Tel.: +47-22-85-13-99, Fax: +47-22-85-13-13
Intake of trans fatty acids from partially hydrogenated vegetable and fish oils and ruminant fat in relation to cancer risk
Version of Record online: 12 AUG 2012
Copyright © 2012 UICC
International Journal of Cancer
Volume 132, Issue 6, pages 1389–1403, 15 March 2013
How to Cite
Laake, I., Carlsen, M. H., Pedersen, J. I., Weiderpass, E., Selmer, R., Kirkhus, B., Thune, I. and Veierød, M. B. (2013), Intake of trans fatty acids from partially hydrogenated vegetable and fish oils and ruminant fat in relation to cancer risk. Int. J. Cancer, 132: 1389–1403. doi: 10.1002/ijc.27737
- Issue online: 24 JAN 2013
- Version of Record online: 12 AUG 2012
- Accepted manuscript online: 23 JUL 2012 05:34AM EST
- Manuscript Accepted: 2 JUL 2012
- Manuscript Received: 13 APR 2012
- Norwegian Cancer Society. Grant Number: HS02-2009-0233
- Throne-Holst Foundation for Nutrition Research
- Trans fatty acids;
- partially hydrogenated vegetable oils;
- partially hydrogenated fish oils;
- ruminant fat;
- cancer risk;
- cohort study
Intake of trans fatty acids (TFA) may influence systemic inflammation, insulin resistance and adiposity, but whether TFA intake influences cancer risk is insufficiently studied. We examined the association between TFA intake from partially hydrogenated vegetable oils (PHVO-TFA), partially hydrogenated fish oils (PHFO-TFA), and ruminant fat (rTFA) and cancer risk in the Norwegian counties study, a large cohort study with a participation rate >80%. TFA intake was assessed three times in 1974–1988 by questionnaire. A total of 77,568 men and women were followed up through 2007, during which time 12,004 cancer cases occurred. Hazard ratios (HRs) and confidence intervals (CIs) were estimated with Cox regression for cancer sites with ≥150 cases during follow-up. Significantly increased or decreased risks were found when comparing the highest and lowest intake categories (HRs, 95% CIs) for PHVO-TFA and pancreatic cancer in men (0.52, 0.31–0.87) and non-Hodgkin lymphoma (NHL) in both genders (0.70, 0.50–0.98); PHFO-TFA and rectal cancer (1.43, 1.09–1.88), prostate cancer (0.82, 0.69–0.96), and multiple myeloma (2.02, 1.24–3.28); and rTFA and all cancers (1.09, 1.02–1.16), cancer of the mouth/pharynx (1.59, 1.08–2.35), NHL (1.47, 1.06–2.04) and multiple myeloma (0.45, 0.24–0.84). Furthermore, positive trends were found for PHFO-TFA and stomach cancer (ptrend = 0.01) and rTFA and postmenopausal breast cancer (ptrend = 0.03). Inverse trends were found for PHVO-TFA and all cancers (ptrend = 0.006) and cancer of the central nervous system in women (ptrend = 0.005). PHFO-TFA, but not PHVO-TFA, seemed to increase cancer risk. The increased risks observed for rTFA may be linked to saturated fat.
Observational studies and randomized trials have found that intake of trans fatty acids (TFA) has unfavorable effects on multiple risk factors for cardiovascular diseases, including blood lipids, lipoproteins and endothelial function.1 For this reason the association between TFA intake and cardiovascular disease risk has been extensively studied.2 TFA intake may also influence factors related to cancer risk such as systemic inflammation, insulin resistance, and adiposity,1, 3 but the association between TFA intake and cancer risk has not been sufficiently studied.4 A few studies on colon/colorectal, breast and prostate cancer have rendered inconclusive results,5, 6 and the number of studies on other cancer sites is limited.
