Food-Frequency Questionnaire Based Estimates of Total Antioxidant Capacity and Risk of Non-Hodgkin Lymphoma

Abstract

Antioxidants, primarily from fruits and vegetables, have been hypothesized to protect against non-Hodgkin lymphoma (NHL). The Oxygen Radical Absorbance Capacity (ORAC) assay, which measures total antioxidant capacity of individual foods and accounts for synergism, can be estimated using a food-frequency questionnaire (FFQ). We tested the hypothesis that higher intake of antioxidant nutrients from foods, supplements, and FFQ-based ORAC values are associated with a lower risk of NHL in a clinic-based study of 603 incident cases and 1007 frequency-matched controls. Diet was assessed with a 128-item FFQ. Logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals adjusted for age, sex, residence and total energy. Dietary intake of α-tocopherol (OR=0.50; p-trend=0.0002), β-carotene (OR=0.58; p-trend=0.0005), lutein/zeaxanthin (OR=0.62; p-trend=0.005), zinc (OR=0.54; p-trend=0.003) and chromium (OR=0.68; p-trend=0.032) were inversely associated with NHL risk. Inclusion of supplement use had little impact on these associations. Total vegetables (OR=0.52; p-trend<0.0001), particularly green leafy (OR=0.52; p-trend<0.0001) and cruciferous (OR=0.68; p-trend=0.045) vegetables, were inversely associated with NHL risk. NHL risk was inversely associated with both hydrophilic ORAC (OR=0.61, p-trend=0.003) and lipophilic ORAC (OR=0.48, p-trend=0.0002), although after simultaneous adjustment for other antioxidants or total vegetables only the association for lipophilic ORAC remained significant. There was no striking heterogeneity in results across the common NHL subtypes. Higher antioxidant intake as estimated by the FFQ-ORAC, particularly the lipophilic component, was associated with a lower NHL risk after accounting for other antioxidant nutrients and vegetable intake, supporting this as potentially useful summary measure of total antioxidant intake.

Although few modifiable risk factors for NHL are known, diet has been hypothesized to play an etiologic role.¹ One of the more consistent epidemiologic findings is a potential protective role of fruits and vegetables in the development of NHL (reviewed in ^{2, 3}). Fruits and vegetables rich in antioxidant nutrients are hypothesized to be chemoprotective against many cancer types including lymphoma, and several mechanisms have been hypothesized, including reduction of reactive oxygen species (ROS) responsible for oxidative DNA damage,⁴ regulation of cell survival and apoptosis pathways,⁵ and protection of immune homeostasis.⁶ Other sources of antioxidants, including whole grains, nuts, chocolate and tea have received less attention and to date have not shown clear associations.^{7, 8} A major limitation of the literature is that total antioxidant capacity, which accounts for multiple antioxidants and synergism within individual foods, has only been assessed in a single study of NHL risk.⁸

Assays to assess the total antioxidant capacity of foods have been developed, but there is no gold standard and comparison of values between different methods and studies has been difficult.⁹ For example, several assays have been proposed to provide an integrated parameter of antioxidant potential in foods and biological samples, including the trolox equivalent antioxidant capacity (TEAC), ferric ion reducing antioxidant power (FRAP), and the oxygen radical absorbance capacity (ORAC), all of which are based on different underlying mechanisms and sources of radicals and oxidants. The TEAC assay relies on a spectrophotometric technique based upon scavenging activity of 2,2′-azinobis-(3-ethylbenzothiazoline-6-sulphonic acid) radicals, which may not be applicable to physiologic conditions.¹⁰ The FRAP assay measures the ability of plasma or other fluid to reduce ferric-tripyridyltriazine to the ferrous form, but the reaction conditions are also not physiologic.¹¹ However, the ORAC assay developed by Cao and colleagues,¹² has the advantage of taking the free radical reaction to completion, using biologically relevant free radicals (peroxyl, hydroxyl, and Cu²⁺), and being one of the most standardized and reproducible methods for characterizing antioxidant activities of foods.^{13, 14} ORAC is useful for determining antioxidant activity in biological samples and can demonstrate an increase in plasma antioxidant capacity in humans after antioxidant-rich meals.^{15, 16} Total antioxidant capacity based on food values that used the ORAC assay was not associated with NHL risk, although there was a suggestive inverse association with hydroxyl radical absorbance capacity and the diffuse large B-cell lymphoma subtype.⁸ That ORAC assay, however, was not based on a more robust fluorescein detection system¹⁷ and did not measure the lipophilic and hydrophilic components separately.¹⁸ In 2007, The United States Department of Agriculture (USDA) published a comprehensive 275-item database of ORAC food values derived from the fluorescein method.¹⁹ Both lipophilic and hydrophilic antioxidant capacities were determined using the ORAC assay on foods procured from the Food and Nutrient Analysis Program.

We tested the hypothesis that higher intake of antioxidants, as estimated from a FFQ for antioxidant nutrients and the FFQ-based ORAC values, would be associated with lower NHL risk. In a secondary analysis, we also assessed heterogeneity of this hypothesis for the most common subtypes of NHL: chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), follicular lymphoma (FL), and diffuse large B cell lymphoma (DLBCL).

Material and Methods

Study population

This study was reviewed and approved by the Human Subjects Institutional Review Board at the Mayo Clinic, and all participants provided written informed consent. Full details of the study design are reported elsewhere.²⁰ Starting on 9/1/02, we offered enrollment to all consecutive cases of pathologically-confirmed lymphoma (including CLL) who were at least 20 years old; a resident of Minnesota, Iowa or Wisconsin at the time of diagnosis; within 9 months of their initial diagnosis at presentation to Mayo Clinic Rochester; no history of HIV; English-speaking; and able to provide written informed consent. A Mayo Clinic hematopathologist reviewed all materials for each case to verify the diagnosis and to classify each case according to the World Health Organization Classification of Neoplastic Diseases of the Hematopoietic and Lymphoid Tissues.²¹ This analysis included all subjects enrolled into the study from 9/1/02 through 2/29/08. Of the 1798 eligible patients identified during this time frame, 1236 (69%) participated, 183 (10%) refused, 39 (2%) were lost to follow-up (i.e., we were unable to contact after multiple attempts), and 340 (19%) had their eligibility expire (i.e., they did not consent within 9 months of diagnosis or despite consenting did not provide all samples and data within 12 months of diagnosis).

