Appendix 1
TOBACCO: A MEDICAL HISTORY
Sir Richard Doll ICRF/MRC/BHF Clinical
Trial Service Unit and Epidemiological Studies Unit, Harkness
Building, Radcliffe Infirmary, Oxford OX2 6HE
INTRODUCTION
At the end of the preface to the splendid history
of smoking that Corti published in Germany in 1930, he wrote that
it would afford him the liveliest pleasure if, after finishing
the book, a reader found himself unable to decide whether the
author was a smoker or not. Admirable though that sentiment is
for a serious historian, I cannot hope to make a similar claim,
so let me say now that I smoked both pipes and cigarettes from
1930 to 1949 and later an occasional cigar (less than one a month)
until 1972, when I learnt that a friend, who smoked many cigars
and had died of a disease that is closely related to smoking,
used to say that it was safe to smoke cigars because I did. Since
then I have not smoked at all. I can, however, say that I was
not antagonistic to tobacco when, in 1947, I began to study its
effects. These are multitudinous, but I shall say little about
most of them, as they are well known. I shall rather concentrate
on the way knowledge developed before the mid 1960s, when smoking
was generally recognised to be seriously harmful to health.
Spread of use of tobacco
To begin at the beginning, we have to go back
some 2,500 years when the peculiar custom of burning tobacco leaves
and inhaling the smoke was adopted by the Mayans in Central America.
At first the leaves were burnt in religious ceremonies and the
priests, who were also physicians, credited the plant with powers
of healing. Later tobacco came to be burnt and the resulting smoke
inhaled for pleasure. The use of tobacco for these purposes spread
north and south and to the Caribbean islands, where leaves were
presented to the Spaniards when they landed at the end of the
15th century. Within a few decades, leaves were brought to Spain
and Portugal, but whether by a Spaniard, Portuguese, or Dutchman
varies with the nationality of the historian. Their use for medical
purposes spread through Europe where tobacco was chewed, taken
nasally as a powder, or applied locally in the treatment of cough,
asthma, headaches, stomach cramps, gout, diseases of women, intestinal
worms, open wounds, and malignant tumours. Although the plant
is now named after Jean Nicot, he did not encounter it until 1559
in Lisbon, where he had been sent on a diplomatic mission. While
there, he became enthused by reports of its healing powers, wrote
about it to the Cardinal of Lorraine, and gave some seeds to a
visiting dignitary from the French Court (Corti, 1931).
Smoking tobacco in pipes became a common habit
only in the last quarter of the 16th century, initially in Britain.
It was introduced by Thomas Harriot, a naturalist and mathematician,
on his return from Virginia, where he had been assigned the task
of investigating anything worthy of note in the new colony, and
it was popularized by his friend Sir Walter Raleigh. Many, however,
thought it a disgusting habit and the use of tobacco in this way
was violently attacked. The opposition was led by James VI of
Scotland, when he succeeded to the throne of the United Kingdom
(as James I) in 1603, and he published a pamphlet against it in
Latin in the same year and anonymously in English, under the title
of "A counterblaste to tobacco" a year later. The pamphlet
was read widely, dutifully praised, and generally ignored. He
tried to persuade Parliament to increase taxation on tobacco but
failed and the main effect of his opposition was to diminish imports
from Virginia and to increase the amount grown at home. By this
time Harriot had died of lip cancer (Körbler, 1952) and Raleigh
might have done so too had he not lost his head for other reasons
in 1618.
Pipe smoking subsequently spread to the Netherlands
at the beginning of the 17th century, where it was recorded in
many old master paintings, and in the succceeding two centuries
it spread throughout Europe and the East. Attempts were made to
ban it in Japan, Russia, Switzerland, and parts of Austria and
Germany, but the prohibition was invariably flouted and control
by taxation came to be preferred.
The use of tobacco as snuff instead of in pipes
became common at the end of the 17th century and cigars, which
had long been smoked in a primitive form in Spain and Portugal,
began to replace snuff a century later. By then cigarettes had
begun to be made in South America and their use had spread to
Spain; but it was not until the Crimean war that they were widely
adopted. Officers returning to Britain made their use fashionable
and by the end of the 19th century cigarettes had begun to replace
cigars. Cigarette consumption increased rapidly in the first world
war, particularly in Britain, and by the end of the second world
war cigarettes had largely replaced all other tobacco products
in most developed countries. By this time, smoking had become
so much the norm, that 80 per cent of middle-aged men in Britain
were regular smokers and some doctors were accustomed to offer
a cigarette to patients, who came to consult them, to put them
at their ease. Women took up smoking in large numbers only later,
at first in the Maori population of New Zealand at the end of
the 19th century and in Britain and the USA in the 1920s, facilitated
in Britain during the second world war when many women began to
work outside the home and had an independent income. In some other
countries, such as France and Spain, women have begun to smoke
only in the last few decades.
