Early Life Predictors of Increased Body Mass Index among Indigenous Australian Children

Handling missing data and measurement error for early-onset myopia risk prediction models

BACKGROUND

Early identification of children at high risk of developing myopia is essential to prevent myopia progression by introducing timely interventions. However, missing data and measurement error (ME) are common challenges in risk prediction modelling that can introduce bias in myopia prediction.

METHODS

We explore four imputation methods to address missing data and ME: single imputation (SI), multiple imputation under missing at random (MI-MAR), multiple imputation with calibration procedure (MI-ME), and multiple imputation under missing not at random (MI-MNAR). We compare four machine-learning models (Decision Tree, Naive Bayes, Random Forest, and Xgboost) and three statistical models (logistic regression, stepwise logistic regression, and least absolute shrinkage and selection operator logistic regression) in myopia risk prediction. We apply these models to the Shanghai Jinshan Myopia Cohort Study and also conduct a simulation study to investigate the impact of missing mechanisms, the degree of ME, and the importance of predictors on model performance. Model performance is evaluated using the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC).

RESULTS

Our findings indicate that in scenarios with missing data and ME, using MI-ME in combination with logistic regression yields the best prediction results. In scenarios without ME, employing MI-MAR to handle missing data outperforms SI regardless of the missing mechanisms. When ME has a greater impact on prediction than missing data, the relative advantage of MI-MAR diminishes, and MI-ME becomes more superior. Furthermore, our results demonstrate that statistical models exhibit better prediction performance than machine-learning models.

CONCLUSION

MI-ME emerges as a reliable method for handling missing data and ME in important predictors for early-onset myopia risk prediction.

Small for gestational age and anthropometric body composition from early childhood to adulthood: the Aboriginal Birth Cohort study

Background

In Australia the estimated rate of small for gestational age (SGA) births is 9% among non-Indigenous births compared to 14% among Aboriginal and Torres Strait Islanders. There is limited research investigating the effect of being born SGA on body composition later in life in Indigenous Australians.

Methods

Using data from the Aboriginal Birth Cohort longitudinal study, we compared the body composition of those born SGA to non-SGA by analysing anthropometric measures (height, weight, waist circumference, fat percentage [FAT%], body mass index [BMI], waist-to-height ratio, and A body shape index [ABSI]) collected at four follow-up periods (from childhood to adult). For cross-sectional analyses, linear regression models were employed to assess factors associated with anthropometric measures. For longitudinal analyses linear mixed models were employed to assess differences in anthropometric measures among SGA versus non-SGA individuals while adjusting for repeated measures.

Results

The analytic baseline cohort were those who participated in Wave 2 (n = 570). In cross-sectional analyses, across all waves those born SGA had smaller anthropometric z-scores compared to non-SGA individuals (β ranging from -0.50 to -0.25). Participants residing in urban environments were significantly larger in Waves 2 to 4 (β ranged 0.26 to 0.65). Those born SGA had higher ABSI scores in Waves 2 and 4 (β 0.26 and 0.37, respectively). In longitudinal analyses, those born SGA had smaller measures of body composition across the life course; these differences were larger in urban communities. In remote communities those born SGA had significantly higher ABSI scores during adolescence and young adulthood, and this difference was not observed in urban communities.

Conclusion

Indigenous Australians born SGA are smaller anthropometrically later in life compared to their non-SGA counterparts. In remote communities, those born SGA had higher levels of central adiposity compared to non-SGA.

Examining relationships between parent-reported factors and recurring ear symptoms among Aboriginal and Torres Strait Islander children

ISSUE ADDRESSED

Aboriginal and Torres Strait Islander child ear health is complex and multiple. We examined relationships between parent-reported sociodemographic, child health, health service access factors and ear symptoms among Aboriginal and Torres Strait Islander children aged 3 to 7 years.

