Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
Published online: October 2, 2014
Ethnic Disparities in General and Abdominal Adiposity at School Age: A Multiethnic Population-Based Cohort Study in The Netherlands Olta Gishti a–c Claudia J. Kruithof a, b Janine F. Felix b, c Hein Raat d Albert Hofman b Liesbeth Duijts b, c Romy Gaillard a–c Vincent W.V. Jaddoe a–c a Generation R Study Group, and Departments of b Epidemiology, c Pediatrics and d Public Health, Erasmus University Medical Center, Rotterdam, The Netherlands
Abstract Background/Aims: Ethnic differences in obesity prevalence have been reported. We examined ethnic differences in general and abdominal fat distribution in school-age children and the influence of parental prepregnancy, pregnancy, and childhood factors on these differences. Methods: We performed a multiethnic population-based prospective cohort study among 5,244 children with information about prepregnancy factors, pregnancy, and childhood factors. At the age of 6 years, the BMI, total fat mass, android/gynoid fat mass ratio, and preperitoneal fat mass were assessed via dual-energy X-ray absorptiometry and abdominal ultrasound. Results: The overweight and obesity prevalences among Dutch children were 10.0 and 2.1%, respectively. Higher prevalences were observed among Cape Verdean (21.0 and 10.3%), Dutch Antillean (18.4 and 13.8%), Moroccan (20.6 and 7.7%), Surinamese-Creole (13.4 and 7.7%), SurinameseHindustani (12.3 and 6.2%), and Turkish (23.8 and 12.0%) children. In the models adjusted for age and sex only, Moroccan, Surinamese-Hindustani, and Turkish children had a higher total fat mass than Dutch children, whereas Surinam-
© 2014 S. Karger AG, Basel 0250–6807/14/0644–0208$39.50/0 E-Mail [email protected] www.karger.com/anm
ese-Creole children had a lower total fat mass. Compared to Dutch children, the android/gynoid fat mass ratio was higher in Moroccan, Surinamese-Hindustani, and Turkish children, whereas the preperitoneal fat mass was higher among Dutch Antillean, Moroccan, Surinamese-Hindustani, and Turkish children (all p < 0.05). Prepregnancy factors explained up to 73% of these differences. In addition to prepregnancy factors, pregnancy factors explained up to 34% of the differences. Childhood factors did not significantly explain these associations. Conclusions: All ethnic minority groups had higher risks of overweight and obesity than Dutch children. Moroccan, Surinamese-Hindustani, and Turkish children also had an adverse body fat profile. Prepregnancy and pregnancy might be critical periods for preventive strategies focused on the reduction of ethnic disparities in childhood adiposity. © 2014 S. Karger AG, Basel
Introduction
The childhood obesity prevalence is increasing worldwide and it is strongly associated with risk factors for cardiometabolic diseases and mortality in adulthood [1]. Major differences in obesity prevalence between ethnic Vincent W.V. Jaddoe, MD, PhD Generation R Study Group (Na 2915) Erasmus University Medical Center PO Box 2040, NL–3000 CA Rotterdam (The Netherlands) E-Mail v.jaddoe @ erasmusmc.nl
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
Key Words Ethnicity · Obesity · Abdominal fat · Total fat · Pediatrics
Methods Design and Population This study was embedded in the Generation R Study, a multiethnic population-based prospective cohort study from fetal life onwards in the city of Rotterdam, the Netherlands [10]. This study was approved by Medical Ethical Committee of the Erasmus University Medical Center. Written consent was obtained from the participating parents [10]. All children were born between 2002 and 2006. The response rate at birth was 61%. In total, 8,305 children participated in measurements until the age of 6 years. Information on ethnicity was available for 7,770 singleton children. For the current analysis, we excluded 1,342 children because of the small country-specific sample sizes (6 months, % Sleep per night, h Outside play for ≥1 h/day, % TV watching for ≥2 h/day, % Childhood body fat outcomes Age, years Height, cm Weight, kg BMI Total body fat mass Android/gynoid fat mass ratio Preperitoneal area, cm2 Overweight, % Obesity, %
Dutch (n = 3,584)
Table 1. Characteristics of the study population (n = 5,244)
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
