Surprisingly, nothing that companies have done in the past 60 years has affected their rates of new-drug production: whether large or small, focused on small molecules or biologics, operating in the twenty-first century or in the 1950s, companies have produced NMEs at steady rates, usually well below one per year. This characteristic raises questions about the sustainability of the industry’s R&D model, as costs per NME have soared into billions of dollars.
What he’s found, actually, is the NME generation at drug companies seems to follow a Poisson distribution, which makes sense. This behavior is found for systems (like nuclear decay in a radioactive sample) where there are a large number of possible events, but where individual ones are rare (and not dependent on the others). A Poisson process also implies that there’s some sort of underlying average rate, and that the process is stochastic – that is, not deterministic, but rather with a lot of underlying randomness. And that fits drug development pretty damned well, in my experience.
But that’s just the sort of thing, as I’ve pointed out, that the business-trained side of the industry doesn’t necessarily want to hear about. Modern management techniques are supposed to quantify and tame all that risky stuff, and give you a clear, rational path forward. Yeah, boy. The underlying business model of the drug industry, though, as with any fundamentally research-based industry, is much more like writing screenplays on spec or prospecting for gold. You can increase your chances of success, mostly by avoiding things that have been shown to actively decrease them, and you have to continually keep an eye out for new information that might help you out. But you most definitely need all the help you can get.
As that Pfizer chart helps make clear, Munos is particularly not a fan of the merge-your-way-to-success idea:
Another surprising finding is that companies that do essentially the same thing can have rates of NME output that differ widely. This suggests there are substantial differences in the ability of different companies to foster innovation. In this respect, the fact that the companies that have relied heavily on M&A tend to lag behind those that have not suggests that M&A are not an effective way to promote an innovation culture or remedy a deficit of innovation.
In fact, since the industry as a whole isn’t producing noticeably more in the way of new drugs, he suggests that one possibility is that nothing we’ve done over the last 50 years has helped much. There’s another explanation, though, that I’d like to throw out, and whether you think it’s a more cheerful one is up to you: perhaps the rate of drug discovery would actually have declined otherwise, and we’ve managed to keep it steady? I can argue this one semi-plausibly both ways: you could say, very believably, that the progress in finding and understanding disease targets and mechanisms has been an underlying driver that should have kept drug discovery moving along. On the other hand, our understanding of toxicology and our increased emphasis on drug safety have kept a lot of things from coming to the market that certainly would have been approved thirty years ago. Is it just that these two tendencies have fought each other to a draw, leaving us with the straight lines Munos is seeing?
Another important point the paper brings up is that the output of new drugs correlates with the number of companies, better than with pretty much anything else. This fits my own opinions well (therefore I think highly of it): I’ve long held that the pharmaceutical business benefits from as many different approaches to problems as can be brought to bear. Since we most certainly haven’t optimized our research and development processes, there are a lot of different ways to do things, and a lot of different ideas that might work. Twenty different competing companies are much more likely to explore this space than one company that’s twenty times the size. Much of my loathing for the bigger-bigger-bigger business model comes from this conviction.
In fact, the Munos paper notes that the share of NMEs from smaller companies has been growing, partly because the ratio of big companies to smaller ones has changed (what with all the mergers on the big end and all the startups on the small end). He advances several other possible reasons for this:
It is too early to tell whether the trends of the past 10 years are artefacts or evidence of a more fundamental transformation of the drug innovation dynamics that have prevailed since 1950. Hypotheses to explain these trends, which could be tested in the future, include: first, that the NME output of small companies has increased as they have become more enmeshed in innovation networks; second, that large companies are making more detailed investigations into fundamental science, which stretch research and regulatory timelines; and third, that the heightened safety concerns of regulators affect large and small companies differently, perhaps because a substantial number of small firms are developing orphan drugs and/or drugs that are likely to gain priority review from the FDA owing to unmet medical needs.
He makes the point that each individual small company has a lower chance of delivering a drug, but as a group, they do a better job for the money than the equivalent large ones. In other words, economies of scale really don’t seem to apply to the R&D part of the industry very well, despite what you might hear from people engaged in buying out other research organizations.
In other posts, I’ll look at his detailed analysis of what mergers do, his take on the (escalating) costs of research, and other topics. This paper manages to hit a great number of topics that I cover here; I highly recommend it.