Objective. To determine the immediate effects of two types of elementary school-based interventions on children with multiple cardiovascular disease (CVD) risk factors.
Design. Randomized, controlled field trial.
Setting. Conducted in 18 randomly selected elementary schools across North Carolina.
Study Participants. Four hundred twenty-two children age 9 ± 0.8 years with at least two risk factors at baseline: low aerobic power and either high serum cholesterol or obesity.
Intervention. Both 8-week interventions consisted of a knowledge and attitude program and an adaptation of physical education. The classroom-based intervention was given by regular teachers to all children in the 3rd and 4th grades. The risk-based intervention was given in small groups only to children with identified risk factors. Children in the control group received usual teaching and physical education.
Outcome Measures. The primary outcome measure was cholesterol; additional measures were blood pressure, body mass index, body fat, eating and activity habits, and health knowledge.
Results. Both interventions produced large reductions in cholesterol (−10.1 mg/dL and −11.7 mg/dL) compared with a small drop (−2.3 mg/dL) in the controls. There was a trend for systolic blood pressure to increase less in both intervention groups than in the controls. Both intervention groups had a small reduction in body fat and higher health knowledge than the control group.
Conclusions. Both brief interventions can improve the CVD risk profile of children with multiple risk factors. The classroom-based approach was easier to implement and used fewer resources. This population approach should be considered as one means of early primary prevention of CVD.
- CVD =
- cardiovascular disease •
- BMI =
- body mass index •
- BP =
- blood pressure •
- pVO2max =
- predicted aerobic power •
- SDR =
- survey data regression •
- MANOVA =
- multivariate analyses of variance •
- SBP =
- systolic blood pressure •
- DBP =
- diastolic blood pressure
There is a growing recognition that the atherosclerotic process can begin in youth, with autopsy evidence confirming that adult atherosclerotic plaques can originate in the first 2 decades of life.1-5 Fatty streaks and atherosclerotic lesions have been found at postmortem in the aorta and coronary arteries of 6- to 30-year-olds; these were related to antemortem cardiovascular disease (CVD) risk factors such as smoking, elevated serum cholesterol, and high body mass index (BMI).3-5 Multiple studies show that risk factors for coronary artery disease and stroke are present in children and adolescents, and that these tend to track, or remain in the same quintile of risk, through adulthood. Risk factors that have been shown to track are lipids,6-8 obesity and blood pressure (BP),8 9 physical inactivity,10-12and smoking.11
Although risk factors are present in children and adolescents, whether these conditions should be treated or even evaluated is controversial. Arguments usually center on screening for cholesterol. Some recommend screening,13-15 whereas others criticize it because of concerns for labeling children prematurely or for beginning a course of medication that could last 50 years or more and cause side effects, perhaps without changing life expectancy.16 17 Although there is little resistance to life-style interventions aimed at reducing obesity, hypercholesterolemia, and hypertension, and at increasing physical activity in children, we do not know the best way to deliver such interventions to children with these risk factors.
Many studies of children with CVD risk factors have tested interventions aimed at only one risk factor such as obesity,18-23 BP,24 25 physical inactivity,11 26 27 or dyslipidemia.28-30However, because these risk factors may occur together,6 31it is logical to test interventions aimed at affecting more than one risk factor. Several large-scale, school-based studies have examined the effects of multi-risk factor interventions on all of the children using a population approach.32-37 Other studies have presented results of a school-based intervention on a subset of children with positive risk factors for CVD, using a risk-based approach.38-42
Only a few studies have incorporated both the population-based and risk-based approaches.43 44 One study, done with adults in South Africa, has specifically reported the effects of a high-risk versus a population-based intervention for risk factors for CVD.45 Although the relative benefits of a population-based versus a risk-based intervention have been discussed,46-48no studies have systematically examined their relative effects in a sample of children with multiple CVD risk factors.
We have previously compared the effects of one of the interventions (the population level, classroom-based approach) on all children, regardless of risk status.33 The purpose of this article is to compare interventions for children with multiple risk factors for future CVD. One intervention is a risk-based, individualized program given only to children with at least two identified risk factors; the other is a classroom-based intervention, in which all children received the intervention but we examined only the effects on the subset of children with the same risk factor profiles as in the risk-based group. The control group is composed of children with similar risk factor profiles who received their usual health and physical education curriculum. Thus, this study determines the relative effects of a classroom-based intervention and an individualized, risk-based intervention on children with multiple CVD risk factors.
