For people who already have established Long COVID, extended courses of Paxlovid have not shown an overall clinical benefit in randomized trials.
That conclusion is now considerably stronger than it was after the first negative study. Three published double-blind, placebo-controlled trials have tested nirmatrelvir-ritonavir, the antiviral combination sold as Paxlovid, for established Long COVID. The trials enrolled 155, 100, and 959 participants and tested treatment lasting 15 or 25 days. Despite using different patient populations and outcome measures, all three failed to demonstrate a statistically significant overall benefit from the antiviral.
That does not establish that viral persistence has nothing to do with Long COVID. It does not prove that no individual patient could respond. And it does not answer the separate question of whether the normal five-day Paxlovid course, taken during an acute COVID-19 infection, can reduce the later risk of developing Long COVID.
Those are three different claims, and the evidence for each is different.
The most defensible conclusion as of September 29, 2026 is narrower but important: Paxlovid is not supported by current randomized evidence as a general treatment for established Long COVID, and the idea that widespread, ongoing, Paxlovid-sensitive viral replication is the dominant reversible cause of Long COVID has become harder to sustain.
The new RECOVER-VITAL trial is much larger than the earlier studies
The immediate reason this question is resurfacing is the RECOVER-VITAL randomized trial indexed by PubMed, which was published in The Lancet Infectious Diseases on August 31 and highlighted by NIH Research Matters on September 29, 2026.
RECOVER-VITAL enrolled 959 adults with Long COVID at 69 U.S. sites. Participants had symptoms lasting at least 12 weeks and were grouped into three major symptom phenotypes: cognitive problems, autonomic symptoms involving regulation of blood pressure or heart rate, and exercise intolerance/post-exertional symptoms.
Within each phenotype, participants were randomized to one of three regimens:
- Paxlovid for 25 days;
- Paxlovid for 15 days followed by 10 days of ritonavir plus placebo; or
- ritonavir plus placebo for 25 days.
The primary question was whether substantially more patients receiving nirmatrelvir would achieve a clinically meaningful improvement by day 90.
They did not.
For the cognitive phenotype, the adjusted difference versus placebo was 3.2 percentage points for the 25-day regimen and -2.2 points for the 15-day regimen. For autonomic symptoms, the differences were -6.4 and -0.1 points. For exercise-related symptoms, they were -7.8 and 0.9 points. Every confidence interval crossed zero, and none of the comparisons was statistically significant. Secondary outcomes likewise showed no treatment advantage.
Participants in all three groups improved over time, but they improved at similar rates. That distinction matters. Improvement after taking a drug is not the same as improvement because of the drug; the randomized placebo comparison is what allows the trial to separate the two.
RECOVER-VITAL was funded by the National Institutes of Health.
Three published randomized Long COVID treatment trials now tell a consistent story
RECOVER-VITAL did not appear in an evidentiary vacuum. Two smaller randomized trials had already tested 15-day courses of Paxlovid for established Long COVID.
| Trial | Participants | Paxlovid regimen | Main outcome | Result |
|---|---|---|---|---|
| STOP-PASC | 155 | 15 days | Pooled severity of six Long COVID symptoms at 10 weeks | No significant benefit |
| PAX LC | 100 | 15 days | PROMIS-29 physical-health score at day 28 | No significant benefit |
| RECOVER-VITAL | 959 | 15 or 25 days | Clinically meaningful improvement in cognitive, autonomic, or exercise phenotypes at day 90 | No significant benefit in any phenotype |
A fourth randomized, double-blind Paxlovid trial, PROLIFIC (NCT05823896), enrolled 219 participants and is listed as completed, but its efficacy results had not been publicly reported as of September 29, 2026. It therefore cannot yet be included in the published efficacy synthesis.
These studies should not be treated as if they were one giant 1,214-person trial. They used different eligibility rules, endpoints, follow-up periods and statistical designs, so simply pooling their results would require a formal meta-analysis.
But their directional convergence is meaningful. Three published randomized experiments asked variations of the same practical question and none found evidence that extended nirmatrelvir produced better overall clinical outcomes than a ritonavir-containing placebo regimen.
