点击显示 收起
1 From the University of Iowa Hospitals and Clinics, Department of Pediatrics, Iowa City; the National Heart, Lung, and Blood Institute, Division of Epidemiology and Clinical Applications, Bethesda, MD; the Northwestern University Medical School, Department of Preventive Medicine, Chicago; the Johns Hopkins Hospital, School of Medicine, Department of Pediatrics, Baltimore; the Maryland Medical Research Institute, Baltimore; the Kaiser Permanente Center for Health Research, Kaiser Foundation Hospitals, Portland, OR; the Johns Hopkins Hospital, Children's Medical and Surgical Center, Baltimore; the Children's Hospital of Alabama, Department of Gastrointestinal Nutrition, Birmingham; the University of Pittsburgh School of Medicine, Department of Clinical Epidemiology and Preventive Medicine, Pittsburgh; and the New Jersey Medical School, Preventive Cardiology Program, Newark.
2 Presented at the symposium Fat Intake During Childhood, held in Houston, June 8–9, 1998. 3 The data presented here were published previously in JAMA 1995;273: 1429–35 and Pediatrics 1997;100:51–9. 4 The Dietary Intervention Study in Children (DISC) was supported by cooperative agreements U01-HL-37947, U01-HL-37948, U01-HL-37954, U01-HL-37962, U01-HL-37966, U01-HL-37975, and U01-HL-38110 from the National Heart, Lung, and Blood Institute, the National Institutes of Health, Bethesda, MD. 5 Address reprint requests to DISC Coordinating Center, Maryland Medical Research Institute, 600 Wyndhurst Avenue, Baltimore, MD 21210. 6 Address correspondence to RM Lauer, Pediatric Cardiology, University of Iowa Hospitals and Clinics-2JPP, 200 Hawkins Drive, Iowa City, IA 52242. E-mail: ronald-lauer{at}uiowa.edu.
ABSTRACT
Background: Few studies have shown the efficacy and safety of lower-fat diets in children.
Objective: Our objective was to assess the efficacy and safety of lowering dietary intake of total fat, saturated fat, and cholesterol to decrease LDL-cholesterol concentrations in children.
Design: A 6-center, randomized controlled clinical trial was carried out in 663 children aged 8–10 y with LDL-cholesterol concentrations greater than the 80th and less than the 98th percentiles for age and sex. The children were randomly assigned to either an intervention group or a usual care group. Behavioral intervention promoted adherence to a diet providing 28% of energy from total fat, <8% from saturated fat, 9% from polyunsaturated fat, and <0.018 mg cholesterol•kJ-1•d-1 (not to exceed 150 mg/d). The primary efficacy measure was mean LDL cholesterol and the safety measures were mean height and serum ferritin concentration at 3 y.
Results: At 3 y, dietary total fat, saturated fat, and cholesterol were lower in the intervention group than in the usual care group (all P < 0.001). LDL cholesterol decreased in the intervention and usual care groups by 0.40 mmol/L (15.4 mg/dL) and 0.31 mmol/L (11.9 mg/dL), respectively. With adjustment for baseline concentration, sex, and missing data, the mean difference between groups was -0.08 mmol/L (95% CI: -0.15, -0.01), or -3.23 mg/dL (95% CI: -5.6, -0.5) (P = 0.016). There were no significant differences between groups in adjusted mean height or serum ferritin.
Conclusion: Dietary changes are effective in achieving modest lowering of LDL cholesterol over 3 y while maintaining adequate growth, iron stores, nutritional adequacy, and psychological well-being during the critical growth period of adolescence.Am J Clin Nutr 2000;72(suppl): 1332S–42S.
Key Words: Efficacy • safety • lower-fat diets • prepubertal children • Dietary Intervention Study in Children • dietary fat • LDL cholesterol
INTRODUCTION
Diets low in total fat, saturated fat, and cholesterol can lower LDL cholesterol in adults (1), but few studies have shown this effect in children (2, 3). Concerns have been raised about the safety of cholesterol-lowering diets in children, particularly during peak growing years (4, 5). These concerns include possible deficits in growth and in nutrient adequacy, specifically in iron intake (5), and potential adverse psychological effects, suspected because of increases in deaths among adults in some cholesterol-lowering trials (6).
In 1987 the National Heart, Lung, and Blood Institute (NHLBI) initiated the Dietary Intervention Study in Children (DISC), a controlled clinical trial to examine the efficacy and safety of long-term dietary intervention for reduction of LDL cholesterol in children followed through puberty (7). We present here the efficacy and safety outcomes of DISC after 3 y of intervention in an intervention and a usual care group. These data were published previously (8, 9).