METHODS

The Longitudinal Study of Indigenous Children is a large child cohort study with annual parent-reported data collection. Generalised linear mixed model analyses examined Wave 1 (1309 children 0-5 years; 2008) predictors of being free of parent-reported ear symptoms in both Waves 2 and 3.

RESULTS

A total of 1030 (78.7%) had no reported ear symptoms in either Wave 2 or 3. In the fully adjusted model, children who had been hospitalised in the past year (aOR = 2.16; 95% CI 1.19-3.93) and those with no ear symptoms (aOR = 2.94; 95% CI, 1.59-5.46) at Wave 1 had higher odds of no ear symptoms in both the subsequent waves. There were also relationships between parent main source of income-government pension or allowance as well as parents who reported no history of their own ear symptoms and higher odds of no ear symptoms in Waves 2 and 3 after partial adjustment for sociodemographic factors.

CONCLUSION

These findings suggest relationships between different sociodemographic and health factors and parent-reported ear symptoms among Aboriginal and Torres Strait Islander children that warrant further investigation. So what? Children with parent-reported ear symptoms during the early years need holistic support to prevent future ear symptoms that impact health, social and educational life trajectories.

Association of large for gestational age with cardiovascular metabolic risks: a systematic review and meta-analysis

OBJECTIVE

The aim of this study was to clarify the relationships among large for gestational age (LGA) and cardiometabolic risk factors.

METHODS

PubMed, Web of Science, and the Cochrane Library databases were searched to identify studies on LGA and outcomes of interest, including BMI, blood pressure, glucose metabolism, and lipid profiles. Data were independently extracted by two reviewers. A meta-analysis was performed using a random-effects model. The Newcastle-Ottawa Scale and funnel graph were used to assess the quality and publication bias, respectively.

RESULTS

Overall, 42 studies involving 841,325 individuals were included. Compared with individuals born appropriate for gestational age, individuals born LGA had higher odds of overweight and obesity (odds ratios [OR] = 1.44, 95% CI: 1.31-1.59), type 1 diabetes (OR = 1.28, 95% CI: 1.15-1.43), hypertension (OR = 1.23, 95% CI: 1.01-1.51), and metabolic syndrome (OR = 1.43, 95%; CI: 1.05-1.96). No significant difference was found in hypertriglyceridemia and hypercholesterolemia. Stratified analyses showed that, compared with individuals born appropriate for gestational age, individuals born LGA had higher odds for overweight and obesity from toddler age to puberty age (toddler age: OR = 2.12, 95% CI: 1.22-3.70; preschool: OR = 1.81, 95% CI: 1.55-2.12; school age: OR = 1.53, 95% CI: 1.09-2.14; puberty: OR = 1.40, 95% CI: 1.11-1.77).

CONCLUSIONS

LGA is associated with increased odds of obesity and metabolic syndrome later in life. Future studies should focus on elucidating the potential mechanisms and identifying risk factors.

Potential Determinants of Cardio-Metabolic Risk among Aboriginal and Torres Strait Islander Children and Adolescents: A Systematic Review

Prevention initiatives during childhood and adolescence have great potential to address the health inequities experienced by Aboriginal and Torres Strait Islander (Indigenous) populations in Australia by targeting modifiable risk factors for cardio-metabolic diseases. We aimed to synthesize existing evidence about potential determinants of cardio-metabolic risk markers—obesity, elevated blood pressure, elevated blood glucose, abnormal lipids, or a clustering of these factors known as the metabolic syndrome (MetS)—for Indigenous children and adolescents. We systematically searched six databases for journal articles and three websites for relevant grey literature. Included articles (n = 47) reported associations between exposures (or interventions) and one or more of the risk markers among Indigenous participants aged 0–24 years. Data from 18 distinct studies about 41 exposure–outcome associations were synthesized (by outcome: obesity [n = 18]; blood pressure [n = 9]; glucose, insulin or diabetes [n = 4]; lipids [n = 5]; and MetS [n = 5]). Obesity was associated with each of the other cardio-metabolic risk markers. Larger birth size and higher area-level socioeconomic status were associated with obesity; the latter is opposite to what is observed in the non-Indigenous population. There were major gaps in the evidence for other risk markers, as well as by age group, geography, and exposure type. Screening for risk markers among those with obesity and culturally appropriate obesity prevention initiatives could reduce the burden of cardio-metabolic disease.

