Linoleic acid intake and cancer risk: a review and meta-analysis 1 3

Transcription

1 Linoleic acid intake and cancer risk: a review and meta-analysis 1 3 Peter L Zock and Martijn B Katan ABSTRACT Replacement of saturated fat by the major dietary polyunsaturated fat linoleic acid reduces blood cholesterol concentrations and the risk of coronary artery disease. However, there is concern that long-term consumption of large amounts of linoleic acid might increase cancer risk. We reviewed the epidemiologic and experimental literature on linoleic acid intake and cancer risk and performed additional meta-analyses of risk estimates from case-control and prospective cohort studies. None of the combined estimates from within-population studies indicated a significantly increased risk of cancer with high compared with low intakes of linoleic acid or polyunsaturated fat. For case-control studies, the combined relative risks were 0.84 (95% CI: 0.71, 1.00) for breast, 0.92 (95% CI: 0.85, 1.08) for colorectal, and 1.27 (95% CI: 0.97, 1.66) for prostate cancer. For prospective cohort studies, combined relative risks were 1.05 (95% CI: 0.83, 1.34) for breast, 0.92 (95% CI: 0.70, 1.22) for colon, and 0.83 (95% CI: 0.56, 1.24) for prostate cancer. Ecologic comparisons of populations showed positive associations between cancer rates and per capita use of animal or saturated fat, but less so with per capita use of vegetable oil or polyunsaturated fat. Controlled studies of coronary artery disease in men did not, except for 1 study, show an increased cancer incidence after consumption of diets with a very high linoleic acid content for several years. Animal experiments indicated that a minimum amount of linoleic acid is required to promote growth of artificially induced tumors in rodents; but above this threshold, linoleic acid did not appear to have a specific tumor-promoting effect. Although current evidence cannot exclude a small increase in risk, it seems unlikely that a high intake of linoleic acid substantially raises the risks of breast, colorectal, or prostate cancer in humans. Am J Clin Nutr 1998;68: KEY WORDS Linoleic acid, n 6 polyunsaturated fat, carcinogenesis, tumors, breast cancer, colon cancer, rectal cancer, prostate cancer, risk factors, humans, meta-analysis INTRODUCTION See corresponding editorial on page 5. Cancer risk in humans may be linked to the composition of the diet (1). In particular, dietary fat intake is often thought to be involved in the etiology of breast and colon cancer (2); both saturated and polyunsaturated fats have been implicated (3, 4). An increased intake of polyunsaturated fat is considered favorable because of its beneficial effects on blood cholesterol concentrations. However, the possible adverse effects of a high polyunsaturated fat intake require scrutiny. The major polyunsaturated fatty acid in most diets is linoleic acid (cis, cis-18:2n 6). Linoleic acid is an essential fatty acid; it is required for the biosynthesis of eicosanoids but cannot be synthesized by the human body. An intake of 2 3% of dietary energy is probably enough to prevent deficiency (5). The average linoleic acid intake in the United States and Western Europe has risen from 3% of energy in the 1950s to 6 7% at present, with a commensurate decrease in saturated fat (6 8). A few populations (Northern Belgium and Israel) habitually consume 8 12% of their total energy intake as linoleic acid (9 12). Replacement of saturated fat with linoleic acid is advocated to improve serum lipoprotein profiles and reduce the risk of coronary artery disease (CAD) (5, 13, 14). Replacement of saturated fat with unsaturated fat does not cause a decrease in HDL cholesterol as occurs when saturated fat is replaced with carbohydrates (15, 16). Some experts have recommended that linoleic acid consumption be raised to 10% of energy intake (5, 13, 14). However, over the past 25 y the wisdom of recommending high intakes of polyunsaturated fatty acids has been increasingly questioned (17 22). The major concern is whether diets high in polyunsaturated fatty acids increase cancer risk because linoleic acid has been linked to the development of cancer in animals (23, 24), and some population comparisons report positive associations with per capita use of polyunsaturated fatty acids (3, 4). An added concern is that polyunsaturated fat may be prone to oxidation, which may play a role in carcinogenesis (25, 26) and may increase the susceptibility of LDL particles to oxidative modification (27). Such concerns led to the current recommendations that the average intake of polyunsaturated fats remain at 7% of total energy and that individual intake should not exceed 10% of total energy intake (28, 1 From the Wageningen Agricultural University, Department of Food Technology and Nutritional Sciences, Division of Human Nutrition and Epidemiology, Wageningen, Netherlands. 2 Supported by the Foundation for Nutrition and Health Research. 3 Reprints not available. Address correspondence to MB Katan, Wageningen Agricultural University, Department of Food Technology and Nutritional Sciences, Division of Human Nutrition and Epidemiology, Bomenweg 2, 6703 HD Wageningen, Netherlands. Received July 30, Accepted for publication December 26, Am J Clin Nutr 1998;68: Printed in USA American Society for Clinical Nutrition

2 29). However, the question remains whether high intakes of linoleic acid do increase the risk of cancer, especially cancer of the breast, colon, and prostate. LINOLEIC ACID AND CANCER RISK 143 METHODS To identify studies of the relation between diet and cancers of the breast, colon and rectum, and prostate we searched the MED- LINE database (National Library of Medicine, Bethesda, MD) for the years and BIOLOGICAL ABSTRACTS for the years using the following key words: diet, fat, fatty acid, cancer, neoplasm, malignancy, carcinogenesis, tumor, carcinoma, and adenoma. We also checked citations in the identified articles. In vitro studies were not considered. We specifically selected epidemiologic studies that provided quantitative estimates of cancer risk and its SE with high compared with low intakes of linoleic acid or polyunsaturated fat. For combined estimates across studies, we used results of published meta-analyses and performed additional meta-analyses when necessary. To this end, we extracted from individual studies the risk estimate that referred to the largest difference in intake and that reflected the greatest degree of control for other environmental and