C:\lea7\n&c441\nc441.vp

NUTRITION AND CANCER, 44(1), 44–51Copyright 2002, Lawrence Erlbaum Associates, Inc.
A Case-Control Study of Dietary
Phytoestrogens and Testicular Cancer Risk
Farzana L. Walcott, Michael Hauptmann, Cherie M. Duphorne,
Patricia C. Pillow, Sara S. Strom, and Alice J. Sigurdson
Abstract: A few dietary studies have found elevated
postnatal endocrine factors likely plays a role (3–19) and testicular cancer risks for higher red meat, fat, and milk in- could include dietary estrogenic plant compounds (phyto- takes and lower intakes of fruits, vegetables, and fiber. Be- estrogens). A few studies have found increased testicular cause hormonal modulation by dietary intake of plant cancer risk among men whose diets were high in fat, red estrogens could affect risk of testicular cancer, we chose to meats, and milk or low in fiber, fruits, and vegetables explore the possible relationship between dietary phytoes- (20–22), suggesting that these diets might also be low in trogens and testicular cancer. We conducted a hospital- phytoestrogens. The fact that dietary factors have been asso- based case-control study of 159 testicular cancer cases di- ciated with testicular cancer risk supports further investiga- agnosed between 1990 and 1996 and 136 adult friend- tion of the constituents in diet that may influence risk, matched controls at the University of Texas M. D. Anderson particularly phytoestrogen intake, because these estrogenic Cancer Center. Amounts of phytoestrogenic compounds in plant compounds are known to modulate hormone levels in foods were added to the National Cancer Institute’s DietSys program and then grouped into prelignans, lignans, flavo- Phytoestrogens are naturally occurring compounds found noids, isoflavonoids, phytosterols, and coumestrol for statis- in many plant foods. Technically, they are defined as plant tical analysis, expressed per 1,000 kcal. The results of substances or plant precursor derivatives that are structurally multivariate logistic regression analysis showed, after ad- or functionally similar to estradiol and consist of a number of justment for age, education, income, ethnicity, cryptorchi- subclasses, including isoflavones, prelignans, lignans, cou- dism, body mass index, baldness unrelated to therapy, severe mestrol, flavonoids, and phytosterols. Phytoestrogens have acne in adolescence, early puberty, daily fiber and fat in- recently become of interest in cancer prevention because of take, and total daily calories, no discernable monotonic in- the broad range of anticarcinogenic properties exhibited by creased or decreased risk estimates across quartiles of these compounds, including antioxidant, antimutagenic, and phytoestrogen intake. A U-shaped pattern was observed for antiproliferative capabilities (29). Furthermore, recent bio- lignans and coumestrol. Further evaluation of this pattern chemical studies have shown that phytoestrogens modulate by cubic spline parameterization did fit the data, but the human sex hormone-binding globulin, aromatase, and b-hy- data were also consistent with no effect. This hypothesis- droxysteroid dehydrogenase, which are critical in steroid me- generating study does not support the premise that dietary phytoestrogens increase or decrease testicular cancer risk For these reasons, we undertook a hypothesis-generating analysis to explore the relationship of dietary phytoestrogensand risk of testicular cancer. We conducted a hospital-basedcase-control study at the University of Texas M. D. Ander- Introduction
son Cancer Center. Specifically, we examined whether tes-ticular cancer risk increased, decreased, or showed no Testicular cancer accounts for ~1% of malignant neo- relationship as dietary phytoestrogen intake increased. Be- plasms in men but is the most common tumor in young adult cause age of onset, pathological features, and clinical treat- men aged 20–34 yr in the United States (1,2). Few risk fac- ment are different depending on tumor type, we also tors, other than cryptorchidism, have consistently been asso- assessed whether risk estimates for phytoestrogen intake ciated with testicular cancer, but an effect of pre- and varied by testicular cancer histopathology.
F. L. Walcott, C. M. Duphorne, P. C. Pillow, S. S. Strom, and A. J. Sigurdson are affiliated with the Department of Epidemiology, University of Texas M. D. Anderson Cancer Center, Houston, TX 77030. M. Hauptmann is affiliated with the Biostatistics Branch, Division of Cancer Epidemiology and Genetics,National Cancer Institute, Bethesda, MD 20892.
same (data not shown) and to avoid the loss of information,we presented the results with the matching dissolved.
