Findings of Efficacy, Safety, and Biomarker Outcomes of Atabecestat in Preclinical Alzheimer Disease: A Truncated Randomized Phase 2b/3 Clinical Trial

Abstract

Importance: Atabecestat, a nonselective oral β-secretase inhibitor, was evaluated in the EARLY trial for slowing cognitive decline in participants with preclinical Alzheimer disease. Preliminary analyses suggested dose-related cognitive worsening and neuropsychiatric adverse events (AEs).

Objective: To report efficacy, safety, and biomarker findings in the EARLY trial, both on and off atabecestat treatment, with focus on potential recovery of effects on cognition and behavior.

Design, setting, and participants: Randomized, double-blind, placebo-controlled, phase 2b/3 study conducted from November 2015 to December 2018 after being stopped prematurely. The study was conducted at 143 centers across 14 countries. Participants were permitted to be followed off-treatment by the original protocol, collecting safety and efficacy data. From 4464 screened participants, 557 amyloid-positive, cognitively normal (Clinical Dementia Rating of 0; aged 60-85 years) participants (approximately 34% of originally planned 1650) were randomized before the trial sponsor stopped enrollment.

Interventions: Participants were randomized (1:1:1) to atabecestat, 5 mg (n = 189), 25 mg (n = 183), or placebo (n = 185).

Main outcomes and measures: Primary outcome: change from baseline in Preclinical Alzheimer Cognitive Composite score. Secondary outcomes: change from baseline in the Cognitive Function Index and the Repeatable Battery for the Assessment of Neuropsychological Status total scale score. Safety was monitored throughout the study.

Results: Of 557 participants, 341 were women (61.2%); mean (SD) age was 70.4 (5.56) years. In May 2018, study medication was stopped early owing to hepatic-related AEs; participants were followed up off-treatment for 6 months. Atabecestat, 25 mg, showed significant cognitive worsening vs placebo for Preclinical Alzheimer Cognitive Composite at month 6 (least-square mean difference, -1.09; 95% CI, -1.66 to -0.53; P < .001) and month 12 (least-square mean, -1.62; 95% CI, -2.49 to -0.76; P < .001), and at month 3 for Repeatable Battery for the Assessment of Neuropsychological Status (least-square mean, -3.70; 95% CI, -5.76 to -1.63; P < .001). Cognitive Function Index participant report showed nonsignificant worsening at month 12. Systemic and neuropsychiatric-related treatment-emergent AEs were greater in atabecestat groups vs placebo. After stopping treatment, follow-up cognitive testing and AE assessment provided evidence of reversibility of drug-induced cognitive worsening and AEs in atabecestat groups.

Conclusions and relevance: Atabecestat treatment was associated with dose-related cognitive worsening as early as 3 months and presence of neuropsychiatric treatment-emergent AEs, with evidence of reversibility after 6 months off treatment.

Trial registration: ClinicalTrials.gov Identifier: NCT02569398.

Conflict of interest statement

Conflict of Interest Disclosures: Dr Sperling has received research funding from Janssen, and has served as a consultant to AC Immune, Biogen, Eisai, Janssen, Roche, and Takeda. Drs Raman and Ernstrom have received research funding from the National Institutes of Health, Janssen, and Eli Lilly. Drs Henley, Shi, Karcher, Raghavan, Bogert, Tymofyeyev, Novak, Thipphawong, Saad, Kolb, Rofael, and Sanga are employees of Janssen Research & Development and may own stock/stock options in Johnson & Johnson. Drs Brashear, Romano, and Streffer were at Janssen Research & Development during the time the study was conducted and may own stock/stock options in Johnson & Johnson. Dr Donohue reported grants from Eli Lilly and the National Institutes of Health during the conduct of the study; grants and other support from Janssen Pharmaceuticals; and personal fees from Eli Lilly, Roche, Neurotrack, Biogen, and Vivid Genomics outside the submitted work; in addition, Dr Donohue is married to an employee of Janssen Research & Development who may own stock/stock options in Johnson & Johnson. Dr Aisen reported personal fees from Merck, Roche, Biogen, ImmunoBrain Checkpoint, and Samus and grants from Janssen Research & Development, Eli Lilly, and Eisai outside the submitted work. Dr Raman reported grants from Janssen during the conduct of the study and grants from Eli Lilly and Eisai outside the submitted work. Dr Streffer reported personal fees from UCB Biopharma outside the submitted work. Dr Romano reported personal fees from Janssen Research & Development outside the submitted work. No other disclosures were reported.

Figures

Figure 1.. Participant Disposition

AE indicates adverse event. aParticipants who discontinued study medication for other reasons before sponsor stopped atabecestat dosing. bParticipants who discontinued the study. Participants could remain in the study after stopping study medication. Most study discontinuations were due to the sponsor stopping the study.

Figure 2.. Least-Square Mean (SE) Difference of Preclinical Alzheimer Cognitive Composite (PACC) (A) and Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) (B) Decline (Intention-to-Treat [ITT] Analysis Set)

A, PACC is defined as the sum of transformed z scores for each of the 4 components of the measure: sum of free and total recall on the Free and Cued Selective Reminding Test, Delayed Paragraph Recall score on Logical Memory from the Wechsler Memory Scale, Digit-Symbol Coding Subtest (Wechsler Adult Intelligence Scale IV); and Mini-Mental State Examination (MMSE) total score. A decrease of 1 baseline standard deviation on each component would result in a 4-point decrease in the composite. B, RBANS consists of 12 subtests measuring the following 5 indices (domains): attention, language, visuospatial/construction, immediate memory, and delayed memory. aP < .001.

Figure 3.. Linear Mixed-Effect Model of Preclinical Alzheimer Cognitive Composite (PACC) Over Time

Mean change in PACC during the indicated epoch of follow-up estimated from a mixed-effect model controlling for follow-up prior to discontinuation (median, 4.7 months), follow-up after discontinuation (median, 3.25 months), total exposure (median, 5.3 months), sex, age, and education. In this model, time 0 is defined as the time of last dose. The dependent variable was PACC score at each visit with fixed effects for time (as a continuous variable) before and after last dose in each group, sex, age, education, and total exposure. Random effects were participant and country-specific random intercepts. The covariance structure was assumed to be autoregressive order one and heterogeneous with respect to time.