Effects of Diet and Sexual Maturation on Low-Density Lipoprotein Cholesterol During Puberty
The Dietary Intervention Study in Children (DISC)
Background The Dietary Intervention Study in Children (DISC) is a multicenter, randomized, controlled clinical trial designed to examine the efficacy and safety of a dietary intervention to reduce serum LDL cholesterol (LDL-C) in children with elevated LDL-C.
Methods and Results The effects of dietary intake of fat and cholesterol and of sexual maturation and body mass index (BMI) on LDL-C were examined in a 3-year longitudinal study of 663 boys and girls (age 8 to 10 years at baseline) with elevated LDL-C levels. Multiple linear regression was used to predict LDL-C at 3 years. For boys, LDL-C decreased by 0.018 mmol/L for each 10 mg/4.2 MJ decrease in dietary cholesterol (P<.05). For girls, no single nutrient was significant in the model, but a treatment group effect was evident (P<.05). In both sexes, BMI at 3 years and LDL-C at baseline were significant and positive predictors of LDL-C levels. In boys, the average LDL-C level was 0.603 mmol/L lower at Tanner stage 4+ than at Tanner stage 1 (P<.01). In girls, the average LDL-C level was 0.274 mmol/L lower at Tanner stage 4+ than at Tanner stage 1 (P<.05).
Conclusions In pubertal children, sexual maturation, BMI, dietary intervention (in girls), and dietary cholesterol (in boys) were significant in determining LDL-C. Sexual maturation was the factor associated with the greatest difference in LDL-C. Clinicians screening for dyslipidemia or following dyslipidemic children should be aware of the powerful effects of pubertal change on measurements of lipoproteins.
The DISC is a multicenter, randomized, controlled clinical trial designed to examine the efficacy and safety of a dietary intervention to reduce serum LDL-C in pubescent children with elevated LDL-C. The children in the intervention group were taught to follow a diet similar to the National Cholesterol Education Program Step II Diet, which is recommended for children with a family history of premature coronary heart disease.1 We recently reported2 that after 3 years of intervention, mean LDL-C was significantly lower in the intervention group than in the usual care group. Compared with the children in the usual care group, the children in the intervention group had significantly lower intakes of dietary total fat (28.6% versus 33.0% of energy), saturated fat (10.2% versus 12.3%), and cholesterol (95.0 versus 112.9 mg/4.2 MJ). The DISC diet was found to be safe, based on no significant differences in height, weight, iron stores (as measured by serum ferritin), sexual development (as measured by Tanner stage), or psychological status between the intervention and usual care groups.
Previous studies found significant declines in LDL-C during pubescence. In our initial report,2 LDL-C decreased 0.303 mmol/L (11.7 mg/dL) in the usual care group compared with 0.398 mmol/L (15.4 mg/dL) in the intervention group. Here, we examine in greater detail the relationship between dietary intake of total fat, saturated fat, and cholesterol at 3 years and the observed levels of LDL-C at 3 years of follow-up and change from baseline to 3 years. Other factors, such as initial LDL-C levels, BMI (a measure of adiposity), sex, or sexual maturation may influence the effect of diet on LDL-C. This report presents the analysis of the relationship between LDL-C observed in both the intervention and usual care groups and dietary intake of fat and these other factors.
The design and main results at 3 years for the DISC study have been reported previously.1 2 Briefly, mass screenings identified potentially eligible children for total cholesterol values at or above the age- and sex-specific 75th percentile and below the 98th percentile. LDL-C and other eligibility criteria were assessed at two additional pre-entry visits. A total of 663 children, for whom the average of the LDL-C measurements from the second and third screening visits fell between the age- and sex-specific 80th and 98th percentiles, was enrolled at six clinical centers. All children were prepubertal at the time of enrollment; boys were between ages 8 years 7 months and 10 years 10 months and girls were between ages 7 years 10 months and 10 years 1 month at entry. The parent or guardian of each child provided informed consent. We randomly assigned eligible children to one of two treatment groups: intervention or usual care.
The goal of the dietary intervention was 28% of energy from total fat, <8% from saturated fat, up to 9% from polyunsaturated fat, and <75 mg/4.2 MJ per day of dietary cholesterol (not to exceed 150 mg/day). The diet goals met age- and sex-specific recommended dietary allowances (RDAs) for energy, protein, and micronutrients.
