Life Style and Life Expectancy1
Hardly a day passes when the public are not given more advice on how to maintain and improve their health and live for a long time. People should eat more fibre but less salt, fat, alcohol and nearly everything else. They should exercise more-but not if the sun is shining on them. Smokers must stop consuming tobacco even if giving up causes stress - which should also be avoided. Et cetera. The avalanche of advice is unending: but what does it mean overall? In principle it should be easy to find out: examine the mountain of evidence that accompanies the avalanche of advice. As might be expected, the practice is more difficult. This chapter is a critical examination of the evidence.
There are basically two kinds of evidence, circumstantial, and controlled trials. Circumstantial evidence can often be quite convincing by itself: The suicide who before your very eyes ingests strychnine and dies at your feet in convulsive agony is quite sufficient evidence to effectively prove the proposition that strychnine is a potent poison even in small quantities. On the other hand, the Prime Minister who assures us that it is the daily drink of his own urine which has preserved his health in his old age, puts up a less convincing case. He may be right, but we are much less inclined to drink urine for our health than we are to shun strychnine for our life.
But suppose we wanted to be very sure that urine did or did not improve health - it would, after all, be a very cheap tonic - how would we go about it? Perhaps we should examine the man concerned with careful medical attention. If it turned out that he had the body of a man 20 years older than himself, with terminal cancer of the pancreas and a heart on its last legs, then in spite of his protestations of good health we might be inclined to proceed no further. But if he had indeed the body of a 20 year old stripling instead, would this be sufficient evidence? No. His preternatural health could be due to any or all of a multitude of his other habits of a lifetime; perhaps the orange he ate religiously every night before going to sleep or his Vegemite on toast every Christmas. (Or none of them: If his centenarian father and grandfather came forward to condemn their offspring's dipsouria, we might suspect a good heredity as the cause of his good health.)
No, to find out as nearly as possible the truth about the benefiial effects of urine we would conduct a controlled trial. Take, say, a thousand healthy volunteers and divide them into two groups which are as similar as possible - age, weight, height, general health, etc. (Healthy individuals are employed even if at risk because of their "unhealthy' life style. The ill constitute a different problem.) The 500 in the FIrst group are then instructed to drink their own urine every morning. (Before volunteering this would have been put to them as a possibility they would have to face.) These people constitute the test group. The other 500 continue their lives as before and constitute a control group.
After, say, 10 years the two groups are compared to determine which contains the healthiest individuals. How is health to be assessed? There are many measures, but perhaps the best is death: the death rates (or the mean ages at death) in the two groups. Other measures are of value and interest but are less convincing. For example, 'feeling well' could be such a measure, but if the first group were all dead within a year, protesting to the last that they "felt well", while the control group were all alive and "feeling no different to before", the death rates would probably deserve more weight than the "feel well" rates.
In this chapter the circumstantial evidence concerning just two factors thought to be connected with health - blood cholesterol and exercise - is examined in some detail. Then controlled trials which have tested the effects of these and other risk factors are critically examined; this forms the greater part of the chapter.
Overall, these controlled trials have been well conducted so that a clear conclusion can be drawn from them: Improvement in lifestyle and reduction of risk factors does not improve health. Struggle though they might, the proponents of a "healthy" lifestyle must face this ineluctable conclusion. Struggle they have. The authors of these controlled trials have tried hard to support their hypotheses. As will be explained, statistical significance levels have been changed, trial endpoints have been changed, unsound statistical tests have been used, and, ultimately some have simply rejected their own findings. In spite of all, the results remain.
Circumstantial Evidence: Cholesterol and Exercise
Is an elevated plasma cholesterol concentration in otherwise healthy people associated with excess mortality and morbidity, particularly that associated with coronary heart disease (CHD)?2 If so, can intervention to reduce plasma cholesterol reduce any such excess? The answer to the first question is by now fairly clear. The answer to the second is the principal subject of this chapter.
Circumstantial evidence has long linked cholesterol and CHD. In 1843, Vogel (cited in Cook, 1958) showed that the atherosclerotic lesion contains cholesterol and this observation was confirmed by later investigations. In 1913, Anitschkow, and Wacker and Hueck (cited in Katz and Stamler, 1953) showed that dietary cholesterol could be important when they induced atherosclerotic lesions in rabbits by feeding them cholesterol in oil.
More recently, epidemiological evidence in the form of prospective studies has appeared. In such studies, a group of people have had their plasma cholesterol measured and they have then been followed for a period of years and correlations between cholesterol levels, mortality and morbidity determined. Most studies have been on men since their substantially higher mortality from CHD produces results more readily (Shurtleff, 1974).
One of the earliest and most in uential of these studies was the Framingham Study which showed a clear positive correlation between plasma cholesterol and CHD (Dawber, Moore and Mann, 1957; Shurtleff, 1974). It was followed by other studies, the results of which are summarised in Table 3.1. It should be borne in mind that Table 3.1 summarises a large amount of information within severe constraints. In particular, where it was found that there was "no increased mortality with increasing plasma cholesterol" the relationship between the two was often very complex none the less. The original papers should be referred to for further information.
Table 3 1
Eight of the studies show a significant and in most cases strong correlation between cholesterol and CHD mortality. None showed a negative correlation. However, the relationship between plasma cholesterol and total mortality is by no means so clear. Because CHD death contributes so much (about one third) to total mortality, a positive correlation between total mortality and cholesterol might have been expected. However, most studies have demonstrated either no correlation or even a negative correlation. Rose and Shipley (1980) have suggested that in some individuals low cholesterol levels are a metabolic consequence of unsuspected cancer. If so, their presence in the study group reduced any tendency to show a positive correlation between cholesterol level and total mortality. There is evidence to oppose this view (Salmond, Beaglehole and Prior, 1985) and the issue is not yet settled.
There are other facts which conflict with the hypothesis that cholesterol and CHD are correlated. At all ages other than 28 to42 men have lower plasma cholesterol levels than women (Adlersberg, Schaefer and Steinberg, 1956; Stenhouse, 1979) yet, at all ages, men have substantially higher CHD mortality than women. The Masai of East Africa have plasma cholesterol levels which are about three standard deviations lower than or about half those of comparable European subjects. Yet their hearts are atherosclerotic and "quite like those described in Western Europe and the U.S." (Mann et al., 1972). The Masai present other problems. Mann et al. found little evidence that the atherosclerotic hearts led to disease and death. Of 600 Masai who were clinically examined, including 350 over the age of 40, only one showed unequivocal electrocardiographic evidence of an infarction. Equally puzzling is the effect of diet on their cholesterol levels. At the age of 12, Masai boys switch from a relatively low-fat diet to a high-fat diet of milk and meat. Their plasma cholesterol levels fall by about 25%. The Masai warrior does not behave like a rabbit. In spite of this discordant evidence the case against cholesterol is sufficiently persuasive to have merited further investigation.
