Spike Protein Detox: What the Studies Actually Show

What spike protein is, and where it comes from

Spike protein is the protein on the outside of the SARS-CoV-2 virus that attaches to the ACE2 receptor on human cells so the virus can enter. You are exposed to it in two ways. An infection makes the virus, and the virus carries spike. An mRNA vaccine delivers a short instruction that tells your own cells to make spike for a period, so the immune system learns to recognize it. In mice given spike mRNA in lipid nanoparticles intravenously, spike expression peaked at about 6 hours, mostly in the liver, while the carrier lipid persisted in tissue for weeks.[20]PubMed The molecule is the same either way, so a blood test for spike cannot tell you which route it came from. History tells you: whether you were infected, and when. That is why one 2026 case report used a second marker, an antibody to the virus's nucleocapsid protein, which an mRNA vaccine does not produce, to argue that the spike it found came from vaccination rather than infection (Hulscher et al. 2026).

How long it stays: the measured record, paper by paper

For most people the measured answer is days to weeks after vaccination and weeks to months after infection, with a few reports of longer. Each study below is listed with what it found and what it did not show, because the second half is where most of the argument happens.

• Ogata 2022, people after mRNA vaccination, blood: vaccine spike antigen in plasma in the days following a dose. It did not follow anyone for years (PMID 34015087).
• Roltgen 2022, seven vaccinated people, lymph node biopsies: vaccine mRNA and spike in lymph node germinal centers up to about 60 days. Nobody was sampled beyond day 60 (PMID 35148837).
• Krauson 2023, autopsy tissue: vaccine mRNA in nearby lymph nodes within about 30 days; spike itself was not seen by stain (PMID 37758751).
• Yonker 2023, teenagers and young adults with myocarditis after a dose, against vaccinated controls: free spike in blood averaged 33.9 pg/mL in the myocarditis group, higher than in the controls. The authors state that cause is unproven.[7]PubMed
• Swank 2023 and Peluso 2024, people with long COVID after infection: spike or viral antigen in the blood of a minority, out to about 12 to 14 months. About 2 percent of one control group also tested positive at baseline. No link between a level and a symptom was shown (PMID 36052466, PMID 38604216).
• Hulscher 2026, one 55-year-old man, three Pfizer doses: plasma spike at 129 femtograms per mL on day 1,173, not detectable on day 1,284; spike mRNA in exosomes at day 1,284; spike in skin biopsies. One patient, no comparison samples (DOI 10.18103/mra.2026.0351).

Two things to hold while reading those

First, detecting a molecule is not the same as showing it did harm. None of these papers ties a measured level to an outcome in a controlled way, and the Swank and Yonker authors say so in their own text. That is not a reason to dismiss the finding; it is the reason the next study needs funding. Second, the case report's authors list their own limits: one DNA signal "cannot be definitively assigned", the genome findings come from a single sample, and a positive-control stain is reported as data not shown. A single patient cannot give a rate. What a single patient can do is establish that something is possible, and that is what this one does.

Can you test for it?

Not routinely. Most labs offer an antibody test, which measures your immune response to spike, not the amount of spike in your blood. A number like 25,000 on a report usually comes from an antibody assay with units such as U/mL or AU/mL. In the 2026 case report the man's spike antibody level was 4,553 U/mL while his free spike protein, measured by a research assay, was 129 femtograms per mL. Different test, different question, different unit, and five orders of magnitude apart. If you have a result in hand, ask the lab three things: is this an antibody or the protein itself, what is the unit, and what is the reference range in people without symptoms. The research assays that measure free spike are not standard clinical tests, and no published study shows that a given level calls for a given treatment. The free Lab Atlas explains what a marker does and does not tell you, for this and for every other test on a panel.

If you were never vaccinated

Infection makes spike. The long COVID papers above followed people who had been infected, and they found antigen in the blood of a minority for months. Whether you were vaccinated, infected, or both, the same molecule is involved and the same question applies: how long does it last in you, and does it matter. A person who never had a noticed infection can still have had one.