Dietary TFA derive from three main sources; partially hydrogenated vegetable oils (PHVO), partially hydrogenated fish oils (PHFO), and ruminant fat. TFA from PHVO (PHVO-TFA) and PHFO (PHFO-TFA) are industrially produced and used in the production of margarines and fats used for baking and deep frying. Intake of PHFO-TFA was high in the United Kingdom, the Netherlands, Germany, Eastern Europe and Norway.7 TFA from ruminant fat (rTFA) occur naturally in small amounts in dairy products and meat from ruminant animals. TFA from these three sources differ in terms of carbon chain length, position, and number of trans bonds. Thus, the health effects may also be different.8, 9 Most previous studies on TFA intake and cancer risk have not distinguished between different sources of TFA, and to our knowledge, no previous study has specifically evaluated the effect of PHFO-TFA on cancer risk.
The Norwegian counties study is a large population-based cohort study of both men and women. The study has a very high participation rate, repeated exposure assessment, including detailed information on TFA intake from different sources,7 and 30 years of complete follow-up. This gives us the unique possibility to investigate the associations between PHVO-TFA, PHFO-TFA and rTFA intake and the risk of several cancers.
Material and Methods
The Norwegian counties study
The Norwegian Counties Study was carried out between 1974 and 1988 in the three counties in Norway: Finnmark, Oppland and Sogn og Fjordane. The study consisted of three health screenings, which were conducted every 5 years in Oppland and Sogn og Fjordane, whereas in Finnmark, screenings II and III were carried out 3 and 13 years, respectively, after screening I. In all three counties, residents aged 35–49 years, and a random sample aged 20–34 years were invited to screening I, in total 34,115 men and 31,509 women.10 All residents were reinvited to screening II if they had been invited to screening I. Previously invited residents were also reinvited to screening III, except that among residents of Oppland and Sogn og Fjordane aged 55 years or older only a random sample was reinvited. In addition, samples of younger residents not previously invited were invited to screening II (ages 17–39 years) and screening III (ages 20–39 years). The attendance rate was 88% at screenings I and II, and 84% at screening III. In total, 92,234 persons attended at least one health screening.
At each health screening, a team of specially trained nurses examined and interviewed the participants. Height, weight and blood pressure were measured, and a nonfasting blood sample was collected and analyzed for triglycerides and total cholesterol. Information on smoking habits and recreational and occupational physical activity was obtained by questionnaire. During the interview, the questionnaire was checked for inconsistencies, and women were asked about menopausal status. The participants' education level was obtained from Statistics Norway.7
The present study was approved by the Norwegian Data Inspectorate and the Regional Committee for Medical and Health Research Ethics, Southeast Norway.
Assessment of dietary intake
A semiquantitative food frequency questionnaire (FFQ) was handed out at screenings II and III. As it was not distributed in all municipalities, only 59% of the participants received the FFQ at screening I. A detailed description of the FFQ, its reproducibility and validity has been published elsewhere.11 The FFQ was primarily designed to detect the main sources of fat in the Norwegian diet and included 80 food items.
The nutrient composition of the food items in the FFQ was obtained from the Norwegian food composition table most relevant for the period of data collection. Food items that contributed to the calculation of nutrient intake included bread, fats on bread, spreads, cheese, milk, cakes, eggs, meat and fish dishes, potatoes and cod liver oil supplement. Standard portion sizes were used for most of the food items. TFA intake was calculated using the fatty acid composition of the food items, margarines and fats used by the baking industry at the time of the screenings. During the study period, PHVO was derived from soybean oil and PHFO from different fish oils. A more detailed description of the production procedures have been published previously.7 We calculated PHFO-TFA intake and PHVO-TFA intake based on the amount consumed of baking fat and different types of margarine. Food items that contributed to intake of baking fat were store-brought bread and cake. Margarine intake included margarine used on bread, for frying fish, and on meat and fish dishes. We did not have information on deep-fried food items or snacks that may contain industrial TFA. However, these foods represented only a small proportion of industrial TFA intake.12 Calculation of rTFA intake was based on consumption of dairy products and ruminant meat products. If more than 10 questions on any FFQ were unanswered, we did not calculate that participant's nutrient intake for the corresponding health screening. Of participants who received the FFQ, nutrient intakes were estimated for 93% at screening I, 82% at screening II, and 84% at screening III. At each health screening, we eliminated nutrient intakes if the ratio of energy intake to basal metabolic rate13 was in the screening-specific top and bottom 1%, leaving 79,227 participants with nutrient intakes from at least one screening (Fig. 1).
