close

Вход

Забыли?

вход по аккаунту

?

Childhood obesity: Effects of physical exercise and dietary behavior

код для вставкиСкачать
ABSTRACT
CHILDHOOD OBESITY: EFFECTS OF PHYSICAL EXERCISE
AND DIETARY BEHAVIOR
By
Brandt Shirer
May 2010
In the early-21st century, we have a well-documented obesity epidemic in the
United States, and it is found increasingly in much of the industrialized world. The
rise in the prevalence of childhood obesity in particular is most striking and
disconcerting for our future. This study seeks to examine the relationships between
physical activity, dietary behavior, and BMI in children and adolescents using a
sample data set of 5,101 survey participants in the 2005-2006 National Health and
Nutrition Examination Survey (NHANES). In general, the findings are consistent
with the findings in the published literature. There does appear to be a statistically
significant relationship between physical activity and BMI; however, the strength of
the relationship and its predictive value are weak. There does not appear to be a
statistically significant relationship between lunches and/or breakfasts eaten at school
and BMI. There does appear to be a statistically significant relationship between BMI
and physical activity combined with dietary behavior, but the linear regression model
tested to have a very weak predictive power. The primary limitations of the study are
1
the broad age range used in both independent variables, and the lack of direct
measurement of physical exercise and dietary behavior used in NHANES. In
conclusion, the data in NHANES is found to be consistent with the evidence in the
published literature, that is, that there is very little evidence to show that increased
physical exercise, decreased sedentary behavior, and/or better dietary intake has a
statistically significant impact on childhood obesity and BMI. What is needed to truly
assess the impact of these variables on BMI is additional scientific evidence: large,
well-designed studies with samples that are representative of the general population,
that include long term follow-up; and interventions which measure direct effect on
physical exercise and diet, and therefore their relationship to BMI.
2
CHILDHOOD OBESITY: EFFECTS OF PHYSICAL EXERCISE
AND DIETARY BEHAVIOR
A THESIS
Presented by the Department of Health Care Administration
California State University, Long Beach
In Partial Fulfillment
of the Requirements for the Degree
Master of Science
Committee Members:
Grace Reynolds, Ph.D. (Chair)
C. Kevin Malotte, Ph.D.
Tony Sinay, Ph.D.
College Designee:
Tony Sinay, Ph.D.
By Brandt Shirer
B.A., 1995, University of California San Diego
May 2010
UMI Number: 1486392
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI
Dissertation Publishing
UMI 1486392
Copyright 2010 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346
TABLE OF CONTENTS
Page
LIST OF TABLES
iv
CHAPTER
1. INTRODUCTION AND LITERATURE REVIEW
1
Overview
Literature Review
Physical Exercise
Dietary Behavior
Physical Exercise and Dietary Behavior
1
4
7
10
13
2. METHODS
17
National Health and Nutrition Examination Survey (NHANES)
Dependent Variable—Body Mass Index (BMI)
Independent Variables—Physical Exercise and Dietary Behavior
Data Analysis Plan
3. RESULTS
17
22
22
25
27
Descriptive Statistics
Bi-Variate Analyses of Corelation
Multi-Variate Analyses of Regression
4. DISCUSSION
27
28
32
34
Overview and Limitations
Findings
Conclusion and Future Direction
REFERENCES
34
35
40
42
iii
LIST OF TABLES
TABLE
Page
1. Bi-Variate Analyses of Corelation
31
2. Multi-Variate Analyses of Regression
33
IV
CHAPTER 1
INTRODUCTION AND LITERATURE REVIEW
Overview
In the early-21st century, we have a well-documented obesity epidemic in the
United States, and it is found increasingly in much of the industrialized world. The
National Center for Health Statistics (2006) reported that nearly one-third of U.S. adults
are obese and two-thirds are overweight or obese. Prevalence rates have skyrocketed
over the past 40 years. The National Health and Nutrition Examination Survey
(NHANES; 2005) showed prevalence of obesity among adults aged 20-74 increased
from 15% in the 1976-80 survey to 32.9% in the 2003-04 survey. The rise in the
prevalence of childhood obesity in particular is most striking and disconcerting for our
future. NHANES (2005) further showed the following increases in obesity prevalence
rates from 1976-80 to 2003-04: 5% to 13.9% in children aged 2-5, 6.5% to 18.8% in
children aged 6-11, and 5% to 17.4% in children aged 12-19.
The current and long-term health impact of the obesity epidemic is significant
and growing. The U.S. Department of Health and Human Services (USDHHS; 2007)
reported that obese individuals have a 10 to 50% increased risk of death from all causes,
compared to healthy weight individuals, contributing to about 112,000 excess deaths
per year. Most of this is due to increased cardiovascular risk. The economic cost
related to obesity is staggering. The USDHHS reported the total cost to be $117 billion
1
($61 billion direct and $56 billion indirect). Ludwig (2007) reported in the New
England Journal of Medicine his four-phased model of the obesity epidemic in the
United States which predicts by mid-century it may shorten life expectancy by 2 to 5
years, an effect equal to that of all cancers combined.
Due to this unprecedented obesity epidemic, much attention was given at all
levels of government, media, and industry to develop an appropriate response. In 2001,
the U.S. Surgeon General issued a report with a call to action to prevent and decrease
obesity and overweight. Since much of the behaviors associated with obesity are
difficult to reverse and begin in childhood, a significant amount of attention focused on
the prevention of childhood obesity. In 2002, the U.S. Congress asked the Institute of
Medicine to develop a prevention focused action plan for childhood obesity. The
Institute of Medicine (IOM; 2004) report made sweeping recommendations for action at
all levels of government, schools, communities, industry, media, and at home by parents
and families.
The response came quickly. The same year, the Federal government passed into
law the Child Nutrition and Women Infant Child Reauthorization Act of 2004 which
mandated that any school participating in the federal school lunch program develop a
comprehensive wellness program aimed at improving the diet and exercise of students.
By 2006, most states had developed comprehensive prevention programs. In California,
Governor Arnold Schwarzenegger provided leadership in his endorsement for the
California Obesity Prevention Plan (CAOPP) by banning all junk food from schools,
extending to high school the law banning sodas in elementary and middle school, and
2
providing funding for fresh fruit and vegetables in school meals (California Department
of Public Health, 2006).
Although well intended, there are a multitude of issues and problems with a
knee-jerk response. The plethora of programs lack oversight at the federal level and
coordination at the state level. In the initial IOM (2004) report, one of the findings was
"presently there is limited experimental evidence regarding the best ways to prevent
childhood obesity and the extent to which various potential factors contribute to weight
gain (Fact Sheet, p. 1). Chaloupka and Johnston (2007), in their introduction to 13
studies about childhood obesity published in the American Journal of Preventive
Medicine, conclude "the growing recognition of the public health and economic
consequences of childhood, adolescent, and adult obesity has led to a variety of policies,
programs, and other interventions to stimulate healthy eating and physical exercise,
often despite the lack of evidence of their impact" (p. S147). Finally, in a systematic
review of childhood obesity prevention programs, Wofford (2008) concluded "Because
the gap between clear articulation of the problem as well as population and the best
strategies to impact the prevention of the problem is evident, health care practitioners
must be involved in well-constructed and evaluation studies that build on the limited
base of current evidence" (p. 5).
In addition to the development of a plethora of programs and initiatives for
which there is no clear evidence of their efficacy, there has been a lack of coordination
and systematic evaluation of the prevention programs. The 2006 IOM follow-up report
assessing the progress of the preceding 2 years found "these interventions, however,
3
generally remain fragmented and small-scale. Moreover, the lack of systematic
monitoring and evaluation has hindered the development of an evidence base to
identify, apply, and disseminate lessons learned" (p. 1). Ludwig (2007) asserts that
there is "lack of anything resembling a comprehensive strategy for encouraging children
to eat a healthful diet and engage in physical activity" (p. 2326).
Literature Review
Within this context, a review of the literature was performed to identify clinical
studies showing the effect of increased physical exercise and/or better diet on childhood
obesity. Medline First Search was utilized, with the following criteria: "randomized
control trial" or "clinical study," within the last 5 years in the United States, English
language only. Due to small number of studies, date was extended to 2000 and the
United Kingdom was included. This resulted in six randomized controlled trials (RCT)
and two prospective cohort or longitudinal studies which were included. In addition,
eight cross sectional and survey studies were included.
