Annals of Epidemiology xxx (2014) 1e4
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Original article
What proportion of cancer deaths in the contemporary United States is attributable to cigarette smoking? Eric J. Jacobs PhD a, *, Christina C. Newton MSPH a, Brian D. Carter MPH a, Diane Feskanich ScD b, Neal D. Freedman PhD c, Ross L. Prentice PhD d, W. Dana Flanders MD, ScD a, e a
Epidemiology Research Program, American Cancer Society, Atlanta, GA Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD d Division of Public Health Sciences, Fred Hutchison Cancer Research Center, Seattle, WA e Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 7 November 2014 Accepted 9 November 2014
Purpose: The proportion of cancer deaths in the contemporary United States caused by cigarette smoking (the population attributable fraction [PAF]) is not well documented. Methods: The PAF of all cancer deaths due to active cigarette smoking among adults 35 years and older in the United States in 2010 was calculated using age- and sex-specific smoking prevalence from the National Health Interview Survey (NHIS) and age- and sex-specific relative risks from the Cancer Prevention Study-II (for ages 35e54 years) and from the Pooled Contemporary Cohort data set (for ages 55 years and older). Results: The PAF for active cigarette smoking was 28.7% when estimated conservatively, including only deaths from the 12 cancers currently formally established as caused by smoking by the US Surgeon General. The PAF was 31.7% when estimated more comprehensively, including excess deaths from all cancers. These estimates do not include additional potential cancer deaths from environmental tobacco smoke or other type of tobacco use such as cigars, pipes, or smokeless tobacco. Conclusions: Cigarette smoking causes a large proportion of cancer deaths in the contemporary United States. Reducing smoking prevalence as rapidly as possible should be a top priority for the US public health efforts to prevent cancer deaths. Ó 2014 Elsevier Inc. All rights reserved.
Keywords: Tobacco Smoking Cigarettes Cancer Mortality Population attributable fraction
Introduction In 1981, Doll and Peto [1] calculated that 30% of all cancer deaths in the United States were caused by smoking. This estimate of the population attributable fraction (PAF) was derived by calculating the difference between the observed number of deaths from smoking-related cancers in the United States in 1978 and the expected number of deaths from these cancers in the absence of smoking, estimated from mortality rates among never smokers in the American Cancer Society’s Cancer Prevention Study I. To our knowledge, the PAF for cigarette smoking and cancer mortality in the United States has not been explicitly estimated since Doll and Peto’s report [1]. A 2004 report by the US Surgeon General [2], however, has been cited as supporting a PAF of approximately 30% [3]. Although the 2004 Surgeon General’s report
* Corresponding author. Epidemiology Research Program, American Cancer Society, National Home Office, 250 Williams Street, Atlanta, GA 30303-1002. Tel.: þ1 (404) 329-7916. E-mail address: [email protected] (E.J. Jacobs).
did not explicitly estimate the proportion of cancer deaths caused by smoking, it did present an estimate of the annual number of cancer deaths caused by cigarette smoking from 1995 to 1999, which divided by the annual number of cancer deaths in the United States during that period [4] is indeed approximately 30%. The PAF in the contemporary United States may differ from earlier estimates because of declines in smoking prevalence, the addition of new cancers to the list of those established as caused by smoking [5] and increases in the absolute risk of lung cancer associated with smoking among women [6]. The purpose of our current analysis was to provide a well-documented estimate of the PAF for cigarette smoking and cancer mortality in the contemporary United States, using the most recent data available. Methods Data sources We estimated the age- and sex-adjusted PAF for cigarette smoking and cancer mortality in the United States in 2010 using the
http://dx.doi.org/10.1016/j.annepidem.2014.11.008 1047-2797/Ó 2014 Elsevier Inc. All rights reserved.
Please cite this article in press as: Jacobs EJ, et al., What proportion of cancer deaths in the contemporary United States is attributable to cigarette smoking?, Annals of Epidemiology (2014), http://dx.doi.org/10.1016/j.annepidem.2014.11.008
2
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
weighted-sum method [7]. This method required three types of information (1) smoking prevalence within age- and sex-specific groups in the United States, (2) relative risks (RRs) for smoking within age- and sex-specific groups, and (3) counts of all cancer deaths within age- and sex-specific groups in the United States. The source of each of these types of information is described in the following. The prevalence of cigarette smoking within sex-specific 5-year age groups was obtained from the 2010 National Health Interview Survey (NHIS) public use data file [8]. The NHIS is an in-person survey of the civilian noninstitutionalized US population conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention [9]. A representative sample of households is selected by a multistage cluster sample design, and US Census Bureau interviewers visit each selected household to administer the survey. The 2010 NHIS included more than 18,000 respondents older than 35 years. RRs for former and current cigarette smoking compared with never cigarette smoking among men and women aged 55 years and older were calculated from new analyses of the Pooled Contemporary Cohort (PCC) data set [6]. The PCC data set, currently maintained at the American Cancer Society, includes study participants drawn from five large contemporary US cohorts; the Cancer Prevention Study-II Nutrition Cohort (CPS-II), the Nurses’ Health Study-I, the Health Professionals Follow-up Study, the Women’s Health Initiative, and the NIH-AARP Diet and Health Study. Details of each cohort’s enrollment, follow-up, and smoking assessments have been previously reported [6]. The PCC study population used in this analysis included 421,378 men and 532,651 women followed from 2000 to 2011. This PCC population used in this analysis is similar to that used in a previous analysis [6] although some additional deaths and follow-up time have now been added. Analyses of the PCC data set were limited to participants aged 55 years or older because there were too few younger participants to reliably estimate RRs. RRs within each sex-specific 5-year age group (55e59, 60e64, 65e69, 70e74, 75e79, 80e84, and 85 years) were estimated in the PCC data set using a Cox proportional hazards model stratified on study cohort [10]. As previously described, smoking status was updated during follow-up using smoking information reported on serial follow-up questionnaires [6]. RRs for former and current cigarette smoking compared with never cigarette smoking among men and women aged 35 to 54 years were estimated from new analyses of the first 6 years of follow-up of the American Cancer Society’s Cancer Prevention Study II (CPS-II) cohort (1982e1988). Smoking status was not updated in CPS-II. As in previous analyses [5], only the first 6 years of follow-up were used to minimize misclassification of smoking status over time. Analyses used here include the same individuals (159,818 men and 283,142 women) included in analyses presented in the 2014 Surgeon General’s report [5]. Too few cancer deaths were available to estimate separate RRs for men and women aged 35 to 39 years and 40 to 44 years. Therefore, we estimated RRs for ages 35 to 49 years combined, adjusted for age in 5-year categories. In both the PCC and CPS-II data sets, RRs were calculated for a composite outcome of mortality from the 12 cancers considered established as caused by smoking in the 2014 Surgeon General’s report [5,11] (lip, oral cavity, pharynx [C00eC14], esophagus [C15], stomach [C16], colon and rectum [C18eC20], liver [C22], pancreas [C25], larynx [C32], trachea, lung, bronchus [C33eC34], cervix uteri [C53], kidney and renal pelvis [C64eC65], urinary bladder [C67], and acute myeloid leukemia [C92.0]). In addition, RRs were also calculated for a second composite outcome of mortality from any cancer other than the 12 established smokingattributable cancers.
