Original article. Dietary glycemic load and colorectal cancer risk

Transcription

1 Annals of Oncology 2: Kluwer Academic Publishers. Printed in the Netherlands. Original article Dietary glycemic load and colorectal cancer risk S. Franceschi, ' 2 L. Dal Maso, L. Augustin, ' 3 E. Negri, 4 M. Parpinel, P. Boyle, 5 D. J. A. Jenkins 3 & C. LaVecchia 46 ^Servizio di Epidemiologia, Centro di Riferimento Oncologico, Istitulo Nazionale Tumori, Avtano, Italy. 2 Field and Intervention Studies Unit. International Agency for Research on Cancer, Lyon, France. 3 Department ofnutritional Sciences. Faculty of Medicine, University of Toronto and the Clinical Nutrition and Risk Factor Modification Center, St. Michael's Hospital, Toronto, Ontario, Canada: "istituto di Ricerche Farmacologiche 'Mario Negri'; s Division of Epidemiology and Biostatistics. European Institute of Oncology, Istituto di Statistica Medica e Biometria. Universita degli Studi di Milano, Milan. Italy Summary Background: Insulin and insulin-like growth factors can stimulate proliferation of colorectal cells. High intake of refined carbohydrates and markers of insulin resistance are associated with colorectal cancer. To test the insulin/colon cancer hypothesis, we determined whether the dietary glycemic index and the glycemic load are associated with colorectal cancer risk. Design: A case-control study on colorectal cancer conducted in Italy. Cases included 25 men and 828 women with histologically confirmed incident cancer of the colon or rectum. Controls were 2073 men and 208 women hospitalized for acute conditions. We calculated average daily dietary glycemic index and glycemic load, and fiber intake from a validated food frequency questionnaire. Introduction Carbohydrate foods consumed in isoglucidic amounts produce different glycemic and insulinemic responses depending upon the nature of the food (e.g., ratio of amylose to amylopectin, amount and type of fiber, particle size of starch granules) and type and extent of food processing (e.g., extrusion, flaking, grinding, and cooking) []. A role of diets with a high glycemic response and consequent hyperinsulinemia in colorectal carcinogenesis has been postulated [2, 3]. Several findings from an Italian case-control study support this hypothesis. Direct associations were found between colorectal cancer and intake of refined carbohydrates and energy [4], number of daily meals [5], body mass index (BMI) in men and, among women, central adiposity [6]. Conversely, inverse associations emerged for intake of vegetables, pulses, and fish [7], fibre [8], unsaturated-to-saturated fat ratio [4], and level of physical activity [9]. In order to test the insulin/colon cancer hypothesis, we decided to evaluate the dietary carbohydrate component using the glycemic index (GI), an indicator of Results: Direct associations with colorectal cancer risk emerged for glycemic index (odds ratio (OR) in highest vs. lowest quintile =.7; 95% confidence interval (CI):.4-2.0) and glycemic load (OR =.8; 95% CI:.5-2.2), after allowance for sociodemographic factors, physical activity, number of daily meals, and intakes of fiber, alcohol and energy. ORs were more elevated for cancer of the colon than rectum. Overweight and low intake of fiber from vegetables and fruit appeared to amplify the adverse consequences of high glycemic load. Conclusions: The positive associations of glycemic index and load with colorectal cancer suggest a detrimental role of refined carbohydrates in the etiology of the disease. Key words: colon cancer, fiber, glycemic load, rectal cancer carbohydrate foods' ability to raise blood glucose levels, and the glycemic load score (GL), as a combination of quality as well as quantity of carbohydrates consumed, and thus a measure of dietary insulin demand. Patients and methods A case-control study of cancer of the colon and rectum was conducted between January 992 and June 996 in six Italian areas: the provinces of Pordenone and Gorizia in North-eastern Italy; the urban areas of Milan and Genoa and the provinces of Forli, in the North; and Latina and the urban area of Naples in the South [7]. Cases had histologically confirmed colorectal cancer diagnosed no longer than one year prior to the interview (median month) and with no previous diagnoses of cancer at other sites. Overall, 225 subjects with cancer of the colon (688 men and 537 women, median age 62, range 9-74 years) and 728 with cancer of the rectum and rectosigmoid junction (437 men and 29 women, median age 62, range years) were included. Controls were in-patients with no history of cancer, in major teaching and general hospitals in the same areas where cases lived, and identified for acute, non-neoplastic. non-gynecological conditions, unrelated to hormonal or digestive tract diseases. All chronic conditions which may lead to lifestyle modifications (e.g., diabetes mcllitus. cardiovascular diseases, etc.) were not eligible causes of hospital Downloaded from by guest on 09 November 208