ATTITUDE TO
SMOKING IN
THE FIRST
HALF OF
THE 20TH
CENTURY
Anti-tobacco movements
By the beginning of the 20th century, the idea
that tobacco might be beneficial had been largely abandoned, except
in so far as it was thought that nicotine might improve some aspects
of cerebral function. Opposition to tobacco, in contrast, had
been formalised in the activities of societies that sought to
discourage smoking on the grounds that nicotine was addicitive.
Tobacco was consequently classed with alcohol and the anti-tobacco
societies were closely associated with the temperance movement.
These societies had little impact in the UK;
but the idea that smoking stunted the growth of children impressed
the Interdepartmental Committee on Physical Deterioration, which
had been appointed to enquire into the reasons for the poor health
of recruits at the time of the Boer war, and their findings contributed
to the introduction of a law in 1908 prohibiting the sale of tobacco
to children under 16 years of age and empowering the police to
seize cigarettes from any child seen smoking in public (Webster,
1984). In the USA the societies got the sale of tobacco prohibited
in 12 states and, in 1919, the temperance movement got the sale
of alcohol prohibited nationally. The law prohibiting the sale
of alcohol was not, however, respected, and the anti-tobacco movement
lost credibility as a result of the backlash against the temperance
movement. Prohibition of the sale of tobacco was consequently
short-lived and was rescinded in the last state (Kansas) in 1927.
In Germany, the situation was different. Starting
with the formation of the "Deutscher Tabakgegnerverein zum
Schutze für Nichtraucher" (German Assocation against
Tobacco for the Protection of Non-smokers) in 1904, the anti-tobacco
movement had a chequered career until the rise of the National
Socialist party in the 1930s (Proctor, 1996, 1997). Hitler was
personally opposed to the use of tobacco and alcohol, which, he
thought, weakened the national will and harmed the national "germ
plasm". When the party came to power in 1933, elementary
schools were required to discuss the dangers of tobacco, government
pamphlets were published warning people against it, and mass meetings
were addressed by the President of the Reich Health Office and
by Nazi medical leaders, in which tobacco and alcohol were attacked
as reproductive poisons and drains on the economy. The Reich Institute
for Tobacco Research developed tobacco with very low levels of
nicotine but it never captured more than a few per cent of the
market. Beginning in 1938, smoking began to be forbidden in more
and more situations: by uniformed police and SS offficers on duty,
by soldiers in the streets, by young people under 18 years of
age in public, and by anyone in air-raid shelters, city trains,
and buses. In 1941 a special institute was established in Jena
University for the investigation of the hazards of smoking (Wissenschaftliches
Institut zur Erforschung der Tabakgefa). The Institute was directed
by Karl Astel, rector of the University and President of Thuringia's
Office of Racial Affairs, and received an initial grant of 100,000
RM from Hitler's personal office. The campaign against tobacco
does not, however, seem to have had much impact on the public,
and the per caput consumption of tobacco increased annually
after the party came to power, to become 18 per cent higher before
supplies were reduced on the outbreak of war (Nicolaides-Bouman
et al, 1993).
Evidence of harmful effects
The anti-tobacco movements were not, in general,
acting on sound medical evidence of harm, for little such evidence
was available to them. Some evidence had, however, been accumulating
since the end of the 18th century. It was of four types: clinical
observations on patients, comparisons of national trends, studies
of the smoking habits of people with and without different diseases,
and laboratory experiments.
