Cigarette Smoking as a Cause of Cancers Other Than Lung Cancer

CHEST

Original Research CANCER

Cigarette Smoking as a Cause of Cancers Other Than Lung Cancer An Exploratory Study Using the Surveillance, Epidemiology, and End Results Program Gabrielle Ray, MPH; Donald E. Henson, MD; and Arnold M. Schwartz, MD, PhD, FCCP

Background: Cigarette smoking is causally related to several cancers, particularly lung cancer, yet for some cancers there are inconsistent associations. This study investigates the association of smoking with other cancers by correlating them with the regional incidence rates for lung cancer, which was used as a proxy for cigarette smoking. This ecologic approach relating cigarette smoking to cancer using a large database avoids the limitations and bias present in case-control and cohort studies. Methods: Based on the assumption that regions with a high rate of lung cancer also have a high rate of cigarette smoking, our original hypothesis is that these high-intensity regions will also have high rates of other cancers if they are associated with cigarette smoking. Linear regression and correlation analysis of regional incidence rates for lung cancer, obtained from the Surveillance, Epidemiology, and End Results (SEER) Program, were plotted with incidence rates of other cancers to determine the association between lung cancer and the other cancers. Results: Cancers that have a strong correlation with cigarette smoking in the literature also demonstrate a strong correlation with lung cancer. These cancers included urinary bladder, laryngeal, esophageal, colorectal, and kidney cancer. A number of cancers showed a weak association with cigarette smoking, such as pancreatic and liver cancer. Other cancers showed no correlation, such as ovarian and prostate cancer. Conclusions: Cancers that respectively showed a strong or absent correlation with lung cancer in the SEER Program were similarly strongly or weakly correlated with cigarette smoking in the literature. Cancers with borderline correlations show ambiguous results or confounding variables in the literature. CHEST 2010; 138(3):491–499 Abbreviations: IARC 5 International Agency for Research on Cancer; SEER 5 Surveillance, Epidemiology, and End Results

smoking remains the leading preventable Cigarette cause of death in the United States, accounting for nearly 20% of all deaths (438,000 persons) each year.1-3 Based on epidemiologic evidence, first-hand

Manuscript received August 11, 2009; revision accepted January 7, 2010. Affiliations: From the Department of Epidemiology and Biostatistics (Ms Ray), School of Public Health and Health Services, George Washington University; the Office of Cancer Prevention and Control (Dr Henson), George Washington University Cancer Institute; and the Department of Pathology (Dr Schwartz), George Washington University Medical Center, Washington, DC. This research was presented in part at the 2007 CHEST Meeting; October 27, 2007; Chicago, IL. Correspondence to: Arnold M. Schwartz, MD, PhD, FCCP, George Washington University Medical Center, Ross Hall, Ste 502, www.chestpubs.org

cigarette smoking is considered causally associated with several different cancers.4 Twelve of these cancers For editorial comment see page 468 are included in the US Surgeon General’s calculation of deaths attributable to cigarette smoking in the United States.4 They include lip, oral cavity, and 2300 Eye St NW, Washington, DC 20037; e-mail: aschwartz@mfa. gwu.edu © 2010 American College of Chest Physicians. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (http://www.chestpubs.org/ site/misc/reprints.xhtml). DOI: 10.1378/chest.09-1909 CHEST / 138 / 3 / SEPTEMBER, 2010

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

491

pharynx; esophagus; stomach; pancreas; larynx; trachea, lung, bronchus; cervix uteri; urinary bladder; kidney and other urinary tract; and acute myeloid leukemia. The report excludes three cancers—liver, nasopharynx, and nasal cavity/paranasal sinuses—which the International Agency for Research on Cancer (IARC) has included on its list of cancers associated with smoking.5 Evidence for smoking as a risk factor for these other cancer sites often remains controversial and inconsistent. Studies have shown, for example, that smoking has both protective and detrimental effects on breast cancer.6,7 The demonstration of the association of cigarette smoking with lung and other visceral cancers has been based on epidemiologic case-control and cohort studies. We have attempted to demonstrate the association of smoking with visceral cancers based on an epidemiologic ecologic approach. Our hypothesis is that cancer rates for non-lung anatomic sites will follow the same geographic trend as cancer rates for lung cancer if cigarette smoking is the cause of cancer for these other sites. Cancer sites with similar trends to lung cancer may be an indication that cigarette smoking is an etiologic agent. In this study, lung cancer is being used as a proxy for smoking, because in the United States about 90% of lung cancer deaths in men and almost 80% of lung cancer deaths in women are associated with smoking.4 Although such a study has limitations, one reason for the project was to explore additional applications of the Surveillance, Epidemiology, and End Results (SEER) database and to test the proof of concept of the ecologic approach linking exposure to disease.

