All rights reserved
Assessment of trans-Fatty Acid Intake with a Food Frequency Questionnaire
and Validation with Adipose Tissue Levels of frans-Fatty Acids
Rozenn N. Lemaitre,1 Irena B. King,2 Ruth E. Patterson,34 Bruce M. Psaty,1-4-5 Mark Kestin,3-6 and
Susan R. Heckbert1'4
Past studies of the association of frans-fatty acid intake with coronary heart disease have been hindered by the lack of a database on the frans-fatty acid content of various foods. The authors used new data from the US Department of Agriculture to estimate frans-fatty acid intake using a self-administered food frequency questionnaire (FFQ), and they assessed the validity of the FFQ by comparing the dietary estimates with frans-fatty acid concentrations in adipose tissue. The 1996 study included 27 women and 24 men aged 51-78 years. The mean consumption of total frans-fatty acids estimated from the FFQ was 2.24 g per day and 5% of total dietary fat. The mean concentration of total frans-fatty acids in buttock adipose tissue was 4.7% of total fatty acids. Pearson correlations between total dietary intake of frans-fatty acids and total frans-fatty acid levels in adipose tissue were 0.67 (95% confidence interval (Cl) 0.36-0.84) among men and 0.58 (95% Cl 0.26-0.79) among women. After adjustment for energy intake, age, and body mass index, the correlation coefficients were 0.76 (95% Cl 0.51-0.89) among men and 0.52 (95% Cl 0.17-0.75) among women. The FFQ validated in this study is an important new tool for assessing usual intake of frans-fatty acids. Am J Epidemiol 1998; 148:1085-93.
adipose tissue; diet; dietary fats; fatty acids; questionnaires
There is convincing evidence that the Western diet plays a major role in atherogenesis (1). Although re-ducing dietary fat intake is believed to have the great-est potential for decreasing the risk of cardiovascular disease and other chronic diseases (2-4), it remains unclear whether the amount of fat or the type of fat is more important in promoting atherosclerotic disease (1). Evidence from feeding and epidemiologic studies suggests a possible role of frans-fatty acids in promot-ing coronary heart disease (5-9).
Trans-fatty acids are unsaturated fatty acids with one or more double bonds in the trans configuration. Trans-fatty acids are formed in the commercial pro-cess of converting vegetable oils into solid fats, a
Received for publication September 8, 1997, and accepted for publication April 14, 1998.
Abbreviations: Cl, confidence interval; FFQ, food frequency ques-tionnaire.
1 Cardiovascular Health Research Unit, Department of Medicine, University of Washington School of Medicine, Seattle, WA.
2 Public Health Sciences Division, Fred Hutchinson Cancer
Re-search Center, Seattle, WA.
3 Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, Seattle, WA.
4 Department of Epidemiology, University of Washington School of Public Health, Seattle, WA.
5 Department of Health Services, University of Washington School of Public Health, Seattle, WA.
6 Nutrition Program, Bastyr University, Seattle, WA.
process known as partial hydrogenation (5). Partially hydrogenated oils are the major source of trans-fatty acids in the US diet. Bacterial fermentation in the stomach of ruminants also produces frans-fatty acids and is responsible for small amounts of frans-fatty acids in meat and dairy products (5).
The results obtained from epidemiologic studies of the association of frans-fatty acid intake with coronary heart disease have been conflicting (9, 10), and the •interpretation of the data is highly controversial (11, 12). Epidemiologic studies have been hindered by the lack of an accurate and comprehensive database on the frans-fatty acid content of foods. Until recently, infor-mation on frans-fatty acid levels in various foods was scarce, scattered, and often outdated. The US Depart-ment of Agriculture recently released a more compre-hensive database on the frans-fatty acid content of foods (13). We took advantage of these new data to estimate frans-fatty acid intake using a food frequency questionnaire (FFQ) especially designed to capture dietary fat consumption (14). In this study, we as-sessed the validity of the FFQ by comparing the esti-mates of frans-fatty acid intake with the adipose tissue concentrations of frans-fatty acids in 51 adults. The study addressed the need for practical and accurate dietary assessment instruments in the measurement of frans-fatty acid intake.
1086 Lemaitre et al.
MATERIALS AND METHODS Study subjects
Fifty-one adult volunteers (27 women and 24 men) aged 51—78 years and residing in the greater Seattle, Washington, area gave informed consent to participate in the study. We recruited participants through local radio advertising and flyers distributed in community senior centers and at the University of Washington. Participants were compensated with $25.
Eligibility criteria included an age between 50 and 79 years and the ability to fill out a dietary question-naire in English. We excluded individuals who re-ported gaining or losing more than 5 percent of their body weight in the previous 2 years and individuals who reported making major changes in their diet in the previous 2 years. The Institutional Review Board of the University of Washington approved the study pro-cedures.
