I have read with interest the recent paper by Liang & Chikritzhs in Addiction and venture to make some comment on it, as it is an area of the alcohol story in which I have a long-standing interest through the British Regional Heart Study (BRHS; http://www.ucl.ac.uk/primcare-popsci/brhs) .
The term ‘sick quitters’ is unfortunate, in that it carries the unintended implication that the person who has quit drinking or reduced their alcohol intake has done so for reasons associated with alcohol usage. Although it is usually made clear that being a ‘sick quitter’ need not be due to alcohol, most readers still assume this to be so. Any study of ex-drinkers or those who have markedly reduced their alcohol intake will show that only a minority have done so through an illness associated directly with the use of alcohol. Most ex-drinkers and most of those who make significant reductions in their intake usually do so as they age and because of physical or mental ill health, medication, economic factors or other changes in life-style conducive to a reduction in alcohol consumption [4,5]. Abandoning this term would be a useful starter in fresh thinking about alcohol usage and health.
There is an assumption, rather too easily made, that most modern epidemiological studies are careful not to include ‘sick quitters’ within the non-drinking category and will use life-long abstainers or light drinkers as their baseline. The problem with this assumption is that different authors vary in their definition of ‘sick quitters’; few studies use regular (i.e. with repeated measurements) light drinkers as the baseline and the use of life-long abstainers has other problems (vide infra).
Liang & Chikritzhs have argued that the phenomenon of the ‘sick quitter’ may be viewed as similar to the loss of subjects to clinical trials.
Subjects who begin as ‘drinkers’ but who ultimately stop drinking while participating in a cohort study are analogous to subjects in clinical trials who drop out of treatment for reasons which are related to the prognosis itself. In clinical trials, subject dropout potentially introduces a bias, because people who complete a particular treatment may, at the outset, be predisposed to have a better outcome. In the same way, people who do not become ex-drinkers may be predisposed to have better health outcomes .
This is a powerful and sustainable argument and should be considered , although reduction of alcohol intake, e.g. becoming an occasional drinker, is a more likely culmination than complete cessation of drinking.
Essential to any understanding of the analysis of the outcome in various drinking categories is a full display of the health and life-style characteristics of the subjects in each drinking category. These apparently crude differences may be crucial to an understanding of why different drinking categories have different outcomes [7,8]. Usually the table showing these features, if it is provided at all, is brief and highly selective, and gives little indication or true measure of the factors, singly or in combination, that might be affecting outcome in these subjects. Remarkably few papers carry a comprehensive table of health characteristics sufficient to enable the reader to assess the risks of morbidity or mortality in the various groups of subjects. In the end, it is this that matters, and not whether or not a person has been a previous drinker at a particular level. While most epidemiological studies make elaborate adjustments for many of the factors involved, this is a process fraught with weaknesses and liable to leave many confounders unconsidered. Most studies which do show the characteristics of the drinking categories reveal that non-drinkers have greater morbidity, more medication and worse life-style characteristics than other groups [7,8]. This situation worsens as the subjects age. Light regular drinkers tend to have a constellation of beneficial factors conducive to a low risk of cardiovascular disease in particular and mortality in general .
A major point in the paper under consideration is that subjects developing ill health and going on to medication will, as a consequence, either reduce their alcohol intake or give it up completely. This has also been shown in other studies, including the BRHS . The protagonists of the ‘protective’ effect of moderate drinking observe that in prospective studies assessing change in alcohol intake at several points in time over many years of follow-up, ‘subjects who decrease their intake are more likely to subsequently develop adverse health outcomes’. This could well be an example of thinking in terms of reverse causation.
The authors record [1; Table 2] that life-time abstainers were more likely than light drinkers to report less than good, very good or excellent health status. This has been the finding in other studies as well, and one of the problems with using life-time abstainers as a baseline is that the reasons for being a life-time abstainer differ in different communities and many who claim to be life-long abstainers are often ex-drinkers of long duration. Life-long abstainers are too unusual and usually too small a group to be used as a baseline [10–12].
If life-long abstainers are not suitable as a baseline group, and the non-drinking category is beset with problems, what should we use as a baseline? In the best epidemiological tradition, perhaps one should look at the category with the healthiest characteristics and the best outcomes in morbidity and mortality: usually the light regular drinkers. The BRHS has, in the past, suggested that the ‘occasional/light drinking category i.e. less than 15 drinks/week, provides a large and satisfactory baseline group for comparative purposes’. However, the ‘occasional’ drinking group may have health characteristics that are less satisfactory than the light regular drinkers of one to two drinks/day and so further analyses, particularly in prospective studies with repeated checks on intake levels, should be carried out to compare the occasional and light regular categories over time . I suspect that the light regular drinking category would be the most reliable to use; one can hazard a guess that all categories below this level will have increased morbidity and mortality, and all above this level will have greater morbidity and mortality.
This issue is much more than an academic argument or debate between epidemiologists. Major public health and social wellbeing issues are involved. That alcohol in moderation may give great pleasure to a great many people is undeniable. That alcohol usage in many situations and in excess amounts is associated with ill health and social disorder is equally undeniable. The protagonists of the ‘protective’ hypothesis should consider that they may be wrong, or at least that the claims made may be exaggerated by errors of methodology [14,15]. They would do well to recall Stephen J. Gould's observation that: ‘We ought to be strongly suspicious of ideas that are enormously comforting’.