CHOLESTEROL-LOWERING DIets

Individual Cholesterol Variation in Response

to a Margarine- or Butter-Based Diet

A Study in Families

Beverley Adams-Huet, MS Anh T. Nguyen, BS

C

-ets have been recommended for the population at large to reduce the incidence of coro-nary heart disease.1_{Changes in the mean} lipid2,3_{and lipoprotein}4,5_{levels can be} re-liably predicted from changes in popu-lation intake of dietary fatty acids and cholesterol. The projected benefits of di-etary modification are substantial and are directly linked to the magnitude of cho-lesterol level reduction.6

Although populational responses to diet can be reliably predicted, it is im-possible to predict how much choles-terol lowering a given individual will achieve as a result of dietary modifica-tion. Individual responses follow a peaked distribution around the mean population response. Theoretically, two thirds of individuals fall within 1 SD of the mean response, with some indi-viduals having little or no cholesterol-lowering response to diet.7,8

Predicting who will and will not re-spond to diet modification would al-low the clinician to target aggressive cholesterol-lowering dietary therapy for those patients who are most respon-sive to diet. It also would allow the cli-nician to differentiate between nonre-sponders who are noncompliant from nonresponders who do not have the biological potential to respond to diet modification.

This study was designed to evaluate whether familial differences explain why individuals respond differently to cholesterol-lowering diets. The

preva-lence of individuals who do not re-spond to diet has been estimated at 17% of institutionalized men on a con-trolled diet9_{and 15% to 20% of}

free-Author Affiliations:Department of Internal Medi-cine (Dr Denke and Ms Adams-Huet) and Center for Human Nutrition (Dr Denke and Ms Nguyen), University of Texas Southwestern Medical Center, Dallas.

Corresponding Author and Reprints:Margo A. Denke, MD, Department of Internal Medicine, Center for Hu-man Nutrition, 5323 Harry Hines Blvd, Room Y3.234, Dallas, TX 75390-9052 (e-mail: mdenke@mednet .swmed.edu).

Context The effectiveness of dietary modification in reducing low-density lipopro-tein cholesterol (LDL-C) levels can be reliably predicted for populations, but not for individuals.

Objective To determine whether individual variation in cholesterol response to di-etary modification is a familial trait.

Design Two-period, outpatient crossover trial conducted from September 1997 to September 1999.

Setting and Participants Fifty-six families from the Dallas–Ft Worth, Tex, area with 2 biological parents and at least 2 children aged 5 years or older volunteered; 46 fami-lies (n=92 adults and n=134 children) completed the study.

Intervention All families followed two 5-week dietary regimens that included indi-vidualized daily dietary prescriptions and emphasized a low–saturated fat diet supple-mented with specially manufactured baked goods and spreadable fat. One regimen used butter only and the other used margarine only.

Main Outcome Measure Mean LDL-C levels during the last 2 weeks of each di-etary period.

Results Margarine intake compared with butter intake lowered LDL-C levels 11% in adults (95% confidence interval [CI], −13% to −9%) and 9% in children (95% CI, −12% to −6%) (P,.001 for both adults and children). The distribution of individual responses were peaked around the mean response. For adults and children together, family membership accounted for 19% of variability in response (P = .007). In chil-dren, family membership accounted for 40% of variability in response of percent change in LDL-C levels (P = .002). Body mass index and change in cholesterol ester (CE) 18: 2/18:1 ratio accounted for 26% of variation, leaving 26% still attributable to family membership. In all participants, BMI predicted response—heavier individuals had higher LDL-C levels, less excursion in CE fatty acids, and less LDL-C response to dietary change. Conclusions Our results suggest that individual variation in response to a cholesterol-lowering diet is a familial trait. Body weight is an important modifiable factor that in-fluences response.

living men,9-11_women,12 _and chil-dren13_{on counseled diets. Although} associations between specific geno-types and dietary responsiveness among unrelated individuals have been re-ported, no factor has been found to be consistently associated with response, and no single factor has explained more than 10% of the individual variation ob-served.

METHODS

The study protocol was approved by the institutional review board at the Uni-versity of Texas Southwestern Medi-cal Center and the Human Subjects Re-view Committee at the Veterans Health Administration North Texas Center, Dallas.

Subjects

Families from the Dallas–Ft Worth metroplex were recruited during their attendance at various museums, health fairs, and church events over 2 years (September 1997 to September 1999). Families were selected by 2 criteria only: interest in participation and having an intact family with 2 biological parents and 2 or more biologically related chil-dren aged 5 years and older. During a follow-up telephone contact, a home visit was scheduled where the study was described, and the age, height, weight, physical activity, and dietary prefer-ences for all family members were corded. If all family members re-mained interested, a date was set to draw ad libitum fasting blood samples and study initiation. Fifty-six families agreed to participate, and 46 families completed the entire diet study. Rea-sons for dropout included difficulty fol-lowing dietary protocol restrictions (5 families), refusal of a child to have blood drawn after the study was initiated (2 families), and change in health status of parent (3 families). Of the 46 fami-lies, 39 were white, 3 African can, 2 Hispanic American, 1 Ameri-can Indian, and 1 Asian AmeriAmeri-can. Baseline characteristics of the 46 fami-lies are detailed inTABLE1; the char-acteristics of the 10 families who dropped out were similar to those who

completed the trial. All children com-pleting the trial received a $50 pay-ment. All family members completing a 3-day diet record received 1 movie pass (value $4) for each of the 4 re-quested records.