The definitions and measurements of obesity and overweight must be identified
to ensure consistency when reviewing the scientific literature. The USDHHS (2007)
defines overweight as "excess body weight compared to set standard" and obesity as
"having an abnormally high proportion of body fat" (p. 3). Various methods exist to
determine if a person's weight is considered overweight or obese. Although the most
technically accurate, percentage of body fat is difficult to obtain, requiring imaging
studies or assessment by professional. The more common method is body mass index
(BMI), which is the index of weight adjusted for height. According to the USDHHS
4
(2007), this is the most commonly used measurement, accepted by the medical and
scientific community.
It is important to note that statistics about overweight and obesity differ over
time. This is because in the past, studies in the United States used the 1959 and 1983
Metropolitan Life Insurance tables of desirable weight-for-height as reference for
overweight. The NHANES II (1976-80) used a statistically derived definition of
"overweight" based on the gender-specific 85th percentile values of BMI for 20 to 29
year olds, which is greater than or equal to 27.4 for women and 27.8 for men.
NHANES II (2006) further defined "severe overweight" based on the 95th percentile
values as greater than or equal to 32.2 for women and 31.1 for men (p. 287).
This issue was put to rest in 1998 when the Clinical Guidelines on the
Identification, Evaluation, and Treatment of Overweight and Obesity in Adults were
issued by the National Institutes of Health (NIH) including the National Institute for
Diabetes and Digestive and Kidney Diseases (NIDDK) and the National Heart Lung
and Blood Institute (NHLBI). Definition for "overweight" according to the guideline is
BMI equal to or greater than 25 but less than 30; "obese" BMI greater than 30
(USDHHS, 2007). These definitions are accepted by the medical and scientific
community at large, and are in synch with international standards set by the World
Health Organization.
The disease etiology of childhood obesity is often taken for granted.
Overweight and obesity occurs, one assumes, because the intake of energy through
calories exceeds the energy expended through activity and exercise. Intuitively this
5
makes sense, and it accounts for the fact that virtually all prevention programs have
focused on increased exercise and better dietary intake (Wofford, 2008). This default
reasoning is also likely because there is not any sound scientific evidence to the
contrary. In a clinical report to the North American Society for Pediatric
Gastroenterology, Hepatology, and Nutrition, S. Baker et al. (2005) reported that "the
molecular biology of obesity is not fully understood" (p. 534). Further, although they
concede "Estimates of the degree to which genetic factors influence body fatness range
from 25% to 80%." they conclude "Nonetheless, changes in environmental and social
factors explain much of the doubling of childhood overweight in the last 30 years" (S.
Baker et al., 2005, p. 534).
It is these environmental and social factors that Phillipas and Lo (2005) term the
"obesogenic" environment of the 21 st century. This environment is also chiefly
characterized by the IOM Report (2004) as the cause for childhood obesity. Since this
environment is the underlying foundation upon which decreased physical exercise and
poorer dietary intake has been built, it is worth describing. According to Phillipas and
Lo (2005), the dietary side of the obesogenic environment includes:
Increased portion sizes served at food outlets; Increased variety and availability
of low-cost, energy-dense foods, especially fast food and snack foods in school
vending machines; Increased consumption of sugar-sweetened beverages such
as soft drinks and juices; Decreased proportion of meals consumed as "family
meals"; Increased consumption of pre-prepared convenience meals; Insufficient
consumption of fruits and vegetables, (p. 78)
6
The physical exercise side of the obesogenic environment includes: "Increased
prevalence of sedentary lifestyles due to video and computer games and TV watching;
Decreased physical education classes in schools; Increased reliance on car, bus,
escalator, elevator and other mechanical modes of transport resulted in less walking,
bicycling, stair climbing and other physical activity" (Phillipas & Ko, 2005, p. 78).
Clearly all of these phenomena present in today's 21 st century lifestyle in the
United States and United Kingdom must account for the decrease in physical activity
and poorer dietary intake. The question remains, however, what do the studies in the
published literature say?
There are a total of 16 studies included in this literature review, of which eight
are descriptive. This includes a total of six cross-sectional and two surveys. The
remaining eight are experimental, including six RCT and two longitudinal cohort
studies. Of the 16 total, roughly half focus on the specific dependent variables of
interest, physical exercise and diet; and the independent variable of interest, obesity or
overweight as measured by BMI. Another way to group the studies is by dependent
variable. Six total studies focused on physical exercise, five focused on diet, and the
remaining five focused on a combination of both. It is this grouping which will serve as
the organization for the literature review.
Physical Exercise
For physical exercise as the dependent variable, the results are fairly consistent.
As we will see, children and adolescents appear to be exercising less, and a few welldesigned studies show an increased likelihood of overweight and obesity with less
7
physical exercise. Of the six total physical exercise studies, four of them focus on the
effect on obesity and overweight as the primary outcome variable, measured by BMI.
Of these four, three conclude that reduced physical exercise may account for an increase
in BMI.
Johnston, Delva, and O'Malley (2007) use the nationally representative
Monitoring the Future (MTF) cross sectional surveys of eighth, 10th, and 12th graders
and the Youth Education and Society (YES) surveys of school administrators from
2003-2005 to show that physical education requirements and participation rates decline
significantly between eighth and 12th grades. With a large sample size of 54,000
students at 500 schools, they find that whereas 87% of eighth graders are required to
take physical education, only 20% of 12th graders are, explaining the participation rate
decline from 90% to 34% (Johnston, Delva, & O'Malley, 2007). Other findings include
lower participation rates for ethnic minorities and lower socioeconomic status (SES).
There is some evidence that the reduced physical activity levels of children and
adolescents may be associated with higher prevalence of overweight and obesity as
measured by BMI and percentage of body fat. There are three studies, two of which are
RCT's which conclude this. Epstein, Paluch, and Dorn (2000) randomly assigned 90
families with obese 8- to 12-year-old children to two different comprehensive
behavioral weight control programs both of which focused on physical exercise and
dietary intake. The primary difference was that one focused on increasing physical
activity and the other on reducing sedentary behavior. At 2 years, follow-up showed
8
that both programs resulted in significant decreases in percent overweight, body fat, and
improved aerobic fitness (Epstein et al., 2000).
Carrel et al. (2005) randomly assigned 50 obese middle school children from
rural U.S. communities and from urban environment at academic medical centers to a
research arm participating in lifestyle-based fitness-focused program or a control group
taking standard gym classes. At nine months, follow-up showed that a statistically
significant reduction in body fat, increased cardiovascular fitness, and improvement in
fasting insulin level for the research arm compared to the control group (Carrel et al.,
2005).
A recent cohort study done by Mitchell et al. (2009) also shows the same
relationship. Using the prospective birth cohort of Avon Longitudinal Study of Parents
and Children (ALSPAC) of 5,434 12-year-olds, they measured BMI and body fat by
Dexa scan, and sedentary behavior by accelerometer. An odds ratio (OR = 1.34) for
sedentary behavior accounts for a large percentage of the variance in obesity. They also
found a negative association between moderate to vigorous physical activity (MVPA)
and obesity (Mitchell et al., 2009).
There were two physical exercise related studies whose results did not provide
any evidence or substantiate the link to obesity. Metcalf, Voss, Hosking, Jeffery, and
Wilkin (2008) used a longitudinal study from England of 212 children aged 5 through 8
years over a 4-year period to show that exercise level above and beyond the government
mandated levels had no effect on BMI or percentage fat, only improved metabolic status
9
(Metcalf et al., 2008). This may be due to a threshold effect because very low exercise
levels were not included.
Finally, Burdette and Whitaker (2005) used a cross sectional survey of 3,141
mothers of 3-year-old children to assess the impact of a mother's perception of
neighborhood safety on obesity level, time spent outdoor playing, and TV viewing time.
BMI was measured with mothers self-reporting time spent in various activities. Their
findings include a significant effect of neighborhood safety on TV viewing time, but not
outdoor playing or obesity levels (Burdette & Whitaker, 2005).
