Symptom control, overall survival and time to treatment failure may be more useful measures in a real-world setting than progression-free survival, FDA oncology director Pazdur says.
If the biopharma industry wants to rely more on real-world data to support regulatory decision-making, it may need to look at endpoints different from those used in traditional clinical trials.
Clinical trial endpoints may not translate well into research conducted in the real-world setting, said participants at a June 16 meeting on real-world evidence sponsored by Friends of Cancer Research and Alexandria Summit.
In the oncology setting, commonly used endpoints such as progression-free survival and response rate may not hold up well in studies conducted outside of a strictly controlled clinical trial, FDA’s top oncology reviewer said.
“Our attempts to try to shoehorn our concepts of a clinical trial may not be entirely appropriate,” FDA’s Pazdur said.
“My plea is not just to take the things that we use in clinical trials, but really to take a look at perhaps novel ways of looking at real-world data,” FDA Office of Hematology and Oncology Products Director Richard Pazdur said. “It is real-world data, it is not a clinical trial, and our attempts to try to shoehorn our concepts of a clinical trial may not be entirely appropriate.”
The need to develop and validate new endpoints will be yet another challenge facing both FDA and product sponsors in the push to leverage vast amounts of electronic medical records and other real-world data to support new indications and other regulatory actions. Other challenges include technological hurdles, data quality, and what some in industry see as a lack of FDA clarity on regulatory expectations for such data (“Real-World Evidence: Efficacy Assessments Await FDA Clarity, Pilot Projects” — “The Pink Sheet,” Mar. 14, 2016).
‘Pressure-Testing’ Real-World Endpoints
Real-world data is a term increasingly used to describe data collected outside the traditional clinical trial setting. This includes claims data, registries and so-called pragmatic clinical trials, which incorporate randomization in routine clinical care settings and rely on data collected during patient care.
FDA Commissioner Robert Califf has said randomized, clinical practice studies would be the most useful type of real-world data for drawing causal inferences about a drug’s risks and benefits, but he noted the value of other types of data generated outside of traditional clinical trials (“FDA’s Califf On Real World Evidence: ‘Use It For The Right Purposes'” — “The Pink Sheet” DAILY, Jun. 17, 2016).
At the FOCR/Alexandria Summit meeting, Amy Abernethy, chief medical officer and senior VP-oncology at the healthcare technology company Flatiron Health Inc., talked about the challenges of capturing endpoints in real-world data based on electronic health records. These complexities include the room for interpretation in radiological reports, variable time points for disease assessments, clinical nuances such as pseudoprogression and mixed response, potential for missing data, and time and complexity of extracting pertinent information from medical records.
She outlined a framework for developing and “pressure-testing” endpoints using electronic health record data, and she applied this framework to assess the validity of a progression endpoint defined in four different ways: response evaluation criteria in solid tumors (RECIST), which is how progression is defined in clinical trials; radiologist interpretation of scans; clinician interpretation of the entire patient chart; and changes in treatment as a signal of worsening disease.
The exercise showed that “real-world progression” based on clinician interpretation could be reliably captured from electronic health records in a population of advanced non-small cell lung cancer patients, Abernethy said. In addition, real-world progression-free survival (rwPFS) correlates well with overall survival at the patient level and is more robust to cohort selection factors than time to treatment failure/time to next treatment, she said.
Symptoms, Survival Are More Real
However, FDA’s Pazdur questioned whether progression was the appropriate endpoint to measure in a real-world setting.
“I don’t know if I would go in that direction,” Pazdur said of an rwPFS endpoint. PFS “is a very difficult endpoint to use in the context of a clinical trial, and then to try to fit it in to real-world data, is that the correct thing to do?”
He suggested there are other endpoints that may better reflect the real-world experience with cancer drugs.
Real-world data can help provide a clearer picture of symptom benefit and control, Pazdur said, pointing out that one of the problems with clinical trials is their eligibility criteria are often restricted so that most patients are asymptomatic when they enter a study.
“I think using real-world data when these drugs actually go out in the real world gives us a unique opportunity to take a look at symptoms and those are, I would think, much more easily quantitated and extractable from the medical records than worrying about progression on an X-ray,” he said. “It’s something much more meaningful to patients.”
Overall survival also would be an important endpoint in the real-world setting because it provides an opportunity to better understand the entire disease process, not just the effects of a single drug, he said.
Pazdur said he would like to see how the survival curve for a drug after several years of real-world use compares to the survival curve from the pre-approval clinical trials. “I don’t know if they’re going to be the same. I don’t know if they’re going to even be close. But what’s really important is what’s really going on in the real world, not what’s in this artificially constructed clinical trial,” he said.
Pazdur acknowledged the agency traditionally has not looked favorably upon time to treatment failure as an endpoint in randomized controlled trials. However, time to next treatment might be better suited to a real-world study because a change in therapy and the reason for that change, such as progression or drug toxicity, likely would be noted in medical records.
While it’s important to look at response rates, assessments based on RECIST may not be practical because most physicians in the real world don’t use such criteria, Pazdur said. “Maybe we just have to get more real about what goes on in the real world and look at a different type of response rate, perhaps a lesser decrease in the size of a lesion.”
The oncology division director said the most important thing about any endpoint and its analysis comes down to magnitude, “especially when we’re dealing with real-world data” and with an endpoint that requires radiographic interpretation.
“When we see usually correlations between PFS and overall survival, or response rates and overall survival, it’s because there’s a large magnitude,” he said. “I do not know if response rates and PFS are true surrogates. I have my significant doubts. I think when we do see correlations it’s because we just have effective drugs, and effective drugs with large magnitudes have effects on multiple endpoints.”
Similarly, large effect sizes will be needed to overcome the “noise” generated in a real-world study so that such data can be used to confirm clinical benefit or support labeling expansions (see related story, “Real-World Evidence May Find A Home On Breakthrough Pathway” — “The Pink Sheet,” Jun. 27, 2016).
Exploration of clinically relevant endpoints is one aspect of a collaboration between FDA and Flatiron that the company announced in May. The project will systematically explore the characteristics and treatment patterns of patients with advanced NSCLC who are receiving immunotherapy to better understand real-world treatment patterns and formulate further regulatory science hypotheses.