Counts of cancer deaths within age- and sex-specific groups in the United States during 2010 (the most recent year for which data were available) were obtained from the NCHS through the online CDC WONDER database [12]. Calculation of the PAF Estimates of PAF can be biased if strong confounding factors are present even when unconfounded adjusted RRs are used [13]. To avoid this potential bias, we calculated PAF within sex-specific 5year age strata, using RRs adjusted only for cohort, and then used the weighted-sum method to obtain an overall estimate of the PAF [7]. We first calculated the PAF conservatively, including only deaths from the 12 established smoking-attributable cancers. We used the standard formula for a multicategory exposure [14] to calculate the PAF within each sex-specific 5-year age strata (denoted by subscript s):
PAFs ¼
i. p0;s þ p1;s RR1;s þ p2;s RR2;s 1 h i p0;s þ p1;s RR1;s þ p2;s RR2;s ;
h
where p0 is the proportion of never smokers, p1 is the proportion of former smokers, p2 is the proportion of current smokers, RR1 is the relative risk for former smokers compared with never smokers, and RR2 is the relative risk for current smokers compared with never smokers. We next multiplied the number of deaths from smokingattributable cancers in the United States in 2010 in each ageesex stratum by the appropriate stratum-specific PAF to obtain the number of cancer deaths caused by smoking. Finally, we summed across all ageesex strata to obtain the total number of cancer deaths caused by smoking and dividing this number by the total number of cancer deaths among adults aged 35 years or older to obtain the overall PAF. We then calculated the PAF more comprehensively, including excess mortality from all cancer sites, rather than only deaths from established smoking-attributable cancers. We used calculations similar to those described previously to obtain the number of cancer deaths within each ageesex stratum from all cancers other than the established smoking-attributable cancers, including cancers of unknown site. We then added the number of deaths from these other cancers to those previously calculated for established smoking-attributable cancers to obtain the total number of cancer deaths in each ageesex stratum. Finally, we summed across all ageesex strata to obtain overall results for cancer deaths among adults aged 35 years and older. Results When calculated conservatively, assuming effects on only the 12 established smoking-attributable cancers, the overall PAF for 2010 was 28.7% (Table 1). The PAF was somewhat higher in men (34.1%) than in women (22.8%). In both men and women, the PAF was slightly lower in younger age groups because of least partly to the higher proportion of never smokers in more recent birth cohorts. In addition, the PAF was lower among men and women older than 80 years due at least partly to the very low prevalence of current smoking among individuals who had survived to this age. When calculated more comprehensively, including excess deaths from all cancers, the PAF was slightly higher overall (31.7%) and in both men (37.6%) and women (25.3%). RRs for the composite outcome of any established smokingattributable cancer are shown in Supplementary Table 1. In
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
3
Table 1 Estimated numbers and percentages of cancer deaths in the United States in 2010 caused by cigarette smoking Smoking prevalence in 2010* Never
Men Age (y) 35e39 0.57 40e44 0.59 45e49 0.57 50e54 0.45 55e59 0.48 60e64 0.39 65e69 0.38 70e74 0.35 75e79 0.33 80e84 0.41 85 0.50 Total d Women Age (y) 35e39 0.71 40e44 0.66 45e49 0.60 50e54 0.57 55e59 0.59 60e64 0.58 65e69 0.54 70e74 0.59 75e79 0.60 80e84 0.67 85 0.75 Total d Men and women
Including only deaths from established smoking-attributable cancersy
Including excess deaths from all cancers
Former
Current
Number of cancer deaths due to smokingz
Number of all cancer deaths
Percentage of all cancer deaths due to smoking
Number of cancer deaths due to smokingz,x
Number of all cancer deaths
Percentage of all cancer deaths due to smoking
0.19 0.21 0.20 0.27 0.31 0.41 0.48 0.54 0.60 0.53 0.49 d
0.25 0.20 0.23 0.28 0.21 0.20 0.14 0.11 0.07 0.05 0.01 d
360 810 2318 7108 9542 13,461 16,212 16,447 14,877 11,923 8332 101,391
1742 3404 8345 16,843 25,848 34,852 39,048 40,637 42,186 40,743 43,893 297,541
20.7 23.8 27.8 42.2 36.9 38.6 41.5 40.5 35.3 29.3 19.0 34.1
643 1242 3199 7548 11,031 14,563 18,227 18,098 16,434 13,132 7663 111,779
1742 3404 8345 16,843 25,848 34,852 39,048 40,637 42,186 40,743 43,893 297,541
36.9 36.5 38.3 44.8 42.7 41.8 46.7 44.5 39.0 32.2 17.5 37.6
0.12 0.13 0.18 0.23 0.24 0.27 0.32 0.28 0.33 0.28 0.23 d d
0.17 0.21 0.22 0.21 0.17 0.15 0.14 0.13 0.07 0.05 0.02 d d
215 498 1327 3091 5872 7752 9300 10,242 9731 8069 5594 61,692 163,083
2228 4435 9,240 15,783 21,051 27,750 31,441 33,509 36,060 38,036 51,117 270,650 568,191
9.6 11.2 14.4 19.6 27.9 27.9 29.6 30.6 27.0 21.2 10.9 22.8 28.7
212 490 1337 3373 5952 8918 10,109 11,273 11,028 8555 7161 68,409 180,188
2228 4435 9240 15,783 21,051 27,750 31,441 33,509 36,060 38,036 51,117 270,650 568,191
9.5 11.0 14.5 21.4 28.3 32.1 32.2 33.6 30.6 22.5 14.0 25.3 31.7
Cigarette smoking prevalence from the 2010 National Health Interview Survey. Established smoking-attributable cancers defined as lip, oral cavity, pharynx (C00eC14), esophagus (C15), stomach (C16), colon and rectum (C18eC20), liver (C22), pancreas (C25), larynx (C32), trachea, lung, bronchus (C33eC34), cervix uteri (C53), kidney and renal pelvis (C64eC65), urinary bladder (C67), and acute myeloid leukemia (C92.0). z Estimated using smoking prevalence from the 2010 National Health Interview Survey and RRs derived from the Cancer Prevention Study-II cohort and the Pooled Contemporary Cohort data set. x In some ageesex strata with smaller numbers, estimated numbers of cancer deaths due to smoking are slightly smaller in analyses of all cancers than in analyses of established smoking-attributable cancers due to statistically nonsignificant RRs of less than 1.0 for other cancers. * y
sensitivity analyses, further adjustment for education and alcohol use resulted in little change in RR estimates. For example, in the PCC data set, the overall RR for current smoking and death from any smoking-attributable cancer was 6.11 (95% confidence interval, 5.89e6.34) when adjusted for age, sex, and cohort and 5.91 (95% confidence interval, 5.69e6.13) when further adjusted for education and alcohol use. RRs for cancer mortality from all other cancers, including cancers of unknown site are shown in Supplementary Table 2. Discussion This analysis used relatively recent national smoking prevalence data and RRs derived from large US cohort studies to estimate the PAF of cancer deaths due to cigarette smoking. We conservatively estimate that, including only deaths from cancers currently formally established as caused by smoking, 28.7% of all cancer deaths in the United States in 2010 were caused by cigarette smoking. Our more conservative estimate for the PAF in 2010 is similar to that which can be calculated for the period from 2005 to 2009 by combining numbers from the 2014 Surgeon General’s report [5] and mortality data available from the NCHS [12]. No estimate of PAF was presented in the Surgeon General’s report. However, according to the estimate of smoking-attributable mortality in the Surgeon General’s report [5], active cigarette smoking caused an average of
163,700 deaths from established smoking-associated cancers among adults aged 35 years and older each year from 2005 to 2009. Dividing this number by the average annual number of all cancer deaths among adults aged 35 years and older during this period according to NCHS (n ¼ 556,406) [12], the resulting PAF is 29.4%. The similarity of our PAF estimate to this simpler estimate is expected. Both PAF estimates used smoking prevalence data from NHIS, and similar, although not identical, RR estimates from CPS-II and the PCC data set. Estimates from the Surgeon General’s report were based on analyses using four broad age categories within each sex, whereas our analysis was based on a weighted-sum approach using finer ageesex strata. Although the two PAF estimates are reassuringly similar, our analysis provides full documentation, in a single source, for the PAF for 2010, the most recent year for which PAF could be calculated. We estimated the PAF in two ways: (1) conservatively, including only deaths from cancers formally established as caused by smoking and (2) more comprehensively, including excess deaths from all cancers. It is uncertain which of these two estimates is more accurate. Our more comprehensive estimate may have overestimated the PAF because of including mortality from cancers that were positively associated with smoking due only to residual confounding. However, our more conservative estimate may have underestimated the PAF because of ignoring truly causal excess mortality associated with categories of cancer not yet formally established as caused by smoking. In particular, cancers of unknown site were not included in
