Jump to: Page Content, Site Navigation, Site Search,
You are seeing this message because your web browser does not support basic web standards. Find out more about why this message is appearing and what you can do to make your experience on this site better.
Richard Peto a Clinical Trial Service Unit and
Epidemiological Studies Unit (CTSU), Radcliffe Infirmary, Oxford OX2
6HE, b Cancer Epidemiology Unit, Radcliffe Infirmary, Oxford
OX2 6HE, c Trent
Institute for Health Services Research, Queen's Medical Centre,
Nottingham NG7 2UH, d Department of Social Medicine, University of Bristol, Bristol BS8 2PR
Correspondence to: S Darby
sarah.darby{at}ctsu.ox.ac.uk
| |
Abstract |
|---|
|
|
|---|
Objective and design:
To relate UK national trends
since 1950 in smoking, in smoking cessation, and in lung cancer to the
contrasting results from two large case-control studies centred around
1950 and 1990.
Medical evidence of the harm done by smoking has been accumulating
for 200 years, at first in relation to cancers of the lip and mouth,
and then in relation to vascular disease and lung cancer.1 The evidence was generally ignored until five case-control studies relating smoking, particularly of cigarettes, to the development of
lung cancer were published in 1950, one in the United
Kingdom2 and four in the United States.3-6
Cigarette smoking had become common in the United Kingdom, firstly
among men and then among women, during the first half of the 20th
century. By 1950 lung cancer rates among men in the United Kingdom had
already been rising steeply for many years, but the relevance of
smoking was largely unsuspected.
2 7
At that time about
80% of men and 40% of women smoked (fig 1 and BMJ's
website, table A). But few of the older smokers had smoked substantial
numbers of cigarettes throughout their adult life, so even male lung
cancer rates were still far from their maximum (except in younger men),
and rates in women were much lower. Over the next few decades, a
substantial decrease occurred in the United Kingdom in the prevalence
of smoking (fig 1), in cigarette tar yields, and, eventually, in lung
cancer rates (fig 2), and by 1990 male lung cancer mortality, although still high, was decreasing
rapidly.8-12
Setting:
United Kingdom.
Participants:
Hospital patients under 75 years of age
with and without lung cancer in 1950 and 1990, plus, in 1990, a matched sample of the local population: 1465 case-control pairs in the 1950 study, and 982 cases plus 3185 controls in the 1990 study.
Main outcome measures:
Smoking prevalence and lung cancer.
Results:
For men in early middle age in the United Kingdom the prevalence of smoking halved between 1950 and 1990 but the
death rate from lung cancer at ages 35-54 fell even more rapidly,
indicating some reduction in the risk among continuing smokers. In
contrast, women and older men who were still current smokers in 1990 were more likely than those in 1950 to have been persistent cigarette
smokers throughout adult life and so had higher lung cancer rates than
current smokers in 1950. The cumulative risk of death from lung cancer
by age 75 (in the absence of other causes of death) rose from 6% at
1950 rates to 16% at 1990 rates in male cigarette smokers, and from
1% to 10% in female cigarette smokers. Among both men and women in
1990, however, the former smokers had only a fraction of the lung
cancer rate of continuing smokers, and this fraction fell steeply with
time since stopping. By 1990 cessation had almost halved the number of
lung cancers that would have been expected if the former smokers had
continued. For men who stopped at ages 60, 50, 40, and 30 the
cumulative risks of lung cancer by age 75 were 10%, 6%, 3%, and 2%.
Conclusions:
People who stop smoking, even well into
middle age, avoid most of their subsequent risk of lung cancer, and
stopping before middle age avoids more than 90% of the risk
attributable to tobacco. Mortality in the near future and throughout
the first half of the 21st century could be substantially reduced by
current smokers giving up the habit. In contrast, the extent to which young people henceforth become persistent smokers will affect mortality
rates chiefly in the middle or second half of the 21st century.
![]()
Introduction
Top
Abstract
Introduction
Participants and methods
Results
Discussion
References

View larger version (16K):
[in a new window]
Fig 1.
