3575

Dietary Factors and the Risk of Endometrial Cancer

Fabio Levi, M.D.," Silvia Franceschi, M.D.,t Eva Negri, Sc.D.,S and Carlo La Vecchia, M.D.*,$

Background. Endometrial cancer is associated with overweight, but little is known on its possible relationship with specific aspects of diet. Methods. The relationship between dietary factors and the risk of endometrial cancer was investigated in a case-control study conducted in Switzerland and Northern Italy on 274 patients with histologically confirmed endometrial cancers and 572 control subjects admitted to the hospital for acute nongynecologic disorders that were not hormone related, metabolic, or neoplastic. Results. Significant direct associations were observed with (1)the total energy intake (odds ratio [OR] for the highest versus the lowest consumption tertile = 2.7) and, after allowance for energy intake, (2)the frequency of consumption of most types of meats, eggs, beans or peas, added fats (OR for total added fat = 2.5), and sugar (OR= 2.5). Significant protection, of the order of 40-60% reduction in the highest versus the lowest consumption tertile, was conferred by elevated intake of most vegetables and fresh fruit and whole grain bread and pasta. This was reflected in the low OR for the highest tertiles of intake of beta-carotene and ascorbic acid (OR for the highest versus the lowest consumption tertile after allowance for energy intake = 0.5). Conclusions. The current study suggests that, aside from the predictable adverse effects of overeating and consequent overweight, some qualitative aspects of the habitual diet may also be associated with the risk of endometrial cancer, chiefly, the intake of animal proteins

and fat (directly) and of fresh fruit, vegetables, and fibers (inversely). Cancer 1993;719575-81. Key words: case-control studies, endometrial cancer, diet, nutrition.

From the *Institut Universitaire de Mkdecine Sociale et PrC.ventive, Lausanne, Switzerland; the tServizio di Epidemiologia, Centro di Riferimento Oncologico, Aviano, Italy; and the Slstituto di Ricerche Farmacologiche "Mario Negri," Milan, Italy. Supported in part by the Swiss National Science Foundation grant 3.866-0.88 and by the Italian National Research Council Applied Projects "Clinical Applications of Oncological Research" and "Prevention and Control of Disease Factors" (contract number 91.00285.9541). The contributions of the Swiss and Italian Leagues Against Cancer, of the Italian Association for Cancer Research, of the Europe Against Cancer Program, and of the Commissions of the European Communities are gratefully acknowledged. Address for reprints: Fabio Levi, Registre Vaudois des Tumeurs, Institut Universitaire de MCdecine Sociale et PrCventive, CHUV Falaises 1, 1011 Lausanne, Switzerland. Accepted for publication January 14, 1993.

Endometrial cancer is strongly related to overweight and, consequently, to dietary habits.',' However, much of the knowledge linking endometrial cancer to diet comes from epidemiologic observations. Strong positive correlations were observed between endometrial cancer incidence and the national disappearance statistics for meat, eggs, milk, proteins, fats and oil, and total caloric intake.3,4However, the issue is clearly complicated by the high correlation between proteins and, particularly, fat and total energy intake, which, in turn, is the principal determinant of ~ b e s i t y . ~ Few analytic studies have directly addressed the role of diet as a cause of endometrial cancer. Preliminary results from a case-control study conducted in the 1970s and based on 24-hour recall suggested that the intake of carbohydrates and total calories (but not of proteins or fat) was higher in cases than in control^.^ A case-control study from Italy found direct associations with subjective scores of fat and oil intake and significant protection by green vegetables and fresh fruit.6 However, information only on a small number of indicator foods was available in that study, and it was not, therefore, possible to estimate (and allow for) the total caloric intake.6 Some of these associations are biologically plausible. Besides the strong relationship between fat, total caloric intake,'*3and obesity (and hence levels of estrogens in postmen~pause),~ has been shown, for init stance, that the levels of estriol, total estrogens, and prolactin were lower and those of sex hormone-binding globulins were higher in postmenopausal vegetarian women.'?These differences were not explicable in terms of differences in body weight only. Such a relationship with estrogen levels, which are the best defined determinant of endometrial cancer risk, may therefore help identify and explain low-risk dietary patterns.

