NIH's RECOVER-NEURO Trial: Three Non-Drug Treatments for Long COVID Cognitive Symptoms All Fail to Beat Control

The NIH RECOVER initiative — the largest long COVID research program in history — has published results from its RECOVER-NEURO clinical trial, evaluating three non-drug treatments for cognitive symptoms in post-acute sequelae of SARS-CoV-2 (PASC). The result was null: none of the three active interventions — BrainHQ, PASC Cognitive Recovery (PASC-CoRE), and transcranial direct current stimulation (tDCS) — outperformed active or placebo controls on the primary cognitive endpoint at 10 weeks.[1]

The RECOVER-NEURO trial enrolled 328 adults with long COVID at 22 US sites using a decentralized design — most participants completed study activities remotely. The three interventions: (1) BrainHQ, an adaptive online brain training platform; (2) PASC-CoRE, a structured online goal management training program; and (3) tDCS, a noninvasive form of brain stimulation applied to the dorsolateral prefrontal cortex. The comparison group received active control tasks — online games or sham brain stimulation.[1]

The primary outcome was the modified Everyday Cognition scale (ECog2). At the end of the 10-week intervention, all five groups — including both control groups — showed improvement over time. No active treatment arm demonstrated a statistically significant advantage over its comparator. Cognitive symptoms improved regardless of which arm patients were assigned to.

This finding contrasts with a smaller UK RCT published the same month in JAMA Network Open, which found that a personalized goal-focused cognitive rehabilitation program (also 10 weeks) produced a large functional benefit (Cohen d = 1.57). The difference likely reflects methodological factors: the UK trial used individualized, self-defined functional goals rather than standardized cognitive training; it also enrolled a smaller, more homogeneous sample and used a self-report functional endpoint rather than a standardized cognitive battery.

Clinical Context

The cognitive sequelae of SARS-CoV-2 infection — commonly called "brain fog" — emerged as among the most disabling features of long COVID from the earliest waves of the pandemic. Cross-sectional data suggested that 10–30% of COVID-19 survivors experienced persistent cognitive complaints, with deficits measurable on objective neuropsychological testing in a substantial minority. The pathophysiology remains contested: viral neuroinvasion, neuroinflammation, autoimmunity, cerebrovascular microinjury, and reactivation of latent herpesviruses have all been implicated.[2] The RECOVER initiative — a $1.15 billion NIH-funded longitudinal cohort spanning more than 100,000 participants — was designed explicitly to fill this mechanistic and therapeutic gap.[3]

Prior small trials of cognitive rehabilitation and neurostimulation in post-infectious fatigue syndromes, including post-COVID, had shown mixed results. Computerized cognitive training platforms such as BrainHQ have demonstrated benefit in aging populations, but their efficacy in neuroinflammatory states is uncertain. Transcranial direct current stimulation (tDCS) modulates cortical excitability and has been trialed in a wide range of neuropsychiatric conditions — with modest, inconsistent effects. The RECOVER-NEURO trial represents the most rigorous head-to-head comparison of these approaches in a well-characterized post-COVID cognitive population, and its null result across all three arms is a meaningful signal — not a data gap — that these interventions do not work for this indication at this intensity and duration.[2]

What This Tells Us

RECOVER-NEURO is a null result, but it is not a nihilistic one. The fact that all groups improved suggests that long COVID cognitive symptoms do improve over time with or without specific intervention — a finding consistent with the natural history literature. It also suggests that placebo and expectancy effects are substantial in this population, which has implications for future trial design.

The trial does not rule out that individualized cognitive rehabilitation (the approach showing benefit in the UK trial) could help. It rules out that generic adaptive brain training or non-individualized goal management is better than active control — at least on the ECog2 at 10 weeks.

Limitations

The trial was conducted entirely remotely, which may have reduced the quality of intervention delivery and the fidelity of tDCS administration. The primary endpoint (ECog2) may not capture the specific functional deficits that matter most to patients. The 10-week follow-up may be insufficient for demonstrating long-term cognitive rehabilitation effects. Patients were not selected by specific biomarkers or cognitive profiles, introducing substantial heterogeneity.