The y-axis is a log scale. The error bars denote 95% confidence intervals. The reference category consists of current smokers. The horizontal dashed line indicates a never-smoker's risk. Adjusted for age (months), follow-up period, history of hypertension, diabetes, high cholesterol levels, body mass index, change in weight from age 18 years to baseline (1980), alcohol intake, physical activity, previous use of oral contraceptives, postmenopausal estrogen therapy use and menopausal status, parental history of myocardial infarction at age 65 years or younger, cigarettes smoked per day during the period prior to quitting, and age at starting smoking.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kenfield SA, Stampfer MJ, Rosner BA, Colditz GA. Smoking and Smoking Cessation in Relation to Mortality in Women. JAMA. 2008;299(17):2037–2047. doi:10.1001/jama.299.17.2037
Author Affiliations: Departments of Epidemiology (Drs Kenfield and Stampfer) and Biostatistics (Dr Rosner), Harvard School of Public Health, and Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School (Drs Kenfield, Stampfer, and Rosner), Boston, Massachusetts; and Department of Surgery, Washington University School of Medicine, St Louis, Missouri (Dr Colditz).
Context Smoking is associated with an increased risk of total and cause-specific death, but the rate of mortality risk reduction after quitting compared with continuing to smoke is uncertain. There is inadequate or insufficient evidence to infer the presence or absence of a causal relationship between smoking and ovarian cancer and colorectal cancer.
Objective To assess the relationship between cigarette smoking and smoking cessation on total and cause-specific mortality in women.
Design, Setting, and Participants Prospective observational study of 104 519 female participants in the Nurses' Health Study with follow-up from 1980 to 2004.
Main Outcome Measure Hazard ratios (HRs) for total mortality, further categorized into vascular and respiratory diseases, lung cancer, other cancers, and other causes.
Results A total of 12 483 deaths occurred in this cohort, 4485 (35.9%) among never smokers, 3602 (28.9%) among current smokers, and 4396 (35.2%) among past smokers. Compared with never smokers, current smokers had an increased risk of total mortality (HR, 2.81; 95% confidence interval [CI], 2.68-2.95) and all major cause-specific mortality. The HR for cancers classified by the 2004 surgeon general's report to be smoking-related was 7.25 (95% CI, 6.43-8.18) and 1.58 (95% CI, 1.45-1.73) for other cancers. Compared with never smokers, the HR for colorectal cancer was 1.63 (95% CI, 1.29-2.05) for current smokers and 1.23 (95% CI, 1.02-1.49) for former smokers. A significant association was not observed for ovarian cancer. Significant trends were observed for earlier age at initiation of smoking for total mortality (P = .003), respiratory disease mortality (P = .001), and all smoking-related cancer mortality (P = .001). The excess risk for all-cause mortality decreases to the level of a never smoker 20 years after quitting, with different time frames for risk reduction observed across outcomes. Approximately 64% of deaths among current smokers and 28% of deaths among former smokers were attributable to cigarette smoking.
Conclusions Most of the excess risk of vascular mortality due to smoking in women may be eliminated rapidly upon cessation and within 20 years for lung diseases. Postponing the age of smoking initiation reduces the risk of respiratory disease, lung cancer, and other smoking-related cancer deaths but has little effect on other cause-specific mortality. These data suggest that smoking is associated with an increased risk of colorectal cancer mortality but not ovarian cancer mortality.
Tobacco use remains the leading preventable cause of death in the United States.1 Globally, approximately 5 million premature deaths were attributable to smoking in 2000.2 The World Health Organization projects by 2030 that tobacco-attributable deaths will annually account for 3 million deaths in industrialized countries and 7 million in developing countries.2,3 The hazards of smoking have been documented over the past 55 years, providing sufficient evidence of a causal relationship between smoking and many types of death.