Dietary Behavior
For dietary intake as the dependent variable, the results are more mixed than the
findings for physical exercise. As we will see, children and adolescents appear to be
eating less healthy food, but only one poorly designed study shows an increased
likelihood of overweight and obesity due to poor diet. Of the five total dietary intake
studies, only one of them focuses on the effect on obesity and overweight as measured
by BMI. It concludes that better dietary intake may account for a reduction in BMI.
Delva, O'Malley, and Johnston (2007) use the nationally representative
Monitoring the Future (MTF) cross sectional surveys of eighth, 10th, and 12th graders
and the Youth Education and Society (YES) surveys of school administrators from
2004-2005 to show that schools offer less healthy food more so than healthy food. With
a large sample size of 37,000 students at 345 schools, they find that 70% of middle
school and 60% high school students eat the lunch provided by the school (Delva,
O'Malley, & Johnston, 2007). Thus the opportunity to target the majority of the
10
nation's students exists. They found that "undisputedly, less-healthy foods are more
available than more-healthy foods in the nation's schools (Delva, O'Malley, &
Johnston, 2007). Of course, these findings pre-date the Child Nutrition and Women
Infant Child Reauthorization Act of 2004 which mandated beginning in school year
2006-07 that all schools that participate in the national school lunch program will
develop and implement wellness programs aimed at improving dietary intake and
increasing physical exercise. As Foerster et al. (2007) notes, these findings "provide a
baseline against which to set national goals and measure success."
The one study that focuses on the effect of dietary intake on overweight and
obesity is from Epstein et al. (2001), an RCT studying 30 pairs of obese parents and
non-obese children aged 6 to 11 years old and the effect of a comprehensive parentfocused behavioral intervention. Randomization determined whether the pair received a
program to focus on increased intake of fruit and vegetables or decreased intake of
high-fat high-sugar foods. At 1 year follow-up, they found that the group focusing on
increasing fruit and vegetables showed significantly greater decrease (Epstein et al.,
2001). This group was also found to have reduced their high-fat high-sugar intake as
well. There are multiple weaknesses in this study which include small sample size,
sample not representative of population, and most significantly BMI is not the primary
independent variable.
There are three additional studies which focus on secondary dependent variables
and their effects on obesity and overweight. Powell, Auld, Chaloupka, O'Malley, and
Johnston (2007) report on the effect of access to food stores on the prevalence of
11
obesity and overweight in students. Using a similar cross sectional model with Dun and
Bradstreet data on type of food store tied to 73,079 students from the Monitoring the
Future (MTF) nationally representative sample of eighth, 10th, and 12th graders by SES,
they report higher BMI and overweight associated with greater availability of
convenience stores and lower BMI and overweight associated with greater availability
of chain supermarkets (Powell, Auld, et al., 2007).
Powell, Chaloupka, and Bao (2007) employ a cross sectional model using Dun
and Bradstreet data of restaurant type in 28,050 zip codes across the country, tying it to
socio-economic data in those zip codes in the 2000 census. They found no difference in
the prevalence of fast-food vs. full-service restaurants by SES or race. However, their
findings help supplement the findings reported earlier about declining nutritional status
of children and adolescents. They report that the number of fast-food restaurants has
doubled in the last decade while the number of full-service restaurants has remained
constant; and that there has been a 71% increase in the prevalence of fast-food
restaurants, from only 17% of total restaurants in 1997 to 30% in 2006 (Powell,
Chaloupka, & Bao, 2007).
Finally, Gleason and Dodd (2009) use the nationally representative sample data
of 2,228 students who took the School Nutrition Dietary Assessment Study in Grades 112 to construct a cross sectional survey assessing participation in school lunch and
breakfast programs on BMI and risk of overweight and obesity. BMI was measured,
with students and parents self-reporting participation in the programs. After controlling
for all other variables, they found no evidence of any relationship between the usual
12
school lunch participation and weight status and BMI. The one significant finding was
that participation in the school breakfast program was actually a protective variable for
BMI, with significantly lower levels (Gleason & Dodd, 2009).
Physical Exercise and Dietary Behavior
For the combined effect of physical exercise and dietary intake as the dependent
variables, the results are fairly consistent. As we will see, children and adolescents
appear to be exercising less and eating less healthy food, and a few well-designed
studies show an increased likelihood of overweight and obesity with less physical
exercise and poorer diet. Of the five total combination studies, four focus on the effect
of obesity and overweight as measured by BMI. Of these four, two conclude that
increased physical exercise and better diet may account for a reduction in BMI.
There are three RCT's and one cohort longitudinal study assessing the combined
effect of exercise and diet on BMI. They vary widely in sample size and methods, and
their conclusions are not consistent. Two of the studies show a positive association and
two show no effect. Both of the larger RCT studies showed no effect. Caballero et al.
(2003) designed an RCT with 1,704 third, fourth, and fifth graders in 41 schools serving
American Indians in the western United States. Randomization resulted in their
assignment to a control group or to the intervention, a school-based program focusing
on dietary intake, physical activity, classroom curricula, and family involvement. At 3
years follow-up, the intervention group had no less body fat than the control group
(Caballero et al., 2003). The only significant finding was a reduction in percentage of
energy from fat in the intervention group. Thus, although the relationship between
13
intake and BMI cannot be ruled out, the intervention did not have significant impact on
intake. McCallum et al. (2007) designed an RCT nested within a cross sectional survey:
2,112 Australian children less than 12-years old participated in BMI survey
measurement; then 163 of them were randomly assigned to get intervention, consisting
of four physician consults over twelve weeks focusing on diet, exercise, and sedentary
behavior, or control which got normal consults. At 9-month and 15-month follow-up,
there was no significant difference between the groups in BMI or self image and health
status which were self-reported (McCallum et al. 2007).
The single smaller RCT did show evidence of a positive association between
increased exercise and improved diet and obesity. Fitzgibbon et al. (2005) designed an
RCT with 300 pre-school children in Head Start Program in Chicago. Randomization
resulted in their assignment to a control group or to the intervention, a school-based
program focusing on dietary intake, physical activity, classroom curricula, and family
involvement. At 2 years follow-up, the intervention group had statistically significant
lower BMI than the control group (Fitzgibbon et al., 2005).
Gable, Chang, and Krull (2007) used data from the Early Childhood
Longitudinal-Kindergarten Cohort with 8,549 obese children and follow-up data
gathering conducted at four time points between kindergarten entry and spring of third
grade. BMI was measured and parents provided feedback on eating and activity factors.
Findings at 4 years included odds ratios for more likely to be overweight for children
who watched more TV (1.02), and ate fewer family meals (1.08); more likely to be
persistently overweight for children who watched more TV (1.03), ate fewer family
14
meals (1.08), and lived in neighborhoods perceived to be less safe for outdoor play
(1.32) (Gable, Chang, & Krull, 2007). All of these odds ratios are quite small, with the
exception of 1.32 for children living in neighborhoods perceived to be less safe for
outdoor play.
As can be seen above, the results are varied with respect to the combination
studies for physical exercise and diet and effect on obesity—two showed no effect and
two suggest that more physical exercise and better dietary intake may result in lower
BMI. Finally, there is one cross sectional study also included in the search which has
significant findings supporting the association. Delva, Johnston, and O'Malley (2007)
use data from the Monitoring the Future (MTF) nationally representative samples of
eighth, 10th, and 12th grade students to assess effect of diet and exercise on prevalence
of obesity. Of the 39,811 students included, they find overall obesity rate of 25% with
minority and low-income males to be statistically significantly more likely to be obese
(Delva, Johnston, & O'Malley, 2007). Environmental factors that have statistically
significant associations with overweight include: frequency of eating breakfast, eating
fruits and vegetables, and exercising regularly—all shown to be inversely related to
overweight; the number of hours spent watching TV—positively related to overweight
(Delva, Johnston, & O'Malley, 2007).
From this review of the literature in late 2009, unfortunately we draw very
similar conclusions to that of the IOM in their update report in 2006 and Wofford in her
systematic review of childhood obesity prevention in 2008. That is, there is much more
research to be conducted. First, in terms of disease etiology, more research in
15
molecular biology and genetics is needed to assess the causal relationship between these
factor and obesity and overweight. Secondly, more well-designed studies are needed to
increase a growing but small evidence base that physical exercise and/or dietary intake
have an effect on obesity and overweight. What is needed to truly assess the impact of
these variables on BMI is additional scientific evidence: large, well-designed studies
with samples that are representative of the general population, that include long term
follow-up; and interventions which measure direct effect on physical exercise and diet,
and therefore their relationship to BMI.
