[Title will be auto-generated] by Erythis, Inc.

Antonia Trichopoulou, Klea Katsouyanni, Sherri Stuver, Lia Tzala, Charalambos Gnardellis, Eric Rimm, Dimitrios Trichopoulos, “Consumption of Olive Oil and Specific Food Groups in Relation to Breast Cancer Risk in Greece” J Natl Cancer Inst 1995; 87: 110

Consumption of Olive Oil and Specific Food Groups in Relation to Breast Cancer Risk in Greece

ABSTRACT: Background: Experimental animal studies suggest that olive oil consumption, as contrasted to consumption of other fat types, does not enhance the occurrence of chemically induced mammary tumors, but human data are sparse. Furthermore, evidence is inconclusive concerning the role of food groups, as distinct from that of major nutrients, in the etiology of breast cancer in women. Purpose: This analysis was conducted to evaluate and quantify the effect of consumption of olive oil, margarine, and a range of food groups on the risk of breast cancer. Methods: Data from a comprehensive, semiquantitative food-frequency questionnaire administered to 820 women with breast cancer and 1548 control women from the study base were used to calculate odds ratios (ORs) and X statistics of linear trend for the consumption of olive oil, margarine, and a series of food groups classified in quintiles. Adjustment for the effects of reproductive risk factors, energy intake, and mutual confounding influences was implemented through unconditional logistic regression modeling. Results: Vegetable consumption and fruit consumption were independently associated with statistically significant reductions of breast cancer risk by 12%

and 8%, respectively, per quintile increase; no significant associations were evident for the other food groups examined. Increased olive oil consumption was associated with significantly reduced breast cancer risk (OR = 0.75 [95% confidence interval = 0.57-0.98] for more than once a day versus once a day), whereas increased margarine consumption was associated with significantly increased risk (OR = 1.05 [95% confidence interval = 1.00-1.10] for an increment of four times a month). The olive oil association was apparently concentrated among postmenopausal women, but the relevant interaction term was not statistically significant; there was no suggestion of interaction with menopausal status for consumption of either vegetables, fruits, or margarine. Conclusions: Although major categories of macronutrients do not show significant associations with breast cancer risk in most studies, including the present one, vegetables and fruits are inversely, significantly, and strongly associated with this risk. There also is evidence that olive oil consumption may reduce the risk of breast cancer, whereas margarine intake appears to be associated with an elevated risk for the disease.

INTRODUCTION The evidence concerning an inverse association between consumption of vegetables and fruits on the one hand and breast cancer risk on the other is generally considered inconclusive [n1-n3], although a meta-analysis [n4] has reported a protective effect. There are several reasons for the widespread hesitation to attribute a major protective role to fruits and vegetables in the context of breast cancer. Early case-control studies on diet and breast cancer [reviewed by Steinmetz and Potter [n3]] did not account for energy intake in the analysis; adjustment for energy intake [n5,n6] controls for the consequences of general overreporting

or underreporting by case patients of their food consumption in case-control investigations [n7]. Furthermore, since the importance of energy intake as a critical factor in epidemiologic analyses was recognized and procedures to account for it were formulated [n5,n6,n8], the emphasis has moved away from food items and food groups to macronutrients and micronutrients believed to represent the biologically active ingredients [n9-n11]. This approach is justifiable in certain geographic and cultural settings where dietary supplements are extensively used [n12]. However, this approach may not be appropriate in other situations, since the

- 1/9 -

protective effect, if any, of vegetables and fruits could be mediated by factors biologically independent of their content in fiber or antioxidant vitamins [n13]. Another reason for the diminished interest in vegetables and fruits as potentially important factors in cancer etiology and prevention is that the scientific and public debate on the nutritional etiology of this disease has focused on the possible role of fat and on the most appropriate research strategies for the demonstration of this postulated role [n10,n14,n15]. A similar issue may have hindered the evaluation of potentially different effects of various types of fat on breast cancer risk. Most major studies and meta-analyses have concentrated on total fat as well as on saturated, monounsaturated, and polyunsaturated fat and have found only minor quantitative differences among the fat type-specific measures of association with breast cancer risk [n10,n16,n17]. However, evidence from experimental studies in animals [n18-n21] suggests that consumption of olive oil per se, as contrasted to

that of other fat types, may not enhance the yield of various types of tumors, including breast cancer. Moreover, in Mediterranean countries monounsaturated fats reflect mostly olive oil, whereas in North America they are derived mostly from other foods, in particular meat. In this article, we have analyzed the data from a large case-control study undertaken in Greece, in order to assess the independent associations with breast cancer risk of consumption of vegetables, fruits, other food groups, and types of fat according to their origin (i.e., olive oil, butter, seed oils, and margarine). In previously published papers based on these data, there were no statistically significant, consistent, or substantial associations of breast cancer risk with any of the macronutrients (saturated, monounsaturated, or polyunsaturated fat, carbohydrates, or protein) [n22], although ethanol intake of more than 30 g/d was associated with a small but statistically significant elevation in risk [n23].

