The association between tobacco smoking and coronary heart disease

Nearly everything that can be said about smoking and heart disease has been. The literature is now so large and so confusing that there is little hope of ever compressing it into a coherent and consistent summary statement. There is support somewhere for almost every shading of opinion, from the view that smoking is directly, causally related to the disease to the idea that there may be no real association between the two. Weighing the evidence presented does not offer a ready solution to the dilemma, for the weight of evidence may shift very rapidly with changing tides of popularity for one view or another. Careful evaluation of the available data is obviously desirable, but all studies have defects and the relative merits of one over another is subjectively assessed and the assessment is subject to imponderable biases of the evaluation. In a general way, a number of criteria for the evaluation of the causal significance of an association can be stated. The strength of the association. There can be little doubt that the importance of an association increases as its strength increases. Exceptions can be noted without detracting from the general applicability of this rule. The specificity of the association. This rule cannot be applied blindly, either, but together with the first rule determine the precision with which one component of an associated pair can be used to predict the occurrence of the other. This is most important in judging the causal significance of the relation between them. In this connection, prediction of an unlikely event is of greater significance than prediction of a common occurrence. The consistency of the association. This concept is related to the prior ones, but includes the notion that different studies, using different methods, done in different populations should yield similar conclusions. The precedence of the causal component of the association. Within the limitations of methods of measurement the component of the association which is presumed to be causal must precede the other. The coherence of association. Along with all the other requirements, an association of a factor with a disease process should be coherent with the known facts of the biology of that disease. This may be impaired by insufficient or incorrect concepts of pathogenesis but eventually all the information must fit into an ordered structure. The application of these rules may be modified by knowledge of the precision with which the variables may be measured and by whether a factor is considered to be necessary or contributory and sufficient or insufficient. All of these considerations must be involved in the judgement of the importance of smoking as a causative factor in the etiology of coronary heart disease. Observations of different aspects of the co-variation of smoking and CHD are commonplace and, while they testify to the presence of an association between the two, are often couched in such broad generalities that they do not generate much confidence in their reliability. These observations have been summarised in Smoking and Health1 and by Bronte-Stewart.2 Briefly, they are these. During the last half-century both cigarette smoking and CHD have been increasing in a generally similar fashion, at least in Great Britain and in the United States. Males smoke more than females and have the higher rates of CHD. The increasing trend of cigarette smoking began first in males. When the CHD mortality rates were lower than they are at present, there was a definite gradient by social class in Great Britain, with the higher rates associated with higher social status. This may have been related to social class variation in cigarette smoking. There is a fair positive correlation between CHD death rates and cigarette consumption for different countries (see Figure 2 in ref 2).2 All of these arguments are tenuous and to them must be added the suggestion that some differences in CHD between different population groups (such as the difference between urban and rural residents) are not in accord with the smoking differentials. These problems of inconsistency of correlation of this kind are not unique to the problem of smoking and CHD. Given the difficulties of measuring these variables precisely, one may conclude that a real association might be present in just such a vague and inconclusive manner. The requirement for a direct measure of the association is that both variables, smoking and coronary heart disease, be determined in the same population. It should be noted here that the relative frequency of smoking in persons with and without CHD is not the same as the relative frequency of CHD in persons who smoke and those who do not. Fortunately, few of the former kind of estimates are to be found and they need not be given special consideration. One, the report by Buechley et al., 3 based on interviews in the California Health Survey, showed that persons reporting CHD smoked an average of 16 cigarettes per day, while a group of matched comparison subjects smoked an average of 10 cigarettes per day. This kind of result would be expected, given an association between smoking and CHD, but does not provide an appropriate measure of the strength of the association; that is, the magnitude