| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Short Communication |
1 Division of Preventive Oncology, Cancer Care Ontario, Toronto, Ontario, Canada; Departments of 2 Public Health Sciences and 3 Nutritional Sciences, Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada; and 4 Postgraduate Program in Epidemiology, Federal University of Pelotas, Pelotas, Brazil
Requests for reprints: Mark P. Purdue, Division of Cancer Epidemiology and Genetics, National Cancer Institute, 6120 Executive Boulevard, EPS 8121, MSC 7240, Bethesda, MD 20892-7240. Phone: 301-451-5036; Fax: 301-402-1819. E-mail: purduem{at}mail.nih.gov
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
There is limited evidence supporting some aspects of diet as risk factors for NHL. Experimental evidence from animal studies suggests that greater fat and protein intake can alter immune function and increase the risk of lymphomas (19-21). Nine epidemiologic studies have investigated the relationship between dietary factors and NHL (17, 22-29). Risk factors suggested from these studies include increased consumption of animal protein, milk, liver, meat, and fat and decreased consumption of fruits, vegetables, and whole grain foods. However, there is considerable inconsistency across the studies in the findings for these dietary components.
Tumors classified as NHL represent several distinct morphologic and histologic entities with different prognoses and responses to treatment; it has been speculated that disease etiology may also vary between these tumors (30, 31). Analyses of incident NHL tumors recorded in the Surveillance, Epidemiology and End Results registry suggest that NHL subtypes have distinct sex, age, race, geographic, and temporal patterns (32, 33). However, eight of the nine studies that investigated diet and NHL did not conduct subtype-specific analysis, probably due to insufficient power.
The National Enhanced Cancer Surveillance System (NECSS) of Canada is one of the largest studies of NHL conducted to date, with information on past diet and other exposures from 1,642 affected individuals (34). We examined these data to explore the associations between dietary factors and NHL by histologic subtype as defined using the Working Formulation (WF) classification system (35).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
Frequency matching was used in the selection of population controls to achieve a similar age and sex distribution to that of all cancer cases. Control selection varied by province, with provincial health insurance records used as a sampling frame for most provinces and with property assessment files (Ontario) or random digit dialing (Alberta and Newfoundland) used in others. In Alberta, a random sample of provincial telephone numbers, including unlisted numbers, was generated. Each randomly selected telephone number was called up to eight times on a pattern structured around call attempts during the day, evenings, and on Saturday. Of the numbers called, 4% were not in service or were businesses, there was a communication barrier in 4%, and no one could be reached for 11%. Of those households contacted, 91% agreed to a census of residents, and 90% of eligible individuals agreed to be sent a questionnaire. In Newfoundland, a random sample of listed and unlisted telephone numbers was also obtained. Exact contact and eligibility rates are unavailable from this province; study personnel estimated that 85% of telephone numbers were reached.
In total, questionnaires were mailed to 8,060 individuals selected as potential controls in the eight provinces. For 7% of individuals, the mailed questionnaire was returned because of a wrong or old address, and no updated address could be found through publicly available sources. In total, 5,039 controls completed and returned the questionnaire (70% of women and 65% of men; 63% of all ascertained controls).
Data Collection
Mailed questionnaires were used to collect information from subjects on suspected risk factors for cancer, with telephone follow-up for clarification of the answers as needed. Information on dietary habits 2 years before interview was collected using a 69-item food frequency questionnaire. Subjects were asked to indicate the frequency with which they consumed a specified portion size of each food; the nine possible answers ranged from <1 serving per month to 
6 servings per day. The dietary questionnaire was adapted from two previously validated instruments: the reduced Block questionnaire (38) and the questionnaire used in the Nurses' Health Study (39). The NCESS questionnaire is less comprehensive than these other instruments, because the study focused quite broadly on environmental causes of cancer rather than focusing specifically on nutrition.
The questionnaire also collected information on subjects' residential and occupational histories and on other possible risk factors, including education, income, ethnicity, height, weight, physical activity, smoking, and exposure to specific occupational carcinogens.
Data Analysis
The topographical and morphologic characteristics of all tumors were classified according to the ICDO-2 (36) based on the contents of pathology reports associated with the original histopathologic review of diagnostic material. Ontario cases had been classified according to the ICDO-1 system; these cases were recoded to ICDO-2 using the IARCtools software application (40). The ICDO-2 coding for Ontario cases was confirmed by a re-review of pathology reports. All tumors were grouped by histologic subtype based on ICDO-2 coding using the method developed by Groves et al. (33). This method, modeled after the WF classification system (35), categorizes tumors into six morphologic categories: small lymphocytic, follicular, diffuse, high grade, peripheral T cell, and not otherwise specified.
