Diet, physical activity, and body size associations with rectal tumor mutations and epigenetic changes

Abstract

Diet and lifestyle factors have been inconsistently associated with rectal tumors. It is possible that evaluation of specific tumor markers with these factors may help clarify these associations. In this study, we examine energy contributing nutrients, dietary fiber, BMI (kg/m2), and long-term physical activity with TP53 mutations, KRAS2 mutations, and CpG Island Methylator Phenotype (CIMP) in 750 population-based cases of rectal cancer compared to healthy controls. We observed that high levels of physical activity reduced the risk of having TP53 and KRAS2 rectal tumor mutations. Dairy products rich in fat were associated with an increased risk of CIMP+ tumors (OR 1.88 95% CI 0.92, 3.84), while low-fat dairy products reduced risk of CIMP+ tumors (OR 0.56 95% CI 0.29, 1.09). Omega-3 fatty acids were associated with a twofold increased risk of a CIMP+ tumor. High levels of vegetable intake reduced risk of both TP53 mutations (OR 0.73 95% CI 0.54, 1.00; p trend 0.02) and KRAS2 mutations (OR 0.60 95% CI 0.40, 0.89; p trend<0.01). High intake of whole grains reduced the likelihood of a TP53 mutation (OR 0.74 95% CI 0.56, 0.99), while high intake of refined grains increased the likelihood of a TP53 mutation (OR 1.41 95% CI 1.02, 1.96). Dietary fiber also was associated with reduced risk of TP53 and KRAS2 rectal tumor mutations. Overall, a prudent dietary pattern significantly reduced the likelihood of a KRAS2 tumor mutation (OR 0.68 95% CI 0.47, 0.98; p linear trend 0.03). These data suggest that diet and lifestyle factors are associated with specific types of rectal tumor mutations and epigenetic changes. Findings need confirmation in other studies.

Diet and lifestyle factors, such as level of BMI and physical activity, have been associated with increased risk of rectal cancer [1–9]. While diet has been associated with rectal cancer, the elements of the diet that are most important differ between studies. The most consistent associations appear to be for vegetable, fiber, and calcium intake [3, 4, 10–12]. Likewise, while some studies show that higher levels of physical activity are associated with lower rectal cancer risk, others show no association [9, 13–16]. BMI has generally not been associated with rectal cancer risk [16–18].

Studies of colon cancer have shown that examination of specific tumor markers can help define disease risk factors [19–21]. Some risk factors such as cigarette smoking align with increased risk of MSI and CIMP+ tumors, while contributing minimally to increased risk of other tumor markers such as KRAS2 or TP53 overall [22, 23]. Because MSI and CIMP+ tumors represent the minority of colon tumors, the risk associated with cigarette smoking and colon cancer at the population level is considerably less than the twofold increase in risk associated with these specific tumor makers. It is possible the evaluation of specific rectal tumor mutations and epigenetic changes may help define the associations between diet, BMI, and physical activity and rectal cancer.

In this study, we evaluate dietary composition by examining both foods and nutrients, with CIMP, TP53, and KRAS2 tumor status in rectal tumors. We also examine associations between BMI and physical levels and these tumor biomarkers. Our hypotheses are that evaluation of diet, BMI, and physical activity with rectal tumor epigenetic and genetic changes will advance our understanding of how these factors are associated with rectal cancer.

Methods

Participants in the study were from the Kaiser Permanente Medical Care Program of Northern California (KPMCP) and the state of Utah. All eligible participants within these defined areas were identified and recruited for the study. Participants with a first primary tumor in the recto-sigmoid junction or rectum were identified between May 1997 and May 2001. Case eligibility was determined by the Surveillance Epidemiology and End Results (SEER) Cancer Registries in Northern California and in Utah. To be eligible for the study, participants had to be between 30 and 79 years of age at time of diagnosis, English speaking, mentally competent to complete the interview, could not have had previous colorectal cancer [24], and could not have known (as indicated on the pathology report) familial adenomatous polyposis, ulcerative colitis, or Crohn’s disease. The race/ethnicity of the rectal study population was reported at the time of interview as 82% white, non-Hispanic, 4.1% African American, 7.6% Hispanic, 4.6% Asian, 0.7% American Indian, and 1% multiple races/ethnicity (data not shown in table).

