Digestive Diseases and Sciences

, Volume 56, Issue 12, pp 3616–3623

Smoking and the Association of Advanced Colorectal Neoplasia in an Asymptomatic Average Risk Population: Analysis of Exposure and Anatomical Location in Men and Women


    • Division of Gastroenterology and the Carole and Ray Neag Cancer Center, Department of MedicineUniversity of Connecticut Health Center
  • Koorosh Moezardalan
    • Division of Gastroenterology, Department of MedicineUniversity of Connecticut Health Center
  • Catherine R. Messina
    • Department of Preventive MedicineStony Brook University
  • Michael Latreille
    • Department of MedicineColumbia University
  • Robert D. Shaw
    • Department of MedicineStony Brook University
Original Article

DOI: 10.1007/s10620-011-1814-8

Cite this article as:
Anderson, J.C., Moezardalan, K., Messina, C.R. et al. Dig Dis Sci (2011) 56: 3616. doi:10.1007/s10620-011-1814-8



Identifying risk factors for advanced colorectal adenomas may aid in colorectal cancer (CRC) screening, especially in light of the American College of Gastroenterology’s recent guidelines, emphasizing cancer prevention through identification and removal of advanced adenomas. Smoking is an important risk factor for advanced adenomas but there is little data regarding levels of exposure for genders.


The aim of this study was to use an existing database to examine the genders separately with respect to exposure level and anatomic location of advanced adenomas. Our database was designed to study smoking in an asymptomatic, screening population. Data included demographics, family history of CRC, smoking exposure (pack-years and years smoked), alcohol, diabetes, medications, exercise and dietary history. We excluded patients with a first degree relative with CRC.


Compared to non-smokers, female smokers had an increased risk for advanced adenomas with an exposure of 10–30 pack-years (adjusted odds ratio [AOR] = 4.11; 95% confidence interval [CI], 1.88–9.01) as well as for ≥30 pack-years (AOR = 2.54; 95% CI, 1.08–5.96) while men had an increased risk with smoking ≥30 pack-years (AOR = 3.10; 95% CI, 1.71–5.65). An increased association with smoking was observed for proximal advanced adenomas (AOR = 4.06; 95% CI, 1.62–10.19) and large hyperplastic polyps in women.


Women smokers had an increased risk for advanced adenomas at a lower exposure level and had a greater risk for proximal lesions. These findings may have an impact on CRC screening for women.


SmokingSexAdenomaColonoscopy and mass screening


Recent colorectal cancer (CRC) screening guidelines divide screening tests into preventative, such as colonoscopy, and detection exams [1]. Since the advanced adenoma is the major target for endoscopic detection and removal, identifying risk factors for these neoplastic lesions as opposed to invasive carcinoma may aid in making CRC screening more efficient.

Smoking is associated with a more advanced stage of CRC [2] and with an increased risk for advanced adenomas for men and women [38]. Thus, smoking may serve as an important factor in triaging patients for earlier screening. The American College of Gastroenterology (ACG) has recently issued guidelines identifying smokers as individuals who may require earlier screening [1].

However, there is little data examining the risk for smoking and colorectal neoplasia associated within each of the genders. Schoenfeld et al. [9, 10] have suggested that women may be at an increased risk for colorectal neoplasia from smoking than men. Furthermore, it has been shown that women may be more susceptible to the effects of tobacco, developing CRC at a younger age [11]. It has also been shown that women who smoke [12] may be at higher risk for proximal lesions [10], perhaps due to an increase in methylation [13] and mis-match repair defects [1416]. Thus, data regarding risk in men and women with regards to smoking would help identify those who need earlier screening as well as those who would benefit from colonoscopy as opposed to a sigmoidoscopy.

Since risk factors are best identified in an asymptomatic population, we used existing data from a screening population whose results have been validated by two meta-analyses [5, 17]. Our goal was to determine the risk for men and women with respect to smoking exposure, as measured in pack-years, and the association with advanced adenomas as well as anatomic location of these lesions. In addition, we hypothesized that female smokers would be at greater risk for proximal advanced neoplasia than non-smoking women.