Follow-up and identification of cancer events
Information on cancer incidence, and death or emigration was obtained through linkages to the Cancer Registry of Norway and Statistics Norway, using the unique identification number assigned to all Norwegian citizens. The Cancer Registry of Norway was used to obtain information on cancer site and histology.
The baseline screening represented the start of follow-up, and was defined as the first health screening a participant attended when (i) they were aged 18 years or above, (ii) fat intake could be estimated, (iii) questions about smoking and physical activity were answered, and (iv) weight and height were measured. We excluded participants whose screenings did not fulfill the criteria for a baseline screening, as well as those without information on education level, those who emigrated or were diagnosed with cancer before start of follow up, and participants who died, emigrated, or were diagnosed with cancer during the first year of follow-up. After these exclusions, 77,568 participants (50.4% men) were included in the analyses (Fig. 1). Of these, 37,439 had one measurement of TFA intake, 24,095 had two measurements and 16,034 had three measurements.
Participants were then followed up until the date of cancer diagnosis, death, emigration or December 31, 2007, whichever occurred first. Participants with cancer at multiple sites with the same date of diagnosis were included as events in analyses of all cancers, but were censored at diagnosis in site-specific analyses. Cancers were categorized using the international classification of diseases, tenth edition, as in “Cancer in Norway.”14 In the present study we considered only cancer sites with 150 events or more during follow-up (Table 1). Breast cancers were defined as premenopausal if at the date of diagnosis, menopause had not yet been reported at a health screening, and the woman was younger than 55 years of age; otherwise they were defined as postmenopausal breast cancers.15 Histological grade was available for 920 (49.8%) prostate cancers, which were classified as low-grade (high or intermediate differentiation, n = 709) or high-grade (low or no differentiation, n = 211). For all the sites we studied, >90% of the cancer cases were morphologically verified, except pancreas and central nervous system (CNS), for which 86.9 and 78.9% of cases, respectively, were morphologically verified.
TFA intake was measured as percent of energy intake (% of kJ day−1 = E%). At each health screening, TFA intake was classified into five categories: < 0.15, 0.15–0.649, 0.65–1.149, 1.15–1.649 or ≥ 1.65 E% for PHVO-TFA; < 0.85, 0.85–1.349, 1.35–1.849, 1.85–2.349 or ≥ 2.35 E% for PHFO-TFA and < 0.4, 0.4–0.549, 0.55–0.699, 0.7–0.849 or ≥ 0.85 E% for rTFA. These categories were chosen to establish intake cut-off points at regular intervals and to define categories with a sufficient number of events. We did not use percentiles to define intake categories because of the changes in intake that occurred between health screenings. Level of physical activity was classified as sedentary, moderately active or active at each health screening by combining the categories of occupational and recreational physical activity.16 Participants were classified as never, former or current smokers, and were classified as former instead of never-smokers at a given screening if they had been previously classified as former or current smoker. Education level was categorized as primary schooling (≤9 years), secondary education (10–12 years) or university level education (≥13 years), according to the most recent information available from Statistics Norway.
Hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between TFA intake and cancer risk were estimated with Cox regression. We used a stratified Cox model with year of birth in 5-year intervals as the strata (<1930, 1930–34, 1935–39, 1940–44 and ≥ 1945) and attained age as the time variable. All analyses were adjusted for gender, except analyses of breast, cervical, endometrial and ovarian cancer, in which only women were included; and analyses of prostate cancer, in which only men were included. The model also included energy intake (kJ day−1, as a continuous variable), level of physical activity, smoking, body mass index (BMI) (kg m−2, as a continuous variable), and education level. These covariates are established risk factors for cancer. All variables, except gender and education level, were modeled as time-dependent. Values for participants were updated at each subsequent screening that fulfilled the same criteria as the baseline screening (see above). For TFA and energy intake, we used the mean of intakes from all preceding attended screenings since start of follow-up.7, 17 For level of physical activity, smoking, and BMI, we used values from the most recent preceding attended screening.7 For cancers of the breast, cervix, endometrium and ovary, additional adjustment for age at first birth (=20, 21–25, 26–30, >30 years, or no children) did not change the results, and this was not included in the final model. In analyses of premenopausal breast cancer, women were censored at onset of menopause. In analyses of postmenopausal breast cancer, onset of menopause was considered the start of follow-up. Analyses of prostate cancer included cases without information on histological grade, but these were censored at diagnosis in analyses of low-grade and high-grade prostate cancer. For cancers of the mouth and pharynx, pancreas and lung, for which smoking is an important risk factor, we performed subanalyses restricted to never-smokers to explore residual confounding due to potentially insufficient adjustment for smoking. We also performed additional analyses including only morphologically verified cancer cases of the pancreas and central nervous system. The results were similar to the results for morphologically verified and unverified cases combined and are therefore not presented.
To test for linear trends, TFA intake was modeled continuously. We tested for interaction between TFA intake and gender. For the events where the interaction test was significant, interaction terms between intake and gender were added to the model. We also tested for interaction between TFA intake and follow-up time (< 10 years or ≥ 10 years after last measurement of nutrient intake). Because the interaction test lacks power,18 it was done only for events with at least 150 cases in each follow-up time period, namely all cancers, colon, rectal, lung, cutaneous malignant melanoma (CMM), postmenopausal breast, endometrial and prostate cancers. We used a likelihood ratio test to test for interaction, with TFA intake modeled continuously. Linear trend was tested with a likelihood ratio test in models without interaction and a Wald test in models with interaction between TFA intake and gender.
All tests were two-sided and p < 0.05 was considered statistically significant. The analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC).
A total of 12,004 men and women were diagnosed with cancer during a mean follow-up of 24.8 years (Table 1).
Baseline mean PHVO-TFA intake was 0.9 E% (median 0.7 E%, range 0.0–6.2 E%). In the highest intake category, 70.5% were men (Table 2). PHVO-TFA intake was inversely related to rTFA intake and positively related to polyunsaturated cis fatty acid intake. The Spearman correlation coefficient between PHVO-TFA intake and polyunsaturated cis fatty acid intake was 0.65. The proportion of current smokers was highest in the highest PHVO-TFA intake category.
For PHVO-TFA, we found significant, inverse trends for all cancers (ptrend = 0.006), pancreatic cancer in men (ptrend = 0.007), CMM in men (ptrend = 0.03), nonmelanoma skin cancer (ptrend = 0.03), cancer of the CNS in women (ptrend = 0.005) and non-Hodgkin lymphoma (NHL) (ptrend = 0.04). Significantly decreased risks in the highest compared to the lowest intake category were observed only for pancreatic cancer and NHL (Table 3). Moreover, for CMM and nonmelanoma skin cancer, we did not find significantly decreased risks in any intake category, and in fact men in intake category 2 (0.15-0.649 E%) were at significantly increased risk of CMM. We found significant interactions between PHVO-TFA intake and gender for pancreatic cancer (pinteraction = 0.02), CMM (pinteraction = 0.04) and cancer of the CNS (pinteraction = 0.02). Results for low-grade and high-grade prostate cancer (data not shown) were similar to results for all prostate cancers. For cancers of the mouth and pharynx, pancreas and lung, results for never-smokers were similar to results for all participants, but the inverse association found for pancreatic cancers in men was no longer significant (data not shown). The associations between intake and cancer risk did not differ by follow-up time (< 10 years or ≥ 10 years after last measurement).
Baseline PHFO-TFA intake was considerably higher than PHVO-TFA intake, mean 1.6 E% (median 1.3 E%, range 0.0–11.7 E%). In the lowest intake category, 64.4% were men (Table 2). The lowest proportion of participants with university level education was found in the highest intake category. With increasing PHFO-TFA intake, PHVO-TFA and rTFA intakes decreased, total fat intake increased and monounsaturated cis fatty acid and carbohydrate intake decreased. Total serum cholesterol increased with increasing PHFO-TFA intake. The proportion of current smokers was highest in the lowest intake category.
