Ways to Improve Attrition in Phase II and Phase III

De-Risking, Dispensing of Aggregate Risk.  At the beginning stages of Discovery, there is scant evidence that a proposed therapy might work. Those blockbusters that command billions of dollars in sales today were in with the pack of wild ideas at the beginning of Drug Discovery. Risk that the proposed therapy might come to fruition is so pervasive in Drug Discovery, it is often not discussed as such – it’s seen as “work”. At the beginning of Discovery the risk of failure in the clinic is part of the aggregate risk that the new project faces, even though work to reduce the clinical risk won’t occur until a candidate is identified. Every experiment or trial performed along the way addresses a single component of the aggregate project risk.  In Development (esp. late stage), there is a mounting body of evidence that the proposed therapy might succeed. Risk is less obvious, so it needs special formal attention. But it can be tightly managed.

A potential outcome of any project risk is project failure.  But project success or failure is not proportional to project risk. Some risk never comes to fruition. Low risk projects can fail or be terminated for reasons not related to risk.  Nonetheless, these days one speaks of de-risking a project, 1 dispersing risk presumably through experiments and investigation that, one-at-a-time, addresses specific risks.  Risk may also be dispersed by sharing cost with partners.  Let us pretend that we could sum up all of the risks that a project faces at Target Discovery, which is all of the risk any project may face going through all the stages of discovery, 100% of aggregate project risk. Then at the end when the project has succeeded in Registration one would say that all of the project risk has been removed, or that the project has been completely de-risked.  Figure 1 shows a hypothetical curve showing the amount of de-risking that may have occurred at each stage of Discovery and Development.  Certainly by Phase II and III a considerable amount of the aggregate project risk has been removed, which is fortunate, as the cost at those stages are considerably higher than any other stage. Still, as we have discussed, project success is also the lowest in Phase II – the risk of failure in Phase II is unacceptably high.  A worthwhile goal would be to perform experiments in earlier stages of Discovery and Development that would improve the probability of success in Stage II and thereby influence the POS in Stage III.  Essentially the “de-risking” curve would be shifted to the left – more risk would be dealt with in earlier stages.

Figure 1, De-Risking Through the Stages of Discovery and Development. A hypothetical curve (in red) depicting the removal of aggregate project risk through the stages of Discovery and Development. Project Cost per stage (from Table 1, Attrition) is displayed (blue bars). Performing experiments in earlier stages that would improve POS in Phase II would essentially shift the hypothetical de-risking curve to the left, shown in green.

Work that can be Done in Earlier Stages. The 46 % of attrition in Development that arises from poor pharmacokinetics (PK), toxicity and formulation (Figure 1, Attrition – Reasons for Failure) are issues that arise in Development but can be prevented in Discovery by improving PK, toxicity and formulation protocols and by preventing drug candidates that show marginal PK, tox and formulation from going forward into Phase I.  Phase I tends to involve healthy male volunteers and begins with low doses to be sure that any toxic reactions are limited.  It is in Phase I that the true NOAEL is determined and the therapeutic index is determined, as the ratio of the dose that produces a toxic effect in 50% of the population (TD50) with the minimum effective dose for 50% of the population (ED50).  While drugs with a 2:1 ratio are known (e.g. digoxin) a considerably wider window of 10:1 or more is typically sought.  It is feasible that the companies knew that there might be adverse effects in humans but were willing to face the risk of failure.

Another 30% of attrition in Development is due to lack of efficacy (Figure 1, Attrition – Reasons for Failure), which would show up in Phase II and III.  Theoretically this might also be preventable in Discovery if animal efficacy models were more predictive of human efficacy.  It may also be due to the dose being limited by concern for toxicity, putting the dose out of the range that would demonstrate efficacy.  Presumably the companies that paid for these clinical trials knew they were taking a chance and were willing to face the risk of failure. 

Two other factors that were not issues in 1991 but arose by 2000 were Cost of goods (COG) and Commercial (Figure 1, Attrition – Reasons for Failure).  COG rose to about 8% in 2000.  While estimates are made even prior to Candidate Selection, it is not until the drug substance is scaled up for late stage clinical trials that true estimates of COG can be made.  Similarly, while market estimates for the drug are made prior to Candidate Selection, it is not until late stage trials that the registerable indication becomes evident.  So the causes of about 30% of attrition may not be fully understood until late stage trials.  

A new stage of R&D has appeared in recent years between Registration and Launch (Figure 1, Attrition) which many call the Reimbursement Stage. Getting approval for Reimbursement from Third Party Providers is far from a slam dunk these days.  In future surveys of Attrition, Reimbursement is likely to get considerable attention.  Given the importance of Reimbursement, ways to reduce attrition in this stage, through work in earlier stages, will likely receive keen attention. 2

Still a full 70% of attrition might be prevented if the companies decided they weren’t going to take the risk with very expensive late stage trials and not advance compounds into the clinic that may not make it.  There tends to be the desire to point to a “robust” pipeline by senior managers, which tends to encourage filling pipeline gaps with drug candidates that are going to fail.  In the event that there are no projects with less risk available to fill the pipeline, this decision alone could potentially become its own cause of attrition! In that event the category “Unknown/Other” might swell to 70%!!  It should be noted also that as long as there is attrition, there will always be reasons for attrition.

The reader may wish to continue this discussion in the next section Living with Attrition – Ways to Compensate for Attrition


  1. Young, M., “Prediction vs Attrition”, Drug Discovery World, p. 9-12, Fall 2008.
  2. “Payers growing influence on R&D decision making”, Bethan Hughes, Nat. Rev. Drug Disc.,  7, p2-3 (2008)