Changes in Alcohol Intake and Mortality: A Longitudinal Population-based Study : Epidemiology

Secondary Logo

Journal Logo

Original Article

Changes in Alcohol Intake and Mortality

A Longitudinal Population-based Study

Grønbæk, Morten*†; Johansen, Ditte*; Becker, Ulrik; Hein, Hans Ole§; Schnohr, Peter; Jensen, Gorm; Vestbo, Jørgen; Sørensen, Thorkild I. A.*

Author Information
Epidemiology 15(2):p 222-228, March 2004. | DOI: 10.1097/01.ede.0000112219.01955.56
  • Free


Prospective studies from many countries have suggested that light to moderate alcohol drinkers have lower risk of death than nondrinkers and heavy drinkers.1–9 Within this nonlinear pattern, rates of coronary heart disease tend to be highest among nondrinkers, whereas rates of cancer and liver cirrhosis are highest among heavy drinkers.2,10–13 These findings suggest that nondrinkers may benefit from changing to light or moderate drinking, whereas heavy drinkers may benefit from reducing their intake. However, these studies were all based on a single assessment of alcohol intake at a given point in time. Whether a change in drinking habits actually results in the expected change in risk remains uncertain. It is possible that the alcohol–mortality relation—particularly in the nondrinking group—is confounded by concurrent diseases, traits, or lifestyle factors associated with both level of alcohol intake and with mortality. Moreover, cumulative or irreversible effects of previous alcohol intake may influence mortality after a change in intake. Large-scale intervention studies may offer an answer, but such studies are not feasible for a variety of logistic and ethical reasons.

One step toward establishment of a causal association, as well as an assessment of its public health implications, may be to compare the health consequences of stable and changing patterns of alcohol intake in prospectively followed populations while taking possible confounders into account. Using this approach, the present study was carried out to assess the influence of a stable alcohol intake on mortality from all causes, from coronary heart disease, and from cancer, and to examine whether individual change in alcohol intake over time affected mortality from these causes.


The Copenhagen Centre for Prospective Population Studies

The study population was from the Copenhagen Centre for Prospective Population Studies. The study included 6148 women and 4805 men from the Copenhagen City Heart Study and 1862 women and 1839 men from the Glostrup Population Studies, all of whom had attended 2 health examinations at approximately 5-year intervals. The cohorts have been described in detail previously.14–17 Of the 19,864 who participated in the first examination, 1303 (7%) died between the first and the second examination, and 2839 (14%) did not attend the second examination. A total of 14,654 participants who contributed with complete information on determinants and covariates of interest were included (93% of those examined twice). Ages ranged from 25 to 98 years.

The participants were asked about their average weekly number of drinks of beer, wine, and spirits. In Denmark, one bottle of beer contains 12 g of alcohol, approximately equivalent to the other types of drinks (one glass of wine [1/6 of a bottle] and 1 glass [4 cL] of spirits).18 Subjects were categorized into 4 groups: less than 1 drink per week (nondrinkers), from 1 to 6 drinks per week (light drinkers), from 7 to 13 drinks per week (moderate drinkers), and more than 13 drinks per week (heavy drinkers). With 4 possible categories at each of the 2 examinations, this provided 16 possible combinations of alcohol intake.

Smoking was categorized as never-smoker, ex-smoker, non-inhaling current smoker, and inhaling current smoker of 1 to 14 g, 15 to 24 g, and more than 24 g of tobacco per day. Educational levels were less than 8 years, 8 to 11 years, and 12 or more years. Categories of physical activity were sedentary (less than 2 hours per week), light (2–4 hours per week), moderate (more than 4 hours per week, noncompetitive), and heavy (more than 4 hours per week, competitive). Body mass index (BMI) was calculated as weight (kg) divided by squared height (m2), and categorized as less than 20.0, 20.0 to 24.9, 25.0 to 29.9, 30.0 to 34.9, and more than 35.0 kg/m2. A variable was created to indicate the cohort from which the subject originated. All categorizations were made for both the first and the second examination.

