Why trust the COVID platforms over the early papers that looked positive?
A review starts with the endpoint, not the brand
Asking 'does it work?' without naming the organism is how a rumor survives. A trial answers a parasite question, a mite question, or a virus question. It does not answer a brand question.
Ivermectin opens a glutamate-gated chloride channel in invertebrate nerve and muscle. Roundworms, filarial worms, and the scabies mite carry that channel. Viruses, bacteria, and human peripheral nerves do not. Stellar results and blank results are the same pharmacology seen from two sides. Sort the claim by organism before you sort it by opinion.
Rank the study type in the same pass. A cell dish is a hypothesis generator. A small open series can fool a feed. A large randomized, placebo-controlled platform is what settles a fight. Parasitic-disease work cleared that bar over decades. Newer viral claims never stayed on it. Keep that ladder in view as the diseases change.
This page is a trial file, not a second monograph and not an origin story. Dates and soil sit in the Kawana paper trail. Chip counts sit on the working card. Here the only question is what each study measured, and whether the measurement survived.
Strongyloides RCTs measured stool cure, not a mood
Intestinal strongyloidiasis is the cleanest single-patient win in the file. Comparative and open-label work behind the US label reported cure in 64-100 percent of infected people after one 200 mcg/kg oral dose. Head-to-heads against thiabendazole and albendazole favored ivermectin on clearance and on how few people quit from side effects. Guidelines did not move it first because it was newer. They moved it because stool went negative more often, with less misery.
Why the bar sits that high: Strongyloides stercoralis can hide for years, then explode when steroids or chemotherapy drop immunity. Hyperinfection still kills. Teams screen before immunosuppression, treat, and confirm clearance on repeat stool - one sample can miss a light load. That workflow rests on repeated comparative success, not a single viral chart.
Claim on the board: 'the old benzimidazoles already did this.' Label the trials. Ivermectin cleared more people and was easier to finish. Strike the 'just newer' line. File first-line status as a measured result. The labeled job is intestinal, nondisseminated disease. Disseminated hyperinfection is a specialist problem, not a one-tablet anecdote.
Onchocerca programs counted larvae, then sight, then maps
One swallow does not retire an adult Onchocerca worm. Trials said so early. The measured job is killing microfilariae - the larvae that crawl skin and eyes - and holding adult output down for months.
Randomized, double-blind work in endemic West Africa, including comparisons with diethylcarbamazine, put more than a thousand people on the record. Skin-snip larval counts fell hard and stayed down for months after a single oral dose. Eye counts moved more slowly, which is why programs watch the anterior chamber and do not treat the tablet as an instant ophthalmic cure. Adult worms in nodules live on. Repeat dosing is biology, not a weak drug.
Proof then scaled past clinic endpoints. Mass donation rounds over decades pushed transmission down region by region across Africa and the Americas. WHO built its onchocerciasis strategy around this molecule because community programs, not a blog series, changed maps. That is a stronger evidence pile than most medicines ever collect.
Strike the 'one dose, forever cured' line. File larval suppression plus scheduled repeats. The origin of the donation that made those repeats possible is in the 1987 gift file. This section only cares that the endpoint was larvae, then blindness averted, then transmission.
Scabies head-to-heads counted mites, then households
Scabies trials favor oral ivermectin once you read the logistics, not the brand loyalty. Two weight-based doses about a week apart clear most ordinary cases because eggs survive the first pass. Permethrin cream still leads many guidelines for uncomplicated disease; head-to-heads show similar cure when cream is applied with care. Pills win when many people must be treated at once and a full-body cream is a fantasy.
Outbreaks in care homes, prisons, and crowded houses are where oral dosing pulls ahead in real life. Community studies report sharp drops in prevalence when the whole group is dosed together. Crusted scabies in immunocompromised hosts is a different measurement: huge mite loads, so programs pair oral drug with a topical and plan multiple rounds. One lonely tablet is a relapse plan.
US oral use for scabies is off-label. That is a regulatory fact, not a verdict that the mite trials are imaginary. File the mite counts. File the second-dose clock. Do not file a claim that a cream is 'always inferior' or that a pill is 'unproven.' Practical clocks live on the working card.
