Table 2 Model terms included in 6 analysis models
Model | Model terms | Note |
|---|---|---|
pMMRM | Treatment, baseline value, randomization stratification variablesa, geographical region, visit, treatment-by-visit interaction | This model is defined in the study SAP. Visit is a discrete model term. |
aMMRM1 | Treatment, baseline value, randomization stratification variables, geographical region, visit, treatment-by-visit interaction, baseline-by-visit interaction | Compared to pMMRM, this model has one additional term: baseline-by-visit interaction. Including baseline-by-visit interaction in the MMRM analysis has been the current practice [9, 10]. |
DPM | Treatment, baseline value, randomization stratification variables, geographical region, week | Week is included as a discrete term. This model assumes the DPR; the rate of the decline of the treatment group to the rate of decline of the placebo group is constant across study weeks. The model is valid and likely more powerful than MMRM if the assumption is true. |
NCS | Treatment, baseline value, randomization stratification variables, geographical region, treatment-by-visit interaction, natural cubic spline (NCS), treatment-by-NCS interaction, baseline-by-NCS interaction | Similar to aMMRM but with visit replaced by a natural cubic spline with 2 degrees of freedom (as a continuous function of week). This model assumes non-linear non-parametric disease progression over time. |
QMM | Treatment, baseline value, randomization stratification variables, geographical region, week, treatment-by-week interaction, baseline-by-week interaction, weekaweek (week2), treatment-by-week2 interaction, baseline-by-week2 interaction | Similar to aMMRM terms but “week” in place of visit with week being included as a continuous term. This model assumes quadratic disease progression over time. |
aMMRM2 | Treatment, baseline value, randomization stratification variables, geographical region, visit, treatment-by-visit interaction, baseline-by-visit interaction, ApoE4 -by-visit interaction, ApoE4-by-treatment interaction, ApoE4-by-treatment-by-visit interaction | Compared to aMMRM1, this model has all 2-way interaction terms and 3-way interaction terms among treatment, ApoE4 status, and visit. This model is good to explore the impact of ApoE4 on treatment comparison when ApoE4 is not balanced between the comparison groups. |