Naimi TS, Xuan Z, Brown DW, Saitz R. Confounding and studies of ‘moderate’ alcohol consumption: the case of drinking frequency and implications for low-risk drinking guidelines. Addiction 2012; pre-publication. doi:10.1111/j.1360-0443.2012.04074.x.
Aims Many observational studies suggest that increased drinking frequency is associated with reduced mortality among those with low-dose alcohol consumption. The purpose of this paper was to examine whether frequent drinkers consume lower-risk amounts during drinking days or have favorable risk factor profiles compared with those who drink less frequently, and discuss implications for the larger debate about the limitations of non-randomized studies about ‘moderate’ drinking and the development of low-risk drinking guidelines.
Methods Data from the 2008 Behavioral Risk Factor Surveillance System survey were used to characterize alcohol consumption characteristics and their relationship with risk factors among adult drinking men who consumed an average of fewer than two drinks per day and adult drinking adult women who consumed an average of less than one drink per day.
Results Those who drank relatively infrequently (14 or fewer days per month) consumed more during drinking days, were more likely to exceed the US Dietary Guidelines drinking limits (41.0% versus 9.7%) and had a larger proportion of drinking days that included binge drinking (13.4% versus 4.3%). Infrequent drinkers also had a higher prevalence of 13 of 15 risk factors assessed. Findings from analyses of those aged ≥ 40 years were similar.
Conclusions Among those with low average alcohol consumption, infrequent drinkers drink more during drinking days and have unfavorable risk factors profiles compared with more frequent drinkers, suggesting that confounding may contribute to favorable associations with ‘moderate’ average alcohol consumption and increased drinking frequency observed in non-randomized studies.
Background: Recent epidemiologic studies on the association between alcohol consumption and health outcomes have often shown that the average amount of alcohol consumed over a week (or other period of time) has less effect on health outcomes than does the pattern of drinking. Most studies conclude that regular, moderate consumption of alcohol, without binge drinking, is the pattern associated with the lowest risk of cardiovascular diseases and total mortality. The results of the present study support this premise by showing that regular moderate drinkers are much less likely to binge drink and show lower levels of essentially all cardiovascular risk factors.
Comments on the authors’ interpretation of their results: The authors of this paper take an unusual turn when it comes to discussing the implications of their results. They tend to down-play any potential health benefits that may be caused by alcohol or the pattern of drinking and infer instead that the favorable risk factors themselves may lead to the drinking pattern.
After presenting their results, the authors take considerable space to point out that essentially all risk factors are lower among regular moderate drinkers. While they show the inverse expected effects of education and income on drinking frequency, they then list ethnicity, being uninsured, being a current smoker, being obese, etc., as also being inversely associated with frequency of consumption. They do not, however, present results (as do most epidemiologic reports) of the effects after adjustment for the socioeconomic factors. Hence, the argument that such a large number of risk factors are simply confounders has little meaning if such factors are all strongly related to socioeconomic status.
Confounding and studies of moderate alcohol consumption: The issue of bias and confounding in observational cohort studies is a real problem and an issue worthy of discussion. The limitations of observational data are well-known and conclusions from observational studies should always be regarded with due suspicion. This paper is a timely contribution to the ongoing debate on the health effects of moderate alcohol consumption. However, Forum reviewers were concerned that the authors of this paper have presented an unbalanced view of the topic.
Forum reviewer Skovenborg has prepared detailed comments dealing with six of the points upon which the authors base their conclusions, as given below:
“(1) The lack of randomized trials of low-dose alcohol consumption: Randomized trials are very expensive, difficult to perform and involve serious ethical considerations. However, the lack of randomized, prospective studies is a general problem in public health policy. To my knowledge there are no long-term, randomized trials with morbidity and mortality outcome involving participants from the general population in areas like regular exercise or healthy diet. Studies of weight loss trials are most often associated with a neutral effect or increased mortality, and there is no evidence for weight loss conferring benefits among healthy obese subjects.1 A considerable number of people (8.4%) have adverse metabolic response to regular exercise.2 In a 24-year prospective follow-up after 6-year physical activity discordance in twin pairs, there was no mortality difference among monozygotic twins, suggesting considerable genetic confounding effects concerning the reduced mortality observed with physical activity.3
“A systematic review of dietary fat intake and prevention of cardiovascular disease found a small (9%) reduction of cardiovascular mortality, but no reduction of total mortality.4 We must conclude that epidemiology in general, and nutritional epidemiology in particular, is a difficult discipline. According to Karin B. Michaels, Harvard School of Public Health,5 randomized controlled allocation of diet is likely to be successful only for severe dietary deficiencies that can be reversed in a short period of time. Public health advice on diet, exercise, weight loss, smoking and alcohol all rest on pretty shaky pillars if randomized trials are to be considered condicio sine qua non.
