Postnatal growth velocity and overweight in early adolescents: A comparison of rural and urban African boys and girls

Original Research Article Postnatal Growth Velocity and Overweight in Early Adolescents: A Comparison of Rural and Urban African Boys and Girls E.D. CHIRWA, 1 * P. GRIFFITHS, 1,2 K. MALETA, 3 P. ASHORN, 4 J.M. PETTIFOR, 1 AND S.A. NORRIS 1 1 Wits/MRC Developmental Pathways for Health Research Unit, Department of Pediatrics, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa 2 Centre for Global Health and Human Development, Loughborough University, Loughborough, Leicestershire, LE 11 3TU, United Kingdom 3 Department of Community Health, College of Medicine, University of Malawi, Malawi 4 Department of International Health, School of Medicine, University of Tampere, Finland and Department of Pediatrics, Tampere University Hospital, Finland Objectives: To compare growth velocity of two African child cohorts and examine the relationship between postnatal growth velocity in infancy/early childhood and the risk of overweight/stunting in early adolescence. Methods: The study used data from two child cohorts from urban (Birth to Twenty Cohort, South Africa) and rural (Lungwena Child Survival Study, Malawi) African settings. Mixed effect modelling was used to derive growth and peak growth velocities. T-tests were used to compare growth parameters and velocities between the two cohorts. Linear and logistic regression models were used to determine the relationship between growth velocity and early adolescent (ages 9–11 years) body mass index and odds of being overweight. Results: Children in the BH cohort were significantly taller and heavier than those in the Lungwena cohort, and exhibited faster weight and height growth velocity especially in the first year of life (P<0.05). No significant association was shown between baseline weight (a w ) and overweight in early adolescence (OR 51.25, CI 50.67, 2.34). The weight growth velocity parameter b w was highly associated with odds of being overweight. Association between overweight in adolescence and weight velocity was stronger in infancy than in early childhood (OR at 3 months 54.80, CI 52.49, 9.26; OR at 5 years 52.39, CI 51.65, 3.47). Conclusion: High weight and height growth velocity in infancy, independent of size at birth, is highly associated with overweight in early adolescence. However, the long term effects of rapid growth in infancy may be dependent on a particular population’s socio-economic status and level of urbanization. Am. J. Hum. Biol. 26:643–651, 2014. V C2014 The Authors American Journal of Human Biology Published by Wiley Periodicals, Inc. Several studies have shown the association between early childhood growth and later health outcomes such as diabetes, cardiovascular diseases, and obesity (Cameron and Demerath, 2002; Cameron et al., 2003; Adair, 2007; Adair et al., 2009). In particular, studies have examined the critical periods in infancy and early childhood that are associated with these health outcomes (Black and Krishnakumar, 1999; McCarthy et al., 2007; Botton et al., 2008; Ridgway et al., 2009). Both growth retardation and rapid growth in the different stages of early life are predictive of the later health outcomes (Cameron and Demerath, 2002; Li, 2003; Cameron et al., 2005; Stein et al., 2010). For example, Flexeder et al. (2012) found that rapid weight gain in infancy is associated with physician-diagnosed asthma in school-aged children (Flexeder et al., 2012). A number of studies have examined the relationship between early growth and later health outcomes in high income as well as low and middle income countries (LMIC; Martorell, 1995; Ong et al., 2000; Li, 2003; Salonen et al., 2009; Mesa et al., 2010; Stein et al., 2010). However, there have also been inconsistent findings regarding the relationship between growth retardation, or stunting and overweight and obesity in later life. Whereas, it has been suggested that childhood under-nutrition predisposes a child to weight gain in later life (Hoffman et al., 2000), other studies have found that childhood stunting was associated with lower body mass index (BMI; Schroeder et al., 1999; Walker et al., 2007). Several studies have also shown the short term and long term benefits of rapid growth for children in resource-poor settings (Victora et al., 2001; Kalanda et al., 2005; Hoddinott et al., 2008; Victora et al., 2008). The short term benefits include, reduced morbidity and mortality, and improved cognitive development, while long term benefits include improved human capital and improved reproductive outcomes in women. Although a wide body of evidence supports the long term benefits and detrimental effects of early rapid growth in LMIC, few studies have looked at this relationship in a sub-Saharan African context, due to the limited number of birth cohort studies. Thus, this study compared the growth velocities of two cohorts from rural and urban African settings, and examined the relationship between size at birth (birth weight), growth velocity in infancy and This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Contract grant sponsors: Consortium for Advanced Research Training in Africa (CARTA) which is—jointly led by the African Population and Health Research Center and the University of the Witwatersrand and funded by the Welcome Trust (UK); Contract grant number: 087547/Z/08/ Z; Department for International Development (DfID) under the Development Partnerships in Higher Education (DelPHE); the Carnergie Corporation of New York; Contract grant number: B 8606; Ford Foundation; Contract grant number: 1100-0399; Google.Org; Contract grant number: 191994; SIDA; Contract grant number: 54100029; MacArthur Foundation; Contract grant number: 10-95915-000-INP. *Correspondence to: Esnat D. Chirwa, MRC/Wits Developmental Pathways for Health Research Unit, Faculty of Health Sciences, University of the Witwatersrand, 7 York Road, Parktown 2193, Johannesburg, South Africa. E-mail: Esnat.Chi[email protected] Received 31 October 2013; Revision received 30 April 2014; Accepted 28 May 2014 DOI: 10.1002/ajhb.22575 Published online 20 June 2014 in Wiley Online