Subjects who developed diseases before the second examination were identified in the National Hospital Patient Register, which includes all hospitalizations in Denmark since 1977, and in the Cancer Register. On this basis, subjects were classified as having had “any disease” (defined as any record apart from pregnancy according to the International Classifications of Diseases, Eighth Revision [ICD-8]), “coronary heart disease” (ICD-8 codes: 410.0–414.9), or “cancer” (ICD-8 codes: 140.0–209.9).

Follow Up

Follow up started at the second examination (1981–1983 for the Copenhagen City Heart Study and 1982 for the Glostrup Population Studies) defined here. The subjects were followed to date of death, disappearance, emigration, or end of follow up, whichever came first. Vital status of the populations was followed until January 9, 1995, using the person identification number in the Civil Registration System. Only 0.7% of the subjects were lost in the follow-up period. The average follow-up period was 10.5 years. Causes of death were obtained from the National Board of Health. The accuracy of the National Hospital Patient Register, the Causes of Death Registry, and the Cancer Register is described elsewhere.19–21

Data Analysis

The purpose of the analysis was to estimate the rate ratios of dying from all causes, and from coronary heart disease and cancer specifically within the 16 categories of alcohol intake at the first and the second examination. Adjustments were made for potentially confounding variables, including cohort of origin. Assuming constant mortality rates within each 10-year age interval, log-linear Poisson regression models were estimated and 95% confidence intervals (CI) were determined as previously described.3 A person observed in more than one age group contributed with corresponding observation time in each group.

Two types of models were constructed. In one model, all 16 possible categories of alcohol intake were included, using stable light alcohol intake at both examinations as the referent. This model describes the effects of changing one's intake between the 2 examinations. The second model includes only people who had a stable intake between the 2 examinations. A comparison of the 2 models enables an assessment of whether changes in mortality after changes in alcohol intake correspond to those expected from the effects of stable intake at the level achieved after the change in intake.

The models included the following variables as measured at the second examination: subjects’ age, sex, smoking habits, educational level, physical activity, and body mass index. To examine whether the effect of changes in alcohol intake could be the result of parallel changes in other lifestyle factors, additional analyses included variables on smoking habits, physical activity, and body mass index at both the first and the second examination. Furthermore, in the analysis of all-cause mortality, the analysis was adjusted for disease states. The analyses of coronary heart disease and cancer mortality were performed with and without exclusion of diseased persons. In the analysis of coronary heart disease mortality, persons with a history of this disease were defined as subjects who, according to the Hospital Discharge Register, had had coronary heart disease diagnosed before the second examination, whereas in the analyses of cancer mortality, persons with a history of this disease were defined as subjects who had had cancer diagnosed before the second examination. The analyses of changes in alcohol intake and all-cause mortality were repeated with the data stratified on age and sex.


Of the 14,654 subjects included in the study, 6459 had changed their alcohol intake between the first and second examination by the categories defined (Table 1). Eighteen percent increased their intake at least one category, 26% reduced their intake, and the remaining were unchanged (of whom one fourth were nondrinkers). Twenty-seven percent of nondrinkers at the first examination increased their intake, whereas 45% of drinkers reduced their intake to the level of nondrinkers. Fifteen percent of participants remained nondrinkers, whereas 14% remained heavy drinkers.

Number of Subjects and Absolute Risks of Death From All Causes, Coronary Heart Disease, and Cancer During Follow-Up After the Second Examination, by Alcohol Intake at the First and Second Examinations

The number of deaths during follow up was 3187 (Table 1). In total, 729 people died of coronary heart disease of whom 606 had no history of this disease at the start of follow up. Death from cancer occurred in 998; of them, 903 had no cancer diagnosis at the start of follow up (Table 1).

Possible confounding variables (sex, age, educational level, smoking habits, physical activity, and body mass index) were unequally distributed across alcohol intake at the first and second examination (Table 2). Most striking was the preponderance of males and current smokers at high and increasing alcohol intake. Participants who increased their intake were more usually young and well educated. At both examinations, abstainers were the lowest educated and least physically active. Median alcohol intake was equally distributed according to intake at second examination; only the stable high-intake group had a higher intake than those who changed from lighter to higher intake.