4
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
our conservative estimate but accounted for 2% to 3% of all cancer deaths in the United States in 2010 [15]. In the PCC data set, current smoking was associated with approximately threefold increased mortality from cancers of unknown site [16]. This substantial association strongly suggests that many of the cancer deaths in the unknown site category were from cancers known to be caused by smoking. If this is true, our conservative estimate may have underestimated the PAF by failing to include deaths from cancers of unknown site. Several additional factors could have contributed to either an overestimate or an underestimate of the PAF reported here. The PAF could have been overestimated if smoking RRs were overestimated because of confounding by factors such as low education, high alcohol use, physical inactivity, or poor diet. Although information on education and alcohol use was available in the CPS-II and PCC data sets, we could not calculate the PAF adjusted for these factors using the weighted-sum method because no information on them was available in the CDC WONDER national mortality database [12]. However, meaningful confounding by education and alcohol use appears unlikely because adjustment for these factors had little effect on smoking RRs. Alternatively, the PAF could have been underestimated because of the opposing bias that occurs as a result of correlations between smoking and potential cancer mortality risk factors (e.g., low education, high alcohol use, poor diet, or physical inactivity) when these factors cannot be accounted for using the weighted-sum method [17]. Finally, the PAF could have been underestimated because of an underestimate of smoking prevalence in the National Health Interview Survey resulting from underreporting of smoking or lower response rates among smokers. We estimated the PAF only for active cigarette smoking. Our PAF estimate does not include cancer deaths caused by other types of tobacco use, such as smokeless tobacco, or smoking cigars or pipes. In addition, our PAF estimate does not include deaths caused by environmental tobacco smoke. The 2014 Surgeon General’s report estimates that environmental tobacco smoke caused an additional 7330 deaths from lung cancer annually from 2005 to 2009 [5], about 1.3% of all cancer deaths during those years [12]. We estimated PAF only among adults aged 35 years and older. We were unable to include younger adults or children because of the absence of reliable RR estimates at these ages. Our PAF estimate is nonetheless a reasonable approximation of the PAF for all cancer deaths in the United States because only about 1% of all cancer deaths occur before the age of 35 years [12]. Despite important declines in smoking prevalence, the PAF for smoking and cancer mortality estimated for 2010 in this analysis is similar to the 30% estimated by Doll and Peto [1] more than 30 years ago. Declines in smoking prevalence have undoubtedly contributed substantially to decreasing both the rate of cancer mortality attributable to smoking and the PAF. Other factors, however, have contributed to increasing the PAF estimate, including the addition of new cancers to the list of those counted as caused by smoking, increases over time in the RR of lung cancer mortality among smokers [6], and decreases in cancer mortality for reasons unrelated to smoking. Our estimate of the PAF therefore does not indicate a lack of progress in reducing cancer mortality due to smoking. However, our estimate of the PAF, the proportion of cancer deaths currently caused by cigarette smoking, is relevant for prioritizing future public health action. Our results indicate that cigarette smoking causes about three in 10 cancer deaths in the contemporary United States. Reducing
smoking prevalence as rapidly as possible should be a top priority for US public health efforts to prevent future cancer deaths. Acknowledgments Support for the National Institutes of Health (NIH)eAARP Diet and Health Study was provided by the Intramural Research Program of the National Cancer Institute, NIH. Support for the American Cancer Society (ACS) Cancer Prevention Study II Nutrition Cohort was provided by the Intramural Research Programs of the ACS. Support for the Nurses’ Health Study and the Health Professionals Follow-up Study was provided by grants (P01 CA87969 and UM1 CA167552, respectively) from the NCI. Support for the Women’s Health Initiative program is provided by contracts (N01WH22110, N01WH24152, N01WH32100e32102, N01WH32105, N01WH32106, N01WH32108, N01WH32109, N01WH32111, N01WH32112, N01WH32113, N01WH32115, N01WH32118,N01WH32119,N01WH32122,N01WH42107eN01WH42126, N01WH42129e N01WH42132, and N01WH44221) from the National Heart, Lung, and Blood Institute. References [1] 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(6):1191e308. [2] U.S. Department of Health and Human Services. The health consequences of smoking: a report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Servies, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2004. [3] American Cancer Society. Cancer facts & figures 2013. Atlanta: American Cancer Society; 2013. [4] Jemal A, Murray T, Ward E, Samuels A, Tiwari RC, Ghafoor A, et al. Cancer statistics, 2005. CA Cancer J Clin 2005;55(1):10e30. [5] U.S. Department of Health and Human Services. The health consequences of smoking - 50 years of progress. A report of the Surgeon General. Rockville, MD: U.S. Department of Health and Human Services, Public Health Service, Office of the Surgeon General; 2014. [6] Thun MJ, Carter BD, Feskanich D, Freedman ND, Prentice R, Lopez AD, et al. 50Year trends in smoking-related mortality in the United States. N Engl J Med 2013;368(4):351e64. [7] Benichou J. A review of adjusted estimators of attributable risk. Stat Methods Med Res 2001;10(3):195e216. [8] National Center for Health Statistics. 2010 National Health Interview Survey (NHIS) Public Use Data Release. Hyattsville, MD: National Center for Health Statistics, Centers for Disease Control and Prevention; 2011 [April 10, 2014]; Available from, http://www.cdc.gov/nchs/nhis/nhis_2010_data_release.htm. [9] Parsons VL, Moriarity C, Jonas K, Moore TF, Davis KE, Tompkins L. Design and estimation for the national health interview survey, 2006-2015. Vital Health Stat Ser 2 2014;(165):1e53. [10] Kleinbaum DG. Survival analysis: a self-learning text. New York: SpringerVerlag; 1996. p. 178e80. [11] World Health Organization. International Statistical Classification of Diseases and Related Health Problems, Tenth Revision. Geneva: World Health Organization; 1992. [12] United States Cancer Statistics. 1999-2010 Mortality, WONDER Online Database. United States Department of Health and Human Services, Centers for Disease Control and Prevention; 2013 [June 17, 2014]; Available from, http:// wonder.cdc.gov/CancerMort-v2010.html. [13] Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health 1998;88(1):15e9. [14] Walter SD. The estimation and interpretation of attributable risk in health research. Biometrics 1976;32(4):829e49. [15] Surveillance, Epidemiology, and End Results (SEER) Program (www.seer. cancer.gov) SEER*Stat Database: Mortality - All COD, Aggregated With State, Total U.S. (1969-2010) , National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch, released April 2013. Underlying mortality data provided by NCHS (www.cdc.gov/nchs). [database on the Internet]. [16] Carter BD, Abnet CC, Feskanich D, Freedman ND, Hartge P, Lewis CE, et al. Smoking and mortality in a large pooled contemporary cohort: beyond established causes. unpublished results. [17] Darrow LA, Steenland NK. Confounding and bias in the attributable fraction. Epidemiology 2011;22(1):53e8.