Data collection
Data collection included administration of the FFQ, collection of an adipose tissue specimen, and measure-ment of weight, height, and waist and hip circumfer-ence. All data were collected during a single clinic visit at the Clinical Research Center of the University of Washington Medical Center (Seattle, Washington) between June 1996 and the end of October 1996.
FFQ trans-tatty acids
The self-administered, semiquantitative FFQ re-quested information on the participant's usual diet during the year prior to the clinic visit. The FFQ was developed at the Fred Hutchinson Cancer Research Center (Seattle, Washington) (15). In format, it is similar to the National Cancer Institute/Block FFQ (16), with modifications to make it more sensitive to fat intake. In a study of 95 women, the correlation between estimates of energy derived from fat that were obtained from the FFQ and from food records (16.3 records per participant (standard deviation 3.2)) was 0.74 (95 percent confidence interval (CI) 0.64-0.82) (14). This correlation is higher than has been reported for other FFQs (16-19) and suggests that our esti-mates of fat intake are fairly precise.
The FFQ is divided into three sections: 1) adjust-ment questions, 2) food line items, and 3) summary questions. There are 15 adjustment questions used to alter how the analysis software calculates the nutrient content of specific food items. Most of the adjustment questions permit more refined analyses of fat intake by asking about food preparation practices and types of added fats. The main section consists of 98 food line
items, with questions on usual frequency of intake (ranging from "never or less than once per month" to "two or more times per day" for foods and "six or more times per day" for beverages) and portion size (small, medium, or large in comparison with the stated medium portion size). The three summary questions ask about usual intake of fruits, vegetables, and fat used in cooking.
For this study of trans-fatty acids, we collected additional detail on types of added fat. For example, while the FFQ asks only about the use of margarine overall, we assessed use of low calorie margarine, stick margarine, and tub margarine separately; and we assessed use of vegetable shortening such as Crisco (Procter and Gamble, Cincinnati, Ohio) separately from use of animal shortening (lard, bacon fat, drip-pings, salt pork, and ham hocks). We collected this information using four questions on 1) fat used to fry foods, 2) fat used when cooking vegetables, potatoes, beans, and rice, 3) fat added after cooking vegetables, potatoes, beans, and rice, and 4) fat usually used on breads, bagels, muffins, tortillas, and rolls. We re-placed participant responses to the less detailed FFQ questions with responses to our more detailed ques-tions on types of added fats.
The FFQ nutrient database is derived from the Uni-versity of Minnesota Nutrition Coding Center nutrient database (20), as described in proceedings from the 1992 National Nutrient Databank Conference (21). We expanded the nutrient database to include total
trans-fatty acids, monounsaturated ?rans-fatty acids
with 18-carbon chains, monounsaturated trans-fatty acids with 16-carbon chains, and diunsaturated
trans-fatty acids with 18-carbon chains. Our data source for
dietary trans-fatty acids was the US Department of Agriculture's Special Purpose Table No. 1, which contains estimates for major sources of rrans-fatty acids (13). We determined grams of trans-fatty acid per gram of fat in the foods analyzed by the Depart-ment of Agriculture and used this ratio to estimate the amounts of trans-fatty acids for foods in the FFQ database. We calculated trans-fatty acid content for 28 FFQ line items. Since most of the FFQ line items contained two or three foods of similar nutrient com-position (e.g., potato salad, macaroni salad, and pasta salad), we calculated the trans-fatty acid content of 83 foods in total.
US Department of Agriculture trans-fatty acid fig-ures were available for most foods in the FFQ database (e.g., margarine, milk, beef). When more than one brand of a food (e.g., bread) was in the Department of Agriculture table, we averaged the trans-fatty acid content of all brands. When Department of Agriculture data were not available for a food, we used the
trans-fatty acid content of a similar food. For example, we used the trans-fatty acid content of French fries for hash browns. A few combination foods (e.g., fried chicken) and baked goods (e.g., muffins) required assumptions about preparation methods and/or reci-pes. A document detailing the procedures used to develop our trans-fatty acid database is available from the authors upon request.
Adipose tissue trans-fatty acids
The adipose tissue samples were obtained by aspi-ration of subcutaneous fat from the upper buttock using a modification of the method of Hirsh et al. (22) that was developed for another study (23). Briefly, 1 ml of sterile saline was injected into the subcutaneous adipose tissue using a sterile 20-ml glass syringe and a 19-gauge needle. The saline was partly aspirated back into the syringe and, with the saline, tiny droplets of fat. A single aspiration procedure was performed on each study participant, and the procedure required no local anesthetic. The saline containing adipose tissue droplets was immediately transferred to a glass tube containing 8 ml of hexane, and the tube was capped with a screw cap lined with Teflon (E. I. du Pont de Nemours and Company, Inc., Wilmington, Delaware). The adipose tissue samples in hexane were stored at -20°C for up to 24 hours and at -70°C until extrac-tion (about 6 months).
