Drinking volume and patterns: back to basics

The thoughtful paper by Wells et al. (2005) illustrates, among other things, three important issues that are common to much of the current research on the health effects of alcohol. First, the change of the paradigm from drinking volume to drinking patterns as the primary determinant of alcohol-related harm needs to be supported by solid evidence. While there is a respectable body of evidence on the importance of drinking pattern (heavy episodic drinking/binge drinking versus regular moderate intake), it is by no means conclusive in the sense that the volume of drinking is no longer important when heavy episodic drinking has been taken into account (Rehm et al. 2003). ‘Negative’ studies, such as this, are essential to remind us of the inconsistent nature of the literature and of the need to continue measuring the volume of alcohol consumption in epidemiological studies. This leads to the second issue—measurement. Despite the progress over the last decades, virtually all studies still rely on self-report. The limitations of self-reported exposures (and outcomes, as in this study), even when the responders have the best will to tell the truth, are well recognised. This is why nutritional epidemiologists are so desperate to use biomarkers. Unfortunately, the biomarkers of alcohol misuse presently available are not sufficiently reliable, particularly to identify episodic heavy drinkers and/or to distinguish them from regular drinkers. Self-reported alcohol consumption is bound to be affected by misclassification, both random and non-random. For example, there is emerging evidence of differential reporting of diet by socioeconomic status (Hebert et al. 2002). It is likely that similar bias occurs with alcohol and that it also relates to characteristics other than social status, possibly including alcohol-related health outcomes. The direction of the bias is often unpredictable, and it is probably one of the reasons why studies on drinking volume versus patterns are inconsistent. The third issue is also methodological. If drinking pattern modifies, rather than explains, the health effects of alcohol, studies need to be large enough to study interactions. It is reasonable to assume that the effects of a given volume are more pronounced in heavy episodic drinkers than in persons who do not drink large amounts per occasion (Rehm et al. 2003). However, demonstration of such interaction (effect modification) requires large sample size. For example, a study with a 1000 participants is probably too small. Once the sample is broken down by, for example, gender, age and binge drinking, the individual cells become too small for meaningful analyses of interactions. The misclassification of alcohol intake would obscure further the underlying associations. Some time ago, Peto suggested that most clinical trials are too small to be useful (Peto 1982). The same may apply to alcohol-related research: there is now a need for large studies and good measurements to disentangle the effects of drinking volume and patterns and their interaction.