Viral persistence has been among the leading hypotheses for long COVID, and it carries an obvious implication — that an antiviral should help. RECOVER-VITAL tested it properly. This double-blind placebo-controlled phase 2 trial at 69 US sites randomised 959 adults with symptoms persisting at least 12 weeks after SARS-CoV-2 infection, grouped into cognitive, autonomic and exercise phenotypes. Participants received nirmatrelvir-ritonavir for 15 days followed by placebo, nirmatrelvir-ritonavir for 25 days, or placebo with ritonavir throughout. Median age was 49 and 67% were women.
There was no benefit anywhere. For the cognitive phenotype the adjusted difference against placebo was 3.2% (95% CI -10.4 to 16.8) for 25 days and -2.2% (-15.5 to 11.1) for 15 days. For autonomic symptoms, -6.4% (-18.5 to 5.7) and -0.1% (-12.5 to 12.3). For exercise, -7.8% (-19.5 to 3.8) and 0.9% (-11.4 to 13.2). Secondary endpoints were also flat. Safety was unremarkable, with no deaths and serious adverse events in 4%.
Extending treatment from 15 to 25 days did not help either, which weakens the standard rescue argument that the course was too short.
The honest reading is that a leading mechanistic hypothesis has failed its most direct test. That does not disprove viral persistence in every patient, but it does mean clinicians should stop offering antiviral courses off-label for long COVID, and should say clearly that no drug currently has evidence in this condition. The management that remains — pacing, graded rehabilitation where post-exertional malaise allows, and treating identifiable components such as orthostatic intolerance — is unglamorous and is what there is.
- Do not prescribe nirmatrelvir-ritonavir for long COVID; neither 15 nor 25 days helped any symptom phenotype
- Say plainly that no drug currently has randomised evidence in long COVID, rather than leaving hope attached to antivirals
- Manage identifiable components separately — orthostatic intolerance, sleep disturbance, deconditioning, mood
- Respect post-exertional malaise when planning rehabilitation; graded exercise is not appropriate for every phenotype
- Note the cohort was 78% White and US-based, so symptom measurement instruments may translate imperfectly
The statistics, in plain English
Six confidence intervals, all crossing zero, and several of them centred on the wrong side of it. The widths — roughly 25 percentage points across — mean the trial cannot exclude a modest benefit, but with 959 participants across three phenotypes and two durations, a clinically useful effect would have been visible somewhere. The point about duration matters statistically too: when a longer course performs no better than a shorter one, the dose-response relationship you would expect from a real drug effect is absent, which strengthens the null well beyond what any single confidence interval shows.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for infectious diseases, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free