To study the association between content of fatty acids from milk fat (14:0, 15:0 and 17:0) in adipose tissue and risk of a first myocardial infarction (MI).
Design and subjects:
A case–control study with 99 patients and 98 population controls both men and postmenopausal women, age 45–75 year. Adipose tissue fatty acids were determined by gas–liquid chromatography.
The content of 14:0, 14:1, 15:0, 17:0 and 17:1 were all significantly higher in adipose tissue of controls than of the patients. Age and sex adjusted odds ratios (OR) for MI were significantly reduced with increasing quartiles of 14:0, 14:1, 15:0 and 17:1 in adipose tissue, but except for 15:0 (OR=0.36, 95% CI 0.13–0.99), the trend was no longer significant after further adjustment for waist-to-hip ratio, smoking and family history for coronary heart disease. Correlations between 14:0 and 15:0 in adipose tissue, and waist-to-hip ratio were significantly negative (r=−0.22 for both, P<0.01).
Our study suggests that intake of dairy fat or some other component of dairy products, as reflected by C15:0 as marker in adipose tissue, may protect persons at increased risk from having a first MI, and that the causal effects may rely on other factors than serum cholesterol.
Throne Holst's foundation for Nutrition Research, Research Council of Norway, The Norwegian Association of Margarine Producers, DeNoFa Fabriker A/S, TINE BA.
Populations with a high intake of saturated fat are shown to have a high mortality of coronary heart disease (CHD) (Renaud and Lanzmann-Petithory, 2001). This is one reason why intake of milk fat has been considered an important factor related to the high incidence of CHD in western countries. A large number of metabolic studies have shown that the saturated fatty acids lauric (12:0), myristic (14:0) and palmitic acid (16:0) increase serum total and LDL cholesterol (Kris-Etherton and Yu, 1997). It has been difficult, however, to demonstrate a clear relationship between intake of saturated fat and risk of cardiovascular disease in large observational studies. A significant positive association was found in two prospective studies (McGee et al., 1984; Kushi et al., 1985), but in others, weak or no association were found (Garcia-Palmieri et al., 1980; Gordon et al., 1981; Shekelle et al., 1981; Kromhout and de Lezenne, 1984; McGee et al., 1984; Kushi et al., 1985; Ascherio et al., 1996; Hu et al., 1997; Pietinen et al., 1997). Other observational studies have shown seemingly paradoxical positive metabolic effects like reduced body mass index (BMI), waist circumference, LDL/HDL ratio, serum triglycerides and blood pressure, associated with dairy products (Smedman et al., 1999; Mennen et al., 2000; Pereira et al., 2002), and Elwood et al. (2004), suggesting that consumption of milk products may be associated with a small reduction in risk of heart disease and stroke.
Pentadecanoic acid (15:0) and heptadecanoic acid (17:0) are characteristic for milk fat. They are synthesized by the bacterial flora in the rumen of ruminants (Wu and Palmquist, 1991), and cannot be synthesized in the human body. Another fatty acid, 14:0, is mainly present in milk fat. In most human tissues, the substrate specificities of the component enzymes involved in chain elongation and chain termination ensure that 16:0 is the major product. However, in the mammary glands, shorter chain fatty acids are synthesized due to the presence of a tissue-specific chain-terminating enzyme, thioesterase II (Thompson and Smith, 1985). All three fatty acids (14:0, 15:0 and 17:0) are also present in ruminant fat (beef and lamb meat) and in fat from fish, but generally, dairy products are the main source.
The fatty acid composition of adipose tissue partly reflects the relative proportion of fatty acids in the diet (London et al., 1991), and Wolk et al. (1998, 2001) have shown that the content of 14:0, 15:0 and 17:0 in adipose tissue and serum lipid fractions are valid biomarkers for long-term intake of dairy fat in populations with high intake of such products. In epidemiologic studies, use of biomarkers may provide a more accurate and objective measure for long-term intake than information from dietary questionnaires.
In this case–control study, we have correlated the content of fatty acids with origin from milk fat in subcutaneous fat to the risk of having a first myocardial infarction (MI) in a moderately high-risk population. This work is part of a larger study investigating the association between fatty acid composition in adipose tissue and serum lipids, and risk of a first MI (Pedersen et al., 2000; Yli-Jama et al., 2002).