Dietary Design

The study was a 2-period, crossover, outpatient diet counseling study de-signed to compare the isocaloric sub-stitution of butter for margarine as the major dietary fat intake. Each dietary period was 5 weeks in duration. Since commercial tub margarine is 60% or less fat by weight and butter is 80% fat by weight, a single batch of 80% fat by weight margarine was produced for this study (courtesy of Unilever, Balti-more, Md). This allowed for a 1-to-1 substitution of the fats for use as a spread as well as in cooking and bak-ing. The margarine contained 38.7%cis -polyunsaturated fatty acids and 7.5%

trans-fatty acids. The butter contained 50.6% saturated fatty acids, of which 8.6% was stearic acid.

Specially formulated products were produced by a local bakery for use in the study. Participants/assessors could not be blinded to the type of fat be-cause of the inherent differences in study products (eg, margarine tub vs butter pat). To avoid accidental sub-stitution of butter fat for margarine fat in cookies and brownies, the bakery was contracted to produce only 1 test prod-uct line at a time. Every 3 months, pro-duction was temporarily halted and switched to the other product line, mak-ing the entry date into the study the cri-terion that assigned diet order. Twenty-three families (n = 104) had diet order butter-margarine, and 23 families (n = 122) had diet order margarine-butter. The 2 diet periods were sepa-rated by at least a 1 month of ad libi-tum diet; families whose participation spanned the Thanksgiving-Christmas-New Year holiday had a 3-month hia-tus between diet periods to avoid prob-lems fitting traditional holiday foods into the dietary regimen.

Based on age, height, weight, physi-cal activity profile, and food

prefer-ences, each family member received a written dietary prescription specify-ing the portion size and frequency of study foods, meat, starchy vegetable, regular vegetable, fruit, and dairy to be consumed every day. Dietary prescrip-tions were adjusted as needed during the first few weeks of each diet period to maintain a constant weight in study participants.

The daily prescription was based on an intake of 25% of energy from test fat. Subjects were instructed to follow the plan as much as possible and to choose low-fat foods for their nonstudy food choices. We expected an inability to ad-here to the protocol for 3 meals each week, with a projected test fat intake of 21% of calories. Families were coun-seled to continue a low–saturated fat diet during their meals that did not con-tain study products. A portable food scale was provided and household cups and bowls were used to instruct fami-lies on how to estimate portion size.

Dietary Evaluation

Compliance to diet was assessed by 3 measures:

Product Inventory for Family. Dur-ing each weekly visit, sufficient study products were delivered to families by investigators to meet 100% adherence Table 1.Baseline Characteristics of Adults and Children From 46 Participating Families*

Characteristics Adults (n = 92) Children (n = 134) Age, y† 41 (4) 12 (4) Sex Males, No. (%) 46 (50) 71 (53) Females, No. (%) 46 (50) 63 (47) BMI, kg/m2_{† 28.8 (6.5) 19.3 (4.3)} Total cholesterol, mg/dL† 184 (27) 152 (25) Triglycerides, mg/dL‡ 122 (74) 74 (41) LDL-C, mg/dL† 121 (27) 95 (22) HDL-C, mg/dL† 44 (11) 46 (11) ApoE, No. (%)§ ApoE 4/4 1 (1) 1 (0.7) ApoE 4/3 27 (29) 36 (27) ApoE 3/3 49 (53) 77 (57) ApoE 3/2 12 (13) 13 (10) ApoE 4/2 3 (3) 5 (4) ApoE 2/2 0 2 (1)

*BMI indicates body mass index; LDL-C, low-density li-poprotein cholesterol; HDL-C, high-density lili-poprotein cholesterol; and ApoE, apolipoprotein E.

†Values are expressed as mean (SD). To convert mg/dL to mmol/L, multiply by 0.0259.

‡To convert mg/dL to mmol/L, multiply by 0.0113. §Percentages may not equal 100 because of rounding.

to the daily dietary prescription. Dur-ing the delivery, an inventory of un-consumed food products was re-corded. Allowing for 3 meals per week without study foods, 85% of the deliv-ered food was expected to be con-sumed.

Daily Check Sheets.All family mem-bers were provided personalized daily check sheets to mark off how much of each study product they consumed each day. Sheets were collected during the weekly visit and recorded, and the grams of test fat consumed were cal-culated. Adherence was estimated as gram intake reported to be consumed divided by gram intake for 100% ad-herence. As with family inventory, the goal intake was 85%.

Three-Day Diet Records.During the third and fourth week of each diet pe-riod, each subject was asked to com-plete a 3-day diet record of total di-etary intake during a weekend day and 2 weekdays. Records were analyzed us-ing Nutritionist IV (San Bruno, Calif ), and the gram intake of test fat was cal-culated. Goal intake of test fat was 21% of total energy.

Laboratory Evaluation

Subjects had blood drawn prior to ini-tiation of the study (ad libitum) and twice during the last 2 weeks of each dietary period. Blood samples were drawn 4 to 7 days apart to minimize the influence of biological variation on mean response.14_{All blood samples} were obtained after having subjects fast for 10 hours; all blood samples were stored using numeric identification code that could be decoded only by the investigators.

Lipoprotein Analysis.Lipid and li-poprotein analyses were performed on each blood sample. Plasma was sepa-rated, and plasma concentrations of to-tal cholesterol (Roche, Indianapolis, Ind) and triglycerides (Sigma Diagnostics, St Louis, Mo) were measured using an en-zymatic procedure. High-density lipo-protein cholesterol (HDL-C) level was measured as the remaining cholesterol in whole plasma after precipitating apo-lipoprotein B–containing apo-lipoproteins

with 6.59 mmol/L of phosphotungstic acid (Dade International, Miami, Fla). To reduce phlebotomy requirements, the LDL-C level was calculated using the Friedewald equation. If a subject had a fasting triglyceride levels greater than 4.52 mmol/L (400 mg/dL), the LDL-C levels were determined by direct assay (Sigma Diagnostics).

Dietary Response.Response to diet was defined as the difference in mean LDL-C level of the margarine intake pe-riod minus the mean LDL-C level of the butter intake period.

