Study: 8 Weeks of Home Cognitive Training Improved Cognition in Older Adults With Diabetes at Higher Dementia Risk

Why diabetes raises the stakes for brain health

According to an increasingly dominant view in dementia research, better management of common modifiable risk factors could prevent more than a third of all dementia cases.^[2] Type 2 diabetes sits near the top of that risk list: population studies repeatedly link it to cognitive decline and to dementia-spectrum disorders.

There is a second, practical problem. Diabetes demands a daily self-management regimen, and subtle cognitive dysfunction in older adults with diabetes is common and tied to worse self-management. Cognition and disease control can drag each other down. This trial, run in the metropolitan Tel-Aviv area, was the first to test whether multidomain computerized cognitive training (CCT) could improve both cognition and diabetes self-management in this population.^[1]

What the study tested

Eighty-five community-dwelling older adults with diabetes (mean age 71.45, SD 4.85; 51 men; 16 years of education on average) enrolled; 84 were randomized in this single-blind trial. All had diabetes but no dementia or Alzheimer's diagnosis. Randomization (1:1) assigned them to one of two home-based programs, both on the commercially available multidomain CogniFit platform:^[1]

Tailored and adaptive CCT (TA-CCT, n = 44): tasks matched to each participant's cognitive profile, difficulty adapting to performance, feedback after every session. Generic CCT (G-CCT, n = 40): same tasks for everyone, fixed difficulty, scores shown only at baseline and end of training.^[1]

Both arms had the same prescribed dose: three training days per week for 8 weeks, two 10–15-minute sessions per day, 48 sessions, plus an optional 3-session booster at month 3. Each arm was further randomized to a global, cognition-specific, or no self-efficacy intervention (fortnightly phone calls plus two short videos).

The primary outcome was a global cognition composite across a named battery: Mini-Addenbrooke's Cognitive Evaluation, L'Hermitte Board, Logical Memory test, Rey Auditory Verbal Learning Test, Rey figure copy, Verbal Fluency, Digit Span, Digit-Symbol Coding, Boston Naming Test, and Trail-Making Task. The main secondary outcome was the Diabetes Self-Management Questionnaire (DSMQ). Psychologists blind to allocation assessed at baseline, post-intervention, and 6 months.

What it found

Cognition improved, and kept improving. Across both arms, global cognition rose from baseline to post-intervention (β = 0.23, Cohen's d = 0.495, p < .001) and again from post-intervention to 6-month follow-up (β = 0.10, d = 0.207, p = .007). Delayed memory (d = 0.540) and memory-and-learning (d = 0.494) showed the largest post-training gains; the non-memory composite improved more modestly (d = 0.189).^[1]

Diabetes self-management improved too. Self-reported diabetes self-management improved after training (β = 0.40, d = 0.249) and held at 6 months, the first reported evidence that, in this trial's participants, multidomain CCT was associated with improvement on a self-reported diabetes self-management outcome, though the change appeared only in self-report.^[1]

Who benefited most? The 15 participants with lower baseline cognition (z ≤ −0.5) improved far more than the 68 with average-to-high baselines: d = 1.072 versus 0.367. Of 75 completers, 13 showed a "clinically meaningful" cognitive improvement (≥0.5 SD) and 16 a clinically meaningful (≥1 SD) self-management improvement.^[1]

Adherence and retention were strong, strongest in the tailored arm. Overall, 89% (76/85) completed post-intervention assessment and 82% (70) the 6-month follow-up; 68% and 71% of the two arms completed at least 80% of the 48 prescribed sessions. TA-CCT participants were more likely to finish post-testing (100% vs 80%, p = .002) and follow-up (93% vs 72%, p = .01), and trained longer (553 vs 412 minutes, p = .009). Yet objective cognitive gains were similar in both arms, tailoring drove engagement, not extra test-score benefit. Dose still mattered: more training time meant higher global cognition at every assessment (p = .0005).^[1]

Limitations, what it does not show

No untreated control group. Both arms received active training, so practice and retest effects cannot be fully ruled out. The authors counter with three observations: the global effect (d = 0.495) was significantly larger than meta-analytic effects for passive (d = 0.12, CI 0.08–0.16) or active (d = 0.18, CI 0.12–0.24) controls; CCT has already beaten both control types in prior studies; and the 4 participants who completed under 20% of the training improved much less (d = 0.17) than the rest (d = 0.49).^[1]

A self-selected, high-functioning sample. Participants were predominantly well-educated (16 years on average) and motivated, limiting generalizability to the wider diabetic population.

Self-report gap. The self-management improvement appeared in self-reports but not informant reports, cognitive and self-management changes were not clearly related (r = −.09, ns), and biological outcomes such as hemoglobin A1c were not measured.^[1]

Tailoring was not the active ingredient. Tailored, adaptive training showed no extra objective cognitive benefit over generic training, in the authors' words, "adaptive difficulty and individual task tailoring may not be critical components of such interventions", even though it clearly improved retention and training time.

Where this fits in the broader evidence

The result extends a substantial prior literature. A 2017 meta-analysis by Hill and colleagues found CCT effective in older adults with mild cognitive impairment,^[4] and a 2018 double-blind RCT by Whitelock and colleagues showed adults with type 2 diabetes improved visuospatial attention after working-memory training.^[5] This trial is the first to test multidomain CCT on both cognition and disease self-management in this population. Its null tailoring result contrasts with a 2017 trial in which tailored, adaptive training beat generic training,^[6] a discrepancy flagged for head-to-head replication. The paper concludes that "CCT represents an intervention that is relatively easy to implement in a range of community and clinical settings at relatively low cost." Browse all peer-reviewed CogniFit research in the research studies index.