Characteristics of All Subjects at the Second Examination (baseline of follow up) by Alcohol Intake (drinks per week) at the First and the Second Examination

Mortality by Stable Alcohol Intake

The relation between stable alcohol intake and all-cause mortality was U-shaped (Fig. 1A). Compared with stable light drinkers, stable nondrinkers and heavy drinkers had similarly increased risks of dying (1.29 [CI = 1.13–1.48] and 1.32 [1.15–1.53], respectively). There was little difference in mortality between light and moderate drinkers (1.05; 0.89–1.25).

Relative risk for all-cause mortality (A), coronary heart disease mortality (B), and cancer mortality (C) for individuals with the same alcohol intake at the first and second examinations, by alcohol intake. Referent group is 1–6 drinks per week. Vertical lines are 95% confidence intervals.

The risk of dying from coronary heart disease declined across levels of stable drinking (Fig. 1B). Relative to stable light drinking, the relative risk was 1.32 (CI = 0.97–1.79) with nondrinking, 0.95 (0.66–1.38) with moderate drinking, and 0.86 (0.62–1.20) with heavy drinking.

Risk shifted in the other direction for cancer mortality (Fig. 1C). Risk was highest among those with the highest levels of stable drinking (1.48; 1.15–1.92), and less for nondrinkers (1.10; 0.86–1.41) and moderate drinkers (1.15; 0.84–1.56).

Mortality by Changes in Alcohol Intake

Table 3 shows the mortality associated with changing alcohol intake relative to a stable alcohol intake. All-cause mortality increased for those who stopped drinking, regardless of their previous level of drinking, and also increased for those who became heavy drinkers. For coronary heart disease, an increase in alcohol consumption was associated with lower mortality, and a reduction in consumption was associated with increased mortality. For cancer, there was an increasing trend in mortality after increased intake and no clear pattern for reduced intake. The estimate in the second row in each cell represents the risk associated with stable intake at the given level; thus, all second rows within a disease category are basically the same 4 relative risks, varying only by which one of the 4 is the referent. A comparison of each of these estimates with those expected from analyses of the effect of a stable intake on the same outcomes (Table 3, second row) shows little or moderate difference between the 2 estimates without any consistent pattern.

Relative Risk of Dying From All Causes, Coronary Heart Disease, and Cancer During Follow Up After the Second Examination, by Alcohol Intake (drinks per week) at Second Examination for Each Level of Alcohol Intake at First Examination With Stable Alcohol Intake as Reference. Estimates in the second rows are the relative risk ratios for those with a stable intake at the level seen at the second examination (also shown in Fig. 1)

The findings regarding changes in alcohol intake were similar in subsets of men and women, and young and old.


In this large, longitudinal population-based study, we found that a stable alcohol intake over an interval of approximately 5 years was associated with a U-shaped risk of all-cause mortality. The risk of dying from coronary heart disease decreased with higher stable consumption, whereas the risk of dying from cancer increased with higher consumption. Subjects who changed their alcohol intake between the 2 examinations acquired the mortality risk associated with their new level of consumption (as estimated among those with stable patterns of alcohol consumption). Thus, the study supports the suggestion that light or moderate alcohol intake compared with nondrinking may have a beneficial effect on health that becomes manifest in reduced mortality from all causes, and from coronary heart disease in particular. The study further suggests that compared with light or moderate drinking, a change to heavy drinking is followed by increased mortality, particularly cancer mortality.

A few small studies have addressed the question of whether a change in alcohol intake influences the risk of coronary heart disease. Most of these studies supported our findings, showing higher risk among non-drinkers compared to ex-drinkers.22–26 The study by Wannamethee and Shaper24 found that middle-aged men who became regular drinkers had a lower risk of major coronary heart disease than occasional drinkers, but an increased risk of noncardiovascular mortality and total mortality. The Honolulu Heart Program used cruder categories of change, but showed a decreased risk of coronary heart disease with an increased intake among younger men.23 The findings of a potentially beneficial effect of alcohol on coronary heart disease risk are supported by findings of a higher level of high-density lipoprotein among alcohol drinkers as compared with nondrinkers, as well as a beneficial effect of alcohol on platelet aggregation.27–29 One study has suggested that alcohol also may play a beneficial role in the secondary prevention of coronary heart disease.30