After 12 years of follow-up in the Nurses' Health Study, Kawachi et al4 described the extent of the increase in risk of total and cause-specific mortality (vascular disease, cancer including and excluding lung cancer, and external causes of death) associated with current smoking and early age at smoking initiation compared with never smokers, as well as the decrease in risk associated with smoking cessation compared with continuing smokers. We have continued to follow up these women over time. With 22 years of follow-up, we are now able to characterize the relationship of smoking with other causes of death, including respiratory diseases and cancers.
The Nurses' Health Study cohort was established in 1976 when 121 700 female US registered nurses aged 30 to 55 years residing in 11 states completed a mailed questionnaire. Participants provided detailed information about medical history and risk factors for cancer, heart disease, and other diseases.5 Since 1976, this information has been updated and extended on biennial follow-up questionnaires. This study was approved by the Partners Human Research Committee (Boston, Massachusetts); completion of the self-administered questionnaire was considered to imply informed consent.
On the initial 1976 questionnaire, participants reported whether they currently smoked or had ever smoked in the past and the age at which they started smoking. Current smokers reported the number of cigarettes smoked per day, and past smokers reported the age at which they stopped smoking and the number of cigarettes smoked per day before quitting. On each subsequent biennial questionnaire, participants reported whether they currently smoked cigarettes, and at the start of each 2-year follow-up cycle, were reclassified by smoking status (never, past, or current), by quantity of cigarettes smoked and duration among current smokers, and by time since quitting among former smokers. For the analysis on current smoking, current smokers were classified into categories of cigarettes smoked per day of 1 to 14, 15 to 24, 25 to 34, and 35 or more and categories of age at initiation of smoking of 17 years or younger, 18 to 21 years, 22 to 25 years, and 26 years or older. For the smoking cessation analysis, past smokers were classified into categories of time since quitting: less than 5 years, 5 to less than 10 years, 10 to less than 15 years, 15 to less than 20 years, and 20 or more years.
The main outcome was death from all causes, occurring after the 1980 questionnaire was returned but before June 1, 2004. Deaths were grouped into 6 broad categories: vascular diseases; respiratory diseases; lung cancer; all smoking-related cancers (cancers denoted by the 2004 surgeon general's report to be caused by smoking6) including those of the lip, mouth, pharynx, esophagus, larynx, pancreas, bladder and kidney, cervix, stomach, trachea and lung, and acute myeloid leukemia; other cancers; and other causes. Vascular deaths were further categorized into deaths due to coronary heart disease and cerebrovascular disease. Respiratory deaths were further categorized into those due to chronic obstructive pulmonary disease (COPD), as well as a broader category that included COPD as well as nonspecified diseases of the respiratory system (this grouping may include some participants with COPD who were not diagnosed).
Deaths were usually reported by families, and deaths among nonrespondents were identified by searching the National Death Index.7 The cause of death was ascertained and the pertinent medical records were obtained. Study physicians reviewed these records and the death certificate to classify individual causes of death. Cause of death was based on death certificate information only for 6.1% of deaths.
Person-years of follow-up accrued from the date of return of the 1980 questionnaire until either the date of death or the end of follow-up (June 1, 2004), whichever came first. We started follow-up in 1980 because alcohol use and physical activity were not ascertained until this follow-up cycle. Person-time for each 2-year follow-up period was equal to the number of months between the return of successive questionnaires. Women did not contribute person-time in follow-up cycles in which they were missing smoking data.
