Most clinical trials include multiple endpoints to demonstrate a drug’s ability to impact one or more characteristics of the disease under study. Such endpoints may include assessments of clinical events (e.g., death, major adverse cardiovascular events, other severe adverse events), symptoms (e.g., pain, nausea, depression), functional effects (e.g., exercise capacity, or physical disease measures, such asVO2 max.), or surrogates of such events or symptoms. In fact, determination of a drug’s effectiveness and safety often cannot be made based on a single endpoint due to the varied ways that a disease may affect a specific patient.
For example, some diseases have more than one critically important symptom. Employing a single efficacy endpoint may not accurately determine whether a drug was effective. In some instances, a drug that fails to impact a combination of several measures could be prematurely judged ineffective. In others, positive results on one endpoint but not others could lead to the drug being judged as effective, when in fact it is not. As the number of endpoints in a trial increases, the likelihood of reaching flawed conclusions about the drug’s efficacy and safety also increases considerably if no statistical adjustment is made for this endpoint multiplicity.
In migraine therapeutics, for instance, while pain is the defining symptom, light sensitivity and nausea are also important manifestations of the disease. In December 2021, the Biohaven (now part of Pfizer) intranasal spray Zavegepant successfully hit two efficacy endpoints in a Phase 3 trial: (i) freedom from pain and (ii) freedom from the most bothersome symptom, at two hours post drug administration. The FDA is expected to approve the drug in early 2023. In contrast, in the case of Biogen’s and Eisai’s Aduhelm (aducanemab) for Alzheimer’s dementia, showing efficacy on multiple fronts proved challenging, with the drug hitting its endpoint on clearance of amyloid beta while being less compelling on endpoints relating to cognitive function.
In 2017, the U.S. FDA began to draft guidance for drug sponsors to address the problems posed by multiple endpoints in the analysis and interpretation of clinical trial results. In October 2022, the agency finalized this guidance document, providing strategies for grouping and ordering endpoints for the analysis of drug effects, clarifying when and how endpoint multiplicity needs to be managed, and adding an appendix to the document with information on specific statistical methods that sponsors might use to help control that risk of erroneous conclusions.
One of the most important topics in the finalized guidance is the clarification of the differences and relationships between primary, secondary and exploratory endpoints. The critical element for grouping endpoints in this way is whether they are intended to support the drug’s approval (primary endpoints) or to demonstrate additional meaningful effects on the patient’s condition (secondary endpoints). Exploratory endpoints are those used to provide insights for further research rather than support conclusions about a drug’s effectiveness or safety, and as such, do not need multiplicity adjustments. The finalized guidelines encourage drug sponsors to create a control plan and a hierarchy of endpoints to help mitigate the risks arising from multiplicity.
The FDA guidance document also discusses the use of different types of primary endpoints, including co-primary endpoints (used when a treatment effect on two or more features of a disease is critically important to determining drug efficacy), multiple primary endpoints (used when a treatment effect on any one of several measures is sufficient to support a conclusion of drug efficacy), and composite endpoints (used when important clinical outcomes are carefully chosen for combination into a single primary endpoint).
Several academic experts in clinical research and biostatistics have commented that while the finalized guidance helps to clarify some terminology and understanding of endpoint hierarchies, and provides resources on specific statistical methods for evaluating and adjusting for multiplicity, the guidance remains somewhat limited due to the need for FDA to be fairly general in its recommendations to accommodate variability in clinical trial design. Thus, the bottom line for drug sponsors is that they should continue to directly communicate with the FDA at an early stage in drug development to ensure that the chosen clinical trial design can potentially support drug approval.