Cancer
Most of the evidence related to cancer. Clinical
observations led Sömmering to write in a prize treatise in
Germany in 1795 that "Carcinoma of the lip is most frequent
when people indulge in tobacco pipes. For the lower lip is particularly
attacked by carcinoma because it is compressed between the pipe
and the teeth". In the next 100 years, pipe smoking, and
specially the smoking of clay pipes, came to be widely accepted
as contributing to the development of cancers of the lip and tongue
and other parts of the mouth (Bouisson, 1858-61; Virchow, 1863-67;
Anon, 1890). In the first half of this century the same cancers
were also found to be characteristically associated with "heavy"
smoking, without reference to method, in cancer clinics in the
USA (Hoffman, 1927; Lombard & Doering, 1928; Potter &
Tulley, 1945). Comparisons were made between patients with different
types of cancer or, in one instance, with life insurance policy
holders (Lombard & Doering, 1928) but without allowing for
differences to age (Table 1). The associations observed were not,
however, taken very seriously and, in so far as pipe-smoking was
thought to be a cause of cancers of the lip and mouth, the risk
was commonly attributed to the heat of the pipe stem rather than
the smoke.
Potter & Tully (1945) also noted the possiblity
of an association between moderate and excessive smoking and cancer
of the respiratory tract (Table 2) which had been considered periodically
since 1898, when Rottmann suggested that a small cluster of cases
of lung cancer in tobacco workers in Leipzig might point to an
occupational hazard, possibly from tobacco dust. At that time,
lung cancer was a rare disease; but it came to be diagnosed progressively
more often over the next five decades and several clinicians and
statisticians in Britain (Tylecote, 1927), Germany (Lickint, 1929;
Fleckseder, 1936), and the USA (Adler, 1912; Hoffman, 1931; Arkin
& Wagner, 1936; Ochsner & DeBakey, 1941) suggested that
cigarette smoking might be a cause, based on the smoking habits
of affected patients and the crude correlation between the increase
in the incidence of the disease and the consumption of cigarettes.
Pathologists, meanwhile, continued to argue
about the reality of the increase. Some, however, had been sufficiently
impressed to try to produce cancer with tobacco tar on the skin
of laboratory animals. Roffo succeeded in doing so in the Argentine
in 1931, using rabbits, but his results were generally dismissed
in the UK and the US on the grounds that the tobacco had been
burnt at unrealistically high temperatures. Experiments in Britain
were negative (Leitch, 1928; Passey, 1929) apart from one which
produced one cancer in 50 animals and led Cooper et al
(1932) to conclude that "tobacco tar is relatively unimportant
in the causation of cancers".
Müller (1939) in Cologne has the credit
for the first case-control study of lung cancer and smoking, even
though the technique he employed was, by modern standards, crude.
Questionnaires were sent to the relatives of people in whom lung
cancer had been diagnosed at autopsy, asking about the subjects'
smoking habits and previous experience to respiratory irritants.
Replies were received relating to 86 returned. Not all the respondents
gave quantitative deails of the amounts smoked and smokers were
classed together in categories based on either quantitative or
qualitative descriptions. The findings for the 86 men are shown
in Table 3, in comparison with those obtained from "the same
number of healthy men of the same ages", but how the healthy
men were selected and how the information was obtained from them
is again not described. The findings, in combination with knowledge
that the use of tobacco had increased five-fold since 1907 and
the results of Roffo's (1937) experiments, led Müller to
conclude that tobacco was an important cause of lung cancer and
the single most important cause of the rising incidence of the
disease. The weakness of the epidemiological method is evident
and the conclusion hardly justified; but the results should certainly
have stimulated research and might have done so in Britain (which,
at that time, had the highest lung cancer rates in the world)
had the war not intervened.
Further research was, however, carried out in
Germany and in the Netherlands. Schairer and Schöniger reported
a case-control study from Astel's Institute in Jena in 1943 and
Wassink reported the results of a Dutch study in 1948. Their findings
are summarized with Müller's in Table 4. The similarity is
impressive. Schairer & Schöniger's work was more convincing
than Müller's, because they gave more details of their methodology
and had an additional control group of men who had died from stomach
cancer. They thought bias was an unlikely explanation of their
findings, that other common explanations for the increase in lung
cancer could be excluded, and that smoking was very likely to
be a cause of the disease.
By 1947, the increase had become so marked in
Britain that the Medical Research Council held a conference to
discuss the reasons for it. Neither of the two German papers was
referred to and Wassink's paper had not then been pubished. The
idea that the increase was due to the increased consumption of
cigarettes, was supported by Kennaway (Cuthbertson, 1968) because
of the probability that the combustion of tobacco would produce
carcinogens and this appealed to Mellanby, then Secretary of the
Medical Research Council, because the mortality from lung cancer
in men was substantially higher in Nottingham, a centre of the
British tobacco industry, than in nearby Leicester (Cuthbertson,
1968). Bradford Hill was consequently asked to carry out a case-control
study to test the idea and the various other hypotheses that had
been suggested. The increasing mortality also led four investigators
to undertake similar studies in the US, the results of which were
all reported together with the British study in 1950 and are described
later.
