Despite the variety of health systems across hundreds of different countries, one feature is near-universal: We all depend on private industry to commercialize and market drug products. And because drugs are such an integral part of our health care system, that industry is generally heavily regulated. Yet despite this regulation, little is publicly known about drug development costs. But aggregate research and development (R&D) data are available, and the pharmaceutical industry spends billions per year.
A huge challenge facing consumers, insurers, and governments worldwide are the acquisition costs of drugs. On this point, the pharmaceutical industry makes a consistent argument: This is a risky business, and it costs a lot to bring a new drug to market. According to PhRMA, the U.S. pharmaceutical industry’s advocacy group, it cost $1.3 billion (in 2005 dollars) to bring a new drug to market. The industry argues that high acquisition costs are necessary to support the multi-year R&D investment, and considerable risks, in to meet the regulatory requirements demanded for new drugs.
But what goes into this $1.3 billion figure? To understand the cost of a new drug, we need to consider both the cost of drugs that were marketed, but also factor in the costs of the failures – those discontinued during development. While most pharmaceutical companies are publicly held, no company produces detailed breakdowns of “per marketed drug” R&D costs, or the specific amounts spent on drugs that were later abandoned. Yet there have been attempts to estimate these values. The most detailed and perhaps controversial paper is a 2003 paper from DiMasi et al, entitled, The Price of Innovation: New Estimates of Drug Development Costs.[Light and Warburton, entitled Demythologizing the high costs of pharmaceutical research. They claim the median R&D cost is a fraction of DiMasi’s estimate: Just $43.4 million. “Big Pharma lies about R&D to justify illicit profits” shouted Natural News. Who’s right?
Drugs can be developed in different ways, but the usual model used describes a series of phases. The pre-clinical development stage constitutes preliminary studies of chemicals that have been synthesized or isolated, and are then screened. This process can take years: Identifying promising leads, validating them, tweaking with their chemical structures, and conducting endless in vitro studies. Only a fraction of drugs that show promise in pre-clinical studies will every progress to clinical trials. Clinical trials are generally grouped into three stages, each one representing an important milestone in a drug’s development. Phase I studies are small studies in healthy volunteers designed to help understand the basic pharmacology and pharmacokinetics in humans: how a drug is absorbed, distributed, metabolized, and eliminated. It’s in Phase II that the drug is tested in groups with the condition of interest. These trials are larger, and may be randomized, with multiple arms, possibly evaluating different dosing regimens. Endpoints are usually related to basic efficacy and safety parameters. Phase III studies are the largest studies, that may be randomized and double-blind, in order to establish a drug’s efficacy against a given condition. Regulators like the FDA will usually require one or more Phase III trials to support an approval to market a drug. In cases where real outcomes need to be measured (like mortality or morbidity), phase III studies can be massive. (Like this one, with over 18,000 participants!).
While the trial pathway is usually illustrated as a straight-line path, that’s a post hoc view: A tree may be a more appropriate model. Clinical trials may be conducted in different doses, treating different patient groups, using different protocols, in order to understand a drug’s effectiveness.
At any one time, multiple drugs may be in development, so only the most promising products may move forward in the development pathway, as subsequent phases of development mean a significant increase in costs. A drug’s development can be discontinued at any point along the path. Developers may identify toxicity issues, or lack of effectiveness issues. Or clinical practice may change, and all of a sudden, the clinical trials are measuring the wrong endpoints in the wrong patients. Decisions are always made in the face of uncertain evidence about efficacy and toxicity, and for every drug that moves forward that is eventually found to fail, it could be that there is an effective drug that didn’t reveal itself as promising, and went back on the shelf.
The development process is laborious and typically takes several years from discovery to clinical trials. The pharmaceutical industry estimates that it takes 10,000 molecules developed to bring a single drug to market. Without validating that number (a whole other post), it’s fair to say that the number of drugs that make it to market is a tiny fraction of the number of products identified or synthesized that enter initial screening. So there will be a substantial investment into drugs that never make it to market. Without including the cost of abandoned drugs into the costs of drugs that are marketed, we’d be underestimating the investment incurred. So any analysis needs to consider this cost, too.
The DiMasi Paper
The DiMasi paper, from the Tufts Center for the Study of Drug Development is the most widely cited paper on drug development costs. While the methodology they use is described in detail, some essential information is unfortunately (though perhaps understandably) opaque. The authors used a sample of drug costs drawn from ten companies that volunteered (out of 24 that were asked) that were willing to provide R&D data on a per-chemical basis. Data were collected, and stratified by development phase. Only the costs of self-originated drugs (i.e., they developed the drug themselves) were included. In total, data on 68 products were collected, and the sample consisted of mostly small-molecule drugs, but also included four recombinant proteins, two monoclonal antibodies, and one vaccine. No further information is provided, so there’s no way to know just how representative this sample is.