SUBJECTS AND METHODS
Design
DISC was a 6-center, randomized controlled trial. The primary efficacy outcome was mean serum LDL cholesterol; the 2 primary safety outcomes were mean height and mean serum ferritin concentration at 3 y. The study was designed to randomly assign 600 subjects to either an intervention or a usual care group (300 per group) to have 90% power in a 2-sided test to detect a difference of 0.14 mmol/L (5.4 mg/dL) in LDL cholesterol, and in a conservative 1-sided test to detect a difference of 0.8 cm in height and 3.5 µg/L in serum ferritin. Computer-generated randomization assignments were provided by the coordinating center to produce approximately equal numbers of participants assigned to the intervention and usual care groups, balanced by age and sex, within each clinical center. The expected difference in LDL cholesterol between the 2 groups was based on the assumption that at year 3, 40% of intervention children would meet the DISC dietary goals, 30% would meet the goals halfway, and 30% would not have changed their dietary intake. It was also assumed that LDL cholesterol would decrease 4.5% in the usual care group because of modest self-selected changes in diet (7).
Secondary efficacy outcomes were mean LDL cholesterol at 1 y and mean total cholesterol at 1 and 3 y. Secondary safety outcomes included red blood cell folate; serum zinc, retinol, and albumin; serum HDL cholesterol; the ratio of LDL to HDL cholesterol (LDL:HDL); triacylglycerol; sexual maturation; and psychosocial health. The study protocol was approved by the institutional review boards of all participating centers and informed consent was obtained from a parent or guardian of each child.
Subjects
Age ranges were selected to accommodate sex differences in sexual maturation and to ensure that participants would be studied during the pubertal growth spurt. Age eligibility for girls was 7 y 10 mo to 10 y 1 mo and that for boys was 8 y 7 mo to 10 y 10 mo.
Children were recruited from public and private elementary schools through mass mailings to members of a health maintenance organization and through pediatric practices. More than 44000 children were prescreened. For prescreening, nonfasting capillary blood cholesterol concentrations were measured with use of desktop cholesterol analyzers. Children whose total cholesterol concentration was 4.5 mmol/L (175 mg/dL) (approximately the 75th age- and sex-specific percentile) (10) were invited to the first screening visit, at which a fasting venous blood sample was obtained for measurement of serum LDL cholesterol. Children with LDL-cholesterol concentrations greater than or equal to the 70th and less than the 99th age- and sex-specific percentiles (10) and who met other DISC eligibility criteria were invited for a second screening visit, at which fasting LDL cholesterol was measured a second time. To include children for whom dietary intervention could be recommended as the main treatment modality while excluding those with severe hypercholesterolemia who might need medication, the final lipid eligibility criterion was the average of the 2 screening LDL-cholesterol values greater than or equal to the 80th and less than the 98th percentiles for age and sex. Children with the following were excluded: medical condition or medication that might affect growth or blood cholesterol, behavior problems in the child or family likely to reduce adherence, onset of puberty, and plans to move within the 3 study years. Other exclusions were described previously (7). Eligible children (n = 663) were randomly assigned to 1 of 2 treatment groups, intervention or usual care.
Measurements
These were described previously (7, 8). Briefly, serum LDL cholesterol, total cholesterol, triacylglycerol, HDL cholesterol, dietary intake, physical activity as ascertained by questionnaire, skinfold thicknesses, body circumferences, and blood pressure were measured at baseline, 1 y, and 3 y. Blood micronutrients, serum albumin, hemoglobin, and psychological assessments were measured at baseline and 3 y. Height, weight, and Tanner stage were measured annually. Data collectors were blinded to group assignment.
The Centers for Disease Control and Prevention (CDC) provided quality-control sera and assigned reference values for lipid measurements. The Johns Hopkins Lipoprotein Analytical Laboratory, a participant in the CDC-NHLBI Lipid Standardization Program (11), served as the central lipid laboratory. LDL-cholesterol values were calculated from cholesterol, triacylglycerol, and HDL cholesterol by the Friedewald equation (12), modified to use the factor triacylglycerol/6.5, which was derived from the Lipid Research Clinics Program Prevalence Studies data and selected to minimize error in calculated LDL-cholesterol values in children (10). Baseline and year-3 lipid values used in data analyses were the average of 2 measurements taken 1 mo apart, except for 88 participants in whom only one year-3 value was available. In these participants, the year-3 value was based on a single measurement.