Vascular Disease
The idea that smoking might be a cause of vascular
disease dates from the end of the last century, when Huchard (1893)
wrote that "The [unfavourable] influences of nicotinism on
the development of arteriosclerosis appears to have been demonstrated,
and this is not surprising since nicotine produces most often
arterial hypertension by vasoconstriction, as the experiments
of Claude Bernard proved." Eleven years later Erb (1904)
found that 25 out of 45 patients with intermittent claudication
were heavy smokers and shortly after that Buerger (1908) noted
that a rare form of peripheral vascular disease that was subsequently
named afer him seldom occurred in non-smokers. Buerger's findings
were repeatedly confirmed in the USA (Weber, 1916; Brown &
Allen, 1925; Allen, Barker & Hines, 1946) and Silbert (1935)
who reported a large series of cases from New York, stated that
he had never seen a case in a non-smoker. Others, however, said
that they had (Christian, 1947) and this, I was told when a medical
student, showed that smoking was not the cause.
Coronary thrombosis was not diagnosed in life
until Herrick diagnosed it in 1912. Subsequently it was reported
progressively more often every year. The correlation between the
increasing number of reports and the increasing consumption of
cigarettes led Hoffman, the American statistician, to suggest,
as early as 1931, that smoking might be responsible for many cases.
Several clinical studies of the relationship with smoking were
published, but the findings were confused and no substantial evidence
was obtained until 1940 when English, Willius & Berkson reported
an association in the records of the Mayo clinic. They compared
the recorded habits of 1,000 patients with the disease with those
of 1,000 other patients matched for sex and for age in three broad
groups, as is shown in Table 5, and subsequently the frequency
of the diagnosis of coronary disease in 1,000 consecutive smokers
with that in 1,000 similarly matched non-smokers, as is shown
in Table 6. The results led them to conclude that the smoking
of tobacco probably had (I quote) "a more profound effect
on younger individuals owing to the existence of relatively normal
cardiovascular systems, influencing perhaps the earlier development
of coronary disease." They eschewed reference to causation,
because the subject would be controversial and that (I quote)
"physicians are not yet ready to agree on this important
subject" (Willius, 1940).
Other conditions
Other conditions included a characteristic type
of blindness, tobacco amblyopia, which was described by Beer in
1817. It occurred principally in heavy pipe smokers in association
with malnutrition and was probably caused by the cyanide in smoke,
when the ability to detoxify it was reduced by deficiency of vitamin
B12 (Heaton et al., 1958; Freeman, 1988). The disease has
been much less common in cigarette smokers and is now extremely
rare.
Peptic ulcers were commonly thought to be aggravated
by smoking, possibly as a result of the action of nicotine on
gastric motility, but the physiological evidence for this and
other mechanisms was inconsistent and never wholly convincing.
Extraordinarily, there was seldom reference
to smoking as a cause of respiratory disease, except by Lickint
(1939) in Germany. In Britain the cough that was so prevalent
in smokers was dismissed as a benign "smokers' cough".
In retrospect, the most important evidence of
the harmful effects of smoking was Pearl's observation in 1938
from a study of the family history records collected at the Johns
Hopkins School of Hygiene and Public Health that (I quote) "The
smoking of tobacco was statistically associated with the impairment
of life duration and the amount or degree of this impairment increased
as the habitual amount of smoking increased." Pearl's unwelcome
finding, shown in Fig 1, was either ignored or dismissed as due
to confounding with some hypothetical other feature.
Medical teaching
Despite the accumulating evidence, academic
departments in general paid little or no attention to smoking
and references to it in medical and surgical textbooks during
and shortly after the war were scarce and brief. In the UK and
the US most mentioned smoking in relation to Buerger's disease
and cancers of the lip and tongue. A few mentioned tobacco amblyopia
and some said that excessive smoking aggravated peptic ulcers
and should be stopped in the treatment of angina. None mentioned
it in relation to coronary thrombosis or cancer of the lung.
More attention was paid to smoking in Germany,
which had been the leading country for medical research. The misuse
of tobacco was sometimes said to cause chronic nicotine poisoning,
with effects in nearly every system in the body (Dening, 1950;
Hoff, 1948).
