Bias for Patient-Reported Outcomes in Open-Label Cancer Trials: How Big of a Concern Is It?

February 28, 2019

The US Food and Drug Administration explored the possibility of such bias in a recent viewpoint published in JAMA Oncology.

A common concern with patient-reported outcomes (PROs) in open-label trials is that a patient’s knowledge of treatment received could influence their view and reporting of their symptoms. With this in mind, members of the US Food and Drug Administration explored the possibility of such bias in a recent viewpoint published in JAMA Oncology. “In light of the anticipated growth in PRO submissions to the FDA,” the authors reasoned, “we outline our thoughts on the ways in which open label bias may arise.”

One such bias is that patients on the investigational arm may be “overly optimistic,” whereas patients on the control arm may be “overly pessimistic,” thus creating a potential bias that favors the new investigational agent. The authors noted that while a meta-epidemiological analysis of trials across several therapeutic areas supports this stance, “There are surprisingly few data to suggest open-label bias regarding PRO measures in cancer clinical trials.”

“The truth is that there’s no evidence at all there’s any kind of bias,” Ethan Basch, MD, MSc, a professor in UNC-Chapel Hill Department of Medicine in the Division of Hematology and Oncology, said during an interview Cancer Network. Like the study authors, he cited a study he co-authored in which no differences were found between cancer trial arms for patient-report symptoms, suggesting no concern for bias.

“There have been some evaluations comparing open-label versus non–open-label studies, and there’s not really, to my eye, any clear evidence that there is any difference in the results,” Basch said. In particular, for patient-reported symptomatic adverse events, he said, a trial being open- or closed-label “shouldn’t matter.” However, the question is “open” whether measuring disease-related symptoms, such as pain related to cancer, and how the symptoms change during treatment should be included in open-label studies.

The viewpoint authors also put forth an area potentially susceptible to bias: measures that are not directly related to treatment, such as emotional function, social function, and global quality of life. They suggested that these measures, in particular, may be “overly optimistic” in the investigation arm and “overly pessimistic” in the control arm.

The viewpoint authors proposed that while it’s important to consider open-label bias for PROs, bias is “not unique” to PROs. For example, progression-free survival and clinician-reported safety data can be biased. Basch agreed, saying, “There are all kinds of biases on the professional side in open-label studies, potentially, that we accept all the time, and there’s no reason for us to think that patients’ potential biases, if they exist, are any bigger of a problem than the clinician biases because, after all, they’re human as well.”

“[PROs] get this sort of scrutiny that other metrics don’t because there’s, I would say, a bit of a bias against the patient’s viewpoint by professionals, and I say that speaking as a professional myself, as an oncologist and a clinical investigator,” Basch continued. “My concern is that [trial] methods are being used as a way to undermine the value of the patient voice in drug development, and that’s a problem because the patients use the drugs, and so what’s more important than the patient experience with the drugs?”