Patients in Database

Community-based primary care practitioners refer all of their patients for CRC screening to our Risk Reduction Clinic. The methods have been described in our previous publications [3, 4]. Information collected includes personal history of colorectal neoplasia, family history of colorectal neoplasia, history of inflammatory bowel disease, use of NSAIDs, calcium supplementation and other medications. Data regarding lifestyle, for the prior ten years as well as any recent changes, are collected using modified questions from the Behavioral Risk Factor Surveillance System State Questionnaire [18]. Smoking is reported as packs per day multiplied by the years smoked (pack-years) as well as the age smoking was started. Smokers were categorized into three groups [3, 4]: (1) non-smokers or those who have never smoked, (2) low exposure smokers or those that smoked and quit over 10 years prior or have smoked less than 10 pack-years, and (3) heavy exposure smokers or those that were currently smoking or quit within the last 10 years and smoked more than 10 pack-years. Alcohol is classified into beer, wine or spirits and is reported as glasses or drinks per week. A beer was defined as a 12-ounce serving; spirits as one ounce of hard liquor per drink and a glass of wine was defined as 4 oz. Patients that had inflammatory bowel disease, gastrointestinal symptoms, previous colorectal neoplasia or lower endoscopy within the past 10 years were excluded.

The most advanced lesion found stratified each patient. Advanced colorectal neoplasia was defined as adenocarcinoma, high-grade dysplasia, villous tissue or adenomatous polyps one centimeter or more in diameter. Size was measured by our pathologist [19] using the World Health Organization classification [20]. The colon was divided into proximal and distal at the splenic flexure, which was considered proximal. A complete colonoscopy required visualization of trans-illumination of the right lower quadrant, ileocecal valve and appendiceal orifice and a good preparation, defined as the ability to visualize the mucosa to detect a lesion of one half of a centimeter or more. We classified classic hyperplastic polyps and sessile serrated adenomas (SSA) together as serrated lesions. Any serrated lesion with dysplasia was removed from this category. Patients who had an incomplete exam had their data collected only if they had a successful repeat exam.

Study Design

Approval from our institutional review board did not require patient consent because the study data had no identifiers. From the existing database, we culled data from patients who were 45 years and older and had no family history of colorectal neoplasia.

The remaining patients were the non-smokers, the low exposure and the heavy exposure smokers with no family history of colorectal neoplasia. We then divided the heavy exposure patients into those that smoked >30 pack-years versus those that smoked ≤ 30 pack-years based on our previous studies [4].

The first analysis we performed was examining the risk associated with advanced neoplasia for each gender with regard to the 30 pack-year category. In this analysis, we had four groups: never smokers, the low exposure smokers, the heavy smokers with 10–30 pack-years and the heavy smokers who smoked >30 pack-years.

We then examined the association between smoking and advanced adenomas for each gender with respect to anatomical location, proximal and distal. For these analyses, we grouped the heavy exposure smokers into one group. Thus the comparison was the heavy smokers, which contained both the 10–30 pack-year exposure and the >30 pack-year groups, versus the non-smokers. We also compared the low exposure smokers to the non-smokers.


All statistics were performed using SPSS software. Baseline characteristics were compared using chi-square or Fisher exact tests for categorical data and Mann–Whitney U or t tests for continuous data. We considered results to be significant if the P value was 0.05. Means, standard deviations and percentages were used in describing the data. The primary outcome was the detection of an advanced neoplasia. Odds ratios (OR) and 95% confidence intervals (CI) for the primary outcome variables were assessed using multivariate logistic modeling, and expressed as functions of the study test variables (age and smoking categories) and of other possible co-variables (e.g., age, drinks per week, exercise, fruits and vegetables, ethnicity and education). For all multivariate analyses, all variables with a p value less than 0.10 were in the final equation. Odds ratios are presented with their 95% confidence intervals. All the tests were two-tailed; tests of significance were evaluated at the P < 0.05 levels.