To adjust for total energy intake, all dietary factors were We identified men with testicular cancer who registered analyzed per 1,000 kcal total energy intake, and total energy at the M. D. Anderson Cancer Center between January 1990 intake was included in all models according to the nutrient and October 1996 through the M. D. Anderson Tumor Reg- density adjustment method described by Willett (41). Before istry and the M. D. Anderson Genitourinary Oncology analysis of phytoestrogens, a basic model was established, Clinic. Controls were adult male friends of the cases, including demographic variables (ethnicity, age, education, matched by ethnicity and age within 5 yr. Potential cases and and income), known or suspected testicular cancer risk, or controls were defined as men who were alive during the beneficial factors (self-reported history of cryptorchidism data-collection phase of the investigation, were between the and early onset of puberty, i.e., pubic hair noticed before age ages of 18 and 55 yr at the time of case diagnosis, and who 13), history of severe adolescent acne (severe enough that a lived in Texas, Louisiana, Arkansas, or Oklahoma. Details physician was seen for the acne), hair loss or balding not due of the case and control selection, inclusion criteria rationale, to cancer therapy, body mass index, and important dietary and participation rates have been published previously (22).
variables (total energy intake, total fat intake, and dietary fi- All men diagnosed with testicular cancer were eligible for ber intake). Body mass index (calculated by dividing weight inclusion, regardless of their ethnicity, tumor stage, and his- in kg by height in m2) and all dietary factors were catego- tology. Pathology reports were reviewed for all cases. We rized into quartiles on the basis of distribution among con- grouped teratoma, embryonal carcinoma, and choriocarci- trols. The units and categories of all variables are shown in noma as nonseminomas, pure seminomas were grouped as Table 1. Other variables analyzed that did not improve the seminomas, and pathology reports with seminomatous and base model fit were daily marijuana smoking and milk and nonseminomatous elements were grouped as mixed germ meat consumption. We chose this approach because we cell tumors. This research was approved by the University of wished to assess the effect of phytoestrogens in the presence Texas M. D. Anderson Internal Review Board and the Uni- of significant variables in our data.
versity of Texas-Houston School of Public Health Commit- All phytoestrogens were collapsed into six groups: pre- tee for the Protection of Human Subjects.
lignans, lignans, flavonoids, isoflavonoids, phytosterols, and Cases and controls completed self-administered ques- coumestrol. The original compounds comprising these tionnaires eliciting information on demographics; lifestyle groupings were prelignans (or lignan precursors of secoiso- habits; medical history, including history of cryptorchidism, lariciresinol and matairesinol), lignans (enterodiol and en- family history of cancer, puberty onset (based on the age the terolactone), flavonoids (quercetin, kaempferol, luteolin, man reported noticing pubic hair), severe adolescent acne apigenin, and myricetin), isoflavonoids (genistein, daidzein, (such that they consulted a physician for it), and balding (un- formononetin, and biochanin A), phytosterols (b-sitosterol, related to case cancer therapy); body size and shape; and campesterol, and stigmasterol), and coumestrol. Odds ratios diet. To assess diet, we used a modified and revised version for categories of phytoestrogen groups were calculated in a of the National Cancer Institute’s Health Habits and History univariate (with addition of 1 phytoestrogen group at a time Questionnaire (HHHQ), which contained 152 foods and to the basic model) and in a multivariate way (with addition beverages (34,35) and has been validated in a range of popu- of all phytoestrogen groups to the basic model).
