Overweight Schoolchildren in NewYork State: Prevalence andCharacteristics

Wendy S. Wolfe, PhD, Cathy C. Campbell, PhD, EdwardA. Frongillo, Jr, PhD,Jere D. Haas, PhD, and ThomasA. Melnik, DrPH

IntroductionChildhood overweight is an increas-

ing public health concern, affecting asmany as 20% to 25% of children in theUnited States.'-3 According to one analy-sis, the prevalence of overweight in 6- to11-year-old children increased by 54%from the early 1960s to the late 1970s.2

Overweight children have an in-creased risk of adult obesity that increaseswith the age of the child and the severityof the overweight.4 In longitudinal stud-ies, 27% of overweight 1- to 5-year-olds,41% to 43% of overweight 3- to 9-year-olds, and 80% to 86% of overweight 10- to13-year-olds have remained overweight asadults.k7 Overweight children also haveincreased risk factors for heart diseaseand diabetes and increased emotionalstress, orthopedic disorders, and respira-tory disease.8

The present study was undertaken todetermine the prevalence of overweightand underweight in elementary schoolchildren in New York State outside ofNew York City and to identify characteris-tics associated with child fatness.

Methods

surveyed. Ages ranged from 6 to 12 years;the mean age of second graders was 7.9(SD = 0.5), and the mean age of fifthgraders was 10.9 (SD = 0.5).

Comparison of the students surveyedwith data on all children in New YorkState outside of New York City suggestedthat the sample was representative.9 As inthe population, the sample was predomi-nantly White, 18% lived with a singleparent, and just over half had motherswho were employed.

Nonparticipation of schools did notappear to relate to the prevalence ofoverweight. Reasons included small schoolsize, factors related to the dental compo-nent of the study,10 and a misperceptionthat the survey was intended only forlow-income children.

Within the 51 schools, only twosamples differed significantly from theirschool population in the proportion ofnon-White students, and two others dif-fered in the proportion of students eli-gible for free or reduced-price schoollunches. Because this number of differ-ences would be expected as a result ofsampling error, no schools were omittedfrom the analysis.'1

Sample Selection andRepresentativeness

A two-stage sampling technique wasused. First, schools were randomly se-lected within 21 strata: 7 geographicalregions in New York State outside ofNewYork City and 3 socioeconomic status(SES) levels within each of these regions(based on 1980 census data). Of the 110schools contacted by letter, 51 (46%)volunteered to participate. Parent con-sent forms with brief questionnaires weregiven to second and fifth graders; 1797(51%) agreed to participate and were

Wendy S. Wolfe, Edward A. Frongillo, Jr, andJere D. Haas are and Cathy C. Campbell waswith the Division of Nutritional Sciences,Cornell University, Ithaca, NY. Thomas A.Melnik is with the Division of Nutrition, NewYork State Department of Health, and theSchool of Public Health, State University ofNew York at Albany.

Requests for reprints should be sent toWendy S. Wolfe, PhD, Division of NutritionalSciences, MVR Hall, Cornell University, Ithaca,NY 14853-4401.

This paper was accepted September 1,1993.

Editores Note. See related editorial byNestle (p 713) in this issue.

American Journal of Public Health 807

TABLE 1 -Prevalence Estimates (%) of Overweight in the Study Sample and Inthe Northeast NHANES II Subsample

NortheastNew York State Outside New York City, 1987/88 United States

1976-1980NHANESI NHANESII (NHANES II)

Percentile BMI TSF AFA BMI TSF BMI TSF

> 85th 28 23 NA 23 19 25 24>90th NA NA 19 17 14 20 15>95th 13 13 12 10 8 8 8

Note. Variation depended on the reference data used for comparison (National Health and NutritionExamination Survey [NHANES] [1971-1974J20 or NHANES II [1976-1980J15), the cutoff level(percentile) used to define overweight, and the indicator used to measure child fatness (bodymass index [BMIJ, triceps skinfold [TSF], or arm fat area [AFA]).

Body Mass Index (kg/m2)

20 4) O0Percentile

Arm Fat Area (mm2

80 100 Pere 60Percentile

100

FIGURE 1-Sample arm fat area and body mass Index cumulative distributionscompared with the first and second National Heaith and NutritionExamination Survey (NHANES 114 and NHANES 1115) nationalreference distributions, by sex.