Clinic-based controls were recruited from Mayo Clinic Rochester patients who had pre-scheduled general medical examinations (i.e., not for a diagnostic work-up or specific, active symptom or disease) in the general medicine divisions of the Department of Medicine from 9/1/02 through 2/29/08. Controls had no history of lymphoma, leukemia or HIV infection and were at least 20 years old; a resident of Minnesota, Iowa or Wisconsin at time of appointment; English-speaking; and able to provide written informed consent. Controls were frequency matched to the case distribution on 5-year age group, sex, and geographic location of residence (8 county groupings based on distance from Rochester, MN and urban/rural status) using a computer program that randomly selects subjects from eligible patients. Of the 1899 eligible subjects identified, 1315 (69%) participated, 548 (29%) refused and 36 (2%) did not provide all samples and data within 12 months of selection.

Data collection

Participants completed a risk-factor questionnaire that included information on demographics, ethnicity, family cancer history, medical history, and selected lifestyle and other factors. In 2003 (after the start of the study), we added a food frequency questionnaire (FFQ) for all study participants, and 748 of 940 cases (79%) and 1139 of 1315 controls (87%) who were offered a FFQ completed it. Cases who did not complete the FFQ were slightly younger than participants (58.3 vs. 61.4 years) but did not differ gender, education level or state of residence; controls who did not complete the FFQ were also younger (56.4 vs. 61.4 years) and did not differ on gender, education level or state of residence.

The FFQ was self-administered, and was most commonly completed at home. It was a modification of the FFQ used in the NCI-SEER Interdisciplinary Case-Control Study of NHL,²² which was based on the Block 1995 Revision of the Health Habits and History Questionnaire.^{23, 24} The FFQ included 103 food items, 25 beverages, cooking doneness for 7 meats, and use of 15 types of supplements. Participants were asked to report “about your usual eating habits, as an adult, before one year ago and not including any recent dietary changes. Please include foods that you ate in a restaurant.” For each food, participants were asked to indicate their usual portion size (small, medium, or large, with a specific amount provided for medium size) and frequency of consumption (never or <1 per month, 1–3 per month, 1 per week, 2–4 per week, 5–6 per week, 1 per day, 2–3 per day, 4–5 per day, 6+ per day). For each supplement used at least once per week for six months, the frequency, duration, and usual dose was ascertained.

Analysis of the FFQ was performed using the Food Processor SQL nutrition analysis software (version 10.0.0., ESHA Research, Salem, OR), under the direction of a dietician (H.O.). This software program provided diet analysis for over 60 nutrients, including carotenoids, fats, and other macro-and micronutrients, and total energy. A database was also developed for analyzing the intake of over 140 multivitamin and mineral supplements using manufacturer’s information.

Oxygen Radical Absorbance Capacity (ORAC) Database

Using food consumption data from the FFQ, we estimated levels of hydrophilic ORAC (H-ORAC), lipophilic ORAC (L-ORAC), and total ORAC using the database for Oxygen Radical Absorbance Capacity (ORAC) of Selected Foods – 2007 developed by USDA (www.ars.usda.gov).¹⁹ The ORAC assay uses fluorescein as the flourescent probe and 2,2′-azobis(2-amidopropane) dihydrochloride as a peroxyl radical generator. The foods that were tested were collected from four different regions and two different seasons in the US market by the National Food and Nutrient Analysis Program. Antioxidants from food sources other than fruits and vegetables (e.g., nuts, grains, wine) were included in ORAC calculations. ORAC measures are reported in μmol of Trolox Equivalents per 100 grams (μmol TE/100g). Total ORAC does not directly sum the hydrophilic and lipophilic components due to different sources of data. In a validation study, the correlation of plasma ORAC values and FFQ-estimated ORAC intake was reasonably strong,²⁵ although this has not been validated in other populations.

Statistical analysis

For the analysis of dietary data, we excluded 19 cases and 3 controls with missing risk factor data, and 91 cases and 129 controls who had >10 missing food items or implausible daily energy intakes (women with less than 600 or more than 5000 kcal/day, and men with less than 800 or more than 6000 kcal/day). This left 638 cases and 1007 controls in the analysis dataset; excluding Hodgkin lymphoma cases left 603 cases.

Individual foods were grouped according to similar nutrient content or genus classification and were expressed as servings per month. Individual nutrients, foods, food groups, and the three ORAC measures were modeled as ordinal variables corresponding to the quartile distribution of intake in the control group. Nutrients from supplements were calculated based on reported dose and frequency of use of multivitamins and individual supplements, and were combined with the nutrient values from food to estimate total intake.

We calculated odds ratios (OR) and 95% confidence intervals (CI) using unconditional logistic regression, and the lowest category of intake was used as the reference. We then calculated a one-degree of freedom trend test, using the ordinal scoring of the intake quartiles, and statistical significance for an association was declared for p<0.05. Polytomous logistic regression was used to simultaneously model the comparison between controls and each of four NHL subtypes – CLL/SLL, follicular lymphoma, DLBCL, and all other. We used a 3 degree-of-freedom Wald test to assess heterogeneity in the trend tests of the four subtypes, and a p-value of <0.05 was declared of interest. All models were adjusted for the design variables of age, sex, residence, as well as total energy. Total energy was modeled categorically based upon the quartile distribution among controls, and was included to adjust for systematic over-and under-reporting of food intake.²⁶ As a sensitivity analysis, we also evaluated whether adjusting for total energy²⁷ using the residual method impacted the results; finding no important differences, we only report the results from the initial analyses.

We evaluated effect modification by gender for all dietary variables, and also evaluated potential confounding by adding educational level (less than high school, high school graduate, some college, college graduate, or graduate/professional school), family history of NHL (first degree relative with NHL versus not), smoking (pack-years), alcohol use (never, former, or current use), and body mass index (continuous) to the basic models. All statistical tests were two-sided, and all analyses were carried out using SAS (SAS Institute, Inc., Cary, NC).

