This article is more than 5 years old. Information may no longer be current.
Surrogate endpoints may not predict OS in oncology clinical trials
Click Here to Manage Email Alerts
A majority of trial-level validation studies of surrogate endpoints demonstrated little correlation with survival rates, according to a systematic review of trial-level meta-analyses.
“Although there is growing recognition that adopting new medical practices based on improvements in surrogate outcomes can lead to misleading conclusions, surrogate endpoints continue to play a prominent role in oncology," Vinay Prasad, MD, MPH, postdoctoral cancer prevention fellow at the NCI, and colleagues wrote. “In the U.S., most new cancer drugs are approved based on surrogate measures, such as response rate and PFS, through the FDA’s Accelerated Approval pathway.”
Prasad and colleagues initiated a literature review of trial-level meta-analyses of randomized clinical trials that quantified the association between surrogate endpoints and OS within oncology.
The researchers used MEDLINE and Google Scholar to identify potentially appropriate meta-analyses. Search terms included “regression” or “correlation,” “surrogate” and “endpoint” and “oncology,” “cancer,” “OS” and “trial level.”
Of 108 articles identified, the researchers read 62 articles in full. Thirty-six of these articles contained 65 specific correlations between surrogate endpoints and OS.
Published literature and meeting abstracts comprised 39% of included articles, whereas systematic reviews accounted for 28%. The included meta-analyses studied surrogate endpoints in the neoadjuvant, adjuvant, locally advanced and metastatic settings.
Few of the meta-analyses used a convenience sample (n = 4; 11%) or included unpublished trials identified in clinical trial registries (n = 5; 14%).
Researchers documented the correlation coefficient (r) for all trial-level analyses.
Fifty-two percent (n = 34) of meta-analyses reported low correlations between surrogate endpoints and OS (r ≤ 0.7), and 25% (n = 16) reported medium-strength correlations (r > 0.7 to r > 0.85).
The remaining fifteen studies reported a high correlation between surrogate endpoints and OS (r ≥ 0.85). Of these, six occurred in the adjuvant setting, six in the metastatic setting and three in the locally advanced setting. The researchers also noted three occurred in the adjuvant colorectal cancer setting.
The researchers identified several limitations of their study, including the lack of external validation of the specific thresholds used to grade the strength of correlation. Further, they could not determine whether correlations improved or weakened based on a comprehensive assessment of clinical trials.
“Most trial-level validation studies of surrogate endpoints in oncology find low- or medium-strength correlations with OS,” Prasad and colleagues concluded. “All validation studies use only a subset of available trials. The evidence that surrogate endpoints predict OS in oncology is limited.”
Rita F. Redberg
Researchers should consider the consequences of expediting patient access to cancer drugs without a clear OS benefit, Rita F. Redberg, MD, MSc, professor of medicine at the University of California, San Francisco and chief editor of JAMA Internal Medicine, wrote in an accompanying editor’s note.
“In 2014, over 60% of approved drugs were already approved using expedited review or orphan drug status, which also has reduced evidence standards for safety and effectiveness,” Redberg wrote. “We must reduce drug approvals based on unreliable surrogates and change practice when critical studies show no survival benefit. … In our rush to find new effective treatments, we should not harm our patients with ineffective toxic ones.” – by Cameron Kelsall
Disclosure: The researchers and Redberg report no relevant financial disclosures.
Click Here to Manage Email Alerts