It was mentioned as contributing to cancers
of the mouth (Lindemann & Lorenz, 1949), tongue and larynx
(Stich & Bauer, 1949) but not in relation to cancer of the
lung, except by Bauer (1949) in his textbook on cancer, who thought
that tobacco might cause a precancerous condition in the bronchi
that other agents converted into cancer.
THE 1950 WATERSHED
In 1950, the situation was radically changed
by the report of the five case-control studies previously referred
to. They differed from the early German studies in that many more
patients were included, the possibility of sustantial bias due
to low response rates was avoided, and much more information was
obtained about past smoking habits, including the method and amount
of smoking and the ages at which smoking had been started and
stopped. Outline results, similar to those shown for the three
pre-1950 studies, are shown in Table 7. All showed a close association
with smoking.
Two studies stood out because of their size,
the precision with which lifelong non-smokers were defined, and
the argument that led to their conclusion. One had been initiated
by Ernst Wynder in 1948, while a summer student at New York University,
on the basis of knowledge (I quote) "that the burning of
tobacco in pipes or as cigars or cigarettes, would lead to the
formation of cancer-causing chemical compounds" (Wynder,
1988). The results he obtained from interviewing 20 patients so
impressed Evarts Graham, the Chief of Surgery at Washington University
School of Medicine, that the study was continued in his surgical
service and a grant for expansion obtained from the American Cancer
Society in the spring of 1949. Analysis led to the conclusion
that "Excessive and prolonged use of tobacco, especially
of cigarettes, seems to be an important factor in the induction
of bronchogenic cancer." (Wynder & Graham, 1950). In
the other, which had been initiated by the British Medical Research
Council's conference in 1947, detailed consideration of the possibility
of confounding, the consistency of the findings in different studies,
the biological relationships with amount and duration of smoking,
the size of the estimated relative risk, and the relationships
over time and place and for each sex led the authors to conclude
that (I quote) "cigarette smoking is a factor, and an important
factor, in the production of carcinoma of the lung" (Doll
& Hill, 1950).
Reaction to findings
This conclusion was accepted by Sir Harold Himsworth,
who had become secretary of the Medical Research Council, but
not generally by medical or statistical scientists and certainly
not by the British Department of Health's Standing Advisory Committee
on Cancer and Radiotherapy (Webster, 1984). Most accepted that
an association had been shown, but not that it implied cause and
effect. Some, however, were even more sceptical, including Berkson
(1955) co-author of the 1940 study of heart disease and the leading
American medical statistician who suggested that the findings
were an artefact due to the combination of lung cancer and smoking
leading to a greater chance of a patient's admission to hospital
than when the disease occurred in a non-smoker. Other sceptics
were the representatives of the tobacco industry, who, in Britain,
sought an interview with the Medical Research Council and were
referred to Professor Hill. The conclusion that cigarette smoking
was a cause of the disease was, they argued, unsustainable for
three reasons: the international correlation between cigarette
consumption and the mortality from lung cancer of about 0.5 was
too low, smoking histories were too unreliable to use as a basis
for an association with disease, and lung cancer, in any case,
was obviously due to atmospheric pollution. To this Hill replied
that a correlation of the size observed with crude international
statistics was, in his experience, unusually high and supported
a causal relationship rather than the reverse: that if smoking
histories were unreliable, this would have weakened a true association
rather than have created a false one; and if they thought that
atmospheric pollution was the main cause of lung cancer they should
go away and prove it, for Hill and I couldn't.
THE
EVIDENCE THAT
LED TO
WIDE ACCEPTANCE
OF MAJOR
HARM FROM
SMOKING
The early cohort studies
Evidence of a different type was, however, clearly
needed, if reactions were to be changed, as, for example, by recording
the smoking habits of large numbers of people and following them
up to see if the risk of lung cancer could be predicted from the
information about the individual's level of smoking. According
to Wynne Griffith (personal communication) the idea that doctors
would make a suitable population to study came to Bradford Hill
one Sunday morning when playing golf and, Wynne Griffith added,
"I don't know what kind of a golfer he (is) but that was
a stroke of genius." It was indeed, for when we wrote to
all the doctors on the British Medical Register in October 1951,
over 40,000 (two-thirds) gave details of their smoking habits
and they have proved so easy to trace that nearly all the men
who were not known to have died could be traced 40 years later
(Doll et al, 1994). The story, however, is apocryphal;
for Sir Austin told me that the idea came to him, in the classical
manner, in his bath.