of the excess risk of smokers over non-smokers. An estimate of the excess risk flows quite naturally from the several prospective studies of mortality and morbidity which have been reported. The simplest and most direct measure of the strength of the association between smoking and CHD is some expression of the ratio of the rates for smokers to the rates for non-smokers. A similar expression, used by some authors, is the ratio of observed to expected cases, when the number of expected cases has been computed by applying the rates for non-smokers to the population of smokers. When this is done, using age-, sex-, or other characteristic-specific rates, the resulting statistic, usually based on 100, has been termed a standardized morbidity or mortality ratio. Tables 1, 2, and 3 present the findings of a number of studies from which the strength of the association can be judged. Ratios of mortality rates due to coronary heart disease of male smokers to non-smokers, by age and amount smoked; selected studies * Persons smoking 1 pack/day or more were compared with those smoking less than 1 pack/day (including non-smokers) Ratios of mortality rates due to coronary heart disease of male smokers to non-smokers, by age and amount smoked; selected studies * Persons smoking 1 pack/day or more were compared with those smoking less than 1 pack/day (including non-smokers) Ratios of prevalence of coronary heart disease of male smokers to non-smokers, by age and amount smoked; selected studies * Persons smoking 40 cigarettes/day or more were compared with those smoking less than 40 cigarettes/day (including non-smokers). Ratios of prevalence of coronary heart disease of male smokers to non-smokers, by age and amount smoked; selected studies * Persons smoking 40 cigarettes/day or more were compared with those smoking less than 40 cigarettes/day (including non-smokers). Ratios of incidence rates of coronary heart disease of male smokers to non-smokers,§ by age and amount smoked; selected studies §Non-smokers (strictly) excl. of cigars and/or pipes. *M.I. only ** All C.H.D. *** M.I., and sudden death. + First figure refers to farmers; second to non-farmers; MI and coronary occlusion only. ++ Calculated from table 8 of authors article N.S. rates 3.3/100; S. rates 5.8/100. $ Editorial note: in the original report the reference for this line in Table 3 is the same as for the previous line; both Doyle JT, Dawber TR, Kannel WB, Heslin DS, Kahn HA. Cigarette smoking and coronary heart disease. Combined experience of the Albany and Framingham studies. New Eng J Med. 1962;266:796-801. However, only 2 tables in this paper show data for MI and only one of these, Table 2, shows data for men aged 40-49. In Table 2 the participants are divided into smokers and non-smokers, but not further categorised by smoking status. Our assumption is that these figures were referenced incorrectly in the original. $$ Editorial note: this article is not referenced. It seems likely that it is Paul O, Lepper MH, Phelan WH, Dupertuis GW, Macmillan A, McKean H, Park H. A longitudinal study of coronary heart disease. Circulation. 1963;28:20-31. Ratios of incidence rates of coronary heart disease of male smokers to non-smokers,§ by age and amount smoked; selected studies §Non-smokers (strictly) excl. of cigars and/or pipes. *M.I. only ** All C.H.D. *** M.I., and sudden death. + First figure refers to farmers; second to non-farmers; MI and coronary occlusion only. ++ Calculated from table 8 of authors article N.S. rates 3.3/100; S. rates 5.8/100. $ Editorial note: in the original report the reference for this line in Table 3 is the same as for the previous line; both Doyle JT, Dawber TR, Kannel WB, Heslin DS, Kahn HA. Cigarette smoking and coronary heart disease. Combined experience of the Albany and Framingham studies. New Eng J Med. 1962;266:796-801. However, only 2 tables in this paper show data for MI and only one of these, Table 2, shows data for men aged 40-49. In Table 2 the participants are divided into smokers and non-smokers, but not further categorised by smoking status. Our assumption is that these figures were referenced incorrectly in the original. $$ Editorial note: this article is not referenced. It seems likely that it is Paul O, Lepper MH, Phelan WH, Dupertuis GW, Macmillan A, McKean H, Park H. A longitudinal study of coronary heart disease. Circulation. 1963;28:20-31. Although the Framingham data are not in complete accord, partly because of a considerable variation in the different estimates emanating from that study, there is fairly good evidence to support the following statements: Cigarette smokers have about twice as much coronary heart disease as non-smokers, whether measured by deaths, prevalence, or the incidence of new events. The excess risk of smokers over non-smokers increases with increasing amount of smoking. This finding is more tenuous than the first one, and cannot be asserted firmly in view of the variable results from different studies. The excess risk of smokers over non-smokers tends to decrease with age, so that over 60 years of age there is probably no difference between the two groups. This conclusion is not supported by all studies, but is a bit more firmly established than No. 2. The excess risk of smokers over non-smokers is confirmed largely (perhaps