Materials and Methods Data Source The data were obtained from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (Bethesda, MD) using SEER*Stat 6.2.4. SEER is an authoritative source of information on cancer incidence and survival in the United States; a listing of the SEER sites is contained within its Web site.8 It currently collects and publishes cancer incidence and survival data from population-based cancer registries covering approximately 26% of the US population.8 Additionally, the SEER Program has a . 95% ascertainment rate.9 Cancer incidence rates between 1973 and 2003 were calculated from SEER*Stat 6.2.4 for 197 counties. Rates are per 100,000 and age-adjusted to the 2000 US Standard Population standard. Between 1973 and 2003, there were 401,163 cases of lung cancer recorded. The number of new cancers may include multiple primary cancers occurring in one patient, but does not include recurrences and metastatic sites.10 Eight counties, seven in Hawaii and one in Connecticut, were excluded from analysis because of insufficient data on lung cancer. The data were also investigated using 10-year intervals to assess the presence of bias based on a period effect. Although the results included fewer cases and more random scatter, the trend of 10-year interval analysis was concordant with the overall 30-year analysis.

The dependent variables were the incidence rates for the nonlung cancers for the 197 counties. The principle independent variable was the incidence rate for lung cancer for these same counties. Scatterplots were created to show the trend for each incidence rate vs the lung cancer incidence rate. Linear regression analysis was performed on all scatter plots. A correlation analysis between lung cancer and each of the other cancers was performed to determine if a statistically significant Pearson correlation coefficient exists. A correlation was considered strong if the R2 value was ⱖ 0.30, weak if the value was between 0.10 and 0.30, and no association if the value was , 0.10. Adult brain cancer, prostate cancer, and mesothelioma were selected as controls, because they have not been reported as being associated with cigarette smoking.4,11,12 For cancers limited to women, lung cancer rates were determined in women only. For cancers limited to men, lung cancer rates were determined in men only. Statistical Analysis All statistical analyses were executed using SAS, version 8.2 (SAS Institute; Cary, NC). Linear regression analysis was performed on the scatter plots. Data points more than two SDs from the mean were rare and were considered outliers and removed. Using the remaining points, correlation analysis was performed to obtain the Pearson correlation coefficient and the R2 value. Additionally, the linear models were extrapolated to the y-axis to estimate the incidence of each cancer in the absence of lung cancer, that is, when there is no smoking in the population and the rate of lung cancer is zero.

Results Correlations The cancers investigated for an association with lung cancer, as a surrogate marker for cigarette smoking, are listed in Table 1. Of the 21 non-lung cancers, five had an R2 value . 0.3, indicating a strong correlation with lung cancer and, by implication, cigarette smoking. Laryngeal cancer had the overall highest R2 value of 0.6726 (Fig 1), indicating that 67% of the variance is due to our model and that this cancer type is linked to smoking. Esophageal cancer had the second highest value with an R2 value of 0.4061 (Fig 2). Urinary bladder cancer, with both sexes considered together, had the third highest R2 value of 0.4204. When stratified by sex, the R2 value for men dropped to 0.3735, and the R2 value for women dropped to 0.3677. Kidney cancer had an R2 value of 0.3493. Also included in this group was colorectal cancer, with an R2 value of 0.3095. Five of the remaining cancers showed a weak correlation with cigarette smoking as demonstrated by an R2 value between 0.1 and 0.3. Pancreatic cancer (Fig 3); liver cancer; premenopausal and postmenopausal breast cancer (Figs 4, 5); cancer of the lip; and cancer of the oral cavity, lip, and pharynx are included in this group. Their R2 values, respectively, are 0.1369, 0.1781, 0.1646, 0.2283, 0.1597, and 0.1641. Premenopausal incidence rates were calculated from women , 50 years of age. Postmenopausal incidence rates were calculated from women aged ⱖ 50 years at the time of incident breast cancer.

492

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

Table 1—Pearson Correlation Coefficient and R2 Values Site Urinary bladder Urinary bladder, women Urinary bladder, men Larynx Esophagus Colorectal Kidney Pancreas Liver Breast cancer, women, age , 50 y Breast cancer, women, age ⱖ 50 y Stomach Trachea, mediastinum, and other respiratory organs Cervix uteri Acute myeloid leukemia Nasopharynx Mesothelioma Prostate Brain, age ⱖ 20 y Ovarian

Incidence Ratesa

Pearson Correlation Coefficient

P Value

R2

Conclusions from Primary Papers

Median Age of Diagnosis

20.77 9.38 36.90 4.55 4.53 57.46 10.56 11.6 3.85 42.83 332.50 9.48 0.250 5.24 3.42 0.693 0.99 62.34 7.6 8.182

0.64839 0.60641 0.61115 0.82013 0.63725 0.55636 0.59104 0.36997 0.422 0.40575 0.4778 0.23077 0.19937 0.27691 0.25164 0.19910 0.28398 20.26347 0.24226 0.13571