Blood micronutrients, including serum ferritin, zinc, retinol, and red blood cell folate, were measured by a CDC laboratory. Serum albumin was measured in the clinical laboratory of the Johns Hopkins Hospital. Two measurements of height and weight were taken at each time point; if the measurements differed by 0.5 cm for height or 0.1 kg for weight, a third measurement was taken. The average of the closest 2 measurements was used. Body mass index (BMI) was calculated from weight and height (kg/m2). Skinfold thicknesses were measured twice at the triceps, subscapular, and suprailiac sites by using Holtain calipers (model 610; Holtain Ltd, Croswell, United Kingdom). Body circumferences were measured at the waist and hip. Sexual maturation was assessed by Tanner staging of pubic hair stage for boys and girls, breast development for girls, and genitalia development for boys (13). Blood pressure was measured by using a Hawksley random-zero sphygmomanometer according to a standardized protocol.
Dietary assessment was conducted by trained and certified dietitians. Three nonconsecutive, 24-h dietary recalls were collected within 2 wk of the clinic visit. The DISC dietary assessment method has been validated (14) and described in detail (7, 15). Nutrient analyses were performed by the Nutrition Coordinating Center (database version 20) at the University of Minnesota. Mean nutrient intakes for the 3 recalls were calculated.
Psychological assessment included the Achenbach Child Behavior Checklist (16), the Kovacs Child Depression Inventory (17), the Spielberger State-Trait Anxiety Inventory for Children (18), the reading and mathematics subtests of the Woodcock-Johnson Psycho-Educational Battery (19), the Moos Family Environment Scale (20), the Eyberg Child Behavior Inventory (21), and the Life Experience Survey (22).
Numerous quality-control procedures were conducted. These included a common protocol for measurements, central training and certification of measurers, site visits to clinical sites, use of central laboratories for blood measurements, central coding and nutrient analyses for dietary assessment, blinded duplicate blood samples and dietary recalls, and data quality-control checks.
Monitoring was implemented for the intervention and usual care groups to review all collected data to detect cases requiring physician referral for inadequate growth or nutrition, delayed sexual maturation, blood chemistries, or psychosocial abnormalities. Additional safety monitoring procedures were initiated for the intervention group, including assessments of dietary intake and height and weight obtained at intervention visits.
Usual care group
The parents or guardians of participants in the usual care group were informed that their child's blood cholesterol concentration was high but were not given specific recommendations to see a physician. At baseline, all DISC families were given publicly available educational publications on heart-healthy eating, and 3-y lipid results were provided for them to share with their regular physicians. In addition, cases in which measurements of LDL cholesterol, height, or ferritin exceeded cutoffs for clinical monitoring were reviewed to assess whether physician referral was warranted on the basis of National Cholesterol Education Program (NCEP) guidelines for drug treatment (23) and clinical judgment. If referral was warranted, the parent or guardian was given the results with a referral letter to take to his or her regular physician.
Intervention group
The primary goal of the intervention was adherence to a diet providing 28% of energy from total fat, <8% from saturated fat, 9% from polyunsaturated fat, and <0.018 mg cholesterol•kJ-1•d-1 (not to exceed 150 mg/d). The diet was designed to meet age- and sex-specific recommended dietary allowances (RDAs) for energy, protein, and micronutrients (24).
Intervention strategies were based on social learning theory (25) and social action theory (26). The intervention program was family oriented. At the first intervention visit, each participant's current eating pattern was assessed and a personalized program was developed. Six weekly and then 5 biweekly group sessions augmented by 2 individual visits of children with their family members were held in the first 6 mo. In the second 6 mo, 4 group and 2 individual sessions were held. During the second and third years, group and individual maintenance sessions were held 4–6 times each year with monthly telephone contacts between sessions. Group sessions, which were led by nutritionists, behaviorists, and health educators, are described in more detail elsewhere (27, 28). At individual sessions with the child and parent or guardian, staff obtained periodic capillary blood cholesterol measures, provided feedback, and answered questions regarding the child's progress toward the dietary goals, nutritional adequacy, and growth.