lations (36,37). The time period assessed by the question- Linear trend of the log odds ratios was evaluated on the naire was the year before cancer diagnosis for cases and the basis of the significance of the slope estimate of the respec- previous year for controls. From this information, we calcu- tive continuous variable. All significance tests were two- lated food consumption and nutrient intake by using DietSys sided, and the significance level was 0.05. Because this was (version 4.0), the nutrient analysis program developed for a hypothesis-generating study and we wanted to reduce as- the National Cancer Institute’s HHHQ (38). The food-fre- sumptions about the nature of the relationships observed, we quency questionnaire was modified to include foods that attempted to more flexibly model the dose response between have been previously reported to be significant sources of phytoestrogens and testicular cancer using a cubic spline phytoestrogens. Also, the database was expanded to include within a logistic regression model to parameterize the spe- phytoestrogen values for foods assessed by the question- cific phytoestrogen under study. Splines are smooth piece- naire. Detailed methodology of the database construction wise polynomial functions of high flexibility with the and application to assess prostate cancer risk have been pub- segments separated by knots (42). In this analysis, two seg- ments were used separated at the median of the phytoestro- Because there were multiple cases who would be excluded in the analysis of matched data because they did not have afriend control, we evaluated the effect of dissolving the match on our crude and adjusted results. To do so, we compared theunconditional and conditional logistic regression point esti- In total, 187 cases and 148 controls participated in the mates for all cases and controls, regardless of matching. Be- study. Twenty-eight cases were excluded because their cause the point estimates in both analyses were essentially the HHHQs were not completed properly, leaving 159 cases for Vol. 44, No. 1
analysis. Among controls, 12 were excluded because of im- mors. The ethnic distribution, age, education, and income of proper completion of the HHHQ, leaving 136 healthy con- men with nonseminoma, seminoma, and mixed germ cell tu- mors were generally similar to those of controls. The only In Table 1, we present selected demographic characteris- exception was for seminoma, which has a known older age tics for cases and controls grouped by histopathological at diagnosis than nonseminoma. History of cryptorchidism type. Eighty-one men were diagnosed with nonseminomas, was more common in all histological groups than in the con- 46 with pure seminomas, and 32 with mixed germ cell tu- trols. Indicators of hormonal status (severe adolescent acne Table 1. Selected Demographic, Risk Factor Characteristics, and Phytoestrogen Intake of Controls and Testicular Cancer
Cases by Histologya
Nutrition and Cancer 2002
Table 1. (Continued)
Total daily dietary fiber intake, g/1,000 kcal a: Values in parentheses are percentages.
and balding as reported on the questionnaire) were similar Univariate and multivariate logistic regression analyses of between cases and controls, except a greater proportion of dietary phytoestrogen groupings (comparing quartiles of con- cases uniformly reported less adolescent acne and balding.
sumption) are presented in Table 3. A U-shaped relationship Body mass index was also similar between cases and con- most consistently describes the univariate point estimates for trols. Total caloric intake among men with testicular cancer prelignans, lignans, isoflavonoids, phytosterols, and coumes- was consistently higher across all histological types than trol for all histological types combined and when analyzed among controls, as was total fat intake, whereas total daily separately. The multivariable estimates, adjusted for age, edu- fiber intake was consistently lower in cases. Among the cation, income, ethnicity, cryptorchidism, early onset of pu- phytoestrogen classes, intake was generally similar in cases berty (self-reported as the age at which pubic hair was and controls, except cases tended to consume less total daily noticed), history of severe adolescent acne, hair loss or bald- ing not due to cancer therapy, body mass index, total energy We present the ranked food items that contributed to the intake, total fat intake, and dietary fiber intake, show an atten- majority of each phytoestrogen for cases and controls in Ta- uation of the U-shaped relationship, but it persists for lignans ble 2. In general, the rank order of foods consumed among and coumestrol. For seminoma, the odds ratios tended to de- cases and controls was very similar, except for genistein, crease as lignan intake increased (P for trend = 0.02), showing daidzein, and enterodiol. For genistein and daidzein, the a relationship pattern slightly different from that of the other cases consumed soy nuts and soy meat substitutes, whereas phytoestrogens and testicular cancer histologies.
controls consumed miso soup. For enterodiol, cases tended The spline models for lignans and coumestrol tended to to eat French fries and controls ate green salad.
support the U shape of the categorical odds ratios. However, Vol. 44, No. 1
Table 2. Major Foods Contributing ³10% to Total Phytoestrogen Intake Among Testicular Cancer Cases and Controls
Mayonnaise/salad dressing (29%), margarine (14%), dark Mayonnaise/salad dressing (24%), dark bread (12%), mar- Mayonnaise/salad dressing (22%), green salad (19%) Green salad (20%), mayonnaise/salad dressing (18%) Cranberry juice (24%), onions (11%), spaghetti with to- Soy nuts (62%), soy meat substitutes (10%) Miso soup (34%), dried soy beans (13%), soy nuts (13%), Miso soup (48%), soy nuts (13%), dried soy beans (12%) Snow peas (60%), refried beans (20%), lima beans/ French fries (14%), other potatoes (12%), green salad Green salad (12%), other potatoes (12%), French fries a: Mammalian lignans enterolactone and enterodiol are produced in the colon from prelignan precursors in foods. These food values were obtained from a previous study using in vitro fermentation of different food products (44).
