QUALITY ASPECTS OF A NOVEL COGNITIVE-FUNCTIONAL COMPOSITE FOR THE MEASUREMENT OF DISEASE PROGRESSION IN ALZHEIMER’S DISEASE: FEASIBILITY, TEST-RETEST RELIABILITY AND PRACTICE EFFECTS

Abstract

To improve the detection of clinically relevant changes in mild cognitive impairment (MCI) and mild Alzheimer's disease (AD), we designed a novel composite measure: the Cognitive-Functional Composite (CFC). As a first validation step, we investigated its feasibility, test-retest reliability and potential practice effects. We recruited patients with MCI or mild dementia due to AD (MMSE ≥18) from four memory clinics in The Netherlands and Scotland (n=48). We included cognitively healthy subjects as control group (n=30). The CFC consists of tests that focus on episodic memory (ADAS-Cog Word Recognition, ADAS-Cog Orientation and ADAS-Cog Word Recall), working memory (Digit Span Backwards) and executive functioning (Category Fluency Test, Controlled Oral Word Association Test and Digit Symbol Substitution Test (DSST)), and an everyday functioning questionnaire (Amsterdam Instrumental Activities of Daily Living Questionnaire, A-IADL-Q). At baseline and after 2–3 weeks follow-up, all subjects underwent the cognitive tests whilst their study partners completed the A-IADL-Q. Feasibility was investigated by interviewing subjects after follow-up. We compared baseline and follow-up scores using paired t-tests with Bonferroni correction for multiple testing, and test-retest reliability with intraclass correlation coefficients (ICC) using a two-factor mixed model and type absolute agreement. Additionally, we applied the Bland-Altman method to explore systematic differences that could point towards practice effects. The patient group (40% female, age 69.9 (SD=7.4), MMSE 25.3 (SD=3.3)) was slightly older than the control group (50% female, age 65 (SD=7.1), p=.006). Overall, subjects experienced test content and materials as feasible. For patients, we only found a significant higher score at follow-up (M=2.4, SD=5.4, corrected p=.04) for the DSST, despite a high absolute agreement (ICC=.881, p<.001). The Bland-Altman plot did not show systematic differences (Figure 1). For all subtests, we found moderate to high test-retest reliability in patients (ICCs ranging from .551 to .945). ICCs were substantially lower in controls (ranging from .048 to .650), probably due to ceiling effects.