Effect of a low-fat diet on fatty acid composition in red cells, plasma phospholipids, and cholesterol esters: investigation of a
the Cancer Prevention Program, Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA (IBK and MK), and the Department of Medicine, Cardiovascular Health Research Unit (RNL) and the School of Public Health and Community Medicine (MK), University of Washington, Seattle, WA
ABSTRACT
Background: The utility of fatty acids (FAs) as biomarkers of total fat intake is unknown.
Objective: We compared FA changes in red cells (RCs), plasma phospholipids (PLs), and cholesterol esters (CEs) in response to a low-fat diet (LFD) and a moderate-fat diet (MFD) and assessed whether individual or combination of FAs predict LFD.
Design: Postmenopausal women (n = 66) were randomly assigned to receive an LFD (17% of energy from fat) or an MFD (34% of energy from fat) for 6 wk. All foods were provided. FAs in diets and blood were determined by gas-liquid chromatography. FA changes between baseline and end of study were compared across diets by using t tests. FA predictors of an LFD were selected by logistic regression.
Results: Many FAs in RCs, PLs, and CEs responded differently to the 2 diets. Changes from baseline with an LFD for palmitic acid (16:0) (3–11% increase), behenic (22:0) and lignoceric (24:0) acids (3–20% decrease, in RCs and PLs only), cis-monounsaturated FA (MUFA) (25–35% increase), linoleic acid (18:2n–6) (11–13% decrease), trans octadecanoic acids (trans 18:1) (7–20% decrease), and n–6 highly unsaturated FA (HUFA) (2–8% increase) were significantly different from changes with an MFD. Individually, 18:2n–6 and trans 18:1 were strong predictors of an LFD [receiver operating characteristic (ROC) curves: 0.92–0.80). A logistic regression model with trans 18:1, 18:2n–6, and vaccenic acid (18:1n–7) predicted an LFD with high specificity and sensitivity (ROC curves: 0.99).
Conclusions: Saturated FA, cisMUFA, n–6 HUFA, and exogenous FAs greatly differed in their response to the LFD and MFD. Parallel responses were observed in RCs, PLs, and CEs. A model with a combination of FAs almost perfectly differentiated the consumption of 34% fat from that of 17% fat.
Key Words: Fatty acids trial biomarkers dietary fats low-fat diet
INTRODUCTION
There is much scientific and public health interest in the role of dietary fat in causing serious chronic diseases such as cardiovascular disease and cancer (1-4). For both observational and intervention studies, measurement of total fat intake is crucial to the validity and interpretation of findings. Given that fat intake of participants can rarely be directly observed, assessments of dietary fat intake rely on self-reports, which often have recall bias and measurement errors, including changes in participants' diets when filling out questionnaires (5, 6). A biological marker of total fat intake would be useful, particularly one that can be ascertained from blood samples. Currently, no reliable biomarkers of total fat intake are known (7-9), and little information exists on fatty acid response to changes in fat quantity.
Blood and adipose tissue reasonably represent dietary intake of exogenous fatty acids (those that are not synthesized in vivo) such as linoleic acid (18:2n–6), trans fatty acids, and long-chain n–3 (10-13). However, most dietary fatty acids can be synthesized de novo; thus, their source of origin is inconclusive. Furthermore, the extent to which individual exogenous fatty acids reflect total dietary fat intake is unknown. The objectives of this study were to investigate fatty acid alterations in response to change in total dietary fat intake, to determine which blood specimen best reflects the change in dietary fat intake, and to investigate whether a combination of fatty acids might be a valid marker of total fat intake. To address these questions, we conducted a randomized dietary trial of low-fat (17% of energy) and moderate-fat (34% of energy) intake among 66 postmenopausal women.
SUBJECTS AND METHODS
Study design and participants
This study was a randomized, controlled, dietary intervention trial conducted at the Fred Hutchinson Cancer Research Center and the Clinical Research Center at the University of Washington in Seattle, WA. The study was approved by the institutional review boards at both institutions.
Participants were healthy postmenopausal women aged 50–79 y. Exclusion criteria were any medical condition that required dietary modification (eg, diabetes, renal disease, or hypercholesterolemia), morbid obesity [body mass index (BMI; in kg/m2) >40], regular consumption of the main meal outside the home 2 times/wk, unsatisfactory completion of food records at any time before randomization, recent weight loss (>10% of usual body weight during the previous 6 mo), or a usual alcohol intake averaging > 2 standard drinks/d (>24 g alcohol).
Potentially eligible women completed two 4-d food records before randomization. Sixty-six women were then randomly assigned (33 in each group) to consume either a low-fat (17% of energy) or a moderate-fat (34% of energy) diet for 6 wk. Reasonable parity between subjects in the 2 dietary groups by age, BMI, and postmenopausal hormone use was ensured by using an adaptive randomization program (14). Sixty-one women completed the study—31 in the low-fat diet group and 30 in the moderate-fat diet group. All 5 participants who dropped out of the study did so for personal, not diet-related reasons.
Clinical measurements
Participants attended a total of 4 clinic visits, 2 at baseline (3 d apart) and 2 at the end of the study (also days apart) after a 12–14-h fast. At each visit, participants were weighed to the nearest 100 g, and 20 mL venous blood was collected into one 10-mL plain evacuated tube and one 10-mL evacuated tube containing K3-EDTA (1 mg/mL). Serum and plasma were divided into aliquots, and red cells were washed twice with 0.9% saline before storage at –70°C until analysis.
Diet design
The study research dietitian designed the low- and high-fat diets by using NDS software (NCC, University of Minnesota, Minneapolis, MN) (15). Women's energy requirements were estimated by standard equations for the basal metabolic rate based on height, weight, and age and multiplied by 1.51 as the activity factor (16). Participants were weighed weekly, and, according to whether they lost or gained >2.0 kg/wk from their initial baseline weight, they were placed on the higher or lower energy level as appropriate.
We designed 7 different breakfasts (15% of energy), lunches (25% of energy), dinners (50% of energy), and snacks (10% of energy) for each diet at 6 energy levels. All food was prepared in advance at the Nutrition Research Kitchen at the University of Washington Clinical Research Center by dietary technicians. Participants were free to chose up to 0.628 MJ (150 kcal)/d from foods low or devoid in fat, and they restricted their intake of alcoholic beverages to a maximum of 6 ounces (170 mL)/wk.
We prepared a weighted composite of each diet on 4 separate occasions for the median energy level, 8.4 MJ, and chemically analyzed a weighted mixture of these 4 composites (Table 1). Protein was estimated by the measuring nitrogen by using the Kjeldahl method and then multiplying by 6.25 (17). Total fat was confirmed by 2 methods: acid hydrolysis procedure and enzymatic digestion with chloroform-methanol extraction (17). Dietary fiber was measaured by gravimetric methods (17). Total carbohydrate was calculated as total intake minus water, ash, protein, fat, and dietary fiber. Fatty acid analyses are described later.
Food records and measures of adherence
The 8-d food records, reflecting the participants' usual diets at baseline, were checked for accuracy by the research dietitian, coded, and analyzed by using the NDS data system (version 2.8; NCC, University of Minnesota). Participants were given a form on which to record their consumption of the foods provided to them during the trial. Any food not consumed was returned to the research kitchen and weighed. Foods consumed in the 0.628 MJ "allowance" were also recorded and analyzed with the NDS data system.
Analyses of fatty acids
Total red cell fatty acids were extracted with isopropanol and chloroform according to the method of Rose and Oklander (18). Total lipids from plasma were extracted by the method of Folch (19) with a 2:1 chloroform:methanol solution and several washes. Plasma phospholipids and plasma cholesterol esters were separated by one-dimensional thin-layer chromatography by using 250-μm Silica Gel G plates (Analtech Inc, Newark, DE) and a 67.5:15:0.75 hexane:ethyl ether:acetic acid (+0.005% butylated hydroxytoluene) development solvent.
