Cancer drugs have a terrible history in clinical trials. The most definitive figure, from development candidates of the 1980s and up to the mid-1990s or so, was a cold, hard, 95% failure rate. That beat even the central nervous system (CNS) drug category, which is a spacious haunted mansion all its own. One reason for this is that all kinds of things get thrown at oncology targets, because there's so much unmet need in the category. Whenever someone comes up with a new technology - monoclonal antibodies, antisense DNA (or RNA interference), disease-altering vaccines, etc. - you can bet that someone's going to try it out on a cancer target. Not all this stuff is going to work, needless to say.
But I wonder if that figure still holds. Starting later on in the 1990s, and gathering speed ever since, a lot of the small-molecule drug candidates in the cancer area have been kinase inhibitors. Now, back when I took my first pharma job, those compounds weren't in very good favor, partly because the key structural motifs that everyone uses today hadn't been worked out yet. If you mentioned kinase inhibitors in the labs, likely as not someone would spit in the sink and say something rude about staurosporine.
That was one of the early potent kinase inhibitors, a fairly nasty natural product. (Note: outdated web page in that link, which fits the subject). All sorts of people worked on staurosporine-like compounds during the 1980s and beyond, and most all those projects came to grief of one sort or another. It gave the whole field an unhealthy look.
There were also good reasons to think that no really selective kinase inhibitors could be discovered (since the enzymes have many structural similarities), and that the resulting broad-spectrum compounds would have just too many side effects to be useful. But molecular biology was uncovering a role for many kinase enzymes in cancer and other disease states, so people kept taking a crack at the area, and finally some far less ugly compound classes were discovered that broke the field open. Once decent compounds were in hand, it was found that they weren't as toxic as everyone had feared. Selectivity was still an issue, but you could sort of tune the structures to inhibit various groups of kinases over others.
I would not want to hazard a guess as to how many kinase inhibitors have gone into development over the past ten or twelve years. It's a pile, for sure - just look at KinasePro and Xcovery to get the idea. I will guess, though, that they haven't failed at quite that horrific 95% rate, and that a 1995-2010 survey of the field will show an improvement. Mind you, the record-holder in the earlier survey was, cardiovascular area, where only about 85% of the compounds collapsed, so don't think I'm talking about a huge increase. But when only one out of twenty of your drugs makes it, getting up to two in twenty means that you have twice as many drugs.