Subjects and Methods During a 3-year period (1989-1991), all women with newly diagnosed breast cancer, admitted in four major hospitals in the greater Athens area, were identified. Of the 873 histologically confirmed cases, 820 were eventually included in the study (94% participation rate). Each case patient was interviewed in the hospital before her first discharge. For each case patient, two control subjects were identified: One was from among hospital visitors in the same hospital (not including firstdegree relatives or women who ever had breast cancer); the other was from among orthopedic patients from the major accident hospital of Athens or Piraeus (depending on the residence of the case patient). Each control subject had to be within +/5 years of the age of the index case patient and from the same area of residence. Eight hundred thirty eligible hospital control subjects were identified, of whom 795 (96%) were interviewed; similarly, 808 eligible visitor controls were identified, of whom 753 (93%) were interviewed. Thus, a total of 1548 control subjects from both series were eventually included in the study. All control subjects were interviewed in the hospital by use of the same questionnaire as used for the case patients; every case-control triplet was interviewed by the same specially trained interviewer. Earlier reports

[n22,n23] have presented additional details concerning study design, data collection, comparability of the control series, and analysis with respect to macronutrients and alcohol intake. The interview covered demographic and reproductive information and also included a validated, semiquantitative food-frequency questionnaire [n24]. Based on an analysis using a series of twelve 24hour diet recalls as the standard, the correlation coefficients for data derived from the foodfrequency questionnaire were around .40 for most macronutrients and food groups, including monounsaturated fat and total fruits and vegetables [n24]. All subjects in the present study were asked to indicate the average frequency of consumption, during a period of 1 year before onset of the present disease or before the interview for visitor control subjects, of 115 food items or beverage categories per month, per week, or per day. For analysis, the frequency of consumption of different food items was quantified approximately in terms of the number of times per month the food was consumed [n25]. Thus, daily consumption was multiplied by 30 and weekly consumption by 4, while a value of 0 was assigned to food items rarely or never consumed.

- 2/9 -

[SEE ORIGINAL SOURCE] Food items were considered in groups as recommended by Davidson and Passmore [n26] and used by Manousos et al. [n27] and Trichopoulos et al. [n28]. The individual values for monthly consumption were added, and the sums were approximately distributed into quintiles based on the distribution of the entire study population. (For starchy roots, i.e., potatoes, the distribution was extremely skewed, making quintile distribution impractical and even quartile distribution suboptimal.) Hsieh et al. [n29] have shown that there is virtually no difference in statistical efficiency if quantiles are based on case patients, control subjects, or both groups combined; however, calculation of quantiles using the entire population (case patients and control subjects) respects the usual tenet of statistical analysis that the null hypothesis is presumed true until there is evidence against it. The food groups formed were as follows: cereals; starchy roots (i.e., potatoes); sugars and syrups; pulses, nuts, and seeds; vegetables; fruits; meats, fish, and eggs; milk and milk products; oils and fats; and nonalcoholic beverages (Table 1). Some cooked meals were allocated to two food groups (one half in each); e.g., moussaka was apportioned into both vegetables and the group containing meats, fish, and eggs, and cheese pie was apportioned into cereals and milk products (Table 1). A number of important items, notably olive oil, butter, margarine, and seed oils, were also considered individually. Table 1. Classification into food groups of food items considered in the food-frequency questionnaire

