How much in revenues do drug companies have to generate to maintain current levels of investment in research and development? Joseph DiMasi and colleagues ¹ try to answer that key question in the estimation of the costs of pharmaceutical research and development (R&D). These researchers used survey data from ten drug companies to compile R&D costs for 68 investigational compounds. Their estimations indicate that for every one new chemical entity brought to market, firms can expect, on average, US$403 million in actual cash outlays for R&D (all values are adjusted for inflation to constant year-2000 US$ unless otherwise indicated).

This estimate includes preclinical and clinical costs for compounds that actually make it to market and for compounds that do not make it to market. Opportunity costs of not investing that $403 million elsewhere--referred to as the cost of capital and estimated1 at 11%--increases the total to $802 million for each new chemical entity. Although some argue against including the cost of capital, it is especially important in answering the above question since firms could invest elsewhere.

When considering the precision of these estimates, there are important issues to keep in mind. First, estimates of the cost of capital do not account for the sometimes substantial tax advantages for R&D expenditures. For example, a 1993 study² used a marginal corporate tax rate of 34% that reduced estimates of drug-company capital expenditures from $259 million to $171 million (adjusted to 1990 US dollars). Second, DiMasi and colleagues' estimates1 are based on survey data from a small number of firms (10 of 24 firms responded) that may well have had a vested interest in the outcome of the study. Third, the underlying data have high variance, so point estimates are inexact representations of costs that a firm may incur. A better representation of the data would be the range DiMasi found using Monte Carlo simulation, in which 1000 cost estimates were generated by simultaneously selecting different values for each variable from its probability distribution. In this random-generation process, Monte Carlo simulation provides a distribution of estimates that captures the uncertainty in the underlying data. In the DiMasi simulation, for example, 95% of cost estimates that included cost of capital fell between $684 million and $936 million.

Even though these Monte Carlo results provide a sense of the variance in the underlying data, they do not account for bias in the underlying data. To show their results are on target, DiMasi and colleagues1 compare their estimates of percentage of expenditure to other sources. However, when large figures are being evaluated, small differences in percentages (eg, 11·8% vs 11·4%) can mask large differences in actual expenditure. They also compare their actual estimates to industry totals as published by the industry group, the Pharmaceutical Research and Manufacturers of America, and found similar results. However, the validity of this comparison is also uncertain since both estimates use data self-reported by the industry.

The pharmaceutical industry uses the high cost of R&D to justify monopoly pricing and continued product protection. Yet these high costs provoke the question: are we overinvesting in this type of research? According to the US National Science Foundation, manufacturers of pharmaceuticals and medicines (as categorised by the Foundation) report industry spending on R&D to be almost $13 billion in 2000 alone. This number was just over $12 billion in 1999 (in 1999 US dollars)³ and has been steadily increasing since the 1960s. Given limited resources, should the goal of policy be to continue the current trend of investment? Are there still sufficient benefits to justify this level of investment? Can society afford to continue paying for expensive new drugs and other medical advancements? These are all open questions.

Some might argue that society can afford it. Sean Tunis, chief medical officer at the US Center for Medicare and Medicaid Services, was quoted in reference to expensive new medical procedures: "if the technology was effective, we would find a way to pay for it. There is no dollar value per life per year at which Medicare would decline to pay."⁴ Yet how much longer can society continue to "find a way to pay for it"? Are there better options? Is it possible to spend more to reduce poverty and improve education and infrastructures, thereby improving people's health and other aspects of their lives? In so doing, could we also reduce the need for new drugs and other medical advancements, and direct more resources toward other technologies?

This is certainly not to belittle the efforts made in pharmaceutical medicine in the past century, but perhaps society needs to think about re-evaluating its investment portfolio. Before that can be done, however, more research is needed to validate these firm-level cost estimates and to quantify the societal costs and benefits of these advances. Only then can we have an informed discussion about setting policies to optimally allocate resources for developing new drugs and other medical advancements.

My opinions and conclusions do not necessarily represent those of the US Bureau of the Census.

T Lynn Riggs

US Bureau of the Census, Center for Economic Studies, Chicago Census Research Data Center, Chicago, IL 60604, USA (e-mail:tlriggs@frbchi.org)

1 DiMasi JA, Hansen RW, Grabowski HG. The price of innovation: new estimates of drug development costs. J Health Econ 2003; 22: 151-85. [PubMed]

2 US Congress, Office of Technology Assessment. Pharmaceutical R&D: costs, risks and rewards. OTA-H-522. Washington, DC: US Government Printing Office, February, 1993.

3 National Science Foundation, Division of Science Resources Statistics. Research and development in industry: 2000. NSF 03-318. Arlington, Virginia: National Science Foundation, May, 2003: http://www.nsf.gov/sbe/srs/nsf03318/start.htm (accessed Nov 3, 2003).

4 Kolata G. Newest treatments create a quandary on Medicare costs. New York Times National Aug 17, 2003: Sect YT:1;13.