How to read this page: Each model section states its core causal claim, then lists the treatments that follow logically from believing that model — with the rationale connecting mechanism to intervention made explicit, and the actual trial evidence (positive, null, or mixed) reported honestly rather than smoothed over. A null trial result for one model's treatment says nothing about whether a different model is correct. Most patients likely have more than one mechanism active at once; see
Long COVID Hypotheses for how these combine, and
Endotypes for how they cluster into patient subgroups.
SARS-CoV-2 antigen (spike/nucleocapsid) or RNA persists in tissue reservoirs after acute infection, providing an ongoing antigenic stimulus that drives chronic immune activation and downstream symptoms.
Treatments this model implies
If persistent antigen is driving symptoms, clearing it with an antiviral should resolve them.
Two completed RCTs treating established Long COVID found no significant symptom benefit; one found brain fog numerically worse on the drug.
A distinct prediction from the same model: if persistence is seeded during acute infection, clearing virus faster and more completely during that window should reduce the rate of later persistence.
A 25,783-participant RCT of molnupiravir given acutely found reduced symptom severity/duration at 3–6 months, though no difference in hospitalization. This answers a different question from the two null trials above — prevention, not treatment of existing Long COVID — and the two should not be conflated.
Persistent T-cell activation/exhaustion, altered B-cell populations, and autoantibodies targeting GPCRs (beta-adrenergic, muscarinic receptors) drive ongoing symptoms independent of any remaining virus.
Treatments this model implies
If autoreactive B-cells producing pathogenic autoantibodies are driving disease, depleting them should improve symptoms.
The largest randomized, double-blind, placebo-controlled trial (ME/CFS, not Long COVID specifically, but the same rationale applies) found no significant benefit over placebo.
Targets the specific proposed autoantibodies directly, rather than depleting B-cells broadly.
A 2025 phase IIa randomized, placebo-controlled, double-blind crossover trial ("reCOVer") reported safety, tolerability, and clinical effects on fatigue and quality of life in Post-COVID syndrome — an early positive signal, not yet a large confirmatory trial.
Proposed to modulate microglial and broader immune activation at low doses, independent of its opioid-receptor activity.
A pilot study combining low-dose naltrexone with NAD⁺ reported improvement in persistent fatigue symptoms after COVID-19 — a small, early signal.
Glutamine fuels lymphocyte and intestinal-epithelial metabolism and maintains gut tight junctions; if gut barrier breakdown is feeding systemic immune activation, restoring glutamine availability should reduce that upstream driver.
Plasma glutamine is depleted in PASC two years post-infection. A small acute-COVID-19 RCT (10g × 3/day) reduced hospitalization duration; multiple RCTs in critical illness/post-surgical settings reduce infection rates and ICU stay. No completed Long-COVID-specific symptom trial was identified — this is extrapolation from a related but not identical population.
One of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol
This is the load-bearing assumption behind both BC007 and part of the case for immunomodulation — worth stating plainly that it is not settled.
One Charité-group study found elevated beta-adrenergic/muscarinic receptor autoantibodies in a CFS subset; a larger 2025 replication using different profiling technology found no significant difference from controls.
Fibrinaloid microclots, platelet activation, and endothelial dysfunction impair peripheral oxygen delivery and extraction, producing tissue hypoxia that drives fatigue and exercise intolerance.
Treatments this model implies
If microclots are mechanically obstructing capillary flow, dissolving or preventing them should restore perfusion.
No completed randomized controlled trial of anticoagulant or antiplatelet treatment based on this hypothesis was identified. Clinical bodies have publicly cautioned against unlicensed anticoagulant/apheresis treatment pending such evidence. This should be read as an untested extrapolation, not a validated treatment.
Proposed fibrinolytic action could degrade microclots directly; separately proposed to degrade spike protein itself, giving it a second rationale under the viral-persistence model too.
In vitro evidence shows nattokinase degrades SARS-CoV-2 spike protein. This is cell-free/enzymatic evidence, not a clinical trial in Long COVID patients.
Citrulline is converted to arginine and then nitric oxide, bypassing first-pass hepatic arginase breakdown; if reduced NO bioavailability from endothelial dysfunction is limiting perfusion, restoring the NO precursor pool should improve vascular flow.
Citrulline reliably raises plasma arginine/NO markers in general and athletic populations. A dedicated Long-COVID-POTS trial (COVIVA, Mount Sinai) is ongoing but not yet completed; the completed COVID-19 trials identified were in acute ARDS, not Long COVID fatigue.
One of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol
Direct substrate for nitric oxide synthase; combined with vitamin C (which reduces oxidative breakdown of NO), the pairing targets both NO production and NO preservation.
A randomized, placebo-controlled trial of L-arginine + vitamin C in Long COVID (n=56, 28 days) found a large reduction in fatigue (8.7% vs. 80.1% placebo, p<0.0001) and improved 6-minute walk distance. This is the single strongest piece of disease-specific RCT evidence behind any treatment on this page — though it tested the combination, not arginine alone.
L-Arginine and Vitamin C are both included in the OSMF-VS4P-001 formulation — see
The Protocol
Impaired oxidative phosphorylation, redox imbalance, and NAD⁺ depletion reduce cellular energy production, directly limiting the capacity for sustained physical and cognitive activity.
