Can brain training prevent dementia? Long-term trial shows speed training with boosters makes a difference

By Vijay Kumar MalesuReviewed by Susha Cheriyedath, M.Sc.Mar 2 2026

A landmark 20-year analysis of the ACTIVE trial suggests that targeted, reinforced speed-based cognitive training may delay dementia diagnosis, offering new insight into how structured mental exercises could support long-term brain health in aging populations.

Study: Impact of cognitive training on claims-based diagnosed dementia over 20 years: evidence from the ACTIVE study. Image Credit: Oksana Tkachova / Shutterstock

In a recent study published in the journal Alzheimer's & Dementia: Translational Research & Clinical Interventions, a group of researchers determined whether domain-specific cognitive training reduces the long-term risk of Alzheimer’s disease and related dementias (ADRD) over 20 years using Medicare claims data.

Background and Rationale

Nearly half of Americans over age 85 are expected to develop dementia during their lifetime, but a question arises: can we train the brain to resist decline? Cognitive training programs promise sharper memory and faster thinking, but debate continues over whether such improvements translate into real-world protection against ADRD. Improvements in thinking, memory, and processing speed have been well documented in the short term, but the long-term prevention of dementia remains uncertain. The study of how mental exercises affect clinical diagnoses of dementia has major implications for changing government policy related to aging, caregiving, and health care costs; thus, more research is needed to identify the types of exercises that offer a long-term basis for protection.

Study Design and Methods

The Advanced Cognitive Training for Independent and Vital Elderly study was a four-arm, multisite, single-blind randomized controlled trial that enrolled 2,802 community-dwelling adults aged 65 years and older between 1998 and 1999. Four groups were created: memory training, reasoning training, processing speed training, and a no-contact control group, and participants were assigned accordingly.

To be eligible, participants must have scored 23 or more on the Mini-Mental State Examination and have had independence with regard to all activities of daily living. People who had a recent stroke within the prior 12 months, were undergoing cancer treatment (chemotherapy or radiation), or had any sensory impairments that would interfere with training participation were excluded from the sample population.

In this analysis, participants' data were linked to Medicare claims from January 1, 1999, to December 31, 2019. The final analytic sample included 2,021 individuals enrolled in traditional Medicare at baseline. ADRD was identified using the Chronic Conditions Warehouse algorithm based on International Classification of Diseases codes. Cause-specific Cox proportional hazards models for the risk of dementia diagnosis, with hazard ratios adjusted for competing risks of death, were estimated and adjusted for age, sex, race, education, marital status, cardiovascular comorbidities, smoking status, and baseline cognitive scores, with additional adjustment for study site and training wave. The booster training sessions (offered at 11 months and 35 months) were analyzed separately among participants who completed at least 8 of the initial 10 training sessions and were therefore eligible for booster randomization.

Long-Term Dementia Risk Results

Over 20 years of follow-up, 48.7% of participants in the control group received a diagnosis of ADRD. Death rate was high across all groups, with 77% dying during follow-up, reflecting the advanced age of the cohort. Baseline demographic and health characteristics were balanced across intervention arms.

When examining the first round of assessments alone, none of the three training arms showed a statistically significant decrease in dementia risk compared to the control group, after adjusting for covariates. There were some indications of a small reduction in risk, approximately 12–15% lower, as suggested by the hazard ratio, but again, none were statistically significant.

The most notable finding emerged when booster sessions were considered. Participants assigned to speed-of-processing training who were randomized to receive booster training experienced a statistically significant 25% lower risk of diagnosed ADRD compared with the control group (adjusted hazard ratio 0.75; 95% confidence interval 0.59 to 0.95). In contrast, speed-trained participants who did not receive booster sessions showed no protective benefit (hazard ratio 1.01; 95% confidence interval 0.81 to 1.27).

Within the speed training arm, participants assigned to booster training had a lower, borderline statistically significant risk compared with those eligible for but not assigned to boosters (hazard ratio 0.81; 95% confidence interval 0.66 to 1.00). Therefore, the results of this study suggest that reinforcement sessions may enhance or sustain training effects, though these findings should not be interpreted as definitively causal, given that booster eligibility required post-randomization session completion and may introduce selection bias. It was also observed that training focused on memory or reasoning skills did not reduce the risk of dementia, independent of booster participation.

Age did not significantly change the effects of training, but younger participants in the memory arm showed a trend toward lower dementia risk, and this association was not statistically significant. Fine-Gray competing risk models produced similar results.

Real-World Implications

From a real-world perspective, these findings are meaningful. Speed-of-processing training emphasized visual attention and rapid information processing, particularly divided attention, skills closely tied to daily tasks such as driving. Previous analyses of the same cohort showed reduced at-fault motor vehicle collisions among speed-trained participants, reinforcing the practical value of this intervention. The current results suggest that sustained, adaptive training targeting attention and processing speed may not only improve everyday functioning but also be associated with delayed clinical dementia diagnosis, though the study relied on claims-based diagnoses rather than adjudicated clinical assessments and may underestimate or misclassify true dementia cases depending on health care utilization and diagnostic coding practices.

Conclusions

Over two decades of follow-up, cognitive training focused on speed of processing, particularly when reinforced with booster sessions, was associated with a significantly reduced risk of ADRD. Memory and reasoning training did not demonstrate comparable long-term protection. These findings suggest that adaptive, attention-based cognitive exercises may help delay dementia diagnosis in older adults. While not a cure, such interventions could extend years of independence and reduce societal burden. However, because the outcome was based on Medicare claims and the analytic sample excluded individuals enrolled in Medicare Advantage at baseline, generalizability may be limited, and further confirmation using clinically adjudicated outcomes is warranted.