Results

Cases and controls in this analysis were well balanced on age, sex, residence, and level of education (Table 1). The most common subtypes were CLL/SLL (35%), FL (23%), and DLBCL (19%).

Table 1.

Characteristics of study participants, Mayo Case-Control Study of NHL, 2002–2008

Characteristic	Cases (N=603) N (%)	Controls (N=1007) N (%)
Age, years
<40	30 (5.0%)	81 (8.0%)
40–49	87 (14.4%)	134 (13.3%)
50–59	136 (22.6%)	210 (20.9%)
60–69	195 (32.3%)	306 (30.4%)
70+	155 (25.7%)	276 (27.4%)
Age, mean ± SD, years	60.9 ± 12.3	60.1 ± 13.7
Gender
Male	346 (57.4%)	535 (53.1%)
Female	257 (42.6%)	472 (46.9%)
Residence
Minnesota	420 (69.7%)	688 (68.3%)
Iowa	102 (16.9%)	186 (18.5%)
Wisconsin	81 (13.4%)	133 (13.2%)
Education Level
Less than high school graduate	25 (4.2%)	27 (2.7%)
High school graduate/GED	117 (19.5%)	223 (22.2%)
Vocational/other post high school	105 (17.5%)	172 (17.1%)
Some college/college graduate	244 (40.7%)	364 (36.3%)
Graduate or professional school	108 (18.0%)	218 (21.7%)
Missing	4	3
NHL Subtype
CLL/SLL	210 (34.8%)
Follicular	138 (22.9%)
DLBCL	116 (19.2%)
Marginal zone	38 (6.3%)
Mantle Cell	24 (4.0%)
T-Cell	17 (2.8%)
Other/NOS	60 (10.0%)

Table 2 reports the results for antioxidant nutrients from foods. Odds ratios reported below in the text are for lowest versus highest level unless noted otherwise. There were inverse associations for α-tocopherol (OR=0.50; p-trend=0.0002), β-carotene (OR=0.58; p-trend=0.0005), lutein/zeaxanthin (OR=0.62; p-trend=0.005), zinc (OR=0.54; p-trend=0.003), and chromium (OR=0.68; p-trend=0.032) with risk of NHL, while there were no statistically significant overall associations for vitamin C, other carotenoids (α-carotene, β-cryptoxanthin and lycopene), manganese, or copper. For α-tocopherol and β-carotene, the inverse associations were observed for FL and DLBCL, but not for CLL/SLL, with statistical evidence for heterogeneity (p=0.016 and p=0.035, respectively). For lutein/zeaxanthin, zinc and chromium, the inverse associations were strongest for DLBCL, weaker for FL, and weakest or null for CLL/SLL; however, there was no statistical evidence for heterogeneity by subtype for any of these antioxidants (all p≥0.2). Finally, while there was no overall association of α-carotene with NHL risk, there was evidence for NHL subtype heterogeneity (p=0.0001), such that there was a strong inverse association for DLBCL, a weaker inverse association for FL, and a positive association for CLL/SLL (Table 2).

Table 2.

Association of intake of antioxidant nutrients from foods with risk of NHL and selected subtypes, Mayo Clinic Case-Control Study, 2002–2008