Data analysis
Primary analyses
For the analysis of primary outcome values, analysis of covariance (ANCOVA) models (29) were used to test each of the primary hypotheses, with baseline value of the outcome variable and sex as covariates. The primary efficacy null hypothesis was that the adjusted mean year-3 LDL-cholesterol concentration did not differ between the usual care and intervention groups. Because we were interested in detecting a difference if either group's mean LDL cholesterol was greater, a two-sided 0.05 significance level was used. For the 2 safety endpoints, one-sided tests with a significance level of 0.05 were used because we were interested in differences in one direction. The safety null hypotheses were that the mean year-3 height and the mean year-3 serum ferritin in the intervention group were equal to or larger than those in the usual care group. Potential interactions between treatment assignment and sex were tested and found to be not significant.
Analyses were performed according to intention-to-treat. It was assumed that missing data in both groups would have come from the same distribution as observed data in the usual care group, so missing year-3 LDL-cholesterol data were estimated by drawing values from the usual care group distribution (30). For comparison, analyses also were conducted by using observed data only. P values, adjusted differences between groups, and confidence intervals around the differences were calculated. For the primary safety outcomes, it was assumed that missing data had the same distribution as observed in the group to which the participant was assigned, so safety outcomes were tested by using the observed data. These different imputation methods were also used because for efficacy we did not want to overstate differences between the 2 groups, whereas for safety we did not want to miss a difference if one existed.
Analyses of secondary outcomes used no imputation for missing values and used ANCOVA models for continuous outcomes and Wilcoxon tests for ordered categorical outcomes. Baseline level and sex were included as covariates.
Secondary analyses of pooled data
The relation between dietary fat intake and nutritional safety and adequacy measures was also examined by pooling data for the intervention and usual care groups. It was possible to pool the data because no interactions between treatment assignment and fat intake were significant in the relations examined. Nutritional safety measures analyzed included 4 anthropometric measures, 14 micronutrients obtained from 24-h recalls, and 8 blood biochemical measures. Separate analyses were performed with use of each measure of nutritional adequacy and safety as the response variable. The independent variable of interest was energy intake from fat, which was represented as "fat energy" [fat in grams multiplied by 9 kcal(38 kJ)/g]. Fat energy was used in the models rather than percentage of energy from fat to avoid ambiguous interpretation associated with nutrient densities (31, 32). Energy intake in the statistical models was controlled for with a variable that included all sources of energy other than fat. Other control variables in all models were sex, treatment group, and indicator variables used to represent visit number (baseline, year 1, and year 3). The first-order interaction terms that were considered a priori to be plausible and important to control for were retained in all models; these included fat intake x treatment, fat intake x sex, fat intake x visit number, and visit number x treatment. For the anthropometric variables (except height), physical activity was controlled for with the overall physical activity score.
To assess dietary adequacy, several dietary measures were examined further by using longitudinal logistic regression to model the relation between fat intake and the probability of meeting two-thirds of the RDAs. In these analyses, significant interactions of energy intake from fat with visit number, sex, or both were observed. When interactions were significant, analyses were performed separately by visit number (baseline, year 1, and year 3) or by visit number and sex. None of the results for the pooled data were adjusted for multiple comparisons because these analyses are secondary to the main safety hypotheses and because, in assessing safety, it is more conservative to risk accepting a false hypothesis than to risk failing to detect a true hypothesis.
RESULTS
Baseline characteristics
Baseline characteristics of the 2 treatment groups were reported previously (7). Of the 663 participants randomly assigned, 334 (179 boys, 155 girls) were assigned to the intervention group and 329 (183 boys, 146 girls) to the usual care group. The mean ages of the children in the 2 groups were similar: 9.7 y for boys and 9.0 y for girls. There were no significant differences between the 2 groups in anthropometric measurements, blood lipids, and blood pressure. Small differences were seen in some nutrients, with the intervention group having a slightly lower intake as a percentage of energy of polyunsaturated fat (P = 0.03) and slightly higher intakes of vitamin B-6 (P = 0.04) and zinc (P = 0.02). The intervention group had a slightly higher proportion with a household income <$20000 (P = 0.002) and scored slightly higher on the Child Depression Inventory (P = 0.09).
Follow-up
For the 3-y lipid assessment, 95.8% of the intervention and 92.1% of the usual care participants were seen. The mean length of follow-up was 36.2 mo (range: 32.5–50.9 mo) for the intervention group and 36.2 mo (range: 32.3–51.5 mo) for the usual care group. Ferritin values were available for 288 intervention (86.2%) and 279 usual care (84.8%) children. Distributions of baseline characteristics (age, sex, family income, height, LDL cholesterol, and BMI) were not significantly different among intervention and usual care children with and without year-3 LDL cholesterol and ferritin values. Attendance at the intervention sessions, which included make-up sessions, averaged 96% during the first 6 mo, 91% during the second 6 mo, 91% during the second year, and 89% during the third year.