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
4.e1
Appendix
Supplementary Table 1 RR for death from any established smoking-attributable cancer by cigarette smoking status* RR (95% confidence interval) Cigarette smoking status
Men Age (y) 35e49y 50e54y 55e59z 60e64z 65e69z 70e74z 75e79z 80e84z 85z Women Age (y) 35e49y 50e54y 55e59z 60e64z 65e69z 70e74z 75e79z 80e84z 85z
Never
Former
Current
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
2.22 2.42 1.61 1.91 2.30 2.54 2.32 2.19 2.06
(1.36e3.63) (1.79e3.27) (1.14e2.29) (1.63e2.23) (2.08e2.54) (2.35e2.74) (2.16e2.49) (2.01e2.40) (1.75e2.43)
2.98 4.51 4.97 4.87 6.43 6.42 5.84 5.82 3.54
(1.89e4.71) (3.39e5.99) (3.46e7.15) (4.07e5.83) (5.71e7.24) (5.82e7.08) (5.24e6.50) (4.93e6.87) (2.15e5.82)
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
1.41 1.75 2.07 2.26 2.19 2.50 2.42 2.27 1.77
(0.97e2.06) (1.31e2.33) (1.43e2.99) (1.91e2.67) (1.97e2.43) (2.31e2.70) (2.25e2.60) (2.08e2.48) (1.51e2.08)
2.66 3.33 6.47 5.97 6.30 6.21 6.25 5.93 5.11
(1.95e3.64) (2.61e4.24) (4.46e9.39) (4.96e7.18) (5.59e7.09) (5.63e6.85) (5.65e6.92) (5.11e6.88) (3.61e7.24)
* Established smoking-attributable cancers defined as lip, oral cavity, pharynx (C00eC14), esophagus (C15), stomach (C16), colon and rectum (C18eC20), liver (C22), pancreas (C25), larynx (C32), trachea, lung, bronchus (C33eC34), cervix uteri (C53), kidney and renal pelvis (C64eC65), urinary bladder (C67), and acute myeloid leukemia (C92.0). y RRs for ages 35 to 54 years are from the Cancer Prevention Study II (1982e1988). RRs for ages 35 to 49 years are age adjusted and combined due to small numbers. z RRs for age 55 years and older from the Pooled Contemporary Cohort (PCC) data set, which includes the Cancer Prevention Study-II Nutrition Cohort (CPS-II), the NIH-AARP Diet and Health Study, the Nurses’ Health Study-I, the Health Professionals Follow-up Study, and the Women’s Health Initiative (WHI). RRs are adjusted for cohort.
Supplementary Table 2 RR for death from cancers other than those formally established as smoking-attributable, including cancers of unknown site, by cigarette smoking status* RR (95% confidence interval) Cigarette smoking status
Men Age 35e49y 50e54y 55e59y 60e64z 65e69z 70e74z 75e79z 80e84z 85z Women Age (y) 35e49y 50e54y 55e59y 60e64z 65e69z 70e74z 75e79z 80e84z 85z
Never
Former
Current
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
2.46 1.02 1.22 1.12 1.23 1.13 1.15 1.15 0.95
(1.46e4.13) (0.72e1.47) (0.78e1.90) (0.94e1.33) (1.10e1.37) (1.05e1.22) (1.07e1.23) (1.05e1.24) (0.82e1.09)
1.82 1.37 1.97 1.39 1.80 1.76 1.43 1.17 0.48
(1.07e3.08) (0.97e1.92) (1.12e3.44) (1.07e1.80) (1.51e2.14) (1.54e2.01) (1.21e1.70) (0.88e1.56) (0.18e1.29)
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
1.16 1.09 0.87 1.20 1.06 1.14 1.20 1.13 1.26
(0.93e1.45) (0.90e1.33) (0.67e1.13) (1.05e1.37) (0.98e1.16) (1.07e1.22) (1.12e1.28) (1.04e1.23) (1.08e1.48)
0.88 1.08 1.22 1.29 1.32 1.32 1.46 0.88 1.46
(0.70e1.11) (0.90e1.31) (0.87e1.70) (1.05e1.58) (1.14e1.53) (1.16e1.50) (1.26e1.69) (0.67e1.16) (0.84e2.54)
* Established smoking-attributable cancers defined as lip, oral cavity, pharynx (C00eC14), esophagus (C15), stomach (C16), colon and rectum (C18eC20), liver (C22), pancreas (C25), larynx (C32), trachea, lung, bronchus (C33eC34), cervix uteri (C53), kidney and renal pelvis (C64eC65), urinary bladder (C67), and acute myeloid leukemia (C92.0). y RRs for ages 35 to 54 years are from the Cancer Prevention Study II (1982-1988). RRs for ages 35 to 49 years are age adjusted and combined due to small numbers. z RRs for age 55 years and older are from the Pooled Contemporary Cohort (PCC) data set, which includes the Cancer Prevention Study-II Nutrition Cohort (CPS-II), the NIHAARP Diet and Health Study, the Nurses’ Health Study-I, the Health Professionals Follow-up Study, and the Women’s Health Initiative (WHI). RRs are adjusted for cohort.