We evaluated the effect of cigarettes smoked per day, age at starting smoking, and time since quitting smoking on total and cause-specific mortality. We decided not to present data on pack-years because this variable combines cigarettes smoked per day and duration of smoking. This results in women with widely different doses and durations being classified as having the same pack-years exposure, and may misrepresent the 2 terms included in this summary measure.8 (The findings for pack-years of smoking in relation to mortality are available from the authors on request.) We chose to evaluate cancers not classified by the 2004 surgeon general's report to be smoking-related if more than 300 cancer-specific deaths were available. Lastly, we evaluated birth cohort effects in our population by evaluating the hazard ratios (HRs) for those born between 1920-1929 and 1930-1939, excluding deaths before age 56 years. We used never smokers as the reference group for the analyses evaluating the HRs for cigarettes smoked per day and age at initiation among current smokers, and current smokers as the reference group for the analysis evaluating the HRs for time since quitting among former smokers.9
For all analyses, we used Cox proportional hazard models conditioned on age in months and follow-up cycle. All multivariate models included history of hypertension, diabetes, and high cholesterol levels, body mass index calculated as weight in kilograms divided by height in meters squared, change in weight from age 18 years to baseline, alcohol intake (categories of nondrinkers and drinkers of 0.1-4.9, 5.0-14.9, and ≥15.0 g/d), physical activity (quintiles based on intensity level and a metabolic equivalent task value calculated from the specific activity engaged in most frequently (1980-1984) and metabolic equivalent task hours per week (1986-2000), previous use of oral contraceptives (never, past, or current), postmenopausal estrogen therapy use (never, past, or current) and menopausal status, and parental history of myocardial infarction at age 65 years or younger. We also adjusted for servings of beef, pork, lamb, or processed meat, total calcium and folate intake, and duration of aspirin use when evaluating the relationship between smoking and colorectal cancer mortality. All variables except height were updated biennially until diagnosis of a nonfatal disease.
Tests for linear trend were calculated, using the Wald test, excluding the reference category and using the midpoint of the categories for cigarettes smoked per day and years since quitting, and the reported age at which the participant started smoking. The exposed attributable fraction was calculated using the HRs for current or former smokers compared with never smokers.
For all analyses, we excluded participants with a prior history of cancer (other than non–melanoma skin cancer), vascular disease, or respiratory disease before baseline. We also excluded those participants (n = 1872) who had smoked but did not provide their age at smoking initiation. There were 104 519 participants included in the analyses of number of cigarettes smoked per day and smoking cessation, and 79 172 participants included in the analyses of age at initiation of smoking because this analysis included only current and never smokers. All analyses were conducted using SAS software version 9 (SAS Institute Inc, Cary, North Carolina). All P values were based on 2-sided tests and were considered statistically significant at P ≤ .05.
Nonfatal diseases that occur during follow-up may affect subsequent smoking and act as an intermediate variable between smoking and mortality. For example, a person experiencing a nonfatal myocardial infarction may reduce their smoking or quit smoking altogether,9,10 and this myocardial infarction (partly induced by smoking) may increase her risk of death. The extent of this type of confounding was evaluated in the first report on smoking and coronary heart disease mortality by performing the G-computational algorithm,11,12 and the risk estimates were identical to the crude estimates of risk. In the analyses presented herein, we attempted to address the problem of confounding by nonfatal diseases by stopping the updating of smoking and all covariates for those participants developing any of the following nonfatal diseases: vascular disease, cancer, or respiratory disease. We used the covariate information provided in the period prior to diagnosis of the nonfatal disease in all subsequent follow-up periods for these participants.
A total of 12 483 deaths occurred in this cohort, 4485 (35.9%) among never smokers, 3602 (28.9%) among current smokers, and 4396 (35.2%) among past smokers. There were 2957 vascular deaths with 1385 deaths due to coronary heart disease and 734 due to cerebrovascular disease, 759 respiratory deaths with 163 due to COPD, 1237 lung cancer deaths, 2104 smoking-related cancer deaths (including lung cancer), 3805 deaths due to other cancers, and 2858 deaths due to other causes. Age-standardized characteristics at baseline are presented by smoking status (Table 1). In 1980, 28% of participants were current smokers, 26% were past smokers, and 46% were never smokers. In 2002, only 8% of those alive were current smokers. Current smokers had less increase in weight since age 18 years, slightly less hypertension and a lower body mass index, more alcohol use (≥15 g/d), and less vigorous weekly exercise than former or never smokers.