Setting and Sample
The Cardiovascular Health In Children study was conducted in 18 schools or school clusters in North Carolina. (A total of 21 schools actually participated in the study. However, several small schools in geographic proximity were pooled and treated as single schools in the sampling frame to create similarly-sized sampling units [two such units were in the selected sample].) To test differences by region and setting, half of the schools chosen were in rural settings and half in urban settings. Urban schools were located in cities with a population of at least 50 000. Rural schools were in towns with populations of less than 2500 that were located in counties with no cities of more than 50 000. In addition, this state has three distinct geographic regions and the rate of heart disease mortality is lower in the western mountain region than in the eastern coastal plain or the central piedmont.49 Thus, the study design incorporated geographic region. Study schools were randomly selected from 33 schools that agreed to participate and met the criteria for being in rural or urban settings. Then three school clusters were randomly selected within each of six strata or blocks (defined by the three state geographic regions and setting of rural or urban) and randomized to one of the two intervention groups or to the control group. Although setting and region were factors in the overall study design for better generalizability, differences at that level are not part of this report.
Overall participant inclusion criteria were: assignment to the 3rd or 4th grade regardless of age; ability to read and write English; no mental, emotional, or physical handicap identified by parents or teachers; no chronic illness such as diabetes or moderate to severe asthma reported by parents, teachers, or child; ability to participate in an exercise program; and at least one relative available to respond to a questionnaire about family history. In each consent packet sent to parents of all 3rd and 4th graders at the participating schools, the cover letter and consent form asked parents not to grant permission for their child to participate if he or she did not meet the inclusion criteria. Some parents ignored this request and those children were discovered during baseline testing and subsequently removed from the study. Because of confidentiality issues with the North Carolina Department of Public Instruction, we were unable to determine how many students actually met our inclusion criteria and therefore the exact size of our potential pool. However, most (60.4%) of the entire group of 3rd and 4th graders in the 18 study schools participated in the study.
An additional inclusion criterion for this report was the presence of low aerobic power (an analog for physical inactivity) and at least one of these other CVD risk factors: high serum cholesterol or obesity. Children were determined to be at-risk because of physical inactivity based on predicted aerobic power (pVO2max) results. Boys with a pVO2max of <39 mL/kg/minute and girls with a pVO2max of <34 mL/kg/minute were considered at risk.50 These cut-points were derived from a meta-analysis of available data on fitness levels in children in the United States.50 Children with a total cholesterol greater than the 75th percentile for age and sex using Lipid Research Clinic standards were considered to have high cholesterol.51Children were classified as obese if they had both a BMI and a sum of triceps and subscapular skinfolds more than the 90th percentiles for age and gender.51 52 Thus, this article only includes those who had either obesity and low fitness, or high cholesterol and low fitness, or all three.
The classroom-based intervention was taught to all the 3rd and 4th graders in the schools randomly assigned to that intervention. Regular classroom teachers used the American Heart Association Lower and Upper Elementary School Site Program Kits53 to provide instruction twice a week for 8 weeks. Content included information about selecting heart-healthy foods, the importance of getting regular physical exercise, the dangers of smoking, and ways to combat pressure to smoke. In addition, children received an aerobically-oriented physical activity program three times a week, taught by physical educators using lesson plans developed by the exercise physiologist coinvestigator and derived from standard physical education curriculum texts. Examples of the aerobic activities used are: jumping rope to music, endless relay, parachute games, and aerobic dance. These knowledge and activity interventions are completely described in another article.33
The children in schools assigned to the risk-based intervention received an intervention only if they were found at baseline to have at least one of the three CVD risk factors described above and/or smoking risk (defined as living with a parent who smokes or only quit in the last 2 years). As a result of the child risk for smoking being based on parents' behavior and the small percentage of children actually reporting smoking at this age (0.7% overall), it was decided not to include those at risk for smoking in this report. Thus, for this report, two risk-based interventions were used: 1) physical activity classes, given to those at-risk because of low aerobic capacity and/or obesity (all children reported herein); plus 2) nutrition classes, given to children who were at risk as a result of elevated cholesterol levels and/or obesity. All interventions for the risk-based groups were conducted in groups of 5 to 8 children, throughout an 8-week period, during regular school hours. The physical activity intervention was taught three times a week by physical educators, using the same activities as those in the classroom-based intervention, but involving only 5 to 8 at-risk children at a time. The nutrition and smoking classes were taught by registered nurses twice a week.
Children in the control group schools did not receive any intervention but had their regular health and physical education classes. However, parents of children in the control group, as well as those in the two intervention groups, received a written report of their child's physical testing results ∼4 weeks after each test.
