Skip to main content
Fig. 10 | BMC Medical Research Methodology

Fig. 10

From: Using linear and natural cubic splines, SITAR, and latent trajectory models to characterise nonlinear longitudinal growth trajectories in cohort studies

Fig. 10

Pooled mean BMC growth trajectory (solid black curves), mean BMC growth velocity (dashed blue curves), and mean age at peak BMC velocity (vertical red lines) from SITAR models applied to individual participant data (ALSPAC, BMDCS and PBMAS). Figure shows mean BMC growth trajectory (solid black curves), mean BMC growth velocity (dashed blue curves), and mean age at peak BMC velocity (vertical red lines) from SITAR models applied to individual participant data (ALSPAC, BMDCS and PBMAS) for females (a) and males (b). Sex-specific individual participant data SITAR models were fitted to ALSPAC, BMDCS and PBMAS combined, to obtained pooled estimates of the timing of peak BMC growth. This analysis included individuals with overlapping measurements (8.8 to 22.1 years) from the 3 cohorts (n=4431 for females and n=4359 for males.) To mitigate the cohort differences in BMC (higher values in ALSPAC due to Lunar machine), we modelled BMC in cohort-specific standardised units (mean=0 and SD=1), and the models were adjusted for cohort (as a fixed effect). Note it is not advised to fit SITAR to SD units as this distorts the underlying biology – though in our example, results are consistent with cohort-specific natural unit results

Back to article page