Data Collection and Measurements

The best way to measure childfatness in the field is unclear. Variousauthors have recommended body mass

index (weight/height2),12'13 triceps skin-fold,13 or arm fat area14 (calculated fromtriceps skinfold and mid-upper-arm cir-cumference: arm fat area [mm2] = arm

area [A] - arm muscle area [M], whereA = rr/4 x [C/IT]2 and M = [C - TrT]2/4'rr [C = arm circumference, T = tricepsskinfold]) as the best indicator of body fatin school-aged children.

In this study, body mass index,triceps skinfold, and arm fat area were allused to measure child fatness. One of twotrained examiners used standard proce-

dures to measure weight, height, mid-upper-arm circumference, and tricepsskinfold on each child. The principalexaminer (first author) measured 1615children between November 1987 andMarch 1988 and also administered a

nonquantitative 24-hour dietary recall.The other examiner measured 182 chil-dren in the spring of 1987.

Measurement errors16 for the fouranthropomnetric indices were acceptable.Intraobserver (test-retest) reliabilities foreach examiner ranged from 97% to 100%.Interobserver reliabilities, based on inde-pendent samples, were 93% to 100%between the principal examiner and an

anthropometric expert and 96% to 100%

between the principal and secondaryexaminers.

The best way to define overweight orunderweight in children is also unclear.Therefore, various percentile cutoffs, aswell as reference data from both the firstand second National Health and Nutri-tion Examination Surveys (NHANES Iand NHANES II), were used for compari-son purposes in deriving prevalence esti-mates. In addition, because obesity ratesare known to be higher in the Northeastthan in other regions of the UnitedStates,2'7 prevalence estimates for 6- to12-year-olds in the Northeast subsampleof NHANES II, with the total NHANESII distribution as the reference, also werecalculated. Logistic regression was used toadjust the seasonality of this subsample tomatch that of the New York sample(November through March), since preva-lence rates also vary with season.2'17

Five diet quality indices were derivedfrom the recall: food diversity (number ofdifferent food items eaten), skipping ofbreakfast, consumption of no vegetableother than potatoes or tomato sauce,number of snack foods eaten, and a foodgroup pattern score ("poor" defined asmissing one or more of the four foodgroups or consuming two or more groupsonly one time each). Children who skippedbreakfast were asked whether this wastypical; most said it was.

On the basis of the parent question-naire, children were classified as being oflow SES if they met at least one of thefollowing criteria: no parent working,eligibility for free or reduced-price schoollunches, or receipt of social or foodassistance. Children were classified asmedium/high SES only if they met noneof these criteria and had values for allthree variables. Twenty-six percent of thechildren were considered low SES. Inschools involved in the National SchoolLunch Program, children were classifiedas participants if their parent respondedpositively to the question "Does yourchild eat school lunch?"

StatisticalAnalysisRegression models were developed

to identify factors characterizing fatterfrom leaner children after adjustment forpossible confounding and interactive ef-fects. Factors examined were height,grade, age within grade, sex, race (Whiteor African American), SES, family struc-ture, school lunch participation, numberof siblings, maternal employment, and thefive dietary indices. ("Whites" includedthe small number ofHispanics and "others";

808 American Journal of Public Health

Wolfe et al.

Boys

Girls

24NYS Sample

(n=895)2D

16 NHANES-II

14

122624 NYS Sample

22 (n=897)24 -

16 NHAN~~~~~ES-II.s

May 1994, Vol. 84, No. 5

Overweight Schoolchildren

the body fatness of these groups did notdiffer significantly from that of Whites.)Because all variables were examined simul-taneously for their effects after adjust-ment for the others, children with missingvalues for any of the variables, includingthose in nine schools that did not partici-pate in the National School Lunch Pro-gram and those in six schools for whichdietary data were not available, wereomitted from the regression analyses.

Regression models were developedfor each of six different anthropometricdependent variables. Two ordinary leastsquares models were developed for bodymass index and arm fat area treated ascontinuous variables. Because their distri-butions were positively skewed and theirvariances increased with age and height,both were transformed into natural loga-rithms to make the regression residualsmore symmetrically distributed.