Library (wileyonlinelibrary.com). V C2014 The Authors American Journal of Human Biology Published by Wiley Periodicals, Inc. AMERICAN JOURNAL OF HUMAN BIOLOGY 26:643–651 (2014) early childhood, and early adolescent obesity. The two cohorts are likely to be at different stages of nutritional transition, considering the rural cohort is from a very low income country while the urban cohort is from a middle income country. Urbanization is generally linked to changes in lifestyle factors that affect obesity risk, such as dietary patterns and sedentary behaviors. Thus, apart from the nutritional differences, there may also be social, cultural, economic and environmental differences between the two cohorts that may affect growth and development of the children and also affect their risk of obesity. Mixed-effects modelling and childhood structural growth model were used to examine the relationship between postnatal growth velocity and obesity or stunting in early adolescence (ages 9–11 years). Mixed-effects modelling is flexible in dealing with unbalanced longitudinal measurements, and takes into account correlations between repeated measurements. The use of the structural growth curve allowed for the estimation of the growth rates at any given age. SUBJECTS AND METHODS Weight and height measurements from 2 African longitudinal cohorts were used. The Bone-Health (BH) Study is a sub-sample of the Birth-to-Twenty (BH) birth cohort in Johannesburg, South Africa, which includes 453 black participants. The cohort has anthropometric measurements at birth, 3 months, 6 months, 1 year, 2 years, 4 years, 5 years, 7/8 years, 9 years, and 10 years. Birth weights were extracted from birth records, while subsequent weight/height measurements were obtained using standard anthropometric techniques (Cameron, 1984). More specific details about this urban cohort are reported elsewhere (Cameron et al., 2003; Cameron et al., 2005). The rural component of the study used the Lungwena Child Survival Study (LCSS), which is a cohort of about 729 children living in Mangochi, a rural district in southern Malawi. The on-going study has growth data of children from birth to 16 or 17 years of age. The anthropometric data in this cohort were collected monthly from birth until 18 months, 3 monthly until 60 months, then at 6 years, 8–9 years, 10 years, 12 years, and 15 years. Weight and height were measured during home visits, using portable spring-scales and self-made length boards, having reading increments of 100 g and 5 mm, respectively. More specific details for the Lungwena cohort are reported elsewhere (Espo et al., 2002; Maleta et al., 2003). In both cohorts, growth velocities were derived from a structural growth model fitted to growth data from birth to 60 months. The exclusion criteria and the overall number of participants available for analysis are shown in Figure 1. Because of differences in the socio-economic status (SES) variables collected in the two cohorts, an SES score was calculated separately for each cohort. The SES variables in the BH cohort included household assets such as car, TV, fridge and washing machine, and household facilities such as type of water system, toilet type and electricity. The SES variables for the Lungwena cohort included ownership of land, farm animals, bicycle, and radio, amongst others and household variables such as maternal and paternal literacy level. In each cohort, an asset score was initially derived based on household assets. Principal component analysis was then used to derive an overall SES score by combining the asset score with other household and community SES variables. Fig. 1. Flowchart of the samples available for analysis from the two cohorts. 644 E.D. CHIRWA ET AL. American Journal of Human Biology A Berkey-Reed first order model (Berkey, 1982) was fitted using mixed effects modelling. The model has the functional form; y5a1bt1cln tðÞ1d=t: where y5weight or height of child at age t, the function parameter ais related to the baseline weight or height at birth, bis related to the linear component of the growth velocity, cis related to the deceleration in growth velocity and drepresents an inflection point that allows growth velocity to peak after birth rather than at birth (MookKanamori et al., 2011). The model was modified as suggested by Simondon et al. (1992), so that it is defined at birth (t50) as follows: y5a1bt1cln t11ðÞ1d=ðt11Þ: Previous analyses have shown this model to have optimum fit to the BH cohort growth data (Chirwa et al., 2014). The model was used to describe growth patterns in early childhood after adjusting for maternal characteristics (maternal height and age), SES and gestational age. Separate models were fitted for boys and girls in each cohort. The first order derivative of the model was then used to derive weight and height velocities over time (Botton et al., 2008; Mook-Kanamori et al., 2011). The growth velocity function was then used to derive the peak weight velocity (PWV), peak height velocity (PHV) and the age at which a child reached its PWV (APWV) and its peak height velocity (APHV). The primary outcomes were BMI and the proportion of children who were overweight in the 9–11 year age group. BMI cut-off charts for children were used to calculate corresponding overweight cut-offs (Cole et al., 2000, 2007). The derived parameter estimates, growth velocity, infant peak velocity and the age at peak velocity were used as predictors of adolescent BMI or obesity. T-tests were used to compare weight, height growth velocity, peak growth velocity between boys and girls within and between cohorts. Linear regression was used to examine the relationship between BMI-for-age z-scores (BMIZ) in late childhood and early adolescence (9–11 years) and predictors, adjusting for birth weight, sex and cohort differences. Logistic regression was then used to explore predictors of obesity, adjusting for cohort differences and birth weight. Analysis was done using Stata Version 11, and all statistical tests were performed at 5% significance level. The BH study was approved by the Human Research Ethics Committee of the University of Witwatersrand, while the LCSS