Hacker News Reader

Comment by asdff | Hacker News Reader

Lay people act like this is a trump card but this is just how science is done. You can't really perform this in humans without having some data that it may very well work. Mice have been pretty good models for cancer biology. Nature Biotechnology is also a good and selective journal.

View on HN

I agree, I'm just mentioning it re: clickbait titles (not calling anyone out) vs. "oh, [xyz] hasn't been cured, there's just some promising results in mice"

Well, the actual title is "Sensitization of tumours to immunotherapy by boosting early type-I interferon responses enables epitope spreading." This is just the reduced version for mass consumption. Keep in mind many people got a C in high school biology.

Out of curiosity, are there examples of drug development that did poorly in animal models, but then did much better in human trials?

I'm guessing this is rarely tested since animal modeling is usually a gating factor for human testing?

Checkpoint inhibitors (which are the primary driver of improved cancer treatment over the last 15 years and generate > $50B/year in sales) generally don't look very good preclinically. Even their clinical data can be hard to interpret prior to a large scale trial, which led to them almost being shelved.

The catch here is that only two targets (PD(L)-1 and CTLA-4) turned out to work well in humans. All of the other immunotherapies that looked mediocre preclinically turned out to also be mediocre or entirely ineffective in humans.

If you ask: out of all the studies which have done well in mice what percent lead to successful drugs for humans: the answer is a very small percent.

This is the nature of research pipelines. If you ask: out of all the clinical trials that have shown promise in early human trials, what percent lead to successful drugs for humans: the answer is a very small percent.

Reply on news.ycombinator.com