Clinical Trials Centre, The University of Hong Kong, Hong Kong SAR, P.R. China. E-mail: jpekarl@hkucc.hku.hk

Abstract

Objective: To develop a probability chart of adult overweight based on childhood body mass index (BMI) values and to evaluate the BMI change during the BMI rebound period during childhood, in different populations, with the use of risk function curves.

Research Methods and Procedures: A longitudinal growth study of 3650 full-term healthy Swedish children followed from birth to 18 years of age. Weight and height values of our subjects were obtained.

Results: A probability chart for reaching a BMI > 23 kg/m^{2} at 18 years of age was constructed for boys and girls. For example, a BMI of 18 kg/m^{2} at 4 years of age is associated with 0.70 probability of attaining a BMI > 23 kg/m^{2} at 18 years of age in boys; a BMI of 16 kg/m^{2} at 4 years of age leads to 0.40 probability of having a BMI > 23 kg/m^{2} at 18 years of age in girls. Children with an obvious BMI rebound before 8 years of age have a high risk of being overweight at 18 years of age. There is a clear trend of BMI increase from the 1970s to the 1990s in U.S. children from a parallel dataset, and Hispanic children are at the highest risk of adult overweight.

Discussion: The probability chart for adult overweight developed here provides a functional method of defining childhood obesity that is based on the risk of long-term ill health rather than on a certain statistical cut-off point. It will help pediatricians or healthcare workers identify those children who are at a high risk of becoming overweight in adulthood, which will allow clinical intervention at younger ages.

The prevalence of obesity is increasing dramatically worldwide in both developed and developing countries (1). Moreover, because the problem seems to be increasing rapidly in children as well as in adults, the health consequence of obesity will only be fully apparent in the next decade or so. Action for the prevention and management of obesity is urgently required on a global basis. Obesity and overweight are related to many diseases in adulthood, such as diabetes, hypertension, dyslipidemia, and cardiovascular diseases (2, 3). Although difficult to treat once developed, obesity and its risk seem preventable through early intervention. Because childhood obesity may persist into adulthood (4, 5, 6), it is necessary to identify high-risk individuals for clinical intervention at an early age. Research, based on a large longitudinal growth study (7), using childhood body mass index (BMI) to predict adult obesity (BMI > 25 kg/m^{2}), has previously been published. However, because the risk factors for obesity-related disorders start to increase from about a BMI of 23 kg/m^{2} in most populations (8, 9, 10), setting a BMI goal of 23 kg/m^{2} would be appropriate from a prevention point of view. The aims of this study were to produce the probability charts for adult overweight based on a cut-off point of BMI > 23 kg/m^{2} and to explore how the risk of adult overweight changes during the BMI rebound period in different ethnic groups and generations.

Research Methods and Procedures

Subjects

The study population was selected from 5111 grade-school children who were born in Gothenburg, Sweden, and surrounding areas. Because more than 98% of children between the ages of 17 and 19 years remain within the school system in Sweden, including those with mild or moderate problems in mental development, we believe that there is no important selection bias in the study population and almost the whole cohort of children were involved. Information on 319 girls and 304 boys was not obtained because of unwillingness or failure to attend the last investigation at school; 76.8% of the remaining (n = 4488) were born in 1974, 16.7% in 1973, 3.0% before 1973, and 3.5% in 1975. Exclusion criteria, such as multiple births, prematurity, growth disorders, and lack of information at birth, were applied in the selection of subjects. The data in the analysis, therefore, represents 3650 full-term (37 to 42 weeks of gestation) and healthy Swedish children, with growth information from birth to 18 years of age (11). The sample size for each gender varied between 582 and 1857 for the various ages; at 18 years of age, 1849 boys and 1801 girls remained.