16
CHAPTER 2
METHODS
Secondary data analysis is the research method selected for this study. The
reasoning is that there is large, representative sample data that is easily obtained and
readily available. Therefore given the limited resources of this study it would not be
prudent or practical to collect primary data for studying the general effect of physical
activity and/or dietary intake on childhood obesity and overweight. As has been stated
previously, large well-designed randomized controlled studies are needed to provide
further insight into this relationship.
National Health and Nutrition Examination Survey (NHANES^
The National Center for Health Statistics (NCHS), part of the Centers for
Disease Control and Prevention (CDCP), has instituted an ongoing program called the
National Health and Nutrition Examination Survey (NHANES). Originally a series of
surveys, in 1999 the NHANES became a continuous program. The NHANES data set
provides an ideal data set to study the relationship between physical activity and/or
dietary intake and childhood obesity. NHANES data (2007) has provided "overweight
prevalence figures (which) have led to the proliferation of programs emphasizing diet
and exercise, stimulated additional research, and provided a means to track trends in
obesity" (p. 4). In addition, due to the obesity epidemic in the industrialized world in
17
recent years, NHANES has been modified to collect more specific data to assist in the
tracking and prevention of obesity and overweight.
The NHANES program and its studies are designed to assess the health and
nutritional status of adults and children in the United States. It is unique in that it
combines data from physical examinations and interviews. It began in the early 1960's
as a series of surveys over multiple years of nationally representative samples of
approximately 5,000 persons in the United States. Different populations and health
topics have been addressed over the years. National Health Examination Surveys
(NHES) I, II, and III were conducted in the years 1959-1962,1963-1965, and 19661970, respectively. NHANES I, II, and III were then conducted in the years 1971-1975,
1976-1980, and 1988-1994, respectively. It was during the NHANES II and III studies
that the marked increase in obesity and overweight prevalence was first recognized, in
the general adult population as well as children and adolescents. In 1999, the survey
became a continuous program, and increments of two years of data are reported in each
release (1999-2000, 2001-2002, etc.) The most recent data available are from
NHANES 2005-2006. This data set will be used in this study.
There are two components to the NHANES survey: physical examination and
interview questionnaires. The physical examination component consists of medical,
dental, and physiological observation, as well as laboratory tests (NHANES, 2006). A
specially designed mobile examination center (MEC) with state of the art equipment is
utilized for the examination. The interview component includes demographic,
socioeconomic, dietary, and health-related questions. The interview occurs in the
18
comfort of the survey participant's home in most cases. Some interviews occur on site
during the examination at the MEC. The data sets are readily available as public
information. It is important to note that the data from the survey are used as the current
national standards for such measurements as height, weight, and blood pressure
(NHANES, 2007).
The sample selected for NHANES is intended to represent the United States
population of all ages. In order to produce reliable statistics, "NHANES over-samples
persons 60 and older, African Americans, and Hispanics" (NHANES, 2006). The
sampling process for NHANES is complex and involves multiple stages. Primary
sampling units (PSU) are counties within the United States. Occasionally small
counties are combined to meet minimum population size. The sampling frame for this
design is all counties in the United States. There is an additional stage of selection in
the probability design for NHANES. Clusters of households are selected, with each
person in the household being screened for demographic characteristics. Then one or
more member of the household is selected for the sample. An example of the size of the
sample: "For NHANES 2003-2004, there were 12,761 persons selected for the sample,
10,122 of those were interviewed (79.3 percent) and 9,643 (75.6 percent) were
examined in the mobile examination center." (NHANES, 2006). NHANES provides
researchers with sample weights to produce unbiased national estimates: "These
sample weights reflect the unequal probabilities of selection, non-response adjustments,
and adjustments to independent population controls" (NHANES, 2006). There are
variables in the SAS statistical computer program to account for these sample weights:
19
"WTINT2YR" is for the interview sample, and "WTMEC2YR" is for the examination
sample (NHANES, 2006).
The data collection effort is highly regulated and adheres to industry standards
for scientific data collection. As stated above, many of the health interviews are
conducted in respondent's homes; the remainder of interviews and the health
examinations are performed in specially designed MEC's which travel to locations
throughout the country. The NHANES team consists of a physician, medical and health
technician, and health and dietary interviewers. All are highly specially trained in the
protocols used in the NHANES program. Many of the staff are bilingual and
communicate in both English and Spanish. In addition, all questionnaires were
translated into Spanish. A custom designed computer system is used to collect and
process all of NHANES data, nearly eliminating the need for paper forms and manual
coding operations (NHANES, 2006). The system allows for interviewers to use
notebook computers with electronic pens for direct data entry at the point of interview.
Data are automatically transmitted into the database through such devices as digital
scales and stadiometers. In addition, there is a computer-assisted personal interview
(CAPI) system which allows respondents to enter their own response by touch-sensitive
computer screens in English and Spanish (NHANES, 2006). This allows for complete
privacy for respondents to answer particularly sensitive questions about health, diet,
alcohol, drug or sexual behaviors. Finally, information is available to NHANES staff
within 24 hours of collection, ensuring the timely collection of high quality accurate
data.
20
In addition to the highly efficient and accurate manner in which data are
collected, the design of the study and its operation also adhere to the rigorous standard
applied to scientific research. The survey is designed to facilitate and encourage
participation: "Transportation to and from the MEC is provided, if necessary.
Participants receive compensation and a report of medical findings is given to each
participant. All information collected in the survey is kept strictly confidential. Privacy
is protected by public laws" (NHANES, 2006). Since the focus of this study is on
childhood overweight and obesity, it is important to note that a responsible adult
provides information for participants less than 16 years of age. All interviewers
complete a comprehensive two-week training program. Many of the interviewers have
previous interviewing experience, and a large percentage of them are bilingual in
English and Spanish (NHANES, 2006).
A significant amount of pilot testing is conducted prior to implementing the
surveys in the field. This pre-testing component was conducted with the field and
examination staff in order to test all systems and procedures prior to NHANES
administration (NHANES, 2006). Several types of quality monitoring methods were
employed to ensure that high quality accurate data were collected for NHANES.
Interviewer monitoring was conducted by National Center for Health Statistics (NCHS)
staff. Retraining on survey procedures was done if necessary. In addition, NCHS used
field staff and interviewer feedback to modify the questionnaires and survey materials.
Finally, survey staff were debriefed and trained annually on any changes made for new
survey content prior to implementation (NHANES, 2006).
21
Dependent Variable—Body Mass Index (BMP
The dependent variable for this study is body mass index (BMI). The definition
of BMI and how it is calculated is reported elsewhere in the study. In the NHANES
codebook, the variable is "BMXBMI" and it is collected for all survey participants (SP)
greater than 2 years of age. For NHANES 2005-2006, there were 8,949 recorded values
with 1,001 missing information for a total of 9,950 (NHANES, 2006). In order to
calculate the BMI of each SP, their weight and standing height are recorded in the
MEC. Weight is measured for all SP of all ages using a digital scale. The weight in
kilograms (kg) is then electronically transferred to the NHANES database via the
computer system. Standing height is measured for all SP 2 years and older using a
stadiometer. SP's under 2 years are measured differently for recumbent length.
Standing height in centimeters (cm) is then electronically transferred to the NHANES
database via the computer system. Since BMI is only reported for SP greater than 2
years in age, infants and toddlers under age 2 will not be included in this study.
NHANES uses the following criteria to screen for weight categories: underweight
(BMI < 18.5), normal or desirable weight (BMI 18.5-24.9), overweight (BMI 25.029.9), and obese (BMI > 30).
