By Warren Whyte, PhD; Jeff Elton, PhD; Charles Lagor, MD, PhD;
and Sudip Bhandari, PhD
Pharma leaders have a critical opportunity to take giant leaps forward to overcome decades of biases in clinical research — most unintentional or non-conscious — that have led to poor representation of certain racial, ethnic, and economic groups in Real-World Data (RWD)-driven research. To realize this opportunity, leaders need to throw out their old playbooks and welcome new data-driven study design approaches. We know which
sub-populations are most at risk for specific diseases. Therefore, we can accept that there is no standard-of-care representation that can be considered complete, or even accurate, without these sub-populations being in our data sets and a focus of our studies.
At ConcertAI, we pursue those goals through a multidisciplinary initiative called Engaging Research to Achieve Cancer Care Equity, or ERACE, which aims to better understand cancer disparities to eliminate them. We are harnessing efforts at the national level, beginning with the push in 2009 for ubiquitous electronic medical records (EMRs) to the 2021 guidance documents on RWD issued for comment. These efforts have set the stage for the growing application of Artificial Intelligence (AI) in clinical trials. In the coming two to three years, every major pharmaceutical company and clinical research organization will be rebuilding and re-syndicating their data. They will also be bringing AI and data science skills to bear for the generation of insights and assurance of generalizability across aggregate and sub-populations.
So, the ‘bar’ on what will be considered research ‘fit for evidence generation’ is going up. Preliminary FDA guidance issued in late 2021 reflects a desire to move RWD from a set of supporting analyses to a ‘strong form of evidence’ that can be an element of regulatory decisions. RWE analyses should be more reflective of all patients, including communities traditionally marginalized and those disproportionally impacted by disease. For the biopharmaceutical companies conducting this research we would like to advance four principles of ‘Equitable RWE’:
Get a new perspective on old data and consider starting anew. The data you have may not be the data you need and may unwittingly perpetuate biases. If you rightly assume that legacy and unintentional bias underlie most existing data, you need to understand the bias, redefine research questions, and then with a new purpose build or partner around the building of data sets appropriate to those questions.Too often post-approval RWE studies emphasized data that was the focus of the original trial population only, whereas the trial may have under accrued specific racial and ethnic subpopulations; largely had academic medical center participation versus community; excluded the morbidly obese, etc. Yet, these can be a significant proportion of the patients being treated post-approval. Getting a new perspective and starting anew may require non-traditional data sources that incorporate social determinants of health, ancestry/nativity information, etc.
Make sure your data and RWE study design follow principles of generalizability, trust, and transparency. Generalizability, trust, and transparency are foundational to research and need to be considered even during feasibility or ad hoc analyses that may inform the study design. Documenting these considerations, steps, and pre-study design results are increasingly a required practice. Data should be representative of the patients impacted by the disease or treated with the specific therapeutic being studied. The study design may need to accommodate the standard of care in the settings where the patients are being treated — even if the approaches, diagnostic technologies, etc. diverge from the trial locations. Patient perspectives and experiences should also be considered. At ConcertAI we have access to almost 6 million patient records, from academic, regional health systems, community centers, 10 different EMR technologies, and possess the ability to integrate all aspects of a patient’s record including EMR, laboratory data, radiological images from PACS, etc. This advances our ability to assure limited to no bias across sub-populations and across the key dimensions of treatment, survival, adverse events, and other clinical outcomes of interest.
Data are never generic, rather they are purposeful to the research question, and improve over time. Medical claims data predominated the data sources used for RWD studies five years ago. EMR data have since displaced claims data in the most rigorous applications focused on clinical versus economic outcomes. Now, EMR, claims, genomic, and radiological imaging data are increasingly accessible, can be confederated, and de-identified. This constant review and discussion should go all the way from study conception and planning, continue during the phase when patients are registering, and last over the entire life cycle of an experimental therapeutic.
Ensure your partners have the same ethical orientation and focus on equity as you do. While health disparities have long been an unfortunate feature of cancer care, the current level of research activity to bring insights and treatment options to all patients, races, and economic groups is simply insufficient. The only way we can take giant leaps forward in eliminating inequities that span the entire care continuum is if we work collectively and cross-disciplinarily. Research is a collaborative endeavor, so having partners aligned with us in investing in research activity to improve care and outcomes in marginalized communities is a necessary catalyst toward much-needed progress.
Through the application of all these principles, we not only accelerate insights, but we make significant progress towards the elimination of the research disparities that underpin some health inequities. The ‘foundation’ of equity is data, the ‘superstructure’ our methods, and the ‘results’ our deep intentionality on those with the highest clinical needs and where we need to reverse historical inequities. In so doing we create ‘equity’ as an integral part of both research culture and intent.
