Bruce Chabner, M.D.
Former Director, DCTD

There's a need to change. The first change was to use human xenographs, that is, human tumors growing in animals. That had some problems. Again, you're using just one tumor to predict the response of a whole variety of human breast cancers. There is no one human breast cancer. There are many different kinds.

And secondly, you were using a ... you were studying it in a mouse, which doesn't handle drugs in the same way as man does. So that presented problems as well. The third problem with animal models was that they were not high-throughput. You could do essentially one experiment per animal or really in a group of animals.

It's much more efficient to try to use a cell-based screen, which we developed in the '80s, or a molecular target, which many of the companies are now relying on for identification of hits. And this only became possible in the molecular targeting once you understood what the targets were. And that information really didn't come out with any kind of profound understanding until the early '80s and then in the '90s it really grew.

So now we have a real transformation. We still depend on animal tumor systems in vivo to tell us that a drug will work in an in vivo model. But it tells us very limited information. It tells us it will work in a mouse. It will work against that one tumor that you're testing. It doesn't guarantee that it will work in people for a lot of different reasons.

There's been an effort to develop more sophisticated animal models that really reflect some of the molecular changes you find in human tumors to engineer these models. So they express high concentration receptors, certain pathways, certain drug resistance markers.

Again, it's one tumor. And the problem we have in man is that there are many different kinds of lung cancers. There are many different kinds of breast cancer. And the model is only as predictive as what you put into it, what molecular changes you make to put in there. And it unfortunately can't predict success in a broad array of tumors. But we're learning, I think, how to do a better job of it. And we're learning the limitations of our modeling.