I think the challenge with @chainsaw's approach is that it doesn't reflect the complexity of finance well enough.
A better way would be to calculate the net present value and split the research costs across 10 years, add a discount rate (say 7%) and projected cashflow rate from the sale in years 10-30 (assuming patent protected and monopoly) and then a generic alternative for years 30-40. If you only generated $1 billion for years 10-30, and .5 billion for years 30-40, you end up with a total revenue of $25 billion on an $11 billion investment. However, the NPV is negative, meaning you will be in the hole about $2.5 billion. The IRR for this project is ~5%. So if the investors only required 5% return, then this is a coin flip of a project.
Now this excludes the foreign markets were price controls and approvals will vary. In that case, those 4 million US souls will pay $26 billion over the course of 30 years of being on the market to company X, which works out to about $6500 (very quick and dirty estimate) per treatment. Assuming no new people get sick. No inflation is assumed as well.
This is also overly simplistic, with numbers, strategy, and distribution of costs all pulled from a hat.