The third and final stage of clinical testing is designed to confirm the efficacy and safety of new drugs under development. Yet McKinsey research finds that, surprisingly, as many as four drugs in ten fail during this phase. To reduce the high number of late failures, pharmaceutical companies should reexamine their approach to the development of drugs and begin differentiating those in their pipelines by risk.

Failures during the last phase of clinical trials exact a heavy toll on drugmakers, since Phase III is both more expensive and more closely scrutinized by investors than the earlier stages of testing. To learn why drugs fail at this stage and the extent to which such failures are under a company’s control, we screened 656 Phase III trials that took place from 1990 to 2002. Our research concentrated on orally ingested, small-molecule drugs (excluding biologics, such as vaccines) produced by large pharma companies. Of these 656 drugs, fully 42 percent failed their trials.¹

For 73 of the failed drugs, enough public information was available to analyze the cause of the failure, so we focused on them. This sample spanned a range of treatment areas, including the central nervous system, infectious disease, endocrinology, oncology, and cardiovascular and respiratory problems. Our subsequent analysis considered the efficacy and safety of the drugs (as compared with placebos) and evaluated them against comparable drugs already on the market.

The most stunning discovery was that 50 percent of the drugs failed in Phase III because they could not be proved effective (Exhibit 1): demonstrating efficacy is, after all, a primary objective of Phase II trials, yet inefficacy was still a major cause of failure in Phase III across all areas of treatment (Exhibit 2). Among the remaining drugs that failed, 31 percent posed safety concerns, while an additional 19 percent were found to be neither safer nor more effective than drugs currently on the market.

Two factors emerged as prime suspects in the efficacy failures. The first was whether the drug had a novel mechanism of action—in other words, did the drug produce its pharmacological effect by using the same biological pathway that other drugs on the market used? The second related to the definition of the endpoint— the way researchers measured or defined clinical results. Drugs that used novel mechanisms of action failed more than twice as often as drugs using established ones. Likewise, drugs with less objective endpoints (say, because of patient-reported results) failed approximately 10 percent more often than those with more objective endpoints (such as glucose levels for diabetes). Drugs combining the two higher-risk factors—novel mechanisms and less objective endpoints—failed 70 percent of the time, compared with 25 percent for drugs with known mechanisms and more objective endpoints.

To improve on such discouraging results, pharmaceutical companies should rethink their approach to the development of drugs and segment those in their pipelines by risk. Drugmakers could, for example, modify their existing stage-gate processes so that more risky compounds (such as those with novel mechanisms of action) receive greater scrutiny in Phase II than compounds with less risky profiles. For the latter, companies should seek ways to shorten the six- to nine-month delay that often occurs between Phases II and III; they might, for instance, include scientists with operations experience on the research teams. The additional costs of such approaches should be offset by the extra revenues gained from reaching pharmacy shelves faster.

Clearly, pharmaceutical executives must also apply more rigor to their decision-making processes and steel themselves to make tougher decisions during Phase II. The results of the study, confirmed by our experience, suggest that drugmakers let too many compounds proceed to Phase III despite negative clinical and statistical evidence. One problem is wishful thinking: teams may work on a given compound for years and therefore run the risk of losing their objectivity. Senior executives can feel pressured by investor expectations and therefore allow projects to proceed despite the risks.

Poorly designed incentives also come into play: some drugmakers reward senior managers for the number of products in the pipeline at one stage or another and not necessarily for making good decisions about them. This should change. Similarly, incentives for scientists should be tailored around faster development times for lower-risk compounds and the need to develop medically unique products.

Adopting a risk-based approach to segmentation could even help drugmakers with licensing. We find that they often use figures on Phase III attrition rates to help them evaluate licensing deals. Instead, they could structure the terms of deals around the risk profile of individual compounds—for instance, by demanding higher payments up front for lower-risk drugs and more substantial royalties down the line for drugs that appear to be more risky.