- Email: [email protected]
Does diet or alcohol explain the French paradox?
Does diet
or
alcohol
explain
the French
paradox?
Introduction
Summary The low rate of coronary heart disease (CHD) in France compared with other developed countries with comparable dietary intake has been called the French paradox. We explored this paradox by looking at alcohol, diet, and mortality data from 21 developed, relatively affluent countries in the years 1965, 1970, 1980, and 1988. We assessed wine, beer, and spirits intake separately. France had the highest wine intake and the highest total alcohol intake, and the second lowest CHD mortality rate. In univariate analyses, ethanol in wine was slightly more inversely correlated with CHD than total wine volume. In multivariate analyses, animal fat tended to be positively correlated, and fruit consumption inversely correlated, with CHD. Beer and spirits were only weakly inversely correlated with CHD. The strongest and most consistent correlation was the inverse association of wine ethanol with CHD. However, wine ethanol was unrelated to total mortality. We conclude that ethanol, particularly wine ethanol, is inversely related to CHD but not to longevity in populations. Although light to moderate alcohol consumption may improve longevity, alcohol abuse—which sharply reduces longevity—is correlated with average alcohol consumption in populations. Thus, while the risk/benefit ratio varies for individuals, the use of alcohol for cardioprotective purposes should not be encouraged as a public health measure.
investigators have noted the strikingly low rate of mortality from coronary heart disease (CHD) in France compared with other developed countries.1,2 St Leger et all assessed 18 developed countries but excluded Japan because of possible genetic factors. They found a strong inverse association between CHD mortality in 1970 and wine consumption. They also found positive associations for saturated and monounsaturated fat intake, and a negative association for polyunsaturated fat. However, in multiple regression analyses wine remained a significant (inverse) predictor of CHD mortality, whilst the Keys equation, accounting for dietary polyunsaturated, monounsaturated and saturated fat, and cholesterol, was not significantly related to CHD mortality. Renaud et aP Several
showed
that
in
17
countries
with
available
wine
consumption data, CHD mortality at all ages in 1987 was positively related to dairy fat with a correlation r=0-73. France, and to lesser degree, Switzerland showed the greatest deviation from the regression line, with low CHD rates despite above-average dairy fat intake. Adjustment for the inverse effect of wine consumption moved France and Switzerland quite close to the regression line, and improved the correlation to 0-87. Artaud-Wild et aP looked at the association between a cholesterol-saturated fat index and CHD mortality in men in 40 countries at varying levels of economic development and found the cholesterol-saturated fat index positively related to CHD mortality in 1977 with a correlation of 078.3 Finland had a higher-than-expected and France a lower-than-expected CHD mortality rate. However, lower milk and higher vegetable consumption in France partly explained this discrepancy. Alcohol was unrelated to CHD after adjustment for CSI. However, wine was inversely (r=-0-67) related to CHD mortality in the 18 countries with a cholesterol-saturated fat index above 20 per 1000 kcal. Thus, these three studies are only partly concordant. Discordant findings appear to reflect different selection of countries, different definitions of dietary and alcohol intake, different gender and age groups, and different study years. To clarify these issues, we analysed data from 21 developed countries, evaluating various dietary components and alcoholic beverages as predictive of CHD and total mortality in 1965, 1970, 1980, and 1988.
Methods
Country selection
Departments of Family and Preventive Medicine, and Medicine, University of California, San Diego School of Medicine, La Jolla, CA 92093-0607 (M H Criqui MD, MPH, B L Ringel MHS)
Correspondence to:
Dr M H Criqui
The countries involved in the analysis met an income cutoff level of US$9500 per capita of annual gross domestic product (GDP), which provided a clear bimodal separation of high-income and low-income countriesWe introduced this criterion to avoid potential ecological confounding by large differences in economic development. Of the 26 countries which met the GDP criterion, the former German Democratic Republic was excluded since
1719
*Annual litres per capita; t% of kilocalories.
Table 1: Nutrient and ethanol intake in 21 countries in years 1965, 1970, 1980, and 1988
provided as GDP per capita, and thus was Kuwait and the United Arab Emirates were comparable. excluded because of inconsistent alcohol data and Singapore was excluded because of inconsistent CHD mortality statistics. Belgium and Luxembourg were grouped as one entity since all food and most alcohol data sources grouped them together. The included were countries Australia, Austria, Belgium, income data
was not
not
Luxembourg, Canada, Denmark, Finland, France, Iceland, Ireland, Israel, Italy, Japan, Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the UK, the US, and (the former) West Germany. Reasonably comparable data were available for 1965, 1970, 1980, and 1988.
Mortality
data
The
for all
source
mortality
data
was
the World Health Statistics
Annuals, which list country-specific, age-specific, sex-specific, and cause-specific mortality rates.5 We analysed data for mortality at ages 35-74 from CHD, cirrhosis, and all causes. We chose an age group of 35-74 to include men and women at risk of CHD, but to exclude the elderly, where attribution of mortality to CHD may be unreliable. The classifications spanned the ICD-7 to ICD-9 codes. The ICD codes for CHD were: ICD7 (A81) "Atherosclerotic and degenerative heart disease", ICD-8 (A83) "Ischaemic heart disease", ICD-9 (270) "Acute myocardial infarction", plus (279) "Other ischaemic heart diseases". The ICD codes for cirrhosis were as follows: ICD-7 (A105) and, ICD-8 (AI02) "Cirrhosis of liver", and ICD-9 (347) "Chronic liver disease and cirrhosis". We used the direct method to adjust all mortality rates for age, with the European age structure given in the 1990 World Health Statistics Annual as standard. Age-adjustment was necessary to compare our mortality rates directly in countries with different age structures. For 1965, 1970, and 1980, the mortality rates for the total population were adjusted for age, whereas in 1988 the mortality rates for each sex were calculated separately, weighted, and combined into a single figure for the total population, since only sex-specific mortality rates were given in 1988.
