Progression of stunting and its predictors among school-aged children ...

European Journal of Clinical Nutrition (2005) 59, 914–922

& 2005 Nature Publishing Group All rights reserved 0954-3007/05 $30.00 www.nature.com/ejcn

ORIGINAL COMMUNICATION Progression of stunting and its predictors among school-aged children in western Kenya JF Friedman1,2*, PA Phillips-Howard3,4, LB Mirel3, DJ Terlouw3, N Okello4, JM Vulule4, WA Hawley3, BL Nahlen3 and F ter Kuile3,4,5 1

US Department of State, Institute of International Education, Fulbright Fellowships, Washington, DC, USA; 2International Health Institute and Department of Pediatrics, Brown University School of Medicine, Providence, RI, USA; 3Division of Parasitic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; 4Centre for Vector Biology and Control Research, Kenya Medical Research Institute, Kisumu, Kenya; and 5Department of Infectious Diseases, Tropical Medicine & Aids, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands

Objectives: The objectives of this study were (1) to assess whether a cohort of school-aged children experiences progression of stunting over a 2-y-period of observation and (2) to identify baseline nutritional and body composition risk factors for the progression of stunting. Methods: As part of a large-scale, randomized controlled trial assessing the impact of insecticide-treated bednets (ITNs) on nutritional status, we longitudinally followed a cohort of school-aged children over a 2-y-period in western Kenya. Anthropometric measurements were made at four time points from which Z-scores for height-for-age (HAZ), weight-for-age (WAZ), and body mass index (BMIZ) were calculated. Two measures of body composition, upper arm fat area and upper arm muscle area, were derived from mid-upper arm circumference (MUAC) and triceps skinfold thickness. Results: Subjects experienced a mean change in HAZ from baseline to 9 months of 0.16 [0.19, 0.13], from baseline to 16 months of 0.18 [0.22, 0.15], and from baseline to 24 months of 0.36 [0.41, 0.31]. Thus, the average individual’s change in HAZ at the three follow-up time points is significantly less than zero, meaning that, on average, the cohort is deviating further from NCHS reference medians over time. The baseline nutritional measure that explained the greatest amount of variance in the progression of stunting was the upper arm muscle area Z-score (F ¼ 8.1; P ¼ 0.005). Conclusions: This longitudinal study provides further evidence from a distinct ecological setting regarding the progression of undernutrition during middle childhood in the developing world. It suggests that school-aged children in the developing world do not experience catch-up growth or remain stable. Rather, they continue to deviate from NCHS standards, accruing greater height deficits with age. In addition, absolute lean body mass explained the most variability in the progression of stunting, which supports cross-sectional findings from other studies.

European Journal of Clinical Nutrition (2005) 59, 914–922. doi:10.1038/sj.ejcn.1602161; published online 1 June 2005 Keywords: malaria; stunting; PEM; malnutrition; school; Kenya

Introduction Until recently, there was a paucity of data addressing the nutritional status of school-aged children in the developing world, with most work focusing on children under five. Studies from the Partnership for Child Development and other sources have recently highlighted the problem of

*Correspondence: J Friedman, International Health Institute, Box G-B495, Brown University, Providence RI, USA. E-mail: [email protected] Received 7 May 2004; revised 4 March 2005; accepted 7 April 2005; published online 1 June 2005

undernutrition among school-aged children in the developing world (Stoltzfus et al, 1997; Partnership for Child Development, 1998a, b; Lwambo et al, 2000). The recognized cost-effectiveness of school-based interventions to reduce the global burden of disease has further focused attention on delivery of health care in the school setting (Musgrove, 1993). Studies conducted across a range of ecologic settings in the developing world demonstrate that children enter middle childhood having already accrued significant deficits in nutritional status as assessed by height-for-age (HAZ) (stunting) and weight-for-age (WAZ) (underweight). (Martorell et al, 1995; Stoltzfus et al, 1997; Partnership for

Progression of stunting among school-aged children JF Friedman et al

915 Child Development, 1998a) Although children experience a slower phase of linear growth during middle childhood, (Tanner, 1990), they will experience one of three patterns of growth during this period. They may experience catch-up growth, defined as height velocity above the statistical limits of normality for age and gender during a defined period of time (Boersma & Wit, 1997). Catch-up growth closes the gap between a child’s HAZ and that of international reference standards. They may remain stable, growing at approximately the same velocity of children in better environments, such that they complete middle childhood with only the deficits accrued during the first few years of life. Finally, they may continue to falter, growing at a slower velocity than children in better environments, such that deficits during middle childhood are added to those accrued during the first few years of life. Although it is commonly believed that stunting occurs mainly in young children (Martorell et al, 1994), recent studies suggest that during middle childhood, children do not catch up or even remain stable, but instead experience progression of stunting. In Sub-Saharan Africa, cross-sectional studies demonstrate an increasing prevalence of stunting with age (Partnership for Child Development, 1998a, b; Lwambo et al, 2000; Monyeki et al, 2000), and longitudinal studies demonstrate that weight and height decline further from the international reference median over time (Stoltzfus et al, 1997). Growth faltering is of great public health significance, as attained height has important implications for adult work capacity (Martorell & Arroyave, 1988; Norgan, 2000) and reproductive outcomes. (Martorell et al, 1981; Eveleth, 1985). It is important to understand the dynamics of ponderal and linear growth during middle childhood, as the timing of nutritional interventions depends on identifying periods of faltering. This should be addressed across different populations that vary genetically and environmentally, such that more general patterns and plans for intervention may be established. Longitudinal studies, in particular, are essential, as growth is a dynamic process, and school-based crosssectional studies examining age and growth may suffer from threats to validity based on cohort effects and reverse causality. Longitudinal studies allow an examination of within-person changes over time, allowing better inference to be made about dynamic processes such as growth. In addition, identification of nutritional and body compositional factors that place children at risk of poor linear growth are important to identify. Cross-sectional studies have demonstrated that children with greater lean body mass are significantly taller than those with less lean body mass. (Frisancho & Garn, 1971; Frisancho et al, 1971). Studies in industrialized countries also suggest that decreased lean body mass is related to growth failure in children with HIV and renal failure (Arpadi et al, 1998; Arpadi et al, 2000; Johnson et al, 2000). As part of a large-scale, randomized controlled trial assessing the impact of insecticide-treated bednets (ITNs)

