Longitudinal Participation and Cognitive Score Interpretation in the J-TRC Webstudy

INTRODUCTION: Web-based registries support remote cognitive assessment for Alzheimer's disease research, but participant-driven reassessment raises questions about long-term data yield and interpretation of repeated scores. We evaluated data contribution, selective observation, and retest-dependent estimates of cognitive change in the Japanese Trial-Ready Cohort (J-TRC) Webstudy. METHODS: We analyzed over six years of J-TRC Webstudy data (launched October 2019; registrations through March 31, 2025; assessments through May 31, 2026). We examined data contribution, inactivity, reactivation, score-associated observation, first-exposure contrasts, and sensitivity to retest models and estimators. RESULTS: Among 14,912 registered accounts, 10,863 participants were eligible; 10,645 (98.0%) completed the Cognitive Function Instrument (CFI), 6,419 (59.1%) completed Cogstate, and 5,792 (53.3%) contributed assessments on multiple days. Among those with sufficient opportunity, 19.4% and 14.1% contributed data at or beyond three and five years. After 365 days without assessment, the estimated probability of resuming within two years was 7.2%. Worse prior cognitive scores were associated with lower odds of subsequent assessment (adjusted odds ratios per standard deviation: CFI, 0.952; Cogstate, 0.845). By assessment 12, returnee changes were +0.43 and +1.14 favorable baseline standard deviations (CFI, Cogstate), versus −0.06 and +0.08 in first-exposure contrasts. Marginal CFI time estimates differed under categorical versus flexible retest adjustment (+0.080 versus −0.227 standard deviations/year). DISCUSSION: High initial completion was accompanied by limited long-term data contribution. Subsequent assessment was associated with prior scores, and estimated time trends varied by retest model. Reporting assessment histories and testing model sensitivity are important for interpreting repeated online cognitive scores.