Trans-fatty acids in the diet

Trans-FA are unsaturated fatty acids with one or more of their double bonds in the ‘trans’ rather than the common ‘cis’ configuration. They occur naturally at low levels in dairy products and meats from ruminant animals, but are also produced during the industrial hydrogenation of vegetable oils. The latter process is responsible for many of the semi-solid and solid fats that are widely available as spreads (for example, margarines) and used as a more solid fat source in commercial baked goods (biscuits, pastries) and catering outlets. The predominant trans-FA isomers are vaccenic acid (trans-18 : 1n-7) and elaidic acid (trans-18 : 1n-9), although other MUFA (14 : 1, 16 : 1, 18 : 1) and PUFA (18 : 2 and 18 : 3) isomers may also be present at detectable levels in both food products and tissue samples.

The average adult (19–64 years) intake of trans-FA in the UK was reported to be 1·2 % energy in 2000–1⁽Reference Henderson, Gregory and Irving⁴⁾. Recent re-estimation of intake, using dietary data from the National Diet and Nutrition Survey (2000–1) and updated trans-FA composition data, has given an estimated average value of 1·0 % energy for current trans-FA intake in the UK adult population⁽Reference Williams, Thompson and Shaw³⁾. This is significantly lower than in the USA, where the average trans-FA intake was 2·6 % energy in 2003⁽⁵⁾. In Europe, intakes in 1995 ranged from 0·5 to 2·1 %, with the lowest intake levels in the Mediterranean countries and the highest intakes in Iceland⁽Reference Hulshof, van Erp-Baart and Anttolainen⁶⁾.

Assessment of trans-fatty acid intake

The most common methods of estimating trans-FA intake are through FFQ or measuring the levels of trans-FA in tissues. A number of fatty acids are primarily sourced from the diet and cannot be synthesised de novo in the body, including the essential fatty acids linoleic acid (18 : 2n-6) and linolenic acid (18 : 3n-3), the long-chain n-3 PUFA (only limited synthetic capacity) and trans-FA. Levels of these fatty acids in tissue lipid pools are considered to be a more reliable assessment of dietary exposure than dietary questionnaires. This may be particularly important for the relative quantification of trans-FA exposure, as the inability of food composition databases to keep pace with rapidly changing compositions of foods, which include partially hydrogenated vegetable oil sources, may result in significant inaccuracies in estimations of dietary trans-FA intake levels. Trans-FA levels in plasma lipid pools such as cholesteryl esters and phospholipids, in platelets and erythrocytes, and in adipose tissue, may be used as indices of integrated exposure over the previous few days, few months and 1–2 years respectively. However, some disease states result in changes in fatty acid metabolism that are reflected in altered tissue compositions, which could confound retrospective case–control comparisons. In addition, because fatty acid compositional data are expressed as percentage of total fatty acids, any increase or decrease in one fatty acid will result in reciprocal changes in one or more of the other fatty acids present. A further limitation is the inability to directly quantify intake from tissue fatty acid composition. Therefore tissue fatty acid composition data cannot be directly translated into public health recommendations for dietary intake levels associated with reduced risk of disease.

Some prospective studies reported in the present review have undertaken only a single measure of diet at baseline, which may lead to misclassification of habitual trans-FA intake in long-term follow-up, due to changes in subjects' habitual diets over time, as well as changes in food formulations. Given the significant recent changes in fat formulation because of the efforts made by food manufacturers to replace trans-FA in oils and spreads, reliance on historical estimates of intake may lead to large inaccuracies in estimating trans-FA-disease associations.

Presentation of statistical data

Throughout the present review, the OR and relative risk (RR) stated are those that have been fully adjusted for potential confounding factors. The P for trend values refer to the level of difference between the highest and lowest tertile, quartile or quintile of intake.

