Skip to main content


Can’t Shake Failure Yet

From the new Endpoints site, here’s a roundup of the worst moments in the clinic for drug development in the first half of 2016. Several of them were covered at the time here as well, but there the whole list is worth a look, because it illustrates the many ways that big trouble can hit.

The basic story is The Drug Didn’t Work, as you’d figure, but there are plenty of variations available. Unexpected tox, higher-than-expected placebo response, and “technically worked but not well enough for anyone to ever buy any of it” are three classics, and no matter what the problem, it can always be exacerbated by overpromising beforehand. That’s a vice common to small companies, but it can happen to the bigger ones, too.

For example, one of the recent events on the list that I haven’t mentioned involves AbbVie’s purchase of Stemcentrx. I first wrote about them here – they’re the Peter Thiel-backed company that’s been going hard after cancer stem cells. Data for their lead program (the antibody-drug conjugate Rova-T) against non-small-cell lung cancer were released recently at the ASCO meeting, but the results were not very impressive (one month survival advantage at best). This was a small trial, and there were indications that Rova-T might be able to perform better, but considering that AbbVie paid nearly $6 billion upfront (with $4 billion in potential milestones), people were expecting a lot more than this.

There will be room for another such list by the end of the year, I’m sure. But it’s worth keeping in mind that clinical failures are (given the state of our knowledge) built into the system we have. We do a very ragged job of picking winners in this business, so if there aren’t a fair number of clinical wipeouts, it would probably mean that we’re not putting enough high-risk high-reward things into the clinic in the first place. (For that matter, even some of the putatively low-risk programs hit the banana peel). We’re a long way from being able to avoid nasty surprises in Phase II and Phase III, a fact that investors should always keep in mind.

17 comments on “Can’t Shake Failure Yet”

    1. Joe says:

      This isn’t slashdot, man….

      1. (Patent) Troll In the Dungeon says:

        I think he/she is just practicing following the America Invents Act: doesn’t matter if you weren’t the first to come up with that post, so long as you file first it’s A-OK!

    2. Anom says:

      Fast follower!!!! 😉

  1. CMCguy says:

    Carroll’s list stopped at just 10 meaning there likely are other failures to include however as FDA indicates 13 approvals so far in 2016 on balance if wish to play nonsensical numbers game without considering details or time shift factors for each category the past 6 months might be deemed a successful period. Because in some ways a number of drugs involved are based on less mature knowledge such as antibody technology there could be a string of early failures or less than dramatic results as part of the learning that seems to be common as ideas get translated to applications

  2. Biologics CMC says:

    Still no love for Forum… I drive by their brand new site and the name just came off the building. Very sad story.

  3. confused says:

    I’ve been reading the blog for a long time now and I recently found myself in the world of medchem (at the bench) but I wanted to ask some questions about these failures. I remember years ago reading an article talking about attrition rates and giving 50% for phase II and 70% for phase III give or take. What sticks in my mind though was the large percentage (over 60%) that failed due to not showing efficacy. For someone who isn’t involved in the later stages of drug development this seems unacceptably high for such a late stage in the development cycle.

    So my question is obviously why this is never picked up earlier for so many drugs? Are animal models so unhelpful for determining efficacy? Or is it that for many diseases you can’t simply can’t use animal models (thinking depression for example)?

    I assume for cancer it’s always going to be difficult because the disease morphology varies largely in a patient set (presumably animal models lack a large degree of variety here?). And I can think of other examples, such as for Alzheimer’s and amyloid plaques, where models may be looking for an unproven theoretical endpoint (that doesn’t imply efficacy). But I would generally of expected animals to provide better evidence of potential success than we observe.

    How often can someone actually be confident in an animal model regarding efficacy?

    Apologies for the long post. Any thoughts greatly welcomed.

    1. Dr. CNS says:

      In a way, you’ve answered your question.
      If the models used had been useful, the drugs would have shown efficacy in late clinical trials…

      If the target engagement had been demonstrated in the clinic, you might be able to disprove the biological hypothesis that lead to the project…

      IMHO, as long as we don’t understand “disease” at the molecular level, not much will work in the clinic, especially in the CNS space… And I mean more than a link to the wrong gene…

      Time to stop looking for the keys under the lamp-post? What is it going to take?

