- Email: [email protected]
Relationships of dietary fat consumption to serum total and low-density lipoprotein cholesterol in hispanic preschool children
PREVENTIVE
MEDICINE
20, 237-249 (l!%‘l)
Relationships of Dietary Fat Consumption to Serum Total and Low-Density Lipoprotein Cholesterol in Hispanic Preschool Children’ STEVEN SHEA, M.D.,*+* CHARLES E. BASCH, PH.D.,S MATILDE IRIGOYEN, M.D.,9 PATRICIA ZYBERT, PH.D.,$ JILL L. RIPS, M.A., M.PHIL. ,$ ISOBEL CONTENTO, PH.D. ,$ AND BERNARD GUTIN, PH.D.S Departments of *Medicine and BPediatrics, and fSchoo1 of Public Health, Columbia University Health Sciences Division, and SCenter for Health Promotion, Teachers College, Columbia University, New York, New York 10027 Background. Studies of the relationship between dietary fat intake and serum lipids in young children have yielded inconclusive results. We studied this relationship in 108 Hispanic children ages 4-5 years. Methods. Four 24-hr recalls approximately 3 months apart and two Willett semiquantitative food frequency questionnaires approximately 6 months apart were obtained by interviewing the children’s mothers. Diet measures were averaged for the multiple administrations of each of these instruments. Results. Based on the 24-hr recalls, children in the highest tertile of total fat consumption (36.2% of total calories) compared with the lowest tertile (30.2% of total calories) had mean total serum cholesterol of 4.32 mmolihter (167 mg/dl) vs 3.91 mmol/Iiter (151 mg/dl) (test for linear trend across tertiles, P < 0.05) and mean low-density lipoprotein cholesterol of 2.74 mmol/Iiter (106 mg/dl) vs 2.29 mmollliter (89 mg/dl) (test for linear trend, P < 0.01). Children in the highest tertile of saturated fat consumption (14.6% of total calories) compared with the lowest tertile (11.2% of total calories) had mean total serum cholesterol of 4.39 mmol/Iiter (170 mg/dl) vs 3.97 mmol&ter (154 mg/dl) (test for linear trend, P < 0.05) and mean lowdensity lipoprotein cholesterol of 2.80 mmolihter (108 mg/dl) vs 2.35 mmol/liter (91 mg/dl) (test for linear trend, P < 0.01). These relationships remained significant when calorieadjusted nutrient intakes were examined and after adjustment in multiple linear regression models for age, sex, and body mass index, with the exception of the association of calorieadjusted total fat with total serum cholesterol level (P = 0.07). Similar results were obtained using the Willett questionnaires. Conclusions. These findings indicate that dietary fat, particularly saturated fat consumption, is an important correlate of blood lipid levels in preschool children. These are also the first reported data indicating that the Willett questionnaire, as a method for measuring the atherogenic components of diet, has criterion-related validity in young children. 0 1991 Academic
Press, Inc.
INTRODUCTION
Increasing recognition has been gained over the last decade for the importance of maintaining moderate serum cholesterol (TC) levels in childhood (1, 2). The ’ Supported by Grants HL35189 and HL21006 from the National Heart, Lung, and Blood Institute. ’ To whom reprint requests should be addressed at Columbia University, Atchley Pavilion 1310, 161 Fort Washington Avenue, New York, NY 10032. 237 0091-7435191$3.00 Copyrisbt 0 1991 by Academic PIWS, Inc. Au rights of reproduction in any form reserved.
SHEA ET AL.
1984 National Institutes of Health (NIH) Consensus Conference, the American Heart Association, and the American Academy of Pediatrics have all issued statements recommending strategies for the detection and management of children over the age of 2 years with high blood TC levels (3-S). The NIH Consensus Conference also recommendedthat U.S. children over the age of 2 years should adopt a diet with total fat intake of less than 30% of total calories, saturated fat intake of less than 10% of total calories, and daily dietary cholesterol intake of 25&300 mg or less (3). For children and adults with hyperlipidemia, dietary treatment is the fust step (4, 6). The goals of dietary intervention among pediatric populations are twofold: to establish healthy eating habits that will persist into adulthood and to prevent or treat hyperlipidemia (3-5). Thus a considerable body of public health policy and clinical practice rests on the link between dietary fat intake and serum lipid levels in childhood. For adults, a great deal of data exist from observational studies in free-living populations (7,8), from dietary manipulation in closely controlled metabolic ward studies (9-16) and from clinical trials (17-20), indicating that dietary fat intake, particularly saturated fat, influences total and low-density lipoprotein cholesterol (LDL-C) levels, but the data addressing this issue in young children are less extensive. Fart-is er al. studied 447 infants in Bogalusa and found a relationship between consumption of cow’s milk, which is high in fat, compared with formula, which is lower in fat, and serum TC at 6 months (21). Berenson et al. studied 440 children from this same cohort at 1 year of age and found correlations between consumption per 1,000 kcal of total fat, animal fat, and dietary cholesterol and serum TC level (22). Nicklas et al. analyzed relationships between nutrient intakes and serum lipid levels in 50 children at ages4 and 7 years. Dietary data were obtained using a 24-hr recall administered to the child’s main caregiver. At age 4, significant relationships were found between dietary cholesterol analyzed as milligrams per 1,000kcal (but not as mg/day or mg/kg of body weight) and both TC and LDL-C, and between saturated fat analyzed as grams per kilogram of body weight (but not as g/day or g/l000 kcal) and TC and LDL-C. At age 7, only the relationship of dietary cholesterol analyzed as milligrams per 1000kcal with serum LDL-C was significant (23). Frank et al. studied 185 children in the Bogalusa cohort at age 10 and found that children in the highest tertile of serum TC level differed from those in the lower tertiles with respect to dietary intake of total fat, animal fat, saturated fat, and unsaturated fat (24). Glueck et al. reported significant relationships between diet and serum lipid levels in 1,234white children ages 6-19 years, but the relationships were weak and limited to age/sexsubgroups. Diet was estimated using a 24-hr recall. Specifically, Glueck et al. found that among boys ages 6-12 years, the dietary polyunsaturated/saturated (P/S) fat ratio was inversely associated with plasma TC, and LDL-C and dietary cholesterol intake was positively associatedwith plasma high-density lipoprotein cholesterol (HDLC) (25). Morrison et al. reported data from 1,669children ages6-19 years in which plasma LDL-C was inversely correlated with the dietary P/S fat ratio (26). Crawford et al., in a study of 84 children at age 6 years, found that the percentage calories from fat and the amount of dietary cholesterol were correlated with serum TC level (27). Vartiainen et al. put 36 children ages 8-18 years on a cholesterol-
DIET
AND
LIPIDS
239
lowering diet in which total fat consumption was reduced from 35 to 24% of total calories and found a mean reduction of 15% in serum TC (28). Ford and her co-workers, in a study of male students 12-18 years of age in a boarding school where the cafeteria diet could be controlled, found that reduction of dietary total and saturated fat and dietary cholesterol produced a 15% mean reduction in serum TC (29, 30). None of these studies used semiquantitative food frequency questionnaires to estimate habitual diet and none adjusted nutrient intake for total calorie consumption. The calculation of nutrient density by expressing nutrient consumption per 1000kcal, while useful for dietary prescription and for comparing diets, does not control for confounding by total calorie consumption. Regression methods, as described by Willett, are necessary for this purpose (31, 32). Only Nicklas et al. (23) and Crawford et al. (27) studied preschool children, and only the studies of Nicklas et al. (23), Frank et al. (24), Vartiainen et al. (28), and Ford et al. (29, 30) showed relationships between saturated fat intake and lipid levels. Finally, none of these studies included significant numbers of Hispanic children or attempted to measurethe diet eaten among Hispanic populations living in the United States. In this article, we report a study of 108 Hispanic preschool children in which the relationships between dietary fat consumption and serum lipid levels were investigated. METHODS Subjects and Setting
Children were recruited between May, 1986,and May, 1987,for the Columbia Study of Childhood Activity and Nutrition, a longitudinal study of healthy preschool children and their parents. Children were eligible for the study if they were in good health and without dietary prescriptions or restrictions, lived in the primary referral area of The Presbyterian Hospital, and were between 36 and 48 months of age at time of entry. Only one child from each family was enrolled; no family had more than one age-eligible child. Families were excluded if the child was not living with the parent or the family did not appear likely to meet the requirements of six visits per year over 3 years to the study field office. Study subjects were drawn from a primarily Hispanic, urban, low-income neighborhood in northern Manhattan, New York City. Additional details of the recruitment procedure have been published elsewhere (33). Of 240 Hispanic families recruited into the study, 24 were lost to follow-up before the children’s serum lipid measurements were made, 106 families declined cholesterol screening for their children, and 2 children did not complete all four 24-hr diet recalls during the year in which the serum lipid sampleswere drawn. Thus there were 108Hispanic children with complete data for the purpose of the present analysis. There were no differences in age, sex, or body mass index (BMI) between these 108children and the 106 children whose parents refused permission for phlebotomy of their children. There were 57 boys and 51 girls. In 2 children, both girls, triglyceride levels were not obtained, so that LDL-C was calculated in 106 children. The 108 study children were all age 4 and 5 years at the time of lipid determination (mean 60.2
240
SHEA ET AL.
months, SD = 5.49), with the exception of one who was 3 years and 11 months old and one who had just turned 6 years old. Measures
Children’s diet was estimated using a 24-hr recall administered on four occasions in the first and third years of the study, approximately 3 months apart to account for seasonal variation, and using the Willett semiquantitative food frequency questionnaire administered on two occasions each year, approximately 6 months apart. Both diet measureswere obtained from the children’s mothers by trained bilingual interviewers using standard questionnaires with probes conceming food preparation methods and brand namesfor foods, three-dimensional food models, food packages, glasses, cups, spoons, and other utensils. For the 24-hr recalls, foods and portion sizes were converted to nutrient intakes using a dietary analysis program (34) based on published nutrient data (35). For the Willett questionnaires, nutrient analysis was done using a computer program supplied by Willett (W. C. Willett, personal communication). Additional foods commonly eaten by the study population but not included in the dietary analysis programs for the 24-hr recalls or the Willett questionnaire were added from another source of nutrient values (36). For the 24-hr recalls, the averageof the four administrations closest in time to drawing the blood samples for lipid analysis was used. The intraclass correlation coefficients for these four measures were 0.28 for total calories, 0.23 for total fat consumption (g/day), 0.29 for saturated fat consumption (g/day), and 0.19 for cholesterol consumption (mg/day). For the Willett questionnaires, the averageof the two administrations closest in time to drawing the blood samples for lipid analysis was used. The intraclass correlation coefficients for these measureswere 0.39 for total calories, 0.38 for total fat consumption (g/day), 0.30 for saturated fat consumption (g/day), and 0.19 for cholesterol consumption (mg/day). Following Willett’s recommendations (31, 32) the natural logarithm transformation of dietary variables was used to improve normality, and regression methods were used to adjust fat and cholesterol consumption for differences in total caloric intake. Specifically, regressions were performed with the natural logarithm of the nutrient of interest as the dependent variable and the natural logarithm of calorie consumption as the independent variable, and the residuals from these models were added to the expected nutrient intake for the mean of the natural logarithm of the caloric intake (31, 32). These methods were used for data derived from the 24-hr recalls as well as the Willett food frequency questionnaires. Untransformed levels of nutrient consumption are presented in Tables 1 and 2, although transformed variables were used in statistical analyses. Fasting venous blood sampleswere analyzed for serum TC and HDL-C (37) and triglyceride levels (38) in a Columbia University research laboratory calibrated to Centers for Disease Control standards. LDL-C was calculated as TC minus the sum of HDL-C and one-fifth the triglyceride level (6). This formula is valid when triglycerides are less than 4.52 mmohliter (400 mg/dl), which was the case for all subjects. Weight was measuredusing a calibrated balance scale. Height was measured using a stadiometer. BMI was calculated as the weight (kilograms) divided by the height (meters) squared.