When a significant difference in an associated risk was reported, the specific factors that were adjusted for in the model have been shown within the tables. Standard confounding factors for which data have normally been adjusted for include age, BMI and family history. In the case of breast cancer, adjustment for a large number of site-specific risk factors relating to lifetime hormonal exposure (for example, age at menses, menopause, hormone replacement therapy use, parity, age at first pregnancy, breast-feeding), adiposity, height and alcohol intake are also normally applied.

Dietary trans-FA intakes positively correlate with dietary SFA and total fat intakes. Since total and SFA intakes have been associated with risk of cancers at some sites, it would be expected that epidemiological studies designed to assess the association between dietary trans-FA intake and cancer risk fully adjust for potential confounding by these dietary variables⁽Reference Margetts and Nelson^{7, 8)}. Details of adjustments made by each study are given in the tables. The tables show the adjusted RR or OR; however, where adjustment for another dietary variable significantly alters the adjusted RR or OR, this is noted within the text.

Animal studies

Selenskas et al. ⁽Reference Selenskas, Ip and Ip⁹⁾ investigated the effect of a high-trans-FA diet on a dimethylbenz[a]anthracene-induced rat mammary tumour model. The diets contained 20 % fat by weight, which was either a partially hydrogenated mixture of 50 % soyabean oil and 50 % cottonseed oil (57·5 % 18 : 1; 22·5 % cis-monoene and 35 % trans-monoene) or a mixture of 58 % olive oil, 40 % cocoa butter and 2 % coconut oil (54·7 % 18 : 1; all in cis configuration). Apart from differences in trans-FA, the diets contained similar levels of other fatty acids. Tumour incidences for the trans- and cis-fatty acid diets were 32 and 40 %, respectively, with no statistical difference between groups.

Using the same blend of trans- and cis-fatty acids, Erickson et al. ⁽Reference Erickson, Schlanger and Adams¹⁰⁾ examined effects of the diets on the growth and metastasis of implanted mammary tumour cells. The study also considered varying amounts of fat, with diets containing either 5 or 20 % fat by weight. Cells were injected into female BALB/c mice either subcutaneously or intravenously. Mice with subcutaneous implants showed no differences in latency period, tumour growth rate or final tumour size, regardless of fat type or amount. However, in mice receiving the intravenous implants, the liver and spleen from those fed the cis-fatty acid diets contained significantly more viable tumour cells than did those from mice fed the trans-FA diets.

The desaturation of trans-18 : 1n-7 to form CLA has been observed in rodents, ruminants and human subjects⁽Reference Lock, Corl and Barbano¹¹⁾, with reported conversion rates in rodents of 5–12 %⁽Reference Santora, Palmquist and Roehrig¹²⁾ and 19 % in humans⁽Reference Turpeinen, Mutanen and Aro¹³⁾. Lock et al. ⁽Reference Lock, Corl and Barbano¹¹⁾ observed that when trans-18 : 1n-7 is converted to CLA it exerts an anticarcinogenic effect against rat mammary tumour initiation and growth. However, the presence of trans-18 : 1n-7 does not affect cancer development when this conversion is blocked.

Ecological studies

The methodology and findings for the human studies examining the association between trans-FA and breast cancer are summarised in Table 1.

Table 1 Human studies investigating the association of trans-fatty acids (trans-FA) with breast cancer

RR, relative risk; FA, fatty acid; N/A, not applicable; Q1, quartile 1 or quintile 1; Q5, quintile 5; EURAMIC, European Community Multicentre Study on Antioxidants, Myocardial Infarction, and Breast Cancer; T1, tertile 1; T3, tertile 3; NR, not reported; HRT, hormone replacement therapy; NHS, Nurses' Health Study; ORDET, Hormone and Diet Etiology of Breast Cancer; NHS II, Nurses' Health Study second phase; EPIC, European Prospective Investigation into Cancer and Nutrition; Q4, quartile 4.

* Risk factors for breast cancer: age at time of study, age at menarche, age at first birth, height, family history of breast cancer and parity.

† Standard confounding factors: energy intake and alcohol consumption.