At the first intervention visit, interventionists assessed the current eating pattern and developed a personalized program for each participant.3 4 Interventionists presented six weekly group sessions to the families assigned to the intervention group, followed by five biweekly group sessions augmented by two individual sessions. In the second 6 months, intervention participants attended four group sessions and two individual sessions. During the second and third years, interventionists held group and individual maintenance sessions four to six times each year with monthly telephone contacts between sessions. The content of these sessions is described in more detail elsewhere.5 6
At baseline, we gave all DISC families educational publications available to the public on heart-healthy eating. We informed parents or guardians of usual care participants that their children’s blood cholesterol level was elevated but gave no specific recommendations to see their physician.
At baseline, 1 year, and 3 years, we obtained measurements of fasting serum LDL-C, total cholesterol, triglycerides, and HDL-C; dietary intake; height; and weight. Baseline and 3-year lipid values are the average of two measurements, performed 1 month apart, except for 88 participants for whom only one 3-year measurement was available. We assessed Tanner stage annually. Data collectors were blinded to treatment group assignment.
The Centers for Disease Control and Prevention provided quality control sera and assigned reference values for lipid measurements. The Johns Hopkins Lipoprotein Analytical Laboratory, a participant in the Centers for Disease Control and Prevention–National Heart, Lung, and Blood Institute Lipid Standardization Program,7 served as the central lipid laboratory. We used a modification of the Friedewald equation8 to calculate LDL-C.1 The mean bias in LDL-C measurements (difference between LDL-C values calculated from reference pools and LDL-C values calculated from DISC laboratory measurements) was −0.8%. Coefficients of variation for LDL-C, calculated from the variances of total cholesterol, HDL-C, and triglyceride measurements on unblinded quality control pools, averaged 3.7%. We determined total cholesterol and HDL-C levels using the cholesterol CHOD-PAP method9 (Boehringer-Mannheim Diagnostics). We analyzed triglycerides enzymatically using a commercially available method (Abbot A-Gent Triglycerides Reagent Set).
Trained and certified dietitians collected three nonconsecutive 24-hour dietary recalls within 2 weeks after the clinic measurement visit. The DISC dietary assessment method has been validated10 and described in detail elsewhere.11 The Nutrition Coordinating Center at the University of Minnesota performed nutrient analyses.
We measured height and weight twice at each visit1 and used the average of the two measurements of height and weight to calculate BMI (kg/m2). We assessed maturation by Tanner staging of pubic hair for boys and girls, breast development for girls, and genitalia development for boys.12 The five Tanner stages are 1 (preadolescent), 2 through 4 (pubertal), and 5 (complete sexual maturation). In the current analysis, we summarize maturation stage by the highest of pubic hair or genitalia stage for boys and by the highest of pubic hair or breast stage for girls.
Of the 663 DISC participants, all had baseline LDL-C data, 623 had 3-year LDL-C data, 596 had complete 3-year nutrition data, 574 had 3-year BMI data, and 566 (312 boys and 254 girls) used in these analyses had all of these data as well as 3-year Tanner staging information.
We examined the associations of LDL-C at 3 years with the independent variables of dietary cholesterol (mg/4.2 MJ), saturated fat (% energy), and polyunsaturated fat (% energy) also measured at 3 years. Adjustment variables were LDL-C at entry (mmol/L) and treatment group. We also investigated the effects on LDL-C of BMI and Tanner stage at 3 years. Because the measures of sexual maturation are not completely comparable between boys and girls, we analyzed the data separately by sex. Tanner stage at 3 years between the two treatment groups was compared within sex using Wilcoxon tests.
We analyzed the data using multiple linear regression first for each independent nutrient variable individually (adjusting for LDL-C at entry and treatment group) and then for all nutrients and other independent variables in the model together. We combined the two treatment groups for these analyses because no significant interactions between treatment and dietary intake were found, indicating that the effect of dietary intake on LDL-C did not differ by treatment group.
