Dietary Pattern and Risk of Hodgkin Lymphoma in a Population-Based Case-Control Study

Abstract

Classic Hodgkin lymphoma (cHL) has few known modifiable risk factors, and the relationship between diet and cHL risk is unclear. We performed the first investigation of an association between dietary pattern and cHL risk in 435 cHL cases and 563 population-based controls from Massachusetts and Connecticut (1997–2000) who completed baseline diet questionnaires. We identified 4 major dietary patterns (“vegetable,” “high meat,” “fruit/low-fat dairy,” “desserts/sweets”) using principal components analysis. We computed multivariable odds ratios and 95% confidence intervals for associations of dietary pattern score (quartiles) with younger-adult (age <50 years), older-adult (age ≥50 years), and overall cHL risk. Secondary analyses examined associations by histological subtype and tumor Epstein-Barr virus (EBV) status. A diet high in desserts/sweets was associated with younger-adult (odds ratio_{(quartile 4 vs. quartile 1)} = 1.60, 95% confidence interval: 1.05, 2.45; P_trend = 0.008) and EBV-negative, younger-adult (odds ratio = 2.11, 95% confidence interval: 1.31, 3.41; P_trend = 0.007) cHL risk. A high meat diet was associated with older-adult (odds ratio = 3.34, 95% confidence interval: 1.02, 10.91; P_trend = 0.04) and EBV-negative, older-adult (odds ratio = 4.64, 95% confidence interval: 1.03, 20.86; P_trend = 0.04) cHL risk. Other dietary patterns were not clearly associated with cHL. We report the first evidence for a role of dietary pattern in cHL etiology. Diets featuring high intake of meat or desserts and sweets may increase cHL risk.

Classic Hodgkin lymphoma (cHL) is a relatively rare cancer with a poorly understood etiology. The epidemiology of cHL is noteworthy for a bimodal age incidence curve in developed countries, with incidence rates peaking in young adulthood (age 15–34 years) and in older adulthood (age ≥50 years) (1, 2). Known risk factors for cHL are largely unmodifiable and include age, sex, a family history of cHL or non-Hodgkin lymphoma, having a compromised immune system, and history of Epstein-Barr virus (EBV) infection. Personal history of infectious mononucleosis, which is often a manifestation of primary EBV infection, has been associated with an elevated risk of developing EBV-positive cHL, as have childhood social environment factors that influence age at first exposure to EBV, such as sibship size, birth order, and nursery school attendance (3–6). Serological studies suggest that severe/chronic EBV infection is associated with cHL risk, particularly with EBV-positive cHL independent of infectious mononucleosis (7, 8). Observations that the distributions of EBV-positive tumors and cHL histological subtypes vary by age group suggest age-related heterogeneity in the role of EBV and other risk factors in cHL etiology (9–11).

A diagnosis of cHL is pathologically confirmed by the presence of malignant Reed-Sternberg cells in affected tissue. The microenvironment of Reed-Sternberg cells, which originate from B lymphocytes, is rich in reactive inflammatory cells and suggests that chronic inflammation may play a role in cHL pathogenesis, possibly in conjunction with EBV infection (12–17). Of note, numerous inflammatory mediators have been identified in Reed-Sternberg cells and Hodgkin lymphoma (HL)-derived cell lines and have been associated with HL risk in human populations (13, 15, 16, 18, 19). It is therefore plausible that modifiable factors that disrupt or modulate the immune system may be associated with cHL risk. Consistent with that hypothesis, we previously observed a 40% decreased risk of cHL in association with regular use of aspirin (20).

Of interest, specific dietary patterns have been associated with levels of inflammatory biomarkers (21). For example, diets rich in fruits, vegetables, and whole grains have been associated with lower levels of proinflammatory markers (22–25). In contrast, a “Western” style diet high in fat, refined grains, red and processed meats, and oils was associated with higher levels of inflammatory markers in the same populations and with inflammation-related chronic diseases (22–31).

Few previous studies have examined associations between dietary factors and cHL risk (32–39). These studies focused largely on individual food items or nutrients and reported inconsistent results. To our knowledge, no study has yet examined an association between dietary pattern and the risk of cHL. We undertook the present population-based case-control analysis to evaluate whether dietary patterns influence the risk of cHL in younger or older adults.

METHODS

Study population

Cases of incident HL were recruited from the greater Boston metropolitan area of Massachusetts and the state of Connecticut from August 1, 1997, to December 31, 2000. Eligible patients were aged 15–79 years, living within the target geographical area and without human immunodeficiency virus (HIV) infection at diagnosis. Cases were identified by using the rapid case ascertainment systems of Harvard and Yale universities with additional support from the Massachusetts and Connecticut state tumor registries. There were 677 eligible cases invited to participate in the study, and 84% (n = 567) consented (32). Certain data used in this study were obtained from the Connecticut Tumor Registry located in the Connecticut Department of Public Health.