Lymphatic filariasis measured a combo, not a solo hero
Mass drug administration for lymphatic filariasis often pairs ivermectin with albendazole. Trials and program data support the pair. They do not support ivermectin as a lone hero for every worm on a travel poster. Tapeworms and flukes need other drugs. Malaria and other protozoa need theirs. The channel is narrow. A 'universal dewormer' claim fails the spectrum table on the reference page.
Smaller files exist for cutaneous larva migrans and, in some settings, head lice. Those studies are real and usually smaller. Oral ivermectin is often second-line to a topical when a topical will reach the bug. Strike the habit of treating every small series as if it were a Strongyloides-sized RCT. File each use at the size of its evidence.
A 2020 dish number never became a human dose
A culture paper in 2020 showed slower coronavirus replication in a dish. That is a fair research start. It is not a treatment. The active concentration sat many times above what a safe human dose produces.
Reaching those dish levels in a person would already be a toxicology problem - confusion, collapse, seizures - not a clever 'higher dose' trick. A molecule that only works at poisonous exposure is not a medicine you can prescribe. Lab hints die at the human step all the time. That is how the ladder is supposed to work.
Early upbeat clinical reports were mostly small, messy, or, in one famous case, pulled over data integrity. Meta-analyses that kept those weak papers produced a wobbling 'signal.' The signal collapsed when the poor studies left the pile. That pattern is ordinary. It is not a cover-up. It is what happens when a feed outruns a protocol.
Regulators stated the human conclusion without poetry. FDA pages at fda.gov say there is no benefit for COVID-19 at doses people can take. Strike the dish-to-dose leap. File the dish as a hypothesis that failed the next rung.
TOGETHER and ACTIV-6 measured hospitals. The gain was empty
| Tier | What it suggested | How it held |
|---|---|---|
| 2020 cell culture | Slower viral replication in a dish | Only at toxic, unreachable levels |
| Small early series | Hints of benefit | Small, messy; one pulled for data |
| Metas of weak trials | Unstable mixed signal | Collapsed when poor studies left |
| TOGETHER, ACTIV-6 | Hospital and recovery endpoints | No meaningful gain vs placebo |
Large randomized platforms answered the human question with the endpoints that matter to a ward. TOGETHER, published in the New England Journal of Medicine in 2022, compared ivermectin with placebo in outpatients and did not find a meaningful drop in hospitalisation or prolonged observation. ACTIV-6, a US platform reported in JAMA, tested time to recovery and related clinical outcomes and did not find a benefit you would take to a bedside.
Those platforms are why this desk strikes the 'they never tested it' line. They tested it. They used placebo. They used thousands of people. They used hospital and recovery endpoints, not a social-media vibe. When a small early paper and a large platform disagree, you follow the platform. That is not politics. That is the ladder.
The episode is a textbook of evidence tiers doing their job. Excitement outran data. Better data closed the file. Correction is not conspiracy. People who needed a worm drug still needed it on Monday. People who wanted an antiviral did not get one from this molecule at a safe dose.
How this desk files a mixed reputation
| Indication | What studies measured | Desk file |
|---|---|---|
| Strongyloidiasis | Stool cure vs older agents | First-line; 64-100% after 200 mcg/kg |
| Onchocerciasis | Skin and eye larvae; MDA maps | Suppresses larvae; drives elimination |
| Scabies | Mite clearance vs permethrin | Comparable; oral wins in outbreaks |
| Lymphatic filariasis | MDA with albendazole | Works in combination, not as a solo |
| COVID-19 | Hospitalisation, recovery (platforms) | No benefit at safe doses |
Stack the indications and one pattern repeats: proof tracks the channel, not the hype. Antiparasitic rows rest on randomized series and, for river blindness, population programs. The COVID row rests on a dish result that never scaled, then failed definitive platforms. One molecule, two verdicts, because the biology differs.
People who dismiss the tablet forget first-line Strongyloides, map-changing onchocerciasis work, and outbreak scabies. Those are not modest wins. People who champion it for everything make the opposite error and invite side effects for no gain. The sensible file sits between: respect what the endpoints support, stop where they stop. If a real new use ever appears, it will arrive the way the worm uses did - in large trials that hold up - not in a lone dish or a wave of hope.
Reader mail
Reader questions on this article
Answered by Dr. Marcus Bellinger, MD · Internal medicine & infectious disease
Named readers asked this desk to file trial claims. Endpoints first, then the strike.
Design beats a like-count. Early upbeat papers were mostly small, some poorly controlled, and at least one high-profile report was retracted over data problems. That pattern manufactures false positives, which is why medicine runs large randomized checks. TOGETHER and ACTIV-6 each tested ivermectin against placebo in thousands of people and found nothing you would take to a ward. When tiers disagree, you follow the top tier. FDA summaries at fda.gov say the same thing in plainer language.
If it slowed the virus in a dish, why does that not count as proof?
Dishes are not bodies. The 2020 culture work needed concentrations far above what a safe human dose produces - already in a toxic range. Reaching those levels in a person risks confusion, collapse, and seizures. An effect that only appears at poisonous exposure is not a treatment. Lab hints die at the human step constantly. That is ordinary pharmacology, not a conspiracy. I file the dish as a failed hypothesis, not as a suppressed cure.
Is ivermectin actually better for Strongyloides, or just newer than the old drugs?
Better by measured stool cure, not by calendar age. Head-to-heads against thiabendazole and albendazole showed more people cleared and fewer dropped out from side effects, which is why guidelines moved it first. The US label cites 64-100 percent cure after one 200 mcg/kg dose in the studies behind approval. That matters because this worm can sit quiet for decades, then kill if steroids arrive. Partial clearance is unacceptable here, so teams often confirm eradication after treatment.
For scabies, do the trials say pills or cream?
They say 'it depends on the room.' Permethrin works well if you can cover from the neck down and keep it on. Pills simplify a household, a care home, or crusted disease where many people need dosing together. Either way, schedule a second dose to catch mites that hatch after the first pass. CDC patient steps at cdc.gov spell out the sequence. I do not file 'pills always win' or 'cream is obsolete.' I file logistics plus a hatch clock.
Why do river-blindness programs last years if the drug is so effective?
Effective on larvae, not on decade-long adult worms sitting in skin nodules. One dose clears circulating microfilariae and quiets adult output for months, which protects eyes and skin. When levels fall, adults pump larvae again. Repeat yearly - or as often as every three months in some individual care - until worms age out or community transmission breaks. Long programs reflect biology. They are not an admission that the tablet failed the larval endpoint.
Are there parasites people assume this treats that the trials never supported?
Plenty. It hits many roundworms and the scabies mite. It does not hit tapeworms or flukes. It does not hit malaria or other protozoa. It does nothing useful to bacteria or viruses at a dose a person can take, because those organisms lack the channel. Calling it a universal antiparasitic is the claim. Matching drug to a named diagnosis is the file. MedlinePlus overviews at medlineplus.gov help with specific names before a visit.
Given the noise, is the safety record actually any good?
For labeled parasitic uses at weight-based human doses, the record is long - including mass campaigns with billions of treatments. Headlines came from veterinary paste, overdoses, and COVID misuse, not from standard Strongyloides or scabies care. One real caution: heavy Loa loa loads in parts of Central and West Africa can trigger severe neurologic reactions, so programs screen there. Correct product, correct dose, correct indication: low drama. Wrong product or no parasite: risk rises fast.
Did anyone actually run a proper COVID trial, or is that a talking point?
Proper trials ran. TOGETHER randomized outpatients and used hospitalisation as a hard endpoint. ACTIV-6 randomized and used recovery time and related outcomes. Both used placebo. Both were large. Both were empty for this drug. 'They never tested it' is the line I strike. 'They tested it and the endpoints did not move' is the file. If a later platform with a different dose or population ever shows a real gain, it will have to beat those studies, not a screenshot.
If I only keep one sentence from this review, what should it be?
Attach every ivermectin claim to an endpoint: stool for Strongyloides, larvae and maps for Onchocerca, mites for scabies, hospital and recovery for COVID. The first three files are thick. The last one is empty at a safe dose. Bring a named parasite to your own clinician. Do not bring a viral post and ask for a tablet. Chip math, if a clinician has already named the worm, is on the working card.
General education from a clinician, not personal medical advice. Bring your own history to your own prescriber.
Last Updated