Subjects and design
A total of 112 cases with a first MI and 107 controls were enrolled to the study during the years 1995–1997, according to the same eligibility criteria. Subjects were Norwegian men and women aged 45–75 years, without previously reported MI or other serious disease (cancer, diabetes, alcohol or drug abuse, major psychiatric disease), which might affect their dietary pattern. All women were postmenopausal. Subjects with body weight changes more than 5 kg during the last year, or using hypolipemic drugs or oestrogen supplement (females) were excluded from the study. Informed consent was obtained from study participants in accordance with the ethical standards of the responsible local Committees on Human Experimentation.
Cases were diagnosed (typical history, ECG and enzyme changes) with a first MI (ICD 9-code 410), and admitted to hospital within 24 h of manifesting symptoms. The cases were recruited from the coronary care units of Ullevål Hospital in Oslo (51 cases) and Østfold Central Hospital in Fredrikstad and Sarpsborg.
Control subjects were 107 healthy men and women without a history of MI, recruited from the study population catchment area and frequency-matched for age in 5-year intervals. The controls were recruited from the population in the catchment areas of patients. As it was thought that population-based samples would result in low response rates, controls were recruited among friends and relatives of cases and project collaborators, state and municipal employees and people attending recreation centres for elderly and retired.
Subcutaneous adipose tissue was taken from buttock by needle aspiration as described by Beynen and Katan (1985). In cases, the adipose samples were taken within 4 days of admission to hospital. The samples were immediately frozen on dry ice and stored at −70°C until analyzed. Needle biopsies were obtained from 112 cases and 107 controls. No fat aspirate or a sample too small for analysis was obtained from 12 cases and nine controls, leaving a total of 100 cases and 98 controls from whom results for fatty acids composition of adipose tissue were obtained. Satisfactory fatty acid results from 99 cases and 98 controls were obtained.
A fasting blood sample was taken from all subjects. Serum was separated after 30 min at room temperature by centrifugation at 2500 g for 10 min and immediately stored at −70°C until analysis. In cases admitted to Østfold Central Hospital, the sample was drawn within 24 h after the acute phase of disease, and in cases admitted to Ullevål University Hospital the sample was drawn within the third day.
Anthropometric measures (weight, height, waist-and-hip circumference) were taken directly from all subjects after a detailed interview on cardiovascular disease risk factors and dietary habits.
Fatty acids in adipose tissue biopsies
The fatty acid analysis of adipose tissue samples has been described in detail in an earlier paper (Pedersen et al., 2000). In short, lipid extraction from adipose tissue and direct methanolysis were modified from Viga and Grahl-Nielsen (1990). Fatty acid composition in the methylated sample was analysed by a Shimadzu GC-17A gas chromatograph equipped with an autoinjector AOC 17 and a flame ionization detector (Shimadzu Corp., Kyoto, Japan). A fused silica capillary column length 100 m, 0.25 mm i.d. was used (SP™-2560, Supelco Inc., Bellefonte, PA, USA). Identification of the different peaks was carried out by comparing the retention times to commercial standards and the percentage distribution was calculated using the Class GC-10 software (Shimadzu Corp., Kyoto, Japan). The interassay coefficient of variation was found to vary between 2.1 and 4.5% for different fatty acids.
Fatty acid values for adipose tissue fatty acids are given as percentage content (mol/100 mol total fatty acids). Summary statistics for all the measured variables were calculated for cases and controls. Differences in mean values or proportions between cases and control groups were tested by Student's t-test and the χ2 test, respectively. Linear trend between selected fatty acids as quartiles among the control subjects was estimated by analysis of variance (GLM univariate procedure in SPSS).
Odds ratios (OR) with 95% confidence intervals (CI) for MI were calculated for the quartiles of the chosen fatty acids by use of logistic regression analysis. The analyses were based on the distribution among control subjects, with the lowest quartile used as reference. Tests for linear trends across quartiles of the fatty acids were performed by assigning the quartiles as continuous variables in the model. Different multivariate models were used to identify as to which effect the potential confounders had on the association between the fatty acids and MI. The first model included age and sex, considering the fact that the patients and the controls were not pair-matched to each other. In the second model, waist–hip ratio was added, and in the third model current smoking (smoker/nonsmoker) and family history of CHD (with/without) were added. Statistical analyses were performed using the statistical package SPSS 11.0 (SPSS Inc., Chicago, IL, USA). P-values 0.05 were considered significant.