Cholesterol Ester and Triglyceride Fatty Acids.Plasma triglyceride and cholesterol ester (CE) fatty acids lev-els were determined in the last blood draw of each diet period by extracting lipids from plasma15_{and separating} lipid classes by thin-layer chromatog-raphy.16

Genotypes for Apolipoprotein E (ApoE) and 7a-Hydroxylase.DNA was extracted from the white blood cell pel-let with DNA-zol Reagent (GIBCO-BRL, Grand Island, NY). 7a -Hydroxy-lase DNA was amplified with 1 pmol/µL of each primer AP2 (59 -TGGTAGG-TAAATTATTAATAGATGT-39) and AE ( 59 A A A T T A A A T G G A T G A A T -CAAAGAGC-39).17 _{The polymerase} chain reaction (PCR)–amplified DNA fragment was digested with 10 units of

BsaI and was electrophoresed. Apolipo-protein E DNA was amplified by PCR in a DNA thermal cycler using oligonucleo-tide primers F4 (59 -ACAGAATTCGC-CCCGGCCTGGTACAC-39) and F6 (59 - TAAGCTTGGCACGGCTGTCCAA-GGA-39).18

Reproducibility of Response To evaluate the reproducibility of in-dividual dietary response, each of the first 20 families that completed the study were asked at the time of their sign-out interview to consider repeat-ing the entire diet study. Two families agreed (n=4 adults and n=8 children). The first family began the second di-etary trial 5 months later and received the study foods in the same diet order as their original study. The second fam-ily repeated the study 3 months later,

receiving the opposite diet order. Body mass index (BMI; calculated as weight in kilograms divided by the square of height in meters) measurements prior to each study were fairly comparable ex-cept in 2 members of the first family who gained 9 kg (20 lb) during the in-terval between their dietary trials.

Statistical Methods

The primary analysis was dietary re-sponse, which by definition could be de-termined only in those families who com-pleted the trial. To avoid underestimating SEs because of correlations between fam-ily members, generalized estimating equations (GEEs) were used to com-pare the 2 diets and construct confi-dence intervals (CIs) to adjust for the lack of independence within families.19 Tri-glycerides levels were log transformed be-cause of skewness; both untransformed and transformed data gave similar sults so only untransformed data are re-ported. Also, GEE was used to assess the reproducibility among the 2 families who agreed to repeat the study.

Mixed linear models were used to assess covariates and variance compo-nents.20,21_{In these models, family} mem-bership was included as a random effect. Bakery run or diet order did not contribute to observed variance. Indi-vidual level covariates considered for the model were either clinically relevant or significantly associated with dietary re-sponse by univariate analysis. These co-variates included age, BMI, ad libitum LDL levels, ApoE genotype, and change in serum fatty acid CE levels. Because of significant interrelationships among these covariates, all possible interac-tions between these covariates also were considered. Apolipoprotein E geno-type was evaluated by combining sev-eral genotypes into 3 categories: ApoE 2,2 plus 2,3; ApoE 3,3 plus 2,4; and ApoE 3,4 plus 4,4. Comparison of these 3 categories was made using 2 dummy variables in the models. Variance com-ponents were used to estimate family in-traclass correlation. Separate regres-sions also were estimated for adults and children. One extreme outlier, a child with a dietary response of –109 mg/dL

(−2.8 mmol/L), was not included in re-gression models.

Statistical analysis was performed us-ing SAS version 8.0 (SAS Institute, Cary, NC). Because of the multiple testing, aP,.01 was assigned as significant.

RESULTS

Dietary Adherence

By all measures, compliance to the di-etary protocol was excellent. Accord-ing to food inventory, families con-sumed a mean (SD) of 88% (12%) and 83% (16%) of test fat delivered during the margarine or butter intake peri-ods, respectively. Daily check sheets showed consistently lower mean (SD) intakes of 75% (18%) and 77% (17%), respectively. The consistently lower in-takes of test fat recorded on daily check sheets may reflect some days in which subjects consumed test fat but did not record their consumption.

The mean intake of macronutrients recorded on the 3-day food records is listed inTABLE2.The goal of test fat in-take of 21% of energy was achieved for the margarine intake period but was sig-nificantly lower for the butter intake pe-riod (18% of calories,P, .01). The lower intake of test fat in the butter in-take period was accounted for by an un-anticipated preference for the marga-rine bread compared with the butter croissants. The 3% less energy from test fat during the butter intake period was offset by small increases in calories from protein (P,.01) and saturated fat (P,.01) from other sources (eg, whole milk and ice cream). Dietary choles-terol was significantly higher (P,.01) for the butter intake period. The over-all dietary goal—to achieve a clini-cally significant difference in intake of cholesterol-raising fatty acids for the 2 diets—was achieved.

Mean Lipoprotein Response to Diet

Consistent with the reported intake, there was no change in body weight be-tween the 2 diet periods (TABLE3). Margarine intake produced signifi-cantly lower total LDL-C levels than butter intake (P,.001). Adults were

more responsive than children on the basis of levels (0.41 mmol/L [16 mg/ dL] vs 0.28 mmol/L [11 mg/dL],P = .03) but not when considered as per-cent reduction of LDL-C levels (11%; 95% CI, − 13% to − 9% mg/dL vs 9%, 95% CI, − 12% to − 6%;P = .17).

No significant differences in HDL-C levels were seen in adults (P= .83) and children (P= .80). A trend for an in-crease in triglycerides during the but-ter intake period was seen in adults (P= .03) but not in children (P= .45). Changes in CE fatty acids mirrored changes in dietary intake, with signifi-cant increases in CE 18:2 content and

decreases in CE 16:0 content in the margarine intake period compared with the butter intake period (P,.005). The primary change in CE content was a substitution for the CE 18:1 content by CE 18:2 content during the margarine intake period, which expressed as a ra-tio fell from percent (SD) of 5.7% (0.9%) in the margarine intake period to 4.9% (0.8%) in the butter intake pe-riod (P,.001).

