How it works

Each page follows one treatment for one condition and shows the public records found for it. This page explains how the status on a page is set and what it does not tell you.

Reported

“A cure in five years”

A news report, 2019

Promised by 2027

Recorded

  • 2020Trial registered
  • 2022Phase 1 trial startedbasis of status
  • 2023Paper published

Where it stands

Phase 1 trial

An example with made-up records, not a page on this site.

How a status is set

A status is derived by fixed rules from the approved evidence events listed on the page. Where no trial is found, the rules also use the kind of study the first reports described. No model and no editor decides a status. Every status lists the events that produced it.

The rules are applied in this order, and the first one that fits sets the status.

  1. If the paper the first news reports were about has a retraction notice, the status is “Origin paper retracted”.
  2. If a regulator has recorded an approval, the status is “Approved”.
  3. Otherwise, trials that have not been terminated, withdrawn or marked as missing their primary endpoint are the ones that count. If any remain, the status follows the furthest one that has started: “Phase 3 trial”, “Phase 2 trial” or “Phase 1 trial”. If none has started, or none has a phase, the status is “Trial registered”.
  4. If no trial remains and a person has marked a trial as missing its primary endpoint, with a source link, the status is “Trial missed its primary endpoint”.
  5. If no trial remains and every registered trial was terminated or withdrawn, the status is “Trials ended early”.
  6. With no trial found at all, the status depends on what the first reports described. If it was confirmed laboratory or animal work, the status is “Preclinical only”. A study in people, or an unconfirmed evidence stage, gives “No trial record found”.

A treatment that is still at a stage open to further evidence (“Preclinical only” through “Phase 3 trial”, and “No trial record found”) can also carry the no-signal note, described below.

What each status means

Preclinical only
The confirmed first-report evidence is laboratory or animal work. No clinical trial was found in the registries checked.
No trial record found
No clinical trial was found in the registries checked. This does not establish whether the treatment has been studied in people.
Trial registered
A clinical trial is registered. None that has started has a phase.
Phase 1 trial
The furthest trial that has started is a Phase 1 trial.
Phase 2 trial
The furthest trial that has started is a Phase 2 trial.
Phase 3 trial
The furthest trial that has started is a Phase 3 trial.
Approved
A regulator has recorded an approval.
Trial missed its primary endpoint
A trial result was reported in which the primary endpoint was not met, and no other trial stands.
Trials ended early
Every registered trial was terminated or withdrawn.
Origin paper retracted
The paper the first news reports were about has been retracted.

Each status links to its entry in the guide, which says what it rests on and what it does not tell you.

No signal

A treatment that has stayed at an open stage for several years, with no new approved record, shows a note such as “No signal for 5 years”. The note means that the most recent record is that old. The count starts from the date of that record, or from the date the treatment was first reported when no record has been found.

It says nothing about work that has not been published or registered, and it is not a status. A new record removes it.

What a trial entry shows

A page lists each of its trials once. Each entry gives what the trial's registry record states: its state, its phase, how many people it enrolled or plans to enrol, when it started, when its main measure is due or was reached, the arms it compares, and whether results have been posted. The arm that tests the page's treatment is shown in bold. Where a trial stopped early and the record says why, the entry repeats the record's words. Under these are the dated changes in the record, such as the day it was registered or completed. These facts are read from the record by fixed rules. No editor and no AI model reads them.

Posted results are the numbers the trial's sponsor gave the registry: each group's value, and any comparison the sponsor posted with it. The site shows the numbers and adds no word on what they mean for the treatment. They do not change a status.

Where the trial's registry record names papers, the entry links them. A paper is not a trial record, and an entry never says what a paper concluded.

An entry does not say where a trial takes place or who may take part in it.

What this site never says

This site gives no verdict on a treatment, a researcher, a company or a news outlet. It gives no medical advice and recommends no treatment. A registered clinical trial is not a treatment option.

Where the data comes from

Every record on a page links to the source that holds it. Records are shown as the registries and databases gave them. The wording of statuses is the site’s. The text of a news article is never copied or stored.

A treatment reported under a brand name is looked up inRxNorm for the name of its ingredient, which is then used to search the trial registry. That source asks for the statement that follows: “This product uses publicly available data from the U.S. National Library of Medicine (NLM), National Institutes of Health, Department of Health and Human Services; NLM is not responsible for the product and does not endorse or recommend this or any other product.”

An optional search of a new report

When you ask for a registry search, OpenAI reads the report's text to identify its treatment and condition. We search ClinicalTrials.gov and linked research through Europe PMC, then use OpenAI to compare the trial records. This is a live search of those research sources, not a general search of the web. Its answer is marked as automatic and unreviewed; it cannot create a page or change a status. We do not store submitted article text or screenshots. Structured answers for independently public reports may be reused for 24 hours; answers based only on browser content stay in that browser. Read what is sent and how the providers handle it before searching.

Related papers come from OpenAlex searches for the treatment and condition. ROR identifies organisations explicitly named in the report. These sections open when you need them. A paper result does not establish that a treatment works, and an organisation record does not verify involvement or endorsement. Neither provider receives the article text, page address or screenshot. Public paper and organisation metadata may be cached for seven days.

Opening the browser extension checks the current page automatically. A tracked article shows the page that follows it. After you enable automatic search, an untracked page starts a live registry search when the service is available. If its text is insufficient, a screenshot of visible content can be sent. Select words first to search only those words, without links or a screenshot. Each candidate's match confidence is the model's estimate that the record concerns the reported treatment and condition. It is not a verified probability or a rating of treatment effectiveness or trial quality. Pages are not searched in the background while the extension is closed.

Ways to keep up

A status can change months or years after a headline. Neither of these needs an account, and neither tells us who you are.

What is counted

Cloudflare, which hosts this site, counts requests and errors to operate the service. We do not add analytics, set tracking cookies, or keep reader profiles. Request logging for our Worker is disabled.

Report an error

An error on any page of this site is reported the same way. Say which page it is on.

Something wrong on this page? Tell us what, and the source that shows it: Report an error (opens your mail)