dietary risk factors. When studies provided estimates of risk for subgroups of subjects, eg, older and younger subjects, we used or calculated the risk estimate for all subjects. For combining risk estimates, we used a random-effects model (30) to take into account both the sampling variance within studies and the variation in the true underlying effects across the studies being combined. Such variation in underlying effects (heterogeneity) seemed plausible because of the differences in populations, designs, and methods among studies. A model assuming equal sampling variances for each study, ie, each study having equal weight, and a fixed-effects model assuming the same underlying effect across studies (homogeneity) yielded similar results. RESULTS Breast cancer Analytic studies within populations Case-control studies. In the 16 case-control studies from which we were able to extract a quantitative estimate of breast cancer risk, risk was not increased with high intakes of linoleic acid (Figure 1). The combined relative risk, involving a total of 6910 cases and 8536 control subjects, was 0.84 (95% CI: 0.71, 1.00). This outcome agrees with results of previous meta-analyses. Boyd et al (47) calculated a combined relative risk of 0.92 (95% CI: 0.79, 1.08) from 9 case-control studies published before Two of these 9 studies (36, 37) showed a lower breast cancer risk with higher polyunsaturated fat intake and the other 7 studies (31 35, 38, 39) showed no significant effect (Figure 1). Howe et al (48) pooled the results of 8 studies that provided data on polyunsaturated fat. Among 5167 postmenopausal cases and control subjects, the univariate relative risks for each extra 45 g/d were 1.25 for polyunsaturated, 1.46 for saturated, and 1.41 for monounsaturated fat. In a model that included all 3 types of fat, the relative risk for polyunsaturated fat was 0.78 (95% CI: 0.51, 1.17). In 4 recent studies not included in previous meta-analyses (40 43), the relative risks FIGURE 1. Relative risks of breast cancer with high compared with low intakes of linoleic acid in case-control studies. Bars are 95% CIs. for high compared with low polyunsaturated fat intakes ranged from 0.7 to 1.3 (Figure 1). The studies mentioned above all assessed fat intake by selfreport of subjects. Biomarkers of intake have also been used. One study measured the fatty acid composition of erythrocyte phospholipids in 46 breast cancer cases and 53 control women in Moscow and found a significantly reduced risk of breast cancer associated with a high proportion of linoleic acid in phospholipids (44) (Figure 1). Erythrocyte phospholipids reflect diet in the past weeks or months (49) and it is possible that patients changed their diets because of their disease or that the disease affected the proportion of linoleic acid in their erythrocytes. A better longterm biomarker is the fatty acid composition of adipose tissue (50). A study among Boston women found no associations between breast cancer risk and n 6 polyunsaturated fatty acids in buttock fat from 380 breast cancer cases and 397 control subjects (45). Similar findings were reported for women in New York (46) (Figure 1). Several case-control studies reported results in terms other than relative risk. A study in Hawaii (51) showed no difference in self-reported linoleic acid intake between breast cancer cases and control subjects. In studies from Finland (52) and Israel (53) there were no significant differences in the linoleic acid content of breast adipose tissue between cases and control subjects. In summary, the case-control studies analyzed showed either no or negative associations between linoleic acid intake and breast cancer risk. Prospective cohort studies. The prospective cohort studies analyzed also did not show positive associations between linoleic acid intake and breast cancer risk (Figure 2). Hunter et al (59) conducted a pooled analysis on standardized data from 7 major cohort studies of fat intake and breast cancer involving a total of 4980 cases among women (41, 54 58, 66). The pooled relative risk, corrected for dietary measurement error, was 1.05 (95% CI: 0.83, 1.34) for each 10-g/d increase in polyunsaturated fat intake (Figure 2). In 3 cohort studies (60 62) that did not meet the criteria of Hunter et al (59), the relative risks for high compared with low polyunsaturated fat intake ranged from 0.73 to 1.23 (Figure 2). In a cohort from California, intake of all types of fat, including linoleic acid, was higher in 15

3 144 ZOCK AND KATAN FIGURE 2. Relative risks of breast cancer with high compared with low intakes of linoleic acid in prospective cohort studies. The pooled estimate was derived from the study by Hunter et al (59). Bars are 95% Cis. The study by Willett et al (58) refers to estimates from the Nurses Health study: A, follow-up from 1980 to 1986; B, follow-up from 1986 to cases than in 575 women who did not develop breast cancer. These differences disappeared after adjustment for total energy intake (67). In a biomarker study that measured phospholipid fatty acids in blood sampled y before the onset of breast cancer, a high proportion of linoleic acid was associated with reduced risk in women younger than 55 y (Figure 2), but not in those older than 55 y (63). Two studies investigated the effect of dietary fat intake after diagnosis on cancer survival. In a study of 161 white and 182 Japanese breast cancer patients in Hawaii (64), the white women with a high polyunsaturated fat intake, not adjusted for energy intake, had a greater risk of death (Figure 2; relative risk: 1.72). In an Australian study (65), the relative risk of death of 412 breast cancer patients was 1.57 at a high compared with a low intake of polyunsaturated fat and 1.14 at a high compared with a low linoleic acid intake (Figure 2). In summary, longitudinal epidemiologic studies (Figure 2) do not suggest that linoleic acid intake has a marked effect on the development of breast cancer. The 1 exception is a new analysis of the Swedish Mammography Screening cohort (68). An earlier analysis showed no relation with polyunsaturated fat intake (41, 59). The new analysis applied a mutual adjustment for intakes of different types of fat and found a relative risk of breast cancer of 1.69 (95% CI: 1.02, 2.78) for each 5-g/d increase in intake of polyunsaturated fat (68). However, other studies also adjusted for mutual confounding among the various types of fat and found no increased risk of breast cancer with higher intakes of