STOP-PASC: 155 patients, 15 days, no symptom benefit
The STOP-PASC randomized clinical trial in JAMA Internal Medicine enrolled 155 people who had experienced moderate-to-severe Long COVID symptoms for at least three months. The average time since their original infection was about 17.5 months.
Participants received either 15 days of nirmatrelvir-ritonavir or placebo plus ritonavir. The primary outcome combined six common symptom categories: fatigue, brain fog, shortness of breath, body aches, gastrointestinal symptoms and cardiovascular symptoms.
At 10 weeks, researchers found no statistically significant treatment benefit on the pooled primary outcome. Nor did they find significant advantages on major measures of fatigue, cognition, dyspnea, physical function or patients’ global assessments of their condition.
Source provenance is worth noting. STOP-PASC was funded by Pfizer, and the paper states that Pfizer collaborated on trial design and execution and reviewed the manuscript.
That funding relationship deserves disclosure, but it cuts against one simplistic criticism of the result: the sponsor’s drug still failed to produce a favorable efficacy finding.
PAX LC: another 100 patients, another negative randomized result
The PAX LC trial indexed by PubMed randomized 100 participants to 15 days of nirmatrelvir-ritonavir or placebo-ritonavir.
Its primary endpoint was change in the PROMIS-29 Physical Health Summary Score at day 28.
The adjusted improvement was 0.45 points in the Paxlovid group and 1.01 points in the placebo group, producing an adjusted between-group difference of -0.55 points, with a 95% confidence interval from -2.32 to 1.21 and a p-value of 0.54. In other words, the study did not detect a benefit.
There were also no significant differences in the secondary efficacy endpoints. Drug-related adverse events were more common with Paxlovid, largely because of altered taste.
PAX LC also received Pfizer funding, along with support from Fred Cohen and Carolyn Klebanoff.
Then came the much larger, NIH-funded RECOVER-VITAL trial, which reached the same basic efficacy conclusion.
That convergence across differently funded studies is more informative than any one trial considered alone.
But some patients have reported dramatic improvement with Paxlovid. How can both things be true?
Anecdotal improvement is one reason extended antiviral treatment was worth testing rather than dismissing.
A 2025 Communications Medicine case series of 13 people with Long COVID documented extended Paxlovid courses lasting 7.5 to 30 days. Outcomes varied considerably. Some participants reported meaningful improvements, some improvements were temporary, and others reported no benefit. Five of the 13 had sustained improvement according to the published case series.
Those observations were scientifically useful because they generated a testable hypothesis.
They could not establish efficacy.
Long COVID symptoms can fluctuate. People may improve or worsen over time. Treatment is often sought during particularly bad periods, creating the possibility of regression toward a person’s usual symptom level. Expectations can affect patient-reported outcomes. Reinfection itself may change symptoms. Other treatments may be changing simultaneously.
The randomized trials deal with those problems by comparing people assigned to the antiviral against people assigned to a blinded control.
And that comparison has repeatedly failed to reproduce an overall Paxlovid effect.
That does not mean the people who reported feeling better were mistaken about feeling better. It means their improvement, by itself, cannot establish that nirmatrelvir caused it.
Does the failure of Paxlovid disprove viral persistence in Long COVID?
No.
This is where the most common interpretation of the trials goes too far.
“Viral persistence” can refer to several biologically different things:
- SARS-CoV-2 RNA or proteins remaining in tissue;
- cells containing viral material that continues to stimulate the immune system;
- low-level or compartmentalized viral replication;
- replication-competent virus persisting as an active infection;
- a persistent virus being not merely present, but actually responsible for ongoing symptoms.
Those propositions are not interchangeable.
A 2024 Science Translational Medicine tissue study found prolonged T-cell activation after SARS-CoV-2 infection and examined rectosigmoid biopsies from five participants with Long COVID. SARS-CoV-2 spike-encoding RNA was found in all five, while double-stranded spike RNA was detected in three, as late as 676 days after infection. The study therefore supplied real evidence that viral material can persist in tissue long after the acute illness.