Independent Variables—Physical Exercise and Dietary Behavior
The independent variables for this study are broken into two categories:
physical activity and diet behavior. The Physical Activity Questionnaire (PAQ) is
"based on the Global Physical Activity Questionnaire (GPAQ) and includes questions
related to daily activities, leisure time activities, and sedentary activities" (NHANES,
22
2006). The GPAQ is produced by the World Health Organization and is a widely
accepted tool for measuring physical activity. Eligible respondents are SP 2 or more
years of age. Proxy respondents answered the questions for SP 2-11 years of age. The
survey was conducted before the physical examination, in the home, using the
computer-assisted personal interviewing (CAPI) system. The three questions of interest
to this study are as follows. Variable "PAD560" is the "Times per week play or
exercise hard" and is quantified numerically. Eligible respondents are SP 2-11 years of
age. The range of values is 0-35. The specific question is: "How many times per week
does {SP} play or exercise enough to make {him/her} sweat and breathe hard?"
(NHANES, 2006) Variable "PAD590" is the "Hours/day watch TV or videos past 30
days" and is coded from 0-6 in increments of hours starting with less than 1 hour and
ending in 5 hours or more hours, including none. Eligible respondents are SP 2-18
years of age. The specific question is: "Over the past 30 days, on average how many
hours per day did {SP} sit and watch TV or videos?" (NHANES, 2006) Variable
"PAD600" is the "Hours/day use computer past 30 days" and is coded 0-6 in increments
of hours starting with less than 1 hour and ending in 5 hours or more, including none.
Eligible respondents are SP 2-18 years of age. The specific question is: "Over the past
30 days, on average how many hours per day did {SP} use a computer or play computer
games outside of school?" (NHANES, 2006). These three independent variables will
provide numerical values which can be statistically analyzed with respect to the
dependent variable of BMI. It is hoped that the relationship can be studied between
BMI and physical exercise and lack of it due to time spent watching TV, videos, and/or
23
the computer. A limitation is the lack of direct measurement of physical exercise for SP
11-18 years of age. For the sample used in this study from NHANES 2005-2006, there
were 4,100 recorded values with 1,001 missing information for a total of 5,102.
The other independent variable is diet behavior and nutrition. The Diet
Behavior and Nutrition (DBQ) questionnaire is administered to eligible respondents of
birth age and up. Proxy respondents answered the questions for SP from birth to 11
years of age. The survey was conducted before the physical examination, in the home,
using the computer-assisted personal interviewing (CAPI) system. There are two sets
of questions of interest to this study. One set applies to the dietary intake of children
aged 4-19 in schools that provide school lunch and breakfast programs. The other set
applies to the dietary intake of children aged 1-11 at home. Because of the importance
of diet and nutrition obtained in both settings, this study will include all three questions.
For dietary intake in school, the questions are as follows. Variable "DBD381"
is the "Times per week complete school lunch" and is quantified numerically. The
range of values is 0-5. The specific question is: "During the school year, about how
many times a week does {SP} usually get a complete school lunch?" (NHANES, 2006)
Variable "DBD411" is the "Times per week complete school breakfast" and is coded
numerically. The range of values is 0-5. The specific question is: "During the school
year, about how many times a week does {SP} usually get a complete breakfast at
school?" (NHANES, 2006) For the sample used in this study from NHANES 20052006, there were 2,918 recorded values with 2,183 missing information for a total
5,101. Since the only NHANES data that were available for analysis were from 2005-
24
2006, it is not possible to examine the effect of the 2004 Child Nutrition Act which
mandated all schools participating in the federal school lunch programs to have adopted
wellness programs aimed at improving nutrition. This is because the mandate was to
begin in the school year 2006-2007.
For dietary intake at home, the questions are as follows. Variable "DBD091" is
the "Times per week meal prepared away from home" and is quantified numerically.
The range of values is 0-21. The specific question is "By meal, I mean breakfast, lunch,
and dinner. During the past 7 days, how many meals did {SP} get that were prepared
away from home in places such as restaurants, fast food places, food stands, grocery
stores, or from vending machines?" (NHANES, 2009) The analysis of these variables
and their relationship to BMI will hopefully address the unhealthy eating habits geared
toward fast-food and meals prepared away from home. A limitation is the lack of direct
measurement of dietary behavior, a better measurement would have been food recall
with mechanism to validate actual intake.
Data Analysis Plan
The data analysis plan consists of descriptive statistics to provide an overview of
the sample population of 5,101 survey participants. The sample was provided as a data
file in the Statistical Package for the Social Sciences (SPSS) computer program for
statistical analysis, version 17. Frequency statistics including mean and standard
deviation will be run for age, gender, race, and annual family income. Differences in
each sub-population for age, gender, and race will be tested using the T test of
independent sample means and Analysis of Variance (ANOVA) tests. In addition,
25
differences in BMI will be tested for each sub-population of race, gender, age, family
income and school attendance.
Bi-variate analysis of correlation will be used to test the relationship between
BMI and each of the independent variables of physical activity: PAD560 (times per
week play or exercise hard), PAD590 (hours/day watch TV or videos last 30 days), and
PAD600 (hours/day use computer last 30 days); and the independent variables of
dietary behavior: DBD381 (times per week complete school lunch), DBD411 (times
per week complete school breakfast), and DBD091 (times per week meal prepared
away from home). In addition, bi-variate analysis of correlation will be used to test the
relationship of the independent variables of physical activity and dietary behavior to
each other.
Finally, using the results of the bi-variate analyses to identify statistically
significant independent variables, a multi-variate analysis of regression will be used to
observe the relationship and predictive power of BMI to these variables (if any).
26
CHAPTER 3
RESULTS
Descriptive Statistics
The sample of 5,101 survey participants (SP) was analyzed using SPSS version
16.0. The range of SP age was 0-18 with a mean of 8.62 and SD of 6.038. With regard
to gender, the distribution of SP was almost equal for males (n = 2,557; 50.1%) and
females (n = 2,544; 49.9%). With regard to race, the distribution of SP was as follows:
Mexican American (n = 1,752; 34.3%), Other Hispanic (n = 186; 3.6%), White nonHispanic (n = 1,374; 26.9%), Black non-Hispanic (n = 1,491; 29.2%), and Other Race
including multi-racial (n = 298; 5.8%). With regard to annual family income, the
distribution of SP was as follows: $0-24,999 (n = 1,713; 33.7%), $25,000-$74,999 (n =
2,096; 41.2%), $75,000+ (n = 1,028; 20.2%), and Refused/Don't Know/Missing (n =
264; 4.9%). With regards to BMI, the mean was 20.70 with SD of 5.988.
Sample data were reviewed for each sub-population in order to confirm whether
or not there were statistically significant differences within the groups. Age by gender
was examined using the T test of independent sample means. There was no statistically
significant difference between boys and girls with t (5,099) = -1.786 (p = .074).
Gender by race was examined using the Chi-squared test with exact tests Monte Carlo
95% confidence interval. There was no statistically significant difference between the
five sub-groups: X2 (4) = 2058.244 (p = .068). Age by race was examined using the
27
ANOVA test with Tukey post-hoc and means plot. There was a statistically significant
difference reported withF(4) = 16.453 (p < 0.001). The Black non-Hispanic sub-group
was older with a mean age of 9.64 and SD of 5.843. The mean was statistically
significant different (higher) than all four other groups, which may help explain
differences in BMI for this group noted later in the study.
Bi-variate Analyses of Correlation
Bi-variate analyses of correlation were conducted to test for any differences in
BMI among the sub-populations of age, gender, race, family income, and school
attendance. There were significant differences amongst all five groups. For age, the
ANOVA test using Tukey post-hoc means plot, produced statistically significant results
F (16) = 129.584 (P < 001). As would be expected, BMI steadily increases with age.
The range of means varies widely between 16.70 for 2 year olds, reaching 19.18 for 9
year olds. By age 17 and 18, average BMI of 25.12 and 26.13 respectively is
considered overweight using the NIDDK definition. For race, the ANOVA test using
Tukey post-hoc means plot, produced statistically significant F (4) = 13.452 (P < .001)
differences between Black non-Hispanic/ White non-Hispanic, Black-Non Hispanic/
Other Race, Mexican American/ White non-Hispanic, and Mexican American/ Other
Race. BMI for Mexican American and Black non-Hispanic were higher than the other
three groups with means of 21.11 and 21.30 respectively. For gender, the t test of
independent sample means produced statistically significant results t (4007) = - 2.370
(P = .018). BMI for females was higher than males with means of 20.92 and 20.47
respectively. For family income, ANOVA test using Tukey post-hoc means plot
28
produced statistically significant F (14) = 3.545 (P < .001) for differences among the
different income levels. BMI for the lowest income levels tended to be higher than
BMI for the upper income levels, with means of 22.70 and 21.74 for incomes levels
between $0-$4,999 and $5,000-$9,999 respectively, and mean of 19.86 for income level
$75,000 and up. Finally, for school attendance, ANOVA test using Tukey post-hoc
means plot produced statistically significant F (2) = 15.682 (P < .001) differences in
those attending school, on vacation, or not attending school. For those not attending
school, although the sample size was small (n = 165) the mean of 24.84 approximated
overweight level.