To test for the risk of overweight orunderweight, four additional models wereexamined with multiple logistic regres-sion; body mass index and arm fat areawere dichotomized into (a) overweight(>90th percentile) vs normal weight(< 90th and > 10th percentile) and (b)underweight ( < 10th percentile) vs nor-mal weight. Body mass index values werecompared with age- and sex-specific refer-ence percentiles from the NHANES IInational probability sample survey,'5 whilearm fat area values were compared withreference percentiles developed by Frisan-cho from NHANES I.14 Although derivedfrom different surveys, these were theonly published national reference percen-tiles available for body mass index andarm fat area when the regression analyseswere done. For consistency in the analy-ses, the 90th percentile was selected as thecutoff for both indices (the 85th percentilewas not available for arm fat area).

For the least squares analyses, astepwise multiple regression procedurewas run in which all main effects wereretained and each conceptually plausibletwo-way interaction effect was tested; theleast significant interaction was removedat each step, and those with a P value of.10 or less were retained.'8"19 This proce-dure was repeated retaining all significanttwo-way interactions and their inclusivemain effects and testing the remainingmain effects (retaining those with a Pvalue of less than .05). The resultingmodels were then tested with logisticregression; variables were removed bymeans of a similar stepwise procedure.

TABLE 2-Sample Distributions of Physical Indices, Compared with Age- andSex-Specific National Reference Percentiles

Second FifthBoys Girls Graders Graders Tolal

Indices and (n = 900), (n = 897), (n = 960), (n = 837), (n = 1797),Percentiles % % % % %

Body mass indexa< 5th 2.3 5.0 4.0 3.2 3.7< 10th 7.0 8.2 6.9 8.3 7.7> 85th 20.9 24.5 23.0 22.4 22.6> 90th 15.6 18.3 18.3 15.5 16.9> 95th 8.7 10.9 8.8 8.3 9.8

Triceps skinfolda< 5th 4.5 5.5 4.9 5.2 5.0<10th 10.7 12.5 11.3 12.0 11.6>85th 15.9 21.8 19.0 18.8 18.9>90th 12.0 16.3 13.4 15.1 14.2> 95th 7.4 9.6 10.7 8.5 8.5

Arm fat areab< 5th 4.3 4.8 3.3 5.9 4.5<10th 8.4 11.4 9.7 10.1 9.9>90th 18.3 20.1 17.9 20.8 19.3>95th 13.2 11.7 13.7 11.1 12.5

Midarmcircumferencea< 5th 4.3 4.1 4.1 4.4 4.2> 95th 8.9 10.0 10.5 8.2 9.4

Height for agec< 5th 4.0 3.5 3.2 4.5 3.8> 95th 7.2 6.0 6.0 5.8 6.6

aCompared with the second National Heafth and Nutrition Examination Survey (NHANES 11).15bCompared with NHANES 1.14cCompared with National Center for Health Statistics percentiles.21

ResultsPrevalence Estimates for Overweightand Underweight

The estimated prevalence of over-weight in the sample (Table 1) varied with(a) the reference data used for compari-son,'1520 (b) the cutoff level used to defineoverweight, and (c) the indicator used tomeasure child fatness. Regardless, theprevalence was higher (by as much astwice) than that expected based on the USreference populations. Prevalence esti-mates differed little from those for theNortheast subsample of NHANES II,however.

The sample anthropometric distribu-tions are shown in Figure 1 and Table2.14,15,21 A higher than expected percent-age of children had values in the upperpercentiles for body mass index, tricepsskinfold, and arm fat area, while thepercentage with values in the lowerextremes (underweight) was similar to orless than expected. Weight status did notvary by grade or sex; within these catego-ries, body mass index measurements were

correlated significantly with both arm fatarea (coefficients of .92 to .94) and tricepsskinfold (.87 to .90).