The fatty acid methyl esters (FAMEs) for all samples, including diet samples, were prepared by direct transesterification by using the method of Lepage and Roy (20). Gas-liquid chromatography was performed on samples dissolved in hexane. The FAME of individual fatty acids of red cell membranes, plasma phospholipids, and cholesterol esters were separated on a gas chromatograph (model 5890B, series II; Hewlett-Packard, Avondale, PA) equipped with a flame ionization detector, automatic sampler (Hewlett-Packard 7673), and CHEMSTATION and MUSTANG software (Hewlett-Packard) as previously described (21).
Quantitative precision and peak identification were evaluated with the use of weighted individual and model mixtures of known FAMEs. This identification was confirmed by a mass spectrophotometric analysis performed by the US Department of Agriculture Lipid Laboratory (Peoria, IL). Our quantitative results were standardized with the National Heart Institute Fatty Acid Standards A, B, C, D, and F (Nu-Check-Prep, Elysian, MN), and response factors were calculated. Fatty acid composition is expressed as relative weight percentages. Pooled red cells and plasma were used as additional controls that were run with each batch of study samples. Interassay CVs were on the average of 3.5% for most of the fatty acids that were present at levels of 1%. All laboratory personnel were blinded with respect to the diet group randomization.
Statistical analyses
The data collected at the 2 baseline visits and at the 2 end-of-study visits were averaged to yield a single value for baseline and for end of study. In descriptive analyses, we compared mean baseline characteristics of women randomly assigned to the low-fat diet and women randomly assigned to the moderate-fat diet by using two-sided t tests and two-sided statistical significance at P value < 0.05. To illustrate the effect of diet on fatty acid composition in the 3 specimens, we computed percentages of change for each fatty acid at the end of the study. To assess the effect of diet on the changes in fatty acid composition, we compared the difference in fatty acid proportions at baseline and at end of study across the diet groups by using t tests and two-sided statistical significance at P value < 0.05. To assess which fatty acid combination predicted randomization to the low-fat diet, we performed stepwise logistic regression with randomization to low fat as the outcome (22). Fatty acid proportions at the end of the study were the predictors. To reduce the number of possible predictors submitted to the stepwise procedure, 3 sets of fatty acids were initially submitted in 3 separate stepwise logistic regression analyses. The first set included fatty acids that originate from the diet, namely trans fatty acids, 18:2n–6, linolenic acid (18:3), and the long-chain saturated fatty acids behenic acid (22:0) and lignoceric acid (24:0). The second set included fatty acids that can be synthesized endogenously from carbohydrates, palmitic acid (16:0), stearic acid (18:0), palmitoleic acid (16:1n–7), vaccenic acid (18:1n–7), and oleic acid (18:1n–9). The third set included polyunsaturated fatty acids (PUFAs) of 20 carbons. With each analysis, fatty acid predictors were selected stepwise by using a P value of entry 0.1. Then, by using the fatty acids that were selected in each of the 3 initial analyses, a last stepwise analysis was performed for a final fatty acid selection. We computed sensitivity and specificity of the final model fitted by the logistic regression. Global performance of the model in predicting randomization to low-fat diet was summarized by the area under the receiver operating characteristics (ROC) curve. These procedures were performed separately for each blood specimen. All statistical analyses were performed by using STATA software (version 8.2; Stata Corp, College Station, TX).
RESULTS
The baseline characteristics of the postmenopausal women in the study are shown in Table 2. None of the differences between dietary groups were statistically significant at P < 0.05. The fat intake of the women at baseline was lower than the typical US intake in both groups (on average, 29% of energy intake). On average, the women were aged 63 y, with a BMI of 26. The baseline diets of women randomly assigned to the to low-fat diet and women randomly assigned to the moderate-fat diet were not significantly different (Table 2). In addition, mean baseline proportions of individual fatty acids, measured in red cells, plasma phospholipids, and plasma cholesterol esters, were not significantly different for the 2 dietary groups (Table 3).
On average, a small weight loss was observed between baseline and the end of study in both groups. No statistically significant differences were apparent when the weight loss was compared between the 2 diets [(± SD) moderate fat: –0.96 ± 0.92 kg; low fat: –1.3 ± 1.40 kg;, P = 0.32).
Baseline fatty acid compositions and changes are shown (Table 3) after 6 wk of randomization to the low-and moderate-fat diets for 3 blood specimens, namely red cells, plasma phospholipids, and plasma cholesterol esters. Changes in many fatty acids differed between the 2 diets, as shown in Table 3. For most fatty acids, the direction of change was consistent among the 3 types of specimens.
Saturated fatty acids
Among the saturated fatty acids, the response of 16:0 to the 2 diets was significantly different with a consistent increase from baseline on the low-fat diet and decrease on the moderate-fat diet in all specimens (Table 3). The long-chain saturated fatty acids 22:0 and 24:0 decreased on the low-fat diet and increased on the moderate-fat diet in red cells and plasma phospholipids, which again resulted in a significant difference between the 2 diets. The proportions of these fatty acids were below detection in the cholesterol esters. The proportions of 18:0 differed in response to the 2 diets only in cholesterol esters (P = 0.03). Total saturated fatty acids in the 3 specimens changed in the direction of 16:0, and the difference in changes across the 2 diets remained significant.
cis Monounsaturated fatty acids
Among cis monounsaturated fatty acids, the greatest difference between the 2 diets was observed in the 16:1 fatty acids. Both the n–7 and n–9 fatty acid families dramatically changed in opposite directions in response to the diets (increasing 14–34% on the low-fat diet and decreasing 4–22% on the moderate-fat diet). The response of 18:1 fatty acids (n–7 and n–9) to the diets mirrored that of the 16:1 families, although the magnitude of the response was smaller.
n–6 Polyunsaturated fatty acids
18:2n–6 Decreased from baseline in red cells, plasma phospholipids, and cholesterol esters in response to the low-fat diet (–11.2%, –12.9%, and –10.8%, respectively) and increased on the moderate-fat diet (2.6%, 3.9%, and 6.2%, respectively) in all 3 specimens. The difference in change from baseline across the 2 diets was highly significant. In contrast, dihomo--linoleic acid (20:3n–6) and arachidonic acid (20:4n–6) increased in response to the low-fat diet in all 3 specimens, which resulted in significant differences between the 2 diets. Another long-chain PUFA, docosatetraenoic acid (22:4n–6), also responded differently to the diets but only in the plasma phospholipids (Table 3). The total n–6 fatty acids, representing the overall change in both 18:2n–6 and long-chain n–6 fatty acids, showed a small but significant difference between the 2 diets.
n–3 Polyunsaturated fatty acids
-Linolenic acid (18:3n–3) decreased from baseline on both diets in red cells, and its response did not differ between the diets. In plasma phospholipids and cholesterol esters, 18:3n–3 decreased on the low-fat diet and increased on the moderate-fat diet, which resulted in a borderline difference in responses (P = 0.06). Docosahexaenoic acid (22:6n–3) consistently increased from baseline on both diets in all 3 specimens.
Total polyunsaturated fatty acids
The total PUFAs (n–6 and n–3 combined) decreased from baseline on the low-fat diet and increased on moderate-fat diet. Thus, the differences in changes across the 2 diets were highly statistically significant in the 3 specimens.
trans Fatty acids
The total trans octadecanoic acids (trans 18:1) consistently decreased on the low-fat diet and increased on the moderate-fat diet in all 3 specimens, which resulted in highly significant differences between the 2 diets. The total trans-hexadecenoic acids (trans 16:1) and linoelaidic acids (trans 18:2), which are present in low proportions, did not change from baseline in response to the low-fat diet; however, trans 16:1 increased from baseline in response to the moderate-fat diet.