Bi-variate analyses of correlation were conducted to test the relationship
between BMI and both sets of independent variables of physical exercise and dietary
behavior. For physical exercise, the results were unexpected. The un-weighted
unadjusted Pearson correlation model was run with BMI and the three independent
variables for physical activity. Only one of the three was statistically significant in the
direction one would expect. The number of hours per day watching TV or videos had a
statistically significant positive relationship to BMI (r = .093, P < .001). This is
intuitively what one would expect. However, the number of hours per day using the
computer had a statistically significant negative relationship to BMI (r = - .038, P =
.016). This is not intuitively what one would expect. Finally, the number of times per
week spent playing or exercising hard did not have a statistically significant relationship
to BMI (r = -.019, P = .380). It is also important to report that none of the three
independent variables had any statistically significant relationship between each other.
29
For dietary behaviors, the results were unexpected. The un-weighted unadjusted
Pearson correlation model was run with BMI and the three independent variables of
dietary behavior. Two of the three variables was not statistically significant, and the
third was statistically significant in the opposite direction. Both the number of times
per week getting a school lunch and number of times per week getting a school
breakfast were not statistically significant (r = .011 and .005, P = .570 and .800)
respectively. The number of times per week eat meals not from home was statistically
significant in the opposite direction one would expect (r = - .049, P = .002). This
negative relationship suggests that individuals who eat out more often actually tend to
have lower BMI than those who eat out less frequently. This is not intuitively what one
would expect. As would be expected, there was a strong positive relationship (r = .257,
P < .001) between the number of times per week get school lunch and number of times
per week get school breakfast.
Finally, an un-weighted unadjusted Pearson correlation model was run to test the
relationship between BMI and all independent variables including physical exercise and
dietary behaviors. The results were the same as reported above with one additional
finding. A statistically significant relationship (r = .058, P < .001) exists between the
number times per week eat meals not from home and number of hours per day use
computer past 30 days. Since both of these independent variables have the unexpected
statistically significant findings, one suspects that a confounding variable such as annual
family income may be at play. More will be discussed in the conclusion section. See
Table 1 for a complete summary of bi-variate analyses of correlation for BMI.
30
TABLE 1. Bi-Variate Analyses of Correlation
Independent Variable
# times/week you play or
exercise hard
Body Mass Index
(kg/m**2)
Pearson Correlation
-.019
Sig. (2-tailed)
.380
2,077.000
N
# hours/day watch TV or
videos past 30 days
Pearson Correlation
Sig. (2-tailed)
N
.093**
.000
3,937.000
# hours/day use computer
past 30 days
Pearson Correlation
Sig. (2-tailed)
N
-.038*
.016
3,937.000
# of times/week get school
lunch
Pearson Correlation
Sig. (2-tailed)
.011
.570
2,795.000
# of times/week get school
breakfast
Pearson Correlation
Sig. (2-tailed)
N
.005
.800
2,398.000
# of times/week eat meals
not from home
Pearson Correlation
Sig. (2-tailed)
N
-.049**
.002
3,926.000
Notes
Correlation is significant at the 0.01 level (2-tailed)
Correlation is significant at the 0.05 level (2-tailed)
Multi-Variate Analyses of Regression
Multi-Variate analyses of regression were then tested for all statistically significant
bi-variate variables. For physical activity, the model consisted of multi-level linear
regression with BMI as dependent variable and statistically significant independent
variables of number of hours per day watch TV or videos and number of hours per day
computer use. Although the results were statistically significant relationship with F (2)
31
= 20.210 (P < .001), the R2 of .010 suggests a very low predictive power. For dietary
behavior, the model consisted of multi-level linear regression with BMI as dependent
variable and statistically significant independent variable of number of times per week
eat meals not from home. Again, although the results were statistically significant
relationship with F (1) = 9.594 (P = .002), the R2 of .002 suggests a very low predictive
power. Finally, combining physical exercise and dietary behavior, the model consists
of multi-level linear regression with BMI as dependent variable and all three
statistically significant independent variables listed above. Similar results are obtained,
with statistically significant relationship of F (3) = 16.486 (P < .001) and low predictive
power of R2 = .012. See Table 2 for a complete summary of multi-variate analyses of
regression for BMI.
32
TABLE 2. Multi-Variate Analyses of Regression
Dependent variable: Body Mass Index (kg/m**2)
Model
Sum of
Mean
Df
Squares
Square
Regression
2
720.992
1441.983
Residual
3934
35.674
140342.594
Total
141784.578
F
20.210
Sig.
.000
3936
Predictors: # hours/day use computer past 30 days, # hours/day watch TV or video
past 30 days
Notes: R2 = .010 (P < .01)
Model
Regression
Residual
Total
Sum of
Squares
343.812
140622.388
140966.201
1
Mean
Square
343.812
3924
335.836
Df
F
9.594
Sig.
.002
2925
Predictors: # times/week eat meals not from home
Notes: R2 = .002 (P = .02)
Model
Regression
Sum of
Squares
1755.492
3
Mean
Square
585.164
35.495
Df
Residual
139210.708
3922
Total
140966.201
3925
F
16.486
Sig.
.000
Predictors: # hours/day use computer past 30 days, # hours/day watch TV or video
past 30 days, and # times/week eat meals not from home
Notes: R2 = .012 (P < .01)
33
CHAPTER 4
DISCUSS ION
Overview and Limitations
This study seeks to examine the relationships between physical activity, dietary
behavior, and BMI in children and adolescents using a sample data set of 5,101 survey
participants in the 2005-2006 NHANES. In general, the findings are consistent with the
findings in the published literature reviewed earlier in this study. First, there does
appear to be a statistically significant relationship between physical activity and BMI,
however the strength of the relationship and its predictive value are weak. This is
consistent with the findings in the published literature with only two small RCT's and
one large cohort study providing evidence of the positive relationship between the two.
Second, there does not appear to be a statistically significant relationship between
dietary behavior and BMI. This is consistent with the findings in the published
literature with only a single RCT with small sample size reporting relationship between
obesity as measured by BMI and diet. In addition, the lack of a significant relationship
between lunches and/or breakfasts eaten at school and BMI is consistent with findings
in the published literature. Finally, there does appear to be a statistically significant
relationship between BMI and physical activity combined with dietary behavior.
However, the linear regression model tested to have a very weak predictive power. This
is consistent with the mixed results in the published literature, of which only one small
34
RCT and one large cohort study found a positive relationship, and two larger RCT
found no relationship between the two.
The primary limitations of the study are the broad age range used in both
independent variables, and the lack of direct measurement of physical exercise and
dietary behavior used in NHANES. For physical activity, the age range for two of the
variables are 2-18 years old: number of hours per day watch TV or video and number
of hours per day use computer. For dietary behavior, the age range for two of the
variables are 4-18 years old: number of times per week get school lunch and number of
times per week get school breakfast. These broad age ranges certainly contribute to the
lack of statistical significance found with respect to BMI. There is also lack of direct
measurement of both independent variables. For physical exercise, there is no direct
measure for SP 11-18 years of age. For dietary behavior, a better measurement would
have been food recall with mechanism to validate actual intake.
Findings
The descriptive statistical findings in this study are in general consistent with the
findings in the published literature. BMI tends to increase with age, and is higher for
ethnic minorities and lower socio-economic status groups. Delva, Johnston, and
O'Malley (2007) found an overall obesity rate of 25% with minority and low-income
males to be statistically significantly more likely to be obese. This is consistent with the
data in 2005-2006 NHANES, with the mean differences for Black non-Hispanic and
Mexican American groups to be statistically significantly higher than White nonHispanic and Other Race a t P < .001 level (1.411, 2.050,1.220, and 1.859) respectively.
