Notes Washed Up in a Bottle

Don. W. MacCorquodale M.D. M.S.P.H.

Periodic Notes From the Field on Philosophy and Science.

ON THE HEALTH EFFECTS OF ALCOHOL

A woman drove me to drink and I never had the courtesy to thank her. -W. C. Fields

ALCOHOL CONSUMPTION AND INCIDENT CORONARY HEART DISEASE

            A cohort study of about 14,500 persons selected from four communities in the United States was conducted to study the association between alcohol consumption and different types of alcoholic beverages and the incidence of coronary heart disease in different race-sex groups (Fuchs F D et al.   Association between alcoholic beverage consumption and incidence of coronary heart disease in White and Blacks.   Am J Epidemiol 2004:   160:   466-474).   About 10,500 participants were white, and almost 4,000 were Black.   The mean period of follow-up was almost 10 years.  

            Alcohol consumption at baseline was determined from an interviewer-administered dietary questionnaire.   Incident CHD cases were ascertained by annual follow-up phone calls, and surveillance of vital records and community hospitals.  

            Proportional hazards regression was used to analyze the association between the incidence of CHD and alcohol consumption while controlling for various risk factors for CHD including age, cigarette-smoking, BMI, LDL cholesterol, educational level, income level, diabetes, etc.

            Black participants had fewer years of formal education, lower incomes, and higher prevalence of obesity, high blood pressure, and diabetes.  

            The hazards ratios incident CHD according to alcohol use for White men and White women are shown in the table below.

            Note that none of the hazards ratios given above are statistically significant.   Note as well the pronounced dose-response effect for increasing alcohol consumption with decreasing hazard ratios for incident CHD in White males.   (If a p value for trend had been performed, it clearly would have been statistically significant.   DWMac)

            Note that statistical significance was only achieved for the hazards ratio for men consuming 140 - < 240 g/week of alcohol, and the confidence intervals are very wide.  

            The hazards ratio of 0.51 for Black women consuming > 1 g/week of alcohol was based on 372 women in the category and 10 incident cases of CHD.

            The authors concluded that there was a positive association between ethanol consumption in Black men and a negative association in White men.   They added that there was also an inverse association for ethanol consumption in White women.   They state that the contrasting findings in Black and White men may be real or may be confounded by “lifestyle characteristics of drinkers.”     

COMMENT:   In 1923 Dr. Raymond Pearl, professor of biology at Johns Hopkins University , reported the findings of a study he conducted of a study of the association of alcohol consumption and mortality.   He found that the age-specific death rate was lower for light drinkers than it was for abstainers, and he also found that the age-specific death rate was highest for those who rank moderately or heavily.  

Since Dr. Pearl reported his findings, there have been slightly more than 50 articles published on the subject of alcohol and mortality.   Almost all of these articles dealt with studies of White men.   With one or two exceptions, those studies showed that all-cause mortality was lowest for light or moderate drinkers, somewhat higher for abstainers, and highest for heavy drinkers.   The lower level of mortality among light and moderate drinkers was overwhelmingly due to reduced numbers of deaths from coronary heart disease.   All-cause mortality almost always was reflected in a J-shaped or U-shaped curve.  

The study described above differs from most similar studies in that the outcome of interest is CHD incidence, not all-cause mortality or CHD mortality.   Since most previous studies have shown a striking effect of light to moderate alcohol intake on CHD mortality, one might reasonably expect a J-shaped or U-shaped association with alcohol use and CHD incidence.   That however is not the case.

The study described above differs strikingly from almost all other studies of this nature in that there is a straightforward negative linear association between alcohol consumption and mortality in White men.   The largest reduction in risk is in men with the highest level of alcohol consumption.  

One of the primary aims of the study was to compare the effect of alcohol use on incident CHD in Blacks and Whites.   I suspect that the Black sample was too small to permit precise estimates of the association.

ALCOHOL CONSUMPTION, COGNITIVE IMPAIRMENT, AND DEMENTIA

            A group of Finnish investigators conducted a cohort study to evaluate the association between alcohol consumption during midlife and mild

cognitive impairment and dementia in old age (Anttila T et al.   Alcohol drinking in middle age and subsequent risk of mild cognitive impairment and dementia in old age:   a prospective population study.   BMJ August 10, 2004; 329:   539-).   A random sample of 2,000 men and women, aged 65-79, in eastern Finland were randomly selected, and of these slightly more than 1,400 individuals who were first enrolled in 1971 and 1979 were asked to participate.   The response rate was 70%, 632 women and 386 men.  