It is fascinating to observe the way medical fashions come and go. As a teenager, Pitt the younger was prescribed a daily bottle of port wine for his gout (Reilly, 1979). Generations later, port was considered a certain cause of gout (for example, Price, 1942: 430). Even tobacco has been considered a sovereign remedy and prophylactic:
Tobacco was then considered an excellent preservative against the plague, which committed dreadful ravages in the reign of Charles the Second; and the Eton boys were ordered to smoke in school daily. Tom Rogers told Hearne "that he was never whipped so much in his life as he was one morning for not smoking". (Lyte, 1899)
As recently as 1942, a distinguished textbook recommended cigarettes as a preventive for asthma (Price, 1942: 1151).
Exercise, particularly strenuous exercise, is now widely regarded as beneficial to health and life expectancy. Yet during the first half of this century the converse was generally believed. Rook (1954) summarises the condition with a characteristically pithy Cambridge story:
"How many regard what they consider to be the folly of undue exertion may be exemplified by the story of the elderly don, himself approaching his hundredth year, who deplored the death of a colleague some three years his junior with the remark that in his youth the dead man had been addicted to climbing mountains, and such exertions must unquestionably have shortened his life."
Rook set out to determine whether exercise did indeed shorten life, an ironic state of affairs given the prevailing views 30 years later.3 Ideally, to answer that question the average age at death of a group of people who have exercised regularly throughout their lives should be compared with that of a matched, similar group who have not. Such an investigation of 70 or 80 years duration is not practicable. Instead, Rook and later investigators assumed that a group of people who are voluntarily and vigorously active in their early life will be likely to continue more active than a similar group who enjoyed a more sedentary youth.
Rook (1954) examined the life expectancy of men who represented Cambridge University in sporting events during the years 1860 to 1900. He compared their mortality with that of honours graduates and that of "non-sportsmen" and "non-intellectuals" of the same period. The intellectuals lived longest with an average age at death of 69.41 years, then the sportsmen at 67.97 years and the random group at 67.43 years. Rook concluded "there was no evidence that the sportsmen died at an earlier age than the group chosen at random; the intellectuals lived longer by a period averaging 1.5 years, but this small difference might well be due to chance". The difference might indeed have been due to chance but Rook, though advised by the Cambridge medical school statistician Dr W.L. Smith, did not bother to apply the usual statistical tests to determine if this was the case. From his Table 4-III it seems that the difference in favour of the intellectuals was not in fact significant. It is important to note that Rook excluded deaths due to accident and war but included suicide. For the intellectuals, 4.7% of deaths were due to accident or war injuries compared with 13.6% for athletes; for the intellectuals 4.1% of deaths were due to suicide compared with 0.9% for athletes. Since the intention of the study was to determine the effects of early strenuous activity on the later condition of the body, this selective inclusion of one form of violent death (suicide) and exclusion of others (accidents and war) seems odd. From the data Rook has provided it is not possible to calculate with certainty the mean ages for non-violent death in the different groups. In particular, Rook does not state explicitly the number of suicides or their age at death. At best, however, the sportsmen lived no longer than the intellectuals and at worst their lives were signi_cantly shorter.
Montoye et al. (1956) carried out a similar study at Michigan State University for students born during the period 1855 to 1919. They defined an athlete to be "a letter winner in a varsity sport". They compared 629 athletes with 563 non-athletes. The mean age at death of athletes was 73.86 years and of non-athletes was 74.24 years. When accidental deaths were excluded athletes lived 74.43 years and non-athletes 74.59 years. These differences of 0.38 and 0.16 years in favour of non-athletes may or may not have been signi_cant but Montoye et al. neither tested for significance nor published sufficient data to allow the readers of their paper to calculate the significance. They simply stated that "there is little difference existing between the two groups". Since their paper is titled "Study of the longevity and morbidity of college athletes" this omission is rather odd. Still odder is the fact that they did test statistically a number of other characteristics, apart from age at death. Their summary ends:
"The results indicated that the longevity of athletes was approximately the same as that of controls and the distribution of causes of death was very similar. A significantly greater percentage of former athletes served in the armed forces and, of those serving a significantly larger proportion served in the Navy and Marine Corps. A significantly greater percentage of former athletes smoked and drank, and their weight in colleges was appreciably greater. There were no significant differences in the number married, weight gain since college days, or strenuousness of activity whilst in the armed Forces.
The commendable zeal with which the authors tested statistically a variety of differences between the two groups is the more striking by comparison with their reluctance to apply the same test to age at death.
Polednak (1972) examined the life expectancy of 6,303 men who attended Harvard College between the years 1880 and 1916. All were sufficiently interested in athletics to have rented a gymnasium locker. He divided them into three groups. "Major" athletes (668 men) had received one or more awards in baseball, football, track, ice hockey, tennis and golf. "Minor" athletes (1501 men) participated in major sports without winning an award or in non-award sports (lacrosse, cricket, basketball, swimming, gymnastics and fencing) or in class sports. Non athletes (4134 men) either had not participated in formal college sports or only as freshmen.
By 1967, 84.5% of the men had died. Polednak summarised his analysis:
"Major athletes (lettermen) had the shortest lives, but differences were small. In mean age at death (natural causes) differences were not statistically significant, but major athletes were consistently the shortest-lived group - by about one to three years in relation to minor athletes and non athletes - in each of three birth decades. Major athletes died significantly earlier than non-athletes from coronary heart disease; they also died more often and earlier from neoplasms, although differences were not statistically significant for this series.
Prout (1972) compared the life expectancy of 172 oarsmen who had rowed during the years 1882 to 1902 at Harvard and Yale with 172 classmates who had not rowed. The mean lifespan of rowers was 67.85 years and of non-rowers 61.55 years, a statistically significant difference in favour of the rowers.
Pekkanan et al. (1987) followed the history of a group of 636 men who were aged 45-64 in 1964. Of these, 260 were classed as highly active physically and 386 were sedentary. After 20 years, 44.6% of low-activity men were dead compared with 46.0% of high-activity men, a small difference in favour of low-activity men which was not significant. However, there was a significant difference in mean age at death, low-activity men dying on average at 67.4 compared with 69.1 for high-activity men. As the authors stated "there is no evidence that high physical activity has the potential for extending the maximum life-span of this male population. Instead a high level of habitual activity reduced premature mortality".