Regression coefficients were adjusted for random error in dietary measurements. Two sources of random measurement error occur in estimating the usual dietary intake for each participant: (1) within-person variation in actual diet and (2) random variation in the accuracy with which the diet is recalled. Measuring multiple factors with error may either decrease or increase (bias toward or away from zero) the regression coefficient estimates.13 14 15 To estimate within-person and between-person variation in recalled dietary intakes, we used three 24-hour dietary recalls for each participant. We corrected the regression coefficient estimates to account for these two sources of variation.14 We estimated the standard errors of the corrected regression coefficients by a technique of repeated (bootstrap) sampling from the observed data. Each bootstrap sample consisted of 312 (for boys) or 254 (for girls) LDL-C values with corresponding values for the independent variables sampled with replacement from the original data. We repeated the regression procedure 500 times with correction for dietary measurement error for each bootstrap sample. Next, we calculated the standard deviations for the resulting 500 sets of estimated regression coefficients. We used these standard deviations as the estimates of the standard error for the coefficients estimated from the original data. We calculated values for P for each coefficient from the z-statistic obtained by dividing each regression coefficient by the bootstrap estimate of its standard error.
In addition, we used linear regression analyses to assess the effect of 3-year change in dietary intakes on the 3-year change in LDL-C (change score models). These regression models included the same independent variables as above, except that the nutrient intakes and BMI at 3 years were replaced, respectively, with change in nutrient intakes and change in BMI from baseline to 3 years. Similarly, the dependent variable was change in LDL-C from baseline to 3 years rather than LDL-C at 3 years. These analyses did not adjust for random error in dietary measurements because of the complexity of adjusting the standard errors of the regression coefficients for the random error in two sets of dietary recalls.
We combined Tanner stages 4 and 5 in the regression analyses because of the small number of participants at Tanner stage 5. To illustrate the changes from preadolescence through puberty, we used participants at Tanner stage 1 as the reference group for Tanner stages in the regression models.
Table 1⇓ presents the characteristics of participants at baseline and 3 years. Table 2⇓ presents the number and percent of participants at each Tanner stage at the 3-year visit. All participants were Tanner stage 1 at entry. We noted no significant differences in stage of maturation between the two treatment groups for either boys or girls.
Figs 1⇓ and 2⇓ show levels of LDL-C at 3 years plotted against percent of energy from total, saturated, and polyunsaturated fat and dietary cholesterol (mg/4.2 MJ) for boys and girls, respectively. Each figure includes linear regression lines to display trends in mean LDL-C versus each dietary measure and the range of the relationship between LDL-C and dietary variables. These figures also include the bivariate Pearson correlation coefficient without adjustment for measurement error or other covariates. The correlation coefficients range from 0 for polyun- saturated fat to +0.15 for dietary cholesterol, both for the usual care boys. Analyses discussed below investigate these relationships after adjustment for other factors.
Sexual maturation and BMI are important biological factors that influence LDL-C levels.16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 Fig 3⇓ presents the mean changes in LDL-C at 3 years by Tanner stage and sex. LDL-C levels were lower at higher levels of puberty. By Tanner stage 4, the mean LDL-C levels were lower by >0.517 mmol/L (20 mg/dL) in boys and by >0.388 mmol/L (15 mg/dL) in girls.
Fig 4⇓ presents LDL-C at 3 years in relation to BMI at 3 years. The Pearson correlation coefficients range from.13 for intervention boys and girls to.25 for usual care girls. Analyses discussed below investigate these relationships after adjustment for other factors.
Table 3⇓ presents results from separate linear regression analyses adjusted for measurement error to examine the dose-response relationship between 3-year levels of LDL-C and individual dietary factors, BMI, and Tanner stage. These models were adjusted for LDL-C at study entry and treatment group. We found no significant relationships between nutrients and LDL-C in girls. In boys, higher dietary cholesterol was associated with higher LDL-C (P<.01). Higher BMI was significantly related to higher LDL-C in both boys and girls. Tanner stage 4+ was significantly related to lower LDL-C in boys, and the trend was in the same direction in girls.