“(2) Levels of evidence: According the Oxford Centre for Evidence-based Medicine 2009, the level of evidence of systematic reviews with homogeneity of cohort studies = level 2a, while the level of cross-sectional surveys = level 4.6 A cross-sectional telephone health survey like BRFSS (used in the present study) is especially vulnerable to information bias, and the low response rate of 54% points to selection bias. In this paper, the authors seem to focus on the splinter in their brother’s eye, but do not perceive the wooden beam in their own eye. Whether it is an appropriate method to discredit level 2a evidence with arguments of level 4 evidence is questionable, to say the least. An example: the participants of several large cohort studies with data on alcohol consumption and health outcomes like the Nurses’ Health Study and the Health Professionals Follow-up Study have comparable socio-economic factors to a large degree. Of the 89,538 participants in the Nurses’ Health Study, 32.2% were non-drinkers, and it is hard to imagine that these 28,835 non-drinking women were a specially disadvantaged group of nurses. Of the 48,493 participants in the Health Professionals Follow-up Study, 23.7% were non-drinkers, and it is also hard to imagine that these 11,519 non-drinking men were a specially disadvantaged group of Health Professionals.7
“(3) Drinking frequency and alcohol intake: The authors of the present paper suggest that frequent drinking may simply be a marker for drinking safer amounts when alcohol is consumed. However, according to the results of a meta-analysis of 4 cohort and 2 case–control studies it may not be that simple. A J-shaped curve, with a nadir around 28 grams of alcohol per week, and last protective dose of 131 grams per week, was obtained including drinkers who consumed alcohol for 2 days a week or less. Conversely, in people who consumed alcohol for more than 2 days a week a significant protective effect was seen even when drinking high amounts of alcohol up to 500 grams per week.”8 For moderate drinkers, there appears to be a very wide range of alcohol consumption for which the risk of mortality is lower than it is for non-drinkers. Based on health effects, data do not support the contention of the authors that “the supposed therapeutic index for alcohol is narrow.”
“(4) Clustering of lifestyle behaviors: Clustering of lifestyle behaviours is a well-known phenomenon; e.g., the results from the U.S. Department of Agriculture’s Continuing Survey of Food Intakes by Individuals showed poor food choices along with lower nutrient intakes to be associated with increases in smoking and drinking.9 The morbidity and mortality of excessive alcohol consumption is confounded by heavy smoking and a poor diet. The clustering of lifestyle behaviours is fertile ground for confounding. The problem is that no one has precise knowledge of the direction of the confounding; e.g., it might be that some of the benefits of regular exercise are due to the fact that regular exercise is associated with regular moderate alcohol intake.10
“(5) Alcohol consumption and CVD-related biomarkers: The discussion of the authors of this paper regarding frequent ethanol consumption and CVD-related biomarkers is superficial. Extensive reviews have found alcohol intake causally related to lower risk of coronary heart disease through many mechanisms, including changes in lipids and haemostatic factors.11
“(6) Alcohol intake, cancer and total mortality: Among the arguments of the authors of this paper is “ . . . the exclusively positive relationship between alcohol consumption and cancer.” The statement is false. A recent meta-analysis found a negative effect of moderate alcohol consumption on the risk of renal cell cancer.12 The key findings of recent analyses based on the Million Women’s Study in the UK are that alcohol consumption appears to lower the risk of several types of lymphoma and plasma cell neoplasms.13 However, the following statement by Naimi et at is entirely correct: “All-cause mortality should be the principal basis of guidelines used in a public health context.” That has been the opinion of our Forum all the way, but a call for all-cause mortality as a proper criterion is rarely heard when the discussion concerns certain outcomes: e.g., alcohol and breast cancer.”