Like Polednak (1972), Paffenberger et al. (1986) examined mortality in Harvard graduates. They sent questionnaires to 16,936 men who had attended Harvard during the period 1916 to 1950 and determined their physical activity patterns. They found consistently that those who exercised more lived longer so that "By the age of 80, the amount of additional life attributable to adequate exercise, as compared with sedentariness, was one to more than two years". As a measure of physical activity they used the subjects' estimate of weekly miles walked, stairs climbed and sports played. From these figures it is possible to estimate the total time expended on exercise during a lifetime. Paffenberger et al. (1986) in their Table 1 imply that the most active subjects were active for about 7 hours a week. Therefore by the age of 80 those subjects had spent about 2 1/2 years being physically active, a period remarkably similar to the improved life expectancy.
Other studies, for example Schnohr (1971) and Karvonen et al. (1974) have compared athletes with the general population. Such work is open to the serious objection that it ignores the "healthy worker effect", defined by Last (1983) as:
"A phenomenon observed initially in studies of occupational diseases. Workers usually exhibit lower overall death rates than the general population, due to the fact that the severely ill and disabled are ordinarily excluded from employment. Death rates in the general population may be inappropriate for comparison if the effect is not taken into account."
To varying degrees the same criticism may be made of those studies which have been described here: those people who are healthiest might be expected to take up sports and athletics. This seems to be re ected in a number of them, where it is found that mortality is higher among non-athletes until middle age after which mortality is lower. For example, Rook (1954) found that at the age of 35, 96.3% of sportsmen were alive compared with 94.8% of intellectuals. Thereafter the reverse was found. At 80, 18.6 % of sportsmen were alive compared with 23.1% of intellectuals.
To summarise, these studies do little to support the hypothesis that exercise increases life expectancy. Perhaps this is not surprising given that the original hypothesis was that exercise decreases life expectancy.
Overall then, the evidence of studies such as those outlined above must be regarded as inconclusive. While it does not support Hippocrates, who wrote of athletes that "no one is in a more risky state of health than they", nor does it exclude the possibility that exercise may be beneficial in increasing life-expectancy. On its own, evidence of the kind so far considered that a particular factor is harmful is no more than circumstantial and suggestive. How then can it be determined whether a fatty diet and high blood cholesterol is bad¯or good¯for you? Whether exercise is good -or bad - for you? Or indeed whether smoking, sunshine, salt or any other factor in everyday life is good or bad? In principle the answer can be obtained very simply by the controlled intervention trial. Take two similar groups of people. Change, so far as possible, the life-style of one group. For example, encourage the group to eat less cholesterol-rich food. Leave the other group to their own devices. After some years compare the two groups. Has the advice worked? Which group has fewer deaths and/or ill subjects?
In practice such experiments are expensive and time-consuming. Thousands of subjects must be followed for many years if sufficient deaths and serious illness are to occur for any difference to become manifest. To make matters worse, the test group may not change their way of life to the extent hoped for¯nor the control group maintain theirs. The difficulties in the conduct and interpretation of such trials are very considerable but within the limits imposed by humanitarian and ethical considerations there seems to be no other approach.
The trials considered here have modi_ed some or all of the following factors in the test group: plasma cholesterol (by diet and/or drugs); smoking; exercise. Tables 3.2a and 3.2b summarise the results of these trials. "Coronary events" is defined by most authors as the sum of both fatal and non-fatal CHD.
The Lipid Research ClinicOs Coronary Primary Prevention Trial (LRC-CPPT)
In 1973, one of the most expensive biological experiments ever devised was begun in the U.S.A. (Lipid Research Clinics Program, 1979). It was intended to test the effect of ingestion of the drug cholestyramine. It was hypothesised that cholestyramine would reduce plasma cholesterol levels and thereby reduce the incidence of CHD. A total of 3806 asymptomatic men with primary hypercholesterolaemia were divided into two groups. The treatment group was given the drug and the control group was given a placebo. The trial continued until 1983, the average time of follow-up being 7.4 years.
The trial was designed to include a number of features outlined in a preliminary report which would make its findings water-tight. "Since the time, magnitude and cost of this study make it unlikely that it could ever be repeated, it was essential to be sure that any observed beneficial effect of cholesterol lowering was a real one" (Lipid Research Clinics Program, 1979). In particular, they chose P<0.01 rather than P<0.054 as the level of statistical signi_cance which would determine whether the drug was bene_cial or harmful. There were to be two primary end points5: CHD death and non-fatal myocardial infarction (MI).
However, when the results of the trial were published these features had vanished. The researchers now accepted as their level of statistical signi_cance "P<0.05, after adjustment for multiple looks" (Lipid Research Clinics Program, 1984). They also adopted as a new primary end point "the combination of definite CHD death and/or definite non-fatal myocardial infarction". With these new criteria they found a significant (P<0.05) 19% reduction in the new primary end point. There was no significant reduction of total mortality or deaths from any specific cause, even with their flexible significance level.
Furthermore, as Pinckney and Smith (1987) have pointed out, the LRC-CPPT committed the statistical solecism of using one-tailed statistical tests which halved their P values. In effect, they assumed that cholestyramine could not possibly increase mortality. This is ironic since in their discussion of deaths from specifc causes they say "The only noteworthy difference (P=0.08) was 11 deaths from accidents and violence in the cholestyramine group, compared with 4 in the placebo group". And in their discussion of side effects they considered the frequency of hospitalisation in the two groups: "The only significant difference was a greater number in the cholestyramine group (40 versus 23) of operations or procedures involving the nervous system". The possible connection between these two statements seems to have eluded the investigators because they state: "Since no plausible connection could be established between cholestyramine treatment and violent or accidental death, it is difficult to conclude that this could be anything but a chance occurrence". The possibility that cholestyramine might reduce life expectancy must be considered in determining the statistical test to be used. This would still be true even if there were not these indications that the drug appeared to have harmful side-effects.
Such rejection of statistically significant results and acceptance of non-significant results is disturbingly common in epidemiological studies. For example, the only randomised, controlled trial of the effects of giving up smoking (Rose et al., 1982) produced only one statistically significant change in mortality. Non-lung cancer deaths increased by 140% (P=0.01) in subjects who gave up smoking. Here also, the authors concluded that this was due to chance, as was their failure to find any reduction in deaths due to lung cancer, or to any other cause, or to all causes. They concluded: "the policy of encouraging smokers to give up the habit should not be changed". This is curious science.