Compared with never smokers, current smokers had an increased risk of dying from any major cause during follow-up (Table 2). Risks increased significantly with number of cigarettes smoked per day for all major causes (except for cerebrovascular deaths; P for trend = .08). The strongest associations for the category of 35 or more cigarettes smoked per day were for COPD (HR, 114.55; 95% confidence interval [CI], 42.81-306.54) and lung cancer (HR, 39.88; 95% CI, 30.14-52.78). Multivariate-adjusted risk estimates for current smoking for cancers included in the all smoking-related cancers group (in addition to lung cancer) are as follows: acute myeloid leukemia (n = 202; HR, 1.72 [95% CI, 1.15-2.58]); bladder and kidney (n = 184; HR, 2.97 [95% CI, 2.00-4.42]); cervix (n = 29; HR, 10.18 [95% CI, 3.46-29.93]); esophagus (n = 44; HR, 7.03 [95% CI, 2.96-16.69]); lip and mouth (n = 30; HR, 4.72 [95% CI, 1.84-12.13]); pharynx (n = 23; HR, 6.01 [95% CI, 1.75-20.69]); pancreas (n = 383; HR, 1.84 [95% CI, 1.39-2.43]); and stomach (n = 108; HR, 1.59 [95% CI, 0.96-2.64]). Only 9 deaths were due to laryngeal cancer, and all were among current smokers.
For the relationship between smoking and cancer mortality, sites with more than 300 cancer-specific deaths and sites not previously studied were explored in this cohort. There were 578 colorectal cancer deaths, 467 ovarian cancer deaths, and 1138 breast cancer deaths; however, a recent Nurses' Health Study analysis found no relationship between current smoking and breast cancer survival among those with breast cancer (relative risk, 1.00; 95% CI, 0.83-1.19).13 Current smokers had an increased risk of colorectal cancer mortality (HR, 1.63; 95% CI, 1.29-2.05) and a slightly elevated risk of ovarian cancer (HR, 1.20; 95% CI, 0.92-1.56) compared with never smokers, but a significant trend for cigarettes smoked per day for colorectal and ovarian cancer mortality was not observed. Overall, approximately 64% of all deaths among current smokers were attributable to cigarette smoking; specifically, 69% of vascular deaths, 90% of respiratory deaths, 95% of lung cancer deaths, 86% of lung and other smoking-related cancer deaths, 37% of other cancer deaths, and 47% of other deaths were attributable to current cigarette smoking.
The HR for total mortality for current smokers who started smoking at age 17 years or younger was 2.93 (95% CI, 2.70-3.18), 22% higher than for those starting at or after 26 years (HR, 2.40 [95% CI, 2.08-2.78]; Table 3). For vascular disease and its subgroups, the HRs for age at starting smoking did not change significantly with increasing age (P = .84), while a significant downward trend was observed with increasing age for respiratory disease (P = .001), lung cancer (P < .001), and smoking-related cancer mortality (P = .001).
The hazards for total mortality in both birth cohorts were similar but because the mean age at smoking initiation was 19.9 years for those born in the 1920s and 19.3 years for those born in the 1930s, we did not expect to observe differences by birth cohort.