Footprints in Time: Physical Activity Levels and Sociodemographic and Movement-Related Associations Within the Longitudinal Study of Indigenous Children

BACKGROUND

Emerging evidence suggests that Indigenous children have higher physical activity levels that non-Indigenous children, yet little is known of the factors that influence these levels or how they may be optimized. This study examines correlates of achieving ≥1 hour/day of physical activity among Indigenous Australian children aged 8-13 years.

METHODS

Data were collected through parental self-report in the Longitudinal Study of Indigenous Children. Proportions of children achieving ≥1 hour/day physical activity, approximating the Australian aerobic physical activity recommendations, were calculated, and associations with sociodemographic, family composition, and movement-related factors were quantified using multiple logistic regression analyses.

RESULTS

Half of the 1233 children achieved ≥1 hour/day physical activity. Children from families with low parental education and unemployment, remote residence, low socioeconomic status, and without a father in the household were more likely to meet the recommendations. Achieving ≥1 hour/day of physical activity was also associated with low levels of playing electronic games and total screen time.

CONCLUSIONS

Sociodemographic correlates of physical activity among Indigenous Australian children run counter to those typically found in non-Indigenous Australian children. Further longitudinal examination of the predictors of these associations would provide a greater understanding of Indigenous physical activity determinants, to inform strategies to facilitate participation.

Developmental Programming of Body Composition: Update on Evidence and Mechanisms

PURPOSE OF REVIEW

A growing body of epidemiological and experimental data indicate that nutritional or environmental stressors during early development can induce long-term adaptations that increase risk of obesity, diabetes, cardiovascular disease, and other chronic conditions-a phenomenon termed "developmental programming." A common phenotype in humans and animal models is altered body composition, with reduced muscle and bone mass, and increased fat mass. In this review, we summarize the recent literature linking prenatal factors to future body composition and explore contributing mechanisms.

RECENT FINDINGS

Many prenatal exposures, including intrauterine growth restriction, extremes of birth weight, maternal obesity, and maternal diabetes, are associated with increased fat mass, reduced muscle mass, and decreased bone density, with effects reported throughout infancy and childhood, and persisting into middle age. Mechanisms and mediators include maternal diet, breastmilk composition, metabolites, appetite regulation, genetic and epigenetic influences, stem cell commitment and function, and mitochondrial metabolism. Differences in body composition are a common phenotype following disruptions to the prenatal environment, and may contribute to developmental programming of obesity and diabetes risk.

Predictors of Overweight and Obesity in American Indian Families With Young Children

OBJECTIVE

To describe sociodemographic factors and health behaviors among American Indian (AI) families with young children and determine predictors of adult and child weight status among these factors.

DESIGN

Descriptive, cross-sectional baseline data.

SETTING

One urban area and 4 rural AI reservations nationwide.

PARTICIPANTS

A total of 450 AI families with children aged 2-5 years participating in the Healthy Children, Strong Families 2 intervention.

INTERVENTION

Baseline data from a healthy lifestyles intervention.

MAIN OUTCOME MEASURES

Child body mass index (BMI) z-score and adult BMI, and multiple healthy lifestyle outcomes.

ANALYSIS

Descriptive statistics and stepwise regression.