The data were modeled through unconditional logistic regression [n30] using the SPSS statistical package (Statistical Package for Social Sciences, Chicago, Ill.). The effects of several established demographic and reproductive risk factors for breast cancer were evident in the present study [n22]. Thus, a core model was used that included age (years), place of birth (urban or rural), Quetelet index (kg/m <2>), parity (parous or nulliparous), age at first pregnancy (years, among parous women), age at menarche (years), and menopausal status (postmenopausal or premenopausal). Exogenous estrogens are used very infrequently in Greece; thus, exposure to oral contraceptives and/or menopausal estrogens was not included in the core model. Since intake of most food items and food groups is positively correlated with total energy intake, caloric adjustment also was utilized [n5,n6,n8,n22]. The statistical significance of the linear trend in quantiles of consumption of a food group was tested by dividing the regression coefficient by its standard error to generate an X statistic [n30]. In addition, the likelihood ratio test, which is the difference of the maximized log-likelihood statistics, was used to assess the statistical significance of additional covariates in a model [n28,n30]. Because odds ratios (ORs) for reproductive variables, macronutrients, and ethanol intake were very similar when cases were compared with either of the control series [n22,n23], all results presented in this article were derived from comparisons between the cases and the combined control series. All P values reported in this article are two-tailed.

Results Table 2 shows the distribution of case patients and control subjects by frequency of consumption of each major food group. The X values for loglinear trend, after adjustment for age, are also given. These data are not directly interpretable because of confounding by demographic and reproductive breast cancer risk factors as well as by the food groups themselves. Table 3 provides the logistic regression-derived coefficients for the food groups, controlling for the core model variables and total energy intake but not for confounding among food groups. Regression coefficients and ORs are expressed by quintile increase of consumption (quartile for potatoes). Only with respect to consumption of fruits and vegetables were there

statistically significant associations, both being inversely related to breast cancer risk (Table 3). Table 4 explores whether there was deviation from the log-linearity assumed in Table 3 for these two food groups, by examining them as quintile categories. The ORs declined in a clearly monotonic way for consumption of both fruits and vegetables. Table 2. Distribution of 820 case patients and 1548 control subjects by frequency of consumption of various food groups

- 3/9 -

[SEE ORIGINAL SOURCE]

Table 3. Logistic regression-derived coefficients for the food groups, controlling for core model variables and total energy intake [SEE ORIGINAL SOURCE] Table 4. Logistic regression-derived ORs, comparing levels of consumption of fruits and vegetables [SEE ORIGINAL SOURCE] Maintaining control for energy intake and the core demographic and reproductive variables, we further investigated whether there was mutual confounding between consumption of fruits and vegetables. The regression coefficient per quintile increase changed from -.11 (OR = 0.89; P = .004) to -.08 (OR = 0.92; P = .03) for fruits and from -.14 (OR = 0.87; P = .0001) to -.12 (OR = 0.89; P = .0007) for vegetables. It appeared that confounding between these two food groups was limited and that fruits and vegetables were both inversely and independently associated with breast cancer risk. We then added to the model that included the core variables, energy intake, fruits, and vegetables the other eight food groups that, individually taken, were not significantly associated with breast cancer risk (Table 3). This procedure allowed us to evaluate whether these eight food groups collectively were confounding the previously demonstrated associations for consumption of fruits and of vegetables. The difference in the log-likelihood was 13.9, which, with 8 degrees of freedom, does not represent a statistically significant reduction. The regression coefficients for either fruits (b = .08, corresponding to an 8% risk reduction per quintile increase) or vegetables (b = -.13, corresponding to a 12% risk reduction per quintile increase) remained virtually unchanged. Furthermore, the addition of any one of the eight food groups did not lead to a statistically significant reduction in deviance. Introduction of total ethanol consumption in the model also had no discernible effect (data not shown); although most women in the study occasionally consumed alcoholic beverages, only 12% drank more than one glass of alcohol per day [n23]. Moreover, alternative and simultaneous inclusion of olive oil, butter, margarine, and seed oils did not appreciably alter the regression coefficients for either fruits or vegetables. We conclude that the inverse associations for consumption of fruits and vegetables are neither mutually confounded nor confounded by the collective influence of the