Dietary Responsiveness

Dietary responsiveness was defined as the mean margarine LDL-C level mi-nus the mean butter LDL-C level. This Table 2.Daily Dietary Intake Calculated From Mean of Paired, 3-Day Food Records*

Nutrient Margarine Intake Period Butter Intake Period

Total energy, mean (SD), cal 2164 (524) 2186 (587) Dietary cholesterol level,

mean (SD), mg/d

249 (111) 330 (119)†

Energy, mean % (SD), from

Carbohydrate 50 (5) 50 (5) Protein 14 (3) 15 (3)‡ Fat 37 (4) 35 (5)† Test fat 21 (5) 18 (6)‡ Saturated fat 9 (2) 16 (4)‡ Monounsaturated fat 14 (2) 14 (2) Polyunsaturated fat 10 (2) 3 (1)† Trans-fat 1.5 (0.4) 0.5 (0.1)‡ Cholesterol-raising fatty acids§ 10 (2) 15 (4)†

*Sixteen subjects failed to complete a food record for a single period, and 1 subject failed to complete a food record for both periods. Data shown are for the 209 remaining paired records.

†P,.001 (using generalized estimating equation that adjusted for family membership). ‡P,.01 (using generalized estimating equation that adjusted for family membership).

§Cholesterol-raising fatty acids are defined as the sum of all saturated fatty acids plustrans-fatty acids minus stearic acid (18:0), a saturated fatty acid known to be neutral in its effects on low-density lipoprotein cholesterol levels.

Table 3.Body Weight, Lipids, and Lipoproteins for Each Dietary Period*

Margarine Butter D(95% CI) PValue† Adults (n = 92) Weight, lb 186 (44) 186 (44) −0.5 (−0.4 to 1.4) .30 Cholesterol, mg/dL 181 (28) 199 (34) −17.6 (−21.0 to −14.0) ,.001 Triglycerides, mg/dL 125 (76) 135 (95) −10.8 (−20.7 to −0.9) .03 LDL-C, mg/dL 116 (27) 131 (32) −15.7 (−18.8 to −12.6) ,.001 HDL-C, mg/dL 46 (12) 46 (12) −0.1 (−1.2 to 0.9) .83 Children (n = 134) Weight, lb 104 (39) 104 (38) −0.37 (−1.3 to 0.6) .43 Cholesterol, mg/dL 150 (25) 161 (32) −11.3 (−15.4 to −7.2) ,.001 Triglycerides, mg/dL 66 (27) 68 (22) −1.8 (−6.6 to 2.9) .45 LDL-C, mg/dL 93 (23) 104 (30) −11.2 (−14.6 to −7.7) ,.001 HDL-C, mg/dL 48 (10) 48 (11) 0.15 (−1.0 to 1.3) .80

*LDL-C indicates low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; and CI, confidence interval. Values are expressed as mean (SD). To convert mg/dL to mmol/L for cholesterol, LDL-C, and HDL-C, mul-tiply by 0.0259, for trigylcerides mulmul-tiply by 0.0113, and lb to kg, multilpy by 0.45.

value was expressed as either an abso-lute value (mmol/L [mg/dL]) or per-centage change from the LDL-C value achieved in the margarine intake pe-riod.

Individual Variation in Dietary Re-sponsiveness.The frequency distribu-tion of individual responses to diet shows that 81% of subjects had lower LDL-C levels in the margarine intake period compared with the butter intake period, and 76% had a 3% or greater reduction in LDL levels (FIGURE).

Reproducibility of Individual Varia-tion in Dietary Responsiveness.Two families (family 1 and family 2) re-peated the diet study and 2 members of family 1 gained 9 kg (20 lb) be-tween dietary challenges. Mean LDL-C levels and the ratio of CE 18:2/18:1 con-tent obtained in the initial butter or margarine intake period were com-pared with those values obtained in the repeat period. The mean difference in change in LDL-C levels between the 2 trials was 0.01 mmol/L (0.5 mg/dL) (P= .78; 95% CI, − 0.08 mmol/L (−3.3 mg/dL] to 0.11 mmol/L [4.4 mg/dL]). Results were fairly reproducible with low responders in the initial period hav-ing low response in the repeat period and vice versa; those subjects who gained weight were less responsive in the second dietary challenge, but their overall response remained in the same rank order.

Familiality Evaluation Using Intraclass Correlation

Self-reported paternity/maternity sta-tus was verified with ApoE and 7a -hydroxylase genotypes expected/ observed in the children compared with their parents. Estimates for propor-tion of variance explained by family membership were made using mixed linear models. Considering adults and children in the same model, family membership accounted for 19% of vari-ability in percent change LDL-C levels (P=.007). When children were consid-ered separately (TABLE4), 40% of the variability in percent change in LDL-C levels was explained by family mem-bership (P= .002).

Predictors of Responsiveness to Diet

No significant correlations were ob-served between variations in compli-ance (estimated by family inventory, daily check sheet, and food record) and variation in response to diet. No signifi-cant correlations were found between di-etary responsiveness and changes in body weight, diet order, age, 7a -hy-droxylase genotype, or sex. Significant associations were found between di-etary responsiveness and BMI, ad libi-tum LDL-C levels, ApoE genotype, CE 18:2 content, CE 18:1 content, and changes in CE 18:2/18:1 ratio. These variables, in turn, were significantly in-tercorrelated, making it difficult to

quan-tify how much each factor influenced re-sponse. Separate prediction models for adults (not genetically related) and chil-dren (2-9 per family genetically re-lated) were as follows: Among chil-dren, family membership explained 42% of the variation in percentage change in LDL-C levels (P=.002) (Table 4). Sig-nificant covariates of response were BMI and change in CE ratio, explaining 26% of variation and leaving 26% still attrib-utable to family membership. The im-portance of BMI was even more strik-ing considerstrik-ing the relative leanness of the children compared with the adults. Among adults, family membership ac-counted for 0% of variation in percent-age change LDL-C levels; BMI, ApoE genotype, and a BMI3ApoE interac-tion accounted for 14% of the variainterac-tion observed.