Trends in prevalence of smoking at ages 35-59 (left) and
60 (right) in men and women in the United Kingdom,
1950-98. Prevalences at ages 25-34 were 80% for men and 53% for women
in 1948-52 and 39% for men and 33% for women in 1998. Further details
are given on the BMJ's website (table A)

View larger version (17K):
[in a new window]
Fig 2.
Trends in mortality from lung cancer in men and
women in the United Kingdom, 1950-97: annual mortality per
105 at ages 35-54 (left) and 55-74 (right) years. Rate in
each 20 year age range is mean of rates in the four component five year
age groups. Age specific rates from 1950-2 to 1993-7 are given on
BMJ's website (tables B and C); at ages 35-54 and 55-74 in
1998 the rates were 17 and 243 (men) and 12 and 20 (women)
In this paper we relate the UK national trends in smoking, in smoking
cessation, and in lung cancer to the contrasting results from two large
case-control studies of smoking and lung cancer in the United Kingdom
that were conducted 40 years apart, centred on the years
1950
2 7
and 1990.8 The 1950 study was
concerned with identifying the main causes of the rise in lung cancer
and showed the predominant role of tobacco. The 1990 study was
concerned not just with reconfirming the importance of tobacco but also with assessing the lesser effects of indoor air pollution of some houses by radon.8 Because there has been widespread
cessation of smoking (indeed, above age 50 there are now twice as many
former cigarette as current cigarette smokers in the United
Kingdom10), the second study was able to assess the long
term effects of giving up the habit at various ages.
| |
Participants and methods |
|---|
|
|
|---|
The 1950 study was conducted in London and four other large towns during 1948-52, and its methods have been described elsewhere. 2 7 It involved interviewing, as potential "cases," patients younger than 75 years of age in hospital for suspected lung cancer and, as "controls," age matched patients in hospital with various other diseases (some of which would, in retrospect, have been conditions associated with smoking). After patients in whom the initial diagnosis of lung cancer was eventually refuted were excluded from the cases, 1465 cases and 1465 controls remained. A preliminary report on 709 case-control pairs was published in 1950, and the full results were published two years later. 2 7
The 1990 study was conducted during 1988-93 in a part of southwest England that had not been included in the 1950 study. Potential cases were patients younger than 75 who were referred with suspected lung cancer to the five hospitals in Devon and Cornwall that investigated lung cancer. For each case a population control was obtained, selected randomly either from lists of the local family health services authority or from electoral rolls, and a hospital control was selected from patients whose current admission was for a disease not thought to be related to smoking. Controls were matched for age, sex, and broad area of residence to the patients with suspected lung cancer. Cases and controls were eligible for the 1990 study only if they were current residents of Devon or Cornwall, had lived in one of these two counties for at least 20 years, and could be interviewed in person by research assistants about smoking habits and other relevant characteristics. The final diagnosis of cases was sought; those who had a smoking related disease other than lung cancer were excluded; and the few who had a disease not known to be associated with smoking were transferred to the hospital control group. Similarly, in 1990 (although not in 1950) the final diagnosis of all the hospital controls was sought, and those whose main reason for being in hospital was a disease known to be related to smoking were excluded from the study.
The distributions of the smoking habits of the population controls and hospital controls in 1990 were closely similar, and the results are presented here with these two control groups combined. Further details of the study design and methods of data collection and analysis have been given elsewhere.8 Information was obtained in the 1990 study about the smoking habits of 667 men and 315 women with a confirmed diagnosis of lung cancer and of 2108 male and 1077 female controls.