other eight food groups or by intake of various types of fat. Table 5 presents the frequency distribution of case patients and control subjects by reported consumption of the main specific types of oils and fats represented in the Greek diet. Since olive oil is abundantly consumed in Greece, only three categories could be constructed: less than once per day, once per day, and more than once per day. In contrast, for consumption of butter, margarine, and seed oils, uniform categorization was used. Table 5. Distribution of 820 case patients and 1548 control subjects by frequency of consumption of specific oils and fats [SEE ORIGINAL SOURCE] Table 6 shows logistic regression-derived coefficients for the specific oil or fat, controlling for the core model variables, total energy intake, and consumption of fruits and vegetables (but not for confounding among the specific oils and fats themselves). The regression coefficients and ORs are expressed per increase of one consumption level (increment of four times per month), except for olive oil where two indicator variables were constructed. There was evidence that frequent consumption of olive oil was associated with reduced risk of breast cancer (OR = 0.75 [95% confidence interval = 0.57-0.98] for more than once per day versus once per day), whereas increasing margarine consumption was associated with increasing risk (OR = 1.05 [95% confidence interval = 1.001.10] for an increment of four times per month) (Table 6). Simultaneous inclusion of all four specific types of fat into the model did not reveal appreciable confounding in the estimates for these oils and fats. Thus, the regression coefficient for consumption of olive oil more frequently than once per day changed from -.286 (P = .04) to -.274 (P = .05) and for margarine consumption from .046 to .047 (P = .05 in both instances). Residual confounding of the apparent effects of olive oil and margarine by vegetables or fruits is unlikely, since the nonparametric correlation coefficient between vegetables or fruits on the one hand and any of the fat types on the other was no higher than + .12. Table 6. Logistic regression-derived coefficients for specific oils and fats, controlling for core model variables, total energy intake, and consumption of fruits and vegetables

- 4/9 -

[SEE ORIGINAL SOURCE] We also tried to assess whether any of the associations reported above were more striking in premenopausal women or in postmenopausal women. Among the study subjects, 270 case patients and 505 control subjects were premenopausal, whereas 550 case patients and 1041 con-

trol subjects were postmenopausal. No statistically significant effect modification of the risk of breast cancer by menopausal status was noted with respect to fruits, vegetables, or types of fat: P values for interaction in all instances were >/.10, although the protective effect of olive oil consumption was apparently concentrated among postmenopausal women (P = .10).

Discussion This analysis is based on data from one of the largest case-control studies that have examined the role of diet in the etiology of breast cancer. The high rates of response among case patients and control subjects alike indicate that overt selection bias has not occurred. Moreover, the OR estimates for the established breast cancer risk factors as well as for macronutrients [n22] and ethanol intake [n23] were similar with the use of either control series. The interviewers were unblinded as to the subject's case or control status, but the dietary questionnaire responses were precoded, requiring little, if any, probing by the interviewer. Moreover, an hypothesis linking types of fat, rather than total fat, to breast cancer has not been widely entertained. There was no evidence in terms of weight loss or otherwise to suggest that women in the study groups changed their diet subsequent to their diseases. The results were essentially identical when women with fractures [43% of the hospital controls [n22]] were excluded to control for a possible association between osteoporosis and diet among the orthopedic patients. Adjustment for energy intake controls for systematic overreporting of cases [n7], unless there is selective systematic overreporting or underreporting with respect to particular foods or food groups; however, the latter situation seems unlikely, since olive oil, margarine, fruits, or vegetables are not generally perceived by the public as having a role in breast cancer. Fruits and vegetables have low nutrient densities, and control for energy intake may be less efficient for these groups than for energy-dense food groups [n31]. However, there was no evidence of underreporting by case patients in this study, and overreporting by women with breast cancer could have only attenuated a genuinely stronger, inverse association between consumption of vegetables and fruits and the disease. Recommendations to reduce total fat intake in order to reduce the incidence of breast cancer have been made in Greece, but they have not emphasized different types of fat, making it unlikely that there would be a fat type-specific dif-

ferential recollection by case patients and control subjects. Lastly, adjustment for socioeconomic and reproductive risk factors as well as for intercorrelations among food groups controls for the mutual confounding effects of the study variables, although in any case-control study, concerns about residual confounding and unidentifiable selection bias cannot be eliminated. The biologic plausibility of a protective effect of vegetables and fruits against breast cancer is strong; Steinmetz and Potter [n13] examined more than a dozen categories of substances in vegetables and fruits that may inhibit the process of carcinogenesis at various stages. The epidemiologic evidence also appears supportive [n3,n4,32], although earlier studies [n33-n35] either were small or did not use a dietary questionnaire sufficiently comprehensive to allow control for energy intake and mutual dietary confounding. Other studies [n9,n11,n36-n40] that have adjusted for energy intake have reported inverse associations between consumption of vegetables and fruits and breast cancer risk, although the authors of these studies have invoked different physiologic processes and effector compounds to explain the empirical associations. The different effects of fat types on the promotion of chemically induced mammary cancer have been studied mostly in experimental animals, particularly in female Sprague-Dawley rats [n2,n20,n21]. The experimental evidence appears to suggest that a minimum of essential fatty acids (about 4% of energy intake as linoleic acid) is required before saturated fat can display its full promoting potential at 20% of total energy intake. It is interesting that, in these studies [n2,n18,n20,n21,n41], olive oil either was a poor promoter or had a protective effect. The epidemiologic data with respect to olive oil consumption and breast cancer appear compatible with the animal evidence. Ecologic analyses do not suggest a positive correlation with breast cancer incidence or mortality for olive oil consumption

- 5/9 -

[n42], as they seem to do for consumption of saturated fat [n14]. Indeed, Greek women with 42% of energy intake from fat, mostly from olive oil [n43,n44), have substantially lower mortality from breast cancer than U.S. women whose energy intake from fat is at present around 35% [n45]. Results of analytic epidemiologic studies are relevant, in this context, only if they refer to olive oil per se or to monounsaturated fat in populations who consume large amounts of olive oil. Studies examining the association of monounsaturated fat consumption with breast cancer in the United States [n10], Finland [n46], and China [n47] may be perfectly valid, but they do not address the olive oil issue. (In the United States, most monounsaturated fat is derived from meat.) Studies in Italy, Spain, and Greece are better suited to examine the role of olive oil, since it is in these countries, together with Libya, that olive oil is consumed in substantial quantities, particularly among the less Westernized segments of the population [n21,n43]. To our knowledge, two case-control studies, one in Spain [n48] and the other in Italy [n49], as well as the present Greek study are the only ones that meet the criteria of examining the role of olive oil per se in olive oil-consuming countries, of being reasonably large, and of being analyzed with procedures allowing control for energy intake and mutual confounding among food groups. The Spanish study [n48] and the Greek study have generated very similar results, suggesting a protective effect of olive oil; the Italian study [n49] found a strong positive association between saturated fat intake and breast cancer risk, but no association with respect to monounsaturated fat intake. The importance of adjustment for energy intake cannot be overemphasized. In a study in France [n50], a positive, but not statistically significant, association was found between breast cancer and consumption of olive oil. However, in that study, total food consumption was positively, significantly, and strongly related to breast cancer risk, indicating that general overreporting by case patients was a distinct possibility.

Previous studies have not systematically examined the role of margarine consumption in the etiology of breast cancer. However, five case-control studies [n51,n55] that were identified in a literature search as having considered margarine consumption all reported positive associations, although these associations were not always statistically significant and were not always substantiated by optimal procedures. It would be premature to speculate in biologic terms what role margarine consumption could play in the etiology of breast cancer. It is possible that the positive association observed in this study is due to chance or mirrors the inverse association noted for olive oil intake. Alternatively, margarine may have distinct biologic properties that have not been fully characterized in experimental research. The priority at this stage is to attempt to replicate this finding in a different population. There is some evidence that premenopausal and postmenopausal women may have breast cancers of different etiologic components [n56], and indeed the international variation in the incidence of this disease is larger among postmenopausal women [n57]. Several studies [n58-n60] have focused on the role of diet in the occurrence of breast cancer among postmenopausal women, although the data remain inconclusive. In the present study, there was no suggestion of an interaction between menopausal status and consumption of vegetables, fruits, or margarine in relation to breast cancer risk. However, the apparent protective effect of olive oil consumption was concentrated among postmenopausal women, although the interaction term was not statistically significant (P = .10). In the absence of an a priori hypothesis focusing on postmenopausal women, it would not be prudent at this time to interpret this apparent effect modification in biologic terms. We conclude that consumption of vegetables and fruits is inversely, significantly, and strongly associated with breast cancer risk and that olive oil consumption may also reduce this risk, whereas margarine intake appears to be associated with an elevated risk for the disease.

SUPPLEMENTARY INFORMATION: <1>Affiliations of authors: A. Trichopoulou, Department of Nutrition and Biochemistry, Athens School of Public Health, Greece; and Department of Nutrition, Harvard School of Public Health, Boston, Mass. K. Katsouyanni, Department of Hygiene and Epidemiology, University of Athens Medical School, Goudi, Greece; and Department of Epidemiology and Center for Cancer Prevention, Harvard School of Public Health.

- 6/9 -

S. Stuver, D. Trichopoulos, Department of Epidemiology and Center for Cancer Prevention, Harvard School of Public Health. L. Tzala, Department of Hygiene and Epidemiology, University of Athens Medical School. C. Gnardellis, Department of Nutrition and Biochemistry, Athens School of Public Health. E. Rimm, Department of Nutrition, Department of Epidemiology, and Center for Cancer Prevention, Harvard School of Public Health. Correspondence to: Dimitrios Trichopoulos, M.D., Department of Epidemiology, Harvard School of Public Health, 677 Huntington Ave., Boston, MA 02115. Supported by grants from the Europe Against Cancer Program of the European Community and from the Central Scientific Health Council of the Greek Ministry of Health (KE UPSILON Y). S. Stuver was supported by Public Health Service National Research Service Award in Cancer Epidemiology 5T32CA09001 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services. The collaboration of many cancer surgeons and physicians in patient accrual is gratefully acknowledged. Manuscript received April 1, 1994; revised October 12, 1994; accepted October 24, 1994.

REFERENCES: 1. [n1.] Committee on Diet, Nutrition, and Cancer, National Academy of Sciences: 2. Diet, Nutrition, and Cancer. Washington, DC: National Academy Press, 1982 3. [n2.] Committee on Diet and Health: National Research Council (US): Diet and Health. Implications for Reducing Chronic Disease Risk. Washington, DC: National Academy Press, 1989 4. [n3.] Steinmetz KA, Potter JD: Vegetables, fruit, and cancer. I. Epidemiology. Cancer Causes Control 2:325-357, 1991 5. [n4.] Block G, Patterson B, Subar A: Fruit, vegetables, and cancer prevention: a review of the epidemiological evidence. Nutr Cancer 18:1-29, 1992 6. [n5] Willett W, Stampfer MJ: Total energy intake: implications for epidemiologic analyses [see comment citations in Medline]. Am J Epidemiol 124:17-27, 1986 7. [n6.] Willett W: Nutritional epidemiology. In Monographs in Epidemiology and Biostatistics, vol 15. New York: Oxford Univ Press, 1990 8. [n7.] Trichopoulos D, Tzonou A, Katsouyanni K, et al: Diet and cancer: the role of case-control studies. Ann Nutr Metab 35 suppl 1:89-92, 1991 9. [n8] Palmgren J, Kushi LH: Re: Total energy intake: implications for epidemiologic analyses [letters: comment]. Am J Epidemiol 133:1291-1293, 1991 10. [n9.] Graham S, Hellmann R, Marshall J, et al: Nutritional epidemiology of postmenopausal breast cancer in western New York [see comment citation in Medline]. Am J Epidemiol 134:552566, 1991 11. [n10.] Willett WC, Hunter DJ, Stampfer MJ, et al: Dietary fat and fiber in relation to risk of breast cancer. An 8-year follow-up [see comment citations in Medline]. JAMA 268:2037-2044, 1992 12. [n11.] Rohan TE, Howe GR, Friedenreich CM, et al: Dietary fiber, vitamins A, C, and E, and risk of breast cancer: a cohort study. Cancer Causes Control 4:29-37, 1993 13. [n12.] Hunter DJ, Manson JE, Colditz GA, et al: A prospective study of the intake of vitamins C, E, and A and the risk of breast cancer [see comment citation in Medline]. N Engl J Med 329:234240, 1993 14. [n13.] Steinmetz KA, Potter JD: Vegetables, fruit, and cancer. II. Mechanisms. Cancer Causes Control 2:427-442, 1991 15. [n14.] Prentice RL, Sheppard L: Dietary fat and cancer: consistency of the epidemiologic data, and disease prevention that may follow from a practical reduction in fat consumption [published erratum appears in Cancer Causes Control 1:253, 1990]. Cancer Causes Control 1:81-97, 1990 16. [n15.] Michels KB, Willett WC: The women's health initiative: daughter of politics or science? In Cancer Prevention (DeVita VT Jr, Hellman S, Rosenberg SA, eds). Philadelphia: Lippincott, 1991, pp 1-11

- 7/9 -

17. [n16.] Howe GR, Hirohata T, Hislop TG, et al: Dietary factors and risk of breast cancer: combined analysis of 12 case-control studies [see comment citation in Medline]. J Natl Cancer Inst 82:561569, 1990 18. [n17.] Boyd NF, Martin LJ, Noffel M, et al: A meta-analysis of studies of dietary fat and breast cancer risk. Br J Cancer 68:627-636, 1993 19. [n18.] Cohen LA, Thompson DO, Choi K, et al: Dietary fat and mammary cancer. II. Modulation of serum and tumor lipid composition and tumor prostaglandins by different dietary fats: association with tumor incidence patterns. J Natl Cancer Inst 77:43-51, 1986 20. [n19.] Cohen LA: Fat and endocrine-responsive cancer in animals. Prev Med 16:468-474, 1987 21. [n20.] Weisburger JH, Wynder EL: Dietary fat intake and cancer. Hematol Oncol Clin North Am 5:7-23, 1991 22. [n21.] Weisburger JH: Mechanisms of macronutrient carcinogenesis. In Macronutrients. Investigating Their Role in Cancer (Micozzi MS, Moon TE, eds). New York: Marcel Dekker, 1992, pp 331 23. [n22.] Katsouyanni K, Trichopoulou A, Stuver S, et al: The association of fat and other macronutrients with breast cancer: a case-control study from Greece. Br J Cancer 70:537-541, 1994 24. [n23.] Katsouyanni K, Trichopoulou A, Stuver S, et al: Ethanol and breast cancer: an association that may be both confounded and causal. Int J Cancer 58:356-361, 1994 25. [n24.] Gnardellis C, Trichopoulou A, Katsouyanni K, et al: Reproducibility and validity of an extensive semi-quantitative food frequency questionnaire among Greek school teachers. Epidemiology. In press 26. [n25.] Katsouyanni K, Skalkidis Y, Petridou E, et al: Diet and peripheral arterial occlusive disease: the role of poly-, mono-, and saturated fatty acids. Am J Epidemiol 133:24-31, 1991 27. [n26.] Davidson SS, Passmore R: Human Nutrition and Dietetics. Edinburgh: Churchill Livingstone, 1979 28. [n27.] Manousos O, Day NE, Trichopoulos D, et al: Diet and colorectal cancer: a case-control study in Greece. Int J Cancer 32:1-5, 1983 29. [n28.] Trichopoulos D, Ouranos G, Day NE, et al: Diet and cancer of the stomach: a case-control study in Greece. Int J Cancer 36:291-297, 1985 30. [n29.] Hsieh CC, Maisonneuve P, Boyle P, et al: Analysis of quantitative data by quantiles in epidemiologic studies: classification according to cases, non-cases, or all subjects? Epidemiology 2:137-140, 1991 31. [n30.] Breslow NE, Day NE: Statistical methods in cancer research. Volume 1 -- The analysis of case-control studies. IARC Sci Publ 32:5-338, 1980 32. [n31.] Serdula M, Byers T, Coates R, et al: Assessing consumption of high-fat foods: the effect of grouping foods into single questions. Epidemiology 3:503-508, 1992 33. [n32.] Lanza E, Shankar S, Trock B: Dietary fiber. In Macronutrients. Investigating Their Role in Cancer (Micozzi MS, Moon TE, eds). New York: Marcel Dekker, 1992, pp 293-319 34. [n33.] Katsouyanni K, Trichopoulos D, Boyle P, et al: Diet and breast cancer: a case-control study in Greece. Int J Cancer 38:815-820, 1986 35. [n34.] La Vecchia C, Decarli A, Franceschi S, et al: Dietary factors and the risk of breast cancer. Nutr Cancer 10:205-214, 1987 36. [n35.] Hislop TG, Kan L, Coldman AJ, et al: Influence of estrogen receptor status on dietary risk factors for breast cancer. Can Med Assoc J 138:424-430, 1988 37. [n36.] Iscovich JM, Iscovich RB, Howe G, et al: A case-control study of diet and breast cancer in Argentina. Int J Cancer 44:770-776, 1989 38. [n37.] Van 't Veer P, Kolb CM, Verhoef P, et al: Dietary fiber, beta-carotene and breast cancer: results from a case-control study. Int J Cancer 45:825-828, 1990 39. [n38.] Zaridze D, Lifanova Y, Maximovitch D, et al: Diet, alcohol consumption and reproductive factors in a case-control study of breast cancer in Moscow. Int J Cancer 48:493-501, 1991 40. [n39] Levi F, La Vecchia C, Gulie C, et al: Dietary factors and breast cancer risk in Vaud, Switzerland. Nutr Cancer 19:327-335, 1993 41. [n40.] Baghurst PA, Rohan TE: High-fiber diets and reduced risk of breast cancer. Int J Cancer 56:173-176, 1994

- 8/9 -

42. [n41.] Cohen LA, Choi K, Weisburger JH, et al: Effect of varying proportions of dietary fat on the development of N-nitrosomethylurea-induced rat mammary tumors. Anticancer Res 6:215-218, 1986 43. [n42.] Rose DP, Boyar AP, Wynder EL: International comparisons of mortality rates for cancer of the breast, ovary, prostate, and colon, and per capita food consumption. Cancer 58:2363-2371, 1986 44. [n43.] Gerber M: Olive oil and cancer. In The Mediterranean Diet and Cancer Prevention (Giacosa A, Hill MJ, eds). Proceedings of a workshop organized by the European Cancer Prevention Organization (ECP) and the Italian League Against Cancer, Cosenza, Italy, June 28-30, 1991. Andover, England: European Cancer Prevention Organization, 1991, pp 127-139 45. [n44.] Trichopoulou A, Toupadaki N, Tzonou A, et al: The macronutrient composition of the Greek diet: estimates derived from six case-control studies. Eur J Clin Nutr 47:549-558, 1993 46. [n45.] Lenfant C, Ernst N: Daily dietary fat and total food-energy intakes -- Third National Health and Nutrition Examination Survey, Phase 1, 1988-91. MMWR 43:116-117, 123-125, 1994 47. [n46.] Knekt P, Albanes D, Seppanen R, et al: Dietary fat and risk of breast cancer. Am J Clin Nutr 52:903-908, 1990 48. [n47.] Yu SZ, Lu RF, Xu DD, et al: A case-control study of dietary and nondietary risk factors for breast cancer in Shanghai. Cancer Res 50:5017-5021, 1990 49. [n48.] Martin-Moreno JM, Willett WC, Gorgojo L, et al: Dietary fat, olive oil intake and breast cancer risk. Int J Cancer 58:774-780, 1994 50. [n49.] Toniolo P, Riboli E, Protta F, et al: Calorie-providing nutrients and risk of breast cancer [see comment citation in Medline]. J Natl Cancer Inst 81:278-286, 1989 51. [n50.] Richardson S, Gerber M, Cenee S: The role of fat, animal protein and some vitamin consumption in breast cancer: a case-control study in southern France. Int J Cancer 48:1-9, 1991 52. [n51.] Nomura A, Henderson BE, Lee J: Breast cancer and diet among the Japanese in Hawaii. Am J Clin Nutr 31:2020-2025, 1978 53. [n52.] Lubin JH, Burns PE, Blot WJ, et al: Dietary factors and breast cancer risk. Int J Cancer 28:685-689, 1981 54. [n53] Simard A, Vobecky J, Vobecky JS: Nutrition and lifestyle factors in fibrocystic disease and cancer of the breast. Cancer Detect Prev 14:567-572, 1990 55. [n54.] Kikuchi S, Okamoto N, Suzuki T, et al: A case-control study of breast cancer, mammary cyst and dietary, drinking or smoking habit in Japan (Japanese). Gan No Rinsho Spec No.:365369, 1990 56. [n55.] D'Avanzo B, Negri E, Gramenzi A, et al: Fats in seasoning and breast cancer risk: an Italian case-control study. Eur J Cancer 27:420-423, 1991 57. [n56.] Newcomb PA, Storer BE, Longnecker MP, et al: Lactation and a reduced risk of premenopausal breast cancer [see comment citations in Medline]. N Engl J Med 330:81-87, 1994 58. [n57.] Kelsey JL, Horn-Ross PL: Breast cancer: magnitude of the problem and descriptive epidemiology. Epidemiol Rev 15:7-16, 1993 59. [n58.] Boyar AP, Rose DP, Loughridge JR, et al: Response to a diet low in total fat in women with postmenopausal breast cancer: a pilot study. Nutr Cancer 11:93-99, 1988 60. [n59.] Prentice R, Thompson D, Clifford C, et al: Dietary fat reduction and plasma estradiol concentration in healthy postmenopausal women. The Women's Health Trial Study Group. J Natl Cancer Inst 82:129-134, 1990 61. [n60.] Kushi LH, Sellers TA, Potter JD, et al: Dietary fat and postmenopausal breast cancer [see comment citation in Medline]. J Natl Cancer Inst 84:1092-1099, 1992

- 9/9 -