Differences in Cognitive Aging Assessments Associated With Retest Effects
Yan Wang, James Russell Pike, Jennifer A. Deal, Frank R. Lin, Nicholas Salvatore Reed, Emmanuel E. Garcia-Morales, Alison R. Huang, Douglas A. Jabs, Alden L. GrossImportance
Retest effects can inflate longitudinal cognitive test performance and complicate estimation of cognitive change. Whether retest effects influence estimated intervention effects in randomized clinical trials remains unclear.
Objective
To quantify the magnitude and heterogeneity of differences associated with retesting in longitudinal cognitive testing and examine whether accounting for retesting alters estimated treatment effects in the Aging and Cognitive Health Evaluation in Elders (ACHIEVE) trial.
Design, Setting, and Participants
This cohort study was an ad hoc secondary analysis of ACHIEVE, a multicenter phase 3 randomized clinical trial conducted from 2017 to 2022 with 3-year follow-up. Community-dwelling older adults with untreated hearing loss were enrolled at 4 US community sites: Forsyth County, North Carolina; Jackson, Mississippi; Minneapolis, Minnesota; and Washington County, Maryland. Analyses were completed in November 2025.
Exposure
Cognitive retesting, defined as a binary indicator (0 for baseline visit; 1 for follow-up visits).
Main Outcomes and Measures
The primary outcome was a global cognitive factor score derived from a neurocognitive battery. Secondary outcomes included domain-specific factor scores (language, executive function, memory) and individual test scores. Retest differences were estimated using linear mixed-effects models adjusted for covariates consistent with the main ACHIEVE analysis. Interactions tested whether retest differences varied by key factors. Model-based trajectories and intervention-control estimates were compared with and without retest adjustment.
Results
A total of 977 older adults (mean [SD] baseline age, 76.3 [4.0] years; 523 [54%] female) contributed 2571 in-person cognitive assessments , including 552 adults (57%) with mild hearing loss and 739 adults (76%) recruited de novo. Retest differences were significant for global cognition (β = 0.11; 95% CI, 0.07 to 0.15), executive function (β = 0.15; 95% CI, 0.10 to 0.20), and memory (β = 0.12; 95% CI, 0.06 to 0.18), approximately offsetting 1 year of model-estimated age-related cognitive decline. Language showed minimal retest differences (β = −0.03; 95% CI, −0.08 to 0.02). Retest differences did not meaningfully vary by hearing loss severity, trial intervention, or recruitment source. Retest adjustment altered estimated cognitive trajectories but did not meaningfully alter intervention outcomes.
Conclusions and Relevance
This cohort study found that retest differences were substantial for global cognition, executive function, and memory but did not meaningfully alter estimated hearing intervention outcomes. Adjustment for retest differences is recommended to improve interpretation and intervention estimation in longitudinal cognitive studies.