The evidence from the "cohort" study
of British doctors mounted quickly, and within two and a half
years the findings with regard to lung cancer had confirmed those
predicted from the case-control studies. This is shown in Table
8, which gives the relative mortality rates for different levels
of smoking, as estimated from the final results of the British
case-control study based on 1,357 deaths from lung cancer in men
(Doll & Hill, 1952), and the first results of the cohort study
based on only 36 such deaths (Doll & Hill, 1954). With so
few deaths in the second study, the confidence limits of the mortality
rates were wide, but even so the trend in mortality with smoking
was significant (P<0.01).
Altogether, however, 789 deaths had been recorded
and it was possible to examine the relationship between smoking
and several other diseases. With 235 deaths attributed to coronary
thrombosis the mortality (standardised for age) increased from
3.9 per 1,000 men per year in lifelong non-smokers to 5.2 per
1,000 in men smoking an average of 25g of tobacco or more per
day. The increase was small but the trend with the amount smoked
was statistically significant and it was concluded that there
was a subgroup of cases in which (I quote) "tobacco has a
significant adjuvant effect".
Two years later these results were confirmed
with larger numbers (Doll & Hill, 1956). More importantly,
they were also confirmed in the much larger study that the American
Cancer Society had started in 1952 specifically, as the principal
investigator told me, to disprove the relationship between smoking
and lung cancer that had been observed in the case-control studies
(Hammond, personal communication). The results, based on nearly
5,000 deaths in the 190,000 American men followed for two years,
are shown in Table 9 for lung cancer and, in four age groups,
for coronary disease (Hammond & Horn, 1954). The investigators
were impressed by the correlations between cigarette smoking and
the mortality from coronary thrombosis in men and women, in urban
and rural areas, and over time, and also by the previous reports
that cigarette smoking caused vasoconstriction and increased heart
rate and blood pressure and they concluded that (I quote) "regular
cigarette smoking causes an increase in death rates from these
two diseases" (that is, from coronary thrombosis and cancer
of the lung) adding that "Probably nicotine is at least partially
responsible for the findings" in relation to the former.
Proof of causation
The conclusion that cigarette smoking was a
major cause of disease had not been easy to accept, as the evidence
was observational in humans and unconfirmed by animal experiment.
Two leading statisticians, moreover, remained unconvinced. In
the USA, Berkson (1958) was disturbed that the relationship with
smoking held to some extent across the board with a variety of
conditions: in 12 of the 15 categories of cause of death for which
data were given in the British study (Doll & Hill, 1956) and
in all the nine categories examined in the American study (Hammond
& Horn, 1957). In Berkson's opinion this raised the suspicion
that there must be something wrong with the method of enquiry
and he suggested that they were the result of the interplay of
various subtle and complicated biases or that they had a constitutional
basis, people who were non- or relatively light smokers, being
the kind who were biologically self-protective and that this (I
quote) "correlated with robustness in meeting mortal stress
from disease generally."
In making this criticism, Berkson (1958) took
no account of the great difference in the relative risks of different
diseases among heavy cigarette smokers compared to non-smokers,
varying in Doll & Hill's (1956) study from 24 to one for lung
cancer to 1.01 to one, not of the fact that tobacco smoke was
not a pure chemical entity, but a mixture of many chemicals, subsequently
shown to number more than 4,000. It was, as Hill (1966) pointed
out, as if he had said that milk could not be a cause of any disease
because it spread tuberculosis, diphtheria, scarlet fever, undulant
fever, dysentery, and typhoid and, he might have added, contributed
to the production of vascular disease and prevented osteoporosis.
In the UK, Fisher, the most eminent theoretical
statistician worldwide, was disturbed that the original finding
(Doll & Hill, 1950) that smokers with lung cancer reported
inhaling less often than smokers without the disease (62 per cent
against 67 per cent) weighed against causation, unless it were
also concluded that (I quote) "inhaling cigarette smoke was
a practice of considerable prophylactic value in preventing the
disease" (Fisher, 1958a) and he argued that secular changes
in smoking habits could not be related to the increase in lung
cancer since "lung cancer has been increasing more rapidly
in men relatively to women" and that "it is notorious,
and conspicuous in the memory of most of us, that over the last
50 years the increase of smoking among women has been great, and
that among men (even if positive) certainly small" (Fisher,
1957).
Neither objection was valid. The effect of inhaling
was impossible to predict without knowing where the smoke droplets
would be deposited and this was uncertain because tobacco aerosols
swell under warm and moist conditions and might, if inhaled deeply,
deposit in the alveoli rather than on the bronchi (Davies, 1949).