RESULTS

Adult and child combined overweight and obesity rates were high: 82% and 40%, respectively. Food insecurity was high (61%). Multiple lifestyle behaviors, including fruit and vegetable and sugar-sweetened beverage consumption, adult physical activity, and child screen time, did not meet national recommendations. Adult sleep was adequate but children had low overnight sleep duration of 10 h/d. Significant predictors of child obesity included more adults in the household (P = .003; β = 0.153), an adult AI caregiver (P = .02; β = 0.116), high adult BMI (P = .001; β = 0.176), gestational diabetes, high child birth weight (P < .001; β = 0.247), and the family activity and nutrition score (P = .04; β = 0.130).

CONCLUSIONS AND IMPLICATIONS

We found multiple child-, adult-, and household-level factors influence early childhood obesity in AI children, highlighting the need for interventions to mitigate the modifiable factors identified in this study, including early life influences, home environments, and health behaviors.

Multiple imputation methods for handling missing values in a longitudinal categorical variable with restrictions on transitions over time: a simulation study

Background

Longitudinal categorical variables are sometimes restricted in terms of how individuals transition between categories over time. For example, with a time-dependent measure of smoking categorised as never-smoker, ex-smoker, and current-smoker, current-smokers or ex-smokers cannot transition to a never-smoker at a subsequent wave. These longitudinal variables often contain missing values, however, there is little guidance on whether these restrictions need to be accommodated when using multiple imputation methods. Multiply imputing such missing values, ignoring the restrictions, could lead to implausible transitions.

Methods

We designed a simulation study based on the Longitudinal Study of Australian Children, where the target analysis was the association between (incomplete) maternal smoking and childhood obesity. We set varying proportions of data on maternal smoking to missing completely at random or missing at random. We compared the performance of fully conditional specification with multinomial and ordinal logistic imputation, and predictive mean matching, two-fold fully conditional specification, indicator based imputation under multivariate normal imputation with projected distance-based rounding, and continuous imputation under multivariate normal imputation with calibration, where each of these multiple imputation methods were applied, accounting for the restrictions using a semi-deterministic imputation procedure.

Results

Overall, we observed reduced bias when applying multiple imputation methods with restrictions, and fully conditional specification with predictive mean matching performed the best. Applying fully conditional specification and two-fold fully conditional specification for imputing nominal variables based on multinomial logistic regression had severe convergence issues. Both imputation methods under multivariate normal imputation produced biased estimates when restrictions were not accommodated, however, we observed substantial reductions in bias when restrictions were applied with continuous imputation under multivariate normal imputation with calibration.

Conclusion

In a similar longitudinal setting we recommend the use of fully conditional specification with predictive mean matching, with restrictions applied during the imputation stage.

Electronic supplementary material

The online version of this article (10.1186/s12874-018-0653-0) contains supplementary material, which is available to authorized users.

Maternal smoking during pregnancy and offspring body mass index and overweight: a systematic review and meta-analysis

Abstract: The present study aimed to conduct a systematic review and meta-analysis to evaluate the evidence on the association of maternal smoking during pregnancy with offspring body composition in childhood, adolescence and adulthood. MEDLINE, Web of Science and LILACS databases were searched. Reference lists were also screened. We included original studies, conducted in humans, that assessed the association of maternal smoking during pregnancy with offspring body mass index (BMI) and overweight in childhood, adolescence and adulthood, published through May 1st, 2018. A meta-analysis was used to estimate pooled effect sizes. The systematic review included 64 studies, of which 37 evaluated the association of maternal smoking during pregnancy with overweight, 13 with BMI, and 14 evaluated both outcomes. Of these 64 studies, 95 measures of effect were extracted and included in the meta-analysis. We verified that the quality of evidence across studies regarding maternal smoking in pregnancy and overweight and BMI of offspring to be moderate and low, respectively. Most studies (44 studies) were classified as moderate risk bias. Heterogeneity among studies included was high and, in the random-effects pooled analysis, maternal smoking during pregnancy increased the odds of offspring overweight (OR: 1.43, 95%CI: 1.35; 1.52) and mean difference of BMI (β: 0.31, 95%CI: 0.23; 0.39). In conclusion, offspring of mothers who smoked during pregnancy have higher odds of overweight and mean difference of BMI, and these associations persisted into adulthood.