Significant positive trends were found for stomach cancer (ptrend = 0.01) and multiple myeloma (ptrend = 0.003, Table 4). Furthermore, participants in all intake categories except category 4 (1.85-2.349 E%) were at significantly increased risk of rectal cancer compared to participants in the lowest intake category. Participants in the highest intake category were also at increased risk of pancreatic cancer, but this was not significant. Significant inverse trends were found for lung cancer in women (ptrend = 0.0003) and prostate cancer (ptrend = 0.002). Moreover, inverse associations were found in analyses of both low-grade and high-grade prostate cancer, but the association was significant for low-grade prostate cancer only (data not shown).
We found a significant interaction between PHFO-TFA intake and gender for lung cancer only (pinteraction = 0.001). No significant interactions were found between PHFO-TFA intake and follow-up time, except for all cancers (pinteraction = 0.004), in which HRs per 0.5 E% increase in intake were 1.02 (95% CI 1.01, 1.04) for follow-up < 10 years after last measurement and 0.99 (95% CI 0.98, 1.00) for follow-up ≥ 10 years after last measurement of nutrient intake. When restricting the analyses of lung cancer to never-smokers, we found a significant, positive association between PHFO-TFA intake and risk in men (HR per 0.5 E% increase in intake was 1.33, 95% CI 1.08, 1.63) and a weak, nonsignificant, positive association in women (HR per 0.5 E% increase in intake was 1.11, 95% CI 0.88, 1.40). For cancers of the mouth and pharynx and pancreas, results in never-smokers were similar to results in all participants (data not shown).
Baseline mean intake of rTFA was 0.6 E% (median 0.5 E%, range 0–2.0 E%), considerably lower than PHVO-TFA and PHFO-TFA intake. The proportion of participants with university level education decreased with increasing rTFA intake (Table 2). rTFA intake was inversely related to both PHVO-TFA and PHFO-TFA intake. Furthermore, rTFA intake was positively related to total fat, saturated fat and monounsaturated cis fatty acid intakes; and inversely related to polyunsaturated cis fatty acid intake. The Spearman correlation coefficient between rTFA and saturated fat intakes was 0.86. Mean total serum cholesterol and serum triglycerides were highest in the highest rTFA intake category. The proportion of current smokers increased, while the proportion of sedentary participants decreased with increasing rTFA intake.
We found significant, positive trends for all cancers, (ptrend = 0.002), cancer of the mouth and pharynx (ptrend = 0.006), nonmelanoma skin cancer (ptrend = 0.04), postmenopausal breast cancer (ptrend = 0.03) and NHL (ptrend = 0.01), with a significantly increased risk in the highest compared to the lowest intake category, except for postmenopausal breast cancer, for which we found a significantly increased risk in intake category 3 (0.55–0.699 E%), but not in the highest intake category (Table 5). We also found an indication of a positive trend for cancer of the CNS (ptrend = 0.05). Significant inverse trends were found for multiple myeloma (ptrend = 0.01) and for CMM in women (ptrend = 0.04), but risk was significantly lower in the highest compared to the lowest intake category only for multiple myeloma. Intake was not associated with either low-grade or high-grade prostate cancer (data not shown). Significant interaction between rTFA intake and gender was only found for CMM (pinteraction = 0.01). Furthermore, significant interactions between intake and follow-up were observed for all cancers (pinteraction = 0.0007) and CMM (pinteraction = 0.03). For all cancers, HRs and 95% CIs per 0.5 E% increase were 0.96 (0.89, 1.03) for follow-up < 10 years and 1.10 (1.06, 1.15) for follow-up ≥ 10 years after last measurement of nutrient intake. For CMM, corresponding HRs and 95% CIs were 0.89 (0.63, 1.27) and 1.39 (1.06, 1.82) for men, and 0.53 (0.34, 0.81) and 0.82 (0.56, 1.20) for women. Intake of rTFA was associated with increased risk of cancer of the mouth and pharynx in never-smokers also, but not significantly. We found no association between rTFA intake and risk of cancers of the pancreas and lung in never-smokers (data not shown).