2 74 Table I. Glycemic index (GI) values for a few common foods and recipes. ' b Foods White bread c Biscuits Sugar Pizza Ice cream Pasta/rice with tomato sauce Apples and pears Peas, beans, chick peas, lentils Vegetable soup with noodles Whole milk Peaches, apricots and prunes Tomatoes GI " Modified by references [7-9]. b Meat- and fish-based recipes and cheese have a GI close to 0. c Standard food admission. Conversely, comorbidity from the above conditions did not represent an exclusion criterion. The control group included 2073 men and 208 women aged 9-74 years (median age 58) belonging to the following diagnostic categories: trauma, mostly fractures and sprains (27%); other orthopedic disorders, such as low back and disc disorders (24%); acute surgical conditions (8%); eye diseases (24%); and other miscellaneous diseases, such as ear, nose, throat, skin and dental conditions (7%). About 4% of cases and controls approached during their hospital stay refused to be interviewed. The same structured questionnaire and coding manual were used in each center, and all interviewers were centrally trained and routinely supervised. The questionnaire included information on socio-demographic characteristics, such as education and occupation, lifetime smoking and alcohol-drinking habits, physical activity, anthropometric measures at various ages, a problem-oriented personal medical history and family history of cancer. Body mass index (BMI) was computed as weight in kilograms divided by height in square meters. I n order to compute waist-to-hip (W. H) ratio, the interviewer measured the circumference of waist (2 cm above the umbilicus) and hip (maximal protrusion). On account of medical or logistic problems, waist and/or hip circumferences were not measured in 25% of colorectal cancer cases and 9% of control subjects [6]. An interviewer-administered food-frequency questionnaire (FFQ) was used to assess subjects' habitual diet, including total energy. Average weekly frequency of consumption of foods or food groups, as well as complex recipes during the two years prior to cancer diagnosis or hospital admission (for controls) was elicited. The FFQ also included 5 questions aimed at assessing individual fat intake pattern and meal frequency. Satisfactory reproducibility [0, ] and validity [2] of the FFQ have been reported. To compute energy and nutrient intake, an Italian food-composition data base was used [3]. Fiber, defined as non starch polysaccharides plus hgnin, was divided according to the type of food from which it originated (i.e.. vegetables and fruit or cereals) [4-6]. The GI is a qualitative assessment of foods based on the incremental glucose response and insulin demand they produce for a given amount of carbohydrates. We expressed GI as a percent of the glycemic response elicited by white bread as a standard food. For each case and control subject, we calculated daily average GI by summing the products of the carbohydrate content per serving for each food or recipe, times the average number of servings of that food per week, times its GI, all divided by the total amount of available carbohydrate weekly intake [7, 8]. A score for the daily average GL was computed as the GI. but without dividing by the total amount of carbohydrates. For these calculations we used the carbohydrate content of 50 foods or recipes since 28 foods or recipes, chiefly meat- and fish-based ones and cheese, contained a negligible amount of carbohydrates [3]. With Table 2. Distribution of 225 cases of colon cancer. 728 of rectal cancer and 454 controls" by sex, age group, years of education, and level of physical activity. Italy, Characteristic Sex Male Female Age group (years) < ^70 Education (years) <7 7-3= 2 2 b Physical activity (at the work place) Low Medium High 2 b Cancer cases Colon, Rectum. n (%) n (%) 688(56) 437(60) 537(44) 29(40) 55 (4) 26 (4) 4(9) 67(9) 32(26) 97(27) 58(42) 306(42) 27(8) 32(8) 62(5) 422(58) 33 (27) 8 (25) 267(22) 22(7) C (36) 23(32) 45(37) 258(35) 330(27) 239(33) C.40 Controls.»(%.) 