Treatments this model implies
CoQ10 is a required electron carrier in the respiratory chain; supplementing it should support oxidative phosphorylation directly if endogenous levels are limiting.
A prospective, randomized, double-blind, placebo-controlled trial in ME/CFS found significant improvement in fatigue perception and quality of life with CoQ10 plus NADH.
Phosphocreatine buffers cellular ATP supply during high demand; if the resting phosphocreatine pool is depleted, supplementation should improve capacity to meet exertional energy demand.
A trial of six-month creatine supplementation reported improved patient- and clinician-reported outcomes and tissue creatine levels in post-COVID-19 fatigue syndrome.
Creatine monohydrate is one of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol for dosing rationale
If NAD⁺ depletion is limiting mitochondrial function, restoring it with a precursor should improve energy metabolism and symptoms together.
A randomized controlled trial in Long COVID confirmed NAD⁺ levels rose with nicotinamide riboside supplementation, but did not find a significant effect on symptoms or cognition — NAD⁺ restoration alone was not sufficient in this trial.
The OSMF-VS4P-001 formulation uses NMN, a related but distinct NAD⁺ precursor, not the nicotinamide riboside tested in the trial above — see
The Protocol
Carnitine shuttles long-chain fatty acids into mitochondria for beta-oxidation; if impaired fat oxidation (documented on CPET in PASC) reflects a carnitine-transport bottleneck, restoring carnitine should restore fat-burning capacity.
A PASC rehabilitation RCT found no significant benefit at 1g/day on dyspnea or functional scores — a real null result at the dose tested. Long-chain acylcarnitines are independently documented as elevated in PASC plasma, consistent with a transport bottleneck, but that finding doesn't establish that supplementation reverses it.
One of the 10 ingredients in the OSMF-VS4P-001 formulation, at a higher (2g/day) dose than the null trial — see
The Protocol
NAC is the rate-limiting precursor for glutathione, the primary cellular antioxidant; if oxidative damage to mitochondrial membranes and proteins is driving dysfunction, replenishing glutathione capacity should reduce that damage.
A small Long COVID case series (n=9) reported normalized von Willebrand factor in NAC users vs. controls and symptom improvement — an early signal, not a powered trial. Broader NAC evidence (COVID-19 mortality meta-analysis, acute-COVID respiratory outcomes) is real but concerns different endpoints and populations than Long-COVID fatigue specifically.
One of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol
Taurine stabilizes inner mitochondrial membranes, modifies mitochondrial tRNAs for respiratory-chain fidelity, and modulates cardiac calcium handling — relevant to both energy production and the autonomic/cardiac symptoms that overlap with this model.
Plasma taurine is inversely correlated with PASC severity in an observational cohort (148 μM severe vs. 225 μM recovered) — a real association, not an interventional result. A dedicated Long-COVID RCT is ongoing but not yet completed; taurine's RCT evidence base is strongest in chronic heart failure, a different population.
One of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol
Serine availability regulates mitochondrial fission/fusion dynamics and feeds one-carbon metabolism for methylation; serine deficiency has been linked mechanistically to mitochondrial fragmentation in cell models.
Serine is reported depleted in PASC and inversely correlated with neuropsychiatric symptom scores in one cohort. Supplementation safety (up to 15g/day) is established in unrelated neurological conditions (ALS, hereditary spastic paraplegia); no completed Long-COVID symptom trial of serine was identified.
One of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol
Vitamin C is a required cofactor for two steps in endogenous carnitine biosynthesis, linking it mechanistically to the fatty-acid-oxidation pathway targeted by carnitine above; it is also the primary aqueous antioxidant and reduces oxidative breakdown of nitric oxide.
The strongest disease-specific evidence is the same Tosato et al. arginine + vitamin C trial cited under the vascular model — vitamin C's individual contribution cannot be isolated from that combination trial. Vitamin C deficiency is independently documented as more severe in hospitalized COVID-19 patients.
One of the 10 ingredients in the OSMF-VS4P-001 formulation — see
The Protocol
Sympathetic overdrive and vagal withdrawal drive orthostatic intolerance (including POTS) and a loss of the vagally-mediated cholinergic anti-inflammatory reflex, linking autonomic and inflammatory symptoms.
Treatments this model implies
If sympathetically-driven tachycardia is the proximate cause of orthostatic symptoms, directly lowering heart rate should relieve them.
A randomized crossover trial comparing ivabradine, propranolol, and placebo in POTS found both active agents effective, with some differences between them — real RCT-level evidence, though not Long-COVID-specific.
If loss of vagal tone is disabling the cholinergic anti-inflammatory reflex, restoring vagal signaling should dampen the downstream inflammatory loops as well as autonomic symptoms.
The cholinergic anti-inflammatory reflex is well-established in general neuroimmunology. Whether restoring it specifically improves Long COVID or ME/CFS symptoms has not been demonstrated in a clinical trial identified in this review.
Why null results don't rule models out or in
A null trial tests one treatment's ability to reverse one proposed mechanism in an unselected population. If only a subset of patients have that mechanism operating (see
Endotypes), diluting them into an all-comers trial can produce a null result even when the model is correct for that subset. Conversely, a positive result doesn't confirm the full causal story — CoQ10 improving fatigue is consistent with the mitochondrial model but doesn't rule out other mechanisms operating alongside it. Treat each row above as one data point, not a verdict on the model as a whole.