School performance and weight status of children and young ... - Nature

International Journal of Obesity (1999) 23, 272±277 ß 1999 Stockton Press All rights reserved 0307±0565/99 $12.00 http://www.stockton-press.co.uk/ijo

School performance and weight status of children and young adolescents in a transitional society in Thailand L Mo-suwan*1, L Lebel2, A Puetpaiboon1 and C Junjana1 1

Department of Pediatrics, Faculty of Medicine, Prince of Songkla University, Hat Yai, Songkhla 90110, Thailand and 2Department of Epidemiology Unit, Faculty of Medicine, Prince of Songkla University, Hat Yai, Songkhla 90110, Thailand

OBJECTIVE: To study the association between current or previous weight status and school performance among children and young adolescents of Hat Yai municipality, southern Thailand. DESIGN: Cross-sectional and longitudinal study SETTING: Primary and secondary schools of Hat Yai municipality, southern Thailand. SUBJECTS: 1207 grades 3 ± 6 and 587 grades 7 ± 9 students. MEASUREMENTS: Body mass index (BMI, kg=m2) calculated from weight and height measurement of subjects in 1992 and 1994; parental education level and occupation, and monthly income, by questionnaire performed in 1992; gradepoint-average (GPA) and grades of mathematics and Thai language from the school records of ®nal examinations in 1994. RESULTS: Overweight subjects (BMI value > 85th percentile of the NHANES-I data for age and gender) in grades 7 ± 9 had a mean GPA 0.20 point (95% con®dence internal (CI) 0.04, 0.37) lower than that of the normal weight children after controlling for gender, age, school and grade. They were twice more likely to have low grades (lower than 2 on the scales of 0 ± 4) of mathematics and Thai language than normal weight children. There were no associations between GPA or individual subject grades and previous BMI status in 1992. Children in grades 7 ± 9 who became overweight over the two years, had a mean GPA of 0.48 point lower than those who did not become overweight (95% CI 0.12, 0.84). In grades 3 ± 6 subjects, however, becoming overweight had no effect on GPA and individual subject scores. CONCLUSIONS: Our study showed that being overweight and becoming overweight during adolescence (grades 7 ± 9) was associated with poor school performance, whereas such an association did not exist in children (grades 3 ± 6). Keywords: child; adolescent; overweight; school performance; transitional society

Introduction Obesity in children is an increasingly important nutritional disorder in developing countries.1,2 Persistence of obesity into adolescence results in increasing risk of adult obesity.3 Apart from health risks, obesity has been shown to be associated with social and economic disadvantages. In one study in the US, obese adolescents had poor high school performance,4 possibly affecting their acceptance into colleges.5 They also had fewer years of schooling than did their non-obese peers.6 Later in adulthood, women who had been overweight during adolescence had lower incomes when compared with their non-obese counterparts, independent of their baseline social status and aptitude-test scores.6,7 Correspondence: Dr Ladda Mo-suwan, Department of Pediatrics, Faculty of Medicine, Prince of Songkla University, Hat Yai, Songkhla 90110, Thailand. Received 24 March 1998; revised 21 September 1998; accepted 19 October 1998

A review of the literature revealed an inverse relationship between obesity and socioeconomic status among women in industrialized societies, whereas in the developing countries the relationship was direct among men, women and children.8 In Thailand, although in transition from an agronomybased to an industrial-based economy,9 the obesitysocioeconomic relation is like that of developing countries. Obesity was more prevalent in the af¯uent urban dwellers.10 From our previous study, children from the high income families were three times more likely to be obese than those from the low income group.11 However, there is little data on the school performance of obese children from developing countries. A report from China showed that the moderately severe obese ( > 50% overweight) school-aged children had average IQ scores 11 points lower than that of the non-obese controls.12 In this report, we examined the association between weight status and school performance among school children residing in an urban area in southern Thailand. We also explored the relation of school performance with the weight status measured two years

School performance and weight status L Mo-suwan et al

earlier, as a stronger test of whether academic achievement is a consequence of obesity.