The prevalence of risk factors in MI cases and control subjects is shown in Table 1. Waist-to-hip ratio, serum triglycerides, percentage of current smokers and family history of CHD were significantly lower in controls than in cases, while total cholesterol, HDL cholesterol and length of education were significantly lower in cases. The lower serum total cholesterol in cases is explained by the lowering of serum lipids after an acute MI (Ahnve et al., 1989). The much larger difference between cases and controls in HDL cholesterol (25%) than in total cholesterol (8%), however, probably reflects lower HDL cholesterol in cases. The larger waist-to-hip ratio in cases was significant both in men and women (P<0.001 for both, data not shown).
Adipose tissue fatty acids in patients and control
With the methods used, about 40 different fatty acids could be identified in the adipose tissue lipid extracts (Table 2). Several fatty acids differed in percentage content (mol/100 mol) between cases and controls. Only those fatty acids with origin from milk fat will be further considered in this paper. Contents of 14:0, 14:1 cis, 15:0, 17:0 and 17:1 cis were all significantly higher in adipose tissue of the controls than of the MI cases.
Contents of 14:0, 15:0 and 17:0 in adipose tissue of controls were positively correlated to each other (Table 3) but not to total very long chain n-3 fatty acids. None of the fatty acids were correlated to total energy intake (data not shown). Pentadecanoic acid was positively correlated to serum cholesterol (Table 3).
Adipose tissue fatty acids and risk of MI
Significant negative trends in ORs adjusted for age and sex were found for 14:0 and 15:0 in adipose tissue, but not for 17:0 (Table 4). The trend remained significant for 15:0 but not for 14:0 after additional adjustment for waist-to-hip ratio, smoking and family history of CHD. The control persons had on average 3 years longer education than the cases, but education was correlated to smoking (r=−0.238, <0.001) and to adjust for education may therefore result in overadjustment. Additional adjustment for education in the 15:0 model, only slightly changed the results (P for trend <0.05).
Significant negative trends in ORs adjusted for age and sex were found for 14:1 and 17:1 in adipose tissue and the trend remained significant after additional adjustment for waist-to-hip ratio, smoking and family history (Table 4).
Smoking is an important risk factor for MI, and the proportion of current smokers was much higher among cases (60%) than among control persons (20%). It is not likely that smoking is biologically connected to fatty acids, and no significant interaction with smoking was found for 15:0 (P=0.10), 14:1 (P=0.66) or 17:1 (P=0.25). When splitting the data into smokers and nonsmokers, the trends in OR over quartiles of different fatty acids were decreasing both for smokers and nonsmokers, but the number of subjects in each group was too small to achieve statistical significance and to draw any conclusions (Table 5).
Correlation between biomarkers of milk fat, and waist-to-hip ratio and BMI
The results showed significant negative correlations between 14:0 and 15:0 in adipose tissue, and waist-to-hip ratio (r=−0.22 for both, P<0.01, n=179). The negative correlations were significant when calculated for case subjects and controls together, but when the data were split into cases and controls, it was significant only for the cases (r=−0.28 for 14:0 and r=−0.22 for 15:0, P<0.05, n=86). Adipose tissue 14:0, 15:0 and 17:0 were significantly negatively correlated to BMI (r=−0.25, r=−0.15 and r=−0.21, respectively, P<0.01, n=180).
Significantly reduced odds for MI were observed for the second, third and fourth quartiles of 15:0 in adipose tissue. Significantly reduced trends in ORs for MI were also observed for 14:1 and 17:1. Age and sex adjusted OR for 14:0 in adipose tissue was significantly reduced in the fourth compared to the first quartile, but the significance disappeared when adjusting for waist-to-hip ratio.
The strength of this study is the use of biomarkers for long-term intake of dairy fat. The subcutaneous fatty acid composition reflects the long-term dietary intake over periods of years prior to the sampling (Katan et al., 1997). It is therefore most unlikely that the reported associations should be due to changes in dietary habits due to the cardiac event, especially as all the patients except two were completely unaware of having CHD prior to the episode, which led to inclusion in the study. Lifestyle factors have been adjusted for in the statistical model, but there may be differences between the two groups in other lifestyle factors than those adjusted for in this study, for instance diet. The content of trans fatty acids in adipose tissue was higher among cases than among controls, which probably reflects a higher intake of margarine among cases (Pedersen et al., 2000). In addition, the control subjects had a significantly higher intake of vegetables, fruit, berries, cereals, fat fish and wine than cases (paper under preparation). It seems that the control subjects had a healthier lifestyle than the MI patients as a total. Thus, residual confounding cannot be excluded.