on nutritional status (Friedman et al, 2003), we followed a cohort of school-aged children over a 2-y-period in western Kenya. The objectives of this study were (1) to assess whether a cohort of school-age children experiences progression of stunting over a 2-y-period of observation and (2) to identify baseline nutritional and body composition risk factors for progression of stunting. Identification of biologic factors related to progression of stunting may help to identify children at greatest risk and glean some insight into possible mechanisms of growth faltering.

Materials and methods Study area and population This study was conducted in the context of a large randomized controlled trial of ITNs for which the primary outcome was under-five mortality (Phillips-Howard et al, 2003b). The study area is in Rarieda Division, western Kenya, which covers an area of 200 km2 within a subsection of Siaya and Bondo Districts. The southern boundary of the study area lies on Lake Victoria. The prevalence of stunting (HAZo2), wasting (WHZo2) and, underweight (WAZo-2) was 30, 4, and 20%, respectively among children under five in this study area (Kwena et al, 2003). The study area is populated predominately by individuals of the Luo ethnic group. Polygamy is a common practice, with a number of wives and their children living in separate houses within the family compound. The principal occupation is subsistence farming. Maize, sorghum, cassava, millet, and a few other vegetables are cultivated and some animal husbandry (primarily cattle) is done. Fishing on Lake Victoria supplements the diet and is used to earn cash. In most years, there is a long rainy season from March to May and a short rainy season from October to December. This pattern of rainfall is sufficient to maintain malaria transmission perennially, and the average number of infective bites ranges from 60 to 300 per person per year (Beier et al, 1994). A study conducted in 15 villages contained within this study area found average monthly Plasmodium falciparum parasite prevalences of 75 and 60% among children aged 5–9 and 10–14 y, respectively (Bloland et al, 1999). Previous studies conducted in this district have demonstrated high prevalences of Schistosoma mansoni (24– 84%), Ascaris (16–43%), Trichuris (24–59%), and hookworm infections (63–75%) among school-aged children (Olsen, 1998; Olds et al, 1999; Brooker et al, 2001). At the outset of the study, 58 primary schools were recorded within the boundaries of the study area. Most schools have a catchment area of three to five villages. School fees are generally equivalent to approximately $8.00 US dollars per trimester. Approximately 80% of children 5– 12 y of age attend primary school at any given time (District Education Officer, personal communication). Discontinuity in school attendance is common. Children may miss entire semesters due to prohibitive school fees or the need for the child’s assistance at home. European Journal of Clinical Nutrition


916 Assembly of subjects Primary schools in the study area were visited in order of selection by a random number generator. Schools were eligible to participate if the headmaster agreed and allowed study staff to explain the study to parents. In addition, eligible schools had a mix of children from ITN and control villages within a range of 40–60%. In all, 13 schools were approached before the desired sample size of approximately 1000 children under age 13 y was achieved with seven primary schools. Ineligibility of the six schools was due to unequal proportions of students from ITN and control villages rather than headmaster refusal. In each of the schools, all students in the appropriate age range participated. The seven recruited schools were visited in October 1996, before ITNs were distributed, to obtain baseline measurements. Follow-up measurements were then obtained at three subsequent time-points: after the long rains in July 1997, just before the subsequent long rains in February 1998, and at the end of the study in October 1998. At each time point, measurements were obtained at all seven schools within a 3-week time frame. In addition, each school was visited a second time at each time point within 2 weeks of the initial measurement to capture any children who were absent at the time of the first visit. With the assistance of teachers, children provided their village of residence and date of birth. Each child was also provided with a form on which a parent recorded the child’s date of birth and the house location code number, which had been written on the door of his/her home during the demographic survey (PhillipsHoward et al, 2003a). This code number indicated the village, compound, and house number. The child’s ITN randomization status, defined on an intention to treat basis (village of residence), and date of birth were determined based upon these data, or by that provided by the child and teacher if the form was not returned. In total, 13 subjects of 1093 initially recruited were over the age of 13.5 y (range 13.6 to 18.3 y) and were excluded from analyses given insufficient numbers in yearly age categories to make reliable estimates of means. This left a cohort of 1080 children between the ages of 4.5 and 13.5 y.

