by James Love
January 13, 1997
The following is an article titled "Call for More Reliable Costs Data on Clinical Trials," Published in the January 13, 1997 issue of the Marketletter, on pages 24 and 25. The article examines evidence from the US Orphan.Drug Tax Credit, from 1983 to 1993, as well as data from 58 NIH funded human-use clinical trials. These data are then compared to the estimates of the costs of human-use clinical trials presented by DiMasi, Lasagna, Grabowski and Hansen, in a series of well known papers, and used in the 1993 OTA study on drug development costs. The highpoints include:
In February 1993, the now defunct US Office of Technology
Assessment issued a report on pharmaceutical R&D, commissioned
to provide lawmakers with an independent analysis of the cost
of developing a new drug (OTA-H-522). While the OTA report is
full of numbers on countless topics, most people are aware of
just one. The OTA said $359 million was the "upper bound
on the full cost of bringing New Chemical Entities to market."
This figure has been used by the multinational pharmaceutical
companies to justify the ever-increasingly prices for new drugs.
However, many readers of the report were struck by
the paucity of independent data collection, and the reliance upon
pharmaceutical company consultants for the core findings. Stripped
to the core, the 1993 OTA report was simply a restatement of a
1991 paper from the Journal of Health Economics, by four economists
with well known ties to the industry.
In that paper, Joseph DiMasi and Louis Lasagna from
Tufts University, Henry Grabowski from Duke and Ronald Hansen
from the University of Rochester put the cost of developing a
new drug was $231 million, in 1987 dollars. Because of the industry's
close ties to the report's authors, and the because the Pharmaceutical
Manufacturing Association (now the Pharmaceutical Research and
Manufacturers of America) published a version of the 1991 study
under a PMA cover, the study was widely seen as an industry estimate.
This paper, often referred to as the "Tufts" study,
was controversial, in part because the figure was far higher than
earlier estimates, and also because it relied upon data from an
unaudited and confidential industry questionnaire.
The OTA did not have access to its own data on R&D
outlays, and Congress never issued a subpoena for the data. Faced
with this lack of information, the OTA hired Mr. DiMasi to recalculate
his early estimates using 1990 dollars, and using a range of industry
"discount rates" to measure the cost of capital. The
$359 million "upper bound" figure was simply the 1991
Tufts study with the numbers adjusted to 1990 dollars, and using
a 14% real rate of capital - up from the 9.5% used in the 1991
report. The basis for the Tufts study and the OTA report was a
data set which consisted of expenditures on human-use clinical
trials and animal testing for 93 NCEs. DiMasi et al reported that
for the sample, the average total out-of-pocket costs of Phase
I, II and III clinical trials were $18.9 million, plus $2.8 million
for research using animals, or $21.7 million. The $21.7 million
was then adjusted for the risk of failure, given an "expected"
cost of clinical tests of $48.1 million per approved drug.
The authors then made some heroic assumptions about
the costs of preclinical research, for which they had no project
level data, and assigned $65.5 million in "expected"
outlays for preclinical research. Both of these numbers were then
increased to reflect the opportunity costs of capital, to obtain
the $231 million and later $359 million figures so widely used
How reasonable are these numbers, given what is known
today? There are several areas where skeptics have questioned
the PhRMA/Tufts/OTA numbers. First, critics say it is troublesome
to apply so much of the industry data, since industry trade associations
have incentives to exaggerate costs of all aspects of R&D.
Secondly, the assumptions regarding preclinical expenditures are
not supported by any project level data. Thirdly, critics say
that much of the costs of preclinical and clinical research is
paid for by taxpayers.
In an effort to get an handle on this issue, we looked
at two sources of public data on drug development costs. The US
National Institutes of Health provided my offices with data from
58 clinical trials funded in the 1996 and 1997 financial year
budget. We also obtained from the US Treasury Department 11 years
of data on the now-defunct Orphan Drug Tax Credit. Data from these
sources suggest direct industry outlays on R&D for many drugs
may be far less than has been imagined.
Orphan Drug Tax Credit
From 1983 until 1994, the US government offered a tax credit to support the development of Orphan Drugs. The tax credit was for 50% of the direct expenditures on human-use clinical trials. From 1983 to 1993, some 93 orphan drugs were approved for marketing. During that same period, companies receive $106.9 million in tax credits. In order to obtain these credits, companies reported direct expenditure on clinical trials of $213.8 million, or $2.3 million per approved drug.
