Supplementary Figure 1. pcVPC of BMI final model (n=1000). Black dots represent BMI observed
measurements. Black lines represent the 5th, 50th and 95th percentiles of observed data. Grey lines and areas represent the 5th, 50th and 95th percentiles and corresponding 95% confidential intervals of simulated data. PI, prediction interval.
Supplementary Figure 2. pcVPC of WHR final model (n=1000). Black dots represent WHR observed
measurements. Black lines represent the 5th, 50th and 95th percentiles of observed data. Grey lines and areas represent the 5th, 50th and 95th percentiles and corresponding 95% confidential intervals of simulated data. PI, prediction interval.
Supplementary Figure 3. Goodness-of-fit plots of BMI structural population model. Loess smooth
curves of the ordinate values are printed in grey. a) Observations vs. individual predictions; line of identity is printed in black. b) Observations vs. population predictions; identity line is printed in black.
c) Conditional weighted residuals (CWRES) vs. population predictions; ordinate value zero is printed in black. d) CWRES vs. time after ART initiation (years); ordinate value zero is printed in black.
Supplementary Figure 4. Goodness-of-fit plots of BMI final covariate population model. Loess
smooth curves of the ordinate values are printed in grey. a) Observations vs. individual predictions;
line of identity is printed in black. b) Observations vs. population predictions; identity line is printed in black. c) Conditional weighted residuals (CWRES) vs. population predictions; ordinate value zero is printed in black. d) CWRES vs. time after ART initiation (years); ordinate value zero is printed in black.
Supplementary Figure 5. Goodness-of-fit plots of WHR structural population model. Loess smooth
curves of the ordinate values are printed in grey. a) Observations vs. individual predictions; line of identity is printed in black. b) Observations vs. population predictions; identity line is printed in black.
c) Conditional weighted residuals (CWRES) vs. population predictions; ordinate value zero is printed in black. d) CWRES vs. time after ART initiation (years); ordinate value zero is printed in black.
Supplementary Figure 6. Goodness-of-fit plots of WHR final covariate population model. Loess
smooth curves of the ordinate values are printed in grey. a) Observations vs. individual predictions;
line of identity is printed in black. b) Observations vs. population predictions; identity line is printed in black. c) Conditional weighted residuals (CWRES) vs. population predictions; ordinate value zero is printed in black. d) CWRES vs. time after ART initiation (years); ordinate value zero is printed in black.
Supplementary Figure 7. Predicted BMI measurements at baseline, 1.5, 3, 5 and 10 years after ART
initiation for the BMI study population (n=1090). The predicted percentage of overweight or obese individuals (according to the WHO classification, BMI ≥ 25 kg/m2) at baseline, 1.5, 3, 5 and 10 years after ART initiation is 30.4, 38.1, 39.4, 41.1 and 47.8% respectively.
Supplementary Figure 8. Predicted WHR measurements at baseline, 2.5, 4 and 10 years after ART
initiation for the WHR study population (n=1000). The predicted percentage of men with abdominal obesity (according to the WHO classification, WHR >0.9) at baseline, 2.5, 4 and 10 years after ART initiation is 52.5, 62.7, 69.5, and 80.6% respectively (left panel). The predicted percentage of women with abdominal obesity (according to the WHO classification, WHR >0.85) at baseline, 2.5, 4 and 10 years after ART initiation is 43.1, 47.0, 52.0 and 65.8% respectively (right panel).