To measure levels of trans-fatty acids in the adipose tissue samples, we evaporated the hexane at 40°C under nitrogen to near dryness and prepared fatty acid methyl esters by direct mzns-esterification, without further extraction, using the method of Lepage and Roy (24). The fatty acid methyl esters of individual fatty acids were injected in a split mode (1:50) and were separated on a gas chromatograph (model 5890B; Hewlett-Packard Company, Avondale, Penn-sylvania). The gas chromatograph was equipped with a flame ionization detector, electronic pressure con-trol, automatic sampler, and Chemstation software (Hewlett-Packard). The fatty acid methyl esters were separated on a 100-m X 0.25-mm internal diameter capillary silica column with a 0.2-jLtm coating (SP2560; Supelco, Bellefonte, Pennsylvania). The car-rier gas was helium at 60 pounds per square inch; makeup gas was nitrogen at 60 pounds per square inch at the tank. Column linear velocity was set at 20.0 cm per second at an oven temperature of 200°C. The injector and the detector port temperatures were both set at 250°C. The oven temperature (170°C at the start) and electronic pressure (52.3 pounds per square inch at the start) were controlled by a set program for a total run of 108 minutes to optimize the separation of
trans-fatty acids.
Quantitative precision and identification were eval-uated with the use of weighted individual and model mixtures of known fatty acid methyl esters. This iden-tification was confirmed by a mass spectrophotometric analysis and was corroborated by the US Department of Agriculture lipid laboratory in Peoria, Illinois. The results were standardized with National Institutes of Health fatty acid standards A, B, C, D, F, GC-87, and GC-68 (Nu-Check-Prep, Elysian, Minnesota), and re-sponse factors were empirically calculated. Store-bought margarine (optimized for trans-fatty acid sep-aration) was used as a quality control sample pool for adipose tissue. Coefficients of variation were less than 5 percent for both within- and between-batch analyti-cal data for most of the ?rans-fatty acids.
The gas chromatographic technique allowed the measurement of 10 trans-fatty acids in adipose tissue samples, two monounsaturated trans-fatty acids of 16-carbon chain length (16:ln7 and 16:ln9), five mono-unsaturated trans-fatty acids of 18-carbon chain length (18:ln6, 18:ln7, 18:ln8, 18:ln9, and 18:lnlO-12), and three diunsaturated fatty acids of 18-carbon chain length with one or two trans double bonds (18:
2n9trans, nUtrans, 18:2n9m, nUtrans, and 18: 2n9trans, n l 2 m ) . We summed data on the individual trans-fatty acids to obtain the adipose tissue
concen-trations of total trans-fatty acids and subclasses of
trans-fatty acids. Adipose tissue trans-fatty acid
con-centrations were expressed as percentage of total fatty acids.
Statistical methods
We performed the statistical analyses with SPSS for Windows, version 5 (SPSS, Inc., Chicago, Illinois). We computed Spearman correlations between dietary
trans-fatty acid estimates from the FFQ and the
adi-pose tissue concentration of frans-fatty acids. We also computed Pearson correlations between the logarithms of dietary trans-fatty acids and the adipose tissue concentration of trans-fatty acids. To adjust for poten-tially confounding factors, we computed partial corre-lations. For all parametric methods, we used loga-rithms of dietary estimates because the distribution of dietary estimates was skewed. Before logarithmic transformation, we added 0.01 to the dietary estimates of 16:1 trans-fatty acids to handle values of zero. The standard deviation of each geometric mean was com-puted as the product of the geometric mean and the standard deviation of the mean of the natural log of the variable (25). We used the adipose tissue concentra-tions of trans-fatty acids without transformation in both nonparametric and parametric methods because they were normally distributed.
pre-1088 Lemaitre et al.
dictors of adipose tissue concentration of total
trans-fatty acids. The dependent variable was the adipose
tissue concentration of total trans-fatty acids. Predic-tors were chosen by stepwise entry using a p value criterion of 0.05 for entry into the model. Final models included only the predictor variables. All of the values for dietary variables in the linear regression proce-dures were logarithmically transformed.
The trans-fatty acid contribution of individual food items to the study subjects' trans-fatty acid intake estimated by the FFQ was calculated as follows (26):
grams of trans-fatty acid provided by a food item summed across the study subjects grams of trans-fatty acid provided by all foods
summed across the study subjects
X 100.
To determine which line items on the FFQ predicted the most variation in trans-fatty acid intake between the study subjects, we used stepwise linear regression with total dietary trans-fatty acids as the dependent variable (27). The predictors were the log-transformed servings of the FFQ line items. We included in the list of predictors only the FFQ line items estimated to contribute at least 0.4 g of trans-fatty acid per serving.