The change in end point, change in significance level, adoption of one-tailed statistical tests and rejection of their own findings do not conceal the fact that the LRC-CPPT project found no beneficial effects on mortality or morbidity as a consequence of cholestyramine-induced reduction of plasma cholesterol. This is an important result in itself and should have been reported. A ten-year study by 214 authors and costing $US150,000,000 deserved better.
The Multiple Risk Factor Intervention Trial (MRFIT)
Over a period of several years, 12,866 men at increased risk of death from CHD were recruited across the USA (Multiple Risk Factor Intervention Trial Research Group, 1982). They had no clinical evidence of CHD but were designated as being at increased risk because their levels of 3 risk factors - cigarette smoking, serum cholesterol and diastolic blood pressure - were sufficiently high to place them in the upper 10% of a risk score distribution based on data from the Framingham study. Half were assigned to a special intervention group (SI) and were treated conventionally with drugs for their hypertension, encouraged to stop smoking and consume less cholesterol and saturated fats. The other men were a "usual care" (UC) group and received no special treatment other than that provided by their personal physicians.
The men were examined annually and followed for an average of 7 years each. The three risk factors declined in both groups but the reductions were larger throughout the trial in the SI group, being significant at P<0.01 at each annual visit. For example, after 6 years, 50% of SI men who were smokers had quit compared with 29% of the UC. Diastolic blood pressure fell in the two groups by 10.5 and 7.3 respectively. Plasma cholesterol fell in the two groups by 12.1 and 7.5 mg/dl respectively, which primarily represented changes in low-density lipoprotein cholesterol and not high-density lipoprotein cholesterol. The unexpected decline in cholesterol in the UC group and a smaller than predicted decline in the SI group meant that the SI-UC difference was about half of that expected. At the end of the follow-up period, the mortality rates were SI 41.2 and UC 40.4/1000. The CHD death rates were 17.9 and 19.3/1000. Neither death rates from CHD nor any other cause were reported as significantly different in the two groups.
The investigators were reluctant to accept their own results and examined sub-groups within their study. They found that SI men with hypertension as an important risk factor and with resting ECG abnormalities had a substantially raised mortality. This, they consider, would have been due to an unexpected deleterious effect of the SI program, particularly toxic effects of the diuretics used. They reported this at a news conference which presented their findings (Kolata, 1982).
After removal of the men with resting ECG abnormalities from their results, total mortality rates for SI and UC men were 37.2 and 39.5/1000 and for CHD 14.8 and 18.8/1000. Again, none of these differences were significant.
The original goal of the MRFIT was to determine whether reduction of the risk factors smoking, cholesterol and elevated blood pressure in high-risk but otherwise healthy men would reduce CHD mortality, non-fatal MI or CHD, cardiovascular mortality and mortality from all causes (Zukel, Paul and Schnaper, 1981). Their paper answers these questions thus:
"In conclusion we have shown that it is possible to apply an intensive long-term intervention program against three coronary risk factors with considerable success in terms of risk factor changes. The overall results do not show a beneficial effect on CHD or total mortality from this multifactor intervention. (Multiple Risk Factor Intervention Trial Research Group, 1982)
In other words, they found that changing the "risk factors" does not apparently change the risks. This necessarily means that the "risk factors" are not as important as was thought. Indeed, it should be concluded that the "risk factors" were no such thing, at least as far as this trial is concerned.
The Oslo Study
From May 1972 to December 1973, all men in Oslo, aged 40-49 years, were invited to take part in a study of health (Leren, Askenvold, Foss et al., 1975). Of 17,965 men examined initially, 1,232 healthy, normotensive (normal blood pressure) men who were considered to be at "high risk" were selected to take part in a controlled intervention trial of the effects of reducing risk factors. "High risk" meant they had serum cholesterol levels of 290-380 mg/dl and "coronary risk scores (based on cholesterol levels, smoking habits and blood pressure) in the upper quartile of the distribution, and systolic blood pressures (mean of two measurements) below 150 mm Hg" (Hjermann et al., 1981).
Half the men (the intervention group) were encouraged to lower their plasma lipid concentrations by change of diet and to stop smoking. The other half (control) were uncounselled. The results of the trial were reported after 5 years (Hjermann et al., 1981) and after 7.5, 8.5 and 10 years (Holme et al., 1985). After 5 years, risk factors had declined in both groups but significantly more in the intervention group (P<0.01). "In the intervention group there was a reduction of 17% of mean serum cholesterol from screening to first follow up. The mean difference in serum cholesterol between the two groups during the 5 years was 13%." And "the mean of the average values of fasting serum triglyceride levels was 20% lower and the mean of non-fasting serum triglyceride levels was 25% in the intervention group than in the control group. Tobacco consumption (expressed as number of cigarettes per man per day; pipe smoking is included taking one pack of pipe tobacco weekly to equal 7 cigarettes daily) fell about 45% more in the intervention group than in the controls". The presumed effects of these changes were reported. Although their original protocol (Leren et al., 1975) had specified P<0.01 as the acceptable level of significance using a t-test, the 5-year report changes the level to P<0.05. With this new level of significance, total mortality was not significantly different in the two groups (intervention group 26/1000, control group 38/1000, P=0.246) but there was a significant difference in sudden coronary death in favour of the intervention group (5/1000 versus 18/1000, P=0.024). Two and a half years after the trial had ended and interviewing with the intervention group had stopped subjects were recalled. The investigators reported (Holme et al., 1985) that risk factors were still different in the two groups, but less so. For cigarette smoking this occurred as a result of the control group continuing to smoke almost as much as before, with the intervention group tending to return to their previous smoking habits. For cholesterol, the control group's values dropped while the intervention group's remained unchanged.
A new endpoint was selected for this extended follow-up period: "major CHD events (non-fatal + fatal MI + sudden death)" and a one-sided test rather than a two-sided test was used. Major CHD events for the 8.5 - 10 year period were "25 and 45 in intervention and control groups, respectively (P ~ 0.02)". Total deaths were 19 and 31 (P ~ 0.05). There was no more detailed presentation of their endpoints.
Given the authors' shifting statistical criteria it is difficult to produce a conclusion as to the benefit of intervention in this trial, but if one must be drawn it is surely that there was no statistically significant benefit. The statements "P ~ 0.02" and "P ~ 0.05" (that is, "the probability is about 0.02" and "the probability is about 0.05") are unusual to say the least.
