Choose your Own Numbers: Crowdsourcing the Cost-to-Produce a new Drug?

Posted in Pharma industry

Slate’s piece on drug costs has really caused some fun and excitement in the life science community.  For those that haven’t followed, Tufts Center for Study of Drug Development published a widely cited number in 2003 of $800M+ to bring a drug to market.  Recent estimates are above $1.2B.  Slate’s piece refers to a new study, by Donald Light and Rebecca Warburton in the London School of Economics journal Biosocieties, that claims the number should be around $50M.  They challenge almost every assumption of the Tuft’s study.

Derek Lowe’s blog has already had some great commentary on it, and I can’t argue with or add much to amplify any of his points. Either the drug industry is actually doing great with R&D productivity and we just didn’t know it, or their assumptions are off.   I suspect that latter.

We can all debate what the right statistics are for the R&D costs to support a drug: direct costs per phase, time per phase, failure rates, etc…   It all depends on the drug itself, its safety profile, how many fumbles it has in development, what indications (diseases) it goes after, what the organizational overhead costs are, how bureaucratic your processes are, etc…  Venture-backed biotechs have a very different cost structure than a Big Pharma.  Some firms build Fords and others build Cadillacs.  So the median and mean are essentially meaningless; the ranges and distributions are more interesting.

With all the good online discussion, I thought I’d have some fun and throw a model into the mix to see if crowdsourcing can generate a better answer.  The link is below the image.

Its like the old “Choose your adventure” books.   Put in the numbers you think are right and it will give you your very own value.  Feel free to share a post with your total value number.

I’ve got a ‘lean’ $25-30M from discovery to Phase 1b/2a PoC model in there – looks like a total adjusted cost of $300M or so.   Still not cheap.

Choose Your Own Drug Cost – Model

Looking forward to hearing from you on the ‘right’ answer from your experience.

ADDITION of MODEL #2 (added to original post on 3/10/2011)

To respond to some good comments and further stimulate the debate, as well as provide an example of a Big Pharma primary care chronic drug package (with some Big Pharma attrition #’s), I’ve added a more “extensive” development cost and timelines and decreased the success rates.  Its now about 1 in 25 from Candidate to Approval, in line with some industry estimates.   The model now spits out $1.6B.   Amazing what a few changes do.  With this Model and the original, we’ve probably got some bookends on reality.

Choose Your Own Drug Cost – Model2

This entry was posted in Pharma industry. Bookmark the permalink.
  • http://twitter.com/exMBB exMBB

    I actually have some older Pharmaprojects that goes back pretty far. I have to dig it up to fill in those numbers. BTW, are you Phase III cost numbers a bit low?

  • Anonymous

    Probably is low – was a placeholder for 2K patients at $25K fully loaded. Certainly too small # for chronic primary care, and probably too small $ for specialty/acute/intensive care. But a guesstimate.

  • Cellbio

    Nice Model. My experience says the success rates are too high. Failure rate in preclinical is a measure of risk tolerance, and that can vary a lot by organization, but I’ve seen the survival rate to be much less than 50% for transition into Ph1. That impacts both preclinical and discovery costs, and opportunity costs, as the teams keep churning out new clinical leads. Also think the clinical success rates, though perhaps in line with historical rates, are a bit optimistic for today. Would love to know if that is true or not. Anyway, never seen an output rate of 10 preclinical projects to market, maybe closer to 100:1? Also can’t imagine getting a PH2a and Ph2b done in 1.5 years. Might need to double the cost and float the time a bit.

    My estimate topped $1B. Maybe high, maybe not, but real life experience, which factors in the human nature of drug development organizations, including risk tolerance, herd mentality, the desire to make well informed decisions with clear facts. This last bit, decision making, plagues organizations on both sides of the issue, failing to move forward in the face of uncertainty, and failing to cut losses early. On the uncertainty side, it influences more active agent comparisons and attempts to launch into well served markets, which I believe is one reason failure rates are rising. On the other side, there is little risk taking without clarity of why something and where something will work. And if we knew that, indeed our drug development costs would be much lower.

  • Respisci

    A comment and some questions: Maybe this is just is our field of respiratory but no one seems to run only one Phase IIa study anymore. A quick check on clinicaltrials.gov will confirm that mulitple PhaseIIa studies being run in various countries before the larger Phase IIb study. The time estimate of 1.5 yrs is therefore an underestimate (even for those big pharma with resources to run studies simultaneously).

    Does your estimate for preclinical include the original short term toxicology study (along with geno and cardio) AND the chronic tox, repro tox etc that are needed for later stage? Or are those costs hidden among the costs of the clinical trial?