DIET AND LIPIDS
241
The study population was divided into tertiles based on both raw and calorieadjusted consumption of total fat, saturated fat, and cholesterol. Tests for linear trend were performed to assess the relationships between nutrient intakes and lipid levels. Differences in mean serum lipid levels among tertiles were tested using Tukey’s studentized range test (39), which adjusts for multiple comparisons. To confirm these bivariate associations, separate multiple linear regression models were constructed with TC and LDL-C as the dependent variables. Age in months, sex, and BMI and raw or calorie-adjusted dietary total fat, dietary total saturated fat, or dietary cholesterol intake were entered into each model. Unstandardized regression coefficients are reported in Table 4. Data were analyzed using SPSS-PC+ (40). The study was approved by the Human Subjects Committee of Teachers College and the Institutional Review Board of the Columbia University Health Sciences Division. RESULTS
Mean serum TC level for the entire sample was 4.10 mmol/liter (158.5 mg/dl) and mean LDL-C level was 2.51 mmolihter (97.2 mg/dl). Serum TC and LDL-C were highly correlated (N = 106; r = 0.95; P < 0.001). Thirty-seven children (34.3%) had serum TC a4.40 mmol/liter (170 mg/dl). Among boys, mean serum TC level was 4.15 mmolfliter (160.5 mg/dl) and mean LDL-C was 2.58 mmol/liter (99.7 mg/dl). Among girls, mean serum TC level was 4.04 mmol/liter (156.2 mg/dl) and mean LDL-C was 2.44 mmol/liter (94.3 mg/dl). Mean consumption of total fat for the entire sample was 33.0% of total calories based on the 24-hr recalls and 33.2% based on the Willett questionnaires; mean consumption of saturated fat was 13.2%of total calories based on the 24-hr recalls and 13.5% based on the Willett questionnaires; and mean consumption of cholesterol was 193.9 mg/lOOOkcal based on the 24-hr recalls and 165.6 mg/lOOOkcal based on the Willett questionnaires. Total calorie intake estimated from the Willett questionnaires was higher than that from the 24-hr recalls. Dietary total fat and saturated fat were highly correlated and both were highly correlated with total calorie consumption (Table 1). Mean consumption of total fat in the lowest tertile was 40.5 g/day (30.2% of total calories) based on data from the 24-hr recalls, compared with 79.3 g/day (36.2% of total calories) in the highest tertile. Differences between the lowest and highest tertile were also approximately twofold for absolute consumption of saturated fat and cholesterol. Calorie-adjustment reduced the difference between lowest and highest tertile substantially (Table 2). The differences in lipid levels between the highest and the lowest tertiles of total fat consumption (as a percentage of the mean of the lowest tertile) were 10.5% for TC and 19.7% for LDL-C. For saturated fat consumption, these differences were 10.6% for TC and 19.2% for LDL-C . Based on estimates derived from the 24-hr recalls, tests for linear trend for serum TC level (Table 2) across tertiles of nutrient consumption were significant for total fat intake (P < 0.05), saturated fat intake (P < 0.05), and calorie adjusted saturated fat intake (P < 0.01). Tests for linear trend for LDL-C level were significant across tertiles of nutrient consumption for total fat intake (P < O.Ol),
242
SHEA ET AL. TABLE 1 CORRELATIONSAMONG NUTWENT CONSUMPTIONLEVELS (iV = 108 SUBJECTS)
CalorieTotal fat
Calorie-
Calorie-
Saturated adjusted adjusted adjusted fat Cholesterol total fat saturated fat cholesterol
Nutrient consumptions” estimated using Total calories 0.93** 0.84** 0.55** 0.91** 0.54** Total fat 0.53** Saturated fat Cholesterol Calorie-adjusted total fat Calorie-adjusted saturated fat
24&r recalls
0.00 0.38** 0.35** 0.07
0.00 0.24* 0.55** 0.13 0.64**
0.00
0.03 0.08 0.83** 0.09 0.15
Nutrient consumptions” estimated using the Willett semiquantitative food frequency questionnaire 0.59** 0.00 0.92** 0.85** 0.00 0.00 Total calories 0.63** 0.38** 0.93** 0.27* 0.10 Total fat 0.72** Saturated fat 0.38** 0.53** 0.26* 0.22 0.40** 0.80** Cholesterol 0.71** 0.27* Calorie-adjusted total fat 0.50** Calorie-adjusted saturated fat QNutrients were transformed using the natural logarithm of the nutrient (grams/day for total fat and saturated fat and mg/day for cholesterol). * P (one-tailed) < 0.01. ** P (one-tailed) < 0.001.
saturated fat intake (P < O.Ol), calorie-adjusted total fat intake (P < 0.05), and calorie-adjusted saturated fat intake (P < 0.01). Simiiar results were obtained from analysesbased on data from the Willett questionnaires (Table 3). Multivariate analysescontrolling for sex, age, and BMI (Table 4) confirmed these relationships with the exception of the association of calorie-adjusted total fat with serum TC level (P = 0.07). Relationships between dietary variables and HDL-C and triglyceride levels were examined using these same methods, and none of the findings was statistically significant. In the multivariate analyses, relationships between dietary cholesterol and serum lipid levels were not statistically significant (Table 4). Relationships between dietary polyunsaturated fat intake and lipid levels were also examined; these relationships were not statistically significant. DISCUSSION
Kimm et al. analyzed dietary patterns in U.S. children using data from the second National Health and Nutrition Examination Survey (NHANES II), which was conducted between 1976 and 1980. These investigators found that among children 2-10 years of age mean total fat consumption was 36% of total calories, mean saturated fat consumption was 13% of total calories, and mean cholesterol consumption was 151 mg/lOOOkcal(41). The findings in our sample of Hispanic children were in close agreement with these values for saturated fat and cholesterol. Total fat consumption in our samplewas slightly lower. These comparisons should be interpreted with caution, since the NHANES II data were obtained
DIET
AND
243
LIPIDS
TABLE 2 MEANS (STANDARD DEVIATIONS) OF SERUM LIPID LEVELS IN HISPANIC PRESCHOOL CHILDREN BY TERTILE OF DIETARY CONSUMPTION PER DAY ESTIMATED USING THE MEAN OF FOUR 24-~~ RECALLS
COLLECTED
DURING
1 YEAR Serum lipids
Dietary
per day Lowest Middle Highest Saturated fat (g) Lowest Middle Highest Cholesterol (mg) Lowest Middle Highest Calorie-adjusted Lowest total fat f,g) Middle Highest Lowest Calorie-adjusted saturated fat (g) Middle Highest Calorie-adjusted Lowest cholesterol (mg) Middle Highest
Total fat (g)
tertile tertile tertile tertile tertile tertile tertile tertile tertile tertile tertite tertile tertile tertile tertile tertile tertile tertile
40.5 54.0 79.3 15.6 22.2 31.4 179.3 281.1 438.9 48.1 55.2 62.8 18.0 22.0 26.5 31.4 47.0 66.6
Total cholesterol (N = 108)
consumption
(7.7) (3.9) (16.3) (3.1) (1.8) (6.3) (36.7) (32.4) (78.2) (3.0) (1.7) (4.0) (2.2) (1.0) (2.3) (5.2) (3.6) (12.8)
Nutrient density 30.2“ 32.7 36.2 11.2” 13.8 14.6 140.2b 187.7 253.6
mmoVliter 3.91 4.07 4.32 3.97 3.93 4.39 3.92 4.16 4.21 3.95 4.04 4.30 3.87 4.06 4.37 3.97 4.21 4.11
(0.74) (0.83) (0.66) (0.76) (0.80) (0.63) (0.79) (0.81) (0.64) (0.73) (0.80) (0.71) (0.62) (0.83) (0.74) (0.71) (0.76) (0.80)
ddl 151.0 157.4 166.9 153.6 151.9 169.9 151.5 160.9 162.9 152.8 156.3 166.2 149.5 156.9 169.0 153.5 162.8 159.1
(28.6) (32.0) (2X4)+’ (29.5)* (30.9)* (24.3)+ (30.6) (31.5) (24.9) (28.2) (31.1) (27.5) (24.1)* (31.9) (28.7)‘* (27.5) (29.5) (30.9)
LDL cholesterol (N = 106) mmoliliter 2.29 2.50 2.74 2.35 2.40 2.80 2.32 2.57 2.65 2.36 2.47 2.71 2.29 2.46 2.81 2.40 2.60 2.54
(0.65) (0.70) (0.62) (0.69) (0.65) (0.61) (0.68) (0.74) (0.57) (0.64) (0.72) (0.64) (0.53) (0.77) (0.61) (0.65) (0.67) (0.71)
mW 88.7 %.6 106.1 90.8 92.8 108.3 89.8 99.2 102.5 91.4 95.5 104.7 88.5 95.0 108.8 92.7 100.6 98.4
(25.1); (27.0) (24. I)#*” (26.6)* (25.2)* (23.5)“* (26.4) (28.8) (21.9)* (24.8) (27.7) (24.8)” (20.4)* (29.9) (23.6)“““ (25.1) (25.7) (27.6)
* Percentage total calories. b mg per 1000 kca.I. * P (two-tailed) c 0.05 vs highest tertile. # P G 0.05 for linear trend test across tertiles. ## P =Z 0.01 for linear trend test across tertiles.
from a representative national sample and the data were collected approximately 10 years earlier than the data in our study. In the Lipid Research Clinics Prevalence Study, at Visit 2, mean plasma cholesterol was 155.2mg/dl for white males ages 5-9 years and 164.2mg/dl for white females of the sameage. Mean plasma LDL-C was 92.5 mg/dl for white males and 100.4mg/dl for white females in this age group (42). These values are comparable to those obtained in our sample of Hispanic children ages 4 and 5 years. Caution is also indicated in interpreting this comparison, since the Lipid Research Clinics obtained a national sample, the children were somewhat older, and the data were obtained in the 1970s.In the study by Nicklas et al. (23) of children age 4 years in the Bogalusa Heart Study, children in the lowest tertile of saturated fat consumption had a mean serum TC level of 155.4mg/dl, compared with 153.6mg/dl in our study, and children in the highest tertile had a mean serum TC level of 169.7 mg/dl, compared with 169.9mg/dl in our study. Thus, the diet and the serum lipid levels found in our study appear generally comparable to those reported in other U.S. populations of similar age. In studies of the relationship between diet and serum lipids, the valid and reliable measurementof diet is essential. Three studies have reported that parental recall correlates well with direct observation of dietary consumption in preschool children (43-45). These studies pertain to short-term recall methods rather
244
SHEA ET AL.