Statistical Analysis First, we compared maternal and childhood characteristics between different ethnic groups using one-way ANOVA and χ2 tests. We examined the relation of the childhood BMI with general and abdominal fat distributions among each ethnic group using linear regression models. Second, we used logistic regression models to examine the ethnic differences in childhood overweight and obesity. These models were first adjusted for the child’s age at the visit and sex (crude models), and additionally for parental prepregnancy factors (maternal age, education, marital status, parity, maternal and paternal prepregnancy BMI, and household income). These models were subsequently adjusted for: (1) pregnancy factors (maternal smoking, alcohol consumption, folic acid supplement use, total calorie intake during pregnancy, gestational weight gain, maternal complications during pregnancy, gestational age at birth, and birth weight) and (2) childhood factors (breast-feeding, age at the introduction of solid foods, and hours of sleep per night, outside play, and TV watching). Pregnancy and childhood factors were added to the prepregnancy models since the main interest was in modifiable factors after the enrolment of pregnant women in health care practices, in addition to factors that could not be modified. These factors were included in the models based on their associations with childhood adiposity outcomes in previous studies [19], a significant association with the outcomes or changes in effect estimates >10%. The associations of each factor with body fat outcomes are presented in online supplementary table 2. Third, we used multiple linear regression models with similar adjustments to assess ethnic differences in childhood body fat outcomes (BMI, total fat mass, android/gynoid fat mass ratio, and preperitoneal fat mass). We log-transformed nonnormally distributed preperitoneal fat mass values. We additionally adjusted all models focused on fat distribution outcomes for each child’s height at measurement [20]. We calculated the exact percentages of change in the effect estimates after adjusting for different factors for childhood body fat outcomes using the following formula: % change = (β2 – β1)/β1 × 100). We tested whether these changes were statistically significant by assessing the heterogeneity between effect estimates. We constructed SD scores [(observed value – mean)/SD] for all variables to enable comparison of the effect sizes of different outcome measures. We did not create ageadjusted SD scores since the outcomes were measured in small age ranges. We tested potential interactions between ethnic groups with the children’s sex and birth weight in relation to fat outcomes. In order to reduce the potential bias associated with missing data, we performed multiple imputations of missing covariates by generating 5 independent data sets using the Markov chain Monte Carlo method, after which the pooled effect estimates were calculated [21]. Imputations were based on the relationships between covariates, determinants, and outcomes. We did not impute missing determinants or outcomes. All analyses were performed using the Statistical Package of Social Sciences version 20.0 for Windows (SPSS Inc., Chicago, Ill., USA).
Ethnicity and Adiposity in Schoolchildren
Results
Subject Characteristics Table 1 shows maternal, paternal, and child characteristics for different ethnic groups. The overweight and obesity prevalences among Dutch children were 10.0 and 2.1%, respectively. Higher prevalences were observed among Cape Verdean (21.0 and 10.3%), Dutch Antillean (18.4 and 13.8%), Moroccan (20.6 and 7.7%), Surinamese-Creole (13.4 and 7.7%), Surinamese-Hindustani (12.3 and 6.2%), and Turkish (23.8 and 12.0%) children. Figure 1 shows the relation between the BMI and body fat outcomes for each ethnic group. Significant interaction terms between ethnicity and BMI were present for associations with the android/gynoid fat mass ratio and preperitoneal fat mass but not for total body fat mass. At any given level of BMI, Moroccan, Surinamese-Hindustani, and Turkish children tended to have a higher total fat mass, whereas Dutch Antillean and Surinamese-Creole children tended to have lower total fat mass, compared to Dutch children (fig. 1a). Compared to Dutch children, Surinamese-Hindustani children had a higher android/gynoid fat mass ratio at any given level of BMI, whereas Turkish children had a higher android/gynoid fat mass at higher levels of BMI. Compared to Dutch children, Cape Verdean and Dutch Antillean children had a lower android/gynoid fat mass ratio for their BMI (fig. 1b). Figure 1c shows that Surinamese-Hindustani children had a higher abdominal preperitoneal fat mass for their BMI, whereas, except for Cape Verdean children, all other ethnic groups had a higher abdominal preperitoneal fat mass only at higher levels of BMI as compared to Dutch children. The specific corresponding effect estimates are given in online supplementary table 3. Risks of Overweight and Obesity in Childhood Figure 2 shows that that in models adjusted for the child’s age and sex only, children from all ethnic minority groups had increased risks of childhood overweight and obesity as compared to Dutch children. The highest risk was present in Turkish children (OR 3.92; 95% CI 3.17–4.84). The associations were largely explained by parental prepregnancy factors and became nonsignificant for Surinamese-Creole and Surinamese-Hindustani children. Additional adjustment for pregnancy and childhood factors did not change the effect estimates further to a large extent.
Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
211
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
betes), and gestational age and weight at birth obtained from medical records. Childhood factors included breast-feeding and the age of introduction of solid foods during infancy, hours of sleep at night at age 2 years, and average hours of outside play and TV watching at age 6 years. This information was assessed via questionnaires [18].
0.40
35
0.35 Android/gynoid fat mass ratio
40
25
20
a
0.25
0.20
0.15
15
10
0.30
13
14
15
16
17
18
19
20
BMI
0.10
b
13
Preperitoneal fat mass
0.8
212
Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
16
17
18
19
20
Dutch Moroccan Turkish Surinamese-Hindustani Cape Verdean Dutch Antillean Surinamese-Creole
0.9
Body Fat Outcomes in Childhood Table 2 shows that in models adjusted for the child’s age and sex only, Cape Verdean, Dutch Antillean, Moroccan, Surinamese-Creole, and Turkish children had a higher BMI compared to Dutch children, with the strongest association present in Turkish children (0.62 SD
15
BMI
1.0
Fig. 1. BMI and specific fat mass measures in different ethnic groups. a Total body fat mass. b Android/gynoid fat mass ratio. c Preperitoneal fat mass. The lines represent the associations of childhood BMI with fat distribution measures among childhood ethnicity groups, obtained from linear regression analyses. The effect estimates for the associations of childhood BMI with fat distribution measures among ethnic groups are given in online supplementary table 3.
14
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
c
13
14
15
16
17
18
19
20
BMI
scores; 95% CI 0.54–0.71). Moroccan, Surinamese-Hindustani, and Turkish children had a higher, and Surinamese-Creole children had a lower, total fat mass than Dutch children (all p < 0.05). Compared to Dutch children, Moroccan, Surinamese-Hindustani, and Turkish children, but not Cape Verdean, Dutch Antillean, or SurinameseGishti /Kruithof /Felix /Raat /Hofman / Duijts /Gaillard /Jaddoe
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
Total fat mass (%)
30
6 5
Crude models Parental prepregnancy factors Pregnancy factors Childhood factors
4 3 2 1 0
Dutch
Creole children, had a higher android/gynoid fat mass ratio (all p < 0.05). Similarly, Dutch Antillean, Moroccan, Surinamese-Hindustani, and Turkish children had a higher abdominal preperitoneal fat mass (all p < 0.05). Parental prepregnancy factors explained up to 73% of the ethnic differences in BMI and fully explained the differences in total body fat mass and android/gynoid fat mass ratio for Dutch Antillean and Moroccan children, respectively. However, significant changes in the effect estimates were present only for Dutch Antillean, Moroccan, and Turkish children. In addition to parental prepregnancy factors, pregnancy factors explained up to 34% of the ethnic differences in childhood general and abdominal fat, but the observed changes in the effect estimates were not significant. Childhood factors did not explain any of these associations. Estimates for the ethnic differences in childhood adiposity outcomes in the different models are given in online supplementary table 4. No significant interactions were present between ethnicity and the child’s sex and birth weight in relation to BMI and total fat mass after taking multiple testing into account.
Cape Verdean
Dutch Antillean
Moroccan Surinamese- SurinameseCreole Hindustani
Turkish
Ethnicity
and obesity than Dutch children. Compared to Dutch children, Moroccan, Surinamese-Hindustani, and Turkish children had an adverse body fat profile, whereas Cape Verdean, Dutch Antillean, and Surinamese-Creole children had a beneficial body fat profile for their BMI. Parental prepregnancy and pregnancy factors explained a larger extent of the observed differences than did factors in childhood.