3576

CANCER June 7, 2993, Volume 71, No. 11

Thus, there is ample scope for further research on the question of diet and endometrial cancer, particularly in consideration of the fact that, although the risk factors for cancer of the corpus uteri are better defined than those of other female hormone-related neoplasms, they cannot by themselves explain the 30-fold difference in incidence observed between various registration

area^.^
Using data from a cooperative case-control study conducted in Switzerland and Italy, we decided, therefore, to reconsider the role of a number of selected indicator foods and nutrients as causes of endometrial cancer.

menstrual and reproductive variables, and use of female hormone preparations. The dietary section of the questionnaire considered 50 indicator foods, including the major sources of energy, starches, fats, fibers, and vitamins A and C in the study areas. The weekly frequencies of intake of each food item before the occurrence of symptoms of the disease were elicited. The categories of weekly frequency of consumption were not defined a priori but were collected as reported by the interviewees.

Data Analysis and Control of Confounding

The frequencies of consumption of various foods, beverages, and nutrients were subdivided into three levels including, as far as possible, equal numbers of cases and controls combined (approximate tertiles, although, because of discrete frequencies, there were uneven distributions for some items). In a few instances, on account of the very limited numbers, intermediate and high tertiles were grouped together. A self-reported subjective score was used to assess the level of intake of whole grain bread and pasta, added fats (i.e., butter, margarine, olive oil, and seed oils), onions, garlic, and salt. The total energy intake was computed approximately by multiplying the average caloric content in kilocalories' 1,12 of one serving of each food or beverage (including alcoholic beverages) by its reported frequency of intake. Indexes of retinol, beta-carotene (provitamin A), and ascorbic acid (vitamin C) intake were derived using nutritional values issued by Italian or English Odds ratios (OR) and the corresponding 95% conTable 1. Distribution of 274 Patients With Endometrial Cancer and 572 Control Subjects According to Study Center, Age, Education, and Body Mass Index
Patients with endometrial cancer
Study center Vaud, Switzerland North Italy Age group (yr) < 45 45-54 55-64 65-74 Education (yr) <8 9-13 2 14 Body mass index (kg/m2) < 20 20-24 25-29 2 30 138 (50.4%) 136 (49.6%) 16 (5.8%) 51 (18.6%) 112 (40.9%) 95 (34.7%) 130 (47.4%) 128 (46.7%) 16 (5.8%) 17 (6.2%) 112 (40.9%) 93 (33.9%) 52 (19.0%)

Subjects and Methods

Study Design
Since January 1988, we have been conducting a cooperative case-control study of endometrial cancer in the Swiss Canton of Vaud and in metropolitan Milan and the Pordenone Province, Friuli-Venezia Giulia Region, Northern Italy. The general design of this investigation has already been described." Briefly, in Vaud, the recruitment of cases was population based because identified cases were cross checked with incident cases reported to the local cancer registry. Overall, more than 80% of identified cases were interviewed. In Milan and Pordenone, case recruitment was hospital based because these areas are not covered by cancer registration schemes. Overall, 274 patients with histologically confirmed endometrial cancers were interviewed. The age range was 31-75 years (median age, 61 years). The control subjects were women aged 75 years or less, who were admitted to the same networks of hospitals in which the cases had been identified and who had primary diagnoses unrelated to any of the known or suspected risk factors for endometrial cancer or to any long-term modification in diet. A specific exclusion was made for women admitted for gynecologic, hormonal, metabolic, or neoplastic diagnoses, or who had undergone hysterectomy. A total of 572 women, aged 30-75 years (median age, 59 years) were interviewed. Of these, 32% were admitted for traumas, 16% had nontraumatic orthopedic diseases, 26% had surgical conditions, and 26% had other miscellaneous disorders, including acute medical conditions and eye, nose, throat, or dental diseases. All interviews were conducted in the hospital. Less than 10% of the subjects approached for interviews refused. The same structured questionnaire was used, including information on sociodemographic factors, personal characteristics and habits (such as smoking and physical exercise), a problem-oriented medical history,

Control subjects
406 (71.0%) 166 (29.0%) 65 (11.4%) 128 (22.4%) 187 (32.7%) 192 (33.5%) 215 (37.6%) 283 (49.5%) 74 (12.9%) 60 (10.5%) 253 (44.2%) 190 (33.2%) 69 (12.1%)