Information from the food frequency questionnaire was summarized into several foods or food groups: milk, cheese, fruit, vegetables, potatoes, legumes/nuts, breads/cereals, meat, fish, eggs, and sweets (Appendix 1). Subgroups were also defined as follows: yellow/orange, cruciferous, leafy, or other vegetable; whole grain, or nonwhole grain bread/cereal; and poultry, nonprocessed beef/pork/lamb, or processed beef/pork/lamb. Estimates of total weekly intake of animal protein, total fat, saturated fat, monounsaturated fat, polyunsaturated fat, and total energy were calculated from nutrient estimates assigned to each dietary item using Canadian nutrient data (41). All dietary variables were categorized with tertile cut points based on the distribution in the controls.
Data analysis was done using Stata (42). The associations between food groups and NHL risk were estimated using odds ratios (OR) and 95% confidence intervals (95% CI) calculated from maximum likelihood estimates using binary logistic regression. OR and 95% CI relating dietary components and histologic subgroups of NHL were calculated using polytomous logistic regression. Likelihood ratio tests assessing the presence of OR heterogeneity across disease subgroups compared the likelihoods of unconstrained polytomous models against those of models constrained to have identical ORs across disease subgroups.
ORs were estimated in two ways: adjusting for age and sex only and adjusting for age, sex, total energy intake, and suspected confounding factors. For analysis of food groups, energy intake was modeled as a continuous variable. Analyses of macronutrient intake employed the multivariate nutrient density method to adjust for energy intake (43). Nondietary variables (education, income adequacy, body mass index, smoking, alcohol consumption, and exposure to herbicides/pesticides) were considered to be confounders if their inclusion resulted in 10% change in the magnitude of ORs relating dietary variables to all NHL. Based on this criterion, only income adequacy (an index of household income adjusted for household size) and alcohol consumption were identified as confounders. In addition to estimating ORs for dietary variables individually, a single multivariable model was fit adjusting for intake of different food groups, total protein and total fat simultaneously, in addition to age, sex, income adequacy, energy intake (modeled using the multivariate nutrient density method), and alcohol consumption.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
60 years. High-grade and peripheral T-cell lymphomas were comparatively more common among young subjects (20% and 19% aged <40 years, respectively); at least half of the cases from these subtypes, along with those of follicular lymphomas, were aged <60 years at diagnosis. No clear differences across subtypes in the distributions of income adequacy and alcohol consumption were apparent.
|
|
|
|
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
An increased risk of NHL was found for high consumption of processed meat, saturated fat, and monounsaturated fat. No association with protein intake was found on adjustment for energy intake and other possible confounders. There is experimental evidence from animal studies that high intake of protein and fat can induce chronic hyperstimulation of the immune system and the development of lymphomas (19-21, 44). The observed association with processed meats may also involve the effect of nitrites, preservatives found in processed meats that are precursors to N-nitroso compounds, known carcinogens in animals (45). Other carcinogens, polycyclic aromatic hydrocarbons, are produced when fish or meat is fried or grilled; there is experimental evidence that high levels of ingested polycyclic aromatic hydrocarbon can induce immunotoxicity and lymphomas in mice (46).
Whereas evidence from animal studies supports an etiologic role for protein, fat, and meat intake, findings from epidemiologic studies of humans are inconsistent. A link between protein intake and NHL is supported by an early ecologic study that reported a positive correlation between per capita protein intake and lymphoma mortality (47). Associations with NHL were inconsistently observed across different types of meat in three hospital-based case-control studies from Italy (17, 22) and Uruguay (27); no analyses involving total protein and fat consumption were done. A population-based case-control study conducted in Nebraska by Ward et al. (25) found no association with consumption of animal products or with intake of animal or vegetable protein (fat was not assessed). However, the evidence from these studies is limited by their retrospective design and lack of adjustment for energy intake. Two published cohort analyses (26, 28) likely offer better insight into the relationship between NHL risk and intake of protein and fat. An analysis by Chiu et al. (26) of data from the Iowa Women's Health Study found an increased risk accompanying high consumption of red meat (particularly hamburgers), animal protein, and saturated and monounsaturated fats. In an analysis of the Nurses' Health Study, Zhang et al. (28) reported an increased risk with high intake of beef, pork, or lamb and saturated and trans-unsaturated fats; however, no association with protein intake was found. The results of studies that have investigated the etiologic importance of fat consumption have been generally consistent in identifying an association with fat intake. By contrast, the epidemiologic evidence investigating protein intake and NHL remains unclear.