But the same study also found increased tissue T-cell activation in many post-COVID participants who did not have Long COVID. And five biopsied Long COVID patients are not enough to establish that persistent viral material is a universal cause of the syndrome.
A Nature Reviews Microbiology review of SARS-CoV-2 persistence makes another crucial distinction: researchers have detected viral RNA and proteins long after acute infection, but replication-competent SARS-CoV-2 has not been established in immunocompetent people with Long COVID in the way it can be demonstrated in prolonged infections among severely immunocompromised patients.
So the existence of persistent viral material and the existence of an ongoing, drug-sensitive infection are separate evidentiary questions.
What viral-persistence theory did the Paxlovid trials actually test?
Nirmatrelvir inhibits the SARS-CoV-2 main protease, interfering with the virus’s ability to replicate.
For extended Paxlovid to substantially improve Long COVID through a viral-reservoir mechanism, several links in a causal chain would need to hold:
- enough actively replicating SARS-CoV-2 would need to persist in a meaningful proportion of patients;
- that virus would need to remain susceptible to nirmatrelvir;
- the drug would need to reach the relevant tissue compartments at sufficient concentrations;
- 15 or 25 days of treatment would need to suppress the process adequately; and
- ongoing replication would need to be sufficiently close to the cause of symptoms that suppressing it produces measurable improvement.
RECOVER-VITAL tested that chain clinically in a broad, symptom-defined Long COVID population.
Its failure does not identify which link failed.
Perhaps persistent material is largely non-replicating. Perhaps only a minority of patients have an antiviral-sensitive reservoir. Perhaps a relevant reservoir is difficult for the drug to reach. Perhaps active infection initiated downstream immune, vascular or neurological changes that no longer depend on continuing replication. Or perhaps viral persistence is simply not driving symptoms in most participants.
Those possibilities remain scientifically open.
But the repeated negative randomized trials still change the evidentiary balance.
They make the simplest and broadest version of the hypothesis less plausible: that ongoing, Paxlovid-sensitive SARS-CoV-2 replication is a common dominant driver of established Long COVID and can generally be reversed with a few additional weeks of the drug.
That is a considerably stronger conclusion than saying merely that “one study was negative.”
Could Paxlovid still work for a specific Long COVID subgroup?
Possibly, but that remains a hypothesis rather than an established treatment strategy.
Long COVID is clinically heterogeneous, and the randomized trials did not enroll only patients proven to harbor an actively replicating SARS-CoV-2 reservoir. If such a biologically identifiable subgroup exists, a treatment effect in that group could disappear when averaged together with patients whose symptoms arise through other mechanisms.
That possibility becomes convincing only if researchers can identify the subgroup before treatment and then show in a randomized trial that those patients respond.
RECOVER researchers are now analyzing collected blood samples for evidence related to viral persistence and treatment response. NIH specifically says additional results from those analyses are planned.
A finding that a prespecified biomarker identifies a strongly responsive subgroup would materially alter the interpretation of RECOVER-VITAL.
Until then, selecting supposed responders based on who happened to improve after taking Paxlovid risks recreating the same problem randomized trials are designed to solve.
Treating Long COVID and preventing Long COVID are different questions
This distinction is particularly important because the evidence becomes more complicated when Paxlovid is taken during the initial COVID-19 infection, before Long COVID develops.
The three negative published trials above studied people who already had persistent symptoms.
They do not establish that acute antiviral treatment cannot prevent some later illness.
A small 2026 randomized trial found a prevention signal
The PANORAMIC Norway randomized trial tested the standard five-day nirmatrelvir-ritonavir course during acute COVID-19 in non-hospitalized adults.
At three months, Long COVID as defined by patient-reported fatigue, dyspnea and/or cognitive symptoms was reported by:
- 17 of 66 patients, or 26%, assigned Paxlovid; and
- 33 of 77 patients, or 43%, assigned placebo.
The estimated relative risk was 0.60, with a 95% confidence interval of 0.37 to 0.98.
That is an intriguing randomized signal.
It is also far from definitive.
Investigators had planned to enroll 2,000 participants but stopped at 144 because of insufficient recruitment. The authors themselves explicitly cautioned that the small sample prevented firm conclusions and called for larger trials.