included in the multiple regression colinearity with animal fat.
analyses because of inverse
Alcohol data For 1965 and 1970, the alcohol data was extracted from the Finnish Foundation for Alcohol Studies statistics7 while data for years 1981 and 1987 the data came from unpublished statistics compiled by the Alcohol Research Foundation in Toronto, Canada. Missing alcohol values were extrapolated from FAO Food balance sheets from the closest available yearbut these could not be used for all relevant years as they did not list different alcoholic beverages separately on an annual basis. We chose these sources as they were the most reliable: there was a very high correlation of data among the Finnish, Canadian, and FAO sources. Units for distilled spirits were provided as litres of absolute alcohol per capita. Wine and beer were converted to litres of absolute alcohol from data from the Finnish Foundation for 1965 and 1970.7 We also used 1970 percentage figures to calculate the 1981 and 1988 absolute alcohol values. This conversion was made to investigate whether ethanol itself, rather than other
Food data We obtained food disappearance data from the food balance sheets of the Food and Agriculture Organization (FAO) of the United Nations to assess averages for years 1964-1966, 1969-1971, 1979-1981, and 1986-1988.ó We assessed animal fat (all dietary fat of animal origin), vegetables, and fruit, each as a percentage of total calorie consumption. Vegetable fat was not
1720
Figure
1:
cirrhosis, 35-74
rates per 100 000 from CHD, other in 21 countries in 1988, ages and all causes
Age adjusted mortality
Table 2: Correlations of wine with rates, ages 35-74
Figure 2: Correlations of age adjusted CHD mortality rates per 100 000 with dietary items in 1988 ingredients (such
as
antioxidants),
was
responsible for mortality
rates.
We classed the Japanese alcoholic wines in this analysis.
Statistical
beverage, sake,
and cider
as
analysis
We made
simple univariate correlations between diet and ageadjusted mortality rates from CHD, and between age-adjusted mortality rates for CHD and cirrhosis. Separate multiple linear regression models examined age-adjusted mortality rates from CHD and all causes as dependent variables and in 1965, 1970, 1980, and 1988. For each model, the independent variables were dietary and alcohol variables from the relevant year. Results Table 1 shows the consumption of ethanol in wine, beer, and spirits in annual litres per capita, and the consumption of animal fat, vegetables, and fruit in kilocalories (1 kilocalorie=4-184 kilojoules) for years 1965, 1970, 1980, and 1988. There was some variability in each independent variable. The items with the largest variability were animal fat and wine ethanol.
Figure 1 shows the age-adjusted mortality rates per 100 000 in 1988 for CHD, cirrhosis, all other causes, and total mortality, ordered by CHD mortality rate. A low CHD mortality rate was associated with a high mortality rate from cirrhosis, and vice versa (r=-0-61). Other causes of mortality were somewhat lower in countries with higher CHD mortality. For example, in 1988 Sweden and Iceland had high CHD rates but low non-CHD and allcause mortality rates. Results from 1965, 1970, and 1980 were similar. Across the four years (1965, 1970, 1980, and 1988), univariate correlations between dietary items and CHD were consistent; figure 2 shows such data for 1988 CHD mortality rates. As expected, animal fat was positively correlated (r=0’40), and vegetables (r=-0°43) and fruit (r=-0’57) inversely correlated with CHD mortality in 1988. Total ethanol correlated to the same range as the dietary items (r=-0-39) but ethanol in wine showed the strongest correlation (r=-0-66). We reasoned that if components of wine other than ethanol (such as antioxidants) protected against CHD, there should be a stronger inverse gradient for litres of wine than for litres of ethanol in wine. However, table 2 shows that in each year studied, ethanol in wine was slightly more inversely correlated with CHD mortality rates than total wine volume (disregarding ethanol content). This association held true with or without the inclusion in the analysis of the most conspicuous outlier, Japan. Because of the intercorrelation of dietary items, we ran multiple linear regression models. Table 3 shows the results for CHD mortality in each of the 4 years studied, with wine ethanol, beer ethanol, spirits ethanol, % kcal animal fat, % kcal vegetables, and % kcal fruit as independent variables. In these analyses, the most consistent food association was animal fat, which was positively associated in 1965, 1970, and 1980, but unrelated in 1988. Fruit showed a positive association in the first two years, but an inverse association in the last two years. Because the only striking inconsistency across the total time was for fruit, we inspected individual data points (table 1). The change from a positive association in 1965 and 1970 to an inverse association in 1980 and 1QRn
*Beta coefficients reflect differences
Table 3:
in
age-adjusted CHD mortality
1QRR
total death rate per 100 000, and for ethanol are per litre per capita per annum, and for nutnents per 1% kcal
Multiple linear regressions of dietary items
on
age-adjusted CHD mortality rates, ages 35-74* 1721
*Beta coefficients reflect differences in total death rate per 100 000, and for ethanol
Table 4:
Multiple linear regressions of dietary items
on
are
age-adjusted
1988 reflected low reported fruit consumption in Japan only in the first two years (with low CHD rates), and high reported fruit consumption in Israel only in the first two
(with high CHD rates). Wine showed the most consistent association, being inversely associated with CHD in each year, with significant associations in 1970 and 1980. Models with additional adjustment for available data on average cigarettes consumed per adult in 1965 and 1970 showed similar results. In 1988, only 20-7% of deaths were due to CHD. Thus, we ran the same dietary items using total mortality as the end point. The results were quite different (table 4). No alcoholic average showed any benefit for total mortality, and there was a suggestion of an overall increased mortality risk for beer. There was a statistically significant benefit for fruit consumption in the last two years studied. The lack of benefit of wine for total mortality reflected positive associations between ethanol and several non-CHD causes of death, which will be the focus of a separate communication.