, .0001 , .0001 , .0001 , .0001 , .0001 , .0001 , .0001 , .0001 , .0001 , .0001 , .0001 .0016 .0056 .0001 .0005 .0058 , .0001 .0003 .0008 .0640

0.4204 0.3677 0.3735 0.6726 0.4061 0.3095 0.3493 0.1369 0.1781 0.1646 0.2283 0.0533 0.0397 0.0767 0.0633 0.0418 0.0806 0.0694 0.0587 0.0184

Y Y Y Y Y ? Y ? ? ? ? Y Y Y Y ? N N N N

73 74 72 65 69 71 65 72 65 … … 71 47.5 48 67 55 74 68 56 63

? 5 uncertain or inconsistent results; N 5 no correlation exists with smoking and cancer; Y 5 correlation exists with smoking and cancer. aPer 100,000 population at risk.

The remaining 10 cancers had an R2 value , 0.1. These included ovarian (Fig 6); brain, age ⱖ 20 years (Fig 7); prostate; mesothelioma (Fig 8); nasopharyngeal; oral cavity and pharynx; acute myeloid leukemia; cervix uteri; trachea, mediastinum and other respiratory organs; and stomach. Lung cancer has a weak to absent correlation with these cancers and, by implication, cigarette smoking appears to have a minimal role in their carcinogenesis. Incidence of Cancer in the Absence of Smoking In those cancers where there exists a significant role of cigarette smoking in the causation and carcinogenesis of cancers, the regression curves may be investigated with respect to a limit point of near-zero smoking effect. If the regression lines were extrapo-

Figure 1. Linear regression of laryngeal cancer vs lung cancer. Linear regression formula and R2 value are noted. www.chestpubs.org

lated toward the ordinate, then the y-intercept may provide an estimate of the incidence of a cancer when the rate of lung cancer approached zero in the population. This extrapolation assumes a linear relationship between the cancer of interest and lung cancer and assumes that non-lung cancer incidences are linear at low levels of lung cancer or smoking in ecological units. This intercept would reflect the incidence of a particular cancer type in a population where there was no smoking, assuming that all cases of incident lung cancer were the result of smoking. These intercept values, which estimate the rate of cancer that is not due to smoking, are presented in Table 2. The quotient of the intercept incidence divided by the range of the incidence for the regression over the full range of smoking represents the percent of the

Figure 2. Linear regression of esophageal cancer vs lung cancer. Linear regression formula and R2 value are noted. CHEST / 138 / 3 / SEPTEMBER, 2010

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

493

Figure 5. Linear regression of breast cancer (age . 50 years) vs lung cancer. Linear regression formula and R2 value are noted. Figure 3. Linear regression of pancreatic cancer vs lung cancer. Linear regression formula and R2 value are noted.

incidence of that cancer that may be attributed to smoking. The percent of the incidence due to smoking for the cancers that our results showed to be strongly correlated with lung cancer ranged from 41.3% to 99.8%. The implication of this analysis is that those cancers with high percentages, for example laryngeal cancer, would be absent from populations without smoking. Those cancers whose percentage is approximately 40% would continue to be present in smoke-free populations because of other significant causative factors. This graphical extrapolative approach works only for cancers with significant regression curves; weaker or nonexistent correlations would be associated with significant errors in the derivation of the intercept and consequently unavoidable errors in the calculation of percentage of the incidence due to smoking. Discussion

gate marker for cigarette smoking. Lung cancer incidence rates were compared with the incidence rates of other cancers in 197 counties included in the SEER Program. SEER has tracked incident cancers in the United States since 1973. As a result, the program has grown, and because of its size ( . 4 million cases), it is appropriate to investigate whether it could be used to substantiate the results of etiologic studies based on more definitive case-control studies. This study has significant relevance to chest physicians, especially those involved in occupational and environmental medicine, in demonstrating the important correlations derived from ecologic studies and their agreement with traditional case-control and cohort studies relating exposure-associated disease using large databases. Ecologic studies do not have the cost, time demands, or recall or selection biases of case-control or cohort studies, although they are limited in their inability to directly correlate individual exposure with disease outcome. Correlations

The purpose of this study was to demonstrate that an epidemiologic ecologic approach has strength and validity in relating putative cancers caused by cigarette smoking relative to lung cancers. This proof of concept graphically related smoking as a risk factor for non-lung cancers by using lung cancer as a surro-

Our study is based on the assumption that a high rate of lung cancer in a county reflects a high rate of smoking, which may also reveal higher rates of other cancers that are also associated with smoking. For the

Figure 4. Linear regression of breast cancer (age , 49 years) vs lung cancer. Linear regression formula and R2 value are noted.

Figure 6. Linear regression of ovarian cancer vs lung cancer. Linear regression formula and R2 value are noted.