Bakker et al. ⁽Reference Bakker, van't Veer and Zock¹⁴⁾ investigated the association of breast cancer incidence and trans-FA status across eleven populations. A statistically significant correlation was found between trans-FA and the incidence of breast cancer, with a Pearson correlation coefficient (r) of 0·89. An increase of 1 g trans-FA per 100 g fatty acids in adipose tissue corresponded to a rise in incidence of 19·3 cases of breast cancer per 100 000 person-years.

Case–control studies

London et al. ⁽Reference London, Sacks and Stampfer¹⁵⁾ analysed the gluteal adipose tissue of 380 US women with newly diagnosed stage I or II breast cancer and 176 with proliferative benign breast disease. Although trans-FA levels in adipose tissue showed no statistically significant association with the risk of breast cancer, this study used controls subjects who may have had breast abnormalities, which may have biased the findings. Petrek et al. ⁽Reference Petrek, Hudgins and Levine¹⁶⁾ also compared the breast and abdomen tissue fatty acid profiles of women with invasive breast cancer with those of women with a negative diagnosis for breast cancer. No differences in trans-FA concentration were found between the case and control groups for either tissue type.

The European Community Multicentre Study on Antioxidants, Myocardial Infarction, and Breast Cancer (EURAMIC) study⁽Reference Kohlmeier, Simonsen and van 't Veer¹⁷⁾ compared gluteal adipose tissue trans-FA levels in 698 cases of postmenopausal primary breast cancer and matched controls. There was a strong positive correlation between the adipose tissue level of trans-FA and breast cancer (OR 1·40). The relationship between trans intake and breast cancer appeared to be modified by PUFA intake. OR after stratification by PUFA tertiles were 3·65 and 0·97 for the lowest and highest PUFA tertile, respectively.

Aro et al. ⁽Reference Aro, Männistö and Salminen¹⁸⁾ reported serum fatty acid levels in 195 cases of breast cancer and 208 population-based controls identified from pre- and postmenopausal Finnish women. The authors reported an inverse association between trans-18 : 1n-7 and risk of breast cancer with an OR for the lowest v. the highest quintiles of 0·2, but the authors failed to report the statistical significance of this trend. Other trans isomers showed no differences between the case and control populations.

Prospective studies

Holmes et al. ⁽Reference Holmes, Hunter and Colditz¹⁹⁾ assessed the links between diet and breast cancer in 88 795 pre- and postmenopausal women over 14 years in the Nurses' Health Study. There was a small inverse association between trans-FA intake and risk of breast cancer, although this was not statistically significant. Additional analysis of the data, including postmenopausal women who had not reported a diagnosis of benign breast disease, also found no association between trans-FA and breast cancer⁽Reference Byrne, Rockett and Holmes²⁰⁾. Similar analysis using 90 655 premenopausal women enrolled in the second phase of the Nurses' Health Study again showed no relationship between the disease and dietary trans-FA intake⁽Reference Cho, Spiegelman and Hunter²¹⁾. The most recent report on the original cohort brings total follow-up to 20 years, and again reports no association between trans-FA and breast cancer⁽Reference Kim, Willett and Colditz²²⁾.

Pala et al. ⁽Reference Pala, Krogh and Muti²³⁾ conducted a prospective study of erythrocyte fatty acids and pre-diagnostic breast cancer (the Hormone and Diet Etiology of Breast Cancer (ORDET) study) in northern Italy. The 4052 postmenopausal participants were followed for an average of 5·5 years, with each case of breast cancer matched with two randomly selected controls. Oleic acid and MUFA were positively associated with the risk of breast cancer, but trans-18 : 1n-9 (the only trans-FA reported) was not associated with breast cancer risk.

Serum fatty acid concentrations from 197 pre- and postmenopausal cases of breast cancer and matched population-based controls in the New York Women's Health Study were assessed by Saadatian-Elahi et al. ⁽Reference Saadatian-Elahi, Toniolo and Ferrari²⁴⁾. The authors found no association between risk of developing breast cancer and serum levels of trans-18 : 1n-9. No other trans-FA were evaluated in the study.