Table 4⇓ presents results from multivariable regression models predicting LDL-C at 3 years from dietary cholesterol, percent of energy from saturated fat and polyunsaturated fat, treatment group, BMI, and Tanner stage at 3 years and from LDL-C at entry. In boys, LDL-C decreased by 0.018 mmol/L (0.7 mg/dL) for each 10 mg/4.2 MJ decrease in dietary cholesterol (P<.05). In girls, no nutrient was significant in the multivariable model. On average, participants assigned to the intervention group had lower LDL-C than the usual care group (0.129±0.054 mmol/L [5.0±2.1 mg/dL] lower in girls, 0.070±0.062 mmol/L [2.7±2.4 mg/dL] lower in boys). The difference between treatment groups after adjustment for the effect of dietary intake and biological variables was significant only in girls.
For both boys and girls, as Tanner stage increased, LDL-C decreased. The effect was larger in boys; on average, LDL-C was 0.603 mmol/L (23.3 mg/dL) lower among boys at Tanner stage 4+ than at Tanner stage 1, and among girls, LDL-C was 0.274 mmol/L (10.6 mg/dL) lower. Although the differences in LDL-C between girls at Tanner stage 1 and all later stages were not statistically significant, overall there was evidence of differences among Tanner stages (P<.05) with a significant difference between stages 4+ and 1.
BMI was a significant predictor of LDL-C at 3 years in both sexes, with an increase of 0.016 mmol/L (0.6 mg/dL) for an increase of 1.0 unit in BMI in boys and 0.028 mmol/L (1.1 mg/dL) for an increase in 1.0 unit in girls.
We found similar results using the change score models. As with the previous models, which adjust for error in dietary measurement, the boys showed a significant effect on LDL-C change from change in dietary cholesterol, baseline LDL-C, change in BMI, and Tanner stages 3 and 4. Girls showed a significant effect on LDL-C change from treatment group, baseline LDL-C, change in BMI, and Tanner stage 4.
To investigate the effect of other measures of adiposity on LDL-C, we also used the sum of three skinfolds (triceps, subscapular, and suprailiac skinfolds) as well as subscapular skinfold separately in the same change score models. The results were the same as with BMI.
Previously, we reported that adolescents in the intervention group in DISC had significantly lower intakes of percent of calories from total fat, saturated fat, monounsaturated fat, and cholesterol (mg/1000 kcal) and significantly greater decrease in LDL-C levels than those in the usual care group.2 Here, we examined the relationship between diet and LDL-C at 3 years as well as change in diet and change in LDL-C. Although differences in LDL-C levels existed between the intervention and usual care groups, unadjusted analyses including only dietary variables did not demonstrate a dose-response relationship between caloric intake from saturated fat, polyunsaturated fat, or dietary cholesterol and LDL-C (Figs 1⇑ and 2⇑) for either boys or girls. This suggested that the relationship among dietary intake, adolescence, and the decline in serum LDL-C levels relationships may be more complex.
In DISC, both the intervention and usual care groups had significant drops in their LDL-C levels, with the boys manifesting a greater drop than the girls. Other investigators have observed significant changes in lipids and lipoproteins during adolescence. Some studies16 17 18 19 reported significant decreases in plasma total cholesterol in adolescent boys but little change in adolescent girls. Others20 21 22 found that girls as well as boys had decreases in plasma total cholesterol during adolescence. These different results may be explained in part by sex differences in lipoprotein fractions. Adolescent boys and girls both appear to have a decrease in LDL-C, whereas boys also have a decrease in HDL-C, which has not been found in girls.16 19 21 23 24 25 26 27
In agreement with others,26 28 29 30 we found that BMI was significantly and positively related to LDL-C levels. Adiposity may also significantly influence the effect of hormones on plasma lipids and lipoproteins during adolescence31 (see below).
After adjustment for baseline LDL-C level and treatment group, we found significant positive effects of dietary cholesterol on LDL-C in boys when we considered dietary factors separately. When dietary factors were analyzed together with BMI and Tanner stage, lower intakes of dietary cholesterol again predicted lower LDL-C levels. In girls, we did not observe any significant effects of dietary nutrients, in either univariate or multivariate analyses, and the signs of the coefficients were opposite to the expected direction. The reasons for this discrepancy between boys and girls are not clear but may be related to differences in hormones, reporting dietary intakes, or lower dietary adherence in the girls.32 The latter explanation is less likely because the adjusted differences in LDL-C between the intervention and usual care groups were greater in girls than in the boys. There may be an interaction of dietary nutrients and error in dietary reporting that is more accentuated in girls than in boys. In the North Karelia Youth study in 13- to 15-year-old Finnish schoolchildren,33 the girls reported greater dietary adherence than the boys.