was approved by the Malawi National Health Science Research Committee. RESULTS Descriptive statistics The proportion of boys in the BH cohort was higher than that of girls (57% vs. 43%), while the proportion of boys and girls in the Lungwena cohort was almost the same (51% vs. 49%). Of the 216 children in the BH cohort, almost half were first born, while only 19% of the 341 children from the Lungwena cohort were first born. The average maternal age for the BH cohort was 25 years (with standard deviation (SD) of 5.9), while the average age in the Lungwena cohort was 26 years (SD 56.5). BH cohort mothers were on average taller than their Lungwena counterparts (158 cm vs. 155 cm). Table 1 shows the mean anthropometric measurements in infancy/early childhood and late childhood/early adolescence between boys and girls in the two cohorts. There were no significant differences in the average size at birth between the cohorts (P>0.05). However, BH boys and girls experienced more rapid weight gain from 3 months TABLE 1. Comparison of physical growth measurements between boys and girls in the two cohorts NBoys Girls nBoys Girls Sig 1 Sig 2 Birth and early childhood Weight (kg) Birth weight 216 3.2 60.5 3.1 60.4 341 3.3 60.5 3.2 60.5 0.115 0.070 3 months 82 6.5 60.8 5.8 60.7 274 6.0 60.8 5.5 60.7 <0.001 <0.001 6 months 52 8.1 60.9 7.4 61.0 286 7.2 61.0 6.6 60.9 0.011 <0.001 1 year 193 9.7 61.4 9.2 61.3 302 8.4 61.1 8.0 61.1 0.012 0.002 2 year 153 11.8 61.7 11.4 61.4 308 10.5 61.4 10.1 61.3 0.126 0.010 4 year 210 15.5 61.9 14.9 61.8 326 14.6 61.5 14.0 61.6 0.021 <0.001 5 year 187 18.6 62.0 17.9 62.1 338 16.0 61.8 15.5 61.9 0.022 0.016 Height (cm) Birth length – – – 341 48.9 62.2 47.9 62.1 - <0.001 3 months 82 60.8 62.9 58.7 62.8 274 57.3 62.4 56.1 62.6 0.001 <0.001 6 months 52 66.3 62.8 63.4 64.1 286 62.5 62.6 60.6 62.6 0.015 <0.001 1 year 187 74.4 63.1 72.6 63.0 302 69.0 62.6 67.8 62.6 <0.001 <0.001 2 years 143 83.6 63.7 82.2 63.1 308 78.2 63.7 76.9 63.4 0.010 0.002 4 years 210 99.2 63.9 97.9 63.9 326 93.9 64.3 92.4 64.4 0.017 0.002 5 years 187 108.8 64.0 107.4 64.2 338 100.9 64.4 99.4 64.7 0.022 0.003 Early adolescence Age (years) 216 10.5 60.3 10.5 60.3 341 10.3 60.3 10.3560.3 <0.001 0.011 Height (cm) 216 137.8 66.1 138.4 65.9 341 129.765.5 128.46 66.0 <0.001 <0.001 Weight (kg) 216 32.966.2 33.4 66.6 341 25.7 63.2 25.22 63.5 <0.001 <0.001 BMI (kg/m 2 ) 216 17.362.5 17.4 62.9 341 15.3 61.2 15.22 61.2 <0.001 <0.001 Overweight (%) 26(21%) 22(24%) 2(1.2%) 0(0%) <0.001 <0.001 Underweight (%) 1(0%) 5(5%) 13(8) 17(10%) <0.001 0.246 Sig 1 : BH boys vs. Lungwena boys. All mean comparisons done using t-test. Sig 2 : BH girls vs. Lungwena girls. All proportions comparisons done using Fishers’ exact test. POSTNATAL GROWTH AND OBESITY IN EARLY ADOLESCENTS 645 American Journal of Human Biology onwards as shown by the significant differences in mean weight from 3 months, with BH boys and girls weighing on average more than their Lungwena counterparts. Although there were no data on birth length for the BH cohort, subsequent measurements showed BH boys and girls were on average significantly taller than their Lungwena counterparts. In early adolescence, there were no significant differences in weight, age, height or BMI between boys and girls within each cohort, but there were significant differences between the cohorts, with BH boys and girls having higher mean BMI, height and weight compared to the Lungwena boys and girls. However, unlike the pattern in infancy/early childhood, girls in the BH cohort were on average taller and weighed more than boys at ages 9/10 years. Figure 1 shows the distribution of BMI for boys and girls in the two cohorts, with BH children having largely higher BMI than Lungwena children. However, there is also wider variation in the BMI values in the BH cohort (Fig. 2). Average model parameter and growth velocity estimates Parameter estimates for weight and height models were derived for each child using the random components of the mixed models. A t-test comparison of these parameter estimates showed significant differences in the average parameter estimates between BH boys and Lungwena boys for both the height and weight growth model (Table 2). Similar results were also found amongst the girls from the two cohorts. BH boys had the highest starting values (a w and a h ), as well as the highest linear growth rates (b w and b h ). However the a h for the height model for both cohorts represents starting height/length at 3 months, due to the BH cohort not having height/length measurements at birth. Within each cohort, there generally were no significant differences in model parameter estimates between boys and girls in both cohort except for b w and b h in the Lungwena cohort. Non-significant difference in the linear component of the velocity curve between boys and girls in the Lungwena cohort was also shown by the non-significant differences in both weight and height velocities. There were significant differences in the weight velocity between boys and girls in the BH cohort except at 24 months. Although boys in the Lungwena cohort tended to have higher weight velocities than girls, there were no significant differences in average weight velocity between boys and girls in the Lungwena cohort from 12 months onwards. Boys in the BH cohort generally exhibited higher height velocities than girls, with significant differences in the average height velocity between boys and girls in the first 2 years of life (P<0.001). As expected and consistent with the changes in average weight and height over time, as shown in Table 1, weight and height velocities were highest in the first 12 months, with growth rates rapidly declining from 12 months (Table 2). There were no significant differences in the average weight velocities between small for gestational age infants (WAZ at birth <22) and appropriate for gestational age (AGA) infants (data not shown). Similar results were found when comparing those with low birth weight (birth weight <2.5 kg) to those with normal birth weight (data not shown). The significant differences in the parameter estimates between BH boys and Lungwena boys as well as between girls in the two cohorts, were also shown by the significant differences in the weight and height velocities between the cohorts, with the BH boys having higher growth rates than their Lungwena counterparts. Similarly, BH girls exhibited higher growth rates than Lungwena girls. A strong positive linear relationship between weight and height velocity (r50.89, P<0.001) was observed. BH girls had the highest infancy PWV (1.39 kg/month). However, no significant difference was observed in the Fig. 2. BMI in early adolescence for boys and girls in the two cohorts. 