TFA from different sources showed different associations with cancer risk. More unfavorable risks were found for PHFO-TFA and rTFA intake than for PHVO-TFA intake. For PHFO-TFA, we observed significant positive trends for stomach cancer and multiple myeloma, with significantly increased risk of multiple myeloma in the highest compared to lowest intake category. Furthermore, we found significantly increased risk of rectal cancer in the highest compared to lowest intake category. Significant inverse trends were observed for prostate cancer and lung cancer in women. However, for lung cancer, we found a significant positive association in men and a nonsignificant positive association in women when restricting the analysis to never-smokers. For PHVO-TFA, we observed significant inverse trends for all cancers, pancreatic cancer in men, CMM in men, nonmelanoma skin cancer, cancer of the CNS in women and NHL, with significantly decreased risks in the highest compared to the lowest intake category for pancreatic cancer and NHL. For rTFA, we found significant positive trends with increased risks in the highest compared to the lowest intake category for all cancers, cancer of the mouth and pharynx, nonmelanoma skin cancer and NHL. Furthermore, we found a significant positive trend for postmenopausal breast cancer. Significant inverse trends were found for multiple myeloma and for CMM in women, with significantly decreased risk in the highest compared to the lowest intake category for multiple myeloma.
To our knowledge, no previous studies have evaluated the effect of TFA from different sources on cancer risk. The different associations between cancer risk and TFA from the three sources in this report may be related to the different chemical structures of each TFA, which may lead to a different site-specific carcinogenic effect. rTFA and PHVO-TFA contain isomers with 18 carbon atoms and almost exclusively trans monoenes (a small amount of trans dienes in PHVO). The spectra of isomers are, however, quite different. The main isomer in PHVO is elaidic acid (18:1trans9) while in ruminant fat it is trans vaccenic acid (18:1trans11) in addition to conjugated linoleic acid (cis9, trans11CLA). PHFO-TFA is a complex mixture of fatty acids of different chain lengths, mainly varying between 14 and 24 carbon atoms, with different number of trans bonds and with different locations in the molecule. It is conceivable that such different chemical structures have different metabolic effects.
We found no associations between PHVO-TFA intake and premenopausal or postmenopausal breast cancer risk or prostate cancer risk. A few nested case-control studies have examined the effects of elaidic acid on breast and prostate cancer risk using biomarkers to assess TFA. The studies on breast cancer are in accordance with our results, indicating no association between elaidic acid and either premenopausal19 or postmenopausal breast cancer risk.19–22 In contrast to our results on prostate cancer, elaidic acid was associated with a significantly increased risk of nonaggressive and total prostate cancer in one study,23 while another study found a nonsignificant positive association between elaidic acid and overall prostate cancer risk.24
We observed inverse associations between PHVO-TFA intake and risk of pancreatic cancer in men, cancer of the CNS in women, NHL and all cancers. Total TFA intake, assessed by FFQ, was not associated with pancreatic cancer risk in three prospective studies,25–27 but a nonsignificant positive association was found in one case-control study.28 Two studies, both using FFQ, examined the association between total TFA intake and NHL risk and found positive associations, but they were significant in only one of the studies.28, 29 None of the previous studies on pancreatic cancer and NHL have differentiated between sources of TFA, and their results are therefore not directly comparable with ours. Until 1997 in Norway, PHVO was used in vegetable oil-based soft margarines, which have a high content of cis monounsaturated and polyunsaturated fatty acids. We observed a positive association between PHVO-TFA intake and polyunsaturated cis fatty acid intake. Polyunsaturated fatty acid intake has been associated with significantly reduced NHL risk in previous studies,30, 31 but not with reduced pancreatic cancer risk.25–27
We found no association between TFA intakes and risk of either colon or rectal cancer. The results from previous studies on colon or colorectal cancer are inconsistent. A few case-control studies have found positive associations between total TFA intake and risk of colon28, 32 or colorectal cancer,33–35 but some of these positive associations were limited to subgroups of participants.32–35 Two case-control studies36, 37 and three cohort studies38–40 found no association between total TFA intake and risk of colon or colorectal cancer. The relation between risk colorectal cancer and specific TFA or TFA from specific sources has not been studied previously.