494

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

Table 2—Incidence of Cancer in the Absence of Lung Cancer

Site Urinary bladder Urinary bladder, women Urinary bladder, men Larynx Esophagus Colorectal Kidney

Figure 7. Linear regression of brain cancer vs lung cancer. Linear regression formula and R2 value are noted.

study, we had to assume that (1) cases of lung cancer were associated with smoking, (2) all smokers were at average risk, and (3) the effects of smoking reflect an average intensity and duration for the population. We recognize, however, that the intensity and duration of smoking vary from individual to individual. Furthermore, we recognized that not all lung cancers are caused by smoking, although the majority of cases, an estimated 90%, are associated with the use of cigarettes. Furthermore, some cases of lung cancer may be the result of passive exposure to cigarette smoke, and we assume that this effect is minor and distributed among all ecologic population groups in proportion to active smoking. These observations may introduce a small bias in the results, but we believe that our conclusions are justified because of the significant correlation observed with published data. By using lung cancer as a proxy for cigarette smoking, we were able to determine which cancers are associated with cigarette smoking and the extent to which their incidence can be attributable to cigarette smoking. One such cancer, laryngeal cancer, has been strongly associated with cigarette smoking for . 50 years.13-15 In concurrence with the literature, we have shown laryngeal cancer to be strongly correlated

Figure 8. Linear regression of malignant mesothelioma vs lung cancer. Linear regression formula and R2 value are noted. www.chestpubs.org

a

Overall Incidence in % of Overall Incidence the Absence of Incidence Due Ratesa Lung Cancer to Smoking 20.8 9.4 36.9 4.6 4.5 57.5 10.6

10.6 4.6 18.6 0.009 1.2 32.6 6.0

49.0 51.1 49.6 99.8 73.3 43.3 43.4

Per 100,000 population at risk.

with smoking. According to our results, laryngeal cancer had the highest correlation with cigarette smoking in comparison with all the other cancers. Esophageal cancer has also been strongly associated with cigarette smoking in the literature.5,16 Our results not only showed the same strong correlation but also indicated that approximately 40% of esophageal cancers can be explained by cigarette smoking. Cigarette smoking has also been consistently associated with an increased risk of urinary bladder cancer.17,18 Our results indicated that 49% of urinary bladder cancer cases are associated with cigarette smoking, a value close to the published figure of 41%.4 Even after stratification by gender, the percentage maintains significance for both men and women at 50% and 51%, respectively. The similarity also suggests that the effect of smoking and other factors on urothelial carcinogenesis may be independent of hormonal modulation. Cigarette smoking is the most consistently established causal risk factor for renal cell cancer.19 It is estimated that 20% to 30% of renal cell cancers in men and 10% to 20% of renal cell cancers in women are caused by cigarette smoking. Our results also indicate a strong positive correlation between kidney cancer and cigarette smoking. The 2004 Surgeon General’s Report concluded that the evidence indicated that there is a possible causal relationship between cigarette smoking and colorectal cancer, but that the evidence was not sufficient.4 In three prospective studies20-22 and in 10 case-control studies,23-32 an increased risk of colorectal adenomatous polyps, the precursor to invasive cancer, were associated with current cigarette smoking. Of these studies, all except one found the relative risk between 1.5 and 3.8.33 Recent metaanalysis has shown that cigarette smoking has a significant association with colorectal cancers. Our results found a relatively strong correlation between colorectal cancer and cigarette smoking with an R2 value of 0.3095. There are a number of other factors that also increase the risk of colorectal cancer and need to be taken into consideration when drawing CHEST / 138 / 3 / SEPTEMBER, 2010

