Composite Dietary Antioxidant Index and the Risk of Colorectal Cancer: Findings from the Singapore Chinese Health Study

Abstract

Colorectal cancer (CRC) is a major contributor to cancer death globally. Several studies showed some protections by certain individual dietary antioxidants against CRC development. Epidemiologic data on the composite dietary antioxidant index (CDAI) in relation to CRC risk are sparse. Using the Singapore Chinese Health Study, an ongoing prospective cohort consisting of 61,321 cancer-free participants aged 45–74 years at baseline, a food-based CDAI was calculated according to a previously established and validated method that included six food-sourced antioxidants including vitamins A, C, and E, manganese, selenium, and zinc. Cox proportional hazard regression method was used to estimate the hazard ratios (HRs) and their 95% confidence intervals (CIs) for CRC associated with various levels of CDAI with adjustment for multiple potential confounders. After an average of 17.5 years of follow-up, 2,140 participants developed CRC. HRs (95% CIs) of CRC for quartiles 2, 3 and 4 of CDAI were 0.94 (0.83–1.07), 0.86 (0.75–1.00) and 0.80 (0.66–0.98), respectively, compared with the lowest quartile (P_trend=0.02). This inverse association between CDAI and CRC risk was more apparent in women or those without a history of diabetes, without family history of CRC, never smokers or overweight/obese individuals. However, none of the heterogeneity tests for the CDAI-CRC risk association reached statistical significance. Our findings suggest that food-based antioxidants may be beneficial for reducing the risk of CRC in the general population.

INTRODUCTION

Approximately 10% of annual global cancer deaths are due to colorectal cancer (CRC).¹ The second and the third most commonly diagnosed cancer is CRC among women and men, respectively.² It is predicted that there will be 2.5 million new cases of CRC worldwide by 2035.^1,3 The age-standardized incidence rate of CRC in Singapore was 39.9 and 28.2 per 100,000 people in males and females, respectively; and the incidence rate increases about 2% annually.⁴ Established risk factors for CRC are alcoholic drinking, intake of red or processed meat, obesity, physical inactivity, and type 2 diabetes.⁵ Possible protective agents for CRC are dietary fibers, whole grains, and calcium supplements.⁶

Antioxidants have critical roles on a cellular level. A low level of antioxidants intake may result in high production of reactive oxygen species (ROS).⁷ Under normal conditions, ROS products are produced in the mitochondrial respiratory chain among aerobic cells.⁸ The excessive generation of ROS, which creates oxidative stress, may damage the cell structure and DNA.⁷ As a result of chronic cellular damage, carcinogenesis processes would be more likely to occur. The tumorigenic features of ROS have been indicated as a possible therapeutic target.⁹

Epidemiological studies on individual dietary antioxidants such as vitamins and minerals and the risk of CRC have been conducted previously and produced inconsistent results.^10–31 Randomized clinical trials evaluating the antioxidant supplementation did not yield consistent data in support of the protective effect against the development of CRC. A meta-analysis including 22 randomized clinical trials (n=161,045) did not show a statistically significant protective effect of individual antioxidants supplements (i.e., vitamin A, vitamin E, selenium, and beta-carotene) against the development of all cancer types including CRC.³⁰ Similarly, another systematic review and meta-analysis did not find supportive evidence of individual or combination of antioxidant supplements on the protection against the development of gastrointestinal cancers including CRC.³¹ In addition, the effect of food-based antioxidant elements, including vitamin A, C, E, manganese, selenium, or zinc on the development of CRC have not been well studied in large prospective cohort studies.

Given the fact that single antioxidants such as vitamin C, vitamin E, selenium, manganese, and zinc may not be representative to capture an individual’s total consumption of antioxidants, a total antioxidant capacity (TAC) was used to evaluate its relationship with the risk of CRC in the Health Professional Follow-up Study (HPFS) among US men.³² The TAC score was derived from the ferric-reducing ability of plasma assay, which may only capture one aspect of the antioxidant activity in vivo. This study found a null association for food and supplement-derived TAC with CRC risk overall, and an moderate inverse association with rectal cancer risk.³² As the association between combined dietary antioxidants and the risk of CRC remains unclear, the goal of the current analysis was to fill such a knowledge gap.