years
Discussion
ecological studies have shown a clear inverse wine consumption and CHD mortality between gradient in developed countries that have an average or greater than average intake of cholesterol and saturated fat. 1,2,3 Their ecological findings are consistent with a large and growing number of case-control and cohort studies within populations that show a protective effect for CHD morbidity and mortality with light to moderate (1 or 2 drinks per day) intake of alcohol.9,10 In addition, statistical pathway analyses in 4 separate studies are consistent with at least 50% of the protective effect being mediated by increases in HDL cholesterol. 11-14 Other investigators have noted the potential beneficial effects of ethanol on thrombosis. IS Wine seemed to have a stronger protective effect for CHD than either beer or spirits (table 3). However, within populations, wine, beer, and spirits in moderate amounts are all protective for CHD. In a California HMO, wine appeared to be more protective than beer, which appeared to be more protective than spirits.’6 Interestingly, white wine appeared more protective than red wine. However, in a study of USA female nurses, the lowest relative risk was for beer drinkers consuming less than 5 g per day, although the confidence intervals for beer, wine, and spirits were wide." In a USA study of male health professionals, spirits showed the most pronounced inverse association. 18 Thus, the overall evidence for greater protection against CHD by wine consumption is weak. Thus a second paradox needs to be resolved-why is there a discrepancy between ecological and cohort studies? The apparent greater benefit of wine could be an artifact. Countries in which wine is the predominant Previous
1722
per litre
per capita per annum, and for nutnents per 1% kcal.
total
mortality rates, ages 35-74*
beverage have the largest overall alcohol consumption. Thus, the much stronger association for wine in ecological studies may simply represent greater variability in the independent variable. Most cohort studies have been reported from countries such as the UK, USA, and Australia, where wine consumption does not exceed other beverages. A second possibility is that people in countries where wine is the principal beverage may have better nutrition, including foods such as olive oil and other components of the Mediterranean diet.19 FAO data do not permit a fair comparison of subtler dietary differences across the 21 countries. Another possibility is that wine may be drunk in a healthier fashion than other alcoholic beverages. The traditional consumption of wine with meals might have a favourable effect on post-prandial hyperlipidaemia.2O This area needs further study. Drinking with meals may also reduce some toxic effects of high alcohol consumption because of reduced absorption. Alternatively, wine might conceivably have other nonethanol beneficial effects. Wine, particularly red wine, has been reported to contain non-ethanol substances which act as antioxidantS,21 or have effects on thrombosis .22,21 These may produce benefits for CHD independent of any ethanol effects. However, the lack of consistent differences for wine as opposed to other alcoholic beverages suggests these effects are small compared with the apparent effect of ethanol through an HDLcholesterol pathway,9 and effects of ethanol on thrombotic factors such as platelet aggregationl5 and fibrinogen.24 The "French paradox" appears soluble. The French have the highest intake of alcohol in the world, and CHD mortality rates lower than any country studied, except for Japan. In Japan, the intake of animal fat has been historically very low. In terms of total mortality, the French ranked only 8th lowest in 1980 and 6th lowest in the world in 1988, congruent with the promotive effects of a high alcohol intake on non-CHD causes of mortality. Although the benefits for CHD itself may be retained at as high as 4 or 5 drinks per day,18,25 it is equally clear that consuming more than 2 drinks of alchol per day is progressively associated with higher risks of all cardiovascular diseases and other causes of mortality. The benefit for total mortality is maximised at one or two drinks per day.9,26 These data are sobering. Although the benefit of wine ethanol for CHD (table 3) appears impressive, alcohol consumption in a country sufficient to produce a large population benefit for CHD is routinely associated with high alcohol abuse. The correlation between mean consumption and extent of abuse in populations has been reported to be r=0-97." Although a strong correlation is predictable, given that alcohol consumption by heavy drinkers strongly influences mean consumption, nonetheless the magnitude of the correlation suggests that countries with high numbers of alcoholic
to moderate drinkers, who have increased longevity, also have significant numbers of alcohol abusers, who have sharply reduced longevity. It is interesting to note that in France, as the total annual litres of ethanol per capita decreased from 18-3 in 1965 to 13-1 in 1988, and was accompanied by a decrease in wine ethanol litres from 13-3 to 9-1, the CHD mortality rate did not increase and actually fell, from 94-9 per 100 000 in 1965 to 71-3 per 100 000 in 1988. However, the mortality rate from cirrhosis fell sharply from 68-5 per 100 000 in 1975 to 37-9 per 100 000. Thus, the French on average consumed far more alcohol than necessary for maximal cardioprotection, and likely continue to do so. In concluson, wine was the strongest dietary correlate of CHD mortality in 21 selected countries, and was consistently inversely related to CHD mortality. However, wine was not related to total mortality because of a positive association with the 79% of mortality not due to CHD. Data from cohort studies have consistently shown the CHD protective effect to plateau at 1-2 drinks per day. The total mortality data demonstrate that the benefits of alcohol preventing CHD at a population level are essentially cancelled by increases in other causes of death and suggest there is a protective effect on total mortality only for light to moderate drinking. Thus although the risk/benefit ratio for alcohol varies with each individual, alcohol consumption for cardioprotective purposes should not be encouraged as a public health measure, given the extent of alcohol abuse and a suggested genetic predisposition to alcoholism in many
light
persons.9,1 1,12,211, 29 This work was supported by a University of California, San Diego Medical Education and Research Foundation summer training grant. We thank Ms Julie Denenberg, MS Patricia Valiton, and Ms Jennifer Hostettler for assistance in computing and manuscript preparation.