Variables and Their Measurement
The primary outcome variables were health behaviors, knowledge, and health status. In addition, selected family influences (parental education and smoking) and child developmental/personal characteristics (age, grade, race, and sex) from the Bruhn and Parcel Development of Positive Health Behavior model54 were measured as potential explanatory variables.
The health behaviors of the child were physical activity and eating habits. For physical activity score the children filled out a revised form of the Know Your Body Health Habits Survey,55 which we adapted for grades 3 and 4. For eating habits, children were given a short list of high- and low-fat foods and asked to circle how often they ate each one (not much, some, or a lot). To encourage truthful answers, children were reassured that parents and teachers would not receive this information.
Health knowledge was assessed only at posttest by a questionnaire based on the Heart Smart test.56 Because the children were such young readers at baseline, we elected not to overburden them with questionnaires at pretest, knowing that randomization at the school level should balance baseline differences between groups. We eliminated some questions that were not relevant to our study but added 5 others about the dangers of smoking. Thus, the Healthy Heart Knowledge test was a 25-item, multiple-choice questionnaire covering nutrition, exercise, smoking, and general heart health.
Health Status Outcomes
All measurements were taken by a team of four trained research assistants. (For details on all instruments, data collection, and quality control procedures, please write to the corresponding author to request our Technical Procedures Report.) Standardized training and interrater reliability testing were conducted. BP was measured on the right upper arm using a calibrated Baumanometer mercury sphygmomanometer and appropriately-sized cuffs according to American Heart Association recommendations for children.57 BP was taken and recorded at least twice at the first, fourth, and fifth Korotkoff phases according to a standardized protocol. The mean of two (or three, if taken) BP measurements was used for analysis. Total serum cholesterol was measured nonfasting with the Reflotron (Boehringer Mannheim Diagnostics, Indianapolis, IN), using rigorous internal and external quality control procedures. pVO2max was determined using a Bodyguard Professional Cycle Ergometer (ΔGLΔND DBS A.S., Monark 543282, Varberg, Sweden), adapted for children, and the Eurofit protocol for children.58 Polar Pacer heart rate monitors (Country Technology, Gay Mills, WI) were used to obtain heart rates continuously during the cycle test. Triceps and subscapular skinfolds were measured in mm on the right side of the body using Lange skinfold calipers (Cambridge Scientific, Cambridge, MD). The skinfold sites were located and measured at least twice according to National Health and Nutrition Evaluation Studies procedures,59 with means at each site summed for analyses. Weight was measured to the nearest 0.1 kg using a calibrated balance beam scale (Detecto Scales, Inc, Webb City, MO) and height was measured to the nearest 0.1 cm using a stadiometer (Perspective Enterprises, Portage, MI). Both measurements were taken with children clothed but shoeless. BMI was calculated as weight divided by squared height (kg/m2).
Parental education, used as an indication of family socioeconomic status, was determined from parents' questionnaires. Highest grade the father completed in school was used unless it was unavailable, in which case the mother's was used. Parental smoking items were included on the parents' questionnaires.
Data Collection Procedures
After obtaining written parental consent and child assent, baseline data were collected from the children in the schools following a systematic geographic pattern. Parental data were collected via mailed questionnaires. The intervention began after baseline testing; posttest data were collected in the same geographic order within 2 weeks of completion of the 8-week intervention.
Questionnaires were administered by the team of trained research assistants. Children completed questionnaires in groups of 10 to 25 in classrooms, libraries, or cafeterias during their regular class periods. One to 3 days after administration of questionnaires, physiologic data were collected in empty classrooms, gyms, or media rooms. Height, weight, and skinfolds were measured at station 1; cholesterol and BP at station 2; and aerobic power at station 3. All variables were measured in the same sequence for each child. After completing all the assessments each child received a small prize. Baseline and posttest procedures were identical and all measures were taken by the same team of research assistants.
Distributions of gender, race (black/white/other), and highest parental education completed (<high school graduate; high school graduate or some college; college graduate) were examined for differences between the intervention groups with χ2tests and adjusted Mantel-Haenszel χ2 tests. Because there was evidence of demographic differences by rural versus urban locale at baseline, adjustments were made for school locale in this analysis. Age differences among intervention groups were examined with analysis of variance and nonparametric Kruskal-Wallis tests.
Intervention Effects at School Level
Two types of analyses were done: one set of analyses at the school level, and another at the individual subject level. Because schools were randomly chosen within six region-by-locale strata and then children were enrolled at each school, analyses were performed at the school level using survey data regression (SDR), which is related to generalized estimating equation models.60 This procedure accounted for the study design as a stratified cluster sample. Sampling weights based on the inverse probability of selecting a school from eligible volunteer schools within each stratum were used. This survey data regression, performed with the Survey Data Analysis software package from Research Triangle Institute (SUDAAN; Cary, NC), accounted for the underlying correlation structure of the stratified cluster sample study design. Models were adjusted for significant demographic factors.