CharactersticsAssociated withChild Fatness

In the least squares models (Table3), arm fat area and body mass index wereexamined as unadjusted raw variables;therefore, most of the explained variationwas attributable to biological variables(R2s for the biological models alone were25.4% and 11.4% for arm fat area andbody mass index, respectively, comparedwith 28.5% and 12.2% for the fullmodels). As expected, fatness increasedsignificantly with grade (age) and height,girls had higher arm fat area indices thanboys,22 and African Americans had lowerarm fat area indices than Whites.23

In the arm fat area model, allsociodemographic variables entered ex-cept maternal employment were relatedto child fatness after adjustmeni for theothers. In the body mass index model,similar results were found, although somerelationships were slightly less significant.

American Journal of Public Health 809May 1994, Vol. 84, No. 5

TABLE 3 Multiple Least Squares Regression Models of Child Fatness as aContinuous Variable

Dependent Variable

Arm Fat Area Body Mass Index(n = 1135) (n = 1138)

Regression RegressionExplanatory Variable Coefficient SE Coefficient SE

Biological variablesGrade (5th vs 2nd) .36* .030 .10* .010Age within grade, ma -.007* .003 .003* .001Height within grade, cma .046* .005 ...b ...

Sex (female vs male) .28* .058 .034 .019Race (African American vs White) -.32* .062 -.040 .021

Sociodemographic variablesSES score (medium/high vs low) -.094 .052 -.031 .018Family structure (single -.15* .062 -.046* .021

parent vs dual parent)School lunch (participants .12* .037 .039* .012vs nonparticipants)

Siblings 1 (single child .10* .046 .020 .015vs 1 or more siblings)

Siblings 2 (3 or more siblings -.084* .042 -.027 .014vs 2 or less)

Breakfast skipping (no breakfast .11 * .056 .063* .019vs some breakfast)

InteractionsSES x Family Structure .22* .084 .072* .028SES x Height -.014* .005 ... ...

SES x Sex -.10 .067 -.032 .022Breakfast x Family Structure -.26* .12 -.14* .038

Intercept 6.76* .060 2.81* .020

Note. SES = socioeconomic status.aCentered around the mean for each grade.bNot entered (already adjusted for in dependent variable).*P < .05.

Low-SES children tended to be fatter (byabout 5%) than medium/high-SES chil-dren, except in single-parent households(an interaction). The inverse relationshipof fatness with SES was stronger in girlsthan in boys; although this interaction wasslightly less significant than the P < .10cutoff, it was kept in the model because ofthe similar findings in adults.

Children in single-parent familiestended to be thinner than those intwo-parent families. Children who partici-pated in school lunch tended to be slightlybut significantly fatter than those who didnot; this relationship held even whenchildren eligible for free or reduced-pricelunches were omitted from the compari-son. Children with no or few siblingstended to be fatter than those with manysiblings. Finally, children who skippedbreakfast tended to be fatter than thosewho ate breakfast, except in single-parenthouseholds. The other dietary indicesexamined did not relate to child fatness.

On the basis of the body mass indexmodel, these relationships translate intolow-SES children being from 0.4 to 1.8 kg(1 to 4 lb) heavier (depending on theirother characteristics), single-parent chil-dren being up to 0.4 kg (1 lb) lighter,school lunch participants being 0.7 to 0.9kg (1.5 to 2 lb) heavier, children with no

siblings being 0.4 to 0.7 kg (1 to 1.5 lb)heavier, and children with three or more

siblings being 0.4 to 0.7 kg (1 to 1.5 lb)lighter than their respective counterparts.

CharacteristicsAssociated withOvenweight and Underweight

On the basis of chi-square tests withno adjustment for other variables, therewas significantly more overweight ( > 90thpercentile for body mass index or arm fatarea) in children with few or no siblings,those with employed mothers, and thosewho ate school lunch. There was more

underweight ( < 10th percentile) in chil-dren with single parents.

The logistic regression models (Table4) mostly supported the findings of theleast squares models. African-Americanchildren were slightly less likely to beoverweight, based on body mass index,and much more likely to be "underfat,"based on arm fat area, than Whitechildren. In two-parent households, me-

dium/high-SES children were about halfas likely as low-SES children to beoverweight and slightly less likely to beunderweight. In single-parent house-holds, on the other hand, SES was notsignificantly related to overweight or

underweight, except for a trend towardmedium/high-SES children being heavierthan low-SES children.