Subjects and methods Subjects

A cohort of 2252 primary schoolchildren (grades 1 ± 6) was recruited in 1991 and 1992 using two-stage sampling. Six schools (two municipality-operated and four private) were randomly selected from 13 primary schools in the Hat Yai metropolitan area, southern Thailand. Then, one or two classrooms of each grade were randomly selected from each school. Baseline demographic and family data including parental education, occupation and monthly income were obtained by questionnaire completed by parents. Weight and height were measured using a Detecto1 beam balance and stadiometer (Detecto Scales Inc., Brooklyn, NY, USA) to the nearest 0.1 kg and 0.5 cm, respectively. Anthropometric data were collected on an annual basis (in January) thereafter. Parental consent was obtained and the study was approved by the Committee for Research in Humans, Faculty of Medicine, Prince of Songkla University. School performance

Compulsory education in Thailand is for six years. Children usually enter school at the age of 7 y old and leave primary school at the age of 13 y. The academic year begins in May and ends in March with a mid-year break in October. A grade system is used to score students' performance. Grades of a learning subject, in an ordinal scale of 0 ± 4, are obtained mainly from written examinations. Teacher's subjective assessment of students' performance accounts for approximately 10% of total scores. Grade-point-average (GPA), a continuous value of 0 ± 4, which re¯ects overall school performance, is calculated from grades of all subjects learned in that academic year. We obtained GPAs of our subjects from the school records at the end of March 1994. Grades of individual learning subject were available only for the secondary school (grades 7 ± 9) children. We used grades of mathematics and Thai language to explore different areas of learning ability in this group of children. Socioeconomic variables

We used the baseline socioeconomic data collected in 1992 and assumed that there were no or relatively minimal changes of grouping in the two years. Parental education was classi®ed as illiterate, primary, secondary and university level and occupation as casual worker, farmer, trader, government service and of®ce worker. Parental income variable was categorized into four classes by using monthly income cut at < 5000, 5 ± < 10 000, 10 000 ±

< 30 000 and 30 000 baht (1 baht 0.04 US dollar at the time of questionnaire). Statistical analysis

Association of school performance in 1994 with current (1994) and previous (1992) weight status was examined. Since primary and secondary schoolers may have a different school performance ± weight status relation, we did a separate analysis of grades 3 ± 6 and grades 7 ± 9 subjects. We used ANOVA and chi-square tests to describe and compare differences of variables. We used body mass index (BMI, the ratio of weight in kilograms (kg) divided by height squared in meters (m2)) to categorize weight status. Subjects were classi®ed as underweight, normal or overweight, if their BMIs' were < 15th percentile ( < P15), P15P85, > P85 of the US First National Health and Nutrition Examination Survey (NHANES I) data for age and gender.13 To examine the association of weight status with GPA, we computed multiple linear regression adjusted for gender, age, parental and family factors. Only factors with signi®cant partial F-test were retained in the model. Since standardized scores are not available for comparison across schools, computation of association was also adjusted for schools and grades. For the grades 7 ± 9 group, to analyze the grades for individual learning subject, dichotomous outcomes were created with scores less than 2 classi®ed as a low grade in that subject. We used logistic regression to examine the relationship between weight status and examination grades of each subject, also adjusted for the aforementioned confounders. Changes of BMI status over the two years from 1992 ± 1994 was categorized into four groups: stable-non-overweight, thinner (move from overweight to non-overweight), stable-overweight and becoming overweight (move from non-overweight to overweight). Multiple linear regression and logistic regression analysis were then performed accordingly to investigate whether the change in BMI status has any association with academic achievement. All analyses were performed using statistical software package STATA version 5.14

Results Of 2252 subjects, follow-up anthropometric measurements were completed for 2003 children in 1994. Schools could provide GPA of 1794 children (89.6%), 1207 grades 3 ± 6 and 587 grades 7 ± 9 students. Since subjects were recruited over two school years (1991=1992), the former grades 1 ± 4 became grades 3 ± 6 and the former grades 4 ± 6 became grades 7 ± 9 in 1994. A number of former grades 4 ± 6 subjects moved to other secondary schools, the total number of schools was ®nally 12.