TABLE 3 MEANT (STANDARD DEVIATIONS) OF SERUM LIPID LEVELS IN HISPANIC F’RESCH~~L CHILDREN BY TERTILE OF DIETARY CONSUMPTION PER DAY ESTIMATED USING THE WILLETT SEMIQUANTITATIVE FOOD FREQUENCY QUESTIONNAIRE Serum lipids Dietary consumption Per day
Lowest tertile Middle textile Highest tertile Saturated fat (g) Lowest tertile Middle tertile Highest tertile Cholesterol (mg) Lowest tertile Middle tertile Highest tertile Calorie-adjusted Lowest tertile total fat (9) Middle tertile Highest tertile Calorie-adjusted Lowest tertile saturated fat (g) Middle tertile Highest tertile Calorie-adjusted Lowest tertile cholesterol (mg) Middle tertiie Highest tertile
Total fat (g)
Nutrient density
70.1 (9.9) 97.8(7.4) 137.7 (31.0) 29.1 (4.0) 40.0 i2.5j 53.7 (8.7) 318.7 (56.1) 435.4 (23.2) 578.9 (75.8) 85.8 (5.3) 97.2 (3.1)
110.2 (8.6) 33.6 39.8 46.0 349.4 432.7 531.6
(3.0) (1.8) (2.7) (62.1) (15.4) (62.1)
30.7" 32.3 36.6 12.5" 13.4 14.6 145.56 167.5 183.9
Total cholesterol (N = 108) mmol/liter 3.98(0.71) 3.91 (0.78) 4.4110.70) 3.93 (0.77) 3.93 (0.70) 4.44(0.70) 3.93 (0.80) 4.11 (0.67) 4.26(0.78) 3.85 (0.73) 4.24 (0.77) 4.21 (0.73) 3.99 (0.76) 3.87 (0.73) 4.43 (0.68) 3.95 (0.76) 4.02 (0.69) 4.32(0.78)
mg/dl 153.8 (27.6)*
151.0 (30.1)’ 170.6 (27.0)* 151.9 (29.9)* 151.9 (26.9)* 171.6 (27.1)"“" 151.8 (31.1) 158.8 (25.8) 164.8 (30.0) 149.0 (28.3) 163.8 (29.6) 162.6(28.3)# 154.4 (29.3)* 149.7 (28.2)' 171.3 (26.4)" 152.8 (29.4)
155.5 (26.8) 167.1 (30.3)'"
LDL cholesterol (N = 106) mmoliliter 2.42(0.62) 2.34 (0.67) 2.79 (0.66) 2.38(0.69) 2.36(0.61) 2.80(0.66) 2.29 (0.67) 2.57 (0.62) 2.68(0.70) 2.24 (0.61) 2.64 (0.62) 2.66 (0.72) 2.38 (0.67) 2.32(0.62) 2.86 (0.62)
mg/dl 93.5 (24.1)* 90.5 (25.9)* 107.9(25.6)" 92.1 (26.5)' 91.3 (23.4)* 108.4 (25.5)"‘" 88.6(26.0)** 99.2(23.8) 103.7 (26.9)* 86.5 (23.7). 102.3 @LO)* 102.8 (27.9)*+ 91.8 (25.8)** 89.5 (24.1)**
2.31 (0.65)
110.5 (23.9)“+“ 89.3 (25.1)*
Z.SO(O.63) 2.73 (0.70)
96.7 (24.2) 105.6(27.0)"+
(1Percentage total calories. b mg per 1000kcal. * P (two-tailed) c 0.05 vs highest tertile. ** P (two-tailed) < 0.01 vs highest tertile. * P s 0.05 for linear trend test across tertiles. ** P < 0.01 for linear trend test across tertiles.
than semiquantitative food frequency questionnaire methods. Studies of adults have shown that food frequency questionnaires tend to overestimate the absolute amounts of nutrients and total calories consumed, compared with 24-hr recalls or food diaries (46). In our study, the estimates of total and saturated fat and cholesterol intake obtained using the Willett questionnaire were substantially higher than estimates obtained from the 24-hr recall (Tables 1 and 2). This was the case for both calorie-adjusted and unadjusted values. Thus the descriptive values for the children’s diet based on the 24-hr recalls are preferred. There was, however, very close agreement between the two methods in estimating nutrient densities, expressed as a percentage of total calories provided by total and saturated fat or as milligrams of cholesterol per 1000kcal. The associations between dietary fat intake and serum lipid levels were somewhat stronger and more consistent using estimates derived from the Willett questionnaires than from the 24-hr recalls. We believe the most likely explanation is related to the higher intraclass correlations for the WilIett questionnaires, compared with the 24-hr recalls, indicating greater internal consistency for the repeat measures.This is expected becausethe purpose of the Willett questionnaire is to measurehabitual diet, while the 24-hr recall measuresrecent diet, and diet varies
245
DIET AND LIPIDS TABLE 4
MULTIPLE LINEAR REGRESSION ANALYSES OF ASSOCIATIONS OF DIETARY VARIABLES” WITH SERUM LIPID LEVELS,~ ADJUSTED= FOR AGE (IN MONTHS), SEX, AND BODY MASS INDEX
Serum total cholesterol P
SE g
P value
Serum LDL cholesterol R2
P
Dietary variables estimated using 24hr recalls 0.23 0.006 0.63 0.08 0.65 Dietary total fat 0.74 0.21 0.001 0.12 0.76 Dietary saturated fat 0.18 0.12 0.29 0.03 0.30 Dietary cholesterol 1.10 0.61 0.07 0.04 1.10 Calorie-adjusted total fat 1.10 0.38 0.005 1.13 0.08 Calorie-adjusted saturated fat 0.71 0.08 0.22 0.01 0.09 Calorie-adjusted cholesterol
SE p
Pvalue
R2
0.20 0.18 0.16 0.54 0.34 0.20
0.002
0.11 0.16 0.05 0.05 0.11 0.02
Dietary variables estimated using the Willett semiquantitative food frequency questionnaire Dietary total fat Dietary saturated fat Dietary cholesterol Calorie-adjusted total fat Calorie-adjusted saturated fat Calorie-adjusted cholesterol
0.52 0.63 0.31 1.36 1.14 0.17
0.24 0.26 0.26 0.61 0.49 0.32
0.04 0.02 0.23 0.03 0.02 0.61
0.05 0.06 0.02 0.05 0.06 0.01
0.44 0.61 0.37 1.51 1.38 0.36
0.22 0.24 0.23 0.54 0.44 0.29
0.04 0.01 0.11 0.01 0.002 0.21
0.05 0.08 0.04 0.08 0.10 0.03
LINutrients were expressed as the natural logarithm of the nutrient (grams/day for total fat and saturated fat and mg/day for cholesterol). b Total and LDL cholesterol were expressed as mmolfliter. c Adjustment was done by multiple linear regression.
within individuals over time. Studies of adults have found that as many as seven 24-hr recalls may be necessary to control fully for intraindividual variation in diet (47), so that four 24-hr recalls administered during a l-year period, as in our study, may not be enough to obtain statistically precise estimates of fat intake. Thus, despite the overestimation of absolute amounts of nutrients consumed, the Willett questionnaire gave more precise estimates. We have emphasized the 24-hr recall data in this report because the Willett questionnaire has been validated for adults (4g-50) but not, to our knowledge, for children, and because the absolute intake of dietary fat and cholesterol as estimated from the 24-hr recalls is closer to what is expected for children of this age (41). The results of our study suggestthat the Willett questionnaire has criterionrelated validity as a method for measuring the atherogenic components of diet in young children and convergent validity with results obtained using the 24-hr recall. Further research is needed to define the validity of the Willett questionnaire as a diet measure in epidemiologic studies in young children. As pointed out by Willett (32), when two dietary variables are highly correlated with each other, such as total fat and saturated fat intake in our study (r = 0.91; Table I), statistical techniques do not establish which one is the true independent variable and which the covariate. The problem of multicolinearity also affects serum TC and LDL-C, which were very highly correlated (r = 0.95). Thus, associations of dietary variables with both blood lipid parameters should be interpreted as confirmatory rather than as two independent findings.
246
SHEA ET AL.
CONCLUSION
No previous studies of the relationship between diet and serum lipid levels in Hispanic children have been reported, despite the growing demographic and public health importance of this population. In 1988 there were approximately 19.4 million Hispanics in the United States (51). Hispanics constitute approximately 8.1% of the U.S. civilian noninstitutionalized population (51) and are the fastest growing minority group (52). The demographic composition of our sample indicates that it is of low socioeconomic status, recently immigrated, and largely Dominican (53). The dietary practices of recently immigrated Hispanics or Hispanics from the Dominican Republic may have unique characteristics. Further research is needed to verify our findings in population-based samples and in non-Dominican Hispanic subpopulations. The findings of our study are consistent with findings from most previous studies of dietary fat consumption and serum lipid levels in children, which found at least some significant relationships (21-30). Other studies, of older children or children of less restricted age range, found no associations between nutrient intake and lipid levels (54, 55) or reported less clear-cut relationships (24-26). Possible explanations for these negative observational studies include less variation in serum lipids or in diet, less precise measuresof diet, or a true lack of relationship in the populations studied. Intervention studies in older children have shown that dietary modification produces changesin blood lipid levels (28-30, 56). The multiple R* for the associationsbetween dietary fat and serum lipid levels were small, indicating the importance of measurement error in diet, due to intraindividual variation or imprecision in the method, measurementerror in the lipid levels, also related to intraindividual variation or methodological imprecision, the influence of nondietary factors, such as genetic predisposition, or combinations of these factors. Despite the modest size of the multiple R* statistics, the differences in mean serum TC between the highest and the lowest tertile of children’s total and saturated fat intake were 10.5 and 10.6%, and the differences for LDL-C were 19.7 and 19.2%. These differences are large enough to have clinical and public health as well as statistical significance. In summary, our study strengthens the existing data that support the role of diet in determining blood lipid levels in children, and it suggeststhat the Willett questionnaire, or other food frequency methods for measuring diet, may have utility in young children. ACKNOWLEDGMENTS We are grateful to Dr. Walter Willett for providing the nutrient analysis programs, to Dr. Henry Ginsberg for performing the serum lipid analyses, and to Drs. Richard Deckelbaum and Dewitt Goodman for their helpful reading of the manuscript.
