Atopic Allergic Conditions and Colorectal Cancer Risk in the Multiethnic Cohort Study

Abstract

Studies have provided evidence of an inverse association between atopic allergic conditions (AACs) and invasive colorectal cancer (CRC) incidence and mortality in predominantly white populations. We examined the association between AACs (asthma, hay fever, or allergy) and CRC among white, African-American, Native Hawaiian, Japanese-American, and Latino men and women in the Multiethnic Cohort Study within Hawaii and Los Angeles, California. The prospective analysis included 4,834 incident CRC cases and 1,363 CRC-related deaths ascertained between 1993 and 2010. We examined associations by ethnicity, location, stage, and potential effect modification by CRC risk factors. AACs were associated with a reduced risk of CRC incidence among both men and women (relative risk (RR) = 0.86, 95% confidence interval (CI): 0.80, 0.92). The reduction in risk was noted in all populations except Latinos and was significant in whites (RR = 0.85, 95% CI: 0.73, 0.98), African Americans (RR = 0.81, 95% CI: 0.70, 0.95), Native Hawaiians (RR = 0.72, 95% CI: 0.54, 0.96), and Japanese Americans (RR = 0.87, 95% CI: 0.78, 0.98). Individuals with AACs also had a 20% reduction in CRC-related mortality (P = 0.001). These findings provide evidence for the potential protective role of the reactive immune system in colorectal cancer.

Colorectal cancer (CRC) is one of the most common cancers with respect to incidence and mortality both globally and nationally (1, 2). There are a number of established environmental and lifestyle risk factors for CRC including diet, body mass index (weight (kg)/height (m)²), smoking status, and type 2 diabetes, as well as protective factors including hormone replacement therapy usage, statin and nonsteroidal antiinflammatory drug usage, and physical activity (3–8). Previous studies have reported differences in CRC risk among racial/ethnic groups after accounting for these various risk factors, including higher risks for Japanese-American men and women and for African-American women compared with whites (9).

Atopic allergic conditions (AACs) such as asthma, hay fever, and food allergies have been examined in connection with cancer risk, as these conditions may indicate a heightened immune response. A more reactive immune system may contribute to lower cancer risk. To date, studies have provided support for a protective association of AACs with pancreatic cancer, glioma, and a number of hematological malignancies, while a positive association has been noted with lung cancer risk (10–12). Studies investigating the association of AACs and cancers of the digestive tract (an important location for manifestation of mucosal inflammation) have had mixed results (12). A meta-analysis of 8 prospective and 8 case-control studies (including both incidence and mortality studies, published during 1985–2006) found no statistically significant evidence of an association between AACs, asthma, and atopy and CRC risk (13). A study that comprised Taiwanese individuals reported an inverse association between allergic rhinitis and rectal cancer, but not colon cancer (14). A more recent study within the Cancer Prevention Study II (CPS-II) cohort reported a nonsignificant 10% decrease in CRC incidence and a statistically significant 21% decrease in CRC mortality for individuals who have both hay fever and asthma, with weaker associations for those individuals reporting only 1 of these conditions (10, 15). The Iowa Women's Health Study also reported that having 2 or more atopic conditions was associated with a 42% reduction in CRC risk (16). Together, results from these more recent prospective studies have provided evidence for a protective association of allergic conditions with CRC risk in primarily white populations. A study is required to examine the association between atopic conditions and CRC in a multiethnic population, as the prevalence of both CRC and atopic diseases differs by race/ethnicity.

To further understand the relationship between AACs and CRC, we conducted a prospective analysis among white, African-American, Latino, Japanese-American, and Native Hawaiian men and women within the Multiethnic Cohort Study. In this study, we examined the association of AACs with CRC risk across populations and whether the association varies by CRC disease subgroups defined by site, location, and stage, as well as by known risk factors of this common cancer.