Statistical methods
Relative and cumulative risks
Relative risks for men and women comparing particular categories
of smoker with lifelong non-smokers in the 1990 study (and the ratios
of the risks in former smokers to those in continuing smokers) were
calculated by logistic regression with adjustment for
age.13 Further adjustment for social class, radon
exposure, and county of residence made no material difference. Relative
risks for men and women in the 1950 study were taken as the odds ratios
indicated by the published frequency distributions of the age matched
cases and controls.7 Relative risks from the studies were
then combined with national lung cancer mortality rates from 1950 and
1990 respectively to estimate the absolute hazards in various
categories of smoker, former smoker, and non-smoker. Because they are
linked to known national rates, these absolute risks are statistically
stable among smokers (and among former smokers), even though the risks
relative to lifelong non-smokers would not be stable as so few
non-smokers develop the disease. Such calculations of absolute risk
allow comparisons between different categories of smoker not only
within this study but also between this and other studies that report
absolute risks.
exp(
5c/105). For the 1950 study
the relative risks were multiplied by 0.6 (men) and 0.5 (women) to
yield the cumulative risk (%) by age 75. These factors were chosen to
ensure that the population weighted means of the cumulative risks for
lifelong non-smokers, former smokers, cigarette smokers, and other
smokers were 4.7% (men) and 0.7% (women) as in the 1950 population.
(The cumulative risk, which depends only on the age specific lung
cancer rates up to age 75 and not on competing causes of death, is
somewhat less than the lifetime risk.)
Use of statistically stable non-smoker rates from a large US
study
The most reliable recent evidence on lung cancer rates
among lifelong non-smokers in developed countries is that from a
prospective study of mortality in one million Americans during the
1980s (see table D on BMJ's website).
14 15
These American rates seem to correspond not only to what normally
happens in the United States but also to what normally occurs in the
United Kingdom, at least among professional men. For, when these
figures were used to predict the total number of deaths from lung
cancer among the non-smokers in a cohort of male British doctors that has been followed prospectively for 40 years from 1951 to
1991,
16 17
the number expected was 19.03; the number
actually observed was 19 (R Doll, personal communication). The American
lung cancer rates for non-smokers suggest cumulative risks by 75 years
of age of 0.44% for men and 0.42% for women.
0.6% (men) and 0.5% (women) in lifelong
non-smokers
may be slightly too high, although the rate in men is
based on only seven cases and was inflated by problems with the 1950 male controls (see Results). We have therefore used the American
results for non-smokers in most of our analyses. This does not affect
the risk ratios comparing smokers and former smokers or the estimated
absolute risks among smokers and former smokers.
| |
Results |
|---|
|
|
|---|
Effects of current smoking in 1990 study
Most of the participants who were still current cigarette smokers
in 1990 would have been cigarette smokers throughout adult life, and
the cumulative risk of lung cancer by age 75 in this group was 15.9%
for men and 9.5% for women (see BMJ's website, table
E). These cumulative risks reflect the death rates from lung cancer of
cigarette smokers in 1990 and were obtained by combining the relative
risks from the 1990 case-control study with national death rates. Had
these men and women smoked as intensively when they were young as
adolescent smokers do nowadays, the cumulative risks might have been
greater. Only 34% of the male and 11% of the female controls who were
current smokers had started before the age of 15 years, and the
case-control comparisons indicate that smokers who had done so had
double the risk of lung cancer of those who had started aged 20 or
older (risk ratios adjusted for age and amount smoked were 2.3 (95%
confidence interval 1.4 to 3.8) for men and 1.8 (0.9 to 3.4) for women).
Effects of cessation in 1990 study
A large number of men and, to a lesser extent, of women had
stopped smoking well before 1990. Hence, particularly for men, robust
estimates can be obtained from the 1990 data of the effects of
prolonged cessation on the avoidance of risk (table 1).
|
|
Comparison of findings for smoking in 1950 and 1990 studies
The hazards at the death rates among current smokers in the 1990 study, when the male lung cancer epidemic was well past its peak, can
be compared with the hazards at the death rates among current smokers
in the 1950 study,
2 7
when the epidemic was still
increasing rapidly, except among men in early middle age (table 2).
