The US$ 1 billion drug. A fairytale.
It takes US$ 802 Mio to take
one drug successfully to market. This was Joseph DiMasi’s central message in
his article in 2003 about R&D costs. This number has been cited many times
and has been abused in many ways. US$ 802 Mio should discourage any rational
investor from investing into drug development. In 2007 DiMasi rerun his study
with new data, especially with a new set of success rates. Overall success
rates increased over the last years and moved from 21% to 30% from IND to
approval, i.e. a drug entering clinical development has a 30% chance to reach
market. But to our big surprise the new number on drug development costs even
increased to an astonishing US$ 1,241 Mio per drug. We expected that an
increase in the success rate translates into lower costs to bring at least one
drug to market. To better understand what is behind DiMasi’s figures we will
deconstruct his misleading number in detail.
DiMasi’s number
does not refer to how much a company has possibly to spend to take its drug to
the market. Neither does he say how much has been spent on the research and
development for drugs that are on the market. No, DiMasi’s numbers are the
average expenses it takes to get one drug onto the market, taking into account all
possible failures too. So, with a 21% success rate it takes already about 5
clinical phase I projects in order to reach the market with one compound. And
in order to inflate the number a little more, DiMasi doesn’t use a
forwardlooking perspective discounting future expenses, but capitalises all
past expenses at the cost of capital of the companies.
For each clinical
phase we have a mean estimate of costs, success rates, and duration.
Furthermore, DiMasi claims that 65% (in 2007, in 2003 it was 69%) of all
R&D expenses are spent for clinical trials. This means that per
investigational new drug (IND) one has to account for US$ 60 Mio (in 2003 only
US$ 26 Mio) for preclinical expenses. This means that a company has to invest
on average US$ 60 Mio for R&D to generate one IND, e.g. to bring one drug
into clinical development.
Table 1: DiMasi's
parameters for clinical phases.
2003 
Phase I 
Phase II 
Phase III 
NDA 
TOTAL 
costs (US$ Mio) 
15.2 
23.5 
86.3 
0.0 
125.0 
success rates 
71% 
44% 
69% 
100% 
21% 
duration (months) 
12.3 
26.0 
33.8 
18.2 
90.3 
2007 





costs (US$ Mio) 
32.3 
37.7 
96.1 
0.0 
166.1 
success rates 
84% 
56% 
64% 
100% 
30% 
duration (months) 
19.5 
29.3 
32.9 
16.0 
97.7 
Table 1 displays
the numbers for clinical development DiMasi has used. While the early
development costs were high for his 2003 calculation, they become exorbitant
for his 2007 calculation. This way the preferential increase of success rates
cannot offset the enormous raise of cost assumptions. According to DiMasi the cost
estimates are taken from a sample of the ten largest pharmaceutical companies.
Of course, this is in no way representative. Biotech companies manage to run
clinical trials at much lower expenses, throughout all phases.
The second point
that inflates DiMasi’s number is the fact that costs are not simply added up,
but weighted by the success rates. Table 2 indicates how many projects it takes
in each phase to reach market with one project on average.
Table 2: Attrition
weights.
2003 
Phase I 
Phase II 
Phase III 
Probability to reach market 
21% 
30% 
69% 
# it takes that one reaches market 
4.7 
3.3 
1.5 
2007 



Probability to reach market 
30% 
36% 
64% 
# it takes that one reaches market 
3.3 
2.8 
1.6 
The last important driver of DiMasi’s number is that he capitalizes all costs at the cost of capital (11% in 2003, 11.5% in 2007). Preclinical costs that lie far in the past are therefore compounded over 10 years at 11%, which means that we do not only account for them 4.6 times because of the attrition weights, but we increase the value by another factor 2.8 because of the capitalisation of costs. An originally relatively reasonable number of US$ 26 Mio then contributes US$ 335 Mio to DiMasi’s number. The final calculation by DiMasi is shown in figure 1.
Figure 1: DiMasi's figures
While the 2003
number requires an average drug reaching US$ 310 Mio peak sales to break even
(which is already far better than the vast majority of all drugs), the 2007
number requires peak sales of US$ 508 Mio. It means that one should abandon any
project that is estimated to reach sales at its peak of less than US$ 508 Mio
(using a standard sales curve). We can of course immediately see that most
biotechs develop drugs with much smaller peak sales. DiMasi’s figures are as
already said in no way usable for other than big pharma companies.
DiMasi’s numbers are correct in the way they are defined, but they have two serious flaws. First they are based on a sample of companies that follow the expensive blockbuster model, i.e. only big pharma is included. Therefore the cost estimates are far from being representative. Second, one wants to know how much he must spend, not how much he has spent. We therefore should rather discount the costs back to today than capitalize it from the past to today. Also, when talking about value, we should adjust all costs by their probability. In reality, no biotech will ever get funded by investors with a sum comparable to DiMasi’s number. Biotechs can and do bring drugs on the market for less than US$ 100 Mio. DiMasi’s number is only interesting for multinational pharmaceutical companies hunting for blockbusters. With this strategy currently failing and companies moving to targeted and personalized medicine, niche and orphan indications, the costs for a successful drug launch need to be much lower resulting from much higher success rates.
Table 3 displays
some more realistic assumptions about drug development. While we stick to
DiMasi’s 2003 numbers for success rates and duration, we have adapted the
clinical costs according to our experience.
Table 3: More realistic
assumptions

Phase I 
Phase II 
Phase III 
NDA 
TOTAL 
costs (US$ Mio) 
5 
12 
68 
3 
88 
success rates 
71% 
44% 
69% 
100% 
21% 
duration 
12.3 
26 
33.8 
18.2 
90.3 
Table 4: Indicative
numbers out of the three data sets

2003 
2007 
Avance 
Capitalized and attrition weighted[1] 
466 
629 
261 
Attrition weighted (out of pocket)[2] 
274 
364 
162 
Summed up[3] 
125 
166 
88 
Risk Adjusted (out of pocket)[4] 
59 
109 
36 
Discounted and risk adjusted[5] 
45 
79 
25 
Required peak sales[6] 
311 
509 
231 
Table 4 finally shows the various values we get out of these three data sets. We focus in this analysis only on the clinical costs. As we can see, the eyecatching US$ 629 clinical spendings of the 2007 data set can also be interpreted as US$ 79 Mio. We however think that this number can even be lowered to US$ 25 Mio, the risk adjusted present value of all investment to come for one specific drug. The pharmaceutical companies might be interested in advertising DiMasi’s high number, entrepreneurs and investors alike might be more interested in the realistic value of the costs for one project (the US$ 25 Mio). The required peak sales allow judging which data set makes most sense.
[1] DiMasi’s approach
[2] DiMasi’s approach but without
capitalisation of costs
[3] All costs summed up as they are,
without adjustment for risk nor time
[4] All costs adjusted by the
probability that they occur (i.e. multiplied by previous success rates)
[5] All costs adjusted for their risk
and discounted at the cost of capital (11%) back to IND
[6] Minimum peak sales to reach an IRR
higher than the cost of capital (preclinical costs included)