A total of 1,505 participants with rectal cancer were identified; of these, 982 were interviewed; reasons for non-response have been detailed [9]. Block retrieval involved obtaining pre-operative biopsy prior to treatment as well as paraffin-embedded tissue from the resection. In some instances, because of radiation prior to resection, tissue was limited from the resection, and therefore biopsy specimens were used for making tumor DNA. In Utah, blocks were requested for all participants except those who refused release of blocks. For those who were not interviewed and had not signed a medical record release, the Utah Cancer Registry retrieved the blocks and released them to the study without key identifiers of name, address, and complete date of birth (year and month of birth were released). At the KPMCP, samples were retrieved from persons who signed a consent form or who had died. For the 1,495 eligible participants with rectal cancer identified at both centers, 239 people identified with rectal cancer had not given consent to have the tissue released (15.9%), and for an additional 234 participants, either tumor tissue could not be obtained or DNA could not be extracted. Tumor DNA was extracted from 81.4% of all participants with rectal cancer identified, of which 750 participants had interview data. Controls were randomly selected from membership lists at KPMCP, social security lists, and driver’s license list (people under 65 years) with the same eligibility criteria; 1,205 controls (68.8% of those selected) participated and are included in these analyses.

Genetic analysis

Tumor DNA was obtained from paraffin-embedded tissue for all samples identified and for which blocks were available. We characterized their genetic profile that include sequence data for exons 5 through 8, or the hotspots of mutations of the TP53 gene; sequence data for KRAS2 codons 12 and 13; and five CpG Island (CIMP) markers MINT1, MINT2, MINT31, CDKN2A (p16), and MLH1 (Table 1). While there no consensus as to the appropriate CIMP panel or method of detection at the time of the study, we have used our panel to demonstrate significant relationships between CIMP and numerous variables, including cigarette smoking and the BRAF V600E mutation, which were independent of micro-satellite instability [23, 25]. This work has helped to support the legitimacy of the CIMP concept [26]. CIMP+ (positive or high) was methylation of two or more of these CpG islands. CIMP− (negative or low) was defined as zero or one of five markers methylated [3]. The use of established assays, selection of genetic loci, and criterion for CIMP+ or CIMP− was based on the pioneering work of other groups that previously defined the CIMP phenotype [27, 28].

Table 1.

Description of the study population

	Cases		Controls
	n	(%)	n	(%)
Total	951	44.1	1,205	55.9
Tumor mutation data	750	78.9
CIMP	74	11.0
KRAS2s
Overall	215	28.9
Transitions	75	12.4
Transversions	75	12.4
TP53
Overall	340	48.3
Transitions	248	40.5
Transversions	69	15.9
Codon 175	37	9.2
Codon 245	19	5.0
Codon 248	36	9.0
Codon 273	29	7.4
Codon 282	11	2.9
	Mean	SD	Mean	SD
Age	61.36	10.9	61.6	11.1
Energy intake	2693.7	1353.4	2598.5	1260.8
Total Fat (% KCAL)	35.1	7.5	34.5	7.8
Omega-3 (% KCAL)	0.78	0.24	0.78	0.26
Total protein (% KCAL)	14.6	2.7	14.6	2.8
Carbohydrate (% KCAL)	49.5	8.8	50.9	9.1
Dietary fiber (g)	25.9	13.4	26.7	13.0
High-fat dairy (serv/day)	1.3	1.5	1.2	1.6
Low-fat dairy (serv/day)	1.3	1.4	1.5	1.7
Fruits (serv/day)	2.0	1.7	2.2	1.9
Vegetables (serv/day)	3.9	2.8	4.2	3.0
Red meat (serv/day)	1.1	1.0	0.9	0.8
Fish (serv/day)	0.28	0.35	0.26	0.34
Whole grains (serv/day)	1.6	1.5	1.8	1.4
Refined grains (serv/day)	4.3	2.8	4.0	3.0