A much larger observational analysis found no overall prevention effect
Meanwhile, a 2025 PLOS Medicine target-trial emulation from the N3C and RECOVER consortia examined electronic health records from 445,738 people who had acute COVID-19 and were eligible for Paxlovid.
That study found no overall reduction in subsequent Long COVID incidence associated with Paxlovid. It did find small estimated protective effects in some subgroups, including people aged 65 or older and people with a Charlson Comorbidity Index of 3–4.
The huge sample size does not make that study equivalent to a randomized trial. Its causal estimates depend on researchers having adequately measured and adjusted for differences between people who did and did not receive Paxlovid. Unmeasured confounding remains possible.
So prevention evidence currently points in two directions:
| Question | Strongest relevant evidence | Current reading |
|---|---|---|
| Can extended Paxlovid treat established Long COVID? | Three published placebo-controlled randomized trials | No demonstrated overall benefit |
| Can five-day Paxlovid during acute infection prevent Long COVID? | One small positive randomized trial plus large mixed/negative observational evidence | Unsettled |
Calling those results contradictory would be misleading. They are testing different interventions at different stages of disease.
These Long COVID results do not erase Paxlovid’s acute-COVID indication
Paxlovid’s established regulatory role is treatment of mild-to-moderate acute COVID-19 in adults at high risk of progression to severe disease, not routine treatment of established Long COVID.
The FDA’s Paxlovid approval information specifies treatment during acute infection and warns that ritonavir can create clinically significant drug interactions because of its strong effects on CYP3A metabolism. FDA’s current COVID-19 drug information continues to list the adult approval.
A negative Long COVID treatment trial therefore does not mean Paxlovid has stopped working for the purpose for which it was approved.
It means a different proposed use — taking substantially longer courses after Long COVID is already established — has now been tested repeatedly without demonstrating overall clinical benefit.
That is also why self-directed extended courses are a poor inference from the viral-persistence hypothesis alone. Beyond the lack of demonstrated efficacy for established Long COVID, ritonavir has substantial interaction considerations that require medication review.
What evidence could still change the conclusion?
The current evidence is strong enough to reject extended Paxlovid as a proven general Long COVID treatment, but it does not close antiviral research.
Several findings could materially change the picture:
- a validated biomarker showing which patients harbor an active, clinically relevant SARS-CoV-2 reservoir;
- a randomized trial enrolling that biomarker-positive population and showing both clinical improvement and reduction of the virologic marker;
- independent replication of that result;
- evidence that a different antiviral, longer regimen or drug with better penetration into a relevant tissue compartment produces an effect;
- or convincing evidence that combination treatment targeting both persistence and downstream immune or inflammatory mechanisms succeeds where antiviral monotherapy did not.
What would not be enough is simply accumulating additional uncontrolled reports of people who improved after treatment. The randomized trials now show why a comparison group matters: substantial symptom improvement can occur in both active-treatment and placebo groups.
The bottom line
As of September 29, 2026, the published randomized evidence gives a reasonably clear answer to one important question:
Extended Paxlovid has not been shown to improve established Long COVID overall.
STOP-PASC found no meaningful benefit from 15 days. PAX LC found no meaningful benefit from 15 days. RECOVER-VITAL, with 959 participants, found no statistically significant benefit from either 15 or 25 days across cognitive, autonomic and exercise-related Long COVID phenotypes.
The accumulated evidence therefore materially weakens the idea that ongoing, Paxlovid-sensitive viral replication is a broadly treatable driver of established Long COVID.
But it does not establish that every form of viral persistence is irrelevant. Persistent SARS-CoV-2 RNA and proteins have been documented, and the relationship between those findings, active replication, immune dysfunction and symptoms remains unresolved. What has failed so far is the crucial bridge from “viral material can persist” to “suppressing SARS-CoV-2 replication with extended Paxlovid improves established Long COVID.”
That bridge has now been tested in three published randomized trials, and none has demonstrated an overall clinical benefit.
The separate question of whether treating acute COVID-19 with Paxlovid can prevent some cases of Long COVID remains open. A small 2026 randomized trial produced a promising signal, while much larger observational evidence found no overall protective effect.
