Statistical analysis

Continuous variables are reported as means and standard deviations (SD) or as medians and 25th and 75th percentiles, as appropriate. Categorical variables are reported as percentages. For each study visit, bivariate analysis was performed to compare characteristics of the two cohorts of patients with mild AD (21–26 MMSE points) and moderate AD (10–20 MMSE points). Bivariate comparisons involved using the t test for normally distributed variables, the Wilcoxon test for continuous parameters that were not normally distributed, or the χ² test for categorical variables.

Change from baseline across the AD severity groups was assessed for the EQ-5D-5L weighted index, the VAS score, DS, RAPA, and the ZBI total score by using the Wilcoxon test. The healthcare resource utilization variables were compared across AD severity.

Multivariable regression including mixed models were developed to explore the relationship between changes in the outcomes including patients’ HRQoL (patients’ EQ-5D index score assessed by caregivers), patients’ dependence (DS) and burden of caregiver (ZBI) during the first year of follow-up and several independent variables. The difference between baseline and 12-month outcomes was considered as the dependent variable in multivariable regression. This involved use of longitudinal data during the first year of data collection. Data after 12-month of follow-up were used for descriptive analysis but not included in the models due to low number of patients. The models adjusted for some time-invariant variables (e.g., age, gender, country, etc.) and clinical variables to examine the relationship between clinical variables and health outcomes. Even though first differences would eliminate the effects of time-invariant covariates, we kept some of the variables in the regression to take into account interaction between those covariates and time. The clinical variables included changes between baseline and the 12-month visit in MMSE score and ADAS-COG score, age (by terciles), gender, country, education level, (former) occupation, main (former) working status, height, relationship with caregivers and AD severity (mild versus moderate). A 12-month timeframe was used to ensure that sufficient patients were included in the follow-up analysis. Only those variables with p-value < 0.20 remained in the adjusted multivariate regression models. Akaike information criterion (AIC) was used to select the most appropriate models.

All analyses implemented were conducted by using SAS statistics software (SAS Institute Inc., Cary, NC), SAS enterprise guide, version 9.4. Statistical significance was set at p < 0.05.