In this population-based study using longitudinal data, the prevalence of overweight/obesity was 11.5–13.7 % in childhood, and increased to 20.1 % in adolescence. These rates in childhood are comparable with rates reported in other studies from Norway and Nordic countries [6, 12, 22, 35], but lower than rates from Norwegian children born after the year 2000 [4, 23], rates reported from Southern European countries [2], and the United States [36]. For adolescents, the rates were almost twice as high as rates from Western-Norway [35], but approximately the same as those in Mid-Norway and other Nordic countries [12, 22, 37] and lower than rates from Southern European countries [2]. The difference in prevalence rates in Norway may be due to differences in prevalence between birth cohorts. Children in our cohort were born in 1992–1994. Cohorts with children born after 2000 have reported higher prevalence rates [23]. The possibility of selection bias in our study, or the study from Western-Norway [35, 38] cannot be ruled out, also as discussed in their paper. Participation proportion in the comparable study was 45 % among adolescents in the upper-secondary school group. The present study included 55 % of adolescents under 18 years in Fit Futures 1. Other suggested explanations have been related to differences between urban and rural areas with higher prevalence in more rural areas [39]. Tromsø is the largest city in North-Norway, however the municipality consists of both urban and more rural areas. Differences in socio-economic status, such as parental educational level and income are other explanations [5, 35, 39]. Unfortunately we lack individual information on socio-demographic factors.
The majority of children remained thin/normal weight between childhood and adolescence. However, we found a moderate indication of tracking of overweight/obesity from childhood to adolescence, as well as from 2–4 years of age to 5–7 years of age. Results from the different tracking analyses were consistent. The strength of the association was strongest between 2–4 and 5–7 years of age (mean time interval 3.4 years). Tracking coefficients were of similar magnitude between 5 and 7 years of age and adolescence (mean time interval 10.6 years) and may be considered a stronger indication of tracking because the risk factor was stable over a longer time interval. Glavin et al. found very high OR for being overweight or obese at 8 years of age among children who were overweight or obese at 4 or 6 years of age (63.8, 95 % CI: 45.5–91.5 and >100, 95 % CI: 90.9- > 100, respectively) [23]. Tracking coefficients are influenced by the time interval considered, and a higher tracking coefficient over a short period isn’t necessarily a stronger indication of tracking than a lower coefficient over a longer time interval [11, 16]. Therefore we consider the findings of moderate tracking coefficients from 2–4 and 5–7 years of age to adolescence to be consistent and important findings in our study. The same pattern has also been found in longitudinal studies from Sweden and Iceland [12, 22]. Our results are similar to the results of the Swedish study, in which 60 % of children who were overweight or obese at 5.5 years, and 44 % of children who were overweight or obese at 2.5 years were also overweight or obese at 20 years of age [22].
Although we found a moderate degree of tracking from 2–4 and 5–7 years of age to 15–17 years age, the proportion of overweight/obese children that became thin/normal weight was higher (60.7 % and 37.0 %) than the proportion of thin/normal weight children that became overweight/obese (17.6 % and 13.3 %) in the same time interval. This finding is in accordance with some studies [12, 22], but not others [20, 23]. Recent systematic reviews and a meta-analysis [10, 11] concluded that reliable studies consistently reported an increased risk for overweight and obese youths to become overweight or obese adults. They further reported that there was strong evidence that persistence of overweight and obesity is moderate. The findings in our study are in line with this conclusion, although the use of different definitions of overweight and obesity makes direct comparisons challenging. BMI at 2–4 years of age was a poor predictor of overweight/obesity at 15–17 years of age. At 2.5 years of age only BMI corresponding to an adult BMI of 30 kg/m2 or higher, predicted overweight/obesity at adolescence with over 50 %. High BMI at 5–7 years of age was a better predictor of overweight/obesity in adolescence with higher probabilities. More severe overweight and obesity in childhood seems to be a stronger predictor of overweight/obesity later in life. This is in accordance with findings in several other studies [10, 11, 22, 36].
Mean BMI in our study decreased between 2–4 and 5–7 years of age, but increased between both ages in childhood and adolescence. This may be explained by the natural variation in BMI during early childhood and the adiposity rebound that occurs between 3 and 7 years of age that is expected in the natural development of BMI [19, 28]. Mean BMI increase in children who were overweight/obese at 5–7 years of age, was significantly higher than that of thin/normal weight children. Change in mean BMI in children who were overweight/obese at 2–4 years of age did not differ significantly compared to thin/normal weight children. This shows the same pattern as the tracking analyses, i.e. that overweight and obesity tend to be more stable later in childhood and overweight and obesity may get more severe with age [11, 22, 36]. In our study the number of obese participants was too low to do separate analyses for the obese group.