Population-based controls were frequency matched to cases by age (within 5 years), sex, and state of residence (Massachusetts or Connecticut) and did not have a personal history of HL. In greater Boston, controls were identified through the “Town Books,” annual records documenting all citizens aged ≥17 years, which are 90% complete (40). Of 720 invited controls in Massachusetts, 51% (n = 367) consented. In Connecticut, 450 eligible controls aged 18–65 years were identified by random-digit dialing, and 61% (n = 276) consented. Of 69 eligible controls in Connecticut aged 66–79 years identified through the Health Care Financing Administration (Medicare), 52% (n = 36) consented to participate (32).

The original research protocol was approved by the institutional review boards of the Harvard School of Public Health, Yale University School of Medicine, Johns Hopkins University School of Medicine, all participating hospitals, the Massachusetts Cancer Registry, and the Connecticut Tumor Registry in the Connecticut Department of Public Health (4, 20, 32). The present analysis of nonidentifiable study data was deemed exempt by the Harvard School of Public Health Human Subjects Committee.

Histopathology

Study pathologists reviewed all available pathology material to confirm an HL diagnosis (4, 41–43). When possible, cHL cases were further subtyped as nodular sclerosis, mixed cellularity, lymphocyte-deleted cHL, or lymphocyte-rich cHL. Sixteen cases with nodular lymphocyte-predominant subtype HL were excluded from the analysis, as this subtype is considered biologically and clinically distinct from cHL. Tumor tissue was analyzed for EBV through in situ hybridization for EBV-encoded RNA transcripts and/or immunohistochemistry to detect the viral latency membrane protein in Reed-Sternberg cells (41, 44, 45). A tumor was considered positive for EBV if at least 1 assay was positive, and negative otherwise (46).

Data collection

Lifestyle information was collected through a structured telephone interview for 97% of study participants, while 3% completed an abbreviated mailed study questionnaire. Additionally, 511 cases (93%) and 648 controls (95%) completed a validated, semiquantitative food frequency questionnaire (FFQ) to assess average consumption of 61 food and beverage items, plus vitamin and mineral supplements, over the year prior to enrollment (47). Participants reported the average frequency of consumption for each food according to commonly used units or portion sizes, which were then converted into standard servings per day. Participants were excluded from the present analysis if they left ≥3 FFQ items blank or reported a total energy intake >3 standard deviations from the sex-specific mean on the natural log scale (n = 77).

After exclusions, we had complete FFQ data on 881 participants. An additional 183 study participants had missing data on only 1 or 2 food items. To avoid unnecessarily reducing statistical power, we imputed a value of 0 servings per day for the 43 foods for which ≥20% of the remaining study population reported 0 servings per day (Web Appendix 1 available at http://aje.oxfordjournals.org/), as missing values for infrequently consumed foods are likely to indicate 0 consumption (48, 49). The cutoff was based on an evaluation of the distribution of missing and reported 0 intake, and changes in the cutoff did not meaningfully affect the results. Missing values on foods for which <20% of the population reported 0 servings per day were retained as missing, and these individuals were excluded from the study population (n = 66). The dietary pattern analysis thus included 435 cases and 563 controls.

Dietary patterns

To identify dietary patterns common to the study population, we conducted a principal components analysis of the 61 food and beverage items included on the FFQ, followed by a varimax orthogonal rotation to improve interpretability and minimize correlation between components. The number of principal components (i.e., eigenvectors) retained in the analysis was determined graphically using the scree test, which plotted the eigenvalues (i.e., the amount of total variance explained by a principal component) by each principal component (50–52). We retained 4 principal components following this assessment, each representing a separate, uncorrelated dietary pattern. Dietary patterns were ranked according to eigenvalue and described through identification of the major foods contributing to the pattern based on each food item's factor loading coefficient. As these variables were not normally distributed, we categorized the scores for each dietary pattern into quartiles. We found similar dietary patterns when the principal components analysis was conducted among the controls only and among cases and controls together; therefore, we retained patterns derived from both cases and controls because of greater stability (53–56).

Statistical analysis

Multivariable unconditional logistic regression models were used to estimate odds ratios and 95% confidence intervals for the association between quartile of a given dietary pattern (lowest quartile as the reference group) and cHL risk. Linear trend across quartiles was assessed in multivariable models by modeling the median values of the quartiles as a semicontinuous variable. All tests of statistical significance were 2 sided. Multivariable models were adjusted for the matching factors, plus total daily caloric intake (kcal/day), body mass index expressed as weight (kg)/height (m)² (<25, 25–29.9, ≥30), and potential confounders previously found to be associated with overall cHL risk: race/ethnicity (non-Hispanic white, other/missing), nursery school/day-care attendance for ≥1 year (yes, no/did not attend/missing), history of smoking ≥10 lifetime packs of cigarettes (yes, no), education (less than high school, high school, more than high school), number of siblings (0, 1, 2, 3, 4, ≥5), alcohol intake (drinker, nondrinker), regular aspirin use (≥2 regular-strength tablets/week; yes, no) and regular acetaminophen use (≥2 regular-strength tablets/week; yes, no). The primary analysis stratified cHL models by age group at diagnosis (<50 vs. ≥50) to explore potential heterogeneity in the associations of dietary patterns with younger- or older-adult cHL. Five controls missing data on smoking status (n = 1) or number of siblings (n = 4) were dropped from multivariable models, as separate categories for missing data introduced instability to the models.

