Lifestyle Factors and Their Combined Impact on the Risk of Colorectal Polyps

Abstract

Understanding patterns of shared and type-specific etiologies for colorectal polyps may provide insights into colorectal carcinogenesis. The authors present the first systematic comparison of risk factors by colorectal polyp type in a large colonoscopy-based case-control study of 3,764 polyp-free controls and 2,543 polyp patients, including 1,444 cases with adenomas only, 662 cases with hyperplastic polyps (HPPs) only, and 437 cases with synchronous HPPs and adenomas. Surveys were completed to obtain information on usual dietary intake and other lifestyle factors. Six lifestyle factors, including cigarette smoking, obesity, no regular use of nonsteroidal anti-inflammatory drugs, high intake of red meat, low intake of fiber, and low intake of calcium, were found to be independently associated with the risk of polyps. The risk of polyps increased progressively with an increasing number of adverse lifestyle factors. Compared with participants with no or only 1 risk factor, odds ratios for those with 5 to 6 risk factors were 2.72 (95% confidence interval: 1.94, 3.79) for adenoma only, 4.12 (95% confidence interval: 2.78, 6.09) for HPPs only, and 9.03 (95% confidence interval: 5.69, 14.34) for synchronous HPPs and adenomas. This study provides strong evidence that lifestyle modification is important for the prevention of colorectal polyps, especially advanced and multiple adenomas, which are established precursors of colorectal cancer.

Colorectal adenomas are precursors of colorectal cancer (1, 2), as they can develop into colorectal cancer mostly through a microsatellite stable pathway in the well-characterized adenoma-carcinoma sequence (3). Hyperplastic polyps (HPPs), on the other hand, have traditionally been considered to have no malignant potential, and thus their etiology has been less thoroughly studied. However, recent evidence has suggested that some HPPs may develop into cancer via the serrated or microsatellite instable pathways (4, 5). It is common for an individual to have synchronous HPPs and adenomas (6, 7). It has been shown that patients with synchronous HPPs and adenomas are more likely to have recurrence of adenomas than are patients who had only adenomas or HPPs (7). However, very few studies have evaluated risk factors for synchronous HPPs and adenomas and their role in the pathogenesis of colorectal cancer.

Risk factors for colorectal cancer and adenomas have been extensively studied (8). Several studies have examined the combined impact of multiple lifestyle factors on the primary prevention of colorectal cancer (9–12). However, to our knowledge, only one study has examined the combined association of multiple lifestyle factors with colorectal adenomas (9), and that study only addressed distal adenomas. Several studies have investigated putative risk factors for HPPs (13–17). There have been very few reports that have addressed risk factors for synchronous HPPs and adenomas (15, 17, 18). To our knowledge, no study has systematically compared the association of risk factors with polyp risk across patients with adenomas only, HPPs only, and synchronous HPPs and adenomas. Understanding patterns of shared and type-specific etiologies for colorectal polyps may provide insights into colorectal carcinogenesis. In the present study, we systematically evaluated the association of known colorectal cancer risk factors in relation to risk of colorectal polyps by histology type using data from the Tennessee Colorectal Polyp Study, a large colonoscopy-based case-control study. Further, we quantified the combined impact of these lifestyle factors on the prevention of polyps.

MATERIALS AND METHODS

Tennessee Colorectal Polyp Study

The Tennessee Colorectal Polyp Study is a colonoscopy-based case-control study conducted in Nashville, Tennessee. Detailed methods used in this study have been described elsewhere (19, 20). Briefly, eligible participants aged 40–75 years were identified from patients scheduled for colonoscopy at an academic medical center (Vanderbilt University Medical Center) and a Veterans' Affairs medical center (Tennessee Valley Health System, Nashville) between February 1, 2003, and March 26, 2010. The study was approved by the Vanderbilt University Institutional Review Board, the Veterans' Affairs Institutional Review Board, and the Veterans' Affairs Research and Development Committee.

Excluded from our study were participants who had genetic colorectal cancer syndromes (e.g., hereditary nonpolyposis colorectal cancer or familial adenomatous polyposis), a prior history of inflammatory bowel disease, adenomatous polyps, or any cancer other than nonmelanoma skin cancer. Of the 10,074 eligible participants, 7,330 (72.8%) provided written informed consent, and 6,638 (65.9%) participated in the study. Most of the participants (2,265 cases and 4,141 controls) were recruited before colonoscopy. There were 894 participants (631 cases and 263 controls) who were recruited after colonoscopy. For 30 cases, we were missing information on time of recruitment. The postcolonoscopy recruitment was designed primarily to increase the sample size for polyp patients, and thus more cases than controls were recruited. Of the 6,638 study participants, 6,331 (95.4%) completed a telephone interview soon after colonoscopy, and 5,489 (82.7% of participants) completed a food frequency questionnaire that was developed (21) and validated (22) for a similar southern United States population. On the basis of the colonoscopy and pathologic findings, participants were assigned as polyp-free controls or patients with HPPs only, adenomas only, or synchronous HPPs and adenomas. Eligible controls were participants who had received a complete colonoscopy reaching the cecum and been found to be polyp-free. Twenty-four cases were excluded because of missing data on smoking status and other major risk factors for colorectal cancer. The present analyses included 1,444 cases with adenomas only, 662 cases with HPPs only, 437 cases with synchronous HPPs and adenomas, and 3,764 polyp-free controls.