In SDR analyses, a step-down procedure was used to control for Type I error, which is finding a statistical difference when one is not actually present, or false significance. First, the overall intervention effect for differences between any of the three groups was assessed. If this was significant, then each of the two intervention groups was compared with the control group with a Bonferroni adjustment (α* = α/2 = 0.025) for the two pairwise comparisons. If either comparison versus the control group was significant at the Bonferroni adjusted level, then the risk-based group was compared with the classroom-based group. All test results are provided parenthetically in the tables for completeness, but are not used for formal inference unless previous steps were statistically significant.
Intervention Effects at the Individual Level
Individual-level analyses were done as a supplement to the SDR to help clarify trends. They might be useful for hypothesis generation or for providing support for additional studies. Because individual variation in biological attributes is high, school-level analyses could have low power to detect important clinical changes in these types of variables. For the individual or subject-level analyses, separate multivariate analyses of variance (MANOVA) assessed the short-range intervention effects as well as other potentially significant demographic factors (gender, race, age, and parental education). One analysis examined changes in cholesterol, skinfolds (log of 10 times the ratio of posttest to pretest), systolic blood pressure (SBP), diastolic blood pressure (DBP), BMI, and aerobic power (pVO2max) (physiologic responses). The second MANOVA examined changes in self-reported activity and eating habits (behavioral responses), as well as knowledge scores at posttesting. All MANOVA used Wilks' λ statistics. The MANOVA was adjusted for study design (region and locale within region).
Because significant MANOVA were found, the intervention effects were examined for each outcome variable using analyses of covariance or analyses of variance adjusting for study design (region and locale within region) and for potentially significant demographic factors (gender, race, parental education, and age at pretest). For the short-range changes reported here (posttest − pretest), analysis of covariance also adjusted for mean pretest values; posttest knowledge scores had no pretest covariates. A step-down procedure was also used here to control type I error in formally interpreting the intervention group comparisons. First, Dunnett's test (at α = 0.05) was used to compare each intervention group to the control group. If either comparison versus the control group was significant, then the risk-based group was compared with the classroom-based group. If neither intervention versus control was significant for a variable, no inferential comparisons were made between the two intervention groups for that variable. Again, the results of these tests have been provided parenthetically in the tables for completeness.
Baseline screening at all sites included 2207 children; posttest evaluation included 2103 of those. The proportion of the 2103 children in the full study with risk factors is shown in Table1. Overall, 43.2% of the children had at least one risk factor. Subjects of this report were the 422 children (19% of the total group) who, at baseline, had low aerobic power and at least one or both of the following additional CVD risk factors: high total cholesterol, or obesity. This group includes children from the risk-based intervention schools as well as those in the two other groups with similar risk profiles who would have been assigned to both the nutrition and physical activity classes if they had been in the risk-based intervention schools (ie, the would-have-been-treated children). The mean age of the children was 9.0 years (standard deviation = 0.8 years) at pretest and 9.3 years (standard deviation = 0.8 years) at posttest. Demographic variables are shown by intervention group in Table 2.
Age, gender, and grade did not vary significantly by intervention group. Racial distribution was significantly different among group assignments (P = .047); there were fewer white children in the control group. There was a significant difference in intervention group by region (P = .011); a greater proportion of those in the coastal area were in the classroom-based intervention, whereas the majority of children in the piedmont region were in the risk-based intervention or control group.
Baseline values of risk factors (unadjusted) are shown for at-risk children in all three groups in Table 3. Cholesterol among children ranged from 111 to 264 mg/dL in the three groups; means were 181.6 mg/dL to 178.4 mg/dL. Mean BMI ranged between 22.9 to 23.5 kg/m2 and BP was ∼110/70 mm Hg. Aerobic power, as expected because of inclusion criteria, was low in all groups, ranging from 29.6 to 30.6 mL/kg/minute. Simple, unadjusted change scores are shown for all groups in Table4.
The MANOVA testing the six physiologic variables for intervention effects and assessing for potentially significant factors were significant for intervention effect (P = .021) and for the demographic variables, race (P = .018) and age (P = .009). For the behavioral measures, MANOVA were also significant for intervention (P = .004) and also for the demographic factors of gender (P < .001), race (P < .001), and parental education (P = .003). Thus, subsequent analyses assessing intervention effects adjusted for those demographic factors. Results of the school-level SDR analyses are shown in Table5, with results of the individual analyses given in Table 6. Because each of these analyses statistically adjusted the results differently, the changes shown in Tables 5 and 6 differ slightly from each other and from the unadjusted changes shown in Table 4.