Fatty acid proportions as a predictor of total fat intake
One purpose of the study was to investigate whether a combination of fatty acids might effectively predict randomization to the low-fat diet. To this end, we selected end-of-study fatty acids that predicted the dietary group assignment by using stepwise logistic regression, as described in Subjects and Methods. For each specimen, we were able to define a multivariate model that predicted randomization to the low-fat diet with high specificity and sensitivity (Table 4). Three fatty acid predictors were common to all specimens: trans fatty acid isomers of trans 18:1, 18:2n–6, and 18:1n–7. Consistent with the univariate analyses (Table 3), lower trans 18:1, lower 18:2n–6, higher 18:1n–7, and, for red cells and cholesterol esters, higher proportions of selected PUFAs predicted randomization to the low-fat diet. In the multivariate models, these fatty acids were all significant predictors in the red cells (Table 4); in the plasma phospholipids, 18:2n–6 was a borderline predictor (P = 0.06); and in the cholesterol ester, trans 18:1 (P = 0.05), 18:2n–6 (P = 0.06), 18:1n–7 (P = 0.06), and 20:3n–6 (P = 0.07) were borderline predictors. As reflected by the area under the ROC curve (0.99 for all specimens), the ability of the models shown in Table 4 to discriminate between the 2 dietary groups was high. Prediction of the low-fat diet was better achieved with combinations of fatty acids than with individual fatty acids. However, 18:2n–6 and trans 18:1 were individually high predictors of low-fat intake. The area under the ROC curve for models with only 18:2n–6 was 0.90 for phospholipids, 0.84 for red cells, and 0.90 for cholesterol ester. Corresponding values for trans 18:1 alone were 0.92, 0.80, and 0.85.
DISCUSSION
In this controlled dietary trial, we have shown that, in weight-stable postmenopausal women, a decrease in dietary fat consumption from a usual diet of 29% energy at baseline to 17% during 6 wk resulted in widespread changes in fatty acid proportions in red cells, plasma phospholipids, and cholesterol esters. These fatty acid changes were absent or in the opposite direction in the women randomly assigned to the moderate-fat diet (34% energy). The magnitude of the difference in fatty acids between the 2 diet groups was partly attributable to a small increase in fat intake from baseline in the moderate-fat group as well as the large decrease in fat intake in the low-fat group. In addition, we found that a combination of fatty acids could be used to predict randomization to low-fat diet with high sensitivity and specificity.
In response to a low-fat diet, we observed definite patterns of fatty acid alterations. Most exogenous fatty acids (originating from the diet only) consistently showed a decrease in the 3 specimens. This was not surprising because the diets were designed to decrease fat across all important dietary fatty acids, and exogenous fatty acids are known to reflect dietary intake (23). In this study, the dietary fats were selected on the basis of the type of fat that is normally consumed in the American diet. The main PUFA in the American diet is 18:2n–6, which was decreased 41% on the low-fat diet, which resulted in an 11–13% decrease in all 3 blood specimens. The decrease in 18:2n–6 in red cells, plasma phospholipids, and cholesterol esters in response to a dietary decrease is well documented (23-25). Although it has been suggested that cholesterol esters adequately capture 18:2n–6 change in the diet (26), we found that red cells and plasma phospholipids were equally informative in this short-term intervention. Our study also showed a decrease in blood proportions of the trans 18:1 fatty acids with reduction of total fat intake. The proportions of total trans 18:1 in the investigated specimens were relatively low, ranging from 0.4% to 1.5% of total fatty acids. Despite these low proportions, trans 18:1 decreased 7–20% in all specimens; the lowest decrease was in the cholesterol esters. The total trans 18:2 decreased slightly with both diets, whereas total trans 16:1 did not decrease in response to the low-fat diet but increased in response to the moderate-fat diet. The total trans 18:2 and trans 16:1 were not measured in the experimental diets, and it is possible that the composition of these fatty acids differed in the 2 diets.
We found systematic increases in many endogenous fatty acids in response to the low-fat diet, despite dietary decrease in these fatty acids. Because we were measuring proportions of fatty acids rather than absolute amounts, a decrease in the proportion of some fatty acids will be compensated, by definition, with an increase in the proportion of other fatty acids. Diets high in carbohydrates and calories lead to de novo synthesis of several fatty acids (27-29). These include saturated fatty acids, namely, myristic acid (14:0) and 16:0 and monounsaturated fatty acids of the n–7 and n–9 series (16:1n–7, 16:1n–9, 18:1n–7, and 18:1n–9). During the de novo synthesis of fatty acids, the fatty acid synthase produces 16:0, which can be either desaturated to 16:1n–7 or elongated to 18:0. In turn, 18:0 is desaturated to 18:1n–9 by -9 desaturase, and 16:1n–7 is elongated to 18:1n–7 by long-chain elongase (30, 31). In agreement with our findings, Raatz et al (32) reported an increase in 18:1n–9 with a low-fat diet; however, they did not report other cis monounsaturated fatty acids. The proportions of 18:0 remained constant despite a 2-fold decrease on the low-fat diet. In support of a lack of response to diet, it was reported that 18:0 does not appear to accumulate in response to dietary treatment (30). Our study also suggests a possible increase in -oxidation to produce higher proportions of 16:1n–9 from endogenously produced 18:1n–9 on a low-fat diet.
Long-chain n–6 PUFAs [(n–6 highly unsaturated fatty acid (HUFA), carbon 20], generally increased on the low-fat diet, despite lower concentrations in the diet. The increase in HUFAs is important because it occurs despite the dilution of the essential fatty acids by the endogenous, carbohydrate-derived fatty acids synthesized on the low-fat diet. These changes likely reflect the influence of 18:2n–6 on fatty acid metabolism, namely, a release of inhibition of -6 desaturase in the presence of decreased concentrations of 18:2n–6. The increase in the 20:3n–6 was relatively higher than an increase in the 20:4n–6, which may contribute to the beneficial effect of the prostaglandin E1-to-prostaglandin E2 ratio.
In this study, the alterations of n–3 fatty acids were inconsistent. This is due partly to the experimental diets. Because the baseline diet of our participants was relatively low in the total n–3 fatty acids, we included 1 canned tuna sandwich/wk in both the moderate- and low-fat diets. This single tuna sandwich was apparently enough to increase eicosapentaenoic acid (20:5n–3) and 22:6n–3 in specimens on both treatment diets. In addition, n–3 fatty acids are known to be highly conserved even when withdrawn from the diet for several weeks (33). It is possible that during our relatively short-term intervention, the n–3 concentrations did not reach steady state equilibrium. Docosapentaenoic acid (22:5n–3), which usually does not come from dietary sources, increased on the low-fat diet, which is analogous to the increase in n–6 HUFAs. This increase, we believe, is due to alterations in metabolism of essential fatty acids. Similar increases in long-chain n–6 and n–3 fatty acids have been reported on a low-fat diet by another study (32). The increase in 22:5n–3 also suggests possible up-regulation of -5 desaturase in response to a decrease in total fat intake.
Although fatty acid changes were generally in the same direction for the 3 specimens, the percentage of change in fatty acid concentrations from baseline on the low-fat diet was often smaller in the red cells, especially for the long-chain PUFAs. Because red cell turnover is 120 d, it is possible that the alterations were not at equilibrium in the red cells after 6 wk of dietary intervention, and a longer intervention might have resulted in larger changes in red cells.
We developed models of combination of 4 or 5 fatty acids that predicted randomization to low-fat diet for each specimen. As in univariate analyses, decreased proportions of exogenous fatty acids and increased proportions of PUFAs and fatty acids biosynthesized from carbohydrates were related to consumption of the low-fat diet. Given collinearity among some of the fatty acids, it is possible that alternate models could be defined. However, 3 fatty acids were common to the 3-specimen specific models, 18:2n–6 and trans 18:1, which are widespread in processed foods, and 18:1n–7, a putative marker of carbohydrate intake. Our findings suggest that these fatty acids may be markers of total fat intake. However, future studies are needed to examine whether these fatty acid changes apply to intermediate concentrations of total fat intake and to circumstances when changes in total fat intake are accompanied with changes in dietary fat composition. Finally, the ability to define models with high specificity and sensitivity for each specimen also suggests that red cells, cholesterol esters, and plasma phospholipids may be used interchangeably for the purpose of monitoring compliance with low-fat diets. However, technically, red cells are less expensive to measure in the laboratory than are plasma phospholipids and cholesterol esters.
