I've had a number of comments on my post below about the changing cancer market, and they deserve some follow-up. One very good point that was made is that this change is really a race between several factors: the narrowing of individual markets as we find out the reasons for efficacy could be balanced out by the ease that this could bring to the design of clinical trials. After all, it should be easy to get good data if you have a clearer idea of which patients are most likely to respond. (Every major drug company is plowing money into biomarker research for just this reason.)
But if you design your trial on the basis of a specific biomarker, I assume that you'll get a specific approval based on that, too. You're trading larger market potential for a higher clinical and regulatory success rate - and believe me, I'm far from suggesting that that's a bad trade. But it'll be an interesting decision in some cases. In the short term, I would expect to see some companies forgoing the use of narrow biomarker-driven trials in an attempt to go for the bigger money. But as the everyone-prescribes-for-everything model for the oncology market breaks down, that'll become less and less of an option.
How many years is that breakdown going to take? It's hard to say. Some areas are going to be way ahead of the rest of the field - for example, it might be happening right in front of us with Iressa, but in other markets we don't have a clue. In general, I'd say that twenty years will see the transformation well underway, but I could be off in either direction. Along the way, there are going to be some nasty surprises for the companies involved as they find that their existing or in-development drugs have suddenly contracted to much smaller roles.
Another thing to think about is how many new drugs we're going to need, versus combinations of existing ones. You'd think that the latter option would cover a lot of ground, but I think that the emergence of resistant tumor lines will keep everyone searching for new compounds. Oncology could end up looking a bit like anti-infectives, although the big difference is the individual occurrence of resistant tumors, rather than the far more challenging spread of resistant organisms.
So you'll end up with courses of treatment that look like this: sequence the patient and a sample of their cancerous cells, which will suggest the first-line therapy. Then if that doesn't knock everything down, you move to the next ones on the list (which will look slightly different under each of those first options), and switch to that, keeping an eye out over the next few years for the need to switch to a third agent or combination, and so on.
How will drug companies deal with these changes? A colleague of mine suggested that the larger outfits will probably fare better, because they'll be able to hold a portfolio of cancer drugs that will add up to something. The risk of changing market sizes will be spread out across several agents. Smaller companies may have to explicitly target smaller markets, which you'd think would lead to fewer stock-market skyrockets like Imclone and OSI. These would also seem to be good candidates for in-licensing from the bigger companies looking to fill gaps in their holdings.