646 E.D. CHIRWA ET AL. American Journal of Human Biology infancy PWV or PHV between sexes within the Lungwena cohort or as well as between girls in the two cohorts. BH boys which had the smallest PWV also had the youngest age at APWV. BH boy’s height velocity also peaked earliest compared to the other three groups. Relationship between birth weight, growth velocity, peak velocity and adolescent BMI There were no significant correlations between birth weight and weight velocity (r520.08, P50.06) or height velocity (r520.05, P50.27). There was a positive but weak relationship between birth weight and adolescent BMI, with birth weight only explaining 1% of the variation in BMI (Table 3). This relationship did not change even after adjusting for sex differences. However, when cohort differences were taken into account, the total variation explained by the model increased from 1% to 24%, signifying the large differences that exist between the two cohorts. The effect of birth weight on BMI was also supported by the results from the relationship between a w for the weight model and BMI, with no significant sex difference being observed. Within each of the three models using parameter estimates from weight models, b w had the strongest linear relationship with BMI compared to the other three parameter estimates, with no linear relationship being observed between adolescent BMIZ and d w . Both c w and d w were non-significant when effect of cohort and sex differences as well as birth weight, were taken into account. There was a negative linear relationship between adolescent BMIZ and APWV, indicating that infants that reached their PWV early were more likely to have high BMI in adolescence. A strong negative correlation was also observed between PWV and APWV (r520.53, P<0.001), indicating that infants with low PWV were more likely to reach their peak later than infants that exhibited high PWV. There was no significant linear relationship between adolescent BMIZ and age at which the infant reached PHV. There was a general decrease in the relationship between weight velocity and adolescent BMIZ over time even after adjusting for birth weight (R 2(3m) 50.38, R 2(60m) 50.29). Even though there was a strong relationship between adolescent BMIZ and height velocity in the first 6 months, even after adjusting for birth weight, as observed from the R 2 values (models 1 and 2), there were no differences in the strength of the relationship over time when cohort and sex differences were taken into account (model 3). TABLE 2. Average differences in parameter estimates, growth velocity and peak velocity between cohort and sex BH cohort Lungwena cohort Coh. dif. Sex dif. Boys Girls Boys Girls M F BH LUN Mean (SD) Mean (SD) Mean (SD) Mean (SD) sig sig sig Sig Parameters (weight) a w 6.70 (0.45) 4.44 (0.34) 5.49 (0.66) 5.13 (0.63) aaaa b w 0.16 (0.02) 0.13 (0.03) 0.14 (0.02) 0.14 (0.03) a 0.023 a 0.816 c w 0.42 (0.02) 1.16 (0.03) 0.54 (0.02) 0.48 (0.03) aaaa d w 23.50 (0.03) 21.36 (0.03) 22.68 (0.03) 22.43 (0.03) aaaa Parameters (height) a h 66.5 (2.07) 56.4 (1.95) 54.1 (2.28) 53.5(2.20) aaa 0.030 b h 0.59 (0.04) 0.52 (0.06) 0.52 (0.06) 0.52 (0.07) a 0.815 a 0.654 c h 1.37 (0.03) 4.42 (0.06) 3.79 (0.07) 3.56 (0.08) aaaa d h 239.8 (0.04) 222.6 (0.05) 215.3 (0.06) 216.8 (0.08) aaaa Peak velocity PWV 1.31 (0.26) 1.39 (0.28) 1.36 (0.24) 1.35 (0.29) 0.067 0.217 0.024 0.579 APWV (months) 2.49 (0.02) 2.86 (0.05) 3.10 (0.06) 2.96 (0.07) aaaa PHV 5.77 (0.37) 5.59(0.58) 5.51 (0.67) 5.53 (0.84) a 0.556 0.006 0.882 APHV (months) 5.00 (0.08) 7.24 (0.34) 6.92 (0.47) 5.43 (0.27) aaaa Weight velocity (kg/months) 3 months 0.48 (0.03) 0.50 (0.03) 0.44 (0.03) 0.41 (0.03) aaaa 6 months 0.29 (0.03) 0.33 (0.03) 0.27 (0.03) 0.26 (0.03) aaaa 12 months 0.21 (0.03) 0.23 (0.03) 0.20 (0.03) 0.19 0.03) aaa 0.056 24 months 0.18 (0.02) 0.18 (0.03) 0.17 (0.02) 0.17 (0.03) aa 0.795 0.431 48 months 0.17 (0.02) 0.16 (0.03) 0.15 (0.02) 0.15 (0.03) a 0.103 0.002 0.796 60 months 0.17 (0.02) 0.15 (0.03) 0.15 (0.02) 0.15 (0.03) a 0.540 a 0.873 Height velocity (cm/monthso) 3 months 3.34 (0.17) 2.99 (0.14) 2.41 (0.10) 2.45 (0.11) aaa 0.002 6 months 1.61 (0.07) 1.62 (0.07) 1.37 (0.08) 1.37 (0.09) aa 0.221 0.468 12 months 0.94 (0.05) 1.00 (0.07) 0.91 (0.07) 0.89 (0.09) aaa 0.144 24 months 0.71 (0.04) 0.73 (0.06) 0.70 (0.07) 0.69 (0.09) 0.134 aa 0.211 48 months 0.64 (0.04) 0.63 (0.06) 0.61 (0.06) 0.60 (0.08) a 0.015 0.019 0.352 60 months 0.63 (0.04) 0.60 (0.06) 0.58 (0.06) 0.58 (0.08) a 0.123 a 0.393 M5BH boys vs. Lungwena boys F5BH girls vs. Lungwena girls BH 5BH boys vs. BH girls LUN 5Lungwena boys vs. Lungwena girls PWV5Peak weight velocity (kg/months) APWV 5Age at peak weight velocity (month) PHV5Peak height velocity (cm/months) APHV5Age at peak height velocity (month) All comparisons done using the t-test. a :P-value <0.001. POSTNATAL GROWTH AND OBESITY IN EARLY ADOLESCENTS 647 American Journal of Human Biology Relationship between birth weight, growth velocity, peak velocity and adolescent obesity Table 4 shows the association between adolescent overweight and growth velocity. The study found no association between sex and being overweight adolescent, even though girls had lower odds of being overweight compared to boys (OR 50.88, P50.671). The Lungwena cohort had lower odds of being overweight than the BH cohort (OR 50.02, P<0.001). The odds did not change even after adjusting for sex differences. Despite the odds of being overweight increasing with increase in birth weight, the association was not significant (OR 51.25, P50.486). Consistent with the observed relationship between BMI and linear growth