Prevention-interventions would certainly benefit from a precise knowledge of the age range when the most pronounced increases in prevalence of overweight and obesity occur in the general population. Data of 15,662 subjects aged 2–18 years were obtained from a national representative health survey (German Health Interview and Examination Survey for Children and Adolescents (KiGGS)) conducted in Germany. Weight, height, and BMI z-scores were calculated relative to the UK 1990 reference, and prevalence of overweight and obesity was defined according to the IOTF (International Obesity Task Force) age- and sex-specific cut-offs. Univariate ANOVAs for overweight, obesity, weight, height, and BMI z-scores as dependent variables were employed to assess significant differences for these measures across various age levels. Significant analysis was followed by post-hoc comparisons using Bonferroni adjustments. The main effect of age was estimated using a multinomial logistic regression model, and by defining the first derivative of a polynomial spline function. Different eclectic slopes over the entire age range from 2 to 18 years have been observed. Prevalence of overweight substantially increases between the 5th and the 8th year (12.5–21.4% P ≤ 0.001). Maximum increase of the polynomial fit was detected at 7.2 years. Our findings suggest a relatively narrow age range at the first school year when overweight in German children especially increases. We therefore propose that psychosocial correlates may be related to the general life-time event around the age of entering school.
The prevalence of overweight and obesity in children has dramatically increased during the last decades and thus is of public health concern (1,2,3). The etiology of overweight and obesity is a complex mixture of genetic, environmental, and psychosocial influences (4,5). However, less is known about the reasons why prevalence rates increased in recent years (3). A first step to answer this question would require a precise determination at which age range in children's life the relevant increase in the prevalence of overweight and obesity occurs. So far, epidemiological studies do not specify in detail whether this increase occurs at a particular age level (1,2,3,5,6,7,8). In addition, these studies showed contradicting results regarding the prevalence of overweight and obesity, because they divided their study cohorts into different age groups by chance or prima vista or based on other unknown criteria. Comparisons between these studies will be difficult and potentially misleading. Arbitrary age classifications are accompanied by methodological problems, including biased observations that could probably influence the study outcomes. Hence, it is necessary to evaluate the data in a way that excludes such bias by haphazardly choosing age groups in order to obtain a more reliable evidence base.
The aim of the current study was to estimate a precise point in children's age when the considerable increase in prevalence of overweight and obesity emerges. We therefore analyzed data of a nationally representative survey in a more comprehensive and continuous way in order to avoid corresponding distortions and to obtain a realistic estimate of the maximum increase in the prevalence rates. We will discuss how this refined data analysis may improve our general understanding of the incidence of childhood overweight and obesity, and potentially helps us to set more cost-effective prevention-interventions.
Participants and Methods
The German Health Interview and Examination Survey for Children and Adolescents (KiGGS, 2003–2006) is the first nationally representative cross-sectional study conducted in Germany. In total, the study includes 17,641 participants (8,985 boys and 8,656 girls) aged 0–18 years from 167 sample points with standardized measurements of height and weight. Detailed descriptions of KiGGS have been previously published (9,10).
This study examined data of the public-use-file (11). BMI was calculated and prevalence of overweight and obesity was defined according to the internationally proposed age- and sex-specific cut-offs (12). Due to methodological difficulties by classifying overweight and obesity in children under 2 years of age, we finally included 15,662 children and adolescents (7,998 boys and 7,664 girls) aged 2–18 years. Age was grouped in bins of half-year steps, and weight, height, and BMI z-scores were calculated relative to the UK 1990 reference (13).
Statistical and mathematical analyses
Statistical analyses were performed using SPSS 17.0 (SPSS, Chicago, IL) and mathematical operations were conducted using MATLAB 7.11.0 (Mathworks, Natick, MA). Data are presented as percentages, means (s.d.), and relative risks (odds ratio; 95% confidence interval (CI)). An α-level of 0.05 was used for all statistical tests. Descriptive analyses and χ2 tests were used to characterize the distribution of the key variables and to estimate the goodness-of-fit.
Separate univariate ANOVAs with post-hoc comparisons (Bonferroni) were computed for the dependent variables. The main effect of age, adjusted for confounding variables (migration status, gender, parental BMI, socioeconomic status), was estimated by using a multinomial logistic regression model. We further fitted a polynomial spline function to the smoothed prevalence data and the first derivative of this function was calculated. The age of year when this derivative was maximal was then determined.
The prevalence of overweight and obesity was 18.3 and 4.8%, respectively. Table 1 shows the prevalence (%) of overweight and means (s.d.) of weight, height, and BMI z-scores between 2 and 18 years. ANOVAs for overweight, weight, and BMI z-scores show statistically significant age effects (P ≤ 0.001). Post-hoc analysis for the dependent variables showed a high-ranked increase in prevalence of overweight, and in weight, and BMI z-scores between 5.0 and 8.5 years (P ≤ 0.01). Basically, prevalence of overweight significantly increased with age from 7.8% at age 2 to a relatively constant rate about 23% at age 8.5 and was seemingly stable thereafter (P ≤ 0.001). The highest significant increase in the prevalence of overweight has been observed between 4.5 and 8.5 years (10.3–22.8%, P ≤ 0.001). The outcome of this analysis did not indicate significant effects for obesity and thus only the results for overweight are described (data not shown). Importantly, increase in weight is not accompanied by a similar increase in height, indicating that excessive weight gain during this critical age period is not based on a premature physical development in the KiGGS cohort compared to the IOTF (International Obesity Task Force) reference.
