Efficacy and Safety of Mobile App–Based Metamemory Cognitive Training for Mild Cognitive Impairment: Multicenter Randomized Clinical Trial

Background Metamemory training (MMT) offers a potential nonpharmacological approach to enhance cognitive function in individuals with mild cognitive impairment (MCI). While digital cognitive training improves accessibility, the effectiveness of mobile app–based MMT has not been evaluated in a randomized clinical trial. Objective We aimed to evaluate the efficacy and safety of a mobile app–based MMT program, ET-101 (Cogthera), compared to a sham device control group in individuals with MCI. Methods This multicenter, randomized controlled trial enrolled participants with MCI, recruited from 7 medical centers, and randomly assigned them to the ET-101 or control group (1:1 ratio). The intervention lasted 12 weeks, with a 12-week follow-up. The ET-101 group received metamemory-based multimemory strategy training and real-time feedback. Assessments of cognition, the daily activities of living, and the quality of life were conducted at baseline, week 12, and week 24. The primary outcome was the proportion of participants who showed cognitive improvement as assessed by the Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog)-14 at weeks 12 and 24. Secondary outcomes included changes in the scores of scales assessing cognition, daily activities, and quality of life. Safety analysis assessed adverse events and their relation to digital therapeutics. Results In the full analysis set, 49 participants were included in the ET-101 group and 50 in the control group. At week 24, the proportion of responders who maintained or improved their ADAS-Cog-14 scores was significantly higher in the ET-101 group than in the control group (P=.002). Additionally, the ET-101 group showed a significant improvement in ADAS-Cog-14 scores at week 24 compared to baseline levels (estimates=–2.53; t265=–3.05; Bonferroni-adjusted P=.003). A subdomain analysis revealed significant improvements in the memory (estimates=–2.50; t264=–4.03; Bonferroni-adjusted P