The unadjusted changes in cholesterol were marked, as cholesterol decreased −10.1 mg/dL in the risk-based intervention group, −11.7 in the classroom-based intervention group, and −2.4 in the control group (Table 4). School-level SDR analyses showed an overall significant difference in cholesterol among intervention groups (Table 5). Cholesterol in the risk-based group was 7.9 mg/dL lower than the controls, and in the classroom-based group it was 8.9 mg/dL lower than the control group. However, differences between the intervention groups and the control group were not statistically significant after the Bonferroni adjustment. In the individual-level analyses, cholesterol in the risk-based group was 7.7 mg/dL lower than in the control group, and in the classroom-based group it was 9.3 mg/dL lower than the control group (Table 6). Each of these were significantly different from the controls, although there was no significant difference between the two intervention groups.
The expected growth-related increase in SBP differed among the three groups. Unadjusted changes were 2.9 mm Hg (classroom-based), 3.3 (risk-based group), and 5.7 (control group) (Table 4). The SDR analyses showed the classroom-based group increased significantly less than the control group (adjusted mean difference from the control group, −2.8 mm Hg); the risk-based group increase, which was also less than the control group (difference from control −2.0 mm Hg), only approached significance (Table 5). Again, the two intervention groups were not significantly different from each other. At the individual level, SBP in the risk-based intervention group was −2.4 mm Hg less than in the control group; this was significant. The classroom-based group was −2.0 mm Hg less than the control group, but this was not statistically significant (Table 6). The difference between the two intervention groups (classroom-based and risk-based) was not significant.
Unadjusted differences in DBP were 5.0 (risk-based group), 4.8 (classroom-based group), and 7.2 (control group) (Table 4). SDR analyses did not show any significant differences between the intervention groups despite mean increases that were less than the controls (−2.8 in classroom-based group and −2.2 in risk-based group) (Table 5). Also, in individual-level analyses, mean DBP rose more in the control group than in either intervention group. These differences were suggestive of significant overall intervention effects (Table 6).
Other Physiologic Variables
Unadjusted changes in aerobic power showed small increases in all three groups (3.7 mg/kg/minute for the risk-based group, 4.4 for the classroom-based group, and 2.7 for the control group), and also small changes in BMI and skinfolds (Table 4). There were no significant differences in the three groups in changes in BMI in either the individual or SDR analyses. The differences in pVO2max and skinfolds were not significant in SDR analyses (Table 5). In the individual-level analyses, pVO2max was significantly higher for those in the classroom-based group (1.67 mL/kg/minute higher) than the controls. Skinfolds were slightly but significantly lower than the controls in both intervention groups (Table 6).
Unadjusted changes in physical activity showed an increase for each intervention group (0.4 for the classroom-based group and 6.4 for the risk-based group) and a decrease for the control group (−1.0) (Table 4) This difference was not significant in SDR analysis. However, with the individual-level analyses physical activity score for the classroom-based intervention group was significantly higher (+8.27) than in the control group (Table 6). Differences among the three groups in eating a high-fat diet were not significant in either analysis.
Unadjusted total knowledge score at posttest was 64% correct for the risk-based group, 68% for the classroom-based group, and 60% for the control group (Table 4). SDR analyses showed significant overall intervention differences in knowledge, attributable to a significant difference between the classroom-based group and control group (adjusted mean difference, 7.2%) (Table 5). SDR analyses for total knowledge, as well as nutrition and exercise knowledge, were very similar to the individual-level results. The mean percent correct was significantly higher in both the risk-based group and the classroom-based group than in the control group but the classroom-based group had a higher mean percent correct (Table 6).
Unadjusted scores in nutrition knowledge at posttest were 75% for the risk-based group, 76% for the classroom-based group, and 60% for the controls (Table 4). SDR found a significant mean difference for both intervention groups in nutrition knowledge, with the risk-based group 7.0% higher than the controls, and the classroom-based group 7.5% higher than the controls (Table 5). Individual level, posttest nutrition knowledge subscale scores had results similar to SDR (Table6).
Exercise knowledge subscale scores at posttest (unadjusted) were all low: 43% for the risk-based group, 50% for the classroom-based group, and 42% for the control group (Table 4). SDR analyses showed the risk-based group was not significantly different from the control group (0.6%), but the classroom-based group was significantly higher than the control group (8.2%) (Table 5). The difference between the two intervention groups was also significant. Results at the individual level were similar (Table 6).