entirely) to myocardial infarction. Data bearing on this point are available in only a few studies and the studies are inconsistent. The specificity of the association must also be considered. Berkson has pointed out very trenchantly the lack of specificity, that in the studies by Doll and Hill, Hammond and Horn, and Dorn, smoking was found to be associated with an increased risk of dying from a variety of diseases. The mortality ratios are summarised in Table 4. From this material, the following alternative conclusions may be drawn: The specificity of the excess risk for all causes where the excess is significant is demonstrated by the failure to find an excess risk for all of the causes. Since an excess risk is demonstrated for a variety of conditions, the excess risk for no single condition can be considered specific. The specificity of the excess risk may be evaluated in terms of relative magnitude for the different causes. Mortality ratios of current cigarette smokers to nonsmokers by specific cause of death; selected studies Mortality ratios of current cigarette smokers to nonsmokers by specific cause of death; selected studies The selection among these alternatives is, in my opinion, a Hobson’s choice. In all the material presented thus far, the consistency of findings in different studies has been remarkable. There is almost complete unanimity that there is an association between smoking and CHD, and disagreement between studies is largely in terms of specifics. This situation would obtain if: There is a real association between smoking and CHD. These studies all incorporated some unknown source of bias which created a spurious association. Prevailing doctrine was established early with subsequent suppression (voluntary, we trust) of contrary findings. Since in epidemiologic circles there are more iconoclasts than icons, the third explanation seems to me unreasonable. Similarly, I cannot envision a pervasive, maleficent bias of the type called for by the second explanation, so until one is presented, I would discard that theory. Thus I am led to believe that the association is real. Whether smoking does or does not precede the development of coronary artery disease is mooted. That it precedes the development of clinically manifest coronary artery disease is established by the prospective studies of morbidity in persons free of CHD insofar as can be determined by clinical examination. Thus in assessing causality, this criterion is not of much value. To summarize up to this point, the following conclusions seem reasonably well established. Smokers are at greater risk of experiencing myocardial infarction than are non-smokers. The excess risk among the middle aged is approximately two-fold. Generally with increasing age the differential diminishes. In seeking an explanation for the increased risk among smokers, four possibilities may be considered. The association is spurious. Smoking causes myocardial infarction. Smoking is causally related to some, one or many precursors of myocardial infarction. Smoking and myocardial infarction are related through some common factor or factors. For the first premise to hold all the studies presented, conducted at different times, in different populations, using different methods, and by different investigators would have to share some undefined, biases leading to a remarkably consistent spurious association. The likelihood of this seems very low. Postulation of a direct causal link between smoking and myocardial infarction suffers from lack of support for any reasonable pathogenetic mechanism. It is conceivable that future research may provide a logical direct link between the two, but for the present, this alternative must be set aside. The third possibility has received considerable attention. Among the factors other than smoking, age, and sex associated with an increased risk of CHD are: High serum lipids Obesity High blood pressure Body habitus Personality type and psychic stress Genetic factors Increased blood coagulability The data bearing on this point are conflicting (Table 1). The weight of evidence seems to favour the idea that serum cholesterol is higher in smokers than in non-smokers and that the differential decreases with age. Acheson and Jessop,4 studying men aged 65 to 85 years, found no association between smoking and serum cholesterol or β/α lipoprotein ratio. Nevertheless, Page5 could demonstrate no acute effect on cholesterol with smoking. From the Framingham data,6 the higher cholesterol of smokers is not sufficient to account for the higher CHD rates of smokers and this would appear to be true for most of the other groups studied. However, Gofman et al.7 computed the expected increase in risk of CHD associated with the higher lipoproteins of smokers to be about 1.4. This they compared with Hammond and Horn’s observed excess risk of 2. This computation calls for great faith in the predictive value of serum lipoprotein determinations. Krut, Perrin, and Bronte-Stewart studied differences in taste perception and food preferences between smokers and non-smokers.8,9 Smokers tended to prefer salty and spiced foods and ate more fat, especially from meat and eggs. The differences were not great but may account, in part, for the observed differences in serum lipids. In contrast to