Doll & Hill (1952), moreover, found that while inhaling was
associated with a diminished risk of cancer in the large bronchi,
it was associated with an increased risk of developing cancer
in the periphery of the lung, which made biological sense. As
for the evidence of secular changes, Fisher (1957) was just wrong;
for he had ignored the cohort effects whereby the risks among
successive cohorts are determined not only by their recent smoking
history but also by their smoking habits in the distant past.
When comparisons are made at appropriate ages and times, the trends
in the sex ratio of the disease mimic the trends in cigarette
consumption by sex over the relevant periods (Doll & Peto,
1981, Appendix E).
Difficulty in reaching a conclusion about a
causal interpretation of the evidence also arose, because different
people gave different meanings to "cause". In saying
that a particular factor is a cause of disease, epidemiologists
have in mind a situation in which, for example, prolonged cigarette
smoking results in a rare disease becoming 10 times as common
as it would have been in the absence of smoking. Cigarette smoking
is not then a necessary cause nor a sufficient cause; but it can
be an important cause (as few people would have developed the
disease if they had not smoked) and this is not contingent on
the absence of other causes. What was claimed was that for several
diseases causation in the sense described was proved beyond reasonable
doubt. The detailed evidence that led to this claim has been reviewed
many times and I note here only the extraordinary strength of
the association with lung cancer, with increased risks of more
than 20 fold in heavy cigarette smokers which alone made the alternative
explanation of confounding virtually impossible, the diminution
of risk with cessation of smoking, and the consistency of the
findings with different methods of investigation and in different
countries and different cultures.
During the 1950s, this epidemiological evidence,
which had been supplemented by many other studies, was supported
by the experimental demonstration that tobacco tars were carcinogenic
when applied regularly for a long time to the skin of laboratory
animals (Engelbreth-Holm & Ahlmann, 1957; Guérin &
Cuzin, 1957; Sugiura, 1956; Wynder et al, 1953, 1955, 1958)
and by the identification of known carcinogens in tobacco smoke
(Cooper & Lindsay, 1955; van Duuren, 1958). Expert committees
appointed to review the evidence were consequently able to reach
positive conclusions. Between 1956 and 1959, the Netherlands Ministry
of Social Affairs and Public Health (1957), the British Medical
Research Council (1957), a study group appointed jointly by the
US National Cancer Institute, the National Heart Institute, and
the American Cancer Society (Study Group on Smoking and Health,
1957), the Swedish Medical Research Council (1958), and the US
Public Health Service (Burney, 1959) all reported that cigarette
smoking was a cause of lung cancer, and a year later an expert
committee of the World Health Organization (1960) did so too.
Public acceptance of causality
Despite their provenance these reports had little
lasting impact on the general public and the situation did not
change materially until after the reports by the Royal College
of Physicians of London in 1962 and the Advisory Committee to
the US Surgeon General in 1964. The first was short and aimed
at interested laymen. The second was long and detailed and was
particularly newsworthy, because the tobacco industry had been
privileged to veto any member of the Committee who had publicly
expressed any views about the subject. Both reports nevertheless
agreed that smoking was a major cause of lung cancer. The Surgeon
General's Committee was also clear that it was a major cause of
chronic bronchitis. Both, however, were cautious about the meaning
of the relationship of smoking to the many other diseases associated
with it.
Following these reports, the idea that smoking
was a major cause of lung cancer ceased to be seriously challenged.
Even the tobacco industry in the UK agreed not to deny the causal
relationship on the advice of Geoffrey Todd, their senior statistician.
Todd had been a representative of the industry who had visited
Doll and Hill in 1952 and had sought to persuade them that their
conclusion was wrong; but he had become convinced that it was
right. In the USA, however, the industry continued to maintain
that all that had been shown was a statistical association and
the causality had not been scientifically proven: that is, until
recently when the smallest manufacturer broke ranks and accepted
that smoking was a cause of the disease.