Assessment of risk factors

In the telephone interview, participants were asked questions about medication use, demographic characteristics, medical history, and selected lifestyle factors, including dietary habits, cigarette smoking, alcohol drinking, and nonsteroidal anti-inflammatory drug (NSAID) use. Regular cigarette smoking was defined as smoking at least 1 cigarette per day continuously for at least 3 months. Former smokers were regular smokers who had stopped for at least 1 year before colonoscopy. Regular alcohol drinking was defined as consumption of 5 or more drinks per week continuously for 12 months. Participants were asked to report whether in the past 15 years they had ever used aspirin (regular or baby aspirin) or nonaspirin NSAIDs for at least 3 days a week for a duration of at least 2 months. For NSAID users, information was obtained regarding NSAID brands, duration of use, and frequency. Regular NSAID users were defined as persons who used the medications at least 3 days per week continuously for at least 1 year (23). Energy intake and dietary factors for each participant were derived from the self-administered semiquantitative 108-item food frequency questionnaire described above (21, 22). Usual dietary intakes of nutrients, including total energy, fat, fiber, calcium, and dietary folate equivalents, were calculated as previously described (24). Dietary intakes were adjusted for total energy intake using the residual method (25, 26), in which intakes of dietary factors of interest were regressed on total energy intake in a liner regression model, and residuals from the model were used to evaluate the association of the dietary factors with polyp risk. Percentile cut points for dietary intake were based on the distributions in the control participants.

Statistical analysis

General linear models and Mantel-Haenszel χ² tests were used to compare the mean difference or distribution of demographic characteristics with known risk factors for colorectal cancer across case and control groups, respectively, with adjustment for age and sex when appropriate. Unconditional logistic regression models were used to estimate odds ratios and their 95% confidence intervals for the association between lifestyle factors and polyp risk. All multivariate analyses were adjusted for demographic variables, which included age, sex, study site, educational attainment, indication for colonoscopy, recruitment before or after colonoscopy, and year of recruitment (2003, 2004, … , 2010). When appropriate in some analyses, additional adjustments were made for known risk factors to evaluate the independent association of the risk factors under investigation. Categories used in the models for these variables are presented in Tables 1 and 2.

Table 1.

Demographic Characteristics of the Study Participants, by Histologic Subtype, the Tennessee Colorectal Polyp Study, 2003–2010

Characteristic	Controls (n = 3,764)		Polyp Type						P Valueb
			Adenoma Only (n = 1,444)		HPP Only (n = 662)		Botha (n = 437)
	%	Mean (SD)	%	Mean (SD)	%	Mean (SD)	%	Mean (SD)
Enrolled before colonoscopy	93.4		71.9		85.2		75.5		<0.001
Study site									<0.001
Vanderbilt University Medical Center	75.4		67.7		61.5		53.8
Tennessee Valley Health System	24.6		32.3		38.5		46.2
Age, years		56.8 (7.7)		58.6 (7.5)		56.7 (7.0)		58.2 (6.6)	<0.001
Female sex	45.1		29.6		33.5		21.7		<0.001
White	88.2		87.8		90.0		91.5		0.149
Indications for colonoscopyc									0.029
Screening	57.0		60.4		56.0		55.6
Family history of colorectal cancer	12.6		11.0		12.5		13.3
Diagnostic/follow up	22.7		21.4		21.5		19.7
Other	7.7		7.2		10.0		11.4
Educational attainmentc									<0.001
High school or less	23.3		29.5		31.4		38.3
Some college	28.3		26.0		29.8		35.6
College graduate	21.4		22.2		19.9		14.3
Graduate or professional education	27.0		22.4		18.9		11.8
Family history of colorectal cancerc	8.8		9.2		8.1		10.1		0.383

Abbreviations: HPP, hyperplastic polyp; SD, standard deviation.

^a Synchronous adenomatous and hyperplastic polyps.

^b Derived from analysis of variance for continuous variables and χ² test for categorical variables.