The control subjects in this study were chosen among friends and relatives of the patients, as well as other available persons from the same area. They had on average 3 years more education than the patients, and only 20% were smokers. A previous study of a random sample of the Norwegian population showed that in an age group comparable to ours, 29% of the men were smokers and the mean length of education was 10.4 years (Johansson et al., 1997). Our control subjects thus had a slightly lower number of smokers and more education than expected. Thus, we cannot exclude the possibility that selection bias in part may explain our results. In order to correct for the differences in risk factors between the groups, we have adjusted for some variables in the multivariate models. Smoking and waist-to-hip ratio were adjusted for, but not other health-related factors associated with education. To adjust for education together with smoking may be considered unreasonable as the association between smoking and education is strong (Jenum et al., 2001). However, the difference in educational length between the cases and controls in our study may also reflect the striking social gradient in risk for CHD (Jenum et al., 1998). Thus, any sampling of patients with MI is likely to result in a population with less education and probably also higher number of smokers than a healthy control group. There is also a strong social gradient with regard to food choice corresponding to what was observed in the present study (Johansson et al., 1999). Another aspect is that the patients may have reduced their milk fat intake because of a family history of CHD.
The association between 14:0, 15:0 and 17:0 in adipose tissue of controls (Table 3) is a clear indication of a common source, namely milk fat. The main sources of these three fatty acids in the Norwegian diet is milk fat and fat from fish, and we cannot exclude that the content of 14:0, 15:0 and 17:0 in adipose tissue also may reflect intake of fat from fish, but this is unlikely since no associations were found between the three fatty acids and VLCn-3 fatty acids, which are exclusively derived from fish or fish oil. Until the end of 1990s, margarines in Norway contained partially hydrogenated fish oil or soybean oil, and the use of coconut or coconut oil, which are good sources of 14:0, has been near to zero in the Norwegian diet.
The control subjects had significantly higher proportions of 14:0, 15:0 and 17:0 in adipose tissue than MI cases, and a significant negative association between MI and 15:0 was found after adjustment for age, sex, smoking, waist-to-hip ratio and education. The mono-unsaturated fatty acids 14:1 cis and 17:1 cis also differed significantly between case subjects and controls with the highest levels in controls (Table 2), and a significant negative association with MI was found after adjustment for age, sex, waist-to-hip ratio, smoking and family history (Table 4). These mono-unsaturated fatty acids are most probably produced in the human body by desaturation of 14:0 and 17:0.
As 14:0, 15:0 and 17:0 in adipose tissue only reflect the intake of dairy fat, the results do not give information about what sort of milk products were consumed by the test persons. We cannot exclude the possibility that other factors than fat in dairy products are responsible for the effect on risk for MI that is shown in our study. For instance, many studies have shown that dietary calcium promotes the excretion of fats as Ca soaps in faeces (Bhattacharyya et al., 1969; Renaud and Lanzmann-Petithory, 2001; Lorenzen et al., 2005). It is also possible that milk products may contain some factors leading to MI, and others that are protective, in different proportions in different milk products. Thus, the effect on risk of MI will most probably be influenced by the choice of milk products in the diet. Dietary data showed that the control persons had a mean daily intake of 39 g cheese and 350 g milk and yoghurt, which is a common intake of dairy products in Norway. The only dairy products that differed between cases and control subjects were cheese and ice cream, with a higher intake among control persons (to be published). Two intervention studies have investigated the effect of different milk products on risk factors for CHD and the results from these studies have shown that cheese is less cholesterol increasing than butter (Biong et al., 2004; Tholstrup et al., 2004). Results from epidemiological studies have shown positive associations between intake of milk and CHD deaths, over different countries and times, while cheese does not appear to be associated, indicating that cheese may differ from other dairy products in association with CHD (Artaud-Wild et al., 1993; Moss and Freed, 2003).