The interrelationship between ad libitum LDL-C levels (a significant predictor in children) and ApoE geno-type (a significant predictor in adults), the relationship between ApoE geno-type, ad libitum LDL-C levels, and per-centage change in LDL-C levels were further evaluated. The ApoE genotype was significantly associated with ad li-bitum LDL-C levels when comparing either ApoE 2,2 plus 3,2 vs 4,4 plus 4,3 or ApoE 3,3 plus 2,4 vs 4,4 plus 4,3 (bothP= .002). However, ApoE geno-type accounted for only 4% of the varia-tion in ad libitum LDL-C levels. Al-though ApoE genotype contributed

Figure.Frequency of Percent Change in Low-Density Lipoprotein Cholesterol Levels in Adults and Children

30 25 20 15 10 5 0 –60 30 40

% Change in LDL-C (Margarine-Butter), mg/dL % Change in LDL-C (Margarine-Butter), mg/dL

Fr equency 30 25 20 15 10 5 0 –60 –50 –40 –30 –20 –10 0 10 20 30 40 –40 –50 –30 –20 –10 0 10 20 A Adults B Children

more than ad libitum LDL-C levels to dietary response in adults, in children the model including ad libitum LDL-C levels accounted for 7% additional vari-ance than the model including ApoE genotype (data not shown).

Since BMI was the only factor con-sistently appearing in all models, the re-lationship between BMI and response was examined further. Compliance to diet was not associated with BMI, and differences in compliance were not seen across categories of BMI (data not shown). Surprisingly, the change in CE ratio paralleled the change in LDL-C levels; obese persons had approxi-mately half the response in LDL-C lev-els and CE ratio than leaner persons (mean [SD] change in LDL-C levels and change in CE 18:2/18:1 ratio for BMI,21,−13 [17] and 0.85 [0.82], and for BMI$30,−9 [17] and 0.42 [0.75]).

COMMENT

The substitution of margarine for butter creates 3 simultaneous changes

in dietary intake that, in turn, alter total cholesterol and LDL-C levels.22 Two changes—reductions in satu-rated fatty acids and dietary choles-terol intake—lower LDL-C levels. A third change—increases intrans-fatty acid intake—raises LDL-C levels and also may lower HDL-C levels.23 In-creases intrans-fatty acid intake could potentially mitigate the benefits of a margarine-based diet.24_{In our study, a} lowtransmargarine-based diet achieved 11% lower LDL-C levels than a butter-based diet, without differences in HDL-C levels. Our findings agree with those from metabolic diet studies evalu-ating greater25_{and lesser}26 percent-ages of energy from butter vs marga-rine, confirming the long-standing advice to the public at large to choose a tub margarine over butter.27

Translating the public benefits of di-etary modification to a given vidual are difficult because of indi-vidual variation in response to dietary change. As previously reported, we

found the distribution of dietary re-sponse to peak around the mean: 19% of individuals had either no change or a paradoxical increase in LDL-C levels in the margarine intake period com-pared with the butter intake period. The differences in responsiveness to diet could be attributable to genetic fac-tors.28_{In our study, we tested this} hy-pothesis by evaluating how individual family members responded to both a lowering and cholesterol-raising diet. Family dietary responsive-ness data allowed for an estimation of the contribution of family member-ship (shared genes plus shared envi-ronment) on responsiveness. A marga-rine vs butter comparison was chosen since these 2 types of fats lend them-selves to simple substitution in both baking and spreads, and previous stud-ies suggested that the response to di-etary cholesterol and saturated fat ap-pear congruent.29

Ideally, a study of dietary respon-siveness would be conducted under

Table 4.Predictive Models Evaluating the Contribution of Family Membership to the Variability in Dietary Responsiveness With or Without Inclusion of Significant Covariates*

Children (n = 133)† Adults (n = 92)

DLDL-C Levels PercentDLDL-C Levels DLDL-C Levels PercentDLDL-C Levels

Variation explained by family membership, %

42 (P= .002) 40 (P= .002) 3 (P= .42) 0

Model Including Significant Covariates‡

Variable

Children (n = 128)§ Adults (n = 92)

DLDL-C Levels Percent_DLDL-C Levels _DLDL-C Levels Percent_DLDL-C Levels Coefficient SE PValue Coefficient SE PValue Coefficient SE PValue Coefficient SE PValue

Intercept −9.71 −1.58 ,.001 −6.56 1.39 ,.001 −24.99 2.91 ,.001 −16.93 2.33 ,.001 BMI, kg/m2 _{0.63 0.23 .008 0.60 0.21 .005 0.60 0.24 .01 0.48 0.30 .11} DCE 18:2/18:1 −5.97 1.23 ,.001 −5.54 1.13 ,.001 . . . . Ad libitum LDL-C, mg/dL −0.15 0.045 .001 . . . . ApoE 2→4 . . . 10.96 4.91 .03 11.78 4.59 .01 ApoE 3→4 . . . 7.94 3.36 .02 2.71 2.98 .37 BMI×APoE 2→4 . . . −1.17 0.63 .07 BMI×ApoE 3→4 . . . 0.19 0.36 .60 Total variation explained by

model covariates, % 31 (P,.001) 26 (P,.001) 12 (P= .003) 14 (P= .002) Remaining variation explained by family membership, % 24 (P= .03) 26 (P= .02) 0 0

*LDL-C indicates low-density lipoprotein cholesterol; BMI, body mass index; CE, cholesterol ester; and ApoE, apolipoprotein E. To convert mg/dL to mmol/L of LDL-C, multiply by 0.0259. Ellipses indicate that this factor was not a significant contributor to the model.

