Send to printer »

Point of View : Jul 1, 2011 (Vol. 31, No. 13)

Drug R&D Costs Questioned

Widely Quoted Average Cost to Bring Drugs to Market Doesn’t Appear to Hold Up to Scrutiny
  • Donald W. Light, Ph.D.
  • ,
  • Rebecca Warburton, Ph.D.

A claim stated over and over becomes a fact. A specific number, like $1.32 billion, sounds scientific, based on fact. Thus, when one of the world’s most extensive press operations, whose full-time staff of journalists write copy for editors, TV producers, government reports, and internet outlets, tells us over and over that the average cost for research (R&D) to bring a new drug to market is $1.32 billion, we assume that the pharmaceutical industry knows what it is talking about.

As a policy researcher and health economist, we discovered that this staggering figure was based on companies reporting closely guarded numbers for their R&D costs to the pharmaceutical industry’s leading policy research center. No one knows what they count as R&D costs, though bits of evidence indicate it’s rather inclusive and varies from company to company.

Companies have strong incentives to make their R&D costs as large as possible to justify higher prices, more government protection from free-market competition, and greater tax breaks. The pharmaceutical industry depends more on government subsidies and protections than almost any other.

We concluded, in the Journal of Health Economics, that this cost data was so subject to internal and external sources of variability that no one should trust any estimate based on it. Yet everyone does, because the high cost of R&D has been repeated like a mantra until it passes for fact.

R&D Estimate

The industry-supported team of health economists behind this estimate has been working on ways to make the estimate of R&D as high as possible since 1976. The $1.32 billion figure (note the false precision to the 1/1,000th place, when no calculation should be more precise than its least precise data point) is based on a small, highly biased sample of drugs first tested in humans between 1983 and 1994 that resulted in an article claiming that in 2000 the average cost of R&D was $802 million.

This figure was then “updated” to $1.32 billion for 2006 by estimating R&D costs had risen 64% since 2000. Reasonable doubts about this staggering figure begin with audited company figures to the IRS of trial costs in the late 1990s averaging just $22.5 million. How did such audited costs become so mythically huge?

First, the $1.32 billion estimate is based on the 22% of drugs with the highest R&D costs, new chemical entities (NCEs) developed in-house by a company. Yet from the beginning, these costs were misrepresented as the average cost for all drugs, even though self-originated NCEs are reported to be 4.4 times more costly than licensed-in discoveries (usually from small biotechs) and 15 times more costly than most newly approved drugs, which are minor variations on existing molecules.

This misstatement was made at a national press conference more than a year before publication in the prestigious Journal of Health Economics. Strangely, the editors did not regard this as prepublication or a clear sign of unscientific bias. In fact, they published the study as their lead article.

More cause for reasonable doubt about the high cost of R&D comes from the trial cost per subject in the study being six times greater than trial costs reported by the National Institutes of Health, and trial sizes being twice as large. Should one then divide the $1.32 billion by 12? Further, using the mean cost rather than the median cost allows a few very expensive drugs to pull up the figure. Taxpayer contributions are not counted, though altogether they allow companies to recover at least half their R&D costs.

If one were to include the large portion of all taxes that pharmaceutical companies do not pay by arranging for profits to accrue in foreign tax havens until Congress agrees to bring them home for a fraction of what other companies pay, tax concessions might pay for nearly all pharmaceutical R&D costs. This unpatriotic practice is called repatriation, while other Americans pay more taxes every year to make up what tax-haven companies do not pay.

Thus, what matters to pharmaceutical companies is their net, median costs, not their gross average costs.

One has further reasonable doubts upon learning that the trials used in the small, biased, nonrandom study on which the $1.32 billion figure is based reportedly took 72 months to complete, when registered data show all trials averaged 36 months. The study also said regulatory review took 18 months, when registered data show it averaged 11 months. Since half of the $1.32 billion comes from compounding the secret costs submitted by companies by 11% for the “cost of capital,” these 42 additional months inflate the R&D cost estimate considerably.

If one is to include the cost of capital at all, the standard rates used by economists are less than half the rate used in this industry-supported study. But while calculating the cost of capital (essentially the estimated profits one would make if one were not to invest in a project but put the money in Treasuries, stocks, or bonds) is a common and reasonable exercise used by companies, should it be a claim against society as a real “cost”?

If Hilton spends seven years and $50 million to build a new hotel, should it add all the returns it would otherwise have made investing the $50 million to your room rate? “Our standard room in older hotels is $200, but in this new hotel it’s $600. You should see how much time and money we put into developing it for you.”

Should Intel’s new chip cost four times more because it put hundreds of millions into developing it?

In a free market, this is impossible—companies have to take their chances in a price-competitive market. But in the pharmaceutical market, it is common practice to charge several times more for new drugs. The allegedly high development costs are used as an excuse for high prices. Yet 85% of new drugs developed are judged by several independent review groups as having few or no advantages over existing drugs. Unfortunately, taxpayers, employers, and insurers often accept high prices uncritically.

The authors of this R&D study had no data on the costs of basic research (the “R” in R&D); so they backed in a large number and then compounded it by 11% per year. No one has reliable data on the costs of discovery and “average” means very little because costs vary so widely. Serendipity plays an important role, and it is often hard to attribute basic research costs to a specific drug.

“R” can take from three months to 30 years, and no one has come up with a meaningful “average” estimate. In another study, we drew on evidence from the National Science Foundation and official reports to show that 84% of the world’s funds for “R” come from public or foundation sources.

In the end, we corrected for what inflators we could using the same data and cost of failures used for the $1.32 billion estimate, and concluded that the net, median costs for R&D in 2000 consisted of an unknown and highly variable “R”, plus about $60 million for “D” that varied from $13 million to $203 million, depending on the type of drug.

Updated to 2006 by 64%, these would be “R” plus $98 million and range from $21 to $333 million. But we don’t trust our estimates to reflect real costs any more than we trust the self-serving estimates of the industry. Our main goal is to educate readers about how the costs of R&D can be manipulated to “show” they are astronomical or very modest, so long as readers and policy makers are gullible and don’t have reasonable doubts about how the estimates are made.

What really matters, however, is not the cost but how pharmaceutical R&D is largely used to develop so many new drugs that offer few or no advantages yet bear risks of harmful side effects. Those side effects are now the fourth leading cause of death. If R&D costs got us better drugs to make us healthier, they might be worth it; but we’re largely getting a proliferation of harmful side effects for little benefit.