REFERENCES 1. Wynder EL, Berenson GS, Strong WB, Williams C, Eds. Coronary artery disease prevention: Cholesterol, a pediatric perspective. Prev Med 1989; l&323-409. 2. Strong WB, Ed. Atherosclerosis: Its pediatric aspects. New York: Grune L Stratton, 1978. 3. National Institutes of Health Consensus Development Conference statement. Lowering blood cholesterol to prevent heart disease. JAMA 1985; 253:208&2086. 4. Diagnosis and treatment of primary hyperlipidemia in childhood. A joint statement for physicians
DIET
AND
LIPIDS
247
by the Committee on Atherosclerosis and Hypertension in Childhood of the Council of Cardiovascular Disease in the Young and the Nutrition Committee, American Heart Association. Circulation 1986; 74:1181A-1188A. 5. American Academy of Pediatrics Committee on Nutrition. Prudent life-style for children: Dietary fat and cholesterol. Pediatrics 1986; 78:521-525. 6. The Expert Panel. Report of the National Cholesterol Education Program Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults. Arch Znt Med 1988; l&3669. 7. Sacks FM, Castelli WP, Donner A, Kass EH. Plasma lipids and lipoproteins in vegetarians and controls. N Engl J Med 1975; 292:1148-l 151. 8. Keys A. Seven Countries: A Multivariate Analysis of Death and Coronary Heart Disease. Cambridge, MA: Harvard Univ. Press, 1980. 9. Connor WE, Hodges RE, Bleiler RE. Effect of dietary cholesterol upon serum lipids in man. JLab Clin Med 1961; 57:331-342. 10. Antonis A, Bersohn I. The influence of diet on serum lipids in South African White and Bantu prisoners. Am J Clin Nutr 1%2; 10:484X?). 11. Hegsted DM, Gandy RB, Nyers ML, Stare FJ. Quantitative effects of dietary fat on serum cholesterol in man. Am J Clin Nutr 1%5; 17:281-295. 12. Mattson FH, Erikson BA, Kligman AM. Effects of dietary cholesterol on serum cholesterol in man. Am J Clin Nutr 1972; 25:589-594. 13. Nestel PJ, Havenstein N, Whyte HM, Scott TJ, Cook LJ. Lowering of plasma cholesterol and enhanced sterol excretion with the consumption of polyunsaturated ruminant fats. N Engl J Med 1973; 228:379-382. 14. Hegsted DM. Serum-cholesterol response to dietary cholesterol: A re-evaluation. Am J Clin Nutr 1986; 443299-305. 15. Mensink R, Katan MB. Effect of a diet enriched with monounsaturated or polyunsaturated fatty acids on levels of low-density and high-density lipoprotein cholesterol in healthy women and men. N Engl J Med 1989; 321:4364+1. 16. Mensink R, Katan MB. Effect of dietary tram fatty acids on high-density and low-density lipoprotein cholesterol levels in healthy subjects. N Engl J Med 1990; 323:43w5. 17. The Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. I. Reduction in incidence of coronary heart disease. JAMA 1984; 251:351-364. 18. The Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. II. The relationship of reduction in incidence of coronary heart disease to cholesterol lowering. JAMA 1984; 251:365-374. 19. Amtzenius AC, Kromhout D, Barth J, et al. Diet, lipoproteins, and the progression of coronary atherosclerosis. N Engl J Med 1985; 312:805-811. 20. Ginsberg HN, Barr SL, Gilbert A, et al. Reduction of plasma cholesterol levels in normal men on an American Heart Association Step 1 diet or a Step 1 diet with added monounsaturated fat. N Engl J Med 1990; 32574-579. 21. Farris RP, Frank GC, Webber LS, Srinivasan SR, Berenson GS. Influence of milk source on serum lipids and lipoproteins during the first year of life, Bogalusa Heart Study. Am J Clin Nutr 1982; 35:42-49. 22. Berenson GS, Blonde CV, Fanis RP, et al. Cardiovascular disease risk factor variables during the fast year of life. Am J Dis Child 1979; 133:104~1057. 23. Nicklas TA, Farris RP, Smoak CG, et al. Dietary factors relate to cardiovascular risk factors in early life. Bogalusa Heart Study. Arteriosclerosis 1988; 8:193-199. 24. Frank GC, Berenson GS, Webber LS. Dietary studies and the relationship of diet to cardiovascular disease risk factor variables in lO-year-old children-The Bogalusa Heart Study. Am J Clin Nutr 1978; 31:328-340. 25. Glueck CJ, Waldman G, McClish DK, et al. Relationships of nutrient intake to lipids and lipoproteins in 1234 white children: The Lipid Research Clinics Prevalence Study. Arteriosclerosis 1982; 2~523-536. 26. Morrison JA, Larsen R, Glatfelter L, et al. Interrelationships between nutrient intake and plasma
248
SHEA ET AL.
lipids and lipoproteins in schoolchildren aged 6 to 19: The Princeton School District Study. Pediatrics 1980; 65:727-734. 27. Crawford PB, Clark MJ, Pearson DL, Huenemann RL. Serum cholesterol of 6-year-olds in relation to environmental factors. J Am Diet Assoc 1981; 7841-46. 28. Vartiainen E, Puska P, Pietinen P, Nissinen A, Leino U, Uusitalo U. Effects of dietary fat modifications on serum lipids and blood pressure in children. Actu Paediatr &and 1986; 75:396-401. 29. McGandy RB, Hall B, Ford C, Stare FJ. Dietary regulation of blood cholesterol in adolescent males: A pilot study. Am J Clin Nutr 1972; 25:61-66. 30. Ford CH, McGandy RB, Stare FJ. An institutional approach to regulation of blood cholesterol in adolescent males. Prev Med 1972; 1:426-445. 31. Willett WC, Stampfer MJ. Total energy intake: Implications for epidemiologic analyses. Am J Epidemiol 1986; 125:17-27. 32. Wiiett W. Nutritional Epidemiology. New York: Oxford Univ. Press, 1989. 33. Shea S, Basch CE, Zybert P, et al. Screening using National Cholesterol Education Program guidelines in a population of urban Hispanic women. Prev Med 1989; l&824-832. 34. Zuckerman AE. Georgetown University Nutrient Analysis Software. Washington, DC: Alan E. Zuckerman, 1987. 35. Adam CF, Richardson M. Nutritional value of food. Home and Garden Bulletin No. 72. Science and Education Administration. Washington, DC: U.S. Govt. Printing Office, 1981. 36. de Reguero LC, de Santiago SMR. Tabla de composition de alimentos de uso coniente en Puerto Rico. Puerto Rico: Universidad de Puerto Rico, 1978. 37. Allain CC, Poon LS, Chan CSG, Richmond W, Fu PC. Enzymatic determination of total serum cholesterol. Clin Chem 1974; 2k470-475. 38. Bucolo G, David H. Quantitative determination of serum triglycerides by the use of enzymes. Clin Chem 1973; 19476-482. 39. Miller R. Simultaneous Statistical Inference. New York: McGraw-Hill, 1966: 37-48. 40. Norusis MJ/SPSS, Inc. SPSSPC+ V.2.0 for the IBM PC/XT/AT and PS12.Base Manual. Chicago, IL: SPSS, Inc., 1988. 41. Kim SYS, Gergen PJ, Malloy M, Dresser C, Carroll M. Dietary patterns of U.S. children: Implications for disease prevention. Prev Med 1990; 19432-444. 42. U.S. Department of Health and Human Services, Public Health Service, National Institutes of health. The Lipid Research Clinics Population Studies Data Book. Vol. 1. The Prevalence Study. NIH Publication No. 80-1527. 1980. Tables 20, 21, 32, and 33. 43. Klesges RC, Brown G, Frank GC. Validation of the 24-hour recall in preschool children. J Am Diet Assoc 1987; 821383-138.5. 44. Eck LH, Klesges RC, Hanson CL. Recall of intake from one meal: Are parents accurate? J Am Diet Assoc 1989; 89:784-789. 45. Basch CE, Shea S, Arliss R, et al. Validation of mothers’ report of dietary intake by 4 to 7 year-old children. Am J Public Health 1990; 81:1314-1317. 46. Sorenson AW, Calkins BM, Connolly MA, Diamond E. Comparison of nutrient intake by four dietary intake instruments. J Nutr Educ 1985; 17:92-99. 47. Balogh M, Kahn H, Medalie JH. Random repeat 24-hour dietary recalls. Am J Clin Nutr 1971; 243304-310. 48. Willett WC, Stampfer MJ, Underwood BA, Speizer FE, Rosner B, Hennekens CH. Validation of a dietary questionnaire with plasma carotenoid and alpha-tocopherol levels. Am J Clin Nutr 1983; 38631439. 49. Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985; 122:51-65. 50. Willett WC, Reynolds RD, Cottrell-Hoehner S, Sampson L, Browne ML. Validation of a semiquantitative food frequency questionnaire: Comparison with a l-year diet record. J Am Diet Assoc 1987; 87~43-47. 51. U.S. Department of Commerce, Bureau of the Census. The Hispanic population in the United States: March 1988(Advance Report). Current Population Reports. Population Characteristics. Series P-20. No. 431.
DIET
AND
LIPIDS
249
52. Markides KS, Coreil J. The health of Hispanics in the Southwestern United States: An epidemiologic paradox. Public Health Rep 1986; 101:253-265. 53. Shea S, Basch CE, Irigoyen M, Zybert P, Rips JL, Content0 IR, Gutin B. Failure of family history to predict high blood cholesterol in a population of Hispanic preschool children. Prev Med 1990; ml43-455.