†One outlier with a change (margarine-butter) in LDL-C of −109 mg/dL (−2.8 mmol/L) was excluded.

‡Regression estimates are computed with random coefficient models to account for correlation among family members; ApoE 2 includes subjects with 2,2 and 3,2; ApoE 3 in-cludes subjects with 3,3 and 4,2; ApoE 4 inin-cludes subjects with 3,4 and 4,4 genotypes.

more strict metabolic control to en-sure standard intake of fats. For the number of subjects in our study, a meta-bolic diet study would have been un-wieldy and costly. Using resources available, we taught families how to measure portion sizes at home and pro-vided detailed low-fat dietary prescrip-tions for each family member. This base diet was supplemented with baked goods and spreadable fats that were pro-vided to the families. Subjects were given an explicit daily and weekly goal for consumption of test fat and test fat products, and this goal and their progress was reviewed during a weekly home visit. Compliance was excel-lent, and the lipid lowering that was achieved matched that of the pre-dicted data derived from other meta-bolic diet studies.

The primary literature linking ge-netic factors with dietary responsive-ness in humans are reports from stud-ies in unrelated individuals in which the genetic analysis occurred after the di-etary trial was completed.30_{One study} in which subjects were preselected based on their genotype, the Apo A-IV allele was associated with variation in response to dietary cholesterol.31_Our study approached the issue of variabil-ity in dietary responsiveness from a dif-ferent angle: we evaluated dietary re-sponsiveness in families who shared lifestyle and dietary habits as well as ge-netic background.

ApoE and 7a-hydroxylase are excel-lent candidate genes for dietary respon-siveness, since both influence LDL-C lev-els,17,32_{and dietary responsiveness is} known to be influenced by baseline LDL-C levels.33_{We did not find a} sig-nificant association between 7a -hy-droxylase genotype and dietary respon-siveness, but we did observe a small effect for ApoE genotype on the dietary response in adults but not in children. Our failure to confirm an association be-tween ApoE genotype and responsive-ness in children may be because of in-adequate power or age differences in the expression of ApoE, since not all stud-ies in children have found an ApoE in-fluence on ad libitum LDL-C

choles-terol levels34-36_{Similar to studies finding} an association between ApoE genotype and response, only 5% to 10% of vari-ance could be attributed to ApoE varia-tion.30

The observation that obese persons are less responsive to diet modifica-tion adds to a growing literature link-ing body weight to lipids and to di-etary response. The linear and positive relationship between body weight and LDL-C levels is present in younger per-sons but appears blunted by age.37,38_In children with familial hypercholester-olemia, body fat is a significant predic-tor of ad libitum LDL-C levels.36_Thus, it should not be surprising that body weight, like ApoE genotype, is an ex-cellent candidate factor for predicting responsiveness. Several studies have ob-served that obese women compared with lean women are less responsive to a cholesterol-lowering diet.39,40 _One study found no difference in response between obese and nonobese men,41_but another observed that nonobese, over-weight men achieved only half of the LDL-C level reduction by diet achieved by lean men.41_{Our findings confirm and} extend the notion that body weight pre-dicts dietary responsiveness in chil-dren as well as adults and for body weight differences even among those who are lean. Excess body weight has no age or sex bias—people who are overweight achieve less of a choles-terol reduction by diet than people who are lean.

We measured CE fatty acids as a bio-logical marker of adherence.42,43_{In our} study, the 18:2 content of the marga-rine diet was far greater than that of the butter diet. The expected increase in the CE content of 18:2 was observed, con-firming adherence. Besides a marker of adherence, we did not anticipate the contribution that other factors make in determining CE content. The CE fatty acid content varies within a relatively narrow range,44_{does not predict} se-rum cholesterol levels,45_{and may be} subject to genetic regulation. In a study of 69 twin pairs and their brothers, monozygotic twins had smaller differ-ences in CE fatty acid content

com-pared with dizygotic twins and broth-ers.46_{Body weight can alter CE fatty acid} content. In the Atherosclerosis Risk in Communities study, even after strati-fying for dietary intakes, the CE satu-rated fatty acid levels were higher and the CE content 18:2 was lower in over-weight men and women than lean men and women.47

When all subjects are stratified by categories of BMI, clear differences in the change in CE fatty acid levels by diet were observed. Although a relative in-crease in CE 18:2 content in the mar-garine intake period was seen in every category of BMI, obese and over-weight persons had less excursion in the CE fatty acid levels during dietary modi-fication than more lean persons. The strong interrelationships between di-etary responsiveness, CE ratio, and BMI raise the hypothesis that the influence of dietary fatty acids on serum choles-terol levels is tempered by the pool of endogenous fatty acids held in adi-pose tissue. If relatively fixed concen-trations of fatty acids in adipose tis-sue48_{mix with dietary fatty acids and} compete for hepatic uptake, even on a low saturated fatty acid diet an obese person’s liver is exposed to more satu-rated fatty acid flux than a lean per-son’s liver. One can only speculate whether differences in fatty acid flux can explain the observations that obese persons are less responsive to diet and have higher cholesterol levels than lean persons.

The unimodal distribution of di-etary responses observed in our study are consistent with our failure to iden-tify a single genetic factor that ac-counts for variation in dietary re-sponse. By studying families, we could determine that 40% of the variability in response to a cholesterol-lowering diet is due to shared traits, whether these are heritable or habitual. Further-more, our study underscores the nearly universal response to a cholesterol-lowering diet in both children and adults. This finding confirms the long standing recommendation promoting a cholesterol-lowering diet for the popu-lation at large.

Author Contributions:Dr Denke designed and ex-ecuted this investigation; Ms Adams-Huet performed all statistical analysis in all phases of the investigation; Ms Nguyen performed phlebotomy, delivered food, di-vided samples, and performed the assays for lipids and lipoproteins, cholesterol ester fatty acids, plasma tri-glyceride fatty acids, free fatty acids, as well as geno-type assays for ApoE and 7a-hydroxylase genotypes. Funding/Support:This work was supported by a grant

from the United Soybean Board, and the National As-sociation of Margarine Manufacturers, and by the Uni-versity of Texas Southwestern GCRC grant (United States Public Health Service grant MO1-RR00633). Disclaimer:The United Soybean Board and National Association of Margarine Manufacturers were influen-tial in the basic dietary comparison of the study (butter vs soybean margarine) but had no other input regard-ing the design of the study, its execution,

interpreta-tion, analysis of the data, writing of the manuscript, or approval of the manuscript text prior to submission. Acknowledgment:We thank Lynne W. Scott, MA, RD, who helped with the initial design of dietary ma-terials, Dick Verstraete who helped perform phle-botomy, Jingping Wang, PhD, who assisted in the ge-netic analysis, and Sharon D. Simpson, MS, RD, who helped with dietary counseling, food delivery, and day-to-day management during part of the study. REFERENCES

1.National Cholesterol Education Program. Report of the expert panel on population strategies for blood cholesterol reduction.Circulation. 1991;83:2154-2232.

2.Keys A, Anderson JT, Grande F. Prediction of serum-cholesterol responses of man to changes in fats in the diet.Lancet. 1957;2:959-966.

3.Hegsted DM, McGandy RB, Myers ML, Stare FJ. Quantitative effects of dietary fat on serum choles-terol in man.Am J Clin Nutr. 1965;17:281-295. 4.Mensink RP, Katan MB. Effect of a dietary fatty ac-ids on serum lipac-ids and lipoproteins: a meta-analysis of 27 trials.Arterioscler Thromb Vasc Biol. 1992;12: 911-919.

5.Yu S, Derr J, Etherton TD, Kris-Etherton PM. Plasma cholesterol-predictive equations demonstrate that stearic acid is neutral and monounsaturated fatty ac-ids are hypocholesterolemic.Am J Clin Nutr. 1995; 61:1129-1139.

6.Wald NJ, Law MR. Serum cholesterol and ische-mic heart disease.Atherosclerosis. 1995;118(suppl 1): S1-S5.

7.Jacobs DR Jr, Anderson JT, Hannan P, Keys A, Black-burn H. Variability in individual serum cholesterol re-sponse to change in diet.Arteriosclerosis. 1983;3: 349-356.

8.Beynen AC, Katan MB, Van Zutphen LF. Hypo- and hyperresponders: individual differences in response of serum cholesterol concentration to changes in diet.Adv Lipid Res. 1987;22:115-171.

9.National Diet-Heart Study Research Group. The Na-tional Diet-Heart Study Final Report.Circulation. 1968; 37(suppl I): I1-I428.

10.Denke MA, Grundy SM. Individual responses to a cholesterol-lowering diet in 50 men with moderate hypercholesterolemia.Arch Intern Med. 1994;154: 317-325.

11.Shenberger DM, Helgren RJ, Peters JR, Quiter E, Johnston EA, Hunninghake DB. Intense dietary coun-seling lowers LDL cholesterol in the recruitment phase of a clinical trial of men who had coronary artery by-pass grafts.J Am Diet Assoc. 1992;92:441-445. 12.Denke MA. Individual responses to a cholesterol-lowering diet in postmenopausal women with mod-erate hypercholesterolemia.Arch Intern Med. 1994; 154:1977-1982.

13.Quivers ES, Driscoll DJ, Garvey CD, et al. Vari-ability in response to a low-fat, low-cholesterol diet in children with elevated low-density lipoprotein cho-lesterol levels.Pediatrics. 1992;89:925-929. 14.Rotterdam EP, Katan MB, Knuiman JT. Impor-tance of time interval between repeated measure-ments of total or high-density lipoprotein cholesterol when estimating an individual’s baseline concentra-tions.Clin Chem. 1987;33:1913-1915.

15.Dole PD, Meinertz H. Microdetermination of long-chain fatty acids in plasma and tissues.J Biol Chem. 1960;235:2591-2599.

16.Lepage G, Roy C. Direct transesterification of all classes of lipids in a one-step reaction.J Lipid Res. 1986; 27:114-120.

17.Wang J, Freeman DJ, Grundy SM, Levine DM, Guerra R, Cohen JC. Linkage between cholesterol 7

ahydroxylase and high plasma low-density lipopro-tein cholesterol concentrations.J Clin Invest. 1998; 101:1283-1291.

18.Hixson JR, Vernier DT. Restriction isotyping of hu-man apolipoprotein E by gene amplification and cleav-age withHhaI.J Lipid Res. 1990;31:545-548. 19. Liang KY, Zeger SL. Longitudinal data analysis us-ing generalized linear models.Biometrika. 1986;73: 13-22.

20.Littell, RC, Milliken GA, Stoup WW, Wolfinger RD. SAS System for Mixed Models. Cary, NC: SAS Insti-tute, Inc; 1986.

21.Singer JD. Using SAS PROC MIXED to fit multi-level models, hierarchical models, and individual growth models.J Educ Behav Stat. 1998;24:323-355. 22.Zock PL, Katan MB. Butter, margarine and se-rum lipoproteins.Atherosclerosis. 1997;131:7-16. 23.Mensink RP, Katan MB. Effect of dietary trans fatty acids on high-density and low-density lipoprotein cho-lesterol levels in healthy subjects.N Engl J Med. 1990; 323:439-445.

24.Ascherio A, Willett WC. Health effects oftrans fatty acids.Am J Clin Nutr. 1997;66(suppl 1):1006S-1010S.

25.Lichtenstein AH, Ausman LM, Jalbert SM, Schaefer EJ. Effects of different forms of dietary hydrogenated fats on serum lipoprotein cholesterol levels.N Engl J Med. 1999;340:1933-1940.

26.Judd JT, Baer DJ, Clevidence BA, et al. Effects of margarine compared with those of butter on blood lipid profiles related to cardiovascular disease risk fac-tors in normolipemic adults fed controlled diets.Am J Clin Nutr. 1998;68:768-777.

27.Krauss RM, Deckelbaum RJ, Ernst N, et al. Di-etary Guidelines for Healthy Americans: a statement for health professionals from the Nutrition Commit-tee, American Heart Association.Circulation. 1996; 94:1795-1800.

28.Tall A, Welch C, Applebaum-Bowden D, Wassef M, and the Working Group. Interaction of diet and genes in atherogenesis. Report of an NHLBI Working Group. Arterioscler Thromb Vasc Biol.1997;17:3326-3331. 29.Katan MB, Berns MA, Glatz JF, Knuiman JT, No-bels A, de Vries JH. Congruence of individual respon-siveness to dietary cholesterol and to saturated fat in humans.J Lipid Res. 1988;29:883-892.

30.Ordovas JM. The genetics of serum lipid respon-siveness to dietary interventions.Proc Nutr Soc. 1999; 58:171-187.

31.McCombs RJ, Marcadis DE, Ellis J, Weinberg RB. Attenuated hypercholesterolemic response to a high-cholesterol diet in subjects heterozygous for the apo-lipoprotein A-VI-2 allele.N Engl J Med. 1994;331: 706-710.

32. Tikkanen MJ, Huttunen JK, Ehnholm, C, Pi-etinen P. Apolipoprotein E4 homozygosity predis-poses to serum cholesterol elevation during high fat diet.Arteriosclerosis. 1990;10:285-288.

33.Denke MA, Frantz ID Jr. Response to a cholesterol-lowering diet: efficacy is greater in hypercholesterol-emic subjects even after adjustment for regression to the mean.Am J Med. 1993;94:626-631. 34.Lehtimaki T, Moilanen T, Viikari J, et al.

Apoli-poprotein E phenotypes in Finnish youths: a cross-sectional and 6-year follow-up study.J Lipid Res. 1990; 31:487-495.

35.Xu C-F, Talmud PJ, Angelico F, Del Ben M, Savill J, Humphries SE. Apolipoprotein E polymorphisms and plasma lipid, lipoprotein, and apolipoprotein levels in Italian children.Genet Epidemiol. 1987;126:202-213.

36.Tonstad S, Leren TP, Sivertsen M, Ose L. Deter-minants of lipid levels among children with heterozy-gous familial hypercholesterolemia in Norway. Arte-rioscler Thromb Vasc Biol.1995;15:1009-1014. 37.Denke MA, Sempos CT, Grundy SM. Excess body weight: an under-recognized contributor to dyslipid-emia in white American women.Arch Intern Med. 1994;154:401-410.

38.Denke MA, Sempos CT, Grundy SM. Excess body weight: an underrecognized contributor to high blood cholesterol in white American men.Arch Intern Med. 1993;153:1093-1103.

39.Cole TG, Bowen PE, Schmeisser D, et al. Differ-ential reduction of plasma cholesterol by the Ameri-can Heart Association phase 3 diet in moderately hy-percholesterolemic, premenopausal women with different body mass indexes.Am J Clin Nutr. 1992; 55:385-394.

40.Hannah JS, Jablonshki KA, Howard BV. The re-lationship between weight and response to cholesterol-lowering diets in women.Int J Obes Relat Metab Dis-ord. 1997;21:445-450

41.Jansen S, Lopez-Miranda J, Salas J, et al. Plasma lipid response to hypolipidemic diets in young healthy non-obese men varies with body mass index.J Nutr. 1998;128:1144-1149.

42.Zock PL, Mensink RP, Harryvan J, de Vries JH, Katan MB. Fatty acids in serum cholesterol esters as quantitative biomarkers of dietary intake in humans. Am J Epidemiol. 1997;145:1114-1122.

43.Sarkkinen ES, Agren JJ, Ahola I, Ovaskainen ML, Uusitupa MI. Fatty acid composition of serum cho-lesterol esters, and erythrocyte and platelet mem-branes as indicators of long-term adherence to fat-modified diets.Am J Clin Nutr. 1994;59:364-370. 44.Moilanen T, Nikkari T, Rasanen L, et al. Plasma cholesteryl ester fatty acids in 3- and 12-year-old Finn-ish children.Atherosclerosis. 1983;48:49-56. 45.Knuiman JT, West CE, Hermus RJ, Hautvast JG. Fatty acid composition of cholesteryl esters in serum in boys from 16 developing and developed coun-tries.Atherosclerosis. 1980;37:617-624.

46.Kang KW, Christian JC, Hedges B, Corey LA, Pow-ell R. Fatty acid variability of plasma lipids and cho-lesteryl esters in adult male twins and their brothers. Lipids. 1977;11:722-726.

47.Ma J, Folsom AR, Shahar E, Eckfeldt JH, for the Atherosclerosis Risk in Communities (ARIC) Study In-vestigators. Plasma fatty acid composition as an in-dicator of habitual dietary fat intake in middle-aged adults.Am J Clin Nutr. 1995;62:564-571. 48.Hudgins LC, Hirsch J. Changes in abdominal and gluteal adipose-tissue fatty acid compositions in obese subjects after weight gain and weight loss.Am J Clin Nutr. 1991;53:1372-1377.