Treatment group assignment had a significant effect on LDL-C levels in girls even when controlling for dietary factors, indicating that the DISC girls in the intervention group had significantly lower levels of LDL-C after 3 years of dietary intervention than the usual care girls. In the boys, the treatment group effect was in the expected direction although not statistically significant. The remaining effect of treatment group while adjusting for diet suggests an intervention effect not detected by the self-reported dietary measurement.
We found that sexual maturation, as judged by Tanner stage, had the most significant influence on LDL-C, with lower LDL-C associated with more advanced sexual maturation. A number of other studies23 26 28 29 also reported significant effects of sexual maturation on plasma lipid and lipoprotein levels during adolescence. For example, Berenson and coworkers23 examined the relationship between Tanner stage and plasma lipid and lipoprotein levels of 4074 children aged 5 to 17 years in Bogalusa, La. In white boys, the mean LDL-C decreased from 2.121 mmol/L (82 mg/dL) at Tanner stage 1 to 1.914 mmol/L (74 mg/dL) at Tanner stage 4 but then increased to 2.043 mmol/L (79 mg/dL) at the most advanced Tanner stage. A similar pattern was seen in white girls, with LDL-C decreasing from 2.250 mmol/L (87 mg/dL) to 1.991 mmol/L (77 mg/dL) at Tanner stage 4 but then increasing to 2.198 mmol/L (85 mg/dL) at the most mature stage.
The magnitude of the effect of sexual maturation was notably larger in boys than in girls (Table 4⇑). Such a marked sex differences may be related to disparate hormonal patterns that emerge during adolescence. Using the hormone data of Bidlingmaier and Knorr, Berenson and coworkers23 found a correlation between lipid changes and increasing testosterone and estrone in boys, and increasing estrone in the girls. Laskarzewski and coworkers31 found that there was a complex interaction between plasma LDL-C and HDL-C levels, BMI, and endogenous testosterone and estradiol levels in adolescent boys. For example, in boys with high estradiol levels, there was a negative and curvilinear relationship between levels of LDL-C and testosterone. For boys with low estradiol, there was an opposite, positive curvilinear relationship between levels of LDL-C and testosterone. The magnitude of these relationships was the greatest in those with the lowest BMI and the least in those with the highest BMI.
Because children in DISC had elevated LDL-C at entry, regression to the mean may account for some of the decrease in LDL-C at 3 years.34 However, this phenomenon would not account for the differences in LDL-C observed across Tanner stage or the multivariate associations with LDL-C. Assessment of dietary intake was obtained by self-report, in which DISC nutritionists collected three 24-hour recalls for each participant at each visit. Although adequate for group descriptions and comparisons, three recalls may not provide the precision required for linear regression analyses.
In summary, the decline in the LDL-C levels seen in the DISC cohort was due to sexual maturation, BMI, dietary intervention (in girls), and dietary cholesterol (in boys). Although the intervention that was used in this study significantly lowered LDL-C when the intervention group was compared with the control group, it was difficult to demonstrate directly that this was due to reported changes in dietary fat and cholesterol intake. Sexual maturation, particularly in boys, was the strongest and most significant factor associated with the decline in LDL-C over 3 years. Clinicians screening for dyslipidemia or following dyslipidemic children should be aware of the powerful effects of pubertal change on measurements of lipoproteins.
Selected Abbreviations and Acronyms
|BMI||=||body mass index|
|DISC||=||Dietary Intervention Study in Children|
|HDL-C||=||high-density lipoprotein cholesterol|
|LDL-C||=||low-density lipoprotein cholesterol|
The 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, National Institutes of Health, Bethesda, Md.
Reprint requests to Bruce A. Barton, PhD, Maryland Medical Research Institute, 600 Wyndhurst Ave, Baltimore, MD 21210.
- Received October 1, 1996.
- Revision received May 1, 1997.
- Accepted May 15, 1997.
- Copyright © 1997 by American Heart Association
Bandura A. Social Learning Theory. Englewood Cliffs, NJ: Prentice Hall; 1977.