^c P values were adjusted for age (40–49, 50–59, 60–64, and ≥65 years) and sex distribution of all study participants.

Table 2.

Associations Between Selected Lifestyle Risk Factors for Colorectal Polyps, the Tennessee Colorectal Polyp Study, 2003–2010

Lifestyle Factor	Controls	Adenoma Only			HPP Only			Bothc			Pheterogeneityd
	No. of Participants	No. of Participantsa	ORb	95% CI	No. of Participantsa	ORb	95% CI	No. of Participantsa	ORb	95% CI
Nondietary factorse	3,764	1,444			662			437
Cigarette smoking											<0.001
Never	2,005	614	1.00	Referent	197	1.00	Referent	95	1.00	Referent
Former	1,282	526	1.21	1.04, 1.41	252	1.93	1.56, 2.38	160	2.07	1.56, 2.75
Current	472	302	1.96	1.61, 2.38	211	4.44	3.47, 5.67	182	6.10	4.51, 8.25
<30 pack-years	214	117	1.85	1.41, 2.41	95	4.74	3.51, 6.42	51	4.67	3.14, 6.95
≥30 pack-years	256	185	2.05	1.62, 2.61	116	4.04	3.02, 5.42	130	7.01	5.02, 9.79
Body mass indexf											0.253
18–24.9	1,146	359	1.00	Referent	167	1.00	Referent	93	1.00	Referent
25–29.9	1,408	575	1.12	0.95, 1.33	239	1.01	0.81, 1.27	174	1.22	0.92, 1.63
≥30	1,162	480	1.29	1.08, 1.54	248	1.36	1.08, 1.70	168	1.56	1.16, 2.10
Ptrend		0.004			0.006			0.003
Regular NSAID use											0.568
Never	2,194	832	1.00	Referent	351	1.00	Referent	235	1.00	Referent
Former	278	99	0.90	0.67, 1.12	51	0.96	0.68, 1.35	27	0.65	0.41, 1.03
Current	1,932	701	0.75	0.65, 0.86	339	0.84	0.69, 1.01	224	0.66	0.52, 0.83
<35 drug indexg	761	271	0.78	0.65, 0.87	115	0.81	0.64, 1.02	86	0.70	0.53, 0.93
35–70 drug indexg	467	161	0.83	0.68, 1.02	93	1.00	0.77, 1.29	58	0.80	0.58, 1.10
≥71 drug indexg	704	269	0.77	0.64, 0.92	131	0.86	0.67, 1.09	80	0.60	0.44, 0.81
Regular alcohol consumption											N/Ah
Never	2,240	746	1.00	Referent	326	1.00	Referent	206	1.00	Referent
Former	809	393	1.05	0.88, 1.24	181	0.86	0.69, 1.08	134	0.75	0.57, 0.98
Current	702	302	1.02	0.86, 1.22	153	1.08	0.86, 1.36	97	0.94	0.71, 1.25
Regular exercise											N/Ah
No	1,577	679	1.00	Referent	314	1.00	Referent	239	1.00	Referent
Yes	2,187	765	0.90	0.79, 1.04	348	0.98	0.82, 1.18	198	0.91	0.73, 1.14
Metabolic equivalent hours/week
<1	540	193	0.94	0.77, 1.16	96	1.07	0.82, 1.39	58	1.09	0.78, 1.51
1–10	551	202	0.97	0.79, 1.18	88	1.04	0.80, 1.37	49	0.94	0.66, 1.35
11–20	525	171	0.88	0.71, 1.09	77	0.88	0.66, 1.18	35	0.67	0.45, 1.00
>20	571	200	0.83	0.67, 1.02	84	0.89	0.68, 1.17	53	0.86	0.61, 1.20
Ptrend		0.059			0.361			0.298
Dietary factorsi	3,269	1,271			565			384
Red meat intake, g/day											0.928
≤10.0	827	253	1.00	Referent	118	1.00	Referent	65	1.00	Referent
10.1–23.0	814	276	1.18	0.95, 1.45	105	0.92	0.69, 1.24	52	0.81	0.54, 1.21
23.1–44.1	814	335	1.37	1.12, 1.68	141	1.13	0.85, 1.49	103	1.36	0.95, 1.94
≥44.2	814	407	1.38	1.13, 1.70	201	1.36	1.04, 1.78	164	1.61	1.15, 2.26
Ptrend			<0.001			0.009			<0.001
Fiber intake, g/day											0.294
≤16.6	818	380	1.00	Referent	175	1.00	Referent	148	1.00	Referent
16.7–19.8	816	342	0.95	0.79, 1.15	167	1.05	0.82, 1.35	103	0.84	0.62, 1.13
19.9–24	817	276	0.78	0.64, 0.95	108	0.76	0.58, 0.99	83	0.77	0.56, 1.05
≥24.1	818	273	0.77	0.63, 0.94	115	0.84	0.64, 1.10	50	0.47	0.33, 0.69
Ptrend			0.002			0.051			<0.001
Calcium intake, mg/day											0.997
≤829.3	817	417	1.00	Referent	182	1.00	Referent	136	1.00	Referent
829.4–971.3	818	336	0.87	0.72, 1.05	158	1.02	0.79, 1.30	107	0.99	0.73, 1.34
971.4–1,169.2	817	250	0.67	0.55, 0.82	118	0.79	0.60, 1.03	77	0.78	0.56, 1.08
≥1,169.3	817	268	0.74	0.60, 0.90	107	0.73	0.56, 0.96	64	0.70	0.50, 0.99
Ptrend			<0.001			0.008			0.021
Dietary folate equivalents, μg/day											N/Ai
≤421.4	817	377	1.00	Referent	180	1.00	Referent	151	1.00	Referent
421.5–489.0	818	349	0.94	0.78, 1.13	160	0.97	0.76, 1.24	97	0.75	0.56, 1.01
489.1–584.3	817	283	0.77	0.63, 0.94	118	0.77	0.59, 1.00	72	0.57	0.41, 0.79
≥584.4	817	262	0.75	0.61, 0.92	107	0.73	0.55, 0.96	64	0.62	0.44, 0.87
Ptrend			0.001			0.008			<0.001
Fat intake, g/day											N/Ai
≤72.9	817	282	1.00	Referent	182	1.00	Referent	136	1.00	Referent
73–82.5	818	312	1.06	0.87, 1.30	158	1.10	0.84, 1.45	107	1.03	0.71, 1.49
82.6–90.8	816	311	1.04	0.85, 1.28	118	1.07	0.81, 1.41	77	1.20	0.84, 1.71
≥90.9	818	366	1.17	0.96, 1.42	107	1.24	0.95, 1.62	64	1.59	1.14, 2.22
Ptrend			0.163			0.154			0.002

Abbreviations: CI, confidence interval; HPP, hyperplastic polyp; NSAID, nonsteroidal anti-inflammatory drug; OR, odds ratio.

^a Counts may not sum to the total because of missing data.

^b Odds ratios were adjusted for age, sex, study sites, race, educational attainment, indication for colonoscopy, recruitment before or after colonoscopy, and year of recruitment. For nondietary factors (except smoking status and NSAID use), odds ratios were additionally adjusted for smoking status and NSAID use. For dietary factors, odds ratios were additionally adjusted for smoking status, NSAID use, and total energy intake. Odds ratios for smoking status and NSAID use were adjusted for each other.

^c Synchronous adenomatous and hyperplastic polyps.

^d P values were derived from using the SAS CATMOD procedure for continuous variables. Odds ratios were adjusted for all the dietary factors list in the table and additionally adjusted for age, sex, study site, race, educational attainment, indication for colonoscopy, total energy intake, recruitment before or after colonoscopy, and year of recruitment.

^e Analysis for nondietary factors was done for all subjects who completed the lifestyle questionnaire (n = 6,307).

^f Weight (kg)/height (m)².

^g The NSAID index was created by multiplying NSAID use frequency (times/week) by years of regular NSAID use among current users.

^h Factors were no longer significantly associated with any polyps risk after inclusion in the final model and thus were excluded from heterogeneity analysis.

ⁱ Analysis for dietary factors was limited to subjects who provided food frequency questionnaire data by residual methods (n = 5,489).

Lifestyle factors that were independently associated with the risk of at least 2 of the 3 case groups at P_trend ≤0.05 were included in a final multivariate model. Lifestyle factors that remained significantly associated with the risk of at least 1 case group at P_trend ≤0.05 in the final model were selected to generate a risk score for evaluation of the combined effect of multiple risk factors on the risk of polyps.

P values for trend tests were derived by entering categorical variables as continuous parameters in the models (27). Likelihood ratio tests for multiplicative interaction were used to compare the models with and without interaction terms. To evaluate heterogeneity among the 3 polyp case groups, we calculated the P for heterogeneity by using a likelihood ratio test in the polytomous model (SAS CATMOD), which tested the null hypothesis that the regression coefficient for each of the risk factors was the same for all 3 case groups. Pair-wise analyses and case-only comparisons between adenoma types were made using unconditional logistic regression models. P values ≤ 0.05 (2-sided probability) were considered statistically significant. All analyses were conducted using SAS statistical software, version 9.2 (SAS Institute, Inc., Cary, North Carolina).