polyunsaturated fat (40, 58). Thus, the bulk of the data still suggest that linoleic acid does not have a marked effect on the risk of breast cancer. The findings on survival in breast cancer patients suggest a possible adverse effect on mammary tumor progression, but the data are limited. Ecologic comparisons between populations Comparisons between countries generally show positive correlations between per capita disappearance of total fat and breast cancer incidence or mortality (69 73). Some studies suggest that these associations are mainly due to the use of animal and saturated fats rather than to the use of vegetables or polyunsaturated fats (72, 74 76), but others show positive associations of breast cancer rates with the use of polyunsaturated fat (3, 4, 77, 78). Carroll (76) reported that mortality from breast cancer is strongly associated with intakes of fat from animal sources but not with the percentage of energy as polyunsaturated fat; in contrast, Prentice et al (3, 77) estimated that a 50% reduction in both saturated and polyunsaturated fats would reduce breast cancer risk by half. Thus, results from ecologic comparisons are not consistent. In view of the uncertainties concerning food disappearance data, population comparisons that use biomarkers of intake may be more reliable. One study assessed intake from linoleic acid in the adipose tissue of subjects from 10 European regions and Israel (79) (Figure 3). Incidence in the Israeli women, whose linoleic acid intakes have been some 10 12% of energy for the past decades (11, 12), was not higher than in women from northwestern Europe, whose intakes were probably 4 7% of energy. The breast cancer rate in the Israeli women may have been additionally inflated by the high frequency of BRCA1 and BRCA2 mutations in Ashkenazi Jews (80). Animal studies The hypothesis that dietary fat or specific fatty acids can cause breast cancer originates from studies in rodents by Tannenbaum in 1942 (81). Since then, a large number of animal experiments showed that the amount and type of fat can markedly influence the growth of induced breast tumors in rodents during the promotion stage, but less so during the initiation stage (for a review see reference 82). In rats, a diet rich in polyunsaturated fat promoted growth of chemically induced (dimethylbenz[a]anthracene or N-nitrosomethylurea) or transplanted tumors more than did a diet rich in saturated fat (24, 83 86). Other studies showed that up to 4 5% of dietary linoleic acid in the diet promotes artificial mammary tumorigenesis in rats but that higher amounts have no additional effects; once the diet contained 4 5% of energy as linoleic acid, tumor yield and growth increased with the total amount of dietary fat, but saturated fats had the same effect as polyunsaturated fats (87 90). On the other hand, in experiments FIGURE 3. Breast cancer incidence in 11 European regions and Israel by the average linoleic acid content of adipose tissue in healthy subjects in the corresponding regions [control subjects from the EURAMIC (European Community Multicenter Study on Antioxidants, Myocardial Infarction, and Breast Cancer) study (79)].

4 LINOLEIC ACID AND CANCER RISK 145 with athymic nude mice that were injected with human breast cancer cells, diets containing 16% or 24% of energy as linoleic acid increased tumor weight and pulmonary metastasis compared with a diet containing the same amount of total fat but only 4% of energy as linoleic acid (91, 92). One study addressed the effect of long-term polyunsaturated fat intake on spontaneous breast tumors in rats and mice (93, 94). In a set of experiments, 3578 female rats and mice received by diet 10% of energy as fat and 2200 animals received by gavage additional corn oil that increased fat intake to 30% of energy. The 2-y incidence of spontaneous breast tumors was the same or somewhat higher with the low-fat diet (2.5% in the rats and 1.7% in the mice) than with the diet high in corn oil (1.5% in rats and 1.3% in mice). In summary, short-term experiments show that a minimum amount of linoleic acid is required to stimulate the growth of artificially induced mammary tumors in rats. Above this threshold, it appears to be the total amount of fat, or dietary energy (95 97), that promotes tumorigenesis, and that linoleic acid is as effective as other types of fatty acids. A high linoleic acid intake did stimulate carcinogenesis in 1 particular mouse model but a long-term high intake of linoleic acid did not increase spontaneous development of breast tumors in rats and mice (93, 94). Colorectal cancer Analytic studies within populations Case-control studies. The case-control studies from which we were able to extract quantitative estimates of risk showed no consistent association between intake of linoleic acid or polyunsaturated fat and colorectal cancer risk (Figure 4). Howe et al (108) combined the results of 13 studies on colorectal cancer and diet. Eleven studies involving a total of 5287 cases and control subjects provided intake data on polyunsaturated fat; 2 studies (103, 104) found a positive association and the other 9 studies (98 102, ) showed no or weakly inverse associations. The pooled odds ratio per 21.3 g polyunsaturated fat/d was 0.92 (95% CI: 0.85, 1.08) for FIGURE 4. Relative risks of colorectal cancer with high compared with low intakes of linoleic acid in case-control studies. The pooled estimates were derived from the study by Howe et al (108). Bars are 95% CIs. all subjects, 1.05 (95% CI: 0.90, 1.23) for the men, and 0.80 (95% CI: 0.66, 0.98) for the women. Case-control studies not included in the meta-analysis by Howe showed relative risks of colorectal cancer ( ) or adenomatous polyps (112) ranging from 0.29 to 1.63 with high compared with low polyunsaturated fat intakes (Figure 4). In case-control studies that did not report estimates of risk, colorectal cancer patients and control subjects consumed the same amount of polyunsaturated fat (113, 114). Three case-control studies used biomarkers to assess linoleic acid intake (11, 114, 115). These studies are not represented in Figure 4 because no estimates of relative risk were given. A small study from Scotland reported a somewhat lower proportion of linoleic acid in red blood cells of 20 colon cancer patients than in 20 control subjects (115); in other studies there were no differences between cases and control subjects in the proportion of linoleic acid in adipose tissue (11, 114) or red blood cells (114). Thus, the case-control