Nutrient intake	Controls	All NHL			CLL/SLL			Follicular NHL			DLBCL
		Cases	OR*	95% CI	Cases	OR*	95% CI	Cases	OR*	95% CI	Cases	OR*	95% CI
Vitamin C (mg/day)
<67	252	174	1.00	reference	58	1.00	reference	48	1.00	reference	39	1.00	reference
67–104	252	137	0.79	0.59–1.06	52	0.91	0.60–1.40	34	0.73	0.45–1.18	16	0.41	0.22–0.76
105–157	252	160	0.92	0.69–1.24	62	1.11	0.73–1.69	36	0.78	0.48–1.27	25	0.65	0.37–1.13
158+	251	131	0.74	0.54–1.02	46	0.81	0.51–1.29	28	0.59	0.34–1.03	25	0.63	0.35–1.14
P-trend			0.16			0.61			0.087			0.21
α-Tocopherol (mg/day)
<4.7	252	193	1.00	reference	61	1.00	reference	52	1.00	reference	43	1.00	reference
4.7–6.4	252	148	0.65	0.47–0.88	52	0.83	0.52–1.31	33	0.59	0.35–0.99	22	0.38	0.21–0.69
6.5–9.0	252	114	0.44	0.31–0.63	46	0.71	0.42–1.19	24	0.39	0.21–0.73	18	0.24	0.12–0.50
9.1+	251	147	0.50	0.34–0.74	59	0.85	0.48–1.49	37	0.55	0.29–1.07	22	0.23	0.11–0.51
P-trend			0.0002			0.52			0.052			0.0002
α-carotene (μg/day)
<242	252	148	1.00	reference	34	1.00	reference	47	1.00	reference	37	1.00	reference
242–461	252	167	1.11	0.84–1.49	57	1.72	1.08–2.75	40	0.86	0.54–1.37	29	0.75	0.45–1.28
462–1106	252	142	0.95	0.70–1.28	67	2.08	1.31–3.30	30	0.64	0.39–1.07	20	0.51	0.28–0.92
1107+	251	145	0.98	0.72–1.33	60	1.94	1.20–3.14	29	0.63	0.37–1.06	19	0.48	0.26–0.89
P-trend			0.64			0.008			0.043			0.008
β-carotene (μg/day)
<1541	252	188	1.00	reference	54	1.00	reference	54	1.00	reference	43	1.00	reference
1541–2707	252	156	0.77	0.58–1.02	62	1.10	0.73–1.66	40	0.70	0.44–1.10	22	0.46	0.27–0.81
2708–4464	252	139	0.69	0.51–0.93	56	1.02	0.66–1.57	25	0.45	0.26–0.75	24	0.50	0.29–0.87
4465+	251	119	0.58	0.42–0.79	46	0.83	0.52–1.32	27	0.46	0.27–0.79	16	0.32	0.17–0.61
P-trend			0.0005			0.39			0.001			0.0006
β-cryptoxanthin (μg/day)
<74	252	171	1.00	reference	60	1.00	reference	50	1.00	reference	35	1.00	reference
74–136	252	133	0.79	0.59–1.06	50	0.85	0.56–1.30	33	0.67	0.42–1.09	18	0.53	0.29–0.96
137–233	252	166	0.97	0.73–1.29	67	1.12	0.75–1.68	38	0.77	0.48–1.24	27	0.78	0.45–1.35
234+	251	132	0.79	0.58–1.07	41	0.70	0.44–1.11	25	0.51	0.30–0.88	25	0.74	0.41–1.33
P-trend			0.31			0.36			0.03			0.48
Lycopene (μg/day)
<2502	252	167	1.00	reference	59	1.00	reference	44	1.00	reference	35	1.00	reference
2502–4222	252	138	0.81	0.61–1.09	57	0.95	0.63–1.44	29	0.66	0.40–1.10	19	0.54	0.30–0.97
4223–7285	252	164	0.95	0.71–1.28	56	0.91	0.59–1.41	45	1.09	0.67–1.76	24	0.67	0.38–1.21
7286+	251	133	0.76	0.55–1.04	46	0.73	0.46–1.17	28	0.66	0.38–1.16	27	0.75	0.41–1.36
P-trend			0.19			0.21			0.41			0.42
Lutein and Zeaxanthin (μg/day)
<694	252	173	1.00	reference	55	1.00	reference	43	1.00	reference	40	1.00	reference
694–1189	252	156	0.91	0.68–1.21	63	1.19	0.79–1.81	45	1.07	0.67–1.70	17	0.42	0.23–0.77
1190–1935	252	158	0.88	0.65–1.18	57	1.03	0.67–1.59	33	0.77	0.46–1.29	27	0.64	0.37–1.10
1936+	251	115	0.62	0.45–0.85	43	0.76	0.47–1.22	25	0.57	0.32–1.00	21	0.48	0.26–0.88
P-trend			0.005			0.20			0.027			0.039
Zinc (mg/day)
<9.9	252	163	1.00	reference	55	1.00	reference	45	1.00	reference	33	1.00	reference
9.9–13.9	252	164	0.86	0.62–1.19	61	1.04	0.65–1.66	37	0.81	0.47–1.39	28	0.68	0.36–1.25
14.0–18.2	252	136	0.62	0.42–0.90	51	0.78	0.45–1.34	34	0.66	0.35–1.24	23	0.45	0.21–0.93
18.3+	251	138	0.54	0.35–0.84	51	0.68	0.36–1.29	30	0.49	0.23–1.06	21	0.31	0.13–0.75
P-trend			0.003			0.16			0.065			0.007
Manganese (mg/day)
<2.4	252	165	1.00	reference	47	1.00	reference	42	1.00	reference	34	1.00	reference
2.4–3.5	252	145	0.83	0.62–1.12	62	1.32	0.85–2.04	35	0.86	0.52–1.43	18	0.49	0.27–0.92
3.6–5.4	252	153	0.88	0.65–1.20	60	1.31	0.83–2.06	34	0.85	0.50–1.43	32	0.89	0.50–1.56
5.5+	251	139	0.78	0.56–1.07	49	1.03	0.63–1.67	35	0.87	0.51–1.50	21	0.55	0.29–1.05
P-trend			0.18			1.00			0.63			0.24
Chromium (μg/day)
<4.5	252	171	1.00	reference	55	1.00	reference	48	1.00	reference	35	1.00	reference
4.5–6.5	252	150	0.82	0.61–1.11	56	1.03	0.67–1.59	36	0.74	0.45–1.22	25	0.62	0.35–1.10
6.6–9.2	252	138	0.70	0.50–0.98	52	0.95	0.58–1.54	25	0.49	0.27–0.88	22	0.47	0.25–0.91
9.3+	251	140	0.68	0.47–0.98	55	0.99	0.58–1.68	36	0.67	0.36–1.24	21	0.40	0.19–0.84
P-trend			0.032			0.90			0.12			0.012
Copper (mg/day)
<1.0	252	169	1.00	reference	55	1.00	reference	43	1.00	reference	35	1.00	reference
1.0–1.3	252	140	0.79	0.57–1.08	53	0.99	0.63–1.57	32	0.82	0.48–1.40	22	0.56	0.30–1.04
1.4–1.8	252	146	0.79	0.56–1.12	61	1.13	0.68–1.85	33	0.88	0.48–1.60	25	0.59	0.30–1.15
1.9+	251	147	0.75	0.50–1.12	49	0.83	0.46–1.50	38	1.02	0.52–2.02	23	0.47	0.21–1.04
P-trend			0.19			0.70			0.92			0.08

^*Adjusted for age, sex, residence, and total energy

The vast majority of cases (82%) and controls (84%) regularly used a multivitamin, and regular multivitamin use was not associated with risk of NHL (data not shown). We also obtained data on the specific use of supplemental vitamin C, α-tocopherol, β-carotene and zinc. When we added levels from multivitamins to dietary intake, the associations for total intake of these antioxidants with NHL risk were unchanged or slightly attenuated from the results for dietary intake only. Contributions of multivitamins to total lutein/zeaxanthin, α-carotene, manganese, chromium and copper were all small and did not affect the results based on dietary intake only (data not shown).

We next evaluated antioxidant intake based on ORAC summary measures with NHL risk (Table 3). There was an inverse association for total ORAC (OR=0.61; p-trend=0.003), and the inverse association was stronger for the lipophilic ORAC (OR=0.48; p-trend=0.0002) relative to the hydrophilic ORAC (OR=0.61; p-trend=0.003). Although the inverse associations were strongest for DLBCL, there were inverse trends for all subtypes, and there was no statistical evidence for etiologic heterogeneity by NHL subtype (all p≥0.4).

Table 3.

