Overall, a total of 1431 adolescents were followed for three rounds that are one year apart spanning a total follow up period of two years starting 2005. Nearly half (49.2%) were females and the rest (50.8%) were males at baseline. Analysis of socio-demographic characteristics by food security status at baseline is presented in Table . It was observed the proportion of food insecure adolescents was significantly high among urban adolescents (23.5%) compared to (20.2%) in the semi-urban areas and 17.9% in the rural areas (P=0.028). Similarly, the proportion of food insecure adolescents was 23% and 21.8%, respectively for low and middle income tertiles compared to 16.8% for high income tertile(P<0.001).
Association between food security and socio-demographic variables at baseline
With regard to exposure of adolescents to food insecurity over time, 15.9% of the girls and 12.2% of the boys (P=0.018) were food insecure both at baseline and year 1 survey, while 5.5% of the girls and 4.4% of the boys (P=0.331) were food insecure in all the three rounds of the survey. In general, a significantly (P=0.045) higher proportion (40%) of girls experienced food insecurity at least in one of the survey rounds compared with boys (36.6%), Table .
Baseline characteristics and repeated measures adolescents who were followed for all the three rounds of survey by sex
As presented in Table , analysis of the trend of food insecurity over the follow period showed doubling of the proportion of adolescents with food insecurity from the baseline (20.5% to 48.4%) on the year 1 survey, which decreased to 27.1% during the year 2 survey.
Trend of food insecurity and proportion of adolescents with height for age z-scores below −1 by food security and round of follow up
Figure shows the average baseline heights of adolescent boys by age and baseline food security status compared to the WHO references for their age. Food insecure boys had their mean height on average 2.22 cm below the heights corresponding to –1SD of the WHO reference for their age, while food secure boys had mean height on average 0.32 cm below the heights corresponding to -1SD of the WHO reference. Similarly, as shown in Figure , the mean baseline heights of food insecure girls was on average 1.16 cm below the heights corresponding to -1SD of the WHO reference, while that of food secure girls was 1.26 cm above the heights corresponding to -1SD of the WHO reference.
Mean baseline height of adolescents boys by age and baseline food security status compared to the WHO reference.
Mean baseline height of adolescent girls by age and baseline food security status compared to the WHO reference.
The average growth of adolescents over the three follow-up surveys is shown in Figure . Both food insecure boys and girls had shorter height than their food secure counterparts at all three measurements. In both boys and girls, the difference between food secure and food insecure decreased over the years although it did not disappear fully.
Growths of adolescents over the two years follow up period by food security status and sex.
As presented in Table , linear mixed effects model showed that food insecurity was negatively associated with height of girls baseline (ß= −0.87, P<0.0001). At baseline, food insecure girls were on average shorter by 0.87 centimeters (cms) compared with food secure girls after controlling for demographic and economic covariates. However, over the follow up period, food insecure girls grew on average by 0.38 cms more per year than food secure girls (ß =0.38, P=0.0663). Other factors positively associated with growth among girls were time (follow up year) and age at baseline. On average, when girls differed with an age of 1 year at baseline, their mean heights differed by 3.57 cms at baseline (ß =3.57, P<0.001). The height of girls increased by 3.44 cms for a unit increase in the follow up year (ß =3.44, P<0.001).
Parameter estimates and standard errors from linear mixed effects model predicting linear growth (height increase) of girls over the follow up period
While place of residence did not have any effect on the baseline heights of girls, there was a significant difference in their growth by place of residence over the follow up period.
Compared to urban girls, the growth rate of rural girls was greater by 1.18 cms for a unit increase in the follow up year (P<0.0001), while, those in the semi-urban areas grew by 0.92 cms more for a unit increase in the follow up year compared to urban girls (P=0.0015), after adjusting for all covariates. Baseline height for age z-score was also positively associated with growth in girls (P<0.0001). For a unit increase in baseline height for age z-score, the height of girls increased by 7.39 cms on average.
Results of linear mixed effects model for boys (Table ) showed that on average when boys differed with an age of 1 year at baseline, their mean heights differed by 3.95 cms (ß=3.95, P<0.001). Food insecurity did not have a significant effect on the height of boys both baseline (ß=−0.34, P=0.3210) and over the follow up period (ß =0.14, P=0.6249). At baseline, food insecure boys were on average shorter by 0.34 centimeters (cms) after controlling for all other covariates. However, over the follow up period, food insecure boys grew by 0.14 cms more per year than food secure boys.
Parameter estimates and standard errors from linear mixed effects model predicting linear growth (height increase) of boys over the follow up period
Other factors that were positively associated with growth among boys were time (follow up year) and baseline height for age z-score. Over the follow up years, the height of boys increased by 1.37 cms for a unit increase in the follow up year (ß =1.37, P<0.0001). Similarly, for a unit increase in the height for age z-score at baseline, height of boys increased by 6.66 cms (ß =6.66, p<0.001). As the mean growth plot of boys over the follow up period did not follow a linear pattern, we used a quadratic time effect in the model which was positively associated with growth (ß =1.97, P<0.0001). Introduction of the quadratic time effect into the model improved the model fit significantly (P<0.001) based on the likelihood ratio test.
Similar to the pattern observed in girls, boys had also different trajectory of growth by residence. Over the follow up period, for a unit increase in the follow up year, rural boys grew by 1.16 cms per year than their peers in the urban areas (P<0.001), while semi-urban boys grew 0.97 cms taller per year than their urban peers (P=0.004).