References St Leger AS, Cochrane AL, Moore F. Factors associated with cardiac mortality in developed countries with particular reference to the consumption of wine. Lancet 1979; i: 1017-20. 2 Renaud S, de Lorgeril M. Wine, alcohol, platelets, and the French paradox for coronary heart disease. Lancet 1992; 339: 1523-26. 3 Artaud-Wild SM, Connor S, Sexton G, Connor W. Differences in coronary mortality can be explained by differences in cholesterol and 1
saturated fat intakes in 40 countries but
paradox. Circulation 1993; 88: 4
not
in France and Finland:
World Tables 1993. The World Bank. Baltimore: The Johns
University Press,
a
2771-79.
Hopkins
1993.
World Health Statistics Annual. Geneva: World Health Organization, various years. 6 Food Balance Sheets. Rome: Food and Agricultural Organization of the United Nations, various years. 7 International Statistics on Alcoholic Beverages: Production, Trade & Consumption 1950-1972. The World Health Organization Regional Office for Europe. The Finnish Foundation for Alcohol Studies. Forssa, Finland: Aurasen Kirjapaino, Vol 27, 1977. 5
World Drink Trends. Toronto: Alcoholism Research Foundation. 1990 edition (unpublished). 9 Criqui MH. The reduction of coronary heart disease with light to moderate alcohol consumption: effect or artifact? Br J Addict 1990; 85: 854-57. 10 Marmot M, Brunner E. Alcohol and cardiovascular disease: the status of the U-shaped curve. BMJ 1991; 303: 565-68. 11 Criqui MH, Cowan LD, Tyroler HA, et al. Lipoproteins as mediators for the effects of alcohol consumption and cigarette smoking on cardiovascular mortality. Results from the Lipid Research Clinical Follow-up Study. Am J Epidemiol 1987; 126: 629-37 12 Langer RD, Criqui MH, Reed DM. Lipoproteins and blood pressure as biologic pathways for effect of moderate alcohol consumption on coronary heart disease. Circulation 1992; 85: 910-15. 13 Suh I, Shaten BJ, Cutler JA, Kuller LH for the Multiple Risk Factor Intervention Trial Research Group. Alcohol use and mortality from coronary heart disease: the role of high-density lipoprotein cholesterol. Am Intern Med 1992; 116: 881-87. 14 Gaziano JM, Buring JE, Breslow JL, et al. Moderate alcohol intake, increased levels of high-density lipoprotein and its subfractions, and decreased risk of myocardial infarction. N Engl J Med 1993; 329: 1829-34. 15 Renaud SC, Beswick AD, Fehily AM, Sharp DS, Elwood PC. Alcohol and platelet aggregation: the Caerphilly Prospective Heart Disease Study. Am J Clin Nutr 1992; 55: 1012-17. 16 Klatsky AL, Armstrong MA. Alcoholic beverage choice and risk of coronary artery disease mortality: do red wine drinkers fare best? Am J Cardiol 1993; 71: 467-69. 17 Stampfer MJ, Colditz GA, Willett WC, Speizer FE, Hennekens CH. A prospective study of moderate alcohol consumption and the risk of coronary disease and stroke in women. N Engl J Med 1988; 319: 267-73. 18 Rimm EB, Giovannucci EL, Willett WC, et al. Prospective study of alcohol consumption and risk of coronary disease in men. Lancet 1991; 338: 464-68. 19 Ulbricht TLV, Southgate DAT. Coronary heart disease: seven dietary factors. Lancet 1991; 338: 985-92. 20 Veenstra J, Ockhuizen T, Van de Pol H, Wedel M, Schaafsma G. Effects of a moderate dose of alcohol on blood lipids and lipoproteins postprandially and in the fasting state. Alcohol & Alcoholism 1990; 25: 371-77. 21 Frankel EN, Kanner J, German JB, Parks E, Kinsella JE. Inhibition of oxidation of human low-density lipoprotein by phenolic substances in red wine. Lancet 1993; 341: 454-57. 22 Demrow HS, Folts JD. Gastric or IV administration of French red wine but not white wine inhibits in vivo platelet activity and thrombosis in stenosed canine coronary arteries. J Am Coll Cardiol 1994; 23: 49A 8
(abstract). 23 Folts JD, Demrow HS, Slane PR. Two glasses of red but not white wine inhibits ex-vivo platelet aggregation and increases bleeding time in human volunteers. J Am Coll Cardiol 1994; 23: 66A (abstract). 24 Meade TW, Chakrabarti R, Haines AP, North WRS, Stirling Y. Characteristics affecting fibrinolytic activity and plasma fibrinogen concentrations. BMJ 1979; 1: 153-56. 25 Jackson R, Scragg R, Beaglehole R. Alcohol consumption and risk of coronary heart disease. BMJ 1991; 303: 211-16. 26 Boffetta P, Garfinkel L. Alcohol drinking and mortality among men enrolled in an American Cancer Society Prospective Study. Epidemiology 1990; 1: 342-48. 27 Rose G. Ancel Keys Lecture. Circulation 1991; 84: 1405-09. 28 Criqui MH. Alcohol—the bottom line: specific health benefits and risks, with a focus on cardiovascular disease. Contemporary Drug Problems (in press). 29 Goodwin DE. Hereditary factors in alcoholism. Hosp Pract 1978; 13: 121-24.
1723
or
alcohol
explain
the French
paradox?
Introduction
Summary The low rate of coronary heart disease (CHD) in France compared with other developed countries with comparable dietary intake has been called the French paradox. We explored this paradox by looking at alcohol, diet, and mortality data from 21 developed, relatively affluent countries in the years 1965, 1970, 1980, and 1988. We assessed wine, beer, and spirits intake separately. France had the highest wine intake and the highest total alcohol intake, and the second lowest CHD mortality rate. In univariate analyses, ethanol in wine was slightly more inversely correlated with CHD than total wine volume. In multivariate analyses, animal fat tended to be positively correlated, and fruit consumption inversely correlated, with CHD. Beer and spirits were only weakly inversely correlated with CHD. The strongest and most consistent correlation was the inverse association of wine ethanol with CHD. However, wine ethanol was unrelated to total mortality. We conclude that ethanol, particularly wine ethanol, is inversely related to CHD but not to longevity in populations. Although light to moderate alcohol consumption may improve longevity, alcohol abuse—which sharply reduces longevity—is correlated with average alcohol consumption in populations. Thus, while the risk/benefit ratio varies for individuals, the use of alcohol for cardioprotective purposes should not be encouraged as a public health measure.
investigators have noted the strikingly low rate of mortality from coronary heart disease (CHD) in France compared with other developed countries.1,2 St Leger et all assessed 18 developed countries but excluded Japan because of possible genetic factors. They found a strong inverse association between CHD mortality in 1970 and wine consumption. They also found positive associations for saturated and monounsaturated fat intake, and a negative association for polyunsaturated fat. However, in multiple regression analyses wine remained a significant (inverse) predictor of CHD mortality, whilst the Keys equation, accounting for dietary polyunsaturated, monounsaturated and saturated fat, and cholesterol, was not significantly related to CHD mortality. Renaud et aP Several
showed
that
in
17
countries
with
available
wine
consumption data, CHD mortality at all ages in 1987 was positively related to dairy fat with a correlation r=0-73. France, and to lesser degree, Switzerland showed the greatest deviation from the regression line, with low CHD rates despite above-average dairy fat intake. Adjustment for the inverse effect of wine consumption moved France and Switzerland quite close to the regression line, and improved the correlation to 0-87. Artaud-Wild et aP looked at the association between a cholesterol-saturated fat index and CHD mortality in men in 40 countries at varying levels of economic development and found the cholesterol-saturated fat index positively related to CHD mortality in 1977 with a correlation of 078.3 Finland had a higher-than-expected and France a lower-than-expected CHD mortality rate. However, lower milk and higher vegetable consumption in France partly explained this discrepancy. Alcohol was unrelated to CHD after adjustment for CSI. However, wine was inversely (r=-0-67) related to CHD mortality in the 18 countries with a cholesterol-saturated fat index above 20 per 1000 kcal. Thus, these three studies are only partly concordant. Discordant findings appear to reflect different selection of countries, different definitions of dietary and alcohol intake, different gender and age groups, and different study years. To clarify these issues, we analysed data from 21 developed countries, evaluating various dietary components and alcoholic beverages as predictive of CHD and total mortality in 1965, 1970, 1980, and 1988.
Methods
Country selection
Departments of Family and Preventive Medicine, and Medicine, University of California, San Diego School of Medicine, La Jolla, CA 92093-0607 (M H Criqui MD, MPH, B L Ringel MHS)
Correspondence to:
Dr M H Criqui
The countries involved in the analysis met an income cutoff level of US$9500 per capita of annual gross domestic product (GDP), which provided a clear bimodal separation of high-income and low-income countriesWe introduced this criterion to avoid potential ecological confounding by large differences in economic development. Of the 26 countries which met the GDP criterion, the former German Democratic Republic was excluded since
1719
*Annual litres per capita; t% of kilocalories.
Table 1: Nutrient and ethanol intake in 21 countries in years 1965, 1970, 1980, and 1988
provided as GDP per capita, and thus was Kuwait and the United Arab Emirates were comparable. excluded because of inconsistent alcohol data and Singapore was excluded because of inconsistent CHD mortality statistics. Belgium and Luxembourg were grouped as one entity since all food and most alcohol data sources grouped them together. The included were countries Australia, Austria, Belgium, income data
was not
not
Luxembourg, Canada, Denmark, Finland, France, Iceland, Ireland, Israel, Italy, Japan, Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the UK, the US, and (the former) West Germany. Reasonably comparable data were available for 1965, 1970, 1980, and 1988.
Mortality
data
The
for all
source
mortality
data
was
the World Health Statistics
Annuals, which list country-specific, age-specific, sex-specific, and cause-specific mortality rates.5 We analysed data for mortality at ages 35-74 from CHD, cirrhosis, and all causes. We chose an age group of 35-74 to include men and women at risk of CHD, but to exclude the elderly, where attribution of mortality to CHD may be unreliable. The classifications spanned the ICD-7 to ICD-9 codes. The ICD codes for CHD were: ICD7 (A81) "Atherosclerotic and degenerative heart disease", ICD-8 (A83) "Ischaemic heart disease", ICD-9 (270) "Acute myocardial infarction", plus (279) "Other ischaemic heart diseases". The ICD codes for cirrhosis were as follows: ICD-7 (A105) and, ICD-8 (AI02) "Cirrhosis of liver", and ICD-9 (347) "Chronic liver disease and cirrhosis". We used the direct method to adjust all mortality rates for age, with the European age structure given in the 1990 World Health Statistics Annual as standard. Age-adjustment was necessary to compare our mortality rates directly in countries with different age structures. For 1965, 1970, and 1980, the mortality rates for the total population were adjusted for age, whereas in 1988 the mortality rates for each sex were calculated separately, weighted, and combined into a single figure for the total population, since only sex-specific mortality rates were given in 1988.
included in the multiple regression colinearity with animal fat.
analyses because of inverse
Alcohol data For 1965 and 1970, the alcohol data was extracted from the Finnish Foundation for Alcohol Studies statistics7 while data for years 1981 and 1987 the data came from unpublished statistics compiled by the Alcohol Research Foundation in Toronto, Canada. Missing alcohol values were extrapolated from FAO Food balance sheets from the closest available yearbut these could not be used for all relevant years as they did not list different alcoholic beverages separately on an annual basis. We chose these sources as they were the most reliable: there was a very high correlation of data among the Finnish, Canadian, and FAO sources. Units for distilled spirits were provided as litres of absolute alcohol per capita. Wine and beer were converted to litres of absolute alcohol from data from the Finnish Foundation for 1965 and 1970.7 We also used 1970 percentage figures to calculate the 1981 and 1988 absolute alcohol values. This conversion was made to investigate whether ethanol itself, rather than other
Food data We obtained food disappearance data from the food balance sheets of the Food and Agriculture Organization (FAO) of the United Nations to assess averages for years 1964-1966, 1969-1971, 1979-1981, and 1986-1988.ó We assessed animal fat (all dietary fat of animal origin), vegetables, and fruit, each as a percentage of total calorie consumption. Vegetable fat was not
1720
Figure
1:
cirrhosis, 35-74
rates per 100 000 from CHD, other in 21 countries in 1988, ages and all causes
Age adjusted mortality
Table 2: Correlations of wine with rates, ages 35-74
Figure 2: Correlations of age adjusted CHD mortality rates per 100 000 with dietary items in 1988 ingredients (such
as
antioxidants),
was
responsible for mortality
rates.
We classed the Japanese alcoholic wines in this analysis.
Statistical
beverage, sake,
and cider
as
analysis
We made
simple univariate correlations between diet and ageadjusted mortality rates from CHD, and between age-adjusted mortality rates for CHD and cirrhosis. Separate multiple linear regression models examined age-adjusted mortality rates from CHD and all causes as dependent variables and in 1965, 1970, 1980, and 1988. For each model, the independent variables were dietary and alcohol variables from the relevant year. Results Table 1 shows the consumption of ethanol in wine, beer, and spirits in annual litres per capita, and the consumption of animal fat, vegetables, and fruit in kilocalories (1 kilocalorie=4-184 kilojoules) for years 1965, 1970, 1980, and 1988. There was some variability in each independent variable. The items with the largest variability were animal fat and wine ethanol.
Figure 1 shows the age-adjusted mortality rates per 100 000 in 1988 for CHD, cirrhosis, all other causes, and total mortality, ordered by CHD mortality rate. A low CHD mortality rate was associated with a high mortality rate from cirrhosis, and vice versa (r=-0-61). Other causes of mortality were somewhat lower in countries with higher CHD mortality. For example, in 1988 Sweden and Iceland had high CHD rates but low non-CHD and allcause mortality rates. Results from 1965, 1970, and 1980 were similar. Across the four years (1965, 1970, 1980, and 1988), univariate correlations between dietary items and CHD were consistent; figure 2 shows such data for 1988 CHD mortality rates. As expected, animal fat was positively correlated (r=0’40), and vegetables (r=-0°43) and fruit (r=-0’57) inversely correlated with CHD mortality in 1988. Total ethanol correlated to the same range as the dietary items (r=-0-39) but ethanol in wine showed the strongest correlation (r=-0-66). We reasoned that if components of wine other than ethanol (such as antioxidants) protected against CHD, there should be a stronger inverse gradient for litres of wine than for litres of ethanol in wine. However, table 2 shows that in each year studied, ethanol in wine was slightly more inversely correlated with CHD mortality rates than total wine volume (disregarding ethanol content). This association held true with or without the inclusion in the analysis of the most conspicuous outlier, Japan. Because of the intercorrelation of dietary items, we ran multiple linear regression models. Table 3 shows the results for CHD mortality in each of the 4 years studied, with wine ethanol, beer ethanol, spirits ethanol, % kcal animal fat, % kcal vegetables, and % kcal fruit as independent variables. In these analyses, the most consistent food association was animal fat, which was positively associated in 1965, 1970, and 1980, but unrelated in 1988. Fruit showed a positive association in the first two years, but an inverse association in the last two years. Because the only striking inconsistency across the total time was for fruit, we inspected individual data points (table 1). The change from a positive association in 1965 and 1970 to an inverse association in 1980 and 1QRn
*Beta coefficients reflect differences
Table 3:
in
age-adjusted CHD mortality
1QRR
total death rate per 100 000, and for ethanol are per litre per capita per annum, and for nutnents per 1% kcal
Multiple linear regressions of dietary items
on
age-adjusted CHD mortality rates, ages 35-74* 1721
*Beta coefficients reflect differences in total death rate per 100 000, and for ethanol
Table 4:
Multiple linear regressions of dietary items
on
are
age-adjusted
1988 reflected low reported fruit consumption in Japan only in the first two years (with low CHD rates), and high reported fruit consumption in Israel only in the first two
(with high CHD rates). Wine showed the most consistent association, being inversely associated with CHD in each year, with significant associations in 1970 and 1980. Models with additional adjustment for available data on average cigarettes consumed per adult in 1965 and 1970 showed similar results. In 1988, only 20-7% of deaths were due to CHD. Thus, we ran the same dietary items using total mortality as the end point. The results were quite different (table 4). No alcoholic average showed any benefit for total mortality, and there was a suggestion of an overall increased mortality risk for beer. There was a statistically significant benefit for fruit consumption in the last two years studied. The lack of benefit of wine for total mortality reflected positive associations between ethanol and several non-CHD causes of death, which will be the focus of a separate communication.
years
Discussion
ecological studies have shown a clear inverse wine consumption and CHD mortality between gradient in developed countries that have an average or greater than average intake of cholesterol and saturated fat. 1,2,3 Their ecological findings are consistent with a large and growing number of case-control and cohort studies within populations that show a protective effect for CHD morbidity and mortality with light to moderate (1 or 2 drinks per day) intake of alcohol.9,10 In addition, statistical pathway analyses in 4 separate studies are consistent with at least 50% of the protective effect being mediated by increases in HDL cholesterol. 11-14 Other investigators have noted the potential beneficial effects of ethanol on thrombosis. IS Wine seemed to have a stronger protective effect for CHD than either beer or spirits (table 3). However, within populations, wine, beer, and spirits in moderate amounts are all protective for CHD. In a California HMO, wine appeared to be more protective than beer, which appeared to be more protective than spirits.’6 Interestingly, white wine appeared more protective than red wine. However, in a study of USA female nurses, the lowest relative risk was for beer drinkers consuming less than 5 g per day, although the confidence intervals for beer, wine, and spirits were wide." In a USA study of male health professionals, spirits showed the most pronounced inverse association. 18 Thus, the overall evidence for greater protection against CHD by wine consumption is weak. Thus a second paradox needs to be resolved-why is there a discrepancy between ecological and cohort studies? The apparent greater benefit of wine could be an artifact. Countries in which wine is the predominant Previous
1722
per litre
per capita per annum, and for nutnents per 1% kcal.
total
mortality rates, ages 35-74*
beverage have the largest overall alcohol consumption. Thus, the much stronger association for wine in ecological studies may simply represent greater variability in the independent variable. Most cohort studies have been reported from countries such as the UK, USA, and Australia, where wine consumption does not exceed other beverages. A second possibility is that people in countries where wine is the principal beverage may have better nutrition, including foods such as olive oil and other components of the Mediterranean diet.19 FAO data do not permit a fair comparison of subtler dietary differences across the 21 countries. Another possibility is that wine may be drunk in a healthier fashion than other alcoholic beverages. The traditional consumption of wine with meals might have a favourable effect on post-prandial hyperlipidaemia.2O This area needs further study. Drinking with meals may also reduce some toxic effects of high alcohol consumption because of reduced absorption. Alternatively, wine might conceivably have other nonethanol beneficial effects. Wine, particularly red wine, has been reported to contain non-ethanol substances which act as antioxidantS,21 or have effects on thrombosis .22,21 These may produce benefits for CHD independent of any ethanol effects. However, the lack of consistent differences for wine as opposed to other alcoholic beverages suggests these effects are small compared with the apparent effect of ethanol through an HDLcholesterol pathway,9 and effects of ethanol on thrombotic factors such as platelet aggregationl5 and fibrinogen.24 The "French paradox" appears soluble. The French have the highest intake of alcohol in the world, and CHD mortality rates lower than any country studied, except for Japan. In Japan, the intake of animal fat has been historically very low. In terms of total mortality, the French ranked only 8th lowest in 1980 and 6th lowest in the world in 1988, congruent with the promotive effects of a high alcohol intake on non-CHD causes of mortality. Although the benefits for CHD itself may be retained at as high as 4 or 5 drinks per day,18,25 it is equally clear that consuming more than 2 drinks of alchol per day is progressively associated with higher risks of all cardiovascular diseases and other causes of mortality. The benefit for total mortality is maximised at one or two drinks per day.9,26 These data are sobering. Although the benefit of wine ethanol for CHD (table 3) appears impressive, alcohol consumption in a country sufficient to produce a large population benefit for CHD is routinely associated with high alcohol abuse. The correlation between mean consumption and extent of abuse in populations has been reported to be r=0-97." Although a strong correlation is predictable, given that alcohol consumption by heavy drinkers strongly influences mean consumption, nonetheless the magnitude of the correlation suggests that countries with high numbers of alcoholic
to moderate drinkers, who have increased longevity, also have significant numbers of alcohol abusers, who have sharply reduced longevity. It is interesting to note that in France, as the total annual litres of ethanol per capita decreased from 18-3 in 1965 to 13-1 in 1988, and was accompanied by a decrease in wine ethanol litres from 13-3 to 9-1, the CHD mortality rate did not increase and actually fell, from 94-9 per 100 000 in 1965 to 71-3 per 100 000 in 1988. However, the mortality rate from cirrhosis fell sharply from 68-5 per 100 000 in 1975 to 37-9 per 100 000. Thus, the French on average consumed far more alcohol than necessary for maximal cardioprotection, and likely continue to do so. In concluson, wine was the strongest dietary correlate of CHD mortality in 21 selected countries, and was consistently inversely related to CHD mortality. However, wine was not related to total mortality because of a positive association with the 79% of mortality not due to CHD. Data from cohort studies have consistently shown the CHD protective effect to plateau at 1-2 drinks per day. The total mortality data demonstrate that the benefits of alcohol preventing CHD at a population level are essentially cancelled by increases in other causes of death and suggest there is a protective effect on total mortality only for light to moderate drinking. Thus although the risk/benefit ratio for alcohol varies with each individual, alcohol consumption for cardioprotective purposes should not be encouraged as a public health measure, given the extent of alcohol abuse and a suggested genetic predisposition to alcoholism in many
light
persons.9,1 1,12,211, 29 This work was supported by a University of California, San Diego Medical Education and Research Foundation summer training grant. We thank Ms Julie Denenberg, MS Patricia Valiton, and Ms Jennifer Hostettler for assistance in computing and manuscript preparation.
References St Leger AS, Cochrane AL, Moore F. Factors associated with cardiac mortality in developed countries with particular reference to the consumption of wine. Lancet 1979; i: 1017-20. 2 Renaud S, de Lorgeril M. Wine, alcohol, platelets, and the French paradox for coronary heart disease. Lancet 1992; 339: 1523-26. 3 Artaud-Wild SM, Connor S, Sexton G, Connor W. Differences in coronary mortality can be explained by differences in cholesterol and 1
saturated fat intakes in 40 countries but
paradox. Circulation 1993; 88: 4
not
in France and Finland:
World Tables 1993. The World Bank. Baltimore: The Johns
University Press,
a
2771-79.
Hopkins
1993.
World Health Statistics Annual. Geneva: World Health Organization, various years. 6 Food Balance Sheets. Rome: Food and Agricultural Organization of the United Nations, various years. 7 International Statistics on Alcoholic Beverages: Production, Trade & Consumption 1950-1972. The World Health Organization Regional Office for Europe. The Finnish Foundation for Alcohol Studies. Forssa, Finland: Aurasen Kirjapaino, Vol 27, 1977. 5
World Drink Trends. Toronto: Alcoholism Research Foundation. 1990 edition (unpublished). 9 Criqui MH. The reduction of coronary heart disease with light to moderate alcohol consumption: effect or artifact? Br J Addict 1990; 85: 854-57. 10 Marmot M, Brunner E. Alcohol and cardiovascular disease: the status of the U-shaped curve. BMJ 1991; 303: 565-68. 11 Criqui MH, Cowan LD, Tyroler HA, et al. Lipoproteins as mediators for the effects of alcohol consumption and cigarette smoking on cardiovascular mortality. Results from the Lipid Research Clinical Follow-up Study. Am J Epidemiol 1987; 126: 629-37 12 Langer RD, Criqui MH, Reed DM. Lipoproteins and blood pressure as biologic pathways for effect of moderate alcohol consumption on coronary heart disease. Circulation 1992; 85: 910-15. 13 Suh I, Shaten BJ, Cutler JA, Kuller LH for the Multiple Risk Factor Intervention Trial Research Group. Alcohol use and mortality from coronary heart disease: the role of high-density lipoprotein cholesterol. Am Intern Med 1992; 116: 881-87. 14 Gaziano JM, Buring JE, Breslow JL, et al. Moderate alcohol intake, increased levels of high-density lipoprotein and its subfractions, and decreased risk of myocardial infarction. N Engl J Med 1993; 329: 1829-34. 15 Renaud SC, Beswick AD, Fehily AM, Sharp DS, Elwood PC. Alcohol and platelet aggregation: the Caerphilly Prospective Heart Disease Study. Am J Clin Nutr 1992; 55: 1012-17. 16 Klatsky AL, Armstrong MA. Alcoholic beverage choice and risk of coronary artery disease mortality: do red wine drinkers fare best? Am J Cardiol 1993; 71: 467-69. 17 Stampfer MJ, Colditz GA, Willett WC, Speizer FE, Hennekens CH. A prospective study of moderate alcohol consumption and the risk of coronary disease and stroke in women. N Engl J Med 1988; 319: 267-73. 18 Rimm EB, Giovannucci EL, Willett WC, et al. Prospective study of alcohol consumption and risk of coronary disease in men. Lancet 1991; 338: 464-68. 19 Ulbricht TLV, Southgate DAT. Coronary heart disease: seven dietary factors. Lancet 1991; 338: 985-92. 20 Veenstra J, Ockhuizen T, Van de Pol H, Wedel M, Schaafsma G. Effects of a moderate dose of alcohol on blood lipids and lipoproteins postprandially and in the fasting state. Alcohol & Alcoholism 1990; 25: 371-77. 21 Frankel EN, Kanner J, German JB, Parks E, Kinsella JE. Inhibition of oxidation of human low-density lipoprotein by phenolic substances in red wine. Lancet 1993; 341: 454-57. 22 Demrow HS, Folts JD. Gastric or IV administration of French red wine but not white wine inhibits in vivo platelet activity and thrombosis in stenosed canine coronary arteries. J Am Coll Cardiol 1994; 23: 49A 8
(abstract). 23 Folts JD, Demrow HS, Slane PR. Two glasses of red but not white wine inhibits ex-vivo platelet aggregation and increases bleeding time in human volunteers. J Am Coll Cardiol 1994; 23: 66A (abstract). 24 Meade TW, Chakrabarti R, Haines AP, North WRS, Stirling Y. Characteristics affecting fibrinolytic activity and plasma fibrinogen concentrations. BMJ 1979; 1: 153-56. 25 Jackson R, Scragg R, Beaglehole R. Alcohol consumption and risk of coronary heart disease. BMJ 1991; 303: 211-16. 26 Boffetta P, Garfinkel L. Alcohol drinking and mortality among men enrolled in an American Cancer Society Prospective Study. Epidemiology 1990; 1: 342-48. 27 Rose G. Ancel Keys Lecture. Circulation 1991; 84: 1405-09. 28 Criqui MH. Alcohol—the bottom line: specific health benefits and risks, with a focus on cardiovascular disease. Contemporary Drug Problems (in press). 29 Goodwin DE. Hereditary factors in alcoholism. Hosp Pract 1978; 13: 121-24.
1723