studies analyzed showed no consistent association between linoleic acid intake and colorectal cancer risk. Prospective cohort studies. Prospective data on fat and colorectal cancer risk are limited; the relative risks (Figure 5) of colorectal cancer or adenomatous polyps in subjects with high compared with low intakes of linoleic acid were measured in 3 large cohorts from the United States (116, 117, ) and 1 from the Netherlands (118). In these 4 studies involving a total of 782 patients with colorectal cancer among persons, the risk of developing colon cancer during 3 6 y of follow-up was not associated with previously reported intakes of linoleic acid or polyunsaturated fat ( ). We calculated a combined relative risk of 0.92 (95% CI: 0.70, 1.22). Risk of adenomatous colorectal polyps in male health professionals who underwent endoscopy was nonsignificantly increased with a high polyunsaturated fatty acid intake (120) (Figure 5). Risk of hyperplastic colorectal polyps in the same population was nonsignificantly decreased with a high polyunsaturated fatty acid intake, as was the risk of hyperplastic colorectal polyps in the Nurses Health Study (121). The combined risk of adenomatous and hyperplastic colorectal polyps with high compared with low polyunsaturated fat intakes, based on a total of 564 cases among persons, was 1.06 (95% CI: 0.55, 2.05). Thus, these prospective cohort studies showed no association of polyunsaturated fat intake with the risk of colorectal cancer. Ecologic comparisons between populations International comparisons showed strong associations between per capita use of total fat and incidence or mortality from colorectal cancer (70, 71, ). However, unlike breast cancer, the association was consistently limited to saturated or animal fats; there was no association with polyunsaturated or vegetable fats (3, 75, 78, ). This finding agrees with hitherto unpublished prospective data from the Seven Countries Study. Between 1958 and 1964, men from 16 cohorts were enrolled in this study (125). Dietary information was collected at baseline from random samples of 8 49 men from each cohort. In 1987, food composites representing intake at baseline were collected locally and analyzed for fatty acids (126). The vital status of all men was verified after 25 y of follow-up (127). Linear regression analysis was used to relate average linoleic acid intake in the 16 cohorts with age-adjusted mortality rates from colorectal cancer [Interna-

5 146 ZOCK AND KATAN FIGURE 5. Relative risks of colon cancer and colorectal polyps with high compared with low intakes of linoleic acid in prospective cohort studies. Bars are 95% CIs. tional Classification of Diseases (ICD) ] and all cancers (ICD ). Intakes of dietary fiber and energy were added to the model as possible confounders. Average linoleic acid intake ranged from 8 g/d in Japan, Rome, and eastern Finland to 22 g/d in Belgrade, Serbia (Figure 6). Mortality from colorectal cancer was not associated with linoleic acid intake. Adjustment for dietary fiber and energy intakes did not change this result. Mortality from all cancers was also not associated with linoleic acid intake (D Kromhout, E Feskens, personal communication, 1996). Thus, population comparisons showed no association of linoleic acid consumption with the risk of colorectal cancer. Animal studies Results from animal experiments that studied the effect of dietary fat on artificially induced colon cancer varied with the model and methods used (for review see references 128 and 129). Diets high in polyunsaturated fat promoted tumor growth in rats after initiation of the tumors, but not during the initiation stage (130). Some studies in rats indicated that polyunsaturated n 6 fatty acids promote the growth of chemically induced colon tumors more so than do saturated fatty acids (23, 131, 132), but others showed no difference (133) or found that n 6 fatty acids promote tumor growth less so than do saturated fatty acids (134). In a meta-analysis of 14 rat studies (135), total fat intake adjusted for energy intake increased tumor growth in Fischer 344 rats, but not in Sprague-Dawley rats. In Fischer 344 rats, both n 6 polyunsaturated and saturated fats increased tumor growth more so than did monounsaturated or n 3 polyunsaturated fats (135). The minimum amount of dietary essential fatty acids needed to induce colon tumors in rats is 1% of energy (136), which is much lower than the 4.5% of energy needed to induce breast tumors (88). In mice, increasing amounts of dietary linoleic acid stimulated growth of transplanted colonic adenocarcinoma tumors up to 4% of energy intake, but not at higher intakes (137). Thus, there is some evidence that n 6 polyunsaturated fat promotes the growth of artificially induced colorectal tumors in rodents, but the data are inconsistent. FIGURE 6. Age-adjusted mortality from colorectal cancer after 25 y of follow-up in the Seven Countries Study, by linoleic acid intake at baseline in A description of the cohorts is as follows: A, US railroad; B, Belgrade, Serbia; C, Crevalcore, Italy; D, Dalmatia, Croatia; E, east Finland; G, Corfu, Greece; K, Crete, Greece; M, Montegiorgio, Italy; N, Zutphen, Netherlands; R, Rome railroad; S, Slavonia, Croatia; T, Tanushimaru, Japan; U, Ushibuka, Japan; V, Velika Krsna, Serbia; W, west Finland; Z, Zrenjanin, Serbia. Data were kindly provided by Daan Kromhout and Edith Feskens of the National Institute of Public Health and the Environment, Bilthoven, Netherlands. Prostate cancer Analytic studies within populations Case-control studies. Case-control studies of the relation between dietary fat and prostate cancer generally show a positive relation with total fat intake (for review see reference 138), but few studies collected data on specific fatty acids. For the 3 casecontrol studies from which we were able to extract quantitative estimates of risk ( ), involving a total of 654 cases and 924 control subjects, we calculated a combined relative risk of prostate cancer of 1.27 (95% CI: 0.97, 1.66) with high compared with low intakes of polyunsaturated fat (Figure 7). One study found an increased risk with high polyunsaturated fat intake in older men, but not in younger men (139). Another study found an increased risk with a high proportion of linoleic acid in erythrocyte membranes and fat tissue (141). A few studies measured polyunsaturated fat intake, but did not report estimates of risk. In 1 study (144, 145), prostate cancer patients consumed the same (144) or somewhat