The methodology for the different calculations is fairly well detailed, but as I noted, the underlying data are not provided. Whether this basket of drugs studied represents a fair measure of the market is impossible to determine. The authors compiled actual costs wherever possible, broken down by development phase. A notable exception is the “preclinical” development period where it’s difficult to draw a direct link between expenditures and a specific drug that ends up being commercialized. For this segment, they inferred, using their own database, costs of $121 million per approved new drug.
To account for the costs of drugs that were abandoned (for any reason) during development, the authors used their own database of investigational drugs to estimate the odds any given new drug would reach a particular development milestone. Setting aside a detailed analysis of the methodology, let’s look at the two biggest cost drivers of the final that have been subject to repeated criticism: Tax issues, and the cost of capital.
A major criticism of the DiMasi paper has been that the preferential taxation provisions for R&D expenses have not been factored into the analysis. Essentially, if R&D costs are given preferential tax treatment, this should reduce the net cost of R&D to the company. I have no particular insight into this issue other than to flag it as one that has caused controversy. Given preferential treatment of R&D expenses isn’t unique to the pharmaceutical sector, the extent to which this biases the validity of this particular analysis isn’t clear to me. But I’m a pharmacist, not a tax expert.
The Cost of Capital
Probably the biggest criticism of the DiMasi paper is that the authors factor in what’s called the cost of capital into the development cost. Looking at the calculations, DiMasi estimated the out-of-pocket costs per new drug at $403 million (2000 dollars). But this is then capitalized, based on the opportunity cost of that investment – at 11%, bringing the “total” cost up to $802 million. Adjusting this cost to 2005 dollars, and we’re at the $1.3 billion that PhRMA is calling “the average cost to develop one new drug.”
The cost of capital can be a bit baffling to understand. If I’m going to invest my money in something now, with a possible payoff down the road, I need to factor in the opportunity cost of something else I could have invested in – but decided not to. It is a true cost, because by choosing to invest in one thing, you’re forgoing the investment in another.
DiMasi uses a cost of capital of 11% – that is, they assumed that the drug developers, by moving forward with the development of a drug, where forgoing investments which would be expected to yield 11%. Is 11% valid? From a personal investment perspective, 11% seems rich. But the cost of capital that companies use is dependent on the risk involved. Different industries have different business risks. The DiMasi paper bases the 11% estimate based (in part) on historic returns in the industry. Given that half of the reported “cost” of a new drug is based on the cost of capital, the value we use use has a massive influence on what the final “cost” of a new drug will be. But is 11% appropriate? Many argue no – that current returns don’t match past returns, and therefore the CoC should be lower. I took a look at a cost of capital table created by Aswath Damodaran, a Professor of Finance at the Stern School of Business at New York University. He calculates that pharma’s cost of capital is 8.59%. But there is no single “right” answer here. It’s an assumption that goes into our calculation.
Other authors have made their own attempt at estimating the cost of a new drug. Paul Adams of the Federal Trade Commission, writing in Health Economics estimates that the DiMasi estimate is low, and the 2003 cost is closer to $1 billion per new drug, but noted there is significant variation between products. A 2006 Congressional Budget Office report on drug development [PDF] largely supports the DiMasi estimates. Most recently, Light and Warburton argued that, “based on independent sources and reasonable arguments, R&D costs companies a median of $43.4 million per new drug, just as company supported analysts can conclude they are over 18 times larger, or $802 million.” This figure seems implausibly small, given a single clinical trial can involve hundreds to thousands of patients. (For a more detailed critique of the Light and Warburton paper, I’ll refer the interested reader to Derek Lowe’s excellent In the Pipeline blog (and its comments) where it was dissected in detail here and here.) One of the best ways to contemplate the costs and calculations is to manipulate the numbers yourself: There’s a model developed by venture capitalist Bruce Booth, where you can enter your own estimates and see what cost it spits out. I tried working with the model for a while, and I couldn’t get it anywhere near $43 million – it was always in the hundreds of millions.
One important factor that isn’t considered in any of these analyses (from what I can see) are the costs of new indications for existing chemical entities. Consider the case of cancer drugs, where drugs are often approved for the treatment of metastatic disease, and only after efficacy is demonstrated, is it studied as a potential “adjuvant” treatment for early stages of disease. Additionally, the DiMasi analysis only looked at drugs developed solely in-house. Given the growing role of smaller biotech companies that develop, and then sell, promising drugs to pharmaceutical companies, the impact on costs isn’t clear. In contrast, the cost of the “me-too” drugs that seem to fill the pharmaceutical marketplace aren’t discussed explicitly, either. When your new drug is a variation on a competitor’s (or your own) product, how does this influence overall R&D expense? Again, it’s not clear.
Is the $1.3 billion new drug a myth? New drugs could be hitting, or even exceeding this mark – it depends on what your assumptions are. When we try to summarize all the variables of drug development into a single number, accounting for the hits and the misses, we can end up with a number that sounds impressive. But is it meaningful? Without transparency, only the manufacturer will know what it cost for their own drugs. It’s probably more important to understand the key drivers of R&D costs, noting that there are a huge number of variables that may influence the final cost of bringing a new drug to market.