Our study has several strengths. We used a randomized design, we confirmed the composition of the diets by chemical analysis, we provided all food to participants for the duration of the study, and we took into account all extra snacks, drinks, and meals that were eaten for a complete knowledge of the participants' diets. The study also has several limitations. The study was conducted among postmenopausal women. Whether the findings can be generalized to men and younger women remains to be established. Participants' weight was strictly maintained. Whether similar changes in fatty acids would result from a low-fat diet accompanied with weight loss when fatty acids might not be synthesized from carbohydrates needs to be investigated. The study was designed to compare large differences in fat intake. Future studies need to examine whether these changes apply to intermediate proportions of total fat intake with changes in fat quality. The fatty acid changes were evaluated after 6 wk, and long-term effects of the diets are not known. The combination of fatty acids that best predicts total fat intake may be specific to the diet of the population studied. For example, if trans fatty acid intake decreases in the United States in response to labeling requirements, it may become less useful as a measure of total fat in the future. Finally, despite the design of the diets to resemble actual food choices, the study participants consumed controlled diets, and our findings should be replicated in a population setting.
In summary, we showed widespread differences in the fatty acid composition of blood specimens in response to a low-fat diet and a moderate-fat diet consumed for 6 wk. We showed that a model with a combination of fatty acids could almost perfectly differentiate the consumption of 34% of energy from fat from that of 17% of energy from fat, and that red cells, plasma phospholipids, or cholesterol esters could be used interchangeably. Our findings represent the first step in the search for a biomarker of total fat intake. Further study is needed to investigate the applicability of the biomarker to other populations and to ascertain whether the biomarker relates to total fat intake in a linear fashion.
ACKNOWLEDGMENTS
We thank everyone involved in the conduct of this study. We are especially grateful to the women who volunteered for this study.
MK and IBK were involved in the design and performance of the study. RNL was involved the statistical analysis of the study. All authors contributed to the writing of the manuscript. None of the authors had any conflict of interest with respect to this study.
REFERENCES
Siscovick DS, Raghunathan TE, King IB, et al. Dietary intake and cell-membrane levels of long-chain n–3 polyunsaturated fatty acids and the risk of primary cardiac arrest. JAMA 1995;274:1363–7.
Satia JA, Patterson RE, Herrero R, et al. Study of diet, biomarkers and cancer risk in the United States, China and Costa Rica. Int J Cancer 1999;82:28–32.
King IB, Kristal AR, Schaffer S, Thornquist M, Goodman GE. Serum. trans-Fatty acids are associated with risk of prostate cancer in Beta-Carotene and Retinol Efficacy Trial. Cancer Epidemiol Biomarkers Prev 2005;14:1–5.
Saadatian-Elahi M, Norat T, Goudable J, Riboli E. Biomarkers of dietary fatty acid intake and the risk of breast cancer: a meta-analysis. Int J Cancer 2004;111:584–91.
Neuhouser ML, Patterson RE, King IB, Horner NK, Lampe JW. Selected nutritional biomarkers predict diet quality. Public Health Nutr 2003;6:703–9.
Willett W. Nutritional epidemiology. New York: Oxford University Press, 1990.
Arab L. Biomarkers of fat and fatty acid intake. J Nutr 2003;133:925S–32S.
Arab L, Akbar J. Biomarkers and the measurement of fatty acids. Public Health Nutr 2002;5:865–71.
Ashley J. Lipid biomarkers of adherence to low fat diets. New York: Plenum Press, 1996.
Baylin A, Kabagambe EK, Siles X, Campos H. Adipose tissue biomarkers of fatty acid intake. Am J Clin Nutr 2002;76:750–7.
Emken EA. Biochemistry of unsaturated fatty acid isomers. J Am Oil Chemists Soc 1982;60:995–1004.
Manku MS, Horrobin DF, Huang Y-S, Morse N. Fatty acids in plasma and red cell membranes in normal humans. Lipids 1983;18:906–8.
Stanford JL, King IB, Kristal AR. Long-term storage of red blood cells and correlations between red cell and dietary fatty acids: results from a pilot study. Nutr Cancer 1991;16:183–8.
Pocock S, Simon R. Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trials. Biometrics 1975;31:103–15.
Sievert YA, Schakel SF, Buzzard IM. Maintenance of a nutrient database for clinical trials. Controlled Clin Trials 1989;10:416–25.
Joint FAO/WHO/UNU Expert Consultation. Energy and protein requirements. World Health Organ Tech Rep Ser 1985;724:
Association of Official Analytical Chemists. Official methods of analysis. 14th ed. Arlington, VA: Association of Official Analytical Chemists, 1987.
Rose HG, Oklander M. Improved procedure for the extraction of lipids from human erythrocytes. J Lipid Res 1965;6:428–31.
Folch J, Lees M, Sloane Stanley GH. A simple method for the isolation and purification of total lipides from animal tissues. J Biol Chem 1957;226:497–509.
Lepage G, Roy CC. Direct transesterification for all classes of lipids in a one-step reaction. J Lipid Res 1986;27:114–20.
Lemaitre RN, King IB, Patterson RE, Psaty BM, Kestin M, Heckbert SR. Assessment of trans-fatty acid intake with a food frequency questionnaire and validation with adipose tissue levels of trans-fatty acids. Am J Epidemiol 1998;148:1085–93.
Hosmer D, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley, 2000.
Dougherty RM, Galli C, Ferro-Luzzi A, Iacono JM. Lipid and phospholipid fatty acid composition of plasma, red blood cells, and platelets and how they are affected by dietary lipids: a study of normal subjects from Italy, Finland, and the USA. Am J Clin Nutr 1987;45:443–55.
Katan MB, van Birgelen A, Deslypere JP, Penders M, van Stavern WA. Biological markers of dietary intake, with emphasis on fatty acids. Ann Nutr Metab 1991;35:249–52.
Zock PL, Mensink RP, Harry V, de Vries JH, Katan MB. Fatty acids in serum cholesteryl esters as quantitative biomarkers of dietary intake. Am J Epidemiol 1997;145:1114–22.
Glatz JFC, Soffers EMF, Katan MB. Fatty acid composition of serum cholesteryl esters and erythrocyte membranes as indicators of linoleic acid intake in man. Am J Clin Nutr 1989;49:269–76.
Holman RT. Control of polyunsaturated acids in tissue lipids. J Am Coll Nutr 1986;5:183–211.
Hudgins LC, Hellerstein M, Seldman C, Neese R, Diakun J, Hirsch J. Human fatty acid synthesis is stimulated by a eucaloric low fat, high carbohydrate diet. J Clin Invest 1996;97:2081–91.
Phinney SD, Odin RS, Johnson SB, Holman RT. Reduced arachidonate in serum phospholipids and cholesteryl esters associated with vegetarian diets in humans. Am J Clin Nutr 1990;51:385–92.
Ntambi J. Regulation of stearoyl-CoA desaturase by polyunsaturated fatty acids and cholesterol. J Lipid Res 1999;40:1549–58.
Ntambi JM, Miyazaki M. Regulation of stearoyl-CoA desaturases and role in metabolism. Prog Lipid Res 2004;43:91–104.
Raatz SK, Bibus D, Thomas W, Kris-Etherton P. Total fat intake modifies fatty acid composition in humans. J Nutr 2001;131:231–4.
