Does SARS-CoV-2 stick around? Circulating antigen, tissue RNA, and T-cell signatures compared across Long COVID, recovered, and uninfected groups – including where studies directly disagree, and what antiviral trials have (and haven't) shown.
Source: "Blinded 2-Year Longitudinal Evaluation of SARS-CoV-2 Antigenemia in Long COVID," Clin Microbiol Infect 2026. This study found no significant association between antigen positivity and symptom number, symptom type, antibody titers, or vaccination status – see the conflicting-evidence card below.
Source: Swank Z et al., "Measurement of circulating viral antigens post-SARS-CoV-2 infection in a multicohort study," Clin Microbiol Infect 2024;30(12):1599–1605, PMID 39389851.
These findings are real and cited, but reported as direction/effect rather than a box-and-whisker plot – either because the underlying measure is categorical (tissue positivity, T-cell fold-change over time) or because two studies directly disagree and averaging them would misrepresent both.
An Australian longitudinal cohort (ADAPT; mild/moderate ancestral-strain infection, compared with matched recovered controls) found spike-specific CD4+ T cells 3.9× higher in Long COVID at 3 months post-infection (falling to 2.2× at 8mo, 1.5× at 24mo), and CD8+ T cells 2.8× higher at 3mo, rising to 5.6× at 8mo, with exhaustion markers (PD-1/TIM-3) elevated at 3–8 months before normalizing by 24 months.
Conflicting finding: A study of 86 healthcare workers with mild/asymptomatic infection found no difference in spike/N antibody, neutralization, or T-cell responses between those with vs. without persistent symptoms at 18 weeks or 1 year – suggesting the T-cell signal may depend on index infection severity and cohort selection.
In a cohort of 107 Long COVID patients, anti-N IgG was significantly lower in those with severe fatigue vs. non-severe fatigue (34 vs. 100 U/mL, p<0.001), and remained an independent predictor of symptom improvement at follow-up (OR 0.337, 95% CI 0.120–0.946, p=0.039). Anti-spike showed no persistent association.
Why this matters: a simple "more antibody = more antigen = worse symptoms" model predicts the opposite direction. The authors interpret this via vaccine-boosted clearance of residual antigen – the empirical relationship between antibody trajectory and persistence is more complicated than it first appears.
Colorectal biopsies in 5 Long COVID patients found single-stranded spike RNA in 4/5 (80%) at 158–676 days post-infection, with double-stranded (replicating) RNA in 3/4. A separate endoscopy study of 249 patients (median 17 months post-infection) found nucleocapsid protein in 31% of biopsies overall (37.3% upper GI vs. 16.9% lower GI).
Limitation: the 5-patient biopsy study had no non-Long-COVID comparator, so no direct symptom association could be statistically tested. The 249-patient endoscopy study did not use validated Long COVID diagnostic criteria, so it cannot establish a symptom link either – both are consistent with persistence, neither proves it drives symptoms.
Every controlled trial testing an antiviral as a treatment for already-diagnosed Long COVID has so far returned a null result.
| Trial | Drug & regimen | N | Outcome measured | Result | Citation |
|---|---|---|---|---|---|
| STOP-PASC | Nirmatrelvir-ritonavir, 15 days | 155 (102 tx / 53 placebo) | Pooled severity of 6 PASC symptoms (fatigue, brain fog, dyspnea, body aches, GI, cardiovascular) at 10 weeks | Null brain fog numerically worse on drug (post-hoc) |
Geng et al., JAMA Intern Med 2024 |
| PAX LC | Nirmatrelvir-ritonavir, 15 days | Decentralized RCT, US-wide | Health outcomes at day 28 | Null more treatment-emergent AEs: 76% vs 55% |
Sawano, Iwasaki, Krumholz et al., Lancet Infect Dis 2025 |
| RECOVER-VITAL | Nirmatrelvir-ritonavir, 15-day & 25-day arms | Large platform trial | PEM, cognitive dysfunction, orthostatic intolerance | Preliminary results posted ~March 2026; not yet peer-reviewed |
NCT05595369 (registry only) |
These studies gave antivirals during acute infection, not to people who already have Long COVID – a different research question from the trials above. The observational evidence gives a more encouraging (though still modest) signal; the one randomized in-hospital trial with 1-year follow-up (SOLIDARITY Finland) was null.
| Study | Drug | N | Outcome | Result | Citation |
|---|---|---|---|---|---|
| PANORAMIC | Molnupiravir, 5 days, given acutely | 25,783 randomized; 23,008 (89.2%) followed | Symptom severity/duration at 3 & 6 months | Modest positive no difference in hospitalization |
Harris et al., Lancet Infect Dis 2024 |
| Retrospective meta-analysis | Nirmatrelvir/ritonavir, molnupiravir, remdesivir, given acutely | 6 cohorts, ~3.35M participants | PASC occurrence | Modest positive 27.5% lower PASC occurrence, OR 0.725 (95% CI 0.409–0.747) as published; the reported interval is not symmetric around the point estimate on the log scale, so treat the precision with caution |
Choi et al., J Clin Med 2023;12(23):7375 |
| SOLIDARITY Finland | Remdesivir, given during hospitalization for acute COVID-19 | 208 randomized, 181 (92%) at 1yr | Self-reported recovery, QoL, symptoms at 1 year | Null 85% vs 86% recovered |
Nevalainen et al., Nat Commun 2022 |
Reusing these data? Please cite the tracker so others can find the source.
Open Source Medicine Foundation. (2026). Viral Persistence in Long COVID. Research Tracker. https://research.opensourcemed.info/long-covid-viral-persistence.html (accessed )
New peer-reviewed studies and trial activity across Long COVID, ME/CFS, PACVS, Lyme and Gulf War Illness, every week. Free, no spam.
Subscribe on Substack Browse the digest archive · RSS