The WHO Clofibrate Trial
During the years 1964 to 1972, 15,745 men, aged 30-59, were recruited from approximately 30,000 volunteers in Edinburgh, Budapest and Prague to take part in a controlled trial of the effects of lowering serum cholesterol as a result of ingestion of the drug clofibrate. All subjects were healthy and, in particular, free from ischaemic heart disease (IHD). The methods and the initial results were reported by Heady (1973) and by the Committee of Principal Investigators (1978, 1980, 1984).
Subjects were divided into 3 equal groups, according to serum cholesterol level. Group I comprised half those in the high third of the cholesterol distribution of 30,000. Group II comprised the other half of the upper third. Group III comprised half of those in the bottom third. Group I was treated with clofibrate and groups II and III given a placebo, the treatment phase of the project lasting, on average, 5.3 years.
When the first results were reported in 1978, serum cholesterol in Group I had fallen by 9% but Group III continued to have the lowest levels. There was no significant difference between IHD mortality rates in Groups I and II (1.6 and 1.4/1000) but there were significant increases in mortality in the intervention group for all causes (4.9 and 3.8/1000), as well as for a variety of other causes.
By 1980, subjects had been followed for an average of 4.3 years from the end of the intervention period. Total mortality in the intervention Group I continued to be higher (8.1/1000) than in the control Group II (6.6/1000) (P<0.01). Mortality from IHD was not significantly higher in the intervention group (3.2/1000) compared with the control (2.9/1000). By 1984, with a mean follow-up time of 7.9 years, there was no significant difference in total mortality between the two groups (8.6 and 7.9/1000) or IHD mortality (3.6 and 3.5/1000).
The authors did find a consistent and significant reduction in non-fatal MI in the intervention group. Even this positive result was obtained only after the abandonment of their original protocol (Heady, 1973): "The 1% level of significance will be used to assess differences". But, as they concluded in their 1984 paper, "The excess of deaths in the 'treated' group has, not unnaturally, diverted attention from this result".
The Coronary Drug Project (CDP)
From 1965 to 1969, over 8,000 men, aged 30-64, with evidence of one or more myocardial infarctions in the past but who were otherwise healthy, were recruited into an investigation of the efficacy of a number of drugs which reduce plasma lipid levels and are used for the treatment of CHD (Coronary Drug Project Research Group, 1970). They were randomly allocated to six groups who received, respectively, low dose oestrogen, high dose oestrogen, clofibrate, dextrothyroxine, niacin and placebo. The primary endpoint was to be the 5 year total death rate. The results were reported over the next 10 years.
High Dose Oestrogen
After an average of 1.5 years follow-up, the investigators reported their preliminary results for the group who received high doses of oestrogen (Coronary Drug Project Research Group, 1970). The effect on plasma lipids was not reported. It was found that "certain non-fatal adverse affects were occurring substantially more often" in the intervention group than in the placebo group. Breast enlargement, breast tenderness, impotence and testicular atrophy had been expected but, in addition, non-fatal MI was significantly (P<0.05) more frequent (6.2% versus 3.2%) as were pulmonary embolism and thrombophlebitis. Coronary death, sudden death and total mortality were also higher but not significantly so, according to their "two approaches for evaluation of statistical significance" which were specially developed by CHD statisticians. The authors concluded; "These findings lessen the potential long-term value of this 5.0 mg dosage level of oestrogen in men with previous MI". The trial was abandoned in 1970.
The results of the low-dosage oestrogen trial (Coronary Drug Project Research Group, 1973) are confusing. Unlike the earlier paper on the effects of high-dosage oestrogen (Coronary Drug Project Research Group, 1970), "classical" statistical tests as well as their specially developed modified sequential testing and a "Bayesian approach yielding a numerical value designated RBO (relative betting odds)" were used. It is not immediately obvious why such a sophisticated analysis was required or even felt desirable.
With their own tests, after an average follow-up of 56 months, there was no significant difference in mortality or morbidity between the treatment and placebo groups. With the more conventional tests, total mortality was not significantly higher in the treatment group, but pulmonary embolism (1.5% versus 0.8 %), all-sites cancer (1.27% versus 0.47%) and lung cancer (0.54% versus 0.14%) were all significantly higher. Changes in plasma lipids were described cursorily as a slight reduction in serum triglyceride. As a result of these findings, the trial was abandoned in 1973.
After an average follow-up of 3 years, the results of treatment with dextrothyroxine were presented (Coronary Drug Project Research Group, 1972). There was a "fall of about 12% in serum cholesterol level and 15% to 20% in fasting triglyceride level" which are presumably significant changes.
As with the oestrogen trials, the statistical treatment is confusing. Total mortality in the test group was 14.8% and in the placebo group 12.5%, a difference which is not significant by either conventional tests or their specially devised tests, nor were mortalities for specific causes different. There was no reduction in major coronary events or cardiovascular death. However, they go on to say "Further analysis by the life-table method revealed that the absolute difference in mortality between these two groups tended to increase progressively with duration of medication (Figure 1). At the 28th month of follow-up and thereafter these differences exceeded two standard errors of the difference". The investigators concluded that, at best, treatment with dextrothyroxine, although it reduced plasma lipids, did not reduce total mortality or mortality for any cause and was possibly harmful. The trial was abandoned in 1971.
The statistical tests used in this part of the study were conventional (Coronary Drug Project Research Group, 1975). Over the five years follow-up period, the mean reduction in plasma lipids in the treatment group relative to the placebo group was a 6.5% fall in cholesterol and a 22.3% fall in triglyceride. Total death rates were not significantly different in the two groups (20.0% clofibrate, 20.9% placebo). The only significant difference was an increase in definite or suspected fatal or non-fatal pulmonary embolism or thrombophlebitis in the clofibrate group (5.2% versus 3.3%, P<0.01). Several side effects were more common in the treated group. The investigators concluded that they had provided "no evidence on which to recommend the use of clofibrate in the treatment of persons with coronary heart disease."
Statistical tests used in this final part of the study were also conventional (Coronary Drug Project Research Group, 1975). The niacin-treated group experienced a mean decrease in plasma cholesterol of 9.9% and in triglyceride of 26.1%. The 5-year mortality rates were not significantly different in the treated group (21.2%) and placebo group (20.9%), nor were there differences for any specific cause of death. However, the incidence of definite, non-fatal MI was 27% lower in the treated group (P<0.005). As with the clofibrate group, a variety of side effects were more common in the treated group. The authors concluded that niacin did not in uence mortality but it might be slightly beneficial in protecting persons to some degree against recurrent non-fatal MI. "However, because of the excess incidence of arrhythmias, gastrointestinal problems and abnormal chemistry findings in the niacin group, great care and caution must be exercised if this drug is to be used for treatment of persons with coronary heart disease."