The first school: proteolytic enzymes

This is the protocol Peter McCullough and colleagues set out in a 2023 review: nattokinase at 2,000 fibrinolytic units twice daily, bromelain 500 mg daily, curcumin 500 mg twice daily.[3]PubMed The review proposes the protocol and calls for trials. It does not report one, and the authors do not claim it does. Nattokinase is an enzyme from fermented soybeans. Here is the whole measured record, positives and nulls together.

• Cell lysate, test tube: nattokinase degraded spike made by transfected cells, from 7.8 ng/mL up to 32 ug/mL. It also degraded ACE2 and an unrelated enzyme, and three of the ten authors work at the enzyme's supplier.[1]PubMed
• Cell culture: a whole natto extract degraded the receptor-binding piece of spike and reduced infection. Nattokinase itself was never isolated from the extract, and the work was funded by a natto maker (PMID 34271432).
• Eleven people, one capsule: a binding assay showed a signal in blood peaking around 13 hours. It did not show that active enzyme reached the blood (PMID 23709455).
• Six trials, 546 people: blood pressure fell about 3.5/2.3 mmHg, limited by small trial sizes.[4]PubMed
• Eighty-six people, eight weeks, double-blind: lower systolic and diastolic pressure than placebo.[5]PubMed
• Two hundred and sixty-five people, 2,000 units a day for three years: no difference in carotid artery thickness or blood pressure (PMID 33843667).
• One hundred and twenty people, 8,000 units a day for six months: no improvement in global cognition.[6]PubMed

What that school has, and what it is still missing

A nattokinase review currently in press at the same journal as the case report states that three laboratories have reported spike degradation and that human pharmacokinetics show systemic exposure.[2]DOI No study has measured spike in a person taking nattokinase. Bromelain, from pineapple, reduced infection of cells at 9.5 to 75 ug/mL in a 2021 letter, while plasma after 3 g a day was about 5 ng/mL, thousands of times lower. A placebo-controlled German registry trial in 77 people stopped for futility. Curcumin binds spike in computer models and cells responded at about 27 micromolar, but one human study found no curcumin in serum after single doses up to 8 g, and its COVID trials are small and mixed. So: real evidence for clotting markers and modest blood pressure effects, and solid bench evidence that nattokinase cuts spike in a dish. What is absent is a human measurement of spike and absorption data showing the active enzyme reaches blood at the doses people take. Safety is case-level and worth knowing: cerebellar bleeding alongside aspirin, one fatal bleed, a clot on a heart valve when it replaced warfarin, and natto's vitamin K2 interfering with warfarin.

The second school: the nicotinic receptor

The idea, set out in a 2020 paper by Jean-Pierre Changeux and colleagues, is that a short stretch of spike resembles snake neurotoxins and may bind nicotinic acetylcholine receptors, which regulate inflammation through the vagus nerve. If so, a molecule that sits on those receptors might compete with spike. It is a hypothesis paper with no new laboratory data (PMID 32720486). The receptor biology underneath it is not in doubt. The alpha-7 subunit is required for the vagus nerve's anti-inflammatory effect in mice,[9]PubMed and cholinergic control of inflammation is an established field.[10]PubMed A review in our library notes that several groups have reported SARS-CoV-2 interacting with the alpha-7 receptor, which the authors suggest may disrupt the vagal reflex that restrains cytokine release.[11]PubMed The open question is whether spike meaningfully occupies those receptors in a living person.