RESULTS
The study subjects included 27 women and 24 men ranging in age from 51 to 78 years. The majority of participants were Caucasian, with a mean body mass index (weight (kg)/height (m)2) of 26 and a mean dietary intake of 6,481 kJ (geometric mean), 47 g of fat (geometric mean), and 29 percent of energy
de-rived from fat. On average, men were older and heavier than women and had a higher body mass index (table 1). Mean reported energy and fat intake were similar for men and women, although the range of intake was greater among the women.
The mean consumption of total trans-fatty acids estimated from the FFQ was 2.24 g per day (geometric mean). Dietary trans-fatty acids represented, on aver-age, 5.2 percent of dietary fat. Table 2 shows mean intakes of total trans-fatty acids and subclasses of
trans-fatty acids by gender. The mean estimates of
total trans-fatty acid intake were higher in men than in women; however, more variability in trans-fatty acid intake was observed among the women.
In adipose tissue, the mean trans-fatty acid concen-tration was 4.65 percent of total fatty acids. Mean concentrations and standard deviations were higher in men than in women (table 2). On average, trans iso-mers with an 18-carbon chain and one double bond (18:1) accounted for 70 percent (range, 61 percent to 78 percent) of total trans-fatty acids in adipose tissue;
trans isomers with an 18-carbon chain and two double
bonds (18:2) accounted for 25 percent (18-33 per-cent); and trans isomers with a 16-carbon chain and one double bond (16:1) accounted for 5.5 percent (2.3-8.1 percent).
The Spearman correlation coefficient between total dietary trans-fatty acid intake as estimated from the FFQ and total trans-fatty acid concentration in adipose tissue was 0.63 among all of the subjects. Similarly, Pearson correlations between the logarithm of total dietary fatty acids and total adipose tissue
trans-fatty acids were 0.60 (95 percent CI 0.39-0.75) for all
TABLE 1. Personal and dietary characteristics of men and women in a study of the association between food frequency questionnaire estimates of fatty acid intake and adipose tissue trans-fatty acid concentrations, Seattle, Washington, 1996
Personal characteristic Age (years) Race (% Caucasian) Weight (kg) Body mass indext Waist circumference (cm) Dietary intake (per day)
Energy (kJ)$ Energy (kcal)$ Fat(g)* Fat (% of energy) Carbohydrates (g)$ Protein (g)$ Men (n = Mean (SD«) 63.0(8.1) 95.8 84.5(12.7) 27.2(4.1) 96.5(10.9) 6,481 (841) 1,549(201) 49.0(11.7) 29.5 (8.9) 208.9 (29.2) 61.7(9.2) 24) Range 5 1 - 7 8 6 5 - 1 1 0 21.2-37.8 81.3-119.4 3,410-14,413 815-3,445 14.2-106.4 12.3-42.7 107.3-458.8 32.5-167.3 Women (n Mean (SD) 57.7 (6.4) 88.9 66.2(10.2) 25.1 (3.4) 79.5 (9.1) 6,481 (1,038) 1,549(248) 44.7 (13.4) 28.1 (10.7) 204.2 (30.6) 61.7(9.9) = 27) Range 51-74 47.0-91.2 20.3-34.8 66.0-101.6 3,163-19,057 756-4,555 8.5-227.1 10.2-52.4 110.4-473.5 30.0-151.2 * SD, standard deviation, t Weight (kg)/height (m)*. X Geometric mean.
TABLE 2. Food frequency questionnaire estimates of (rans-fatty acid intake and adipose tissue concentrations of trans-fatty acids for men and women, Seattle, Washington, 1996
Dietary estimates of trans-fatty acid intake (g/day)t Total
18:1 18:2 16:1 Total (% of fat)
Adipose tissue concentrations of frans-fatty acids (% of total fatty acids) Total 18:1 18:2 16:1 Men (n = 24) Mean (SD*) 2.57 (0.69) 2.24 (0.60) 0.29 (0.08) 0.016 (0.002) 5.46(1.68) 4.93(1.40) 3.50(1.08) 1.17(0.34) 0.26 (0.07) Range 0.7-8.6 0.6-7.8 0.09-0.83 0-0.020 3.5-10.6 1.74-7.67 1.16-5.70 0.45-1.83 0.13-0.39 Women (n = 27) Mean (SD) 1.99(0.88) 1.78(0.78) 0.24 (0.09) 0.014(0.002) 4.98 (2.39) 4.40(1.17) 3.09 (0.95) 1.08(0.28) 0.24 (0.03) Range 0.3-22.8 0.2-20.6 0.04-2.33 0-0.020 1.8-10.7 2.77-7.63 1.92-5.67 0.66-1.75 0.17-0.31 * SD, standard deviation, t Geometric mean.
