Two randomized trials published in the past six weeks have tested whether drugs taken during acute SARS-CoV-2 infection can prevent long COVID. One trial succeeded by its own statistical standard. The other did not. Together they offer the first randomized evidence that early pharmacologic intervention can reduce the risk of persistent post-infection illness; the pattern of results reveals how deeply the measurement of long COVID shapes what researchers can claim about it.
The first trial, published July 16 in The Lancet Infectious Diseases, tested the antiviral nirmatrelvir-ritonavir, sold as Paxlovid, in 144 non-hospitalized adults aged 18 to 65 with acute COVID-19 and symptoms for five days or fewer. Conducted across three sites in Norway, the PANORAMIC Norway trial was led by Oddvar Oppegaard and Nina Langeland at the University of Bergen. Participants received either 300 milligrams of nirmatrelvir with 100 milligrams of ritonavir or placebo twice daily for five days.
Among the 66 participants assigned to nirmatrelvir-ritonavir, 17 (26 percent) reported persistent fatigue, shortness of breath, or cognitive symptoms at three months. Among the 77 assigned to placebo, 33 (43 percent) did. After imputing data for the single missing outcome in the placebo group, the relative risk was 0.60 with a 95 percent confidence interval of 0.37 to 0.98 and a p-value of 0.039.
The trial was originally designed to enroll 2,000 participants. The steering committee stopped it early after recruitment proved impossible to sustain, a consequence of declining COVID-19 testing rates and the widespread perception that the acute phase of the pandemic had passed. With 144 participants, the confidence interval is wide, and the p-value hovers just below the conventional 0.05 threshold.
Five participants in the nirmatrelvir-ritonavir group discontinued treatment due to adverse events. The most common side effect was dysgeusia or ageusia, reported by 57 of 66 participants (86 percent) in the active group compared with 14 of 78 (18 percent) in the placebo group. Nausea or vomiting occurred in 19 (29 percent) of those taking the drug versus eight (10 percent) taking placebo. No severe adverse events were reported.
The second trial, published June 4 in Clinical Infectious Diseases, tested metformin, a generic diabetes drug first approved in the 1950s, in 2,983 adults aged 30 and older from 90 sites across the United States. Conducted within the ACTIV-6 platform and led by Carolyn Bramante at the University of Minnesota, the quadruple-blinded trial randomized participants to 14 days of immediate-release metformin or placebo, with a tapering dose schedule starting at 500 milligrams once daily and escalating to 1,000 milligrams twice daily.
The population was predominantly immune. Eighty-three percent of participants had prior vaccination or prior infection, and the trial enrolled during a period of low disease severity. The median age was 47, and 63 percent were female. Only 2 percent had diabetes. The event rate for long COVID was low: 79 participants (2.6 percent) reported symptoms attributed to COVID-19 at day 180, with no deaths in either group.
The primary outcome measured whether participants reported any COVID-19 symptoms at day 180. In the metformin group, 33 of 1,439 (2.3 percent) reported symptoms, compared with 46 of 1,544 (3.0 percent) in the placebo group. The covariate-adjusted risk difference was 0.8 percentage points lower with metformin, with a 95 percent credible interval of -2.2 to 0.6. The posterior probability of efficacy was 0.83. The prespecified threshold for declaring efficacy was 0.975. The primary outcome was not met.
The secondary outcome told a different story. On day 180, 8 participants (0.56 percent) in the metformin group reported having received a clinician diagnosis of long COVID, compared with 18 (1.17 percent) in the placebo group. The risk ratio was 0.495 with a 95 percent credible interval of 0.155 to 0.995, and the posterior probability of efficacy was 0.96. This 50 percent reduction in clinician-diagnosed long COVID replicated the finding from the earlier COVID-OUT trial, also led by Bramante, which reported a 41 percent reduction using the same metformin regimen.
The divergence between the primary and secondary outcomes in ACTIV-6 reflects a fundamental ambiguity in how long COVID is defined. The primary outcome measured patient-reported symptoms, a broad and subjective endpoint that captures any persisting discomfort a participant attributes to their infection. The secondary outcome measured whether a medical provider had diagnosed long COVID, a narrower and more clinically significant endpoint that requires the participant to seek care and the provider to make the attribution. In a largely immune, low-risk population with low symptom rates, the patient-reported measure lacked the statistical power to distinguish the treatment from placebo, even though the clinician-diagnosis measure showed a clear signal.
Together, the two trials suggest that early intervention can reduce long COVID risk, but through different mechanisms and with different levels of evidence. The Norwegian trial shows that suppressing viral replication during the acute phase with a direct antiviral can cut the risk of persistent symptoms by 40 percent, albeit in a small sample with a strikingly high placebo rate. The ACTIV-6 trial shows that metformin, which acts through anti-inflammatory and host-directed mechanisms rather than direct antiviral activity, reduces the risk of a formal long COVID diagnosis by half in a broad outpatient population, though the effect did not reach the primary endpoint.
Both trials point in the same direction. Neither, on its own, is definitive. The Norwegian trial is too small to generalize from. The ACTIV-6 trial saw its primary signal drown in a low-event-rate population. What they offer together is a starting point: evidence that the biology of long COVID is not fixed after the acute phase, and that drugs selected on different principles (direct antiviral and host-directed) can alter its course.
References
Oppegaard, O., Blomberg, B., Cox, R. J., et al. Nirmatrelvir for acute COVID-19 to prevent long COVID (PANORAMIC Norway): a double-blind, randomised, placebo-controlled trial. The Lancet Infectious Diseases (2026). DOI: 10.1016/S1473-3099(26)00244-6
Bramante, C. T., Stewart, T. G., Boulware, D. R., et al. Metformin on the presence of COVID-19 symptoms 6 months after infection: the ACTIV-6 randomized clinical trial. Clinical Infectious Diseases (2026). DOI: 10.1093/cid/ciag335