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Original article
What proportion of cancer deaths in the contemporary United States is attributable to cigarette smoking? Eric J. Jacobs PhD a, *, Christina C. Newton MSPH a, Brian D. Carter MPH a, Diane Feskanich ScD b, Neal D. Freedman PhD c, Ross L. Prentice PhD d, W. Dana Flanders MD, ScD a, e a
Epidemiology Research Program, American Cancer Society, Atlanta, GA Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD d Division of Public Health Sciences, Fred Hutchison Cancer Research Center, Seattle, WA e Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 7 November 2014 Accepted 9 November 2014
Purpose: The proportion of cancer deaths in the contemporary United States caused by cigarette smoking (the population attributable fraction [PAF]) is not well documented. Methods: The PAF of all cancer deaths due to active cigarette smoking among adults 35 years and older in the United States in 2010 was calculated using age- and sex-specific smoking prevalence from the National Health Interview Survey (NHIS) and age- and sex-specific relative risks from the Cancer Prevention Study-II (for ages 35e54 years) and from the Pooled Contemporary Cohort data set (for ages 55 years and older). Results: The PAF for active cigarette smoking was 28.7% when estimated conservatively, including only deaths from the 12 cancers currently formally established as caused by smoking by the US Surgeon General. The PAF was 31.7% when estimated more comprehensively, including excess deaths from all cancers. These estimates do not include additional potential cancer deaths from environmental tobacco smoke or other type of tobacco use such as cigars, pipes, or smokeless tobacco. Conclusions: Cigarette smoking causes a large proportion of cancer deaths in the contemporary United States. Reducing smoking prevalence as rapidly as possible should be a top priority for the US public health efforts to prevent cancer deaths. Ó 2014 Elsevier Inc. All rights reserved.
Keywords: Tobacco Smoking Cigarettes Cancer Mortality Population attributable fraction
Introduction In 1981, Doll and Peto [1] calculated that 30% of all cancer deaths in the United States were caused by smoking. This estimate of the population attributable fraction (PAF) was derived by calculating the difference between the observed number of deaths from smoking-related cancers in the United States in 1978 and the expected number of deaths from these cancers in the absence of smoking, estimated from mortality rates among never smokers in the American Cancer Society’s Cancer Prevention Study I. To our knowledge, the PAF for cigarette smoking and cancer mortality in the United States has not been explicitly estimated since Doll and Peto’s report [1]. A 2004 report by the US Surgeon General [2], however, has been cited as supporting a PAF of approximately 30% [3]. Although the 2004 Surgeon General’s report
* Corresponding author. Epidemiology Research Program, American Cancer Society, National Home Office, 250 Williams Street, Atlanta, GA 30303-1002. Tel.: þ1 (404) 329-7916. E-mail address: [email protected] (E.J. Jacobs).
did not explicitly estimate the proportion of cancer deaths caused by smoking, it did present an estimate of the annual number of cancer deaths caused by cigarette smoking from 1995 to 1999, which divided by the annual number of cancer deaths in the United States during that period [4] is indeed approximately 30%. The PAF in the contemporary United States may differ from earlier estimates because of declines in smoking prevalence, the addition of new cancers to the list of those established as caused by smoking [5] and increases in the absolute risk of lung cancer associated with smoking among women [6]. The purpose of our current analysis was to provide a well-documented estimate of the PAF for cigarette smoking and cancer mortality in the contemporary United States, using the most recent data available. Methods Data sources We estimated the age- and sex-adjusted PAF for cigarette smoking and cancer mortality in the United States in 2010 using the
http://dx.doi.org/10.1016/j.annepidem.2014.11.008 1047-2797/Ó 2014 Elsevier Inc. All rights reserved.
Please cite this article in press as: Jacobs EJ, et al., What proportion of cancer deaths in the contemporary United States is attributable to cigarette smoking?, Annals of Epidemiology (2014), http://dx.doi.org/10.1016/j.annepidem.2014.11.008
2
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
weighted-sum method [7]. This method required three types of information (1) smoking prevalence within age- and sex-specific groups in the United States, (2) relative risks (RRs) for smoking within age- and sex-specific groups, and (3) counts of all cancer deaths within age- and sex-specific groups in the United States. The source of each of these types of information is described in the following. The prevalence of cigarette smoking within sex-specific 5-year age groups was obtained from the 2010 National Health Interview Survey (NHIS) public use data file [8]. The NHIS is an in-person survey of the civilian noninstitutionalized US population conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention [9]. A representative sample of households is selected by a multistage cluster sample design, and US Census Bureau interviewers visit each selected household to administer the survey. The 2010 NHIS included more than 18,000 respondents older than 35 years. RRs for former and current cigarette smoking compared with never cigarette smoking among men and women aged 55 years and older were calculated from new analyses of the Pooled Contemporary Cohort (PCC) data set [6]. The PCC data set, currently maintained at the American Cancer Society, includes study participants drawn from five large contemporary US cohorts; the Cancer Prevention Study-II Nutrition Cohort (CPS-II), the Nurses’ Health Study-I, the Health Professionals Follow-up Study, the Women’s Health Initiative, and the NIH-AARP Diet and Health Study. Details of each cohort’s enrollment, follow-up, and smoking assessments have been previously reported [6]. The PCC study population used in this analysis included 421,378 men and 532,651 women followed from 2000 to 2011. This PCC population used in this analysis is similar to that used in a previous analysis [6] although some additional deaths and follow-up time have now been added. Analyses of the PCC data set were limited to participants aged 55 years or older because there were too few younger participants to reliably estimate RRs. RRs within each sex-specific 5-year age group (55e59, 60e64, 65e69, 70e74, 75e79, 80e84, and 85 years) were estimated in the PCC data set using a Cox proportional hazards model stratified on study cohort [10]. As previously described, smoking status was updated during follow-up using smoking information reported on serial follow-up questionnaires [6]. RRs for former and current cigarette smoking compared with never cigarette smoking among men and women aged 35 to 54 years were estimated from new analyses of the first 6 years of follow-up of the American Cancer Society’s Cancer Prevention Study II (CPS-II) cohort (1982e1988). Smoking status was not updated in CPS-II. As in previous analyses [5], only the first 6 years of follow-up were used to minimize misclassification of smoking status over time. Analyses used here include the same individuals (159,818 men and 283,142 women) included in analyses presented in the 2014 Surgeon General’s report [5]. Too few cancer deaths were available to estimate separate RRs for men and women aged 35 to 39 years and 40 to 44 years. Therefore, we estimated RRs for ages 35 to 49 years combined, adjusted for age in 5-year categories. In both the PCC and CPS-II data sets, RRs were calculated for a composite outcome of mortality from the 12 cancers considered established as caused by smoking in the 2014 Surgeon General’s report [5,11] (lip, oral cavity, pharynx [C00eC14], esophagus [C15], stomach [C16], colon and rectum [C18eC20], liver [C22], pancreas [C25], larynx [C32], trachea, lung, bronchus [C33eC34], cervix uteri [C53], kidney and renal pelvis [C64eC65], urinary bladder [C67], and acute myeloid leukemia [C92.0]). In addition, RRs were also calculated for a second composite outcome of mortality from any cancer other than the 12 established smokingattributable cancers.
