Commentary
Misleading Congress about Drug
Development: Reply
Joseph A. DiMasi
Tufts University
Ronald W. Hansen
University of Rochester
Henry G. Grabowski
Duke University
Abstract The review essay by Donald Light about a Congressional Budget Office
report on pharmaceutical research and development (R&D) (Light 2007) contains
factual errors, leaves the reader uninformed about rebuttal responses to criticisms
made in the review about studies of R&D costs, and draws erroneous conclusions
about the nature of industry economics.
The review essay by Donald Light about a Congressional Budget Office
report on pharmaceutical research and development (R&D) (Light 2007)
contains factual errors, leaves the reader uninformed about rebuttal
responses to criticisms made in the review about studies of R&D costs,
and draws erroneous conclusions about the nature of industry economics.
We briefly address these problems.
Nearly all of what Light writes in his section on development costs
in relation to our study of pharmaceutical R&D costs (DiMasi, Hansen,
and Grabowski 2003) has been published in two comments in the journal in which our study appeared (Light and Warburton 2005a, 2005b).
Light, however, does not inform readers that we have rebutted all of these
points in two responses in that journal (DiMasi, Hansen, and Grabowski
2005a, 2005b). Since we have already published detailed responses to the
criticisms, here we simply direct readers to those publications. Light does
reference one of our responses, but he makes no mention that it contains
rebuttals to his criticisms. Instead, he cites it, incorrectly, as indicating
that we are working on estimates of R&D costs for what he calls “me-too”
Journal of Health Politics, Policy and Law, Vol. 33, No. 2, April 2008
DOI 10.1215/03616878-2007-063 © 2008 by Duke University Press
JHPPL332_07C_DiMasi.indd 319
2/24/08 3:38:58 PM
320 Journal of Health Politics, Policy and Law
derivative products. Other factual misstatements in Light (2007) include
the claim that the CEO of Merck “hosted and launched” the first public
presentation of our study of pharmaceutical R&D costs (his role was only
that of an invited guest speaker at the forum in which we presented our
results) and the implication that F. M. Scherer’s (2001) study on the relationship between pharmaceutical industry profits and R&D expenditures
was supported by the Merck Foundation and the National Pharmaceutical
Council (Scherer did not receive any financial support for his study).
R&D Costs
In his section on pharmaceutical R&D costs, Light also touts the alternative estimates offered in a report by the Congress Watch division of
the advocacy group Public Citizen (2001). However, the Public Citizen
analysis is deeply methodologically flawed in ways that make their estimates seriously biased downward. Since we have already written about
the report’s flaws in detail, we direct readers to our paper (DiMasi, Hansen, and Grabowski 2004).
One flaw in the Public Citizen report is echoed several times in Light
(2007) and so is worth addressing in some detail here. Light maintains that
our estimate of the cost to get a new molecular entity (NME) to market
is not representative of new drug development costs because our sample
excluded the (presumably lower) costs associated with all of the other new
drug application (NDA) approvals made by the Food and Drug Administration (FDA) for a drug after it had obtained initial marketing approval
(i.e., NME approval). These approvals, for the most part, are line extensions. Firms in this and in many other industries fill out their product lines
for a basic product to better serve varied consumer needs and desires (e.g.,
oral solutions for children or adults who have difficulty swallowing tablets
or capsules, more convenient dosing regimens, or different side-effect or
efficacy profiles). In the case of drugs and the FDA statistics on NDA
approvals, many of the approvals, in fact, are obtained long after the drugs
have lost patent expiration by firms other than the sponsor of the original NME (often by small, specialty pharmaceutical firms or generic drug
companies). Even leaving aside this substantive point about the makeup of
the approval statistics, it is not appropriate to treat the incremental costs of
later regulatory approvals associated with an active ingredient as though
they were independent of the costs of obtaining the original approval. This
would make no more sense than it would to take all of the R&D costs
associated with developing a new automobile model and dividing that by
JHPPL332_07C_DiMasi.indd 320
2/24/08 3:38:59 PM
Commentary 321 the number of trim lines of that model that the manufacturer happens to
offer for sale and then use that as a measure of the R&D cost of developing
a new automobile model.
Light incorrectly claims that we are currently working on cost estimates
for individual line extensions and will then add them later to preoriginal
approval costs. In fact, our published study of R&D costs (DiMasi, Hansen, and Grabowski 2003) already contains an estimate of the average cost
for all postoriginal approval R&D, which was then appropriately added
to preapproval costs to get an estimate of full life-cycle R&D cost per
approved new drug ($897 million in year 2000 dollars on a capitalized
basis taken at the point of original marketing approval).1
Light (2007) also repeatedly makes the claim that $300 million is
the median R&D cost per approved new drug for our data. There is no
such figure in our study, nor does Light indicate how such a figure can be
derived. In fact, it is impossible to construct a meaningful overall median
value for these data because the R&D expenditures include costs for the
bulk of new drugs that fail in testing at some point in the preclinical or
clinical development process and therefore do not have costs for all development phases and because some of the early expenditures cannot be allocated to individual compounds.
Light (2007) also questions the data that we had on clinical trial sizes
by comparing our average number of subjects per approved new drug to
much lower figures given in Love (2003) and later updated on the Listserv belonging to the same group (Consumer Project on Technology, now
called Knowledge Ecology International [KEI]). Light portrays these
alternative averages as though they had been reported by the FDA itself. In
fact, what KEI did was to treat whatever sample sizes they found reported
in the “Clinical Studies” sections of the labels for FDA-approved drugs
as if they represented the total number of subjects that had been included
in industry clinical trials for the drug. What the FDA generally discusses
in these sections of the label are the results for a few efficacy trials that
were considered pivotal for the indication that was approved. The section
does not necessarily note all phase 3 trials and typically excludes phase
1 and phase 2 trials. The result is that the KEI averages substantially
underestimate the total number of subjects in industry trials. In a posting
on the same Listserv that contained the KEI averages, one of us made
these points and provided illustrative examples, using information in other
1. If one then wants to compare these costs to net sales, then one should look at total lifecycle sales for all product presentations containing the active ingredient.
JHPPL332_07C_DiMasi.indd 321
2/24/08 3:38:59 PM
322 Journal of Health Politics, Policy and Law
parts of the FDA-approved label, that demonstrate unequivocally that the
KEI numbers can be substantially off the mark (DiMasi 2006). Light also
notes the clinical trial sizes postulated in a report on the development of a
hypothetical drug for tuberculosis (Global Alliance for TB Drug Development 2001). However, this is a special case that was never meant to represent drug development as a whole, and there are numerous reasons why it
should not be used to imply anything about clinical trial sizes or development costs in general. Our report critiquing the Public Citizen report also
discusses the Global Alliance for TB Drug Development report (DiMasi,
Hansen, and Grabowski 2004). In addition, as noted in our study (DiMasi,
Hansen, and Grabowski 2003), our data on clinical trial sizes were consistent with those from another source whose estimates are not subject to the
same methodological errors made by KEI (PAREXEL 2002).
Profitability
In his discussion of industry profitability, Light (2007) argues that the
standard economic treatment of R&D costs as investment expenditures
that create intangible assets is wrong simply because R&D expenditures
are expensed on tax returns. One has nothing to do with the other. It is
widely recognized that a more appropriate tax structure would require
that R&D expenditures be capitalized and amortized over the life cycles
of the products created as a result of that R&D. However, in light of the
substantial uncertainties and arbitrariness that would arise in making a
forward-looking assessment of the benefits that might be linked to current R&D expenditures, Generally Accepted Accounting Principles and
other accounting guidelines take the most conservative route and allow
for expensing of R&D expenditures. The desire of accounting guideline
setters to be practical with regard to the tax treatment of R&D expenditures says nothing about how a proper economic analysis of past R&D
expenditures should be conducted.
Finally, Light misconstrues industry data when he argues here and
originally in Light and Lexchin (2005) that lower prices in the United
Kingdom and Canada have no negative impact on R&D and that these
countries cannot be called “free riders” because local R&D expenditures
as a share of local sales is purportedly greater in these countries than in
the United States.2 The markets for pharmaceuticals are typically global,
2. Foreign free-rider arguments are more accurately referred to as “cheap-rider” arguments
(Scherer 1993).