lower amounts of linoleic acid than did control subjects (145). Another study found no association between polyunsaturated fat intake and prostate cancer (146). In a study from Scotland, the polyunsaturated fat intake of 20 patients was higher than that of 20 control subjects according to a food-frequency questionnaire, but not according to the proportion of linoleic acid in red blood cells (115). Thus, the limited amount of data from case-control studies show either no or positive associations between linoleic acid intake and prostate cancer risk. Prospective cohort studies. Several prospective cohort studies have investigated dietary factors in relation to prostate cancer (for review see reference 138), but only 2 (142, 143) have investigated the relation between prostate cancer and polyunsaturated fat intake (Figure 7). In 2 cohorts, 1 of US health professionals (142) and 1 of US physicians (143), linoleic acid showed no or a weakly negative association with prostate cancer. The combined relative

6 LINOLEIC ACID AND CANCER RISK 147 FIGURE 7. Relative risks of prostate cancer with high compared with low linoleic acid intakes in case-control and prospective cohort studies. Bars are 95% CIs. risk, involving a total of 399 cases of prostate cancer among men, was 0.83 (95% CI: 0.56, 1.24). However, both studies found an increased risk of prostate cancer with higher intakes of -linolenic acid (18:3 n 3). Thus, these cohort studies suggest that prostate cancer risk may be associated with -linolenic acid, but not with linoleic acid. Ecologic comparisons between populations Population comparisons show positive associations between total fat disappearance and mortality or incidence rates of prostate cancer (3, 71, 75, 78, 147, 148). One study suggested that this association was due to animal rather than to vegetable fat (75), but others showed positive associations with both saturated and polyunsaturated fats (3, 78). Thus, there is some, but inconsistent, evidence from population comparisons that linoleic acid may be associated with prostate cancer risk. Animal studies There are few animal models for the study of prostate cancer (149). In 1 study of a genetically susceptible strain of rats treated with testosterone, animals consuming a diet to which corn oil had been added up to a level of 40% of energy from total fat spontaneously developed prostate tumors, whereas animals consuming a control diet with 10% of energy from fat did not (150). In another study of chemically induced prostate tumors in testosterone-treated rats, the amount or type of fat in the diet had no effect on tumor growth (151). Thus, animal experiments do not provide much evidence for a possible link between linoleic acid and prostate cancer risk. Cancer cases in controlled trials of diets high in linoleic acid Diets high in linoleic acid (> 12% of energy intake) have been used in intervention trials to determine their possible relation to CAD (Table 1). These trials may show whether a higher than normal incidence of cancer occurs when people consume high amounts of linoleic acid for 1 7 y. Men in the Los Angeles Veterans Administration trial (17, 157) who consumed a diet high in linoleic acid (14% of total energy) for 6.5 y had higher cancer mortality than men who consumed the standard institutional diet, which was high in saturated fat and low in linoleic acid (4% of total energy). Eighteen of the 31 deaths from carcinoma during the diet phase in the experimental group occurred in men who did not adhere closely to the diet (adherence score of < 50%). Ederer et al (152) combined the Los Angeles cancer data with those of 4 other trials that studied the effects of diets with a high content of linoleic acid ( ) (Table 1). When data from the Veterans Administration trials were excluded, the relative risks in the groups with a high linoleic acid intake as compared with the control groups were 0.75 for cancer incidence and 0.62 for cancer mortality. When data from the Veterans Administration trial were included, the relative risk became 1.15 for cancer incidence and 1.08 for cancer mortality. In a more recent, large, 4.5-y trial with institutionalized men and women, 23 cancer deaths occurred among 4541 participants who consumed a diet containing 15% of energy as linoleic acid, whereas 20 of the 4516 subjects consuming a diet containing 5% of energy as linoleic acid died of cancer (159). Thus, cancer mortality was not higher in the high linoleic acid group than in the control group. However, unlike in previous trials, in this trial there was no effect of linoleic acid intake on the incidence of CAD. This finding sheds some doubt on whether the duration and the intensity of the dietary treatment were sufficient. In summary, 1 trial of diet and CAD showed an increased incidence of cancers with a very high intake of linoleic acid during several years, whereas 5 other trials of diet and CAD did not. However, each of these trials had serious methodologic limitations in terms of adherence, duration, and small number of cancer cases. DISCUSSION A definitive answer to the question of whether high intakes of linoleic acid increase the risk of cancer requires a randomized, controlled trial in which thousands of people consume a diet either high or low in linoleic acid for their entire lives. Because such a trial is obviously not feasible, one has to rely on other sources of evidence for the answer. Current evidence is summarized in Table 2. We found no strong evidence indicating that a diet high in linoleic acid or polyunsaturated fat increases the risk of breast, colorectal, or prostate cancer. Analytic studies within populations showed no consistent relation between linoleic acid intake and cancer risk. Some ecologic comparisons between populations showed associations between vegetable or polyunsaturated fat intake and incidence or mortality rates for breast and prostate cancers, but not for colorectal cancer. With 1 exception, trials studying the effect of linoleic acid on incidence of CAD in men do not suggest that a high linoleic acid intake for 1 7 y raises cancer risk. Several, but not all, animal experiments have indicated that linoleic acid promotes the growth of artificially induced breast and colorectal tumors in rodents. Evidence from analytic studies within populations The findings from case-control and prospective cohort studies do not categorically exclude an influence of linoleic acid on cancer risk. There are several reasons why true underlying effects may have been missed or obscured. Recall bias may occur in case-control studies. For example, if patients underreported their linoleic acid intake, a positive association between linoleic acid intake and cancer incidence may have been obscured. However, case-control studies that objectively assessed intake by examining tissue fatty acid contents also