273


274

We included only schools with more than 10 students in the analysis, resulting in dropping 30 subjects. The ®nal numbers of subjects were 1193 students in grades 3 ± 6 from 6 schools and 571 students in grades 7 ± 9 from 7 schools. Table 1 shows the mean GPA of grades 3 ± 6 and grades 7 ± 9 subjects by baseline (1992) family data. Several of these variables were associated. Parental occupation was closely associated with education. Likewise, mother's education and occupation were strongly associated with those of the father. Monthly income was also correlated with both parental education and occupation. All of these associations among variables were statistically signi®cant at the P < 0.001 level using chi-square tests. BMI status and GPA are shown in Table 2. The current prevalence of overweight using the 85th percentile cut-off of the NHANES-I BMI values for age and gender13 of grades 3 ± 6 and 7 ± 9 was 20.9% and 15.2%, respectively, and differed signi®cantly (w2 (2) 48, P < 0.001). The percentage of grades 3 ± 6 subjects becoming overweight was twice that of the grades 7 ± 9 subjects (w2 (3) 182, P < 0.001). For overall school performance, there were no demonstrable difference of mean GPA among the students in grades 3 ± 6 with different BMI status, whereas the overweight grades 7 ± 9 group had an average GPA 0.23 point lower than that of the normal weight and underweight groups (F(2,568) 1.13, P 0.03). For the grades 3 ± 6 group, a regression analysis showed that there was no signi®cant association of Table 1

GPA with current BMI status, the adjusted regression coef®cients and 95% con®dence intervals (95% CI) were 7 0.03 ( 7 0.13,0.06) and 0.06 ( 7 0.05,0.16) for underweight and overweight groups, respectively. The association with the previous (1992) BMI status was signi®cant for the overweight group, regression coef®cients were 7 0.03 (95% CI 7 0.14, 0.07) Table 2 Body mass index (BMI) statusa and grade-pointaverage (GPA)b of students in grades 3 ± 6 and 7 ± 9 groups

Gender male, n(%) 1994 Age (y), mean (s.d.) Current (1994) BMI status, n (%) Underweight Normal Overweight Previous (1992) BMI status, n (%) Underweight Normal Overweight Change of BMI category over two years, n (%) Stable-non overweight Thinner Stable-overweight Becoming overweight 1994 GPA of subjects by 1994 BMI category, mean (s.d.) Underweight Normal Overweight

Grades 7 ^ 9

556 (46.6) 10.7 1.1

247 (43.3) 13.8 0.9

275 (23.1) 668 (56) 250 (20.9)

69 (12.1) 416 (72.7) 87 (15.2)

265 (22.2) 724 (60.7) 204 (17.1)

123 (21.5) 350 (61.2) 99 (17.3)

919 24 180 70

456 28 71 16

(77.0) (2.0) (15.1) (5.9)

3.06 0.76 3.11 0.75 3.15 0.71

(79.9) (4.9) (12.4) (2.8)

2.60 0.78 2.60 0.77 2.37 0.71

a

BMI status categorized by using NHANES-I data for age and gender: < P15 underweight, P15-P85 normal, > P85 overweight; bGPA in a continuous scale of 0 ± 4.

Mean grade-point-average (GPA) by baseline (1992) family data of subjects in grades 3 ± 6 and grades 7 ± 9 groups Grades 3 ^ 6

Variables Fathers' occupation No Casual Farmer Trader Government of®cial Of®ce worker Mothers' occupation No Casual Farmer Trader Government of®cial Of®ce worker Fathers' education No Primary Secondary Higher Mothers' education No Primary Secondary Higher Parental monthly incomec < 5000 baht 5 ± < 10 000 baht 1 ± < 30 000 baht 30 000 baht a

Grades 3 ^ 6

a

n (%)

Mean GPA (s.d.)