After removing the patients outlined above, we had 1,996 total patients who were 45 years and older with no family history of colorectal neoplasia. Tables 1 and 2 show selected characteristics as categorized by smoking exposure. Women in the two smoking categories were more likely to have a lower BMI than men in the same respective smoking exposure groups. Overall women had a lower average BMI than men (27.7 vs. 28.6, respectively; P < 0.001). They were also more likely than men to drink less alcohol and eat more fruits and vegetables when comparing each of the respective three groups. Women who smoked more than 30 pack-years were also more likely to report regular aspirin use than the women in the other smoking categories (P < 0.02).
Table 1

Demographics and lifestyle characteristics of 1,082 male screening patients

Smoking group

Non-smokers (men), n = 559

Low exposure (men), n = 297

10–30 pack-years (men), n = 105

>30 pack-years (men), n = 121

Age (avg ± SD)

59.9 ± 9.5

62.1 ± 8.6

56.9 ± 8.2

60.2 ± 7.7

BMI (avg ± SD)

28.4 ± 5.0

28.8 ± 4.8

28.1 ± 3.8

29.1 ± 6.0

Caucasian (%)

95.2% (532)

98.7% (296)

92.4% (97)

96.7% (117)

Pack-years (avg ± 95% CI)


27.2 ± 2.7

21.4 ± 7.7

69.6 ± 70.9

Daily aspirin or NSAID use

32.1% (180)

33.3% (99)

24.8% (26)

30.6% (37)

Education (1 year of college)

78.8% (441)

74.1% (220)

65.7% (69)

70.2% (85)

Fruits/vegs > 2 servings/day

49.5% (276)

45.1% (134)

37.1% (39)

26.4% (32)

Alcohol > 7 servings/week

17.0% (95)

24.2% (72)

21.9% (23)

24.8% (30)

Exercise > 1 h/week

49.8% (279)

56.4% (167)

35.2% (37)

38.8% (47)

Diabetes mellitus

7.0% (39)

4.7% (14)

6.7% (7)

5.0% (6)

Hormone replacement therapy

Table 2

Demographics and lifestyle characteristics of 914 female screening patients

Smoking group

Non-smokers (women), n = 530

Low exposure (women), n = 233

10–30 pack-years (women), n = 82

> 30 pack-years (women), n = 69

Age (avg ± SD)

59.5 ± 9.0

61.4 ± 8.0

55.9 ± 7.5

60.7 ± 8.4

BMI (avg ± SD)

27.7 ± 6.7

28.1 ± 6.9

26.5 ± 5.0

28.6 ± 7.1

Caucasian (%)

505 (95.3%)

238 (99.2%)

80 (97.6%)

69 (100.0%)

Pack-years (avg ± 95% CI)


23.0 ± 6.2

20.9 ± 7.0

68.6 ± 69.2

Daily aspirin or NSAID use

27.6% (146)

27.5% (64)

24.4% (20)

42.0% (29)

Education (1 year of college)

73.0% (387)

66.1% (154)

72.0% (59)

59.4% (41)

Fruits/vegs > 2 servings/day

58.1% (308)

60.5% (141)

52.4% (43)

55.1% (38)

Alcohol > 7 servings/week

6.4% (34)

11.2% (26)

13.4% (11)

10.1% (7)

Exercise > 1 h/week

45.1% (239)

46.4% (108)

43.9% (36)

33.3% (23)

Diabetes mellitus

9.2% (49)

11.2% (26)

3.7% (3)

0% (0)

Hormone replacement therapy

15.3% (81)

19.3% (45)

18.3% (15)

21.7% (15)

We then examined the risk for advanced neoplasia associated with smoking for men and women. Smokers were categorized into four groups as described in the Methods section. After multivariate analyses, we observed an increased risk of advanced neoplasia for men with an exposure of >30 pack-years only (adjusted odds ratio [AOR] = 3.10; 95% confidence interval [CI], 1.71–5.65). In addition to smoking, age (AOR = 1.05; 95% CI, 1.02–1.07 per year) and a history of diabetes mellitus (AOR = 4.29; 95% CI, 2.15–8.55) were included in the final equation for men. After multivariate analyses, we observed an increased risk of advanced neoplasia for women with an exposure of both the 10–30 pack-year group (AOR = 4.11; 95% CI, 1.88–9.01) as well as the >30 pack-years (AOR = 2.54; 95% CI, 1.08–5.96). In the final equation, we included age (AOR = 1.03; 95% CI, 1.00–1.06 per year) and obesity (BMI ≥ 30) (AOR = 2.24; 95% CI, 1.27–3.96). These data are shown in Table 3.
Table 3