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

495

conclusions from our results. Other etiologic factors that have been suggested are obesity, level of physical activity, and alcohol consumption.34-36 It is possible that some persons who smoke also eat poorly, are not physically active, and consume alcohol. Therefore, the high incidence may in part be due to these other confounding factors. Our method classified five cancers as being weakly associated with cigarette smoking. Most cohort and case-control studies have found an association between pancreatic cancer and cigarette smoking.5 According to our results, the correlation of pancreatic cancer with lung cancer was intermediate or weak. This weak correlation may be reflected by a low relative risk of pancreatic cancer among smokers compared with nonsmokers. In the 2004 Report of the Surgeon General, the relative risk of dying from pancreatic cancer for both men and women smokers was 2.3 in comparison with nonsmokers.4 There have also been studies that found no increased risk of pancreatic cancer for smokers.37,38 In accordance with the majority of studies, our results showed that pancreatic cancer is weakly associated with cigarette smoking, with an R2 value of only 0.1369. As our data indicate, only 14% of incident pancreatic cancer can be attributable to cigarette smoking. We found liver cancer to be weakly associated with cigarette smoking. Most published studies concerning hepatocellular carcinoma have found similar results, with relative risk estimates ranging from 1.5 to 2.5.39 However, liver cancer is not currently listed by the Surgeon General as being associated with cigarette smoking because of two strong confounding factors, which include hepatitis B and C infection and alcohol consumption.40,41 Evidence has shown that 80% to 95% of all hepatocellular carcinomas, the most common form of liver cancer, are causally related to hepatitis B virus and hepatitis C virus.42 Our results indicate that approximately 18% of the variance of incident liver cancer may be attributable to cigarette smoking. Once again, cigarette smoking may be associated with confounding variables that include alcohol consumption and hepatitis infection. The relationship between cigarette smoking and breast cancer has also been controversial. Based on the literature, no consistent increase or decrease in risk of overall breast cancer incidence due to smoking has been found, nor has menopausal status been shown to have an effect.5 Despite these results, the relationship between smoking and breast cancer continues to be studied because carcinogens in tobacco smoke cause mammary cancer in rodents43,44 and because DNA adducts containing polycyclic aromatic hydrocarbons have been found in exfoliated ductal epithelial cells in human breast milk.45,46 Our results showed a weak correlation with cigarette smoking for

both premenopausal and postmenopausal breast cancer patients, with about 16% and 23% associated with cigarette smoking. Research has shown that cigarette smoking is a likely cause of cancers of the oral cavity, lip, and pharynx. The average relative risk of oropharyngeal cancer among current cigarette smokers is about 10.0 in men and 5.0 in women compared with lifelong nonsmokers.4 Our results, however, did not coincide with the literature. We found a negative correlation for lip, oral cavity, and pharyngeal cancer with cigarette smoking. This finding may result from the low overall incidence for this group of cancers and consequently the errors associated with small sample populations. We found the remaining 10 cancers to have minimal or no correlation with cigarette smoking. Ovarian, prostate, mesothelioma, and brain cancer (age ⱖ 20 years), all showed little to no correlation with cigarette smoking, in accordance with the literature. Stomach, nasopharynx, acute myeloid leukemia, cervix uteri, oral cavity and pharynx, and trachea, mediastinum, and other respiratory organs all showed little to no correlation with cigarette smoking; results for some of these cancers are not completely concordant with published data. According to IARC, . 20 cohort studies and 40 casecontrol studies have reported an association between cigarette smoking and stomach cancer.5 It has been estimated that 17% of stomach cancers among men and 11% among women are attributable to smoking; however, our results indicated that smoking may only explain approximately 5% of stomach cancers for men and women.47 Helicobacter pylori infection has been identified as a possible confounder for stomach cancer. The effect of cigarette smoking on uterine cervical cancer historically has been uncertain because of the presence of human papillomavirus infection as a confounder. Only in 2002 did IARC determine there was an association between uterine cervical cancer and cigarette smoking. Our results indicate that a small percentage, approximately 7% to 8%, of uterine cervical cancers may be associated with cigarette smoking, although we need to point out that the median age of diagnosis for cervical cancer is 48 years. Therefore, most cases of cervical cancer occur before lung cancer, which has a median age of diagnosis at 70 years. In addition, cigarette smoking may be a confounding variable and linked to those with multiple partners and the risk of acquiring human papillomavirus infection. The incidence of cancers of the nasopharyngeal area and trachea, mediastinum, and other respiratory organs showed little to no correlation with cigarette smoking, despite their association in the literature. This result may be the effect of the very low incidence

496

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

rates for both of these cancer groups. Our results also showed that acute myeloid leukemia was not correlated with smoking even though smoking was considered a cause in the Surgeon General’s Report. Perhaps a failure to find a correlation may relate to the difficulty of drawing conclusions about cause on cancers that are relatively uncommon in the population. Incidence of Cancer in the Absence of Smoking By extending the linear regression to the ordinate, we were able to determine the incidence of cancer when the incidence rate of lung cancer approaches zero within a population, that is, in the absence of smoking (Table 2). The incidence for the particular cancer type then reflects other etiologies and causative factors that are present in the population. This approach may only be performed in those cancers with a significant association with smoking, given a significant regression line and a statistically significant ordinate intercept. The other regression curves with weak or no association with smoking are characterized by data scatter, large residuals, and errors about their regression lines; consequently, the derived ordinate intercept has a large uncertainty and error. Tumors of the urinary bladder, larynx, esophagus, colon and rectum, and kidney all strongly correlated with cigarette smoking, demonstrated a range of 43.3% to 99.8% for the contribution of smoking to their respective cancer incidence. Interestingly, these results suggest that cancers of the larynx would be essentially absent in populations without smoking, whereas, cancers of the large bowel, rectum, and kidney would continue to be present because, in part, of environmental exposures to a variety of chemical agents. Limitations One possible limitation of this study is that cancers associated with cigarette smoking have different latent periods. The median age at diagnosis of lung cancer is 70 years.48 A number of other cancers that correlated with lung cancer have a similar median age of diagnosis: esophageal cancer is 69 years and pancreatic cancer is 72 years.48 Other cancers have a significantly different median age of diagnosis. For example, the median age of diagnosis for nasopharyngeal cancer is 55.0 years.48 The median age of diagnosis may affect the interpretation of the results and therefore should be taken into consideration. A second limitation is that SEER does not provide information about different etiologic agents that contribute to cancer rates, and we were therefore unable to control for them.8 For example, occupational exposure to aromatic amines is an etiologic agent for uriwww.chestpubs.org