Table 2. Odds Ratios of Endometrial Cancer According to Approximate Frequency Consumption Tertiles of Various Foods and Beverages (1988-1991)
Frequency consumption tertile* No. of cases: no. of controls Food or beverage
Pasta Rice Polenta White bread Whole grain bread and pasta Biscuits Pastry Beef Pork Poultry Fish Liver Raw ham Boiled ham Salami and sausages Canned meat Other meats Milk Cheese
1 (low)

Odds ratiot
I+
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

2 (intermediate)
103:177 96:235 119:208 84:211 80:192 79:187 83:255 84:199 91:174 118:222 112:245 110:217 73:153 89:179 117:207 53:51 129:262 133:242 92:206 90:195 101:220 99:227 162:335 99:256 1012 2 3 88:240 109:225 78:162 129:226 123:230 130:308 54:218 115:253 117:220 87:165 88:145 69:144 114:183 64:142 111:229 150:370 191:404 122:276 79:160 88:168 51:96 127:230 20:33 91:225 18:40 40:71 133:345

3 (hixh) 79:104 102:153 46:93 103:1801 69:233 79:126 75:118 126:174 121:228 82:164 83:177 49:81 119:181 97:118 99:213
-

2 1.61 1.05 1.25 1.oo 0.62 1.16 0.76 1.34 1.79 1.33 0.92 1.40 1.41 1.49 1.80 1.83 1.96 1.44 1.03 1.46 1.47 1.61 0.90 0.78 0.66 0.70 0.81 1.04 0.86 1.37 0.64 0.67 0.89 1.16 0.89 1.19 0.77 2.12 1.13 0.56 1.13 1.33 1.23 1.72 1.18 1.26 1.84 1.71 1.03 0.99 1.46 1.14

3 1.45 1.18 1.18 2.19 0.46 1.69 1.60 2.26 2.16 0.87 0.93 1.37 1.29 2.13 1.62 2.15 1.62 1.36 2.58 2.06 2.35 0.64 1.20 0.63 1.00 0.92 0.97 0.46 1.58 0.70 0.71 0.62 0.78 0.63 0.93 0.63 3.71 2.90 0.82 4.71 1.67 2.17 2.49 1.22 1.70 2.43 5.24 2.07

Chisquare (tiend)
3.18 0.63 1.15 12.845 14.835 7.145 3.9211 19.219 14.105 0.32 0.13 3.40 1.77 16.035 4.3011 7.085 10.085 5.9611 3.04 23.008 9.345 l5.09 4.14 0.43 3.9711 0.71 0.11 0.02 11.175 5.1611 4.2311 4.7511 5.4611 0.42 6.5511 0.36 6.695 29.785 0.73 3.71 8.165 3.55 6.675 22.075 1.11 1.32 13.02 1.76 5.6711 0.27 6.2711

Eggs Potatoes Beans and peas Vegetables (total) Cruciferous vegetables Carrots Spinach Tomatoes Green peppers Artichokes Salad Onions Garlic Fresh fruit (total) Apples Pears Citrus fruits Melons Butter Margarine Olive oil Seed oils Salt Pepper Sugar Coffee Decaffeinated coffee Tea Cola-containing beverages Wine Beer Liquor Meal freauencv

92:291 76:184 113:271 86:181 125:147 116:259 116:199 64:199 62:170 74: 186 79:190 115:274 82:238 88:275 58:152 220:521 41:134 85:221 58:132 90:238 84:162 56:204 42:63 109:175 7335 145:254 74:128 101:220 98:199 117:283 101:161 150:180 61:111 120:241 98:178 68:116 111:161 136:368 208:429 140:306 101:186 45:117 121:246 68:256 81:176 223:476 147:342 254:539 105353 253:529 229:498 113:184

103:176 56:109 124:234 94:139 89:190 119:141 70: 174 70:141 100:264 41:78 91:219 95:190 47:147 34:50 43:103 30:74 98:208 37:111 89:129 118:311 94:267 24:21 2:l 23:37 23: 16 38:51 31:50 127:156 105:228
-

78:94 3:3 5:3 29:43

* For some items, the sum of the strata does not add up to the total because of missing values

t Estimates from multiple logistic regression equations, including terms for study center and age.
$ Reference category. g P < 0.01. )I P < 0.05.