Nutritional assessment The height and weight of children were measured with them wearing light clothing and no shoes, according to procedures described by Jelliffe, (1966). The child’s height was recorded to 0.1 cm using a wooden height board with a sliding headpiece parallel to the base. The same height board, which was constructed by a local carpenter according to United Nations specifications (1986) was used throughout the study. Weight was measured to 0.1 kg on a Seca model 770 scale (Seca Inc., Columbia, MD, USA). Mid-upper arm circumference (MUAC), a composite measure of upper arm muscle and subcutaneous fat reserves, was measured to 0.1 cm at the midpoint of the left upper arm, between the acromion European Journal of Clinical Nutrition

process and the tip of the olecranon, using a nonstretch ‘‘Zerfuss’’ insertion tape (Ross, USA) (Gibson, 1990). Triceps skinfold thickness, a measure of subcutaneous fat stores, was measured to 0.2 mm at the same location using Holtan skinfold calipers (Crymych, UK) (Gibson, 1990). Triceps skinfold measurements were made in duplicate by one of two trained field staff who performed the triceps skinfold measure throughout the study. The mean of these two measurements was used in all statistical analyses. HAZ, weight-for-height (WHZ), and WAZ Z-scores were calculated from Center for Disease Control (National Center for Health Statistics)/World Health Organization (1977/ 1985) reference values using EpiInfo 2002 software. Body mass index (BMI), which is the weight in kilograms divided by height in square meters, was also calculated. EpiInfo 2000 software and Centers for Disease Control (2000) reference data were used to calculate BMI Z-scores (BMIZ). BMIZ was the primary weight-for-stature measure used, as EpiInfo will not calculate WHZ for girls 410 y of age and boys 411.5 y of age due to variability in this index based on pubertal status. Stunting, wasting and underweight were defined as HAZ, BMIZ, and WAZo2 s.d. from the reference median, respectively. In addition, two measures of body composition, upper arm fat area and upper arm muscle area, were derived from MUAC and triceps skinfold thickness as described by Frisancho (Frisancho, 1981). Upper arm muscle area provides a summary measure of protein and bone or ‘lean body mass’ that is an indicator of protein reserves (Zemel et al, 1997; Frisancho, 1981). Upper arm fat and muscle area Z-scores were calculated based on age and gender specific means and standard deviations from the National Health and Nutrition Examination Surveys I and II (Frisancho, 1990). Estimates of percent body fat and lean body mass were also determined based on percent of upper arm attributable to fat and muscle, respectively (Frisancho, 1990).

Confounding/mediating covariates Age was calculated from date of birth. If month and year, or only year (N ¼ 2), were available for a child’s date of birth, the 15th of the known month or the mid-year day were used, respectively. Age data were available for 878 subjects (81%). For children whose parents provided the study number from the door of the house, socioeconomic status and hygiene variables were available, including whether animals were owned, the number of cows owned, and whether there was a pit latrine (data available for 77.3% of the cohort). The effect of seasonality was also addressed using a covariate that captured whether the follow-up measurement was made during a rainy or dry season.

Informed consent Village-based barazas were held at the time of ITN distribution to discuss the project, provide information, and answer questions in the native language, Dholuo. The ITN trial was


917 approved by the institutional review boards of the Kenya Medical Research Institute (KEMRI) and the Centers for Disease Control and Prevention (CDC), Atlanta, Georgia, USA. Informed consent was obtained from all parents after explanation of the study procedures. In addition, information sessions were conducted at all seven schools explaining the details of this project.

Data analysis Most statistical analyses were performed using SAS version 8.2. (Cary, NC, USA). Yearly age categories were created such that a given age category contains children who are within 6 months of the specified age, for example age category five contains children aged 4.5–5.5 y of age. If age categories did not contain at least six observations, means were not depicted graphically. Bivariate analyses were performed to assess baseline nutritional differences between the girls and boys and to identify covariates for inclusion in multivariate models, using SUDAAN 8.0 to adjust for clustering at the school level. For continuous covariates, which were all normally distributed, confidence intervals (CI) and significance tests (Student’s T-test) are presented, and for all other data, proportions and w2-square tests of association are reported. A mixed effects model that included data from all four time points was created with HAZ as the outcome (Singer, 1998). Covariates that were significant in bivariate analyses (Po0.10) were included for evaluation in the multivariate model. Baseline HAZ was included as a covariate, rather than in the response vector, such that the model is adjusted for the baseline measure. The model, therefore, represents change in an individual’s HAZ over time. An attempt to account for the two levels of clustering (school and village) in one model was made; however, this model did not converge. Therefore, clustering was taken into account at the school level, in accordance with study design. This was done by specifying school as a random intercept term in the model. In addition, we adjusted the standard error to account for repeated measures on children, assuming that observations closer together in time were more highly correlated than those further apart (autoregressive covariance structure). School level random intercepts were evaluated in each model. We adjusted for ITN status in all multivariate analyses. Model diagnostics were performed to assess collinearity and homoskedasticity. The least squares means was used to produce population estimates of mean HAZ at 9 and 24 months, adjusting for baseline upper arm muscle area, baseline HAZ, and other covariates. In multivariate models, covariates that were significant at Po0.05 were considered statistically significant.