nary bladder cancer.49,50 Exposure to this agent may increase the rate of urinary bladder cancer in a county, which therefore may not be attributable to cigarette smoking. Examples of other contributors to a number of cancers include obesity and alcohol consumption.51,52 A third limitation of the results is the methodologic approach of an ecologic study in which individual cases are not monitored, unlike those in traditional case-control or cohort studies. Rather, our approach is to identify a population at risk based on its incidence of lung cancer and use those data as a surrogate for smoking exposure within that population. The consequence is to investigate a dose-response relationship within a population as opposed to stratifying individuals with respect to exposure history and disease occurrence. The conclusions drawn from this analysis are exploratory and hypothesis-generating and suggest areas for further investigation and analysis. The strength of the ecologic approach rests on the robust nature of the database and the ease of operation. The reproducibility of our findings and their strong correlation with traditional case-control studies supports the validity of selective and limited use of the ecologic approach to exposure-related disease associations. The strength of the method of regression correlation of non-lung cancer types with lung cancer is the large numbers obtained from SEER registry and the reliability of the diagnostic data. Other approaches to investigate cancer linkage to smoking are based on odds ratios of case-control studies or relative risk calculations based on cohort analysis. These approaches require extensive questionnaires and are plagued by selection and memory recall bias. In well-performed studies, in addition to the identification of a linkage between smoking and cancer incidence, these studies may investigate the dose-response relationship of smoking with the particular cancer type, a technique not available in our approach. In addition, the comparisons of the results of our approach, regression analysis and the calculation of the relative cancer incidence in the absence of smoking, are not readily interchangeable with those results obtained using odds ratios and relative risk.

Conclusions There are studies that show a strong association between cigarette smoking and a cancer, such as lung and laryngeal cancers. In those cases, this regression analytical approach shows strong correlations of the incidence of non-lung cancers with lung cancers. For other cancers, epidemiologic studies have had weak, inconsistent, and unreliable results because of low-incidence CHEST / 138 / 3 / SEPTEMBER, 2010

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

497

cancers, small population numbers, and difficulties in study design and implementation. This ecologic study shows that using a large database may provide a resource for investigating disease relationships as surrogate markers for exposure. This approach enhances the ability to investigate a possible association between cigarette smoking and non-lung cancers where the epidemiologic evidence has been weak, inconclusive, or unreliable. Overall, this method has shown that cancers known to have a strong correlation with lung cancer are likely to be strongly correlated with cigarette smoking. Cancers with either a weak correlation or no correlation with lung cancer are likely to be either weakly correlated or not correlated with cigarette smoking. This method also confirmed that the SEER Program may be an adjunct for epidemiologic studies investigating cause. Acknowledgments Author contributions: Ms Ray: contributed to the creation, design, execution, and composition of the study and to the graphs and statistical analysis. Dr Henson: contributed to the creation, design, execution, and composition of the study. Dr Schwartz: contributed to the creation, design, execution, and composition of the study. Financial/nonfinancial disclosures: The authors have reported to CHEST that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

References 1. Centers for Disease Control and Prevention. Annual smokingattributable mortality, years of potential life lost, and economic costs—United States, 1995-1999. MMWR Morb Mortal Wkly Rep. 2002;51(14):300-303. 2. Centers for Disease Control and Prevention. Annual smokingattributable mortality, years of potential life lost, and productively losses-United States, 1997-2001. MMWR Morb Mortal Wkly Rep. 2005;54(25);625-628. 3. Centers for Disease Control and Prevention. National Center for Health Statistics; Health, United States, 2004 With Chartbook on Trends in the Health of Americans. Hyattsville, MD: US Dept of Health and Human Services, CDC; 2004. 4. US Department of Health and Human Services. The Health Consequences of Smoking: A Report of the Surgeon General. Atlanta, GA: US Department of Health and Human Services, Office of the Surgeon General; 2004. 5. International Agency for Research on Cancer. Tobacco Smoke and Involuntary Smoking. IARC Monograph on the Evaluation of Carcinogenic Risks to Humans. Vol 83. Lyon, France: IARC; 2004. 6. MacMahon B, Trichopoulos D, Cole P, et al. Cigarette smoking and urinary estrogens. N Engl J Med. 1982;307(17):1062-1065. 7. Hiatt RA, Fireman BH. Smoking, menopause, and breast cancer. J Natl Cancer Inst. 1986;76(5):833-838. 8. Surveillance Epidemiology and End Results. Overview of the SEER Program. http://seer.cancer.gov/about/. Accessed January 29, 2007.