54. Weidman WH, Elveback LR, Nelson RA, Hodgson PA, Ellefson RD. Nutrient intake and serum cholesterol level in normal children 6 to 16 years of age. Pediatrics 1978; 61:354-359. 55. Prewitt TE, Haynes SG,Graves K, Haines PS, Tyroler HA. Nutrient intake, lipids, and lipoprotein cholesterols in black and while children: The Lipid Research Clinics Prevalence Study. Prev Med 1988; 12247-262. 56. Walter HJ, Hofman A, Vaughan RD, Wynder EL. Modification of risk factors for coronary heart disease: Five-year results of a school-based intervention trial. N Engl J Med 1988; 318:10931100. Received April 23, 1990 Revised September 18, 1990 Accepted September 18, 1990
MEDICINE
20, 237-249 (l!%‘l)
Relationships of Dietary Fat Consumption to Serum Total and Low-Density Lipoprotein Cholesterol in Hispanic Preschool Children’ STEVEN SHEA, M.D.,*+* CHARLES E. BASCH, PH.D.,S MATILDE IRIGOYEN, M.D.,9 PATRICIA ZYBERT, PH.D.,$ JILL L. RIPS, M.A., M.PHIL. ,$ ISOBEL CONTENTO, PH.D. ,$ AND BERNARD GUTIN, PH.D.S Departments of *Medicine and BPediatrics, and fSchoo1 of Public Health, Columbia University Health Sciences Division, and SCenter for Health Promotion, Teachers College, Columbia University, New York, New York 10027 Background. Studies of the relationship between dietary fat intake and serum lipids in young children have yielded inconclusive results. We studied this relationship in 108 Hispanic children ages 4-5 years. Methods. Four 24-hr recalls approximately 3 months apart and two Willett semiquantitative food frequency questionnaires approximately 6 months apart were obtained by interviewing the children’s mothers. Diet measures were averaged for the multiple administrations of each of these instruments. Results. Based on the 24-hr recalls, children in the highest tertile of total fat consumption (36.2% of total calories) compared with the lowest tertile (30.2% of total calories) had mean total serum cholesterol of 4.32 mmolihter (167 mg/dl) vs 3.91 mmol/Iiter (151 mg/dl) (test for linear trend across tertiles, P < 0.05) and mean low-density lipoprotein cholesterol of 2.74 mmol/Iiter (106 mg/dl) vs 2.29 mmollliter (89 mg/dl) (test for linear trend, P < 0.01). Children in the highest tertile of saturated fat consumption (14.6% of total calories) compared with the lowest tertile (11.2% of total calories) had mean total serum cholesterol of 4.39 mmol/Iiter (170 mg/dl) vs 3.97 mmol&ter (154 mg/dl) (test for linear trend, P < 0.05) and mean lowdensity lipoprotein cholesterol of 2.80 mmolihter (108 mg/dl) vs 2.35 mmol/liter (91 mg/dl) (test for linear trend, P < 0.01). These relationships remained significant when calorieadjusted nutrient intakes were examined and after adjustment in multiple linear regression models for age, sex, and body mass index, with the exception of the association of calorieadjusted total fat with total serum cholesterol level (P = 0.07). Similar results were obtained using the Willett questionnaires. Conclusions. These findings indicate that dietary fat, particularly saturated fat consumption, is an important correlate of blood lipid levels in preschool children. These are also the first reported data indicating that the Willett questionnaire, as a method for measuring the atherogenic components of diet, has criterion-related validity in young children. 0 1991 Academic
Press, Inc.
INTRODUCTION
Increasing recognition has been gained over the last decade for the importance of maintaining moderate serum cholesterol (TC) levels in childhood (1, 2). The ’ Supported by Grants HL35189 and HL21006 from the National Heart, Lung, and Blood Institute. ’ To whom reprint requests should be addressed at Columbia University, Atchley Pavilion 1310, 161 Fort Washington Avenue, New York, NY 10032. 237 0091-7435191$3.00 Copyrisbt 0 1991 by Academic PIWS, Inc. Au rights of reproduction in any form reserved.
SHEA ET AL.
1984 National Institutes of Health (NIH) Consensus Conference, the American Heart Association, and the American Academy of Pediatrics have all issued statements recommending strategies for the detection and management of children over the age of 2 years with high blood TC levels (3-S). The NIH Consensus Conference also recommendedthat U.S. children over the age of 2 years should adopt a diet with total fat intake of less than 30% of total calories, saturated fat intake of less than 10% of total calories, and daily dietary cholesterol intake of 25&300 mg or less (3). For children and adults with hyperlipidemia, dietary treatment is the fust step (4, 6). The goals of dietary intervention among pediatric populations are twofold: to establish healthy eating habits that will persist into adulthood and to prevent or treat hyperlipidemia (3-5). Thus a considerable body of public health policy and clinical practice rests on the link between dietary fat intake and serum lipid levels in childhood. For adults, a great deal of data exist from observational studies in free-living populations (7,8), from dietary manipulation in closely controlled metabolic ward studies (9-16) and from clinical trials (17-20), indicating that dietary fat intake, particularly saturated fat, influences total and low-density lipoprotein cholesterol (LDL-C) levels, but the data addressing this issue in young children are less extensive. Fart-is er al. studied 447 infants in Bogalusa and found a relationship between consumption of cow’s milk, which is high in fat, compared with formula, which is lower in fat, and serum TC at 6 months (21). Berenson et al. studied 440 children from this same cohort at 1 year of age and found correlations between consumption per 1,000 kcal of total fat, animal fat, and dietary cholesterol and serum TC level (22). Nicklas et al. analyzed relationships between nutrient intakes and serum lipid levels in 50 children at ages4 and 7 years. Dietary data were obtained using a 24-hr recall administered to the child’s main caregiver. At age 4, significant relationships were found between dietary cholesterol analyzed as milligrams per 1,000kcal (but not as mg/day or mg/kg of body weight) and both TC and LDL-C, and between saturated fat analyzed as grams per kilogram of body weight (but not as g/day or g/l000 kcal) and TC and LDL-C. At age 7, only the relationship of dietary cholesterol analyzed as milligrams per 1000kcal with serum LDL-C was significant (23). Frank et al. studied 185 children in the Bogalusa cohort at age 10 and found that children in the highest tertile of serum TC level differed from those in the lower tertiles with respect to dietary intake of total fat, animal fat, saturated fat, and unsaturated fat (24). Glueck et al. reported significant relationships between diet and serum lipid levels in 1,234white children ages 6-19 years, but the relationships were weak and limited to age/sexsubgroups. Diet was estimated using a 24-hr recall. Specifically, Glueck et al. found that among boys ages 6-12 years, the dietary polyunsaturated/saturated (P/S) fat ratio was inversely associated with plasma TC, and LDL-C and dietary cholesterol intake was positively associatedwith plasma high-density lipoprotein cholesterol (HDLC) (25). Morrison et al. reported data from 1,669children ages6-19 years in which plasma LDL-C was inversely correlated with the dietary P/S fat ratio (26). Crawford et al., in a study of 84 children at age 6 years, found that the percentage calories from fat and the amount of dietary cholesterol were correlated with serum TC level (27). Vartiainen et al. put 36 children ages 8-18 years on a cholesterol-
DIET
AND
LIPIDS
239
lowering diet in which total fat consumption was reduced from 35 to 24% of total calories and found a mean reduction of 15% in serum TC (28). Ford and her co-workers, in a study of male students 12-18 years of age in a boarding school where the cafeteria diet could be controlled, found that reduction of dietary total and saturated fat and dietary cholesterol produced a 15% mean reduction in serum TC (29, 30). None of these studies used semiquantitative food frequency questionnaires to estimate habitual diet and none adjusted nutrient intake for total calorie consumption. The calculation of nutrient density by expressing nutrient consumption per 1000kcal, while useful for dietary prescription and for comparing diets, does not control for confounding by total calorie consumption. Regression methods, as described by Willett, are necessary for this purpose (31, 32). Only Nicklas et al. (23) and Crawford et al. (27) studied preschool children, and only the studies of Nicklas et al. (23), Frank et al. (24), Vartiainen et al. (28), and Ford et al. (29, 30) showed relationships between saturated fat intake and lipid levels. Finally, none of these studies included significant numbers of Hispanic children or attempted to measurethe diet eaten among Hispanic populations living in the United States. In this article, we report a study of 108 Hispanic preschool children in which the relationships between dietary fat consumption and serum lipid levels were investigated. METHODS Subjects and Setting
Children were recruited between May, 1986,and May, 1987,for the Columbia Study of Childhood Activity and Nutrition, a longitudinal study of healthy preschool children and their parents. Children were eligible for the study if they were in good health and without dietary prescriptions or restrictions, lived in the primary referral area of The Presbyterian Hospital, and were between 36 and 48 months of age at time of entry. Only one child from each family was enrolled; no family had more than one age-eligible child. Families were excluded if the child was not living with the parent or the family did not appear likely to meet the requirements of six visits per year over 3 years to the study field office. Study subjects were drawn from a primarily Hispanic, urban, low-income neighborhood in northern Manhattan, New York City. Additional details of the recruitment procedure have been published elsewhere (33). Of 240 Hispanic families recruited into the study, 24 were lost to follow-up before the children’s serum lipid measurements were made, 106 families declined cholesterol screening for their children, and 2 children did not complete all four 24-hr diet recalls during the year in which the serum lipid sampleswere drawn. Thus there were 108Hispanic children with complete data for the purpose of the present analysis. There were no differences in age, sex, or body mass index (BMI) between these 108children and the 106 children whose parents refused permission for phlebotomy of their children. There were 57 boys and 51 girls. In 2 children, both girls, triglyceride levels were not obtained, so that LDL-C was calculated in 106 children. The 108 study children were all age 4 and 5 years at the time of lipid determination (mean 60.2
240
SHEA ET AL.
months, SD = 5.49), with the exception of one who was 3 years and 11 months old and one who had just turned 6 years old. Measures
Children’s diet was estimated using a 24-hr recall administered on four occasions in the first and third years of the study, approximately 3 months apart to account for seasonal variation, and using the Willett semiquantitative food frequency questionnaire administered on two occasions each year, approximately 6 months apart. Both diet measureswere obtained from the children’s mothers by trained bilingual interviewers using standard questionnaires with probes conceming food preparation methods and brand namesfor foods, three-dimensional food models, food packages, glasses, cups, spoons, and other utensils. For the 24-hr recalls, foods and portion sizes were converted to nutrient intakes using a dietary analysis program (34) based on published nutrient data (35). For the Willett questionnaires, nutrient analysis was done using a computer program supplied by Willett (W. C. Willett, personal communication). Additional foods commonly eaten by the study population but not included in the dietary analysis programs for the 24-hr recalls or the Willett questionnaire were added from another source of nutrient values (36). For the 24-hr recalls, the averageof the four administrations closest in time to drawing the blood samples for lipid analysis was used. The intraclass correlation coefficients for these four measures were 0.28 for total calories, 0.23 for total fat consumption (g/day), 0.29 for saturated fat consumption (g/day), and 0.19 for cholesterol consumption (mg/day). For the Willett questionnaires, the averageof the two administrations closest in time to drawing the blood samples for lipid analysis was used. The intraclass correlation coefficients for these measureswere 0.39 for total calories, 0.38 for total fat consumption (g/day), 0.30 for saturated fat consumption (g/day), and 0.19 for cholesterol consumption (mg/day). Following Willett’s recommendations (31, 32) the natural logarithm transformation of dietary variables was used to improve normality, and regression methods were used to adjust fat and cholesterol consumption for differences in total caloric intake. Specifically, regressions were performed with the natural logarithm of the nutrient of interest as the dependent variable and the natural logarithm of calorie consumption as the independent variable, and the residuals from these models were added to the expected nutrient intake for the mean of the natural logarithm of the caloric intake (31, 32). These methods were used for data derived from the 24-hr recalls as well as the Willett food frequency questionnaires. Untransformed levels of nutrient consumption are presented in Tables 1 and 2, although transformed variables were used in statistical analyses. Fasting venous blood sampleswere analyzed for serum TC and HDL-C (37) and triglyceride levels (38) in a Columbia University research laboratory calibrated to Centers for Disease Control standards. LDL-C was calculated as TC minus the sum of HDL-C and one-fifth the triglyceride level (6). This formula is valid when triglycerides are less than 4.52 mmohliter (400 mg/dl), which was the case for all subjects. Weight was measuredusing a calibrated balance scale. Height was measured using a stadiometer. BMI was calculated as the weight (kilograms) divided by the height (meters) squared.