Interpretation of the Main Findings
The results of this multiethnic population-based prospective cohort study in the Netherlands showed that all ethnic minority groups had higher risks of overweight
Obesity prevalence varies between ethnic groups [2, 22]. Higher prevalences of overweight and obesity have been reported among black and Hispanic children as compared to white children in the USA [3]. Less is known about the ethnic differences in specific European countries. Large studies in the UK have suggested that South Asian children have a lower BMI compared to white European children [23, 24]. In the Netherlands, the largest ethnic minority groups are Cape Verdean, Dutch Antillean, Moroccan, Turkish, Surinamese-Creole, and Surinamese-Hindustani [5, 6]. Previously, it has been shown that Turkish and Moroccan children have higher prevalence rates of overweight and obesity [7]. Also, a study among 7,801 children aged 5 years reported that not Surinamese children but children from other ethnic minority groups have a higher risk of being overweight than
Ethnicity and Adiposity in Schoolchildren
Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
Discussion
213
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
OR (95% CI) for overweight and obesity
Fig. 2. Ethnic differences in the risk of overweight and obesity in children (n = 5,244). Values are OR (95% CI) that reflect the risk of combined overweight and obesity for each childhood ethnicity group. Crude models are adjusted for the child’s age at the visit, sex, and height (fat mass outcomes only). Parental prepregnancy factor models were additionally adjusted for maternal age, education, marital status, parity, maternal and paternal prepregnancy BMI, and household income. These models were subsequently additionally adjusted for pregnancy factors (maternal smoking, alcohol consumption, folic acid supplement use, total calorie intake and gestational weight gain during pregnancy, maternal complications during pregnancy, gestational age, and weight at birth) and for childhood factors (breast-feeding, age at the introduction of solid foods, and hours of sleep per night, outside play, and TV watching).
Table 2. Associations of ethnicity with childhood adiposity outcomes after adjustment for prepregnancy, pregnancy, and childhood fac-
tors (n = 5,244) Ethnic groups Dutch BMI (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change Total body fat mass (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change Android/gynoid fat mass ratio (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change Preperitoneal fat mass area (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change
3,584 reference reference reference reference 3,498 reference reference reference reference 3,498 reference reference reference reference 2,901 reference reference reference reference
Cape Verdean
Dutch Antillean
Moroccan
SurinameseCreole
SurinameseHindustani
Turkish
195 0.46 0.32 to 0.60* –43.8 6.3 –6.3
196 0.52 0.39 to 0.66* –48.1e 0 –3.7
378 0.39 0.29 to 0.50* –64.1e 0 –7.1
194 0.25 0.11 to 0.39* –56.0 NA NA
196 –0.15 –0.29 to –0.01* 73.3 34.6 3.8
501 0.62 0.54 to 0.71* –45.2e –2.9 –2.9
194 0.46 0.33 to 0.58* –53.3 –16.7 –3.3
192 0.18 0.05 to 0.30* –144.4e NA NA
368 0.55 0.46 to 0.64* –50.9e –14.8 0
189 –0.09 –0.21 to 0.04 NA NA NA
192 0.64 0.51 to 0.76* –23.4 –10.2 –2.0
482 0.77 0.69 to 0.85* –37.7e –15.8 –2.1
194 0.00 –0.14 to 0.14 NA NA NA
192 0.12 –0.02 to 0.26 NA NA NA
368 0.14 0.03 to 0.38* –150.0e NA NA
189 0.02 –0.12 to 0.15 NA NA NA
192 0.28 0.15 to 0.42* –35.7 –22.2 –5.6
482 0.67 0.58 to 0.76* –31.3e –13.0 –2.2
157 0.14 –0.00 to 0.29 NA NA NA
157 0.17 0.02 to 0.31* –102.4 NA NA
325 0.46 0.35 to 0.56* –41.3e –11.1 –3.7
151 0.03 –0.12 to 0.18 NA NA NA
145 0.31 0.16 to 0.46* –32.3 –14.3 –4.76
393 0.72 0.63 to 0.82* –29.2e –9.8 0
NA = Not available. * p < 0.05. a Values are linear regression coefficients (95% CI) for the crude models, adjusted for the child’s age at the visit, sex, and height (fat mass outcomes only). b If these associations were significant, percentages of change in the effect estimates after additional adjustment for parental prepregnancy factors (maternal age, education, marital status, parity, maternal and paternal prepregnancy BMI, and household income) were calculated. If significant, these models were subsequently additionally adjusted for: c pregnancy factors (ma-
ternal smoking, alcohol consumption, folic acid supplement use, total calorie intake and gestational weight gain during pregnancy, maternal complications during pregnancy, gestational age, and weight at birth), and d childhood factors (breast-feeding, age at introduction of solid foods, hours of sleep per night, playing outside, and TV watching) and the percentages of change in the effect estimates were calculated. The change in the effect estimates was calculated as (β2 – β1)/β1 × 100). e p value for the percentages of change in the estimates