• For: Oliveira 2021, Lagoumintzis 2021 and Farsalinos 2020, computer models and docking, predict binding of a spike segment. The authors call it a theoretical model (PMID 33609494, PMID 32823591).
• For, with a twist: Chrestia 2022, O'Brien 2023 and Hone 2024 and 2026, in cells and oocytes using short spike peptides, saw alpha-7 currents change, mostly as negative or allosteric modulation rather than at nicotine's own site (PMID 35859025, PMID 37061001).
• Against: Godellas 2022, intact human alpha-7 receptors on cells, with whole S1 and S1-S2 and both positive and scrambled controls, found spike did not appreciably compete with the toxin at the receptor and saw no clear effect on currents. NIH-funded (PMID 36279466).
• Nicotine itself, in cells and mice: no protection in one cell test, higher ACE2 and infection in lung cells in another, and less brain virus but no change in illness in mice (PMID 35980910, PMID 33850935).

Nicotine in people: the three pieces that exist

This is the part that circulates furthest from its source, so here it is at full strength with its own authors' words.

• Leitzke 2023: a single-author hypothesis article describing four people with long COVID treated with nicotine patches at 7.5 mg a day, 15 mg for two of the men. No control group, symptoms self-scored, and one man stopped after ten hours with vomiting and diarrhea. The author writes that the results "do not allow for general conclusions" and that a psychosomatic component "cannot be safely ruled out" (PMID 36650574).
• Leitzke 2025: a retrospective survey of 231 people who chose to use patches, with 73.5 percent reporting improvement. The paper describes itself as retrospective and not randomized controlled (PMID 40011942).
• Labro 2022: the only randomized trial, in 218 ventilated COVID patients in intensive care. Day-28 mortality was 28 percent on nicotine and 28 percent on placebo, and the authors concluded there is no indication for nicotine in that setting. It was not a long COVID trial and does not speak to one (PMID 35676335).

Ingredient by ingredient, at the doses actually studied

Formulas built on the nicotinic idea share a common ingredient list. Most print a proprietary blend without per-ingredient amounts, which means no dose comparison against any study below is possible. What follows is what each ingredient has been measured doing, and what it interacts with, so the list can be taken to a pharmacist in one piece.

• Lobelia: its alkaloid lobeline binds nicotinic receptors, and at the human alpha-7 subtype specifically it acts as an antagonist rather than an agonist (PMID 9833639). Oral absorption is poor. In 750 smokers it had no effect on cessation (PMID 19663757); in nine adults with ADHD there was a modest working-memory signal.[13]PubMed A PubMed search pairing lobeline, lobelia, mucuna, wormwood or thujone with spike returns zero records.
• Licorice: glycyrrhizin inhibited the viral main protease in cell work at 0.44 mg/mL,[15]DOI while a second group found the protease was not inhibited until 1 millimolar and reported an effect on entry through ACE2 instead, noting blood levels near 1 micromolar. Docking work adds binding predictions.[16]DOI Two small human trials of licorice extract found no difference on most outcomes. Around 100 mg of glycyrrhizin a day is where blood pressure and potassium effects begin.
• Mucuna: not a nicotinic agent. Its active is L-dopa, the direct precursor of dopamine. L-dopa content of commercial products varies from 2 to 241 mg per serving. It interacts with MAO inhibitors (PMID 35939305).
• Lemon balm: extracts bind nicotinic and muscarinic receptors in human brain tissue in laboratory tests, though binding varies considerably by plant batch. Human trials are small and mixed.
• Citicoline: supplies choline, which is a weak cholinergic agonist about a tenth as potent as acetylcholine and needing concentrations well above blood levels.[12]PubMed Twelve weeks improved episodic and composite memory in healthy older adults;[14]PMC the large 2,298-person stroke trial was null.
• Wormwood: thujone blocks GABA-A and inhibits human alpha-7 non-competitively. The European guidance is about 3 mg of thujone a day for up to two weeks. The artemisinin literature is a different plant.
• Copper as cuprous nicotinic acid: a copper compound of niacin. Copper is an essential mineral with an adult recommended intake of 0.9 mg a day and an upper limit of 10 mg.