The results of the Coronary Drug Project can be summed up briefly: At best, the drugs employed did no harm. Again it must be emphasised that the subjects were essentially in good health in spite of their earlier heart troubles. For the ill, the drugs' worth could well outweigh any harmful side-effects.
The North Karelia Project
Coronary heart disease is exceptionally common in Finland and particularly so in the county of North Karelia. This study (Puska, Tuomilehto, Salonen et al., 1979; Salonen, Puska and Mustaniemi, 1979) aimed at reducing risk factors and, as a consequence, at reducing mortality and morbidity. Five thousand people were recruited in North Karelia as subjects. Seven thousand were recruited as subjects in the control county of Kuopio, selected because of its similarity to North Karelia. The Karelian test subjects were encouraged to reduce their cardiovascular disease risk factors over the period of 1972-7 and their progress was compared with the control subjects.
Initially, high mean values for most risk factors were common in North Karelia. Significantly more cigarettes were consumed per smoker, serum cholesterol was higher and systolic blood pressure was higher. The prevalence of smoking (proportion of the population smoking) was the same. Only diastolic blood pressure was lower. By 1977, all risk factors were the same or, in the case of systolic and diastolic blood pressures, lower in North Karelia than in Kuopio.
In spite of the initial differences in risk factors, the initial total mortalities were the same in test and control areas (test 13.8 versus control 13.6/1000 men) as were cardiovascular mortalities (7.7 versus 7.7/1000 men). Even before the trial started, therefore, a disinterested observer would have had reason to doubt the connection between "risk factors" and actual risks. At the end of the trial, despite the lowered risk factors in North Karelia, total mortalities (11.6 versus 11.4/1000 men) and cardiovascular mortalities (6.3 versus 5.8/1000 men) were still the same. "All" acute MI is described as significantly lower in North Karelia (11.1 versus 12.8/1000 men) but since "definite" acute MI was if anything, higher, such a result in a non-blind trial is very suspect.
The results for women were similar. Initially, risk factors were worse for North Karelian women. Serum cholesterol and systolic blood pressures were higher. By 1977, risk factors were the same (amount of tobacco consumed per day, serum cholesterol) or lower (prevalence of smoking, systolic and distolic blood pressures.)
Again, in spite of the difference in risk factors, initial total mortalities in North Karelia and Kuopio were the same (4.8 versus 5.0/1000), as were cardiovascular mortalities, nor was there any difference at the end of the trial (total mortality 3.9 versus 3.8/1000 women). There was, however a significant reduction in "all" acute MI (2.3 versus 3.8/1000 women). As with the men, this result must necessarily be treated with caution since the North Karelian study was not a blind trial. "Definite" acute MI was not reduced (1.1 versus 0.8/1000 women).
With no change in total mortality or CHD mortality it must be concluded that this trial of the effects of intervention to reduce CHD found no benefit.
The Finnish Businessmen's Study
This was a randomised five-year multifactorial prevention trial of vascular disease (Miettinen, Huttunen, Naukkarinen et al., 1985). An intervention group of 612 forty-eight year old businessmen, considered to be at high risk of cardiovascular disease, were encouraged to change their diet (particularly with regard to fat intake) to reduce smoking and to take more exercise. Where blood pressure and serum lipid levels did not fall sufficiently, the subjects were treated with a variety of drugs, mainly diuretics and beta-blockers and probucol and clofibrate. A similar group of 610 men was uncounselled and untreated except that 15% of them received anti-hypertensive drugs. After 5 years, most risk factors, including weight, blood pressure, serum cholesterol and triglycerides and tobacco consumption had improved signi_cantly (P<0.01) in the intervention group compared with the control group.
At the end of 5 years, total mortality in the intervention group was 10/612 and in the control group 5/610, a non-significant difference. There were no significant differences in mortality from specific causes nor in morbidity except for non-fatal stroke which was more common in the control group (8 versus 0, P<0.01). In brief, improvement in lifestyle did not reduce CHD deaths or total deaths.
The Finnish Mental Hospital Study
This was a controlled study of the effects of a cholesterol lowering diet on 10,000 people in two Finnish mental hospitals (Miettinen, Turpeinen, Karvonen et al., 1972; Turpeinen, Karvonen, Pekkarinen et al., 1979). In Hospital 1, from 1959 to 1965, patients were placed on a cholesterol-lowering diet while Hospital 2 served as a control. Milk was replaced by soybean oil in skim milk and butter and margarine were replaced by a "soft" margarine with a high content of polyunsaturated fatty acids. The overall fat content of the diet was unchanged. In 1965, Hospital 2 patients became the test group and Hospital 1 the control. During the 12 years of the study death rates during hospitalisation were examined.
Biopsy revealed that the fatty acid composition of adipose tissue changed to re ect the change in diet, and serum cholesterol levels fell by 18% and 13% during the trial periods in the 2 hospitals. Pooled results from the 2 hospitals showed that the CHD mortality fell from 14.08 in the control period to 6.61/1000 person years (P<0.002) during the test period. Total mortalities for control and test periods were not significantly different (control 39.50 and test 34.84/1000), nor were any other cause-specific mortalities.
The reductions in serum cholesterol in the two hospitals during the test periods were 12% and 13%. In both test and control periods, serum cholesterol was higher in women than in men. Comparing control and test periods, there was no significant change in total mortality (control 29.01 versus test 30.87/1000), CHD mortality (control 7.90 versus 5.21/1000) or any other cause-specific mortality.
These results are consistent with and support the idea that a cholesterol-lowering diet can reduce the incidence of CHD. This is in fact the only trial to produce a significant reduction in CHD, if only in men.
The WHO Collaborative Trial
In this trial, (WHO European Collaborative Group, 1974, 1982, 1983, 1986) 60,881 working men, aged 40-59 were recruited from 80 factories in Belgium, Italy, Poland and the U.K. Half received an appropriate treatment of advice for a cholesterol lowering diet, cessation of cigarette smoking, daily physical exercise, weight reduction and hypotensive therapy.