Results At the baseline measurement, 1080 children were eligible to participate. The following numbers of subjects were available for measurement at the three follow-up time points: 789

(73%) after the long rains in July 1997, 724 (67%) just before the subsequent long rains in February 1998, and 605 (56%) at the end of the study in October 1998. A total of 835 subjects had complete data for use in multivariate models (77.3%). Using the last measurement time point to define lost to followup, there were no statistically significant baseline differences with respect to age or nutritional status (HAZ, WAZ, or BMIZ) between those who were followed for the whole study and those who were ultimately lost to followup. There were gender differences, however, with girls more likely than boys to be lost to follow-up (48.4 vs 39.2% lost to follow-up respectively; w2- Po0.003). Baseline characteristics of the cohort are shown in Table 1, pooled and split by gender and age group. At study inception, the children were 4.9–13.5 y of age (mean (95% CI) of 9.1 [8.9, 9.2] y) and 52.0% were female. The observed standard deviations of the Z-scores for HAZ and WAZ were 1.24 and 0.90, respectively, falling within the WHO recommended ranges and suggesting the quality of the data was good (WHO, 1995). Relative to a healthy reference population, the cohort was malnourished at baseline with mean Z-scores for nutritional outcomes around one standard deviation below the median. Overall, 18.3, 16.0, and 17.1% were stunted, undernourished, and wasted (HAZ, WAZ, and BMIZo2 s.d. from the NCHS median), respectively. A greater proportion of boys than girls were stunted (21.6 vs 15.3%; P ¼ 0.02). Cross-sectional prevalence of stunting and wasting by gender and age category are shown in Figure 1. In bivariate analyses, baseline HAZ was inversely related to age, with yearly differences in HAZ of 0.29 s.d. (Po0.0004) with each increase in age category. Baseline HAZ was higher for girls than boys (adjusted Z-score difference 0.19 s.d.; P ¼ 0.10). Changes from baseline results, examining the intraindividual progression of stunting over 9 and 24 months, are presented in Figure 2. Similar trends were observed at 16 months and this time point is omitted for clarity. The mean intra-individual change in HAZ from baseline for the whole cohort was 0.16 (0.19, 0.13) at 9 months, 0.18 (0.22, 0.15) at 16 months, and 0.36 (0.41, 0.31) at 24 months. Thus between baseline and 9 months and baseline and 24 months, the cohort deviated further from the NCHS reference median. In addition, the prevalence of stunting was 18.3, 20.3, 19.7, and 26.1% at time points 1, 2, 3, and 4, respectively. Longitudinal multivariate models were made to investigate baseline nutritional and body compositional predictors of the progression of stunting over time. Models included baseline HAZ, gender, age, time, ITN status, and a single baseline nutritional measure. The baseline measure that explained the greatest amount of variance in the progression of stunting was the upper arm muscle area Z-score (F ¼ 8.1; P ¼ 0.005). The final longitudinal model containing all covariates significantly related to the progression of stunting is shown in Table 2. Baseline upper arm muscle area Z-score was significantly and positively associated with HAZ at European Journal of Clinical Nutrition

918

European Journal of Clinical Nutrition

Table 1 Baseline characteristics of study participants by sex adjusted for clustering by school Pooled (N ¼ 1080) Demographic/socio-economic status Age (months)a Animals owned by compound Pit latrine on compound Child sleeps under bednet

108.5 318/835 575/835 437/835

a

Girls (N ¼ 562)

Ages 4.5–10 years N ¼ 606 Ages 10–13.5 years N ¼ 272

(106.7, 110.4) 109.4 (105.7, 113.0) 107.7 (105.1, 110.2) (38.1) 155/402 (38.6) 163/433 (37.6) (68.9) 280/402 (69.7) 295/433 (68.1) (52.3) 209/402 (52.0) 228/433 (52.7)

NS NS NS NS

98.2 206/549 383/549 292/549

(1.06, 0.83) 1.04 (1.28, 0.81) 0.85 (0.92, 0.77) (18.3) 92/427 (21.6) 68/448 (15.2) (1.30, 0.97) 1.18 (1.41, 0.96) 1.08 (1.22, 0.93) (16.0) 73/427 (17.1) 66/445 (14.8) (14.4, 15.2) 14.8 (14.5, 15.2) 14.7 (14.3, 15.1) (1.37, 0.98) 1.21 (1.52, 0.91) 1.14 (1.27, 1.00) (17.1) 75/420 (17.9) 71/436 (16.3) (17.1, 17.5) 17.1 (16.9, 17.4) 17.5 (17.3, 17.7) (1.32, 1.10) 1.23 (1.38, 1.08) 1.19 (1.28, 1.10)

0.10 0.02 NS NS 0.07 NS NS 0.03 NS

0.67 79/605 0.94 66/600 14.6 1.01 37/553 17.0 1.23

(1.15, 0.98)

0.92 (0.98, 0.86)

1.07 (1.15, 0.98) 0.0001

Mean (95% confidence interval); P-value from difference in means adjusted for clustering by school. Number yes/denominator for covariate (%). P-value from Fisher’s Exact.

b

P

(96.7, 99.6) (37.5) (70.0) (53.2)

(0.83, 0.51) (13.0) (1.15, 0.73) (11.0) (14.3, 15.0) (1.81, 1.26) (6.3) (16.8, 17.2) (1.40, 1.07)

0.95 (1.16, 1.02)

131.5 102/251 176/251 136/251

1.56 81/270 1.55 73/272 15.0 1.54 5/129 18.0 1.18

(130.3, 132.7) (40.6) (70.1) (54.2)

P o0.0001 NS NS NS

(1.73, 1.39) 0.0002 (30.0) o0.0001 (1.69, 1.41) 0.0003 (26.8) o0.0001 (14.6, 15.4) 0.007 (1.81, 1.26) 0.001 (3.9) NS (17.8, 18.3) o0.0001 (1.33, 1.03) NS

1.09 (1.16, 1,02)

0.004


Nutritional status Height-for-age Z-score a 0.94 Stunted (HAZ o 2.0 s.d.)b 160/875 Weight-for-age Z-score a 1.13 Undernourished (WAZ o 2.0 s.d.)b 139/872 2 a Body mass index (Kg/m ) 14.8 Body mass index Z-score a 1.17 Wasted (BMI Z-score o 2.0 s.d.)b 146/856 Mid-upper arm circumference (cm)a 17.3 Mid-upper arm circumference for 1.21 height Z-score a 1.067 Triceps skinfold Z-score a

Boys (N ¼ 518)

Figure 1 (a) Prevalence of stunting by age and gender. (b) Prevalence of wasting by age and gender. Each data point represents between 17 and 102 observations, with the exception of girls of age category 5 for stunting and wasting (each 10 observations) and boys of age category 5 for stunting and wasting (each six observations).

follow up (b ¼ 0.055; P ¼ 0.04). Baseline age was significantly related to progression of stunting, with older children experiencing sharper declines in HAZ (b ¼ 0.005; Po0.0001) over time. An interaction term was added to


919

Figure 2 Longitudinal mean change in HAZ from baseline to 9 and 24 months. Each data point represents between 12 and 83 observations, with the exception of age category 12 at 24 months, which represents eight observations. Note: Similar trends were observed at 16 months and this time point is omitted for clarity.

Table 2 Longitudinal predictors of Height-for-age Z-score (HAZ) over timea Covariate

Beta (95% confidence interval) b

0.84 HAZ at baseline (s.d. ) Time (months) 0.007 Sex (male ¼ 1, female ¼ 2) 0.05 Upper arm muscle area Z-score (s.d.) 0.055 Age (months) 0.005

P

(0.81, 0.87) o0.0001 (0.01, 0.004) o0.0001 (0.10, 0.004) 0.08 (0.004, 0.106) 0.04 (0.006, 0.003) o0.0001

a Hierarchical model adjusting for clustering by school and repeated measures for students. Model also includes a covariate adjusting for ITN status. b Standard deviation.

the model to investigate whether the relationship between age and HAZ was the same for older and younger children (age cutoff o 9 y at baseline). The interaction term was not statistically significant, suggesting that the relationship between age and progression of stunting was not limited to older children, who might experience the onset of puberty and its associated growth spurt later than the populations on whom the reference curves are based. SES was not associated with the progression of stunting. Female gender conferred a slightly higher risk of progression of stunting; however, this did not reach statistical significance (b ¼ 0.05; P ¼ 0.08). Finally, the season of follow-up measurement (rainy vs dry), did not have a statistically significant impact on HAZ. The mean HAZ at follow up adjusted for baseline HAZ, age, sex, and ITN status split by children with baseline upper arm muscle area Z-scores less than or greater than the 25%ile and 75%ile of the distribution, respectively, is shown in Figure 3. Similar trends were observed at 16 months and this time

Figure 3 Mean adjusted HAZ Z-score at follow up time points split by baseline upper arm muscle area Z-score. HAZ 795% confidence intervals adjusted for baseline HAZ age, sex, and ITN status. Note: Similar trends were observed at 16 months and this time point is omitted for clarity.

point is omitted for clarity. Change in HAZ was statistically significantly greater among those with higher vs lower upper arm muscle area Z-scores at baseline (HAZ 1.044 vs 1.917 at 9 months; Po0.001 and HAZ 1.114 vs 1.301 at 24 months; P ¼ 0.04).

Discussion Results from this work confirm findings from many other studies addressing age and gender patterns of undernutrition in Sub-Saharan Africa. It provides further evidence from a distinct ecological setting regarding the persistent problem of undernutrition during middle childhood in the developing world. The longitudinal design of this study, encompassing a relatively long period of follow up, also allows us to draw conclusions about within-individual progression of stunting during the school-age years. Finally, the longitudinal nature of this work allowed identification of baseline nutritional and body compositional factors associated with the progression of stunting. To our knowledge, this is the first longitudinal study conducted in the developing world to address baseline nutritional factors associated with the progression of stunting. European Journal of Clinical Nutrition


920 Both cross-sectional and longitudinal data from this study demonstrate the progression of height deficit accrual during middle childhood. The increasing prevalence of stunting at baseline by increasing age category suggests this. The longitudinal data, which is less susceptible to threats to validity such as cohort effects and selection bias, also suggest that children in this age group continue to accrue yearly deficits in height compared with international reference standards. The 95% confidence limits for change in HAZ from baseline to both 9 and 24 months did not include zero, suggesting that children deviate further from NCHS medians with age. Although it is believed that the greatest degree of stunting occurs during the first 2 y of life (Martorell et al, 1994), a time period of extremely rapid linear growth, data from this study and other recent studies suggest that growth stunting progresses through the school-age years (Neumann & Harrison, 1994; Stoltzfus et al, 1997; Partnership for Child Development, 1998a, b; Lwambo et al, 2000). This is not surprising given children in better environments continue to grow, though at a slower pace, throughout middle childhood. This study, together with other longitudinal studies (Neumann & Harrison, 1994; Stoltzfus et al, 1997), suggests that school-aged children in the developing world do not experience significant ‘type A’ catch-up growth (accelerated growth velocity following an insult to growth) (Boersma & Wit, 1997) or remain stable. Rather, data from this study suggests that school-aged children, who remain in the same environments, continue to deviate from NCHS standards, accruing greater height deficits with age. It is possible that these deficits are made up through Type B catch-up growth, a later onset of pubertal growth spurt, which allows a longer period of middle childhood growth. In this study, we were also able to explore nutritional and body compositional covariates related to the progression of stunting. We found that upper arm muscle area Z-score, an age and gender adjusted measure of absolute lean body mass, explained the most variability in the progression of stunting, whereas percent lean body mass and measures of absolute body fat were not significantly related to the progression of stunting. Similarly, cross-sectional studies conducted in Central America found that children with greater lean body mass as measured by upper arm muscle diameter are significantly taller than those with less lean body mass, whereas fat mass is not related to stature (Frisancho & Garn, 1971; Frisancho et al, 1971). More recent studies of children with chronic diseases suggest that deficits in the fat free mass compartment, rather than the fat compartment, are related to growth failure in children with HIV (Arpadi et al, 1998, 2000), and treatment-related improvements in linear growth in renal failure are related to increases in fat free mass and actual decreases in total fat (Johnson et al, 2000). These findings as well as our study highlight the importance of the status of specific body compartments, rather than summary measures of body size, in relation to linear growth. Further, the longitudinal nature of our study allows the establishEuropean Journal of Clinical Nutrition

