This article is more than 5 years old. Information may no longer be current.
Research collaboration designed to better define real-world effectiveness of immunotherapies
Click Here to Manage Email Alerts
A research collaboration between the FDA and Flatiron Health, a health care technology company, aims to evaluate patient data collected outside of clinical trials to provide new insights that could benefit patients with cancer.
The first component of the collaboration focuses on use of immunotherapy among patients with advanced non–small cell lung cancer.
As few as 4% of patients with cancer are represented in clinical trials. Thus, clinicians must recommend treatments for their patients based on safety and efficacy data that only reflect a small proportion of the population.
“Through this research collaboration, we can augment our understanding of the safety and effectiveness of lung cancer therapies outside of clinical trials, in a manner consistent with supporting a learning health system where new knowledge and discovery can be byproducts of the health care delivery experience,” Kristofer Baumgartner, a spokesman for the FDA, told HemOnc Today. “This collaboration is part of an organized effort by the Office of Hematology and Oncology Products to proactively investigate the utility of novel pipelines of data — such as real-world data from electronic health records — in advancing regulatory science.”
Researchers will evaluate de-identified, HIPAA–compliant data from patients treated outside clinical trials to determine how the safety and efficacy data from studies translate to a real-world population. Investigators also will explore analytic approaches, clinical endpoints and safety assessment methods that are relevant to a real-world population.
“This initial collaboration with the FDA asks fundamental questions,” Amy P. Abernethy, MD, PhD, chief medical officer and senior vice president of oncology for Flatiron Health, as well as adjunct professor in the department of medicine and member in the Duke Clinical Research Institute at Duke University School of Medicine, told HemOnc Today. “These questions include whether the depth and quality of real-world data are complete enough for the FDA to analyze to make a variety of different decisions, and how these complex datasets can be best analyzed.”
Need in lung cancer
By evaluating real-world populations, Abernethy and colleagues hope to help the FDA with regulatory decisions.
“Clinical trial infrastructure is fundamental in helping to understand safety and efficacy in a very detailed way,” Abernethy said. “However, clinical trial populations tend to exclude patients who have co-occurring disease, such as heart disease or HIV, and be very specifically focused on people who have the wherewithal and the physical functioning to get through the required tests. This results in a narrow population.”
Researchers chose the population with lung cancer as the entryway into this initiative due to the potential to make the greatest impact on the most patients.
Lung cancer is the leading cause of cancer death in the United States. More than 220,000 Americans were diagnosed with lung cancer in 2015, and 160,000 people died of the disease. NSCLC accounts for up to 85% of lung cancer cases.
“Even as a melanoma specialist, I knew that this research would have the greatest effect in lung cancer,” Abernethy said.
Advancements in immunotherapy have sparked hope among clinicians, researchers and patients across a range of cancer types, particularly lung cancer.
In 2015, the FDA approved two immunotherapies — pembrolizumab (Keytruda, Merck) and nivolumab (Opdivo, Bristol-Myers Squibb) — for the treatment of advanced lung cancer.
Pembrolizumab was approved for PD-L1–expressing squamous and nonsquamous NSCLC based on efficacy data from 61 patients positive for PD-L1. The overall response rate in this cohort was 41%.
The FDA based the initial approval of nivolumab for lung cancer on results of a phase 3 trial composed of 272 patients with squamous NSCLC. Nivolumab extended median OS by 3.2 months compared with standard chemotherapy (9.2 months vs. 6 months; HR = 0.59; 95% CI, 0.44-0.79). The FDA later expanded the approval to include nonsquamous NSCLC based on data from 582 patients. These data showed a 27% reduced risk for death with nivolumab (median OS, 12.2 months vs. 9.4 months; HR = 0.72; 95% CI, 0.6-0.88).
The FDA–Flatiron Health initiative will explore the safety and effectiveness of immunotherapy in a much larger cohort of patients than were initially treated in the clinical trials. The first cohort of 1,578 patients treated with immunotherapy provided by Flatiron are a subset of a larger cohort of nearly 30,000 patients with lung cancer for whom data have been captured.
“Analyzing and examining raw data from electronic health records can help the FDA formulate an approach in defining the utility of real-world evidence for use in the regulatory setting,” Baumgartner said. “This project provides an opportunity for the FDA to examine issues surrounding electronic health records data quality and veracity, as part of ongoing efforts to develop methods and best practices to reduce uncertainties associated with drawing conclusions based on data generated outside of clinical trials.”
Real-world endpoints
Abernethy and colleagues also seek to establish real-world endpoints that would provide a reliable assessment of treatment impact. They can include OS, time to treatment failure, duration of therapy, progression of disease and tumor response.
These endpoints must be based on routinely captured clinical data; tied to source evidence; shown to be meaningful according to a predefined experimental validation framework; and accepted by oncologists, researchers, regulatory bodies and industry.
Capturing an endpoint based on real-world data from electronic health records can be confounded by variable time points in disease assessment, the potential for missing data, lack of access to prior scans, and the number of sources of evidence per event. Researchers required an approach that accounts for the complexities of the real world, is scalable, is replicable across abstractors, and is portable to multiple settings and electronic health records.
Reliable real-world endpoints could benefit retrospective analyses, postapproval regulatory submissions, payer discussions and prospective pragmatic trials.
The first endpoint Abernethy and colleagues evaluated was real-world progression.
RECIST criteria — used in clinical trials — define progression as changes in tumor size based on radiologic evidence, but they may not be a practical solution using electronic health record data. In the real world, progression is defined as a change in tumor burden based on a radiologist’s interpretation of scans, or a worsening of disease based on a clinician’s interpretation of the entire patient chart.
Because numerous ways exist to define progression based on available data, Abernethy and colleagues proposed a real-world endpoint that would be a consensus of these various inputs. They evaluated three approaches to establish the first progression event after first-line therapy: a clinician assessment, radiologist assessment or hybrid approach.
The analysis included data from 200 patients (median age, 65.5 years; 50% female) with advanced NSCLC who received a diagnosis in 2011 or later. All patients completed at least one line of therapy and had initiated a subsequent line.
Researchers reviewed all relevant documents — such as clinician notes, and radiology and pathology reports — for evidence of potentially eligible progression events.
Median follow-up was 13.2 months.
Use of a first clinician-confirmed event compared with first radiology-based event to indicate progression identified fewer patients who had a progression event (n = 173 vs. n = 180), more patients who had treatment change within 60 days of progression (n = 126 vs. n = 122), greater percentage of progression events associated with treatment change (73% vs. 68%) and a longer median real-world PFS (5.49 months vs. 4.9 months).
Of the 173 patients with progression events by clinician and radiology methods, the progression dates were a match for 156 (90%) patients. In 16 of 17 of the cases that did not match, the clinician made an assessment of stable disease, and all of these patients subsequently progressed based on follow-up scans.
Based on these data, researchers determined clinician-confirmed progression is the best approach to assess real-world progression.
They then analyzed 727 patients (median age, 65 years; 48% female) with NSCLC — all of whom were diagnosed in 2011 or later and had initiated at least one line of therapy — to evaluate the correlation between real-world endpoints and OS using Spearman’s rank correlation.
Data showed OS correlated best with real-world PFS (Spearman’s rho = 0.749) rather than time to next treatment (Spearman’s rho = 0.685) and time to treatment failure (Spearman’s rho = 0.515).
Based on both analyses, Abernethy and colleagues proposed to define real-world progression as “all distinct episodes in which the treating clinician concludes that there has been overall growth or worsening of the disease of interest.”
Next steps
Researchers plan to analyze the association between treatment effects on real-world PFS and OS in a population of patients treated at cancer centers. The analysis will replicate a clinical trial cohort using real-world data to compare actual outcomes with published endpoints.
The first pilot study will evaluate approved PD-1 inhibitors in patients with highly mutated tumors.
Eventually, this methodology can be extended past lung cancer into larger patient samples, other tumor types and other treatment settings. Further, researchers can evaluate endpoints other than real-world PFS, such as real-world tumor response and a composite of PFS, quality of life and opiate use.
These research efforts stand to expand the safety profile of a therapy, identify populations with enhanced benefits or risks for an approved therapy, build evidence for supplemental packaging to expand the therapy’s indication and serve as a postmarket confirmation of benefit.
“We can start to characterize how people are similar and how they are different from the population represented in the clinical trial, and we can start to evaluate how those differences impact the performance of a specific drug,” Abernethy said. “I think we are going to see label expansion — decisions made on drugs that are already approved.
“The goal is for clinical trial data and real-world data to become married as a way to totally understand the total view of how interventions work for patients, and how both clinical trial data and real-world data are important,” she added. – by Nick Andrews
References:
Borghaei H, et al. N Engl J Med. 2015;doi:10.1056/NEJMoa1507643.
Brahmer J, et al. N Engl J Med. 2015;doi:10.1056/NEJMoa1504627.
Flatiron Health. Development of real-world endpoints. Presented at: Alexandria Summit; June 16, 2016; New York.
Garon EB, et al. N Engl J Med. 2015;doi:10.1056/NEJMoa1501824.
For more information:
Amy P. Abernethy, MD, PhD, can be reached at amy@flatiron.com.
Kristofer Baumgartner can be reached at kristofer.baumgartner@fda.hhs.gov.
Disclosure: Abernethy reports an employment role with Flatiron Health. Baumgartner reports an employment role with the FDA.