Association of total, hydrophilic and lipophilic Oxygen Radical Absorbance Capacity (ORAC) with risk of NHL and selected subtypes, Mayo Clinic Case-Control Study, 2002–2008

μmolTE/100g	Controls	All NHL			CLL/SLL			Follicular NHL			DLBCL
		Cases	OR*	95% CI	Cases	OR*	95% CI	Cases	OR*	95% CI	Cases	OR*	95% CI
Total ORAC
<5490	252	172	1.00	reference	51	1.00	reference	47	1.00	reference	36	1.00	reference
5490–8653	252	164	0.95	0.71–1.26	70	1.42	0.94–2.14	36	0.79	0.49–1.27	23	0.63	0.36–1.11
8654–12914	252	152	0.84	0.63–1.13	55	1.07	0.69–1.67	33	0.71	0.43–1.18	29	0.76	0.44–1.32
12915+	251	114	0.61	0.44–0.84	42	0.80	0.49–1.30	30	0.63	0.37–1.09	17	0.43	0.22–0.83
P-trend			0.003			0.22			0.087			0.026
Hydrophilic ORAC
<5266	252	168	1.00	reference	50	1.00	reference	45	1.00	reference	35	1.00	reference
5266–8363	252	170	1.00	0.75–1.33	74	1.52	1.01–2.29	37	0.85	0.52–1.37	24	0.68	0.39–1.19
8363–12518	252	153	0.87	0.64–1.17	52	1.03	0.66–1.61	37	0.83	0.51–1.37	29	0.79	0.45–1.37
12519+	251	111	0.61	0.44–0.85	42	0.82	0.50–1.33	27	0.60	0.34–1.05	17	0.45	0.23–0.86
P-trend			0.003			0.18			0.087			0.033
Lipophilic ORAC
<201	252	188	1.00	reference	50	1.00	reference	56	1.00	reference	42	1.00	reference
201–292	252	143	0.70	0.52–0.93	61	1.19	0.77–1.82	28	0.47	0.29–0.78	20	0.43	0.24–0.76
293–425	252	158	0.74	0.55–1.00	71	1.37	0.88–2.14	32	0.53	0.32–0.89	23	0.46	0.25–0.83
426+	251	113	0.48	0.34–0.69	36	0.65	0.38–1.13	30	0.46	0.25–0.82	20	0.35	0.18–0.70
P-trend			0.0002			0.32			0.009			0.003

^*Adjusted for age, sex, residence, and total energy.

The next set of analyses evaluated the association of food groups and individual foods with high levels of antioxidants with risk of NHL. There was no association with fruit intake but there was a strong inverse association with total vegetable intake (OR=0.52; p-trend=0.0003) (Table 4). The latter association was due to inverse associations with intake of green leafy (OR=0.52; p-trend<0.0001) and cruciferous (OR=0.68; p-trend=0.045) vegetables, and not red/yellow/orange vegetables (OR=0.80; p-trend=0.26) or legumes (data not show). For green leafy vegetables, each of the food items (cooked spinach, raw spinach, mustard greens and green salad) was inversely associated with risk (data not shown). For cruciferous vegetables, broccoli and radishes were the food items with the strongest inverse association for that group (data not shown). With respect to NHL subtypes, the inverse association with total vegetables was strongest for DLBCL and FL, and was attenuated and not statistically significant for CLL/SLL. The association for cruciferous vegetables was strongest for DLBCL, while all three subtypes showed a strong inverse association with green leafy vegetables. Red/yellow/orange vegetables were only statistically significantly associated with DLBCL. From a statistical perspective however, there was no evidence for heterogeneity of any subtype-specific risks with these dietary variables (all p>0.2).

Table 4.

Association of selected food high in antioxidants with risk of NHL and selected subtypes, Mayo Clinic Case-Control Study, 2002–2008