rates in weight (b w )asshownin Table 3, the study also found strongest association between being overweight and linear growth rate in weight (b w ), even after adjusting for cohort differences and birth weight (OR 52.05, P-value <0.001). While there was a strong association between being an overweight adolescent and a child’s estimated baseline weight (a w ), this relationship was not significant when cohort differences were taken into account. After adjusting for cohort differences and birth weight, only the linear growth rate function (b w )wasfoundtobeassociated with adolescent overweight. The study also found strong association between adolescent overweight and linear growth rates in height, with children exhibiting faster height growth rates being more likely to be overweight in adolescence. However, this relationship was non-significant when cohort differences were taken into account. Consistent with weight model parameters, baseline height (a h ) and the decrease in height velocity over time (c h ) were also not associated with being overweight. The logistic regression models showed a stronger association between overweight in early adolescence (ages 9–11 years) and weight gain in infancy than with weight gain in early childhood. At 3 months, every 1 SD increase in weight velocity had an eightfold odds of being overweight in early adolescence. These odds reduced with age such that by the time a child is 5-years-old, every 1 SD increase in weight velocity resulted in almost threefold odds of being overweight. Despite there being a decrease in the odds of being overweight after adjusting for cohort differences and birth weight, the trend over time was the same. The same trend was observed with height velocity. However, there were no significant association between height velocity and being overweight after adjusting for birth weight and cohort differences. No association was found between obesity and PHV, age at PHV or age at PWV, TABLE 3. Relationship between adolescent BMI and growth velocity, peak velocity and Reed1 model parameters, adjusting for sex and cohort differences Main predictor Model 1 Model 2 Model 3 b(SE) R 2 b(SE) R 2 b(SE) R 2 Sex : boys Ref girls 20.18 (0.09) a 0.01 Cohort: BH Ref LUN 21.07 (0.08) 0.22 Birth weight 0.27 (0.10) 0.01 Parameters (weight) a w 0.34 (0.05) 0.08 0.33 (0.05) 0.08 0.25 (0.06) a 0.27 b w 21.2 (1.55) 0.25 21.3 (1.61) 0.25 18.9 (1.45) 0.42 c w 0.96 (0.18) 0.05 1.00 (0.18) 0.07 0.31 (0.21) a 0.25 d w 20.05 (0.07) a 0.001 20.04 (0.07) a 0.01 0.06 (0.09) a 0.25 Parameters (height) a h 0.08 (0.01) 0.15 0.07 (0.01) 0.15 0.001(0.013) a 0.25 b h 4.32 (0.62) 0.08 4.18 (0.63) 0.08 2.20 (0.60) 0.26 c h 20.24 (0.04) 0.05 20.24 (0.04) 0.06 0.017 (0.05) a 0.25 d h 20.05 (0.01) 0.16 20.05 (0.002) 0.18 0.001 (0.01) a 0.25 Infant peak velocity PWV 1.85 (0.16) 0.20 1.84 (0.16) 0.20 1.86 (0.14) 0.44 PHV 0.35 (0.07) 0.05 0.33 (0.07) 0.05 0.23 (0.06) 0.26 APWV 21.33 (0.19) 0.08 21.39 (0.19) 0.10 1.32 (0.29) 0.27 APHV 20.004 (0.05) a 0.001 20.008 (0.05) a 0.01 0.05 (0.04) a 0.24 Weight velocity 3 months 14.8 (0.80) 0.38 14.7 (0.81) 0.38 12.3 (1.05) 0.40 6 months 18.3 (1.03) 0.36 18.2 (1.04) 0.36 14.8 (1.23) 0.40 12 months 21.7 (1.28) 0.34 21.7 (1.31) 0.34 17.4 (1.37) 0.42 24 months 23.5 (1.43) 0.33 23.7 (1.47) 0.33 19.3 (1.42) 0.43 48 months 23.3 (1.50) 0.30 23.5 (1.55) 0.30 19.6 (1.44) 0.43 60 months 22.5 (1.52) 0.29 23.1 (1.57) 0.29 19.5 (1.44) 0.43 Height velocity 3 months 1.31(0.10) 0.24 1.32 (0.10) 0.25 0.81 (0.25) 0.26 6 months 3.68 (0.29) 0.22 3.68 (0.29) 0.24 1.86 (0.50) 0.26 12 months 3.79 (0.54) 0.08 3.72 (0.55) 0.09 1.50 (0.54) 0.26 24 months 3.67 (0.65) 0.05 3.54 (0.65) 0.06 2.22 (0.59) 0.26 48 months 4.07 (0.66) 0.06 3.93 (0.66) 0.07 2.32 (0.61) 0.26 60 months 3.98 (0.65) 0.06 3.83 (0.67) 0.07 2.30 (0.61) 0.26 a Effect of main predictor was not significant. Model 1: Adolescent BMI vs. main predictor. Model 2: Adolescent BMI vs. main predictor (adjusting for birth weight). Birth weight was non-significant when adjusted for the main predictors in Model 2. Model 3: Adolescent BMI vs. main predictor (adjusting for birth weight, sex and cohort differences). R 2 5total variation in BMI explained by the overall model. 648 E.D. CHIRWA ET AL. American Journal of Human Biology when adjusted for cohort difference. However children with high PWV were more likely to be overweight in adolescence even after adjusting for cohort differences. DISCUSSION This study has been able to demonstrate a positive linear relationship between rapid weight gain in infancy and early childhood and early adolescent BMI, as well as shown a relationship between rapid growth in the early years and the odds of being overweight/obese in early adolescence. The high odds ratio and R 2 values between growth velocity in the first year of life (the period which was also characterised by high growth velocity) and adolescent BMI, highlight the association between rapid weight gain and obesity in early adolescence. The decreasing trend in the OR values in later early childhood highlights the critical period during infancy that is highly associated with adolescence/adult obesity. This supports what prior studies in the BH cohort and others elsewhere have found, albeit using different methods or measures (Cameron et al., 2003; Ekelund et al., 2007; McCarthy et al., 2007; Botton et al., 2008; Adair et al., 2009; Demerath et al., 2009; Stein et al., 2010). In a study of the relationship between rapid weight gain in the first 2 years of life and obesity in childhood in BH children with appropriate birth weight for gestational age (AGA), Cameron et al. (2003), using weight-for-age z-scores, found that children that exhibited rapid growth in infancy were significantly taller, and weighed more in childhood. Our study has been able to demonstrate this using parameter estimates from the Reed1 model, with the parameter b w , a function related to growth velocity being highly positively associated with early adolescent BMI. Consistent with the study by Mook-Kanamori et al., our study found high PWV to be highly associated with early adolescence overweight. The significance of the relationship between rapid