Published online: October 2, 2014
Ethnic Disparities in General and Abdominal Adiposity at School Age: A Multiethnic Population-Based Cohort Study in The Netherlands Olta Gishti a–c Claudia J. Kruithof a, b Janine F. Felix b, c Hein Raat d Albert Hofman b Liesbeth Duijts b, c Romy Gaillard a–c Vincent W.V. Jaddoe a–c a Generation R Study Group, and Departments of b Epidemiology, c Pediatrics and d Public Health, Erasmus University Medical Center, Rotterdam, The Netherlands
Abstract Background/Aims: Ethnic differences in obesity prevalence have been reported. We examined ethnic differences in general and abdominal fat distribution in school-age children and the influence of parental prepregnancy, pregnancy, and childhood factors on these differences. Methods: We performed a multiethnic population-based prospective cohort study among 5,244 children with information about prepregnancy factors, pregnancy, and childhood factors. At the age of 6 years, the BMI, total fat mass, android/gynoid fat mass ratio, and preperitoneal fat mass were assessed via dual-energy X-ray absorptiometry and abdominal ultrasound. Results: The overweight and obesity prevalences among Dutch children were 10.0 and 2.1%, respectively. Higher prevalences were observed among Cape Verdean (21.0 and 10.3%), Dutch Antillean (18.4 and 13.8%), Moroccan (20.6 and 7.7%), Surinamese-Creole (13.4 and 7.7%), SurinameseHindustani (12.3 and 6.2%), and Turkish (23.8 and 12.0%) children. In the models adjusted for age and sex only, Moroccan, Surinamese-Hindustani, and Turkish children had a higher total fat mass than Dutch children, whereas Surinam-
© 2014 S. Karger AG, Basel 0250–6807/14/0644–0208$39.50/0 E-Mail [email protected] www.karger.com/anm
ese-Creole children had a lower total fat mass. Compared to Dutch children, the android/gynoid fat mass ratio was higher in Moroccan, Surinamese-Hindustani, and Turkish children, whereas the preperitoneal fat mass was higher among Dutch Antillean, Moroccan, Surinamese-Hindustani, and Turkish children (all p < 0.05). Prepregnancy factors explained up to 73% of these differences. In addition to prepregnancy factors, pregnancy factors explained up to 34% of the differences. Childhood factors did not significantly explain these associations. Conclusions: All ethnic minority groups had higher risks of overweight and obesity than Dutch children. Moroccan, Surinamese-Hindustani, and Turkish children also had an adverse body fat profile. Prepregnancy and pregnancy might be critical periods for preventive strategies focused on the reduction of ethnic disparities in childhood adiposity. © 2014 S. Karger AG, Basel
Introduction
The childhood obesity prevalence is increasing worldwide and it is strongly associated with risk factors for cardiometabolic diseases and mortality in adulthood [1]. Major differences in obesity prevalence between ethnic Vincent W.V. Jaddoe, MD, PhD Generation R Study Group (Na 2915) Erasmus University Medical Center PO Box 2040, NL–3000 CA Rotterdam (The Netherlands) E-Mail v.jaddoe @ erasmusmc.nl
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
Key Words Ethnicity · Obesity · Abdominal fat · Total fat · Pediatrics
Methods Design and Population This study was embedded in the Generation R Study, a multiethnic population-based prospective cohort study from fetal life onwards in the city of Rotterdam, the Netherlands [10]. This study was approved by Medical Ethical Committee of the Erasmus University Medical Center. Written consent was obtained from the participating parents [10]. All children were born between 2002 and 2006. The response rate at birth was 61%. In total, 8,305 children participated in measurements until the age of 6 years. Information on ethnicity was available for 7,770 singleton children. For the current analysis, we excluded 1,342 children because of the small country-specific sample sizes (6 months, % Sleep per night, h Outside play for ≥1 h/day, % TV watching for ≥2 h/day, % Childhood body fat outcomes Age, years Height, cm Weight, kg BMI Total body fat mass Android/gynoid fat mass ratio Preperitoneal area, cm2 Overweight, % Obesity, %
Dutch (n = 3,584)
Table 1. Characteristics of the study population (n = 5,244)
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