2073(50) 208(50) 347 (8) 732(8) 244(30) 356(33) 475() 2276(55) 56(28) 693(7) 378(33) 476(36) 299(3) u Some figures do not add up to the total because of some missing values. b Compared to the control group, adjusted for age, sex, and center. c P < respect to GI values we chiefly used international tables [8]. In order to take into account Italian cooking habits (e.g., pasta 'al dente'), Italian sources were used for a few local recipes [9]. Food items for which a GI had not been determined were assigned the GI of the nearest comparable food (e.g., tangerines were assigned the GI of oranges). As an example, Table shows GIs estimated for a few common foods and recipes. Odds ratios (ORs) and the corresponding 95% confidence interval for quintiles of GI and GL were derived using unconditional multiple logistic regression models [20]. The regression equations included terms for age, sex, study center, years of education, occupational physical activity, number of daily meals, and intakes of fiber, alcohol and energy. Adjustment for energy was made using the residuals method [2]. GI and GL were also introduced as continuous variables, and the unit of measurement were 5 and 50, respectively. Results After allowance for age, sex, and center, patients with cancer of the colon, but not rectum, were significantly more educated and reported lower levels of physical activity than controls (Table 2). GI was positively correlated to GL (Pearson correlation coefficient, r = 0.59), intake of bread (r = 0.66), cakes and sweets (r = 0.28), table sugar (r ), and energy (r = 0.27), but negatively correlated to intake of fruit (r = -0.9) and vegetables (r = -0.2). Correlation of GI with other dietary and non-dietary factors, including intake of different types of fat, alcoholic beverages, daily number of meals, BMI, level of physical activity, and smoking habits were weak (i.e., r < 0.0). Downloaded from by guest on 09 November 208

3 75 Table 3. Odds ratios (ORs) and 95% confidence intervals (CIs) a of colorectal cancer by quintile of energy-adjusted daily glycemic index and glycemic load score, by subsite(s) and overall. Italy, Cancer site(s) Number OR (95% Quintile CI) Xi (trend) P Continuous " Glycemic index Upper limit Colon 225 Rectum 728 Colon and rectum 953 Glycemic load Upper limit Colon 225 Rectum 728 Colon and rectum S.4 (.2-.8).0 (0 8-.4).3 (.-.5) 90.2 (.0-.6).2 (0.9-.5) 2 (.0-.5) (.4-2.).4 (.-.9).6 (.3-.9) (.0-.5).0 (0.7-.3) I.I (0.9-.4) (.3-2.0).3 (.0-.7).5 (.2-.8) (.3-2.0).3 (.0-.7).5 (.2-.8) _.9 (.5-2.4).4 (.-.9).7 (.4-2.0) _.9 (.5-2.4).5 (.-.9).8 (.5-2.2) 29.5 P < P = P < P < ^ = P < 0.00 (5 units x day).8 (.-.26).2 (.04-.2) 6 (.0-.22) (50 units x day).26 (.6-.36).09 ( ).20 (.3-29) " Adjusted for age, sex, study center, years of education, occupational physical activity, number of daily meals, intakes of fiber, alcohol and energy. b Reference category. Colorectal cancer risk increased with an increase in dietary GI (OR for highest versus lowest quintile:.7; 95% CI:.4-2.0) and GL (OR =.8; 95% CI:.5-2.2) (Table 3). The ORs were somewhat higher for colon cancer (.9 for GI and.9 for GL in the highest quintile) than rectal cancer (.4 and.5, respectively). Findings were similar in separate strata of age, but the association for the combination of colorectal cancer was somewhat stronger among women (OR in highest quintile = 2.0 for GI and 2.6 for GL) than among men (ORs.5 for either GI and GL). This difference was probably accounted for by a greater proportion of subjects with colon cancer in the female than in the male sex. No material interaction or effect modification was observed between GI or GL and level of physical activity, alcohol drinking and unsaturated-to-saturated fatty acid ratio. One hundred sixteen (6%) cases and one hundred eighty-five (4%) control subjects reported a history of non-insulin dependent diabetes. OR in the highest intake quintile of GI and GL were, respectively, 2. and 2.4 among diabetic individuals and.7 and.8 among nondiabetic individuals. We examined the joint effect of GL and different fiber intake by cross-classifying patients with both variables (Figure la and b). GL was directly associated with colorectal cancer risk in each stratum of fiber from vegetables and fruit (median intake.9 g per day). Similarly, low fiber intakes showed elevated ORs in each stratum of dietary GL (Figure la). The OR for the combination of the highest tertile of GL and the lowest tertile of fiber from vegetables and fruit compared with the opposite extreme was.9 (95% CI:.5-2.4). GL