The design deserves closer attention than the verdict. Three arms against placebo: the full combination, nirmatrelvir on its own, ritonavir on its own [5]. That structure buys something a simple drug-versus-placebo comparison does not. A null on the combination alone always leaves the question of whether one ingredient blunted the other. Separating them closes that off, and a benefit showing up in a single-component arm would have pointed away from the replication blockade the combination is built to produce [3].
The durations tell you what the investigators believed. Fifteen to twenty-five days [5] against the five days used for acute infection [6] is three to five times the standard course [10]. If replicating virus is still sitting in tissue months later, the obvious first move is the same block held for longer. It was tried, and co-principal investigator Lindsey Baden's summary was that treating viral persistence with this antiviral showed no evidence of clinical benefit [7].
The thing this doesn't tell you is how firmly the null is pinned down. The STAT item reports no participant count, no primary endpoint, no effect size, and no follow-up duration [11]. A null across sixty people and a null across six hundred are different objects, and only one of them constrains anything. Those numbers are in the Lancet Infectious Diseases paper [9]; they are the first thing to read.
Nor does this falsify persistence as a mechanism. Nirmatrelvir stops the virus from multiplying [3]. It does not clear viral protein that is sitting in tissue without replicating, and it does nothing about an immune response provoked by such a protein. Long Covid is also a label stretched over a heterogeneous group of patients [2], and a trial that does not select for people with measured antigen or viral RNA can dilute a real subgroup effect toward zero. What was tested, and what came back flat, is this drug, at these durations, in whoever was enrolled [1].
Even so, the burden has moved. My reading, with its condition stated: outside a trial, an extended antiviral course for long Covid is no longer defensible on the persistence rationale, and it becomes defensible again only if a study enriches for participants who have the thing the drug is supposed to remove and shows separation there. Until then, clinicians are back to treating symptoms.
Worth saying plainly: a publicly funded platform study [8] that tests its most plausible candidate and reports that the candidate did not work is doing its job. The result costs the field a hypothesis it can now stop paying for.
