Table 1.

Candidate linear mixed effect models in the simulation studies to measure treatment effects.

Models	Intercept and baseline FRS	Extra baseline variables	Time & time treatment interactions
Random forest predictions	$Y_{\mathit{ij}}‐Y_{i0}={\beta }_0+{\beta }_1{Y}_{i0}+$	${\beta }_2{\text{RFpred}}_{\mathit{ij}}$	$+{\mathrm{\Sigma }}_{t=1}^J{\alpha }_{t}I\left(t=j\right)+{\mathrm{\Sigma }}_{t=1}^J{\gamma }_{t}I\left(t=j\right)\times I\left(\text{Trt}\right)+{\epsilon }_{\mathit{ij}}$
Linear mixed effects model predictions		${\beta }_2{\text{LMMpred}}_{\mathit{ij}}$
Super learner predictions		${\beta }_2{\text{SLpred}}_{\mathit{ij}}$
Single best predictor		${\beta }_2{\text{preslope}}_i$
Standard baseline covariates		${\beta }_2{\text{SiteOnset}}_i+{\beta }_3\text{El}$
No extra covariates		–
Full model including random forest predictions, preslope, and standard baseline		${\beta }_2{\text{RFpred}}_{\mathit{ij}}+{\beta }_2{\text{preslope}}_i+{\beta }_4{\text{SiteOnset}}_i+{\beta }_5\text{El}$

For 6‐month studies, we consider follow‐up time to be baseline, Month 1, Month 2, Month 3, and Month 6. For 12‐month studies, we consider follow‐up time to be baseline, Month 1, Month 6, and Month 12. For 18‐month studies, we consider follow‐up time to be baseline, Month 1, Month 6, Month 12, and Month 18. The following models are estimated with unstructured correlation structure.