Absolute risks in smokers unaffected by biases in 1950 male
controls
The findings in the earlier study were reported for categories of
smoking that differ slightly from those now considered appropriate, but
this probably makes little difference. In addition, the hospital
controls in the earlier study included an unknown, but appreciable,
proportion of patients who were in hospital for conditions that were
subsequently shown to be related to smoking but were not known to be so
in 1950. This means that the proportion of smokers was higher than in
the general population and also that the relative risks estimated from
the 1950 study for different levels of smoking were too low. Both
effects will have been relatively unimportant for women, as few women
at that time had been smoking long enough to have been admitted to
hospital because of a smoking related disease. Even for men, they will have had little effect on the calculated absolute risk among smokers. If, for example, the male rate of hospital admission for the control diseases was about 1.5 times as great among smokers as among
non-smokers, then correction for this would multiply the relative risk
of lung cancer in male smokers by about 1.5 and would indicate that the percentage of current smokers in the study areas was not 86%, but
about 80% (which was about the percentage in the country as a whole).
But this correction would have no material effect on the cumulative
risk calculated for cigarette smokers (and little effect on that
calculated for other smokers or former smokers), as the weighted
average has to remain 4.7% to match the 1950 male death rates. It
would merely reduce the cumulative risk calculated for male non-smokers
from 0.6% to about 0.4%, thereby bringing it closer to that in US
non-smokers.
|
Changes in prevalence of smoking
One clear difference between the 1950 and 1990 study results in
table 2 is that many of the controls in the 1990 study had given up
smoking, so there was a large decrease in the prevalence of smoking
between the two studies. (In both 1950 (after correction) and 1990, the
prevalence of smoking among controls resembled that in national
surveys.) The reduction in the proportion currently smoking cigarettes
was smaller in women than in men. Among women who still smoked in 1990, a higher proportion smoked heavily than was the case in 1950, and a
substantially larger proportion had started before the age of 20 (68%
in 1990 and 24% in 1950 among women, compared with 83% and 76%
respectively among men). Moreover, the way that women smoke a cigarette
has become more like the way men do.22 Nevertheless, among
women old enough to be in the 1990 study more than half of those who had been cigarette smokers had given up the habit, and an even greater
proportion of the men had done so. A recent national survey confirms
that among men and women aged over 50 in the United Kingdom, the number
of former cigarette smokers is double the number of continuing
cigarette smokers.10 But those who are continuing smokers
nowadays may well have smoked substantial numbers of cigarettes throughout adult life, whereas national cigarette sales during the
first few decades of the last century
9 18
show that few of the older smokers in 1950 can have done so.
Changes in lung cancer rates among continuing smokers
Another clear difference between the two studies is that the
cumulative risk of lung cancer among smokers increased substantially.
The increase occurred not only among women (among whom the cumulative
risk for cigarette smokers was 1.0% in 1950 and 9.5% in 1990) but
also among men (among whom it increased from 5.9% at 1950 cigarette
smoker lung cancer rates to 15.9% at 1990 rates). As lung cancer
mainly occurs above the age of 55, the increase in the cumulative risk
is mainly because current smokers aged 55-74 in 1950 were less likely
to have smoked a substantial number of cigarettes throughout adult life
than current smokers in 1990.
18 19
Among younger men,
however, the death rate from lung cancer decreased more rapidly than
the prevalence of smoking (figs 1 and 2 ), indicating lower death rates
from lung cancer in 1990 than 1950 among male cigarette smokers in
early middle age.
| |
Discussion |
|---|
|
|
|---|
Prolonged cigarette smoking
The 1990 study provides reliable evidence, particularly among men,
about the absolute effects of prolonged cigarette smoking and about the
effects of prolonged cessation (table 1, fig 3). Information about the
effects of prolonged cigarette smoking could not have been obtained in
1950 because the habit became widespread in the United Kingdom (firstly
among men and then among women) only during the first half of the 20th century. By 1950 the increase in smoking was too recent to have had its
full effects on disease rates, except perhaps among men in early middle
age. The fact that by 1990 many of the current smokers would have
smoked substantial numbers of cigarettes throughout adult life is the
chief reason for the large increase in the cumulative risk of lung
cancer among continuing smokers.19 For the same reason,
increases in the risks associated with smoking were also seen between
the first 20 years (1951-71) and the next 20 years (1971-91) of follow
up in the prospective study of smoking and death among British
doctors,17 and between the two large prospective studies
carried out by the American Cancer Society in the 1960s and
1980s.