That is where the evidence currently stops.
References and Further Reading
Randomized Trials of Paxlovid for Established Long COVID
Nirmatrelvir-Ritonavir Targeting Viral Persistence in Post-COVID-19 Condition: RECOVER-VITAL — The Lancet Infectious Diseases, 2026. The 959-participant, NIH-funded randomized trial comparing 15- and 25-day Paxlovid regimens against placebo-ritonavir across cognitive, autonomic and exercise-related Long COVID phenotypes.
Nirmatrelvir-Ritonavir and Symptoms in Adults With Postacute Sequelae of SARS-CoV-2 Infection: STOP-PASC — JAMA Internal Medicine, 2024. Randomized 155 participants to 15 days of nirmatrelvir-ritonavir or placebo-ritonavir and found no significant overall symptom benefit. The article also provides detailed funding and sponsor-role disclosures.
Nirmatrelvir-Ritonavir Versus Placebo-Ritonavir in Individuals With Long COVID: PAX LC — The Lancet Infectious Diseases, 2025. A 100-participant randomized trial that found no significant improvement in its primary physical-health outcome after 15 days of treatment.
Earlier Clinical Observations
Impact of Extended-Course Oral Nirmatrelvir-Ritonavir in Established Long COVID: A Case Series — Communications Medicine, 2025 publication of a patient-led case series involving 13 people. Useful for understanding the anecdotal and heterogeneous responses that helped motivate randomized testing.
Prevention of Long COVID
Nirmatrelvir for Acute COVID-19 to Prevent Long COVID: PANORAMIC Norway — The Lancet Infectious Diseases, 2026. Small placebo-controlled randomized trial reporting fewer Long COVID endpoints after acute Paxlovid treatment, but recruitment stopped at 144 participants rather than the planned 2,000.
Effect of Paxlovid Treatment During Acute COVID-19 on Long COVID Onset: N3C and RECOVER Target-Trial Emulation — PLOS Medicine, 2025. Large observational causal-inference analysis of 445,738 patients that found no overall reduction in PASC incidence, while identifying small estimated effects in certain subgroups, including patients aged 65 or older and those with a Charlson Comorbidity Index of 3–4.
Viral Persistence and Mechanism
Tissue-Based T Cell Activation and Viral RNA Persist for Up to Two Years Following SARS-CoV-2 Infection — Science Translational Medicine, 2024. Provides tissue-level evidence of prolonged immune activation and SARS-CoV-2 RNA persistence, including rectosigmoid biopsy findings in a small Long COVID subset.
The Consequences of SARS-CoV-2 Within-Host Persistence — Nature Reviews Microbiology, 2025. Reviews the distinction between detection of persistent viral material, prolonged replication in immunocompromised hosts, and the still-unresolved question of replication-competent virus as a cause of Long COVID.
Current Trial and Regulatory Context
PROLIFIC: ImPROving Quality of LIFe in the Long COVID Patient — ClinicalTrials.gov, NCT05823896. Registry record for a completed 219-participant randomized, double-blind trial of 15-day nirmatrelvir-ritonavir versus placebo-ritonavir for established Long COVID. Efficacy results had not been publicly reported as of September 29, 2026.
NIH Research Matters: Antiviral Drug Did Not Improve Long COVID Symptoms — National Institutes of Health, September 29, 2026. Summarizes RECOVER-VITAL and notes that investigators are continuing to analyze blood samples for evidence related to viral persistence and treatment response.
FDA Approval of Paxlovid for Acute COVID-19 in High-Risk Adults — U.S. Food and Drug Administration. Establishes Paxlovid’s approved adult indication and summarizes the important drug-interaction considerations associated with ritonavir.
Editorial currency note: Evidence reviewed through September 29, 2026. RECOVER investigators have said additional biomarker analyses examining possible viral persistence and treatment response are pending. The completed PROLIFIC trial also had not publicly reported efficacy results as of this review. Those findings could refine the interpretation for specific biological subgroups or add another randomized result but do not alter the currently published randomized clinical evidence.