Diet and lifestyle data

Trained and certified interviewers collected diet and lifestyle data as previously outlined [29, 30]. The referent year for the study was the calendar year approximately two years prior to date of diagnosis (cases) or selection (controls). Information was collected on demographic factors such as age, sex, and study center; physical activity was determined by a detailed physical activity questionnaire that obtained information on activity patterns 10 and 20 years ago as well as activity during the referent year [31, 32]; body size, including usual adult height and weight two and five years prior to diagnosis; cigarette smoking history; family history of colorectal cancer in first degree relatives; medical and reproductive history including use of hormone replacement therapy (HRT); and use of aspirin and non-steroidal anti-inflammatory drugs on a regular basis (labeled NSAIDs). Approximately 90% of participants self-identified as non-Hispanic white.

Dietary intake was ascertained using an adaptation of the CARDIA diet history [30, 33, 34]. Participants were asked to recall foods eaten, the frequency at which they were eaten, serving size, and if fats were added in the preparation. Nutrient information was obtained by converting food intake data into nutrient data using the Minnesota Nutrition Coding Center (NCC) nutrient database version 30. We assessed both food and nutrient intake. Servings per day of foods in specific groups were evaluated, including red meat, vegetables, grains, and fruits. We also evaluated macronutrient intake and dietary fiber.

Statistical analysis

Dietary variables were assessed by fertile of intake, based on the distribution of the controls for men and women separately and include nutrients from foods only. These sex-specific tertiles are noted as T1, T2, and T3 in the tables. Energy-contributing nutrients were analyzed as percent of total energy; foods were analyzed as servings per day. Servings were defined as number of ½ cup servings of vegetables; 8 oz servings of milk; 2 oz. of cheese; 3 oz of meat or fish; ½ cup of cereal or grains, or 1 slice of bread. Dietary patterns were developed as previously described using foods commonly eaten and factor analysis [35]. With these methods, we identified two common dietary patterns, the Western diet that is comprised of meat, fried food, and refined grains and the Prudent diet that is characterized by fruits and vegetables, whole grains, and fish and chicken. BMI was assessed using <25 as the referent group and overweight (25–29) and obese (≥30). Long-term vigorous physical activity score was evaluated as those who reported little or no physical activity over the past 20 years (referent group), those who reported moderate amounts of intense activity over the past years (roughly those reporting less than 3 h per week), and those with higher amounts of intense activity (roughly those reporting 3.5 or more hours per week).

All statistical analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC). Tumors were defined by specific mutations detected as any TP53 versus no TP53 mutation, any KRAS2 mutations versus no KRAS2 mutation, or CIMP+ versus negative. CIMP+ was defined as at least two of five markers methylated. For TP53 and KRAS2 mutations, we also examined specific types of mutations such as transversion and transition mutations since other studies have shown specific mutations to have etiologic associations [19, 36]. Because of relatively few participants with MSI and BRAF (22 and 27 participants, respectively), we were unable to examine dietary and other factors with these mutations. Population-based controls were used as the comparison group to estimate associations for the population overall while examining multiple outcomes defined by tumor status. Multiple logistic regression models were used to compare all interviewed participants, regardless of whether or not tumor tissue was obtained, to controls. Multinomial generalized estimating equations (GEE) were used to assess associations for the population overall comparing specific types of mutations to controls. Participants could contribute one to three observations in the GEE models depending upon an individual’s number of tumor mutations (CIMP, KRAS2, TP53). The GEE is used to adjust for non-independent observations and was executed in PROC GENMOD as described by [37, 38] All logistic regression models were adjusted for age, sex, and other factors related to overall rectal cancer risk in our study, including recent used of aspirin/NSAIDs, long-term vigorous physical activity, pack-years of cigarettes smoked, dietary calcium per 1,000 calories, and total energy intake per 1,000 calories. Statistical significance was tested at the 0.05 level. Interaction was assessed by determining whether the interaction term significantly improved the overall fit of the model by comparing the likelihood ratio of a model with the interaction term as an ordered categorical variable and a model without the interaction term using a chi-square test with one degree of freedom. Data are presented for men and women combined since no statistically significant interaction by sex was observed.

