinkers (<30 units/week) there is a difference in right-side hippocampal volume but not left-side hippocampal volume compared to non-drinkers. In very heavy drinkers (>30 units/week) there is a difference in both left- and right-side hippocampal volume compared to non-drinkers.
In both cases although the changes are statistically significant (meaning that the abnormalities almost certainly exist), there is no evidence on how clinically significant the change is – the authors do not say whether the rate of loss due to alcohol is a lot or a little compared to the rate of loss due to other factors (such as old age). Table 3 implies that the lost brain volume due to being a very heavy drinker compared to an abstainer is around 1% of total brain volume, which is approximately equivalent to the amount lost per year in a 60 year-old. However, there is no evidence in this study linking this loss to any negative cognitive effects, even though these effects were tested for.
There are two very great strengths of this study; the fact that data on the participants goes back 30 years and the fact that a dose-response relationship is found for its key conclusion (which is that there is a difference in brain structure between moderate drinkers and light drinkers).
The biggest limitation of the study is that the population is nonrandomised and not very representative – as they are all civil servants from the 1980s the population tends to be more male, more middle class and more intelligent than the general population. Although the authors make a very considered and sustained effort to correct for confounding factors, this is incredibly difficult to do.
The study doesn’t have a lot to say about the relationship between women’s brains and drinking; this is not a criticism of the study but an observation that there were fewer women than men working in the civil service in the 1980s and fewer of these women drank heavily, making it difficult to draw robust conclusions.
MRI – Magnetic Resonance Imaging (a method of scanning the brain).
Dose-response – If X causes Y, then we would normally expect giving more X to cause more Y. If the relationship being examined has this property, it is said to be ‘dose dependent’, and the dose-response is how responsive Y is to changes in X.
Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.