Incidence rates are based on the second examination in 11 151 men
and 8563 women from Copenhagen, Denmark.
Godtfredsen NS, Prescott E, Osler M. Effect of Smoking Reduction on Lung Cancer Risk. JAMA. 2005;294(12):1505–1510. doi:10.1001/jama.294.12.1505
Author Affiliations: The Copenhagen Centre
for Prospective Population Studies, Danish Epidemiology Science Centre at
the Institute of Preventive Medicine, Copenhagen University Hospital, Copenhagen,
Denmark (Drs Godtfredsen and Prescott); Department of Cardiology and Pulmonary
Medicine, Hvidovre Hospital, Hvidovre, Denmark (Dr Godtfredsen); Department
of Cardiology, Rigshospitalet, Copenhagen, Denmark (Dr Prescott); and Department
of Social Medicine, Institute of Public Health, University of Copenhagen,
Copenhagen, Denmark (Dr Osler).
Context Many smokers are unable or unwilling to completely quit smoking. A proposed
means of harm reduction is to reduce the number of cigarettes smoked per day.
However, it is not clear whether this strategy decreases the risk for tobacco-related
Objective To assess the effects of smoking reduction on lung cancer incidence.
Design, Setting, and Participants Observational population-based cohort study with up to 31 years of follow-up
from the Copenhagen Centre for Prospective Population Studies, which administrates
data from 3 longitudinal studies conducted in Copenhagen and suburbs, the
Copenhagen City Heart Study, the Copenhagen Male Study, and the Glostrup Population
Studies, Denmark. Participants were 11 151 men and 8563 women (N = 19 714)
aged 20 to 93 years, who attended 2 consecutive examinations with a 5- to
10-year interval between 1964 and 1988. Participants underwent a physical
examination and completed self-filled questionnaires about lifestyle habits.
The study population was divided into 6 groups according to smoking habits:
continued heavy smokers (≥15 cigarettes/d), reducers (reduced from ≥15
cigarettes/d by minimum of 50% without quitting), continued light smokers
(1-14 cigarettes/d), quitters (stopped between first and second examination),
stable ex-smokers, and never smokers.
Main Outcome Measure Incident primary lung cancer cases assessed by record linkage with the
National Cancer Registry until December 31, 2003.
Results There were 864 incident lung cancers during follow-up. Using Cox regression,
the adjusted hazard ratio (HR) for lung cancer in reducers was 0.73 (95% confidence
interval [CI], 0.54-0.98) compared with persistent heavy smokers. The HR for
light smokers was 0.44 (95% CI, 0.35-0.56); for quitters, HR 0.50 (95% CI,
0.36-0.69), for stable ex-smokers, HR 0.17 (95% CI, 0.13-0.23), and for never
smokers, HR 0.09 (95% CI, 0.06-0.13).
Conclusion Among individuals who smoke 15 or more cigarettes per day, smoking reduction
by 50% significantly reduces the risk of lung cancer.
Lung cancer remains the leading cause of death from cancer worldwide
with an estimated 90% of the cases being tobacco related. Consequently, efforts
to prevent smoking uptake and to encourage smoking cessation are crucial in
cancer control. However, the overall prevalence of cigarette smoking is still
high, and the efficacy of smoking cessation intervention is limited.1- 3 This has led tobacco
researchers to investigate possible alternative methods for diminishing the
harmful effects of smoking, so-called harm reduction. One element in harm
reduction, which is gaining increasing attention, concerns reducing the number
of cigarettes smoked per day.
Despite comprehensive reviews on the issue,4,5 there
are limited data in the literature on the effects of smoking reduction with
respect to end points such as morbidity and mortality from tobacco-related
diseases. We have previously shown that heavy smokers (>15 cigarettes/d) who
reduce their tobacco intake by at least 50% do not decrease their risk of
fatal or nonfatal myocardial infarction, hospitalization for chronic obstructive
pulmonary disease (COPD), or all-cause mortality compared with heavy smokers
who do not change smoking habits.6- 8 Compensatory
smoking, ie, increasing the puff volume per cigarette, is a possible explanation
for the lack of benefit.
Evidence regarding changes in smoking habits and lung cancer incidence
is—with the exception of the impact of smoking cessation—also
sparse. Whether shifting to low-tar and low-nicotine yield cigarettes decreases
lung cancer risk is still controversial, as is the issue concerning the effect
of smoking other tobacco types than cigarettes (cigar and/or pipe; dark vs
blond tobacco).9- 20 The
general tendency in the numerous case-control studies is toward a more favorable
risk profile from low-tar cigarettes and smoking of a cigar or pipe, whereas
most cohort studies have not found any substantial reductions in risk. This
can probably be attributed to differences in adjustment for inhalation habits
and smoking intensity and selection bias in the case-control design.