ment of a temporal sequence of events that does not suffer from threats to validity such as reverse causality, which may make inferences in the aforementioned cross-sectional studies more difficult. Identification of specific nutritional risk factors for the progression of stunting will help to more accurately identify children at risk and may suggest possible mechanisms as discussed below. Chronic undernutrition, as evidenced by a high proportion of stunted children, was prevalent in this cohort (18.3% at baseline). Wasting, as assessed by BMIZ, was also prevalent (17.1%). Of note, wasting as determined by WHZ, as compared with BMIZ, was much less prevalent (6.0% at baseline), as has been demonstrated in other studies in the developing world (Stoltzfus et al, 1997; Partnership for Child Development, 1998a, b; Lwambo et al, 2000). Studies in SubSaharan Africa suggest that the prevalence of undernutrition may be greater in children not enrolled in school (Fentiman & Hall, 1997; Beasley et al, 2000); therefore, the proportions found in this school-based sample likely underestimate the degree of undernutrition in this age group. In addition, boys in this cohort were more likely to be stunted than girls, a finding reported in several other studies (Stoltzfus et al, 1997; Partnership for Child Development, 1998a, b; Lwambo et al, 2000) In our 2 y follow-up, however, girls experienced greater height deficit accrual compared to a healthy reference population than did boys. The paucity of longitudinal data extending into early adulthood, and of descriptive studies addressing sex differences in the prevalence of stunting among adults in the developing world, leaves unclear whether these differences persist into adulthood. There are several limitations to this study that should be addressed. First, this was a school-based cohort and some findings may be related to selection bias. For example, the fact that girls were better off nutritionally, and older children experienced more significant declines in HAZ over time, may reflect the fact that children attending school who are younger and female come from families with greater resources. Families with less disposable income may choose to send only boys and older children to school. Further, our finding that girls were more likely to be lost to follow up, which is often due to missing semesters based on scarce family resources, suggests that families may be more likely to withdraw females from school and possibly less likely to initially enroll them, if resources are inadequate. Second, as our cohort did not include older volunteers completing puberty, we cannot address whether deficits accrued during middle child impact final adult height. If children experience increased growth velocity during adolescence (Type A catch-up growth), or begin their adolescent growth spurt later allowing an extended period of middle childhood growth (Type B catch-up growth), or a combination of these (Type C catch-up growth), they may recapture some of the height deficits lost during early and middle childhood (Boersma & Wit, 1997). More studies are needed to address the dynamics of growth during and just after adolescence to address this question. Finally, absolute lean body mass,


921 which was related to the progression of stunting, may be a proxy for an unmeasured factor that is a more biologically relevant predictor of the progression of stunting. Unmeasured factors that may be related both to lean body mass and progression of stunting include micronutrient status (Brown et al, 2002) and host elaboration of pro-inflammatory cytokines. Proinflammatory cytokines cause specific lean body mass depletion (Li et al, 1998; Kotler, 2000) and may impede growth through both their anorexia inducing effects and direct inhibitory effects on bone growth (Bertolini et al, 1986; Skerry, 1994; Goldring & Goldring, 1996) Malaria and the helminth infections prevalent in this study area, are known to cause elaboration of proinflammatory cytokines (Kloos et al, 1982; Kwiatkowski, 1990; Kwiatkowski et al, 1990; Zwingenberger et al, 1990; Kern et al, 1992; Abdel Azim et al, 1995), and children who make relatively more of these in response to infection may have both lower lean body mass and more marked growth faltering. That children continue to experience growth faltering during middle childhood indicates that they have insufficient intake of micronutrients, macronutrients or both to meet the demands of normal growth. Parasitic diseases may contribute to growth faltering by interfering with nutrient absorption or through the elaboration of proinflammatory cytokines, which cause anorexia and may have direct inhibitory effects on bone growth. In addition to the potential long-term consequences of stunting, the same deficiencies that cause growth faltering during middle childhood may have other important functional consequences for the school-aged child, most importantly, potential effects on cognitive function and learning. Although this and other studies indicate that school-aged children represent a nutritionally vulnerable population, more studies are needed to elucidate the primary mechanisms leading to growth faltering, such that recommendations for cost-effective interventions can be made. Acknowledgements We thank the Centers for Disease Control and Prevention/ Kenya Medical Research Institute field teams and support staff for their assistance with this project. We thank Marlene Barber for support with data management. We thank the students, teachers and headmasters for their time and participation. We are grateful to Hemal Kanzaria for careful review of this manuscript. We also thank the director of the Kenya Medical Research Institute for his permission to publish this work.