15 20
Prolonged cessation
In the 1990 study we were able to assess the effects of prolonged
cessation among those who had smoked cigarettes for many years.
Although efforts to change from cigarettes to other types of tobacco,
or from smoking substantial numbers of cigarettes to smoking smaller
numbers, seemed to confer only limited benefit (table 2), stopping
smoking confers substantial benefit. Figure 3 indicated that even
people who stop smoking at 50 or 60 years of age avoid most of their
subsequent risk of developing lung cancer, and that those who stop at
30 years of age avoid more than 90% of the risk attributable to
tobacco of those who continue to smoke (see fig 3 and
BMJ's website, table G). In the United Kingdom widespread
cessation has roughly halved the number of cases of lung cancer that
would now be occurring, as by 1990 it had already almost halved the
number that would have occurred in the study (table 1).
Past and future trends in total mortality attributable to tobacco
Despite cessation of smoking and improvements in cigarette
composition, lung cancer is still the chief neoplastic cause of death
in the United Kingdom, and tobacco causes even more deaths from other
diseases than from lung cancer.
14 15
The changes since
1950 in tobacco-attributable mortality from diseases other than lung
cancer can be estimated indirectly from national mortality
statistics.
14 15
Such estimates indicate that in 1965 the
United Kingdom probably had the highest death rate from tobacco related
diseases in the world, but that since then the number of deaths in
middle age (35-69) from tobacco has decreased by about half, from
80 000 in 1965 to 43 000 in 1995. Nevertheless, cigarette smoking
remains the largest single cause of premature death in the United
Kingdom and eventually kills about half of those who persist in the
habit.17 The 1990 study assessed the effects of stopping
smoking only on lung cancer, but a comparably large benefit of stopping
was found for all cause mortality in the prospective study of smoking
and death among British doctors.17 This reinforces similar
evidence from many other countries that even in middle age those who
stop smoking avoid most of their subsequent risk of being killed by tobacco.
|
What is already known on this topic
Smoking is a cause of most deaths from lung cancer in the United Kingdom Early studies could not reliably assess the effects of prolonged cigarette smoking or of prolonged cessation What this study addsIf people who have been smoking for many years stop, even well into middle age, they avoid most of their subsequent risk of lung cancer Stopping smoking before middle age avoids more than 90% of the risk attributable to tobacco Widespread cessation of smoking in the United Kingdom has already approximately halved the lung cancer mortality that would have been expected if former smokers had continued to smoke As most current smokers in the United Kingdom have consumed substantial numbers of cigarettes throughout adult life, their risks of death from lung cancer are greater than earlier studies had suggested Mortality from tobacco in the first half of the 21st century will be affected much more by the number of adult smokers who stop than by the number of adolescents who start |
| |
Acknowledgments |
|---|
We thank the individuals and research assistants who took part in both studies; the staff in hospitals, general practices, and the South and West Cancer Intelligence Unit; Cathy Harwood and Anthea Craven for secretarial assistance; and Jillian Boreham for graphics.
Contributors: RD planned the 1950 study with A Bradford-Hill and planned the 1990 study with SD. SD, RD, HD, PS, and EW conducted and analysed the 1990 study in the Imperial Cancer Research Fund Cancer Epidemiology Unit. RP, SD, and RD planned and wrote the paper. SD is the guarantor.
| |
Footnotes |
|---|
Funding: direct support to Clinical Trial Service Unit from the Medical Research Council (which also funded the 1950 study), the British Heart Foundation, and the Imperial Cancer Research Fund. The 1990 study was funded by the Imperial Cancer Research Fund; the National Radiological Protection Board; the Department of Health; the Department of the Environment, Transport and the Regions; and the European Commission.
Competing interests: None declared.
Further data are available on the
BMJ's website
| |
References |
|---|
|
|
|---|
| 1. |
Doll R.
Uncovering the effects of smoking: historical perspective.