We found NHL risk to be positively associated with high intake of cheese and eggs but not milk. None of the other studies that analyzed consumption of cheese and eggs reported any association with NHL risk (17, 22, 25, 26). The levels of consumption of these foods did not appreciably differ between those studies and ours. It is possible that our observed associations arose due to chance. High consumption of milk was associated with an increased risk of lymphatic cancers in a Norwegian cohort study (23), and a similar association with NHL risk was reported in the two Italian case-control studies (17, 22). Three other studies conducted in Uruguay and the United States found no association (25-27).
In our analysis, high consumption of dessert foods was weakly associated with elevated NHL risk. Zhang et al. (28) also identified positive associations with consumption of different dessert foods; desserts were not examined in the other studies. The investigators speculated that the increased risk may be attributable to the high levels of trans-unsaturated fat present in these foods. Dessert foods are also high in simple sugars, the consumption of which triggers insulin secretion. High insulin levels have been linked to an increased risk of cancers of the breast, colon, prostate, and lung (48). No such relationship with NHL has been reported; however, individuals diagnosed with diabetes have been found in some studies to have an increased risk of NHL (49-51).
Consumption of fruit and most types of vegetables was generally not found to be associated with NHL risk in our data. Six previous NHL studies have investigated fruit and vegetable consumption (17, 24-27, 29). High fruit intake was found to be associated with a low risk of NHL by Ward et al. (25) and Chiu et al. (26); no relationship was found in the other four studies. Evidence suggesting a protective effect from high consumption of at least some types of vegetables consumption has emerged from three studies (17, 25, 29), whereas other studies have reported no association (26) or weak evidence of a positive association with NHL risk (24, 27). There was no consistency in findings across the three cohort studies that investigated fruits, vegetables, and NHL risk (24, 26, 29). Although overall vegetable intake was not found to be a risk factor in our study, high consumption of items categorized as "other vegetables" (i.e., other than cruciferous, leafy, or yellow/orange vegetables) was positively associated with NHL. Vegetables naturally contain the suspected carcinogen nitrate (52). However, it is unlikely that nitrate intake underlies these associations, as nitrate-rich vegetables included in our questionnaire (cabbage, spinach, and other greens) were categorized as cruciferous or leafy vegetables and were not associated with NHL risk.
No difference in consumption of fresh fish was found between cases and controls. Investigators conducting a separate case-control analysis of NHL and fish consumption using the NECSS data reported a weak, nonsignificant inverse association with NHL risk accompanying consumption of 4 servings of fish per week (OR, 0.88; 95% CI, 0.71-1.10; ref. 53). This slightly different finding is likely due to differences between analyses in the choice of cut points for fish consumption, model covariates, and subject inclusion criteria. The findings from past epidemiologic studies of NHL do not suggest an association with fish intake (17, 26, 54).
The findings from the published studies relating diet to NHL risk are limited to varying degrees by a variety of methodologic issues, including a retrospective study design, use of hospital controls, accuracy in measuring past or current diet, and an absence of adjustment for energy intake. The results from the two methodologically strongest studies (Iowa Women's Study and Nurses' Health Study) are fairly consistent in reporting increased risk from red meat and fat intake but are contradictory with respect to the effects of protein, fruit, and vegetable intake (26, 28, 29). In addition, epidemiologic studies of diet (except ecologic analyses) are often limited in their ability to detect dietary effects because of the narrow range of dietary intake reported within study populations, such that only large associations are apparent (55). This issue may have contributed to the inconsistency observed across studies investigating diet and NHL. Such limitations may be particularly acute in our analysis, given our decision to categorize intake levels into only three groups to minimize problems of sparse cell counts in the analysis of disease subtypes. On the other hand, this problem may be offset to some extent by the large sample size of the NECSS, which provided this analysis with reasonable statistical power to detect weak associations.
It is possible that measurement error in the assessment of past diet may have affected the study findings. We believe such measurement error is most likely to be nondifferential in nature, given that diet is not a widely accepted risk factor for NHL and that the project was presented to subjects as a study of health and the environment. The usual effect of such error is to bias observed associations toward the null, which may partly explain the relatively weak dietary associations observed in this study. However, we cannot rule out the possibility that the dietary assessments of some cases were influenced by recent changes in their eating patterns due to the effects of the disease or its treatment. Such nondifferential measurement error could introduce bias toward or away from the null.