Dietary intake
The mean percentage of energy from total fat and saturated fat and mean cholesterol intake decreased in the intervention group at 1 and 3 y, whereas intake changed little in the usual care group (Table 1). The percentage of energy from monounsaturated fat was lower in the intervention group than in the usual care group at 1 and 3 y. The percentage of energy from polyunsaturated fat was not significantly different between the groups at 1 y and was slightly lower in the intervention group at 3 y. The mean percentage of energy from protein and carbohydrate was higher in the intervention group than in the usual care group at 1 and 3 y. Total energy intake was lower in the intervention group than in the usual care group at 1 and 3 y.
View this table:
TABLE 1. Macronutrient and cholesterol intakes assessed by multiple 24-h recalls1
Primary analyses
LDL-cholesterol concentrations at 1 and 3 y decreased in both groups, with greater decreases in the intervention group (Table 2). The mean difference between the 2 groups in LDL cholesterol at 3 y, after values for missing data were imputed and baseline LDL-cholesterol concentrations and sex were adjusted for, was -0.08 mmol/L (95% CI: -0.15, -0.01), or -3.23 mg/dL (95% CI: -5.6, -0.5) (P = 0.016). As shown in Table 2, the results were essentially the same without imputation for missing values. The mean difference in height between the intervention and usual care groups was not significant at 1 y or at 3 y (Table 3). Serum ferritin values decreased in both the intervention and usual care groups at 1 y (-1.5 and -1.3 µg/L, respectively) and at 3 y (-6.5 and -4.9 µg/L, respectively), but the adjusted mean differences between groups were not significant (Table 4).
View this table:
TABLE 2. Serum lipids and lipoproteins1
View this table:
TABLE 3. Anthropometric measures1
View this table:
TABLE 4. Biochemical safety measures1
Total cholesterol decreased in both groups at 1 and 3 y, with greater decreases in the intervention group (Table 2). At 1 y the intervention group showed a small decrease in HDL cholesterol whereas the usual care group showed little change, resulting in a small difference between groups. At 3 y mean HDL-cholesterol concentrations were not significantly different between groups. There were small decreases in LDL:HDL at 1 and 3 y in both groups but no significant differences between groups. There were increases in triacylglycerol in both groups at 1 y and greater increases at 3 y, but the mean differences between groups were not significant.
There were no significant differences in mean weight or BMI between the groups (Table 3). Additionally, the sum of skinfold thicknesses was not significantly different between the groups. Waist-to-hip ratio was lower in the intervention group than in the usual care group at 1 y but was not significantly different at 3 y. Mean systolic and diastolic blood pressures were not significantly different between the 2 groups at 1 and 3 y (data not shown).
Except for serum retinol at 1 y, which was higher in the intervention group than in the usual care group, there were no significant differences between the 2 groups in hemoglobin (data not shown), mean serum concentrations of retinol, zinc, or albumin, or in red blood cell folate at 1 and 3 y (Table 4). There were also no significant differences between the 2 groups in mean intakes of vitamin A, vitamin C, vitamin B-6, calcium, or iron or in the proportion who met or exceeded two-thirds of the RDAs for age and sex of protein, vitamin A, vitamin C, vitamin B-6, calcium, or iron at 1 and 3 y for (data not shown). The intervention group's mean intake of zinc decreased at 1 y by 0.5 mg compared with an increase of 0.6 mg in the usual care group (adjusted difference: -0.6 mg, P = 0.03). At 3 y the adjusted mean difference (-0.3 mg) was not significant.
No significant differences in sexual maturation were seen between the groups at either 1 or 3 y (data not shown). With the exception of the Kovak's Child Depression Inventory, which showed a lower adjusted mean depression score for the intervention children at 3 y (P = 0.03), there were no significant differences between the groups in psychosocial assessments (data not shown).
Secondary analyses of pooled data
Relation of dietary fat to growth and micronutrients
The percentage of energy from fat ranged from a low of 12.2% to a high of 52.0%. Energy intake from fat was not associated with height, weight, BMI, or the sum of skinfold thicknesses (all P > 0.20) (Table 5). Lower energy intake from fat was associated with higher concentrations of red blood cell folate and hemoglobin and with a trend for higher concentrations of serum ferritin (Table 5). Serum concentrations of ß-carotene, retinol, vitamin E, zinc, and albumin were not related to fat intake. The interactions tested were not significant.