RESULTS

Distributions of characteristics for the study groups are presented in Table 1. More controls than cases were recruited before colonoscopy than after and enrolled from the academic medical center than from the Veterans' Affairs medical center. Compared with controls, cases with polyps were more likely to be male and have lower educational levels. Cases with adenomas only or synchronous HPPs and adenomas were older than controls, whereas HPP-only cases were similar in age to controls. Case-control distributions of race, family history of colorectal cancer, and indications for colonoscopy were comparable in general, although the statistical test was significant for indications for colonoscopy because of the large sample size.

Table 2 shows the associations of lifestyle factors with colorectal polyp risk by case group. For all 3 polyp case groups, elevated risks of polyps were observed in association with cigarette smoking (for both former and current smokers and by pack-years of current smoking) and obesity (body mass index (BMI; weight (kg)/height (m)²) ≥30). On the other hand, current regular NSAID use was found to be associated with a reduced risk. Regular alcohol consumption and regular exercise were not associated with the risk of polyps. High intakes of fiber, calcium, and folate (dietary folate equivalents) were related to significantly decreased risk in all 3 polyp case groups (P_trend ≤ 0.05). Red meat intake was significantly associated with increased risks in all 3 polyp case groups (P_trend ≤ 0.05). High consumption of total dietary fat intake was only significantly related to the elevated risk for the group with synchronous HPPs and adenomas (P_trend = 0.003). After adjustment for other dietary factors, the associations of folate and fat intake with the risk of all 3 polyp types were no longer statistically significant (data not shown). Except for smoking, no significant heterogeneity was present across the 3 polyp case groups. To identify specific differences between each pair, pair-wise analyses were performed. The difference in association with cigarette smoking was statistically significant for any of the pair-wise comparisons among the 3 case groups. Stratified analyses were performed to evaluate the potential modifying impacts of age, sex, and reason for colonoscopy referral on the association of lifestyle factors and polyp risk. Although none of the interaction tests was statistically significant, the inverse association between dietary fiber intake and polyp risk tended to be stronger in females than in males, whereas the positive association between red meat intake and polyp risk tended to be weaker in persons who received colonoscopy for cancer screening than in persons who received it for other purposes (data not shown).

Lifestyle factors that were statistically significant in at least 2 of the 3 polyp groups were included in a single model to evaluate their independent impact. Factors that remained statistically significant for at least 1 polyp group were selected to construct a risk score. These factors included cigarette smoking, obesity (BMI ≥30), NSAID use status, dietary red meat intake, dietary calcium intake, and dietary fiber intake. Table 3 shows the association of colorectal polyp risk with the cumulative impact of multiple lifestyle factors as measured by the risk score. The risk score was derived by assigning a score of 1 to each of the independent risk factors and then summing the individual scores. A risk score of 5–6 was related to an approximately 4-fold elevated risk of overall colorectal polyps, a 2.72-fold elevated risk of adenomas only, a 4.12-fold elevated risk of HPPs only, and 9.03-fold elevated risk of synchronous HPPs and adenomas. The trend tests of the risk score were statistically significant for all 3 polyp case groups.

Table 3.

Risk Scores for Colorectal Polyps, the Tennessee Colorectal Polyp Study, 2003–2010

Risk Scorea,b	No. of Controls(n = 3,233)	All Polyps (n = 2,159)			Case Types
					Adenoma Only (n = 1,229)			HPP Only (n = 553)			Bothd (n = 377)
		No. ofCases	ORc	95% CI	No. ofCases	ORc	95% CI	No. ofCases	ORc	95% CI	No. ofCases	ORc	95% CI
0–1	913	239	1.00	Referent	159	1.00	Referent	92	1.00	Referent	43	1.00	Referent
2	885	290	1.32	1.11, 1.57	223	1.26	1.03, 1.55	129	1.35	1.01, 1.80	71	1.54	1.03, 2.30
3	817	336	1.64	1.38, 1.95	289	1.58	1.29, 1.95	144	1.57	1.18, 2.10	97	2.11	1.43, 3.12
4	482	275	2.39	1.98, 2.90	290	2.22	1.77, 2.80	124	2.22	1.63, 3.04	99	3.51	2.34, 5.25
5–6	136	89	3.95	3.02, 5.16	195	2.72	1.94, 3.79	64	4.12	2.78, 6.09	67	9.03	5.69, 14.34
Ptrend			<0.001			<0.001			<0.001			<0.001

Abbreviations: CI, confidence interval; HPP, hyperplastic polyp; OR, odds ratio.

^a Range for score was 0–6 possible points for 6 risk factors, comprising smoking status (never, 0 points; ever or current, 1 point), body mass index (weight (kg)/height (m)²) (<30, 0 points; ≥30, 1 point), regular nonsteroidal anti-inflammatory drug use (never, 1 point; ever or current, 0 points), dietary calcium equivalents (<971.4 mg/day, 1 point; ≥971.4 mg/day, 0 points), dietary fiber intake (<19.9 g/day, 1 point; ≥19.9 g/day, 0 points), and red meat intake (<23.0 g/day, 0 points; ≥23.0 g/day, 1 point).

^b Analysis was limited to subjects who provided food frequency questionnaire data by residual methods and for whom a risk scored could be assigned (n = 5,392).

^c Odds ratios were adjusted for age, sex, study site, race, educational attainment, indication for colonoscopy, total energy intake, recruitment before or after colonoscopy, and year of recruitment.

^d Synchronous adenomatous and hyperplastic polyps.

Table 4 summarizes the associations of the lifestyle risk score with adenoma risk by selected adenoma characteristics. A positive association between the risk score and adenoma risk was seen regardless of stage (advanced or nonadvanced), multiplicity (single or multiple), or polyp size (diameter <1 cm vs. ≥1 cm). However, the association was stronger for advanced adenomas than for nonadvanced adenomas, for multiple adenomas than for a single adenoma, and for large adenomas (diameter ≥1 cm) than for small adenomas (diameter <1 cm). All tests for heterogeneity were statistically significant.

Table 4.

Risk Scores in Adenomas Risk, by Stage and Multiplicity, the Tennessee Colorectal Polyp Study, 2003–2010

Risk Scorea	No. of Controls (n = 3,233)	Stage of Adenomas
		Nonadvanced (n = 1,220)			Advanced (n = 386)
		No. of Casesb	ORc	95% CI	No. of Casesb	ORc	95% CI
Risk score
0–1	913	222	1.00	Referent	60	1.00	Referent
2	885	274	1.23	1.00, 1.52	87	1.51	1.06, 2.16
3	817	334	1.61	1.31, 1.99	99	1.83	1.28, 2.60
4	482	280	2.25	1.78, 2.83	94	3.09	2.12, 4.51
5–6	136	110	3.39	2.46, 4.67	46	5.69	3.55, 9.10
Ptrend		<0.001			<0.001
Pheterogeneityd		0.050
		Multiplicity of Adenomas
		Single (n = 1,117)			Multiple (n = 489)
		No. ofCasesb	ORc	95% CI	No. ofCasesb	ORc	95% CI
Risk score
0–1		200	1.00	Referent	82	1.00	Referent
2		269	1.37	1.10, 1.70	92	1.12	0.81, 1.56
3		304	1.65	1.33, 2.05	129	1.67	1.22, 2.28
4		237	2.18	1.71, 2.77	137	3.07	2.21, 4.27
5–6		107	3.66	2.65, 5.06	49	4.22	2.71, 6.58
Ptrend		<0.001			<0.001
Pheterogeneityd		0.021
		Size of Adenomas
		< 1 cm (n = 1,285)			≥1 cm (n= 321)
		No. ofCasesb	ORc	95% CI	No. ofCasesb	ORc	95% CI
Risk score
0–1		237	1.00	Referent	45	1.00	Referent
2		283	1.19	0.97, 1.47	78	1.82	1.23, 2.70
3		354	1.60	1.30, 1.97	79	1.96	1.32, 2.91
4		297	2.24	1.78, 2.81	77	3.46	2.28, 5.25
5–6		114	3.28	2.39, 4.5	42	7.21	4.36, 11.92
Ptrend		<0.001			<0.001
Pheterogeneityd		0.023

Abbreviations: CI, confidence interval; OR, odds ratio.

^a Range for score was 0–6 possible points for 6 risk factors, comprising smoking status (never, 0 points; ever or current, 1 point), body mass index (weight (kg)/height (m)²) (<30, 0 points; ≥30, 1 point), regular nonsteroidal anti-inflammatory drug use (never, 1 point; ever or current, 0 points), dietary calcium equivalents (<971.4 mg/day, 1 point; ≥971.4 mg/day, 0 points), dietary fiber intake (<19.9 g/day, 1 point; ≥19.9 g/day, 0 points), and red meat intake (<23.0 g/day, 0 points; ≥23.0 g/day, 1 point).

^b A total of 85 subjects were excluded from the analysis because of missing data.

^c Adjusted for age, sex, race, study site, educational attainment, indications for colonoscopy, total energy intake, recruitment before or after colonoscopy, and year of recruitment.

^d Tests for heterogeneity were case-only comparisons.

To further investigate the heterogeneous association between the lifestyle risk score and risk of polyps, we conducted pair-wise analyses by case groups (Table 5). Compared with the adenomas-only group or the HPP-only group, in the synchronous HPPs and adenomas group, the lifestyle risk score was associated with an elevated risk (P_trend ≤ 0.05). Excluding cigarette smoking from the risk score attenuated the association, although the positive association remained statistically significant. No significant difference in association with risk scores was found between the HPP-only and adenoma-only case groups.

Table 5.

Case-Only Comparison of Risk Scores^a in Different Polyp Groups, the Tennessee Colorectal Polyp Study, 2003–2010

	HPP Only Versus AD Only				Bothb Versus HPP Only				Bothb Versus AD Only
	HPP Only (n = 565)	Both (n = 384)	Both (n = 384)	AD Only (n = 1,271)	ORa	95% CI	HPP Only (n = 565)	ORa	95% CI	AD Only (n = 1,271)	ORa	95% CI
Risk scorec
0–1	92	159	1.00	Referent	43	92	1.00	Referent	43	159	1.00	Referent
2	129	223	1.07	0.77, 1.48	71	129	1.15	0.71, 1.86	71	223	1.27	0.83, 1.94
3	144	289	1.01	0.74, 1.40	97	144	1.36	0.85, 2.16	97	289	1.42	0.94, 2.14
4	124	290	0.99	0.70, 1.39	99	124	1.65	1.02, 2.67	99	290	1.65	1.09, 2.51
5–6	64	195	1.56	1.01, 2.40	67	64	2.24	1.31, 3.83	67	195	3.51	2.17, 5.69
Ptrend				0.260				<0.001				<0.001
Risk score 2d
0–1	112	269	1.00	Referent	57	112	1.00	Referent	57	269	1.00	Referent
2	140	318	0.99	0.73, 1.34	91	140	1.25	0.85, 1.82	91	318	1.24	0.81, 1.90
3	174	370	1.01	0.75, 1.36	118	174	1.32	0.91, 1.90	118	370	1.26	0.83, 1.90
4–5	128	273	0.98	0.71, 1.35	111	128	1.66	1.14, 2.42	111	273	1.75	1.13, 2.71
Ptrend			0.985				0.009				0.016

Abbreviations: AD, adenoma; CI, confidence interval; HPP, hyperplastic polyp; OR, odds ratio.

^a Adjusted for age, sex, race, study site, educational attainment, indications for colonoscopy, total energy intake, recruitment before or after colonoscopy, and year of recruitment.

^b Synchronous adenomatous and hyperplastic polyps.

^c A total of 79 subjects were excluded from the analysis because of missing data. Range for score was 0–6 possible points for 6 risk factors, comprising smoking status (never, 0 points; ever or current, 1 point), body mass index (weight (kg)/height (m)²) (<30, 0 points; ≥30, 1 point), regular nonsteroidal anti-inflammatory drug use (never, 1 point; ever or current, 0 points), dietary calcium equivalents (<971.4 mg/day, 1 point; ≥971.4 mg/day, 0 points), dietary fiber intake (<19.9 g/day, 1 point; ≥19.9 g/day, 0 points), and red meat intake (<23.0 g/day, 0 points; ≥23.0 g/day, 1 point).

^d Risk score 2 is the risk score excluding smoking; 77 subjects were excluded from analysis because of missing data. Other risk factors are scored as indicated in Table 3.

DISCUSSION

In the present large colonoscopy-based case-control study, we systematically evaluated the association of major known risk factors for colorectal cancer with colorectal polyp risk by histologic types. Six factors, including cigarette smoking, BMI, NSAID use, and dietary intakes of red meat, fiber, and calcium, were significantly and independently associated with colorectal polyp risk. To our knowledge, this colonoscopy-based study is the first and largest to quantify the combined impact of multiple lifestyle factors in relation to colorectal polyp risks.

Cigarette smoking (28, 29), high BMI (30, 31), and high red meat intake (32, 33) are well-established risk factors and regular NSAID use (34, 35) is a well-established protective factor for colorectal cancer and adenomas. Dietary fiber (36, 37) and high dietary calcium intake (38, 39) have been also associated with decreased risks of colorectal cancer and adenoma in previous studies. In the present study, we found that except for cigarette smoking, other individual risk factors were similarly associated with the risks for adenomas only, HPPs only, and synchronous HPPs and adenomas. For cigarette smoking and other lifestyle factors combined, the strongest association was found for synchronous HPPs and adenomas, followed by HPPs only and then adenomas only (P_{heterogeneity} ≤ 0.05). These findings were supported by results reported from previous studies (15, 17). Our results suggest that although the etiologic profiles for these 3 polyp phenotypes are similar in general, the associations are heterogeneous. The stronger association of cigarette smoking and other lifestyle factors combined with synchronous HPPs and adenomas or with HPP only than with adenomas only remains to be understood.

