Table 1.
Construct | Measure | Collection | Brief protocol | |
---|---|---|---|---|
Exposures | LSAC (Waves 2–8) | Check Point (Wave 6.5) | ||
Disadvantage | Neighbourhood | • | Neighbourhood disadvantage was measured with the census-based Socio-Economic Indexes for Areas (SEIFA) Index of Relative Socioeconomic Disadvantage (IRSD) which is updated every four years based on Australian Census data.18 The IRSD is based on the statistical area (SA1) where the child's family live and is a weighted combination of census-collected variables that indicate social and material disadvantage of the neighbourhood (e.g., % of people unemployed, % of occupied private dwellings with no cars). Because the study period spans three census waves (2006, 2011, 2016), the IRSD census variables and the SA1 boundaries are slightly different over time. At each time point, scores are standardised to have a mean of 1000 (national average) and a standard deviation of 100; low scores indicate high disadvantage, with higher scores indicating less disadvantage. In the main analyses, IRSD is converted into cohort-specific quintiles with 1 being high neighbourhood disadvantage, and 5 being low neighbourhood disadvantage (i.e., better/advantaged IRSD). Guided by previous research,19 for Aim 2 analysis, we then categorised as ‘most disadvantaged’ (quintile 1–2), ‘average’ (quintile 3), or ‘least disadvantaged’ socioeconomic conditions (quintile 4–5). |
|
Family | • | Family disadvantage (socioeconomic position (SEP))20 was a composite of parent/adult-reported combined household income, current or most recent occupation of each parent, and highest achieved educational qualification of each parent. Each component was scaled, and an unweighted average calculated over 3 values in a single-parent household or over 5 values in a dual-parent household. The unweighted average variable at each LSAC wave was standardised within the wave (mean 0, SD 1); low z-scores indicate high disadvantage, with higher z-scores indicate less disadvantage. In the main analyses, SEP is converted into cohort-specific quintiles with 1 being high family disadvantage, and 5 being low family disadvantage (i.e., better/advantaged SEP). For Aim 2 analysis, we then categorised as ‘most disadvantaged’ (quintile 1–2), ‘average’ (quintile 3), or ‘least disadvantaged’ socioeconomic conditions (quintile 4–5). |
||
Genetic Risk | Polygenic risk score (PRS) for BMI | • | Participants' genetic samples were isolated using Illumina Infinium® Global Screening Array-24 v1.0. Following which the PRS was developed with a scoring algorithm derived from a genome-wide association study (GWAS) of 340,000 participants using Khera et al.‘s methods.21 To create the PRS,22 we matched 2,048,277 individual nucleotide polymorphisms from CheckPoint participants to the GWAS, and then multiplied the number of risk alleles for these variants by estimated effect sizes. These values were summed to create a PRS for each participant that is a proxy for polygenic risk for high BMI in the past, present, and future. Following the creation of the PRS, we derived genetic principal components from a Principal Components Analysis of the CheckPoint genetics dataset. Because the PRS was created with summary data from European populations,21 to derive variables to control for population structure (i.e., systematic differences between different sub-populations within the cohort, see Supplementary Table S1 for further detail) for our causal analysis (Aim 2), we selected the top five components based on a scree plot and initial regression models showing that further components did not contribute to the models.22 This PRS accounts for 12% of the variance in BMI z-scores within CheckPoint children (11–12 years), and 9% of the variance in CheckPoint adults BMI.22 For Aim 1 analyses, PRS is converted to quintiles with 1 being lower risk, and 5 being higher risk; for Aim 2 analysis the PRS is divided at the median to create two groups indicating lower vs. higher polygenic risk. | |
Outcomes | LSAC (Waves 2–8) | Check Point (Wave 6.5) | ||
Body Mass Index Overweight and Obesity Status |
Objective height & weight (child) Self-reported height & weight (adult/parent) |
• | During LSAC home visits children's height and weight were measured by trained interviewers. Weight was measured in light clothing without shoes using HoMedics digital scales (Waves 2–3), or Tanita body fat scales (Waves 4–8). Height was measured to the nearest 0.1 cm using an Invicta stadiometer (Waves 2–3), or a laser stadiometer (Waves 4–8). Two measurements were taken, and a third if these differed by > 0.5 cm; the average of the two closest measures was used. Adults' height and weight were self-reported at all waves. For the main analysis raw BMI was calculated as weight(kg)/(height(m)2) and for children we also provide z-scores for sample characteristics (Table 2) according to the US Centers for Disease Control (CDC) reference values. For main analysis, CDC cut-offs were used to determine overweight/obesity among children/adolescents at the ≥85th percentile; among adults overweight/obesity was BMI ≥25 kg/m2. |
Abbreviations: LSAC: Longitudinal Study of Australian Children; SEP: socio-economic position; SEIFA: Socio-Economic Index for Areas; IRSD: Index of Relative Socioeconomic Disadvantage; PRS: polygenic risk score; GWAS: Genome-Wide Association Study; CDC: Center for Disease Control.