4 (0.4) 227 (23.4) 22 (2.3) 387 (39.9) 258 (26.6) 71 (7.3)

2.85 1.11 3.05 0.72 2.70 0.67 3.24 0.69 3.16 0.77 3.23 0.75

185 160 20 412 188 41

(18.4) (15.9) (1.9) (40.9) (18.7) (4.1)

3.15 0.73 2.95 0.76 2.81 0.85 3.16 0.73 3.32 0.68 3.32 0.68

19 218 296 438

(1.9) (22.4) (30.5) (45.1)

2.09 0.74 3.04 0.71 3.13 0.73 3.24 0.74

35 378 197 396

(3.5) (37.6) (19.6) (39.4)

3.03 0.76 3.01 0.76 3.17 0.69 3.29 0.69

187 373 352 112

(18.3) (36.4) (34.4) (10.9)

2.97 0.72 3.06 0.74 3.30 0.69 3.29 0.67

Grades 7 ^ 9 b

P-value

0.0019

0.0000

0.0113

0.0000

0.0000

n (%)

Mean GPA (s.d.)

2 90 11 224 92 38

(0.4) (19.7) (2.4) (49.6) (20.1) (8.3)

2.31 0.49 2.39 0.69 2.48 0.64 2.70 0.73 2.61 0.73 2.56 0.80

86 58 10 213 86 15

(18.4) (12.4) (2.1) (45.5) (18.4) (3.2)

2.58 0.74 2.29 0.72 2.02 0.47 2.63 0.76 2.70 0.72 2.81 0.83

13 120 158 167

(2.8) (26.2) (34.5) (36.5)

2.78 0.63 2.46 0.73 2.58 0.75 2.70 0.76

24 182 101 163

(5.1) (38.7) (21.5) (34.7)

2.59 0.66 2.45 0.73 2.66 0.78 2.71 0.77

63 157 188 64

(13.3) (33.3) (39.8) (13.6)

2.31 0.68 2.54 0.77 2.64 0.79 2.89 0.69

P-valueb 0.0201

0.0003

0.0431

0.0130

0.0006

Numbers of children in each variable with non missing values; bChi-square test; c1 baht 0.04 US dollar at the time of study.


and 0.12 (95% CI 0.005, 0.24) for underweight and overweight groups, respectively. However, the marginal association of GPA with the overweight was no longer signi®cant after adjustment, regression coef®cient 0.09 (95% CI 7 0.01,0.19). Changes in BMI status had no signi®cant effect on GPA for the grades 3 ± 6 group, adjusted regression coef®cient with the non-overweight as the reference level 0.03 with 95% CI between 7 0.14 and 0.19 for the becoming overweight group, 0.09 (95% CI 7 0.02, 0.19) for the stable overweight, and 0.15 (95% CI 7 0.13, 0.43) for the thinner group. Table 3 shows the results of analysis for the association of GPA with BMI status and change of BMI category over the past two years of the grades 7 ± 9 group. It demonstrated that the overweight children had on average a GPA 0.24 (95% CI 0.06, 0.41) point lower than that of the normal weight children when no other possible confounding factors were taken into account. After adjustment for gender, age group, school and grade, the relationship with obesity was still signi®cant and only slightly changed, suggesting there was little confounding by these other variables. The adjusted mean GPA of overweight children was 0.20 point lower than normal weight children (95% CI 0.04, 0.37). However, all of these factors in the adjusted model explained approximately one ®fth of the variability of GPA (r2 of the adjusted model 0.18). Boys had a mean GPA 0.30 point lower than girls (95% CI 0.15,0.46). In particular, we could not detect any association between GPA and the previous (1992) BMI status, the adjusted regression coef®cients with the normal weight group as the reference were 7 0.02 (95% CI 7 0.17,0.14) and 7 0.09 (95% CI 7 0.26,0.08) for the underweight and overweight groups, respectively. The mean GPA of children in grades 7 ± 9 shows a signi®cant association with the changes of weight status over the past two years. After adjustment for confounders, children who became overweight had a mean GPA of 0.48 point lower than those who remained non-overweight. Overweight students in grades 7 ± 9 had a greater percentage with low grades of Thai language and

mathematics than the non-overweight subjects (Table 4). A crude analysis shows that overweight children were twice more likely to have low mathematics and Thai grades than normal weight children. Using logistic regression analysis, the associations still remain signi®cant after adjustment with variables similar to the GPA analysis. Males were about six times more likely than females to have low Thai grades (odds ratio (OR) 5.84, 95% CI 2.84, 11.99), but performance on mathematics was not different between genders (OR 0.85, 95% CI 0.49, 1.45). No association between individual subject grades and previous BMI status in 1992 could be detected. Likewise, there was no association between individual subject scores and changes of BMI categories over the two years.

Discussion Our study shows that in young adolescence (grades 7 ± 9), school performance as measured by GPA was associated with the current BMI status. Overweight in this group was related to a low mean GPA and signi®cant risks of having low Thai language and mathematics scores. An upward change in BMI status was associated with a greater risk of having a low mean GPA. Such effects of overweight on GPA scores were not seen in the younger children (grades 3 ± 6). The inverse relationship of obesity and being overweight, with socioeconomic indices, including academic achievement, observed in several studies,4 ± 8,15 could be viewed as either causal or consequent. To count low socioeconomic status (SES) as a cost of obesity, requires that low status be a consequence of obesity that would not occur if obesity were prevented.16 Follow-up studies of adolescent and young adult cohorts for 7 ± 12.5 y6,7,15 suggest that this may be the case. In these studies the association between baseline obesity and subsequent socioeconomic attainment was controlled for other possible determinants of

Table 3 Association of grade-point-average (GPA) with 1994 body mass index (BMI) status and changes of BMI status over the two years of the grades 7 ± 9 group by multiple linear regression analysis 1994 Mean GPA 1994 BMI statusa Underweight Normal Overweight Change of BMI status over two yearsb Thinner Stable-non overweight Stable-overweight Becoming overweight

2.60 0.78 2.60 0.77 2.37 0.71 2.68 0.74 2.59 0.77 2.43 0.69 2.11 0.73

Regression coefficient (95% CI)

P

0.06 ( 7 0.12, 0.25)

NS

7 0.20 ( 7 0.04, 7 0.37)

0.017

0.14 ( 7 0.12, 0.41)

NS

7 0.14 ( 7 0.32, 0.04) 7 0.48 ( 7 0.12, 7 0.84)

NS 0.008

95% CI 95% con®dence interval; NS non-signi®cant; anormal weight group reference level, regression analysis adjusted for age, gender, school, and grade; bstable-non overweight group reference level, regression analysis adjusted for age, gender, school, and grade.

275


276

Table 4 Association of low Thai language scores and low mathematics scores with 1994 body mass index (BMI) status and changes of BMI category over the two years, for the grades 7 ± 9 subjects by logistic regression analysis Subjects with low grades n/total(%)

Odds ratio (95% CI)

10/58 (17.2) 46/284 (16.2) 19/68 (27.9)

1.09 (0.48, 2.45) 1 1.98 (1.02, 3.85)

15/45 (33.3) 73/241 (30.3) 28/61 (45.9)

1.11 (0.53, 2.32) 1 2.08 (1.13, 3.83)

3/17 53/269 14/38 5/11

(15.0) (16.5) (26.9) (31.3)

0.69 (0.18,2.64) 1 1.67 (0.81,3.45) 3.07 (0.93,10.15)

NS

3/17 85/269 20/45 8/16

(17.7) (31.6) (44.4) (50.0)

0.37 (0.09,1.36) 1 1.84 (0.94,3.62) 2.35 (0.82,6.75)

NS

Low Thai language grades BMI statusa Underweight Normal Overweight Low mathematics grades BMI statusb Underweight Normal Overweight Low Thai language grades Change of BMI category over two yearsc Thinner Stable- non overweight Stable- overweight Becoming overweight Low mathematics grades Change of BMI category over two yearsd Thinner Stable-non overweight Stable-overweight Becoming overweight

P-value

NS 0.043 NS 0.018

NS NS

NS NS

NS non signi®cant; aNormal weight group reference level, regression analysis adjusted for age, gender, school, and grade; bnormal weight group reference level, regression analysis adjusted for age, gender, school, grade and father's education; cstable-non overweight group reference level, regression analysis adjusted for age, gender, school, and grade; dstable-non overweight group reference level, regression analysis was adjusted for age, gender, school, and grade.