Counts of cancer deaths within age- and sex-specific groups in the United States during 2010 (the most recent year for which data were available) were obtained from the NCHS through the online CDC WONDER database [12]. Calculation of the PAF Estimates of PAF can be biased if strong confounding factors are present even when unconfounded adjusted RRs are used [13]. To avoid this potential bias, we calculated PAF within sex-specific 5year age strata, using RRs adjusted only for cohort, and then used the weighted-sum method to obtain an overall estimate of the PAF [7]. We first calculated the PAF conservatively, including only deaths from the 12 established smoking-attributable cancers. We used the standard formula for a multicategory exposure [14] to calculate the PAF within each sex-specific 5-year age strata (denoted by subscript s):
PAFs ¼
i. p0;s þ p1;s RR1;s þ p2;s RR2;s 1 h i p0;s þ p1;s RR1;s þ p2;s RR2;s ;
h
where p0 is the proportion of never smokers, p1 is the proportion of former smokers, p2 is the proportion of current smokers, RR1 is the relative risk for former smokers compared with never smokers, and RR2 is the relative risk for current smokers compared with never smokers. We next multiplied the number of deaths from smokingattributable cancers in the United States in 2010 in each ageesex stratum by the appropriate stratum-specific PAF to obtain the number of cancer deaths caused by smoking. Finally, we summed across all ageesex strata to obtain the total number of cancer deaths caused by smoking and dividing this number by the total number of cancer deaths among adults aged 35 years or older to obtain the overall PAF. We then calculated the PAF more comprehensively, including excess mortality from all cancer sites, rather than only deaths from established smoking-attributable cancers. We used calculations similar to those described previously to obtain the number of cancer deaths within each ageesex stratum from all cancers other than the established smoking-attributable cancers, including cancers of unknown site. We then added the number of deaths from these other cancers to those previously calculated for established smoking-attributable cancers to obtain the total number of cancer deaths in each ageesex stratum. Finally, we summed across all ageesex strata to obtain overall results for cancer deaths among adults aged 35 years and older. Results When calculated conservatively, assuming effects on only the 12 established smoking-attributable cancers, the overall PAF for 2010 was 28.7% (Table 1). The PAF was somewhat higher in men (34.1%) than in women (22.8%). In both men and women, the PAF was slightly lower in younger age groups because of least partly to the higher proportion of never smokers in more recent birth cohorts. In addition, the PAF was lower among men and women older than 80 years due at least partly to the very low prevalence of current smoking among individuals who had survived to this age. When calculated more comprehensively, including excess deaths from all cancers, the PAF was slightly higher overall (31.7%) and in both men (37.6%) and women (25.3%). RRs for the composite outcome of any established smokingattributable cancer are shown in Supplementary Table 1. In
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
3
Table 1 Estimated numbers and percentages of cancer deaths in the United States in 2010 caused by cigarette smoking Smoking prevalence in 2010* Never
Men Age (y) 35e39 0.57 40e44 0.59 45e49 0.57 50e54 0.45 55e59 0.48 60e64 0.39 65e69 0.38 70e74 0.35 75e79 0.33 80e84 0.41 85 0.50 Total d Women Age (y) 35e39 0.71 40e44 0.66 45e49 0.60 50e54 0.57 55e59 0.59 60e64 0.58 65e69 0.54 70e74 0.59 75e79 0.60 80e84 0.67 85 0.75 Total d Men and women
Including only deaths from established smoking-attributable cancersy
Including excess deaths from all cancers
Former
Current
Number of cancer deaths due to smokingz
Number of all cancer deaths
Percentage of all cancer deaths due to smoking
Number of cancer deaths due to smokingz,x
Number of all cancer deaths
Percentage of all cancer deaths due to smoking
0.19 0.21 0.20 0.27 0.31 0.41 0.48 0.54 0.60 0.53 0.49 d
0.25 0.20 0.23 0.28 0.21 0.20 0.14 0.11 0.07 0.05 0.01 d
360 810 2318 7108 9542 13,461 16,212 16,447 14,877 11,923 8332 101,391
1742 3404 8345 16,843 25,848 34,852 39,048 40,637 42,186 40,743 43,893 297,541
20.7 23.8 27.8 42.2 36.9 38.6 41.5 40.5 35.3 29.3 19.0 34.1
643 1242 3199 7548 11,031 14,563 18,227 18,098 16,434 13,132 7663 111,779
1742 3404 8345 16,843 25,848 34,852 39,048 40,637 42,186 40,743 43,893 297,541
36.9 36.5 38.3 44.8 42.7 41.8 46.7 44.5 39.0 32.2 17.5 37.6
0.12 0.13 0.18 0.23 0.24 0.27 0.32 0.28 0.33 0.28 0.23 d d
0.17 0.21 0.22 0.21 0.17 0.15 0.14 0.13 0.07 0.05 0.02 d d
215 498 1327 3091 5872 7752 9300 10,242 9731 8069 5594 61,692 163,083
2228 4435 9,240 15,783 21,051 27,750 31,441 33,509 36,060 38,036 51,117 270,650 568,191
9.6 11.2 14.4 19.6 27.9 27.9 29.6 30.6 27.0 21.2 10.9 22.8 28.7
212 490 1337 3373 5952 8918 10,109 11,273 11,028 8555 7161 68,409 180,188
2228 4435 9240 15,783 21,051 27,750 31,441 33,509 36,060 38,036 51,117 270,650 568,191
9.5 11.0 14.5 21.4 28.3 32.1 32.2 33.6 30.6 22.5 14.0 25.3 31.7
Cigarette smoking prevalence from the 2010 National Health Interview Survey. Established smoking-attributable cancers defined as lip, oral cavity, pharynx (C00eC14), esophagus (C15), stomach (C16), colon and rectum (C18eC20), liver (C22), pancreas (C25), larynx (C32), trachea, lung, bronchus (C33eC34), cervix uteri (C53), kidney and renal pelvis (C64eC65), urinary bladder (C67), and acute myeloid leukemia (C92.0). z Estimated using smoking prevalence from the 2010 National Health Interview Survey and RRs derived from the Cancer Prevention Study-II cohort and the Pooled Contemporary Cohort data set. x In some ageesex strata with smaller numbers, estimated numbers of cancer deaths due to smoking are slightly smaller in analyses of all cancers than in analyses of established smoking-attributable cancers due to statistically nonsignificant RRs of less than 1.0 for other cancers. * y
sensitivity analyses, further adjustment for education and alcohol use resulted in little change in RR estimates. For example, in the PCC data set, the overall RR for current smoking and death from any smoking-attributable cancer was 6.11 (95% confidence interval, 5.89e6.34) when adjusted for age, sex, and cohort and 5.91 (95% confidence interval, 5.69e6.13) when further adjusted for education and alcohol use. RRs for cancer mortality from all other cancers, including cancers of unknown site are shown in Supplementary Table 2. Discussion This analysis used relatively recent national smoking prevalence data and RRs derived from large US cohort studies to estimate the PAF of cancer deaths due to cigarette smoking. We conservatively estimate that, including only deaths from cancers currently formally established as caused by smoking, 28.7% of all cancer deaths in the United States in 2010 were caused by cigarette smoking. Our more conservative estimate for the PAF in 2010 is similar to that which can be calculated for the period from 2005 to 2009 by combining numbers from the 2014 Surgeon General’s report [5] and mortality data available from the NCHS [12]. No estimate of PAF was presented in the Surgeon General’s report. However, according to the estimate of smoking-attributable mortality in the Surgeon General’s report [5], active cigarette smoking caused an average of
163,700 deaths from established smoking-associated cancers among adults aged 35 years and older each year from 2005 to 2009. Dividing this number by the average annual number of all cancer deaths among adults aged 35 years and older during this period according to NCHS (n ¼ 556,406) [12], the resulting PAF is 29.4%. The similarity of our PAF estimate to this simpler estimate is expected. Both PAF estimates used smoking prevalence data from NHIS, and similar, although not identical, RR estimates from CPS-II and the PCC data set. Estimates from the Surgeon General’s report were based on analyses using four broad age categories within each sex, whereas our analysis was based on a weighted-sum approach using finer ageesex strata. Although the two PAF estimates are reassuringly similar, our analysis provides full documentation, in a single source, for the PAF for 2010, the most recent year for which PAF could be calculated. We estimated the PAF in two ways: (1) conservatively, including only deaths from cancers formally established as caused by smoking and (2) more comprehensively, including excess deaths from all cancers. It is uncertain which of these two estimates is more accurate. Our more comprehensive estimate may have overestimated the PAF because of including mortality from cancers that were positively associated with smoking due only to residual confounding. However, our more conservative estimate may have underestimated the PAF because of ignoring truly causal excess mortality associated with categories of cancer not yet formally established as caused by smoking. In particular, cancers of unknown site were not included in
