top of page

New Study: Brain Training and Dementia

17 hours ago
6 min read

Does brain training prevent dementia? The largest randomized trial ever run on cognitive training followed its participants for 20 years, and it found a lower rate of dementia diagnosis in one group: people assigned to speed-of-processing training who were also assigned to booster sessions. The same training without boosters showed no reduction. That distinction is the whole finding, and it goes missing in most accounts of it.


An older couple seen from behind, walking together on a path through an autumn park.

The Trial, and Its Four Groups


ACTIVE, Advanced Cognitive Training for Independent and Vital Elderly, was a randomized controlled trial funded by the National Institute on Aging. It enrolled 2,802 community-dwelling older adults in 1998 and 1999 and randomly assigned them to one of four groups: memory training, reasoning training, speed-of-processing training or a no-contact control group.


Initial training ran in small groups over five to six weeks, up to 10 sessions of 60 to 75 minutes each. Participants who completed at least eight of those 10 sessions became eligible for booster training, and half of the eligible group was randomly assigned to receive it. Boosters came at 11 months and again at 35 months after baseline, up to four 75-minute sessions at each point.


That design is the reason this trial can answer a question most cognitive training research cannot. The groups were assigned at random, the follow-up is long, and the analysis was preregistered at AsPredicted.org under number 67523 before the data were examined.


Twenty Years Later


A 20-year analysis published in Alzheimer's & Dementia: Translational Research & Clinical Interventions in February 2026, with Norma B. Coe of the University of Pennsylvania as first author and Marilyn S. Albert of the Johns Hopkins School of Medicine as corresponding author, linked 2,021 participants to Medicare claims and looked for diagnoses of Alzheimer's disease and related dementias.


Here is the result, group by group, as adjusted hazard ratios with 95% confidence intervals, measured against the control group. A hazard ratio below 1.00 means a lower rate of diagnosis, and a confidence interval that crosses 1.00 means the result did not reach statistical significance.


  • Speed-of-processing training with boosters: 0.75, 95% CI 0.59 to 0.95

  • Speed-of-processing training without boosters: 1.01, 95% CI 0.81 to 1.27

  • Memory training with boosters: 0.86, 95% CI 0.68 to 1.07

  • Memory training without boosters: 0.84, 95% CI 0.66 to 1.06

  • Reasoning training with boosters: 0.83, 95% CI 0.66 to 1.04

  • Reasoning training without boosters: 0.96, 95% CI 0.76 to 1.21


One line in that set reached significance. The authors state plainly that there was no main effect of memory or reasoning training on risk of ADRD.


There is a second number worth having, and it is the one that gets skipped most often. When each training group is compared against control as a whole, without splitting by booster assignment, no group reached significance: memory 0.85, 95% CI 0.70 to 1.02; speed 0.87, 95% CI 0.72 to 1.05; reasoning 0.88, 95% CI 0.73 to 1.07. The headline result lives inside a subgroup, and the subgroup is defined by booster assignment.


Booster Sessions


In the speed-of-processing arm, 105 of 264 participants assigned to boosters received a dementia diagnosis over 20 years. In the control group, 239 of 491 did, which is 48.7%.


Set that against the arm that trained and received no boosters: a hazard ratio of 1.01 with a confidence interval running from 0.81 to 1.27. That is the same rate as doing nothing, measured in the same trial, on the same training, in the same population.


The authors also compared boosted participants against those who were eligible for boosters and were not randomized to them, which holds the completion effect constant. That comparison gave a hazard ratio of 0.81, 95% CI 0.66 to 1.00, with the upper bound sitting on the line.


Anyone repeating this result owes the reader the booster schedule. Up to eight reinforcement sessions, delivered at 11 months and 35 months, are part of the thing that was tested. A reader told that a few hours of training lowered dementia risk, with no mention of reinforcement nearly three years later, has been given a different intervention from the one the trial ran.


Medicare Claims


The outcome here is a diagnosis recorded in Medicare claims. The researchers say plainly that adjudicated diagnoses are not possible from claims data, so the analysis used a diagnostic-code-based algorithm, and they report its sensitivity and specificity at 85% to 90%. What an instrument reads is always part of what its number means, which is the subject of our page on measuring brain health.


