Critique 077: Swedish study supports a “U-shaped” association of alcohol consumption with risk of pre-diabetes and diabetes mellitus — 29 March 2012

Cullmann M, Hilding A, Östenson CG.  Alcohol consumption and risk of pre-diabetes and type 2 diabetes development in a Swedish population.  Diabet Med 2012;29:441–452.  DOI: 10.1111/j.1464-5491.2011.03450.x

Authors’ Abstract

Aims  Alcohol is a potential risk factor of Type 2 diabetes.  However, more detailed information on effects of alcohol types and early phases of Type 2 diabetes development seems warranted.  The aim of this study was to investigate the influence of alcohol consumption and specific alcoholic beverages on the risk of developing pre-diabetes and Type 2 diabetes in middle-aged Swedish men and women.

Methods  Subjects, who at baseline had normal glucose tolerance (2,070men and 3,058 women) or pre-diabetes (70men and 41 women), aged 35–56 years, were evaluated in this cohort study.  Logistic regression was performed to estimate the risk [odds ratio (OR) and 95%confidence interval (CI)] to develop pre-diabetes and Type 2 diabetes at 8–10 years follow-up, in relation to self-reported alcohol intake at baseline.  Adjustment was performed for several risk factors.

Results  Total alcohol consumption and binge drinking increased the risk of pre-diabetes and Type 2 diabetes in men (OR 1.42, 95% CI 1.00–2.03 and OR 1.67, 95% CI 1.11–2.50, respectively), while low consumption decreased diabetes risk in women (OR 0.41, 95% CI 0.22–0.79).  Men showed higher risk of pre-diabetes with high beer consumption (OR 1.84, 95% CI 1.13–3.01) and of Type 2 diabetes with high consumption of spirits (OR 2.03, 95% CI 1.27–3.24).  Women showed a reduced risk of pre-diabetes with high wine intake (OR 0.66, 95% CI 0.43–0.99) and of Type 2 diabetes with medium intake of both wine and spirits (OR 0.46, 95% CI 0.24–0.88 and OR 0.55, 95% CI 0.31–0.97, respectively), whereas high consumption of spirits increased the pre-diabetes risk (OR 2.41, 95% CI 1.47–3.96).

Conclusion  High alcohol consumption increases the risk of abnormal glucose regulation in men.  In women, the associations are more complex: decreased risk with low or medium intake and increased risk with high alcohol intake.

Forum Comments

Stampfer and colleagues1 reported in 1988 that moderate drinkers in the Nurses’ Health Study had a much lower risk for development of diabetes than did abstainers, due partially to a lower risk of obesity among moderate drinkers.  Since then, prospective observational studies have been remarkably consistent in showing an inverse association of moderate alcohol intake with the risk of developing type 2 diabetes mellitus (DM).2-7  Some, but not all such studies, have suggested that higher intake of alcohol may increase such risk. 

The present analyses were based on a cohort study of men and women in Sweden whose alcohol intake was assessed only at baseline.  Men reporting a total alcohol intake ≥ 22.14 g/day (about two typical “drinks” by US standards) and women reporting ≥ 8.76 g/day (about 2/3 of a typical drink) were classified as having “high” alcohol consumption. 

Subjects were assessed 8-10 years after baseline for the presence of pre-diabetes (pre-DM, fasting glucose 6.1-6.9 or 2 hours glucose 7.8-11.0 mmol/l) or type 2 diabetes mellitus (DM, fasting glucose ≥7.0 and/or 2 hour glucose ≥ 11.1. mmol/l).  The authors report that < 20% of invited subjects did not participate in the follow-up examination, indicating very good compliance

The cohort was not population based, but had been enriched with persons who had a positive family history of diabetes.  The study reported the development of DM and of pre-DM among subjects with normal glucose values at baseline, and also reported on the progression from pre-DM to DM among subjects with impaired values at baseline.  Apparently, analyses were based on all subjects, including those within “intervention communities” (those receiving a program to prevent the development of diabetes) and those in “control communities” (that did not receive the community intervention).

Overall, 105 men and 57 women developed DM de novo, and 175 men and 98 women progressed to DM from pre-DM.  A total of 240 men and 98 women with previously normal values met diagnostic criteria for pre-DM at the end of the follow-up period. 

Specific comments on the present study:  The data collection for this study was apparently carried out between 1992 and 1998.  One reviewer suggested that a “shift in drinking pattern in this Swedish population over 8-10 years is possible (a change to more wine than spirits and perhaps beer).  It would have been helpful to have an estimate of alcohol intake at the end of the study period.”