Voorrips et al. ⁽Reference Voorrips, Brants and Kardinaal²⁵⁾ analysed the data from dietary questionnaires of 941 cases of breast cancer and 1598 subcohort controls in the Netherlands Cohort Study. There was a significant positive association between increasing trans-FA intake and risk of breast cancer (P for trend = 0·01), although the 95 % CI for the lowest v. the highest quintile of dietary intake of trans-FA encompassed 1·0 (RR 1·30; 95 % CI 0·93, 1·80). Further analysis to consider the effect of trans-18 : 1n-7 showed similar results, with a highly significant trend across the quintiles of intake (P for trend = 0·006). However, the 95 % CI for individual quintiles were consistently non-significant (RR 1·34; 95 % CI 0·98, 1·82 for the lowest v. the highest quintile).

Rissanen et al. ⁽Reference Rissanen, Knekt and Jarvinen²⁶⁾ studied the relationship between serum trans-FA and risk of breast cancer in 127 incident breast cancer cases and 242 matched population-based controls from a prospective study of 8196 women in Finland. Higher serum trans-18 : 1n-7 levels were associated with an increased risk of breast cancer (OR 3·69), increasing after further adjustment for BMI, serum cholesterol, alcohol intake, education, exercise and parity (OR 4·23). The relationship appeared to be stronger in postmenopausal than in premenopausal women, but the trend failed to reach statistical significance in either group. There was no significant correlation with total MUFA trans-FA.

A subset of women was randomly selected from a large trial (266 064 women) of breast self-examination in Shanghai by Shannon et al. ⁽Reference Shannon, King and Moshofsky²⁷⁾. Erythrocyte fatty acids were analysed in 322 cases of breast cancer and 367 controls, matched for age and menstrual status. A strong positive association was found between the concentration of trans-18 : 1n-7 in the erythrocytes and breast cancer (OR 2·21). No other trans-FA were included in the study.

The association between serum phospholipid trans-FA and breast cancer risk during 7 years of follow-up was assessed in the French component of the European Prospective Investigation into Cancer and Nutrition (EPIC) study⁽Reference Chajes, Thiebaut and Rotival²⁸⁾ (19 934 women). The 363 cases of incident invasive breast cancer were matched to two controls by age, menopausal status at blood collection, whether the subjects were fasting before blood collection, date and collection centre. Increasing levels of serum phospholipid trans-16 : 1n-7, and combined trans-16 : 1n-7 and trans-18 : 1n-9 were correlated with risk of breast cancer (OR 2·24 and 1·70, respectively). The levels of cis isomers were unrelated to breast cancer risk.

Summary of evidence for an association between trans-fatty acids and breast cancer

Prospective cohort, case–control and ecological studies reported between 1994 and 2006 were reviewed. Most of the prospective studies assessed trans-FA exposure by dietary questionnaire; one nested case–control study was based on analysis of erythrocyte fatty acids while another two analysed serum. The five case–control studies used either serum or adipose tissue fatty acids as the measure of exposure. Data from the prospective studies that evaluated dietary trans-FA intake are summarised in Fig. 1.

Fig. 1 Risk of breast cancer from prospective epidemiological studies that evaluated dietary intake of trans-fatty acids (trans-FA): Voorips et al. ⁽Reference Voorrips, Brants and Kardinaal²⁵⁾ (■) and Cho et al. ⁽Reference Cho, Spiegelman and Hunter²¹⁾ (Δ). The current mean trans-FA intake for women in the UK has recently been estimated to be 1·0 % of total energy or 1·7 g/d⁽Reference Williams, Thompson and Shaw³⁾ (). Risk of breast cancer is plotted as the relative risk (RR), with vertical bars representing the 95 % CI for intake ranges above reference (RR = 1) in each study. Data from Cho et al. ⁽Reference Cho, Spiegelman and Hunter²¹⁾ were converted from percentage of total energy to g/d based on an energy intake of 1700 kcal/d (7113 kJ/d).