9. Zippin C, Lum D, Hankey BF. Completeness of hospital cancer case reporting from the SEER Program of the National Cancer Institute. Cancer. 1995;76(11):2343-2350. 10. Surveillance Epidemiology and End Results. Available data and statistics. http://seer.cancer.gov/faststats/statistics.html. Accessed May 29, 2007. 11. Zheng T, Cantor KP, Zhang Y, Chiu BC, Lynch CF. Risk of brain glioma not associated with cigarette smoking or use of other tobacco products in Iowa. Cancer Epidemiol Biomarkers Prev. 2001;10(4):413-414. 12. Hickey K, Do KA, Green A. Smoking and prostate cancer. Epidemiol Rev. 2001;23(1):115-125. 13. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health. The health consequences of smoking: a report of the Surgeon General. http://cdc.gov/tobacco/data_statistics/ sgr/2004/complete_report/index htm. Accessed May 29, 2007. 14. Wynder EL, Bross IJ, Day E. A study of environmental factors in cancer of the larynx. Cancer. 1956;9(1):86-110. 15. Talamini R, Bosetti C, La Vecchia C, et al. Combined effect of tobacco and alcohol on laryngeal cancer risk: a case-control study. Cancer Causes Control. 2002;13(10):957-964. 16. Znaor A, Brennan P, Gajalakshmi V, et al. Independent and combined effects of tobacco smoking, chewing and alcohol drinking on the risk of oral, pharyngeal and esophageal cancers in Indian men. Int J Cancer. 2003;105(5):681-686. 17. Brennan P, Bogillot O, Cordier S, et al. Cigarette smoking and bladder cancer in men: a pooled analysis of 11 case-control studies. Int J Cancer. 2000;86(2):289-294. 18. Zeegers MP, Goldbohm RA, van den Brandt PA. A prospective study on active and environmental tobacco smoking and bladder cancer risk (The Netherlands). Cancer Causes Control. 2002;13(1):83-90. 19. McLaughlin JK, Lipworth L. Epidemiologic aspects of renal cell cancer. Semin Oncol. 2000;27(2):115-123. 20. Giovannucci E, Colditz GA, Stampfer MJ, et al. A prospective study of cigarette smoking and risk of colorectal adenoma and colorectal cancer in U.S. women. J Natl Cancer Inst. 1994;86(3):192-199. 21. Giovannucci E, Rimm EB, Stampfer MJ, et al. A prospective study of cigarette smoking and risk of colorectal adenoma and colorectal cancer in U.S. men. J Natl Cancer Inst. 1994;86(3):183-191. 22. Nagata C, Shimizu H, Kametani M, Takeyama N, Ohnuma T, Matsushita S. Cigarette smoking, alcohol use, and colorectal adenoma in Japanese men and women. Dis Colon Rectum. 1999;42(3):337-342. 23. Kinlen LJ, Rogot E. Leukaemia and smoking habits among United States veterans. BMJ. 1988;297(6649):657-659. 24. Cope GF, Wyatt JI, Pinder IF, Lee PN, Heatley RV, Kelleher J. Alcohol consumption in patients with colorectal adenomatous polyps. Gut. 1991;32(1):70-72. 25. Monnet E, Allemand H, Farina H, Carayon P. Cigarette smoking and the risk of colorectal adenoma in men. Scand J Gastroenterol. 1991;26(7):758-762. 26. Zahm SH, Heineman EF, Vaught JB. Soft tissue sarcoma and tobacco use: data from a prospective cohort study of United States veterans. Cancer Causes Control. 1992;3(4):371-376. 27. Olsen J, Kronborg O. Coffee, tobacco and alcohol as risk factors for cancer and adenoma of the large intestine. Int J Epidemiol. 1993;22(3):398-402. 28. Boutron MC, Faivre J, Dop MC, Quipourt V, Senesse P. Tobacco, alcohol, and colorectal tumors: a multistep process. Am J Epidemiol. 1995;141(11):1038-1046. 29. Martínez ME, McPherson RS, Annegers JF, Levin B. Cigarette smoking and alcohol consumption as risk factors

498

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

30.

31. 32. 33. 34.