Statistical Analysis First, we compared maternal and childhood characteristics between different ethnic groups using one-way ANOVA and χ2 tests. We examined the relation of the childhood BMI with general and abdominal fat distributions among each ethnic group using linear regression models. Second, we used logistic regression models to examine the ethnic differences in childhood overweight and obesity. These models were first adjusted for the child’s age at the visit and sex (crude models), and additionally for parental prepregnancy factors (maternal age, education, marital status, parity, maternal and paternal prepregnancy BMI, and household income). These models were subsequently adjusted for: (1) pregnancy factors (maternal smoking, alcohol consumption, folic acid supplement use, total calorie intake during pregnancy, gestational weight gain, maternal complications during pregnancy, gestational age at birth, and birth weight) and (2) childhood factors (breast-feeding, age at the introduction of solid foods, and hours of sleep per night, outside play, and TV watching). Pregnancy and childhood factors were added to the prepregnancy models since the main interest was in modifiable factors after the enrolment of pregnant women in health care practices, in addition to factors that could not be modified. These factors were included in the models based on their associations with childhood adiposity outcomes in previous studies [19], a significant association with the outcomes or changes in effect estimates >10%. The associations of each factor with body fat outcomes are presented in online supplementary table 2. Third, we used multiple linear regression models with similar adjustments to assess ethnic differences in childhood body fat outcomes (BMI, total fat mass, android/gynoid fat mass ratio, and preperitoneal fat mass). We log-transformed nonnormally distributed preperitoneal fat mass values. We additionally adjusted all models focused on fat distribution outcomes for each child’s height at measurement [20]. We calculated the exact percentages of change in the effect estimates after adjusting for different factors for childhood body fat outcomes using the following formula: % change = (β2 – β1)/β1 × 100). We tested whether these changes were statistically significant by assessing the heterogeneity between effect estimates. We constructed SD scores [(observed value – mean)/SD] for all variables to enable comparison of the effect sizes of different outcome measures. We did not create ageadjusted SD scores since the outcomes were measured in small age ranges. We tested potential interactions between ethnic groups with the children’s sex and birth weight in relation to fat outcomes. In order to reduce the potential bias associated with missing data, we performed multiple imputations of missing covariates by generating 5 independent data sets using the Markov chain Monte Carlo method, after which the pooled effect estimates were calculated [21]. Imputations were based on the relationships between covariates, determinants, and outcomes. We did not impute missing determinants or outcomes. All analyses were performed using the Statistical Package of Social Sciences version 20.0 for Windows (SPSS Inc., Chicago, Ill., USA).
Ethnicity and Adiposity in Schoolchildren
Results
Subject Characteristics Table 1 shows maternal, paternal, and child characteristics for different ethnic groups. The overweight and obesity prevalences among Dutch children were 10.0 and 2.1%, respectively. Higher prevalences were observed among Cape Verdean (21.0 and 10.3%), Dutch Antillean (18.4 and 13.8%), Moroccan (20.6 and 7.7%), Surinamese-Creole (13.4 and 7.7%), Surinamese-Hindustani (12.3 and 6.2%), and Turkish (23.8 and 12.0%) children. Figure 1 shows the relation between the BMI and body fat outcomes for each ethnic group. Significant interaction terms between ethnicity and BMI were present for associations with the android/gynoid fat mass ratio and preperitoneal fat mass but not for total body fat mass. At any given level of BMI, Moroccan, Surinamese-Hindustani, and Turkish children tended to have a higher total fat mass, whereas Dutch Antillean and Surinamese-Creole children tended to have lower total fat mass, compared to Dutch children (fig. 1a). Compared to Dutch children, Surinamese-Hindustani children had a higher android/gynoid fat mass ratio at any given level of BMI, whereas Turkish children had a higher android/gynoid fat mass at higher levels of BMI. Compared to Dutch children, Cape Verdean and Dutch Antillean children had a lower android/gynoid fat mass ratio for their BMI (fig. 1b). Figure 1c shows that Surinamese-Hindustani children had a higher abdominal preperitoneal fat mass for their BMI, whereas, except for Cape Verdean children, all other ethnic groups had a higher abdominal preperitoneal fat mass only at higher levels of BMI as compared to Dutch children. The specific corresponding effect estimates are given in online supplementary table 3. Risks of Overweight and Obesity in Childhood Figure 2 shows that that in models adjusted for the child’s age and sex only, children from all ethnic minority groups had increased risks of childhood overweight and obesity as compared to Dutch children. The highest risk was present in Turkish children (OR 3.92; 95% CI 3.17–4.84). The associations were largely explained by parental prepregnancy factors and became nonsignificant for Surinamese-Creole and Surinamese-Hindustani children. Additional adjustment for pregnancy and childhood factors did not change the effect estimates further to a large extent.
Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
211
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
betes), and gestational age and weight at birth obtained from medical records. Childhood factors included breast-feeding and the age of introduction of solid foods during infancy, hours of sleep at night at age 2 years, and average hours of outside play and TV watching at age 6 years. This information was assessed via questionnaires [18].
0.40
35
0.35 Android/gynoid fat mass ratio
40
25
20
a
0.25
0.20
0.15
15
10
0.30
13
14
15
16
17
18
19
20
BMI
0.10
b
13
Preperitoneal fat mass
0.8
212
Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
16
17
18
19
20
Dutch Moroccan Turkish Surinamese-Hindustani Cape Verdean Dutch Antillean Surinamese-Creole
0.9
Body Fat Outcomes in Childhood Table 2 shows that in models adjusted for the child’s age and sex only, Cape Verdean, Dutch Antillean, Moroccan, Surinamese-Creole, and Turkish children had a higher BMI compared to Dutch children, with the strongest association present in Turkish children (0.62 SD
15
BMI
1.0
Fig. 1. BMI and specific fat mass measures in different ethnic groups. a Total body fat mass. b Android/gynoid fat mass ratio. c Preperitoneal fat mass. The lines represent the associations of childhood BMI with fat distribution measures among childhood ethnicity groups, obtained from linear regression analyses. The effect estimates for the associations of childhood BMI with fat distribution measures among ethnic groups are given in online supplementary table 3.
14
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
c
13
14
15
16
17
18
19
20
BMI
scores; 95% CI 0.54–0.71). Moroccan, Surinamese-Hindustani, and Turkish children had a higher, and Surinamese-Creole children had a lower, total fat mass than Dutch children (all p < 0.05). Compared to Dutch children, Moroccan, Surinamese-Hindustani, and Turkish children, but not Cape Verdean, Dutch Antillean, or SurinameseGishti /Kruithof /Felix /Raat /Hofman / Duijts /Gaillard /Jaddoe
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
Total fat mass (%)
30
6 5
Crude models Parental prepregnancy factors Pregnancy factors Childhood factors
4 3 2 1 0
Dutch
Creole children, had a higher android/gynoid fat mass ratio (all p < 0.05). Similarly, Dutch Antillean, Moroccan, Surinamese-Hindustani, and Turkish children had a higher abdominal preperitoneal fat mass (all p < 0.05). Parental prepregnancy factors explained up to 73% of the ethnic differences in BMI and fully explained the differences in total body fat mass and android/gynoid fat mass ratio for Dutch Antillean and Moroccan children, respectively. However, significant changes in the effect estimates were present only for Dutch Antillean, Moroccan, and Turkish children. In addition to parental prepregnancy factors, pregnancy factors explained up to 34% of the ethnic differences in childhood general and abdominal fat, but the observed changes in the effect estimates were not significant. Childhood factors did not explain any of these associations. Estimates for the ethnic differences in childhood adiposity outcomes in the different models are given in online supplementary table 4. No significant interactions were present between ethnicity and the child’s sex and birth weight in relation to BMI and total fat mass after taking multiple testing into account.
Cape Verdean
Dutch Antillean
Moroccan Surinamese- SurinameseCreole Hindustani
Turkish
Ethnicity
and obesity than Dutch children. Compared to Dutch children, Moroccan, Surinamese-Hindustani, and Turkish children had an adverse body fat profile, whereas Cape Verdean, Dutch Antillean, and Surinamese-Creole children had a beneficial body fat profile for their BMI. Parental prepregnancy and pregnancy factors explained a larger extent of the observed differences than did factors in childhood.