To our knowledge, no previous studies have examined PHFO-TFA intake alone in relation to cancer risk. We found an inverse association between PHFO-TFA intake and lung cancer risk in women. The proportion of current smokers was highest in the lowest category of PHFO-TFA intake, and a possible explanation could be that our adjustment for smoking was insufficient when studying lung cancer. In never-smokers, we found a significant, positive association between PHFO-TFA intake and lung cancer risk in men and a positive, nonsignificant association in women. Thus, a protective effect of PHFO-TFA intake on lung cancer risk seems unlikely. A previous case-control study reported no association between total TFA intake and lung cancer risk.28
We found inverse associations between PHFO-TFA intake and prostate cancer risk that was significant for low-grade, but not for high-grade prostate cancer. Previous findings on TFA intake and prostate cancer are conflicting and illustrate the complexity of the matter. Brasky et al. reported that serum levels of 18:1 TFA and 18:2 TFA were associated with significantly decreased risk of high-grade prostate cancer, but not low-grade prostate cancer.41 In contrast, Chavarro et al. found no association between serum levels of TFA and risk of aggressive prostate cancer, but 18:1 TFA and 18:2 TFA were associated with a significantly increased risk of non-aggressive prostate cancer.23 A study where low-grade prostate cancers were excluded found no association for either 16:1 TFA, 18:1 TFA or 18:2 n-6 TFA.42 A Dutch study found no association between total TFA intake and risk of either localized or advanced prostate cancer.43 Moreover, one study found that the association between TFA and prostate cancer risk was modified by a genetic polymorphism.44 We did not have information about family history or genetic polymorphisms in the present study.
For rTFA, we found no associations with premenopausal breast cancer or prostate cancer risk, but a significant, positive association with postmenopausal breast cancer risk. Trans vaccenic acid has been reported to be associated with a non-significant increased risk of premenopausal breast cancer45 and with both significantly increased45, 46 and decreased risk47 of postmenopausal breast cancer. One study examined trans vaccenic acid and prostate cancer risk, and found a significant, positive association.24 Trans vaccenic acid can be desaturated to C18:2 cis-9, trans-11 conjugated linoleic acid (CLA) in humans.48 Also, preformed CLA is found in small amounts in ruminant fat. Animal models have suggested that CLA may have anticancer properties.48 However, results from epidemiological studies in humans on CLA and cancer risk are inconsistent. A significant inverse association was found in one study on colorectal cancer49 and one study on breast cancer.47 Other studies on breast cancer have found no association45, 50, 51 or a significant, positive association.46
Intakes of rTFA and saturated fat will always be highly correlated because the sources of rTFA are also important sources of saturated fat. In the present study, the Spearman correlation coefficient between these two intakes at baseline screening was 0.86. Because of the high correlation, we could not include saturated fat and rTFA intakes in the same model, as it would have caused serious colinearity problems, making the results impossible to interpret. Because we cannot distinguish between the effects of rTFA and saturated fat, it is possible that the positive associations observed between rTFA intake and cancer risk are caused by saturated fat, and not rTFA.
We found significant trends for PHVO-TFA intake and risk of CMM and nonmelanoma skin cancer, but the results from categorical analyses were not consistent with these trend tests. Moreover, rTFA intake was significantly inversely associated with CMM risk and positively associated with nonmelanoma skin cancer risk. However, this cohort is not ideal for studying risk factors for skin cancer, as we lack information on ultraviolet exposure, the main risk factor for skin cancer and residual confounding may be a problem.52 We are not aware of any other studies on skin cancer and TFA intake.
