Almost every disagreement about repurposed drugs — ivermectin, fenbendazole, mebendazole — is not really a disagreement about biology. It is a disagreement about evidence quality. One person has read a study showing a compound killing cancer cells; another has read that no human trial has demonstrated benefit. Both are describing real documents. They are simply describing documents that carry very different weight.

The skill that resolves this is not scientific training. It is knowing how to open a paper and ask a short list of questions. This article teaches that list.

1. What kind of study is this?

This single question does more work than all the others combined. The word "study" covers wildly different things, arranged in a rough hierarchy of confidence:

  • In vitro — cells in a dish. Cheap, fast, and the origin of most exciting headlines. Enormous numbers of substances kill cancer cells in a dish, including bleach. It tells you a mechanism might exist. It tells you almost nothing about whether it works in a person.
  • In vivo (animal) — usually mice. A meaningful step up, but the graveyard of oncology is full of compounds that cured mice and did nothing in humans. Estimates commonly cited in the literature suggest the large majority of drugs that succeed in animals still fail in human trials.
  • Case report / case series — one patient, or a handful. Genuinely useful for generating hypotheses. Incapable of establishing that a treatment caused an outcome, because there is no comparison group and no control for the possibility that something else did.
  • Observational study — watching what happens to people who did or did not take something. Better, but vulnerable to confounding: the people who choose a treatment often differ systematically from those who do not.
  • Randomised controlled trial (RCT) — participants assigned by chance to treatment or control. Randomisation is the mechanism that lets you say caused rather than associated. This is the standard.
  • Systematic review / meta-analysis — a structured synthesis of all the RCTs. The top of the hierarchy, if the underlying trials are sound.

When you encounter a claim, locate it on this ladder first. A great deal of confident internet argument consists of citing an in-vitro paper as though it were an RCT.

2. What phase is the trial?

Human trials proceed in phases, and the phase tells you what question was being asked:

  • Phase I — is it safe, and at what dose? Small, often no control group. Not designed to show whether it works.
  • Phase II — does it appear to do anything? Larger, sometimes controlled. Promising Phase II results frequently evaporate in Phase III.
  • Phase III — does it work better than the existing standard, in a large randomised population? This is where efficacy is established or destroyed.

A common error is reading a Phase I safety result as evidence of effectiveness. It is not. It was never asking that question.

3. What was the endpoint?

An endpoint is the thing the researchers measured, and swapping one endpoint for another can transform a disappointing result into an exciting-sounding one.

Overall survival — did people live longer? — is the endpoint that matters most and is hardest to fake. Watch carefully for surrogate endpoints: tumour shrinkage, progression-free survival, or a change in a blood marker. These are quicker and cheaper to measure, and they sometimes track survival. But sometimes they do not. Tumours can shrink without patients living any longer.

Ask: did the study measure the outcome I actually care about, or a stand-in for it?

4. Relative or absolute?

This is the most common way statistics mislead honest readers.

Suppose an outcome occurs in 2 people out of 100 in the control group and 1 person out of 100 in the treatment group. You can describe that result two ways, both true:

"A 50% reduction in risk." (relative)
"A 1 percentage point reduction in risk." (absolute)

The first is what appears in press releases. The second is what tells you whether the effect matters to you. When you see a dramatic percentage, look for the raw numbers — how many people, out of how many, in each group. If the paper does not make those findable, treat that as information about the paper.

5. How many people, and compared to what?

Small studies produce erratic results simply through chance. A trial of 20 people can show a large apparent effect that vanishes when the study is repeated at scale. And a study with no control group cannot tell you what would have happened anyway — which is fatal in oncology, where spontaneous remissions, concurrent standard treatment, and natural variation in disease course all exist.

If a case report describes someone taking a compound alongside chemotherapy and improving, the honest reading is that we cannot separate the contributions. That is not a criticism of the patient or the doctor. It is simply what a case report can and cannot establish.

6. Who paid, and what happened to the negative results?

Funding source and conflicts of interest are disclosed in the paper — usually at the end. This cuts in every direction: industry funding is a real source of bias, and so is a researcher or advocate with a reputational stake in a compound being vindicated.

The subtler problem is publication bias: positive results get published, null results often do not. This means that browsing the published literature systematically overstates how well things work. Trial registries such as ClinicalTrials.gov exist partly to counter this — a trial registered but never reporting results is itself a signal worth noticing.

7. Where was it published?

Peer review is an imperfect filter, not a guarantee. But there is a real difference between an established journal with a rigorous review process and a predatory journal that will publish anything for a fee. If you have never heard of the journal, check whether it is indexed in PubMed, and look at whether the publisher is listed among known predatory operations.

Putting it together

Run the list. What kind of study? What phase? What endpoint? Relative or absolute? How many people, compared to what? Who paid? Where published?

Apply this honestly and something uncomfortable happens: the evidence stops arranging itself neatly on either side of the argument. The literature on repurposed antiparasitics contains genuine mechanistic findings, genuine laboratory signals, and genuine case reports — and it contains no completed Phase III trial establishing human efficacy in cancer. Both halves of that sentence are true, and a reader who can hold both is better equipped than a reader who has picked a team.

That is the whole point of learning to read the literature. Not to arrive at a predetermined conclusion faster, but to be able to see what is actually there.

Disclaimer: This article is for educational purposes only and is not medical advice. It does not diagnose, prescribe, or recommend any treatment. No human efficacy has been established for these compounds in cancer treatment, and none is approved by any regulator for that purpose. Consult a qualified oncologist about your care.