SES such as, parental SES and test scores for intelligence. Our ®ndings of no association for poor academic performance with the previous BMI status, however, do not support these studies. If obesity results in stigmatization and social handicap17,18 causing poor school performance, our subjects who were obese in 1992 should have been more likely to have lower school grades than those who were not and the formerly overweight subjects who did not change over the two years of follow-up, not the newly overweight, should have performed worst. It is possible that our follow-up period of only two years is too short for the adverse effect of obesity on the academic achievement to be expressed. Or it re¯ects the different weight ± school performance relation of the transitional society of the developing countries. In our study, BMI status and parental SES account for 18% of the variability in GPA scores, suggesting the existence of other factors. It is possible that obesity and low school performance share some causes. Probable common factors are heredity and social environment. In¯uences of genetic factors on obesity are well established.19,20,21,22 Genetic and familial environmental in¯uences on educational attainment has also been demonstrated through the study of adoptees and their adoptive and biological fathers.23 This study of Danish adoptees revealed that SES of biological father in¯uenced the SES and years of schooling of offspring with whom they had no contact, probably through the inheritance of some personal characteristics. A report from the Colorado Adoption Project showed that cognitive ability and

academic achievement at the age of 7 y were correlated and genetic in¯uences accounted for 33 ± 64% of these observed correlations.24 These genetic and social environmental factors may be the main predictors of the association between overall school performance and BMI status in our study. Whether negative attitude and social discrimination against fat persons affect grade assignment is also a concern. Since subjective assessment accounted for only 10% of the ®nal score, it is less likely that teachers' biases would contribute to poor performance of the overweight subjects observed in our study. If discrimination and stigmatization did exist, children who remained overweight over the two years should have been worst affected and obesity associated poor school performance should have been demonstrated for both grades 3 ± 6 and grades 7 ± 9 groups. On the contrary, we only found an inverse relationship between academic performance and overweight in the grades 7 ± 9 students (aged 12 ± 18 y). This ®nding con®rms the late expression, more strongly in adulthood, of this relationship found in other studies.6,7 Adolescents have been shown to use eating as a means of coping with stress.25 Putting on weight, as well as the poor academic performance, in our grades 7 ± 9 group could possibly be the consequence of social environmental factors, for example, stress in school. The grades 3 ± 6 students, being younger, probably have other ways of dealing with the stress. Our previous study (unpublished observations) has shown that children younger than 10 y were less concerned about obesity than older children.


Conclusion In a transitional society setting, we found that the socioeconomic-obesity relation was typical of developing countries,11 whereas the school performanceobesity relation was similar to that of industrialized countries. We have shown that being overweight during adolescence (grades 7 ± 9, 12 ± 18 y) is associated with poor school performance. This unwanted association may be persistent if our society adopts the social value of slimness as it is now being highly publicized in most commercial advertisements. Long term follow-up of these subjects will clarify whether obesity during this period, in a transition society like Thailand, will have an unwanted social and economic consequence or not. Acknowledgements

We are grateful to the teachers of the studied schools for their cooperation in this study. We would like to thank Mrs Pranom Boonpan for her assistance in data collection. This study was supported by a grant from the Songklanagarind Hospital Foundation. References

1 Popkin BM, Richards MK, Montiero CA. Stunting is associated with overweight in children of four nations that are undergoing the nutrition transition. J Nutr 1996; 126: 3009 ± 3016. 2 Mo-suwan L, Junjana C, Puetpaiboon A. Increasing obesity in school children in a transitional society and the effect of the weight control program. Southeast Asian J Trop Med Public Health 1993; 24: 590 ± 594. (Erratum: Southeast Asian J Trop Med Public Health 1994; 25: 224) 3 Charney M, Goodman HC, Mcbridge M, Lyon B, Pratt R. Childhood antecedents of adult obesity: Do chubby infants become obese adults. N Engl J Med 1976; 295: 6 ± 9. 4 Canning H, Mayer J. Obesity: an in¯uence on high school performance. Am J Clin Nutr 1967; 20: 352 ± 354. 5 Canning H, Meyer J. Obesity ± its possible effect on college acceptance. N Engl J Med 1960; 275: 1172 ± 1174. 6 Sargent JD, Blanch¯ower DG. Obesity and stature in adolescence and earnings in young adulthood. Arch Pediatr Adolesc 1994; 148: 681 ± 687. 7 Gortmaker SL, Must A, Perrin JM, Sobol AM, Dietz WH. Social and economic consequences of overweight in adolescence and young adulthood. N Engl J Med 1993; 329: 1008 ± 1012.