35
The mean BMI for the lowest annual family income levels were significantly higher
than those in the upper-middle and upper levels.
Physical activity and BMI were found to have a statistically significant
relationship by using the un-weighted and unadjusted Pearson correlation test. The
number of hours per day watching TV or videos had a statistically significant positive
relationship to BMI (r = .093, P < .001). This is intuitively what one would expect.
However, the number of hours per day using the computer had a statistically significant
negative relationship to BMI (r = - .038, P = .016). This is not intuitively what one
would expect. When both variables are placed in a multi-variate linear regression
model, the statistical significance remains (P < .001). However the predictive power is
weak (R2 = .010). These findings are consistent with what was found in the published
literature. Two RCT's with small sample size not representative of the overall
population, and one cohort study showed a positive relationship between physical
exercise and BMI. Epstein, Paluch, and Dorn (2007) studied 8 to 12-year-old children
(N = 90), and found that both programs resulted in significant decreases in percent
overweight and body fat. Carrel et al. (2005) studied obese middle school children (N =
50), and found that a statistically significant reduction in body fat at 9 months. A large
cohort study of 12-year-olds (N = 5,434) done by Mitchell et al. (2009) found that
sedentary behavior accounts for a large percentage of the variance in obesity as
measured by BMI. On the other hand, two descriptive studies showed no effect.
Metcalf, Voss, Hosking, Jeffery, and Wilkin (2008) used a small (N = 21) longitudinal
data of 5 to 8 year olds to show that exercise beyond the government mandated levels
36
had no effect on BMI. Burdette and Whittaker (2005) studied a large (N = 3,141) crosssectional study of 3-year-olds but found no relationship between neighborhood safety,
time playing outside, or obesity levels.
So the data are mixed for physical activity. This researcher found very little
evidence to support the relationship between increased physical activity and lower BMI.
Certainly there are other studies and the literature base is constantly growing. The
relationship does exist for less sedentary behavior. But even within that sub-category,
the results are not consistent. Whereas the number of hours per day spent watching TV
or videos was positively associated with BMI, the number of hours per day spent using
the computer was negatively associated with BMI. The confounding variable may well
be socio-economic status, as the upper-middle and upper class levels are more likely to
have better access to computers.
Dietary behavior and BMI were not found to have a statistically significant
relationship by using the un-weighted and unadjusted Pearson correlation test. The
number of times per week eating meals prepared away from home was the only variable
that had a statistically significant relationship to BMI (r = -.049, P = .002). This
negative relationship is not intuitively what one would expect. Conventional thinking is
that meals cooked at home would tend to be healthier and more balanced than meals
prepared away from home at a restaurant, whether full service or fast food, and that less
of them would be associated with higher BMI. However the opposite was found to be
the case. The possible confounding variable would again be socio-economic status, as
middle-upper and upper income level groups may have better access to eating out more
37
often, and have lower BMI. When placed in a multi-variate linear regression model, the
statistical significance remains (P = .002). However the predictive power is weak (R2 =
.002). These findings are consistent with what was found in the published literature.
Only one small RCT showed a positive relationship between dietary behavior and BMI.
In a small (N = 30) RCT of pairs of obese parents and non-obese children aged 6 to 11
years old, Epstein et al. (2001) found that both increased intake of fruits and vegetables
and decreased intake of high-fat high-sugar foods resulted in significantly reduced BMI.
In addition, there is no evidence in the published literature to support any relationship
between frequency of eating school lunches and / or breakfasts and BMI. Our findings
from NHANES show no statistically significant relationship between either variable and
BMI. This is consistent with Gleason and Dodd (2009) who constructed a large (N =
2,228) cross-sectional survey of students in Grades 1-12 to assess the effect of
participation in school lunch and breakfast programs on BMI, and found no evidence
between them.
What is concerning about the results for dietary behavior and BMI is that they
are not changing over time. Whereas NHANES 2005-2006 preceded the Federal
mandate for all schools participating in the Federal school lunch program to develop
wellness programs aimed at improving dietary behavior and physical exercise, Gleason
and Dodd's study took place afterwards. This is not the promising data we would hope
for with the implementation of a Federal program aimed at improving the dietary habits
of our school children. So the data are at best mixed for dietary behavior. Clearly there
38
is little to no evidence to support the relationship between improved dietary intake and
lower BMI.
Finally, there does appear to be a statistically significant relationship between
BMI and physical activity combined with dietary behavior. When the three statistically
significant variables were taken and input into a multi-variate linear regression model, it
tested to have a very weak predictive power. The number of hours per day watching
TV or videos, the number of hours per day using the computer, and the number of times
per week eat meals prepared away from home were utilized. The results (R2 = .012, P <
.001) are significant but with weak predictive power. This is consistent with the mixed
results in the published literature, of which only one small RCT and one large cohort
study found a positive relationship, and two larger RCT found no relationship between
the two. In a small (N = 200) RCT of pre-school aged children, Fitzgibbon et al. (2005)
showed that assignment to a school-based program focusing on dietary intake and
physical activity with family involvement resulted in significantly lower BMI levels at
two years follow-up. In addition, Gable, Chang, and Krull (2007) used a large (N =
8,549) longitudinal study of obese children between kindergarten and third grade to
show children who watched more TV and ate fewer family meals were more likely to
be overweight and persistently overweight. Although the odds ratios were small, the
relationship existed nonetheless.
On the other hand, the two large RCT showed no effect of physical activity and
dietary behavior on BMI. Caballero et al. (2003) used a large (N = 1,704) RCT of third
to fifth graders from 41 schools serving American Indians to study the impact of a
39
school-based program focusing on physical activity and dietary intake on BMI, and
found no impact from the intervention. McCallum et al. (2007) designed an RCT
nested within a large (N = 2,112) cross sectional survey of children less than 12 years
old to study impact of physician consults focusing on physical activity and dietary
behavior on BMI, and found no significant difference between the controls and
intervention group.
So the data are mixed for the combination of physical activity and dietary
behavior and its effect on BMI. Clearly there is very little evidence to support the
relationship between the two variables and lower BMI. The studies that do show
statistically significant results focus on the younger populations of pre-school and K-3
aged children for which the obesity prevalence rate is lower. What is needed is
additional evidence focused on maintaining this decrease in BMI as the children
progress into puberty and adolescence, when the prevalence rates begin to increase
dramatically.
Conclusion and Future Direction
In conclusion, it is concerning to this researcher that the data in NHANES is
found to be consistent with the evidence in the published literature regarding the effects
of physical exercise and dietary behavior on BMI. That is, that there is very little
evidence to show that increased physical exercise, decreased sedentary behavior, and /
or better dietary intake has a statistically significant impact on childhood obesity and
BMI. It has been almost 10 years since the initial call in 2001 to action by the Surgeon
General of the United States to address the obesity epidemic. Despite both IOM reports
40
in 2004 and 2006, and the plethora of federal and state initiatives focusing on improving
children's physical exercise and dietary behavior, and the results still do not show any
significant impact. This questions whether the assumed etiology of childhood obesity
has been correct. Society in general, and the scientific community in particular, has
perhaps made the incorrect assumption that the independent variables of physical
exercise and dietary behavior direct affect BMI. What is needed to truly assess the
impact of these variables on BMI is additional scientific evidence: large, well-designed
studies with samples that are representative of the general population, that include long
term follow-up. In addition, after controlling for sex, race, and socio-economic status,
studying specific age groupings will be helpful in determining exactly when the onset of
obesity occurs, and precisely what specific interventions are effective in reducing BMI.
Certainly the contributing factors in the pre-school age range will differ from those in
the K-6 and adolescent age ranges. These differences must be accounted for and
understood in order to more strategically target specific age groups with effective
interventions. It is hoped that the scientific community will recognize the need for
further study in this area so that the battle against the childhood obesity epidemic can be
effectively won over time.
41
REFERENCES
42
REFERENCES
Baker, J. L., Olsen, L. W., & Sorensen, T. I. (2007). Childhood body-mass index and
the risk of coronary heart disease in adulthood. The New England Journal of
Medicine, 35(23), 2329-37.
Baker, S., Barlow, S., Cochran, W., Fuchs, G., Klish, W., Krebs, N., ... Udall, J.