Composite Dietary Antioxidant Index (CDAI) is a summary score of multiple dietary antioxidants including vitamins A, C, and E, manganese, selenium, and zinc that represents an individual’s antioxidant profile. The CDAI was developed according to their aggregate effect on anti-inflammation based on markers such as IL-1β and TNF-α,⁸ both of which are pro-inflammatory and associated with many health outcomes including CRC.^33,34 Previous study found an inverse association between the combined dietary antioxidant index and the risk of lung cancer³⁵, but similar efforts in CRC has not yet conducted. In addition, we proposed to derive the CDAI from food only given the lack of protective effect of antioxidant supplementation on cancer risk in numerous clinical trials as described above. The objective of the current study was therefore to evaluate the association between the food-derived CDAI and CRC risk in the Chinese Singapore Health Study (SCHS), an ongoing prospective cohort study of more than 63,000 Chinese Singaporeans. We also conducted subgroup analysis stratified by selected characteristics of study participants that may influence the purported association between CDAI and CRC risk.

METHODS

Study Population

Data for the current analysis was obtained from the SCHS, which was described in detail elsewhere.³⁶ Briefly, the SCHS is an ongoing population-based prospective cohort study that recruited 63,257 men and women, aged 45–74 from two main dialect groups of Chinese in Singapore (i.e., Hokkiens and Cantonese) who resided in the government-built housing estates between April 1993 and December 1998. The two Chinese dialect groups, the Hokkiens, and the Cantonese, were originated from the Fujian and Guangdong provinces, respectively, in southern China. Written informed consent was obtained from all study participants. The SCHS has been approved by the Institutional Review Boards of the National University of Singapore and the University of Pittsburgh.

The primary goal of the Singapore Chinese Health Study was to investigate the role of dietary and other lifestyle factors, environmental exposures, and genetic factors in the development of cancer.³⁶ Any individual with a history of cancer at baseline was ineligible for the cohort study. After excluding 1,936 individuals who had a history of invasive cancer, the present analysis included 61,321 participants who were free of cancer at baseline.

Dietary Assessment

Dietary assessment in SCHS was performed using a semi-quantitative food frequency questionnaire (FFQ) that was validated against a series of 24-hour dietary recall interviews³⁷ and selected biomarker studies on random subsets of cohort participants (n=810).^38,39 The FFQ contained 165 food items and food groups commonly consumed by the Chinese in Singapore. Study participants were asked how frequently (in 8 categories: ranging from “never or hardly ever” to “two or more times a day”) they consumed the food or food group followed by a question on the portion size, which was assisted with photographs showing three portion sizes (i.e., small, medium, and large). The average daily intake of approximately 100 nutrients and non-nutrient compounds was calculated for each study participant using the Singapore Food Composition Database.³⁷ Between April 1994 and March 1997, the FFQ was validated against two 24-hour dietary recalls, one on a weekday and the other on a weekend, among a random sample of 810 participants. The two 24-hour dietary recall surveys were approximately two months apart. The correlation coefficients among most of the calorie-adjusted nutrients derived from the 24-hour dietary recalls and the FFQ ranged from 0.24 to 0.79.¹⁷

Assessment of Other Covariates

Information on demographics, lifestyle factors, and history of disease and family history of cancer of participants was also collected through in-person interviews using the same questionnaire as the FFQ. Based on the collected information, the following covariates were created for the present analysis: cigarette smoking (never or ever smokers), history of type 2 diabetes mellitus (no, yes), and family history of CRC (no, yes). In addition, weekly physical activity (no, yes) was defined as having at least 1 of any weekly physical activities according to the self-reported number of hours per week spent on leisure-time strenuous sports (jogging, cycling on hills, tennis, swimming, tennis, or aerobics), vigorous work (moving heavy items, shoveling, or related labor work), and moderate activities (walking, cycling on flat ground, Tai Chi, and Chi Kung).⁴⁰ The body mass index (BMI) was calculated by weight in kg divided by height in meter squared (kg/m²). According to the recommendations by the World Health Organization (WHO) for Asian populations,^41,42 individuals with BMI 23 to <27.0 and 27.0 or higher were classified as being overweight and obese, respectively. Alcohol consumption was categorized into heavy or light/none drinking. A heavy drinking was defined as alcohol consumption of 15 or more drinks/week for men or 8 or more drinks/week for women, according to the criteria recommended by the US Centers for Disease Control and Prevention (CDC).⁴³

CDAI Calculation

The development of CDAI was described previously,³³ and validated using serum anti-inflammatory makers in another prospective cohort study.⁸ For each subject, the CDAI was calculated by summing consumption of 6 antioxidants (i) including vitamins A, C, and E, manganese, selenium, and zinc from food sources only (i.e., dietary supplements were excluded) as follows:

$$\mathit{CDAI}={\sum }_{i=1}^6\frac{x_i-{\mu }_i}{S_i}$$

Here, x_i was the daily intake of antioxidant i; μ_i was the mean of x_i over the entire cohort for antioxidant i; S_i was the standard deviation for μ_i.

It is also noted that 7.7% of the SCHS participants used any dietary supplements of antioxidants (i.e., vitamin A, vitamin C, vitamin E, selenium, and zinc). We did not collect information on use of manganese supplement, which was expected to be similarly low use in our study population. We conducted analyses and did not find any significant association between the use of any dietary supplements and risk of CRC (Supplementary Table 4).

Ascertainment of Colorectal Cancer Cases

All incident cases of cancer and death among the cohort participants were identified through the record linkage analysis with the nationwide Singapore Cancer Registry and the Singapore Registry of Births and Deaths, respectively. Colon cancer was defined by the International Classification of Disease-Oncology 2^nd Edition (ICD-O-2) C18 and rectal cancer by ICD-O-2 C19–C20.⁴⁴ The ascertainment of incident cancer and death cases among the cohort participants was virtually complete given that only 56 (<0.1%) original participants were known to be lost to follow-up due to their migration out of Singapore.

Statistical Analysis

We used the chi-square test for the frequencies of categorical or nominal variables or t-test for continuous variables between the CRC status or across quartiles of CDAI. Person-years for each participant at risk were calculated from the date at baseline interview to the date of CRC diagnosis, death, migration out of Singapore, or December 31, 2015, whichever occurred first.

Cox proportional hazard regression model was used to estimate the hazard ratio (HR) and its 95% confidence interval (CI) for colon and rectal cancers combined and separately associated with higher quartiles of CDAI relative to the lowest quartile. The linear trend was tested for both continuous values and quartile in ordinal values (1, 2, 3, and 4) with the risk of CRC. Proportional hazard assumption was tested using the Schoenfeld residuals test, and no violation was found. Potential confounders were adjusted in the Cox proportional hazard regression model, including daily calorie intake (Kcal/d), age (years), sex (male vs. female), dialect group (Hokkien vs. Cantonese), level of education (no formal education, primary school, secondary or higher education), year of enrollment into the study (1993–1995 vs. 1996–1998), BMI (<20, 20–23.9, 24–27.9, or ≥28 kg/m²), cigarette smoking status (ever vs. never), alcohol consumption (non/light vs. heavy drinking), history of type 2 diabetes (yes vs. no), and family history of cancer (yes vs. no).

We used the same Cox regression method for analysis stratified by level of BMI <23 and ≥23 kg/m²), presence or absence type 2 diabetes or family history of CRC, cigarette smoking (never, ever), and heavy alcohol drinking (no, yes). An interaction term for CDAI with these stratifying variables was included in the multivariable Cox regression models to examine the potential interaction effect on CRC risk. Sensitivity analysis was performed on the subset of data after excluding person-years and CRC cases occurred within the first two years post enrollment.

All statistical analyses were performed using SAS statistical package, version 9.4 (SAS Institute, Cary, NC). All P values are two-sided with considering P<0.05 as statistically significant.

RESULTS

After a mean (standard deviation-SD) follow-up of 17.5 (5.4) years, we identified 2,140 incident CRC cases among 61,321 participants who were free of cancer at baseline. The mean (SD) age of participants at enrollment was 59.5 (8.0) and 56.3 (8.0) years for CRC cancer cases and non-case, respectively. Participants who developed CRC were older, more likely to be male, ever smokers, and heavy drinkers. CRC cases also were more likely to have a family history of CRC, but lower intake of vitamins A, C, and E (all P’s <0.01) (Supplementary Table 1).

The proportions of men, Cantonese, ever smokers, heavy drinkers, people with higher education or weekly physical activities or any use of supplement antioxidant increased with increasing quartile of CDAI (Table 1). On the other hand, the frequency of participants with a history of type 2 diabetes decreased with increasing CDAI. Individuals with higher CDAI consumed more calories.

Table 1.