Stat Methods Med Res
1998;
7:
87-117 |
| 2. | Doll R, Hill AB. Smoking and carcinoma of the lung. Preliminary report. BMJ 1950; ii: 739-748. |
| 3. | Wynder EL, Graham EA. Tobacco smoking as a possible etiologic factor in bronchogenic carcinoma. JAMA 1950; 143: 329-336. |
| 4. | Levin ML, Goldstein H, Gerhardt PR. Cancer and tobacco smoking. JAMA 1950; 143: 336-338. |
| 5. | Mills CA, Porter MM. Tobacco smoking habits and cancer of the mouth and respiratory system. Cancer Res 1950; 10: 539-542. |
| 6. | Schrek R, Baker LA, Ballard GP, Dolgoff S. Tobacco smoking as an etiologic factor in disease. Cancer Res 1950; 10: 49-58. |
| 7. | Doll R, Hill AB. A study of the aetiology of carcinoma of the lung. BMJ 1952; ii: 1271-1286. |
| 8. | Darby S, Whitley E, Silcocks P, Thakrar B, Green M, Lomas P, et al. Risk of lung cancer associated with residential radon exposure in southwest England: a case-control study. Br J Cancer 1998; 78: 394-408[Medline]. |
| 9. | Nicolaides-Bouman A, Wald N, Forey B, Lee PN. International smoking statistics. Oxford: Oxford University Press, 1993. |
| 10. | Office for National Statistics. Results from the 1998 general household survey. London: Stationery Office, 2000. |
| 11. | Office for National Statistics. Mortality statistics, general: 1993, 1994 and 1995. London: Stationery Office, 1997. |
| 12. | World Health Organization. Mortality statistics for the United Kingdom, 1950-1997. http://www.who.int/whosis/mort. |
| 13. | Stata Corporation. Stata Statistical Software: Release 5.0. College Station, TX: Stata Corporation, 1997. |
| 14. | Peto R, Lopez A, Boreham J, Thun M, Heath C. Mortality from tobacco in developed countries: indirect estimation from national vital statistics. Lancet 1992; 339: 1268-1278[CrossRef][Medline]. |
| 15. | Peto R, Lopez A, Boreham J, Thun M, Heath C. Mortality from tobacco in developed countries: indirect estimation from national vital statistics. Oxford: Oxford University Press, 1994. |
| 16. | Doll R, Peto R. Mortality in relation to smoking: 20 years' observations on male British doctors. BMJ 1976; ii: 1525-1536. |
| 17. |
Doll R, Peto R, Wheatley K, Gray R, Sutherland I.
Mortality in relation to smoking: 40 years' observations on male British doctors.
BMJ
1994;
309:
901-911 |
| 18. | Wald N, Kiryluk S, Darby S, Doll R, Pike M, Peto R. UK smoking statistics. Oxford: Oxford University Press, 1988. |
| 19. | Doll R, Peto R. The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J Natl Cancer Inst 1981; 66: 1191-1308. |
| 20. |
Thun MJ, Day-Lalley CA, Calle EE, Flanders WD, Heath CA.
Excess mortality among cigarette smokers: changes in a 20-year interval.
Am J Pub Health
1995;
85:
1223-1230 |
| 21. | Peto R. Overview of cancer time-trend studies in relation to changes in cigarette manufacture. In: Zaridze D, Peto R, eds. Tobacco: a major international health hazard. Lyons: International Agency for Research on Cancer, 1986:211-226. (IARC scientific publication No 74.) |
| 22. | Doll R, Gray R, Hafner B, Peto R. Mortality in relation to smoking: 22 years' observation on female British doctors. BMJ 1980; 280: 967-971. |
| 23. | Doll R, Darby S, Whitley E. Trends in mortality from smoking-related diseases. In: Charlton J, Murphy M, eds. The health of adult Britain, 1841-1994. , Vol 1 London: Stationery Office, 1997:128-155. |
| 24. |
Peto R, Chen ZM, Boreham J.
Tobacco the growing epidemic.
Nature Med
1999;
5:
15-17[CrossRef][Medline].
|
| 25. | Peto R, Lopez AD. The future worldwide health effects of current smoking patterns. In: Koop CE, Pearson CE, eds. Global health in the 21st century. New York: Jossey-Bass, 2000. |
(Accepted 7 July 2000)