3578

CANCER June 1, 2993, Volume 71, No. 11

fidence intervals of endometrial cancer were derived from multiple logistic-regression equations, with maximum-likelihood Several models were fitted, starting with the simplest model, which included the study center and age, to the more complex ones, including simultaneously all potential nondietary confounding factors (i.e., education, body mass index, parity, estrogen replacement therapy, and level of physical activity) in addition to total energy intake.15,16 Because the most (and only) substantial modification in OR was derived from inclusion of total energy intake, the results of the equations including age and study center, with or without total energy intake, were chosen for presentation. The dose gradients in the logistic models were based on the chi-square values for trend, computed as the difference between the deviance of the models with and without the variable of interest.
Results

Table 1 gives the distribution of endometrial cancer cases and control subjects by study center, age, educa-

tion, and body mass index. It shows a significant excess of overweight and obese women in cases compared with controls (x: for trend, adjusted for age and study center, 7.52; P < 0.01). The distribution of cases and controls by approximate consumption tertiles and the corresponding OR for investigated foods and beverages are given in Table 2. Cases tended to consume more frequently the following: white bread (OR for highest versus lowest tertile = 2.2), biscuits (OR = 1.7), pastry (OR = 1.6), and a number of different types of meat, such as beef (OR = 2.3), pork (OR = 2.2), boiled ham (OR = 2.1), salami and sausages (OR = 1.6), and other meats (OR = 2.1). Also frequent consumption of milk (OR = 1.6), eggs (OR = 2.6), potatoes (OR = 2.1), beans or peas (OR = 2.4), salads (OR = 1.6), several types of added fats (chiefly butter; OR = 3.7) and seed oils (OR = 4.7), sugar (OR = 2.5), and tea (OR = 1.8)seemed to be associated with a significantly increased risk of endometrial cancer. Conversely, whole grain bread and pasta (OR for highest versus lowest tertile = 0.5); total vegetables (OR = 0.6), especially carrots (OR = 0.6), and artichokes

Table 3. Foods and Beverages Significantly Associated With the Risk of Endometrial Cancer After Allowance for Total Energy Intake (1988-1991)
Odds ratio* Frequency consumption tertile Food or beverage
Direct Beef Pork Boiled ham Canned meat Other meats

I t (low)
1 1
1

2 (intermediate)

3 (high)

Chi-square (trend)
12.424 6.525 12.34$ 5.165 5.575 13.523 9.103 23.68$ 5.445 13.123 10.28$ 16.483 15.10$ 8.133 18.103 6.86$ 7.34$ 13.53$ 13.323 13.64i

Eggs Beans and peas Butter Seed oils Sugar Tea Inverse Whole grain bread and pasta Vegetables (total) Carrots Artichokes Onions Garlic Fresh fruit (total) Pears Melon

1 1 1 1 1 1
1 1

1.22 1.60 1.45 1.69 1.80 1.30 1.43 1.98 1.06 1.55 1.73 0.68 0.67 0.58 0.78 0.63 0.63 0.85 0.79 0.70

1.95 1.74 1.96

-

1.83 2.13 1.98 3.26 3.97 2.07

-

1 1 1 1 1 1 1 1 1

0.43 0.38 0.51 0.34 0.59 0.61 0.45 0.50 0.50

* Estimates from multiple logistic regression equations, including terms for age, study center. and total energy intake.

t Reference category.
$ P < 0.01. 5 P < 0.05.

Dietary Factors and the Risk of Endometrial CA/Levi et al.

3579

(OR = 0.5); and fresh fruit (OR = 0.5), especially pears (OR = 0.6) and melons (OR = 0.6), exerted a beneficial influence. Finally, elevated OR were seen for the highest consumption tertile of wine (OR for highest versus lowest tertile = 1.7), and liquors (OR = 5.2) and for a more frequent eating pattern (OR for 2 4 versus I2 meals per day = 2.1; Table 2). Foods and beverages that seemed to increase significantly the risk of endometrial cancer were further considered after allowance was made for the total energy intake. The strength of the direct association with cancer risk was substantially diminished for several items, whereas significantly elevated OR persisted for most meats, eggs, beans and peas, butter, seed oils, sugar, and tea (Table 3). None of the inverse associations with whole grain bread and pasta and various vegetables and fresh fruit was appreciably lowered by the allowance for total energy intake (Table 3). The simultaneous inclusion of major nondietary confounding factors (i.e., education, body mass index, parity, estrogen replacement therapy, and level of physical activity) did not modify further any of the OR adjusted for the total energy intake (not shown). Table 4 shows the distribution and the corresponding OR for the tertiles of the consumption of several nutrients, which were computed from the available list of foods and beverages. A high intake of retinol, total added fats, and total alcohol increased the endometrial cancer risk, but these direct associations were substan-