Distributions of Baseline Characteristics among Study Participants by Diet Composite Dietary Antioxidant Index, the Singapore Chinese Health Study, 1993–2015

	Total # Subjects	Composite Dietary Antioxidant Index in Quartile				P-value
Characteristics		Q1(−13.3, −4.9)	Q2(−4.9, −2.3)	Q3(−2.3, 1.0)	Q4(1.0, 58.6)
Number of subjects	61,321	15,330	15,330	15,331	15,330
Mean age (±SD)a, year	56.4 (8.0)	58.6 (8.2)	56.7 (7.9)	55.6 (7.8)	54.6 (7.5)	<0.001
Sex (%)
Male	27,293 (44.5)	4,953 (32.3)	6,091 (39.7)	7,376 (48.1)	8,873 (57.9)	<0.001
Female	34,028 (55.5)	10,377 (67.7)	9,239 (60.3)	7,955 (51.9)	6,457 (42.1)
Dialect
Cantonese	28,325 (46.2)	5,571 (36.3)	6,860 (44.8)	7,626 (49.7)	8,268 (53.9)	<0.001
Hokkien	32,996 (53.8)	9,759 (63.7)	8,470 (55.2)	7,705 (50.3)	7,062 (46.1)
Highest level of education (%)
No formal education	16,661 (27.2)	6,592 (43.0)	4,428 (28.9)	3,284 (21.4)	2,357 (15.4)	<0.001
Primary school	27,224 (44.4)	6,269 (40.9)	7,033 (45.9)	7,164 (46.7)	6,758 (44.1)
Secondary school or higher	17,436 (28.4)	2,469 (16.1)	3,869 (25.2)	4,883 (31.9)	6,215 (40.5)
Mean BMI (±SD), Kg/m2	23.1(3.3)	23.0 (3.2)	23.1 (3.3)	23.1 (3.3)	23.2 (3.3)	0.41
Smoking status (%)
Never smoker	42,583 (69.4)	10,812 (70.5)	10,913 (71.2)	10,655 (69.5)	10,203 (66.6)	<0.001
Ever smoker	18,738 (30.6)	4,518 (29.5)	4,417 (28.8)	4,676 (30.5)	5,127 (33.4)
Alcohol consumption
Heavy drinker	1,007 (1.6)	195 (1.3)	185 (1.2)	257 (1.7)	379 (2.4)	<0.001
Non-heavy drinker	60,314 (98.4)	15,135 (98.7)	15,145 (98.8)	15,074 (98.3)	14,960 (97.6)
Weekly physical activityb
No	41,083(67.0)	11,920 (77.8)	10,668 (69.6)	9,726 (63.4)	8,769 (57.2)	<0.001
Yes	20,238(33.0)	3,410 (22.2)	4,662 (30.4)	5,605 (36.6)	6,561 (42.8)
Diabetes history (%)
No	55,852 (91.1)	13,764 (89.8)	13,949 (91.0)	14,055 (91.7)	14,084 (91.9)	<0.001
Yes	5,469 (8.9)	1,566 (10.2)	1,381 (9.0)	1,276 (8.3)	1,246 (8.1)
Family history of colorectal cancer (%)
No	59,972 (97.8)	15,082 (98.4)	15,015 (98.0)	14,960 (97.6)	14,915 (97.3)	<0.001
Yes	1,349 (2.2)	248 (1.6)	315 (2.0)	371 (2.4)	415 (2.7)
Any use of Supplement Antioxidant
No	56,606(92.3)	14,593(95.2)	14,218(92.8)	13,977(91.2)	13,818(90.1)	<0.001
Yes	4,715(7.7)	737(4.8)	1,112(7.2)	1,354(8.8)	1,512(9.9)
Mean energy intake (±SDa), Kcal/day	1,556.6 (566.2)	1008.1 (220.7)	1,343.5 (223.3)	1,646.6 (277.8)	2,228.3 (548.8)	<0.001

^aBMI, body mass index; SD, standard deviation

^bsee definition in the Methods

The association between CDAI and the risk of CRC is presented in Table 2. Higher levels of CDAI in either quartile or continuous values were associated with a lower risk of CRC. Compared to the lowest quartile, HRs of CRC for the highest quartile was 0.80 (0.66–0.98) (P_trend=0.02) after adjustment for age, sex, dialect, education level, BMI, smoking status, alcohol consumption, history of type 2 diabetes, physical activity, family history of CRC, and total calorie (energy) intake. When data were analyzed separately by subsite, higher CDAI quartiles were mainly associated with lower risk of colon cancer (HR_Q4vs.Q1=0.77, 95% CI: 0.60–1.00) than rectal cancer (HR_Q4vs.Q1 =0.85, 95% CI: 0.61–1.17) (Table 2).

Table 2.