Table 1. Mean (s.d.), weight, height, BMI z-scores, and prevalence (%) of overweight of the entire KiGGS cohort (N = 15,662)
Weight and BMI z-score did not increase significantly for all consecutive age steps, but increased substantially from 2 to 12 years. Significant mean differences in weight and BMI z-scores for certain age periods were observed (data not shown). In particular, highly significant differences in mean BMI z-scores were found between the 5th (z-score: −0.08, 95% CI: −0.19 to 0.01) and the 7th year (0.24, 0.13–0.33, P for difference 0.007).
These data were confirmed when the German reference values according to Kromeyer—Hauschild instead of the IOTF reference were employed (data not shown).
The multinomial regression model did not indicate a significant gender-effect (odds ratio = 1.00, 95% CI: 0.92–1.11). Influence of parental BMI was highly significant (1.10, CI: 1.09–1.12). Socioeconomic status and migration status also contributed significantly to risk of overweight (1.13, CI: 1.03–1.26; 1.23, CI: 1.08–1.40). Analysis stratified by age predicted the highest risk of being overweight at the age of 7 years (2.54, CI: 0.36–17.7).
Figure 1 summarizes the mathematical operations based on fitting a polynomial to the smoothed continuous raw data with the maximum slope of this fitted function including the bootstrap sampling distribution of the age estimate associated with this maximum (Figure 1a,b). The maximum slope of the first derivative was observed at the age of 7.2 years (Figure 1c; vertical line).
The purpose of this study was to explore in detail the prevalence of childhood overweight and obesity. We examined weight, height, BMI z-scores, and prevalence of overweight to determine a critical period of age in the development of overweight among German children and adolescents. Results obtained from the entire KiGGS cohort show highly significant increases in weight as well as in BMI z-scores. Most crucially, the prevalence of overweight predominantly appeared between 5.0 and 8.5 years. However, the present study confirmed a similar increase in the BMI z-scores between 6.0 and 9.5 years (0.22; P ≤ 0.001; data not shown) in direct comparison to the study of Hughes et al. They discussed critically the way of analyzing epidemiological survey data to itemize the increase rates of overweight and obesity, although no continuous analysis for age was employed (14). In principle, our findings are consistent with those of Schaffrath-Rosario et al. who described an asymmetric upward shift of the BMI distribution starting at about 6 years of age (10). However, our results indicate that, in Germany, at the age of 7.2 years there may be the highest increase in the prevalence of childhood overweight.
The strength of our study is that we reduced the observation bias to a minimum by reducing the age groups to a more comprehensive and continuous way; and second, by analyzing data with our method, we possibly could detect the most appropriate time to set effective future prevention-interventions, that is during the kindergarten years. Therefore, our analysis makes a crucial contribution to the implementation of targeted prevention-interventions in both, behavioral and situational prevention and the outcome of this analysis might be helpful in the view of standardization, sustainability, and cost effectiveness (15). However, the major limitation of our study is its cross-sectional design, which does not enable us to determine a causal relationship between the described influencing factors and the prevalence of overweight in children (12,16). Therefore, variables such as socioeconomic status and migration status that did not show a significant association with overweight in our analysis. Such an association, however, has to be assessed in a longitudinal design to clarify their role for the development of overweight. Nevertheless, longitudinal studies will benefit from the knowledge when the most pronounced increase of overweight occurs as it was revealed in the present study in a refined analysis.
In addition to the identified early-life determinants of childhood overweight and obesity (e.g., parental overweight, maternal smoking, rapid infant growth, and lack of daily physical activity), the roles of psychosocial stressors have received increasing recognition in research in recent years (4,5,17,18,19). Several studies have presented conflicting evidence as to whether psychological stress and early stressful life events plays an important role in the development of childhood overweight and obesity (5,18,19). We therefore hypothesize that school entry in Germany, as an early stressful life event, may be responsible for the rapid increase in prevalence of overweight. We probably found evidence that the increase in prevalence of overweight in Germany can be refined to an age around the stressful life event “entering school” that is independent of socioeconomic status. In Germany, children typically start school between 5.5 and 6.5 years of age and it seems reasonable to assume that there may be a connection between leaving kindergarten, starting school, and the development of overweight.
We conclude that the increase in the prevalence of overweight in Germany can be linked to an obviously distinguished age around 7.2 years and may be associated with the general life-time event of “entering school”. Hopefully, we can provide a novel direction for effective prevention-interventions that will focus on the time from birth up to the seemingly important age range between 5 and 8 years. May be these findings can be generalized to other countries with similar structures in education.
The dataset was generated by the Robert Koch Institute and the KiGGS study was funded by the German Ministry of Health, the Ministry of Education and Research, and the Robert Koch Institute.