One of four case–control studies showed a strong positive association between trans levels in tissues and risk of breast cancer. One study reported a negative association between serum vaccenic acid and breast cancer but failed to report values for statistical significance. Three of the ten prospective studies have shown a positive association, with two of these based on measurement of serum or erythrocyte trans-FA levels and one based on dietary trans-FA assessment. For the latter study, the OR was 1·3 for a highest v. lowest quintile range of 1·5–3·6 g/d, which is similar to the current range of intakes in the UK⁽Reference Williams, Thompson and Shaw³⁾. However, there was no significant association found in the three large studies (the Nurses' Health Study, the Hormone and Diet Etiology of Breast Cancer (ORDET) study and the New York Women's Health Study), which have had up to 20 years of follow-up. Overall, the evidence for an association between trans-FA intake and breast cancer is weak. There are limited data from animal studies, and the findings do not provide any evidence to support an effect of trans-FA consumption on mammary tumorigenesis.

Animal studies

Watanabe et al. ⁽Reference Watanabe and Koga²⁹⁾ used the dimethyihydrazine model to examine the effect of trans-FA on colon cancer in Fischer rats. A partially hydrogenated maize oil and olive oil were used at 10 % by weight in the diet, and fed for 15 months. The partially hydrogenated maize oil contained 42 % trans-18 : 1 and 27·2 % cis-18 : 1 fatty acids, whereas the olive oil contained 74·1 % cis-18 : 1. No statistically significant difference in colon tumour incidence was observed, with incidence rates of 35·3 and 31·3 % in animals receiving partially hydrogenated maize oil and olive oil, respectively.

A similar study with a strain of animals especially susceptible to colon cancer (Wistar–Furth–Osaka) was carried out by Sugano et al. ⁽Reference Sugano, Watanabe and Yoshida³⁰⁾. High-trans-FA partially hydrogenated maize oil was compared with high-18 : 1 safflower-seed oil at 5 % of energy. The incidence of dimethylbenz[a]anthracene-induced tumours in the small and large intestines were 63 and 75 %, respectively in the animals fed partially hydrogenated maize oil and 65 and 71 %, respectively, in the group fed high-18 : 1 safflower-seed oil.

Hogan & Shamsuddin⁽Reference Hogan and Shamsuddin³¹⁾ fed inbred female F344 rats a diet containing 25 % trans-18 : 1 fat or 25 % cis-18 : 1 fat, and injected the animals weekly with azoxymethane to induce large-intestinal carcinomas. Although more animals receiving the diet containing trans-fat developed tumours than those receiving the cis-fat diet, this difference was not statistically significant. Identical numbers of animals from each group developed extracolonic neoplasms.

Reddy et al. ⁽Reference Reddy and Tanaka³²⁾ studied the effect of increasing dietary levels of trans-fat on azoxymethane-induced colon carcinogenesis in rats. Three diets were prepared, each containing 23·5 % fat by weight, but with varying amounts of the trans-fat mix and oleic acid-rich safflower-seed oil (Oleinate), the latter used to balance the amount of 18 : 1 across the diets. The three diets were referred to as low-trans-fat (5·9 % trans-fat+11·7 % Oleinate +5·9 % maize oil), intermediate-trans-fat (11·7 % trans-fat+5·9 % Oleinate+5·9 % maize oil) and high-trans-fat (17·6 % trans-fat+5·9 % maize oil). For the low-, intermediate- and high-trans-fat diets, there was no significant differences between diets, with incidences of colon tumours of 63, 67 and 57 %, respectively, while incidences of small-intestinal tumours were 40, 43 and 37 %, respectively.

Ecological studies

An ecological investigation of the association between colon cancer and trans-FA status found a statistically significant correlation between colon cancer and the level of trans-FA in adipose tissue (r 0·93)⁽Reference Bakker, van't Veer and Zock¹⁴⁾.

