A 1,154-patient hospitalized-COVID-19 cohort has produced the largest longitudinal look yet at what happens to the dormant viruses everyone already carries once SARS-CoV-2 infection sets in — and the picture it draws is more specific, and less EBV-centric, than the story that has circulated since early in the pandemic.
The finding, in brief (as of 2026-08-07)
A paper titled “Virus reactivation in acute and long COVID-19,” published in Nature (DOI: 10.1038/s41586-026-10740-z) and widely reported starting 2026-08-05, draws on the NIH/NIAID-funded IMPACC (Immunophenotyping Assessment in a COVID-19 Cohort) study: 1,154 hospitalized COVID-19 patients across 20 U.S. centers, more than 200,000 biological samples, and over a year of multi-omic follow-up. Within the first 40 days after hospital admission, the team detected reactivation of 11 distinct dormant viruses. The most frequently reactivated were Epstein-Barr virus (EBV), herpes simplex virus 1 (HSV-1), cytomegalovirus (CMV), and viruses in the family Anelloviridae — a group of small, non-pathogenic circular DNA viruses that persist latently in an estimated 90% of people and remain, by the researchers’ own description, poorly characterized.
Two findings from that dataset are the actual story. First, reactivation of Anelloviridae specifically — not EBV — was the signal associated with long-term physical disability and long COVID outcomes. Second, EBV and CMV reactivation tracked with markers of systemic inflammation, not the immune-suppression mechanism that has been the default explanation for why a latent herpesvirus would wake up during severe illness.
What a cohort this size lets you see that a smaller study can’t
Most of what’s been published on viral reactivation and long COVID to date has looked at one or two viruses at a time, usually EBV, in cohorts an order of magnitude smaller than this one. That approach can establish an association — elevated anti-EBV antibody titers correlate with long COVID symptoms — but it cannot easily distinguish a specific causal signal from a general marker of a dysregulated immune system, because it has no comparison set of other reactivating viruses measured the same way, in the same patients, on the same timeline.
IMPACC’s design changes that. Because the cohort tracked reactivation across 11 viruses simultaneously, with the same multi-omic sampling protocol, in the same 1,154 patients, the researchers could ask a sharper question than “does EBV reactivate in long COVID” — they could ask which of several co-occurring reactivation events actually tracks with which outcome, and by what apparent mechanism. That’s a design advantage a single-virus study structurally cannot offer, independent of sample size alone. See our guide to cohort study design and our comparison of observational studies vs. randomized designs for the general version of this trade-off: observational cohorts trade randomization for scale and follow-up depth, and scale is exactly what let this analysis separate an inflammation-driven pattern (EBV, CMV) from a distinct, disability-associated one (Anelloviridae).
What specifically is now superseded
The EBV-reactivation-causes-long-COVID narrative that has circulated since early in the pandemic was never absurd — smaller studies had real, repeated evidence of elevated EBV markers in long COVID patients, and immune suppression during acute severe illness is a well-established mechanism for herpesvirus reactivation in other contexts (transplant medicine, critical illness). What this cohort supersedes is not that EBV reactivation happens — it clearly does, and frequently — but the assumption that (a) EBV reactivation is the primary or most specific viral signal tied to long COVID, and (b) immune suppression is the mechanism driving it. On this dataset, the disability-associated signal sits with Anelloviridae, and the EBV/CMV reactivation that does occur correlates with inflammation rather than suppression. That is a meaningfully different mechanistic story, not a minor refinement of the old one — a research audience relying on the EBV framing for hypothesis generation, biomarker selection, or grant narrative should treat it as outdated as of this paper.
What remains genuinely open
The researchers’ own stated limitation is the one to hold onto: the IMPACC cohort was hospitalized, unvaccinated, and infected predominantly with ancestral SARS-CoV-2 strains. Whether the same reactivation pattern — Anelloviridae tracking disability, EBV/CMV tracking inflammation rather than suppression — holds in a largely vaccinated population with milder infections and current variants is not established by this paper and cannot be assumed from it. A single large cohort, however well designed, is also a single cohort; the anellovirus-long COVID association is a strong signal in this dataset, not yet a replicated, mechanistically resolved finding. Anelloviridae themselves remain poorly understood as a virus family generally, which limits how far any single study — including this one — can go in explaining why their reactivation would track with disability rather than being an incidental bystander marker of overall immune disruption.
Why the framing matters, not just the finding
The gap between “EBV reactivation is associated with long COVID” and “EBV reactivation, via immune suppression, drives long COVID” is exactly the kind of overconfident narrowing that a five-times-smaller evidence base tends to produce, and that a single large, well-instrumented cohort can correct — but correcting it shouldn’t mean swapping one tidy mechanistic story (EBV/suppression) for another (anellovirus/disability) with equal confidence. The honest read of this paper is that it substantially weakens the specific EBV-suppression mechanism as previously framed, raises Anelloviridae as a genuinely new and better-supported candidate signal, and leaves the underlying biology of anellovirus reactivation itself as an open question. For institutions and funders tracking where post-acute COVID research is heading, that combination — a large-cohort correction plus an explicitly unresolved mechanism — is the more accurate summary than either “long COVID’s viral cause has been found” or “nothing has really changed.”
Sources
- Maguire et al., “Virus reactivation in acute and long COVID-19,” Nature (2026). DOI: 10.1038/s41586-026-10740-z; PubMed: PMID 42557313.