View this table:
TABLE 5. Three 24-h diet recalls: relations between energy intake from fat and anthropometric measures and blood biochemical determinations1
Relation between dietary fat and self-reported nutrient intake
Lower energy intake from total fat was associated with higher dietary intakes of folate, vitamin C, and vitamin A, and a trend for higher iron intake (Table 5). Lower fat intake was associated with lower intakes of calcium, zinc, magnesium, phosphorus, vitamin E, vitamin B-12, thiamine, niacin, and riboflavin. There was no relation between energy intake from fat and vitamin B-6 intake. The only significant interaction occurred for vitamin E, for which the coefficients differed between the 3 visits (significant fat intake x visit number interaction).
Nutrient intake from supplement use was not included in these analyses because nutritional adequacy of the diet was the primary focus. On average, 18% of the DISC children were taking supplements at each visit. Treatment assignment, visit number, sex, and energy intake from fat were not related to the probability of taking supplements. In addition, no interactions were observed between energy intake from fat and supplement use with respect to blood biochemical measures. Results were consistent when analyses relating dietary fat intake with blood biochemical measures were performed in the subgroup of children who were not taking supplements. Thus, relations between energy intake from fat and nutritional biochemical measures were not modified by the use of supplements.
Relation between dietary fat and meeting two-thirds of the RDAs
The risk of consuming less than two-thirds of the RDA was examined for 9 nutrients: calcium, zinc, magnesium, phosphorus, vitamin E, vitamin B-12, thiamine, niacin, and riboflavin. Lower fat intake was not related to the risk of consuming less than two-thirds of the RDA for most of the 9 nutrients. The exceptions were as follows: calcium in girls at baseline only [odds ratio (OR): 1.011; P < 0.0003], vitamin E for all visits (baseline OR: 1.009; year 1 OR: 1.007; year 3 OR: 1.007; P < 0.0001 for all visits), and zinc at all visits for boys (baseline OR: 1.004, P < 0.05; year 1 OR: 1.003, P < 0.02; and year 3 OR: 1.004, P < 0.003) and girls (baseline OR: 1.007, P < 0.001; year 1 OR: 1.008, P <0 .0006; and year 3 OR: 1.005, P < 0.003).
DISCUSSION
This large randomized controlled trial on the efficacy and safety of a dietary intervention to reduce LDL cholesterol in pubescent children had characteristics not present in most previous studies (2, 3, 33, 34), including a long follow-up, a high follow-up rate, and a randomized controlled design that enabled the effects of the intervention to be distinguished from developmental or secular trends or other causes. The DISC participants had mean baseline LDL-cholesterol concentrations at about the 95th percentile of the Lipid Research Clinics Prevalence Study (10). Children with extreme LDL-cholesterol concentrations (at or above the 98th percentile) or with premature coronary heart disease (CHD) in a parent were excluded. Thus, it is likely that few, if any, of the participants had familial hypercholesterolemia.
The DISC diet was similar to the NCEP Step II diet recommended for children with a family history of CHD (23). The intervention group met or exceeded the NCEP Step II diet recommendations for total fat and cholesterol and met the NCEP Step I diet recommendation for saturated fat.
The efficacy of the intervention in DISC was determined by comparing at 3 y the mean LDL-cholesterol value adjusted for baseline LDL cholesterol and sex between the intervention and usual care groups. In this trial, mean LDL cholesterol was significantly lower at 3 y in the intervention group than in the usual care group. The magnitude of the net difference in LDL-cholesterol concentrations between the groups was similar to that reported in randomized controlled studies in schools (35, 36), communities (37), and clinics (2).
The 0.09-mmol/L (3.3-mg/dL) difference in LDL cholesterol observed at 3 y between the intervention and usual care groups was less than the 0.14-mmol/L (5.4-mg/dL) difference targeted when DISC was planned. The mean dietary saturated fat and cholesterol intakes in the intervention group were consistent with the design assumptions about adherence to the diet. The smaller-than-expected difference in LDL cholesterol between the 2 groups was due mostly to the decrease in LDL cholesterol in the usual care group, a phenomenon also seen in several intervention trials in both the pediatric (2, 35, 37) and adult (38) literature. The decrease in LDL cholesterol in the DISC children may have been due to a combination of dietary modification in some adolescents and effects of puberty on LDL-cholesterol concentrations (39).