There were significant improvements in some risk factors in this controlled, school-based study of children with multiple risk factors for CVD. Both the risk-based and the classroom-based interventions produced large decreases in total serum cholesterol, −10.1 mg/dL and −11.7 mg/dL, respectively, although the control group only had a −2.5 mg/dL change. There was also a trend for SBP and DBP to increase less in intervention than in control children. Further, in the individual-level analyses, there was an increase in aerobic power in the classroom-based group and a decrease in skinfolds in both intervention groups. Posttest health knowledge was significantly higher in both intervention groups than in the control groups and there was an increase in physical activity in the classroom-based group.
Because health knowledge was only measured at posttest and no baseline statistical comparisons are reported for the other pre- to postvariables, it could be argued that differences between groups are merely because of unchanged baseline differences. For the knowledge test, because the children were such young readers at baseline, we elected not to overburden them with questionnaires at pretest and took the risk of being criticized for not measuring at baseline as well as at posttest. For knowledge and the other outcome variables, randomization to treatment at the site level is expected to balance variables, both measured and unmeasured, among groups. Also, knowledge was compared with a control group.
In a previous article examining all children in the classroom-based and the control groups, we reported that children in the intervention group had lower total cholesterol, greater knowledge, and a significant increase in self-reported physical activity than children in the control group.33 Further, intervention children had a greater reduction in body fat and increase in aerobic power and a smaller rise in diastolic BP than control children. When compared with the total group of Cardiovascular Health in Children study participants, the responses in this group of at-risk children to the interventions were greater for cholesterol, SBP, and self-reported physical activity, and similar for the other variables.
The large drop in total serum cholesterol in both intervention groups after this brief, 8-week intervention is rather surprising. The reduction in cholesterol in these at-risk children was nearly twice as great as that we previously reported for all of the children in the schools, in which those in the classroom-based intervention group had a mean reduction of −5.27 mg/dL in total cholesterol.33Although the magnitude of the cholesterol change could theoretically be attributable to the phenomenon of regression to the mean, all three groups would have the same regression toward the mean, so statistical comparisons to the control group and other between-group comparisons are still valid. Although the size of the effect may be overestimated, this remains an internally consistent way to statistically judge interventions versus the control and to each other.
In addition, our findings are similar to some of the other studies on reducing hyperlipidemia in children. Alexandrov et al44studied 12-year-old Russian boys annually for 3 years. The children were in two districts, intervention and reference; those in the intervention district received a two-level intervention. Children with dyslipidemia or high BP or obesity were invited to attend a single counseling session with their parents and the families were given booklets on nutrition. The mean total cholesterol of the boys in the intervention district decreased during 3 years from 175.7 mg/dL at baseline to 166.1 and 157.0 at years 2 and 3, whereas the mean cholesterol of boys in the reference district went from 166.8 mg/dL at baseline to 174.2 at year 2 and 165.5 at year 3. Thus, their long-term changes were similar to our short-term results. The large Dietary Intervention Study in Children found less marked reductions in total cholesterol, which was a secondary outcome.28 At baseline, cholesterol was 200.0 mg/dL in both intervention and control groups; after 1 year it was significantly lower in the intervention group (191.4 mg/dL) than in the control group (197.4). This decrease, the result of intensive dietary-only interventions, was less than the reduction of >10 mg/dL we saw in both of our intervention groups, suggesting that an activity component should be part of an intervention to reduce cholesterol. Other studies with at-risk children, done through clinics, report significant changes in low density lipoprotein-cholesterol but do not report changes in total cholesterol.29 30
Our study produced only small reductions in body fat and no change in BMI, perhaps because of its short duration or because of confounding by multiple examiners. However, our team of examiners received standardized training on height, weight, skinfolds, and BP measurements before data collection and the same team conducted both baseline and posttesting. In addition, they were subjected to repeated interrater reliability testing and results indicated that differences between raters were minimal, so examiner technique is not likely to be a confounding factor.