weight gain in infancy and adolescent BMI was also highlighted by the negative associated between APWV and adolescent BMI, indicating that infants that reached infant PWV early were more likely to have high BMI. Apart from that, our study has also explored the relationship using height velocity and extended the period to early adolescence (9– 11 years). Our study has also been able to show similar association between rapid growth and adolescent BMI in a rural population, which is from a predominately malnourished population, with high levels of stunting and underweight. The critical period of development is the same in both cohorts. However, the rapid infant growth in this rural population seems to have beneficial effects, as it protects the adolescent child from the effects of undernutrition,with few cases of obesity. There are several hypothesised biological relationships between prenatal and postnatal growth and obesity in later life, and a large body of evidence supports these hypothesised relationships (Ong and Loos, 2006; Adair, 2007; Ekelund et al., 2007; Jones-Smith et al., 2007; McCarthy et al., 2007; Chomtho et al., 2008; Druet et al., 2012). These previous studies showed that either small size at birth, small size at birth combined with fast growth or fast growth itself, have effects on later life health outcomes. Studies have also shown that low-birth-weight infants usually exhibit rapid growth during the first year of life (Ong, 2006; Adair, 2007; Johnson et al., 2012). However, our study which used data from two cohorts from different settings in terms of environmental and socioeconomic factors, found no relationship between size at birth (birth weight) and growth velocity in both cohorts. Similarly, we found no association between size at birth and overweight in early adolescence. Both birth weight and its estimated parameter (a w ) were not associated with adolescent overweight. The non-significant relationship between birth weight and growth velocity as well as with overweight in early adolescence could also be due to the limited range of birth weight measurements, since our sample excluded preterm babies. Even though there was a wide variation in the age of initial weight measurements for the Lungwena cohort, for babies not delivered in a health facility, the mixed effects model adjusted for the age at which the measurements were taken. However, the effect of the differences in the environmental and socio-economic factors in the two cohorts were shown by the differences in the growth rates and the postnatal prevalence of stunting/underweight and overweight in the cohorts as well as the significance of cohort term in the models. Despite there being no significant differences in birth weights between the two cohorts, the urban BH children exhibited more rapid weight gain in the first year TABLE 4. Odds ratios from the relationship between overweight and growth velocity, peak velocity and Reed 1 parameters, adjusting for sex and cohort difference Main predictor Model 1 Model 2 Model 3 a OR (95% CI) OR (95% CI) OR (95% CI) Sex :boys Ref :girls 0.88 (0.49, 1.58) b Cohort: BH Ref LUN 0.02 (0.005, 0.09) Birth weight 1.25 (0.67, 2.34) b Parameters (weight) a w 1.53 (1.19, 1.97) 1.55 (1.20, 2.02) 1.12 (0.89, 1.40) b b w 2.21 (1.65, 2.96) 2.27 (1.68, 3.08) 2.05 (1.46, 2.87) c w 1.46 (1.24, 1.72) 1.48 (1.25, 1.75) 1.08 (0.92, 1.27) b d w 1.06 (0.87, 1.29) b 1.07 (0.88, 1.31) b 1.05 (0.92, 1.21) b Parameters (height) a h 1.79 (1.46,2.21) 1.79 (1.45, 2.22) 0.88 (0.67, 1.15) b b h 1.71 (1.21, 2.42) 1.70 (1.20,2.41) 1.14 (0.77, 1.70) b c h 0.74 (0.62, 0.89) 0.74 (0.62, 0.89) 1.05 (0.90, 1.24) b d h 0.53 (0.43, 0.65) 0.53 (0.43, 0.65) 1.10 (0.84, 1.45) b Infancy peak velocity PWV 1.42 (1.08, 1.87) 1.42 (1.07, 1.88) 1.68 (1.17, 2.42) PHV 1.58 (0.99, 2.52) b 1.56 (0.97, 2.50) b 1.44 (0.76, 2.74) b APWV 0.06 (0.01, 0.12) 0.03 (0.01, 0.12) 4.31 (0.78, 23.7) b APHV 0.99 (0.74, 1.35) b 0.99 (0.74, 1.34) b 1.17 (0.88, 1.55) b Weight velocity 3 months 7.49 (4.50, 12.46) 7.96 (4.68, 13.52) 4.80 (2.49, 9.26) 6 months 4.07 (2.82, 5.86) 4.09 (2.84, 5.90) 2.60 (1.77, 3.83) 12 months 3.51 (2.48, 4.94) 3.58 (2.52, 5.08) 2.46 (1.89, 3.61) 24 months 3.06 (2.20, 4.25) 3.18 (2.26, 4.47) 2.44 (1.68, 3.55) 48 months 2.67 (1.95, 3.66) 2.78 (2.01, 3.86) 2.41 (1.68, 3.64) 60 months 2.60 (1.90, 3.56) 2.71 (1.96, 3.75) 2.39 (1.65, 3.47) Height velocity 3 months 3.02 (2.19, 4.15) 3.01 (2.19, 4.14) 0.87 (0.50, 1.50) b 6 months 5.49 (3.27, 9.22) 5.52 (3.29, 9.24) 1.62 (0.61, 4.32) b 12 months 2.25 (1.60, 3.15) 2.25 (1.60, 3.16) 1.35 (0.89, 2.03) b 24 months 1.55 (1.13, 2.14) 1.54 (1.12, 2.13) 1.34 (0.88, 2.04) b 48 months 1.51 (1.10, 2.09) 1.50 (1.09, 2.07) 1.32 (0.86, 2.03) b 60 months 1.55 (1.12, 2.14) 1.53 (1.11, 2.13) 1.26 (0.82, 1.94) b a Not adjusted for sex due to limited number of children in Lungwena cohort with outcome. b Effect of main predictor was not significant. Model 1: Overweight vs. main predictor. Model 2: Overweight vs. main predictor (adjusting for birth weight). Model 3: Overweight vs. main predictor (adjusting for birth weight and cohort differences). POSTNATAL GROWTH AND OBESITY IN EARLY ADOLESCENTS 649 American Journal of Human Biology of life. This rapid weight gain was associated with a high prevalence of overweight adolescents in this population. The differences in the prevalence of overweight adolescents in the two cohorts, considering the non-significant differences in their size at birth, highlights the significance of rapid weight gain rather than birth size, in the relationship between early growth and obesity in adolescence, in this particular setting. These results are in support of the “fast growth and obesity” hypothesis, rather than the “size at birth” or the “size at birth and fast growth” hypotheses. The relationship between faster growth velocity and obesity/overweight in later life, independent of birth weight, has been hypothesised to be mainly due to over-nutrition (Jones-Smith et al., 2013). The more rapid weight gain in the BH cohort relative to the Lungwena cohort may be due to nutritional and environmental differences, among other factors. The slower weight velocity