            A self-administered questionnaire was administered, and it included questions on health behavior, medical history, and socioeconomic factors.   A venous blood specimen was taken to determine serum cholesterol and apolipoprotein E genotypes.  

            Participants were classified as those who had never consumed alcohol, those who drank “infrequently” (less than once a month), and those who drank “frequently” (several times a month).  

            Cognitive function was assessed in 1998.   The association between midlife alcohol consumption and subsequent mild cognitive impairment and dementia was investigated by means of multiple logistic regression.   Infrequent drinkers were used as the reference group.   Sixty-one participants, 5.8%, met the criteria for mild cognitive assessment, and 48, 4.6%, met the requirements for dementia.     

             The results, odds ratios, are shown in the table below.

Model 1:   Adjusted for age, sex, and education

Model 2:   Adjusted for age, sex, education, BMI, serum cholesterol, blood pressure, history of myocardial infarction, and history of stroke

            Among carriers of the apolipoprotein e4 allele, the risk of dementia increased with increasing alcohol use after adjustment.   Among non-carriers of the apolipoprotein e4 allele, the risk of dementia did not change significantly with increasing alcohol consumption.  

ALCOHOL CONSUMPTION AND COGNITIVE FUNCTION

            The longitudinal Whitehall Study was established in 1985 to examine the socioeconomic gradient in health and disease among slightly more than 10,000 British civil servants, and 73 percent of those eligible agreed to participate.   The median length of follow-up from phase I to phase V was 11 years, slightly more than 6,500 participants underwent a clinical examination, which included cognitive study (Britton A et al.   Alcohol consumption and cognitive function in the Whitehall Study.   Am J Epidemiology 2004:   160:   240-247).

            Alcohol consumption was assessed at baseline and subsequent study phases by asking participants the number of alcoholic drinks they had consumed during the last 7 days.   A standard measure of spirits and a glass of wine are considered in the UK to contain 8 g of alcohol, and a pint of beer to contain 16 g of alcohol.  

            Participants were also asked about whether or not they currently smoked cigarettes, cigars, etc., and the number of items they smoked per day or when they stopped smoking.   Salary and work role were used to define socioeconomic position.   Binary logistic regression was used to assess the association between alcohol consumption and cognitive status.  

            When compared with   those drinkers who consumed an average of less than 1 g of alcohol per week, those who drank 1 g or more of alcohol per week were significantly less likely to be in the lowest quintile of each of the cognitive tests at phase 5.   Those participants drinking at the highest levels reported in this study had the lowest odds ratios.   Never drinkers had a twofold risk of being in the lowest quintile.   Results were similar for men and women.

                        The odds ratios for men of being in the lowest cognitive function quintile by baseline alcohol consumption, 1985-1988, are show in the table below.   Since the findings are so very similar for each test of cognitive function, only the findings for three of five such tests reported are shown.

Alcohol consumption              Cognitive test

            The findings shown above are much the same after adjustment for age, smoking, physical and mental health, and employment grade.   The findings for women – with and without adjustment as shown previously – are again much the same.  

            The authors concluded that “for middle-aged subjects, increasing levels of alcohol consumption were associated with better functioning regarding some aspects of cognition.”  

            COMMENT:   The Finnish study suggests you're damned if you do and you're damned if you don't – as far as drinking and mild cognitive impairment is concerned.   The risk for abstainers and the risk for “frequent” drinking were almost the same.   The British study suggests that the more you drink, the more likely you are to escape cognitive impairment – even if you drink more than 241 g/week.   The latter is heavy drinking by almost any standard.

            Small wonder that “black box epidemiology” or risk factor epidemiology has been under attack for some years now.   Of course, it also has its defenders including the notables, Sander Greenland and Kenneth Rothman.   Kenneth Rothman speaking at a meeting of epidemiologists once observed that if risk factor epidemiology had done nothing more in the last decade than demonstrate the role of folic acid in preventing neural tube defects, that alone would justify it.   Sander Greenland is the principle author of a paper in a recent edition of the American Journal of Epidemiology in which he makes a powerful case for the usefulness of risk factor epidemiology.

            I would like to add that we are not likely to see a randomized clinical trial of alcohol as a preventive against cognitive impairment in our lifetimes.   That leaves us with risk factor epidemiology.   The latter will be even more useful when combined with considerations of the role of genetics in disease causation and prevention.