Not only duration of the intervention but also the subcomponents affect the success of obesity interventions. A meta-analysis of 41 intervention studies to reduce obesity in children found that interventions that incorporate both a behavioral and knowledge component, particularly those that include exercise in the intervention, are likely to produce weight loss; also, children that weighed more at the beginning were more likely to lose weight.21 Resnicow46 also reviewed a variety of high-risk interventions for obesity in children, with treatment times varying from 10 weeks to 18 months. He reported significant treatment effects for intervention children; however, only two of the six studies had true random assignment to intervention group. Those studies also showed that treatment effects were larger for heavier children. Epstein et al19 say that their series of studies to reduce obesity in children indicate that results are better if there is direct involvement of at least one parent, if physical activity is part of the program, and if family and friend support is positive. Our interventions did include physical activity, but did not involve parents. In addition, our interventions did nothing to manipulate diet other than provide information to the child about wise food choices. Because it is likely that the child does not have much control over what food is purchased or prepared in the household, this is probably a factor in the lack of change in this variable. Also, the intervention may not have been long enough to produce a greater decrease in obesity especially because we did not include any dietary manipulation and we used a strict definition for obesity. We wanted to be sure we targeted truly overly fat children so we incorporated skinfold measurements with BMI measurements. Although our definition has not been used before and is extremely stringent, we still found ∼26% of our overall sample to be obese.33 Thus, the fact that the interventions did not impact on these obese children is most likely a result of several factors in combination, but particularly the lack of parental involvement, especially with respect to diet.
The BP results are rather intriguing because there are slight differences in results from the two different types of analyses. Both showed an overall intervention effect, with SBP rising more in the control children than the other two groups. However, the individual-level analyses showed that the difference was greater for the risk-based versus control children, whereas the SDR indicated that the difference was larger between classroom-based group versus the control group. Diastolic pressure rose more than SBP, and there were no significant intervention differences in DBP. There are few studies of interventions to reduce BP in children with which to compare results. A feasibility study reported that girls with hypertension will take part in special physical education classes designed to include 30 minutes of aerobic activity, but did not provide intervention results.24 Hansen et al25 studied 137 children aged 9 to 11 years; half were hypertensive and half normotensive. Children were randomly assigned within those categories to receive an 8-month intervention (3 extra days of physical education at school each week) or to serve as controls. In the intervention groups, as compared with the controls, both SBP and DBP fell by 6.5 and 4.1 mm Hg, and maximum oxygen uptake increased. Both changes were greater than we found, most likely because of the greater length of the intervention. The Heart Smart program included a small component for high-risk 4th- and 5th-grade children and their parents. Interventions were given to 12 of the 19 families in this small controlled study.40There were more changes in parents than in children. The only change noted in the children was a tendency toward decreased DBP during the time period in the intervention children and an increase in control children.
There are few school-based studies of effects of a physical activity intervention in at-risk children. A small study by Ignico and Mahon39 examined the effects of a 10-week after-school physical fitness program on 18 low-fit children; there was a comparison group of 10 children. At pretest the age and weight of the children were comparable with ours, but their pVO2max was higher. There was a very small drop in total cholesterol in the intervention children (0.7 mg/dL) and an increase in the comparison group (+9.3 mg/dL); these differences between groups were similar in magnitude to our study but different in direction, because there was a reduction in cholesterol in all of our groups, with a much larger decrease in the two intervention groups. Ignico and Mahon39 found no change in body fatness and no significant increase in pVO2max, whereas we found small decreases in skinfolds in both intervention groups and an increase in pVO2max in the classroom-based intervention, possibly because of our larger sample size.
Very few studies have compared a population approach with a high risk approach. One example is the Coronary Risk Factor Study, which examined interventions in three white communities in South Africa.45Two communities received a variety of media-based messages encouraging physical activity and heart-healthy eating and restaurants cooperated by featuring low-fat and low-salt foods. A second community received all of the above plus a variety of other interventions that included small-group interventions to high-risk individuals and active follow-up of hypertensives. Children in both intervention communities were also told the results of their initial screening. The control community received only this screening information. The program lasted 4 years. Total cholesterol was reduced approximately the same in all three communities, but BP and smoking decreased more in the intervention areas. The authors concluded that a small mass-media-based program can improve the community CVD risk profile, but adding components of a high-risk strategy does not necessarily lead to further benefit to the community or to the high-risk individuals. Although their study was community-based and focused on adults, and our study was school-based and focused on children, the results and conclusions are remarkably similar; however, they used only 1 community for each group, whereas we had six schools in each group.
Other researchers have suggested that provision of the results of a school-site screening for CVD risk factors can actually serve as a motivation for action and thus may not be an appropriate control.46 If this is true, our study would have been biased toward null results because we provided feedback to all parents, including those in the control group. However, we still had strong differences in cholesterol and knowledge and small differences in SBP in both intervention groups as compared with controls.