Only 1 cohort study and 1 case-control study have actually investigated
individual changes in smoking behavior and subsequent lung cancer risk.19,21 In the former, switching from cigarette
smoking to pipes or cigars decreased lung cancer mortality by 46% compared
with continuing cigarette smokers, while the case-control study showed nonsignificantly
decreased lung cancer risks with change in cigarette type or a reduction in
cigarettes per day by more than 25%. Our observational study of smoking reduction
and mortality indicated a nonsignificant trend toward a reduced risk of mortality
from tobacco-related cancer.6 From a public
health perspective, the recent results from the California tobacco control
program support that extensive smoking restrictions are followed by a decline
in lung cancer incidence.22 A newly published
intervention study has investigated the relationship between short-term (up
to 26 weeks) smoking reduction and a tobacco-specific lung carcinogen.23 The uptake in urine of the metabolites of the lung
carcinogen was significantly reduced in participants who reduced smoking;
however, the percentage reduction in levels of the metabolites was modest
and did not correspond with the percentage decrease in cigarettes per day.
The objective of the present cohort study was to analyze the association
between smoking reduction and incidence of the major histologic subtypes of
primary lung cancer.
The Copenhagen Centre for Prospective Population Studies contains pooled
data from 3 longitudinal population-based studies carried out in Copenhagen,
Denmark and its vicinity during 1964 through 1993. A total of 30 911
persons participated at least once in 1 of the substudies, and the majority
of participants were examined on 3 occasions with 1- to 10-year intervals.
Mean response rate was 77%. The studies were approved by local ethics committees
and all participants signed an informed consent prior to participation. All
studies have been previously described.24- 26
In the present study of changes in smoking habits, we required complete
information on smoking habits in at least 2 consecutive examinations, preferably
the first (baseline) and the second (follow-up). Thus, the present population
was composed of 19 714 study participants (8563 women and 11 151
men). Participants were observed by use of a central personal registration
number from the second examination until 2004, ie, up to 31 years of follow-up
with a mean of 18 years. Follow-up was complete for all participants except
those who emigrated or disappeared (<1%). Table 1 shows the distribution of the population on substudy, age,
and years of examination.
Smoking status and changes in smoking habits in this study are based
on self-reports. At each examination, participants were asked whether they
smoked or not and if affirmative, about amount, duration, inhalation, and
preferred type of tobacco. Ex-smokers were asked about years of active smoking
and participants from the Copenhagen City Heart Study were asked about time
since quitting (>5 years or <5 years prior to examination). For standardization
of the tobacco variables in the pooled population, participants were categorized
as heavy smokers (≥15 g/d), light smokers (1-14 g/d), ex-smokers, and never
smokers. Tobacco consumption was calculated by equating a cigarette to 1 g
of tobacco, a cheroot to 3 g, and a cigar to 5 g. Pack-years of smoking were
calculated at the baseline examination as number of cigarettes per day multiplied
by number of years of smoking divided by 20. Due to the imprecise information
on exact dates for changes in smoking (quitting or reducing), it was not possible
to calculate pack-years at follow-up.
Our definition of smoking reduction is in accordance with the clinical
studies of heavy smokers who reduce, in which a chosen set point of 50% reduction
or more in the amount smoked is sought and achieved.27- 29 It
is well known that the heavier smokers are generally more addicted to nicotine
than light smokers, and hence, find it more difficult to quit. To measure
a substantial reduction in tobacco consumption, we only considered participants
who were heavy smokers at first examination. Smoking reduction was then defined
as smokers of 15 g of tobacco or more per day at first examination, who at
the second examination reported a decrease of 50% or more without quitting.
We also categorized 2 groups of former smokers: those who were ex-smokers
both at baseline and follow-up and those who were smokers at baseline but
had stopped smoking at follow-up. The study population as a whole was then
divided into the following categories: continued heavy smokers, reducers,
continued light smokers, quitters, continued ex-smokers, and never smokers.
New cases of lung cancer from study entry, ie, date of second examination
until December 31, 2003, were identified by record linkage with the Danish
National Cancer Registry. In validation studies, the cancer registry has been
found to register more than 95% of all cancers with a nearly 100% accuracy
for cancers with a high mortality rate such as lung cancer.30 We
obtained the year and month of diagnosis and included the following histologic
types: squamous cell carcinoma, adenocarcinoma, anaplastic carcinoma (small
cell lung cancer), and primary lung cancer with unspecified histology.
Multivariate Cox proportional hazards regression models were applied
to study the association between smoking reduction, smoking cessation, no
change in smoking habits, and lung cancer incidence. Age was chosen as the
underlying timescale, thus allowing for delayed entry. Factors such as education,
alcohol consumption, and body mass index (calculated as weight in kilograms
divided by the square of height in meters) did not differ between the 2 main
smoking groups of interest (reducers and sustained heavy smokers), and were
therefore, not included in the analyses. We included the following confounders
in the model: cohort of origin, sex, duration of smoking, inhalation habits,
and type of tobacco smoked. Pack-years of smoking were not included because
of the method our dependent variable was constructed (the 6 smoking groups),
which would have resulted in over-adjustment. Tests of interaction mainly
between sex and baseline smoking habits were performed. The assumption of
proportional hazards was tested graphically and with a formal statistical
test as proposed by Grambsch and Therneau.31 All
analyses were performed using Stata software version 7.0 (Stata Corp, College
The demographic and smoking characteristics by smoking group are shown
in Table 2. Reducers decreased their
tobacco consumption from a mean of 22.2 g per day to a mean of 8.5 g per day,
stable heavy smokers consumed approximately 20 cigarettes per day at both
examinations, quitters consumed an average of 14.5 g of tobacco per day at
baseline, and stable light smokers consumed 9 g of tobacco per day at both
examinations. Compared with continued heavy smokers, the reducers were significantly
older, a larger proportion were men, and they had smoked slightly more and
longer; however, number of pack-years was significantly smaller (27 vs 31).
Furthermore, a smaller proportion of reducers inhaled and were less likely
to smoke cigarettes only.
During follow-up, 864 participants were diagnosed with primary lung
cancer: 360 cases were among women and 504 were among men. Table 3 shows the distribution of cancer type among the smoking
groups. There were 229 cases of squamous cell carcinoma, 234 cases of adenocarcinoma,
179 cases of small cell lung cancer, and 222 lung cancers with unspecified
histology. Overall, the cancers are evenly distributed among the groups of
current smokers with a tendency of a larger proportion of small cell lung
cancer with increasing tobacco exposure.The age-standardized incidence rates
of lung cancer according to the smoking groups of interest for men and women
are shown in the Figure. Overall, the
rates are higher for men than for women except in never smokers. There is
a dose response relationship between increasing smoking intensity and lung
cancer incidence rates except in male quitters, whose rates are slightly higher
than continued light smoking men.
Results of the Cox regression analyses are shown in Table 4. Reducing tobacco consumption from approximately 20 cigarettes
per day to less than 10 was associated with a 27% (95% confidence interval
[CI], 2%-46%) reduction in lung cancer risk compared with unchanged heavy
smoking. Participants who were continued light smokers or who quit smoking
between baseline and follow-up reduced their lung cancer risk by 56% and 50%,
respectively, compared with persistent heavy smokers. Risk of lung cancer
among the stable ex-smokers was 83% lower than among the heavy smokers, but
still significantly higher than among the never smokers. Omitting the first
2 years after follow-up in the analyses to control for any possible ill-quitter
or ill-reducer effect did not significantly change results, neither did omission
of participants according to a chronic baseline disease index based on self-reported
chronic respiratory conditions or hospital admission for lung disease before
study entrance. We also performed analyses confined to cigarette smokers,
thus excluding smokers of any other tobacco product, but this did not change
the associations. In the multivariate analyses, inhaling the smoke, longer
duration of smoking, and being of male sex were independent risk factors for
lung cancer, whereas smoking other types of tobacco than cigarettes (or mixed),
and cohort of origin were not.
In this large, population-based study with extensive follow-up, we found
that reporting a large reduction in tobacco consumption between baseline and
follow-up was associated with a decreased risk of subsequent lung cancer.
However, a mean decrease of 62% in the amount smoked corresponded with a 27%
reduction in risk of lung cancer, whereas participants who were light smokers
throughout the study or who stopped smoking had considerably lower risk. This
indicates that risk reduction is disproportionally smaller than the corresponding
Undoubtedly, smoking cessation decreases the risk of lung and other
tobacco-related cancers,32,33 and
our study also demonstrates that participants who were already ex-smokers
when enrolled in the study had a much lower lung cancer risk than those who
reported quitting between baseline and follow-up. Recently, Peto et al34 showed that cessation of smoking before middle age
is associated with a more than 90% reduction in tobacco-attributable cancer
risk. Regarding smoking reduction and tobacco-related cancer, there are no
data from other population-based studies. Results from 2 intervention studies
examining carcinogenic biomarkers23,35 and
other short-term smoking-reduction studies using nicotine replacement therapy36,37 clearly demonstrate the major problems
in harm reduction. Only a minority of the smokers are actually able to achieve
and sustain a considerable reduction in cigarettes per day, and equally important,
even with ad libitum nicotine replacement therapy,
substantial compensatory smoking occurs as measured by the most common biomarkers
of tobacco exposure. Our results are in accordance with these studies.
In our study, we did not have repeated measurements of biomarkers of
smoking. Carbon monoxide in expired air was measured in the Copenhagen City
Heart Study III (1991-1993) in approximately 10 000 participants, and
serum cotinine was obtained from 3300 participants in the Copenhagen Male
Study in 1985. In this subset of smokers, we found that the reducers had significantly
decreased levels of carbon monoxide and cotinine, respectively, compared with
the heavy smokers but still higher than the stable light smokers, despite
similar exposure in these 2 groups. This indicates that in observational studies
not using nicotine replacement therapy, compensatory mechanisms occur. Furthermore,
our results could be biased from differential misclassification especially
if the reducers tended to underreport their consumption or changed their smoking
habits during follow-up. However, due to the study design and our biomarker
results, we do not believe this to be of major importance. Data on smoking
behavior from the second examination to the third indicate that approximately
50% have continued as light smokers, 20% have quit entirely, and the remaining
30% have relapsed to heavy smoking. Hence, misclassification of smokers is
likely to affect results in both directions (underestimation and overestimation
of associations). However, also a proportion of the reference group has presumably
reduced or quit smoking during follow-up leading to an underestimation of
the effect of smoking reduction.
We have previously investigated the end points all-cause and cause-specific
mortality, fatal or nonfatal myocardial infarction, and a first hospitalization
for COPD in this cohort, and as mentioned earlier, we did not find any risk
reduction with smoking reduction compared with continued heavy smoking. However,
there was a trend toward a declining risk of lung cancer deaths following
reduced smoking.6 It is likely that lung cancer,
which in particular demonstrates a dose-response relationship with smoking
amount, will develop to a lesser extent after smoking reduction in contrast
with other tobacco-related disorders due to DNA repair. On the other hand,
given the long duration of excess lung cancer risk in former smokers, a rapid
decline in risk for the reducers to the level of light smokers would not have
been expected even if there were no signs of compensatory smoking as discussed
This study has the advantages of a reasonably large size, a large proportion
of smokers at baseline, and a long, up-to-date and almost 100% complete follow-up
of participants. When studying changes in risk factors in an observational
setting, the question of unmeasured confounding becomes relevant. We have
focused on the comparison between the reducers and the sustained heavy smokers,
but our study design does not enable us to examine reasons for this change
in smoking behavior. For instance, we do not know if this group actually comprises
the participants who are unable or unwilling to quit altogether. We have previously
shown from the Copenhagen City Heart Study38 that
heavy smokers who reduce their smoking generally have a less healthy lifestyle
than continued heavy smokers. However, these differences were not as pronounced
in the pooled study population and we also used another definition of smoking
reduction in the present study.
In conclusion, smoking reduction from an average of 20 cigarettes per
day to less than 10 cigarettes per day reduced the lung cancer risk by approximately
25%. Presumably, the discrepancy between reported amount of reduction and
calculated risk reduction can be largely explained by compensatory smoking.
More data from long-term studies of smoking reduction are warranted, but for
the present, smoking cessation and not smoking reduction should still be advocated
as the ultimate method of reducing harm from smoking, especially since diseases
such as COPD and myocardial infarction, which have a larger public health
effect than lung cancer, have not shown any reductions in risks after smoking
Corresponding Author: Nina S. Godtfredsen,
MD, PhD, Department of Cardiology and Pulmonary Medicine 253, H:S Hvidovre
Hospital, Kettegård allé 30, DK-2650 Hvidovre, Denmark (email@example.com).
Author Contributions: Dr Godtfredsen had full
access to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Godtfredsen, Prescott,
Acquisition of data: Godtfredsen.
Analysis and interpretation of data: Godtfredsen,
Drafting of the manuscript: Godtfredsen.
Critical revision of the manuscript for important
intellectual content: Prescott, Osler.
Statistical analysis: Godtfredsen.
Obtained funding: Godtfredsen, Prescott, Osler.
Study supervision: Prescott, Osler.
Financial Disclosures: None reported.
Funding/Support: This study was supported by
grants from the Danish Ministry of Health, the Health Insurance Foundation,
the Danish Lung Foundation, and the Wedell-Wedellsborg Foundation.
Role of the Sponsors: The funding organizations
did not have any role in either design or conduct of the study or preparation,
review, or approval of the manuscript.
Acknowledgment: We would like to thank the
steering committee for the Copenhagen Centre for Prospective Population Studies.