Case–control studies

The association between colorectal adenomatous polyps and the consumption of foods containing partially hydrogenated oils was examined by McKelvey et al. ⁽Reference McKelvey, Greenland and Chen³³⁾. Dietary intake was obtained from a self-administered FFQ. While there was evidence of a positive association between total dietary trans-FA and adenomas, this did not reach statistical significance.

Slattery et al. ⁽Reference Slattery, Benson and Ma³⁴⁾ studied 1993 cases with colon cancer and 2410 population-based controls matched for age and sex. Dietary information was assessed from a detailed diet history questionnaire. A significant positive association was found between trans-FA consumption and colon cancer risk in women (OR 1·5), with a positive but not statistically significant association in men. Postmenopausal women who were not taking hormone replacement therapy had a twofold increase in risk from high levels of trans-FA in the diet, while the risk of developing colon cancer was unaffected by dietary trans-fat in women on hormone replacement therapy.

Nkondjock et al. ⁽Reference Nkondjock, Shatenstein and Maisonneuve³⁵⁾ compared the data from the dietary questionnaires of 402 cases of colorectal cancer with 668 population-based controls in Montreal, but found no effect of trans-FA in either men or women. A similar study involving 1455 cases and matched population-based controls in Scotland (UK)⁽Reference Theodoratou, McNeill and Cetnarskyj³⁶⁾ also observed no association between total trans-FA consumption and colorectal cancer after adjustment for intake of energy and total fatty acids.

Prospective studies

Lin et al. ⁽Reference Lin, Zhang and Cook³⁷⁾ used dietary and health questionnaires from 37 547 women in the Women's Health Survey to examine associations between diet and colorectal cancer. The authors found no statistically significant link between the consumption of trans-FA and colorectal cancer through their standard multivariate risk analysis, although the RR for trans-FA intake became stronger when adjusted for the consumption of other types of fat and cholesterol. There was a strong positive association between the intake of fried foods away from home and colorectal cancer risk (RR 1·86 (95 % CI 1·09, 3·16); P for trend = 0·01).

Summary of evidence for an association between trans-fatty acids and colorectal cancer

One prospective cohort, one ecological and four case–control studies reported between 1997 and 2007 were reviewed. Apart from the ecological study, all studies used dietary assessment as the measure of exposure to trans-FA. Data from these case–control and prospective cohort studies are summarised in Fig. 2.

Fig. 2 Risk of colorectal cancer from case–control and prospective epidemiological studies that evaluated dietary intake of trans-fatty acids (trans-FA): McKelvey et al. ⁽Reference McKelvey, Greenland and Chen³³⁾ ( × ), Slattery et al. ⁽Reference Slattery, Benson and Ma³⁴⁾ (men □; women ■), Lin et al. ⁽Reference Lin, Zhang and Cook³⁷⁾ (◇) and Theodoratou et al. ⁽Reference Theodoratou, McNeill and Cetnarskyj³⁶⁾ (▲). The current mean trans-FA intake in the UK has recently been estimated to be 1·0 % of total energy or 1·7 g/d for women () and 1·0 % of total energy or 2·4 g/d for men ()⁽Reference Williams, Thompson and Shaw³⁾. Risk of colorectal cancer is plotted as the relative risk (RR), with vertical bars representing the 95 % CI for intake ranges above reference (RR = 1) in each study. Data from Lin et al. ⁽Reference Lin, Zhang and Cook³⁷⁾ were converted from percentage of total energy to g/d based on an energy intake of 1700 kcal/d (7113 kJ/d) (study only included women).

There are limited data available to assess evidence for an association between trans-FA intakes and incidence of colon cancer. Only one case–control study has shown a positive association, and this was significant only in women⁽Reference Slattery, Benson and Ma³⁴⁾. In this study, the RR was 1·5 for a highest v. lowest quintile range of < 2·6 to>5·1 g trans-FA per d (based on an energy intake of 1700 kcal/d (7112 kJ/d)), which is notably higher than the recent estimate of mean intake in the UK for women of 1·7 g/d⁽Reference Williams, Thompson and Shaw³⁾.

