Differential responsiveness of outcome measures according to biomarker inclusion criteria: implications for trial design

Background Accurate selection of outcome measures, sensitive to the earliest cognitive changes, is essential to demonstrate treatment efficacy in secondary prevention trials in preclinical Alzheimer’s disease(AD). These outcome measures may differ according to biomarker inclusion criteria. We aimed to determine sensitivity to change over time, of commonly used cognitive measures. Method Participants with longitudinal neuropsychological assessments with available baseline biomarkers were included, from the SCIENCe project. We created different (overlapping) target groups based on inclusion criteria in pre‐clinical AD trials; (1)amyloid PET or CSF (A+, n = 114), (2)CSF p‐tau (T+, n = 124), (3)AT(N)+ (n = 41), (4)APoE4 carriers(E4+, n = 195) and (5)APoE4 carriers with amyloid(E4+A+,n = 65). As a control population, we included 150 controls with normal biomarkers. Linear mixed models, adjusting for age, sex and education, were used to investigate the sensitivity to change of 14 individual (standardized) neuropsychological tests and two composites with time as the variable of interest. Composites were based on Preclinical Alzheimer Cognitive Composite; PACC(1) and the Cognitive‐Functional Composite; CFC(2). Result A total of 556 participants where included(Age = 61.7±8.7, 44%F, follow‐up 3.1y ±2.7y)(Table 1). Linear mixed models in controls showed improvement over time in several tests, perhaps reflecting a learning effect: Stroop II/III, Wordlist Immediate Recall, Wordlist Delayed Recall and Letter Fluency. In contrast, all target groups showed significant decline in MMSE(Beta[‐0.09–0.26]), Digit‐Span‐Backwords (Beta[‐0.04‐ ‐0.06]), Trailmaking‐A (Beta[‐0.06–0.19]),‐B (Beta[‐0.15–0.29]), Stroop‐I (Beta[‐0.04–0.09]), Word‐List‐Recognition (Beta[‐0.09–0.16]) and Category‐Fluency (Beta[‐0.06–0.12])(Table 2). Linear mixed models showed a steeper decline on tests in AT(N)+ group. The composite measures showed significant decline in A+, AT(N)+ and E4+A+, (Beta[‐0.08–0.26]). With a sensitivity analyses we limited the follow up time to 2 years, leading to fewer neuropsychological tests showing decline and no decline in the composite measures. Conclusion We demonstrated that in preclinical AD, commonly‐used neuropsychological tests capture cognitive decline dependent on biomarker profiles. Tests for memory, attention and executive functioning were sensitive to decline, as were composite measures. When follow‐up time was limited to 2 years, fewer tests were sensitive to change including the composite measures. Therefore, careful selecting cognitive outcome measures and an appropriate period of time is important when defining the target population for the clinical trial.