Pharmaceutical companies that license externally generated drug candidates enjoy far greater R&D productivity than those that rely solely on internally generated ones, a recent study has found. Licensed compounds cost an average of $5 million to $9 million less to acquire at the preclinical stage than internal candidates cost to develop (Exhibit 1), are twice as successful in clinical trials, and achieve similar commercial results. So if pharma companies applied to their own programs the same rigor and best practices they use when licensing drugs from other companies, they could significantly increase the productivity of their research.

To understand the relationship between research productivity and licensing, we compared the cost, clinical-success rates, and commercial value of internally generated and licensed drugs. Because information on the amount of money pharmaceutical companies spend on discovery at the project level isn't publicly available, we used the current research pipelines of pharmaceutical companies for our sample and applied standard attrition rates to estimate the number of all preclinical projects from 1995 to 2001. By dividing this number into the cumulative amount spent on discovery during that period—estimated at 25 to 35 percent of total R&D spending—we calculated the cost of each project. The result of this calculation, an average of $21 million to $29 million, varies markedly by therapeutic area (from $18 million to $41 million) but is consistent with industry estimates and has been confirmed by discovery managers.

Compare that sum with our estimate of the average present-value cost—$14 million to $19 million—of licensed preclinical candidates. (These figures are based on 77 such deals from 1995 to 2001 and adjusted for standard attrition.) Interviews with licensing managers confirmed the accuracy of this range for preclinical deals and suggested a few reasons for the disparity at big companies: larger, more specialized discovery teams that cost more per compound to operate; heavier investment in discovery technology, adding to fixed costs; and much higher overhead rates.¹

We then compared the relative success of licensed and internally generated drugs by studying clinical-trial attrition rates for 1,448 compounds from 1991 to 2000. Licensed compounds were successful in 27 percent of all cases (Phases I through III), internal candidates in only 14 percent—results consistent with the findings of other researchers.² We attribute this gap to the way projects are evaluated: external licensing deals are subjected to considerable scrutiny, while internal projects can move forward on the strength of inertia or politics. Many companies unintentionally lower the bar when they evaluate their own compounds. Not surprisingly, internal compounds go on to fail more often in Phase II trials (Exhibit 2). As the head of oncology at one company put it, "We frequently get projects from our discovery group that we would never in-license."

Finally, we evaluated the revenues generated by 71 internal and 73 licensed compounds³ that the top 15 pharmaceutical companies launched from 1997 to 2001. We found virtually no difference in average cumulative revenues during the first four years after a product's launch (a proxy for its total value): $914 million and $975 million for internal and external compounds, respectively (Exhibit 3). The net result is that internal candidates cost more, are less likely to survive clinical trials, and achieve no more commercial success than licensed ones do. No wonder the price of licensed candidates continues to rise, as a number of recent studies have demonstrated.⁴

Our research has a number of implications for R&D managers seeking to keep their research pipelines competitive and to bring equilibrium to the process of evaluating internal and external opportunities. Although some small companies are moving toward a virtual R&D model, with licensed products only, this option may not be practical for most of the established companies. The market for compounds remains relatively illiquid, and companies developing drugs don't share all relevant information with companies seeking to license them.

What is to be done? Currently, different managers handle internal and external projects. The disparities this approach creates could be overcome by developing a gatekeeper operation to assess both on an "apples-with-apples" basis. In addition, such a group could evaluate overall spending on the discovery and early-development portfolio to ensure that the balance between internal and licensed projects was appropriate.

A central group could also use the licensing process to develop more robust market intelligence that would guide the long-term research agenda. Most large pharmaceutical corporations have a poor knowledge of what the roughly 1,600 companies conducting early-stage preclinical research are doing. By connecting the information gathered during the licensing process with conversations about internal research priorities—and by using flexible terms, deals that include a number of compounds over several years, or minority equity positions to learn even more—pharmaceutical managers can gain valuable market insight into how they ought to shape their research portfolios.