This article provides fuel to the debate over whether to use a population approach or a risk-based approach for promoting cardiovascular health in children and adolescents. The high-risk strategy is familiar to physicians and nurse practitioners and is relatively easy for them to incorporate into their practice.48 It can also be useful when working with high risk families,40 and has been shown to be an effective method for dealing with small groups of children with obesity,20 21 high BP,25 and hyperlipidemia.28 However, studies have shown that pediatricians and family practice physicians do not routinely assess family history for CVD and seldom emphasize CVD risk factors in their routine practice with children and adolescents.61 62
Rose63 pointed out that “medical thinking has primarily been concerned with responding to the needs of sick individuals.” However, most factors related to illness are located on a continuum, so those that distinguish sick (top end of the distribution) from well (lower end of the distribution) are probably a mix of genetic and environmental factors that vary in the population. Although it may be difficult to locate all the relevant genetic factors, if we can identify environmental factors that impact on a large proportion of the population, we may be able to induce a large change in the health of many people by making a relatively small but consistent change in the environment.47 An example of such an environmental change is this school-based intervention that provided classroom instruction and a physical activity program to children.
We have shown that both a risk-based and a population-based environmental intervention can produce an improvement in the CVD risk profile of children with multiple risk factors. Because both approaches seem to work, one could choose either if both were equally feasible and had similar costs. In practical terms, the risk-based intervention was much more difficult to incorporate into the school day, required hiring of additional nurses to teach the children, and thus, used more resources.
This relatively inexpensive population or classroom-based intervention was at least as effective in improving the CVD risk profiles of the children with multiple CVD risk factors as the more intensive, risk-based intervention. The classroom-based approach, even for these at-risk children, is preferred because it is easier to implement, and fits logically into the school curriculum. Also, it avoids stigmatization of at-risk children, uses positive peer pressure, and avoids the potential problem of misclassifying children at risk. It has the added benefit of providing some benefits to children from all risk groups, even those at low risk.33 The findings of this study are consistent with the recent study of Tosteson et al,64 who concluded that population-based programs can lengthen life and save resources, and recommend that “population-wide programs should be a part of any national health strategy to reduce coronary heart disease.” The classroom-based (population) strategy presented here should be tested more widely to provide further information about its potential for primary prevention of CVD in adulthood.
This work was supported by federal Grant 5R01 NR01837 from the National Institute of Nursing Research, National Institutes of Health.
- Received September 25, 1997.
- Accepted January 12, 1998.
Reprint requests to (J.S.H.) School of Nursing, CB# 7460, 506 Carrington Hall, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7460.
- Research Group (Pathobiological Determinants of Atherosclerosis in Youth)
- Raitakari OT,
- Porkka KVK,
- Rasanen L,
- Ronnemaa T,
- Viikari JS
- Dennison BA,
- Straus JH,
- Mellits ED,
- Charney E
- American Academy of Pediatrics
- Haddock CK,
- Shadish WR,
- Klesges RC,
- Stein RJ
- Ewart CK,
- Loftus K,
- Hagberg JM
- Hansen HS,
- Froberg K,
- Hyldebrandt N,
- Nielsen JR
- Kuh DJL,
- Cooper C
- Polonsky SM,
- Bellet PS,
- Sprecher DL
- Shannon BM,
- Terhskovec AM,
- Martel JK,
- et al.
- Harrell JS,
- McMurray RG,
- Bangdiwala SI,
- Frauman AC,
- Gansky SA,
- Bradley CB
- Whitaker RC,
- Wright JA,
- Finch AJ,
- Psaty BM
- Rowland TW,
- Boyajian A
- ↵Heiss, G. Cardiovascular Disease Mortality in North Carolina. In: The 62nd American Heart Association Annual Meeting, Delegate Assembly and Scientific Sessions; November 13–16, 1989; New Orleans, LA
- ↵Anonymous. The Lipid Research Clinics Population Studies Data Book: Volume I: The Prevalence Study 1980. Washington DC: US Government Printing Office; NIH publication No. 80–1527
- Johnston FE,
- Hamill PV,
- Lemeshow S
- ↵American Heart Association. The AHA Schoolsite Program: Heart Decisions Middle School Package Technical Report. Dallas, TX: American Heart Association; 1989
- ↵Klissouras V, Tokmakidis S. Evaluation of physical fitness of school children: The Eurofit test. In: Proceedings of the XVth Meeting of the International Council for Physical Fitness Research. Olympia Seminar; August 21–25, 1982:198–212; Magglingen, Switzerland
- Najjar MF,
- Rowland M
- Jennings RB,
- Leon SP
- Arneson T,
- Luepker R,
- Pirie P,
- Sinaiko A
- ↵Rose, G. Strategies of prevention: the individual and the population. In: Coronary Heart Disease Epidemiology. Oxford, England: Oxford Medical Publications; 1992:311–324
- Tosteson ANA,
- Weinstein MC,
- Hunink MGM,
- et al.
- Copyright © 1998 American Academy of Pediatrics