Interpretation of the Main Findings
The results of this multiethnic population-based prospective cohort study in the Netherlands showed that all ethnic minority groups had higher risks of overweight
Obesity prevalence varies between ethnic groups [2, 22]. Higher prevalences of overweight and obesity have been reported among black and Hispanic children as compared to white children in the USA [3]. Less is known about the ethnic differences in specific European countries. Large studies in the UK have suggested that South Asian children have a lower BMI compared to white European children [23, 24]. In the Netherlands, the largest ethnic minority groups are Cape Verdean, Dutch Antillean, Moroccan, Turkish, Surinamese-Creole, and Surinamese-Hindustani [5, 6]. Previously, it has been shown that Turkish and Moroccan children have higher prevalence rates of overweight and obesity [7]. Also, a study among 7,801 children aged 5 years reported that not Surinamese children but children from other ethnic minority groups have a higher risk of being overweight than
Ethnicity and Adiposity in Schoolchildren
Ann Nutr Metab 2014;64:208–217 DOI: 10.1159/000365022
Discussion
213
Downloaded by: NYU Medical Center Library 128.122.253.228 - 5/20/2015 10:07:48 AM
OR (95% CI) for overweight and obesity
Fig. 2. Ethnic differences in the risk of overweight and obesity in children (n = 5,244). Values are OR (95% CI) that reflect the risk of combined overweight and obesity for each childhood ethnicity group. Crude models are adjusted for the child’s age at the visit, sex, and height (fat mass outcomes only). Parental prepregnancy factor models were additionally adjusted for maternal age, education, marital status, parity, maternal and paternal prepregnancy BMI, and household income. These models were subsequently additionally adjusted for pregnancy factors (maternal smoking, alcohol consumption, folic acid supplement use, total calorie intake and gestational weight gain during pregnancy, maternal complications during pregnancy, gestational age, and weight at birth) and for childhood factors (breast-feeding, age at the introduction of solid foods, and hours of sleep per night, outside play, and TV watching).
Table 2. Associations of ethnicity with childhood adiposity outcomes after adjustment for prepregnancy, pregnancy, and childhood fac-
tors (n = 5,244) Ethnic groups Dutch BMI (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change Total body fat mass (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change Android/gynoid fat mass ratio (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change Preperitoneal fat mass area (SDS) n Crude modela β 95% CI Parental prepregnancy factorsb, % change + Pregnancy factorsc, % change + Childhood factorsd, % change
3,584 reference reference reference reference 3,498 reference reference reference reference 3,498 reference reference reference reference 2,901 reference reference reference reference
Cape Verdean
Dutch Antillean
Moroccan
SurinameseCreole
SurinameseHindustani
Turkish
195 0.46 0.32 to 0.60* –43.8 6.3 –6.3
196 0.52 0.39 to 0.66* –48.1e 0 –3.7
378 0.39 0.29 to 0.50* –64.1e 0 –7.1
194 0.25 0.11 to 0.39* –56.0 NA NA
196 –0.15 –0.29 to –0.01* 73.3 34.6 3.8
501 0.62 0.54 to 0.71* –45.2e –2.9 –2.9
194 0.46 0.33 to 0.58* –53.3 –16.7 –3.3
192 0.18 0.05 to 0.30* –144.4e NA NA
368 0.55 0.46 to 0.64* –50.9e –14.8 0
189 –0.09 –0.21 to 0.04 NA NA NA
192 0.64 0.51 to 0.76* –23.4 –10.2 –2.0
482 0.77 0.69 to 0.85* –37.7e –15.8 –2.1
194 0.00 –0.14 to 0.14 NA NA NA
192 0.12 –0.02 to 0.26 NA NA NA
368 0.14 0.03 to 0.38* –150.0e NA NA
189 0.02 –0.12 to 0.15 NA NA NA
192 0.28 0.15 to 0.42* –35.7 –22.2 –5.6
482 0.67 0.58 to 0.76* –31.3e –13.0 –2.2
157 0.14 –0.00 to 0.29 NA NA NA
157 0.17 0.02 to 0.31* –102.4 NA NA
325 0.46 0.35 to 0.56* –41.3e –11.1 –3.7
151 0.03 –0.12 to 0.18 NA NA NA
145 0.31 0.16 to 0.46* –32.3 –14.3 –4.76
393 0.72 0.63 to 0.82* –29.2e –9.8 0
NA = Not available. * p < 0.05. a Values are linear regression coefficients (95% CI) for the crude models, adjusted for the child’s age at the visit, sex, and height (fat mass outcomes only). b If these associations were significant, percentages of change in the effect estimates after additional adjustment for parental prepregnancy factors (maternal age, education, marital status, parity, maternal and paternal prepregnancy BMI, and household income) were calculated. If significant, these models were subsequently additionally adjusted for: c pregnancy factors (ma-
ternal smoking, alcohol consumption, folic acid supplement use, total calorie intake and gestational weight gain during pregnancy, maternal complications during pregnancy, gestational age, and weight at birth), and d childhood factors (breast-feeding, age at introduction of solid foods, hours of sleep per night, playing outside, and TV watching) and the percentages of change in the effect estimates were calculated. The change in the effect estimates was calculated as (β2 – β1)/β1 × 100). e p value for the percentages of change in the estimates