Previous studies of diets low in saturated fat, total fat, and cholesterol, particularly those high in polyunsaturated fat, reported significant decreases in HDL cholesterol with consumption of these diets (1, 33, 40). In DISC, HDL cholesterol decreased in both the intervention and usual care groups. The decrease was slightly larger in the intervention group than in the usual care group at 1 y (P < 0.03), but there was no significant difference between the 2 groups at 3 y. The lack of a differential effect of the DISC diet on HDL cholesterol may be because polyunsaturated fat constituted only 5–6% of energy in both groups.
LDL:HDL has been used to assess relative amounts of atherogenic LDL cholesterol and antiatherogenic HDL cholesterol (41). The ratio decreased in both groups. There was a somewhat greater decrease in the intervention group than in the usual care group, but the difference was not significant. Implications for CHD risk of a lack of significant change in LDL:HDL in children are unclear. However, LDL cholesterol has been established as an independent risk factor for CHD (42). Because there is a much larger body of evidence supporting the causal relation between LDL cholesterol and CHD than between HDL cholesterol and CHD (43), it is likely that a decrease in LDL cholesterol without a concomitant significant decrease in LDL:HDL does decrease CHD risk.
Although a diet low in fat and high in carbohydrate can increase total triacylglycerol (1, 43), the DISC diet did not increase the triacylglycerol concentration in the intervention group above that in the usual care group. The increase in both groups was most likely related to the well-documented rise that occurs with age (10).
The difference in LDL cholesterol between the intervention and usual care groups was not large, and the magnitude of the effect on subsequent CHD incidence is uncertain. Long-term follow-up studies will be necessary to determine whether the intervention had long-term effects on LDL cholesterol. However, it could be hypothesized that small differences in LDL cholesterol could affect the pathogenesis of atherosclerosis, particularly in early stages during childhood, or could lesson the slope of the rise in LDL-cholesterol concentrations with age. It is important to note that a small downward shift of the population distribution of a risk factor can have a considerable public health effect on disease outcome, particularly for prevalent conditions such as CHD, and on the proportion above a clinical cutoff requiring drug treatment in later life.
Despite reported lower energy intakes in the DISC intervention group, mean weight and BMI were not significantly different between groups. These results suggest underreporting by the intervention group, which is important if this biased the reporting of macronutrients. Lissner and Lindroos (44) suggest that underreporting tends to be proportional for macronutrients. Because predicted changes in total cholesterol concentrations based on Key's formula (45) were not significantly different from observed changes at 6 mo (46), it is likely that the reported proportion of energy from macronutrients reflected true relative intakes.
DISC was planned to have high statistical power to detect a small difference in height between the intervention and usual care groups. Growth was comparable in the 2 groups despite a reduction in fat intake by the intervention group. In addition to group averages, individual children were monitored for slow growth by using the third percentile or less of height velocity with use of Tanner and Whitehous's data (47) as the cutoff. Over 3 y, 19 intervention and 28 usual care children were identified as requiring further evaluation for growth. This number is commensurate with approximately the lowest 3% of the population distribution of growth velocity. Although the DISC trial showed no diminution of growth during the children's peak growing years, we cannot ascertain the effect of the trial on ultimate adult height until further follow-up of the participants is analyzed.
Concern has been expressed that cholesterol-lowering diets may reduce iron intake because of lower animal food intake (5). Serum ferritin decreased in both groups and the intervention group had slightly lower mean ferritin concentrations than did the usual care group at 3 y (P = 0.08). The -2.1-µg/L difference in mean serum ferritin concentrations decreased to -1.5 µg/L (P = 0.12) when a high outlying value of 215 µg/L (normal range: 5–150 µg/L) for a usual care child was removed from the analysis. In both groups, mean serum ferritin concentrations remained above the 75th percentile for age and sex (48). Four children, 3 in the intervention group and 1 in the usual care group, required referral for evaluation of low serum ferritin. Review of the 24-h dietary recalls revealed no evidence of diet causing these low ferritin values.
Psychological assessment revealed no group differences with the exception of the depression score, which was lower at year 3 in the intervention group than in the usual care group. This unexpected finding suggests that the behavioral approaches used in the DISC intervention, such as enhancing social support through group sessions, may have had a salutary effect.