Association Between Diet Composite Dietary Antioxidant Index (CDAI) and Risk of Colorectal, Colon and Rectal Cancer in the Singapore Chinese Health Study, 1993–2015

CDAI in quartile	Colorectal cancer			Colon Cancer		Rectal Cancer
	Person-year	Cases	HRa (95% CI)	Cases	HRa (95% CI)	Cases	HRa (95% CI)
Overall
Q1 (−13.3, −4.9)	258,240	582	1.00	372	1.00	210	1.00
Q2 (−4.9, −2.3)	268,428	545	0.94 (0.83–1.07)	354	0.97 (0.83–1.14)	191	0.89 (0.72–1.10)
Q3 (−2.3, 1.0)	271,862	507	0.86 (0.75–1.00)	325	0.89 (0.74–1.06)	182	0.82 (0.65–1.05)
Q4 (1.0, 58.6)	275,831	506	0.80 (0.66–0.98)	304	0.77 (0.60–1.00)	202	0.85 (0.61–1.17)
Ptrend			0.02		0.05		0.21
Continuous scale (per SD increase)			0.92 (0.83–1.02)		0.87 (0.76–0.99)		1.02 (0.87–1.19)
Men
Q1 (−13.0, −4.9)	75,239	215	1.00	120	1.00	95	1.00
Q2 (−4.9, −2.3)	99,408	261	0.99 (0.82–1.20)	149	1.00 (0.78–1.29)	112	0.98 (0.73–1.31)
Q3 (−2.3, 1.0)	125,041	292	0.94 (0.76–1.15)	183	1.02 (0.78–1.33)	109	0.83 (0.60–1.15)
Q4 (1.0, 58.6)	154,817	347	0.93 (0.71–1.21)	204	0.92 (0.65–1.30)	143	0.95 (0.63–1.45)
Ptrend			0.50		0.72		0.56
Continuous scale (per SD increase)			0.98 (0.85–1.12)		0.87 (0.72–1.04)		1.16 (0.94–1.42)
Women
Q1 (−13.3, −4.9)	183,001	367	1.00	252	1.00	115	1.00
Q2 (−4.9, −2.3)	169,020	284	0.89 (0.75–1.06)	205	0.96 (0.78–1.18)	79	0.75 (0.55–1.03)
Q3 (−2.3, 1.0)	146,821	215	0.78 (0.63–0.97)	142	0.78 (0.60–1.02)	73	0.79 (0.54–1.15)
Q4 (1.0, 55.1)	121,014	159	0.66 (0.48–0.90)	100	0.64 (0.43–0.94)	59	0.70 (0.41–1.18)
Ptrend			0.007		0.02		0.19
Continuous scale (per SD increase)			0.86 (0.74–1.00)		0.86 (0.72–1.03)		0.86 (0.67–1.11)
Pheterogeneity			0.41		0.14		0.50

^aAdjusted for age, sex, dialect, year of interview, education level, BMI levels (<18.5, 18.5–<23.0, 23.0–<27.0, ≥27.0), smoking status, alcohol consumption, history of type 2 diabetes, physical activity, family history of colorectal cancer, and total energy intake for total subjects.

In the subgroup analyses, the inverse association between CDAI and CRC risk was more apparent in women (HR_Q4vs.Q1=0.66, 95% CI: 0.48–0.90) (Table 2), individuals without a history of type 2 diabetes (HR_Q4vs.Q1=0.77, 95% CI: 0.62–0.94), BMI ≥23 kg/m² (HR_Q4vs.Q1=0.75, 95% CI: 0.57–0.99), never smokers (HR_Q4vs.Q1=0.69, 95% CI: 0.54–0.89), and those without a family history of CRC (HR_Q4vs.Q1=0.80, 95% CI: 0.65–0.97) (Table 3). However, all the differences in the CDAI-CRC risk associations between these subgroups examined were not statistically significant (all P_{heterogeneiry} > 0.15) (Tables 2 and 3).

Table 3.

Association Between Diet Composite Dietary Antioxidant Index (CDAI) and the Risk of Colorectal Cancer (CRC) Stratified by Selected Characteristics in the Singapore Chinese Health Study, 1993–2015