DIET AND LIPIDS
241
The study population was divided into tertiles based on both raw and calorieadjusted consumption of total fat, saturated fat, and cholesterol. Tests for linear trend were performed to assess the relationships between nutrient intakes and lipid levels. Differences in mean serum lipid levels among tertiles were tested using Tukey’s studentized range test (39), which adjusts for multiple comparisons. To confirm these bivariate associations, separate multiple linear regression models were constructed with TC and LDL-C as the dependent variables. Age in months, sex, and BMI and raw or calorie-adjusted dietary total fat, dietary total saturated fat, or dietary cholesterol intake were entered into each model. Unstandardized regression coefficients are reported in Table 4. Data were analyzed using SPSS-PC+ (40). The study was approved by the Human Subjects Committee of Teachers College and the Institutional Review Board of the Columbia University Health Sciences Division. RESULTS
Mean serum TC level for the entire sample was 4.10 mmol/liter (158.5 mg/dl) and mean LDL-C level was 2.51 mmolihter (97.2 mg/dl). Serum TC and LDL-C were highly correlated (N = 106; r = 0.95; P < 0.001). Thirty-seven children (34.3%) had serum TC a4.40 mmol/liter (170 mg/dl). Among boys, mean serum TC level was 4.15 mmolfliter (160.5 mg/dl) and mean LDL-C was 2.58 mmol/liter (99.7 mg/dl). Among girls, mean serum TC level was 4.04 mmol/liter (156.2 mg/dl) and mean LDL-C was 2.44 mmol/liter (94.3 mg/dl). Mean consumption of total fat for the entire sample was 33.0% of total calories based on the 24-hr recalls and 33.2% based on the Willett questionnaires; mean consumption of saturated fat was 13.2%of total calories based on the 24-hr recalls and 13.5% based on the Willett questionnaires; and mean consumption of cholesterol was 193.9 mg/lOOOkcal based on the 24-hr recalls and 165.6 mg/lOOOkcal based on the Willett questionnaires. Total calorie intake estimated from the Willett questionnaires was higher than that from the 24-hr recalls. Dietary total fat and saturated fat were highly correlated and both were highly correlated with total calorie consumption (Table 1). Mean consumption of total fat in the lowest tertile was 40.5 g/day (30.2% of total calories) based on data from the 24-hr recalls, compared with 79.3 g/day (36.2% of total calories) in the highest tertile. Differences between the lowest and highest tertile were also approximately twofold for absolute consumption of saturated fat and cholesterol. Calorie-adjustment reduced the difference between lowest and highest tertile substantially (Table 2). The differences in lipid levels between the highest and the lowest tertiles of total fat consumption (as a percentage of the mean of the lowest tertile) were 10.5% for TC and 19.7% for LDL-C. For saturated fat consumption, these differences were 10.6% for TC and 19.2% for LDL-C . Based on estimates derived from the 24-hr recalls, tests for linear trend for serum TC level (Table 2) across tertiles of nutrient consumption were significant for total fat intake (P < 0.05), saturated fat intake (P < 0.05), and calorie adjusted saturated fat intake (P < 0.01). Tests for linear trend for LDL-C level were significant across tertiles of nutrient consumption for total fat intake (P < O.Ol),
242
SHEA ET AL. TABLE 1 CORRELATIONSAMONG NUTWENT CONSUMPTIONLEVELS (iV = 108 SUBJECTS)
CalorieTotal fat
Calorie-
Calorie-
Saturated adjusted adjusted adjusted fat Cholesterol total fat saturated fat cholesterol
Nutrient consumptions” estimated using Total calories 0.93** 0.84** 0.55** 0.91** 0.54** Total fat 0.53** Saturated fat Cholesterol Calorie-adjusted total fat Calorie-adjusted saturated fat
24&r recalls
0.00 0.38** 0.35** 0.07
0.00 0.24* 0.55** 0.13 0.64**
0.00
0.03 0.08 0.83** 0.09 0.15
Nutrient consumptions” estimated using the Willett semiquantitative food frequency questionnaire 0.59** 0.00 0.92** 0.85** 0.00 0.00 Total calories 0.63** 0.38** 0.93** 0.27* 0.10 Total fat 0.72** Saturated fat 0.38** 0.53** 0.26* 0.22 0.40** 0.80** Cholesterol 0.71** 0.27* Calorie-adjusted total fat 0.50** Calorie-adjusted saturated fat QNutrients were transformed using the natural logarithm of the nutrient (grams/day for total fat and saturated fat and mg/day for cholesterol). * P (one-tailed) < 0.01. ** P (one-tailed) < 0.001.
saturated fat intake (P < O.Ol), calorie-adjusted total fat intake (P < 0.05), and calorie-adjusted saturated fat intake (P < 0.01). Simiiar results were obtained from analysesbased on data from the Willett questionnaires (Table 3). Multivariate analysescontrolling for sex, age, and BMI (Table 4) confirmed these relationships with the exception of the association of calorie-adjusted total fat with serum TC level (P = 0.07). Relationships between dietary variables and HDL-C and triglyceride levels were examined using these same methods, and none of the findings was statistically significant. In the multivariate analyses, relationships between dietary cholesterol and serum lipid levels were not statistically significant (Table 4). Relationships between dietary polyunsaturated fat intake and lipid levels were also examined; these relationships were not statistically significant. DISCUSSION
Kimm et al. analyzed dietary patterns in U.S. children using data from the second National Health and Nutrition Examination Survey (NHANES II), which was conducted between 1976 and 1980. These investigators found that among children 2-10 years of age mean total fat consumption was 36% of total calories, mean saturated fat consumption was 13% of total calories, and mean cholesterol consumption was 151 mg/lOOOkcal(41). The findings in our sample of Hispanic children were in close agreement with these values for saturated fat and cholesterol. Total fat consumption in our samplewas slightly lower. These comparisons should be interpreted with caution, since the NHANES II data were obtained
DIET
AND
243
LIPIDS
TABLE 2 MEANS (STANDARD DEVIATIONS) OF SERUM LIPID LEVELS IN HISPANIC PRESCHOOL CHILDREN BY TERTILE OF DIETARY CONSUMPTION PER DAY ESTIMATED USING THE MEAN OF FOUR 24-~~ RECALLS
COLLECTED
DURING
1 YEAR Serum lipids
Dietary
per day Lowest Middle Highest Saturated fat (g) Lowest Middle Highest Cholesterol (mg) Lowest Middle Highest Calorie-adjusted Lowest total fat f,g) Middle Highest Lowest Calorie-adjusted saturated fat (g) Middle Highest Calorie-adjusted Lowest cholesterol (mg) Middle Highest
Total fat (g)
tertile tertile tertile tertile tertile tertile tertile tertile tertile tertile tertite tertile tertile tertile tertile tertile tertile tertile
40.5 54.0 79.3 15.6 22.2 31.4 179.3 281.1 438.9 48.1 55.2 62.8 18.0 22.0 26.5 31.4 47.0 66.6
Total cholesterol (N = 108)
consumption
(7.7) (3.9) (16.3) (3.1) (1.8) (6.3) (36.7) (32.4) (78.2) (3.0) (1.7) (4.0) (2.2) (1.0) (2.3) (5.2) (3.6) (12.8)
Nutrient density 30.2“ 32.7 36.2 11.2” 13.8 14.6 140.2b 187.7 253.6
mmoVliter 3.91 4.07 4.32 3.97 3.93 4.39 3.92 4.16 4.21 3.95 4.04 4.30 3.87 4.06 4.37 3.97 4.21 4.11
(0.74) (0.83) (0.66) (0.76) (0.80) (0.63) (0.79) (0.81) (0.64) (0.73) (0.80) (0.71) (0.62) (0.83) (0.74) (0.71) (0.76) (0.80)
ddl 151.0 157.4 166.9 153.6 151.9 169.9 151.5 160.9 162.9 152.8 156.3 166.2 149.5 156.9 169.0 153.5 162.8 159.1
(28.6) (32.0) (2X4)+’ (29.5)* (30.9)* (24.3)+ (30.6) (31.5) (24.9) (28.2) (31.1) (27.5) (24.1)* (31.9) (28.7)‘* (27.5) (29.5) (30.9)
LDL cholesterol (N = 106) mmoliliter 2.29 2.50 2.74 2.35 2.40 2.80 2.32 2.57 2.65 2.36 2.47 2.71 2.29 2.46 2.81 2.40 2.60 2.54
(0.65) (0.70) (0.62) (0.69) (0.65) (0.61) (0.68) (0.74) (0.57) (0.64) (0.72) (0.64) (0.53) (0.77) (0.61) (0.65) (0.67) (0.71)
mW 88.7 %.6 106.1 90.8 92.8 108.3 89.8 99.2 102.5 91.4 95.5 104.7 88.5 95.0 108.8 92.7 100.6 98.4
(25.1); (27.0) (24. I)#*” (26.6)* (25.2)* (23.5)“* (26.4) (28.8) (21.9)* (24.8) (27.7) (24.8)” (20.4)* (29.9) (23.6)“““ (25.1) (25.7) (27.6)
* Percentage total calories. b mg per 1000 kca.I. * P (two-tailed) c 0.05 vs highest tertile. # P G 0.05 for linear trend test across tertiles. ## P =Z 0.01 for linear trend test across tertiles.
from a representative national sample and the data were collected approximately 10 years earlier than the data in our study. In the Lipid Research Clinics Prevalence Study, at Visit 2, mean plasma cholesterol was 155.2mg/dl for white males ages 5-9 years and 164.2mg/dl for white females of the sameage. Mean plasma LDL-C was 92.5 mg/dl for white males and 100.4mg/dl for white females in this age group (42). These values are comparable to those obtained in our sample of Hispanic children ages 4 and 5 years. Caution is also indicated in interpreting this comparison, since the Lipid Research Clinics obtained a national sample, the children were somewhat older, and the data were obtained in the 1970s.In the study by Nicklas et al. (23) of children age 4 years in the Bogalusa Heart Study, children in the lowest tertile of saturated fat consumption had a mean serum TC level of 155.4mg/dl, compared with 153.6mg/dl in our study, and children in the highest tertile had a mean serum TC level of 169.7 mg/dl, compared with 169.9mg/dl in our study. Thus, the diet and the serum lipid levels found in our study appear generally comparable to those reported in other U.S. populations of similar age. In studies of the relationship between diet and serum lipids, the valid and reliable measurementof diet is essential. Three studies have reported that parental recall correlates well with direct observation of dietary consumption in preschool children (43-45). These studies pertain to short-term recall methods rather
244
SHEA ET AL.