this, Blackburn et al.10 found only a very slight excess of fat intake among smokers. Table 1 presents the data on weight of smokers and non-smokers. There is no explanation here for the serum lipid findings. Brozek and Keys11 reported that persons who stopped smoking gained an average of more than eight pounds when measured 1 and 2 years later, and Blackburn, et al.10, report that this was accompanied by a small rise in cholesterol and a suggestive rise in blood pressure. Generally smokers do not differ significantly from non-smokers by casual blood pressure measurements (Table 1). Thomas12 found no differences by smoking group in the acute effect of smoking one cigarette. Thomas12 classed medical students as nonectomorph, intermediate, and ectomorph according to their ponderal indices. Smokers showed a slight tendency to cluster in the nonectomorph group at the expense of the ectomorph. This was not consistent by degree of smoking and was produced largely by pipe and cigar smokers. Seltzer,13 studying the association of morphologic measures and smoking in Harvard graduates whose physical measures were taken in 1942, found smokers to be taller, heavier, and generally of larger build than non-smokers. From largest men to smallest, the groups ordered as follows: cigar smokers, pipe smokers, cigarette smokers, non-smokers. There was no consistent gradation by amount of cigarette smoking. Blackburn et al.10 presented an “anthropometric index,” defined as the bicristal plus the biacromial diameters divided by the standing height, for middle aged businessmen. There was no difference between smokers and non-smokers. Heath14 presented data on certain psychologic characteristics of the group of Harvard graduates studied by Seltzer. His findings are summarised in Table 2. “A comparison of the case records of the five heaviest smokers … with those of five non-smokers chosen at random showed marked contrasts in personalities. The smokers showed great energy, restlessness, seeking for danger, and a kind of independence which keeps them actively engaged in some enterprise which appeals, and they had difficulties with marriage. The non-smokers were steady, dependable, and hard workers, with stable marriages and histories of specialized, noncombat war duties, and they led rather quiet progressive lives.” From interviews of a probability sample of residents of Buffalo, New York, Lilienfeld15 matched non-smokers with smokers for age, sex, race and social status. The questionnaire included 31 items said to discriminate between “neurotic” and normal persons. Smokers differed from non-smokers to a significant degree on 19 of the 31 on the “neurotic” Smokers were found to more more more more and in more than Thomas12 the medical students on of when with and few marked differences were found between smokers and non-smokers, the former reported more for of questionnaire and less on only 2 or There were no marked differences between the two groups on a and compared smoking of men and classed by type as and not A higher of the group smoked and their average cigarette consumption was much There are four on and smoking. presented data on the greater of smoking in than in and this with a second which showed the same and further showed that among the there was no difference between those and those et also found greater in smoking of over Blackburn et studied the of men and students in two small using a over found no acute effect associated with smoking. and observed increased of in a two following smoking was but not significantly groups of smokers and non-smokers were compared for 8 of and no difference was Although not all of the studies are in accord on all the following conclusions seem Smokers generally have higher serum cholesterol than non-smokers. This difference is not sufficient to account for the differential in CHD rates between smokers and non-smokers. The to this association is not and the association may through the common association of cholesterol and smoking with CHD. Data are insufficient to the association of other serum lipids with smoking. are associated with smoking. There is an factor involved in the of smoking Smokers are more more more less and less well than non-smokers. Thus smokers have been found to differ from non-smokers according to a variety of some of which may be related to the development of CHD. the only in which smoking may be in a causal to CHD is through the effect of smoking on blood by a direct effect of smoking on or by of increased This is but is an for further all of this has taken any are by a between two Smoking is in some involved in the of CHD, or Smoking is associated with CHD in some way, not as a causative with these may with the of The is Since the data are to the we must more These many of and may be to be the for an must to knowledge of CHD. The is the of This be determined by and as well as assessment of the For I believe the second is but I have no notion of the to which this is by factors other than I can this only to my The social is the one, because this must determine and I believe the social must be to the first alternative as a and That this is not must be should the of to determine However, the social that is called for is a to smoking using the not of the and This is a which taken social as an the are