Hartmuller VW, Snetselaar L, Van Horn L, Steinmuller P, Smith K, Gernhofer N, Evans M, Lasser V, Cecil M, Brown KM, Craddick S, Stevens VJ, von Almen K, Chiostri J, Moag-Stahlberg A. Creative approaches to cholesterol lowering used in the Dietary Intervention Study in Children (DISC). Top Clin Nutr. 1994;10:71-78.
Stevens VJ, Obarzanek E, Franklin FA, Steinmuller P, Snetselaar L, Lavigne J, Batey D, von Almen K, Hartmuller V, Reimers T, Lasser VI, Craddick S, Gernhofer N. A family-oriented nutritional and behavioral approach to cholesterol reduction. J Nutr Educ. 1995;27:133-140.
Friedewald WT, Levy RI, Frederickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma without use of the preparative ultracentrifuge. Clin Chem. 1972;18:499-502.
Bachorik PS, Walker RE, Virgil DG. High density lipoprotein cholesterol in heparin MnCl2 supernates determined with the Dow enzymatic method after precipitation of Mn+2 with HCO3. Clin Chem. 1984;30:839-842.
Tanner JM. Growth at Adolescence. 2nd ed. Oxford, UK: Blackwell Scientific Publications; 1962:28-39.
Fuller WA. Measurement Error Models. New York, NY: John Wiley & Sons; 1987.
Liu K. Measurement error and its impact on partial correlation and multiple linear regression analyses. Am J Epidemiol. 1988;127:864-874.
Dyer AR, Shipley M, Elliot P. Urinary electrolyte excretion in 24 hours and blood pressure in the INTERSALT Study, I: estimates of reliability. Am J Epidemiol. 1994;139:927-939.
Twisk JW, Kemper HC, Mellenbergh GJ. Longitudinal development of lipoprotein levels in males and females aged 12-28 years: the Amsterdam Growth and Health Study. Int J Epidemiol. 1995;24:69-77.
Lauer RM, Lee J, Clarke WR. Factors affecting the relationship between childhood and adult cholesterol levels: the Muscatine Study. Pediatrics. 1988;82:309-318.
Guo S, Beckett L, Chumlea WC, Roche AF, Siervogel RM. Serial analysis of plasma lipids and lipoproteins from individuals 9-21 years of age. Am J Clin Nutr. 1993;58:61-67.
Morrison JA, deGroot I, Edwards BP, Kelly KA, Mellies MJ, Khoury P, Glueck CJ. Lipids and lipoproteins in 927 schoolchildren, ages 6 to 17 years. Pediatrics. 1978;62:990-995.
Pagnan A, Ambrosia GB, Vincenzi M, Mormino P, Maiolino P, Gerin L, Barbieri E, Cappelletti F, Dal Palu C. Precursors of atherosclerosis in children: the Cittadella Study: follow-up and tracking of total serum cholesterol, triglycerides, and blood glucose. Prev Med. 1982;11:381-390.
Berenson GS, Srinivasan SR, Cresanta JL, Foster TA, Webber LS. Dynamic changes of serum lipoproteins in children during adolescence and sexual maturation. Am J Epidemiol. 1981;113:157-170.
Cobbaert C, Deprost L, Mulder P, Rombaut K, Gijsels G, Kesteloot H. Pubertal serum lipoprotein(a) and its correlates in Belgian schoolchildren. Int J Epidemiol. 1995;24:78-87.
Webber LS, Srinivasan SR, Wattigney WA, Berensen GS. Tracking of serum lipids and lipoproteins from childhood to adulthood: the Bogalusa Heart Study. Am J Epidemiol. 1991;133:884-899.
Frerichs RR, Webber LS, Srinivasan SR, Berensen GS. Relation of serum lipids and lipoproteins to obesity and sexual maturity in white and black children. Am J Epidemiol. 1978;108:486-496.
Tell GS, Mittelmark MB, Vellar OD. Cholesterol, high density lipoprotein cholesterol, and triglycerides during puberty: the Oslo Youth Study. Am J Epidemiol. 1985;122:750-761.
Davis CE. The effect of regression to the mean in clinical and epidemiological studies. Am J Epidemiol. 1976;104:493-498.