In a randomized trial of 58 people with long COVID, NR raised blood NAD⁺ but did not significantly improve cognition, fatigue, sleep or mood compared with placebo. Some symptoms improved in exploratory before-and-after analyses. Those observations cannot carry the same weight as the controlled comparison. [1]
That difference is especially important when symptoms fluctuate and participants are understandably looking for something that helps. A plausible mitochondrial mechanism is worth testing, but it does not establish a treatment. The general NAD evidence guide explains why a blood-marker increase is only one step in that process.
How the trial was designed
The single-center Boston study ran for 24 weeks and used a placebo lead-in followed by two sequences. Thirty-seven people were assigned to NR for 20 weeks. Twenty-one received placebo for ten weeks and then NR for ten weeks. The design gave everyone exposure to NR, but only the earlier treatment period supplied the clean concurrent NR-versus-placebo comparison. [1]
| Study and participants | Design and outcome | Finding | Main limit |
|---|---|---|---|
| 58 adults with long COVID [1] | Randomized, blinded, placebo lead-in; NR-NR versus placebo-NR | Blood NAD⁺ increased; no significant between-group cognitive benefit. | Single center; self-reported infection history; limited diversity. |
| Symptoms and safety [1] | Fatigue, mood and sleep questionnaires; adverse-event monitoring | No significant controlled symptom benefit; one unrelated serious event. | Differential dropout and small samples reduce precision. |
| Exploratory pooled NR exposure [1] | Within-person changes after NR, pooled across sequences | Some measures improved from baseline. | Post hoc, no multiplicity correction; not equivalent to placebo-controlled efficacy. |
The primary cognitive measures included a questionnaire about everyday cognition, a broader neuropsychological battery and a timed executive-function task. Fatigue, mood and sleep questionnaires were secondary measures. Multiple measures are useful because a person’s perceived cognitive difficulty and performance on a test are related but not interchangeable.
The result to lead with
NR increased NAD⁺ roughly 2.6- to 3.1-fold after five to ten weeks in the group assigned NR from the start. The cognitive comparisons were not statistically significant, with reported p values from 0.47 to 0.74. Neither were differences in fatigue (p = 0.59), sleep (p = 0.69), anxiety (p = 0.84) or depression (p = 0.20). [1]
It would therefore be inaccurate to say that this trial proved NR relieved long-COVID brain fog. It would also be inaccurate to say it had no biological effect: the blood NAD⁺ result was clear. The clinical question and the biochemical question simply had different answers.
Similar distinctions appear in the older-adult memory trials. Those populations cannot be combined as though every cognitive complaint has the same cause, but they illustrate why an appealing biomarker result should not replace a cognitive endpoint.
Why the exploratory findings are weaker
Researchers combined participants’ first ten weeks of NR exposure and compared their symptoms with their own baseline. Executive functioning, fatigue, sleep and depression measures improved in parts of that analysis. It was post hoc and not adjusted for the number of comparisons. [1]
Without a concurrent randomized comparison for that pooled question, improvement can reflect time, repeated testing, changing background care or fluctuations in symptoms. These are reasons for a larger confirmatory trial. They are not grounds to dismiss participants’ experiences, and they are not proof that the supplement caused those experiences.
Missing participants make certainty harder
Dropout was uneven. The group assigned NR from the beginning lost 32.4% of participants by ten weeks and 51.4% by twenty weeks; the placebo-then-NR group lost 14.3% at each point. Reasons included reinfection, adverse events, medication changes and personal circumstances. It would be wrong to attribute every withdrawal to NR, but it would also be wrong to ignore that imbalance. [1]
The main repeated-measures model handled missing observations under a missing-at-random assumption without imputing replacement scores. A separate analysis examined protocol completers. Neither approach can recover information that was never collected or guarantee that dropouts were unrelated to outcomes. Smaller final samples also make it harder to detect a true effect.
Safety and practical meaning
The authors reported no significant difference in adverse events between NR and placebo phases. One serious event was considered unrelated to the study treatment. The trial received support that included Niagen Bioscience as well as academic and philanthropic sources. These details belong beside the efficacy results, not in place of them. [1]
This was an oral NR study, not an IV NAD⁺ or home-injection study. It cannot establish the effectiveness of those services for long COVID. The safety overview discusses why formulation and population matter. People interested in contributing to better evidence can use the trial participation guide; research enrollment is different from buying a product marketed with research language.