    Lastly, What about actual CMC? The cost of manufacturing a drug, performance of stability studies, and the investigation of either different delivery devices (again, sorry for my respiratory focus) or formulations need also to be considered. There is scale up as the drug moves from Phase I to Phase III, to change the manufacturing process from gram quantities to kg to tonnes. This takes resources, people and $$$. And as anyone who has submitted an application for approval to run a clinical trial, the regulatory agencies look at the CMC sections very closely so don’t tell me that this isn’t part of drug development.

  • Pingback: Demythologizing the high costs of pharmaceutical research « osberta's Blog

  • Anonymous

    Your questions are good ones. These costs differ greatly by how many Phase 2 studies are required in the ‘campaign’, how difficult/easy the synthesis is, and how potent (for dose calc’s) it is. It’s a huge range. I’d be curious to know what the #’s look like from your vantage point. Is respiratory closer to $1B than $300M? Probably.

  • Bruce

    CellBio, thx for the comments. PC certainly could be much higher – guess it depends on how much pre-designation exploratory tox and safety screening work has been done. Fair point time too, especially in a world where multiple 2a’s are being done. At the end of the indication is a huge driver of time and value. BTW, thanks for posting. I know ‘In the Pipeline’ feels like home for you, but appreciate the post!

  • Respisci

    Even within Respiratory it depends on the indication. Clinical trials in asthma, while costly aren’t as much as trials for COPD. If we could pull in a drug for $300M, I can imagine some very happy directors/VPs and investors. $1B, although a scary number, is a realistic estimate.

  • Scotsilv

    Perhaps a graph of probability/likelihood vs. cost would be better than a single figure…

  • Pingback: The Cost to Develop a Drug: Your Own Numbers? | Pharma Marketer

  • MoreMolecules

    This is like asking “how blue is the sky”? Drug discovery and developement is specific for each individual company, drug and disease.

    Why not ask “how much did it cost yuou to bring a drug to the market-start to approval?’

    Don’t hold your breath waiting.

  • John Lamattina

    Any drug that needs to have a preNDA morbidity/mortality trial will have costs that blow the socks off these estimates. This would include any drug to treat heart disease, stroke, diabetes, obeisty and even osteoporosis. In addition, the FDA is looking at the cardiovascualr impact of any new OA drug. These studies on average require 15,000 patients studied over 3 years at a cost of about $20K/patient. Thus, this single type of study in these important therapeutic areas costs $300,000,000. PreNDA outcomes studies are now also needed new new cancer drugs in areas where treatments already exist. Forget about your category on “total coasts including failures” The need to show outcome results – as well as the need to run Phase 3 differentiation studies in all disease categories – runs the costs of developing SUCCESSFUL drugs well above your $334,000,000 estimate.

  • Anonymous

    Totally agree, John. The whole point of putting out the spreadsheet wasn’t to say these were definitive numbers, but to say that each drug is likely to be different based on indication etc… For a specialty market in a small population these costs would work. I’m certain there are a number of growth factors, small cancer drugs, and orphan products that have gotten approved on less. But that’s the point I’m making – the distribution range of drug costs is huge and variable. I’d love to see the spreadsheet with your experience in it. In fact, I built such a spreadsheet for your leadership team during ATF2 and if I recall it was well north of $1.3B for Pfizer’s big primary care MM indications! Good to hear from you.

  • http://twitter.com/FundMyMR FundMyMedResearch

    This would possibly be interesting to evolve your spreadsheet by replacing the costs of projects for clinical phases by the two parameters: number of patients and cost per patient.

    This would help distinguish between the “big” (15,000 patients) and expensive ($20K per patient) studies, and the other ones.

    I think your work helps clarify the debate and show that the assertion that “any drug development costs at least $1.6B” is only a partial representation of the reality.

  • PWoitach

    It the risk of demonstrating a profound grasp of the obvious, we’ve modeled this for people as part of planning portfolio direction and used data showing that attition rates vary across TAs and across stages within TAs. It’s not a recent artcle but Kola & Landis’ piece in Nature Reviews Drug Discovery (August 2004) “Can the pharmaceutical industry reduce attrition rates?” has some good data on this that are driven by some things that may not change over time. These models, above are good and also really highlight the benefit of leveraging CMC across indications…more in terms of time since the clinic drives the cost but the ability to get more shots on goal from a discovery and preclincal program can reduce the time and risk adjusted cost at launch. It also points to the great vulue in leveraging safety and cmc data on existing drugs from one geographic market to another to solve niche problems in other indications.