(2005). Overweight children and adolescents: A clinical report of the North
American Society for Pediatric Gastroenterology, Hepatology and Nutrition.
Journal of Pediatric Gastroenterology and Nutrition, 40(5), 533-43.
Burdette, H. L., & Whitaker, R. C. (2005). A national study of neighborhood safety,
outdoor play, television viewing, and obesity in preschool children. Pediatrics,
116(3), 657-62.
Caballero, B., Clay, T., Davis, S. M., Ethelbah, B., Rock, B. H., Lohman, T., ...
Stevens, J. (2003). Pathways: A school-based, randomized controlled trial for
the prevention of obesity in American Indian schoolchildren. The American
Journal of Clinical Nutrition, 78(5), 1030-8.
California Department of Public Health. (2006). California obesity prevention plan: A
vision for tomorrow, strategic actions for today. Retrieved from www.cdph
.cdi.gov/programs/Documents/COPP.pdf
Carrel, A. L., Clark, R. R., Peterson, S. E., Nemeth, B. A., Sullivan, J., & Allen, D. B.
(2005). Improvement of fitness, body composition, and insulin sensitivity in
overweight children in a school-based exercise program: A randomized,
controlled study. Archives of Pediatrics & Adolescent Medicine, 159(10), 963-8.
Centers for Disease Control and Prevention, National Center for Health Statistics.
(2005). National health and nutrition examination survey: survey results.
Retrieved from http://www.cdc.gov/nchs/nhanes/nhanes20052006/nhanes05_06.htm.
Centers for Disease Control and Prevention, National Center for Health Statistics.
(2006a). National health and nutrition examination survey: analytic and
reporting guidelines. Retrieved from http://www.cdc.gov/nchs/nhanes/
nhanes2005-2006/nhanes05 06.htm.
43
Centers for Disease Control and Prevention, National Center for Health Statistics.
(2006b). National health and nutrition examination survey: documentation,
codebook, and frequencies. Retrieved from http://www.cdc.gov/nchs/nhanes/
nhanes2005-2006/nhanes05_06.htm.
Centers for Disease Control and Prevention, National Center for Health Statistics.
(2006c). Health, United States, 2006, chartbook on trends in the health of
Americans. Retrieved from http://www.cdc.gov/nchs/data/hus/hus06.pdf.
Centers for Disease Control and Prevention, National Center for Health Statistics.
(2007). National health and nutrition examination survey: survey overview.
Retrieved from http://www.cdc.gov/nchs/nhanes/ nhanes20052006/nhanes05_06.htm.
Chaloupka, F. J., & Johnston, L. D. (2007). Bridging the Gap: Research informing
practice and policy for healthy youth behavior. American Journal of Preventive
Medicine, 35(4), SI47-61.
Delva, J., Johnston, L. D., & O'Malley, P. M. (2007). The epidemiology of overweight
and related lifestyle behaviors: Racial/ethnic and socioeconomic status
differences among American youth. American Journal of Preventive Medicine,
53(4), S178-86.
Delva, J., O'Malley, P. M , & Johnston, L. D. (2007). Availability of more-healthy and
less-healthy food choices in American schools: A national study of grade,
racial/ethnic, and socioeconomic differences. American Journal of Preventive
Medicine, 35(4), S226-39.
Epstein, L. H., Gordy, C. C , Raynor, H. A., Beddome, M., Kilanowski, C. K., &
Paluch, R. (2001). Increasing fruit and vegetable intake and decreasing fat and
sugar intake in families at risk for childhood obesity. Obesity Research, 9(3),
171-8.
Epstein, L. H., Paluch, R. A., Gordy, C. C , & Dorn, J. (2000). Decreasing sedentary
behaviors in treating pediatric obesity. Archives of Pediatrics & Adolescent
Medicine, 154(3), 220-6.
Fitzgibbon, M. L., Stolley, M. R., Schiffer, L., Van Horn, L., Kaufer-Christoffel, K., &
Dyer, A. (2005). Two-year follow-up results for Hip-Hop to Health Jr.: A
randomized controlled trial for overweight prevention in preschool minority
children. The Journal of Pediatrics, 146(5), 618-25.
44
Foerster, S. B., Silver, L. D., Kohatsu, N. D., Frieden, T. R., Bassett, M. T., & Horton,
M. B. (2007). Childhood obesity on the front lines. American Journal of
Preventive Medicine, 33(4), SI75-7.
Gable, S., Chang, Y., & Krull, J. L. (2007). Television watching and frequency of
family meals are predictive of overweight onset and persistence in a national
sample of school-aged children. Journal of the American Dietetic Association,
107(1), 53-61.
Gleason, P. M., & Dodd, A. H. (2009). School breakfast program but not school lunch
program participation is associated with lower body mass index. Journal of the
American Dietetic Association, 109(2), SI 18-28.
Institute of Medicine. (2004). Preventing childhood obesity: health in the balancereport brief. Retrieved from http://www.iom.edu/~/media/Files/Report %20
Files/2004/Preventing-Childhood-Obesity-Health-in-the-Balance/Childhood
Obesity4pagerfixforwebpdf.ashx
Institute of Medicine. (2006). Progress in preventing childhood obesity: how do we
measure up?—report brief. Retrieved from http://www.iom.edu/Reports/2006/
Progress-in-Preventing-Childhood-Obesity—How-Do-We-Measure-Up.aspx
Johnston, L. D., Delva, J., & O'Malley, P. M. (2007). Sports participation and physical
education in American secondary schools: Current levels and racial/ethnic and
socioeconomic disparities. American Journal of Preventive Medicine, 33(4),
S195-208.
Ludwig, D. S. (2007). Childhood obesity~the shape of things to come. The New
England Journal of Medicine, 357(23), 2325-7.
McCallum, Z., Wake, M., Gerner, B., Baur, L. A., Gibbons, K., Gold, L., ... Waters, E.
(2007). Outcome data from the LEAP (Live, Eat and Play) trial: A randomized
controlled trial of a primary care intervention for childhood overweight/mild
obesity. International Journal of Obesity, 31(4), 630-6.
Metcalf, B. S., Voss, L. D., Hosking, J., Jeffery, A. N., & Wilkin, T. J. (2008). Physical
activity at the government-recommended level and obesity-related health
outcomes: A longitudinal study (Early Bird 37). Archives of Disease in
Childhood, 93(9), 772-7.
Mitchell, J. A., Mattocks, C , Ness, A. R., Leary, S. D., Pate, R. R., Dowda, M., ...
Riddoch, C. (2009). Sedentary behavior and obesity in a large cohort of
children. Obesity, /7(8), 1596-602.
45
Phillipas, N. G., & Lo, C. W. (2005). Childhood obesity: Etiology, prevention, and
treatment. Nutrition in Clinical Care: An Official Publication of Tufts
University, 8(2), 77-88.
Powell, L. M., Auld, M. C , Chaloupka, F. J., O'Malley, P. M., & Johnston, L. D.
(2007). Associations between access to food stores and adolescent body mass
index. American Journal of Preventive Medicine, 33(4), S301-7.
Powell, L. M., Chaloupka, F. J., & Bao, Y. (2007). The availability of fast-food and
full-service restaurants in the United States: Associations with neighborhood
characteristics. American Journal of Preventive Medicine, 33(A), S240-5.
Wofford, L. G. (2008). Systematic review of childhood obesity prevention. Journal of
Pediatric Nursing, 25(1), 5-19.
U.S. Congress. (2004). Child nutrition and women infant child reauthorization act of
2004. Section 204 of Public Law 108-26. June 30, 2004.
U.S. Department of Health Services, National Institute of Diabetes and Digestive and
Kidnsey Diseases, Weight-control Information Network. (2007). Statistics
related to overweight and obesity. Retrieved from http://www.win.niddk.nih.
gov/statistics/index.htm.
U.S. Department of Health Services, HHS News. (December 13, 2001). Overweight
and obesity threaten U.S. health gains-communities can help address the
problem, surgeon general says. Retrieved from http://www.cga.ct.gov/ph/
medicaid/mmcc/special/Forum/Q AObesity.pdf.
46
Документ
Категория
Без категории
Просмотров
0
Размер файла
1 921 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа