This was a paper for an April 2, 2000 Geneva meeting of the MSF
Working Group on R&D.
How Much does it cost to develop a new drug
James Love
Consumer Project on Technology
http://www.cptech.org
April 2, 2000
Draft - Please do not quote without permission from author
How much does it cost to develop at new drug? In 1988, the
National Cancer Institution estimated it spent about $1
million to fund develop of a drug through Phase II trials, and
another $1.6 million to $4 million for Phase III trials, for
cost of $2.6 to $5 million. From 1983 to 1993, the
pharmaceutical industry reported spending $2.3 million on
clinical trials for each new FDA approved orphan drug. In
1991, NCI said it was spending more than $30 million on
clinical trials for the cancer drug Taxol. In 1999, PhRMA
testified that the average cost of drug development was $.5
billion, and that only 1 in 5,000 compounds are successful.
Bristol-Myers Squibb now claims that it has spent more than $1
billion to develop Taxol, a drug it did not invent. While the
costs of drug development are less of a mystery than one might
think, clearly there is considerable confusion about the costs
of research and development for a new drug.
These are some of the sources of this confusion.
1. Allowances for Risk and the Opportunity Cost of Capital
Some estimates are based upon the direct costs of drug
development, without accounting for the risk of failures, and
without adjustments for inflation or the cost of capital.
Others include both these items. For example, the US
government's investments in Taxol development were reported as
the nominal costs of research on Taxol, and did not reflect
the costs of government research on other products that were
unsuccessful. Nor did the NCI include adjustments for
inflation or the opportunity costs of capital. Studies
promoted by PhRMA include adjustments for risk and for the
opportunity cost of capital -- the latter using very high
capital costs.
2. Definitions of "development" vary.
One important question is what does it cost to discover and
test a drug, through the point where the drug is approved for
marketing. This is how I would normally define the costs of
drug development. But the pharmaceutical companies often use
a much different definition. For example, when asked how much
of its own money it spent to develop Ceredase, a drug invented
on NIH grants, Genzyme Corporation included the costs of
building a factory for commercial production. When BMS was
asked in 1993 how much it spent on Taxol development, it said
more than $114 million dollars. BMS did not invent Taxol,
did not sponsor the clinical trials used for the FDA approval,
and did not even know how to manufacture the drug. BMS
included in its "development" costs the projected costs of a
long term supply contract for production of Taxol from Hauser
Chemicals (a former NCI contractor), and a long term agreement
with a firm to grow Yew trees for Paclitaxel production. In
other cases, companies may include such items as the costs of
marketing a new drug, an item certainly part of a business
plan, but not a research expense. There is also money spent
to find new uses for a drug, once it is on the market, and
investments in clinical trials that have marketing objectives,
such as to justify inclusion of the drug in national or
private formularies.
3. Companies take credit for things they do not do.
While in some cases, a company may do all of the important
stages of research and development for a drug, including both
clinical and pre-clinical investments, but in other cases
governments or private donors fund important parts of
research. Studies have shown that the government role is
particularly true for innovative drugs for severe illnesses,
and less significant for so called "me too" drugs, or for
drugs such less important public health problems, as hair
loss.
When the government funds the discovery of a drug, such as in
the case of AIDS drugs like AZT, ddI, ddC, d4T, Ziagen and
Norvir, the taxpayers have paid for the most expensive part of
the R&D process. In the often quoted 1991 study by Joseph
DiMasi and his colleagues, the costs of preclinical
expenditures were 67 to 73 percent of the total development
costs, depending upon the assumptions regarding the
opportunity costs of capital. So when the government is
responsible for the pre-clinical discovery, 2/3 to 3/4 of the
costs of the drug are already paid for, according to the
DiMasi analysis.
This is also true for each stage of the clinical process. For
example, d4T was discovered on an NIH grant, and its use for
HIV/AIDS was discovered at Yale, on a government grant. Then
the NIH was a sponsor of the first Phase I trial on d4T. When
companies enter the R&D pipeline at the Phase II or Phase III
stage, the are entering at the tail end of a process, avoiding
investments, risk and time -- the three critical elements in a
cost study.
It is also the case that it is often less expensive to develop
the second or third drug within a therapeutic class, than the
first, because the risks are less. Once AZT was identified as
a product that would treat HIV/AIDS, there was a rush to test
other nucleoside analogue reverse transcriptase inhibitors.
Similarly, after the first protease inhibitors or
non-nucleoside reverse transcriptase inhibitors where shown to
be effective for HIV/AIDS, the risks were lowered for the next
versions of drugs in these therapeutic classes.
4. Samples can be skewed and average can be misleading
In his 1991 study of the costs of clinical trials, DiMasi
indicated that the median costs of clinical trials were about
60 percent of the average costs. DiMasi also reported that
the costs of successful trials were considerably higher than
the costs of unsuccessful trials, on average, suggesting
companies invest more when risks are lower.