Compared with children who did noteat school lunch, children who ate schoollunch were about half as likely to beunderweight and somewhat, but not sig-nificantly, more likely to be overweight.Children with no siblings were somewhatmore likely, and those with three or more

siblings significantly less likely, to beoverweight than those with one or twosiblings. In two-parent families, childrenwho skipped breakfast were almost twiceas likely as breakfast eaters to be over-

weight; they were less likely to be under-weight, as measured by arm fat area, thanthose who ate breakfast.

DiscussionBecause the exact degree of child

fatness related to later health problems isunknown, there is no health-based defini-tion of child overweight. Its diagnosis thusremains statistical rather than pathologi-cal, and the prevalence in a given popula-tion depends on the reference data, thecutoff level, and the indicator used.

The most useful reference data are

those representative of the US popula-tion. Therefore, the sample measure-

ments were compared with the nationalNHANES I and NHANES II survey data.Because children in the nation as a wholebecame fatter during the 1970s,2 theprevalence of overweight in the sample ishigher when compared with NHANES Inational percentiles than with NHANESII percentiles, even though both data setsare representative samples of healthyindividuals in the United States.

As far as cutoff level, values over the95th percentile clearly represent a nutri-tional concern. Many also recommend use

of the 85th percentile to define over-

weight in children, as is used increasinglyin adults.2426 All three of the fatnessindicators resulted in similar prevalences

810 American Journal of Public Health

Woffe et al.

May 1994, Vol. 84, No. 5

Overweight Schoolchildren

TABLE 4-Odds Ratios (95% Confidence Intervals) of Being Overweight or Underweight Compared with Normal Weight, Basedon Logistic Regression, for Each Explanatory Characteristic

Dependent Variables

Overweight Underweight(> 90th Percentile) (10th Percentile)

Explanatory Body Mass Arm Fat Body Mass Arm FatCharacteristic Index Area Index Area

African American vs White 0.43a (0.19, 0.98) 0.87 (0.44,1.73) 1.57 (0.72, 3.43) 3.63a (1.94, 6.80)

Medium/high vs low SESb 0.53a (0.35, 0.80) 0.59a (0.40, 0.87) 0.57 (0.30,1.08) 0.79 (0.42,1.47)In 2-parent familiesIn 1-parent families 2.01 (0.93, 4.36) 1.14 (0.53, 2.43) 0.76 (0.26, 2.25) 0.72 (0.28,1.83)

Single-parent vs dual-parent familybIn low-SES families 0.54 (0.27,1.08) 0.62 (0.33,1.16) 0.90 (0.38, 2.12) 1.28 (0.59, 2.81)In medium/high-SESfamilies 2.05a (1.12, 3.77) 1.20 (0.66, 2.18) 1.21 (0.48, 3.07) 1.17 (0.52, 2.63)

School lunch participants 1.48 (0.96, 2.27) 1.40 (0.95, 2.08) 0.66 (0.38,1.14) 0.51a (0.32, 0.82)vs nonparticipants

Single child vs 1-2 siblings 1.46 (0.91, 2.34) 1.35 (0.88, 2.08) 1.51 (0.79, 2.88) 0.96 (0.50,1.83)

3 or more siblings vs 1-2 0.60a (0.37, 0.98) 0.50a (0.31, 0.82) 0.66 (0.32,1.37) 0.90 (0.50,1.61)

No breakfast vs somebIn 2-parent families 1.95a (1 .17, 3.25) 1.51 (0.92, 2.46) 1.36 (0.59, 3.17) 0.23a (0.06, 0.97)In 1-parent families 0.64 (0.20, 2.03) 0.43 (0.12,1.54) 0.69 (0.15, 3.28) 0.61 (0.17, 2.22)

Note. The ratios for each explanatory characteristic have been adjusted for the other characteristics. SES = socioeconomic status.aSignificant based on 95% confidence interval not including odds ratio of 1.0.bSignfficant interaction with following variable.

of overweight, although body mass indextended to give higher prevalences thantriceps skinfold or arm fat area.

Regardless of the definition, theprevalence of overweight in the sample issubstantially higher than that in thenation in 1974 or 1980. This is a concernbecause it may result in increased rates ofadult obesity and rising health care costs,supporting the need for obesity preven-tion efforts starting in childhood. Whilethe results probably reflect a greatercurrent prevalence of child overweight inNew York State than in the nation overall,similar to that found in the Northeastsubsample of NHANES II, children na-tionally may well have increased in fatnessbetween the most recent national surveyand the present New York survey.

The prevalence of underweight inthe study sample was similar to that in thenation in 1980, suggesting less of aproblem than for overweight. However,this finding does not imply adequatedietary quality or the absence of at leastepisodic hunger (food insecurity) or thatsome children are not experiencing healthproblems due to being underweight.

The multivariate analyses identifiedseveral factors related to the degree offatness in children. The lower arm fat areabut not body mass index values of African

Americans relative to Whites, similar toskinfold vs body weight findings fromother studies,2226 may relate to racialdifferences in fat patterning rather than todifferences in overall fatness.

The relationship between SES andchild overweight is unclear in the litera-ture. An inverse relationship like thatrevealed in this study was found in threeother large US studies, including theNational Health Examination Survey.27-29However, a positive relationship wasfound in the Ten State Nutrition Survey(except for an inverse relationship in lateteen girls)30 and the Tecumseh commu-nity study.31 No relationship was found inthree other large US studies.32-" In anextensive review of studies of all sizes inthe United States and other developedcountries, no clear trend emerged.35 How-ever, most of the relationships in girlswere in the inverse direction, similar tothe trend found in the present study andcorresponding to differences found inadults.35 This suggests that the trendtoward females of lower SES being fattermay begin in childhood.

The interaction between SES andfamily structure, such that low-SES chil-dren tended to be fatter than higher SESchildren except in single-parent house-holds, may be due to single-parent low-

SES families being even poorer (in allnecessary resources) than two-parent low-SES families. Although further study andbetter measures of SES are needed, thissuggests a nonlinear relationship betweenSES and overweight, such that overweightis lowest in the very poor and in thewealthy and highest in the somewhatpoor. Of four other studies that examinedthe relationship between child overweightand single parenthood, two showed apositive association3637 and two showedno association3839; however, none con-trolled for SES.

A positive relationship between childfatness and participation in school lunchalso was found in the national evaluationof the School Lunch Program.40 Thecausal nature of this relationship is un-clear, however. For example, childrenwho choose not to eat school lunch couldbe leaner because they are pickier eatersor have more health-conscious parents.On the other hand, school lunches havebeen suggested to be high in fat,4lA3 whichcould lead to greater fatness in children.

The inverse relationship between thenumber of children in the family and childfatness has been found in a number ofother studies,364445 but the mechanismremains unclear. Children with moresiblings may be more physically active.

American Journal of Public Health 811May 1994, Vol. 84, No.5

Wolfe et al.

The positive association betweenbreakfast skipping and overweight alsowas found in a study of adolescents,46 butthe pathway is unclear. Those who skipbreakfast may eat more later in the day orbe less active, or already-overweight chil-dren may tend to skip breakfast in anattempt to lose weight.

Maternal employment was not re-lated to child overweight, as found else-where.3844 Other factors reported to re-late to child overweight but not examinedhere include activity level and televisionviewing.47

In conclusion, the prevalence ofoverweight among school-aged children inNew York State outside ofNew York City(14% to 28%, depending on the defini-tion) is high in comparison with nationalreference data. Children who tended tobe fatter were members of low-SES,two-parent households, those with few orno siblings, those who ate school lunch,and those who skipped breakfast. Thesevariables may be useful for targetinginterventions to prevent child overweight,although further research is needed tobetter understand the relationships andtheir mechanisms and to identify otherrelated factors. O

AcknowledgmentsSupport for this work was provided by theNutrition Surveillance Program of the NewYork State Department of Health and aNational Institutes of Health nutrition traininggrant (DK07158).

We would like to thank Dr Jay Kumarand Annette Palk of the New York StateDental Bureau for sample selection and recruit-ment, Dr Tom Santner and Gail Rothe for helpwith data analysis, Janet Weber for the initialdata collection, Dr Mary Fran Sowers and DebSpicer for help in the initial design of theproject, and Gayle Leitch Kelly for her adminis-trative support.