Association between smoking exposure and advanced neoplasia



Low exposure

10–30 pack-years

>30 pack-years


Multi-variate OR (95% CI)


Multi-variate OR (95% CI)


Multi-variate OR (95% CI)


Multi-variate OR (95% CI)

Mena (n = 1,082)

6.4% (36/559)


9.1% (27/297)

1.34; 0.79–2.29

8.6% (9/105)

1.61; 0.74–3.50

16.5% (20/121)

3.10; 1.71–5.65

Womenb (n = 914)

4.5% (24/530)


4.7% (11/233)

1.00; 0.48–2.09

13.4% (11/82)

4.11; 1.88–9.01

11.6% (8/69)

2.54; 1.08–5.96

OR odds ratio, CI confidence interval

aCovariates in the multivariate analysis include history of diabetes mellitus and age

bCovariates in the multivariate analysis include age and obesity (BMI ≥ 30)

Our next outcome was the risk of distal advanced neoplasia associated with smoking for men and women. We examined smoking in three groups as described in the methods. Men and women who smoked were at risk for distal advanced neoplasia. Age (AOR = 1.02; 95% CI, 1.00–1.05 per year) and a history of diabetes mellitus (AOR = 4.17; 95% CI, 1.93–9.01) were included in the final equation for men while age (AOR = 1.04; 95% CI, 1.00–1.08 per year) was the variable included for women. These data are shown in Table 4.
Table 4

Risk of distal advanced adenomas in men and women smokers



Low exposure

Heavy exposure


Multivariate (OR; 95% CI)


Multivariate (OR; 95% CI)


Multivariate (OR; 95% CI)

Mena (n = 1,082)

4.1% (23/559)


4.7% (14/297)

1.07; 0.53–2.17

10.2% (23/226)

2.83; 1.54–5.20

Womenb (n = 914)

3.2% (17/530)


2.6% (6/233)

0.76; 0.29–1.95

7.3% (11/151)

2.53; 1.15–5.55

OR odds ratio, CI confidence interval

aCovariates in the multivariate analysis were age and a history of diabetes mellitus

bCovariate in the multivariate analysis was age

The risk of smoking and proximal advanced neoplasia was also examined. In the model for men, we included age (AOR = 1.10; 95% CI, 1.06–1.13 per year) and a history of diabetes mellitus (AOR = 2.56; 95% CI, 0.72–9.10). With regard to the model for women, we included age (AOR = 1.05; 95% CI, 1.01–1.10 per year), NSAID use (AOR = 0.32; 95% CI, 0.11–0.96), obesity (BMI ≥ 30) (AOR = 2.22; 95% CI, 1.00–4.90) and fruit and vegetable intake (AOR = 0.41; 95% CI, 0.18–0.93) (see Table 5).

We also examined the risk for right-sided large (≥1 cm) serrated polyps. There was a higher prevalence of these lesions in female smokers (5/151; 3.3%) than non-smokers (3/530; 0.6%) (P = 0.015). Included in these lesions were three confirmed SSAs in the smoking group and one SSA in the non-smoking group. The other four were presumed to be hyperplastic polyps. After adjusting for age (AOR = 1.01; 95% CI, 0.94–1.09 per year), we observed that female smokers were more likely than non-smokers to have these right-sided lesions (AOR = 6.08; 95% CI, 1.43–25.82). In males there was a trend toward significance with regard to the difference between male smokers (5/226; 2.2%) and non-smokers (3/559; 0.5%) (P = 0.05). Included in these eight serrated polyps were three confirmed SSAs.
Table 5

Risk of proximal advanced adenomas in men and women smokers



Low exposure

Heavy exposure


Multivariate (OR; 95% CI)


Multivariate (OR; 95% CI)


Multivariate (OR; 95% CI)

Mena (n = 1,082)

2.7% (15/559)


5.4% (16/297)

1.95; 0.93–4.06

4.4% (10/226)

2.18; 0.94–5.07

Womenb (n = 914)

1.9% (10/530)


3.0% (7/233)

1.57; 0.58–4.21

6.6% (10/151)