and fiber from cereals (median intake 5.7 g per day) produced a different picture (Figure lb). The OR for the combination of the highest tertile of GL and the lowest one of cereal fiber was 0.8 (95% CI: ), whereas for the highest tertile of GL and the highest of cereal fiber was.5 (95% CI:.0-2.). In fact GL and fiber from cereals were strongly correlated (r - 0.8). We also assessed the combined influence of GL and the best indicators of visceral adipose tissue in our study, i.e., W: H ratio in women and BMI in men (Table 4). The OR was 2.8 (95% CI:.9-4.) in the highest GL tertile in combination with the highest tertile of W: H ratio in women. Among men, the OR for the combination of the highest tertile of BMI and the highest GL tertile was.8 (95% CI:.3-2.7). Discussion High dietary GI and GL have been associated with increased risk of non-insulin dependent diabetes [22,23] and coronary heart disease [24], but the influence of GI and GL on colorectal cancer risk has been little studied [25]. Our present findings show that a diet that increases glycemic response is involved in the etiology of cancers of the colon-rectum, particularly of those which arise from the colon. The associations with GI and GL were independent from the presence of diabetes. Low intake Downloaded from by guest on 09 November 208

4 76 a) Fiber from vegetables and fruit High (>245) Medium! ) Low (<77) Glycemic load b) Fiber from cereals High (>245) Medium (77-245) Low (< 77) Glycemic load Figure I. Odds ratio* of colorectal cancer by tertile of energy-adjusted glycemic load and fiber intake from different sources. Italy, 'Adjusted for age. sex. study center, years of education, occupational physical activity, meal frequency, alcohol consumption, and energy intake. * Reference category. of fiber from vegetables and fruit and elevated BMI or W: H ratio appeared to contribute in enhancing colorectal cancer risk when the diet was high in GL. The positive associations we observed between GI and GL and colorectal cancer were among the strongest reported so far for any dietary factor [7, 25], and were consistent in different strata of age, sex, and various risk covariates. The strong risk increase among individuals on a high-gl-low-fiber diet is in agreement with the property of fiber to flatten the glucose-response and has also been observed in individuals with non-insulin dependent diabetes [22, 23]. Fiber from vegetables and fruit, which accounted for 2/3 of fiber intake, seemed to exert a beneficial effect on colorectal cancer risk and, possibly, to mitigate the impact of high dietary GL. It is possible that fiber from cereals did not appear to be protective in our [8] as well as in other study populations [26] because it was derived chiefly from highly refined cereal products [27]. Data on the relationship between different types of fiber and glycemic response and insulin demand depend on the nature of fiber and its relation to the nutrients contained in the food [28]. When taken out of their botanical structure, isolated cereal fiber is not seen to reduce the rate of carbohydrate absorption while viscous fiber (i.e., guar) has been shown to decrease the rate of glucose absorption [29]. Whereas consumption of sugar and cakes is relatively low, Italy shows the highest consumption of carbohydrates from refined cereals among affluent countries [30], up to more than 300 g/day in one fifth of this study population. Interestingly, the highest intake quintile of white bread, whose GI is 00, was associated with an energy-adjusted OR of.6 (95% CI:.3-.9) whereas the highest intake quintile of pasta, whose GI is only approximately 60, showed an OR of.2 (95% CI:.0-.5). Visceral adipose tissue has been associated with increased insulin resistance and colorectal cancer risk [3]. Table 4. Odds ratios (ORs) and 95% confidence intervals (CIs)" of colorectal cancer by tertile of energy-adjusted daily glycemic load and waist-to-hip ratio in females, and body mass index in males. Italy Females Waist-to-hip ratio < >0.86 All" Males Body mass index < > All" OR (95% CI) Glycemic load >27 All" T.5.9 l c (.-2.) (.4-2.8) ( ) (.5-3.) (.3-2.7) (.2-2 0) (0.9-.8) (.-2.4) (.9-4.) (.2-2.) r 0.9 (0.6-3). (0 8-.5) l c.4 (.-.8) 0.9 (0.7-.3).2 (0.8-.6).3 (0.9-.8). (0.9-.4).7 (.3-2 2).2 (0.9-7).3 (0.9-.8).8 (.3-2 7).4 (.- 8) l c. (0.9-.3).3 (.-.6) a Adjusted for age, study center, years of education, occupational physical activity, number of daily meals, intakes of fiber, alcohol and energy, and recent weight loss. b Adjusted also for glycemic load, waist-to-hip ratio or body mass index as appropriate. c Reference category. Downloaded from by guest on 09 November 208