Forum reviewer Stockley adds: “In addition to reducing the risk of cardiovascular disease and certain cancers, light-to-moderate alcohol consumption reduces the risk of dying from all or any causes (all-cause mortality). A recent study of 16,958 US individuals followed for 18 years by the US Centers for Disease Control and Prevention (CDC) examined the relationship between four low-risk behaviors and mortality. Moderate consumption of alcohol was considered as one of “four healthy lifestyle behaviors that exert a powerful and beneficial effect on mortality.”14 The other low-risk behaviors were non smoking, eating a healthy diet, and physical activity. These CDC study authors stated that: “The number of low-risk behaviors was inversely related to the risk for mortality. This analysis also showed that alcohol consumption at 2.5–14.9 g/day was consistently associated with a 14–25% reduction in the risk of all outcomes assessed compared with abstaining from alcohol.”
In her appraisal, Dr Stockley continues: “Lee et al15 showed that although light-to-moderate drinkers may have better risk factor profiles than non-drinkers, including higher socioeconomic status and fewer functional limitations (such as activities of daily living, instrumental activities of daily living and mobility), which explain some of the survival advantage associated with alcohol consumption, light-to-moderate drinkers still maintain their survival advantage even after adjustment for these factors. Further, Sun et al16 recently showed that in addition to lower mortality, women moderate alcohol consumers surviving to age 70 years and older generally had less disability and disease, and more signs of ‘successful ageing.’ For ‘regular’ light-to-moderate alcohol consumers (on 5-7 days/week), there was an approximately 50% greater chance of such successful ageing compared with non-drinkers.”
Other Forum comments on confounding: Forum reviewer De Gaetano notes a paper by Naimi and colleagues published in 200517 that was also based on data from the Behavioral Risk Factor Surveillance System. Mukamal, Ding and Djoussé re-analyzed the same data in 2006 and published a seminal (in our opinion) paper about the role of confounders in the studies of alcohol in moderation.18 These latter authors concluded that “Both moderate drinking and physical activity are associated with healthier lifestyle characteristics after adjustment for age and sex, although these associations are attenuated after modest multivariable adjustment for race and education alone and their absolute magnitude is modest. Neither moderate drinking nor physical activity have been proven to prevent cardiovascular disease in randomized trials, and hence observational studies of both factors may be susceptible to uncontrolled confounding. Nonetheless, our results provide no evidence that moderate drinking is unique in this regard, at least when compared to physical activity, which is widely assumed to prevent cardiovascular disease.”18
The authors of the present paper state that socioeconomic factors such as lower levels of education are determined early in life and “are not plausibly due to a lack of alcohol consumption in later life and yet are important predictors of morality.” It is clear from most studies that adults with less education tend to be more likely to be non-drinkers, or episodic or binge drinkers, than do their better educated peers.
Forum reviewer Zhang commented: “My major concern is the high non-response rate (46%). With such a high non-response rate, it is hard to comment on the validity of study findings. On the other hand, residual confounding is always an issue in observational studies. I do believe that moderate drinkers may have many ‘favorable risk profiles,’ as do wine drinkers in many studies. It is easy to criticize study findings from observational studies, stating that they are suffering from residual confounding, but that does not move the scientific research forward. The real issue is how big such residual confounding effect is when assessing moderate drinking in relation to CHD or other outcome variables; this is a challenging but more relevant question.”
Other concerns of Forum reviewers: There are numerous statements by the authors that require some unusual assumptions to accept. For example, the authors state that “. . . ethanol’s short half-life and the infrequency with which most people drink raises questions about whether protective associations of low-dose ethanol for CVD mortality might be attributable to CVD-related biomarkers.” They then point to a few intervention studies of the effects of alcohol on HDL cholesterol saying that all of these studies have been based on daily administration of alcohol, rather than in a pattern more consistent with the usual intake of the population. They also use data based on pharmacologic approaches for raising HDL, saying that such drugs do not decrease mortality; the relevance of this to alcohol and HDL are unclear. Using such reasoning, they conclude that “ . . . it seems unlikely that changes in ethanol-related biomarkers explain population-level protective associations for CVD outcomes.” This statement ignores not only thousands of animal experiments but the results of hundreds of well-done intervention studies in humans showing that moderate alcohol intake favorably modifies most risk factors.
References from Forum review
1. Harrington M, Gibson S, Cottrell RC. A review and meta-analysis of the effect of weight loss on all-cause mortality risk. Nutrition Research Reviews 2009;22:93-108.