4.06; 1.62–10.19

OR odds ratio, CI confidence interval

aCovariates in the multivariate analysis included age and diabetes mellitus

bCovariates in the multivariate analysis included age, obesity (BMI ≥ 30), NSAID or aspirin use and fruit and vegetable intake


Overall, our data underscores the importance of smoking as a risk factor for colorectal neoplasia as observed recently with regard to advanced adenomas and colorectal cancer [21, 22]. However, we also made important observations regarding tobacco exposure and the risk for advanced colorectal neoplasia in men and women. In our sample, women who smoked between 10 and 30 pack-years were more than twice as likely to have advanced neoplasia as non smokers. This was not observed in men who required more than 30 pack-years to have a statistically significant increased risk. The women who smoked 10–30 pack-years had a higher odds ratio than the women in the higher exposure group. This may be due to the higher use of NSAIDs or aspirin in this latter group. NSAID or aspirin use was a significant co-factor in the model for women and proximal advanced lesions. We also observed that female smokers, but not males, were more likely to have proximal advanced neoplasia.

With regard to the difference between the genders and tobacco exposure, Onega et al. [23] examined the data from the New Hampshire Colonoscopy Registry and observed that smoking more than 15 pack-years was associated with a modest increase risk with adenomas for men and women. Only men had an increased risk for an exposure of less than 15 pack-years. The difference in findings between their study and the present one may be due to the smoking categorizations used in the two studies. In this present study, the low exposure group was removed to eliminate the effect of the people who quit over 10 years prior to screening. This was based on our previous published findings that the patients in the low exposure group had a risk for adenomas that was similar to the non-smokers [3, 4]. The inclusion of the people who quit several years prior to screening may explain the difference in findings between the women and the men. Specifically, the group of female smokers in the study by Onega et al. may have included more patients who quit smoking prior to screening than the male smokers. In addition, the inclusion of patients who quit may explain the overall modest increase of adenomas associated with smokers as compared to the higher risk observed in the present study or other published data [5].

In their study of over 3,000 patients in Germany, Hoffmeister et al. [21] observed an increased risk for smoking and advanced neoplasia in men and not women. They divided the tobacco users into current and former smokers. Whereas a large percentage of the men who were current or former smokers had more than 20 pack-years of smoking, the percentage of women with this exposure level was much lower. This exposure difference may have accounted for their gender-related findings.

Zisman et al. [11] also observed that female smokers were diagnosed with CRC at a younger age than male smokers. An increased susceptibility of women to tobacco occurs in other cancers as well. Stabile et al. observed a higher risk of oral, lung and bladder cancer for women than men for a similar tobacco exposure [24].

We also found that women who smoked had a significant risk for proximal advanced neoplasia than non-smokers. This relationship was not observed in men. One previous study demonstrated that smokers were at risk for isolated proximal neoplasia but did not examine gender differences [25]. Limberg et al. observed an increased risk for proximal colorectal cancer in women who smoked in Iowa Women’s Health Study [12]. One possible explanation includes the increased rate of microsatellite instability in women, which is increased in the colorectal cancers of smokers [16, 26]. There may be an increase of hyper-methylation abnormalities in women smokers [13, 2729], which may increase the risk for proximal neoplasia as well as serrated polyps. A recent case control study examining SSAs, most of which were proximal, observed an increased risk associated with smoking [30]. To examine this issue, we examined the risk for large proximal serrated lesions in our sample and observed that the risk for these lesions was increased in female smokers only.

In our study, as has been done in recently published studies [31, 32], we classified the proximal serrated lesions together due to the difficulties of differentiating the SSAs from the classic hyperplastic polyps. Serrated polyps can be classified as typical hyperplastic polyps, atypical polyps such as SSAs and dysplastic such as the traditional serrated adenomas and mixed polyps [33]. While our methods likely allowed us to select out the TSAs and the mixed polyps, the sample we examined may have included hyperplastic polyps as well as SSAs. Therefore, we acknowledge that these polyps in our study were a heterogeneous group. Although one half were SSAs which are associated with the BRAF pathway, the other lesions included in the group were read as hyperplastic polyps which may be associated with a different pathway [34]. However, in light of a recent study demonstrating the association of all large proximal serrated lesions with synchronous advanced neoplasia [32], our findings may have implications for screening female smokers.