7 148 ZOCK AND KATAN TABLE 1 Cancer incidence in clinical trials of high linoleic acid diets and coronary artery disease in men 1 Duration of study Number of cancer cases Study Linoleic acid intake (diet + follow-up) Experimental Control Relative risk (95% Cl) 2 % of energy y Oslo Diet-Heart Study (153) /206 7/ (0.5, 3.8) Medical Research Council soybean oil (154) /199 8/ (0.1, 1.2) Finnish Mental Hospital (155) /327 4/ (0.2, 3.1) Faribault (156) /167 1/ (0.0, 5.5) Los Angeles Veterans (157) /424 52/ (0.9, 1.9) Combined / / (0.7, 1.4) 1 Adapted from the study by Ederer et al (152), who estimated a combined relative risk of 1.15 (95% Cl: 0.81, 1.63) by the method of Gart (158). 2 High compared with low linoleic acid intakes. 3 Mean. 4 Crude risk ratio of the pooled data. found no associations between linoleic acid intake and cancer risk. Furthermore, prospective cohort studies, which are not subject to recall bias, also showed no associations. Thus, recall bias is not the most likely explanation for the absence of associations in case-control studies. Errors in measurements of dietary intake in epidemiologic studies are generally large. Random (nondifferential) measurement error attenuates estimates of relative risk toward 1. Most of the studies reviewed here did not correct risk estimates for this type of error, but Hunter et al (59) found that correction had little effect on the outcome of their pooled analysis of cohort studies on breast cancer. This suggests that this type of error cannot totally explain the absence of associations. However, Prentice (160) criticized the existing methods for correction and argued that dietary measurement error may also be systematic (ie, not random). He concluded that dietary self-report instruments may be inadequate for analytic studies. On the other hand, assessment of dietary intake with biomarkers avoids such systematic error, and, except for 1 study (141), studies that used biomarkers also found no (45, 46) or negative (44, 63, 161) associations between linoleic acid intake and cancer incidence. It is unclear to what extent dietary measurement error may have affected the estimates of risk in studies that relied on self-reported intake. A narrow range of linoleic acid intake in the population may be another reason why associations were not found. In the studies reviewed here, linoleic acid intake in the highest category was typically twice that in the lowest category of intake, with differences ranging from 5 to 25 g/d. This is comparable with the normal range of intake of 4 10% of daily energy. However, it cannot be excluded that cancer risk within populations would be affected by larger differences in linoleic acid intake. Yet another reason for not finding a significant association in an individual epidemiologic study is the limited number of subjects, which results in low statistical power. However, the combined risk estimates presented here, involving large numbers of patients, also did not show associations between linoleic acid intake and cancer risk. Confounding must also be considered. For example, people who consume high amounts of linoleic acid may also consume high amounts of vegetables and fruit, which might obscure increases in colorectal cancer risk caused by a high linoleic acid intake. Possible confounders in studies of breast cancer are body weight and intake of total energy or of other fatty acids. Such confounders are often not taken into account or cannot be completely adjusted for. Thus, bias of risk estimates by confounding factors cannot be excluded. Also, linoleic acid intake reflects the use of vegetable oils such as rapeseed and soybean oils. It remains possible that associations reported for linoleic acid or polyunsaturated fat are affected by other substances from these oils. Publication bias is not plausible because studies showing positive associations between linoleic acid intake and cancer risk would be more likely to be published than would studies showing negative associations. Many of the analytic studies reviewed measured the intake of all polyunsaturated fats and not linoleic acid intake per se. Polyunsaturated fat in human diets consists mainly of linoleic acid, but it also includes -linolenic acid and long-chain n 3 fatty acids from fish oil. It has been suggested that these n 3 fatty acids may have specific effects on cancer risk ( ). If so, then a risk estimate for polyunsaturated fat would not be the same as a risk estimate for linoleic acid. However, studies that measured polyunsaturated fat intake and studies that specifically measured linoleic acid intake showed no consistent associations with cancer risk. Therefore, it is plausible that differences in polyunsaturated fat intake reflected differences in linoleic acid intake, and that n 3 fatty acids did not materially affect the risk estimates for polyunsaturated fat. Thus, there are several methodologic reasons analytic studies within populations may have missed a possible association between linoleic acid intake and cancer risk. The question becomes whether such studies, when applying similar methods in similar populations, can at all detect associations between fatty acid intake and disease risk. For example, do within-population studies reveal the expected relation between linoleic acid intake and the risk of CAD? Indeed, a substantial proportion of prospective cohort studies did show inverse associations between polyunsaturated fat intake and the risk of CAD ( ), although the remainder of the studies did not ( ). Thus, epidemiologic studies of this type should be able to detect an association between linoleic acid intake and disease risk, if one exists. In addition, several of the studies described above showed associations between cancer risk and food components other than linoleic acid, such as animal fat and red meat (107, 116). Therefore, if a substantial association between linoleic acid intake and cancer risk exists, it is unlikely that virtually all analytic studies would fail to find such an association purely because of methodologic limitations. However, the evidence from these analytic studies cannot exclude the possibility of a small increase in cancer risk with high intakes of linoleic acid. Evidence from other types of studies Ecologic comparisons Ecologic studies compare average cancer rates of countries or regions rather than risks of individuals. Such studies have major