Prisco D, Filippini M, Francalanci I, et al. Effect of n–3 polyunsaturated fatty acid intake on phospholipid fatty acid composition in plasma and erythrocytes. Am J Clin Nutr 1996;63:925–32., http://www.100md.com(Irena B King1, Rozenn N Lemaitre1 and Ma)
ABSTRACT
Background: The utility of fatty acids (FAs) as biomarkers of total fat intake is unknown.
Objective: We compared FA changes in red cells (RCs), plasma phospholipids (PLs), and cholesterol esters (CEs) in response to a low-fat diet (LFD) and a moderate-fat diet (MFD) and assessed whether individual or combination of FAs predict LFD.
Design: Postmenopausal women (n = 66) were randomly assigned to receive an LFD (17% of energy from fat) or an MFD (34% of energy from fat) for 6 wk. All foods were provided. FAs in diets and blood were determined by gas-liquid chromatography. FA changes between baseline and end of study were compared across diets by using t tests. FA predictors of an LFD were selected by logistic regression.
Results: Many FAs in RCs, PLs, and CEs responded differently to the 2 diets. Changes from baseline with an LFD for palmitic acid (16:0) (3–11% increase), behenic (22:0) and lignoceric (24:0) acids (3–20% decrease, in RCs and PLs only), cis-monounsaturated FA (MUFA) (25–35% increase), linoleic acid (18:2n–6) (11–13% decrease), trans octadecanoic acids (trans 18:1) (7–20% decrease), and n–6 highly unsaturated FA (HUFA) (2–8% increase) were significantly different from changes with an MFD. Individually, 18:2n–6 and trans 18:1 were strong predictors of an LFD [receiver operating characteristic (ROC) curves: 0.92–0.80). A logistic regression model with trans 18:1, 18:2n–6, and vaccenic acid (18:1n–7) predicted an LFD with high specificity and sensitivity (ROC curves: 0.99).
Conclusions: Saturated FA, cisMUFA, n–6 HUFA, and exogenous FAs greatly differed in their response to the LFD and MFD. Parallel responses were observed in RCs, PLs, and CEs. A model with a combination of FAs almost perfectly differentiated the consumption of 34% fat from that of 17% fat.
Key Words: Fatty acids trial biomarkers dietary fats low-fat diet
INTRODUCTION
There is much scientific and public health interest in the role of dietary fat in causing serious chronic diseases such as cardiovascular disease and cancer (1-4). For both observational and intervention studies, measurement of total fat intake is crucial to the validity and interpretation of findings. Given that fat intake of participants can rarely be directly observed, assessments of dietary fat intake rely on self-reports, which often have recall bias and measurement errors, including changes in participants' diets when filling out questionnaires (5, 6). A biological marker of total fat intake would be useful, particularly one that can be ascertained from blood samples. Currently, no reliable biomarkers of total fat intake are known (7-9), and little information exists on fatty acid response to changes in fat quantity.
Blood and adipose tissue reasonably represent dietary intake of exogenous fatty acids (those that are not synthesized in vivo) such as linoleic acid (18:2n–6), trans fatty acids, and long-chain n–3 (10-13). However, most dietary fatty acids can be synthesized de novo; thus, their source of origin is inconclusive. Furthermore, the extent to which individual exogenous fatty acids reflect total dietary fat intake is unknown. The objectives of this study were to investigate fatty acid alterations in response to change in total dietary fat intake, to determine which blood specimen best reflects the change in dietary fat intake, and to investigate whether a combination of fatty acids might be a valid marker of total fat intake. To address these questions, we conducted a randomized dietary trial of low-fat (17% of energy) and moderate-fat (34% of energy) intake among 66 postmenopausal women.
SUBJECTS AND METHODS
Study design and participants
This study was a randomized, controlled, dietary intervention trial conducted at the Fred Hutchinson Cancer Research Center and the Clinical Research Center at the University of Washington in Seattle, WA. The study was approved by the institutional review boards at both institutions.
Participants were healthy postmenopausal women aged 50–79 y. Exclusion criteria were any medical condition that required dietary modification (eg, diabetes, renal disease, or hypercholesterolemia), morbid obesity [body mass index (BMI; in kg/m2) >40], regular consumption of the main meal outside the home 2 times/wk, unsatisfactory completion of food records at any time before randomization, recent weight loss (>10% of usual body weight during the previous 6 mo), or a usual alcohol intake averaging > 2 standard drinks/d (>24 g alcohol).
Potentially eligible women completed two 4-d food records before randomization. Sixty-six women were then randomly assigned (33 in each group) to consume either a low-fat (17% of energy) or a moderate-fat (34% of energy) diet for 6 wk. Reasonable parity between subjects in the 2 dietary groups by age, BMI, and postmenopausal hormone use was ensured by using an adaptive randomization program (14). Sixty-one women completed the study—31 in the low-fat diet group and 30 in the moderate-fat diet group. All 5 participants who dropped out of the study did so for personal, not diet-related reasons.
Clinical measurements
Participants attended a total of 4 clinic visits, 2 at baseline (3 d apart) and 2 at the end of the study (also days apart) after a 12–14-h fast. At each visit, participants were weighed to the nearest 100 g, and 20 mL venous blood was collected into one 10-mL plain evacuated tube and one 10-mL evacuated tube containing K3-EDTA (1 mg/mL). Serum and plasma were divided into aliquots, and red cells were washed twice with 0.9% saline before storage at –70°C until analysis.
Diet design
The study research dietitian designed the low- and high-fat diets by using NDS software (NCC, University of Minnesota, Minneapolis, MN) (15). Women's energy requirements were estimated by standard equations for the basal metabolic rate based on height, weight, and age and multiplied by 1.51 as the activity factor (16). Participants were weighed weekly, and, according to whether they lost or gained >2.0 kg/wk from their initial baseline weight, they were placed on the higher or lower energy level as appropriate.
We designed 7 different breakfasts (15% of energy), lunches (25% of energy), dinners (50% of energy), and snacks (10% of energy) for each diet at 6 energy levels. All food was prepared in advance at the Nutrition Research Kitchen at the University of Washington Clinical Research Center by dietary technicians. Participants were free to chose up to 0.628 MJ (150 kcal)/d from foods low or devoid in fat, and they restricted their intake of alcoholic beverages to a maximum of 6 ounces (170 mL)/wk.
We prepared a weighted composite of each diet on 4 separate occasions for the median energy level, 8.4 MJ, and chemically analyzed a weighted mixture of these 4 composites (Table 1). Protein was estimated by the measuring nitrogen by using the Kjeldahl method and then multiplying by 6.25 (17). Total fat was confirmed by 2 methods: acid hydrolysis procedure and enzymatic digestion with chloroform-methanol extraction (17). Dietary fiber was measaured by gravimetric methods (17). Total carbohydrate was calculated as total intake minus water, ash, protein, fat, and dietary fiber. Fatty acid analyses are described later.
Food records and measures of adherence
The 8-d food records, reflecting the participants' usual diets at baseline, were checked for accuracy by the research dietitian, coded, and analyzed by using the NDS data system (version 2.8; NCC, University of Minnesota). Participants were given a form on which to record their consumption of the foods provided to them during the trial. Any food not consumed was returned to the research kitchen and weighed. Foods consumed in the 0.628 MJ "allowance" were also recorded and analyzed with the NDS data system.
Analyses of fatty acids
Total red cell fatty acids were extracted with isopropanol and chloroform according to the method of Rose and Oklander (18). Total lipids from plasma were extracted by the method of Folch (19) with a 2:1 chloroform:methanol solution and several washes. Plasma phospholipids and plasma cholesterol esters were separated by one-dimensional thin-layer chromatography by using 250-μm Silica Gel G plates (Analtech Inc, Newark, DE) and a 67.5:15:0.75 hexane:ethyl ether:acetic acid (+0.005% butylated hydroxytoluene) development solvent.
