The potential benefits (and pitfalls) of real-world evidence in oncology
Years ago, on a visit to Cambridge in the U.K., I walked across one of the many parks in that beautiful city.
The park separated a middle-class, residential community from central Cambridge. Those of you who have visited will know that the central part of the city is dominated by Cambridge University — one of the oldest academic institutions in the world — with stunning buildings dating back many centuries and a distinct, rarified academic atmosphere.
In the center of the park, someone had spray-painted a lamppost with the words “reality checkpoint.” I understood exactly what he or she meant — the center of the park marked the transition from academia to the “real world” outside.
In recent years, much has been written about the highly selected nature of patients entered on clinical trials in cancer. Exclusions based on age, comorbidity, poor performance status, specific patterns of disease spread and so on provide a “clean” study population for the purposes of evaluating the safety and efficacy of a new intervention, but don’t readily translate into the real world of cancer care.
In the real world, our patients typically are older, often have significant comorbidities, and may have a remote history of another cancer or evidence of spread of their disease to the central nervous system. Until recently, most of these factors would have excluded a patient from a trial, leaving unanswered the question of whether the anticipated outcome for these patients would be the same as for the selected patients in the trial.
Many organizations and advocacy groups have highlighted the need to relax eligibility criteria for clinical trials for two main purposes. Firstly, to give as many patients as possible access to potentially effective new treatments, and secondly, to generate meaningful results from clinical trials that have widespread applicability to the entire cancer population.
Both goals are highly appropriate and have the potential to test new treatments in a rigorous fashion, according to the gold standard of a prospective interventional trial, but with eligibility criteria more representative of the true patient population encountered in the community.
That said, approval of new cancer treatments — through phase 1 to phase 3 trials — is, as we all know, a slow process, until recently very focused on safety before efficacy. It can take many years for an effective new treatment to reach the patients who need it.
Fortunately, changes in the regulatory agencies have accelerated the process for approval, and many highly effective new drugs are now reaching patients in a much timelier fashion. To accelerate this process further, the FDA has expressed interest in using real-world evidence in place of data from prospective studies as part of its approval process.
Mining big data
In many respects, there’s nothing new about the idea of real-world evidence.
For many years, hypothesis-generating clinical research has been developed from large case registries in many different settings, allowing researchers to explore outcomes from large data sets and pose questions that can then be addressed in prospective studies.
In fact, in many cases, data from these registries resulted in changes in practice and establishment of new standards of care without confirmation in prospective studies — most of the current practice in stem cell transplantation, for example, is based on these kinds of data, as relatively few randomized trials have been performed in this setting.
The potential pitfall of such registry-based data is that they don’t overcome the intrinsic selection bias involved in registry studies — the patient has to meet certain criteria to receive the intervention, which may not be so different from those required to get onto a study.
The promise of real-world evidence is that it has the potential to overcome some of this inherent bias, partly because of the volume and granularity of data that are now available from very large data sets derived from electronic health records and other sources.
The advent of “big data” in oncology has made this possible. Many of us are now contributing data to these projects, and the possibility of mining these data sets to identify patients, for example, who might substitute for the control arm of a randomized trial is intriguing and could certainly accelerate the approval process for some new treatments.
The FDA has signaled that it is interested in exploring real-world evidence for this further — something that to date has only been applied to post-registration and marketing studies of new agents.
There is little doubt that big data provides an exciting opportunity, through real-world evidence, to get new treatments through the approval process and get benefit to patients more quickly.
It will be important to ensure the quality of the data and that definitions around what constitutes real-world data, and how these data sets are put together, are very closely defined.
As an example, two studies presented in December at ASH Annual Meeting and Exposition explored the outcome for patients with diffuse large B-cell lymphoma undergoing chimeric antigen receptor T-cell therapy with a commercially available CAR T-cell product outside of a clinical trial due to failed eligibility. There is no intended criticism of the studies — they were carefully conducted, retrospective studies looking at the early outcomes from CAR T cells, compared with results reported in the formal ZUMA-1 phase 2 trial.
Both studies were framed as real-world evidence. However, even a superficial look at the patient characteristics shows that these patients had a median age about 10 years younger than typically seen in DLBCL, and most had a favorable performance status of 0 or 1 — this hardly represents the real world of DLBCL, where most of our patients are older and have comorbidities. We must be careful not to let the new concept of real-world evidence drift back to what we previously defined as historical controls.
We will also have to accept that as we gain access to real-world evidence, others — such as insurance companies — will, too.
Once a new treatment is approved using these data, they may be in a position to determine the likelihood that an individual patient will benefit based on experience with a similar cohort in the real-world data set. Of course, this ultimately may be the right decision if the data are reliable.
The real world in oncology is difficult to define, and varies depending on our practice setting, demographics and access to care, but there is no doubt that well-constructed real-world evidence — if used appropriately — could be a huge benefit to approval of new treatments for our patients.
It will be important that real-world evidence is tightly defined and applied to prevent the new jargon creeping into the vocabulary of less well-derived and -defined comparator data.
We will need to be sure that historical controls and matched analyses are not redefined as real-world evidence to gain newfound respectability.
The following were presented at ASH Annual Meeting and Exposition; Dec. 1-4, 2018; San Diego:
Jacobson CA, et al. Abstract 92.
Nastoupil LJ, et al. Abstract 91.
For more information:
John Sweetenham, MD, FRCP, FACP, is HemOnc Today’s Chief Medical Editor for Hematology. He also is senior director of clinical affairs and executive medical director of Huntsman Cancer Institute at The University of Utah. He will start a position as associate director for clinical affairs at Harold C. Simmons Comprehensive Cancer Center at UT Southwestern Medical Center on April 15. He can be reached at firstname.lastname@example.org.
Disclosure: Sweetenham reports no relevant financial disclosures.