2. Bouhard C, Blair SN, Church TS, Earnest CP, Hagberg JM, et al. Adverse metabolic response to regular exercise: Is it a rare or common occurrence? PLoS ONE 2012;7(5): e37887. doi:10.1371/journal.pone.0037887.
3. Waller K, Kujala UM, Rantanen T, Kauppinen M, Silventoinen K, Koskenvuo M, Kaprio J. Physical activity, morbidity and mortality in twins: a 24-year prospective follow-up. Eur J Epidemiol 2010;25:731–739.
4. Hooper L, Summerbell CD, Higgins JPT, Thompson RL, Capps NE, Smith GD, Riemersma RA, Ebrahim S. Dietary fat intake and prevention of cardiovascular disease: systematic review. BMJ 2001;322:757-763.
5. Michels KB. Nutritional epidemiology-past, present, future. Int J Epidemiol 2003;32:486–488.
7. Colditz G, et al. Alcohol intake in relation to diet and obesity in women and men. Am J Clin Nutr 1991;54:49-55.
8. Bagnardi V, Zatonski W, Scotti L, La Vecchia, C, Corrao G. Does drinking pattern modify the effect of alcohol on the risk of coronary heart disease? Evidence from a meta-analysis. Epidemiol Community Health 2008;62:615-619.
9. de Vries H, Kremers SP, Smeets T, Brug J, Eijmael K. The effectiveness of tailored feedback and action plans in an intervention addressing multiple health behaviours. American Journal of Health Promotion 2008; 22:417-425.
10. Pedersen JØ, Berit Lilienthal Heitmann BL, Schnohr P, Grønbæk M. The combined influence of leisure-time physical activity and weekly alcohol intake on fatal ischaemic heart disease and all-cause mortality. Eur Heart J 2008;29: 204-212.
11. Rimm EB, Fosher K, Criqui M, et al. Moderate alcohol intake and lower risk of coronary heart disease: meta-analysls of effects on lipids and haemostatic factors. BMJ 1999;319:1523-1528.
12. Bellocco R, Pasquali E, Rota M, et al. Alcohol drinking and risk of renal cell carcinoma: results of a meta-analysis. Ann Oncol 2012;23: 2235-2244.
13. Kroll ME, Murphy F, Pirie K, Reeves GK, Green J, Beral V, for the Million Women Study Collaborators. Alcohol drinking, tobacco smoking and subtypes of haematological malignancy in the UK Million Women Study. British Journal of Cancer 2012;107:879-887.
14. Earl S. Ford ES, Zhao G, Tsai J, Li C. Low-risk lifestyle behaviors and all-cause mortality: Findings from the National Health and Nutrition Examination Survey III Mortality Study. AJPH 2011; doi:10.2105/AJPH.2011.300167.
15. Lee SJ, Sudore RL, Willaims BA, et al. Functional limitations, socioeconomic status, and all-cause mortality in moderate alcohol drinkers. J Am Geriatr Soc 2009;57:955-962.
16. Sun Q, Townsend MK, Okerecke OI, et al. Alcohol consumption at midlife and successful ageing in women: A prospective cohort analysis in the Nurses’ Health Study. PLoS Med 2011;8(9):e1001090. doi:10.1371/journal.pmed.1001090.
17. Naimi TS, Brown DW, Brewer RD, Giles WH, Mensah G, Serdula MK, Mokdad AH, Hungerford DW, Lando J, Naimi S, Stroup DF: Cardiovascular risk factors and confounders among nondrinking and moderate-drinking U.S. adults. Am J Prev Med 2005;28:369-373.
18. Mukamal KJ, Ding EL, Djoussé L. Alcohol consumption, physical activity, and chronic disease risk factors: a population-based cross-sectional survey. BMC Public Health 2006, 6:118; doi:10.1186/1471-2458-6-118.
Comments on this paper have been provided by the following members of the International Scientific Forum on Alcohol Research:
Erik Skovenborg, MD, Scandinavian Medical Alcohol Board, Practitioner, Aarhus, Denmark
Creina Stockley, PhD, MBA, Clinical Pharmacology, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia
Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway
R. Curtis Ellison, MD, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA
Giovanni de Gaetano, MD, PhD, Research Laboratories, Catholic University, Campobasso, Italy
Andrew L. Waterhouse, PhD, Marvin Sands Professor, Department of Viticulture and Enology, University of California, Davis; Davis, CA, USA
Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA
David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa