How this register works
A fair thing to want before believing a website about cancer. This page is the whole of the method, including the parts where it has been wrong.
Where the grades come from
Every grade here is the one The Cancer Evidence Handbook gives. None of them is worked out again on this site, and none is adjusted. 79 graded claims across 69 interventions, checked against the book's own list on every build; if a grade here ever disagreed with the book, the build would stop rather than publish it.
The letters describe how strong the evidence is. They do not describe how much anyone should want it to be true, and they are not a safety rating.
- Grade A
- Randomised human evidence of benefit.
- Grade B
- Good human evidence, with a limit worth knowing.
- Grade C
- Human evidence exists and does not settle it.
- Grade D
- No human evidence of benefit, or none that survived testing.
- Grade E
- Human evidence points the wrong way.
- Grade X
- Documented harm, with no benefit to weigh against it.
What every entry has to carry
These are checked mechanically, and a page that fails one is not published.
- A source you can look up. Every graded claim cites something with a permanent identifier — a DOI, a PubMed number or a trial registry number. Where a source exists but cannot be looked up, the entry says so in place of pretending otherwise.
- Nothing invented to fill a gap. A citation is recorded exactly as the source displayed it. No year, page or identifier is ever supplied from memory or from a pattern, and where a paper was reached through something that cited it rather than read directly, the entry says that too.
- Absence stated as absence. Where nothing has been measured, the entry says nobody has measured it, rather than leaving a silence that reads like reassurance.
- A question, not an instruction. Each claim ends in something you could ask the people treating you. That is the whole output of this site.
What this site will not do
- It carries no doses, amounts, frequencies or schedules, for anything, at any grade. Not one, anywhere, and the build refuses to publish a page containing one.
- It gives no verdicts. There is no traffic light here and no score, because a letter describing evidence and a letter telling you what to do are different things and only one of them is honest.
- It cannot see you. Nothing is personalised, nothing is stored, and there is no account. Questions you add to your list live in the tab and are gone when you close it.
- It does not check your medicines. What it records about interactions is at the level of a class of treatment, and its output is a request for a pharmacist, who has your actual list.
- It runs no artificial intelligence. Nothing on this site generates an answer when you ask it something.
How it has been wrong
What changed carries every grade that has moved, dated, with the evidence that moved it — and every reason that has been withdrawn while the grade stayed where it was. The second kind is the harder one to notice and the more useful one to read: it is the register saying that something it told you was not true, on a page it did not have to write.
Grades that were looked at again and left alone are logged too. An entry reconsidered and held is the method working, and a log that only recorded movements would hide that.
When each entry was last read
Every entry carries the date somebody last went through it, at the top of its own page. The oldest here is 2026-09-16 and the most recent is 2026-09-22. An entry that has not been read for more than seven months stops the build, so nothing on this site can quietly go stale without somebody being told.
Who wrote it
Suryaprakash Tripathy, who wrote the book this accompanies. The register is the part of that work that has to keep moving: the book is fixed once printed, and the evidence is not.