confidence intervals (not shown) are very wide. The esti- planation would be that the associations are spurious be- mated spline is significantly different from a constant line cause of the limitations of small sample size and inadequate for coumestrol (P = 0.02) and almost significantly different power. On the other hand, it may be that the relationship be- for lignans (P = 0.07), but not for other phytoestrogen tween estrogenic agents and testicular cancer risk is not straightforward. In a recent study of prenatal diethylstilbes-trol (DES) exposure, counterintuitive results were also re-ported. Among a cohort of men exposed in utero to DES, Discussion
risk of testicular cancer was elevated in men exposed tolower, rather than higher, levels of DES (19).
There are few published studies on the relationship be- There is limited evidence that diet may modulate testicu- tween diet and testicular cancer risk and no studies on phyto- lar cancer risk. Previous epidemiological studies on diet and estrogen consumption and risk of testicular cancer. To our testicular cancer have found a protective effect for consump- knowledge, this is the first exploratory analysis of testicular tion of green vegetables on risk for testicular cancer (43) and cancer risk and phytoestrogen intake. The results of the pres- increased risk of disease with consumption of milk, red ent study do not show a linear trend or dose-response pattern meat, total fat, and heterocyclic amines (20,21). A previ- associated with phytoestrogen intake and testicular cancer ously published analysis from this study group observed an risk. A U-shaped pattern might best describe the relationship increased testicular cancer risk associated with high con- between testicular cancer and lignans and coumestrol, but sumption of fat and meat (22) after adjustment for caloric in- these patterns are only marginally convincing, as evidenced by the weakly suggestive spline modeling results. Prelignans There are several limitations in our study. Because tes- and phytosterols showed a modestly reduced risk for testicu- ticular cancer is relatively rare, to accrue sufficient case lar cancer with moderate intake, but not convincingly so.
numbers, we included men diagnosed up to 6 yr before the One explanation for these results is the high level of correla- questionnaire was administered. The greatest concern is the tion between these compounds, particularly prelignans and potential for bias in dietary recall inherent in case-control phytosterols (data not shown), and so it might be difficult to studies, although Willett (41) suggests that diet can be ade- differentiate the effects of one from the other. Another ex- quately recalled for up to 10 yr with acceptable levels of Nutrition and Cancer 2002
Vol. 44, No. 1
misclassification. We used adult friends of cases as controls, 6. Brown LM, Pottern LM, and Hoover RN: Prenatal and perinatal risk because we thought population-based controls would not re- factors for testicular cancer. Cancer Res 46, 4812–4816, 1986.
7. Moss A, Osmond D, Baccheti P, Torti FM, and Gurgin V: Hormonal flect the population from which the cases arose, despite the risk factors in testicular cancer: a case-control study. Am J Epidemiol problem that friends might be too “similar” to cases. We do 124, 39–52, 1986.
not believe that “overmatching” occurred, because even 8. Swerdlow AJ, Huttly SR, and Smith PG: Prenatal and familial associa- though our controls were not too dissimilar from cases, they tions of testicular cancer. Br J Cancer 55, 571–577, 1987.
tended to be slightly older and better educated than cases.
9. Prener A, Hsieh CC, Engholm G, Trichopoulos D, and Jensen OM: Birth order and risk of testicular cancer. Cancer Causes Control 3,
However, this implies that the associations between diet and testicular cancer may be spurious because of the inclusion of 10. United Kingdom Testicular Cancer Study Group: Aetiology of tes- more highly educated controls. For example, the more edu- ticular cancer: association with congenital abnormalities, age at pu- cated controls may be more conscious of fat intake and berty, infertility, and exercise. Br Med J 308, 1393–1399, 1994.
methods to reduce dietary fat consumption. However, any 11. Gallagher RP, Huchcroft S, Phillips N, Hill GB, Coldman AJ, et al.: Physical activity, medical history, and risk of testicular cancer (Alberta conscious dietary alterations based on news reports of soy and British Columbia, Canada). Cancer Causes Control 6, 398–406,
product benefits (e.g., reduction of prostate cancer risk) postdated data collection, so little impact on phytoestrogen 12. Braun MM, Ahlbom A, Floderus B, Brinton LA, and Hoover RN: Ef- fect of twinship on incidence of cancer of the testis, breast, and other In conclusion, a comprehensive hormonal mechanism for sites (Sweden). Cancer Causes Control 6, 519–524, 1995.