Data Collection

The data were collected from April to November 1992. Four trained investigators from the International Pediatric Growth Research Center, Department of Pediatrics, University of Gothenburg, and two school nurses comprised the study group. The weight and height of all the children were measured in a standardized way using a calibrated Harpenden stadiometer. Information relating to each child at birth and throughout the prenatal period, such as size at birth, length of gestation, and any health problems, was obtained from the Swedish Birth Register at the Swedish National Board for Health and Welfare. The other health records, including height and weight measurements of children from birth to 18 years of age were taken from the notes made by doctors and nurses at the Child Health Centers or at schools. A total of 99.2% of our study population had reached their final height at 18 years of age when the study team visited the schools (11).

All growth data were computerized and quality-control analysis was performed. For the study group, 52,078 height and weight measurements were available from birth to the last examination, and the average number of measurements for each child was 14.4.

This study was approved by the Ethical Committee of the Medical Faculty, the University of Gothenburg, Sweden, and the Swedish Data Authorities. Written informed consent was received from the children and from those parents whose children were below 16 years of age.

Data Analysis

The classification of BMI for adults from the International Obesity Task Force was chosen as a references in a number of previous studies. The classification is as follows: underweight, BMI < 18.5 kg/m^{2}; normal range, 18.5 to 24.9 kg/m^{2}; overweight, 25.0 to 29.9 kg/m^{2}; and obese, ≥30.0 kg/m^{2} (12). An international workshop on childhood obesity also established percentile cut-off points for obesity in childhood and adolescence that corresponded to BMIs of 25 and 30 kg/m^{2} in young adults (12). However, in most populations, increased risk factors for obesity-related disorders start from a BMI of 23 kg/m^{2} (8, 9, 10). Therefore, having a BMI goal of 23 kg/m^{2} would be ideal. In this work, a BMI > 23 kg/m^{2} was defined as the cut-off point of overweight at 18 years of age when making the prediction model.

BMI change between 3 and 8 years of age is important for predicting adult overweight (13). An illustration showing the mean BMI change during this period was made according to six BMI groups at 18 years of age. A comparison of the risk pattern of adult overweight during this period was also made among different ethnic groups and generations. Swedish (14) and U.S. (15) BMI reference values (5th, 50th, and 95th percentiles) represent the BMI level of the children living in the 1970s. The BMI values of the U.S. white and Hispanic children were taken from the large epidemiological study based on children living in the U.S. in the 1990s (16).

All statistics were computed using SAS software (Version 6.10; SAS Institute Inc., Cary, NC) (17).

Results

Probability Chart of Adult Overweight at 18 Years

Figure 1, A and B, shows the probability charts starting from 3 years of age for the prediction of subjects overweight at 18 years of age. Overweight at 18 years of age is defined here as having a BMI > 23 kg/m^{2} in both sexes. This means that some are overweight and obese, with a BMI > 30 kg/m^{2}. Based on this chart, we can estimate that a BMI of 18 kg/m^{2} at 4 years of age is associated with a 0.70 probability (70% risk) of attaining a BMI > 23 kg/m^{2} by 18 years of age for boys (Figure 1A). Another example is as follows: girls with a BMI of 16 kg/m^{2} at 4 years of age have a 40% risk of having a BMI > 23 kg/m^{2} at 18 years of age(Figure 1B).

Construction of the Probability Charts

The first step in producing the probability charts shown in Figure 1, A and B, is based on the percentage of boys or girls with a BMI > 23 kg/m^{2} at 18 years of age in relation to their BMI values during the pediatric years (data not shown). For those boys with a BMI > 6 kg/m^{2} at 4 years of age, 30.8% reached a BMI > 23 kg/m^{2} at 18 years of age. However, for boys with a BMI > 19 kg/m^{2} at 4 years of age, 89.5% reached a BMI > 23 kg/m^{2} at 18 years of age. Note that the percentage represents the number of children becoming overweight at 18 years of age for a certain childhood BMI cut-off point.