Shaper and Wannamethee31 have suggested that the apparent increased risk of coronary heart disease among nondrinkers is mainly the result of confounding by associated conditions and lifestyle habits. Aging and its accompanying accumulation of poor health may cause a gradual decrease in consumption. However, the authors did not assess the risk of dying as a function of these changes. In the present study, we have taken into account both previous and prevailing diseases, as well as important lifestyle factors and related conditions, that may confound the analysis. The socialized structure of the Danish healthcare services, with free access to hospital treatment (if judged relevant by general practitioners), makes it unlikely that serious comorbidity associated with excess mortality would not be detected by using hospitalization as a marker of such diseases. Obviously, confounding factors that we have not accounted for in this study such as diet may still be important; however, to explain our findings, such confounding factors must have both a close relation to the changes in alcohol intake and a strong effect on mortality.32

In this prospective study, participants were asked twice about their average intake of alcohol before the follow-up period. Self-reported alcohol consumption can be subject to misclassification, which also may apply to changes in alcohol intake. However, intake as reported in questionnaires appears valid compared with a thorough dietary interview.33 The change in alcohol intake may have occurred at any time between the first and second examination, and those reporting a stable intake may have experienced changes as well. However, these potential differences in exposure among members of the same category would tend to diminish the observed differences in mortality compared with the true ones.

This study included only subjects who attended both the first and the second examination. Subjects who were alive but did not attend the second examination had a greater mortality than responders to both examinations. If those who increase or reduce their alcohol intake from a light drinking level (associated with the nadir of the risk function) die before the second examination, then this may have weakened the observed associations. If those who continue at either a very low or a high level of intake after the first examination die before the second examination, that would also weaken the observed effect.

In conclusion, the present study supports a U-shaped relation between alcohol intake and all-cause mortality with particular benefits regarding coronary heart disease. Given the difficulty of clinical trials on this topic, more careful observational studies are needed. These should include detailed information on changes in drinking patterns over time, the reasons for those changes, and more extensive capture of morbidity and mortality end points.