TABLE 3 MEANT (STANDARD DEVIATIONS) OF SERUM LIPID LEVELS IN HISPANIC F’RESCH~~L CHILDREN BY TERTILE OF DIETARY CONSUMPTION PER DAY ESTIMATED USING THE WILLETT SEMIQUANTITATIVE FOOD FREQUENCY QUESTIONNAIRE Serum lipids Dietary consumption Per day
Lowest tertile Middle textile Highest tertile Saturated fat (g) Lowest tertile Middle tertile Highest tertile Cholesterol (mg) Lowest tertile Middle tertile Highest tertile Calorie-adjusted Lowest tertile total fat (9) Middle tertile Highest tertile Calorie-adjusted Lowest tertile saturated fat (g) Middle tertile Highest tertile Calorie-adjusted Lowest tertile cholesterol (mg) Middle tertiie Highest tertile
Total fat (g)
Nutrient density
70.1 (9.9) 97.8(7.4) 137.7 (31.0) 29.1 (4.0) 40.0 i2.5j 53.7 (8.7) 318.7 (56.1) 435.4 (23.2) 578.9 (75.8) 85.8 (5.3) 97.2 (3.1)
110.2 (8.6) 33.6 39.8 46.0 349.4 432.7 531.6
(3.0) (1.8) (2.7) (62.1) (15.4) (62.1)
30.7" 32.3 36.6 12.5" 13.4 14.6 145.56 167.5 183.9
Total cholesterol (N = 108) mmol/liter 3.98(0.71) 3.91 (0.78) 4.4110.70) 3.93 (0.77) 3.93 (0.70) 4.44(0.70) 3.93 (0.80) 4.11 (0.67) 4.26(0.78) 3.85 (0.73) 4.24 (0.77) 4.21 (0.73) 3.99 (0.76) 3.87 (0.73) 4.43 (0.68) 3.95 (0.76) 4.02 (0.69) 4.32(0.78)
mg/dl 153.8 (27.6)*
151.0 (30.1)’ 170.6 (27.0)* 151.9 (29.9)* 151.9 (26.9)* 171.6 (27.1)"“" 151.8 (31.1) 158.8 (25.8) 164.8 (30.0) 149.0 (28.3) 163.8 (29.6) 162.6(28.3)# 154.4 (29.3)* 149.7 (28.2)' 171.3 (26.4)" 152.8 (29.4)
155.5 (26.8) 167.1 (30.3)'"
LDL cholesterol (N = 106) mmoliliter 2.42(0.62) 2.34 (0.67) 2.79 (0.66) 2.38(0.69) 2.36(0.61) 2.80(0.66) 2.29 (0.67) 2.57 (0.62) 2.68(0.70) 2.24 (0.61) 2.64 (0.62) 2.66 (0.72) 2.38 (0.67) 2.32(0.62) 2.86 (0.62)
mg/dl 93.5 (24.1)* 90.5 (25.9)* 107.9(25.6)" 92.1 (26.5)' 91.3 (23.4)* 108.4 (25.5)"‘" 88.6(26.0)** 99.2(23.8) 103.7 (26.9)* 86.5 (23.7). 102.3 @LO)* 102.8 (27.9)*+ 91.8 (25.8)** 89.5 (24.1)**
2.31 (0.65)
110.5 (23.9)“+“ 89.3 (25.1)*
Z.SO(O.63) 2.73 (0.70)
96.7 (24.2) 105.6(27.0)"+
(1Percentage total calories. b mg per 1000kcal. * P (two-tailed) c 0.05 vs highest tertile. ** P (two-tailed) < 0.01 vs highest tertile. * P s 0.05 for linear trend test across tertiles. ** P < 0.01 for linear trend test across tertiles.
than semiquantitative food frequency questionnaire methods. Studies of adults have shown that food frequency questionnaires tend to overestimate the absolute amounts of nutrients and total calories consumed, compared with 24-hr recalls or food diaries (46). In our study, the estimates of total and saturated fat and cholesterol intake obtained using the Willett questionnaire were substantially higher than estimates obtained from the 24-hr recall (Tables 1 and 2). This was the case for both calorie-adjusted and unadjusted values. Thus the descriptive values for the children’s diet based on the 24-hr recalls are preferred. There was, however, very close agreement between the two methods in estimating nutrient densities, expressed as a percentage of total calories provided by total and saturated fat or as milligrams of cholesterol per 1000kcal. The associations between dietary fat intake and serum lipid levels were somewhat stronger and more consistent using estimates derived from the Willett questionnaires than from the 24-hr recalls. We believe the most likely explanation is related to the higher intraclass correlations for the WilIett questionnaires, compared with the 24-hr recalls, indicating greater internal consistency for the repeat measures.This is expected becausethe purpose of the Willett questionnaire is to measurehabitual diet, while the 24-hr recall measuresrecent diet, and diet varies
245
DIET AND LIPIDS TABLE 4
MULTIPLE LINEAR REGRESSION ANALYSES OF ASSOCIATIONS OF DIETARY VARIABLES” WITH SERUM LIPID LEVELS,~ ADJUSTED= FOR AGE (IN MONTHS), SEX, AND BODY MASS INDEX
Serum total cholesterol P
SE g
P value
Serum LDL cholesterol R2
P
Dietary variables estimated using 24hr recalls 0.23 0.006 0.63 0.08 0.65 Dietary total fat 0.74 0.21 0.001 0.12 0.76 Dietary saturated fat 0.18 0.12 0.29 0.03 0.30 Dietary cholesterol 1.10 0.61 0.07 0.04 1.10 Calorie-adjusted total fat 1.10 0.38 0.005 1.13 0.08 Calorie-adjusted saturated fat 0.71 0.08 0.22 0.01 0.09 Calorie-adjusted cholesterol
SE p
Pvalue
R2
0.20 0.18 0.16 0.54 0.34 0.20
0.002
0.11 0.16 0.05 0.05 0.11 0.02
Dietary variables estimated using the Willett semiquantitative food frequency questionnaire Dietary total fat Dietary saturated fat Dietary cholesterol Calorie-adjusted total fat Calorie-adjusted saturated fat Calorie-adjusted cholesterol
0.52 0.63 0.31 1.36 1.14 0.17
0.24 0.26 0.26 0.61 0.49 0.32
0.04 0.02 0.23 0.03 0.02 0.61
0.05 0.06 0.02 0.05 0.06 0.01
0.44 0.61 0.37 1.51 1.38 0.36
0.22 0.24 0.23 0.54 0.44 0.29
0.04 0.01 0.11 0.01 0.002 0.21
0.05 0.08 0.04 0.08 0.10 0.03
LINutrients were expressed as the natural logarithm of the nutrient (grams/day for total fat and saturated fat and mg/day for cholesterol). b Total and LDL cholesterol were expressed as mmolfliter. c Adjustment was done by multiple linear regression.
within individuals over time. Studies of adults have found that as many as seven 24-hr recalls may be necessary to control fully for intraindividual variation in diet (47), so that four 24-hr recalls administered during a l-year period, as in our study, may not be enough to obtain statistically precise estimates of fat intake. Thus, despite the overestimation of absolute amounts of nutrients consumed, the Willett questionnaire gave more precise estimates. We have emphasized the 24-hr recall data in this report because the Willett questionnaire has been validated for adults (4g-50) but not, to our knowledge, for children, and because the absolute intake of dietary fat and cholesterol as estimated from the 24-hr recalls is closer to what is expected for children of this age (41). The results of our study suggestthat the Willett questionnaire has criterionrelated validity as a method for measuring the atherogenic components of diet in young children and convergent validity with results obtained using the 24-hr recall. Further research is needed to define the validity of the Willett questionnaire as a diet measure in epidemiologic studies in young children. As pointed out by Willett (32), when two dietary variables are highly correlated with each other, such as total fat and saturated fat intake in our study (r = 0.91; Table I), statistical techniques do not establish which one is the true independent variable and which the covariate. The problem of multicolinearity also affects serum TC and LDL-C, which were very highly correlated (r = 0.95). Thus, associations of dietary variables with both blood lipid parameters should be interpreted as confirmatory rather than as two independent findings.
246
SHEA ET AL.
CONCLUSION
No previous studies of the relationship between diet and serum lipid levels in Hispanic children have been reported, despite the growing demographic and public health importance of this population. In 1988 there were approximately 19.4 million Hispanics in the United States (51). Hispanics constitute approximately 8.1% of the U.S. civilian noninstitutionalized population (51) and are the fastest growing minority group (52). The demographic composition of our sample indicates that it is of low socioeconomic status, recently immigrated, and largely Dominican (53). The dietary practices of recently immigrated Hispanics or Hispanics from the Dominican Republic may have unique characteristics. Further research is needed to verify our findings in population-based samples and in non-Dominican Hispanic subpopulations. The findings of our study are consistent with findings from most previous studies of dietary fat consumption and serum lipid levels in children, which found at least some significant relationships (21-30). Other studies, of older children or children of less restricted age range, found no associations between nutrient intake and lipid levels (54, 55) or reported less clear-cut relationships (24-26). Possible explanations for these negative observational studies include less variation in serum lipids or in diet, less precise measuresof diet, or a true lack of relationship in the populations studied. Intervention studies in older children have shown that dietary modification produces changesin blood lipid levels (28-30, 56). The multiple R* for the associationsbetween dietary fat and serum lipid levels were small, indicating the importance of measurement error in diet, due to intraindividual variation or imprecision in the method, measurementerror in the lipid levels, also related to intraindividual variation or methodological imprecision, the influence of nondietary factors, such as genetic predisposition, or combinations of these factors. Despite the modest size of the multiple R* statistics, the differences in mean serum TC between the highest and the lowest tertile of children’s total and saturated fat intake were 10.5 and 10.6%, and the differences for LDL-C were 19.7 and 19.2%. These differences are large enough to have clinical and public health as well as statistical significance. In summary, our study strengthens the existing data that support the role of diet in determining blood lipid levels in children, and it suggeststhat the Willett questionnaire, or other food frequency methods for measuring diet, may have utility in young children. ACKNOWLEDGMENTS We are grateful to Dr. Walter Willett for providing the nutrient analysis programs, to Dr. Henry Ginsberg for performing the serum lipid analyses, and to Drs. Richard Deckelbaum and Dewitt Goodman for their helpful reading of the manuscript.