References1. Dietz WH. Childhood obesity: susceptibil-

ity, cause, and management. J Pediatr.1983;103:676-686.

2. Gortmaker SL, Dietz WH, Sobol AM,Wehler CA. Increasing pediatric obesity inthe United States.AmJDis Child. 1987;141:535-540.

3. Ross JA, Gilbert G. The National Childrenand Youth Fitness Study: a summary offindings. J Phys Educ Recreation Dance.1985;56(suppl 1):45-50.

4. Rimm IJ, Rimm A. Association betweenjuvenile onset of obesity and severe adultobesity in 73,532 women. Am J PublicHealth. 1976;66:479-489.

5. Gain SM, LaVelle M. Two-decade fol-low-up of fatness in early childhood. Am IDis Child. 1985;139:181-185.

6. Cronk CE, Roche AF, Chumlea WC, KentR. Longitudinal trends of weight/stature2

in childhood in relationship to adulthoodbody fat measures. Hum Biol. 1982;54:751-764.

7. Abraham S, Nordsieck M. Relationship ofexcess weight in children and adults. PublicHealth Rep. 1960;75:263-273.

8. Dietz WH. Obesity in infants, children,and adolescents in the United States, I.Identification, natural history, and afteref-fects.NutrRes. 1981;1:117-137.

9. Wolfe WS, Campbell CC. Food pattern,diet quality and related characteristics ofschool children in New York State. JAmDietAssoc. 1993;93:1280-1284.

10. Kumar J, Green E, Wallace W, Bustard R.Changes in dental caries prevalence inupstate New York school children. JPublicHealth Dent. 1991;51:158-163.

11. Snedecor GW, Cochran WG. StatisticalMethods. 7th ed. Ames, Iowa: Iowa StateUniversity Press; 1980.

12. Cole TJ. A critique of the NCHS weight forheight standard. Hum Biol. 1985;57:183-196.

13. Roche AF, Siervogel RM, Chumlea WC,Webb P. Grading body fatness from lim-ited anthropometric data. Am J Clin Nutr.1981;34:2831-2838.

14. Frisancho AR. New norms of upper limbfat and muscle areas for assessment ofnutritional status. Am J Clin Nutr. 1981;34:2540-2545.

15. Anthropometric reference data and preva-lence of overweight, United States, 1976-80. Vital Health Stat [11]. 1987;238. DHHSpublication PHS 87-1688.

16. Marks GC, Habicht J-P, Mueller WH.Reliability, dependability, and precision ofanthropometric measurements: the Sec-ond National Health and Nutrition Exami-nation Survey 1976-1980. Am J Epidemiol.1989;130:578-587.

17. Dietz WH, Gortmaker SL. Factors withinthe physical environment associated withchildhood obesity.Am J Clin Nutr. 1984;39:619-624.

18. Kleinbaum DG, Kupper LL, MorgensternH. Epidemiologic Research. Belmont, Calif:Lifetime Learning Publications; 1982.

19. SAS User's Guide: Statistics, Version 5Edition. Cary, NC: SAS Institute Inc; 1985.

20. Must A, Dallal GE, Dietz WH. Referencedata for obesity: 85th and 95th percentilesof body mass index (wt/ht2) and tricepsskinfold thickness.Am J Clin Nutr. 1991;53:839-844.

21. Hamill PV, Drizd TA, Johnson CL, et al.Physical growth: National Center for HealthStatistics percentiles. Am J Clin Nutr.1979;32:607429.

22. Tanner JM. Foetus into Man: PhysicalGrowth from Conception to Maturity. Cam-bridge, Mass: Harvard University Press;1978.

23. Harsha DW, Voors AW, Berenson GS.Racial differences in subcutaneous fatpattern in children aged 7-15 years. Am JPhysAnthropol. 1980;53:333-337.

24. The Surgeon General's Report on Nutritionand Health. Washington, DC: US Dept ofHealth and Human Services; 1988. DHHSpublication PHS 88-50210.

25. Healthy People 2000: NationalHealth Promo-tion and Disease Prevention Objectives.Washington, DC: Govt Printing Office, USDept of Health and Human Services; 1990.DHHS publication PHS 91-50212.

