The claim that "probability of meaningful transfer between animals to humans can be astoundingly low, far below 50%" requires more scrutiny. Let's say drug b is being tested for treating condition A. Assume that it's found effective in rats. It seems the claim then is that probability of it also being effective in humans is less than 50% and sometimes 0%. But this in itself isn't a good enough argument against animal testing. If one can show that there were drugs c, d, e.. etc that were found ineffective in rats and as a result did not have to be considered for human trials. If there is a good enough probability that not effective in rats A implies not effective in humans, it will still lead to huge benefits to science and clinical trials.