RWD in clinical trials: External control arms take the lead

The US Food and Drug Administration (FDA) has now published four draft guidances spelling out how it will consider real-world data (RWD) in regulatory decision-making, from the use of electronic health records and registries to RWD in non-interventional studies. More guidance documents are planned, including details on study designs that incorporate RWD for external control arms.
 
External control arms, which use data collected outside of the current trial to provide a comparator group, are already being used in regulatory applications. This type of approach is particularly attractive for studies of rare disease treatments and in oncology, where single-arm trials are sometimes the most practical study design or the only ethical option. However, no marketing applications in oncology have used an external control arm as part of the primary efficacy analysis. Instead, external control arm data has been used to establish the natural history of the disease, to isolate the treatment effect, or for a comparative efficacy analysis, according to a January 2022 review in Annals of Oncology authored by experts from the FDA.
 
“What is changing now is using [external controls] as a proper comparator arm. To do that, you have to maintain the standard of a clinical trial,” said Laura Fernandes, PhD, senior statistical director at COTA, an oncology data company. “That is where the field is now moving, where we are trying to use real-world data, especially from electronic health records to fulfill the need of a control arm in a clinical trial.”
 
Contemporaneous data
 
External control arms can include data from prior clinical trials, or RWD that comes from registries, electronic health records (EHRs) or medical claims. While most of the experience with external control arms comes from using datasets from previously conducted trials, RWD sources can offer an advantage because they can produce contemporaneous data to reflects the current practice of medicine.
 
Data from previous trials may be helpful because it conforms most closely to the exclusion/inclusion criteria of a trial, Fernandes said, but it might also reflect outdated practice standards related to imaging or diagnostic testing. Contemporaneous data from an EHR or a registry is more like to reflect the current standard of care for diseases with many new advances, such as lung cancer. The applicability of the data source depends on the disease and the research question, she explained.
 
‘Fit for purpose’ data
 
As FDA begins to consider more regulatory applications that incorporate external control arms, either as primary comparator or in a supportive data role, the agency is looking for the data to be “fit for purpose,” Fernandes said. “The FDA wants the EHR data to look like clinical trial data, but you cannot convert an apple into an orange,” she said. Instead, the draft guidance documents have focused on data standards and ensuring that data sets are well curated and well characterized. (RELATED: FDA drafts data standards guidance for RWD, Regulatory Focus 22 October 2022)
 
There are challenges in finding real-world data that can be compared with trial data such as differences in the frequency and timing of imaging in actual practice, as well as missing data on medical adherence. “That’s a struggle between expectation and reality,” Fernandes said.
 
Real-world data is “messy” because it was created for a different purpose, explained Nancy Dreyer, PhD, chief scientific officer and senior vice president for IQVIA Real World Solutions. This means that using an external control arm will introduce some error, she said, but if the effect size is large enough the benefit or the harm will still show up. This makes external control arms useful options for new cancer treatments but not a good option for treatments likely to show small incremental benefit.
 
“The smaller the expected benefit from the treatment, the riskier it is to use an external comparator because it is less likely to be informative,” Dreyer said.
 
It can be difficult to find EHR data that is fit for regulatory purposes because the outcomes researchers are seeking may not be available in one place or may exist only in the provider notes, which are not always available. Additionally, there is a lack of standardization among health care providers on subjective measures such as pain improvement. “A lot of outcomes aren’t recorded or aren’t recorded in a way that would be useful,” Dreyer said.
 
Dreyer said registry data tends to be more useful as an external comparator because of the systematic data collection of various endpoints, but it may not have the follow up that researchers need.
 
Linking data sources

While individually all the data sources are imperfect, linking data together offers potential benefits. Dreyer said “linkage” of data sources, such as EHRs and registries to clinical trials subjects, was a theme of the FDA draft guidance documents. Data linkage has the potential to show the longitudinal patient journey and to check the accuracy of self-reported information.
 
“Linkage is the hot topic,” she said. “FDA has asked that anyone who wants to bring that type of information forward should show how accurate their linkage is.”
 
Moving forward, RWD is likely to play a larger role in assessing benefits and risks for drugs that receive accelerated approval or in diverse patients who were not included in pivotal clinical trials, Dreyer said. “It’s too simple to say [the real-world data] is not good enough,” she said.
 
Draft guidance on EHR, claims data, draft guidance on registries, draft guidance on data standards, draft guidance on RWE/RWD in regulatory decision making, Annals of Oncology review