Beyond OS: Common Sense Endpoints for Cancer Drugs

The rapid advance of oncology drug development requires evolving beyond using overall survival in randomized trials as the only acceptable endpoint.

Maurie Markman, MD

The rapid advance of oncology drug development requires evolving beyond using overall survival (OS) in randomized trials as the only acceptable endpoint, according to Maurie Markman, MD.

“I am not in any way suggesting that we shouldn’t be using rigorous endpoints. I’m not suggesting that we shouldn’t be objective. I’m not suggesting that we shouldn’t have higher standards. I’m simply saying they cannot be phase III randomized trials looking at hundreds and hundreds of patients and looking at an overall survival endpoint. It would take us decades, which means we would not have new drugs available to our patients,” Markman, national director for Medical Oncology, Cancer Treatment Centers of America, said in a presentation at the 33rd Annual Miami Breast Cancer Conference®.

Although OS in a randomized trial has historically been considered the “gold-standard endpoint” in oncology, Markman said the paradigm must shift now that the availability of multiple lines of therapy is increasingly the norm, rather than the exception, particularly in the advanced setting.

“Therapies utilized in disease management following completion of a given trial may significantly impact overall survival for patients on that study completely independent of the impact of the trial-based strategy itself, seriously interfering with any rational interpretation of the impact of the investigative regimen on survival,” said Markman.

A meta-analysis by Flaherty, et al (Lancet Oncol. 2014;15[3]:297-304) demonstrates the confounding affect of future treatment lines on OS data. The analysis evaluated the correlation between progression-free survival (PFS) and OS in dacarbazine-controlled randomized trials evaluating BRAF inhibitors in metastatic melanoma.

The results showed that initially, when the BRAF inhibitors became available, there was a 0.96 correlation coefficient between PFS and OS, because there were no effective second-line therapies available, said Markman. However, once the drugs were on the market for a few years, the correlation coefficient between OS and PFS fell to 0.55, as multiple lines of treatment, including immunotherapies, became available.

“These data show that when you have effective second-line therapies versus when you don’t, it impacts the relationship between OS and PFS—nobody would say these frontline therapies are not effective because you don’t see that correlation anymore.”

He added, “Imatinib has never been shown in a phase III randomized trial to improve overall survival when used as frontline therapy in chronic myeloid leukemia. Why? Because in the trials comparing it to interferon, imatinib was used as second line after the patient had failed interferon, and guess what? It worked very well, so overall survival wasn’t improved.”

Markman also highlighted the infeasibility of OS as an endpoint in terms of the number of patients required for the research. Broglio and Berry (J Natl Cancer Inst. 2009;101[23]:1642-1649) conducted a statistical analysis in which they determined the number of patients required to convert a 3-month PFS improvement into an OS advantage in a hypothetical population of 280 patients with a nonspecified type of cancer. The researchers found that it would require 350 patients to demonstrate a 2-month median OS benefit and 2400 patients to show a 24-month median OS benefit.

“This is a stunning example of how it is irrational to continue to talk about randomized phase III trials looking at a survival endpoint when we are increasingly converting advanced cancers into chronic illnesses,” said Markman.

These issues with OS, combined with new realities in clinical research—increased understanding of molecular biology of malignant disease, heterogeneity within a group of cancers and a single cancer, smaller molecularly-defined subgroups—mandate novel study designs and endpoints to evaluate the efficacy and value of novel cancer agents, according to Markman.

Markman discussed potential nonrandomized study endpoints worthy of consideration in specific clinical trial settings, including a substantial RECIST response rate of long duration, and survival or time to disease progression compared to a robust historical database.

Another approach would employ a personalized medicine approach, examining time to disease progression for an individual patient following treatment with a novel agent compared to the time to progression for the patient on the preceding regimen.

“We can ask for doubling of that time or tripling of that time, or quadrupling of that time. We can make it as rigorous as we want to say it meets our criteria. But we need some way to aggressively, realistically, and rapidly demonstrate utility of these agents so we can decide whether or not we want to bring them into our armamentarium for the benefit of our patients.”


View more from the 33rd Annual Miami Breast Cancer Conference