Over 6 years, all risk factors fell more in the intervention than in the control group, although serum cholesterol fell only by 1.2%. The estimated change in all risk factors (multiple logistic function, MLF) was -11.1% with a range "from a reduction of 28% in Italy down to 4% in UK and Cracow". (The MLF is a mathematical measure of all the risk factors investigated here.) At the end of the 6 years, the intervention group had 6.9% fewer fatal CHD, 14.8% fewer non-fatal MI and 5.3% fewer total deaths but none of these changes were statistically significant. The authors in their final report (WHO European Collaborative Group, 1986) claimed that "the pooled result was non-significant because the large UK Centre had little success in sustained risk factor control (D mlf -4%), so that its failure to reduce CHD was not surprising". This cannot be the explanation. Not only was there an increase in mortality in the UK, but the same risk factor reduction was observed in Poland with a quite different result: "a large overall benefit to the intervention group". (WHO European Collaborative Group, 1986; cf. tables 2 and 4 in WHO European Collaborative Group, 1983). One cannot have it both ways. If the 22% reduction in Polish CHD is to be credited to the drop in risk factors, then the 14% increase in the UK should also be credited to the same drop in risk factors.
The Anti-Coronary Club Program
In 1957, 814 New York men were recruited to take part in the anti-coronary club program (Christakis, Rinzler, Archer et al., 1966; Rinzler, 1968). They were placed on a diet rich in polyunsaturated fats. A similar group of 463 men formed a control group. After 4 years, the risk factors for CHD¯obesity, hypertension and hypercholesterolaemia¯all diminished signi_cantly in the treated group as compared with the control. Serum cholesterol in the test group fell from 260 to 230 mg/100 ml after 1 year and remained low, that in the control group was unchanged at 250 mg/100 ml.
After 4 years, there were 27 deaths (3.3%) in the test group and 6 (1.3%) in the control group. These results are not stated explicitly and have been extracted from the text (Christakis et al., 1966). There were 9 CHD deaths (1.1%) in the test group and none in the control group (0%). The authors made nothing of these alarming findings and concluded that "a statistically significant difference was observed between the two groups in morbidity from new CHD". Two years later, further results were presented (Rinzler, 1968) but the authors refrained from presenting any mortality figures.
The Veterans Administration Study
Middle-aged and elderly male veterans living in the "Los Angeles domicile" were asked, in 1959, to take part in this study (Dayton, Pearce, Hashimoto et al., 1969; Dayton and Pearce, 1969). After randomisation, 424 men formed a test group and 424 a control group. The geometric mean age was 65.5. Over 8 years, the test group was given a diet which was low in cholesterol and high in unsaturated fats but otherwise similar to the conventional food received by the control group. Serum cholesterol in the test group, relative to the control, fell by a mean value of 2.7% and serum total lipids fell by 6.6%. Biopsy of adipose tissue revealed a 3-fold increase in linoleic acid concentration in test men considered to be "good adherers" to the diet.
After 8 years, there were 177 deaths in the control group and 174 deaths in the experimental group. This result is presented in Dayton et al., (1969, Table 27) and although Dayton and Pearce (1969) give the figures reversed, i.e. 177 test deaths and 174 control deaths, it appears to be correct. The difference in either case was not significant. There were significantly (P<0.05) fewer deaths from acute atherosclerotic events in the test group (48) compared with the control group (70). However, Mann (1977) has analysed their results further and concluded that there were also significantly more deaths from cancer (P<0.05) in the test group (32) compared with the control group (17) so that one difference cancelled out the other. (This is also relevant to the earlier discussion of cholesterol and cancer. See "Cholesterol" section, this chapter.)
In brief, the improved diet produced no change in total mortality or CHD mortality.
In 1970 middle aged men of Goteborg were assigned to either an intervention group or one of two control groups (Wilhelmsen et al., 1986). The participants were 47-55 years of age with a mean of about 51. There were approximately 10,000 men in each group. Men in the intervention group who were found to be at risk because they had elevated serum cholesterol (>300 mg per 100 ml), elevated blood pressure (systolic >175 mm Hg or distolic >115 mm Hg) or smoked more than 15 cigarettes per day were treated and counselled appropriately: cholesterol was lowered by change of diet and where this failed after 6 months "clofibrate and /or nicotinic acid (according to lipoprotein type) was given, but clofibrate was stopped when the adverse effects of this treatment were reported".
Men with high blood pressure were treated with conventional antihypertensive drugs including beta-blockers and diuretics. Smokers were invited to join anti-smoking groups and in some cases were given nicotine chewing gum.
When examined after 4 and 10 years, each of the 3 risk factors was on average lower in the intervention group than in the 2 control groups although the difference was more pronounced at 4 years than 10: risk factors were reduced in the control groups as well as in the intervention group, though less so.
After a mean follow up period of 11.8 years there was "no effect on coronary heart disease, a slight non-significant decrease in fatal stroke, and a slight (far from significant) effect on total mortality". In all there were 1293 deaths in the intervention group and 1304 and 1332 deaths in the two control groups. The authors concluded: "Strategies other than intervention on high risk individuals must be chosen if a major impact on disease incidence is to be achieved in the general population". An unusually frank statement.
Sydney Diet-Heart Study
In 1966, 458 men aged 30D59 with clinical coronary disease took part in a controlled trial of the effects of change in dietary fat intake (Woodhill et al., 1977). A control group of 237 men was given no specific dietary instruction "apart from restriction of calories if thought to be overweight". In addition they were "allowed to use polyunsaturated margarine instead of butter if they wished".
The intervention group comprised 221 men who were "advised and tutored individually to reduce saturated fat intake to approximately 10% of calories and dietary cholesterol to 300 mg or less per day. They were encouraged to use food containing polyunsaturated fatty acids to 15% or more of their daily calories". At follow up, serum cholesterol was significantly lower in the intervention group (250.2 mg/100 ml), than in the control group (262.3 mg/100 ml), a difference of 5%. After 5 years "Survival was significantly better in the P [control] group, although the difference was not marked". In fact, 17.6% of the intervention group died compared with 11.8% in the control group. 94% of the deaths were from cardiovascular disease, 3% from cancer and 3% from motor accidents. The authors concluded that their subjects were "not a good choice for testing the lipid hypothesis" and that weight loss and other changes "may well have more important beneficial effects than changes in dietary lipids". Bearing in mind their mortality results one cannot but admire the authors' mastery of the understatement.
The authors further complained that, because the control group had of their own volition also "improved" their life-style "that the difference in diet between P [control] and F [test] groups was smaller than we had hoped". Given the result of the trial one might reasonably regret, for the sake of all subjects, that there was any change in diet at all.
