<i>J</i>‐SHAPE OR LINEAR RELATIONSHIP BETWEEN ALCOHOL CONSUMPTION AND DEPRESSION: DOES IT MATTER?

The prospective cohort study of women by Alati et al. (2005) reported a J-shaped association between alcohol consumption and depressive symptoms at 5 years and linear associations at both baseline and at 14 years. Also, it reported that the depression prevalence rate among former drinkers was similar to that of abstainers at all measurement points, suggesting that ‘sick quitting’ is not responsible for the J-shaped association. Although this paper made for interesting reading, we do have some concerns with respect to design and interpretation of the study. First and foremost, we suspect that the same people were responsible both for the J-shaped curve at 5 years and the linear curve at baseline and 14 years. Overall, only 10% of the sample were abstainers at all three time-points, and very few respondents (approximately 1%) changed from being moderate or heavy drinkers to abstainers from cycle to cycle, so it is likely that the different kinds of risk relationships were caused by the same people in terms of fluctuation of classification between abstainer and light drinkers. The categorization of abstainers versus light drinkers at any given time-point may have been arbitrary for very light drinkers, and we hypothesize that the majority of movers into or out of the abstainer group was caused by the same people, mainly very light drinkers. It is highly unlikely that very light drinking, such as drinking small amounts less than weekly, has any health effect compared to abstention (Rehm et al. 2003a). Collapsing the abstainers and light drinkers of Alati et al. (2005) into one category, the relationship between volume of alcohol consumption and depressive symptoms showed a linear trend at all three time-points, although confidence intervals overlapped (see Table 1). Thus, Table 2 in Alati et al. (2005) seems to be based on overall unstable relationships. This assertion is further reinforced by the fact that, when confounding was controlled, almost all the discussed relationships disappear. Secondly, we are somewhat worried by the strict exclusion criteria applied by Alati et al. (2005), where only 63% of the initial sample were included in the statistical analyses of the study. The authors give evidence that those excluded differed regarding their characteristics on key variables studied. Overall, the analysis sample consisted of less heavy drinkers and less depressed people compared to those excluded. We do not know whether the relationship in the full sample between alcohol and depression would have been different, but this could be easily tested with different analytical strategies. The analytical strategies are the last point of our concern. Although the authors had a longitudinal sample where they could perform more definitive analyses on the causality, or at least the temporal succession of characteristics (for an introduction to potential designs, see Finkel 1995), they chose to limit themselves to three cross-sectional analyses. This restriction is wasting the wealth of the data available where incidence of depression, for example, could be associated with drinking patterns at the previous measurement point, and heavy drinking could be associated with depressive symptoms at the previous time-point. For some of these analyses, imputation of missing values or Generalized Estimating Equations (GEE) techniques may be helpful, instead of working on a restricted sample of less than two-thirds of the overall sample, with unclear possibilities to generalize the results. So, overall, do these results matter? We fear they do not, either in terms of scientific contribution or in impact on policy, as the authors suggest. The analyses shown were inconclusive, but with access to this large and longitudinal data set, other analyses may be possible to gain more meaningful insight into the relationship between alcohol use and depression. Finally, some words of caution regarding conclusions about abstainers. While the majority of the adult general population abstains globally from alcohol (Rehm et al. 2003b), abstaining in different cultures is associated with different meaning and, thus, results on abstaining should always be considered culture-specific. If being an abstainer means to be part of a small minority, sometimes stigmatized, as in some regions of Europe, or to be part of the normal way of life in a culture where more than 90% abstain, as in some regions of India, this clearly has different meanings as well as associations with other variables influencing health (e.g. socio-economic status, life-style, religion, etc.). As a result, many conclusions about abstention and health outcomes, especially without underlying biological processes, may simply reflect relationships independent of alcohol consumption.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.