References Abdel Azim A, Sedky HA, el-Tahawy MA, Fikry AA & Mostafa H (1995): Serum levels of tumor necrosis factor in different stages of schistosomal infection. J. Egypt Soc. Parasitol. 25, 279–287. Arpadi SM, Cuff PA, Kotler DP, Wang J, Bamji M, Lange M, Pierson RN & Matthews DE (2000): Growth velocity, fat-free mass and energy intake are inversely related to viral load in HIV-infected children. J. Nutr. 130, 2498–2502.

Arpadi SM, Horlick MN, Wang J, Cuff P, Bamji M & Kotler DP (1998): Body composition in prepubertal children with human immunodeficiency virus type 1 infection. Arch. Pediatr. Adolesc. Med. 152, 688–693. Beasley NM, Hall A, Tomkins AM, Donnelly C, Ntimbwa P, Kivuga J, Kihamia CM, Lorri W & Bundy DA (2000): The health of enrolled and non enrolled children of school age in Tanga, Tanzania. Acta. Trop. 76, 223–229. Beier JC, Oster CN, Onyango FK, Bales JD, Sherwood JA, Perkins PV, Chumo DK, Koech DV, Whitmire RE & Roberts CR et al. (1994): Plasmodium falciparum incidence relative to entomologic inoculation rates at a site proposed for testing malaria vaccines in western Kenya. Am. J. Trop. Med. Hyg. 50, 529–536. Bertolini DR, Nedwin GE, Bringman TS, Smith DD & Mundy GR (1986): Stimulation of bone resorption and inhibition of bone formation in vitro by human tumour necrosis factors. Nature 319, 516–518. Bloland PB, Boriga DA, Ruebush TK, McCormick JB, Roberts JM, Oloo AJ, Hawley W, Lal A, Nahlen B & Campbell CC (1999): Longitudinal cohort study of the epidemiology of malaria infections in an area of intense malaria transmission II. Descriptive epidemiology of malaria infection and disease among children. Am. J. Trop. Med. Hyg. 60, 641–648. Boersma B & Wit JM (1997): Catch-up growth. Endocr. Rev. 18, 646–661. Brooker S, Miguel EA, Waswa P, Namunyu R, Moulin S, Guyatt H & Bundy DA (2001): The potential of rapid screening methods for Schistosoma mansoni in western Kenya. Ann. Trop. Med. Parasitol. 95, 343–351. Brown KH, Peerson JM, Rivera J & Allen LH (2002): Effect of supplemental zinc on the growth and serum zinc concentrations of prepubertal children: a meta-analysis of randomized controlled trials. Am. J. Clin. Nutr. 75, 1062–1071. Eveleth P (1985): Nutritional implications of differences in adolescent growth and maturation and in adult body size. In: Nutritional Adaptation in Man, eds Blaxter K and Waterlow J, pp 31–43. London: John Libbey. Fentiman A & Hall A (1997): Factors Influencing School Enrollment and the Health of Children. Accra, Ghana: UNICEF. Friedman JF, Phillips-Howard PA, Hawley WA, Terlouw DJ, Kolczak MS, Barber M, Okello N, Vulule JM, Duggan C & Nahlen BL et al (2003): Impact of permethrin-treated bed nets on growth, nutritional status, and body composition of primary school children in western Kenya. Am. J. Trop. Med. Hyg. 68, 78–85. Frisancho AR (1981): New norms of upper limb fat and muscle areas for assessment of nutritional status. Am. J. Clin. Nutr. 34, 2540–2545. Frisancho AR (1990): Anthropometric Standards for the Assessment of Growth and Nutritional Status. Ann Arbor: University of Michigan Press. Frisancho AR & Garn SM (1971): Skin-fold thickness and muscle size: implications for developmental status and nutritional evaluation of children from Honduras. Am. J. Clin. Nutr. 24, 541–546. Frisancho AR, Garn SM & McCreery LD (1971): Relationship of skinfolds and muscle size to growth of children. I. Costa Rica. Am. J. Phys. Anthropol. 35, 85–90. Gibson R (1990): Principles of Nutritional Assessment. New York: Oxford University Press. Goldring SR & Goldring MB (1996): Cytokines and skeletal physiology. Clin. Orthop. 324, 13–23. Jelliffe DB (1966): The assessment of the nutritional status of the community (with special reference to field surveys in developing regions of the world). Monogr. Ser. World Health Organ. 53, 3–271. Johnson VL, Wang J, Kaskel FJ & Pierson RN (2000): Changes in body composition of children with chronic renal failure on growth hormone. Pediatr. Nephrol. 14, 695–700. Kern P, Hemmer CJ, Gallati H, Neifer S, Kremsner P, Dietrich M & Porzsolt F (1992): Soluble tumor necrosis factor receptors correlate