4
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
our conservative estimate but accounted for 2% to 3% of all cancer deaths in the United States in 2010 [15]. In the PCC data set, current smoking was associated with approximately threefold increased mortality from cancers of unknown site [16]. This substantial association strongly suggests that many of the cancer deaths in the unknown site category were from cancers known to be caused by smoking. If this is true, our conservative estimate may have underestimated the PAF by failing to include deaths from cancers of unknown site. Several additional factors could have contributed to either an overestimate or an underestimate of the PAF reported here. The PAF could have been overestimated if smoking RRs were overestimated because of confounding by factors such as low education, high alcohol use, physical inactivity, or poor diet. Although information on education and alcohol use was available in the CPS-II and PCC data sets, we could not calculate the PAF adjusted for these factors using the weighted-sum method because no information on them was available in the CDC WONDER national mortality database [12]. However, meaningful confounding by education and alcohol use appears unlikely because adjustment for these factors had little effect on smoking RRs. Alternatively, the PAF could have been underestimated because of the opposing bias that occurs as a result of correlations between smoking and potential cancer mortality risk factors (e.g., low education, high alcohol use, poor diet, or physical inactivity) when these factors cannot be accounted for using the weighted-sum method [17]. Finally, the PAF could have been underestimated because of an underestimate of smoking prevalence in the National Health Interview Survey resulting from underreporting of smoking or lower response rates among smokers. We estimated the PAF only for active cigarette smoking. Our PAF estimate does not include cancer deaths caused by other types of tobacco use, such as smokeless tobacco, or smoking cigars or pipes. In addition, our PAF estimate does not include deaths caused by environmental tobacco smoke. The 2014 Surgeon General’s report estimates that environmental tobacco smoke caused an additional 7330 deaths from lung cancer annually from 2005 to 2009 [5], about 1.3% of all cancer deaths during those years [12]. We estimated PAF only among adults aged 35 years and older. We were unable to include younger adults or children because of the absence of reliable RR estimates at these ages. Our PAF estimate is nonetheless a reasonable approximation of the PAF for all cancer deaths in the United States because only about 1% of all cancer deaths occur before the age of 35 years [12]. Despite important declines in smoking prevalence, the PAF for smoking and cancer mortality estimated for 2010 in this analysis is similar to the 30% estimated by Doll and Peto [1] more than 30 years ago. Declines in smoking prevalence have undoubtedly contributed substantially to decreasing both the rate of cancer mortality attributable to smoking and the PAF. Other factors, however, have contributed to increasing the PAF estimate, including the addition of new cancers to the list of those counted as caused by smoking, increases over time in the RR of lung cancer mortality among smokers [6], and decreases in cancer mortality for reasons unrelated to smoking. Our estimate of the PAF therefore does not indicate a lack of progress in reducing cancer mortality due to smoking. However, our estimate of the PAF, the proportion of cancer deaths currently caused by cigarette smoking, is relevant for prioritizing future public health action. Our results indicate that cigarette smoking causes about three in 10 cancer deaths in the contemporary United States. Reducing
smoking prevalence as rapidly as possible should be a top priority for US public health efforts to prevent future cancer deaths. Acknowledgments Support for the National Institutes of Health (NIH)eAARP Diet and Health Study was provided by the Intramural Research Program of the National Cancer Institute, NIH. Support for the American Cancer Society (ACS) Cancer Prevention Study II Nutrition Cohort was provided by the Intramural Research Programs of the ACS. Support for the Nurses’ Health Study and the Health Professionals Follow-up Study was provided by grants (P01 CA87969 and UM1 CA167552, respectively) from the NCI. Support for the Women’s Health Initiative program is provided by contracts (N01WH22110, N01WH24152, N01WH32100e32102, N01WH32105, N01WH32106, N01WH32108, N01WH32109, N01WH32111, N01WH32112, N01WH32113, N01WH32115, N01WH32118,N01WH32119,N01WH32122,N01WH42107eN01WH42126, N01WH42129e N01WH42132, and N01WH44221) from the National Heart, Lung, and Blood Institute. References [1] 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(6):1191e308. [2] U.S. Department of Health and Human Services. The health consequences of smoking: a report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Servies, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2004. [3] American Cancer Society. Cancer facts & figures 2013. Atlanta: American Cancer Society; 2013. [4] Jemal A, Murray T, Ward E, Samuels A, Tiwari RC, Ghafoor A, et al. Cancer statistics, 2005. CA Cancer J Clin 2005;55(1):10e30. [5] U.S. Department of Health and Human Services. The health consequences of smoking - 50 years of progress. A report of the Surgeon General. Rockville, MD: U.S. Department of Health and Human Services, Public Health Service, Office of the Surgeon General; 2014. [6] Thun MJ, Carter BD, Feskanich D, Freedman ND, Prentice R, Lopez AD, et al. 50Year trends in smoking-related mortality in the United States. N Engl J Med 2013;368(4):351e64. [7] Benichou J. A review of adjusted estimators of attributable risk. Stat Methods Med Res 2001;10(3):195e216. [8] National Center for Health Statistics. 2010 National Health Interview Survey (NHIS) Public Use Data Release. Hyattsville, MD: National Center for Health Statistics, Centers for Disease Control and Prevention; 2011 [April 10, 2014]; Available from, http://www.cdc.gov/nchs/nhis/nhis_2010_data_release.htm. [9] Parsons VL, Moriarity C, Jonas K, Moore TF, Davis KE, Tompkins L. Design and estimation for the national health interview survey, 2006-2015. Vital Health Stat Ser 2 2014;(165):1e53. [10] Kleinbaum DG. Survival analysis: a self-learning text. New York: SpringerVerlag; 1996. p. 178e80. [11] World Health Organization. International Statistical Classification of Diseases and Related Health Problems, Tenth Revision. Geneva: World Health Organization; 1992. [12] United States Cancer Statistics. 1999-2010 Mortality, WONDER Online Database. United States Department of Health and Human Services, Centers for Disease Control and Prevention; 2013 [June 17, 2014]; Available from, http:// wonder.cdc.gov/CancerMort-v2010.html. [13] Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health 1998;88(1):15e9. [14] Walter SD. The estimation and interpretation of attributable risk in health research. Biometrics 1976;32(4):829e49. [15] Surveillance, Epidemiology, and End Results (SEER) Program (www.seer. cancer.gov) SEER*Stat Database: Mortality - All COD, Aggregated With State, Total U.S. (1969-2010) , National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch, released April 2013. Underlying mortality data provided by NCHS (www.cdc.gov/nchs). [database on the Internet]. [16] Carter BD, Abnet CC, Feskanich D, Freedman ND, Hartge P, Lewis CE, et al. Smoking and mortality in a large pooled contemporary cohort: beyond established causes. unpublished results. [17] Darrow LA, Steenland NK. Confounding and bias in the attributable fraction. Epidemiology 2011;22(1):53e8.