CDAI in quartile	Person-year	CRC Cases	HR (95% CI)a
By Diabetes History
No history of type 2 diabetes
Q1 (−13.3, −4.9)	237,453	525	1.00
Q2 (−4.9, −2.3)	248,794	505	0.96 (0.84–1.09)
Q3 (−2.3, 1.0)	253,902	459	0.84 (0.72–0.98)
Q4 (1.0, 58.6)	256,831	453	0.77 (0.62–0.94)
Ptrend			0.007
Continuous scale (per SDb)			0.93 (0.84, 1.04)
Had history of type 2 diabetes
Q1 (−12.1, −4.9)	20,787	57	1.00
Q2 (−4.9, −2.3)	19,633	40	0.78 (0.50–1.20)
Q3 (−2.3, 1.0)	17,960	48	1.04 (0.65–1.67)
Q4 (1.0, 37.2)	19,000	53	1.18 (0.63–2.20)
Ptrend			0.51
Continuous scale (per SD)			0.81 (0.59–1.11)
Pheterogeneity			0.16
By BMI Level
BMI<23 kg/m2
Q1 (−13.0, −4.9)	117,670	263	1.00
Q2 (−4.9, −2.3)	128,101	259	0.96 (0.80–1.15)
Q3 (−2.3, 1.0)	132,431	258	0.96 (0.78–1.19)
Q4 (1.0, 58.6)	134,609	248	0.87 (0.65–1.16)
Ptrend			0.43
Continuous scale (per SD)			0.96 (0.84–1.11)
BMI ≥23 kg/m2
Q1 (−13.3, −4.9)	140,570	319	1.00
Q2 (−4.9, −2.3)	140,326	286	0.93 (0.78–1.10)
Q3 (−2.3, 1.0)	139,431	249	0.78 (0.64–0.96)
Q4 (1.0, 55.1)	141,222	258	0.75 (0.57–0.99)
Ptrend			0.01
Continuous scale (per SD)			0.88 (0.76–1.01)
Pheterogeneity			0.37
By Smoking Status
Never Smokers
Q1 (−13.3, −4.9)	191,506	385	1.00
Q2 (−4.9, −2.3)	199,601	358	0.93 (0.80–1.09)
Q3 (−2.3, 1.0)	196,040	318	0.83 (0.69–0.99)
Q4 (1.0, 58.6)	190,259	288	0.69 (0.54–0.89)
Ptrend			0.005
Continuous scale (per SD)			0.90 (0.79–1.02)
Ever Smokers
Q1 (−13.0, −4.9)	66,733	197	1.00
Q2 (−4.9, −2.3)	68,826	187	0.96 (0.78–1.19)
Q3 (−2.3, 1.0)	75,822	189	0.93 (0.73–1.19)
Q4 (1.0, 46.4)	85,572	218	1.01 (0.73–1.40)
Ptrend			0.92
Continuous scale (per SD)			0.98 (0.83–1.15)
Pheterogeneity			0.35
By Family History of CRC
No
Q1 (−13.3, −4.9)	253,917	567	1.00
Q2 (−4.9, −2.3)	262,598	540	0.97 (0.86–1.10)
Q3 (−2.3,1.0)	265,244	484	0.86 (0.74–0.99)
Q4 (1.0,58.6)	268,337	482	0.80 (0.65–0.97)
Ptrend			0.01
Continuous scale (per SD)			0.92 (0.83–1.01)
Yes
Q1 (−10.5, −4.9)	4,323	15	1.00
Q2 (−4.9, −2.3)	5,829	5	0.23 (0.08–0.65)
Q3 (−2.3,1.0)	6,618	23	1.02 (0.47–2.20)
Q4 (1.0,31.6)	7,494	24	0.98 (0.36–2.64)
Ptrend			0.42
Continuous scale (per SD)			1.14 (0.68–1.90)
P heterogeneity			0.25

^aAdjusted for age, sex, dialect, year of interview, education level, physical activity, family history of colorectal cancer, total energy intake, history of type 2 diabetes, BMI level, smoking status, and alcohol consumption, if applicable.

^bSD, standard deviation.

In the sensitivity analysis for the subset after removing CRC cases and person-years that occurred within the first 2 years post-enrollment, the results derived from the subset were comparable with those of the entire dataset. For example, the multivariable-adjusted HR of CRC for highest relative to the lowest quartile of CDAI was 0.81 (95% 0.66–0.99, P_trend =0.02) (Supplementary Table 2).

DISCUSSION

The current analysis in an ongoing prospective cohort study of 61,321 Chinese men and women in Singapore showed that a higher CDAI, which summarized all major food-based antioxidants, was associated with a lower risk of developing CRC overall. The inverse CDAI-CRC risk associations were mainly observed among female participants, overweight or obese individuals (i.e., BMI≥23kg/m²), never smokers, individuals without a history of type 2 diabetes, or those without a family history of colorectal cancer.

Oxidative stress is an imbalance between free radicals and antioxidants in cells and tissues. Free radicals are oxygen-containing molecules with an uneven number of electrons, which allows them to easily react with other molecules such as DNA, resulting in DNA damages.^45,46 The accumulation of DNA damages can alter regular gene expressions and may initiate the carcinogenesis process.⁴⁵ Antioxidants are molecules that can donate an electron to a free radical, which becomes more stable and less reactive, whereby provides a protective effect against the oxidative stress initiated carcinogenesis process.⁴⁷ CDAI, a composite measurement of dietary total antioxidant consumption from food, is inversely correlated with circulating levels of inflammatory cytokines such as IL-1β and TNF-α,⁸ suggesting that CDAI may exert its effect through anti-inflammatory pathway.