1. Fuchs CS, Stampfer MJ, Colditz GA, et al. Alcohol consumption and mortality among women. N Engl J Med. 1995;332:1245–1250.
2. Doll R, Peto R, Hall E, et al. Mortality in relation to consumption of alcohol: 13 years’ observations on male British doctors. BMJ. 1994;309:911–918.
3. Grønbæk M, Deis A, Sørensen TIA, et al. Influence of sex, age, body mass index, and smoking on alcohol and mortality. BMJ. 1994;308:302–306.
4. Cullen KJ, Knuiman MW, Ward NJ. Alcohol and mortality in Busselton, Western Australia. Am J Epidemiol. 1993;137:242–248.
5. Klatsky AL, Armstrong MA, Friedman GD. Alcohol and mortality. Ann Intern Med. 1992;117:646–654.
6. Farchi G, Fidanza F, Mariotti S, et al. Alcohol and mortality in the Italian rural cohorts of the seven countries study. Int J Epidemiol. 1992;21:74–81.
7. Boffetta P, Garfinkel L. Alcohol drinking and mortality among men enrolled in an American cancer society prospective study. Epidemiology. 1990;1:342–348.
8. Thun MJ, Peto R, Lopez AD, et al. Alcohol consumption and mortality among middle-aged and elderly US adults. N Engl J Med. 1997;337:1705–1714.
9. Mukamal KJ, Conigrave KM, Mittleman MA, et al. Roles of drinking pattern and type of alcohol consumed in coronary heart disease in men. N Engl J Med. 2003;348:109–118.
10. Maclure M. Demonstration of deductive meta-analysis: ethanol intake and risk of myocardial infarction. Epidemiol Rev. 1993;15:328–351.
11. Hein HO, Sørensen H, Suadicani P, et al. Alcohol consumption, Lewis phenotypes, and risk of ischaemic heart disease. Lancet. 1993;341:392–396.
12. Garg R, Wagener DK, Madans JH. Alcohol consumption and risk of ischemic heart disease in women. Arch Intern Med. 1993;153:1211–1216.
13. Criqui MH. Alcohol consumption, blood pressure, lipids, and cardiovascular mortality. Alcohol Clin Exp Res. 1986;10:564–569.
14. Appleyard M, Hansen AT, Schnohr P, et al. The Copenhagen City Heart Study. A book of tables with data from the first examination (1976–78) and a five year follow-up (1981–83). Scand J Soc Med. 1989;170(suppl 41):1–160.
15. Hagerup L, Eriksen M, Schroll M, et al. The Glostrup Population Studies. Collection of epidemiologic tables. Reference values for use in cardiovascular population studies. Scand J Soc Med. 1981;Suppl 20:1–112.
16. Schroll M. The World Health Organization MONICA project (monitoring trends and determinants in cardiovascular disease): a major international collaboration. J Clin Epidemiol (Oxford). 1988;41:105–114.
17. Hein HO, Suadicani P, Gyntelberg F. Alcohol consumption, serum low density lipoprotein cholesterol concentration, and risk of ischaemic heart disease: six year follow up in the Copenhagen male study. BMJ. 1996;312:736–741.
18. Grønbæk M, Iversen L, Olsen J, et al. Genstandsgrænser. Ugeskr Laeger. 1997;159:5939–5945.
19. Andersen TF, Madsen M, Jorgensen J, et al. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull. 1999;46:263–268.
20. Juel K, Helweg-Larsen K. The Danish registers of causes of death. Dan Med Bull. 1999;46:354–357.
21. Storm HH, Michelsen EV, Clemmensen IH, et al. The Danish Cancer Registry—history, content, quality and use. Dan Med Bull. 1997;44:535–539.
22. Sesso HD, Stampfer MJ, Rosner B, et al. Seven-year changes in alcohol consumption and subsequent risk of cardiovascular disease in men. Arch Intern Med. 2000;160:2305–2312.
23. Goldberg RJ, Burchfiel CM, Reed DM, et al. A prospective study of the health effects of alcohol consumption in middle-aged and elderly men. The Honolulu Heart Program. Circulation. 1994;89:651–659.
24. Wannamethee SG, Shaper AG. Taking up regular drinking in middle age: effect on major coronary heart disease events and mortality. Heart. 2002;87:32–36.
25. Jackson R, Scragg R, Beaglehole R. Alcohol consumption and risk of coronary heart disease. BMJ. 1991;303:211–216.
26. Lazarus NB, Kaplan GA, Cohen RD, et al. Change in alcohol consumption and risk of death from all causes and from ischaemic heart disease. BMJ. 1991;303:553–556.
27. Suh I, Shaten J, Cutler JA, et al. Alcohol use and mortality from coronary heart disease: the role of high-density lipoprotein cholesterol. Ann Intern Med. 1992;116:881–887.
28. Diehl AK, Fuller JH, Mattock MB, et al. The relationship of high density lipoprotein subfractions to alcohol consumption, other lifestyle factors, and coronary heart disease. Atherosclerosis. 1988;69:145–153.
29. Renaud SC, Beswick AD, Fehily AM, et al. Alcohol and platelet aggregation: the Caerphilly Prospective Heart Disease Study. Am J Clin Nutr. 1992;55:1012–1017.
30. Muntwyler J, Hennekens CH, Buring JE, et al. Mortality and light to moderate alcohol consumption after myocardial infarction. Lancet. 1998;352:1882–1885.
31. Shaper AG, Wannamethee SG. The J-shaped curve and changes in drinking habit. Novartis Found Symp. 1998;216:173–188.
32. Grønbæk M, Sørensen TIA. Is the effect of wine on health confounded by diet? [Letter]. Epidemiology. 2002;13:236.
33. Grønbæk M, Heitmann BL. Validity of self-reported intakes of wine, beer and spirits in population studies. Eur J Clin Nutr. 1996;50:487–490.
© 2004 Lippincott Williams & Wilkins, Inc.