Some time ago, I took nominations for Least Useful Animal Models. There were a number of good candidates, many of them from the CNS field. A recent report makes me think that these are even stronger contenders than I thought.
The antidepressant reboxetine (not approved in the US, but sold in a number of other countries by Pfizer) was recently characterized by a German meta-analysis of the clinical data as "ineffective and potentially harmful". Its benefits versus placebo (and SSRI drugs) have been overestimated, and its potential for harm underestimated. It was approved in Europe in 1997, and provisionally by the FDA in 1999, although that was later rolled back when more studies came in that showed lack of efficacy.
Much has been made of the fact that Pfizer had not published many of the studies they conducted on the drug. These do seem, however, to have been available to regulatory authorities, and were the basis for the FDA's decision not to grant full approval. As that BMJ link discusses, though, there's often not a clear pathway, especially in the EU, for a regulatory agency to go back and re-examine a previous decision based on efficacy (as opposed to safety).
So the European regulatory agencies can be faulted for not revisiting their decision on this drug in a better (and quicker) fashion, and Pfizer can certainly be faulted for letting things stand (in the face of evidence that the drug was not effective). All this is worrisome, but these are problems that are being dealt with. Since 2007, for example, trials for the FDA have been required to be posted at clinicaltrials.gov, although the nontranparency of older data can make it hard to compare newer and older treatments in the same area.
What's not being dealt with as well is an underlying scientific problem. As this piece over at Scientific American makes plain, reboxetine, although clinically ineffective, works just fine in all the animal models:
And this is a rough moment for scientists studying depression. Why? Because reboxetine works beautifully in our animal models. It’s practically a poster-child antidepressant. It produces acute effects in tests such as forced-swim tests and tail-suspension tests (which use changes in struggle as a measure of antidepressant efficacy). It produces neurogenesis in the hippocampus, which is thought to be correlated with antidepressant effects. When behavioral pharmacologists are doing comparisons between older antidepressants and newer ones, reboxetine is often used as a positive control, a drug known to have an effect in the behavioral test of choice.
But it doesn’t work in patients. And patients are what matters. Now, scientists are stuck with a difficult question: What went wrong?
A very good question, and one without any very good answers. And this certainly isn't the first CNS drug to show animal model efficacy but do little good in people. So, how much is the state of the art advancing? Are we getting anywhere, or just doing the same old thing?