tially weakened by the allowance for total energy intake, which was itself associated with an OR (highest versus lowest tertile) of 2.7 (95% confidence interval, 1.8-3.9; Table 4). Similar to what is noticed in Table 3, the total energy intake made the beneficial effect of nutrients of plant origin emerge more clearly (OR for highest versus lowest tertile for both beta-carotene and ascorbic acid, allowing for total energy intake, 0.5; Table 4). When results were examined separately in the two study centers (i.e., Switzerland and Italy), they appeared consistent. The adverse effect of various meats, eggs, and added fats and the beneficial influence of whole grain bread and pasta and various fruits were of similar magnitude. However, direct associations with a few items, including white bread (but not pasta), biscuits, pastry, milk, cheese, potatoes, tea, wine, and total energy intake tended to be somewhat stronger in the Swiss data than in the Italian data. Conversely, a relatively more marked protection from various vegetables emerged from the Italian results. The intake of olive oil was inversely associated with the risk of endometrial cancer in the Swiss study population, whereas in Italian women its effect resembled that of other added fats more closely. Some of these differences are likely to be derived from some imbalances in the distribution of cases and controls by the approximate consumption tertile of a few food items in the two study centers. The allowance for total energy intake, however, diminished

Table 4. Odds Ratios of Endometrial Cancer According to Consumption Tertiles of Various Nutrients and Total Energy Intake (1988-1991)
Consumption tertile* No. of cases: no. of controls Nutrient
Retinol Beta-carotene Ascorbic acid Total added fats Total alcohol Total energy([
1 (low)

Odds ratio
3 (high) 115:168 89:197 87:195 89:116 97:131 118:166 Model Model Model Model Model Model Model Model Model Model Model
1 2 1 2 1 2 1 2 1 2 2

2 (intermediate)
80:201 96:182 88:192 108:229 77:204 93:189

It
1 1 1 1 1 1 1 1 1 1 1

2
1.14 0.97 1.02 0.85 0.94 0.76 1.77 1.63 0.94 0.78 1.76

Chi-square (trend)
13.104 3.77 1.53 11.63$ 1.15 13.613 27.71$ 19.54$ 5.405 0.60 27.6011

79:203 89:193 99:185 77:227 100:237 63:217

1.94 1.45 0.79 0.49 0.82 0.46 2.91 2.52 1.59 1.19 2.72

* Tertile cutpoints for: retinol: 43,500, 149,500 IU/month; beta-carotene: 86,300, 133,500 IU/month; ascorb~c acid: 2.7, 3.7 g/month; total alcohol: 0, I 0 ml/day; total energy: 1782, 2188 kcal/day; total added fats: combined score of butter, margarine, olive oil, and seed oil. t Reference category. $ P < 0.01. 5 P i0.05. 11 Odds ratio for intermediate and high tertile, after allowance for major confounding factors, 1.79 and 2.65; chi-square (trend) = 8.94. Model 1: estimates from multiple logistic equations. including terms for study center and age; Model 2: like Model 1, plus total energy intake.