The fatty acid methyl esters (FAMEs) for all samples, including diet samples, were prepared by direct transesterification by using the method of Lepage and Roy (20). Gas-liquid chromatography was performed on samples dissolved in hexane. The FAME of individual fatty acids of red cell membranes, plasma phospholipids, and cholesterol esters were separated on a gas chromatograph (model 5890B, series II; Hewlett-Packard, Avondale, PA) equipped with a flame ionization detector, automatic sampler (Hewlett-Packard 7673), and CHEMSTATION and MUSTANG software (Hewlett-Packard) as previously described (21).
Quantitative precision and peak identification were evaluated with the use of weighted individual and model mixtures of known FAMEs. This identification was confirmed by a mass spectrophotometric analysis performed by the US Department of Agriculture Lipid Laboratory (Peoria, IL). Our quantitative results were standardized with the National Heart Institute Fatty Acid Standards A, B, C, D, and F (Nu-Check-Prep, Elysian, MN), and response factors were calculated. Fatty acid composition is expressed as relative weight percentages. Pooled red cells and plasma were used as additional controls that were run with each batch of study samples. Interassay CVs were on the average of 3.5% for most of the fatty acids that were present at levels of 1%. All laboratory personnel were blinded with respect to the diet group randomization.
Statistical analyses
The data collected at the 2 baseline visits and at the 2 end-of-study visits were averaged to yield a single value for baseline and for end of study. In descriptive analyses, we compared mean baseline characteristics of women randomly assigned to the low-fat diet and women randomly assigned to the moderate-fat diet by using two-sided t tests and two-sided statistical significance at P value < 0.05. To illustrate the effect of diet on fatty acid composition in the 3 specimens, we computed percentages of change for each fatty acid at the end of the study. To assess the effect of diet on the changes in fatty acid composition, we compared the difference in fatty acid proportions at baseline and at end of study across the diet groups by using t tests and two-sided statistical significance at P value < 0.05. To assess which fatty acid combination predicted randomization to the low-fat diet, we performed stepwise logistic regression with randomization to low fat as the outcome (22). Fatty acid proportions at the end of the study were the predictors. To reduce the number of possible predictors submitted to the stepwise procedure, 3 sets of fatty acids were initially submitted in 3 separate stepwise logistic regression analyses. The first set included fatty acids that originate from the diet, namely trans fatty acids, 18:2n–6, linolenic acid (18:3), and the long-chain saturated fatty acids behenic acid (22:0) and lignoceric acid (24:0). The second set included fatty acids that can be synthesized endogenously from carbohydrates, palmitic acid (16:0), stearic acid (18:0), palmitoleic acid (16:1n–7), vaccenic acid (18:1n–7), and oleic acid (18:1n–9). The third set included polyunsaturated fatty acids (PUFAs) of 20 carbons. With each analysis, fatty acid predictors were selected stepwise by using a P value of entry 0.1. Then, by using the fatty acids that were selected in each of the 3 initial analyses, a last stepwise analysis was performed for a final fatty acid selection. We computed sensitivity and specificity of the final model fitted by the logistic regression. Global performance of the model in predicting randomization to low-fat diet was summarized by the area under the receiver operating characteristics (ROC) curve. These procedures were performed separately for each blood specimen. All statistical analyses were performed by using STATA software (version 8.2; Stata Corp, College Station, TX).
RESULTS
The baseline characteristics of the postmenopausal women in the study are shown in Table 2. None of the differences between dietary groups were statistically significant at P < 0.05. The fat intake of the women at baseline was lower than the typical US intake in both groups (on average, 29% of energy intake). On average, the women were aged 63 y, with a BMI of 26. The baseline diets of women randomly assigned to the to low-fat diet and women randomly assigned to the moderate-fat diet were not significantly different (Table 2). In addition, mean baseline proportions of individual fatty acids, measured in red cells, plasma phospholipids, and plasma cholesterol esters, were not significantly different for the 2 dietary groups (Table 3).
On average, a small weight loss was observed between baseline and the end of study in both groups. No statistically significant differences were apparent when the weight loss was compared between the 2 diets [(± SD) moderate fat: –0.96 ± 0.92 kg; low fat: –1.3 ± 1.40 kg;, P = 0.32).
Baseline fatty acid compositions and changes are shown (Table 3) after 6 wk of randomization to the low-and moderate-fat diets for 3 blood specimens, namely red cells, plasma phospholipids, and plasma cholesterol esters. Changes in many fatty acids differed between the 2 diets, as shown in Table 3. For most fatty acids, the direction of change was consistent among the 3 types of specimens.
Saturated fatty acids
Among the saturated fatty acids, the response of 16:0 to the 2 diets was significantly different with a consistent increase from baseline on the low-fat diet and decrease on the moderate-fat diet in all specimens (Table 3). The long-chain saturated fatty acids 22:0 and 24:0 decreased on the low-fat diet and increased on the moderate-fat diet in red cells and plasma phospholipids, which again resulted in a significant difference between the 2 diets. The proportions of these fatty acids were below detection in the cholesterol esters. The proportions of 18:0 differed in response to the 2 diets only in cholesterol esters (P = 0.03). Total saturated fatty acids in the 3 specimens changed in the direction of 16:0, and the difference in changes across the 2 diets remained significant.
cis Monounsaturated fatty acids
Among cis monounsaturated fatty acids, the greatest difference between the 2 diets was observed in the 16:1 fatty acids. Both the n–7 and n–9 fatty acid families dramatically changed in opposite directions in response to the diets (increasing 14–34% on the low-fat diet and decreasing 4–22% on the moderate-fat diet). The response of 18:1 fatty acids (n–7 and n–9) to the diets mirrored that of the 16:1 families, although the magnitude of the response was smaller.
n–6 Polyunsaturated fatty acids
18:2n–6 Decreased from baseline in red cells, plasma phospholipids, and cholesterol esters in response to the low-fat diet (–11.2%, –12.9%, and –10.8%, respectively) and increased on the moderate-fat diet (2.6%, 3.9%, and 6.2%, respectively) in all 3 specimens. The difference in change from baseline across the 2 diets was highly significant. In contrast, dihomo--linoleic acid (20:3n–6) and arachidonic acid (20:4n–6) increased in response to the low-fat diet in all 3 specimens, which resulted in significant differences between the 2 diets. Another long-chain PUFA, docosatetraenoic acid (22:4n–6), also responded differently to the diets but only in the plasma phospholipids (Table 3). The total n–6 fatty acids, representing the overall change in both 18:2n–6 and long-chain n–6 fatty acids, showed a small but significant difference between the 2 diets.
n–3 Polyunsaturated fatty acids
-Linolenic acid (18:3n–3) decreased from baseline on both diets in red cells, and its response did not differ between the diets. In plasma phospholipids and cholesterol esters, 18:3n–3 decreased on the low-fat diet and increased on the moderate-fat diet, which resulted in a borderline difference in responses (P = 0.06). Docosahexaenoic acid (22:6n–3) consistently increased from baseline on both diets in all 3 specimens.
Total polyunsaturated fatty acids
The total PUFAs (n–6 and n–3 combined) decreased from baseline on the low-fat diet and increased on moderate-fat diet. Thus, the differences in changes across the 2 diets were highly statistically significant in the 3 specimens.
trans Fatty acids
The total trans octadecanoic acids (trans 18:1) consistently decreased on the low-fat diet and increased on the moderate-fat diet in all 3 specimens, which resulted in highly significant differences between the 2 diets. The total trans-hexadecenoic acids (trans 16:1) and linoelaidic acids (trans 18:2), which are present in low proportions, did not change from baseline in response to the low-fat diet; however, trans 16:1 increased from baseline in response to the moderate-fat diet.