We believe it is unlikely that the observed study findings can be explained by confounding. OR estimates were adjusted for alcohol consumption and income adequacy, as these variables were found to influence the magnitude of some variable estimates. Occupational exposure to pesticides and herbicides has been previously identified as a risk factor for NHL but did not seem to confound the dietary associations with NHL in this study. Given that some past studies had been restricted to women (26, 28, 29) and that two studies found evidence of differential risk between sexes (25, 27), it is possible that the combined analyses done in this project may have obscured sex-specific differences. However, a reanalysis stratified by sex suggested no such differences.
If NHL is truly a collection of etiologically distinct lymphoid tumors, then it is possible that the inconsistent evidence in the published literature relating NHL and diet may be due to differences between studies in the distribution of NHL subtypes. The NECSS, with 1,642 cases and 5,039 controls, is one of the largest case-control studies of NHL developed to date and provided an opportunity to explore evidence of etiologic heterogeneity in NHL. Differences in sex and age distributions between subtypes consistent with those identified in previous analyses of Surveillance, Epidemiology and End Results data (32, 33) were apparent in the study data. This analysis generally found no evidence of heterogeneity across disease subtypes with respect to dietary associations. Associations with consumption of some vegetables and fats were found to be significantly different across subtypes. However, given the large number of diet/subtype comparisons made in this analysis, the possibility that these statistically significant findings arose by chance cannot be ruled out. Conversely, some aspects of study design may have limited the ability of this analysis to identify evidence of etiologic heterogeneity. The NECSS was only powered to detect main effects and not to detect heterogeneity in associations across disease subtypes (34); consequently, it is impossible to rule out the existence of such heterogeneity with any certitude.
This analysis may also have been limited in its ability to detect etiologic heterogeneity due to errors in classification of histologic subtype. Disease classification was based on the histopathologic tumor characteristics described in the original pathology reports rather than from review by a single expert pathologist. Given the demonstrated error in classifying NHL cases using the WF criteria (57-63% agreement among expert pathologists; refs. 56, 57) and in assigning ICDO codes to cases (77% agreement among Surveillance, Epidemiology and End Results Program coders; ref. 58), it is possible that a proportion of cases were assigned an incorrect subtype. Such misclassification, if independent of exposure, would lead to an attenuation of estimated subtype-specific associations. This is of particular concern for high-grade tumors, because the proportion of cases classified as high grade in our study is only half of the corresponding proportion from NHL cases registered in Surveillance, Epidemiology and End Results between 1978 and 1995 (5% versus 10%; ref. 33). These differences suggest that a high proportion of high-grade tumors in our study may have been classified into other categories (most likely not otherwise specified); as a result, the ORs for high-grade tumors should be interpreted with caution. Measurement error in the dietary assessment is likely another important source of OR attenuation. The absence of a difference in the distribution of histologic subtypes between Ontario participants and nonparticipants (
2 = 6.30, df = 5; P = 0.28) suggests that selection bias is unlikely to have affected the subtype-specific results.
It is possible that etiologic heterogeneity may exist between groups of NHL tumors characterized by differences other than their WF classification. A limitation of the WF classification system is that it was not designed to categorize tumors into separate disease entities. Instead, tumors are assigned to subgroups based on their expected clinical outcome. Furthermore, the WF system relies on histologic characteristics of tumors for classification; immunophenotypic and genetic characteristics are not taken into account. Since the creation of the WF in 1982, newer classification systems (the Revised European-American Classification of Lymphoid Neoplasms and the WHO Classification; ref. 59) have been developed that incorporate tumor immunophenotypic and genetic features and are based on a better understanding of lymphoid neoplasia. It would have been preferable to use such classification systems in our study; however, only a minority of tumors had been classified according to these systems and that information had not been collected as part of the NECSS. Ward et al. (25) investigated dietary factors separately by histologic and immunophenotypic type. No differences in dietary associations between different histologic types or between B-cell and T-cell lymphomas were found, although the study was underpowered to detect such heterogeneity.
