A friendly debate held at the European Hematology Association 2023 Congress focused on the potential for real-world data to improve the generalizability of clinical trial results but also acknowledged the limitations and current challenges with these data.
A friendly debate held at the European Hematology Association 2023 Congress focused on the potential for real-world data to improve the generalizability of clinical trial results but also acknowledged the current challenges in collecting and standardizing these data.
Moderator Philippe Rousselot, MD, of the Université de Versailles Saint-Quentin-en-Yvelines in Paris, set the stage by discussing the considerable attrition seen in clinical trials, where the rule is that 1 patient is analyzed for every 10 identified. Involving patients more and incorporating their input in a trial may be one way to improve that rate, he said.
Further, the real world carries much more variability, as clinical trials have eligibility criteria that often exclude older patients or those with comorbidities, which affects the generalizability of clinical trial results. In other words, Rousselot said, there is a continuum with sensitivity at one end, where clinical trials aim to detect an intervention’s superiority, and applicability at the other, with pragmatic trials closer to approximating the real world.
With this tension in mind, he asked the audience, can we use real-world data instead of clinical trials? The poll responses were split with 45% voting yes and 55% voting no.
There to argue the “no” position was Martin Dreyling, MD, professor at Ludwig Maximilian University in Munich, Germany. “There is no alternative to clinical trials,” he said, presenting a slide of a pyramid ranking meta-analyses as the highest-quality evidence, with those meta-analyses composed of individual randomized controlled trials. To prove the point that high-quality evidence is of utmost importance, he asked the audience to raise their hands if they don’t believe in evidence-based medicine; no one did.
Case studies do have a place in formulating the hypotheses investigated in trials, he said, but only randomized prospective trials can balance for the known and unknown risk factors that determine outcomes. Cohort studies can account for known risk factors, but no one knows everything about medicine.
His work with the European Consortium on Reducing Bureaucracy in Clinical Trials has informed his view that instead of doing away with clinical trials because they don’t translate to the real world, we instead need trial participant pools that represent the general population.
On the contrary, argued Carsten Niemann, MD, PhD, clinical associate professor at Copenhagen University in Denmark, we have to use real-world data to provide evidence of how something works for actual patients. He picked as an example the TRIANGLE trial—not coincidentally, led by Dreyling—of autologous stem cell transplantation vs ibrutinib in mantle cell lymphoma and asked the audience if these participants represent their own patients with the disease. Those who don’t meet clinical trial inclusion criteria have much poorer survival outcomes, and half of Danish patients with multiple myeloma wouldn’t meet typical trial criteria, he noted.
Clinical practice has historically been informed by “the medical art” of incorporating education, training, experience with patients, and discussion with colleagues, Niemann said, but now with the advent of time-series data and genomics, we are moving into the age of the data-driven medical art. Using technology to identify patterns within broad sources of data can overcome the criticisms of real-world data as low-quality, anecdotal evidence.
He highlighted his work with the DALY-CARE cohort of Danish lymphoid cancer research, which links a biobank of tissue samples, national health registries, electronic health record data, and other data sources together to determine which patients benefit from which intervention and inform how to treat them.
Niemann proposing inverting the evidence pyramid shown by Dreyling so that the top spot is held by the reality for real-world patients, followed by the truth for clinical trial patients, and finally at the bottom are the accepted beliefs in medical textbooks.
“We need to represent our patient population and we need to do pattern recognition to adjust for all the problems with real-world data,” he concluded.
Moving into the debate portion, Rousselot asked the participants about their predictions for the next 10 years as targeted therapies increase the number of available treatment options and decrease the number of patients eligible for each trial.
Dreyling said that it may have been easier to perform trials back when there was 1 drug for treating myeloma and 1 for treating lymphoma, but now that we have the ability to segment patients into subsets like p53 alterations within rare diseases, it makes clinical trial design more complicated but still feasible if the framework is improved to reduce bureaucracy.
In response, Niemann argued that we also need evidence for approaches that haven’t been tested in clinical trials, such as medical devices and decision support tools. New technologies, often supported by artificial intelligence, can connect the patterns leading to outcomes that can be extrapolated to the real-world population outside of clinical trials.
Regulatory bodies have an important role in this aim as well, he added, as data-driven pattern recognition will be more feasible if the FDA and European Medicines Agency make it mandatory to collect and standardize these data.
In response to an audience question, both speakers expressed skepticism of real-world evidence coming from social media, in part because of the potential bias: Patients who are able to log in and report their experiences are more likely to be having a better outcome.
There are also legal considerations around both types of data, the panelists agreed, which can make pharmaceutical companies hesitant to release their clinical trial data for real-world research. But these data are necessary, Niemann said, “to make sure that we can actually improve treatment for our patients, that we can identify the subgroups that we need to model and identify and treat differently.”
Another audience member raised concerns about reliability in real-world data collection, but Niemann noted that adverse events reported in clinical trials can vary in classification, whereas medical histories can show, for instance, the timing of infection based on blood cultures being drawn and antibiotics being prescribed.
Polled again on whether real-world data can replace clinical trials, the audience shifted more toward Dreyling’s perspective, with 25% saying yes and 75% no.
Satisfied that the attendees at least saw the advantages of both, the 2 experts shook hands, concluding the debate.