13. Akre O, Ekbom A, Hsieh CC, Trichopoulos D, and Adami HO: testicular cancer has not been clearly established, and it Testicular nonseminoma and seminoma in relation to perinatal charac- should be remembered that phytoestrogens have been noted teristics. JNCI 88, 883–889, 1996.
for actions that suggest estrogenicity or antiestrogenicity, as 14. Petridou E, Roukas KI, Dessypris N, Aravantinos G, and Bafaloukos well as antioxidant effects. Therefore, the mechanism of ac- D: Baldness and other correlates of sex hormones in relation to tion of phytoestrogens in humans and cancer risk remains testicular cancer. Int J Cancer 71, 982–985, 1997.
15. Weir HK, Kreiger N, and Marrett LD: Age at puberty and risk of highly speculative. As steroidogenic organs, the human tes- testicular germ cell cancer (Ontario, Canada). Cancer Causes Control tes are prime targets for hormonal modulation by exogenous 9, 253–258, 1998.
hormones. Although we did not find an effect of dietary 16. Akre O, Ekbom A, Sparen P, and Tretli S: Body size and testicular can- phytoestrogen intake on testicular cancer risk because the cer. JNCI 92, 1093–1096, 2000.
human diet is the primary means of exposure to exogenous 17. Weir HK, Marrett LD, Kreiger N, Darlington GA, and Sugar L: Pre- natal and peri-natal exposures and risk of testicular germ-cell cancer.
hormonal substances, further research on the impact of ex- Int J Cancer 87, 438–443, 2000.
posure to phytoestrogens and other hormonally active agents 18. Srivastava A and Kreiger N: Relation of physical activity to risk of in the diet on testicular cancer risk is warranted.
testicular cancer. Am J Epidemiol 151, 78–87, 2000.
19. Strohsnitter WC, Noller KL, Hoover RN, Robboy SJ, and Palmer JR: Cancer risk in men exposed in utero to diethylstilbestrol. JNCI 93,
Acknowledgments and Notes
20. Davies TW, Palmer CR, Ruja E, and Lipscombe JM: Adolescent milk, dairy product and fruit consumption and testicular cancer. Br J Cancer The authors thank the men who participated in the study. This work 74, 657–660, 1996.
was supported by the University of Texas M. D. Anderson Cancer Center 21. De Stefani E, De Stefani R, and Alvaro L: Risk factors for testicular Education Program in Cancer Prevention, which is supported by National cancer: a case-control study in Uruguay. Abstr 32nd Annu Conf Int Cancer Institute Grant R25-CA-57730. Address correspondence to A.
Assoc Cancer Registries, 17 August 1998, Atlanta, GA, PC-5.
Sigurdson, Radiation Epidemiology Branch, Div. of Cancer Epidemiol- 22. Sigurdson AJ, Chang S, Annegers JF, Duphorne CM, Pillow PC, et al.: ogy and Genetics, National Cancer Institute, NIH, 6120 Executive Blvd., A case-control study of diet and testicular carcinoma. Nutr Cancer 34,
EPS 7092, MSC 7238, Bethesda, MD 20892-7238. Phone: (301) 594–7911. FAX: (301) 402–0207. E-mail: [email protected].
23. Hamalainen EK, Adlercreutz H, Puska P, and Pietinen P: Decrease of serum total and free testosterone during a low-fat high-fiber diet. J Submitted 18 March 2002; accepted in final form 10 June 2002.
Steroid Biochem 18, 369–370, 1983.
24. Howie BJ and Schultz TD: Dietary and hormonal interrelationships among vegetarian Seventh-Day Adventists and non-vegetarian men.
References
Am J Clin Nutr 42, 127–134, 1985.
25. Rose DP: Dietary fiber, phytoestrogens, and breast cancer. Nutrition 8,
1. Greenlee RT, Hill-Harmon MB, Murray T, and Thun M: Cancer statis- tics, 2001. CA Cancer J Clin 51, 15–36, 2001.
26. Rose DP: Diet, hormones, and cancer. Annu Rev Publ Health 14, 1–17,
2. Surveillance, Epidemiology, and End Results (SEER) Cancer Statistics Review, 1973–1998, Ries LAG, Eisner MP, Kosary CL, Hankey BF, 27. Thompson LU: Antioxidants and hormone-mediated health benefits of Miller BA, et al. (eds). Bethesda, MD: National Cancer Institute, 2001.
whole grains. Crit Rev Food Sci Nutr 34, 473–497, 1994.