E.J. Jacobs et al. / Annals of Epidemiology xxx (2014) 1e4
4.e1
Appendix
Supplementary Table 1 RR for death from any established smoking-attributable cancer by cigarette smoking status* RR (95% confidence interval) Cigarette smoking status
Men Age (y) 35e49y 50e54y 55e59z 60e64z 65e69z 70e74z 75e79z 80e84z 85z Women Age (y) 35e49y 50e54y 55e59z 60e64z 65e69z 70e74z 75e79z 80e84z 85z
Never
Former
Current
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
2.22 2.42 1.61 1.91 2.30 2.54 2.32 2.19 2.06
(1.36e3.63) (1.79e3.27) (1.14e2.29) (1.63e2.23) (2.08e2.54) (2.35e2.74) (2.16e2.49) (2.01e2.40) (1.75e2.43)
2.98 4.51 4.97 4.87 6.43 6.42 5.84 5.82 3.54
(1.89e4.71) (3.39e5.99) (3.46e7.15) (4.07e5.83) (5.71e7.24) (5.82e7.08) (5.24e6.50) (4.93e6.87) (2.15e5.82)
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
1.41 1.75 2.07 2.26 2.19 2.50 2.42 2.27 1.77
(0.97e2.06) (1.31e2.33) (1.43e2.99) (1.91e2.67) (1.97e2.43) (2.31e2.70) (2.25e2.60) (2.08e2.48) (1.51e2.08)
2.66 3.33 6.47 5.97 6.30 6.21 6.25 5.93 5.11
(1.95e3.64) (2.61e4.24) (4.46e9.39) (4.96e7.18) (5.59e7.09) (5.63e6.85) (5.65e6.92) (5.11e6.88) (3.61e7.24)
* Established smoking-attributable cancers defined as lip, oral cavity, pharynx (C00eC14), esophagus (C15), stomach (C16), colon and rectum (C18eC20), liver (C22), pancreas (C25), larynx (C32), trachea, lung, bronchus (C33eC34), cervix uteri (C53), kidney and renal pelvis (C64eC65), urinary bladder (C67), and acute myeloid leukemia (C92.0). y RRs for ages 35 to 54 years are from the Cancer Prevention Study II (1982e1988). RRs for ages 35 to 49 years are age adjusted and combined due to small numbers. z RRs for age 55 years and older from the Pooled Contemporary Cohort (PCC) data set, which includes the Cancer Prevention Study-II Nutrition Cohort (CPS-II), the NIH-AARP Diet and Health Study, the Nurses’ Health Study-I, the Health Professionals Follow-up Study, and the Women’s Health Initiative (WHI). RRs are adjusted for cohort.
Supplementary Table 2 RR for death from cancers other than those formally established as smoking-attributable, including cancers of unknown site, by cigarette smoking status* RR (95% confidence interval) Cigarette smoking status
Men Age 35e49y 50e54y 55e59y 60e64z 65e69z 70e74z 75e79z 80e84z 85z Women Age (y) 35e49y 50e54y 55e59y 60e64z 65e69z 70e74z 75e79z 80e84z 85z
Never
Former
Current
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
2.46 1.02 1.22 1.12 1.23 1.13 1.15 1.15 0.95
(1.46e4.13) (0.72e1.47) (0.78e1.90) (0.94e1.33) (1.10e1.37) (1.05e1.22) (1.07e1.23) (1.05e1.24) (0.82e1.09)
1.82 1.37 1.97 1.39 1.80 1.76 1.43 1.17 0.48
(1.07e3.08) (0.97e1.92) (1.12e3.44) (1.07e1.80) (1.51e2.14) (1.54e2.01) (1.21e1.70) (0.88e1.56) (0.18e1.29)
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
(ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref) (ref)
1.16 1.09 0.87 1.20 1.06 1.14 1.20 1.13 1.26
(0.93e1.45) (0.90e1.33) (0.67e1.13) (1.05e1.37) (0.98e1.16) (1.07e1.22) (1.12e1.28) (1.04e1.23) (1.08e1.48)
0.88 1.08 1.22 1.29 1.32 1.32 1.46 0.88 1.46
(0.70e1.11) (0.90e1.31) (0.87e1.70) (1.05e1.58) (1.14e1.53) (1.16e1.50) (1.26e1.69) (0.67e1.16) (0.84e2.54)
* Established smoking-attributable cancers defined as lip, oral cavity, pharynx (C00eC14), esophagus (C15), stomach (C16), colon and rectum (C18eC20), liver (C22), pancreas (C25), larynx (C32), trachea, lung, bronchus (C33eC34), cervix uteri (C53), kidney and renal pelvis (C64eC65), urinary bladder (C67), and acute myeloid leukemia (C92.0). y RRs for ages 35 to 54 years are from the Cancer Prevention Study II (1982-1988). RRs for ages 35 to 49 years are age adjusted and combined due to small numbers. z RRs for age 55 years and older are from the Pooled Contemporary Cohort (PCC) data set, which includes the Cancer Prevention Study-II Nutrition Cohort (CPS-II), the NIHAARP Diet and Health Study, the Nurses’ Health Study-I, the Health Professionals Follow-up Study, and the Women’s Health Initiative (WHI). RRs are adjusted for cohort.