35. 36. 37. 38. 39. 40. 41.

for colorectal adenomatous polyps. J Natl Cancer Inst. 1995; 87(4):274-279. Longnecker MP, Chen MJ, Probst-Hensch NM, et al. Alcohol and smoking in relation to the prevalence of adenomatous colorectal polyps detected at sigmoidoscopy. Epidemiology. 1996;7(3):275-280. Potter JD, Bigler J, Fosdick L, et al. Colorectal adenomatous and hyperplastic polyps: smoking and N-acetyltransferase 2 polymorphisms. Cancer Epidemiol Biomarkers Prev. 1999;8(1):69-75. Almendingen K, Hofstad B, Trygg K, Hoff G, Hussain A, Vatn MH. Smoking and colorectal adenomas: a case-control study. Eur J Cancer Prev. 2000;9(3):193-203. Kato I, Tominaga S, Matsuura A, Yoshii Y, Shirai M, Kobayashi S. A comparative case-control study of colorectal cancer and adenoma. Jpn J Cancer Res. 1990;81(11):1101-1108. Garland C, Shekelle RB, Barrett-Connor E, Criqui MH, Rossof AH, Paul O. Dietary vitamin D and calcium and risk of colorectal cancer: a 19-year prospective study in men. Lancet. 1985;325(8424):307-309. Colditz GA, Cannuscio CC, Frazier AL. Physical activity and reduced risk of colon cancer: implications for prevention. Cancer Causes Control. 1997;8(4):649-667. Kune GA, Vitetta L. Alcohol consumption and the etiology of colorectal cancer: a review of the scientific evidence from 1957 to 1991. Nutr Cancer. 1992;18(2):97-111. Clavel F, Benhamou E, Auquier A, Tarayre M, Flamant R. Coffee, alcohol, smoking and cancer of the pancreas: a casecontrol study. Int J Cancer. 1989;43(1):17-21. La Vecchia C, Liati P, Decarli A, Negri E, Franceschi S. Coffee consumption and risk of pancreatic cancer. Int J Cancer. 1987;40(3):309-313. González CA, Pera G, Agudo A, et al. Smoking and the risk of gastric cancer in the European Prospective Investigation Into Cancer and Nutrition (EPIC). Int J Cancer. 2003;107(4):629-634. International Agency for Research on Cancer. Tobacco Smoking. IARC Monographs on the Evaluation of Carcinogenic Risks Chem to Humans. Vol 83. Lyon, France: IARC; 1986. US Department of Health and Human Services. Reducing the Health Consequences of Smoking: 25 Years of Progress.

www.chestpubs.org

42. 43.

44. 45.

46.

47. 48. 49. 50. 51.

52.

A Report of the Surgeon General. Rockville, MD: US Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention and Health Promotion, Office on Smoking and Health; 1989. International Agency for Research on Cancer. Hepatitis Viruses. Lyon, France: IARC; 1994. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans; vol 59. Ambrosone C, Shields P. Smoking as a risk factor for breast cancer. Totowa, NJ: Humana Press; 2001:519-536. Bowcock A, ed. Breast Cancer: Molecular Genetics, Pathogenesis, and Therapeutics; vol 146. Hecht SS. Tobacco smoke carcinogens and breast cancer. Environ Mol Mutagen. 2002;39(2-3):119-126. Gorlewska-Roberts K, Green B, Fares M, Ambrosone CB, Kadlubar FF. Carcinogen-DNA adducts in human breast epithelial cells. Environ Mol Mutagen. 2002;39(2-3): 184-192. Thompson PA, DeMarini DM, Kadlubar FF, et al. Evidence for the presence of mutagenic arylamines in human breast milk and DNA adducts in exfoliated breast ductal epithelial cells. Environ Mol Mutagen. 2002;39(2-3):134-142. Trédaniel J, Boffetta P, Buiatti E, Saracci R, Hirsch A. Tobacco smoking and gastric cancer: review and meta-analysis. Int J Cancer. 1997;72(4):565-573. Ries LAG, Harkins D, Krapcho M, et al (eds). SEER Cancer Statistics Review, 1975-2003. Bethesda, MD: National Cancer Institute. Bethesda, MD; 2006. Miyakawa M, Tachibana M, Miyakawa A, et al. Re-evaluation of the latent period of bladder cancer in dyestuff-plant workers in Japan. Int J Urol. 2001;8(8):423-430. Vineis P, Pirastu R. Aromatic amines and cancer. Cancer Causes Control. 1997;8(3):346-355. Patel AV, Rodriguez C, Bernstein L, Chao A, Thun MJ, Calle EE. Obesity, recreational physical activity, and risk of pancreatic cancer in a large U.S. Cohort. Cancer Epidemiol Biomarkers Prev. 2005;14(2):459-466. Polednak AP. Geographic pattern of cancers related to tobacco and alcohol in Connecticut: implications for cancer control. Cancer Detect Prev. 2004;28(4):302-308.

CHEST / 138 / 3 / SEPTEMBER, 2010

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 15, 2010 © 2010 American College of Chest Physicians

499