(http://seer.cancer.gov/Publications/CSR1973_1998/) 28. Adlercreutz CH, Goldin BR, Gorbach SL, Hockerstedt KAV, Wata- 3. Henderson BE, Benton B, Jing J, Yu MC, and Pike MC: Risk factors nabe S, et al.: Soybean phytoestrogen intake and cancer risk. J Nutr for cancer of the testis in young men. Int J Cancer 23, 598–602, 1979.
125, 757S–770S, 1995.
4. Schottenfeld D, Warshauer ME, Sherlock S, Zauber AG, Leder M, et 29. Knight DC and Eden JA: A review of the clinical effects of phyto- al.: The epidemiology of testicular cancer in young adults. Am J estrogens. Obstet Gynecol 87, 897–904, 1996.
Epidemiol 112, 232–246, 1980.
30. Adlercreutz H, Bannwart C, Wahala K, Makela T, Brunow G, et al.: In- 5. Depue RH, Pike MC, and Henderson BE: Estrogen exposure during hibition of human aromatase by mammalian lignans and isoflavonoid gestation and risk of testicular cancer. JNCI 71, 1151–1154, 1983.
phytoestrogens. J Steroid Biochem Mol Biol 44, 147–153, 1993.
Nutrition and Cancer 2002
31. Keung WM: Dietary estrogenic isoflavones are potent inhibitors of b- tered three times over three different seasons. Nutr Cancer 25, 305–
hydroxysteroid dehydrogenase of P. testosteronii. Biochem Biophys Res Commun 215, 1137–1144, 1995.
38. HHHQ-DIETSYS Analysis Software, version 4.0. Bethesda, MD: 32. Schottner M, Spiteller G, and Gansser D: Lignans interfering with 5a- SEER–Scientific Systems, National Cancer Institute, 1998. (http:// dihydrotestosterone binding to human sex hormone-binding globulin.
www-seer.ims.nci.nih.gov/ScientificSystems/DIETSYS) J Nat Prod 61, 119–121, 1998.
39. Pillow PC, Duphorne CM, Chang S, Contois JH, Strom SS, et al.: De- 33. Krazeisen A, Breitling R, Moller G, and Adamski J: Phytoestrogens in- velopment of a database for assessing dietary phytoestrogen intake.
hibit human 17b-hydroxysteroid dehydrogenase type 5. Mol Cell En- Nutr Cancer 33, 3–19, 1999.
docrinol 171, 151–162, 2001.
40. Strom SS, Yamamura Y, Duphorne CM, Spitz MR, Babaian RJ, et al.: 34. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, et al.: A Phytoestrogen intake and prostate cancer: a case-control study using a data-based approach to diet questionnaire design and testing. Am J new database. Nutr Cancer 33, 20–25, 1999.
Epidemiol 124, 453–469, 1986.
41. Willet WC: Nutritional Epidemiology, 2nd ed. Oxford, UK: Oxford 35. Block G, Coyle LM, Hartman AM, and Scoppa SM: Revision of di- etary analysis software from the health habits and history question- 42. Greenland S: Dose-response and trend analysis in epidemiology: alter- naire. Am J Epidemiol 139, 1190–1196, 1994.
natives to categorical analysis. Epidemiology 6, 356–365, 1995.
36. Block G, Thompson FE, Hartman AM, Larkin GA, and Guire KE: 43. Gallagher RP, Huchcroft S, and Phillips N: Physical activity and di- Comparisons of two dietary questionnaires validated against multiple etary factors in testicular cancer (abstr). Epidemiology S27, 108, 1994.
dietary records collected during a 1-year period. J Am Diet Assoc 92,
44. Thompson LU, Robb P, Serriano M, and Cheung F: Mammalian lignan production from various foods. Nutr Cancer 16, 43–52, 1991.
37. Hartman AM, Block G, Chan W, Williams J, and McAdams M: Repro- ducibility of a self-administered diet history questionnaire adminis- Vol. 44, No. 1

Source: http://bioinformatics.nki.nl/~hauptmann/pubs/walcott_NutrCancer_2003.pdf