I've mentioned before that one of our big problems in the drug industry seems to be finding compounds that work in man. I know, that sounds pretty obvious, but the statement improves when you consider the reasons why compounds fail. Recent studies have suggested that these days, fewer compounds are failing through some of the traditional pathways, like unexpectedly poor blood levels or severe toxicity.
In the current era, we seem to be getting more compounds that make it to man with reasonable pharmacokinetics (absorption from the gut, distribution and blood levels, etc.) and reasonably clean toxicity profiles. Not all of them, by any means - there are still surprises - but the stuff that makes it into the clinic these days is of a higher standard than it was twenty years ago. But that leaves the biggest single reason for clinical failure now as lack of efficacy against the disease.
That failure is the sum of several others. We're attacking some diseases that are harder to understand (Alzheimer's, for example), and we're doing so with some kind of mechanistic reason behind most of the compounds. Which is fine, as long as your understanding of the disease is good enough to be pretty sure that the mechanism is as important as you think it is. But the floor is deep with the sweepings of mechanistically compelling ideas that didn't work out at all in the clinic - dopamine D3 ligands for schizophrenia, leptin (and galanin, and neutropeptide Y) for obesity, renin inhibitors for hypertension. I'm tempted to add "highly targeted angiogenesis inhibitors for cancer" to the list. The old-fashioned way of finding a compound that works, and no matter how, probably led to fewer efficacy breakdowns (for all that method's other problems.)
Another basic problem is that our methods of evaluating efficacy, short of just giving the compound to a sick person and watching them for a while, aren't very reliable. If I had to pick the therapeutic area that's most in need of a revamp, I'd have to say cancer. The animal models there are numerous, rich in data, and will tell you things that you want to hear. It's just that they don't seem to do a very good job telling you about what's going to work in man. I will point out that Iressa, for one, works just fine in many of the putatively relevant models.
The journal Nature Reviews: Drug Discovery (which is probably the best single journal to read for someone trying to understand pharma research) published a provocative article a couple of years ago on this subject. The author (the now late) David Horrobin, compared some parts of modern drug discovery to Hesse's Glass Bead Game: complex, interesting, internally consistent and of no relevance to the world outside. They got a lot of mail. Now the journal has promised a series of papers over the next few months on animal models and their relevance to human disease, and I'm looking forward to them. We need to hit the reset button on some of our favorites.
1. Kevin on January 18, 2005 10:31 PM writes...
Doesn't Novartis have a renin inhibitor, SPP100 (Aliskiren) in phase III? And I read somewhere that Merck is also hot on the renin trail...
Permalink to Comment2. PacRim Jim on January 19, 2005 10:19 AM writes...
We humans are nowhere near being able to accurately model human physiology. By the time we are able to do so, we'll be dead, anyway. ;>)
Permalink to Comment3. Derek Lowe on January 19, 2005 9:05 PM writes...
Hmm, I should have checked my databases before slamming renin. I just know that most everyone has had a crack at it (for at least twenty years now) without much to show for it. Maybe I should have made fun of elastase, instead.
Permalink to Comment4. jim harris on January 20, 2005 7:52 AM writes...
Renin inhibitors of the past (and perhaps the future) died because variability in pharmacokinetics led to lack of efficacy (there were pharmacoeconomic issues as well because some patients required grams of API to show an effect, but they did show an effect). I respectfully disagree that PK failures have decreased ... they often get categorized as efficacy failures. In my world, high PK variability, leading to low efficacy relative to a reference therapy, is a PK failure. We do not have tools necessary to choose respectable half-life and low variability compounds for development. [Many vendors are available to sell you some snake oil, however.]
Permalink to Comment5. Daen de Leon on January 20, 2005 4:54 PM writes...
I'd venture that it's not just the animal models that need some rethinking. A couple of years ago I wrote a literature review on mechanisms of resistance to cisplatin for my master's. I found that out of 30 randomly sampled papers no less than 26 different cell lines had been used to study the development of cisplatin resistance in vitro, hardly any pair of which show the same two patterns of resistance. An extreme example, the A2780 cell line, displays six or seven of the eight resistance modes that I identified. The researchers admit that this unstoppable cell line "may not necessarily reflect the underlying clinical drug resistance phenotype". Indeed.
Permalink to Comment6. Kevin Foley on January 22, 2005 7:49 AM writes...
The Nature Reviews series should be interesting. But I was amused by the first article they selected, which describes the uses of zebrafish as a model system. Now I have nothing against these little striped fishies (I'm a geneticist at heart, myself), but it does seem to me that if one really wants to improve the predictability of animal models, moving even further away from humans is not the best solution. Hopefully the next article is not on C. elegans (worms).
Everyone knows there's a problem, but the biggest failing in this field is a lack of research. The average academic lab is too busy filling in publishable little black boxes to focus on the big picture of drug development. And when their new transgenic or knockout mouse suddenly develops symptoms of a disease, they are quick to trumpet a "new human disease model", but with out the careful comparative analyses that are really needed to support such a statement. More importantly, the proper validation of a new model using a variety of successful and unsuccessful drugs, measuring their affects on relevant physiological parameters and biomarkers, and the comparison of these results with those from previously accepted models, is completely lacking. Likewise, industry is too busy and short sighted to expend much effort in this area, choosing instead to go with whatever is the current “gold standard “model for a particular disease that the FDA is used to seeing. We really need a more systematic (NIH sponsored?) effort. Some mention of this problem was made in last year’s “NIH Roadmap.”
Incidentally, for those who are interested in the subject, IBC is having a conference on “Early Efficacy Assessment” this June in Boston. Looks an interesting selection of talks (at least for the session I’m chairing!).
Permalink to Comment