JHPPL332_07C_DiMasi.indd 322
2/24/08 3:38:59 PM
Commentary 323 as is the R&D that creates products for those markets. This renders comparisons between local R&D expenditures and local sales meaningless.3 A
relevant metric would be based on contributions to global quasi rents (revenues minus variable costs). Specifically, the metric would be a country’s
net sales as a share of global net sales considered on a per capita basis. It
is widely thought that the United States contributes most of the pharmaceutical industry’s profits in absolute terms, let alone on a per capita basis.
Furthermore, Light’s arguments suggesting that drug companies could
make up for the imposition of European and Canadian level prices in the
United States by reducing marketing expenditures while maintaining or
increasing R&D expenditures are specious. If firms can increase profits
by simply reducing marketing expenditures when prices are lowered, then
they could also increase profits by doing so now at current prices. In other
words, Light’s implication could only be true if firms do not currently set
their marketing expenditures so as to maximize profits. We doubt that
Light would make that argument.
References
DiMasi, J. A. 2006. Reply: Size of Trials by Status (S or P) — Some 2004 FDA and
PAREXEL Data Compared. July 27. lists.essential.org/pipermail/ip-health/2006
-July/009888.html.
DiMasi, J. A., R. W. Hansen, and H. G. Grabowski. 2003. The Price of Innovation: New
Estimates of Drug Development Costs. Journal of Health Economics 22:151 – 185.
——— . 2004. Assessing Claims about the Cost of New Drug Development: A Critique of
the Public Citizen and TB Alliance Reports. November 1. csdd.tufts.edu/_documents/
www/Doc_231_45_735.pdf.
——— . 2005a. Extraordinary Claims Require Extraordinary Evidence: Reply. Journal of Health Economics 24:1034 – 1044.
——— . 2005b. Setting the Record Straight on Setting the Record Straight: Response to
the Light and Warburton Rejoinder. Journal of Health Economics 24:1049 – 1053.
Global Alliance for TB Drug Development. 2001. The Economics of TB Drug Development. New York: Global Alliance for TB Drug Development.
Light, D. W. 2007. Misleading Congress about Drug Development. Journal of Health
Politics, Policy and Law 32:895 – 913.
3. Ironically, at later points in both Light (2007) and Light and Lexchin (2005), the authors
seem to acknowledge the global nature of pharmaceutical research and development (R&D) and
sales, even though this conflicts with their earlier free-rider arguments.
JHPPL332_07C_DiMasi.indd 323
2/24/08 3:38:59 PM
324 Journal of Health Politics, Policy and Law
Light, D. W., and J. Lexchin. 2005. Foreign Free Riders and the High Price of U.S.
Medicines. BMJ 331:958 – 960.
Light, D. W., and R. Warburton. 2005a. Extraordinary Claims Require Extraordinary
Evidence: Comment. Journal of Health Economics 24:1030 – 1033.
——— . 2005b. Setting the Record Straight in the Reply by DiMasi, Hansen, and
Grabowski. Journal of Health Economics 24:1045 – 1048.
Love, J. 2003. Evidence Regarding Research and Development Investments in Innovative and Non-Innovative Medicines. September 22. www.cptech.org/ip/health/rnd/
evidenceregardingrnd.pdf.
PAREXEL. 2002. Emerging Clinical Trial Enrollment Benchmarks: Trial Size Statistics for New Drugs Approved in 2001 (2000, 1999, 1998). In PAREXEL’s Pharmaceutical R&D Statistical Sourcebook 2002/2003, ed. M. P. Mathieu, 108 – 111.
Waltham, MA: PAREXEL International.
Public Citizen. 2001. Rx R&D Myths: The Case Against the Drug Industry’s R&D
Scare Card. July 23. www.citizen.org/documents/rdmyths.pdf.
Scherer, F. M., 1993. Pricing, Profits, and Technological Progress in the Pharmaceutical Industry. Journal of Economic Perspectives 7 (3): 97 – 115.
——— . 2001. The Link between Gross Profitability and Pharmaceutical R&D Spending. Health Affairs 20 (5): 216 – 220.
JHPPL332_07C_DiMasi.indd 324
2/24/08 3:38:59 PM