Results

The study population is described in Table 1. Of the 750 participants with rectal cancer for whom tumor data were available, 11.0% were CIMP+, 28.9% had a KRAS2 mutation, and 48.3% had a TP53 mutation. The mean age was similar for both participants and controls. Participants consumed slightly more fat and red meat and fish, while controls consumed slightly more dietary fiber, carbohydrates, and plant foods. Too few non-Hispanic white participants were available to determine tumor marker distribution by ethnicity.

BMI was not associated with rectal cancer overall or with any specific type of tumor mutation (Table 2). Although data are shown for men and women combined, there were no sex-specific association. On the other hand, long-term participation in vigorous physical activity statistically significantly reduced risk of rectal cancer overall as well as for TP53 and KRAS2 mutations specifically. Physical activity was not statistically significantly associated with CIMP + tumors; associations were similar for men and women.

Table 2.

Associations^a between BMI and long-term vigorous physical activity and rectal tumor mutations

	Control	All Casesb			CIMP+			TP53 Mutation			KRAS2 Mutation
	n	n	OR	(95% CI)	n	OR	(95% CI)	n	OR	(95% CI)	n	OR	(95% CI)
BMI (kg/m2)
≤25	393	303	1.00		34	1.00		116	1.00		75	1.00
25 to < 30	510	367	0.92	(0.75, 1.14)	20	0.45	(0.26, 0.79)	121	0.83	(0.63, 1.08)	76	0.85	(0.62, 1.18)
≥30	278	265	1.22	(0.97, 1.54)	19	0.73	(0.42, 1.28)	99	1.15	(0.87, 1.53)	59	1.07	(0.76, 1.52)
p trend			0.11			0.27			0.34			0.74
Long-term Activity
Low	241	270	1.00		20	1.00		97	1.00		58	1.00
Moderate	424	319	0.66	(0.52, 0.83)	23	0.88	(0.49, 1.59)	116	0.80	(0.60, 1.06)	87	1.06	(0.76, 1.49)
High	524	351	0.60	(0.47, 0.75)	30	1.07	(0.61, 1.89)	124	0.73	(0.55, 0.97)	66	0.67	(0.47, 0.97)
p trend			< .01			0.56			< .01			< .01

^aAdjusted for age, sex, recent aspirin or NSAID use, long-term activity level (BMI only), BMI (long-term activity only), pack-years of cigarette smoking, dietary calcium, and energy intake

^bCases include all cases identified

High-fat dairy products were associated with a trend toward increased risk of CIMP+ tumors (p = 0.05), while low-fat dairy products had the opposite effect (Table 3). High levels of vegetable intake were statistically significantly associated with reduced risk of rectal tumors overall as well as with TP53 and KRAS2 mutations specifically. Whole grain intake was statistically significantly associated with reduced risk of TP53 mutations, while refined grains were statistically significantly associated with increased risk of TP53 mutations. Western dietary pattern was not statistically significantly associated with rectal tumor mutations, while higher levels of consumption of a prudent dietary pattern was statistically significantly associated significantly with reduced risk overall and with KRAS2 mutations specifically.

Table 3.

Associations^a between foods and dietary patterns and rectal tumor mutations

Servings per day	Control	All casesb			CIMP+			TP53 Mutation			KRAS2 Mutation
	n	n	OR	(95% CI)	n	OR	(95% CI)	n	OR	(95% CI)	n	OR	(95% CI)
Dairy
High fat
T1	393	255	1.00		17	1.00		88	1.00		58	1.00
T2	401	356	1.31	(1.06,1.63)	29	1.68	(0.91, 3.11)	138	1.43	(1.08, 1.89)	72	1.06	(0.75, 1.50)
T3	395	329	1.16	(0.91,1.49)	27	1.88	(0.92, 3.84)	111	1.06	(0.77, 1.46)	81	1.18	(0.81, 1.71)
p trend			0.21			0.05			0.48			0.32
Low fat
T1	394	366	1.00		32	1.00		127	1.00		74	1.00
T2	403	305	0.90	(0.73,1.12)	22	0.68	(0.40, 1.18)	112	0.97	(0.74, 1.28)	71	1.13	(0.80, 1.59)
T3	392	269	0.84	(0.63,1.11)	19	0.56	(0.29, 1.09)	98	0.92	(0.64, 1.32)	66	1.30	(0.84, 2.02)
p trend			0.19			0.12			0.70			0.38
Fruit
T1	390	349	1.00		26	1.00		126	1.00		73	1.00
T2	405	298	0.87	(0.71,1.08)	23	0.94	(0.54, 1.64)	105	0.85	(0.64, 1.13)	63	0.90	(0.64, 1.27)
T3	394	293	0.89	(0.71,1.12)	24	1.04	(0.58, 1.87)	106	0.87	(0.65, 1.18)	75	1.09	(0.77, 1.55)
p trend			0.31			0.86			0.45			0.77
Vegetables
T1	393	344	1.00		28	1.00		130	1.00		81	1.00
T2	402	321	0.87	(0.71,1.08)	24	0.95	(0.55, 1.66)	112	0.82	(0.62, 1.07)	77	0.91	(0.66, 1.25)
T3	394	275	0.73	(0.57,0.93)	21	1.07	(0.57, 2.00)	95	0.73	(0.54, 1.00)	53	0.60	(0.40, 0.89)
p trend			0.01			0.72			0.02			< .01
Red meat
T1	393	268	1.00		24	1.00		97	1.00		61	1.00
T2	404	298	1.01	(0.80,1.26)	29	1.18	(0.69, 1.99)	130	1.19	(0.90, 1.58)	66	0.95	(0.67, 1.34)
T3	392	374	1.16	(0.91,1.50)	20	0.93	(0.44, 1.95)	110	0.86	(0.62, 1.18)	84	1.13	(0.78, 1.63)
p trend			0.24			0.79			0.49			0.70
Fish
T1	393	290	1.00		21	1.00		106	1.00		59	1.00
T2	406	317	1.03	(0.83,1.28)	27	1.28	(0.73, 2.26)	113	0.95	(0.72, 1.25)	86	1.34	(0.96, 1.87)
T3	390	333	1.11	(0.89,1.39)	25	1.32	(0.73, 2.40)	118	1.07	(0.80, 1.42)	66	1.02	(0.71, 1.47)
p trend			0.34			0.34			0.55			0.83
Whole grains
T1	394	363	1.00		23	1.00		131	1.00		71	1.00
T2	406	295	0.83	(0.67,1.03)	26	1.23	(0.70, 2.17)	111	0.87	(0.67, 1.15)	58	0.85	(0.60, 1.21)
T3	389	282	0.83	(0.67,1.04)	24	1.24	(0.69, 2.22)	95	0.74	(0.56, 0.99)	82	1.29	(0.93, 1.80)
p trend			0.10			0.53			0.11			0.30
Refined grains
T1	390	281	1.00		26	1.00		93	1.00		61	1.00
T2	404	289	0.93	(0.74,1.17)	21	0.89	(0.49, 1.61)	99	1.04	(0.77, 1.42)	55	0.85	(0.58, 1.24)
T3	395	370	1.15	(0.90,1.48)	26	1.19	(0.61, 2.35)	145	1.41	(1.02, 1.96)	95	1.27	(0.88, 1.85)
p trend			0.28			0.39			0.02			0.10
Western diet
T1	393	287	1.00		25	1.00		103	1.00		60	1.00
T2	406	290	0.93	(0.74,1.16)	29	1.23	(0.74, 2.06)	101	0.90	(0.67, 1.20)	70	1.05	(0.74, 1.49)
T3	390	363	1.08	(0.83,1.40)	19	0.87	(0.42, 1.77)	133	1.15	(0.83, 1.58)	81	1.06	(0.72, 1.57)
p trend			0.60			0.89			0.52			0.75
Prudent diet
T1	393	346	1.00		30	1.00		130	1.00		81	1.00
T2	402	314	0.86	(0.70,1.07)	22	0.82	(0.47, 1.44)	102	0.77	(0.58, 1.01)	74	0.92	(0.66, 1.28)
T3	394	280	0.77	(0.61,0.97)	21	0.92	(0.51, 1.65)	105	0.84	(0.64, 1.12)	56	0.68	(0.47, 0.98)
p trend			0.03			0.47			0.13			0.03

^aAdjusted for age, sex, recent aspirin or NSAID use, long-term activity level, pack-years of cigarette smoking, dietary calcium, and energy intake

^bIncludes cases without tumor marker data

Energy-providing nutrients appeared to have minimal influence on rectal tumor mutations, with a few exceptions (Table 4). Most notably, higher levels of omega-3 fatty acids were statistically significantly associated with CIMP+ tumors. High levels of animal protein statistically significantly increased risk of rectal tumors, while high levels of vegetable protein were statistically significantly associated with reduced risk of rectal tumors. Dietary fiber was statistically significantly associated with reduced risk of rectal tumors overall as well as with reduced risk of TP53 and KRAS2 tumor mutations specifically. No significant interactions were detected by sex or age group.

Table 4.

Associations^a between nutrients and rectal tumors

	Control	All Casesb			CIMP+			TP53 Mutation			KRAS2 Mutation
	n	n	OR	(95% CI)	n	OR	(95% CI)	n	OR	(95% CI)	n	OR	(95% CI)
Calories (kcal)
T1	394	305	1.00		24	1.00		108	1.00		61	1.00
T2	402	299	0.96	(0.77, 1.19)	27	1.12	(0.65, 1.91)	111	0.98	(0.74, 1.30)	73	1.19	(0.84, 1.67)
T3	393	336	1.07	(0.86, 1.33)	22	0.94	(0.52, 1.67)	118	1.05	(0.79, 1.38)	77	1.26	(0.90, 1.76)
p trend			0.55			0.92			0.63			0.18
Fats (% kcal)
Saturated
T1	392	264	1.00		20	1.00		103	1.00		61	1.00
T2	407	338	1.15	(0.93, 1.44)	26	1.24	(0.68, 2.24)	127	1.11	(0.84, 1.47)	67	0.96	(0.68, 1.37)
T3	390	338	1.17	(0.93, 1.47)	27	1.45	(0.79, 2.63)	107	0.90	(0.67, 1.20)	83	1.22	(0.86, 1.72)
p trend			0.18			0.23			0.73			0.32
Monounsaturated
T1	392	281	1.00		18	1.00		100	1.00		65	1.00
T2	406	339	1.11	(0.90, 1.38)	34	1.81	(1.02, 3.20)	130	1.15	(0.87, 1.52)	72	0.92	(0.66, 1.30)
T3	391	320	1.00	(0.79, 1.26)	21	1.21	(0.62, 2.35)	107	0.93	(0.69, 1.27)	74	0.96	(0.67, 1.37)
p trend			0.99			0.65			0.71			0.74
Polyunsaturated
T1	393	312	1.00		24	1.00		105	1.00		64	1.00
T2	404	305	0.86	(0.69, 1.07)	28	1.13	(0.66, 1.95)	115	0.97	(0.73, 1.28)	74	1.02	(0.73, 1.43)
T3	392	323	0.93	(0.74, 1.16)	21	0.91	(0.50, 1.66)	117	1.01	(0.76, 1.36)	73	1.00	(0.70, 1.42)
p trend			0.51			0.79			0.98			0.86
Trans
T1	394	259	1.00		24	1.00		88	1.00		61	1.00
T2	405	364	1.26	(1.01, 1.56)	23	0.83	(0.46, 1.49)	139	1.41	(1.06, 1.87)	73	0.99	(0.70, 1.39)
T3	390	317	1.10	(0.88, 1.38)	26	1.00	(0.56, 1.79)	110	1.10	(0.82, 1.49)	77	1.08	(0.76, 1.54)
p trend				0.45		0.97			0.60			0.71
Omega-3c
T1	394	263	1.00		14	1.00		92	1.00		52	1.00
T2	400	356	1.28	(1.03, 1.59)	32	2.09	(1.13, 3.87)	121	1.12	(0.85, 1.50)	86	1.46	(1.03, 2.07)
T3	395	321	1.14	(0.92, 1.43)	27	1.92	(1.00, 3.66)	124	1.18	(0.88, 1.58)	73	1.19	(0.83, 1.72)
p trend			0.27			0.05			0.13			0.29
Protein (% kcal)
Animal
T1	396	276	1.00		25	1.00		104	1.00		66	1.00
T2	402	342	1.28	(1.03, 1.60)	23	0.88	(0.50, 1.56)	118	1.15	(0.87, 1.52)	83	1.30	(0.93, 1.81)
T3	391	322	1.38	(1.08, 1.75)	25	0.98	(0.54, 1.78)	115	1.28	(0.94, 1.73)	62	1.07	(0.72, 1.58)
p trend			< .01			0.93			0.11			0.48
Vegetable
T1	393	373	1.00		29	1.00		130	1.00		83	1.00
T2	405	322	0.88	(0.71, 1.08)	29	1.02	(0.62, 1.68)	111	0.89	(0.68, 1.16)	77	0.98	(0.72, 1.34)
T3	391	245	0.74	(0.59, 0.92)	15	0.59	(0.31, 1.10)	96	0.91	(0.69, 1.20)	51	0.75	(0.52, 1.07)
p trend			< .01			0.08			0.25			0.07
Carbohydrates (% kcal)
T1	394	355	1.00		28	1.00		125	1.00		83	1.00
T2	404	322	1.01	(0.81, 1.24)	21	0.76	(0.43, 1.34)	117	1.05	(0.80, 1.38)	74	0.99	(0.72, 1.36)
T3	391	263	0.89	(0.71, 1.12)	24	0.93	(0.53, 1.63)	95	0.94	(0.70, 1.27)	54	0.80	(0.55, 1.15)
p trend			0.34			0.74			0.50			0.24
Dietary fiber (g)
T1	393	348	1.00		30	1.00		128	1.00		76	1.00
T2	402	316	0.79	(0.63, 0.99)	25	0.93	(0.53, 1.63)	115	0.82	(0.62, 1.08)	71	0.86	(0.61, 1.21)
T3	394	276	0.58	(0.43, 0.77)	18	0.81	(0.36, 1.82)	94	0.58	(0.39, 0.84)	64	0.68	(0.43, 1.07)
p trend			< .01			0.32			< .01			0.03

^aAdjusted for age, sex, recent aspirin or NSAID use long-term activity level, pack-years of cigarette smoking, dietary calcium, and energy intake

^bIncludes cases without tumor marker data

^cOmega-3 fatty acids include pfa_18_3, pfa_18_4, pfa_20_4, pfa_20_5, pfa_22_5, pfa_22_6; pfa_20_4, and pfa_20_5 are omega-6 fatty acids that are broken down to Omega-3 fatty acids in the body

Discussion

These data suggest that dietary components and physical activity are associated with rectal cancer overall and more specifically with specific tumor mutations. Long-term vigorous physical activity was associated with reduced risk of having TP53 and KRAS2 tumor mutations; BMI was not associated overall with rectal cancer nor with specific types of tumor mutations. In general, associations were strongest when examining foods rather than nutrients, with the strongest statistically significant associations for vegetables, grain products, and dietary fiber.