To our knowledge, this is the first prospective cohort study that evaluated the association between CDAI and CRC risk. Prior studies for the effect of individual antioxidants on the development of CRC have been conducted, including Vitamin A, Vitamin C, Vitamin E, selenium, manganese, and zinc. Vitamins C and E are recommended to be protective agents against cancer.^10–13 Minerals such as selenium, manganese, and zinc are also considered to be essential elements in CRC prevention,^11,18,19 whereas several studies did not find a significant protective effect of vitamin A against the development of CRC.^15–17 A prospective cohort study reported an inverse relationship between high selenium intake and CRC risk.⁴⁸ A similar inverse association was also observed among people who had higher serum selenium levels.²⁵ Dietary zinc was associated with reduced risk of CRC and rectal cancer in the Nurses’ Health Study (NHS) and Health Professional Follow-up Study (HPFS) ²⁸, and in a meta-analysis involving eight prospective studies among men and women.²⁹ The effect of manganese was mostly assessed in combination with other minerals on the risk of CRC. In the Iowa Women Health Study (55–69-year-old US women, n=35,221), it was reported that higher intake of calcium, magnesium, manganese, zinc, selenium, potassium, and iodine, amalgamated with lower intake of iron, copper, phosphorus, and sodium may link with lower risk of CRC.⁴⁹ In the same study, Bostick et al. found that high intake of Vitamin E was associated with a lower risk of colon cancer in women.¹⁴ Our findings, especially in women, are consistent with these prior studies. However, some studies also reported null associations for CRC or precursor disease on some of the antioxidants including dietary zinc,⁵⁰ selenium, and Vitamin E supplementation,⁵¹ and vitamin C, and E supplementation.⁵² The inconclusive results from prior studies might be explained by the use of dietary supplements instead of food-sourced antioxidants, and different study populations with different lifestyle and dietary patterns.⁵³

In the analysis of individual CDAI components and CRC risk, we found that higher levels of vitamins A and C were significantly associated with a lower risk of CRC while selenium was associated with a marginally increased risk of CRC (Supplementary Table 3). It should be noted that while the effect of individual dietary antioxidants had a modest effect, their aggregated effect as a composite score demonstrated a greater effect on CRC risk. Our findings suggest that the CDAI more precisely captured an individual’s dietary antioxidant profile that could reduce the misclassification of exposure and manifest their protective effect. These results also pointed out greater importance of a comprehensive measurement than an individual assessment of dietary antioxidants for assessing their roles in the development of CRC.

Epidemiological studies on dietary total antioxidants and CRC risk are limited. In the prospective study of HPFS among 47,339 men in the US,³² total antioxidant capacity was calculated using the method based on the ferric-reducing ability of plasma assay at the presence of the antioxidant. This study found that total antioxidant capacity from foods only was associated significantly with a lower risk of rectal cancer but not with colon cancer or CRC. Inclusion of dietary supplements in the total antioxidant capacity diminished the inverse association with rectal cancer risk, becoming statistically non-significant. Two meta-analyses of clinical trials did not find a significant effect of antioxidant supplements (single agent or the combination of vitamins A, C, and E, beta-carotene, and selenium) on reducing the occurrence of CRC and/or recurrence of colorectal adenoma.^54,55 The lack of effect of antioxidant supplements in those clinical trials is consistent with the findings from the HPFS.³² Furthermore, the clinical trial studies were usually conducted among high-risk individuals with detected colorectal adenoma and had a relatively short duration of intervention and follow-up, which might not be able to observe the long-term effect of antioxidant supplements against the development of CRC or recurrence of colorectal adenoma. These prior studies, particularly clinical trials overall did not show a protective effect of antioxidant supplements on the risk of CRC. Our findings are inconsistent with prior studies.

Our study shows a more apparent protective effect of CDAI among women, never smokers, diabetes-free, and overweight/obese participants. The biological mechanisms for the observed differential associations within these particular subgroups are not completely understood. Female participants who were 45–79 years old at baseline already or soon experienced menopause, when the estrogen level dramatically decreased and oxidative stresses increased. The consumption of foods containing a high level of antioxidants may enhance their anti-inflammatory effect in postmenopausal women.^56,57 A similar explanation would be applicable to the CDAI-CRC risk association observed in overweight or obese individuals. The lack of association for CDAI with CRC risk among smokers may be explained by the fact that antioxidants may promote tumorigenesis in smokers, as observed in clinical trials.⁵⁸ The lack of a CDAI-CRC risk association among diabetic patients might be due to the dietary change after baseline assessment and/or small sample size. Future studies are thus warranted to elucidate the biological mechanisms for the differential effects of dietary antioxidants on CRC protection, especially in women and obese individuals.