Intergenerational and early life influences on the well-being of Australian Aboriginal and Torres Strait Islander children: overview and selected findings from Footprints in Time, the Longitudinal Study of Indigenous Children

Footprints in Time: The Longitudinal Study of Indigenous Children (LSIC) is a national study of 1759 Australian Aboriginal and Torres Strait Islander children living across urban, regional and remote areas of Australia. The study is in its 11th wave of annual data collection, having collected extensive data on topics including birth and early life influences, parental health and well-being, identity, cultural engagement, language use, housing, racism, school engagement and academic achievement, and social and emotional well-being. The current paper reviews a selection of major findings from Footprints in Time relating to the developmental origins of health and disease for Australian Aboriginal and Torres Strait Islander peoples. Opportunities for new researchers to conduct further research utilizing the LSIC data set are also presented.

Fathers Count: The Impact of Paternal Risk Factors on Birth Outcomes

Objective To determine the contribution of paternal factors to the risk of adverse birth outcomes. Methods This is a retrospective cross-sectional analysis using birth certificate data from 2004 to 2015 retrieved from the Finger Lakes Regional Perinatal Data System. Primiparous women with singleton pregnancies were analyzed in the study. Two multivariate logistic regression models were conducted to assess potential paternal risk factors including age, race/ethnicity, and education on four birth outcomes, including preterm birth (PTB), low birthweight (LBW), high birthweight (HBW), and small for gestational age (SGA). Results A total of 36,731 singleton births were included in the analysis. Less paternal education was significantly related to an elevated risk of PTB, LBW, and SGA, even after adjustment for maternal demographic, medical, and lifestyle factors (P < 0.05). Paternal race/ethnicity was also significantly associated with all four birth outcomes (P < 0.05) while controlling for maternal factors. Older paternal age was associated with increased odds (OR 1.012, 95% CI 1.003-1.022) of LBW. Maternal race/ethnicity partially mediated the association of paternal race/ethnicity with HBW and SGA. Maternal education partially mediated the relationship between paternal education and SGA. Conclusion Paternal factors were important predictors of adverse birth outcomes. Our results support the inclusion of fathers in future studies and clinical programs aimed at reducing adverse birth outcomes.

Body mass index trajectories of Indigenous Australian children and relation to screen time, diet, and demographic factors

OBJECTIVE

Limited cross-sectional data indicate elevated overweight/obesity prevalence among Indigenous versus non-Indigenous Australian children. This study aims to quantify body mass index (BMI) trajectories among Indigenous Australian children aged 3-6 and 6-9 years and to identify factors associated with the development of overweight/obesity.

METHODS

Three-year BMI change was examined in up to 1,157 children in the national Longitudinal Study of Indigenous Children. BMI trajectories among children with normal baseline BMI (n = 907/1,157) were quantified using growth curve models.

RESULTS

Baseline prevalences of overweight/obesity were 12.1% and 25.4% among children of mean age 3 and 6 years, respectively. Of children with normal baseline BMI, 31.9% had overweight/obesity 3 years later; BMI increased more rapidly for younger versus older (difference: 0.59 kg/m2 /year; 95% CI: 0.50-0.69), female versus male (difference: 0.15 kg/m2 /year; 95% CI: 0.07-0.23), and Torres Strait Islander versus Aboriginal (difference: 0.36 kg/m2 /year; 95% CI: 0.17-0.55) children. Results were consistent with less rapid rates of BMI increase for children with lower sugar-sweetened beverage (including fruit juice) and high-fat food consumption. Children's BMI was lower in more disadvantaged areas.

CONCLUSIONS

Overweight/obesity is common, and increases rapidly, in early childhood. Interventions are required to reduce the overweight/obesity prevalence among Indigenous Australian children in the first 3 years of life and to slow the rapid overweight/obesity onset from age 3 to 9 years.