Our study has several strengths which include that our study was performed in an asymptomatic population whose previous results regarding gender and tobacco have been validated in recent meta-analyses of colorectal neoplasia [5, 17]. The endoscopic results of screening populations better represent the prevalence of adenomas than symptomatic patients. In addition, since all the patients had complete colonoscopies, we can be reasonably sure of the adenoma status and accurate locations in smokers and non-smokers. These patients, referred from their primary care practitioner and not recruited for a study, may be more representative of the general screening population. In addition, we controlled for several important factors known to increase the risk of advanced neoplasia. Hoffmeister et al. [21] observed recently in a screening population that smoking, male gender and family history of CRC were the most important predictors of advanced neoplasia. We removed those patients with a family history of colorectal cancer to ensure that the population represented an average risk sample to further examine the effect of smoking, especially on the proximal side. Our observations were associated with an odds ratio greater than two-fold decreasing the likelihood that confounding variables may explain the findings. Finally, our classification of smoking allowed us to account for those patients who quit as well as the total pack-years of exposure. Whereas Onega et al. [23] classified smokers based on total exposure and Hoffmeister et al. [21] based their categories on former or current, we feel that our methods allowed us to account for both factors. This enabled us to make observations regarding smoking and advanced neoplasia.

Our results have many implications for CRC screening. Recent guidelines have divided the screening tests into those that detect cancer and those that prevent cancer, such as the colonoscopy. Since the presumed mechanism for CRC prevention from colonoscopy is the detection and removal of advanced neoplasms, identifying risk factors for these growths may be important for screening.

Furthermore, there has been a concern regarding insufficient resources to screen patients with colonoscopy. One author has suggested an alternative strategy such as a sigmoidoscopy first and a colonoscopy at a later age [35]. Our data would suggest that perhaps women who smoke may require screening with a colonoscopy and not a sigmoidoscopy given their predisposition to advanced proximal neoplasia.

Another has suggested commencing screening some risk groups at different ages [36]. For example, women may lag men by 5 years with respect to their risk for colorectal neoplasia. Based on our observed gender difference, smoking may aid in triaging patients for screening. There is also evidence to demonstrate that smokers may be at risk for more advanced disease as well as be diagnosed at an earlier age than non-smokers [11]. Phipps et al. [37] showed that smokers who were diagnosed with CRC had an increased mortality as compared to non-smokers. A recent study of patients undergoing screening colonoscopy observed that smoking was one of the most predictive factors of detecting advanced adenomas [21]. In addition, the recently published ACG guidelines have introduced the concept of tailored screening by suggesting that smokers, with >20 pack-years, may benefit from earlier CRC screening. Our data demonstrating a gender difference with regard to tobacco exposure may have implications for colorectal cancer screening. For example, a lower exposure threshold would be set for women than men.

In addition, interval neoplasia may have a proximal predilection [38, 39] and women may be at more risk than men [40, 41]. It is possible that smoking may play a role in women. We observed that female and not male smokers were at risk for proximal advanced neoplasia. It has been shown that smoking may be associated with flat polyps [42] as well as serrated tissue [43], both of which can be proximal in location and difficult to detect and thus may explain the right-sided paradox. Serrated polyps have an increased frequency of BRAF mutations [44] which along with other methylation abnormalities have been observed in interval cancers [45, 46] and in smokers [47]. We observed that women who smoked had an increased risk for large proximal serrated polyps. However, our sample of hyperplastic lesions was small and heterogeneous. Although many of our large proximal serrated polyps were SSAs, the rest were likely of the hyperplastic variant and thus may not be associated with the BRAF pathway. Thus, future studies will need to examine whether women smokers are more likely to have a decreased protective effect proximally from colonoscopy.

Our data demonstrates an increased risk for women with respect to a low tobacco exposure and also with regard to the proximal colon. Studies examining interval cancers should include a focus on women who have had significant tobacco exposure. Further data could help to substantiate these findings and perhaps alter current screening practices.

Conflict of interest


Copyright information

© Springer Science+Business Media, LLC 2011