In conclusion, we found a positive association between NHL risk and higher intake of saturated and monounsaturated fat and particular food items (cheese, eggs, processed meat, and sweet dessert foods). We did not find any clear evidence of heterogeneity between histologic subtypes of NHL in their associations with components of dietary intake. However, given the potential for misclassification in assessing both past dietary intake and histologic subtype, we cannot rule out the existence of such heterogeneity. To effectively investigate the existence of etiologic heterogeneity within NHL subgroups, studies including adequate numbers of cases and a standardized assessment of the histologic, immunophenotypic, and possibly molecular characteristics of tumors are needed. Meta-analyses of data pooled from different studies of NHL may represent an opportunity to conduct such investigations.
|
Appendix 1: List of dietary groups, subgroups, constituents, and weekly serving size as listed on the NECSS food frequency questionnaire |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
|
|
Acknowledgments |
|---|
|
Footnotes |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: The Canadian Cancer Registries Epidemiology Research Group comprises a principal investigator from each of the Provincial Cancer Registries: Bertha Paulse, M.Sc., B.N., Newfoundland Cancer Foundation; Ron Dewar, M.Sc., Nova Scotia Cancer Registry; Dagny Dryer, M.D., Prince Edward Island Cancer Registry; Nancy Kreiger, Ph.D., Cancer Care Ontario; Erich Kliewer, Ph.D., CancerCare Manitoba; Diane Robson, B.A., Saskatchewan Cancer Foundation; Shirley Fincham, Ph.D., Division of Epidemiology, Prevention and Screening, Alberta Cancer Board; and Nhu Le, Ph.D., British Columbia Cancer Agency. The group also includes Drs. Yang Mao and Kenneth Johnson from the Environmental Risk Assessment and Case Surveillance Division, Cancer Bureau, Center for Chronic Disease Prevention and Control, Health Canada.
Received 12/11/03; revised 3/29/04; accepted 5/ 4/04.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1: List of...References |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. H Eckel Egg consumption in relation to cardiovascular disease and mortality: the story gets more complex Am. J. Clinical Nutrition, April 1, 2008; 87(4): 799 - 800. [Full Text] [PDF]
|
|
|
|
 |
|
 S. Koutros, Y. Zhang, Y. Zhu, S. T. Mayne, S. H. Zahm, T. R. Holford, B. P. Leaderer, P. Boyle, and T. Zheng Nutrients Contributing to One-Carbon Metabolism and Risk of Non-Hodgkin Lymphoma Subtypes Am. J. Epidemiol., February 1, 2008; 167(3): 287 - 294. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. F. Skibola Obesity, Diet and Risk of Non-Hodgkin Lymphoma Cancer Epidemiol. Biomarkers Prev., March 1, 2007; 16(3): 392 - 395. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. E Kelemen, J. R Cerhan, U. Lim, S. Davis, W. Cozen, M. Schenk, J. Colt, P. Hartge, and M. H Ward Vegetables, fruit, and antioxidant-related nutrients and risk of non-Hodgkin lymphoma: a National Cancer Institute-Surveillance, Epidemiology, and End Results population-based case-control study Am. J. Clinical Nutrition, June 1, 2006; 83(6): 1401 - 1410. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
U. Lim, S. Weinstein, D. Albanes, P. Pietinen, L. Teerenhovi, P. R. Taylor, J. Virtamo, and R. Stolzenberg-Solomon Dietary factors of one-carbon metabolism in relation to non-hodgkin lymphoma and multiple myeloma in a cohort of male smokers. Cancer Epidemiol. Biomarkers Prev., June 1, 2006; 15(6): 1109 - 1114. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Polesel, R. Talamini, M. Montella, M. Parpinel, L. Dal Maso, A. Crispo, M. Crovatto, M. Spina, C. La Vecchia, and S. Franceschi Linoleic acid, vitamin D and other nutrient intakes in the risk of non-Hodgkin lymphoma: an Italian case-control study Ann. Onc., April 1, 2006; 17(4): 713 - 718. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. Cross, M. H. Ward, M. Schenk, M. Kulldorff, W. Cozen, S. Davis, J. S. Colt, P. Hartge, J. R. Cerhan, and R. Sinha Meat and meat-mutagen intake and risk of non-Hodgkin lymphoma: results from a NCI-SEER case-control study Carcinogenesis, February 1, 2006; 27(2): 293 - 297. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
U. Lim, M. Schenk, L. E. Kelemen, S. Davis, W. Cozen, P. Hartge, M. H. Ward, and R. Stolzenberg-Solomon Dietary Determinants of One-Carbon Metabolism and the Risk of Non-Hodgkin's Lymphoma: NCI-SEER Case-Control Study, 1998-2000 Am. J. Epidemiol., November 15, 2005; 162(10): 953 - 964. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Pre |
melon)
cup)