      1. Anon says:

        Success rates used to be so much higher back in the good old days, when there were no other drugs available and we were allowed to test compounds on foreigners, soldiers and prisoners. Sigh.

      1. Allison says:

        I’d like to add.

        Think about all the drug candidates that may have been fine for humans but killed animals and so they were discontinued. All you need to do is look at certain foods we can’t give dogs (for instance: onions,  grapes, chocolate,  xylitol) to know that’s a real possibility.

        I really like the microfluidics “organ on a chip.” I’m especially hoping for when they can start combining the chips to mimic a very basic human system.

        1. sep says:

          Hm, maybe, but I’m sceptical. If we can’t understand disease at the gene/pathway level, why do you think we can reduce something as complex as toxicity? I wouldn’t want to be the volunteer.

    2. Walther White says:

      For Phase III, you have to show more than efficacy–you need to show a substantial improvement over the standard of care (i.e., the existing treatment). And sometimes you also have to factor in additional FDA guidelines over what that substantial improvement should be, as well as additional safety requirements.

  4. Andy says:

    Maybe the next question (for us non med chemists) then, is how do projects ever start?
    How can the early stages of a project be defined with any certainty?
    Or is it implicit in all early stage projects that a certain assay (or whatever) is hopefully a suitable proxy for the disease, but everyone knows it’s not guaranteed?
    So how the hell do projects not get politicised? If they can’t be based on anything other than tenuous reasoning?
    How do you guys sleep at night? Does it not worry you to death that a whole project could be a total waste of time?

    1. a grad student says:

      As a non-med chemist, I can’t say whether there’s a particular assay with which to start, because no assay is 100% accurate. But as an academic, I know that if I lose sleep worrying about whether my project will work (let alone be useful to future iterations of humanity), nothing will get done. If I worry too much about whether I’ve started at the right place, nothing will get started.

      I think it’s the same for med chem. You need a sound idea based on existing knowledge to start, sure, but the point of the research is to expand that knowledge so that we can explore more ideas and ways to synthesize that knowledge into usable results. You will run into ideas that seem like they should work and don’t.

      As I’ve come to learn reading Derek’s posts, late clinical failures are extremely costly in terms of time, brainpower, and money. But that is not a reason for pharma to stop looking for drugs, because then they’ll stop finding them. I doubt that’s a better direction for everyone.

      1. Dr CNS says:

        @ a grad student,

        Yes, you are right… somewhat… There are options between 1) running a project based on the knowledge we have when we start and 2) not doing anything at all. It is not a binary choice.

        Indeed, for the new targets we follow these days it is expected we know little when we start. But it is also expected we learn (hopefully a lot!) as we move forward. This new knowledge should be the basis for making the key decision on whether to move or not to clinical testing.

        In my view, part of the problem is the “mechanical” nature in which projects are managed. When you stop learning, or face a problem you cannot figure out how to solve, it’s time to stop.

  5. JAM says:


    [This comes from 10 years experience in drug discovery at both a large pharma and a large biotech]

    A snarky answer to your question of how do projects ever get started: that what med chemists (and everyone else at biotech/pharmas) gets paid for!

    Actually, before the chemists get heavily involved, teams consisting mainly of biologists perform “target ID” and “target validation”. Those processes are imperfect due to 1) the boundaries of current knowledge of biology; 2) the complexity of that knowledge and the tools via which it is assessed (e.g. literature search); and the imperfect nature of cellular and animal assays.

    Although we call it “target validation”, a target isn’t ultimately validated until a successful Phase III trial. But using genetics, genomics and pre-clinical assays one can build pretty convincing-sounding stories – convincing enough that our superiors assign millions of dollars in resources to pursue the stories further.

    Derek has written many times about the misaligned incentives in drug discovery and development. Teams and individuals are incentivized to *advance* drug programs – it’s pretty rare to get a promotion/bonus for *killing* a program. That only compounds the inherent optimism we all have that stems from our desire that our work be “productive”. Pretty much everyone working on a drug program is convinced it is going to work – until it doesn’t.

    The optimism is certainly not evidence-based: last year Derek did a back of the envelope calculation of the probability that an individual med chemist would synthesize a successful drug candidate AT ANY POINT IN HER CAREER (see URL for this comment). The answer: miniscule.

Comments are closed.