C60, gold and structured water, measured rather than waved off

C60 is a sixty-carbon cage molecule dissolved in oil. A 2012 rat study reported nearly doubled lifespan at 1.7 mg/kg (PMID 22498298); a 14-day oral rat study at 3.8 mg/kg found no harm, funded by a C60 maker; a mouse genotoxicity study found none.[17]PubMed Against that: C60-treated mice lived no longer than water-treated mice,[18]DOI online oils were off their label by up to 38.5 percent and light-degraded oil killed mice, and a review found no long-term human safety research.[19]DOI Human oral studies of C60: none found. If a serving were pure oil at 0.8 mg/mL, a 1 ml dose would sit roughly 150 times below the rat lifespan dose per kilogram. Gold has a real medical precedent. Auranofin, an oral gold drug at 6 mg a day, is an approved arthritis treatment carrying a gold-toxicity warning and diarrhea in 47 percent of patients. Gold nanoparticles by mouth caused no toxicity in 60- and 90-day rodent studies, and an ALS gold nanocrystal trial showed no benefit. The phrases monoatomic gold and ORMUS return zero PubMed records each, and no independent assay of such a product was found. Structured water: work on an exclusion zone of ordered water exists, and so do papers explaining the same observations through ion exchange and diffusion. No human trial of drinking structured water was found.

Ivermectin, fasting and autophagy

Ivermectin binds spike in computer models and blocked a virus in one cell assay, but the drug level needed in that assay exceeds what blood reaches at approved doses. The large randomized trials, Lopez-Medina, TOGETHER, ACTIV-6 twice and COVID-OUT, found no benefit on their main outcomes; one open-label trial, PRINCIPLE, found recovery about two days faster by self-report; meta-analyses split, and the Cochrane review of eleven trials found little or no effect.[8]PubMed No study measured spike with ivermectin. Fasting and autophagy: the cell-biology papers contradict each other on whether spike turns autophagy on or off. A four-day human study saw a blood autophagy marker rise (PMID 31151228), while another found markers up in mouse liver but not in the muscle of mice or people (PMID 35660501). No study measured spike in a fasting person.

Combining them, and the interactions that actually matter

No study has tested a combination against spike, so no pairing is evidence-based for that purpose. What can be said is what each agent does to something else, because that is where combinations bite. Nattokinase, bromelain and curcumin each thin the blood a little and stack with aspirin or anticoagulants. Mucuna's L-dopa interacts with MAO inhibitors. Licorice raises blood pressure and lowers potassium at 100 mg of glycyrrhizin a day or more. Citicoline and lemon balm act on the same acetylcholine system the nicotinic school is aiming at. A pharmacist can check a full list against a prescription in about a minute. That is worth doing before any stack, and it costs nothing.

How to read a spike-detox claim on any product

By law a dietary supplement may describe how an ingredient affects the structure or function of the body, but may not claim to diagnose, treat, cure or prevent a disease. The FDA rule lists the ways a statement becomes a disease claim, including having a role in the body's response to a disease or to a vector of disease, and naming the symptoms of a condition. A page saying a product detoxes a viral protein, or relieves persistent symptoms, is making a claim of that kind. The Federal Trade Commission's guidance says health claims need competent and reliable scientific evidence, that randomized human trials are the most reliable form, and that dish and animal findings are not sufficient on their own. Read every protocol against that standard, including the ones in this post. The reason to know the rule is not to be timid about the science. It is so you can tell the difference between a company that has evidence and one that is counting on you not to check.

What would settle this

One study. Take people with a measured spike level, randomize them to a protocol or placebo, measure spike again at a fixed interval, and publish whatever comes out. It is not an expensive trial by the standards of the field, and the assays already exist, because the persistence papers used them. Until someone runs it, every protocol on the market, including the ones described here, rests on mechanism, tradition and uncontrolled reports rather than a measurement. That is worth saying plainly in both directions: the institutions that will not fund the study and the sellers who do not need it to sell are, between them, the reason the question is still open.