Search

Multivariate Gaussian mixtures are a class of models that provide a flexible parametric approach for the representation of heterogeneous multivariate outcomes. When the outcome is a vector of repeated measurements taken on the same subject, there is often inherent dependence between observations. However, a common covariance assumption is conditional independence-that is, given the mixture component label, the outcomes for subjects are independent. In this paper, we study, through asymptotic bias calculations and simulation, the impact of covariance misspecification in multivariate Gaussian mixtures. Although maximum likelihood estimators of regression and mixing probability parameters are not consistent under misspecification, they have little asymptotic bias when mixture components are well separated or if the assumed correlation is close to the truth even when the covariance is misspecified. We also present a robust standard error estimator and show that it outperforms conventional estimators in simulations and can indicate that the model is misspecified. Body mass index data from a national longitudinal study are used to demonstrate the effects of misspecification on potential inferences made in practice., (Copyright © 2013 John Wiley & Sons, Ltd.)

It has been hypothesized that environmental exposures at key development periods such as in utero play a role in childhood growth and obesity. To investigate whether in utero exposure to endocrine-disrupting chemicals, dichlorodiphenyltrichloroethane (DDT) and its metabolite, dichlorodiphenyldichloroethane (DDE), is associated with childhood physical growth, we took a novel statistical approach to analyze data from the CHAMACOS cohort study. To model heterogeneity in the growth patterns, we used a finite mixture model in combination with a data transformation to characterize body mass index (BMI) with four groups and estimated the association between exposure and group membership. In boys, higher maternal concentrations of DDT and DDE during pregnancy are associated with a BMI growth pattern that is stable until about age five followed by increased growth through age nine. In contrast, higher maternal DDT exposure during pregnancy is associated with a flat, relatively stable growth pattern in girls. This study suggests that in utero exposure to DDT and DDE may be associated with childhood BMI growth patterns, not just BMI level, and both the magnitude of exposure and sex may impact the relationship.

ObjectiveTo investigate long-term body mass index (BMI) changes associated with childbearing.Design and methodsAdjusted mean BMI changes were estimated by race-ethnicity, baseline BMI, and parity using longitudinal regression models for 3,943 young females over 10 and 25 year follow-up from the ongoing 1979 National Longitudinal Survey of Youth cohort.ResultsEstimated BMI increases varied by group, ranging from a low of 2.1 BMI units for white, non-overweight nulliparas over the first 10 years to a high of 10.1 BMI units for black, overweight multiparas over the full 25-year follow-up. Impacts of parity were strongest among overweight multiparas and primaparas at 10 years, ranges 1.4-1.7 and 0.8-1.3 BMI units, respectively. Among non-overweight women, parity-related gain at 10 years varied by number of births among black and white but not Hispanic women. After 25 years, childbearing significantly increased BMI only among overweight multiparous black women.ConclusionChildbearing is associated with permanent weight gain in some women, but the relationship differs by maternal BMI in young adulthood, number of births, race-ethnicity, and length of follow-up. Given that overweight black women may be at special risk for accumulation of permanent, long-term weight after childbearing, effective interventions for this group are particularly needed.

ObjectivesTo address molecular mechanisms underlying obesity development, we examined patterns of critical metabolism-related hormones, adiponectin and leptin (adipokines), over childhood.Subjects and designPlasma adiponectin and leptin were measured in 80 Mexican-American children at birth and again at 2, 5, and 9 years from the ongoing prospective cohort followed by the Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS). We used a mixture modeling approach to identify patterns in adipokine trajectories from birth to 9 years.ResultsLeptin was positively related to child body size within all ages, however adiponectin had inverse and weaker associations with BMI at 2, 5, and 9 years. Correlations between adipokine levels over the 0-2, 2-5, and 5-9-year periods increased for both leptin (r = 0.06, 0.31 and 0.62) and adiponectin (r = 0.25, 0.41 and 0.46). Our mixture modeling approach identified three trajectory clusters for both leptin (1L [slowly-rising], 2L [rapidly-rising], and 3L [stable]) and adiponectin (1A [steep-dropping and rebounding], 2A [moderately-dropping], and 3A [stable]). While leptin groups were most separated over the 2-9-year period, adiponectin trajectories displayed greatest heterogeneity from birth to 2 years. Children in the rapidly-rising 2L group had highest BMI and waist circumference at 9 years. Further, children with greater birth weight had increased odds of belonging to this high risk group (OR = 1.21 95% CI 1.03, 1.43, compared to stable group 3L). Children whose mothers consumed more sugar-sweetened beverages during pregnancy were at risk of being in the steep-dropping 1A group (OR = 1.08, 95% CI 1.01, 1.17, compared to stable group 3A).ConclusionOur results highlight developmental differences in leptin and adiponectin over the childhood period. Leptin closely reflects child body size however factors affecting adiponectin and long-term consequences of its changes over infancy need to be further explored.

Per- and polyfluoroalkyl substances (PFAS) are environmentally persistent, potential metabolic disruptors of concern for infants. Mothers participating in the New Hampshire Birth Cohort Study (NHBCS) provided a plasma sample during pregnancy to measure concentrations of seven PFAS, and infant weight and length were abstracted from well-child visits between birth and 12 months. Sex-specific growth patterns of child body mass index (BMI) were fit using a growth mixture model (GMM) and the relative risk ratios (RRR) and 95% Confidence Intervals (95% CI) for the association of maternal plasma PFAS with BMI growth patterns during infancy were estimated by using multinomial logistic model for the group probabilities in the GMM. Four growth patterns were identified: Group 1) a steep increase in BMI during the first 6 months, then a leveling off; Group 2) a gradual increase in BMI across the year; Group 3) a steep increase in BMI during months 1–3, then stable BMI; and Group 4) a gradual increase in BMI with plateau around 3 months (reference group). For boys, higher maternal pregnancy perfluorooctanoate concentrations were associated with a 60% decreased chance of being in group 3 as compared to group 4, after adjusting for potential confounding variables (RRR = 0.4; 95% CI: 0.1, 0.9). For girls, higher maternal perfluorooctane sulfonate (PFOS) concentrations during pregnancy were associated with a higher likelihood of following the growth pattern of groups 2 (RRR = 2.5; 95% CI: 1.0, 6.1) and 3 (RRR = 2.8; 95% CI: 1.0, 7.6) as compared to group 4, adjusting for potential confounding variables. In this cohort, sex-specific associations of maternal plasma PFAS concentrations during pregnancy with growth patterns during the first year of life were observed, with greater BMI growth observed among infant girls born to mothers with higher pregnancy concentrations of PFOS. • Sex-specific growth patterns in infancy were fit using a growth mixture model. • Associations of PFAS with growth in infancy were sex-specific. • Among girls with higher prenatal PFOS, steeper BMI growth occurred during infancy. [ABSTRACT FROM AUTHOR]