Our results regarding the development of BMI differ from those on the development of BMI SDS, as BMI SDS decreased in overweight/obese children and increased in the thin/normal weight children in the same time interval. This indicates that both groups transitioned to more normal weight over time compared to the IOTF reference population [28]. This may also be seen as an expression of the statistical phenomenon; the regression towards the mean that may occur in repeated measurements [40]. Other studies that have looked at the development of overweight and obesity retrospectively have found opposite results [20, 23]. BMI SDS is commonly used to look for long-term trends in growth [6, 23, 28] and we wanted to look at the natural development prospectively. This may be a basis for comparison for intervention studies. It has been suggested that short-term change in adiposity in children is best evaluated by BMI and that BMI SDS is less suitable because it depends on baseline BMI [25, 26]. The children in our study were not severely overweight/obese at baseline. The prevalence of obesity was 1.5 % (n = 8) and 4.3 % (n = 23) in our study sample (boys and girls combined), with a BMI SDS of 1.73 and 2.03 among the overweight/obese at 2–4 and 5–7 years of age, respectively. The weight classes thinness and normal weight were merged in our study, which may explain the increase in BMI SDS in this combined group as the prevalence of thinness decreased with age. BMI SDS may also depend on the growth reference used. We used the international reference as recommended [24, 28], but we also repeated the analyses using the 1990 UK reference population [30]. No difference in development of BMI SDS was present when comparing our data with the 1990 UK reference population [30]. The pattern shown in Fig. 3 might be differences in growth pattern between different populations since Norway, and in general Nordic populations, have lower prevalence of overweight and obesity than other populations.
More girls than boys were categorized as overweight/obese in childhood. At adolescence this tendency changed and more boys than girls were categorized as overweight/obese. This is in accordance with findings from several other studies [4–6, 22, 35] and is an interesting finding that should be explored in further research. However, the gender difference was statistically significant only at 5–7 years of age and was not investigated further in this study.
Strengths of this study were the population-based design, the high participation proportion in Fit Futures 1 (>92 %) [27], longitudinal data over more than a decade, and a fairly large study sample for the tracking analyses (n = 532). The prevalence rate of overweight/obesity was comparable with most other study populations in Norway and the Nordic countries born before 2000. Excluded youths, in the core age group under 18 years of age, had statistically significantly higher BMI at adolescence (boys: 22.7 kg/m2, girls: 22.6 kg/m2) compared to the study sample (boys: 22.1 kg/m2, girls: 22.2 kg/m2). Nevertheless we consider the groups to be essentially the same, as the differences were small [27]. We therefore consider the results to be representative for Norwegian children and adolescents. Data collected in Fit Futures 1 were standardized measures and a calibrated scale was used, therefore the data are considered valid [27]. In this study we used IOTF cut-off values and LMS values based on an international reference population, as recommended [28]. This may ease comparisons between studies and is an advantage of our study. We also believe that the use of several methods showing consistency between findings strengthens the reliability of our findings.
However, our study also has some limitations. The magnitude of the tracking coefficient can be affected by inaccuracies in the measures of height and weight collected from childhood health records. Several factors can affect this accuracy e.g. inter-observer variability and the accuracy and quality control of different scales [41]. Usually these factors are anticipated to be non-differential errors that lead to both higher and lower height and weight measures, and that thus affect all weight classes in approximately equal amounts [5, 6]. Data from routine measurements are commonly used in studies of overweight and obesity in children [5, 6, 12, 22, 23]. Despite the fairly large study sample, the obese group was too small to investigate any differences between the obese group and the other weight classes separately. Other studies have indicated that the degree of tracking is higher with more severe overweight and obesity [11, 22]. Another weakness of this study is the lack of explanatory variables that were available from childhood. This gives us limited possibilities to explain and adjust our findings for influential factors that have been identified in other studies, like maternal BMI, maternal smoking, parental education level and other factors [35, 42]. This is especially a limitation with regard to our description of the natural development of BMI. We don’t have information of potential influential factors like dietary habits, physical activity level or any weight controlling interventions in this period from childhood to adolescence.
This study adds to the knowledge base of tracking of overweight and obesity using data from early childhood to adolescence in a Norwegian cohort. From a public health perspective, the positive finding was that the vast majority of normal weight children remained of normal weight up to adolescence. Since high BMI alone, especially at 2–4 years of age, only is of moderate predictive value for overweight and obesity at adolescence, additional risk factors as parental factors must be taken into account to identify children at high risk [35, 42]. Children at high risk of becoming overweight/obese in later life may be identified before they reach school age. Early childhood and preschool age stands out as an important time to focus both on individual preventive initiatives targeting those at high risk as well as all children. The need for early intervention has also been pointed out by other researchers [43]. Childhood health controls are a natural meeting point between public health nurses, children, and their parents. Our study also showed that the prevalence of overweight and obesity increased significantly (doubled in boys) between childhood and adolescence. In addition, children changed weight classes in both directions. Therefore a broad focus on general preventive efforts in society, targeting all children, seems highly appropriate.