While the total number of subjects in the study was reasonable, when evaluating subsets by category of alcohol intake and certain outcomes, the numbers in individual cells were often very small, leading to very wide confidence limits.  As stated by one Forum reviewer: “The numbers were indeed relatively small to allow the numerous sub-analyses performed in this study.  Possible differences between men and women are of interest, as well as among different alcoholic beverages, but should be based on much higher and consistent figures.”  Said another: “I am not comfortable with the large number of sub-group analyses with small numbers of participants and wide confidence intervals.  Further, it seems strange that women in the ‘high wine consumption’ quartile are consuming as low as 1.66 grams of alcohol per day, which amounts to just one glass of wine per week, hardly a dose of ‘alcohol medicine’ that can be expected to have any physiological impact on insulin sensitivity.”

The authors apparently looked only at p-values in interpreting their results.  They did not mention the very clear “U-shaped” curve seen in their study.  In almost every comparison, the highest risks of pre-DM and DM were seen in the “abstainer” category, with lower risks for most categories of alcohol consumers.  It should be pointed out, however, that the number of abstaining subjects was small, and apparently this group included life-time abstainers and ex-drinkers.  There was a suggestion in some comparisons that the risk was higher among male subjects (at least those consuming beer or spirits) in the highest drinking category.

One reviewer observed: “Men who had the highest alcohol consumption had the highest consumption of beer and spirits, were more likely to smoke, were more obese, and had a lower educational level.  These associations were less pronounced in women.  The authors write that they adjusted for these confounders, but I don’t think it is possible to make a complete adjustment.  Smoking is predisposing for diabetes, in addition to obesity and high carbohydrate intake.  There are no diet data, a weakness of the study.  I assume that those participants who consumed more beer and spirits also had the most unhealthy diet – also an important factor.”

Some cohort members exposed to a diabetes prevention program:  This cohort was apparently involved in a community-wide diabetes prevention intervention program, and it appears that these analyses include those persons who lived in the three (of five) communities that were involved in the intervention.  The authors state that this intervention program “was community-based and focused on the whole adult population in the three intervention municipalities (i.e. not implemented on an individual level).  Activities to prevent Type 2 diabetes, aiming at risk factors such as obesity, low physical activity, dietary habits and tobacco use, were initiated.”  It is unclear whether or not these community-based interventions had any effect on the subsequent development of pre-DM or DM in these communities.  If so, it would be interesting to know if the results of the associations with alcohol were the same in the control communities as in the communities receiving the program.  Added one Forum reviewer: “I prefer population-based cohort studies with a good description of the population and the sample taken.  The present study is very likely derived from an intervention program, but we are left in the dark how the cohort was created and what interventions were performed.”

Increasing the number of subjects with a positive family history of diabetes:  Forum member Tedd Goldfinger has pointed out a concern about interpreting the results from a sample of the population that had been “enriched” with subjects having a positive family history of diabetes.  When seeking causative factors for a disease, the relative importance of any single factor (e.g., alcohol) depends on other factors in the causative pathway that may be present in the population being studied.  We can assume that subjects with a positive family history have a number of factors, including genetic ones, that predispose them to developing diabetes. Among such subjects, any environmental factor could have less of an impact on the disease than would be present in the general population if one uses strength of relative risk as a measure of impact.  This phenomenon was described well by Ken Rothman in one of his text-books on epidemiologic methods. Rothman gave an example of this by commenting on environmental risk factors for lung cancer.  Smoking is by far the most “important” risk factor, followed perhaps by radon exposure.  However, in a non-smoking population, radon may then become the predominant causative factor for lung cancer.  The strength of a cause will vary from population to population depending on the presence or absence of other causative factors.8..This example illustrates that what we mean by “strength of effect” is not a biologically stable characteristic of a causative factor, and the increases and decreases in the relative risk of diabetes shown in this analysis are probably less than they would be in a population-based study. 

Overall, data from the present study support most previous research showing a reduction in the risk of DM to be associated with moderate drinking, with possibly some increase with greater alcohol intake.  Some observational studies suggest greater effects in women than men, while others suggest greater effects among men.  It is likely that both men and women show protection against developing DM from moderate drinking, although the levels of intake for specific positive or negative effects may differ by gender.  As in many other studies, the present study suggests larger beneficial effects of wine (and possibly of beer) than of spirits in terms of their effects on risk of diabetes.


1.  Stampfer MJ, Colditz GA, Willett WC, et al.  A prospective study of moderate alcohol drinking and risk of diabetes in women.  Am J Epidemiol 1988;128:549-558.

2.  Rimm EB, Chan J, Stampfer MJ, et al.  Prospective study of cigarette smoking, alcohol use, and the risk of diabetes in men.  BMJ 1995;310:555-559.

3.  Perry IJ, Wannamethee SG, Walker MK, et al.  Prospective study of risk factors for development of non–insulin dependent diabetes in middle aged British men.  BMJ 1995;310:560-564.

4.  Howard AA, Arnsten JH, Gourevitch MN.  Effect of alcohol consumption on diabetes mellitus: a systematic review. Ann Intern Med 2004; 140: 211–219.

5.   Koppes LL, Dekker JM, Hendriks HF, Bouter LM, Heine RJ.  Moderate alcohol consumption lowers the risk of type 2 diabetes: a meta-analysis of prospective observational studies.  Diabetes Care 2005;28:719–725.

6.  Carlsson S, Hammar N, Grill V.  Alcohol consumption and type 2 diabetes. Meta-analysis of epidemiological studies indicates a U-shaped relationship.  Diabetologia 2005;48:1051–1054.

7.  Baliunas DO, Taylor BJ, Irving H, Roerecke M, Patra J, Mohapatra S, et al.  Alcohol as a risk factor for type 2 diabetes: a systematic review and meta-analysis. Diabetes Care 2009;32:2123–2132.

8.  Rotham KJ.  Epidemiology. An Introduction.  Oxford University Press, New York, 2002, pages 11-12. 

Forum Summary

Subjects in a cohort in Sweden, some of whom had been exposed to a community intervention program to prevent diabetes, were evaluated 8-10 years after baseline for the presence of diabetes mellitus or impaired glucose metabolism (“pre-diabetes”) in relation to a baseline report of alcohol consumption.  Approximately 2,000 men and 3,000 women had a normal glucose tolerance test at baseline; of these 105 men and 57 women developed type II diabetes.  Of subjects with pre-diabetes at baseline, 175 men and 98 women progressed to diabetes.  The authors report that total alcohol consumption and binge drinking increased the risk of pre-diabetes and diabetes in men, while low consumption decreased diabetes risk in women.  However, the authors did not discuss the fact that in essentially all comparisons, the highest risk of diabetes or pre-diabetes was among abstainers.

Forum reviewers had some concerns about the study.  For example, the study included some subjects who had been exposed to an intervention trial to prevent diabetes, yet no information is given on potential effects of the intervention.  It was not a population-based group.  Also, the sample was “enriched” with subjects who had a positive family history of diabetes, which may make it more difficult to judge the effects of environmental factors.  Ex-drinkers and never drinkers were included in the abstainer group. 

It appears that the authors focused only on the “statistically significant” results rather than commenting on the overall pattern of association (lower risk of developing diabetes for moderate drinkers than for abstainers and heavier drinkers).  Further, the number of subjects in many of the sub-groups was very small, making it difficult to define specific cut-points for effects of alcohol on risk. 

Nevertheless, reviewers considered that, overall, these analyses support the usual findings from previous research of a “U-shaped curve” for alcohol and diabetes for both men and women.  There appears to be a reduction in risk with moderate alcohol intake and possibly an increased risk for heavier drinking. 

                                                                                   *                      *                      *

Comments on this paper were provided by the following members of the International Scientific Forum on Alcohol Research:

Luc Djoussé, MD, DSc, Dept. of Medicine, Division of Aging, Brigham & Women’s Hospital and Harvard Medical School, Boston, MA, USA

David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa

Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA

Tedd Goldfinger, DO, FACC, Desert Cardiology of Tucson Heart Center, Dept. of Cardiology, University of Arizona School of Medicine, Tucson, Arizona, USA

Giovanni de Gaetano, MD, PhD, Research Laboratories, Catholic University, Campobasso, Italy

Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway

Andrew L. Waterhouse, PhD, Marvin Sands Professor, Department of Viticulture and Enology, University of California, Davis; Davis, CA, USA

David Vauzour, PhD, Senior Research Associate, Department of Nutrition, Norwich Medical School, University of East Anglia, Norwich, UK

Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA

Erik Skovenborg, MD, Scandinavian Medical Alcohol Board, Practitioner, Aarhus, Denmark

Fulvio Ursini, MD, Dept. of Biological Chemistry, University of Padova, Padova, Italy

R. Curtis Ellison, MD, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA

Ulrich Keil, MD, PhD, Institute of Epidemiology and Social Medicine,  University of Münster, Münster, Germany

Creina Stockley, MSc, MBA, Clinical Pharmacology, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia