1017 TABLE I-COUNTRIES EXAMINED IN THIS STUDY
Public Health
FACTORS ASSOCIATED WITH CARDIAC MORTALITY IN DEVELOPED COUNTRIES WITH PARTICULAR REFERENCE TO THE CONSUMPTION OF WINE A. S. ST. LEGER
A. L. COCHRANE* F. MOORE
Medical Research Council Epidemiology Unit, 3AS
Cardiff CF2
Deaths from ischæmic heart-disease in 18 developed countries are not strongly associated with health-service factors such as doctor and nurse density. There is a negative association with gross national product per capita and a positive but inconsistent association with saturated and monounsaturated fat intake. The principal finding is a strong and specific negative association between ischæmic heart-disease deaths and alcohol consumption. This is shown to be wholly attributable to wine consumption.
Summary
INTRODUCTION
IN a previous study1.2we reported on the associations between age-specific total mortality and a variety of economic, health-service, and dietary variables in 18 developed countries. Our conclusion was that, in general, the health-service factors showed negligible association with mortality, the dietary factors were of little importance, and the only consistent pattern was a negative association between gross national product (G.N.P.) per
capita and mortality. Age-specific total mortality is, however, a crude measure by which to compare countries and so we have embarked upon a comparison of particular diseases. Here we shall present results for heart-disease. *Present address: Rhoose Farm House, Rhoose.
METHODS
The 18 countries used in this study are listed in table i. The criteria for their selection are stated in our previous work.’1 Our mortality data consists of death-rates for males and females separately in the 55-64 age-group from hypertensive disease, ischsemic heart-disease (I.H.D.), cerebro-vascular disease, and bronchitis plus emphysema and asthma; and in the 25-34 age group death-rates from road accidents. These death-rates were obtained from the World Health Statistics Annual for 1970 (codes A82, A83, A85, A93, and AE138, respectively). The latter two causes of death were included in the study for comparison with heart-disease. It has been suggested3 that France underreports i.H.D. deaths and we made allowance for this by analysing our data with France included and excluded. Our health-service, economic, and demographic variables have been described elsewhere,’ as have cigarette consumption per capita per annum and total alcohol consumption. The only additional variables we used in this study were: (1) Consumption (litres per capita per annum) of absolute alcohol attributed separately to beer, wine, and spirits.4 (2) Average daily intake of total calories and carbohydrate, and of total, saturated, monounsaturated, and polyunsaturated fat.5 Fat intake was also expressed as an energy percentage of total calorie intake. These figures were averaged over the period 1954-62 which is reasonable in relation to mortality in 1970. (3) We also obtained daily cholesterol consumption5 which, together with total saturated fats and polyunsaturated fats, we inserted into Keys6 predictive equation for serum-cholesterol. As in our previous study we adopted a pragmatic approach to the statistical analysis. We examined scatter diagrams of one variable plotted against another and product moment correlation coefficients. The variables showing the most convincing associations with the death-rates or with other "explanatory"
TABLE 11-CORRELATION COEFFICIENTS BETWEEN DEATH-RATES AND CERTAIN VARIABLES
1018 variables were selected for examination by regression analysis. No emphasis has been placed on formal significance testing but in the table presenting a summary of the regression analyses we find it helpful to indicate where the t value for inclusion of an "explanatory" variable in the equation exceeds 2. RESULTS
Table n displays the coefficients of correlation between the death-rates and a selection of our "explanatory" variables. In interpreting this table we have conof the coefficients but sidered not only the also their consistency of sign and consistency between the sexes. Doctor density has a negative correlation with the deaths from heart-disease and bronchitis. Nurse density shows a less consistent pattern, there being particularly large positive correlations with deaths from I.H.D. G.N.P. per capita is quite strongly negatively correlated with all the death-rates except road accidents. Population density is mostly negatively correlated with the death-rates and this correlation is fairly large with I.H.D. deaths. Cigarette consumption is positively associated with all causes of death. Total calorie intake is strongly positively associated with I.H.D. and bronchitis deaths. Total fat intake is quite strongly positively associated with I.H.D. mortality and negatively associated with cerebrovascular disease deaths in males. Both saturated fats and monounsaturated fats are strongly positively associated with I.H.D. mortality and not strongly positively associated elsewhere. Polyunsaturated fats are moderately to weakly negatively associated with deaths other than road accidents. Keys’ prediction of average serum-cholesterol is very strongly positively associated with I.H.D. deaths and shows little association elsewhere. Not included in the table are total fat, saturated fat, monounsaturated fat and polyunsaturated fat intake expressed as an energy percentage of total calorie intake. Their pattern of association with I.H.D. deaths is almost identical to that of these fats expressed in absolute
magnitude
terms.
Alcohol consumption is strongly negatively associated with I.H.D. deaths, moderately negatively with bronchitis, and moderately positively with road-accident deaths. The large correlation with I.H.D. deaths prompted us to break alcohol consumption into its three main components-wines, beers, and spirits. As can be seen from table n, wine appears to account for the entire
Relationship between I.H.D. mortality-rate and wine consumption.
men
aged 55-64
alcohol effect and the correlation coefficient with I.H.D. deaths is of the same magnitude as that of Keys’ prediction. The figure shows the association between I.H.D. deaths in males aged 55-64 and the logarithm of wine
consumption (litres per capita per annum). Regression analysis was done with various subsets of the independent variables. The most important findings were:
and nurse density showed relatively little association with the death-rates when other variables were taken into account. (2) G.N.P. per capita was strongly negatively related to mortality and this effect was independent of other variables.
(1) Doctor
(3) Cigarette consumption
was
strongly positively
related to bronchitis mortality and showed appreciable association with hypertensive disease deaths in both sexes and I.H.D. deaths in females. (4) Alcohol consumption was very strongly negatively associated with I.H.D. mortalityin both males and females. Of the three alcohol components wine had the strongest association with I.H.D., its magnitude being the same as that of total alcohol. The wine effect was strong
TABLE III-REGRESSION ANALYSIS: DEATH RATE AS DEPENDENT VARIABLE
* value for inclusion of the variable in the regression set exceeds 2. tthe numbers in the table are standardised regression effects (see text).
in
1019 no matter what other variables were in the regression equation-in other words it was not explained away by G.N.P., cigarettes, or diet. (5) The dietary variables showed less clearcut associations with mortality in the presence of other variables than they did when considered singly. Total fat, saturated fat, and monounsaturated fat were positively associated with I.H.D. mortality when in the presence of
other variables but on the whole these associations were weak. Polyunsaturated fat tended to be strongly negatively related to I.H.D. mortality. Keys’ prediction seemed to carry most of the information present singly or collectively in the fats. The size of its association with I.H.D. mortality depended heavily on the nature of other variables in the regression equation. Excluding France from the analysis did not substantially alter the results. Table III shows some regression analyses using a particular set of independent variables. These variables were chosen for their intrinsic interest (doctors and nurses) or for their importance. The figures in the table are standardised regression effects which have been explained in detail by St. Leger and Sweetnam.7 In essence they represent the proportional change of deathrate to a one standard deviation increase in the independent variable from its mean value. Any two standardised regression effects in the table may be compared, if desired. Most of them are subject to considerable random variation so greater weight should be placed on pattern and consistency than on magnitude alone. DISCUSSION
The results have to be interpreted in the light of the general validity and known pitfalls of this type of study and its method of analysis. This has been discussed elsewhere’ but here we must stress that all our findings, despite the apparent complexity of the analysis, are statistical associations only and are not themselves sufficient evidence to impute causality to any of the observed relationships. Nevertheless these associations deserve further comment and should be viewed within the wider context of other findings. We found negligible association between health service indices-e.g., doctor and nurse density-and mortality from heart-disease or the comparison mortalities from bronchitis and road-traffic accidents. This is in accord with our previous findings regarding overall age-
specific mortality. As before’ increasing per capita wealth appears to have a protective influence not explained by any of the other variables. The association between cigarette consumption and bronchitis is in accord with reasonable expectation and lends support to the validity of our data. The association between cigarette consumption and hypertensive disease and I.H.D. is again roughly in agreement with expectation. The
correlations between I.H.D. death-rate and fats accord with those presented by Stamler et al.5 This is to be expected since our data are similar to theirs. But we find a little surprising their statement that: "It is valid at the present time to conclude that-at a high level of probability-a cause-and-effect relationship has been demonstrated between dietary lipid (specifically, saturated fat and cholesterol) and
dietary
raw
widespread premature coronary heart disease even though substantial direct proof from definitive, largescale long term mass field trials is still to be obtained." Since this was written, large-scale field trials have been completed. The clofibrate trial8 certainly showed a reduction of serum-cholesterol and non-fatal myocardial infarcts in the treated group but there was no difference in I.H.D. mortality and there was an increase in total mortality in that group. Furthermore the evidence from a substantial number of studies9 on the effect of diet, or changes in diet, on the development of I.H.D. is confused and contradictory. The strong negative partial association in our data between I.H.D. death-rate and polyunsaturated fat intake is inconsistent with current opinion because, in the absence of a strong positive partial association between I.H.D. death-rate and saturated fats, this finding is more consistent with the view that polyunsaturated fats are themselves "beneficial" rather than merely apparently beneficial as would be the case were they merely replacing "harmful" saturated fats in the diet. We admit that this argument, being based as it is on association and partial association, must be treated with caution. By far the most interesting result to emerge from our analysis was the strong, specific association between I.H.D. deaths and alcohol consumption, more particularly with wine. This association was not explained away by by fat consumption, Keys’ prediction, or any of the other variables we examined. Correlation studies are bound from time to time to throw up curious but fundamentally insignificant associations but there is reason to believe this one worthy of attention. Klatsby et aI. 10 in the Kaiser-Permanente study found a statistically significant negative association between alcohol consumption and subsequent myocardial infarction in 464 patients. Hennekens et al.," after allowing for covariates, found a negative association between light alcohol intake and fatal myocardial infarction in a large group of cases and controls. Yano et al. 12 found a negative association between moderate alcohol consumption and I.H.D. in a sample of Japanese men. There are other studies13,14 whose results broadly agree with these and, in addition, there is evidence15,16 of a direct effect of alcohol on blood-lipids and other blood components. There is also some evidence 17,18 which conflicts with the hypothesis that alcohol has a protective effect against I.H.D. but nevertheless this hypothesis is at present more strongly supported than refuted in the literature as a whole. Our findings suggest that the apparent relationship between I.H.D. death-rates and alcohol consumption may, be completely explained by wine consumption but we must observe that the experience of Yano et al. 12 suggests that beer may be equally important. If wine has a protective effect against i.H.D. death then this is, in view of our results, more likely to be due to constituents other than alcohol. Wines are rich in aromatic compounds and other trace components which give them their distinct character and it may be to these that we should look for the protective effect. All this is, of course, speculation and wine drinking is said to be related to a relaxed way of living but we firmly believe that our evidence, and that from other studies, justifies an experimental approach to this question. An initial step would be to examine the effect of alcohol and, in
1020
blood-lipids, platelet aggregation, and on such other blood constituents as may plausibly be involved in the pathogenesis of atheroma. The results of this may then justify a randomised controlled trial of the preventive or therapeutic effects of moderate wine consumption on I.H.D. Such a study would, however, pose severe ethical and practical difficulties.
particular wine,
on
If wine is ever found to contain a constituent protective against I.H.D. then we consider it almost a sacrilege that this constituent should be isolated. The medicine is already in a highly palatable form (as every connoisseur will confirm). We can only regret that we are as yet unable to give information to our friends about the relative advantages of red, white or rose wine. Requests for reprints should be addressed
to
REFERENCES common.
81, 294. 11. Hennekens, C. H., Robner, B., Cole, D. Am J. Epidemiol 1978, 107, 196. 12. Yano, K., Rhoads, G. G., Kagan, A. New Engl. J. Med., 1977, 297, 405 13. Berecochea, J. E Report of "Health Consequences of Drinking Practices kind of beverage and subsequent mortality" prepared for the Wine Institute, California. Berkeley, 1978. 14. Stason, W. B., Neff, R. K., Miettmen, O. S., Jick, H. Am. J Epidemol
1976, 104, 603. Myrhed, M., Berglund, L., Bottiger, L. E. Acta med. scand. 1977, 202, 11 16. Castelli, W. P., Gordon, T., Hjortland, M. D., Kagan, A., Doyle, J. T, Hames, C. G., Hulley, S. B., Zukel, W. J. Lancet, 1977, ii, 153. 17. Paul, O., Lepper, M. H., Phelan, W. H., Dupertius, G. W., MacMillan, A, McKean, H., Park, H. Circulation, 1963, 28, 20. 18. Wilhelmsen, L., Wedel, H., Tibblin, G. Circulation, 1973, 43, 950. 15.
A.S.St.L.
1. Cochrane, A. L., St.Leger, A. S., Moore, F. J. Epidemiol. 1978, 32, 200
2. St.Leger, A S., Cochrane, A. L , Moore, F. Lancet, 1978, ii, 1153 3. Aubenque, M., Damiani, P., Deruffe, L. J. Soc. stat. Paris, 1978, 3, 1 4. Produktschap voor Gedistilleerde Dranken, Schiedam, Netherlands, 1975 5. Stamler, J, Stamler, R., Shekelle, R. B. in Ischæmic Heart-Disease edited by J. H. de Haas, H. C. Hemker, and H. A. Snellen); p 84, Leiden, 1970 6. Keys, A., Anderson, J. T., Grande, F. Lancet, 1957, ii, 959. 7. St.Leger, A. S., Sweetnam, P. M. Int. J. Epidemtol. (in the press). 8. Oliver, M. F., Heady, J. A., Morris, J. N., Cooper, J. Br. Heart J. 1978, 40, 1069. 9. Glueck, C. J., Mattson, F., Bierman, E. L. New Engl. J. Med. 1978, 298, 1471. 10. Klatsby, A. L., Friedman, G. D., Siegelaub, A. B. Ann. intern. Med, 1974,
Hlth,
Points of View
professor protects his interests by keeping his sub)ect unto itself. Setting up a school from scratch, is, naturally, a different matter. Razing the existing building and its incumbents was a solution occasionally discussed by the student committee, usually towards the end of the evening when beer was to the fore and ideals the an
IN SEARCH OF THE HOLY GRAIL PETER BARRITT Yeovil District Hospital, Yeovil, Somerset
THE publication of Sir George Pickering’s Quest for Excellence in Medical Education jogged a few memories because I met him on his fact-finding tour when I was a student representative at Birmingham. The memories are none too fond, I must confess. Hurried lunches, interminable meetings, polite silences, impolite silences, feelings of transparency ... all part of the regular life of a student representative on the curriculum review body of the day. My favourite senior physician used to say that if they changed the curriculum much more it would end up the same as when he
trained. The very existence of the student representatives was thorn in the flesh of the distinguished and erudite professors of the faculty who usually had a fairly shrewd idea of what was good for students. In true democratic style we aired our views and they were probably as unrepresentative of student interests as those of our elders and betters. Attempts at introducing democracy were made but they failed because, on the whole, the silent majority didn’t care a stuff about the curriculum as long as they passed through it uneventfully. a
Plans to break down established barriers between preclinical and clinical subjects are doomed to failure. There are two main reasons for this. Firstly, these two sections of the medical school have no idea what each teaches. Clinicians have an inherent interest in maintaining the status quo, since they can then hold forth on how little students seem to learn these days, and so on. Secondly, because funding and the political prestige of a department depend on the amount of teaching time,
island
somewhere behind.
Many of the problems of medical education stem from the enormous increase in the number of students being trained. Medicine, I believe, should be an apprenticeship and it seems sad that students can pass through a medical school without getting to know a few clinicians well. Groups of ten students lead to artificial relationships - the slothful undetected and the loud supreme.
review body’s new curriculum, I would have contained little pxdiatrics and remember, less obstetrics. The argument ran that these subjects were essentially postgraduate and would be learnt during vocational training. Only a teaching hospital could produce a theory like that; for anyone who has worked as a house officer in these specialties in peripheral hospitals would immediately recognise the fallacy of the argument. As an S.H.O. there you are immediately thrown into complex situations, sometimes with only a consultant above you. There is no time to start too much work and not enough learning basics then staff. In academic centres, where house officers are glonfied medical students with keen registrars breathing down their necks, there is no such problem. More important is that many profound medical advances have still not filtered through to some peripheral hospitals. In obstetrics, for example, the active management of labour, fetal monitoring, diabetic control in pregnant. resuscitation of the newborn, and the care of neonates may be mere pipe-dreams in smaller units. If one didn’t learn about these at the centre of excellence, one might never hear of them at all. The
faculty
...
The medical training of the average student consists of his medical education and subsequent experience in non-teaching hospitals. The impact of the vocationalh trained has yet to be felt in general practice, but it will
Public Health
FACTORS ASSOCIATED WITH CARDIAC MORTALITY IN DEVELOPED COUNTRIES WITH PARTICULAR REFERENCE TO THE CONSUMPTION OF WINE A. S. ST. LEGER
A. L. COCHRANE* F. MOORE
Medical Research Council Epidemiology Unit, 3AS
Cardiff CF2
Deaths from ischæmic heart-disease in 18 developed countries are not strongly associated with health-service factors such as doctor and nurse density. There is a negative association with gross national product per capita and a positive but inconsistent association with saturated and monounsaturated fat intake. The principal finding is a strong and specific negative association between ischæmic heart-disease deaths and alcohol consumption. This is shown to be wholly attributable to wine consumption.
Summary
INTRODUCTION
IN a previous study1.2we reported on the associations between age-specific total mortality and a variety of economic, health-service, and dietary variables in 18 developed countries. Our conclusion was that, in general, the health-service factors showed negligible association with mortality, the dietary factors were of little importance, and the only consistent pattern was a negative association between gross national product (G.N.P.) per
capita and mortality. Age-specific total mortality is, however, a crude measure by which to compare countries and so we have embarked upon a comparison of particular diseases. Here we shall present results for heart-disease. *Present address: Rhoose Farm House, Rhoose.
METHODS
The 18 countries used in this study are listed in table i. The criteria for their selection are stated in our previous work.’1 Our mortality data consists of death-rates for males and females separately in the 55-64 age-group from hypertensive disease, ischsemic heart-disease (I.H.D.), cerebro-vascular disease, and bronchitis plus emphysema and asthma; and in the 25-34 age group death-rates from road accidents. These death-rates were obtained from the World Health Statistics Annual for 1970 (codes A82, A83, A85, A93, and AE138, respectively). The latter two causes of death were included in the study for comparison with heart-disease. It has been suggested3 that France underreports i.H.D. deaths and we made allowance for this by analysing our data with France included and excluded. Our health-service, economic, and demographic variables have been described elsewhere,’ as have cigarette consumption per capita per annum and total alcohol consumption. The only additional variables we used in this study were: (1) Consumption (litres per capita per annum) of absolute alcohol attributed separately to beer, wine, and spirits.4 (2) Average daily intake of total calories and carbohydrate, and of total, saturated, monounsaturated, and polyunsaturated fat.5 Fat intake was also expressed as an energy percentage of total calorie intake. These figures were averaged over the period 1954-62 which is reasonable in relation to mortality in 1970. (3) We also obtained daily cholesterol consumption5 which, together with total saturated fats and polyunsaturated fats, we inserted into Keys6 predictive equation for serum-cholesterol. As in our previous study we adopted a pragmatic approach to the statistical analysis. We examined scatter diagrams of one variable plotted against another and product moment correlation coefficients. The variables showing the most convincing associations with the death-rates or with other "explanatory"
TABLE 11-CORRELATION COEFFICIENTS BETWEEN DEATH-RATES AND CERTAIN VARIABLES
1018 variables were selected for examination by regression analysis. No emphasis has been placed on formal significance testing but in the table presenting a summary of the regression analyses we find it helpful to indicate where the t value for inclusion of an "explanatory" variable in the equation exceeds 2. RESULTS
Table n displays the coefficients of correlation between the death-rates and a selection of our "explanatory" variables. In interpreting this table we have conof the coefficients but sidered not only the also their consistency of sign and consistency between the sexes. Doctor density has a negative correlation with the deaths from heart-disease and bronchitis. Nurse density shows a less consistent pattern, there being particularly large positive correlations with deaths from I.H.D. G.N.P. per capita is quite strongly negatively correlated with all the death-rates except road accidents. Population density is mostly negatively correlated with the death-rates and this correlation is fairly large with I.H.D. deaths. Cigarette consumption is positively associated with all causes of death. Total calorie intake is strongly positively associated with I.H.D. and bronchitis deaths. Total fat intake is quite strongly positively associated with I.H.D. mortality and negatively associated with cerebrovascular disease deaths in males. Both saturated fats and monounsaturated fats are strongly positively associated with I.H.D. mortality and not strongly positively associated elsewhere. Polyunsaturated fats are moderately to weakly negatively associated with deaths other than road accidents. Keys’ prediction of average serum-cholesterol is very strongly positively associated with I.H.D. deaths and shows little association elsewhere. Not included in the table are total fat, saturated fat, monounsaturated fat and polyunsaturated fat intake expressed as an energy percentage of total calorie intake. Their pattern of association with I.H.D. deaths is almost identical to that of these fats expressed in absolute
magnitude
terms.
Alcohol consumption is strongly negatively associated with I.H.D. deaths, moderately negatively with bronchitis, and moderately positively with road-accident deaths. The large correlation with I.H.D. deaths prompted us to break alcohol consumption into its three main components-wines, beers, and spirits. As can be seen from table n, wine appears to account for the entire
Relationship between I.H.D. mortality-rate and wine consumption.
men
aged 55-64
alcohol effect and the correlation coefficient with I.H.D. deaths is of the same magnitude as that of Keys’ prediction. The figure shows the association between I.H.D. deaths in males aged 55-64 and the logarithm of wine
consumption (litres per capita per annum). Regression analysis was done with various subsets of the independent variables. The most important findings were:
and nurse density showed relatively little association with the death-rates when other variables were taken into account. (2) G.N.P. per capita was strongly negatively related to mortality and this effect was independent of other variables.
(1) Doctor
(3) Cigarette consumption
was
strongly positively
related to bronchitis mortality and showed appreciable association with hypertensive disease deaths in both sexes and I.H.D. deaths in females. (4) Alcohol consumption was very strongly negatively associated with I.H.D. mortalityin both males and females. Of the three alcohol components wine had the strongest association with I.H.D., its magnitude being the same as that of total alcohol. The wine effect was strong
TABLE III-REGRESSION ANALYSIS: DEATH RATE AS DEPENDENT VARIABLE
* value for inclusion of the variable in the regression set exceeds 2. tthe numbers in the table are standardised regression effects (see text).
in
1019 no matter what other variables were in the regression equation-in other words it was not explained away by G.N.P., cigarettes, or diet. (5) The dietary variables showed less clearcut associations with mortality in the presence of other variables than they did when considered singly. Total fat, saturated fat, and monounsaturated fat were positively associated with I.H.D. mortality when in the presence of
other variables but on the whole these associations were weak. Polyunsaturated fat tended to be strongly negatively related to I.H.D. mortality. Keys’ prediction seemed to carry most of the information present singly or collectively in the fats. The size of its association with I.H.D. mortality depended heavily on the nature of other variables in the regression equation. Excluding France from the analysis did not substantially alter the results. Table III shows some regression analyses using a particular set of independent variables. These variables were chosen for their intrinsic interest (doctors and nurses) or for their importance. The figures in the table are standardised regression effects which have been explained in detail by St. Leger and Sweetnam.7 In essence they represent the proportional change of deathrate to a one standard deviation increase in the independent variable from its mean value. Any two standardised regression effects in the table may be compared, if desired. Most of them are subject to considerable random variation so greater weight should be placed on pattern and consistency than on magnitude alone. DISCUSSION
The results have to be interpreted in the light of the general validity and known pitfalls of this type of study and its method of analysis. This has been discussed elsewhere’ but here we must stress that all our findings, despite the apparent complexity of the analysis, are statistical associations only and are not themselves sufficient evidence to impute causality to any of the observed relationships. Nevertheless these associations deserve further comment and should be viewed within the wider context of other findings. We found negligible association between health service indices-e.g., doctor and nurse density-and mortality from heart-disease or the comparison mortalities from bronchitis and road-traffic accidents. This is in accord with our previous findings regarding overall age-
specific mortality. As before’ increasing per capita wealth appears to have a protective influence not explained by any of the other variables. The association between cigarette consumption and bronchitis is in accord with reasonable expectation and lends support to the validity of our data. The association between cigarette consumption and hypertensive disease and I.H.D. is again roughly in agreement with expectation. The
correlations between I.H.D. death-rate and fats accord with those presented by Stamler et al.5 This is to be expected since our data are similar to theirs. But we find a little surprising their statement that: "It is valid at the present time to conclude that-at a high level of probability-a cause-and-effect relationship has been demonstrated between dietary lipid (specifically, saturated fat and cholesterol) and
dietary
raw
widespread premature coronary heart disease even though substantial direct proof from definitive, largescale long term mass field trials is still to be obtained." Since this was written, large-scale field trials have been completed. The clofibrate trial8 certainly showed a reduction of serum-cholesterol and non-fatal myocardial infarcts in the treated group but there was no difference in I.H.D. mortality and there was an increase in total mortality in that group. Furthermore the evidence from a substantial number of studies9 on the effect of diet, or changes in diet, on the development of I.H.D. is confused and contradictory. The strong negative partial association in our data between I.H.D. death-rate and polyunsaturated fat intake is inconsistent with current opinion because, in the absence of a strong positive partial association between I.H.D. death-rate and saturated fats, this finding is more consistent with the view that polyunsaturated fats are themselves "beneficial" rather than merely apparently beneficial as would be the case were they merely replacing "harmful" saturated fats in the diet. We admit that this argument, being based as it is on association and partial association, must be treated with caution. By far the most interesting result to emerge from our analysis was the strong, specific association between I.H.D. deaths and alcohol consumption, more particularly with wine. This association was not explained away by by fat consumption, Keys’ prediction, or any of the other variables we examined. Correlation studies are bound from time to time to throw up curious but fundamentally insignificant associations but there is reason to believe this one worthy of attention. Klatsby et aI. 10 in the Kaiser-Permanente study found a statistically significant negative association between alcohol consumption and subsequent myocardial infarction in 464 patients. Hennekens et al.," after allowing for covariates, found a negative association between light alcohol intake and fatal myocardial infarction in a large group of cases and controls. Yano et al. 12 found a negative association between moderate alcohol consumption and I.H.D. in a sample of Japanese men. There are other studies13,14 whose results broadly agree with these and, in addition, there is evidence15,16 of a direct effect of alcohol on blood-lipids and other blood components. There is also some evidence 17,18 which conflicts with the hypothesis that alcohol has a protective effect against I.H.D. but nevertheless this hypothesis is at present more strongly supported than refuted in the literature as a whole. Our findings suggest that the apparent relationship between I.H.D. death-rates and alcohol consumption may, be completely explained by wine consumption but we must observe that the experience of Yano et al. 12 suggests that beer may be equally important. If wine has a protective effect against i.H.D. death then this is, in view of our results, more likely to be due to constituents other than alcohol. Wines are rich in aromatic compounds and other trace components which give them their distinct character and it may be to these that we should look for the protective effect. All this is, of course, speculation and wine drinking is said to be related to a relaxed way of living but we firmly believe that our evidence, and that from other studies, justifies an experimental approach to this question. An initial step would be to examine the effect of alcohol and, in
1020
blood-lipids, platelet aggregation, and on such other blood constituents as may plausibly be involved in the pathogenesis of atheroma. The results of this may then justify a randomised controlled trial of the preventive or therapeutic effects of moderate wine consumption on I.H.D. Such a study would, however, pose severe ethical and practical difficulties.
particular wine,
on
If wine is ever found to contain a constituent protective against I.H.D. then we consider it almost a sacrilege that this constituent should be isolated. The medicine is already in a highly palatable form (as every connoisseur will confirm). We can only regret that we are as yet unable to give information to our friends about the relative advantages of red, white or rose wine. Requests for reprints should be addressed
to
REFERENCES common.
81, 294. 11. Hennekens, C. H., Robner, B., Cole, D. Am J. Epidemiol 1978, 107, 196. 12. Yano, K., Rhoads, G. G., Kagan, A. New Engl. J. Med., 1977, 297, 405 13. Berecochea, J. E Report of "Health Consequences of Drinking Practices kind of beverage and subsequent mortality" prepared for the Wine Institute, California. Berkeley, 1978. 14. Stason, W. B., Neff, R. K., Miettmen, O. S., Jick, H. Am. J Epidemol
1976, 104, 603. Myrhed, M., Berglund, L., Bottiger, L. E. Acta med. scand. 1977, 202, 11 16. Castelli, W. P., Gordon, T., Hjortland, M. D., Kagan, A., Doyle, J. T, Hames, C. G., Hulley, S. B., Zukel, W. J. Lancet, 1977, ii, 153. 17. Paul, O., Lepper, M. H., Phelan, W. H., Dupertius, G. W., MacMillan, A, McKean, H., Park, H. Circulation, 1963, 28, 20. 18. Wilhelmsen, L., Wedel, H., Tibblin, G. Circulation, 1973, 43, 950. 15.
A.S.St.L.
1. Cochrane, A. L., St.Leger, A. S., Moore, F. J. Epidemiol. 1978, 32, 200
2. St.Leger, A S., Cochrane, A. L , Moore, F. Lancet, 1978, ii, 1153 3. Aubenque, M., Damiani, P., Deruffe, L. J. Soc. stat. Paris, 1978, 3, 1 4. Produktschap voor Gedistilleerde Dranken, Schiedam, Netherlands, 1975 5. Stamler, J, Stamler, R., Shekelle, R. B. in Ischæmic Heart-Disease edited by J. H. de Haas, H. C. Hemker, and H. A. Snellen); p 84, Leiden, 1970 6. Keys, A., Anderson, J. T., Grande, F. Lancet, 1957, ii, 959. 7. St.Leger, A. S., Sweetnam, P. M. Int. J. Epidemtol. (in the press). 8. Oliver, M. F., Heady, J. A., Morris, J. N., Cooper, J. Br. Heart J. 1978, 40, 1069. 9. Glueck, C. J., Mattson, F., Bierman, E. L. New Engl. J. Med. 1978, 298, 1471. 10. Klatsby, A. L., Friedman, G. D., Siegelaub, A. B. Ann. intern. Med, 1974,
Hlth,
Points of View
professor protects his interests by keeping his sub)ect unto itself. Setting up a school from scratch, is, naturally, a different matter. Razing the existing building and its incumbents was a solution occasionally discussed by the student committee, usually towards the end of the evening when beer was to the fore and ideals the an
IN SEARCH OF THE HOLY GRAIL PETER BARRITT Yeovil District Hospital, Yeovil, Somerset
THE publication of Sir George Pickering’s Quest for Excellence in Medical Education jogged a few memories because I met him on his fact-finding tour when I was a student representative at Birmingham. The memories are none too fond, I must confess. Hurried lunches, interminable meetings, polite silences, impolite silences, feelings of transparency ... all part of the regular life of a student representative on the curriculum review body of the day. My favourite senior physician used to say that if they changed the curriculum much more it would end up the same as when he
trained. The very existence of the student representatives was thorn in the flesh of the distinguished and erudite professors of the faculty who usually had a fairly shrewd idea of what was good for students. In true democratic style we aired our views and they were probably as unrepresentative of student interests as those of our elders and betters. Attempts at introducing democracy were made but they failed because, on the whole, the silent majority didn’t care a stuff about the curriculum as long as they passed through it uneventfully. a
Plans to break down established barriers between preclinical and clinical subjects are doomed to failure. There are two main reasons for this. Firstly, these two sections of the medical school have no idea what each teaches. Clinicians have an inherent interest in maintaining the status quo, since they can then hold forth on how little students seem to learn these days, and so on. Secondly, because funding and the political prestige of a department depend on the amount of teaching time,
island
somewhere behind.
Many of the problems of medical education stem from the enormous increase in the number of students being trained. Medicine, I believe, should be an apprenticeship and it seems sad that students can pass through a medical school without getting to know a few clinicians well. Groups of ten students lead to artificial relationships - the slothful undetected and the loud supreme.
review body’s new curriculum, I would have contained little pxdiatrics and remember, less obstetrics. The argument ran that these subjects were essentially postgraduate and would be learnt during vocational training. Only a teaching hospital could produce a theory like that; for anyone who has worked as a house officer in these specialties in peripheral hospitals would immediately recognise the fallacy of the argument. As an S.H.O. there you are immediately thrown into complex situations, sometimes with only a consultant above you. There is no time to start too much work and not enough learning basics then staff. In academic centres, where house officers are glonfied medical students with keen registrars breathing down their necks, there is no such problem. More important is that many profound medical advances have still not filtered through to some peripheral hospitals. In obstetrics, for example, the active management of labour, fetal monitoring, diabetic control in pregnant. resuscitation of the newborn, and the care of neonates may be mere pipe-dreams in smaller units. If one didn’t learn about these at the centre of excellence, one might never hear of them at all. The
faculty
...
The medical training of the average student consists of his medical education and subsequent experience in non-teaching hospitals. The impact of the vocationalh trained has yet to be felt in general practice, but it will