in the Lungwena cohort, from as early as 3 months, could be due to poor maternal nutritional status and the early introduction of complementary foods. As Lungwena is predominately a poor rural community, the complementary foods used are likely to be of poor nutritional content and to expose the infants to pathogens (Espo et al., 2002). Our results are in general consistent with study by Adair et al. (2013), which also looked at association between weight and height gain, and adult health outcomes in five cohorts from LMIC (Adair et al., 2013). The study found a positive relationship between weight gain and adult BMI, with the strength of the relationship increasing with age at which measurement was taken. However, unlike our study, they also found positive relationship between birth weight and adult BMI, and they also found a decreasing relationship between height gain and BMI. The variations in these results could be due to the differences in the age ranges used as well as the limited amount of observations at 3 and 6 months in the BH cohort of our study. Apart from looking at cohorts from different SES and environmental settings, the other strength of this study is in the use of mixed effects modelling to model growth trajectories and to derive growth velocities. Mixed effects modelling allowed us to compare growth velocity at any age even though some of the data collection waves in the two cohorts were at different times. Our results are in general, consistent with results from other studies that have used mixed effects. In a study of Dutch children, Mook-Kanamori et al., also using the Berkey-Reed model and mixed effects modelling, found that rapid weight gain in the first months was more associated with risk of overweight than catch-up growth (‘size at birth and fast growth hypothesis’) during the first 2 years (Mook-Kanamori et al., 2011). Similarly, Botton et al. (2008) using the adapted Jenns-Bayley model and using mixed effects modelling, also found increased risk of obesity due to rapid growth in the first 6 months in French children. However, their study also found that this risk started increasing again from 3 years. However, they derived their growth velocity from a model fitted from birth to 10 years, which may have made it possible to pick out the increase in growth velocity from 3 years. Our study fitted the growth model up to 5 years only. The main limitation of the study is unavailability of data on adolescent factors associated with BMI in one of the cohort, which could have been adjusted for in the relationship between postnatal growth and adolescent BMI/ overweight. The other limitation for the study is the amount of missing weight and height measurements during the first year of life in the BH cohort which could have affected the fitness of the growth model used for estimating growth velocity. In conclusion, although our results support the hypothesis that rapid growth in infancy increases the risk of overweight/ obesity in later life, the long term effects of infancy rapid growth are dependent on the particular population’s stage of nutrition transition. For a population in early stages of nutrition transition or with poor nutritional status, rapid growth in early childhood may have long term beneficial effects as was evidenced by the almost non-existent prevalence of overweight in the Lungwena cohort, despite some children exhibiting rapid growth in early childhood. Conversely, for populations undergoing rapid nutrition transition as is the case with the BH cohort, rapid growth has detrimental long term effects, as was evidenced by the prevalence of overweight and obesity in early adolescence. To further explore the relationship between postnatal growth velocity and later health outcome, we would recommend modelling growth into adolescence and also include pubertal stages and SES factors during adolescence that are highly associated with BMI, such as dietary patterns and physical activity behaviors of the adolescents in the cohorts. Further studies in similar cohorts in LMIC might also help in explaining the effect of shifts in dietary and sedentary behaviors associated with urbanization. LITERATURE CITED Adair L. 2007. Size at birth and growth trajectories to young adulthood. Am J Hum Biol 19:327–337. Adair LS, Fall CH, Osmond C, Stein AD, Martorell R, Ramirez-Zea M, Sachdev HS, Dahly DL, Bas I, Norris SA, Micklesfield L, Hallal P, Victora CG. 2013. Associations of linear growth and relative weight gain during early life with adult health and human capital in countries of low and middle income: findings from five birth cohort studies. Lancet 382: 525–534. Adair LS, Martorell R, Stein AD, Hallal PC, Sachdev HS, Prabhakaran D, Wills AK, Norris SA, Dahly DL, Lee NR, Victora CG. 2009. Size at birth, weight gain in infancy and childhood, and adult blood pressure in 5 lowand middle-income-country cohorts: when does weight gain matter? Am J Clin Nutr 89:1383–1392. Berkey CS. 1982. Comparison of two longitudinal growth models for preschool children. Biometrics 38:221–234. Black MM, Krishnakumar A. 1999. Predicting longitudinal growth curves of height and weight using ecological factors for children with and without early growth deficiency. J Nutr 129 (2S Suppl):539S–543S. Botton J, Heude B, Maccario J, Ducimetiere P, Charles M. 2008. Postnatal weight and height growth velocities at different ages between birth and 5y and body composition in adolescent boys and girls. Am J Clin Nutr 87:1760–1768. Cameron N. 1984. The measurement of human growth. London: Croom Helm. Cameron N, Demerath EW. 2002. Critical periods in human growth and their relationship to diseases of aging. Am J Phys Anthropol Suppl 35: 159–184. Cameron N, Pettifor J, De Wet T, Norris S. 2003. The relationship of rapid weight gain in infancy to obesity and skeletal maturity in childhood. Obes Res 11:457–460. Cameron N, Wright MM, Griffiths PL, Norris SA, Pettifor JM. 2005. Stunting at 2 years in relation to body composition at 9 years in African urban children. Obes Res 13:131–136. Chirwa ED, Griffiths PL, Maleta K, Norris SA, Cameron N. 2014. Multilevel modelling of longitudinal child growth data from the Birth-toTwenty Cohort: a comparison of growth models. Ann Hum Biol 41:166– 177. Chomtho S, Wells JC, Williams JE, Lucas A, Fewtrell MS. 2008. Associations between birth weight and later body composition: evidence from the 4-component model. Am J Clin Nutr 88:1040–1048. 