We observed a significant 13% reduction in the risk of all-cause mortality within the first 5 years of quitting smoking compared with continuing to smoke, and the excess risk decreased to the level of a never smoker 20 years after quitting (Table 4 and Figure), with some causes taking more or less time. Significant trends were observed with increasing years since quitting for all major cause-specific outcomes. A more rapid decline in risk after quitting smoking compared with continuing to smoke was observed in the first 5 years for vascular diseases compared with other causes. Much of the reduction in the excess risk for these causes of death were realized within the first 5 years for coronary heart disease and cerebrovascular disease. Sixty-one percent of the full potential benefit of quitting in regard to coronary heart disease mortality and 42% of the full potential benefit of quitting in regard to cerebrovascular mortality was realized within the first 5 years of quitting smoking, when comparing HRs for recent quitters of less than 5 years with long-term quitters of 20 years or greater. For death due to respiratory disease, an 18% reduction in risk of death was observed 5 to 10 years after quitting smoking, with the risk reaching that of a never smoker's risk after 20 years. This time frame for risk reduction was similar to that observed for COPD. For lung cancer mortality, a significant 21% reduction in risk was observed within the first 5 years compared with continuing smokers, but the excess risk did not disappear for 30 years. Past smokers with 20 to less than 30 years of cessation had an 87% reduction in risk of lung cancer mortality compared with current smokers (HR, 0.13; 95% CI, 0.10-0.18) while those with 30 or more years of cessation had a 93% reduction in risk (HR, 0.07; 95% CI, 0.05-0.10). When including the other smoking-related cancers, the excess risk approached a never smoker's risk more than 20 years after quitting smoking. Although the test for trend was not significant, quitting for more than 20 years was associated with a significant 30% reduction in colorectal cancer mortality (HR, 0.70; 95% CI, 0.53-0.93) compared with continuing to smoke. Approximately 28% of all deaths among past smokers were attributable to cigarette smoking; specifically, 24% of vascular deaths, 75% of respiratory deaths, 81% of lung cancer deaths, 57% of lung and other smoking-related cancer deaths, 8% of other cancer deaths, and 21% of other deaths were attributable to former cigarette smoking.
This report adds to the growing evidence on the relationship between smoking and mortality.14,15 The original report from the Nurses' Health Study on smoking and cause-specific mortality included 2847 deaths and evaluated the 5 mortality-specific outcomes: total mortality, total cardiovascular diseases, total cancer including and excluding lung cancer, and external causes of injury. This updated report on smoking and mortality in the Nurses' Health Study cohort includes an additional 16 years of follow-up, 12 483 deaths, and new estimates for coronary heart disease, cerebrovascular disease, respiratory disease, COPD, lung cancer, smoking-related cancers, colorectal cancer, ovarian cancer, and other causes. Because smoking behavior changes over time, updating participants' smoking status every 2 years enables more accurate evaluation of the detrimental effects from long-term smoking and the risk reduction over time from sustained cessation. The 9636 additional deaths that have accrued over time also allow for better precision in estimating the extent of risks associated with smoking and smoking cessation on causes of death previously studied.
As expected, smoking was associated with an increased risk of cause-specific mortality, with HRs 8 to 14 times higher for lung cancer mortality and COPD mortality compared with total mortality. The relationship between an increasing risk of death with increasing number of cigarettes smoked per day varied by outcome. The trend was less pronounced for deaths due to vascular disease, suggesting that the first few cigarettes account for most of the increased risk; in contrast, an increase in the number of cigarettes smoked per day substantially increased the risk of death from respiratory disease.
Cohort studies consistently support an increased risk of colorectal cancer associated with current smoking, but only after accounting for an induction period of 30 to 40 years.16 Our mortality estimates are higher for current smoking and similar for former smoking compared with estimates from the American Cancer Society Cancer Prevention Study II, which reported HRs of 1.41 (95% CI, 1.26-1.58) and 1.22 (95% CI, 1.09-1.37) for current and past smoking status, respectively, among women.17 The 2004 Surgeon General's report concluded that the evidence is suggestive but not sufficient to infer a causal relationship between smoking and colorectal cancer,6 mainly because of the possibility that the higher death rates from colorectal cancer may be due to less screening in smokers and a later stage of disease at diagnosis. However, we observed only modest differences in colorectal cancer screening in our cohort. In 1992, 5% of smokers reported screening by sigmoidoscopy and 31% by the stool occult blood test in the past 2 years, vs 10% and 42%, respectively, for past smokers, and 8% and 38%, respectively, for never smokers. There was a small difference in the percentage of never smokers compared with past and current smokers who had an advanced stage of colorectal cancer at diagnosis. It is unlikely that these small differences in screening and stage at diagnosis explain the smoking and colorectal cancer mortality relationship.
