news

In oncology research, compound outcome measures, such as progression-free survival (PFS) and disease-free survival (DFS), are increasingly replacing the traditional endpoints of overall survival (OS) and have become a key trial basis for drug approval by the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA). These measures improve clinical trial efficiency and reduce costs by combining multiple events (e.g., tumor growth, new disease, death, etc.) into one time-to-event endpoint, but they also create problems.

Changes in endpoints of antitumor clinical trials

In the 1970s, the FDA used an objective response rate (ORR) when approving cancer drugs. It was not until the 1980s that the Oncology Drugs Advisory Committee (ODAC) and the FDA recognized that improvements in survival, quality of life, physical function, and tumor-related symptoms were not consistent with ORR correlations. In oncology clinical trials, OS is a better clinical endpoint for measuring direct clinical benefit. Nevertheless, ORR remains a common alternative clinical endpoint when considering accelerated approval of cancer drugs. In single-arm trials in patients with refractory tumors, ORR is also specifically considered as the primary clinical endpoint.

Between 1990 and 1999, 30 percent of FDA-approved cancer drug trials used OS as the primary clinical endpoint. As targeted therapies have evolved, the primary clinical endpoints used to evaluate anti-cancer drugs have also changed. Between 2006 and 2011, that number dropped to 14.5 percent. As the number of clinical trials with OS as the primary endpoint has decreased, the use of composite endpoints such as PFS and DFS has become more frequent. Funding and time constraints are driving this shift, as OS requires longer trials and more patients than PFS and DFS. Between 2010 and 2020, 42% of randomized controlled trials (RCTS) in oncology have PFS as their primary endpoint. 67% of anti-tumor drugs approved by the FDA between 2008 and 2012 were based on alternative endpoints, 31% of which were based on PFS or DFS. The FDA now recognizes the clinical benefits of DFS and PFS and allows them to be used as primary endpoints in trials seeking regulatory approval. The FDA also announced that PFS and other alternative endpoints can be used to accelerate the approval of drugs for serious or life-threatening diseases.

Endpoints will evolve not only as new therapies are developed, but also as imaging and laboratory testing methods improve. This is evidenced by the replacement of the World Health Organization (WHO) criteria with the RECIST criteria for the Assessment of Efficacy in Solid Tumors (RECIST). As clinicians learn more about tumors, patients once considered stable may be found to have micrometastases in the future. In the future, some endpoints may no longer be applied, and new endpoints may emerge to safely accelerate drug approval. The rise of immunotherapy, for example, has led to the development of new evaluation guidelines such as irRECIST and iRECIST.

Composite end point overview

Composite endpoints are widely used in clinical studies, especially in oncology and cardiology. Composite endpoints improve statistical power by increasing the number of events, reducing the required sample size, follow-up time, and funding.
The most widely used composite endpoint in cardiology is major adverse cardiovascular events (MACE). In oncology, PFS and DFS are often used as proxies for overall survival (OS). PFS is defined as the time from randomization to disease progression or death. Solid tumor progression is usually defined according to RECIST 1.1 guidelines, including the presence of new lesions and the enlargement of target lesions. Event-free survival (EFS), DFS, and relapse-free survival (RFS) are also common composite endpoints. EFS is used in trials of neoadjuvant therapy, and DFS is used in clinical studies of adjuvant therapy.

Different effects in different therapies on compound endpoints

Reporting only compound outcomes can also lead to assuming that the treatment effect applies to each component event, which is not necessarily true. A key assumption in the use of composite endpoints is that the treatment will alter the components in a similar way. However, the effects of antitumor therapy on variables such as primary tumor growth, metastasis, and mortality sometimes go in the opposite direction. For example, a highly toxic drug may reduce tumor spread but increase mortality. This was the case in the BELLINI trial of patients with relapsed/refractory multiple myeloma, where PFS improved but OS was lower due to higher treatment-related infection rates.

In addition, there is preclinical data suggesting that using chemotherapy to shrink the primary tumor accelerates distant spread in some cases because chemotherapy selects stem cells that are more likely to trigger metastasis. The directionality hypothesis is unlikely to hold when there are a large number of events in the composite endpoint, as is the case with some definitions of PFS, EFS, and DFS. For example, allogeneic hematopoietic stem cell transplantation therapy trials often use a composite endpoint that includes death, cancer recurrence, and graft-versus-host disease (GVHD), known as GVHD free RFS (GRFS). Therapies that reduce the incidence of GVHD may increase the rate of cancer recurrence, and vice versa. In this case, GVHD and relapse rates must be analyzed separately to accurately measure the risk-benefit ratio of treatment.

Routine reporting of different event rates for complex outcomes ensures that the effects of treatment on each component are in the same direction; Any “qualitative heterogeneity” (i.e., differences in directionality) leads to ineffective use of composite endpoints.

The EMA recommends “individual analysis of individual event types using descriptive summary tables and, where appropriate, competitive risk analysis to explore the impact of treatment on each event”. However, due to the insufficient statistical power of many studies, significant differences in component events in composite outcomes could not be detected.

Lack of transparency in reporting composite endpoint events

In cardiology trials, it is common practice to provide the incidence of each component event (such as stroke, myocardial infarction, hospitalization, and death) along with the MACE composite endpoint. However, for PFS and other composite endpoints in oncology clinical trials, this criterion does not apply. An analysis of 10 recent studies published in five top oncology journals that used PFS as an endpoint found that only three (6%) reported deaths and events of disease progression; Only one study distinguished between local progression and distant metastasis. In addition, one study distinguished between local and distant progression, but did not provide the number of deaths before the disease progressed.

The reasons for the differences in reporting standards for the composite endpoints in cardiology and oncology are unclear. One possibility is that composite endpoints such as PFS and DFS are efficacy indicators. MACE originated from safety outcomes and was first used in the study of complications of percutaneous coronary intervention. Regulatory agencies have high standards for reporting safety results, so there is a need for detailed documentation of adverse events in clinical trials. When MACE was widely used as an endpoint of efficacy, it may have become common practice to provide quantities of each event. Another reason for the different reporting standards is that PFS is considered to be a collection of similar events, while MACE is considered to be a collection of distinct events (e.g., stroke vs. myocardial infarction). However, primary tumor growth and distant metastases differ significantly, especially in terms of clinical impact. All of these explanations are speculative, but obviously none of them justify an incomplete report. For oncology trials that use composite endpoints, especially when the composite endpoint is the primary endpoint or is used for regulatory purposes, and when the composite endpoint is present as a secondary endpoint, transparent component event reporting must become the norm.