The claims window ran from Jan. 1, 1999, to Dec. 31, 2019. Participants enrolled in Medicare Advantage at baseline were excluded, which removed 725 people, or 26% of the sample. The authors note that those participants tend to be healthier as a group and that the exclusion could have biased the results toward the null.


Their other stated limitations are worth carrying alongside the result. They write that although the intervention arms are balanced on observable characteristics, the potential for post-randomization selection bias remains, and they name the potential for a Type I error. The study was adjusted for age, race, sex, education, marital status, smoking status, baseline cognitive scores, study site and a range of cardiovascular conditions including ischemia, heart attack, atrial fibrillation, congestive heart failure, diabetes and hypertension.


The researchers and the journal are the authority on all of this.


Does Brain Training Prevent Dementia?


One form of training, reinforced over three years, was associated with a lower rate of dementia diagnosis in one randomized trial. The paper's own highlights put it as no prior cognitive training intervention having been shown to reduce risk of ADRD over a 20-year period, which makes it a first. The claim the evidence carries is that specific one: that arm, on that schedule, measured that way.


Brief online cognitive tests have started to offer more than a score. Some now return a risk profile, comparing a few minutes of performance against findings from a long-term study and telling the reader where they sit. A number produced that way deserves three questions, alongside the five questions worth asking about any validated cognitive test. First, which arm of which trial is the comparison drawn from, since a result from a reinforced subgroup does not describe an unreinforced one. Second, does the reader resemble the people in the study, which here means adults who were 65 and older in 1998. Third, was the comparison validated prospectively, meaning tested on people other than those it was built from.


A score on a brief cognitive test measures performance on that test at that moment. It is not a diagnosis, a prediction or a medical assessment of any individual.


Brain Training and Blood Biomarkers


Neurova Academy is facilitator-led cognitive training, a session a day for five days, about 30 minutes a session, with a certified facilitator in the room. ACTIVE studied speed-of-processing training, which is a different program, and the figures above describe that trial.


Programs running Neurova Academy can add the iXpressGenes CoreSignals Panel, which reads 22 markers of RNA gene expression from whole blood and is designed to track inflammatory and immune signals associated with stress and recovery. iXpressGenes describes its own work in three steps: detect, recommend and verify. A practitioner measures a starting point, guides an intervention, then confirms what changed. The third step is why we pair a training program with a certified laboratory.


Before a cell builds a protein it copies the instructions from the gene into RNA, so counting those copies reads what cells are doing at the moment of the draw. A protein level reads what has accumulated in the blood over time. Most of the RNA in whole blood comes from white blood cells, which makes a small draw a practical window onto immune and inflammatory activity. Expression also moves quickly, so a single draw is a snapshot, and the value comes from comparison: the same person, measured the same way, at planned points before and after.


In a 12-participant observational study tracking 22 RNA transcripts, with blood drawn at baseline, after the program and two weeks later, 58% of participants achieved meaningful inflammatory reduction, and strong responders showed a 62% to 86% reduction in inflammatory burden. The panel reads inflammation and immune regulation, so that is what those numbers describe. A controlled validation study is underway.


Our measurement page covers the panel in full, including the timescale it works on, our claims page carries these commitments, and a separate page sets out what FDA clearance covers and what general wellness means. If you find anything on this site that contradicts any of them, write to info@neurovalabs.com and we will fix the other one.


Neurova Academy is a general wellness and cognitive training program. It is not a medical device and is not offered to diagnose, treat, cure, or mitigate any disease or condition, including Alzheimer's disease, dementia, concussion, traumatic brain injury, or CTE. Outcomes described reflect observational pilot findings and are not a prediction of results for anyone else. Laboratory testing provided by iXpressGenes, a CLIA-certified, CAP-accredited laboratory.

Media contact: Nathaniel Free, chief communications officer, nathaniel.free@neurovalabs.com

Comments


Commenting on this post isn't available anymore. Contact the site owner for more info.
bottom of page