Fatty acid proportions as a predictor of total fat intake
One purpose of the study was to investigate whether a combination of fatty acids might effectively predict randomization to the low-fat diet. To this end, we selected end-of-study fatty acids that predicted the dietary group assignment by using stepwise logistic regression, as described in Subjects and Methods. For each specimen, we were able to define a multivariate model that predicted randomization to the low-fat diet with high specificity and sensitivity (Table 4). Three fatty acid predictors were common to all specimens: trans fatty acid isomers of trans 18:1, 18:2n–6, and 18:1n–7. Consistent with the univariate analyses (Table 3), lower trans 18:1, lower 18:2n–6, higher 18:1n–7, and, for red cells and cholesterol esters, higher proportions of selected PUFAs predicted randomization to the low-fat diet. In the multivariate models, these fatty acids were all significant predictors in the red cells (Table 4); in the plasma phospholipids, 18:2n–6 was a borderline predictor (P = 0.06); and in the cholesterol ester, trans 18:1 (P = 0.05), 18:2n–6 (P = 0.06), 18:1n–7 (P = 0.06), and 20:3n–6 (P = 0.07) were borderline predictors. As reflected by the area under the ROC curve (0.99 for all specimens), the ability of the models shown in Table 4 to discriminate between the 2 dietary groups was high. Prediction of the low-fat diet was better achieved with combinations of fatty acids than with individual fatty acids. However, 18:2n–6 and trans 18:1 were individually high predictors of low-fat intake. The area under the ROC curve for models with only 18:2n–6 was 0.90 for phospholipids, 0.84 for red cells, and 0.90 for cholesterol ester. Corresponding values for trans 18:1 alone were 0.92, 0.80, and 0.85.
DISCUSSION
In this controlled dietary trial, we have shown that, in weight-stable postmenopausal women, a decrease in dietary fat consumption from a usual diet of 29% energy at baseline to 17% during 6 wk resulted in widespread changes in fatty acid proportions in red cells, plasma phospholipids, and cholesterol esters. These fatty acid changes were absent or in the opposite direction in the women randomly assigned to the moderate-fat diet (34% energy). The magnitude of the difference in fatty acids between the 2 diet groups was partly attributable to a small increase in fat intake from baseline in the moderate-fat group as well as the large decrease in fat intake in the low-fat group. In addition, we found that a combination of fatty acids could be used to predict randomization to low-fat diet with high sensitivity and specificity.
In response to a low-fat diet, we observed definite patterns of fatty acid alterations. Most exogenous fatty acids (originating from the diet only) consistently showed a decrease in the 3 specimens. This was not surprising because the diets were designed to decrease fat across all important dietary fatty acids, and exogenous fatty acids are known to reflect dietary intake (23). In this study, the dietary fats were selected on the basis of the type of fat that is normally consumed in the American diet. The main PUFA in the American diet is 18:2n–6, which was decreased 41% on the low-fat diet, which resulted in an 11–13% decrease in all 3 blood specimens. The decrease in 18:2n–6 in red cells, plasma phospholipids, and cholesterol esters in response to a dietary decrease is well documented (23-25). Although it has been suggested that cholesterol esters adequately capture 18:2n–6 change in the diet (26), we found that red cells and plasma phospholipids were equally informative in this short-term intervention. Our study also showed a decrease in blood proportions of the trans 18:1 fatty acids with reduction of total fat intake. The proportions of total trans 18:1 in the investigated specimens were relatively low, ranging from 0.4% to 1.5% of total fatty acids. Despite these low proportions, trans 18:1 decreased 7–20% in all specimens; the lowest decrease was in the cholesterol esters. The total trans 18:2 decreased slightly with both diets, whereas total trans 16:1 did not decrease in response to the low-fat diet but increased in response to the moderate-fat diet. The total trans 18:2 and trans 16:1 were not measured in the experimental diets, and it is possible that the composition of these fatty acids differed in the 2 diets.
We found systematic increases in many endogenous fatty acids in response to the low-fat diet, despite dietary decrease in these fatty acids. Because we were measuring proportions of fatty acids rather than absolute amounts, a decrease in the proportion of some fatty acids will be compensated, by definition, with an increase in the proportion of other fatty acids. Diets high in carbohydrates and calories lead to de novo synthesis of several fatty acids (27-29). These include saturated fatty acids, namely, myristic acid (14:0) and 16:0 and monounsaturated fatty acids of the n–7 and n–9 series (16:1n–7, 16:1n–9, 18:1n–7, and 18:1n–9). During the de novo synthesis of fatty acids, the fatty acid synthase produces 16:0, which can be either desaturated to 16:1n–7 or elongated to 18:0. In turn, 18:0 is desaturated to 18:1n–9 by -9 desaturase, and 16:1n–7 is elongated to 18:1n–7 by long-chain elongase (30, 31). In agreement with our findings, Raatz et al (32) reported an increase in 18:1n–9 with a low-fat diet; however, they did not report other cis monounsaturated fatty acids. The proportions of 18:0 remained constant despite a 2-fold decrease on the low-fat diet. In support of a lack of response to diet, it was reported that 18:0 does not appear to accumulate in response to dietary treatment (30). Our study also suggests a possible increase in -oxidation to produce higher proportions of 16:1n–9 from endogenously produced 18:1n–9 on a low-fat diet.
Long-chain n–6 PUFAs [(n–6 highly unsaturated fatty acid (HUFA), carbon 20], generally increased on the low-fat diet, despite lower concentrations in the diet. The increase in HUFAs is important because it occurs despite the dilution of the essential fatty acids by the endogenous, carbohydrate-derived fatty acids synthesized on the low-fat diet. These changes likely reflect the influence of 18:2n–6 on fatty acid metabolism, namely, a release of inhibition of -6 desaturase in the presence of decreased concentrations of 18:2n–6. The increase in the 20:3n–6 was relatively higher than an increase in the 20:4n–6, which may contribute to the beneficial effect of the prostaglandin E1-to-prostaglandin E2 ratio.
In this study, the alterations of n–3 fatty acids were inconsistent. This is due partly to the experimental diets. Because the baseline diet of our participants was relatively low in the total n–3 fatty acids, we included 1 canned tuna sandwich/wk in both the moderate- and low-fat diets. This single tuna sandwich was apparently enough to increase eicosapentaenoic acid (20:5n–3) and 22:6n–3 in specimens on both treatment diets. In addition, n–3 fatty acids are known to be highly conserved even when withdrawn from the diet for several weeks (33). It is possible that during our relatively short-term intervention, the n–3 concentrations did not reach steady state equilibrium. Docosapentaenoic acid (22:5n–3), which usually does not come from dietary sources, increased on the low-fat diet, which is analogous to the increase in n–6 HUFAs. This increase, we believe, is due to alterations in metabolism of essential fatty acids. Similar increases in long-chain n–6 and n–3 fatty acids have been reported on a low-fat diet by another study (32). The increase in 22:5n–3 also suggests possible up-regulation of -5 desaturase in response to a decrease in total fat intake.
Although fatty acid changes were generally in the same direction for the 3 specimens, the percentage of change in fatty acid concentrations from baseline on the low-fat diet was often smaller in the red cells, especially for the long-chain PUFAs. Because red cell turnover is 120 d, it is possible that the alterations were not at equilibrium in the red cells after 6 wk of dietary intervention, and a longer intervention might have resulted in larger changes in red cells.
We developed models of combination of 4 or 5 fatty acids that predicted randomization to low-fat diet for each specimen. As in univariate analyses, decreased proportions of exogenous fatty acids and increased proportions of PUFAs and fatty acids biosynthesized from carbohydrates were related to consumption of the low-fat diet. Given collinearity among some of the fatty acids, it is possible that alternate models could be defined. However, 3 fatty acids were common to the 3-specimen specific models, 18:2n–6 and trans 18:1, which are widespread in processed foods, and 18:1n–7, a putative marker of carbohydrate intake. Our findings suggest that these fatty acids may be markers of total fat intake. However, future studies are needed to examine whether these fatty acid changes apply to intermediate concentrations of total fat intake and to circumstances when changes in total fat intake are accompanied with changes in dietary fat composition. Finally, the ability to define models with high specificity and sensitivity for each specimen also suggests that red cells, cholesterol esters, and plasma phospholipids may be used interchangeably for the purpose of monitoring compliance with low-fat diets. However, technically, red cells are less expensive to measure in the laboratory than are plasma phospholipids and cholesterol esters.