3580

CANCER \une 1, 2993, Volume 71, No. 1 3

these apparent differences between study centers substantially. Discussion The amount of evidence on the role of body weight and estrogen replacement therapy' in the causation of endometrial cancer contrasts with the scarcity of data on the effect of dietary habits per se. In addition to investigating the frequency of consumption of several indicator foods and beverages, the current study provides the first set of data on dietary habits and the risk of endometrial cancer where an allowance for total energy intake was possible. It suggests that, aside from the predictable adverse effect of overeating and consequent obesity, some qualitative aspects of the habitual diet may also be associated with cancer risk, chiefly, the intake of animal proteins and fat (directly) and the consumption of fiber and fresh fruit and vegetables (inversely). Indeed, the allowance for the total energy intake reduced the apparent influence of frequent intake of several foods that initially emerged as risk factors for endometrial cancer, chiefly, starchy foods. It modified the elevated OR from high consumption of most types of meat, eggs, added fats, and sugar very little, however, and, if anything, increased the protection from the frequent intake of whole grain bread and pasta and most vegetables and fresh fruit. In terms of various nutrients, the halved risk of endometrial cancer in women in the highest tertile of intake of beta-carotene and ascorbic acid, but not retinol, after an allowance for energy intake, well summarizes the favorable effect of most foods of plant origin, although the current investigation, and most epidemiologic work on cancer, lacks the precision to determine whether the risk is modified by such micronutrients or by other components associated with a diet high in fresh fruit and vegetables. Indeed, fibers are another good candidate in the light of the strong protection afforded in this study by whole grain bread and pasta and of the reported ability of fibers to shield the estrogens excreted in the bile from bacterial deconjugation and reabsorption, thus lowering serum estrogen levels.' Also, the quantity and type of dietary fat seem to influence estrogen metabolism, and studies in human and animals have established that estrogen reabsorption in & e bowel is enhanced by diets rich in beef or h fat.' Furthermore, a decrease in circulating estradiol levels was observed with dietary fat r e d ~ c t i o n . 'In the ~ current investigation, almost all red meats (i.e., beef and pork) and added fats seemed to increase the risk of developing endometrial cancer independently from the reported level of energy intake.

The increased risk of endometrial cancer in women in the highest tertile of consumption of various alcoholic beverages (i.e., more than one glass per day) was not confirmed by the assessment of the influence of alcohol consumption after an allowance for the total energy intake. These results argue against the hypothesis that alcohol ingestion may reduce a woman's risk of endometrial cancer." They are compatible with two previous studies that showed there was no19 or, if anything, an adverse effect6 of alcohol intake. Certain limitations of the current study must be borne in mind. The choice of hospital controls, for instance, has long been debated,20particularly in relation to the analysis of life-style habits or diet. Great attention was, however, paid to obtaining unselected groups of cancer cases and controls from the same catchment areas. We excluded diseases potentially leading to changes in dietary habits from the comparison group and made sure that no inconsistencies in results existed across the major diagnostic categories of controls and study centers. Participation was high, and information bias should not have distorted these results because the role of diet as a cause of endometrial cancer was ignored by both study subjects and interviewers. Of greater concern was the limited food list that provided information only on the frequency of consumption of certain foods and beverages that were of interest a priori. Consequently, only an approximate estimate was possible for the total energy intake. However, the most important findings of this study (i.e., the association with various red meats and added fats [a direct relationship] and with whole grain bread and pasta and fresh fruit and vegetables [an inverse relationship]) were only slightly modified by the allowance for major dietary and nondietary potential confounding factors, including energy intake. A last comment concerns the comparison of endometrial cancer and the more common and thoroughly investigated breast cancer in regard to dietary factors. The association of endometrial cancer risk with the intake of energy, fat, proteins, fresh fruits, and vegetables are in the same direction as those for cancer of the breast, although they are somewhat stronger.' This difference may simply reflect chance or it may indicate that "recent diet" (investigated by the present study) is a better predictor for endometrial cancer risk than for breast cancer risk. A direct association between endometrial cancer risk and weight gain among women aged 60 years or more, independent of that with obesity, has been also interpreted as evidence that the net energy balance over most recent time could play a role.'l A few differences, however, have been established between cancer of the breast and the endometrium from an etio-

Dietary Factors and the Risk of Endometrial CAILevi e t al.