922 with parasitemia and disease severity in human malaria. J. Infect. Dis. 166, 930–934. Kloos H, Sidrak W, Michael AA, Mohareb EW & Higashi GI (1982): Disease concepts and treatment practices relating to schistosomiasis haematobium in upper Egypt. J. Trop. Med. Hyg. 85, 99–107. Kotler DP (2000): Cachexia. Ann. Intern. Med. 133, 622–634. Kwena AM, Terlouw DJ, de Vlas SJ, Phillips-Howard PA, Hawley WA, Friedman JF, Vulule JM, Nahlen BL, Sauerwein RW & ter Kuile FO (2003): Prevalence and severity of malnutrition in pre-school children in a rural area of western Kenya. Am. J. Trop. Med. Hyg. 68, 94–99. Kwiatkowski D (1990): Tumour necrosis factor, fever and fatality in falciparum malaria. Immunol. Lett. 25, 213–216. Kwiatkowski D, Hill AV, Sambou I, Twumasi P, Castracane J, Manogue KR, Cerami A, Brewster DR & Greenwood BM (1990): TNF concentration in fatal cerebral, non-fatal cerebral, and uncomplicated Plasmodium falciparum malaria. Lancet 336, 1201–1204. Li YP, Schwartz RJ, Waddell ID, Holloway BR & Reid MB (1998): Skeletal muscle myocytes undergo protein loss and reactive oxygen-mediated NF-kappaB activation in response to tumor necrosis factor alpha. FASEB. J. 12, 871–880. Lwambo NJ, Brooker S, Siza JE, Bundy DA & Guyatt H (2000): Age patterns in stunting and anaemia in African schoolchildren: a cross-sectional study in Tanzania. Eur. J. Clin. Nutr. 54, 36–40. Martorell R & Arroyave G (1988): Malnutrition, work output, and energy needs. In: Capacity for work in the Tropics eds Collins K and Roberts D Cambridge: Cambridge University Press. Martorell R, Delgado HL, Valverde V & Klein RE (1981): Maternal stature, fertility and infant mortality. Hum. Biol. 53, 303–312. Martorell R, Khan LK & Schroeder DG (1994): Reversibility of stunting: epidemiological findings in children from developing countries. Eur. J. Clin. Nutr. 48 (Suppl 1), S45–57. Martorell R, Schroeder DG, Rivera JA & Kaplowitz HJ (1995): Patterns of linear growth in rural Guatemalan adolescents and children. J. Nutr. 125, 1060S–1067S. Monyeki KD, Cameron N & Getz B (2000): Growth and nutritional status of rural South African children 3–10 years old: The Ellisras growth study. Am. J. Human Biol. 12, 42–49. Musgrove P (1993): Investing in health: the 1993 World Development Report of the World Bank. Bull Pan. Am. Health Organ. 27, 284–286. Neumann CG & Harrison GG (1994): Onset and evolution of stunting in infants and children. Examples from the Human Nutrition Collaborative Research Support Program. Kenya and Egypt studies. Eur. J. Clin. Nutr. 48 (Suppl 1), S90–102. Norgan NG (2000): Long-term physiological and economic consequences of growth retardation in children and adolescents. Proc. Nutr. Soc. 59, 245–256.


Olds GR, King C, Hewlett J, Olveda R, Wu G, Ouma J, Peters P, McGarvey S, Odhiambo O & Koech D et al (1999): Double-blind placebo-controlled study of concurrent administration of albendazole and praziquantel in schoolchildren with schistosomiasis and geohelminths. J. Infect. Dis. 179, 996–1003. Olsen A (1998): The proportion of helminth infections in a community in western Kenya which would be treated by mass chemotherapy of schoolchildren. Trans. R. Soc. Trop. Med. Hyg. 92, 144–148. Partnership for Child Development (1998a): The anthropometric status of schoolchildren in five countries in the partnership for child development. Proc. Nutr. Soc. 57, 149–158. Partnership for Child Development (1998b): The health and nutritional status of schoolchildren in Africa: evidence from school-based health programmes in Ghana and Tanzania. Trans. R. Soc. Trop. Med. Hyg. 92, 254–261. Phillips-Howard PA, Nahlen BL, Alaii JA, ter Kuile FO, Gimnig JE, Terlouw DJ, Kachur SP, Hightower AW, Lal AA & Schoute E et al (2003a): The efficacy of permethrin-treated bed nets on child mortality and morbidity in western Kenya I. Development of infrastructure and description of study site. Am. J. Trop. Med. Hyg. 68, 3–9. Phillips-Howard PA, ter Kuile FO, Nahlen BL, Alaii JA, Gimnig JE, Kolczak MS, Terlouw DJ, Kariuki SK, Shi YP & Kachur SP et al (2003b): The efficacy of permethrin-treated bed nets on child mortality and morbidity in western Kenya II. Study design and methods. Am. J. Trop. Med. Hyg. 68, 10–15. Singer JD (1998): Using SAS PROC MIXED to fit multilievel models, hierarchical models, and individual growth models. J. Edu. Behav. Stat. 24, 323–355. Skerry TM (1994): The effects of the inflammatory response on bone growth. Eur. J. Clin. Nutr. 48 (Suppl 1), S190–7. discussion S198. Stoltzfus RJ, Albonico M, Tielsch JM, Chwaya HM & Savioli L (1997): Linear growth retardation in Zanzibari school children. J. Nutr. 127, 1099–1105. Tanner JM (1990): Fetus into Man. Cambridge: Harvard University Press. United Nations specifications (1986): Assessing the nutritional status of young children in household surveys. United Nations, New York. World Health Organization (1995): Physical status: the use of and interpretation of anthropometry. Geneva. Zemel BS, Riley EM & Stallings VA (1997): Evaluation of methodology for nutritional assessment in children: anthropometry, body composition, and energy expenditure. Annu. Rev. Nutr. 17, 211–235. Zwingenberger K, Irschick E, Vergetti Siqueira JG, Correia Dacal AR & Feldmeier H (1990): Tumour necrosis factor in hepatosplenic schistosomiasis. Scand. J. Immunol. 31, 205–211.