The present study had several strengths. This was the first prospective study examining the association between CDAI and the risk of CRC in an Asian population. The prospective study design allowed the collection of information on dietary and other lifestyle factors before the occurrence of CRC, minimizing the potential impact of CRC progression on dietary habits. A comprehensive evaluation for CDAI with other modifiable factors and medical history allowed for the adjustment for multiple potential confounders in the statistical analysis. Our study had a large sample size with long-term follow-up, which provided sufficient statistical power to evaluate the associations between CDAI and the risk of CRC. The original CDAI was developed and assessed with the circulating level of inflammatory markers, which reflect the biological effects of dietary antioxidants on oxidative-induced systemic inflammation, one of carcinogenesis pathways. On the other hand, circulating levels of individual antioxidants reflect their most recent intake from all sources, including supplements, which have not shown a protective effect, or in some case, showed detrimental effect, in the development of various cancers such as bladder cancer or thyroid cancer³⁰. Thus, the blood levels of antioxidants may not correctly reflect antioxidants derived from food sources, resulting in inconsistent associations with the risk of cancer, in particular, CRC.

In the current analysis, we focused on the CDAI score with elements derived from food sources only, which was, as expected, positively correlated with total energy intake and higher levels of physical activity (Table 1). These relationships may provide some explanations for the lack of association between CDAI and BMI.

The present study also has several limitations. The one-time assessment of dietary intake using FFQ at baseline did not capture the dietary changes over time. Given the prospective study design, the changes in the dietary habits after baseline administration of FFQ would result in non-differential misclassification, which were more likely to underestimate the true association towards the null result.⁵⁹ In addition, even though multiple confounders were adjusted in statistical analysis, we cannot completely rule out the residual confounding on the CDAI-CRC risk association. Also, potential reversal causation due to occult disease might occur; however, the analysis by excluding incident cases of CRC and person-years within the first two years of observation produced similar results, suggesting the observed association was less likely to be explained by the reversal causality. Although any subject with a history of cancer were excluded from the cohort study according to eligibility criteria at baseline, we did not screen for the presence of cardiovascular disease at enrollment given the cancer focus of the study and prohibitive cost for screening such a large cohort with more than 60,000 subjects. Given lack of direct biological link between cardiovascular disease and CRC, it is less likely that the presence of cardiovascular disease would confound the observed CDAI-CRC association. Further, a stronger CDAI-CRC association observed in participants free of diabetes at baseline suggests that the overall effect of CDAI on CRC risk may be potentially underestimated.

Overall, we found that a higher level of CDAI was significantly associated with a reduced risk of CRC in the Singaporean Chinese population. The inverse association was mainly presented in female participants, participants free of type 2 diabetes, never smokers, or overweight and obese individuals as well as those without a family history of CRC. Our findings suggest that dietary antioxidants may offer protection against the development of CRC in this Asian population.

WHAT IS NEW?

In this large prospective Asian study, the authors applied a recently developed composite dietary antioxidant index (CDAI) for the assessment of colorectal cancer incidence and found a statistically significant 20% reduction in colorectal cancer risk overall. These findings support the notion that food-based antioxidants may offer a beneficial effect and provide a potential strategy for the primary prevention of colorectal cancer.

Abbreviations:

BMI

body mass index

CI

confidence interval

CRC

colorectal cancer

CDAI

composite dietary antioxidant index

FFQ

food frequency questionnaire

HR

hazard ratio

HPFS

Health Professional Follow-up Study

NHS

Nurses’ Health Study

ROS

reactive oxygen species

RCT

randomized clinical trials

SCHS

Singapore Chinese Health Study

SD

standard deviation

TAC

total antioxidants capacity

Data Availability Statement

De-identified data relevant to the report will be available for sharing upon request through the University of Pittsburgh for researchers who meet the criteria for access to confidential data. Data are accessible to the corresponding author and also are available from the University of Pittsburgh Institutional Data Access/Ethics Committee with the following contact information: 3500 Fifth Avenue, Hieber Building Main Office, Suite 106 Pittsburgh, PA 15213. Main Phone: (412) 383–1480. Main Fax: (412) 383–1508. askirb@pitt.edu.