26. National Research Council, Committee onDiet and Health. Diet and Health: Implica-tions for Reducing Chronic Disease RiskWashington, DC: National Academy Press;1989.

27. Determinants ofChildren 's Health. Washing-ton, DC: National Center for HealthServices Research; 1981. DHHS publica-tion PHS 81-3309.

28. Stunkard AJ, D'Aquili E, Fox S, FilionRDL. The influence of social class onobesity and thinness in children. JAMA.1972;221:579-584.

29. Huenemann RL, Shapiro LR, HamptonMC, Mitchell MW. A longitudinal study ofgross body composition and body conforma-tion and their association with food andactivity in a teenage population.Am J ClinNutr. 1966;18:325-338.

30. Garn SM, Clark DC. Nutrition, growth,development and maturation: findings ofthe Ten State Nutrition Survey of 1968-70.Pediatrics. 1975;56:306-319.

31. Garn SM, Hopkins PJ, Ryan AS. Differen-tial fatness gain of low income boys andgirls.AmJ Clin Nutr. 1981;34:1465-1468.

32. Khoury PR, Morrison JA, Laskarzenski P,et al. Relationships of education andoccupation to coronary heart risk factors inschoolchildren and adults: the PrincetonSchool District study. Am J Epidemiol.1981;113:378-395.

33. Hunter S, Frerichs PR, Webber LS, Beren-son GS. Social status and cardiovasculardisease risk factor variables in children: theBogalusa Heart Study. J Chronic Dis.1979;32:441-449.

34. Malina RM. Patterns of development inskinfolds of Negro and White Philadelphiachildren. Hum Biol. 1966;38:89-103.

35. Sobal J, Stunkard AJ. Socioeconomic sta-tus and obesity: a review of the literature.PsycholBull. 1989;105:260-275.

36. Jacoby A, Altman DG, Cook J, HollandWW. Influence of some social and environ-mental factors on the nutrient intake andnutritional status of schoolchildren. Br JPrev Soc Med. 1975;29:116-120.

37. Garman AR, Chinn S, Rona RJ. Compara-tive growth of primary schoolchildren fromone and two parent families. Arch DisChild. 1982;57:453-458.

38. Bowering J, Wynn RL. Nutritional statusof preschool children from intact anddivorced families. Home Econ Res J. 1986;15:132-140.

39. Darke SJ, Disselduff MM, Try GP. Anutrition survey of children from one-parent families in Newcastle upon Tyne in1970. BrJNutr. 1980;44:237-241.

40. Vermeersch J, Hanes S, Gale S. Thenational evaluation of school nutritionprograms: program impact on anthropo-metric measures. Am J Clin Nutr. 1984;40:414-424.

41. Dillon MS, Lane HW. Evaluation of theoffer vs. serve option within self-serve,choice menu lunch program at the elemen-tary school level. JAm DietAssoc. 1989;89:1780-1785.

42. Coale EK, Bedford MR. Fat-controlledmenus in a school lunch program. Sch FoodServRes Rev. 1984;8:37-41.

43. Morris PM. Heading for a Health Crisis:Eating Patterns of America's School Chil-dren. Washington, DC: Public Voice forFood and Health Policy; 1991.

812 American Journal of Public Health May 1994, Vol. 84, No. 5

44. Rona RJ, Chinn S. National Study ofHealth and Growth: social and biologicalfactors associated with weight-for-heightand triceps skinfold of children from ethnicgroups in England.Ann Hum BioL 1987;14:231-248.

45. Samuelson G, Grahnen H, Arvidsson E.

An epidemiological study of child healthand nutrition in a northern Swedish county,VI: relationship between general and oralhealth, food habits, and socioeconomicconditions. Am J Clin Nutr. 1971;24:1361-1373.

46. Resnicow K. The relationship between

Overweight Schoolchildren

breakfast habits and plasma cholesterollevels in schoolchildren. J Sch Health.1991;61:81-85.

47. Gortmaker SL, Dietz WH, Cheung LWY.Inactivity, diet, and the fattening ofAmerica. J Am Diet Assoc. 1990;90:1247-1255.

May 1994, Vol. 84, No. 5 American Journal of Public Health 813