Our study has several strengths. We used a randomized design, we confirmed the composition of the diets by chemical analysis, we provided all food to participants for the duration of the study, and we took into account all extra snacks, drinks, and meals that were eaten for a complete knowledge of the participants' diets. The study also has several limitations. The study was conducted among postmenopausal women. Whether the findings can be generalized to men and younger women remains to be established. Participants' weight was strictly maintained. Whether similar changes in fatty acids would result from a low-fat diet accompanied with weight loss when fatty acids might not be synthesized from carbohydrates needs to be investigated. The study was designed to compare large differences in fat intake. Future studies need to examine whether these changes apply to intermediate proportions of total fat intake with changes in fat quality. The fatty acid changes were evaluated after 6 wk, and long-term effects of the diets are not known. The combination of fatty acids that best predicts total fat intake may be specific to the diet of the population studied. For example, if trans fatty acid intake decreases in the United States in response to labeling requirements, it may become less useful as a measure of total fat in the future. Finally, despite the design of the diets to resemble actual food choices, the study participants consumed controlled diets, and our findings should be replicated in a population setting.
In summary, we showed widespread differences in the fatty acid composition of blood specimens in response to a low-fat diet and a moderate-fat diet consumed for 6 wk. We showed that a model with a combination of fatty acids could almost perfectly differentiate the consumption of 34% of energy from fat from that of 17% of energy from fat, and that red cells, plasma phospholipids, or cholesterol esters could be used interchangeably. Our findings represent the first step in the search for a biomarker of total fat intake. Further study is needed to investigate the applicability of the biomarker to other populations and to ascertain whether the biomarker relates to total fat intake in a linear fashion.
ACKNOWLEDGMENTS
We thank everyone involved in the conduct of this study. We are especially grateful to the women who volunteered for this study.
MK and IBK were involved in the design and performance of the study. RNL was involved the statistical analysis of the study. All authors contributed to the writing of the manuscript. None of the authors had any conflict of interest with respect to this study.
REFERENCES
Siscovick DS, Raghunathan TE, King IB, et al. Dietary intake and cell-membrane levels of long-chain n–3 polyunsaturated fatty acids and the risk of primary cardiac arrest. JAMA 1995;274:1363–7.
Satia JA, Patterson RE, Herrero R, et al. Study of diet, biomarkers and cancer risk in the United States, China and Costa Rica. Int J Cancer 1999;82:28–32.
King IB, Kristal AR, Schaffer S, Thornquist M, Goodman GE. Serum. trans-Fatty acids are associated with risk of prostate cancer in Beta-Carotene and Retinol Efficacy Trial. Cancer Epidemiol Biomarkers Prev 2005;14:1–5.
Saadatian-Elahi M, Norat T, Goudable J, Riboli E. Biomarkers of dietary fatty acid intake and the risk of breast cancer: a meta-analysis. Int J Cancer 2004;111:584–91.
Neuhouser ML, Patterson RE, King IB, Horner NK, Lampe JW. Selected nutritional biomarkers predict diet quality. Public Health Nutr 2003;6:703–9.
Willett W. Nutritional epidemiology. New York: Oxford University Press, 1990.
Arab L. Biomarkers of fat and fatty acid intake. J Nutr 2003;133:925S–32S.
Arab L, Akbar J. Biomarkers and the measurement of fatty acids. Public Health Nutr 2002;5:865–71.
Ashley J. Lipid biomarkers of adherence to low fat diets. New York: Plenum Press, 1996.
Baylin A, Kabagambe EK, Siles X, Campos H. Adipose tissue biomarkers of fatty acid intake. Am J Clin Nutr 2002;76:750–7.
Emken EA. Biochemistry of unsaturated fatty acid isomers. J Am Oil Chemists Soc 1982;60:995–1004.
Manku MS, Horrobin DF, Huang Y-S, Morse N. Fatty acids in plasma and red cell membranes in normal humans. Lipids 1983;18:906–8.
Stanford JL, King IB, Kristal AR. Long-term storage of red blood cells and correlations between red cell and dietary fatty acids: results from a pilot study. Nutr Cancer 1991;16:183–8.
Pocock S, Simon R. Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trials. Biometrics 1975;31:103–15.
Sievert YA, Schakel SF, Buzzard IM. Maintenance of a nutrient database for clinical trials. Controlled Clin Trials 1989;10:416–25.
Joint FAO/WHO/UNU Expert Consultation. Energy and protein requirements. World Health Organ Tech Rep Ser 1985;724:
Association of Official Analytical Chemists. Official methods of analysis. 14th ed. Arlington, VA: Association of Official Analytical Chemists, 1987.
Rose HG, Oklander M. Improved procedure for the extraction of lipids from human erythrocytes. J Lipid Res 1965;6:428–31.
Folch J, Lees M, Sloane Stanley GH. A simple method for the isolation and purification of total lipides from animal tissues. J Biol Chem 1957;226:497–509.
Lepage G, Roy CC. Direct transesterification for all classes of lipids in a one-step reaction. J Lipid Res 1986;27:114–20.
Lemaitre RN, King IB, Patterson RE, Psaty BM, Kestin M, Heckbert SR. Assessment of trans-fatty acid intake with a food frequency questionnaire and validation with adipose tissue levels of trans-fatty acids. Am J Epidemiol 1998;148:1085–93.
Hosmer D, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley, 2000.
Dougherty RM, Galli C, Ferro-Luzzi A, Iacono JM. Lipid and phospholipid fatty acid composition of plasma, red blood cells, and platelets and how they are affected by dietary lipids: a study of normal subjects from Italy, Finland, and the USA. Am J Clin Nutr 1987;45:443–55.
Katan MB, van Birgelen A, Deslypere JP, Penders M, van Stavern WA. Biological markers of dietary intake, with emphasis on fatty acids. Ann Nutr Metab 1991;35:249–52.
Zock PL, Mensink RP, Harry V, de Vries JH, Katan MB. Fatty acids in serum cholesteryl esters as quantitative biomarkers of dietary intake. Am J Epidemiol 1997;145:1114–22.
Glatz JFC, Soffers EMF, Katan MB. Fatty acid composition of serum cholesteryl esters and erythrocyte membranes as indicators of linoleic acid intake in man. Am J Clin Nutr 1989;49:269–76.
Holman RT. Control of polyunsaturated acids in tissue lipids. J Am Coll Nutr 1986;5:183–211.
Hudgins LC, Hellerstein M, Seldman C, Neese R, Diakun J, Hirsch J. Human fatty acid synthesis is stimulated by a eucaloric low fat, high carbohydrate diet. J Clin Invest 1996;97:2081–91.
Phinney SD, Odin RS, Johnson SB, Holman RT. Reduced arachidonate in serum phospholipids and cholesteryl esters associated with vegetarian diets in humans. Am J Clin Nutr 1990;51:385–92.
Ntambi J. Regulation of stearoyl-CoA desaturase by polyunsaturated fatty acids and cholesterol. J Lipid Res 1999;40:1549–58.
Ntambi JM, Miyazaki M. Regulation of stearoyl-CoA desaturases and role in metabolism. Prog Lipid Res 2004;43:91–104.
Raatz SK, Bibus D, Thomas W, Kris-Etherton P. Total fat intake modifies fatty acid composition in humans. J Nutr 2001;131:231–4.
Prisco D, Filippini M, Francalanci I, et al. Effect of n–3 polyunsaturated fatty acid intake on phospholipid fatty acid composition in plasma and erythrocytes. Am J Clin Nutr 1996;63:925–32., http://www.100md.com(Irena B King1, Rozenn N Lemaitre1 and Ma)