Because the distributions of EBV-positive tumors and cHL histological subtypes suggest age-related heterogeneity in the role of cHL risk factors, we conducted secondary analyses to examine the association of dietary pattern with cHL risk separately by tumor EBV status (positive, negative) or histological subtype (mixed cellularity, nodular sclerosing). Age-stratified models for tumor EBV status limited covariable adjustment to matching factors, total caloric intake, and body mass index, while models for older-adult mixed cellularity subtype were not adjusted for nursery school attendance because of sparse categories. All analyses were conducted by using SAS, version 9.2, statistical software (SAS Institute, Inc., Cary, North Carolina).

RESULTS

Of the 435 cases in the final analysis, 83% were aged <50 years and 17% were aged ≥50 years at diagnosis; 68% of cases were diagnosed with nodular sclerosing cHL, and 17% of cases had EBV-positive tumors (Table 1). The study population was largely Caucasian and highly educated.

Table 1.

Descriptive Characteristics of the Study Population, a Case-Control Study of Hodgkin Lymphoma in Connecticut and Massachusetts, 1997–2000

Characteristic	Cases (n = 435)		Controls (n = 563)
	No.	%	No.	%
Sex
Male	220	51	306	54
Female	215	49	257	46
State of residence
Massachusetts	242	56	315	56
Connecticut	193	44	248	44
Age, yearsa
<50	361	83	451	80
≥50	74	17	112	20
Ethnicity
Caucasian	403	93	493	88
Non-Caucasian/missing	32	7	70	12
Body mass indexb
<25	232	53	264	47
25–29.9	135	31	199	35
≥30	68	16	100	18
Smoked ≥10 packs of cigarettes in lifetime
Yes	214	49	250	44
No	221	51	312	55
Missing	0	0	1	0.2
Day care or nursery school
Attended for ≥1 year	48	11	89	16
Attended for <1 year	89	20	107	19
Did not attend/missingc	298	69	367	65
Education
Less than high school	34	8	36	6
Completed high school	113	26	130	23
College or higher	288	66	396	70
Number of siblings
0	26	6	37	7
1–2	248	57	299	53
3–4	114	26	156	28
≥5	47	11	67	12
Unknown	0	0	4	1
Average aspirin use during the past 5 years
≥2 tablets per week	53	12	96	17
<2 tablets per week	371	85	440	78
Missing	11	3	27	5
Average acetaminophen use during the past 5 years
≥2 tablets per week	110	25	94	17
<2 tablets per week	312	72	438	78
Missing	13	3	31	5
Diagnosed with mononucleosis
Yes	58	13	81	14
No	368	85	478	85
Unknown	9	2	4	1
Alcohol intaked
Nondrinker	151	35	155	28
Drinker	284	65	408	72
Total caloric intakee,f	1,607 (442–4,119)		1,604 (433–4,398)
Histological type
Nodular sclerosing	295	68
Multiple cellularity	51	12
Other/undefined	89	20
Tumor EBV status
Positive	73	17
Negative	255	59
Unknown	107	24

Abbreviation: EBV, Epstein-Barr virus.

^a Age at diagnosis among cases and age at study entry among controls.

^b Expressed as weight (kg)/height (m)².

^c Unable to differentiate between participants who did not attend day care or nursery school and those missing this information.

^d Alcohol intake as reported on food frequency questionnaire.

^e Total caloric intake reported as the mean and range of values in kcal/day.

^f Values are expressed as mean (range).

The principal components analysis identified 4 distinct dietary patterns. The most prominent dietary pattern (i.e., the pattern that explained the greatest amount of total variance) was characterized by high intake of vegetables (eigenvalue: 4.6; 7.6% of total variance), with the highest positive loading factors observed for broccoli, cabbage/cauliflower, cooked spinach/collard greens, and string beans. The second most prominent dietary pattern represented a high meat diet (eigenvalue: 4.1; 6.7% of total variance), with high positive loading factors observed for hamburger, bacon, and beef, pork, or lamb meat. Other factors moderately correlated with the high meat pattern included refined or complex carbohydrates (white bread, potatoes) and sugar-sweetened beverages (Web Table 1). The third dietary pattern (eigenvalue: 2.1; 3.4% of total variance) featured high loading factors for fruits (fresh apples/pears, oranges, other fruit), as well as low-fat dairy products (yogurt, cheese). The fourth dietary pattern (eigenvalue: 1.8; 3.0% of total variance) was characterized by high intake of desserts and sweets (ready-made pie, ready-made cake, home-baked pie, ice cream). Together, the 4 dietary patterns accounted for 20.7% of the variability in the original dietary variables. Additional details on each dietary pattern are in Web Table 1.