Food or food group (servings per month)	Controls	All NHL			CLL/SLL			Follicular NHL			DLBCL
		Cases	OR*	95% CI	Cases	OR*	95% CI	Cases	OR*	95% CI	Cases	OR*	95% CI
Total fruit
<36.3	252	163	1.00	reference	54	1.00	reference	46	1.00	reference	33	1.00	reference
35.4–65.9	252	157	0.99	0.74–1.31	60	1.15	0.76–1.74	36	0.79	0.49–1.27	26	0.81	0.47–1.40
66.0–102.2	252	142	0.91	0.68–1.21	55	1.08	0.71–1.65	30	0.68	0.41–1.12	26	0.82	0.47–1.42
102.3+	251	140	0.89	0.66–1.21	49	0.97	0.62–1.51	34	0.79	0.48–1.29	20	0.63	0.34–1.15
P-trend			0.38			0.84			0.25			0.15
Total Vegetables
<42.0	255	184	1.00	reference	55	1.00	reference	45	1.00	reference	45	1.00	reference
42.0–69.8	250	151	0.81	0.61–1.08	63	1.17	0.77–1.77	42	0.95	0.60–1.52	17	0.36	0.20–0.66
69.9–109.7	251	157	0.79	0.59–1.06	59	1.05	0.68–1.62	38	0.83	0.51–1.36	20	0.40	0.22–0.72
109.8+	251	110	0.52	0.37–0.72	41	0.68	0.42–1.11	21	0.43	0.24–0.79	23	0.42	0.23–0.77
P-trend			0.0003			0.13			0.009			0.004
Green leafy vegetables
<4.2	288	230	1.00	reference	76	1.00	reference	53	1.00	reference	45	1.00	reference
4.2–12.8	270	174	0.80	0.62–1.04	67	0.93	0.64–1.34	43	0.86	0.56–1.34	31	0.73	0.45–1.18
12.9–21.1	204	90	0.53	0.39–0.72	34	0.61	0.39–0.95	24	0.64	0.38–1.08	11	0.33	0.17–0.66
21.2+	245	108	0.52	0.39–0.69	41	0.59	0.39–0.91	26	0.57	0.34–0.95	18	0.44	0.25–0.79
P-trend			<0.0001			0.005			0.016			0.001
Cruciferous Vegetables
<2.7	269	171	1.00	reference	54	1.00	reference	36	1.00	reference	39	1.00	reference
2.7–6.2	245	155	1.01	0.76–1.33	65	1.35	0.90–2.03	38	1.19	0.73–1.94	26	0.74	0.43–1.26
6.3–14.6	248	167	1.07	0.81–1.41	61	1.25	0.83–1.89	47	1.47	0.92–2.36	20	0.56	0.31–0.99
14.7+	245	109	0.68	0.50–0.92	38	0.76	0.48–1.21	25	0.79	0.45–1.37	20	0.55	0.30–0.98
P-trend			0.045			0.30			0.75			0.019
Red, yellow and orange vegetables
<15.9	254	163	1.00	reference	53	1.00	reference	40	1.00	reference	37	1.00	reference
15.9–27.2	252	143	0.87	0.65–1.17	53	1.01	0.66–1.55	35	0.90	0.55–1.48	26	0.68	0.40–1.17
27.3–45.2	250	159	0.95	0.70–1.27	63	1.19	0.77–1.82	38	0.99	0.60–1.64	18	0.45	0.25–0.84
45.3+	251	137	0.80	0.58–1.10	49	0.91	0.57–1.46	33	0.87	0.50–1.50	24	0.58	0.32–1.06
P-trend			0.26			0.90			0.72			0.033
Whole Grains
<3.6	268	141	1.00	reference	53	1.00	reference	34	1.00	reference	23	1.00	reference
3.6–23.3	243	136	1.05	0.78–1.41	44	0.90	0.58–1.40	36	1.16	0.70–1.91	27	1.27	0.71–2.27
23.4–29.9	261	170	1.26	0.94–1.67	59	1.17	0.77–1.76	41	1.29	0.79–2.10	28	1.27	0.71–2.27
30+	235	155	1.28	0.95–1.72	62	1.37	0.90–2.08	35	1.24	0.74–2.08	27	1.37	0.75–2.49
P-trend			0.055			0.081			0.35			0.32
Peanuts
<1.8	290	165	1.00	reference	58	1.00	reference	38	1.00	reference	33	1.00	reference
1.8–6.2	225	132	1.05	0.78–1.40	38	0.87	0.55–1.36	38	1.32	0.81–2.14	25	0.99	0.57–1.72
6.3–14.9	258	180	1.23	0.94–1.62	68	1.35	0.91–2.00	40	1.23	0.76–1.99	34	1.16	0.69–1.94
15.0+	234	125	0.91	0.68–1.23	54	1.14	0.75–1.75	30	1.02	0.60–1.72	13	0.47	0.24–0.93
P-trend			0.98			0.22			0.92			0.12
Chocolate candy
<0.9	344	202	1.00	reference	73	1.00	reference	49	1.00	reference	37	1.00	reference
0.9–2.0	216	127	1.01	0.76–1.34	44	0.97	0.64–1.48	37	1.25	0.78–1.98	22	0.96	0.55–1.69
2.1–6.4	209	120	0.97	0.72–1.29	42	0.94	0.62–1.44	34	1.16	0.72–1.88	19	0.85	0.47–1.53
6.5+	238	153	1.08	0.80–1.44	59	1.16	0.77–1.76	26	0.77	0.45–1.32	27	1.06	0.60–1.87
P-trend			0.73			0.57			0.49			0.99
Green Tea
Never	745	470	1.00	reference	173	1.00	reference	108	1.00	reference	85	1.00	reference
<3	114	46	0.65	0.45–0.94	14	0.54	0.30–0.96	18	1.10	0.64–1.89	8	0.62	0.29–1.32
3+	148	86	0.91	0.68–1.22	31	0.89	0.58–1.36	20	0.93	0.56–1.54	12	0.70	0.37–1.32
P-trend			0.21			0.27			0.87			0.17
Black Tea
None	769	482	1.00	reference	173	1.00	reference	113	1.00	reference	87	1.00	reference
Any	238	120	0.81	0.63–1.04	45	0.85	0.59–1.22	33	0.94	0.62–1.43	18	0.67	0.40–1.14
P-trend			0.10			0.38			0.79			0.14
White wine
None	677	434	1.00	reference	153	1.00	reference	93	1.00	reference	79	1.00	reference
<5	217	126	0.89	0.69–1.14	43	0.85	0.59–1.24	45	1.50	1.01–2.21	23	0.89	0.55–1.46
5+	113	42	0.55	0.38–0.80	22	0.81	0.50–1.33	8	0.50	0.24–1.07	3	0.22	0.07–0.70
P-trend			0.003			0.29			0.66			0.013
Red Wine
None	648	412	1.00	reference	145	1.00	reference	96	1.00	reference	73	1.00	reference
<5	201	132	1.03	0.80–1.33	46	1.02	0.70–1.47	38	1.29	0.85–1.94	25	1.10	0.68–1.79
5+	158	58	0.54	0.39–0.75	27	0.71	0.45–1.12	12	0.50	0.27–0.94	7	0.37	0.17–0.83
P-trend			0.002			0.21			0.16			0.047
Beer
None	560	334	1.00	reference	114	1.00	reference	78	1.00	reference	71	1.00	reference
<5	226	136	0.95	0.74–1.23	55	1.12	0.78–1.61	35	1.09	0.71–1.68	15	0.49	0.27–0.88
5+	221	132	0.93	0.71–1.20	49	1.00	0.68–1.46	33	1.06	0.68–1.67	19	0.62	0.36–1.07
P-trend			0.54			0.91			0.73			0.029

^*Adjusted for age, sex, residence, and total energy

Other foods with high antioxidants, including whole grain bread, peanuts, chocolate candy, green tea, and black tea were not significantly associated with NHL risk overall or by common subtypes (Table 4). However, intake of both white wine (OR=0.55; p-trend=0.003) and red wine (OR=0.54; p-trend=0.002) were inversely associated with NHL risk overall and with DLBCL; weaker inverse associations that were not statistically significant were observed for FL and CLL/SLL. Beer intake was associated only with a reduced risk of DLBCL.

We conducted several additional analyses. We found no evidence for heterogeneity by sex (data not shown). Further adjustment of all models for family history of NHL, smoking, alcohol use, and body mass index did not alter these results (data not shown).