The mechanisms by which TFA may produce carcinogenic effects are poorly understood.1 However, TFA intake may promote chronic low-grade inflammation, a risk factor for the development of several cancers.4 Positive associations between TFA intake and biomarkers of inflammation have been found in both clinical trials53, 54 and observational studies.55, 56 Furthermore, fatty acids may exert immunomodulatory effects on immune cells.57, 58 A possible link between fatty acids, inflammation and immunity are the eicosanoids, which are generated from polyunsaturated fatty acids with 20 carbon atoms57 and are involved in both inflammatory and immune responses. PHFO contain a complex mixture of modified, very long-chain fatty acids, in addition to trans monoenes, trans dienes and trans trienes.59 It is thus likely that PHFO-TFA intake results in adversely modified eicosanoids that have adverse effects on inflammatory and immune responses. This may explain why we observed positive associations between cancer risk and PHFO-TFA intake, but not PHVO-TFA intake.
Strengths of our study include the prospective design, participation rates above 80%, and practically complete follow-up through linkage with high-quality national registries. Indeed, reporting of definite malignant neoplasms to the Cancer Registry of Norway is required by law, and for the period 2001–2005, the overall completeness of the Norwegian Cancer Registry was 98.84%.60 For the sites included in the present report, completeness ranges from 93.81% for cancers of the CNS to 99.98% for endometrial cancer.60 Even though most of our cases were morphologically verified, we cannot rule out the possibility that some cancer cases may have been misclassified. The proportion of cases that was verified was somewhat lower for cancers of the pancreas and CNS than for the other sites. However, limiting the analyses of these cancers to verified cases gave similar results, thus misclassification does not seem to have influenced the results. Furthermore, the FFQ was designed specifically to measure fat intake in the Norwegian diet, and we have accurate information from the margarine industry on the compositions of the different types of margarine that were produced at the time the screenings took place. Intake of industrially produced TFA has decreased substantially in Norway, and because we have up to three measurements of intake for each participant, we have been able to take at least some of the changes that have occurred into account. Although intake was not measured after 1988, it does not seem to have influenced the results. Indeed, the associations between PHVO-TFA intake and cancer risk did not differ by follow-up time (<10 years and ≥10 years after last measurement of nutrient intake), and for PHFO-TFA intake the associations did not differ for any individual cancer site, but did differ for all cancers combined. Measurement error in the exposure variables is inevitable in epidemiological studies on diet and cancer. Intakes of fat and energy measured with the FFQ have been compared to a 24-hr recall interview in a sample of participants,11 but a limitation of our study is that this was not done for TFA intake. Moreover, we have not measured biomarkers of TFA intake. Another limitation is that for some sites we have not been able to control for important potential confounding variables, e.g., alcohol intake for cancer of the mouth and pharynx and ultraviolet exposure for CMM and nonmelanoma skin cancer.
Because we have studied several end points and three different exposures, we have performed a large number of statistical tests. Even if all our null hypotheses are true, i.e., there is no association between TFA intake and risk of any cancer, we would expect to wrongly reject the null hypothesis for one in twenty tests. In conclusion, our results indicate that PHFO-TFA have more adverse effects on cancer risk than PHVO-TFA. The high correlation between rTFA and saturated fat raises the possibility that the increased risk observed for rTFA is linked to saturated fat.
The authors thank all the participants in the Norwegian Counties Study and the professional staff who made this study possible, including the former Norwegian Health Screening Service (now part of the Norwegian Institute of Public Health), which initiated the Norwegian Counties Study in 1974 and continued data collection throughout the following years. They also thank Mills DA, Norway, for providing information on the fatty acid composition of margarines and fats.
- 4World Cancer Research Fund/American Institute for Cancer Research. Food, nutrition, physical activity, and the prevention of cancer. Washington DC: AICR, 2007.
- 7A prospective study of intake of trans-fatty acids from ruminant fat, partially hydrogenated vegetable oils, and marine oils and mortality from CVD. Br J Nutr 2011. [Epub ahead of print]., , , et al.
- 13World Health Organization. Energy and protein reqirements, Report No: 724. Geneva: WHO, 1985.
- 14Cancer Registry of Norway. Cancer in Norway 2009—cancer incidence, mortality, survival and prevalence in Norway. Oslo: Cancer Registry of Norway, 2011. Available at: http://www.kreftregisteret.no/en/General/News/Cancer-in-Norway-2009/
- 18Essential medical statistics, 2nd ed. Malden: Blackwell Science, 2003., .