METHODS

The Multiethnic Cohort Study of Diet and Cancer includes over 215,000 men and women from Hawaii and California and comprises primarily 5 self-reported racial/ethnic populations: whites, African Americans, Latinos, Japanese Americans, and Native Hawaiians (17). Potential cohort members were recruited primarily through Department of Motor Vehicle license files, as well as through Health Care Financing Administration (Medicare) files for additional African-American individuals. Recruited participants were between 45 and 75 years of age and completed a detailed, 26-page self-administered questionnaire at entry into the cohort (baseline data, 1993–1996), which included basic demographic information, lifestyle factors (e.g., diet, exercise, medication use, and smoking history), and chronic medical conditions. AAC status was defined on the basis of a self-report of having been previously told by a physician that the respondent had asthma, hay fever, skin allergy, food allergy, or any other allergy (asked as a single question on the baseline questionnaire). Cohort linkage with the Surveillance, Epidemiology, and End Results (SEER) registries, which cover all of Hawaii and California, was used to identify incident cancer cases within the cohort, as well as additional details about CRC status (including location and stage). Deaths within the cohort were determined via linkage with death certificate files and supplemented with data from the National Death Index in both Hawaii and California. In this study, participants were followed until 1 of the following events: the identification of incident CRC cases, death, or end of follow-up (December 31, 2010, in both Hawaii and California). Individuals were excluded if they were not one of the 5 main ethnic groups (n = 13,988) or had a previous diagnosis of CRC by questionnaire or tumor registry linkage (n = 2,554), resulting in 199,112 individuals in the final analysis; 2,554 male and 2,280 female incident cases of invasive CRC were identified, and 735 men and 628 women died of CRC during the follow-up period (1993–2010).

Statistical analysis

Age-standardized CRC incidence rates were computed on the basis of the number of cases and person-years accumulated within 5-year age groups, using the US Census 2000 standard population. Cox regression was used to estimate all adjusted hazard ratios (reported as “relative risks”) for the association of AACs with CRC incidence and mortality. Individuals were censored at the time of death due to other causes or December 31, 2010. All models were adjusted as strata variables for age at entry into the cohort, sex, and ethnicity (white, African-American, Native Hawaiian, Japanese-American, Latino). The following covariates were included as terms in all log-linear models: smoking status (never, past, current); pack-years (<20, ≥20, or missing); educational level (≤12 years, some college/vocational, college graduate); body mass index (<23, 23–24.9, 25–29.9, 30–34.9, ≥35); and aspirin usage (none, past, current). Individuals missing information for these covariates were assigned a missing indicator variable in the analysis. Additionally, the following covariates were also examined: vigorous physical activity, alcohol intake, saturated fat intake, nonsaturated fat intake, dietary fiber intake, energy intake, and a family history of CRC. These variables were not included in the final model, because they did not alter the estimated relationship between CRC risk and status of AACs.

Relative risks were examined separately for males and females, for colon and rectal cancer, by site (left vs. right), and by stage (localized vs. regional/distant). Tests of heterogeneity by tumor characteristic were performed by using a fixed-effects meta-analysis. Individuals were excluded from subgroup analyses if they had multiple tumor site diagnoses on the same visit. When examining mortality, we separated subgroup analyses by incident cancer location only. We also examined whether the associations between AAC status and CRC risk were heterogeneous by sex, ethnicity, family history, body mass index, age, smoking status, aspirin usage, and educational group level using likelihood ratio tests for interaction between AACs and each of these variables. Finally, we also examined AAC status and CRC risk stratified by history of endoscopy (defined as colonoscopy or sigmoidoscopy), which was self-reported on the second questionnaire (approximately 6 years after baseline, 1999–2001). This analysis of those who returned the second questionnaire and answered the question about history of endoscopy contained 155,915 individuals (78.3% of the cohort), including 2,723 cases that were diagnosed after the collection of the second questionnaire (56.3% of total cases). All P values reported are 2-sided.

RESULTS

The mean age at entry for men (n = 89,496) was 60.7 (standard deviation, 8.9) years, and for women (n = 109,616) it was 60.0 (standard deviation, 8.8) years, with these ages being similar between AAC and non-AAC groups. The prevalence of AACs varied among ethnicities, from 16% and 24% in the Latino male and female populations to 24% and 37% in the white male and female populations. Age-standardized CRC incidence rates were higher in all non-AAC versus AAC groups (Table 1).

Table 1.

Baseline Characteristics of Analyzed Individuals by Sex and Status of AACs in the Multiethnic Cohort Study, 1993–2010

	All Individuals				Males Only				Females Only
	AACs		Non-AACs		AACs		Non-AACs		AACs		Non-AACs
	No.	%	No.	%	No.	%	No.	%	No.	%	No.	%
No. of subjects	51,973	26	147,139	74	17,997	20	71,499	80	33,976	31	75,640	69
Age at baseline, yearsa	59.38 (8.9)		60.81 (8.8)		59.53 (9.0)		60.97 (8.8)		59.30 (8.8)		60.66 (8.8)
Ethnicity
White	15,142	31	33,882	69	5,472	24	17,151	76	9,670	37	16,731	63
African-American	8,841	26	25,511	74	2,227	18	10,286	82	6,614	30	15,225	70
Native Hawaiian	4,064	28	10,304	72	1,307	21	4,930	79	2,757	34	5,374	66
Japanese-American	14,709	26	41,333	74	5,538	21	20,853	79	9,171	31	20,480	69
Latino	9,217	20	36,109	80	3,453	16	18,279	84	5,764	24	17,830	76
CRC cases	1,053	2.0	3,781	2.6	436	2.4	2,118	3.0	617	1.8	1,663	2.2
CRC incidence ratesb	109.4		131.1		139.0		153.6		94.3		111.1
Family history of CRCc		8.7		7.5		8.0		6.9		9.1		8.1
Endoscopy prevalencec,d,e
Yes		33.1		25.9		34.9		27.4		32.1		24.4
No		47.7		51.6		43.9		47.9		49.8		55.0
Body mass indexc,e,f
<23		23.8		22.2		16.0		15.7		28.4		28.8
23–24.9		17.9		18.1		19.9		19.3		16.8		17.0
25–29.9		35.7		38.6		45.1		46.4		30.7		31.1
30–34.9		13.9		13.6		13.6		13.5		14.0		13.6
≥35		7.1		5.6		4.4		4.0		8.3		7.0
Smoking statusc,e
Never		45.5		42.6		30.3		29.2		53.9		55.5
Past		40.5		38.6		54.4		49.9		32.6		27.6
Current		13.2		17.0		14.8		19.6		12.4		14.6
Educational levelc,e
≤12 years		38.4		46.4		34.2		43.0		40.9		49.7
Some college/vocational		31.4		28.0		30.7		28.8		31.8		27.2
College graduate		29.5		24.3		34.6		26.9		26.7		21.6
Physical activity, hours/dayc,e
Never		46.9		42.8		31.9		31.1		54.9		53.9
>0–0.21		16.0		15.5		16.1		15.4		15.8		15.5
>0.21–0.71		17.3		17.7		22.8		22.0		14.4		13.6
>0.71		15.6		18.6		26.4		27.6		9.9		10.0
Aspirin usec,e
Never		57.1		58.2		55.1		56.8		58.3		59.7
Past		18.7		17.2		17.8		17.0		19.1		17.4
Current		20.5		20.0		24.4		22.7		18.3		17.3

Abbreviations: AAC, atopic allergic condition; CRC, colorectal cancer; SD, standard deviation.

^a Mean (SD).

^b Expressed as incidence rate, age standardized (5-year groups) to US Census 2000 standard population with ages 45 years or above, per 100,000 person-years.

^c Age standardized (5-year age groups) and ethnicity/race standardized to the total population included in the study.

^d Reported on the second questionnaire (refer to Methods).

^e Does not total to 100% because of missing values.

^f Body mass index: weight (kg)/height (m)².

In both men and women, individuals with AACs were more likely to have higher educational levels (P < 0.001 among all ethnicities). Age- and race-adjusted family history of CRC was higher in all AAC groups than non-AAC groups (8.0% vs. 6.9% in men and 9.1% vs. 8.1% in women). In both sexes, AAC sufferers were more likely to be past smokers and less likely to be current smokers (P < 0.001 for all ethnic groups). Among both men and women, those with AACs reported a greater frequency of previous endoscopy than did non-AAC individuals in each population (men: AAC, 34.9% vs. non-AAC, 27.4%, P < 0.003; women: AAC, 32.1% vs. non-AAC, 24.4%, P < 0.001) (Table 1).

The risk of CRC, after adjustment for potential confounders, was 14% lower among those with AACs compared with those without AACs (relative risk (RR) = 0.86, 95% confidence interval (CI): 0.80, 0.92) (Table 2). This inverse relationship was found in each ethnic population, ranging from 0.72 (95% CI: 0.54, 0.96) in Native Hawaiians to 0.96 (95% CI: 0.81, 1.14) in Latinos. The association was statistically significant in whites (RR = 0.85; P = 0.03), African Americans (RR = 0.81; P = 0.01), and Japanese Americans (RR = 0.87; P = 0.01), and the test of heterogeneity in the risk estimates among ethnicities was not statistically significant (P_ethnicity = 0.36). The association with AACs was observed in both men (RR = 0.86, 95% CI: 0.77, 0.95) and women (RR = 0.86, 95% CI: 0.78, 0.94; P_{heterogeneity} = 0.94) (Table 2).

Table 2.

Estimated Relative Risk of CRC Incidence Associated With AAC Status by Ethnicity and Subgroup in the Multiethnic Cohort Study, 1993–2010

Subgroup	AAC Casesa	Non-AAC Casesa	RRb	95% CI	Pcethnicity
All
All	1,053	3,781	0.86d	0.80, 0.92	0.36
Whites	241	709	0.85	0.73, 0.98
African Americans	214	783	0.81	0.70, 0.95
Native Hawaiians	63	242	0.72	0.54, 0.96
Japanese Americans	363	1,325	0.87	0.78, 0.98
Latinos	172	722	0.96	0.81, 1.14
Mene
All	436	2,118	0.86d	0.77, 0.95	0.65
Whites	100	404	0.85	0.68, 1.06
African Americans	55	340	0.76	0.57, 1.01
Native Hawaiians	26	133	0.79	0.51, 1.20
Japanese Americans	175	797	0.89	0.76, 1.05
Latinos	80	444	0.94	0.74, 1.19
Womene
All	617	1,663	0.86d	0.78, 0.94	0.49
Whites	141	305	0.85	0.70, 1.04
African Americans	159	443	0.84	0.70, 1.01
Native Hawaiians	37	109	0.65	0.45, 0.96
Japanese Americans	188	528	0.86	0.72, 1.01
Latinos	92	278	0.99	0.78, 1.25
Right Colonf
All	501	1,600	0.94d	0.85, 1.04	0.23
Whites	120	332	0.88	0.71, 1.09
African Americans	115	400	0.84	0.68, 1.04
Native Hawaiians	26	83	0.84	0.54, 1.32
Japanese Americans	153	493	0.99	0.82, 1.19
Latinos	87	292	1.19	0.93, 1.52
Left Colonf
All	290	1,122	0.79d	0.70, 0.91	0.94
Whites	66	187	0.89	0.67, 1.19
African Americans	56	207	0.82	0.61, 1.10
Native Hawaiians	21	75	0.81	0.49, 1.32
Japanese Americans	104	439	0.74	0.59, 0.92
Latinos	43	214	0.79	0.56, 1.10
Rectumf
All	238	977	0.78d	0.67, 0.90	0.53
Whites	51	170	0.76	0.56, 1.05
African Americans	36	149	0.74	0.51, 1.07
Native Hawaiians	14	80	0.48	0.27, 0.85
Japanese Americans	97	379	0.84	0.67, 1.05
Latinos	40	199	0.85	0.60, 1.20
Localizedg
All	481	1,683	0.87d	0.78, 0.96	0.89
Whites	115	315	0.90	0.72, 1.12
African Americans	95	311	0.90	0.71, 1.13
Native Hawaiians	31	118	0.72	0.48, 1.07
Japanese Americans	170	624	0.85	0.72, 1.01
Latinos	70	315	0.90	0.69, 1.17
Regionalg
All	532	1,970	0.84d	0.76, 0.92	0.16
Whites	118	371	0.80	0.65, 0.99
African Americans	101	430	0.71	0.57, 0.88
Native Hawaiians	32	115	0.76	0.51, 1.14
Japanese Americans	185	674	0.89	0.75, 1.05
Latinos	96	380	1.01	0.80, 1.26

Abbreviations: AAC, atopic allergic condition; CI, confidence interval; CRC, colorectal cancer; RR, relative risk.

^a Subgroup totals may not add up to total number of cases as individuals were excluded if they had multiple tumors at the time of diagnosis.

^b Adjusted for age at entry (continuous), smoking status (status and pack-years), educational level, body mass index, nonsteroidal antiinflammatory drug use, and sex (except for sex-specific analyses).

^c Adjusted additionally for ethnicity.

^d Test for heterogeneity by ethnicity using likelihood ratio tests.

^e P_{heterogeneity} by sex = 0.94 (test for interaction via likelihood ratio test).

^f P_{heterogeneity} by stage = 0.04 (via fixed-effects meta-analysis).

^g P_{heterogeneity} by location = 0.64 (via fixed-effects meta-analysis).

When stratifying by cancer site, a greater reduction in risk was noted for rectal cancer (RR = 0.78, 95% CI: 0.67, 0.90) and left-sided CRC (RR = 0.79, 95% CI: 0.70, 0.91) compared with right-sided CRC (RR = 0.94, 95% CI: 0.85, 1.04; P_{heterogeneity} for site = 0.04) (Table 2). An association with AACs was also found for both localized (RR = 0.87, 95% CI: 0.78, 0.96) and regional (RR = 0.84, 95% CI: 0.76, 0.92; P_{heterogeneity} for stage = 0.64) CRC.

In analyses stratified by CRC risk factors, we observed no significant interaction between AAC status and the following CRC risk factors: age of entry, family history of CRC, body mass index, aspirin usage, smoking status, and level of education (Web Table 1 available at http://aje.oxfordjournals.org/). However, there was suggestive evidence of effect modification by smoking status, with a greater reduction in risk observed in current smokers (RR = 0.74, 95% CI: 0.61, 0.90) compared with past smokers (RR = 0.86, 95% CI: 0.77, 0.95) and never smokers (RR = 0.90, 95% CI: 0.80, 1.00; P_interaction = 0.23).

To examine the impact of endoscopy on these results, given the higher endoscopy rates in those reporting AACs, we performed an analysis stratified by a report of endoscopy on the second questionnaire (1999–2001; refer to Methods). The relative risk in this analysis subgroup was similar to that in the entire cohort (RR = 0.86). Among individuals that reported a history of having had an endoscopy (n = 54,431; 847 cases), the relative risk was 0.88 (95% CI: 0.75, 1.02), while in individuals who reported never having a history of endoscopy (n = 100,328; 1,876 cases), the relative risk was 0.86 (95% CI: 0.77, 0.97) (Web Table 2).

The inverse association between AAC status and CRC risk was also noted with CRC mortality (RR = 0.80, 95% CI: 0.70, 0.91) (Table 3). The association was similar across ethnic groups (P_ethnicity = 0.14) and was statistically significant in whites and Native Hawaiians. When stratified by sex, males experienced a greater reduction in risk (RR = 0.67, 95% CI: 0.55, 0.83) compared with females (RR = 0.91, 95% CI: 0.76, 1.08; P_{heterogeneity} = 0.041). Across all ethnic populations, the risk of mortality due to regional/aggressive CRC associated with AACs was significantly reduced (RR = 0.80, 95% CI: 0.69, 0.94; P_ethnicity = 0.36). Mortality risk estimates between regional/aggressive and localized CRC (RR = 0.75, 95% CI: 0.51, 1.11) were similar (P_{heterogeneity} = 0.68), although there was a limited number of cases in the localized group (n = 166).

Table 3.

Estimated Relative Risk of CRC Mortality Associated With AAC Status by Ethnicity and Subgroup in the Multiethnic Cohort Study, 1993–2010

Subgroup	AAC Casesa	Non-AAC Casesa	RRb	95% CI	Pethnicityc
All
All	275	1,088	0.80d	0.70, 0.91	0.14
Whites	63	226	0.74	0.56, 0.98
African Americans	70	284	0.74	0.57, 0.97
Native Hawaiians	10	65	0.44	0.23, 0.87
Japanese Americans	84	297	0.93	0.73, 1.19
Latinos	48	216	0.91	0.66, 1.24
Mene
All	101	634	0.67d	0.55, 0.83	0.65
Whites	26	138	0.67	0.44, 1.02
African Americans	16	125	0.58	0.35, 0.98
Native Hawaiians	4	39	0.40	0.14, 1.13
Japanese Americans	37	192	0.81	0.56, 1.15
Latinos	18	136	0.69	0.42, 1.14
Womene
All	174	454	0.91d	0.76, 1.08	0.21
Whites	37	88	0.82	0.55, 1.20
African Americans	54	155	0.82	0.60, 1.12
Native Hawaiians	6	26	0.46	0.19, 1.12
Japanese Americans	47	105	1.09	0.77, 1.55
Latinos	30	80	1.15	0.75, 1.75
Right Colonf
All	128	434	0.89d	0.73, 1.09	0.39
Whites	34	90	0.97	0.65, 1.45
African Americans	33	141	0.68	0.46, 0.99
Native Hawaiians	5	24	0.54	0.20, 1.43
Japanese Americans	35	105	1.12	0.75, 1.65
Latinos	21	74	1.11	0.68, 1.81
Left Colonf
All	68	259	0.85d	0.65, 1.12	0.78
Whites	12	48	0.70	0.37, 1.33
African Americans	16	57	0.83	0.47, 1.45
Native Hawaiians	3	15	0.73	0.21, 2.57
Japanese Americans	25	86	0.93	0.59, 1.47
Latinos	12	53	0.96	0.51, 1.81
Rectumf
All	51	278	0.62d	0.46, 0.84	0.51
Whites	11	49	0.64	0.33, 1.25
African Americans	9	52	0.60	0.29, 1.23
Native Hawaiians	1	19	0.13	0.02, 0.97
Japanese Americans	19	91	0.70	0.43, 1.16
Latinos	11	67	0.70	0.37, 1.34
Localizedg
All	32	134	0.75d	0.51, 1.11	0.06
Whites	9	29	0.88	0.41, 1.87
African Americans	7	34	0.59	0.26, 1.34
Native Hawaiians	0	10
Japanese Americans	12	32	1.21	0.62, 2.39
Latinos	4	29	0.59	0.21, 1.69
Regionalg
All	209	821	0.80d	0.69, 0.94	0.36
Whites	44	163	0.72	0.52, 1.01
African Americans	49	208	0.71	0.52, 0.98
Native Hawaiians	10	47	0.59	0.30, 1.18
Japanese Americans	68	242	0.92	0.70, 1.22
Latinos	38	161	0.95	0.67, 1.36