Overall the epidemiological data are limited, with weak evidence to support an adverse effect of trans-FA intakes on colon cancer. The animal studies, which included very high exposures to trans-FA, provided no evidence that trans-FA can enhance the development of colon tumours.

Ecological studies

The ecological study by Bakker et al. ⁽Reference Bakker, van't Veer and Zock¹⁴⁾ examined the association between prostate cancer and trans-FA status measured in adipose tissue, but found no statistically significant correlation.

Case–control studies

Liu et al. ⁽Reference Liu, Schumacher and Plummer³⁸⁾ assessed the potential modification of prostate cancer risk by a functional polymorphism in the RNASEL gene (R462Q). The study involved 1012 cases of prostate cancer and matched controls. Among Caucasians (n 834), a statistically significant positive association between prostate cancer and the intake of trans-FA (OR 2·77) was evident. The association remained strongly significant for all groups of trans isomers (16 : 1, 18 : 1 and 18 : 2; all P < 0·005). There was no significant association in African American individuals. When the data for Caucasians were stratified by genotype, the association between total and isomer trans-FA intake and prostate cancer was strongest among men with the QQ/RQ genotype (OR 4·80). For men with the RR genotype, neither total trans-FA intake nor intake of any group of trans isomers was associated with prostate cancer.

Prospective studies

The Netherlands Cohort Study recruited 58 279 men between 55 and 69 years of age, with an average follow-up of 6·3 years⁽Reference Schuurman, van den Brandt and Dorant³⁹⁾. No association was found between total trans-FA consumption estimated through dietary questionnaires and development of prostate cancer.

Serum phospholipid trans-FA levels were compared in subjects that had been recruited for the β-Carotene and Retinol Efficacy Trial (CARET), a randomised trial of supplemental β-carotene and retinol for the prevention of lung cancer among 18 314 heavy smokers and asbestos-exposed workers⁽Reference King, Kristal and Schaffer⁴⁰⁾. A sample of 272 men that developed prostate cancer and 426 matched controls showed increasing prostate cancer risk with higher levels of serum trans-18 : 1n-7 (OR 1·69). Other trans-C18 fatty acids had positive trends with P values of between 0·07 and 0·12, but none of the trans-C16 fatty acids were close to statistical significance.

Chavarro et al. ⁽Reference Chavarro, Stampfer and Campos⁴¹⁾ analysed the trans-FA levels in whole blood from 476 men diagnosed with prostate cancer and controls matched according to age and smoking status at recruitment. The subjects were part of the Physicians' Health Study, a randomised trial of aspirin and β-carotene in which subjects received either the test compound or a placebo. There was no association between trans-FA levels and total prostate cancer risk, but stratification according to type of tumour showed significant correlations between total and isomer levels of trans-FA and non-aggressive tumours (OR 2·21). This association was stronger in overweight and obese men than in men with a normal BMI at baseline, and in men receiving placebo compared with men receiving aspirin. However, these interactions did not reach statistical significance. There was no association between trans-FA and aggressive prostate cancer for any trans-FA isomer or for any stratification of the cohort.

Summary of evidence for an association between trans-fatty acids and prostate cancer

One ecological, one case–control and three prospective studies were reviewed. Two of the prospective studies used serum phospholipids or whole-blood levels as the measure of exposure to trans-FA; the others used dietary assessment. Data from these case–control and prospective cohort studies that evaluated dietary trans-FA intake are summarised in Fig. 3.