One expects some lag time between the beginning of the tax credit and the drug approvals, so we looked at the results beginning in 1989, the seventh year of the tax credit program, until 1993. During this five-year period, 60 orphan drugs were approved, while tax credits of $86.6 million were taken, or about $2.9 million per approved drug. This amount was fairly consistent from year to year. In 1995 dollars, the amount expended on human-use clinical trials was $3.2 million per approved drug, from 1989 to 1993.
What are we to make of these data (table below)?
DiMasi et al say that the average out-of-pocket cost of human-use
clinical trials is $18.9 million in 1987 dollars, or $24.5 million
in 1995 dollars. With these numbers of adjusted for failures,
they calculated the average "expected cost" of human-use
clinical trials (per approved drug) at $42.3 million in 1987 dollars,
or $54.8 million in 1995 dollars. In the study, the average was
significantly higher than the median, and for the median the expected
cost of human-use clinical trials would be $23.7 and $30.7 million.
But no matter how you slice it, it is hard to reconcile the DiMasi
data with the data from the Orphan Drug Tax Credit. The differences
are very large.
There are three possible explanations. First, it is possible that
not all companies conducting human-use clinical trials claimed
the Orphan Drug Tax Credit on tax returns, for example if the
company was not profitable. However, this cannot explain the entire
disparity. In 1987, the OTA said some six firms with assets greater
than $250 million claimed $4.7 million in orphan drug tax credits.
This implied expenditures of $9.4 million, or $1.6 million per
firm. In 1987, nine orphan drugs were approved for marketing,
and over the next five years an average of 12 drugs were approved
for marketing every year. The amount of reported expenditures,
even from these six firms, seems low.
Another explanation is that the government, rather than the companies,
actually paid for the orphan drug human-use clinical trials. Indeed,
it would seem as though this must be the case, since the reported
expenditures under the Orphan Drug tax credit are less than 8%
of the expenditures predicted by the PhRMA/Tufts/OTA data. Taken
by itself, this would suggest that 92% of the costs of human-use
clinical trials for orphan drugs is paid for by the taxpayers,
rather than the drug firms.
The third scenario is that the PhRMA/Tufts/OTA numbers are too high. Based on clinical trial data from the NIH, we calculated the expected costs, using the same "hazard rates" for success that were used in the DiMasi study. Thus, for example, a dollar spent on Phase I trials is multiplied by a factor of 4.35 to reflect the risks of failure. The average and median out-of-pocket costs for combined Phase I, II and III trials were $7 and $4.6 million for the NIH-funded trials, and using 1995 dollars, the PhRMA/Tufts/OTA figures were $24.5 and $14.3 million, three times as high.
|No of Observations||NIH||P/T/O (95 $)|
|Average cost of trials||6,986||24,467|
|Average expected cost||16,106||54,816|
|Median expected cost||11,490||30,675|
*[Note, due to my own editing mistake, the original published version of this table had reversed the figures for the NIH's average expected cost and the median cost. The numbers in the text and in the revised table above are correct.]
One partial explanation for the differences between the PhRMA/Tufts/OTA data and the NIH budget numbers could be the treatment of overheads. The NIH numbers appear to reflect direct expenditures on clinical trials, while the PhRMA/Tufts/OTA may include generous overhead allowances. In a small number of cases, the NIH trials involve cooperative R&D agreements, which may involve some industry cost sharing, such as with Taxol (paclitaxel), where Bristol-Myers Squibb provided the NIH with 17 kilos of Taxol for use in NIH-sponsored trials. This is an area for further research. It is also possible that the PhRMA/Tufts/OTA data was biased, given the incentives to the industry to overestimate private-sector R&D costs.
Our initial view is that the huge disparities between the Orphan
Drug Tax Credit data and PhRMA/Tufts/OTA estimates of drug development
costs are explained by a combination of all three factors -- some
unclaimed tax credits, a large role by the government in the development
of orphan drugs, and overstating of costs by the industry in the
PhRMA/Tufts/OTA study. The differences also point to the need
for disclosure to the public of more reliable data from the industry,
so policy makers can better evaluate industry R&D costs.
Director of Economic Studies
Center for Study of Responsive Law
P.O. Box 19367, Washington, DC 20036
Voice 202.387-8030; Fax 202.234-5176
CPT's Home Page
Return to: CPT's Pharmaceutical Page