No significant differences in blood biochemical measures or dietary micronutrient intakes were found between children in the intervention and usual care groups (8). Thus, by pooling the intervention and usual care groups in secondary analyses, we examined further the nutritional adequacy and safety of a reduced-fat diet as determined by anthropometric and blood biochemical measures and self-reported dietary intakes. These results showed that growth and biochemical measures of nutritional status were not adversely related to lower dietary fat (9). These findings are consistent with other observational studies reporting no adverse associations with growth and nutritional biochemical measures in children consuming vegetarian diets (49–52) and in nonvegetarian children consuming lower-fat diets (53–55), and with other cholesterol-lowering intervention studies in children that reported no adverse effects on growth or serum measures of micronutrients (56–58). In addition, the results showed that lower fat intakes did not result in significant differences in body fat.
In the present study, however, lower fat intakes, independent of total energy intake, were associated with self-reported lower intakes of 9 micronutrients from among 14 investigated. Of these 9, assessment of nutritional adequacy showed that energy intake from fat was associated with increased risk of not meeting two-thirds of the RDA for zinc and vitamin E and inconsistently for calcium (at baseline in girls). Fat intake, however, was not related to serum zinc or vitamin E concentrations and there was no evidence of zinc or vitamin E deficiency in any individual child on the basis of the distribution of serum concentrations. The mean zinc, vitamin E, and calcium intakes of the DISC children were similar to those reported in US children (59), particularly when expressed in terms of 4200 kJ (1000 kcal). Because the risk of not meeting two-thirds of the RDA for calcium occurred only at baseline in girls, the intervention did not result in further risk of inadequate calcium intake. Underreporting of food intake, including animal products and high-fat foods, may account for the relations observed with dietary measures and not with biochemical measures. Also, because of a paucity of source data, 74% of the vitamin E data in the nutrient database were imputed. Although the reported dietary intake data suggest that lower fat intakes may result in lower than recommended intakes of zinc and vitamin E, there was no evidence of any adverse effects on biochemical and anthropometric objective measures of nutritional status.
In conclusion, the results of DISC provide evidence from a large-scale, long-term, randomized controlled clinical trial that a properly designed dietary intervention is effective in achieving modest lowering of elevated LDL cholesterol over 3 y while maintaining adequate growth, iron stores, nutritional adequacy, and psychological well-being during the critical growth period of adolescence. These results indicate that children who require dietary change to lower their LDL cholesterol may safely and successfully do so under supervision. An important public health inference from the DISC results is that current dietary recommendations for healthy children, which are less restricted in total fat than the DISC diet, can be advocated safely, particularly when children receive health care in which their growth and development are followed.
ACKNOWLEDGMENTS
The investigators express appreciation to the DISC Data and Safety Monitoring Committee: John C LaRosa (chair), Phyllis E Bowen, Allan Drash, Robert J Hardy, Judith K Ockene, Carol Whalen, and Richard Grimm. The DISC Collaborative Research Group consists of the following: The Johns Hopkins University School of Medicine, Baltimore, Peter O Kwiterovich Jr (principal investigator), Ginny Hartmuller; Northwestern University Medical School, Chicago, Linda Van Horn (principal investigator), Katherine K Christoffel, Niki Gernhoffer, Samuel Gidding, John V Lavigne; the University of Iowa School of Medicine, Iowa City, Ronald M Lauer (principal investigator), Linda Snetselaar, Patti Steinmuller, Lynette Stickney; the New Jersey Medical School, Newark, Norman L Lasser, (principal investigator), Rhonda E Greenberg, Patricia Kennedy, Vera I Lasser; Children's Hospital, New Orleans, Alan M Robson (principal investigator), Frank A Franklin Jr, Kristian Von Almen; Kaiser Permanente Center for Health Research, Portland, OR, Victor J Stevens (principal investigator), Shirley Craddick, Merwyn R Greenlick, Jacob A Reiss; Maryland Medical Research Institute, Baltimore (Coordinating Center), Bruce A Barton (principal investigator), Kathleen Brown, Paul L Canner, Sue YS Kimm, Robert McMahon; National Heart, Lung, and Blood Institute (Project Office), Bethesda, MD, Eva Obarzanek (project officer), Jeffrey A Cutler, Marguerite A Evans, Marilyn Farrand Zukel, Sally A Hunsberger, Edward Lakatos, Nancy C Santanello, Denise G Simons-Morton; Central Lipid Laboratory, Lipid Research Atherosclerosis Unit, and Nonlipid Laboratory, The Johns Hopkins Hospital Clinical Laboratory, Baltimore, Paul S Bachorik (principal investigator); the Centers for Disease Control and Prevention (Micronutrient Laboratory), Elaine W Gunter (chief); the Nutrition Coordinating Center (Nutrition Coding Center), University of Minnesota, Minneapolis, I Marilyn Buzzard (director).
REFERENCES