H-ORAC and L-ORAC were strongly correlated (r=0.72), and so we did not try to include them in the same model. The top 10 foods that contributed to H-ORAC values were beer, red wine, apple juice, berries, apples, prune juice, citrus fruits, green tea, strawberries, and black tea, all of which either had null or weak (and not statistically significant) inverse associations with NHL risk, except for red wine which was strongly and inversely associated with risk. Individual adjustment of H-ORAC by α-tocopherol, β-carotene, or lutein/zeaxanthin each modestly attenuated the association with NHL, while individual adjustment for dietary vitamin C, α-carotene, β-cryptoxanthin, zinc, or chromium did not change the associations. Simultaneous adjustment for all of these factors greatly attenuated the association of H-ORAC with NHL (OR=1, 1.06, 0.99, 0.79 for quartiles 1–4, respectively; p-trend=0.24). This attenuation was observed for the NHL subtypes.

The top 10 foods that contributed to L-ORAC values were chocolate bars, string beans/green beans, bananas, peas, citrus fruits, peanuts, other beans, green salad, fresh spinach, and berries, and most of these were individually null or weakly inversely associated with NHL risk except for fresh spinach, which was strongly and inversely associated with NHL risk (Table 4). Adjustment of L-ORAC by α-tocopherol, β-carotene, or lutein/zeaxanthin individually only slightly attenuated the association with NHL, while adjustment for dietary vitamin C, α-carotene, β-cryptoxanthin, zinc, or chromium individually did not change the associations. In contrast to H-ORAC, simultaneously adjustment for all of these factors only slightly attenuated the association of L-ORAC with NHL (ORs=1, 0.74, 0.84, 0.59 for quartiles 1–4, respectively; p-trend=0.027); the inverse association for DLBCL (ORs=1, 1.47, 0.55, 0.46 for quartiles 1–4, respectively; p-trend=0.053) was strongest, followed by FL (ORs=1, 0.51, 0.62, 0.59 for quartiles 1–4, respectively; p-trend=0.12) and weakest for CLL/SLL (ORs=1, 1.23, 1.53, 0.77 for quartiles 1–4, respectively; p-trend=0.84), although there was no statistical evidence for heterogeneity (p=0.35).

There was also a relatively strong correlation of total vegetables with H-ORAC (r=0.61) and L-ORAC (r=0.60). In secondary analyses, we assessed the impact of simultaneous adjustment of H- or L-ORAC and total vegetables (also adjusting for the design factors), keeping in perspective the intercorrelation of these factors. After adjustment for total vegetable intake, the association of H-ORAC with NHL risk attenuated (ORs=1, 1.05, 1.04, 0.82 for quartiles 1–4, respectively) and was no longer statistically significant (p-trend=0.39), while the association of L-ORAC was essentially unchanged (ORs=1, 0.72, 0.78, 0.51 for quartiles 1–4, respectively; p-trend=0.0066). In a multivariate model with H-ORAC, the association of total vegetables with NHL risk did not change (ORs=1, 0.78, 0.82, 0.54 for quartiles 1–4 respectively; p-trend=0.008), while in a multivariate model with L-ORAC, the association of total vegetables with NHL risk attenuated slightly (ORs=1, 0.85, 0.92, 0.62 for quartiles 1–4 respectively; p-trend=0.046).

Discussion

We found an inverse association for dietary intake of several antioxidants with risk of NHL, including α-tocopherol, β-carotene, lutein/zeaxanthin, zinc and chromium, while estimates for NHL risk with total intake of antioxidants that included diet and supplements were similar to or weaker than the associations for dietary intake only, suggesting a greater relevance of dietary intake. Vegetable (particularly green leafy vegetables) but not fruit intake was inversely associated with NHL risk. Total antioxidant intake as estimated by the FFQ-based ORAC was also inversely associated with lower risk of NHL, and this association remained for L-ORAC after adjustment for dietary antioxidant intake or vegetable intake. All associations held after multivariable adjustment for potential confounding factors, and were generally similar for DLBCL and FL, while results for CLL/SLL were more variable.

To our knowledge, this is only the second report to evaluate a summary index of antioxidant intake with NHL risk. Using dietary values based on an older version of the ORAC assay, Chang and colleagues⁸ found that the total antioxidant score was not associated with NHL risk overall or for the three common subtypes. There were also no associations for peroxyl or hydroxyl radicals, except perhaps a suggestive inverse association for hydroxyl radicals with DLBCL (RR=0.68 for ≥2.2 versus <0.9 μM Trolox equivalent/g; p-trend=0.08). Our analysis used a more robust methodology and was able to estimate separately the antioxidant capacity of the hydrophilic and lipophilic components of foods.²⁸ Our results suggest that the ORAC measures added additional information beyond the traditional nutrient and food approaches to assessing antioxidant intake and NHL risk, and the antioxidant capacity of the lipophilic component of foods may be more strongly protective against NHL, although this needs to be replicated. The lipophilic components have different physiochemical properties from the hydrophilic components, and thus different functions and sites of action in vivo,²⁸ and may offer novel insights into protection against lymphomagenesis.

Our study also extends previous results from epidemiologic studies that have reported an inverse association of vegetable intake with NHL risk, including 6 case-control^{22, 29–33} and 2 cohort^{7, 34} studies, although there are exceptions. Rohrmann et al. could not confirm an inverse association with vegetable intake and lymphoma risk in 849 cases of NHL evaluated as a part of the European Prospective Investigation into Cancer and Nutrition (EPIC) study, although the risk of DLBCL did tend to be lower in those with the highest intake of vegetables (HR=0.49, 95% CI 0.23–1.02).³⁵ A recently published multiethnic cohort study also did not demonstrate an inverse association with vegetable intake with lymphoma risk except in Caucasian women (HR=0.56, p-trend 0.04).³⁶ The California Teachers Study found a weak and not statistically significant inverse association for vegetable intake (RR=0.82 for ≥2 versus ≤0.6 servings/day, 95% CI 0.65–1.03; p-trend=0.21).⁸ A non-significant elevation in risk with intake of vegetables in women was reported by a small case-control study from Uruguay (75 cases, OR=2.6, 95% CI 0.9–6.7), although the FFQ used in that study was designed primarily to identify patterns in alcohol, yerba maté, tobacco, and meat consumption with the risk of NHL.³⁷

Our study confirms the finding of dietary antioxidant nutrients associated with reduced NHL risk, including β-carotene,^{38, 39} although other studies have observed inverse associations for other carotenoids,^{7, 32, 40} α-tocopherol,³⁸ lutein/zeaxanthin,²² and zinc.²² The finding of heterogeneity of NHL subtypes with respect to dietary α-carotene intake is potentially interesting. The Iowa Women’s Health Study (IWHS) cohort demonstrated the most significant inverse association with FL, not DLBCL, as in this study.⁷ The association of CLL/SLL with carotene-rich foods in the IWHS cohort was not significant,⁴¹ although there were only 58 cases compared with 215 cases in the present report.

We did not confirm the finding of manganese as protective for NHL as in the IWHS.⁷ To our knowledge, the association of chromium with reduced NHL risk has not been reported previously and will require replication. Among vegetables, green leafy vegetables, radishes, and broccoli have relatively high chromium content, potentially explaining this association.⁴² We did not observe overall associations of vitamin C, β-cryptoxanthin, lycopene, or copper with NHL risk.

Our findings that supplemental intake of antioxidant nutrients add little after accounting for dietary intake is consistent with several recent studies that addressed this issue,^{7, 22} and consistent with other studies that find no impact of supplement use.^{43, 44}

Overall, relatively little heterogeneity between NHL subtypes was observed. Associations when present tended to be strongest in DLBCL, attenuated in FL, and null in CLL/SLL. With the exception of dietary α-tocopherol and β-carotene, there was no statistical evidence for heterogeneity between these three most common subtypes and food/nutrient intake.

Our study has several strengths, including the relatively large number of NHL cases (>600) making it among the largest studies of its type which allows evaluation of etiologic heterogeneity for the three common NHL subtypes. Cases and controls were drawn from the same epidemiologic catchment area, and response rates were reasonable at 69% in both cases and controls. Although only 79% of the cases and 87% of the controls completed a FFQ, beyond the somewhat younger age those who completed the FFQ (for both cases and controls), there were no other systematic differences by sex, education level or state of residence. Pathology was reviewed and diagnoses confirmed centrally. Dietary assessment was comprehensive (>120 items), used an established method, and included vitamin/mineral supplement intake. We were able to adjust for total energy intake and other potentially confounding factors. As a basis for estimating total antioxidant intake from diet, the ORAC assay overcomes prior limitations since it is based on an assay that is biologically relevant to in vivo antioxidant mechanisms, accounts for synergy within foods, and can be used to provide estimates of intake from FFQs.

There are also limitations. Clinic-based case-control studies are susceptible to bias, although our study has been carefully designed and assessed to enhance internal and external validity.²⁰ More generally, case-control studies carry a greater potential for recall and selection bias when evaluating dietary hypotheses.⁴⁵ It is somewhat reassuring that our findings for fruit and vegetable intake are broadly consistent with published case-control and most prospective cohort associations of diet and lymphoma risk. Another potential source of error lies in the dietary assessment made through a single self-reported questionnaire, although random misclassification would tend to attenuate associations in both cases and controls. Cases and controls are primarily from one geographic location, which may limit the generalizability to other populations. While we accounted for other potential risk factors for development of NHL (family history, tobacco use, alcohol use, and body mass index), there may be other unmeasured confounders. Finally, there are limitations to the FFQ-ORAC. Foods with the highest values of ORAC (on a per unit basis) were not captured by the FFQ. These include cinnamon, oregano, cloves, dried parsley and basil for H-ORAC and cloves, tumeric, curry powder, ginger and oregano for L-ORAC. However, absolute intake from these sources would be expected to be very small relative to foods consumed in larger quantities. Variation of antioxidant levels in fruits and vegetables exists due to multiple factors including cultivar, growing and harvesting conditions, and food processing/preparation. For example, organically grown blueberries have a significantly higher ORAC than blueberries grown in conventional culture.⁴⁶ Although consumption of certain foods can raise the demonstrable antioxidant capacity in plasma,⁴⁷ the degree of potential discrepancy between antioxidant capacity as estimated by the ORAC and in vivo antioxidant effects remains unknown.

In summary, our findings add to the body of evidence supporting a protective role of antioxidant nutrients in the development of NHL. The protective benefit appears to be primarily derived from food and not from supplements, suggesting that any association with antioxidants is more likely to be mediated through foods. The inverse association of lipophilic ORAC with NHL risk after accounting for intake of specific antioxidant nutrients further suggests that there may be synergy between antioxidant compounds within foods that are not captured by traditional approaches. These results also suggest that the summary ORAC measure captures additional information on antioxidant intake beyond traditional nutrient and food-based approaches. Finally, our results, in the context of the broader literature, suggests that prevention strategies should be aimed at improving antioxidant intake from foods rather than supplements, and understanding mechanistically why food-derived antioxidants appear to confer superior protection against lymphomagenesis.

Novelty and impact of the paper.

The Oxygen Radical Absorbance Capacity (ORAC) assay, which measures total antioxidant capacity of individual foods and accounts for synergism, can be estimated using a food-frequency questionnaire (FFQ). Higher antioxidant intake as estimated by the FFQ-ORAC, particularly the lipophilic component, was associated with a lower NHL risk after accounting for other antioxidant nutrients and vegetable intake, supporting this as potentially useful summary measure of total antioxidant intake.

Abbreviations

NHL

non-Hodgkin lymphoma

ORAC

Oxygen Radical Absorbance Capacity

FFQ

food-frequency questionnaire

ORs

odds ratios

ROS

Oxygen Radical Absorbance Capacity

TEAC

trolox equivalent antioxidant capacity

FRAP

ferric ion reducing antioxidant power

USDA

United States Department of Agriculture

CLL/SLL

chronic lymphocytic leukemia/small lymphocytic lymphoma

FL

follicular lymphoma

DLBCL

diffuse large B cell lymphoma

H-ORAC

hydrophilic ORAC

L-ORAC

lipophilic ORAC

CI

confidence interval