Longer Paxlovid courses did not beat placebo for long COVID in a 964-person trial

Results posted to the federal ClinicalTrials.gov registry on 27 March show that extended courses of the antiviral Paxlovid did not meaningfully improve long COVID symptoms compared with placebo, in one of the largest randomized trials yet completed on antiviral treatment for the condition [s1].

What the trial tested

RECOVER-VITAL, run under the NIH's RECOVER long COVID research initiative, enrolled 964 participants and randomized them to one of three arms: nirmatrelvir 300mg plus ritonavir 100mg twice daily for 25 days; the same combination for 15 days followed by ritonavir alone plus a nirmatrelvir-matching placebo for 10 more days; or ritonavir plus nirmatrelvir-matching placebo for the full 25 days, serving as the control arm [s1]. Nirmatrelvir-ritonavir is the generic combination sold as Paxlovid, FDA-authorized for treating acute COVID-19; RECOVER-VITAL tested whether a longer, higher-dose course could clear residual viral reservoirs some researchers have hypothesized underlie persistent long COVID symptoms.

The trial enrolled participants starting in July 2023 and reached primary completion in December 2024, with the full trial closing out in March 2025 [s1] — a multi-year federal research effort whose results only became public more than a year after the last data were collected.

Participants were assessed within three separate symptom clusters, each with its own validated measurement instrument and its own primary endpoint at day 90: cognitive dysfunction, measured by the PROMIS Cognitive 8a questionnaire, with improvement defined as at least a 5-point rise in T-score from baseline; autonomic dysfunction, measured by the Orthostatic Hypotension Questionnaire, with improvement defined as at least a 1-point drop in the symptom-severity item; and exercise intolerance — post-exertional malaise, the worsening of symptoms after physical or mental exertion that is one of long COVID's most disabling features — measured by a modified DePaul Symptom Questionnaire [s1].

What the results showed

Across all three symptom clusters, the differences between the Paxlovid arms and placebo were small and did not favor sustained treatment. In the cognitive dysfunction cluster, 57.8% of participants in the 25-day Paxlovid arm improved, compared with 52.4% in the 15-day arm and 54.6% in the placebo arm [s1]. In the autonomic dysfunction cluster, the placebo arm actually had a numerically higher improvement rate — 69.5% — than either the 25-day arm (63.1%) or the 15-day arm (69.4%) [s1]. In the exercise intolerance cluster, 25.5% of the 25-day Paxlovid group improved, versus 34.2% of the 15-day group and 33.3% of the placebo group [s1].

No consistent pattern favors either Paxlovid regimen over placebo across the three clusters, and the one cluster where placebo scored highest — autonomic dysfunction — undercuts a simple story of drug benefit obscured by dosing or timing. Read together, the results describe a trial that failed to demonstrate the benefit it was designed to detect.

Why the trial was worth running despite the outcome

The scientific premise behind RECOVER-VITAL — that residual SARS-CoV-2 viral reservoirs might persist in tissue and drive some long COVID symptoms, and that clearing them with a longer antiviral course might resolve those symptoms — was a serious, biologically grounded hypothesis, not a long shot. Long COVID has no FDA-approved treatment, and the viral-persistence hypothesis has been one of the more mechanistically plausible explanations put forward for a subset of cases, distinct from competing hypotheses centered on autoimmune activation, microclotting, or lasting organ damage from the acute infection. A trial of this size — powered to detect meaningful effects across three separate symptom domains — was a genuine test of that hypothesis, not a token effort.

The negative result is informative precisely because the hypothesis was credible. It does not disprove viral persistence as a contributor to long COVID in general, but it does undercut the specific proposition that a longer, standard-dose course of an already-available antiviral is an effective treatment for the condition once it has become chronic — as opposed to Paxlovid's established role in treating acute infection, which this trial did not test and does not call into question.

Limitations to weigh

This is a results posting to a clinical trial registry, not yet a peer-reviewed journal publication with the full statistical analysis, subgroup breakdowns, and discussion of limitations that a published paper would include. The percentages reported are unadjusted rates within a modified intent-to-treat population — participants who were randomized, received at least part of one dose, and had data collected for the relevant symptom cluster — and the registry posting does not include confidence intervals or formal significance testing in the figures released so far. Long COVID itself is also a heterogeneous condition, likely encompassing multiple underlying mechanisms; a null result for one candidate mechanism in one trial population does not rule out benefit in a more precisely defined subgroup, such as patients with laboratory evidence of persistent viral antigen.

What to watch

A full peer-reviewed publication of RECOVER-VITAL's results, expected to follow the registry posting, should clarify statistical significance and any subgroup findings the topline percentages don't capture. More broadly, the trial's outcome narrows — without eliminating — one of the leading mechanistic hypotheses for long COVID, at a point when the RECOVER initiative's broader slate of trials, testing other candidate treatments across the condition's various proposed mechanisms, remains the primary pipeline for identifying any effective therapy.

Sources

• RECOVER-VITAL: Platform Protocol, Appendix to Measure the Effects of Paxlovid on Long COVID Symptoms (NCT05965726) — Study Results — ClinicalTrials.gov, U.S. National Library of Medicine, results posted 27 March 2026