8 Sobal J, Stunkard AJ. Socioeconomic status and obesity: A review of literature. Psychol Bull 1989; 105: 260 ± 275. 9 Kachondham Y, Winichagoon P, Tontisirin K. Nutrition and health in Thailand: trends and actions. UN ACC=SCN Country Case Study supported by UNICEF United Nation Children's Fund. A case study for the XV Congress of the International Union of Nutrition Sciences, September 26 ± October 1, 1993, Adelaide. 10 Tanpaichitr V, Kulapongse S, Pakpeankitvatana R, Leelahagul P, Tamwiwat C, Lochaya S. Prevalence of obesity and its associated risks in urban Thais. In: Oomura Y, Tarui S, Inoue S, Shinazu T (eds) Progress in Obesity Research 1990. London: John Libbey, 1991, 649 ± 653. 11 Mo-suwan L, Geater AF. Risk factors for childhood obesity in a transitional society in Thailand. Int J Obes 1996; 20: 697 ± 703. 12 Li X. A study of intelligence and personality in children with simple obesity. Int J Obes 1995; 19: 355 ± 357. 13 Must A, Dallal GE, Dietz WH. Reference data for obesity: 85th and 95th percentiles of body mass index (wt=ht2) and triceps skinfold thickness. Am J Clin Nutr 1991; 53: 839 ± 846. (Erratum: Am J Clin Nutr 1991; 54: 773) 14 StataCorp. Stata statistical software: Release 5.0. College Station. Stata Corporation: Texas. 1997. 15 Sonne-Holm S, Sorensen TIA. Prospective study of attainment of social class of severely obese subjects in relation to parental social class, intelligence, and education. BMJ 1986; 292: 586 ± 589. 16 Sorensen TIA. Socio-economic aspects of obesity: causes or effects? Int J Obes 1995; 19 (Suppl 6): S6 ± S8. 17 De Jong W. The stigma of obesity: the consequences of naive assumptions concerning the causes of physical deviance. J Health Soc Behav 1980; 21: 75 ± 87. 18 Berscheid E, Walster E. Physical attractiveness. Adv Exp Soc Psychol 1974; 7: 157 ± 215. 19 Stunkard AJ, Foch TT, Hrubec Z. A twin study of obesity. JAMA 1986; 256: 51 ± 54. 20 Stunkard AJ, Sorensen TIA, Harris C, Teasdale TW, Chakraborty R, Schull WJ, Schulsinger F. An adoption study of human obesity. N Engl J Med 1986; 314: 193 ± 198. 21 Sorensen TIA, Price RA, Stunkard AJ, Schulsinger F. Genetics of obesity in adult adoptees and their biological siblings. BMJ 1989; 298: 87 ± 90. 22 Stunkard AJ, Harris JR, Pedersen NL, McClearn GE. The body-mass index of twins who have been reared apart. N Engl J Med 1990; 322: 1483 ± 1487. 23 Teasdale TW, Sorensen TIA. Educational attainment and social class in adoptees: genetic and environmental contribution. J Biosoc Sci 1983; 15: 509 ± 518. 24 Wadsworth SJ, DeFries JC, Fulker DW, Plomin R. Cognitive ability and academic achievement in the Colorado Adoption Project: a multivariate genetic analysis of parent-offspring and sibling data. Behav Genet 1995; 25: 1 ± 15 25 Mellbin T, Vuille JC. Rapidly developing overweight in school children as an indicator of psychosocial stress. Acta Paediatr Scand 1989; 78: 568 ± 575.

277