3581
12. Fidanza F, Versiglioni N. Tabelle di composizione degli alimenti. In: Fidanza F, Liguori G, editors. Nutrizione umana. Naples: Idelson, 1988:677-730. 13. Breslow NE, Day NE. Statistical methods in cancer research. vol. 1. Lyon, France: International Agency for Research on Cancer, 1980. IARC scientific publication no. 32. 14. Baker RJ, Nelder JA. The GLIM system. Release 3. Oxford, England: Numerical Algorithms Group, 1978. 15. Willett W, Stampfer MJ. Total energy intake: implications for epidemiologic analyses. Am J Epidetniol 1986; 124:17-27. 16. Pike M, Bernstein L, Peters RK. Total energy intake: implications for epidemiologic analyses [letter]. Ani ] Epideiniol 1989; 12911312-3. 17. Prentice R, Thompson D, Clifford C, Gorvbach S, Goldin B, Byar D. Dietary fat reduction and plasma estradiol concentration in healthy post-menopausal women. ] Natl Cancer Jmt 1990; 82:129-34. 18. Webster LA, Weiss NS, The Cancer and Steroid Hormone Study Group. Alcoholic beverage consumption and the risk of endometrial cancer. Jnt ] Epideniiol 1989; 18:786-91. 19. Williams RR, Horm JW. Association of cancer sites with tobacco and alcohol consumption and socioeconomic status of patients: interview study from the Third National Cancer Survey. J Nafl Cancer Jnsf 1983; 71:681-6. 20. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Znst 1959; 22: 719-48. 21. Le Marchand L, Wilkens LR, Mi MP. Early-age body size, adult weight gain and endometrial cancer risk. I t i t J Cancer 1991; 48:807-11. 22 Franceschi 5.Reproductive factors and cancers of the breast, ovary and endometrium. EurJ Cancer C h i Oncol 1989; 25:193343. 23. Gurpide E. Endometrial cancer: biochemical and clinical correlates. ] Nut/ Cancer Inst 1991; 83:405-16. 24. Ballard-Barbash R, Schatzkin A, Taylor PR, Kahle LL. Association of change in body mass with breast cancer. Caticer Rcs 1990; 50:2152-5. 25. Folsom AR, Kaye SA, Prineas JD, Potter JD, Gapstur SM, Wallace RB. Increased incidence of carcinoma of the breast associated with abdominal adiposity in postmenopausal women. Am 1 Epidemiol 1990; 131:794-803. 26. Austin H, Austin JM, Partridge EE, Hatch KD, Shingleton HM. Endometrial cancer, obesity, and body fat distribution. Cancer Res 1991; 51:568-72.

logic viewpoint, notably in regard to the importance of conditions or events occurring in the premenopausal period, such as overweight, age at first birth, and frequency of anovulation,22the effect of antiestrogens like t a m ~ x i f e n ,and, possibly, the association between ~~ body fat distribution and cancer r i ~ k . ~ * - ~ the comThus, parative study of dietary influences may provide a useful framework for interpreting existing epidemiologic knowledge of female tumors and designing broadspectrum interventions.
References
1. La Vecchia C. Nutritional factors and cancer of the breast, endometrium and ovary. Eur ] Cancer 1989; 25:1945-51. 2. Parazzini F, La Vecchia C, Bocciolone L, Franceschi S. The epidemiology of endometrial cancer. Gynecol Otrcol 1991; 41:l-16. 3. Armstrong BK. The role of diet in human carcinogenesis. In: Watson JD, Winsten JA, editors. Origin of human cancer. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, 1977:55763. 4. Kolonel LN, Hankin JH, Lee J, Chu SY, Nomura AMY, Ward Hinds M. Nutrient intakes in relation to cancer incidence in Hawaii. Br ] Cancer 1981; 44:332-9. 5. Armstrong BK. Diet and hormones in the epidemiology of breast and endometrial cancer. N u t r Cancer 1979; 1:90-5. 6. La Vecchia C, Decarli A, Fasoli M, Gentile A. Nutrition and diet in the etiology of endometrial cancer. Cancer 1986; 57:1248-53. 7. Cauley JA, Gutal JP, Kuller LH, LeDonne D, Powell JG. The epidemiology of serum sex hormones in postmenopausal women. Arn / Epiderniol 1989; 129:1120-31. 8. Gorbach SL, Goldin BR. Diet and the excretion and enterohepatic cycling of estrogens. Preo Mrd 1987; 16:525-31. 9. Muir CS, Waterhouse JAH, Mack T, Powell J, Whelan 5, editors. Cancer incidence in five continents. vol. 5 . Lyon, France: International Agency for Research on Cancer, 1987990-1. IARC scientific publication no. 88. 10. Levi F, La Vecchia C, Gulie C, Negri E, Monnier V, Franceschi S, et al. Oral contraceptives and the risk of endometrial cancer. Cancer Causes Control 1991; 2:99-103. 11. Paul AA, Southgate DAT, editors. McCance and Widdowsons the composition of foods. 4th ed. London: HMSO, 1978:37278.