8 LINOLEIC ACID AND CANCER RISK 149 TABLE 2 Summary of the evidence concerning the relation between linoleic acid intake and cancer risk 1 Cancer site Type of study Breast Colon Prostate All Epidemiologic studies Within-population studies (combined relative risk) Case-control study Cohort study Comparisons between populations +/= = +/= Controlled clinical trials on coronary artery disease Animal studies +/= +/= NA = 1 +/=, inconsistent positive association; =, no association; NA, not available. limitations. There are multiple differences between populations that may affect cancer risk and there is no or only a limited possibility to adjust for these confounders. Also, the use of per capita use data to assess dietary intake is questionable, and the quality of mortality or incidence data from national cancer registrations may be different for different countries, which may bias the associations. Nevertheless, ecologic comparisons offer the advantage of involving large differences in intakes between populations. Furthermore, some ecologic comparisons assess intake by aggregating individual intake data and in this way eliminate random dietary measurement error. Two ecologic comparisons that assessed intake by aggregating individual intake or biomarker data showed no association between linoleic acid and cancer risk despite large differences in linoleic acid intake (79) (Figures 3 and 6). Studies that assessed linoleic acid intake from per capita use data indicated that saturated and animal fats rather than polyunsaturated fat or linoleic acid may play a role in increased cancer risk. Thus, the ecologic comparisons we analyzed did not suggest a major influence of linoleic acid on the risk of breast or colorectal cancer. Controlled trials of high linoleic acid diets Controlled trials that were designed to study the effects of linoleic acid on CAD did not show an increased number of cancer cases in groups of subjects that consumed high amounts of linoleic acid (16% of energy intake) for 1 7 y. These trials involved mostly men and therefore provide no information on breast cancer. Also, the number of cases was small and the exposure period may have been too short to show any effect of linoleic acid intake on cancer incidence. Thus, the strength of the evidence from controlled trials of high linoleic acid diets is limited. Nevertheless, except for 1 trial (17), the findings do not support the hypothesis that a high linoleic acid intake increases cancer risk. Animal experiments The relevance of short-term experiments in animals with artificially induced tumors to the development of human cancers over decades is unclear. Profound and consistent effects of linoleic acid on initiation of tumorigenesis have not been shown (82). Linoleic acid promotes tumor growth in rodents only under certain conditions and in certain models (129, 175). For example, breast tumors induced by hormones were less responsive to dietary fat than were those induced by chemicals (176), and dietary fat increased colorectal tumor growth in 1 strain of laboratory rats but not in another (135). Nevertheless, there appears to be a minimum required intake of linoleic acid for tumor development in rodents. However, this requirement is about as low or lower than intakes recommended to prevent deficiency of essential fatty acids, and reducing linoleic acid intake to below these amounts is neither realistic nor desirable. We feel that experiments in rodents are of limited value for addressing the question of whether a life-long high intake of linoleic acid increases the risk of spontaneous cancers in humans. Implications The available evidence does not suggest that a high intake of linoleic acid substantially raises the risk of breast, colorectal, or prostate cancer. Nevertheless, a small increase in risk cannot be excluded. When applied to the total population, a small increase in risk may still have a large effect on public health. For example, an increase in breast cancer risk of 10% would increase the number of breast cancer cases occurring in the United States each year by (177). A relevant question for public health policy is whether a possible adverse effect of linoleic acid would be outweighed by beneficial effects on CAD risk. Replacement of saturated fatty acids with oleic acid instead of linoleic acid would also reduce CAD risk and at the same time avoid worries about any possible increase in cancer risk; however, it is not clear whether oleic acid is as effective as linoleic acid in improving the serum lipid profile and reducing CAD risk (15, 178). Therefore, future studies should determine the benefit of linoleic acid compared with that of oleic acid on CAD risk. If such studies show that linoleic acid and oleic acid have similar effects on CAD risk, then it may still be prudent to restrict linoleic acid intake. However, the currently available evidence does not provide a compelling argument for such a restriction. We thank E Feskens, D Kromhout, and M Jansen for providing data from the Seven Countries Study, and N Boyd, L Holmberg, G Howe, D Hunter, R Kaaks, P Toniolo, and A Wolk for supplying additional information about their studies. REFERENCES 1. Doll R, Peto R. The causes of cancer: quantitative estimates of the avoidable risks of cancer in the United States today. J Natl Cancer Inst 1981;66: National Research Council, Committee on Diet, Nutrition, and Cancer. Diet, nutrition, and cancer. Washington, DC: National Academy of Sciences, 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. Cancer Causes Control 1990;1: Sasaki S, Horacsek M, Kesteloot H. An ecological study of the relationship between dietary fat intake and breast cancer mortality. Prev Med 1993;22: Joint Experts. Fats and oils in human nutrition. Report of a joint expert consultation. Rome: FAO/WHO, 1994: McDowell MA, Briefel RR, Alalmo K, et al. Energy and macronutrient intakes of persons ages 2 months and over in the United States: third National Health and Nutrition Examination Survey, phase 1, Vital Health Stat 1994;255: Ministry of Health, Welfare, and Sports, Ministry of Agriculture, Nature Management, and Fisheries. Zo eet Nederland, Resulaten van de Voedselconsumptiepeiling. (The second Dutch