CURRENT KNOWLEDGE
OF EFFECTS
In the three subsequent decades, cigarette smoking
has been found to be positively associated with nearly 50 diseases
or causes of death and to be negatively associated with eight
or nine. In a few instances the associations are due to confounding
with other factors, but the great majority arise because tobacco
smoke is a contributory cause. Pace Berkson, this is not
surprising; not only because of the complexity of tobacco smoke,
but also because many of the diseases are different clinical manifestations
of common processes, such as DNA damage, vascular occlusion, and
damage to small airways. Most of the associations have been demonstrated
in cohort studies, which have now also been carried out in Canada
(Best et al, 1961), China (Chen et al, 1997), Japan
(Akiba & Hirayama, 1990), Norway (Lund & Zeiner-Heinricksen,
1981), and Sweden (Cederlöf et al, 1975) as well as
in Britain and the USA and have been extended to cover the last
two decades, when most smokers have been smoking cigarettes for
nearly all their smoking lives (Doll et al, 1994; Thun
et al, 1995). Other associations have been demonstrated
in case-control studies.
Harmful effects
The morbid effects that are caused in part by
cigarette smoking are listed in Tables 10-13. Those that are five
or more times more common in cigarette smokers than in non-smokers
are marked with an asterisk. To these have to be added a proportion
of the deaths from the causes shown in Table 14 that are principally,
or in some instances wholly, associated with smoking through confounding
with other aetiological agents, and possibly a small proportion
of childhood cancers due to mutations in paternal sperm (Sorahan
et al, 1997).
Beneficial effects
Finally, there are eight or nine diseases that
may be alleviated or prevented by some of the chemicals in tobacco
smoke. These are shown in Table 15. Most are uncommon or seldom
fatal and the combined impact on mortality of their reduction
in incidence as a result of smoking is less than 1 per cent of
the increase in mortality caused by smoking. Whether Alzheimer's
disease is, in fact, inversely related to smoking is uncertain.
It has appeared to be so in case-control studies but not in a
cohort study (Ott et al, 1998) and the inverse relationship
may be an artefact due to studying prevalent cases rather than
incident cases. In our study of British doctors, the mortality
attributed to dementia as a whole was slightly higher in smokers
than in lifelong non-smokers (Doll et al, 1994) possibly
due to an increased risk of vascular dementia in smokers.
Total effect on risk of death
In sum, the total effect of cigarette smoking
appears to double the risk of death in middle and old age in both
sexes. Some six per cent of the excess mortality in men is, however,
due to diseases that were listed in Table 14 as caused by factors
with which smoking is confounded and this might be thought to
reduce the risk that the avoidance of smoking could avoid. In
fact, it does not, for confounding can operate in both directions
and confounding with the consumption of alcohol reduces the effect
of smoking because alcohol reduces the risk of vascular disease
and this, in developed countries, is the principal cause of death.
This more than compensates for the attribution to smoking of the
excess mortality from other causes with which smoking is associated
through confounding (such as cirrhosis of the liver and accidents)
and the estimate that prolonged cigarette smoking causes the risk
of death to be doubled is likely to be too small rather than too
great. On the assumption that it doubles the risk it will cause
one regular cigarette smoker in four to die because of his smoking
habit under 70 years of age, losing on average 20 years expectation
of life, and one in four to die later, losing on average eight
years. ENVOI
In retrospect, it may be surprising that resistance
to the idea that smoking caused so much disease was initially
so strong. Three factors, at least, contributed to it. One was
the ubiquity of the habit, which was as entrenched among male
doctors and scientists as among other men and had dulled the sense
that tobacco might be a major threat to health. Another was the
novelty of the epidemiological techniques, which had not previously
been applied to any important extent to the study of non-infectious
disease. The findings were consequently undervalued as a source
of scientific evidence. A third was the primacy given to Koch's
postulates for determining causation. The evidence that lung cancer
occurred in non-smokers was consequently taken to show that smoking
could not be the cause and the possibility that it might be a
cause was inappropriately doubted. The manner in which lung cancer
was linked to smoking was not, however, unique. All the other
major diseases related to smoking were found to be so by epidemiological
enquiry and laboratory evidence of physiological effects that
provided plausible mechanisms by which smoking might cause them
was obtained only later and, in some instances, is still awaited.
All the diseases related to smoking that cause
large numbers of deaths should by now have been discovered, but
further effects like age-related macular degeneration that was
firmly linked to smoking only two years ago (Christen et al,
1996; Seddon et al, 1996; Vingerling et al, 1996),
may well be revealed by cohort studies that are able to link individuals'
morbidity data with their personal characteristics through personal
identity numbers. That so many diseases—major and minor—should
be related to smoking is one of the most astonishing findings
of medical research in this century; less astonishing perhaps
than the fact that so many people have ignored it—but then
perhaps they don't enjoy life as much as I do.
|