650 E.D. CHIRWA ET AL. American Journal of Human Biology Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. 2000. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 320:1240–1243. Cole TJ, Flegal KM, Nicholls D, Jackson AA. 2007. Body mass index cut offs to define thinness in children and adolescents: international survey. BMJ 335:194. Demerath EW, Jones LL, Hawley NL, Norris SA, Pettifor JM, Duren D, Chumlea WC, Towne B, Cameron N. 2009. Rapid infant weight gain and advanced skeletal maturation in childhood. J Pediatr 155:355–361. Druet C, Stettler N, Sharp S, Simmons RK, Cooper C, Smith GD, Ekelund U, Levy-Marchal C, Jarvelin MR, Kuh D, Ong KK. 2012. Prediction of childhood obesity by infancy weight gain: an individual-level meta-analysis. Paediatr Perinat Epidemiol 26:19–26. Ekelund U, Ong KK, Linne Y, Neoyius M, Brage S, Dunger DB, Wareham NJ, Rossner S. 2007. Association of weight gain in infancy and early childhood with metabolic risk in young adults. J Clin Endocrinol Metab 92:98–103. Espo M, Kulmala T, Maleta K, Cullinan T, Salin ML, Ashorn P. 2002. Determinants of linear growth and predictors of severe stunting during infancy in rural Malawi. Acta Paediatr 91:1364–1370. Flexeder C, Thiering E, Bruske I, Koletzko S, Bauer CP, Wichmann HE, Mansmann U, Von Berg A, Berdel D, Kramer U, Schaaf B, Lehmann I, Herbarth O, Heinrich J. 2012. Growth velocity during infancy and onset of asthma in school-aged children. Allergy 67:257–264. Hoddinott J, Maluccio JA, Maluccio JA, Behrman JR, Flores R, Martorell R. 2008. Effect of a nutrition intervention during early childhood on economic productivity in Guatemalan adults. Lancet 371:411–416. Hoffman DJ, Sawaya AL, Verreschi I, Tucker KL, Roberts SB. 2000. Why are nutritionally stunted children at increased risk of obesity? Studies of metabolic rate and fat oxidation in shantytown children from Sao Paulo, Brazil. Am J Clin Nutr 72:702–707. Johnson W, Vazir S, Fernandez-Rao S, Kankipati VR, Balakrishna N, Griffiths PL. 2012. Using the WHO 2006 child growth standard to assess the growth and nutritional status of rural south Indian infants. Ann Hum Biol 39:91–101. Jones-Smith JC, Fernald LC, Neufeld LM. 2007. Birth size and accelerated growth during infancy are associated with increased odds of childhood overweight in Mexican children. J Am Diet Assoc 107:2061–2069. Jones-Smith JC, Neufeld LM, Laraia B, Ramakrishnan U, Garcia-Guerra A, Fernald LC. 2013. Early life growth trajectories and future risk for overweight. Nutr Diabetes 3:e60. Kalanda BF, van Buuren S, Verhoeff FH, Brabin BJ. 2005. Catch-up growth in Malawian babies, a longitudinal study of normal and low birthweight babies born in a malarious endemic area. Early Hum Dev 81:841–850. Li H, Stein AD, Banrhart HX, Ramakrishnan U, Martorell R. 2003. Association between prenatal and postnatal growth and adult body size and composition. Am J Clin Nutr 77:1498–1505. Maleta K, Virtanen SM, Espo M, Kulmala T, Ashorn P. 2003. Seasonality of growth and the relationship between weight and height gain in children under three years of age in rural Malawi. Acta Paediatr 92:491– 497. Martorell R, Schroeder DG, Rivera JA, Kaplowitz HJ. 1995. Patterns of linear growth in rural Guatemalan adolescents. J Nutr 125 (4 Suppl): 1060s–1067s. McCarthy A, Hughes R, Tilling K, Davies D, Smith GD, Ben-Shlomo Y. 2007. Birth weight; postnatal, infant, and childhood growth; and obesity in young adulthood: evidence from the Barry Caerphilly Growth Study. Am J Clin Nutr 86:907–913. Mesa JM, Araujo C, Horta BL, Gigante DP. 2010. Growth patterns in early childhood and the onset of menarche before age twelve. Rev Saude Publica 44:249–260. Mook-Kanamori DO, Durmus B, Sovio U, Hofman A, Raat H, Steegers EA, Jarvelin MR, Jaddoe VW. 2011. Fetal and infant growth and the risk of obesity during early childhood: the Generation R Study. Eur J Endocrinol 165:623–630. Ong KK. 2006. Size at birth, postnatal growth and risk of obesity. Horm Res 65 (Suppl 3):65–69. Ong KK, Loos RJ. 2006. Rapid infancy weight gain and subsequent obesity: systematic reviews and hopeful suggestions. Acta Paediatr 95:904–908. Ong KK, Ahmed ML, Emmet PM, Preece MA Dunger DB. 2000. Association between postnatal catch-up growth and obesity in childhood: prospective cohort study. BMJ 320:967–971. Ridgway CL, Ong KK, Tammelin T, Sharp SJ, Ekelund U, Jarvelin MR. 2009. Birth size, infant weight gain, and motor development influence adult physical performance. Med Sci Sports Exerc 41:1212–1221. Salonen MK, Kajantie E, Osmond C, Forsen T, Yliharsila H, PaileHyvarinen M, Barker DJ, Eriksson JG. 2009. Childhood growth and future risk of the metabolic syndrome in normal-weight men and women. Diabetes Metab 35:143–150. Schroeder DG, Martorell R, Flores R. 1999. Infant and child growth and fatness and fat distribution in Guatemalan adults. Am J Epidemiol 149: 177–185. Simondon KB, Simondon F, Delpeuch F, Cornu A. 1992. Comparative study of five growth modelsaplied to weight data from congolese infants between birth and 13 months of age. Am J Hum Biol 4:327–335. Stein AD, Wang M, Martorell R, Norris SA Adair LS, Bas I, Sachdev HS, Bhargava SK, Fall CH, Gigante DP, Victora CG. 2010. Growth patterns in early childhood and final attained stature: data from five birth cohorts from lowand middle-income countries. Am J Hum Biol 22:353– 359. Victora CG, Barros FC, Horta BL, Martorell R. 2001. Short-term benefits of catch-up growth for small-for-gestational-age infants. Int J Epidemiol 30:1325–1330. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, Richter L, Sachdev HS. 2008. Maternal and child undernutrition: consequences for adult health and human capital. Lancet 371:340–357. Walker SP, Chang SM, Powell CA. 2007. The association between early childhood stunting and weight status in late adolescence. Int J Obes 31: 347–352. POSTNATAL GROWTH AND OBESITY IN EARLY ADOLESCENTS 651 American Journal of Human Biology