REFERENCES 1. Wynder EL, Berenson GS, Strong WB, Williams C, Eds. Coronary artery disease prevention: Cholesterol, a pediatric perspective. Prev Med 1989; l&323-409. 2. Strong WB, Ed. Atherosclerosis: Its pediatric aspects. New York: Grune L Stratton, 1978. 3. National Institutes of Health Consensus Development Conference statement. Lowering blood cholesterol to prevent heart disease. JAMA 1985; 253:208&2086. 4. Diagnosis and treatment of primary hyperlipidemia in childhood. A joint statement for physicians
DIET
AND
LIPIDS
247
by the Committee on Atherosclerosis and Hypertension in Childhood of the Council of Cardiovascular Disease in the Young and the Nutrition Committee, American Heart Association. Circulation 1986; 74:1181A-1188A. 5. American Academy of Pediatrics Committee on Nutrition. Prudent life-style for children: Dietary fat and cholesterol. Pediatrics 1986; 78:521-525. 6. The Expert Panel. Report of the National Cholesterol Education Program Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults. Arch Znt Med 1988; l&3669. 7. Sacks FM, Castelli WP, Donner A, Kass EH. Plasma lipids and lipoproteins in vegetarians and controls. N Engl J Med 1975; 292:1148-l 151. 8. Keys A. Seven Countries: A Multivariate Analysis of Death and Coronary Heart Disease. Cambridge, MA: Harvard Univ. Press, 1980. 9. Connor WE, Hodges RE, Bleiler RE. Effect of dietary cholesterol upon serum lipids in man. JLab Clin Med 1961; 57:331-342. 10. Antonis A, Bersohn I. The influence of diet on serum lipids in South African White and Bantu prisoners. Am J Clin Nutr 1%2; 10:484X?). 11. Hegsted DM, Gandy RB, Nyers ML, Stare FJ. Quantitative effects of dietary fat on serum cholesterol in man. Am J Clin Nutr 1%5; 17:281-295. 12. Mattson FH, Erikson BA, Kligman AM. Effects of dietary cholesterol on serum cholesterol in man. Am J Clin Nutr 1972; 25:589-594. 13. Nestel PJ, Havenstein N, Whyte HM, Scott TJ, Cook LJ. Lowering of plasma cholesterol and enhanced sterol excretion with the consumption of polyunsaturated ruminant fats. N Engl J Med 1973; 228:379-382. 14. Hegsted DM. Serum-cholesterol response to dietary cholesterol: A re-evaluation. Am J Clin Nutr 1986; 443299-305. 15. Mensink R, Katan MB. Effect of a diet enriched with monounsaturated or polyunsaturated fatty acids on levels of low-density and high-density lipoprotein cholesterol in healthy women and men. N Engl J Med 1989; 321:4364+1. 16. Mensink R, Katan MB. Effect of dietary tram fatty acids on high-density and low-density lipoprotein cholesterol levels in healthy subjects. N Engl J Med 1990; 323:43w5. 17. The Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. I. Reduction in incidence of coronary heart disease. JAMA 1984; 251:351-364. 18. The Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. II. The relationship of reduction in incidence of coronary heart disease to cholesterol lowering. JAMA 1984; 251:365-374. 19. Amtzenius AC, Kromhout D, Barth J, et al. Diet, lipoproteins, and the progression of coronary atherosclerosis. N Engl J Med 1985; 312:805-811. 20. Ginsberg HN, Barr SL, Gilbert A, et al. Reduction of plasma cholesterol levels in normal men on an American Heart Association Step 1 diet or a Step 1 diet with added monounsaturated fat. N Engl J Med 1990; 32574-579. 21. Farris RP, Frank GC, Webber LS, Srinivasan SR, Berenson GS. Influence of milk source on serum lipids and lipoproteins during the first year of life, Bogalusa Heart Study. Am J Clin Nutr 1982; 35:42-49. 22. Berenson GS, Blonde CV, Fanis RP, et al. Cardiovascular disease risk factor variables during the fast year of life. Am J Dis Child 1979; 133:104~1057. 23. Nicklas TA, Farris RP, Smoak CG, et al. Dietary factors relate to cardiovascular risk factors in early life. Bogalusa Heart Study. Arteriosclerosis 1988; 8:193-199. 24. Frank GC, Berenson GS, Webber LS. Dietary studies and the relationship of diet to cardiovascular disease risk factor variables in lO-year-old children-The Bogalusa Heart Study. Am J Clin Nutr 1978; 31:328-340. 25. Glueck CJ, Waldman G, McClish DK, et al. Relationships of nutrient intake to lipids and lipoproteins in 1234 white children: The Lipid Research Clinics Prevalence Study. Arteriosclerosis 1982; 2~523-536. 26. Morrison JA, Larsen R, Glatfelter L, et al. Interrelationships between nutrient intake and plasma
248
SHEA ET AL.
lipids and lipoproteins in schoolchildren aged 6 to 19: The Princeton School District Study. Pediatrics 1980; 65:727-734. 27. Crawford PB, Clark MJ, Pearson DL, Huenemann RL. Serum cholesterol of 6-year-olds in relation to environmental factors. J Am Diet Assoc 1981; 7841-46. 28. Vartiainen E, Puska P, Pietinen P, Nissinen A, Leino U, Uusitalo U. Effects of dietary fat modifications on serum lipids and blood pressure in children. Actu Paediatr &and 1986; 75:396-401. 29. McGandy RB, Hall B, Ford C, Stare FJ. Dietary regulation of blood cholesterol in adolescent males: A pilot study. Am J Clin Nutr 1972; 25:61-66. 30. Ford CH, McGandy RB, Stare FJ. An institutional approach to regulation of blood cholesterol in adolescent males. Prev Med 1972; 1:426-445. 31. Willett WC, Stampfer MJ. Total energy intake: Implications for epidemiologic analyses. Am J Epidemiol 1986; 125:17-27. 32. Wiiett W. Nutritional Epidemiology. New York: Oxford Univ. Press, 1989. 33. Shea S, Basch CE, Zybert P, et al. Screening using National Cholesterol Education Program guidelines in a population of urban Hispanic women. Prev Med 1989; l&824-832. 34. Zuckerman AE. Georgetown University Nutrient Analysis Software. Washington, DC: Alan E. Zuckerman, 1987. 35. Adam CF, Richardson M. Nutritional value of food. Home and Garden Bulletin No. 72. Science and Education Administration. Washington, DC: U.S. Govt. Printing Office, 1981. 36. de Reguero LC, de Santiago SMR. Tabla de composition de alimentos de uso coniente en Puerto Rico. Puerto Rico: Universidad de Puerto Rico, 1978. 37. Allain CC, Poon LS, Chan CSG, Richmond W, Fu PC. Enzymatic determination of total serum cholesterol. Clin Chem 1974; 2k470-475. 38. Bucolo G, David H. Quantitative determination of serum triglycerides by the use of enzymes. Clin Chem 1973; 19476-482. 39. Miller R. Simultaneous Statistical Inference. New York: McGraw-Hill, 1966: 37-48. 40. Norusis MJ/SPSS, Inc. SPSSPC+ V.2.0 for the IBM PC/XT/AT and PS12.Base Manual. Chicago, IL: SPSS, Inc., 1988. 41. Kim SYS, Gergen PJ, Malloy M, Dresser C, Carroll M. Dietary patterns of U.S. children: Implications for disease prevention. Prev Med 1990; 19432-444. 42. U.S. Department of Health and Human Services, Public Health Service, National Institutes of health. The Lipid Research Clinics Population Studies Data Book. Vol. 1. The Prevalence Study. NIH Publication No. 80-1527. 1980. Tables 20, 21, 32, and 33. 43. Klesges RC, Brown G, Frank GC. Validation of the 24-hour recall in preschool children. J Am Diet Assoc 1987; 821383-138.5. 44. Eck LH, Klesges RC, Hanson CL. Recall of intake from one meal: Are parents accurate? J Am Diet Assoc 1989; 89:784-789. 45. Basch CE, Shea S, Arliss R, et al. Validation of mothers’ report of dietary intake by 4 to 7 year-old children. Am J Public Health 1990; 81:1314-1317. 46. Sorenson AW, Calkins BM, Connolly MA, Diamond E. Comparison of nutrient intake by four dietary intake instruments. J Nutr Educ 1985; 17:92-99. 47. Balogh M, Kahn H, Medalie JH. Random repeat 24-hour dietary recalls. Am J Clin Nutr 1971; 243304-310. 48. Willett WC, Stampfer MJ, Underwood BA, Speizer FE, Rosner B, Hennekens CH. Validation of a dietary questionnaire with plasma carotenoid and alpha-tocopherol levels. Am J Clin Nutr 1983; 38631439. 49. Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985; 122:51-65. 50. Willett WC, Reynolds RD, Cottrell-Hoehner S, Sampson L, Browne ML. Validation of a semiquantitative food frequency questionnaire: Comparison with a l-year diet record. J Am Diet Assoc 1987; 87~43-47. 51. U.S. Department of Commerce, Bureau of the Census. The Hispanic population in the United States: March 1988(Advance Report). Current Population Reports. Population Characteristics. Series P-20. No. 431.
DIET
AND
LIPIDS
249
52. Markides KS, Coreil J. The health of Hispanics in the Southwestern United States: An epidemiologic paradox. Public Health Rep 1986; 101:253-265. 53. Shea S, Basch CE, Irigoyen M, Zybert P, Rips JL, Content0 IR, Gutin B. Failure of family history to predict high blood cholesterol in a population of Hispanic preschool children. Prev Med 1990; ml43-455.
54. Weidman WH, Elveback LR, Nelson RA, Hodgson PA, Ellefson RD. Nutrient intake and serum cholesterol level in normal children 6 to 16 years of age. Pediatrics 1978; 61:354-359. 55. Prewitt TE, Haynes SG,Graves K, Haines PS, Tyroler HA. Nutrient intake, lipids, and lipoprotein cholesterols in black and while children: The Lipid Research Clinics Prevalence Study. Prev Med 1988; 12247-262. 56. Walter HJ, Hofman A, Vaughan RD, Wynder EL. Modification of risk factors for coronary heart disease: Five-year results of a school-based intervention trial. N Engl J Med 1988; 318:10931100. Received April 23, 1990 Revised September 18, 1990 Accepted September 18, 1990