The proportions of 15:0 and 17:0 in serum lipids have in two studies been shown to be valid markers for dairy fat intake (Smedman et al., 1999; Wolk et al., 2001). The results from the present study are in line with previous findings from a Swedish case–control study (Warensjo et al., 2004), which showed that high proportions of 15:0 and 17:0 in serum phospholipids are associated with reduced risk of first acute MI. In that study, correlations between clinical variables associated with the metabolic syndrome (MS) and MI, and markers of milk fat intake were investigated, and significant negative correlations between markers and serum triacylglycerols, cholesterol, insulin, leptin, PAI-1 ag and BMI were found. The same negative correlations between biomarkers of milk fat intake and clinical variables were found in a prospective cohort study of elderly men (Smedman et al., 1999). In that study, negative associations were found between 15:0 in serum cholesterol esters and phospholipids, and BMI, body weight, Apo B, hip- and waist circumference, whereas positive correlations were found for Apo AI and HDL cholesterol. Another study investigated correlations between 15:0 in cholesterol esters and metabolic variables in young adolescents, and found negative correlations between 15:0 and serum cholesterol in both girls and boys (Samuelson et al., 2001).
The lack of consistency in earlier studies investigating the association between intake of saturated fat and MI may be partly explained by the large measurement error inherent in dietary studies. Other confounding factors in epidemiologic studies may be recall bias and genetic variants and different levels of intake of saturated fat and dairy fat among different ethnic groups. In the present and the Swedish studies, these confounders are removed because of use of biomarkers, and the Swedish population is highly comparable with the Norwegian population both in genetic characteristics, dietary habits, lifestyle and dairy intake.
Another study – the Cardia Study – investigating the relation between intake of dairy products and cardiovascular disease has found that dietary patterns characterized by increased dairy consumption have a strong inverse association with MS among overweight adults and may reduce risk of type 2 diabetes and cardiovascular disease (Pereira et al., 2002). Individuals with MS have an increased risk of developing CHD (Reaven, 1993), and many cases in the present study had clinical characteristics typical for MS (elevated serum triglycerides, blood pressure, waist-to-hip ratio, and reduced serum HDL) (Table 1). The higher intake of dairy fat, reflected by the proportions of 14:0, 15:0 and 17:0 in adipose tissue, and the reduced risk of having a first MI among the control persons, may indicate that a diet with a certain content of dairy fat, or some other component of milk, may protect against developing MS. A positive correlation was found between 15:0 in adipose tissue and serum total cholesterol of control persons (Table 3). This indicates that the possible positive effect of dairy products on MS may be attributed to other risk factors than total cholesterol.
Obesity, and especially central obesity, is part of MS (Reaven, 1993). Cases in our study had significant higher waist-to-hip ratio than controls, and biomarkers of milk fat intake showed significant negative correlation to waist-to-hip ratio and BMI. Even though the correlations were weak, they indicate that intake of dairy products, reflected by markers in adipose tissue, may represent a weak protection against overweight, especially central obesity. This is supported by epidemiologic and experimental studies, suggesting that dairy products may have favourable effects on body weight in children (Carruth and Skinner, 2001) and adults (Davies et al, 2000; Lin et al, 2000; Zemel et al, 2000). In these studies, calcium is suggested as the factor responsible for the effect on body weight.
In summary, our results show that persons with relatively ample dairy fat consumption as reflected by markers in adipose tissue, in particular C15:0, may possibly be protected against a first MI. We cannot conclude whether a high intake of dairy fat or a proportional low intake of other types of fat, for example, trans fatty acids, is responsible for the effect. Neither can we conclude that the effect is caused by the fat itself or by some other components of the dairy products. This needs to be further elucidated. Owing to the nature of case–control studies, here visualised by selection bias, confounding and the fact that dairy products have different composition, we cannot conclude that the reduced risk of MI is caused by dairy products alone, but rather that the control persons in our study have had a healthy lifestyle protecting them from having a first MI, and that dairy fat has been a part of this lifestyle. Results from this study also indicate that the causal effects may be on some other factors than serum total cholesterol. Further investigation is needed to verify our results and to clarify whether there are differences between different dairy products on risk of MI.
We thank Kari Almendingen, Hege Møklebust Rebnord, Ragnhild Lekven Fimreite and Thomas S Haugen for their effort in the main study.
About this article
Evaluation of urinary iodine excretion as a biomarker for intake of milk and dairy products in pregnant women in the Norwegian Mother and Child Cohort Study (MoBa)
European Journal of Clinical Nutrition (2009)