The 2004 Surgeon General's report also concluded that the evidence was inadequate to infer a causal relationship between smoking and ovarian cancer.6 Although we observed a positive but nonsignificant relationship between current smoking and ovarian cancer morality, we found no significant trend with increasing cigarettes smoked per day, and age of smoking initiation, nor an association between smoking cessation and ovarian cancer mortality, even 20 years after quitting. Previous studies suggest an increased risk of ovarian cancer incidence associated with current smoking for mucinous epithelial tumors.18,19
Smoking cessation was beneficial for each cause-specific mortality outcome examined. Unlike the Cancer Prevention Study I, which did not update smoking status during follow-up and found that the risks associated with lung cancer and COPD mortality remained even after 20 years, we observed a monotonic decrease in risk compared with current smoking with increasing years of smoking cessation, with risks equivalent to those of never smokers after 20 years for COPD and after 30 years for lung cancer.20 By stopping the updating of covariates after diagnosis, we minimized the bias due to symptom-induced smoking cessation or reducing smoking levels (the “ill quitter” effect). Inability of other studies to update smoking exposure over time or use smoking information just before diagnoses may obscure the harms of continuing smoking and the benefits of cessation because current smokers may quit smoking over time and some past smokers may resume smoking.
In the British Doctors Study, men born in the 1920s were assumed to have more intense early cigarette exposure than earlier birth cohorts, and coupled with improvements in treatment, an estimated two-thirds of those persistent smokers were likely to die from smoking.21 We did not see differences in early cigarette exposure among women born between 1920 and 1929 compared with women born between 1930 and 1939, translating into similar hazards for total mortality in both groups. However, youth are starting to smoke at younger ages, and one national survey found that 13% of eighth grade students first smoked by age 11 years,22 and 22% of all high school students reported being current smokers.23 It is likely that deaths attributable to smoking will increase over time unless there is a substantial increase in cessation.
In summary, our findings indicate that 64% of deaths in current smokers and 28% of deaths in past smokers are attributable to smoking. Quitting reduces the excess mortality rates for all major causes of death examined. Most of the excess risk of vascular mortality due to smoking may be eliminated rapidly upon cessation and within 20 years for lung diseases, in which the damaging effects of smoking are greatest. Early age at initiation is associated with an increased mortality risk so implementing and maintaining school tobacco prevention programs, in addition to enforcing youth access laws, are key preventive strategies.24,25 Effectively communicating risks to smokers and helping them quit successfully should be an integral part of public health programs.
Corresponding Author: Stacey A. Kenfield, ScD, Channing Laboratory, 181 Longwood Ave, Room 452, Boston, MA 02115 (firstname.lastname@example.org).
Author Contributions: Drs Kenfield and Colditz had full access to all of the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.
Study concept and design: Kenfield, Stampfer, Rosner, Colditz.
Acquisition of data: Kenfield, Stampfer, Colditz.
Analysis and interpretation of data: Kenfield, Stampfer, Rosner, Colditz.
Drafting of the manuscript: Kenfield.
Critical revision of the manuscript for important intellectual content: Kenfield, Stampfer, Rosner, Colditz.
Statistical analysis: Kenfield, Rosner.
Obtained funding: Stampfer, Colditz.
Administrative, technical, or material support: Colditz.
Study supervision: Stampfer, Colditz.
Financial Disclosures: None reported.
Funding/Support: The project was supported by grants R25CA098566 and T32CA09001 from the National Institutes of Health, the Association of Schools of Public Health, and the Legacy Foundation.
Role of the Sponsor: The funding sources had no role in the design or conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval or the manuscript.
Disclaimer: The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute of the National Institutes of Health, the Association of Schools of Public Health, the Legacy Foundation, the Legacy Foundation staff, or the Legacy Foundation's Board of Directors.
Additional Contributions: We thank the participants and staff of the Nurses' Health Study for their valuable contributions. Special thanks to Weiliang Qiu, PhD (Harvard Medical School and Brigham and Women's Hospital, Boston, Massachusetts) for his graphical expertise. Dr Qiu was not compensated for his contribution.