BMI has not been associated with rectal cancer in most studies. Likewise, we did not observe an association between BMI and specific tumor mutations for rectal cancer. This is in contrast to our studies of colon cancer where we showed that obesity increased risk of colon cancer for both men and estrogen-positive women and the associations were strongest for people who had a KRAS2 mutation.

Few studies have shown that involvement in more physical activity, especially activity performed at a more vigorous level of intensity reduces risk of rectal cancer. Our data suggest a reduced risk of TP53 and KRAS2 mutations among those with the highest level of long-term vigorous physical activity. It is possible that our detailed assessment of activity enabled us to capture a dimension of activity that is most important for rectal cancer. TP53 and KRAS2 were detected in over 75% of rectal tumors. Thus, failure to detect associations with physical activity in other studies is not likely from different associations with between specific tumor mutations.

Of interest was our finding that high intake of low-fat dairy products reduced risk of CIMP+ tumors, while high intake of high-fat dairy products and omega-3 fatty acids increased the likelihood of developing a CIMP+ tumor. These associations with dietary fat were the only dietary factors consistently associated with CIMP+ rectal tumors of the nutrients considered in this study. Our prior assessment of CIMP+ tumors and colon cancer did not show any meaningful association between fat and CIMP+ tumors [21]. The biological mechanism for this association is not clear and certainly needs replication in other studies.

We observed that high intake of vegetables associated with reduced risk of TP53 and KRAS2; consumption of high levels of dietary fiber and eating a prudent dietary pattern in general were associated with reduced risk of KRAS2 mutations. The inverse association between high intake of vegetables overall and TP53 may indicate antioxidant properties of vegetables [39]. Our previous work has shown that antioxidants and factors associated with decreased inflammation are more likely reduce the risk of TP53 mutations [40]. Additionally, other studies suggest that TP53 mutations are linked to oxidative stress [39, 41]. Few studies have reported on acquired mutations and rectal tumors and those that have do not report on associations between vegetable intake and KRAS2 mutations [42]. Our previous work on colon cancer and KRAS2 mutations did are not reveal similar associations [36], although many differences in risk factors have been observed for colon and rectal cancers [3, 5, 35, 43, 44].

There are both limitations and strengths in the study. Because CIMP+ tumors rarely occur in rectal tumors, we were limited in our ability to examine these associations fully. Likewise, while we were able to obtain tumor blocks from most individuals, we were not able to obtain sufficient DNA from all blocks, which reduced the numbers available for analysis. However, while we did not obtain DNA from all blocks, this represents the largest study of rectal cancer to date. The study is rich in its availability of detailed diet and lifestyle data as well as information on the most commonly occurring mutations and epigenetic changes in rectal tumors. However, there are few studies have that examined the associations reported here; hence confirmatory studies are needed. Because of the number of comparisons made, it is possible that some associations were chance findings, further supporting the need for confirmatory studies.

In summary, our data suggest that being physically active reduces the risk of both TP53 and KRAS2 mutations in rectal tumors. Diets rich in fats, especially diets rich in omega-3 fatty acids, may increase risk of CIMP+ tumors. High levels of vegetables reduced the likelihood of having a TP53 and KRAS2 mutation. Additionally, high dietary fiber and a more prudent dietary pattern appeared to reduce the risk of a KRAS2 mutation. Although these findings need confirmation in other large studies of rectal cancer specifically, they nonetheless provide valuable insight into dietary and lifestyle associations with rectal cancer.