Helsinki Heart Study
This study was intended to determine the effectiveness of the drug gemfibrozil in preventing coronary heart disease in middle-aged men who had raised non-high-density lipoprotein cholesterol levels (Frick et al., 1987). In a controlled trial, 2051 men were given the drug and a similar group of 2030 were given a placebo over a period of 5 years. This resulted in a sustained change in plasma fats in the treated group: "The total cholesterol level was initially reduced by 11%, the level of LDL cholesterol by 10%, that of non-HDL cholesterol by 14% and that of triglycerides by 43%." There was also an increase in the HDL cholesterol, which is thought to be beneficial.
During the course of the trial the mortality rate in the treated group was 21.9/1000 and in the placebo group 20.7/1000. This was not a statistically significant difference. Nor were there significant differences for any of the specific causes of death. There were significantly fewer (27.3/1000) "cardiac end points" (non-fatal MI, fatal MI, sudden cardiac death and unwitnessed death) in the treated group compared with the control (41.4/1000). Against this must be set the significantly greater number of gastrointestinal operations in the test group 81 versus 53).
Overall then this study found no reduction in either total mortality or CHD mortality after treatment with the cholesterol-lowering drug gemfibrozil.
The Whitehall Study
Smoking is a possible factor which has been tested but once. In 1968, 1445 British civil servants who smoked were recruited to take part in a trial of the effects of giving up smoking (Rose and Hamilton, 1978; Rose et al., 1982). Half were advised and counselled to give up smoking, the other half were not.
After one year reported cigarette consumption in the intervention group was one-quarter of that in the normal-care group. Over 10 years the difference in cigarette consumption between the two groups continued although it diminished.
Cigarette consumption is widely believed to be injurious to health. Just as widely believed is the notion that people who give up smoking become healthier and live longer. No such result was found in this study. After 10 years, 17.2% of the test group had died, as had 17.5 % of the control group. There was no statistically signi_cant change in mortality from CHD, lung cancer or any other cause except for cancer at sites other than the lung where 28 deaths (3.9% occurred) in the group who gave up smoking compared with 12 (1.7%) in those who did not. The authors believed this signi_cant result was due to chance and give their reasons for so believing in one page of their seven page paper. Equally, they believed that their failure to find any reduction in total mortality or mortality from lung cancer or any other specific cause was also due to chance, so they could conclude the summary to their paper with the sentence: "The policy of encouraging smokers to give up the habit should not be changed". I am aware of no other controlled trial of the effects of giving up smoking. Curiously, even this lone study has been omitted from the Surgeon-General's report on passive smoking (Koop, 1986), although he had earlier (Koop, 1982) praised it for "pointing up the positive consequences of cessation in an authoritative manner".
Tables 3.2a and 3.2b (at the end of this chapter) summarise a vast amount of information regarding subjects in intervention studies. The studies constitute over a million subject-years of experience conducted at a cost of billions of dollars. Has intervention to "improve" life-style increased life expectancy? The answer is clear: In not one of the 19 studies is there any beneficial effect. On the contrary, in 3 of them there was a significant reduction in life expectancy. Since the thrust of recent campaigns aimed at changing our life-style is the elimination of premature and unnecessary death, it is evident that such campaigns are founded upon a false premise, or at least fail to take into account the inescapable conclusion of these studies so that the exhortations are quite misleading and pointless. If anything, so-called improvements in our life-style are deleterious, certainly with respect to life expectancy. Could it be however that these subjects felt better even if they lived no longer? Were they "healthier"in a way not directly connected with mortality? Such a question is much harder to answer and is not the prime consideration in these trials, but the likely answer is no. I have two reasons for saying this. First, subjects in studies which involved drugs often experience unpleasant and unwelcome side effects of the drugs, as described earlier. Second, subjects in studies which required changes in diet, smoking and exercise habits may be assumed to have responded like most of us: we may change our diet, smoke less and exercise more if we believe it will improve our overall life expectancy or reduce our mortality from other causes but the change per se is rarely welcome. Dieters and ex-smokers are notorious recidivists. Exercisers may be an exception. Many people claim they require regular exercise for their well-being. This may be the case and indeed it is difficult to see how such a belief could be refuted. The sight of exhausted joggers cluttering lunchtime streets is common enough today in cities around the world. It would be interesting to know how many would persist if the orthodox medical view of the nineteenth century were to revive and they found themselves faced with public health campaigns warning that exercise was shortening their lives.
Even though none of these trials has found increased life expectancy with Oimproved life-styleO, it is worth summarising the resorts to which some of the researchers have been driven to make a case where none really exists:
1. Change of statistical significance level (i.e., P-value)
The Lipid Research ClinicOs Coronary Primary Prevention Trial
The Oslo Study
The WHO Clofibrate Trial
2. Change of trial endpoint
The Lipid Research Clinic's Coronary Primary Prevention Trial
3. Unsound use of a one-tailed statistical test
The Lipid Research Clinic's Coronary Primary Prevention Trial
4. Rejection of significant results
The Lipid Research Clinic's Coronary Primary Prevention Trial
The Whitehall Study
The Anti-Coronary Club Program
Sydney Diet-Heart Study
These exercises in scholarly gymnastics attempt to defend what cannot be proven from their material. This is an odd approach to science.
It needs to be emphasised, once again, that the subjects in the trials discussed here were essentially in good health and the results therefore apply to the healthy. The acutely ill must be considered separately. For example, someone with congenital hypercholesterolaemia may have a plasma cholesterol level many times greater than the population average and be threatened by terminal heart disease while still a teenager. Such a person may well require draconian treatment. That does not mean that the rest of us should also be subject to such treatment or any at all. For the healthy, intervention to "improve" life-style does not reduce the overall death rate nor deaths from particular causes and the only likely consequence is an increase in anxiety and the unnecessary elimination of some of life's pleasure.
1 Chapter 3 is a revised and enlarged con ation of Johnstone (1988) and Johnstone (1989). See also Johnstone (1981, 1982).
2 There is considerable variation in terminology. Coronary heart disease (CHD) atherosclerotic heart disease (ASHD) and ischaemic heart disease (IHD) as well as other terms have been used according to investigator preference. See Dawber (1980) and National Heart Foundation (1967) for a discussion of this point. Serum cholesterol, plasma cholesterol and unspeci_ed or ObloodO cholesterol are terms which have all been used in the literature but the difference between serum and plasma levels is small and the terms are here used interchangeably according to author preference. Plasma levels are about 3% higher than serum (Myant, 1981).
3 RookOs 1954 paper and some of those discussed below are reprinted in Polednak, (1979).
4 P<0.01 means that the probability of being wrong is less than 1 in 100; P<0.05 means it is less than 1 in 20.
5 An endpoint is an awaited event. For example, in a trial to determine whether a treated group will live longer than an untreated group, the end point is death.