What Price Drug Development?

UPDATE 3/18/2011 - see links to Tufts Center response and PharmaExec's op-ed at the end of of the post.

There have been several pieces in the trade and popular press this week about the cost of clinical development of new drugs.  This discussion is really about drugs, not devices, as not only the costs but the way devices are researched, evaluated and cleared can be quite different.  Here is a rundown:

Slate published a piece by Timothy Noah on March 3rd called The Make Believe Billion, which takes a critical (does that surprise you?) look at Big Pharma's oft-quoted and not-often challenged claim that it costs $800 million (in 2000 dollars) and takes 12 years to bring a new drug to the patient.  These figures, adjusted upwards for inflation, are used by the Big Pharma lobby with roaring success to justify the high cost of brand-name prescription medicines.  The two numbers come from this 2003 study published in the Journal of Health Economics.  The lead author is economist Joseph DiMasi who was working for the Tufts University Center for Drug Development, an academic, non-profit study group that welcomes corporate sponsorship.  Because of the Tufts connection of DiMasi, most of the time that you see the $800M/12 year figures cited, the source of the data is often cited as the Tufts Center.

The Slate piece discovers a new challenge to the $800M claim in a study (PDF) published last month in the London School of Economics' BioSocieties journal by Princeton's Donald Light and University of Victoria's Rebecca Warburton.  Following along in the vein of The $800 Million Pill by Merrill Goozner, the Light study takes apart the source data the Tufts group used to develop their results, citing various problems.

First, the Light paper challenges the sampling of data used to feed the model.  In the Tufts study, financial information was requested from 24 companies, 10 of whom ultimately submitted data via confidential survey.  Because the data were collected in a confidential form, they are unverifiable by independent means, say the Light group.  And because the results are based on data submitted by self-selected companies that agreed to share their data, there is no true random sampling from a larger population of data.  The Tufts study does not describe any efforts to verify the reported data, so there is no way to determines whether the R&D costs are uniformly reported from sponsor to sponsor.  Additionally it is not clear whether or how much of the data reported are R&D costs for self-originated new chemical entities (NCEs), which according tot he same Tufts group are 4.4 times more costly to develop than in-licensed NCEs, which are in turn 3.4 times more costly than non-NCE variations on existing drugs.  Only 35% of new drugs developed during the reporting period (1990-2000) were NCEs, and even fewer were self-originated, or developed in-house.

The second challenge of the Light paper comes from the exclusion of discovery costs reported as part of the R&D figures used for the $800M calculation.  The reasons for this are multiple: discovery costs are much harder to tease out as discrete portions of R&D budgets because there are no recognizable chunks of work like phase 1, 2 or 3; often several compounds or targets are researched together under one budget line item; basic research is often carried out by entities other than pharma companies, like universities or the government; and sometimes new therapies are discovered quite by accident, like penicillin.  To solve this problem, the Tufts group simply factored in an average cost of $121 million and 52 months of development time, without any data to support the figures.

Thirdly, the Light group argue against the Tufts claim that tax breaks and taxpayer subsidies for pharma companies should not be discounted from the total R&D cost.  The Light paper argues that this might be reasonable if R&D were depreciated over time like over long-term investments, but they are not: R&D costs are deducted from profits each year that they are incurred.  This effectively creates a 100% tax deduction every year.  According to a 1993 US Office of Technology review of pharma research costs, the net savings was nearly 50%, and this was at a time when the top marginal tax rate was 46%.  Now that the top rate is 35%, the Tax Policy Center puts the average tax savings on R&D per year at about 39%, yet these discounts do not appear in the Tufts study.

The fourth large challenge to the Tufts study is in the assumption that half of R&D costs are accounted for by the 'cost of capital', in other words, returns that did not materialize from investing that same money in the market instead of using it for research, assuming an 11% return in the period under study.  Calculating the cost of capital is a common method for corporations to decide whether to pursue new projects, but the Light authors claim it is a bit ingenuous for the pharma industry to then claim that the cost of capital, what the pharma company would have made on the market instead of a new drug, should be considered part of the cost of R&D, especially when developing a new drug is the business of the pharma company, not to invest money.  And when that cost of capital calculation ends up as half of the claimed $800 million R&D costs.

There are some other challenges in the Light paper too, about inflation of individual per-patient trial costs, exaggerated time for total development (52 months preclinical, 72 months for trials and 18 months for review, or 11.8 years), and the claim that only 1 in 10,000 compounds ever make it to the marketplace, the use of median costs instead of means to counter the shifting effect of outlier data points, i.e., more expensive or large trials.  When all is said and done, the Light group claim that the average cost of the "D" part of bringing a drug to market is more like $59 million based on the 2000 data in the Tufts study.  In today's dollars that would be $75 million average, or $55 million median.

Not so fast, says research scientist Derek Lowe in his blog In the Pipeline.  He challenges that it is a lot harder than people think to come up with compounds and targets that actually have some potential clinical benefit while remaining safe enough for human use, and that you can't simply not count discovery research costs simply because they are harder to identify.  From my own perspective on the clinical side of things, during the time represented by the Tufts study, we were routinely running $20 million/year clinical budgets to develop an antiretroviral medication for CMV, and the study sizes were relatively low, just a few thousand patients.  The clinical development period of that product was around 5 years if memory serves, and of course it was an accelerated program since it was largely developed for HIV+, full-blown AIDS patients.

Later this week we saw two other articles make the rounds:  one from CenterWatch that pleaded the pharma manufacturers case that if only they could get longer data exclusivity, companies would be able to innovate again and patients would live longer.  Extending data exclusivity to 12 years would net the pharma companies 5% additional profits, resulting in 228 new drugs over the next 50 years and increasing the average person's lifespan by...wait for it...1.7 months.  The article is here.

And lastly, today's paper had a piece that got some pretty strong reactions from all sides, where KV Pharma announced that injections of their drug Makena, which now cost $10-$20 per dose, next week will cost $1500, the total cost to the patient ranging from $27,000 depending on length of gestation.  The company's CEO defends the price increase by pointing out that PICU care can cost $51,000 or more, and Makena can defray some of that cost.  The reason for the change?  Makena is a newly approved form of progesterone that until now was provided by compounding labs for literally a few cents worth of chemicals.  Compounding will no longer be legal now that an approved version is available.  KV Pharma has had all kinds of problems recently, having entered into a consent decree for making and distributing adulterated and unapproved drugs, pleading guilty to two criminal fraud for failing to report oversize tablets to the FDA, but all that seems to be in the past now.  Their stock is up from $1.50 in early February to close at $12.64 today.


3/18/2011 - Tufts Center response is here; an op-ed on the PharmaExec blog is here.