5 77 Men have more visceral adipose tissue than women. Thus, BMI was a good surrogate of visceral adipose tissue in the male sex while W: H ratio had to be considered among women [6]. Assessment of BMI and W: H ratio in colorectal cancer patients is hampered by weight loss in the proximity of cancer diagnosis. We had, however, information on BMI changes over time, and we were able to allow for recent weight loss [6]. In addition to dietary findings, several lines of evidence suggest a role of insulin in colorectal carcinogenesis. Non-insulin dependent diabetes mellitus is associated with increased risk of colorectal cancer [32, 33]. Insulin is considered to be a growth factor and has been shown to increase insulin-like growth factor (IGF)-I, which is involved in long-term regulation of energy metabolism [2]. Insulin and IGF receptors have been found in both normal and malignant cells of colonic mucosa, and can stimulate proliferation of human colorectal cells [2, 34]. Tall individuals [35] and those with acromegaly, who have elevated levels of growth hormone and IGF-I [36], have an increased incidence of colorectal cancer. Circulating levels of IGF-I and II and IGF binding protein (IGF BP)-III, which modulate access of IGF-I to IGFreceptors, were related to subsequent colorectal cancer risk [37, 38]. We used a validated questionnaire [0, 2] and had access to a large number of colorectal cancer patients and control subjects who, in the vast majority, agreed to participate in the study. The control group included a broad range of acute conditions unrelated to diet modifications. Reported associations with GI and GL were not restricted to any major nosological category of control, and held true when cases and controls with a history of diabetes were excluded. Allowance was possible for a number of potential confounding factors, including number of daily meals, energy and alcohol intake, and physical activity. Weaknesses of the present findings include, however, the limitations of current GI estimates. Many of them have been derived from small samples and variations of estimates in mixed-meal situations are unclear [7,8]. Notwithstanding, the present findings lend indirect support to the hypothesis that increased glycemic response and insulin resistance are associated with colorectal cancer, most notably with that of the colon. Acknowledgements This work was supported by the contributions of the Italian Association for Research on Cancer, Milan, Italy. The authors wish to thank Drs A. Giacosa, M. Montella, E. Conti, and Prof. D. Amadori for their help in study coordination, and Mrs L. Mei for editorial assistance. References. Jenkins DJ, Kendall CW, Axelsen M et al. Viscous and nonviscous fibres, nonabsorbable and low glycaemic index carbohydrates, blood lipids and coronary heart disease. Curr Opin Lipidol 2000: : McKeown-Eyssen G. Epidemiology of colorectal cancer revisited: Are serum triglycerides and/or plasma glucose associated with risk? Cancer Epidemiol Biomarkers Prev 994; 3: Giovannucci E. Insulin and colon cancer. Cancer Causes Control 995; 6: Franceschi S, La Vecchia C, Russo A et al. Macronutrient intake and risk of colorectal cancer in Italy. Int J Cancer 998: 76: Favero A, Franceschi S, La Vecchia C et al. Meal frequency and coffee intake in colon cancer. Nutr Cancer 998: 30: Russo A, Franceschi S, La Vecchia C et al. Body size and colorectal cancer risk. Int J Cancer 998; 78: Franceschi S. Favero A. La Vecchia C et al. Food groups and risk of colorectal cancer in Italy. Int J Cancer 997; 72: Negri E, Franceschi S, Parpinel M, La Vecchia C. Fiber intake and risk of colorectal cancer. Cancer Epidemiol Biomarkers Prev 998:7: Tavani A, Braga C. La Vecchia C et al. Physical activity and risk of cancers of the colon and rectum: An Italian case-control study. Br J Cancer 999: Franceschi S, Negri E, Salvini S et al. Reproducibilily of an Italian food frequency questionnaire for cancer studies: Results for specific food items. Eur J Cancer 993; 29A: Franceschi S, Barbone F, Negri E et al. Reproducibility of an Italian food frequency questionnaire for cancer studies. Results for specific nutrients. Ann Epidemiol 995; Decarh A, Franceschi S, Ferraroni M et al. Validation of a foodfrequency questionnaire to assess dietary intakes in cancer studies in Italy. Results for specific nutrients. Ann Epidemiol 996; 6: Salvini S, Parpinel