As a second step in the probability chart development, we fitted a fourth degree polynomial function to the percentage values to represent the various cut-off points for gender and age. Details of the curve-fitting have been published elsewhere (7). From these age- and gender- specific polynomial functions, we could determine the pediatric BMI value associated with 30%, 40%, 50%, 60%, 70%, and 80% risk of adult overweight in girls and with 40% to 90% risk in boys. Values below 3 years of age were not estimated in this work, because the risk of adult overweight in infancy is different and relatively less than that in childhood.

The last step in the construction of the probability charts was to fit a third degree polynomial function to the values derived in step two. The function was, for instance, fitted to the BMI values associated with a 30% risk of adult overweight from 3 to 17 years of age for girls. The same procedure was also used for the other BMI-related risk values for boys and girls separately.

Table 1 gives the parameters of the fitted function with the corresponding R^{2} values. These functions were used in the final construction of the probability charts shown in Figure 1, A and B. Functions of the probability curve, 0.30 for boys and 0.90 for girls, were not provided due to unsatisfactory curve-fitting.

Table 1. Probability functions of adult overweight (body mass index > 23 kg/m^{2})

Probability of adult overweight

Slope

Intercept α

β1

β2

β3

F value

p value

R^{2}

Body mass index value (corresponding to a certain risk) =α + β1 * (Age) + β2 * (Age)^{2} +β3 * (Age)^{3}.

Boys

0.40

20.87251

−1.95369

0.17778

−0.00400

123.99

0.0001

0.97

0.50

22.31625

−2.15114

0.20031

−0.00476

192.87

0.0001

0.98

0.60

22.86116

−2.10884

0.20143

−0.00493

184.56

0.0001

0.98

0.70

23.54430

−2.13409

0.20985

−0.00536

160.51

0.0001

0.98

0.80

23.94907

−2.06655

0.21155

−0.00565

162.47

0.0001

0.98

0.90

24.55617

−2.07487

0.22710

−0.00655

111.74

0.0001

0.97

Girls

0.30

21.09806

−2.24491

0.20092

−0.00446

117.27

0.0001

0.97

0.40

21.75067

−2.18873

0.20835

−0.00499

227.12

0.0001

0.98

0.50

21.85760

−1.98609

0.19631

−0.00484

232.04

0.0001

0.98

0.60

21.74615

−1.71618

0.17566

−0.00441

163.79

0.0001

0.98

0.70

21.34483

−1.37979

0.15160

−0.00396

142.56

0.0001

0.97

0.80

21.18030

−1.12931

0.13951

−0.00399

84.28

0.0001

0.96

Change of the Risk of Adult Overweight during the BMI Rebound Period

The smoothed mean values of BMI from birth to 8 years of age are shown for six BMI groups at 18 years of age with the probability chart of adult overweight for boys (Figure 2A) and girls (Figure 2B). The six BMI groups are defined as follows:≤18 kg/m^{2}, 18 to 21 kg/m^{2}, 21 to 23 kg/m^{2}, 23 to 25 kg/m^{2}, 25 to 27 kg/m^{2}, or>27 kg/m^{2} at 18 years of age. The difference of the mean BMI among these six BMI groups became obvious from 2 years of age. No sign of BMI rebound before 8 years of age was observed for both boys and girls in the two lower BMI groups (≤18 kg/m^{2} and 18 to 21 kg/m^{2}). The BMI rebound was shown clearly, however, for both sexes in the three higher BMI groups, which had a BMI > 23 kg/m^{2} at 18 years of age. The risk of adult overweight between 3 and 8 years of age also changed gradually for these six BMI groups, with a larger risk curve in the higher BMI groups.

Comparison of the Risk Pattern of Adult Overweight in Different Populations and Generations

BMI percentile values of U.S. children living in the 1970s and the 1990s and Swedish children born in the 1970s were illustrated in combination with the probability chart of adult overweight (Figure 3, A and B). Note that U.S. studies are cross-sectional in nature, whereas the Swedish study is longitudinal. The difference of risk pattern of adult overweight was not obvious among these four study groups for the 5th and 50th percentile values. It is notable that all U.S. 95th percentile values lie above the Swedish 95th percentile, which varied between a 0.70 and 0.80 probability curve for adult overweight in both sexes. The BMI values of the U.S. children in the 1990s, represented by the white and the Hispanic children, are much higher than U.S. reference values, which represent the children in the 1970s. According to the probability chart of adult overweight, 5% of U.S. children today are above a 70% risk of adult overweight at 5 years of age, and the risk increases steeply for older children. The U.S. Hispanic children have the highest risk of adult overweight among the four study groups during this critical period.

Discussion

A set of practical and simple probability charts for overweight at 18 years of age is presented based on the large longitudinal growth study worldwide, consisting of 3650 full-term babies followed from birth to 18 years of age. Obvious BMI rebound before 8 years of age associated with a high risk of adult overweight was also shown in our study.

The prediction of adulthood obesity during the pediatric years is urgently needed because of the dramatic increasing prevalence of adult obesity (18) and the relatively unsuccessful treatment of obese adults (19). Because obesity in childhood is associated with less morbidity than in adulthood, evaluation of childhood obesity should be based on the risk of short-term or long-term ill health, rather than on a certain statistical cut-off point (20). Very little data is available linking age-specific childhood BMI values to the risk of short-term or long-term ill health (21). A set of prediction charts for obtaining a BMI > 23 kg/m^{2} at 18 years of age was created for boys and girls separately in our study. These charts provide a novel and easy way of estimating the risk of young adults (18 years of age) being overweight based on the individual BMI value from 3 years of age and older. The data used in our study came from a large, longitudinal, population-based growth study of Swedish children followed from birth to 18 years of age. This study represents the largest current longitudinal growth study worldwide. The representativeness and uniqueness of this data set have been discussed in previous publications (7, 14).

Another attempt to predict adult obesity is related to the nadir of the BMI-for-age curve, which usually occurs between 3 and 8 years of age. Rolland-Cachera et al. (13) reported that most children follow the percentile line pattern of the BMI chart; a decrease followed by a rise in BMI values during this period. The corresponding time of the minimum BMI value during this period is called the adiposity rebound age; the earlier the adiposity rebound, the greater the risk of adult obesity (22). We depict BMI change during this period using the probability chart developed in our study. It clearly can be seen that a group with a rebound before 8 years of age will have a higher risk of adult obesity, reaching a BMI > 23 kg/m^{2} at 18 years of age. Therefore, maintaining less increase in BMI, or following the same risk curve before 8 years of age, is important to prevent overweight at 18 years of age.

The secular trend of BMI increase in the U.S. population can also be clearly seen when the BMI level in the 1970s is compared with that in the 1990s. Today, 5% of U.S. children are above 70% risk of adult overweight at 5 years of age, and Hispanic children are at the highest risk of adult overweight. Because the long-term treatment programs developed for obese children are relatively expensive and time-consuming, children who are at a high risk of adult overweight should be treated as the target group for intervention.

The global extrapolation of current results, which are based on a Swedish sample, might require further discussion. With the dramatic increase in overweight and obesity worldwide, it is possible that the secular changes over time in Sweden, as well as elsewhere, could have or may change the BMI prediction model presented here.

Because the risk of obesity-related disorders starts to increase from a BMI of 23 kg/m^{2} in most populations, a BMI goal of 23 kg/m^{2} would be ideal from a preventative point of view. A set of practical, simple, and novel probability charts for adult overweight are reported here. These charts will provide a useful tool for pediatricians to identify those children who are at a high risk of becoming overweight in adulthood so that clinical intervention can be started as early as possible.

Acknowledgments

This study was supported by grants from the Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, and from Pharmacia Upjohn, Inc., Stockholm, Sweden.