Historically, HPPs were generally not regarded as precancerous; therefore, there has been limited research into risk factors for these types of polyps (13–16, 40–42). There has now been some evidence suggesting that some HPPs may progress to colorectal cancer (5). Recent studies have further suggested that some HPPs may develop into cancer via serrated or microsatellite instability pathways (4, 5, 43). These pathways differ from the common adenoma-carcinoma sequence, which is mostly through a microsatellite stable pathway (3, 5). Approximately 10%–15% of colorectal cancers show microsatellite instability, which is characterized by defective nucleotide mismatch repair (44). Epidemiologic studies on colorectal cancers have also shown that there are differences in the risk factor profiles between microsatellite instability tumors and microsatellite stable tumors (45–47). Some studies have shown that cigarette smoking was more strongly associated with colorectal cancer characterized by microsatellite instability and somatic v-raf murine sarcoma viral oncogene homolog B1 mutation (45, 47), both hallmarks of serrated hyperplastic neoplasm (5), than was cancer without these characteristics or microsatellite stable colon tumors (45–48). As in our study, cigarette smoking has been consistently associated with increased risks of HPPs (13–17, 20). We found a strong association of cigarette smoking with synchronous HPPs and adenomas, and this association may be explained in part by the stronger association of cigarette smoking with HPPs than with adenomas. However, it appears that other lifestyle factors in combination also showed a stronger association with synchronous HPPs and adenomas than with the HPP-only group or the adenoma-only group. This was reflected in the case-case analysis presented in Table 5 after exclusion of cigarette smoking from the lifestyle risk score. Therefore, it is possible that synchronous HPPs and adenomas might be a distinct phenotype of colorectal neoplasm (15), which may develop into cancer via both microsatellite instability and microsatellite stable pathways. This hypothesis warrants further investigation.

Individuals with multiple adenomas are more likely to be advised by their health care providers to have a follow-up colonoscopy in a shorter interval than are persons without multiple adenomas (49, 50). Thus, preventing multiple polyps could potentially be important in reducing the public health and health care burden due to polyps. In this context, it is particularly intriguing that our results showed that lifestyle factors are strongly associated with synchronous HPPs and adenomas and that synchronous adenomas and a high proportion of synchronous polyps could be prevented by modifying lifestyle. In addition, we also found that the association between these risk factors was stronger for clinically important adenomas (advanced or large than nonadvanced or small adenomas). It might be premature to recommend NSAID use solely for colorectal polyp/cancer prevention because of their known side effects. However, similar to the proportion in a national survey (51), about 50% of our participants have already regularly taken this medication for other reasons, so prevention of polyps may be a serendipitous effect.

There are a number of important strengths of this large study. Colonoscopy was used to classify the case-control status, minimizing misclassification error from incomplete examination of the entire colon. The majority of study participants (87.8%) were recruited before colonoscopy and thus before polyp diagnosis, which reduces possible selection bias. Exclusion of those participants recruited after colonoscopy (n = 894) did not appreciably change the association. Because it was difficult to conduct the lengthy interview in the clinics, all interviews were done via the telephone soon after colonoscopy. However, polyps are benign lesions, and thus any lifestyle change or differential recall after polyp diagnosis is unlikely to be substantial.

For ease of interpretation, we created the risk score using arbitrary weights for each of the factors included. However, in the present study, the estimates using the full-weighted score (created based on the effect size for each variable) and nonweighted score (created by simply counting the number of risk factors) were similar, suggesting that a weighted approach may not improve the estimates appreciably. In addition, we did not present the fully weighted risk points to limit the dependence of the results to that observed only in our study population. Interestingly, the distributions of lifestyle factors (ever smoking, obesity, NSAID use, and physical activity) in our controls were similar to those shown in the 2009 National Health Interview Survey (52) and the 2000 National Health and Nutrition Examination Survey (53). Therefore, the estimate from our study should be close the level estimated using exposure prevalence rates from the general US population.

In conclusion, we found that in this large case-control study, with the exception of cigarette smoking, the strength of associations with other lifestyle risk factors was similar for adenomas and HPPs. Lifestyle factors were more strongly related to synchronous HPPs and adenomas than adenomas only or HPPs only, suggesting that synchronous HPPs and adenomas may be a distinct case group with a somewhat different etiologic profile. These findings shed light on the etiology and pathogenesis of colorectal polyps/cancers and provided additional evidence that lifestyle modification is important for the early prevention of colorectal cancer.