Fig. 3 Risk of prostate cancer from case–control and prospective epidemiological studies that evaluated dietary intake of trans-fatty acids (trans-FA): Schuurman et al. ⁽Reference Schuurman, van den Brandt and Dorant³⁹⁾ (all men ♦) and Liu et al. ⁽Reference Liu, Schumacher and Plummer³⁸⁾ (African-American ▲; Caucasian ■; Caucasian Q allele □; Caucasian no Q allele ░). The current mean trans-FA intake in the UK for men has recently been estimated to be 1·0 % of total energy or 2.4 g/d ()⁽Reference Williams, Thompson and Shaw³⁾. Risk of prostate cancer is plotted as the relative risk (RR), with vertical bars representing the 95 % CI for intake ranges above reference (RR 1) in each study. Data from Schuurman et al. ⁽Reference Schuurman, van den Brandt and Dorant³⁹⁾ were converted from percentage of total energy to g/d based on an energy intake of 2400 kcal/d (10 042 kJ/d). Reference points for Caucasians and genotypes from Liu et al. ⁽Reference Liu, Schumacher and Plummer³⁸⁾ are all at 2·25 g/d.

There are very limited data available on which to base conclusions concerning possible adverse effects of trans-FA on prostate cancer. Two nested case–control studies have demonstrated a positive association between trans-FA intake and prostate cancer⁽Reference King, Kristal and Schaffer^40, Reference Chavarro, Stampfer and Campos⁴¹⁾, although in the most recent study this correlation was only for non-aggressive tumours. A large case–control study has observed a moderate increase in risk for the Caucasian population, but a very marked increase in risk for the RNASEL QQ/RQ genotype (about 35 % of the population)⁽Reference Liu, Schumacher and Plummer³⁸⁾. The RNASEL gene is involved in protein coding and is a mediator of interferon action. Mutations in this gene have been associated with predisposition to prostate cancer, and the gene has been identified as a candidate for the hereditary prostate cancer 1 (HPC1) allele. The finding of an association between this gene, trans-FA and prostate cancer warrants further study.

Non-Hodgkin's lymphoma

It has been suggested that a higher intake of dietary fats could decrease immune response, leading to increased risk of developing non-Hodgkin's lymphoma (NHL). In the Nurses' Health Study, Zhang et al. ⁽Reference Zhang, Hunter and Rosner⁴²⁾ examined the links between NHL and amount and composition of dietary fat. After 14 years of follow-up, a strong positive relationship was found between total trans-FA intake and an increased risk of NHL (RR 2·4). The statistically significant association between trans-FA consumption and development of NHL remained after further adjustments for other types of fat, protein, alcohol, and fruit and vegetable intake. The lowest to highest quintiles of trans-FA intake were 1·3–3·2 g/d, which are similar or slightly higher than the range of UK intakes⁽Reference Williams, Thompson and Shaw³⁾. The authors also considered the importance of the source of the trans-FA, and found that the link between NHL and trans-FA consumption was stronger for vegetable fat sources (RR 1·9) than animal fat sources (RR 1·4). However, it must be noted that this study based trans-FA intake on the dietary questionnaire completed in 1980, which is likely to introduce significant error due to the changes in food manufacturing and personal food choices over the course of the study.

Ovarian cancer

Data from the Nurses' Health Study were also used to assess the possible link between diet and risk of ovarian cancer⁽Reference Bertone, Rosner and Hunter⁴³⁾. In the 80 258 pre- and postmenopausal women included in the analysis, 301 cases of ovarian cancer were diagnosed. There was no association between consumption of trans-FA and development of the disease.

Pancreatic cancer

Michaud et al. ⁽Reference Michaud, Giovannucci and Willett⁴⁴⁾ used the dietary and health data from the Nurses' Health Study to determine whether there was an association between diet and the risk of developing pancreatic cancer, but found no significant relationship.

Summary of evidence for an association between trans-fatty acids and other forms of cancer

There are limited data available to assess evidence for an association between trans-FA intakes and other forms of cancer. Data from the Nurses' Health Study provide preliminary evidence of a potential association between trans-fat consumption and NHL. However, further studies are required to enable the findings to be verified.