The magnitude of most observed associations between dietary pattern and cHL risk did not change notably between univariable and multivariable models (Table 2). The risk of younger-adult cHL was increased among individuals with a dietary pattern high in desserts and sweets (multivariable odds ratio_{(quartile 4 vs. quartile 1)} = 1.60_, 95% confidence interval: 1.05, 2.45; P_trend = 0.01), whereas the risk of older-adult cHL was not associated with this dietary pattern (P_trend = 0.67). In contrast, the risk of older-adult cHL was significantly increased among individuals consuming a high meat dietary pattern (odds ratio = 3.34, 95% confidence interval: 1.02, 10.91; P_trend = 0.04), but this dietary pattern was not associated with younger-adult cHL (P_trend = 0.85). The dietary pattern high in fruit and low-fat dairy intake and the dietary pattern high in vegetable intake were not significantly associated with risk of overall, younger-adult, or older-adult cHL.

Table 2.

Associations Between Dietary Pattern and Risk of Classic Hodgkin Lymphoma, Overall and by Age Group, Connecticut and Massachusetts, 1997–2000

cHL by Age Group and Quartile	Univariable				Multivariablea				Multivariableb
	No. of Cases	No. of Controls	Odds Ratios	95% CI	No. of Cases	No. of Controls	Odds Ratios	95% CI	No. of Cases	No. of Controls	Odds Ratios	95% CI
	High Vegetable Dietary Pattern
Overall cHL
Quartile 1	118	131	1.00	Referent	118	131	1.00	Referent	118	129	1.00	Referent
Quartile 2	113	137	0.92	0.64, 1.30	113	137	0.94	0.66, 1.34	113	137	0.98	0.67, 1.41
Quartile 3	107	143	0.83	0.58, 1.18	107	143	0.85	0.59, 1.22	107	141	0.87	0.60, 1.27
Quartile 4	97	152	0.71	0.50, 1.01	97	152	0.73	0.50, 1.07	97	151	0.78	0.52, 1.16
Ptrendc			0.05				0.09				0.19
Age <50 yearsd
Quartile 1	108	119	1.00	Referent	108	119	1.00	Referent	108	117	1.00	Referent
Quartile 2	97	115	0.93	0.64, 1.35	97	115	0.93	0.64, 1.36	97	115	0.93	0.63, 1.38
Quartile 3	84	107	0.87	0.59, 1.27	84	107	0.86	0.59, 1.27	84	106	0.85	0.57, 1.27
Quartile 4	72	110	0.72	0.49, 1.07	72	110	0.74	0.49, 1.10	72	109	0.76	0.49, 1.16
Ptrendc			0.10				0.13				0.17
Age ≥50 years
Quartile 1	10	12	1.00	Referent	10	12	1.00	Referent	10	12	1.00	Referent
Quartile 2	16	22	0.87	0.30, 2.51	16	22	0.94	0.32, 2.81	16	22	1.12	0.35, 3.51
Quartile 3	23	36	0.77	0.29, 2.06	23	36	0.68	0.24, 1.90	23	35	0.72	0.25, 2.11
Quartile 4	25	42	0.71	0.27, 1.89	25	42	0.51	0.17, 1.49	25	42	0.68	0.22, 2.13
Ptrendc			0.46				0.14				0.34
	Western-Style Dietary Pattern
Overall cHL
Quartile 1	108	141	1.00	Referent	108	141	1.00	Referent	108	139	1.00	Referent
Quartile 2	109	141	1.01	0.71, 1.44	109	141	1.09	0.76, 1.57	109	141	1.09	0.75, 1.59
Quartile 3	104	146	0.93	0.65, 1.33	104	146	1.06	0.72, 1.56	104	144	1.06	0.72, 1.59
Quartile 4	114	135	1.10	0.77, 1.57	114	135	1.38	0.87, 2.20	114	134	1.39	0.86, 2.26
Ptrendc			0.71				0.25				0.26
Age <50 yearsd
Quartile 1	91	107	1.00	Referent	91	107	1.00	Referent	91	105	1.00	Referent
Quartile 2	88	105	0.99	0.66, 1.47	88	105	1.03	0.68, 1.55	88	105	1.02	0.67, 1.56
Quartile 3	86	122	0.83	0.56, 1.23	86	122	0.90	0.58, 1.38	86	121	0.89	0.57, 1.39
Quartile 4	96	117	0.97	0.65, 1.42	96	117	1.11	0.66, 1.86	96	116	1.08	0.63, 1.87
Ptrendc			0.67				0.92				0.99
Age ≥50 years
Quartile 1	17	34	1.00	Referent	17	34	1.00	Referent	17	34	1.00	Referent
Quartile 2	21	36	1.17	0.53, 2.58	21	36	1.20	0.52, 2.76	21	36	1.14	0.48, 2.72
Quartile 3	18	24	1.50	0.65, 3.49	18	24	1.79	0.70, 4.54	18	23	1.78	0.67, 4.74
Quartile 4	18	18	2.00	0.83, 4.80	18	18	2.93	0.97, 8.86	18	18	3.34	1.02, 10.91
Ptrendc			0.10				0.05				0.04
	Fruit/Low-Fat Dairy Dietary Pattern
Overall cHL
Quartile 1	118	131	1.00	Referent	118	131	1.00	Referent	118	129	1.00	Referent
Quartile 2	103	147	0.78	0.55, 1.11	103	147	0.75	0.52, 1.07	103	147	0.70	0.48, 1.01
Quartile 3	108	142	0.84	0.59, 1.20	108	142	0.78	0.54, 1.13	108	140	0.80	0.55, 1.18
Quartile 4	106	143	0.82	0.58, 1.17	106	143	0.74	0.49, 1.11	106	142	0.70	0.45, 1.07
Ptrendc			0.38				0.17				0.18
Age <50 yearsd
Quartile 1	89	103	1.00	Referent	89	103	1.00	Referent	89	101	1.00	Referent
Quartile 2	90	114	0.91	0.62, 1.36	90	114	0.90	0.60, 1.34	90	114	0.86	0.57, 1.30
Quartile 3	89	110	0.94	0.63, 1.39	89	110	0.89	0.59, 1.36	89	109	0.94	0.61, 1.46
Quartile 4	93	124	0.87	0.59, 1.28	93	124	0.85	0.54, 1.33	93	123	0.83	0.51, 1.34
Ptrendc			0.53				0.49				0.55
Age ≥50 years
Quartile 1	29	28	1.00	Referent	29	28	1.00	Referent	29	28	1.00	Referent
Quartile 2	13	33	0.38	0.17, 0.87	13	33	0.33	0.14, 0.78	13	33	0.24	0.10, 0.62
Quartile 3	19	32	0.57	0.27, 1.24	19	32	0.53	0.23, 1.18	19	31	0.43	0.18, 1.04
Quartile 4	13	19	0.66	0.28, 1.59	13	19	0.50	0.19, 1.33	13	19	0.42	0.15, 1.19
Ptrendc			0.33				0.17				0.13
	Desserts/Sweets Dietary Pattern
Overall cHL
Quartile 1	92	157	1.00	Referent	92	157	1.00	Referent	92	155	1.00	Referent
Quartile 2	106	144	1.26	0.88, 1.80	106	144	1.21	0.84, 1.74	106	144	1.16	0.80, 1.69
Quartile 3	120	130	1.58	1.10, 2.25	120	130	1.56	1.09, 2.24	120	128	1.49	1.03, 2.17
Quartile 4	117	132	1.51	1.06, 2.16	117	132	1.59	1.10, 2.30	117	131	1.50	1.02, 2.21
Ptrendc			0.01				0.005				0.02
Age <50 yearsd
Quartile 1	80	135	1.00	Referent	80	135	1.00	Referent	80	133	1.00	Referent
Quartile 2	89	125	1.20	0.82, 1.77	89	125	1.16	0.79, 1.72	89	125	1.10	0.73, 1.65
Quartile 3	101	97	1.76	1.19, 2.60	101	97	1.73	1.17, 2.57	101	96	1.63	1.08, 2.46
Quartile 4	91	94	1.63	1.10, 2.44	91	94	1.69	1.12, 2.54	91	93	1.60	1.05, 2.45
Ptrendc			0.004				0.002				0.008
Age ≥50 years
Quartile 1	12	22	1.00	Referent	12	22	1.00	Referent	12	22	1.00	Referent
Quartile 2	17	19	1.64	0.63, 4.29	17	19	1.33	0.49, 3.59	17	19	1.43	0.48, 4.23
Quartile 3	19	33	1.06	0.43, 2.60	19	33	0.90	0.35, 2.26	19	32	0.78	0.28, 2.21
Quartile 4	26	38	1.25	0.53, 2.97	26	38	1.07	0.43, 2.70	26	38	0.94	0.34, 2.61
Ptrendc			0.90				0.87				0.60

Abbreviations: cHL, classic Hodgkin lymphoma; CI, confidence interval.

^a Logistic regression models adjusted for sex, age (overall only), state of residence, total daily caloric intake, and body mass index.

^b Logistic regression models additionally adjusted for race, alcohol intake, smoking, education, attended nursery school >1 year, number of siblings, regular aspirin use, and regular acetaminophen use.

^c Test for trend across quartiles from logistic regression model, modeling the median of each quartile as a semicontinuous variable.

^d Age defined as age at diagnosis (years) for cases and age at study entry for controls.

When cHL risk was examined jointly by age group and tumor EBV status, younger adults consuming a diet rich in desserts and sweets had more than a 2-fold increased risk of EBV-negative cHL (multivariable odds ratio_{(quartile 4 vs. quartile 1)} = 2.11, 95% confidence interval: 1.31, 3.41; P_trend = 0.001) (Table 3). Among the markedly smaller stratum of older adults, we observed a positive association between the high meat dietary pattern and tumor EBV-negative cHL that appeared strong but imprecise (odds ratio = 4.64, 95% confidence interval: 1.03, 20.86; P_trend = 0.04) among 35 cases. There were no significant associations between dietary pattern and tumor EBV-positive cHL in either age group or overall (Web Table 2).

Table 3.

Association Between Dietary Pattern and Risk of Classic Hodgkin Lymphoma by Tumor Epstein-Barr Virus Status, Overall and by Age Group for Selected Dietary Patterns, Connecticut and Massachusetts, 1997–2000

cHL by Age Group and Quartile	High Meat Dietary Pattern				Desserts/Sweets Dietary Pattern
	No. of Cases	No. of Controls	Odds Ratioa	95% CI	No. of Cases	No. of Controls	Odds Ratioa	95% CI
	EBV-Positive Tumors
Overall
Quartile 1	17	141	1.00	Referent	19	157	1.00	Referent
Quartile 2	15	141	0.91	0.43, 1.94	22	144	1.31	0.67, 2.55
Quartile 3	16	146	0.99	0.46, 2.14	15	130	0.96	0.47, 1.98
Quartile 4	25	135	1.74	0.73, 4.17	17	132	1.06	0.52, 2.16
Ptrendb			0.16				0.93
Age <50 yearsc
Quartile 1	12	107	1.00	Referent	16	135	1.00	Referent
Quartile 2	10	105	0.92	0.37, 2.30	15	125	1.04	0.49, 2.23
Quartile 3	10	122	0.84	0.33, 2.18	13	97	1.16	0.53, 2.55
Quartile 4	22	117	2.33	0.83, 6.52	10	94	0.87	0.37, 2.04
Ptrendb			0.07				0.79
Age ≥50 years
Quartile 1	5	34	1.00	Referent		22	1.00	Referent
Quartile 2	5	36	0.68	0.16, 2.85	7	19	2.85	0.60, 13.60
Quartile 3	6	24	1.33	0.32, 5.60	2	33	0.42	0.06, 2.80
Quartile 4	3	18	0.68	0.10, 4.50	7	38	1.27	0.27, 5.94
Ptrendb			0.91				0.71
	EBV-Negative Tumors
Overall
Quartile 1	59	141	1.00	Referent	52	157	1.00	Referent
Quartile 2	67	141	1.23	0.79, 1.89	62	144	1.24	0.80, 1.92
Quartile 3	65	146	1.24	0.78, 1.96	67	130	1.53	0.99, 2.36
Quartile 4	64	135	1.47	0.84, 2.57	74	132	1.77	1.15, 2.74
Ptrendb			0.21				0.008
Age <50 yearsc
Quartile 1	53	107	1.00	Referent	45	135	1.00	Referent
Quartile 2	56	105	1.14	0.70, 1.83	54	125	1.23	0.77, 1.97
Quartile 3	56	122	1.03	0.62, 1.70	59	97	1.80	1.13, 2.88
Quartile 4	55	117	1.15	0.62, 2.11	62	94	2.11	1.31, 3.41
Ptrendb			0.77				0.001
Age ≥50 years
Quartile 1	6	34	1.00	Referent	7	22	1.00	Referent
Quartile 2	11	36	1.81	0.57, 5.74	8	19	1.00	0.29, 3.48
Quartile 3	9	24	2.67	0.73, 9.73	8	33	0.67	0.21, 2.18
Quartile 4	9	18	4.64	1.03, 20.86	12	38	0.77	0.24, 2.46
Ptrendb			0.04				0.60

Abbreviations: CI, confidence interval; EBV, Epstein-Barr virus.

^a All logistic regression models adjusted for sex, state of residence, total caloric intake, and body mass index.

^b Test for trend from logistic regression model, modeling the median of each quartile as a semicontinuous variable.

^c Age defined as the age at diagnosis for cases and the age at study entry for controls.

In histological subtype-specific analyses, consuming a diet high in desserts and sweets was associated with a nonsignificant 43% increased risk of nodular sclerosing cHL overall (n = 295 cases) (Table 4). Odds ratios for the association between a diet high in desserts and sweets and nodular sclerosing cHL risk were similar in magnitude for younger- and older-adult nodular sclerosing cHL. The analysis of 51 mixed cellularity cHL cases yielded a suggestive, nonsignificant inverse association with the dietary pattern high in fruit and low-fat dairy intake. Because of small numbers, we were unable to cross-classify any case group by EBV status and histological subtype.

Table 4.

Associations Between Dietary Pattern and Classic Hodgkin Lymphoma by Histological Subtype, Overall and by Age Group, Connecticut and Massachusetts, 1997–2000

cHL by Age Group and Quartile	Diet Pattern
	High Vegetable				Western Style				Fruit/Low-Fat Dairy				Desserts/Sweets
	No. of Cases	No. of Controls	Odds Ratio	95% CI	No. of Cases	No. of Controls	Odds Ratio	95% CI	No. of Cases	No. of Controls	Odds Ratio	95% CI	No. of Cases	No. of Controls	Odds Ratio	95% CI
	Nodular Sclerosinga
Overallb
Quartile 1	84	131	1.0	Referent	76	141	1.0	Referent	74	131	1.0	Referent	60	157	1.0	Referent
Quartile 2	74	137	0.88	0.58, 1.34	76	141	1.07	0.70, 1.62	78	147	0.80	0.53, 1.21	82	144	1.34	0.88, 2.04
Quartile 3	74	143	0.83	0.55, 1.26	68	146	1.05	0.67, 1.65	73	142	0.77	0.50, 1.19	74	130	1.29	0.84, 1.98
Quartile 4	63	152	0.69	0.45, 1.06	75	135	1.39	0.81, 2.41	70	143	0.69	0.42, 1.13	79	132	1.43	0.93, 2.21
Ptrendc			0.07				0.24				0.15				0.16
Age <50 yearsd
Quartile 1	81	119	1.0	Referent	68	107	1.0	Referent	62	103	1.0	Referent	56	135	1.0	Referent
Quartile 2	68	115	0.90	0.58, 1.40	65	105	1.04	0.65, 1.66	70	114	0.90	0.57, 1.42	73	125	1.25	0.80, 1.94
Quartile 3	60	107	0.88	0.55, 1.39	63	122	0.93	0.57, 1.52	64	110	0.90	0.55, 1.45	67	97	1.48	0.93, 2.35
Quartile 4	51	110	0.76	0.47, 1.25	64	117	1.06	0.58, 1.94	64	124	0.76	0.45, 1.30	64	94	1.66	1.03, 2.69
Ptrendc			0.24				0.92				0.33				0.04
Age ≥50 years
Quartile 1	3	12	1.0	Referent	8	34	1.0	Referent	12	28	1.0	Referent	4	22	1.0	Referent
Quartile 2	6	22	1.51	0.25, 9.16	11	36	1.04	0.32, 3.40	8	33	0.26	0.07, 0.94	9	19	3.70	0.69, 19.86
Quartile 3	14	36	1.63	0.32, 8.22	5	24	0.95	0.21, 4.35	9	32	0.40	0.12, 1.32	7	33	0.94	0.18, 5.03
Quartile 4	12	42	1.22	0.22, 6.78	11	18	3.13	0.62, 15.83	6	19	0.41	0.10, 1.72	15	38	1.58	0.32, 7.75
Ptrendc			0.99				0.15				0.28				0.93
	Mixed Cellularity
Overall
Quartile 1	11	131	1.0	Referent	7	141	1.0	Referent	23	131	1.0	Referent	15	157	1.0	Referent
Quartile 2	14	137	0.91	0.38, 2.17	13	141	1.88	0.70, 5.04	8	147	0.35	0.14, 0.83	4	144	0.33	0.11, 1.06
Quartile 3	14	143	0.95	0.40, 2.26	13	146	1.77	0.64, 4.92	11	142	0.56	0.25, 1.28	19	130	1.91	0.89, 4.10
Quartile 4	12	152	0.69	0.28, 1.73	18	135	2.66	0.83, 8.53	9	143	0.38	0.15, 1.01	13	132	1.29	0.56, 2.98
Ptrendc			0.43				0.15				0.07				0.20
Age <50 yearsd
Quartile 1	7	119	1.0	Referent	4	107	1.0	Referent	12	103	1.0	Referent	10	135	1.0	Referent
Quartile 2	11	115	1.37	0.48, 3.94	7	105	1.10	0.29, 4.21	5	114	0.51	0.16, 1.59	3	125	0.37	0.09, 1.47
Quartile 3	9	107	1.31	0.42, 4.04	8	122	0.98	0.25, 3.80	8	110	0.98	0.34, 2.82	12	97	2.72	1.00, 7.40
Quartile 4	6	110	0.84	0.24, 2.95	14	117	1.30	0.29, 5.79	8	124	0.61	0.19, 2.02	8	94	1.69	0.57, 4.99
Ptrendc			0.68				0.72				0.54				0.14
Age ≥50 yearse
Quartile 1	4	12	1.0	Referent	3	34	1.0	Referent	11	28	1.0	Referent	5	22	1.0	Referent
Quartile 2	3	22	0.20	0.03, 1.43	6	36	3.59	0.63, 20.43	3	33	0.17	0.04, 0.85	1	19	0.14	0.01, 1.70
Quartile 3	5	36	0.27	0.05, 1.50	5	24	4.57	0.69, 30.22	3	32	0.27	0.05, 1.31	7	33	0.89	0.18, 4.37
Quartile 4	6	42	0.19	0.03, 1.22	4	18	6.78	0.72, 63.88	1	19	0.07	0.01, 0.81	5	38	0.72	0.13, 3.90
Ptrendc			0.20				0.12				0.02				0.91

Abbreviation: CI, confidence interval.

^a Histological subtype determined by pathology review.

^b Logistic regression models adjusted for sex, state of residence, total daily caloric intake, body mass index, race, alcohol intake, smoking, education, attended nursery school >1 year, number of siblings, regular aspirin use, and regular acetaminophen use.

^c Test for trend from logistic regression model, modeling the median of each quartile as a semicontinuous variable.

^d Age defined as the age at diagnosis for cases and the age at study entry for controls.

^e Model for older-adult mixed cellularity subtype adjusted for sex, state of residence, total daily caloric intake, body mass index, race, alcohol intake, smoking, education, number of siblings, regular aspirin use, and regular acetaminophen use.

DISCUSSION

In this population-based case-control study, we observed positive associations between a diet high in desserts and sweets and cHL risk, particularly with total and EBV-negative younger-adult cHL and with nodular sclerosing cHL. We also observed positive associations between a high meat dietary pattern and risk of older-adult cHL, and particularly EBV-negative older-adult cHL, although power was limited in some of the jointly classified strata. A suggestive, albeit imprecise inverse association between a diet characterized by high intake of fruit and low-fat dairy products and mixed cellularity cHL risk warrants further study. We did not observe any statistically significant associations between cHL risk and a dietary pattern high in vegetable intake.

In this analysis, we characterized dietary patterns through a principal components analysis, which agnostically defined dietary patterns that best captured the food consumption habits reported in the study FFQ and accounted for the variability among participants' diet. The 4 selected dietary patterns together accounted for 20.7% of the total variability in the original diet variables, reflecting the considerable variability of individual diets. The amount of total explained variability is consistent with other epidemiologic studies of dietary patterns derived from principal components analysis, which explained between 19% and 24% of variability in those study populations (23, 29, 57).

We observed several significant positive associations between cHL risk and a diet high in desserts and sweets, especially among younger adults. An experimental study in overweight adults suggested that a high-sucrose diet may increase levels of the inflammatory marker C-reactive protein (58). However, a recent 6-month randomized trial of high glycemic index diet, low glycemic index diet, and low-fat diet among overweight and obese adults observed no significant changes in inflammatory or metabolic risk biomarkers among the diet groups (59). In contrast, other studies have suggested that low glycemic load diets may be associated with reduced levels of inflammatory markers including C-reactive protein (60, 61). We also observed significant positive associations between cHL risk and a high meat dietary pattern, in particular among older adults. Previous studies have linked a diet high in fat, refined grains, red and processed meats, and oils to higher levels of proinflammatory biomarkers, including C-reactive protein and interleukin 6 (22–25), which have also been associated with HL risk (14, 18, 62), suggesting a possible biological link between a Western-style dietary pattern and cHL risk through inflammation. An inflammation-related mechanism may underlie the observed associations between cHL risk and both the desserts/sweets and high meat dietary patterns; however, it is unclear why this mechanism would vary between age groups.

Existing studies of diet and HL risk have yielded inconsistent results, although some elements of a Western-style diet have been associated previously with an increased HL risk. Previous analyses of cHL and diet in our study population showed a significant positive association between EBV-negative cHL and multivitamin, supplemental vitamin B₆, and supplemental vitamin B₁₂ use (32); a significant positive association of younger-adult cHL risk with higher saturated fat intake in women; and a significant inverse association of younger-adult cHL risk with higher monounsaturated fat intake, also among women (33). An analysis of the prospective California Teachers Study cohort found a significant 2-fold increased risk of HL with high consumption of dark/whole grain bread, and no association with various phytocompounds (34). Case-control studies of dietary factors and HL risk in Italy reported suggestive inverse associations with higher intake of whole grains (36) and no association with fruit, vegetable, or red meat intake (37, 38), while higher intake of liver and ham increased HL risk (39), although case numbers were generally small. A multisite case-control study observed an increased risk of HL associated with dietary vitamin A intake in men (63).

Overall, our findings of an increased risk of EBV-negative disease related to high consumption of a diet rich in desserts and sweets (for younger adults) and meat (for older adults) suggest that diet may influence HL risk independently of EBV infection. Alternatively, any association of diet with EBV-positive cHL risk may be weak in comparison to the influences of altered host control of EBV infection and related immune dysregulation (8). This hypothesis is supported by the lack of any strong association between dietary pattern and EBV-positive disease in our study population, although numbers of EBV-positive cases were small, and statistical power was limited to detect such an association.

Although our study population was one of the largest case-control studies of HL to date, certain subgroup analyses had limited statistical power, most notably those stratified by cHL histological subtype or related to EBV-positive cHL. The effect estimates from those models are highly imprecise and must be interpreted with particular caution until confirmed in other populations. Cases may have recalled their diet habits differently from controls, leading to recall bias. In addition, as diet was recalled for the year prior to index date or diagnosis, we cannot rule out reverse causality or that the reference period did not adequately capture diet habits in the time window of greatest relevance to disease initiation or promotion (although this time period is not known). To avoid unnecessarily excluding 117 people who missed only 1 or 2 FFQ items, we imputed values of 0 servings/day for 43 foods that were generally less commonly consumed in the study population. Some of these omissions may not truly represent 0 servings/day, and thus we may have introduced a small amount of misclassification into the diet scores. Control participation rates varied slightly by state and recruitment method and were lower than among cases. However, we replaced nonparticipating Massachusetts controls with individuals from the same residential area to limit the potential for selection bias, and Chang et al. (20) demonstrated that the income distribution across census tracts of Massachusetts controls who participated was representative of the source population (33). Data to evaluate potential selection bias among Connecticut controls were not available.

Our study is the first to examine a potential association between dietary pattern and cHL risk by using a validated FFQ to retrospectively assess diet. The large study population allowed us to collect information on, and statistically adjust for, a number of potential confounding factors, including those identified in prior analyses of this study population. In addition, we conducted analyses separately by age group, tumor EBV status, and histological subtype, which may help to elucidate etiologically distinct subtypes of cHL. In conclusion, this study is the first to suggest that dietary pattern may play a role in the development of cHL, with apparent variation by age group, tumor EBV status, and possibly also by histological subtype. These results require confirmation in additional populations, including in studies with prospectively assessed dietary intake.