9 150 ZOCK AND KATAN National Food Consumption Survey, 1992.) The Hague: Dutch Bureau for Nutrition Education, 1993 (in Dutch). 8. Gregory J, Foster K, Tyler H, Wiseman M. The dietary and nutritional survey of British adults. A survey of dietary behavior, nutritional status and blood pressure of adults aged 16 to 64 living in Great Britain. London: Office of Population Censuses and Surveys, Ministry of Agriculture, Fisheries and Food and the Department of Health, Seidell JC, Cigolini M, Deslypere J-P, Charzewska J, Ellsinger B- M. Polyunsaturated fatty acids in adipose tissue in European men aged 38 years in relation to serum lipids, smoking habits, and fat distribution. Am J Epidemiol 1991;134: Kesteloot H, Geboers J, Joossens JV. On the within-population relationship between nutrition and serum lipids: the B.I.R.N.H. study. Eur Heart J 1989;10: Berry EM, Zimmerman J, Peser M, Ligumsky M. Dietary fat, adipose tissue composition, and the development of carcinoma of the colon. J Natl Cancer Inst 1986;77: Friedlander Y, Kark JD, Kaufmann NA, Stein Y. Coronary heart disease risk factors among religious groupings in a Jewish population sample in Jerusalem. Am J Clin Nutr 1985;42: Steinberg D, Blumenthal S, Carleton RA, et al. Consensus conference. Lowering blood cholesterol to prevent heart disease. JAMA 1985;253: The Expert Panel. Report of the National Cholesterol Education Program Expert Panel on Detection, Evaluation and Treatment of High Blood Cholesterol in Adults. Arch Intern Med 1988;148: Mensink RP, Katan MB. Effect of dietary fatty acids on serum lipids and lipoproteins: a meta-analysis of 27 trials. Arterioscler Thromb 1992;12: Katan MB, Grundy SM, Willett WC. Clinical debate: should a lowfat, high-carbohydrate diet be recommended for everyone? Beyond low-fat diets. N Engl J Med 1997;337: Pearce ML, Dayton S. Incidence of cancer in men on a diet high in polyunsaturated fat. Lancet 1971;1: Frantz ID Jr. Lipids and atherosclerosis. Cancer Res 1981;41: Grundy SM, Denke MA. Dietary influences on serum lipids and lipoproteins. J Lipid Res 1990;31: Felton CV, Crook D, Davies MJ, Oliver MF. Dietary polyunsaturated fatty acids and composition of human atherosclerotic plaques. Lancet 1994;344: Rothacker DQ, Ellis PK. Linoleic acid supplementation in slimming products is unnecessary and may pose a potential added risk to the user population. Int J Obes 1995;19: Groves BA. Exit linoleic acid too? Lancet 1996;347:58 (letter). 23. Broitman SA, Vitale JJ, Vavrousek-Jakuba E, Gottlieb LS. Polyunsaturated fat, cholesterol and large bowel tumorigenesis. Cancer 1977;40: Carroll KK, Khor HT. Effects of level and type of dietary fat on incidence of mammary tumors induced in female Sprague-Dawley rats by 7,12-dimethylbenzanthracene. Lipids 1971;6: Welsch CW. Enhancement of mammary tumorigenesis by fat: review of potential mechanisms. Am J Clin Nutr 1987;45: Fang JL, Vaca CE, Valsta LM, Mutanen M. Determination of DNA adducts of malonaldehyde in humans: effects of dietary fatty acid composition. Carcinogenesis 1996;17: Steinberg D, Parthasarathy S, Carew TE, Khoo JC, Witztum JL. Beyond cholesterol. Modifications of low-density lipoprotein that increase its atherogenicity. N Engl J Med 1989;320: National Cholesterol Education Program. Second report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel II). Circulation 1994;89: Krauss RM, Deckelbaum RJ, Ernst N, et al. Dietary guidelines for healthy American adults. A statement for health professionals from the Nutrition Committee, American Heart Association. Circulation 1996;94: Dersimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7: Katsouyanni K, Willett W, Trichopoulos D, et al. Risk of breast cancer among Greek women in relation to nutrient intake. Cancer 1988;61: Rohan TE, McMichael AJ, Baghurst PA. A population-based casecontrol study of diet and breast cancer in Australia. Am J Epidemiol 1988;128: Toniolo P, Riboli E, Protta F, Charrel M, Cappa APM. Calorie-providing nutrients and risk of breast cancer. J Natl Cancer Inst 1989;81: Shun-Zhang Y, Rui-Fang L, Da-Dao X, Howe GR. A case-control study of dietary and nondietary risk factors for breast cancer in Shanghai. Cancer Res 1990;50: van t Veer P, Kok FJ, Brants HAM, Ockhuizen T, Sturmans F, Hermus RJJ. Dietary fat and the risk of breast cancer. Int J Epidemiol 1990;19: Lee HP, Gourley L, Duffy SW, Esteve J, Lee J, Day NE. Dietary effects on breast-cancer risk in Singapore. Lancet 1991; 337: Zaridze D, Lifanova Y, Maximovitch D, Day NE, Duffy SW. Diet, alcohol consumption and reproductive factors in a case-control study of breast cancer in Moscow. Int J Cancer 1991;48: Richardson S, Gerber M, Cenée S. The role of fat, animal protein and some vitamin consumption in breast cancer: a case control study in southern France. Int J Cancer 1991;48: Ingram DM, Nottage E, Roberts T. The role of diet in the development of breast cancer: a case-control study of patients with breast cancer, benign epithelial hyperplasia and fibrocystic disease of the breast. Br J Cancer 1991;64: 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 1994;70: Holmberg L, Ohlander EM, Byers T, et al. Diet and breast cancer risk. Results from a population-based, case-control study in Sweden. Arch Intern Med 1994;154: Yuan J-M, Wang Q-S, Ross RK, Henderson BE, Yu MC. Diet and breast cancer in Shanghai and Tianjin, China. Br J Cancer 1995;91: Franceschi S, Favero A, Decarli A, et al. Intake of macronutrients and risk of breast cancer. Lancet 1996;347: Zaridze DG, Chevchenko VE, Levtshuk AA, Lifanova YE, Maximovitch DM. Fatty acid composition of phospholipids in erythrocyte membranes and risk of breast cancer. Int J Cancer 1990;45: London SJ, Sacks FM, Stampfer MJ, et al. Fatty acid composition of the subcutaneous adipose tissue and risk of proliferative benign breast disease and breast cancer. J Natl Cancer Inst 1993;85: Petrek JA, Hudgins LC, Levine B, Ho M, Hirsch J. Breast cancer risk and fatty acids in the breast and abdominal adipose tissues. J Natl Cancer Inst 1994;86: Boyd NF, Martin LJ, Noffel M, Lockwood GA, Tritchler DL. A meta-analysis of studies of dietary fat and breast cancer risk. Br J Cancer 1993;68: Howe GR, Hirohata T, Hislop TG, et al. Dietary factors and risk of breast cancer: combined analysis of 12 case-control studies. J Natl Cancer Inst 1990;82: Katan MB, Deslypere JP, van Birgelen APJM, Penders M, Zegwaard M. Kinetics of the incorporation of dietary n 3 fatty acids into serum cholesteryl esters, erythrocyte membranes, and adipose tissue an 18-month controlled trial in man. J Lipid Res 1997;38: