Limits...
Childhood body mass index trajectories: modeling, characterizing, pairwise correlations and socio-demographic predictors of trajectory characteristics.

Wen X, Kleinman K, Gillman MW, Rifas-Shiman SL, Taveras EM - BMC Med Res Methodol (2012)

Bottom Line: Among 3,289 children seen at 81,550 pediatric well-child visits from infancy to 18 years between 1980 and 2008, we fit individual BMI trajectories using mixed effect models with fractional polynomial functions.BMI trajectories did not differ by birth year or type of health insurance, after adjusting for other socio-demographics and birth weight z-score.Future research should evaluate associations of these novel BMI trajectory characteristics with adult outcomes.

View Article: PubMed Central - HTML - PubMed

Affiliation: Obesity Prevention Program, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA 02215, USA. xiaozhongwen@hotmail.com

ABSTRACT

Background: Modeling childhood body mass index (BMI) trajectories, versus estimating change in BMI between specific ages, may improve prediction of later body-size-related outcomes. Prior studies of BMI trajectories are limited by restricted age periods and insufficient use of trajectory information.

Methods: Among 3,289 children seen at 81,550 pediatric well-child visits from infancy to 18 years between 1980 and 2008, we fit individual BMI trajectories using mixed effect models with fractional polynomial functions. From each child's fitted trajectory, we estimated age and BMI at infancy peak and adiposity rebound, and velocity and area under curve between 1 week, infancy peak, adiposity rebound, and 18 years.

Results: Among boys, mean (SD) ages at infancy BMI peak and adiposity rebound were 7.2 (0.9) and 49.2 (11.9) months, respectively. Among girls, mean (SD) ages at infancy BMI peak and adiposity rebound were 7.4 (1.1) and 46.8 (11.0) months, respectively. Ages at infancy peak and adiposity rebound were weakly inversely correlated (r = -0.09). BMI at infancy peak and adiposity rebound were positively correlated (r = 0.76). Blacks had earlier adiposity rebound and greater velocity from adiposity rebound to 18 years of age than whites. Higher birth weight z-score predicted earlier adiposity rebound and higher BMI at infancy peak and adiposity rebound. BMI trajectories did not differ by birth year or type of health insurance, after adjusting for other socio-demographics and birth weight z-score.

Conclusions: Childhood BMI trajectory characteristics are informative in describing childhood body mass changes and can be estimated conveniently. Future research should evaluate associations of these novel BMI trajectory characteristics with adult outcomes.

Show MeSH

Related in: MedlinePlus

Fitted BMI trajectories of 8 randomly selected children, one from each quartile of residual BMI variance. Lower residual BMI variance indicates a better fit of an individual's data points to the individual-specific model. A) 1st quartile - a boy (residual BMI variance = 0.21), B) 2nd quartile - a boy (0.60), C) 3rd quartile - a boy (1.08), D) 4th quartile - a boy (1.26), E) 1st quartile - a girl (0.23), F) 2nd quartile - a girl (0.69), G) 3rd quartile - a girl (0.85), H) 4th quartile - a girl (1.59).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3375197&req=5

Figure 3: Fitted BMI trajectories of 8 randomly selected children, one from each quartile of residual BMI variance. Lower residual BMI variance indicates a better fit of an individual's data points to the individual-specific model. A) 1st quartile - a boy (residual BMI variance = 0.21), B) 2nd quartile - a boy (0.60), C) 3rd quartile - a boy (1.08), D) 4th quartile - a boy (1.26), E) 1st quartile - a girl (0.23), F) 2nd quartile - a girl (0.69), G) 3rd quartile - a girl (0.85), H) 4th quartile - a girl (1.59).

Mentions: Among the 8 candidate variance-covariance structures, the autoregressive structure had the lowest BIC in the model 8 candidate polynomials, and was thus chosen for further selection of the best mean structure from the 219 candidate models. The mean of BIC values of these candidate models was 168,029 (SD, 7,841) for boys, and 158,490 (SD, 8,006) for girls. Table 2 shows goodness of fit for the best models by degree. For boys, the best model (lowest BIC) was "BMI = 96.8 - 3.6*Age(-2) + 51.0*Age(-1) - 134.9*Age(-0.5) - 24.4*ln(Age) + 4.6*Age0.5, and for girls it was "BMI = 90.8 - 3.2*Age(-2) + 47.0*Age(-1) - 125.4*Age(-0.5) - 22.7*ln(Age) + 4.3*Age0.5. Overall, these two 5th-degree models fit BMI trajectories of most children with reasonable accuracy, according to the distribution of residual BMI variances: inter-quartile range 0.49-1.18 BMI units for boys and 0.51-1.15 for girls (Figure 2). Figure 3 shows observed BMI values and individual-specific fitted BMI trajectories of 8 children randomly selected within quartile of residual BMI variance by sex.


Childhood body mass index trajectories: modeling, characterizing, pairwise correlations and socio-demographic predictors of trajectory characteristics.

Wen X, Kleinman K, Gillman MW, Rifas-Shiman SL, Taveras EM - BMC Med Res Methodol (2012)

Fitted BMI trajectories of 8 randomly selected children, one from each quartile of residual BMI variance. Lower residual BMI variance indicates a better fit of an individual's data points to the individual-specific model. A) 1st quartile - a boy (residual BMI variance = 0.21), B) 2nd quartile - a boy (0.60), C) 3rd quartile - a boy (1.08), D) 4th quartile - a boy (1.26), E) 1st quartile - a girl (0.23), F) 2nd quartile - a girl (0.69), G) 3rd quartile - a girl (0.85), H) 4th quartile - a girl (1.59).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3375197&req=5

Figure 3: Fitted BMI trajectories of 8 randomly selected children, one from each quartile of residual BMI variance. Lower residual BMI variance indicates a better fit of an individual's data points to the individual-specific model. A) 1st quartile - a boy (residual BMI variance = 0.21), B) 2nd quartile - a boy (0.60), C) 3rd quartile - a boy (1.08), D) 4th quartile - a boy (1.26), E) 1st quartile - a girl (0.23), F) 2nd quartile - a girl (0.69), G) 3rd quartile - a girl (0.85), H) 4th quartile - a girl (1.59).
Mentions: Among the 8 candidate variance-covariance structures, the autoregressive structure had the lowest BIC in the model 8 candidate polynomials, and was thus chosen for further selection of the best mean structure from the 219 candidate models. The mean of BIC values of these candidate models was 168,029 (SD, 7,841) for boys, and 158,490 (SD, 8,006) for girls. Table 2 shows goodness of fit for the best models by degree. For boys, the best model (lowest BIC) was "BMI = 96.8 - 3.6*Age(-2) + 51.0*Age(-1) - 134.9*Age(-0.5) - 24.4*ln(Age) + 4.6*Age0.5, and for girls it was "BMI = 90.8 - 3.2*Age(-2) + 47.0*Age(-1) - 125.4*Age(-0.5) - 22.7*ln(Age) + 4.3*Age0.5. Overall, these two 5th-degree models fit BMI trajectories of most children with reasonable accuracy, according to the distribution of residual BMI variances: inter-quartile range 0.49-1.18 BMI units for boys and 0.51-1.15 for girls (Figure 2). Figure 3 shows observed BMI values and individual-specific fitted BMI trajectories of 8 children randomly selected within quartile of residual BMI variance by sex.

Bottom Line: Among 3,289 children seen at 81,550 pediatric well-child visits from infancy to 18 years between 1980 and 2008, we fit individual BMI trajectories using mixed effect models with fractional polynomial functions.BMI trajectories did not differ by birth year or type of health insurance, after adjusting for other socio-demographics and birth weight z-score.Future research should evaluate associations of these novel BMI trajectory characteristics with adult outcomes.

View Article: PubMed Central - HTML - PubMed

Affiliation: Obesity Prevention Program, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA 02215, USA. xiaozhongwen@hotmail.com

ABSTRACT

Background: Modeling childhood body mass index (BMI) trajectories, versus estimating change in BMI between specific ages, may improve prediction of later body-size-related outcomes. Prior studies of BMI trajectories are limited by restricted age periods and insufficient use of trajectory information.

Methods: Among 3,289 children seen at 81,550 pediatric well-child visits from infancy to 18 years between 1980 and 2008, we fit individual BMI trajectories using mixed effect models with fractional polynomial functions. From each child's fitted trajectory, we estimated age and BMI at infancy peak and adiposity rebound, and velocity and area under curve between 1 week, infancy peak, adiposity rebound, and 18 years.

Results: Among boys, mean (SD) ages at infancy BMI peak and adiposity rebound were 7.2 (0.9) and 49.2 (11.9) months, respectively. Among girls, mean (SD) ages at infancy BMI peak and adiposity rebound were 7.4 (1.1) and 46.8 (11.0) months, respectively. Ages at infancy peak and adiposity rebound were weakly inversely correlated (r = -0.09). BMI at infancy peak and adiposity rebound were positively correlated (r = 0.76). Blacks had earlier adiposity rebound and greater velocity from adiposity rebound to 18 years of age than whites. Higher birth weight z-score predicted earlier adiposity rebound and higher BMI at infancy peak and adiposity rebound. BMI trajectories did not differ by birth year or type of health insurance, after adjusting for other socio-demographics and birth weight z-score.

Conclusions: Childhood BMI trajectory characteristics are informative in describing childhood body mass changes and can be estimated conveniently. Future research should evaluate associations of these novel BMI trajectory characteristics with adult outcomes.

Show MeSH
Related in: MedlinePlus