Abstract
Objective: To detect treatment effects in severe patients with Alzheimer’s disease (AD) using a composite score of Severe Impairment Battery (SIB) items. Background: While the SIB detects treatment effects in patients with moderate-to-severe AD who perform at floor level on other cognitive scales, traditional total SIB scoring may lack sensitivity in severe AD. Design/Methods: A quantitative-SIB composite score (qSIB-total) and a severe SIB composite score (qSIB-severe) were developed using partial least squares (PLS) regressions and tested in separate datasets (Training-dataset and Test-dataset, respectively) utilizing an MMRM approach. The Training-dataset (N=966) was pooled from three 24-week phase 3 memantine trials in moderate-to-severe AD patients. A composite of weighted SIB items (qSIB-composite) with variable importance projection (VIP) ≥0.80 was derived using PLS regression. The qSIB-composite and qSIB-severe were tested in the Test-dataset (N=661) from an independent study of memantine in moderate-to-severe AD patients using MMRM regression. Results: For all patients, the SIB total score identified significant treatment effects in memantine-vs placebo-treated patients at weeks 18 and 24 (P<0.05). The qSIB-total scores did not improve sensitivity. The qSIB-severe composite included 5 items (language, memory, praxis, attention, and orientation) with weights of 0.009, 0.003, 0.021, 0.020, and 0.044, respectively, and a minimum VIP=0.8101. Compared with SIB total score, the qSIB-severe score improved sensitivity (lower P-values) in severe AD patients (baseline MMSE<9) at weeks 8 (qSIB-composite, P=0.6304; total SIB, P=0.7057), 12 (P=0.6305; P=0.8403), 18 (P=0.0204; P=0.0491), and 24 (P=0.002; P=0.0241). Conclusions: In severe AD patients, qSIB-severe provided additional measurement sensitivity and differentiation between memantine- and placebo-treated versus total SIB score. The qSIB-total did not improve sensitivity, indicating that SIB total score is optimized for combined moderate-to-severe patients. Development and utilization of quantitatively optimized composite scales from established clinical trial measures may improve scale sensitivity in patient subgroups, particularly when floor effects are present.