According to DiMasi, we’re now looking at $2.6 billion to develop a new drug.
Now DiMasi has been performing this analysis and refreshing the data for decades. The arguments against “the number” are well-known:
1. “The Number” is a combination of out-of-pocket costs and opportunity costs.
2. The calculation is based on a limited survey of development programs from pharma (although we can never be sure who was surveyed and what the raw data look like, for obvious competitive reasons). In fact, a lack of detail on the raw data makes the entire methodology difficult to place into context.
3. The number fails to take into account government spending (i.e., grant) which can be a critical component of any new drug development program.
Here are a few items we noticed from the summary presentation (we cannot reproduce the slides here due to copyright, but you can find them here):
1. R&D Expenditures, adjusted for inflation, (page 6) increased rapidly from 1983-2006, then flattened. Meanwhile, approvals have remained in the 15-30 p.a. range (with a spike in the late 1990s).
2. The dataset includes drugs which entered FIM trials from 1995-2007. We would argue that including data from that far back could skew the end result. However, we have no idea which indications are included, the number of patients per trial, cost per subject, etc., etc. Earlier versions of this analysis had data by therapeutic category (see, for example, DiMasi, et al., 1995).
The 1995 paper (email us if you want it) has a great deal of detail. For example, mean Phase III costs back then for NSAID development were $35.9 million, versus $18 million for cardiovascular. Would the same proportions carry over with the current data set? Probably not.
3. Post-Approval costs (page 22) are enormous; $466 million out of pocket per approved new compound. But again, who knows what this number really means?
Overall, DiMasi paints a picture where drug development costs (especially clinical costs) increased quite a bit during this study period, while drug approvals actually declined. This we already knew.
However, with the lack of clarity around the data set, it is quite difficult to draw substantive conclusions beyond this. Perhaps a future publication will shed more light on what these numbers are and what they really mean.