# Table 3 Associations between participant’s size at birth and MRI percent water

Relative change in MRI percent water, geometric meansa (95 % CI)
Complete datab Imputed datac (n = 491)
Absolute size vs. rate of growth n = 455
Model 1 Birthweight (per 1 SD 472.6 g) 1.03 (1.02–1.05) 1.03 (1.02–1.05)
Model 2 Birthweight (per 1 SD 472.6 g) 1.04 (1.02–1.06) 1.04 (1.02–1.06)
Gestational age (weeks) <39 1 (ref) 1 (ref)
39 0.97 (0.92–1.02) 0.97 (0.92–1.02)
40 0.97 (0.92–1.02) 0.97 (0.92–1.02)
41+ 0.96 (0.92–1.01) 0.96 (0.92–1.01)
LR test/Wald test p valued   0.519 0.477
Which measure best captures linear (skeletal) growth? n = 356
Birth length (per 1 SD 2.3 cm) 1.00 (0.98–1.02) 1.01 (0.99–1.03)
Head circumference (per 1 SD 1.2 cm) 1.02 (1.00–1.05) 1.02 (1.00–1.04)
Linear growth vs. adiposity n = 361
Model 1 Birthweight (per 1 SD 472.6 g) 1.03 (1.01–1.05) 1.03 (1.02-1.05)
Model 2 Birthweight (per 1 SD 472.6 g) 1.03 (1.00–1.06) 1.03 (1.01–1.06)
Head circumference (per 1 SD 1.2 cm) 1.01 (0.98–1.03) 1.00 (0.97–1.03)
LR test/Wald test p valued (n = 353) 0.671 0.917
Model 1 Birthweight (per 1 SD 472.6 g) 1.03 (1.01–1.05) 1.03 (1.02–1.05)
Model 2 Birthweight (per 1 SD 472.6 g) 1.03 (1.01–1.05) 1.03 (1.02–1.05)
Ponderal Index (per 1 SD 4.1 g/cm3) 1.01 (0.99–1.02) 1.00 (0.99–1.02)
LR test/Wald test p valued 0.577 0.654
1. Abbreviations: MRI Magnetic resonance imaging, LR Likelihood ratio test, ref Reference category
2. aMRI percent water was log-transformed for the analysis, and exponentiated estimated regression parameters, with 95 % CIs calculated by exponentiating the original 95 % CIs, are presented. Models adjusted for age, BMI z-score and menstrual phase/hormonal contraceptive use at the time of MRI scan. Bold indicates 95 % CI do not cross the null (1.00)
3. bAnalysis restricted to those with non-missing data for all variables included in each model
4. cSee Statistical methods section of main text
5. dLR test performed on the complete record data, while a Wald test was performed on the imputed data (and summarised using Rubin’s rule), to test the null hypothesis that the inclusion of the additional variable in model 2 did not improve the fit to the data