In my examination of the US Orphan Drug Tax Credit, the
reported industry outlays on clinical trials were only $2.3
million per approved drug, a number adjusted for risk but not
opportunity costs. This was a small fraction (after
adjustments for inflation, about 6 percent) of the "averages"
used in the DiMasi study. There were probably several reasons
for the differences, including the fact that the US government
had paid for all or part of the costs of clinical trials for
many of the orphan drugs. But PhRMA claimed the difference
was because orphan drugs were much cheaper, because of the
smaller sizes of clinical trials and the fast track regulatory
procedure. The category of orphan drugs is larger than one
might think, and several "blockbuster" drugs have qualified
for orphan status. Indeed, during the period of 1983 to 1993,
all HIV/AIDS drugs qualified as orphan drugs. And in 1998,
about half of all US FDA approvals for new molecular entities
(NME) were classified as orphans.
One can also look at other assumptions in the DiMasi study and
see issues regarding averages. DiMasi 1991 study used the
following periods for the period between the beginning of a
stage of development and the approval of a drug:
Table 1
DiMasi's 1991 estimate of Time to Market
Start to Phase
NDA Length
Preclinical 11.8 3.6
Phase I 8.2 1.3
Phase II 6.9 2
Phase III 5 3
NDA Review 2.5 2.5
The data on time to market are quite important when one is
calculating the opportunity cost of capital, which is in fact,
the lion's share of costs in studies by DiMasi, the OTA and
others.
When we looked at development times for HIV/AIDS drugs, the
DiMasi/OTA assumptions did not seem realistic. For all 14
HIV/AIDS drugs approved by the US FDA (In the class of nucleoside
analogue reverse transcriptase inhibitors, protease inhibitors,
and non-nucleoside reverse transcriptase inhibitors), the average
time from filing a patent to NDA approval was only 4.4 years.
(Table 14) The longest period between filing for a patent and FDA
approval was 8 years, for Ziagen, a drug discovered by the
University of Minnesota, now sold by Glaxo. The period was much
shorter for many drugs, and in particular, the first drugs in a
therapeutic class. (See Table 14)
For AZT, the first nucleoside analogue reverse transcriptase
inhibitor, the time between filing the patent and product
approval was 1.5 years. For protease inhibitors Ritonavir/Norvir
and Indinavir/Crixivan the period was .9 and 2.9 years,
respectively. For the three non-nucleoside reverse transcriptase
inhibitors, the period of 2.9 to 3.3 years.
When asked about AIDS drugs, DiMasi said his own data suggested
AIDS drugs were about the same as other drugs in terms of
development time. But his averages were based upon data points
that could be misleading. For AZT, a drug that was on the market
1.5 years after the patent was filed, DiMasi used the earlier
1964 date of discovery of the compound, for the beginning of
preclinical research, giving AZT more than 20 years. The same
was done for ddI and other drugs, even though the initial
discovery of the compounds were done by government funded
researchers.
How important are data on the timing of development? When an
estimate of drug development costs includes the opportunity cost
of capital, the investments are increased at a compound rate of
interest. At 14 percent, the cost of capital for preclinical
research in the most frequently quoted estimate from the 1993 OTA
report, the cost basis increases by 48 percent in 3 years,
doubles in 5.3 years and triples in 8.4 years. In DiMasi's 1991
paper, he presents one set of estimates using a 15 percent
discount rate, and 11.8 years to market. In 1998 dollars, the
total cost of development was $502 million, of which $347 million
were capital costs.
Table 2
Years Cost basis will double
under different opportunity cost of
capital assumptions
Cost of
Capital 5% 9% 14%
Years 14.2 8.0 5.3
By shaving years off the estimated time for development of the 14
existing HIV/AIDS drugs, the estimated costs of development fall
sharply. When one considers the fact that in many cases the
company was not even the source of the pre-clinical research, the
bottom follows out of the cost basis.
Another area where there are large differences in costs concerns
the size and duration of clinical trials. The 1993 OTA report on
the costs of drug development presented this data on the size of
clinical trials. The data were used to indicate the size and
upward trend of clinical trials.
Table 3
OTA estimates of Mean Enrollment in Clinical Trials Prior to DNA
1978 to 1983, 1986 to 1990
1978-83 1986-90
Antihypertension drugs 1,791 2,485
Antimicrobial 1,885 3,461
Nonsterodidal antiflammatory 3,036 3,575
Source: Chapter 3, page 65, 1993 OTA report, Table 3-8: Mean
Enrollment in Clinical Trials Prior to New Drug Application,
1978-83, 1986-90. 1993, the Office of Technology Assessment,
Pharmaceutical R&D;: Costs, Risks, and Rewards, OTA-H-522, GPO
stock #052-003-01315-1, NTIS order #PB93-163376.
We are examining the data from NDA applications HIV/AIDS drugs,
to determine the size of the trials. There are the preliminary
data from two therapeutic classes of HIV/AIDS drugs.
Table 4
Number of Patents in Clinical trials
discussed in FDA NDA approval
(data complete for two classes of HIV/AIDS drugs)
Protease Inhibitors
Saquinavir 1,265
Ritonavir 1,583
Indinavir 1,262
Nelfinavir 605
Amprenavir 736
Average: 1,109
Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTI)
Nevirapine 549
Delavirdine 2,452
Efavirenz 928
Average: 1,310
Note that all of these trials are smaller than the 1986 to 1990
averages of the three drug categories examined by OTA. It was
also the case that the US government sponsored some of these
trials. For example, for all three of the NNRTI drugs, the US
government sponsored or cosponsored trials used in application
for an NDA.
Table 10
DiMasi, et al's 1991 Estimates of the Costs of Drug Development
In 1998 dollars
with with with
Capital Capital Capital
out of pocket Costs Costs Costs
per approval @ 5% @ 9% @ 15%
------------ ----------- --------- ----------
Preclinical 89 57.9% 145 63.3% 212 67.5% 365 72.7%
Clinical 65 42.1% 84 36.7% 102 32.5% 137 27.3%
Total 155 100.0% 229 100.0% 313 100.0% 502 100.0%
Table 11
DiMasi et al 1991 estimates: out of pocket and capital costs
compared
1998 dollars, capital costs at 9 percent
out of Capital out of Capital
pocket Costs pocket Costs
p/app @9% total p/app @9% total
--------------------- --------------------
Preclinical 89 122 212 28% 39% 67%
Clinical 65 37 102 21% 12% 32%
Total 155 159 313 49% 51% 100%
Table 12
DiMasi et al 1991 estimates: out of pocket and capital costs
compared
1998 dollars, 15 percent capital costs
out of Capital out of Capital
pocket Costs pocket Costs
p/app @15% total p/app @15% total
----------------------- --------------------
Preclinical 89 276 365 18% 55% 73%
Clinical 65 72 137 13% 14% 27%
Total 155 347 502 31% 69% 100%
Table 13
US R&D spending by Pharmaceutical Companies and US DHHS
(billions of US dollars)
PhRMA IRS* DHHS**
Survey Returns R&D
1991 $ 7.9 $4.4 $ 9.8
1992 $ 9.3 $5.1 $ 9.1
1993 $10.5 $5.9 $10.5
1994 $11.1 $6.6 $10.4
*Form 6765 qualifying expenditures
** NSF's Survey of Federal Funds for Research and Development
Table 14
14 HIV/AIDS drugs
sponsorship of clinical trials, CRISP grants
and time to Market
(Survey November 1999)
gov/non-gov patent
trials % CRISP app
in ACTIS gov grants to NDA
Nucleoside Analogue Reverse Transcriptase Inhibitors (RT)
Zidovudine (AZT)/Retrovir 314/189 62% 1,462 1.5 yrs
Didanosine (ddI)/Videx 131/59 69% 66 4.2
Zalcitabine (ddC)/Hivid 43/26 62% 6 4.9
Stavudine (d4T)/Zerit 53/56 49% 158 7.5
Lamivudine (3TC)/Epivir 65/54 55% 191 6.8
Abacavir/Ziagen 19/29 40% 2 8.0
Protease Inhibitors
Saquinavir/Invirase 25/32 44% 29 5.0
Ritonavir/Norvir 30/28 52% 43 0.9
Indinavir/Crixivan 53/54 50% 91 2.9
Nelfinavir/Viracept 40/41 49% 25 5.7
Amprenavir/Agenerase 14/11 56% 1 5.4
Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTI)
Nevirapine/Viramune 42/23 65% 151 2.9
Delavirdine/Rescriptor 20/20 50% 53 3.1
Efavirenz/Sustiva 22/17 56% 1 3.3
AVG: 55% 4.4 yr
Table 15
U.S. Orphan Drug Credit and
Pre-Tax Private Sector Expenditures on Clinical Trials
(thousands of US dollars)
Pre-tax
Year Credit Approvals Expenditure $/drug
1983 236 2 472 236
1984 105 3 210 70
1985 204 6 408 68
1986 6,530 5 13,060 2,612
1987 5,154 9 10,308 1,145
1988 8,053 8 16,106 2,013
1989 14,190 10 28,380 2,838
1990 15,637 12 31,274 2,606
1991 18,475 12 36,950 3,079
1992 17,826 13 35,652 2,742
1993 20,486 13 40,972 3,152
83-93 106,896 93 213,792 2,299
Source: US IRS