Study: 8 weeks of personalized brain training improved all 5 cognitive domains in adults with long COVID brain fog

Cognitive dysfunction, commonly called brain fog, was among the most prevalent persistent symptoms of post-acute sequelae of COVID-19 (PASC, or long COVID): a 2022 Global Burden of Disease analysis in JAMA (Global Burden of Disease Long COVID Collaborators) estimated that a persistent cognitive symptom cluster affected a large share of people following symptomatic COVID-19.

A 2023 pilot study published in the International Journal of Environmental Research and Public Health was, to the authors' knowledge, the first effort to investigate whether personalized computerized cognitive training (CCT) could ameliorate this cognitive dysfunction. After 8 weeks of self-administered training at home, participants' scores rose in all five cognitive domains measured, attention, memory, coordination, perception and reasoning, and the size of the gain scaled with the amount of training completed.

What the study tested

The study was a single-center, before-and-after feasibility trial conducted entirely online between October 2021 and February 2022, registered at ClinicalTrials.gov (NCT05571852) and reported per the CONSORT extension for pilot and feasibility trials. Recruitment was led by the CIR Long COVID Research Center with Nebrija University (Madrid, Spain), with participants invited through Long COVID ACTS, Spain's main PASC patient association.

Of 262 eligible volunteers, 103 completed both the initial and final cognitive assessments, and 73 also completed the required minimum of 10 training sessions and formed the analyzed sample: mean age 46.1 years (SD = 7.6), 66 of 73 female. All were adults reporting concentration problems or brain fog at least 3 months after COVID-19 infection, screened by a trained neuroscientist.^[1]

The baseline picture was severe. 45 participants (61.6%) reported having lost at least 50% of their pre-infection health level, and 65 (89.0%) rated their health as poor or fair. On the named outcome battery, the Cognitive Assessment Battery (CAB)™ PRO, a 30–35-minute online battery of 17 tests scored against a normative dataset of 1,282,242 healthy test-takers, the sample sat below the 50th percentile in all five cognitive domains before training.^[1]

The intervention: 8 weeks of gamified cognitive training sessions of around 10 minutes each, completed at home as often as participants wished. Each session combined two tasks drawn from a pool of 12 activities, with task selection and difficulty chosen automatically for each person based on the strengths and weaknesses detected at baseline. The primary outcome was the change in z-scores across the five cognitive domains; total minutes of training served as a secondary (adherence) measure.

What it found

Cognitive performance increased significantly after training (main effect of test moment: F(1,72) = 104.36, p < 0.001, η²_partial = 0.592; mean z-score gain = 0.63, t(72) = 10.2). Gains were significant in every individual domain, and a non-parametric Kruskal–Wallis analysis confirmed all five comparisons (all p < 0.01).^[1]

In raw score terms, the mean increase over baseline was 31% for attention, 37% for memory, 52% for coordination, 42% for perception, and 26% for reasoning. Gains were not equal across domains (interaction F(4,288) = 9.48, p < 0.001): coordination showed the largest standardized improvement (z = 1.01 [95% CI: 0.75–1.26]) and reasoning the smallest (z = 0.36 [0.19–0.52]).^[1]

Adherence was substantial: participants completed a mean of 51 training sessions (median = 44; range 10–251) and invested a mean of 435 minutes of training (median = 358; range 78–2,448) over the 8 weeks, evidence that people with brain fog could realistically operate an unsupervised digital program at home.

More training, more improvement

A covariate analysis found a significant interaction between test moment and minutes of training (F(1,70) = 6.60, p = 0.012): cognitive improvement increased as a function of the amount of training, similarly across all five domains. Age, by contrast, did not significantly moderate the effect (all Fs < 1.88, p-values > 0.11).^[1]

Whether someone rated their health positively or negatively, and whether their infection was more or less than a year in the past, made no difference to whether they improved (all Fisher's exact p-values > 0.21).

Limitations, what this does NOT show

The authors are explicit that the findings "should be taken with the necessary caution." The honest list:

No control group. As a before-and-after feasibility pilot, the design cannot separate the training effect from practice effects, natural recovery, or placebo response. The authors state that randomized controlled trials "are required in the near future to compare the outcomes of the intervention group to a group that did not receive the CCT."

Self-selected, mostly female sample with an unknown denominator. Only 73 of the 262 initial volunteers were analyzed, the common denominator of potentially eligible participants is unknown, and 66 of 73 participants were women, consistent with the documented gender bias in long COVID, but limiting generalization.^[1]

No full neuropsychological assessment. PASC was self-reported (cognitive symptoms ≥3 months post-infection). Whether effects are sustained after training ends, and what the optimal dose is, also remain untested.

Where this fits in the broader evidence

Earlier meta-analyses had reported significant cognitive improvements with computerized cognitive training in people with cognitive dysfunction associated with stroke (Zhou et al., 2021), Parkinson's disease (Gavelin et al., 2022), and other conditions, over training periods that typically ranged from 4 to 12 weeks, this 2023 pilot was, to the authors' knowledge, the first to extend that line of research to long COVID. For the full list of peer-reviewed research on CogniFit's assessment and training technology, see the research index; for CCT evidence in other conditions, see CogniFit's pages on stroke and multiple sclerosis.