Associations of age- and sex-standardized IGF z-scores with the flamingo test from logistic regression models.

*IGFBP-2 and insulin were natural log transformed before standardizing due to skewed distributions.†Sample sizes are based on complete-case analyses for all confounders: Boyd Orr IGFs n = 392, insulin n = 358; CaPS IGFs n = 725, insulin n = 959.‡Data were combined if there was no evidence the association differed between the cohorts.§An odds ratio (OR) >1 indicates a relative increase in odds of poor balance; an odds ratio ¶Minimal adjusted models control for age at performance test and sex (Boyd Orr); confounder adjusted models additionally control for research centre, adult social class, diet variables (milk intake, calcium, protein, fat and total energy), exercise, smoking, alcohol consumption; fully adjusted models additionally control for body mass index (BMI) at the time of IGF measurement and co-morbidities at the time of the physical performance tests.