Investigators Question the Use of PFS as a Surrogate End Point
Investigators challenge the assumption that PFS and HRQoL are linked, and call into question the use of PFS as a primary end point in oncology trials.
At least 12 drugs have been approved by the US Food and Drug Administration (FDA) between 2005 and 2010 using progression-free survival (PFS) as a primary end point. But, according to the findings of a new study, a PFS benefit doesn't always confer an overall survival (OS) benefit — nor does it correlate with improved patient-reported health-related quality of life (HRQoL) scores.
The authors of the study, which was published in JAMA Internal Medicine, wrote that the prevailing thought about surrogate end points — that they should correspond to an improved survival outcome, or at the very least, should correspond to how a patient feels — may not fulfill either of these functions at all. “Currently, evidence suggests that PFS serves as a valid surrogate for OS only in limited scenarios, both being variable and unpredictable,” they wrote.
And, because HRQoL is generally hampered by a medication's adverse effects, the association between PFS and improved HRQoL “is far from self-evident,” they added, and “the convenience of shorter and smaller trials made possible by measuring PFS is driving its use, rather than compelling evidence of its adequate surrogacy for either OS or HRQoL.” In fact, according to the researchers, there may be only 1 example of a strong PFS-HRQoL association from the past — but this specific study, run by the Agency for Healthcare Research and Quality, yielded inconclusive results.
Because of the uncertainty surrounding whether a true benefit exists for PFS, it is possible patients are receiving expensive and toxic medications that could prolong life — but the quality of the life that additional time affords may be poor.
To prove this suspicion, the researchers reviewed randomized controlled trials (RCTs) on cancer drugs (biologics or chemotherapeutics used as first-line anticancer medications) in which disease-related outcomes were reported by participants. They specifically examined trials that were published and registered in the International Prospective Register of Ongoing Systematic Reviews from January 2000 to May 2016. RCTs reporting an OS benefit were excluded from the original study protocol but included in the final analysis.
Through the use of 10 pairs of independent reviewers, studies were screened for eligibility and reviewers collaborated to discuss inclusions and exclusions. From 35,960 potential citations, the authors found only 38 trials (reported across 52 articles, across 12 cancer types) that met their criteria for eligibility, which included the reporting of HRQoL, PFS, or both. And, of the trials that did report HRQoL, the researchers found that “many of these trials fail to publish or report their HRQoL data to allow for quantitative analysis,” which prevented many of the trials that reported HRQoL from even being included in the authors' final analysis. And across the 38 included studies, there were 6 different HRQoL instruments that were used to assess the patient quality of life scores.
The researchers admitted their findings are probably more applicable to cytotoxic agents than to all classes of anticancer medications, as 24 of the 38 studies examined involved traditional chemotherapeutics. In addition, they noted that their findings may lack statistical power, as so few studies measured PFS and HRQoL as outlined by the current study's eligibility criteria.
Skeptics could say that HRQoL is not worthwhile to measure because it is self-reported by patients, and therefore, as a result, could be unreliable information.
“In the era of patient-centered health care, patient-reported outcome measures [are] of utmost importance,” corresponding author of the current study, Feng Xie, PhD, professor of health economics at McMaster University in Canada, wrote to Cancer Therapy Advisor. “HRQoL is patient-reported and sometimes there might be reliability or validity issues, but that is our (not patients') problem. We, as researchers, should do a better job to create measures that can reliably and validly capture patient's true feeling[s].”
Clinical trialists might be better served by measuring HRQoL robustly and adequately, Dr Xie and colleagues wrote. Rather than considering reports of HRQoL as add-ons, or “secondary” end points, HRQoL should take more of a leading role than a supporting role in the final presentation of an experimental drug's efficacy and safety.
Though the researchers concluded that PFS is often used as an end point for purely practical reasons, such as to expedite a trial's completion or to reduce the confounding by crossover designs and the administration of post-progression therapies, they expressed doubt that the disease control or purported stability awarded by a PFS benefit serves to improve a patient's experience.
“OS and/or HRQoL are what matter, nothing else,” Dr Xie said.
The conclusion that PFS correlates poorly with QoL is similar to what Bishal Gwayali, MD, PhD, Brigham and Women's Hospital and Harvard Medical School in Boston, Massachusetts, and colleagues discovered in a previous study; this team presented their findings at the 2018 American Society of Clinical Oncology meeting.2 He said that the findings in the current study surprised him because he and his colleagues examined papers only between 2010 to 2015 and found 147 trials that reported QoL. In contrast, the JAMA Internal Medicine paper only discovered 38 trials that met their QoL criteria from the period of 2000 to 2016. “This is an important issue because the QoL data are usually not published simultaneously, and therefore, such correlation analyses necessarily suffer from publication biases,” Dr Gyawali noted.
“It's also a surprise to me that the original plan was to exclude trials with positive OS,” Dr Gyawali added. “That might make sense from regulatory perspective; you would approve a drug that improves OS irrespective of QoL. But for tests of surrogacy, we need to include all data points [that are] available. If we are to examine [a] PFS-QoL correlation, what happens to OS should be immaterial, similar to [how] we disregard QoL when we do [a] PFS-OS correlation. To the authors' credit, they did include OS for the final analysis, but I am surprised that removing ‘positive OS trials' as an exclusion criteria would add only 8 more trials.”
“Finally, the use of TTP [time to progression] as equivalent to PFS is also of concern, despite the authors' acknowledgment that they are not the same,” Dr Gyawali emphasized. “Whether TTP is a correlate for QoL should be a separate study.”
- Kovic B, Jin X, Kennedy SA, et al. Evaluating progression-free survival as a surrogate outcome for health-related quality and life in oncology [published online October 1, 2018]. JAMA Intern Med.doi: 10.1001/jamainternmed.2018.4710
- Gyawali B and Hwang T. Prevalence of quality of life (QoL) outcomes and association with survival in cancer clinical trials. J Clin Oncol. 2018;36(no. 15_suppl):6573-6573.