subjects, 0.58 (95 percent CI 0.26-0.79) for women, and 0.67 (95 percent CI 0.36-0.84) for men (table 3 and figure 1). In women, the Pearson correlation co-efficient between total dietary frans-fatty acids and total frans-fatty acids in adipose tissue decreased slightly after adjustment for the logarithm of energy intake (table 3). In contrast, in men, the Pearson cor-relation coefficient for total frans-fatty acids increased after adjustment for energy intake and after further adjustment for age and body mass index. Among the subclasses of frans-fatty acids, Pearson correlations between log dietary intake and adipose tissue concen-trations were similar for total frans-fatty acids and
18:1 frans-fatty acids and were lower for 18:2 and 16:1 frans-fatty acids (table 3).
We used linear regression to identify predictors of the adipose tissue concentration of total frans-fatty acids. Factors that were not selected by stepwise re-gression included age, gender, weight, height, body mass index, waist circumference, hip circumference, and dietary fat, protein, carbohydrate, cholesterol, and fiber. Significant predictors were dietary frans-fatty acids and energy intake, which predicted 41 percent of the variability in adipose tissue concentration of
trans-fatty acids (table 4). No significant interactions were
found between dietary frans-fatty acids and weight,
TABLE 3. Correlation between food frequency questionnaire estimates of trans-fatty acid intake and adipose tissue frans-fatty acid concentrations, expressed as percentage of total fatty acids, in men and women, Seattle, Washington, 1996
frans-fatty acids Total 18:1 18:2 16:1 Total 18:1 18:2 16:1 Spearman coefficient 0.55** 0 . 5 8 * * 0.38 0.48* 0 . 6 7 * * * 0 . 6 8 * * * 0.50* 0.54** r 0.58 0.56 0.45 0.51 0.67 0.67 0.51 0.49 Unadjusted 95% Clf Pearson coefficient! Adjusted for energy intake r 95% CI Women (n = 27) 0.26-0.79 0.23-0.78 0.08-0.71 0.16-O.74 0.51 0.16-0.74 0.46 0.10-O.71 0.45 0.08-0.71 0.42 0.05-0.69 Men (n = 24) 0.36-0.84 0.36-0.84 0.13-0.76 0.11-0.75 0.71 0.43-0.86 0.69 0.40-0.85 0.56 0.20-0.79 0.42 0.02-0.70
Adjusted for energy intake, age, and body mass index
r 0.52 0.48 0.47 0.30 0.76 0.75 0.58 0.42 95% CI 0.17-0.75 0.12-0.73 0.11-0.72 -0.09 to 0.61 0.51-0.89 0.50-0.88 0.23-0.80 0.02-0.70 * p < 0.05; ** p < 0.01; *** p < 0.001.
t Estimates of frans-fatty acid intake were logarithmically transformed. $ CI, confidence interval.
1090 Lemaitre et al. Women 10 ^ 8
I •
CD CO o a.5
4 _ 2 S o -0.5 (0.32) 0 (D 0.5 (3.16) 1 (10) 1.5 (31.6)Total TFA estimated in the diet, Iog10 (back transformations, grams per day)
Men 10 ^ 8 0 3 I/I a>
s
a. !: 2I
-0.5 (0.32) 0 (1) 0.5 (3.16) (10) 1.5 (31.6)Total TFA estimated in the diet, Iog10 (back transformations, grams per day) FIGURE 1. Scatterplot of the relation between log total dietary
frans-fatty acid (TFA) intake, as estimated by food frequency ques-tionnaire, and total adipose tissue concentration of frans-fatty acids in women (n = 27) and men (n = 24), Seattle, Washington, 1996.
body mass index, or energy intake. However, there was a significant interaction with gender (p = 0.02); therefore, correlation coefficients and results of re-gression analyses are stratified by gender in tables 3 and 4.
Beyond the validation of the FFQ trans-fatty acid estimates, we thought it would be of interest to deter-mine which foods listed on the FFQ were major con-tributors to the dietary trans-fatty acid estimates of the study subjects. The foods with the highest estimated amounts of total trans-fatty acids per serving are listed in the first column of table 5. The foods with the highest amounts of trans-fatty acids were apple or cherry pie with a double crust (7.0 g per serving) and biscuits (4.7 g per serving). In the second column of table 5, we list the foods that contributed the most to the study subjects' ^raws-fatty acid intake as estimated by the FFQ. The list of foods is specific to these study subjects and might differ in other study populations. Stick margarine made the biggest contribution (20 percent of total trans-fatty acid intake), followed by biscuits and muffins (11 percent). The third column in table 5 lists the foods that contributed the most to the variation in trans-fatty acid intake between the study subjects. The answer to one of the FFQ questions—the frequency with which fat was added to vegetables, potatoes, rice, and beans, together with the serving size of the added fat—explained 45 percent of the between-subject variation in trans-fatty acid intake. The FFQ line items listed in table 5 together explained 80 percent of the variation in estimated trans-fatty acid intake.
DISCUSSION
In this study, we estimated usual intake of
trans-fatty acids with an FFQ using information on the trans-fatty acid content of selected foods recently
re-leased by the US Department of Agriculture. Intake of
trans-fatty acids as estimated by the FFQ was highly
correlated with an objective marker of trans-fatty acid intake, the adipose tissue concentration of trans-fatty acids (Pearson correlations were 0.5 in women and 0.7 in men). These correlations suggest that the FFQ cap-tured most of the trans-fatty acid-containing foods consumed by the study subjects, that the information available from the Department of Agriculture was sufficient, and that the assumptions made in estimating the trans-fatty acid content of the FFQ line items were adequate. Since the FFQ assessed trans-fatty acid in-take in the previous year, while the adipose tissue level reflected trans-fatty acid intake over the previous 2-3 years (28, 29), the high correlation between the two measures also suggests no major change in the
trans-fatty acid content of foods consumed over the previous
few years.
The mean estimates of the study subjects'
trans-fatty acid intake (2.2 g per day; 5.2 percent of fat
intake) were lower than estimates based on food dis-appearance or food availability, reported at 8-15 g per
TABLE 4. Results from multivariate linear regression of adipose tissue trans-fatty acid concentrations, expressed as percentage of total fatty acids, on food frequency questionnaire estimates of trans-fatty acid intake, Seattle, Washington, 1996
Predictor*
Log dietary fran^-fatty acidsf Log kcal Constant All subjects 3.0903 (<0.0001) - 3 . 2 2 7 9 (0.04) 13.8569
Regression coefficient (p value) Women (n = 27) 2.3134 (0.007) -2.5253 (0.27) 11.7435 Men (n = 24) 4.8207 (0.0002) -3.9507 (0.08) 15.5873 * R* in the multivariate model was 0.41 for all subjects, 0.37 for women, and 0.53 for men.
t Log of grams per day.
day (30). However, the estimates obtained from this study were similar to previously reported mean
trans-fatty acid intakes estimated with an FFQ (9, 23, 31).
Using the FFQ developed for the Nurses' Health Study and estimates of the trans-fatty acid content of foods that predated the estimates now available from the Department of Agriculture, a mean total trans-fatty acid intake of 3.4 g per day (5.8 percent of fat intake) was reported for postmenopausal women (23), and a mean total intake of 3.0 g per day (4.7 percent of fat intake) was reported for men (31). Mean adipose tis-sue concentrations of trans-fatty acid in this study were similar to previously reported values (10).
The reason why we observed higher correlation coefficients in men than in women in this study is not known. Possibilities include a more stable weight in the men, a more stable diet in the men, or greater reliability of the men in assessing dietary intake in this study. The correlation coefficients we observed in women were similar to those reported in the study of postmenopausal women (23). In that study, a Spear-man correlation of 0.5 was reported between adipose tissue concentrations of trans-fatty acids and total dietary trans-fatty acid levels estimated from an FFQ and expressed as percentage of fat intake. However,
using the same FFQ with male subjects, a Spearman correlation of 0.3 between adipose tissue trans-fatty acids and total dietary trans-fatty acids was reported (31), in sharp contrast to the correlation of 0.7 we observed among men. The improvement in the corre-lation coefficient in this study might be explained by a number of differences between the two studies, includ-ing different FFQs, different study populations, and, perhaps most importantly, different estimates of the
trans-fatty acid content of foods. In this study, we used
information recently released by the US Department of Agriculture (13). This information was not available at the time of previous validation studies. The quality of the newly available trans-fatty acid data may have allowed for a more precise estimate of dietary intake in our study.
Measurement error in an exposure of interest de-creases the power to detect associations with disease risk (32). For example, if one standard deviation of the dietary trans-fatty acid measure was associated with a true fivefold increase in disease risk but the trans-fatty acid measure was contaminated with 70 percent ran-dom error (as suggested by a 0.3 correlation with a gold standard), the relative risk that would be observed would in fact be only 1.6. Contamination of the
trans-TABLE 5. Major contributors to dietary trans-fatty acid (TFA) intake as estimated by food frequency questionnaire, Seattle, Washington, 1996
10 highest sources ofTFAs Food
Double-crust dessert pies Biscuits
Pot pie, beef
Single-crust dessert pies Frosted cakes (yellow or
devil's food) Fried chicken, breaded Danish pastry Fried fish, breaded Crackers Stick margarine TFA (g/serving) 7.0 4.7 4.2 3.5 3.5 2.8 2.5 2.4 2.3 1.9 10 highest contrtoutors to TF/ intake in the study subjects
Line item*
Stick margarine Biscuits and muffins Crackers
Dessert pies Dark breads Cookies Tub margarine
Doughnuts, cakes, pastries Cookies, low fat
Fried fish % of total TFA intake 19.7 11.2 6.7 6.7 6.5 5.7 5.2 3.9 3.5 2.7
10 highest contributors to the variation in TFA intake between subjects
Line item*
Fat added to vegetables, potatoes, rice, or beans Cookies
Biscuits and muffins Hamburger Crackers Dark breads
Doughnuts, cakes, pastries Beef
Fried chicken, breaded Mayonnaise Cumulative 0.45 0.55 0.65 0.70 0.71 0.76 0.79 0.81 0.81 0.81 A line item is a single question on the food frequency questionnaire and might refer to a single food or a group of foods.
1092 Lemaitre et al.
fatty acid measure with 30-50 percent random error, as for the FFQ estimate we validated, would result in observed relative risks of 3.1 and 2.2, respectively. Also, if dietary trans-fatty acid was a strong con-founder of the association of another risk factor with disease risk, minimizing measurement error in dietary
trans-fatty acid would be essential in order to
ade-quately control for its confounding effect (33). A strength of this study is that we designed it to minimize "noise" in our exposure (dietary trans-fatty acid) and our gold standard (adipose tissue trans-fatty acid). Specifically, we selected subjects who reported having had a stable diet and body weight over the past 2 years. These criteria were important, because studies have consistently demonstrated that the composition of stored exogenous fatty acids in adipose tissue re-flects fatty acid composition in the diet over a period of years (28, 29). Therefore, we excluded subjects who had made major changes in their diet, because their current diet would not accurately represent their long term diet, which is the primary determinant of tissue
trans-fatty acid. Similarly, subjects with stable body
weight were chosen in order to minimize variations in the rate of incorporation of fatty acids in adipose tissue, thereby minimizing the impact of minor dietary changes. This validation study provides strong evi-dence that the FFQ can estimate dietary trans-fatty acid accurately. However, correlations between di-etary and adipose tissue trans-fatty acid are likely to be attenuated in studies conducted in the general pop-ulation, where the exclusions used in this study would not be applicable.
Given the absence of trans-fatty acids in nutrient databases, the FFQ we validated is an important new tool for assessing usual intake of trans-fatty acids. Although a biomarker such as adipose tissue concen-tration of trans-fatty acids might be preferred to di-etary assessment, an FFQ is cheaper and likely to result in higher response rates in epidemiologic stud-ies. Equally important, the availability of two totally different and highly correlated measures of trans-fatty acid intake is invaluable for epidemiologists interested in studying the potential association of trans-fatty acids with coronary heart disease.
ACKNOWLEDGMENTS
This work was supported by a Pilot and Feasibility Grant from the Clinical Nutrition Research Unit, Division of Me-tabolism, Endocrinology, and Nutrition, University of Washington (supported by National Institute of Diabetes and Digestive and Kidney Diseases grant P30-DK35816). In addition, a portion of this work was conducted through the Clinical Research Center of the University of
Washing-ton Medical Center (supported by National Institutes of Health grant M01-RR00037).
REFERENCES
1. Gaziano JM, Manson JE. Diet and heart disease: the role of fat, alcohol, and antioxidants. Cardiol Clin 1996;14:69-83. 2. Office of the Surgeon General, US Public Health Service. The
Surgeon General's report on nutrition and health. Washington, DC: US GPO, 1988. (DHHS publication no. (PHS) 8 8 -50210).
3. Committee on Diet and Health, National Research Council. Diet and health: implications for reducing chronic disease risk. Washington, DC: National Academy Press, National Acad-emy of Sciences, 1989.
4. US Public Health Service. Healthy People 2000: national health promotion and disease prevention objectives. Washington, DC: US GPO, 1991. (DHHS publication no. (PHS) 9 1 -50212).
5. Mensink RP, Katan MB. Trans monounsaturated fatty acids in nutrition and their impact on serum lipoprotein levels in man. Prog Lipid Res 1993,32:111-22.
6. Khosla P, Hayes KC. Dietary /ra/w-monounsaturated fatty acids negatively impact plasma lipids in humans: critical re-view of the evidence. J Am Coll Nutr 1996;15:325-39. 7. Willett WC, Stampfer MJ, Manson JE, et al. Intake of trans
fatty acids and risk of coronary heart disease among women. Lancet 1993;341:581-5.
8. Ascherio A, Hennekens CH, Buring JE, et al. Trans-fatty acids intake and risk of myocardial infarction. Circulation 1994;89: 94-101.
9. Pietinen P, Ascherio A, Korhonen P, et al. Intake of fatty acids and risk of coronary heart disease in a cohort of Finnish men: The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study. Am J Epidemiol 1997;145:876-87.
10. Kris-Etherton PM, Emken EA, Allison DB, et al. Trans fatty acids and coronary heart disease risk. Am J Clin Nutr 1995; 62(suppl):655S-708S.
11. Willett WC, Ascherio A. Trans fatty acids: Are the effects only marginal? Am J Public Health 1994;84:722-4. 12. ASCN/AIN Task Force on trans Fatty Acids. Position paper
on trans fatty acids. Am J Clin Nutr 1996;63:663-70. 13. Exler J, Lemar L, Smith J. Fat and fatty acid content of
selected foods containing trans-fatty acids. (Special purpose table no. 1). Washington, DC: US Department of Agriculture, 1996.
14. Martin LJ, Lockwood GA, Kristal AR, et al. Assessment of a food frequency questionnaire as a screening tool for low fat intakes. Control Clin Trials 1997;18:241-50.
15. Thompson FE, Byers T. Dietary assessment resource manual. J Nutr 1994;124(suppl):2245S-317S.
16. Block G, Woods M, Potosky A, et al. Validation of a self-administered diet history questionnaire using multiple diet records. J Clin Epidemiol 1990;43:1327-35.
17. Willett WC, Reynolds RD, Cottrell-Hoehner S, et al. Valida-tion of a semi-quantitative food frequency quesValida-tionnaire: com-parison with a 1-year diet record. J Am Diet Assoc 1987;87: 43-7.
18. Munger RG, Folsom AR, Kushi LH, et al. Dietary assessment of older Iowa women with a food frequency questionnaire: nutrient intake, reproducibility, and comparison with 24-hour dietary recall interviews. Am J Epidemiol 1992,136:192-200. 19. Rimm EB, Giovannucci EL, Stampfer MJ, et al. Reproduc-ibility and validity of an expanded self-administered semi-quantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135:1114-26.
20. Schakel SF, Sievert YA, Buzzard IM. Sources of data for developing and maintaining a nutrient database. J Am Diet Assoc 1988;88:1268-71.
21. Kristal AK, Shattuck AL, Williams AE. Food frequency ques-tionnaires for diet intervention research. Proceedings of the 17th National Nutrient Databank Conference, Washington, DC, June 1992. Washington, DC: International Life Sciences Institute, 1994:110-25.
22. Hirsch J, Farquhar JW, Ahrens EH, et al. Studies of adipose tissue in man: a microtechnic for sampling and analysis. Am J ClinNutr 1960;8:499-511.
23. London SJ, Sacks FM, Caesar J, et al. Fatty acid composition of subcutaneous adipose tissue and diet in postmenopausal US women. Am J Clin Nutr 1991;54:340-5.
24. Lepage G, Roy CC. Direct transesterification of all classes of lipids in a one-step reaction. J Lipid Res 1986;27:114-20. 25. Bickel PJ, Doksum KA. Mathematical statistics: basic ideas
and selected topics. San Francisco, CA: Holden-Day, Inc, 1977.
26. Block G, Dresser C, Hartman A, et al. Nutrient sources in the American diet: quantitative data from the NHANES II survey. II. Macronutrients and fats. Am J Epidemiol 1985;122:27-40. 27. Willett WC. Food frequency methods. In: Willett WC, ed.
Nutritional epidemiology. New York, NY: Oxford University Press, 1990:69-91.
28. Dayton S, Hashimoto S, Dixon W, et al. Composition of lipids in human serum and adipose tissue during prolonged feeding of a diet high in unsaturated fat. J Lipid Res 1966;7:103-111. 29. Beynen AC, Hermus RJ, Hautvast JG. A mathematical rela-tionship between the fatty acid composition of the diet and that of the adipose tissue in man. Am J Clin Nutr 1980,33:81-5. 30. Hunter JE, Applewhite TH. Reassessment of trans fatty acid
availability in the US diet. Am J Clin Nutr 1991;54:363-9. 31. Hunter DJ, Rimm EB, Sacks FM, et al. Comparison of
mea-sures of fatty acid intake by subcutaneous fat aspirate, food frequency questionnaire, and diet records in a free-living pop-ulation of US men. Am J Epidemiol 1992;135:418-27. 32. Armstrong BK, White E, Saracci R. Exposure measurement
error and its effects. In: Principles of exposure measurement in epidemiology. New York, NY: Oxford University Press, 1992. 33. Marshall JR, Hastrup JL. Mismeasurement and the resonance of strong confounders: uncorrelated errors. Am J Epidemiol 1996;143:1069-78.