With major advances in the scientific community’s understanding of cancer, the cancer treatment landscape has evolved, resulting in significant changes in the cancer therapy landscape and in turn the primary endpoints elected for cancer clinical trials.

Where overall survival was the dominant and standard endpoint of clinical trials in the 1980s and 1990s, the past decade has seen a regression in overall survival as a primary endpoint and a rise in progression-free survival (PFS).

Given the number of primary endpoint changes that are often redacted, experts emphasise the need for transparency when changing primary endpoints in oncology clinical trials.

Changes must be justified

Not all primary endpoint changes are problematic, but not all endpoint changes are intrinsic, says Dr Ethan Ludmir, assistant professor of gastrointestinal radiation oncology and biostatistics at the University of Texas MD Anderson Cancer Center.

By itself, an endpoint modification is not problematic because a clinical trial is a coalescence of ideas and midstream changes. However, a noted change in and of itself raises a perception of bias in an active trial that is open and accruing and even more so in a trial that has completed accrual.

There are several incidents where the primary endpoints appear to have changed after trial accrual is complete, according to research published in May 2023.  

The level of appropriateness for a primary endpoint change depends on the justification for change in the context in which the trial is being conducted, explains Dr Christopher Booth, professor of oncology and public health sciences at Queen's University.

If the original primary endpoint is no longer the most appropriate endpoint due to new knowledge or events, then changing the endpoint may be appropriate, Booth elaborates. However, the reevaluation should occur in conjunction with a data safety monitoring committee and input from trial leadership.

Some endpoints are changed for valid reasons, such as in response to regulatory considerations and conversations with groups such as the US Food and Drug Administration (FDA), adds Ludmir. Therefore, it is imperative to see how a trial has changed over its lifecycle to make that determination.

“Changing the primary endpoint or statistical analysis plan during the conduct of a trial is problematic and will add uncertainty to results. In some cases, we [FDA] may refuse to accept the change, but in others, we may feel the change will still lead to interpretable data and the evidence we need to make a decision,” a spokesperson for the Oncology Center of Excellence told Clinical Trials Arena.

Powering trials for efficacy

The issue with a changing endpoint is less about the limitations of statistical analyses to evaluate the changes and more about the type of endpoint change, such as an alteration of a surrogate endpoint like response rate versus changing a non-surrogate like overall survival, says Ludmir. Furthermore, an investigator can make several assertions based on what the trial designates as a true primary endpoint and what measure has sufficient statistical power in inferential hypothesis testing.

A study is designed to have patients randomised to a new treatment versus a control to demonstrate that the treatment shows an advantage in the outcome a trialist would like to detect; more simply, giving the study “power”, explained an anonymous biostatistician with clinical trial experience. Changing the primary endpoint gives the study more power to detect the difference, making it more efficient, and can increase the likelihood of proving a drug effective.

“When you co-primary an endpoint, then you're statistically designing it from a power calculation standpoint, to detect something. And if you're treating overall survival potentially as a secondary endpoint, then that question of whether overall survival is appropriately powered or not, is a different sort of thing,” explains Ludmir.

Powering a trial with a co-primary of PFS and overall survival (OS) is different than powering a trial with a PFS primary and an OS secondary, adds Ludmir. At the core of many of these changes is the concern that arises when progression-free survival is initially co-primaried with overall survival and one or the other is subsequently dropped to a secondary.

The culprit endpoints: overall survival and progression-free survival

The conversion of a primary endpoint of overall survival to progression-free survival raises concerns about whether trials are diluting the potential effect of the intervention and whether they are missing an opportunity to understand whether the new treatment actually helps people live any longer and better, or if the predominant benefit is a reduction of tumour growth on a CAT scan, says Booth. Shifting a primary endpoint from overall survival to progression-free survival is the most common scenario and also the most problematic one.

“Progression-free survival is often thought to be a surrogate for overall survival. But in most cases, it's actually not a very good surrogate for overall survival,” says Booth.

PFS is a composite endpoint, which includes survival; but most of the events in PFS are growth of tumours on a CAT scan, he explains. In essence, trialists have moved away from asking: "does this treatment help the patient live longer or feel better?" to an endpoint with more uncertainty surrounding whether a treatment actually helps people live longer and better lives.

“We now have an ecosystem where the default standard for most clinical trials is to use alternative endpoints rather than overall survival. I think we need to have a conversation about that because overall survival shouldn’t be the primary endpoint of every trial, but we've shifted too far in the other direction,” says Booth.

Most cancer clinical studies use PFS as an evaluation measure, which is a necessary measure to prove that the treatment in question is effective, says the anonymous biostatistician.

Most commonly, overall survival data will be projected in the same direction as PFS, but it won’t demonstrate as much of an advantage over the control treatment, they add.

The problem is when the data from PFS and overall survival endpoints conflict and the treatment demonstrates a strong advantage in PFS while looking worse in overall survival.

It is a real challenge to understand why the discrepancies occur because there are many reasons—environmental and treatment-related complications—as to why the data is conflicted. Even if the treatment gives a strong advantage in PFS, if the overall survival appears worse, chances of approval become shakier, they explain. And approval is what drives Pharma.