Abbreviations: AAC, atopic allergic condition; CI, confidence interval; CRC, colorectal cancer; RR, relative risk.

^a Subgroup totals may not add up to total number of cases as individuals were excluded if they had multiple tumors at the time of diagnosis.

^b Adjusted for age at entry (continuous), smoking status (status and pack-years), educational level, body mass index, nonsteroidal antiinflammatory drug use, and sex (except for sex-specific analyses).

^c Adjusted additionally for ethnicity.

^d Test for heterogeneity by ethnicity using likelihood ratio tests.

^e P_{heterogeneity} by sex = 0.041 (test for interaction via likelihood ratio test).

^f P_{heterogeneity} by stage = 0.015 (via fixed-effects meta-analysis).

^g P_{heterogeneity} by location = 0.68 (via fixed-effects meta-analysis).

DISCUSSION

In this prospective analysis of 5 racial/ethnic populations, past AACs (asthma, allergies, or hay fever) were found to be associated with a statistically significant 14% decrease in risk of developing CRC. We observed a similar inverse association among both men and women for colon and rectal cancer, as well as for localized and regional disease, with some suggestion of lower risk for left-sided CRC and rectal cancer versus right-sided colon cancer (P_{heterogeneity} = 0.04). The inverse association was noted in all ethnic populations (RRs = 0.72–0.96), although the magnitude of the association was weaker in Latinos for overall disease as well as for all disease subgroup analyses. Consistent patterns of association were also observed with CRC mortality, with AACs associated with a reduction in risk of CRC-related death in all populations. Males had a greater reduction in risk (RR = 0.67) compared with females (RR = 0.91).

In previous studies, a meta-analysis that combined studies of CRC incidence and mortality reported no statistically significant association between asthma or any allergy and CRC incidence/mortality. Findings in the current study on CRC incidence, overall and by subsite, are also directionally consistent and similar in magnitude to more recent associations between CRC incidence and hay fever and/or asthma reported in the CPS-II Nutrition and Taiwanese cohorts. These studies had substantially fewer incident CRC cases (442 cases in exposed groups within the CPS-II Nutrition Cohort and about 400 cases in the Taiwanese cohort), and not all results were statistically significant (14, 15). A 26% reduction in CRC risk associated with allergies was also reported in the Iowa Women's Health Study (16). Our findings with CRC mortality were also similar to that in the CPS-II mortality cohort, which reported a 21% decrease in CRC-related mortality associated with both hay fever and asthma (with weaker associations among individuals that reported only 1 of these conditions) (15). Despite there being evidence of similarities in results across studies, it is difficult to draw direct comparisons between the above studies and the current study, as the methods of categorizing exposures are different. For example, the CPS-II and Taiwanese studies examined 1 or more of the following: allergic rhinitis (hay fever), atopic dermatitis, or asthma, although neither study examined all allergies together, as was done in the current study.

The biological basis for the apparent protective association of AACs with CRC is not known but could be due to antitumor effects of type I immunoglobulin E-mediated immune activity, in which migrating eosinophils, mast cells, and other immune cells are present within the highly vascularized and rapidly overturning intestinal mucosa. Atopic allergens can enter the system at numerous points, including the respiratory system (resulting in asthma) and the digestive system (promoting food allergies), where mast cells and eosinophils are ubiquitous (18). This can lead to type I immunoglobulin E-mediated hypersensitivity reactions, such as asthma or allergic rhinitis (hay fever), due to the overstimulation of eosinophils and mast cells, as well as an excessive T-helper cell type II response (through activation by immunoglobulin E and tumor cell complexes). Mast cells can release factors that enhance permeability and inflammation when degranulated by immunoglobulin E and tumor cell complexes. Other immune responses activated in this scenario can include immunoglobulin E antibody-dependent cellular cytotoxicity directed at precancerous and cancerous cells and direct anticancer effects (19). For example, eosinophil levels, which are inversely associated with CRC, are thought to have cytotoxic effects and can mediate antitumor-like activity on precancerous cells that arise in rapidly developing tissues, such as the gut lining (16, 20, 21). Asthmatic persons are already known to have increased eosinophilic presence in the respiratory mucosal lining, and together with mast cell degranulation, this may reflect an increased presence of leukocytes in the gut of individuals with atopic conditions (22). Another potential theory, known as the prophylaxis hypothesis, contends that allergic responses and inflammation in mucosal surfaces can lead to more rapid clearance of mutagenic triggers (12).

A limitation of this study is that the exposure was nonspecific and based on one's self-report of any 1 of the following atopic allergic conditions: asthma, hay fever, skin allergy, food allergy, or any other allergy. Consequently, we were unable to analyze each atopic allergic condition separately or to examine whether a participant with more than 1 condition may represent a more exposed group. However, asthma, hay fever, and other allergic conditions have a similar biological process based on atopy (hypersensitization toward an external trigger) and are conditions that are highly correlated in adults (23–25). Another limitation of the current study is that the analysis utilized baseline data in the cohort and thus did not capture new adult-onset AACs, which would most likely result in nondifferential misclassification of AAC individuals into the non-AAC group and a subsequent underestimate of the association of AACs with CRC risk.

Individuals with AACs were noted to have higher endoscopy rates than those without AACs (33.1% vs. 25.9%). As endoscopy usage has the potential to both reduce the risk of CRC and increase the rate of CRC capture, we performed an analysis stratified by endoscopy status reported on the second questionnaire. However, when examining the population that did not report having an endoscopy, we found that the risk of incident CRC was still significantly decreased (RR = 0.86), with an inverse association observed in all ethnic groups. This indicates that the reduction in risk associated with AACs is independent of endoscopy status.

We conducted analyses stratified by CRC risk factors to elucidate subgroups in which AACs may be more protective, as well as potential biological CRC disease mechanisms that may be influenced by AACs. In these analyses, AAC status was associated with lower risk in current smokers compared with past and never smokers (RR = 0.74 vs. 0.86 and 0.90). One potential explanation may involve the modulation of the T-helper cell type I and T-helper cell type II responses and resultant modulation of inflammatory cytokine profiles in response to smoke exposure. Another possibility is that this association may be due to a greater number of premalignant cells in the gut serving as targets for the inflammation process to act upon in the environments of smokers, compared with those where the burden of carcinogens is lower as in past and/or never smokers. This suggestive effect modification needs additional investigation. In contrast, no evidence of interaction by smoking status was observed in the CPS-II cohorts, although the authors reported on only the combined condition of hay fever and asthma (15).

Future work to understand the mechanistic role of AACs in CRC development should include examination of the association between AACs and polyps (detailed data on which are not available in the Multiethnic Cohort Study). Specifically, it would be interesting to examine whether AACs reduce CRC risk by preventing polyp formation, or if AACs reduce CRC risk by preventing polyp progression to CRC. Finally, the analysis of serum-specific immune markers, such as immunoglobulin E or eosinophil count, may provide additional insight into the association between AACs and CRC (12). However, self-reported history of allergic diagnoses may capture aspects of immune function that are not highly associated with a single measurement of immunoglobulin E (26).

In this large prospective study among 5 racial/ethnic populations, we observed a significant inverse association between AACs and the development of CRC in both men and women. These findings support results from previous prospective studies linking atopic allergic conditions, such as asthma, hay fever, and food allergies, and the development of colorectal cancer, and they further highlight the importance of the immune system in the pathogenesis of CRC.