  • Anonymous

    Hi Folks, I would love to see any well-documented examples of a molecule getting from start to POC success for $30M…

  • Huge fan of BB

    Interesting debate. The lower numbers are absurdly low. In my experience most (successful) discovery phase projects will take about 3-5 years to get from idea to preclinical candidate, at a cost of approximately 15-20 million dollars. This reflects activity of one full biology lab at the basic level for 3 years, ($3 million), two- to three full chemistry labs for 2-3 years ($6-8), a high throughput screen ($1 million), half of an in vivo and half of an in vitro lab for 3 years ($3 million) and some additional costs amounting to another $2-3 million. This number can be smaller for ‘me-too’ projects, but ‘me-too’ projects often require much greater sophistication and costs to ensure ‘best in class’ molecules and to deal with blind alleys that come from patents that publish along the way. That investment just gets you the molecule to start development. At the back end, phase 3 can easily go into the 300+ million range for chronic indications, as your second model indicates.

    Getting to PoC for 30 million is achievable, but ONLY if you count the relatively straightforward successes. Unfortunately, for every successful program you’ll have five to ten or more that don’t pan out along the way.

  • Anonymous

    Tom, thx for the note. As per our offline discussion, the lower numbers in the first table weren’t for Big Pharma – they were for more virtually integrated biotech going after non-primary care indications. There are a lot of such companies that have or aim to get to (positive or negative) PoC data in the clinic spending less than $30M. Several in our portfolio have projects that have cost much less than that. Getting to a Dev Can on $3-5M within 24 months of a project initiation is certainly not out of the question, especially if one has leveraged global labor cost differentials. Having 15 chemists in Sandwich is about the same cost as 45 in Shanghai or New Delhi. Inflation is happening in those regions, but its still a huge advantage in drug discovery costs.

  • Anonymous

    Great suggestion – certainly worth doing if one were to extend this model. I wouldn’t say $20K per patient is expensive though. A critical care patient or late stage patient will often cost 2-3x in a clinical trial once all the medical care, imaging and analysis are factored in. $20K per patient is probably on the low end. Hence why 15K patient trials cost $300M for the Biggest Pharma companies.

  • Anonymous

    Yes, won’t hold my breath. Would be nice if the industry was more transparent on this stuff, but its not going to happen.

  • Anonymous

    Agree – some sort of distribution is much more appropriate. Doubt its a normal distribution either. Probably skewed right, where the huge outliers like Torcetrapib’s $800M+ in direct costs. Adjusting for failures and cost of capital, if all the projects were like Torcetrapib that cost-to-NDA would be $3B+…

  • Anonymous

    Mark, as discussed in email, the preclinical IND-enabling work through PoC can easily be done on less than $15-20M for most indications ($3-5M Preclinical, $4-6M Phase 1, $6-12M Phase 2). Drug discovery for a team of 15 (primarily chemists) in the US is $4.5M a year ($300K a piece) so for 24 months is $9M. But in high quality CROs offshore, those teams are $75-90K per FTE, so it lead generation and optimization prior to Candidate selection can be done on less than $3M. Virtually-enabled, globally distributed R&D.

    There are multiple examples of this efficiency. But you won’t find them in legacy Pharma.

  • Anonymous

    Probably not high from a Pharma primary care perspective, for sure. Agree that duration and attrition rates were probably aggressive on the first model. Thanks for plugging in your numbers.

  • http://www.crowdsourcing.org Ruth Baxter

    I am not really sure if many people realize just how unique the whole concept of crowdsourcing is and will continue to be for a very long time. When I read your article and thought just how this can be applied to the cost of drugs, I really found the numbers a bit low and also some issues were not taken into consideration to give the final price. I am sure that there are those that would argue with me about this as well and then this is where the whole debate will begin. I hope people particpate in your Choose Your Own Drug Cost – Model2, because we really need this information to fully understand all the costs that go into making a drug. You may have a look our comments to your article here:http://www.crowdsourcing.org/document/choose-your-own-numbers-crowdsourcing-the-cost-to-produce-a-new-drug/2904 or just browse our website here: http://crowdsourcing.org if you wish.

  • Pingback: The drug development discord: does it cost less than we think? - SmartPlanet

  • Pingback: 404 Not Found

  • Pingback: What does a new drug cost? | eimmuneboosters.com

  • http://profiles.google.com/robwinder Rob Winder

    Looking at comments from yesterday’s PhRMA meeting Sidney Taurel said that the rate for phase II drugs going to phase III used to be about 70%, but it is now down to 50%. That makes a huge differnce in your models – an extra $100m (33% increase) in model 1 and an extra $500m (30% increase) in model 2.

  • Pingback: How Much Does It Cost To Bring One Drug To Market? - Better Health

  • http://www.semlerresearch.com clinical studies india

    Of course that range and distributions are more interesting..  I think in this post only median and mean are meaningless..