M, Gnagnarella P. Maisonneuve P. Turrini A. Banca Dati di composizione degh ahmenti per studi epidemiologici in Italia. Milano, Italy: Istituto Europeo di Oncologia Holland B, Unwin ID, Buss DH. Cereals and cereal products. The third supplement to McCance & Widdowson's. The composition of foods (4th edition). Nottingham, UK: The Royal Society of Chemistry and Ministry of Agriculture, Fisheries and Food Holland B, Unwin ID, Buss DH. Vegetables, herbs and spices. The fifth supplement to McCance & Widdowson's. The composition of foods (4th edition). Nottingham. UK. The Royal Society of Chemistry and Ministry of Agriculture, Fisheries and Food Holland B, Unwin ID. Buss DH. Fruit and nuts. The first supplement to the fifth edition of McCance & Widdowson's. The composition of foods. Nottingham, UK: The Royal Society of Chemistry and Ministry of Agriculture, Fisheries and Food Wolever TM, Nguyen PM, Chiasson JL et al Determinants of diet glycemic index calculated retrospectively from diet records of 342 individuals with non-insulin-dependent diabetes mellitus. Am J Clin Nutr 994; 59: Foster-Powell K, Miller JB. International tables of glycemic index. Am J Clin Nutr 995; 62: 87S-93S. 9. Bnghenti F. Casiraghi MC. Influenza dei processi di trasformazione sulla risposta glicemica ad alimenti amidacei. Giornale Itahano di Nutrizione Clinica e Preventiva 992; : Breslow NE, Day NE. Statistical methods in cancer research. Vol. I. The analysis of case-control studies. IARC Scientific Publications no. 32 Lyon: International Agency for Research on Cancer Willett WC. Stampfer MJ. Total energy intake: Implications for epidemiologic analysis. Am J Epidemiol 986; 24: Salmeron J. Ascherio A, Rimm EB et al. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care 997:20: Salmeron J, Manson JE. Stampfer MJ et al. Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA 997; Downloaded from by guest on 09 November 208

6 Liu S, Stampfer MJ, Manson JE et al. A prospective study of dietary glycemic load, carbohydrate intake and risk of coronary heart disease in US women. Am J Clin Nutr 2000; 7: Slattery ML, Benson J, Berry DTet al. Dietary sugar and colon cancer. Cancer Epidemiol Biomark Prev 997; 6: Fuchs CS, Giovannucci EL, Colditz GA et al. Dietary fiber and the risk of colorectal cancer and adenoma in women. N Engl J Med 999; 340: Favero A. Salvini S, Russo A et al Sources of macro- and micronutrients in Italian women: Results from a food frequency questionnaire for cancer studies. Eur J Cancer Prev 997; 6: Jenkins DJA. Axelsen M, Kendall CWC et al. Dietary fiber, lente carbohydrates and the insulin resistant diseases. Br J Nutr 2000; 83 (Suppl ): S Leclere CJ, Champ M, Boillot J et al. Role of viscous guar gums in lowering the glycemic response after a solid meal. Am J Clin Nutr 994; 59: World Cancer Research Fund in association with American Institute for Cancer Research. Food, Nutrition and the Prevention of Cancer: A Global Perspective. Washington. DC: American Institute for Cancer Research Schoen RE.TangenCM, Kuller LH et al. Increased blood glucose and insulin, body size, and incident colorectal cancer. J Natl Cancer Inst 999: La Vecchia C, Negri E, Decarli A, Franceschi S. Diabetes melhtus and colorectal cancer risk. Cancer Epidemiol Biomarkers Prev 997:6: Hu FB, Manson JE, Simin L et al. Prospective study of adult onset diabetes mellitus (type 2) and risk of colorectal cancer in US women. J Natl Cancer Inst 999; 9: Lahm H, Suardet L, Laurent PL et al. Growth regulation and costimulation of human colorectal cancer cell lines by insulin-like growth factor I, II and transforming growth factor alpha. Br J Cancer 992; 65: Hebert PR, Ajani U, Cook NR et al. Adult height and incidence of cancer in male physicians (United States). Cancer Causes Control 997; 8: Ron E, Gridley G, Hrubec Z, Page W, Arora S, Fraumeni JF Jr. Acromegaly and gastrointestinal cancer [published erratum appeared in Cancer 992; 69: 549]. Cancer 99, 68: Ma J, Pollak MN. Giovannucci E et al. Prospective study of colorectal cancer risk in men and plasma levels of insulin-like growth factor (IGF)-I and IGF-binding protein-3. J Natl Cancer Inst 999; 9: Manousos O, Souglakos J, Bosetti C et al. IGF- and IGF-II in relation to colorectal cancer. Int J Cancer 999; 83: 5-7. Received 0 August 2000; accepted 29 September Correspondence to Dr L. Dal Maso Servizio di Epidemiologia Centro di Riferimento Oncologico Istituto Nazionale Tumon Via Pedemontana Occ 3308 Aviano (PN) Italy epidemiology@ets.it Downloaded from by guest on 09 November 208