Fred Ashbury, PhD, Dawn Walter, MPH, and Shawn Lach, BSBA
Real-world data (RWD) are increasingly used to support agile decision making in the health sector, particularly to improve quality and timeliness of clinical decisions, to identify potential therapeutic options, and to evaluate clinical and related interventions. In healthcare, RWD are typically defined as data from electronic medical record platforms (EMRs), payor and administrative claims, cancer registry data, as well as information that patients upload from personal health devices.
Fred Ashbury (FA): Given the ascendance of RWD in oncology, what benefits might be expected to support cancer care?
Dawn Walter (DW)/Shawn Lach (SL): RWD can play an important role in cancer research by augmenting randomized controlled trials (RCTs). The RCT is the gold standard study design for determining the safety and efficacy of novel cancer treatments, but it has many challenges1. RCTs are expensive, take a long time to implement, and typically experience significant lag time in reporting results to the medical community. Comprehensive RWD, on the other hand, can be less costly to appropriate and analyze, and new insights can be gleaned more quickly.
Another benefit of RWD is the applicability to wider populations. Currently, only a small percentage of cancer patients enroll in RCTs and because of this the generalizability of RCT results to certain populations (patients of older age, of certain racial and ethnic groups, or with more complex disease progression, for example) is questionable. RWD are based on broader and more diverse experiences, and as such, can better illustrate the difference between a new cancer treatment’s efficacy in a tightly controlled RCT vs. its effectiveness in real-world clinical use.
In addition to amplifying research findings, RWD can be used to help improve quality by allowing for measures of appropriate decision-making.2 In a world where new cancer diagnostic tests, biomarkers, and therapies are entering the market at a rapid rate, RWD can be used, for example, to determine providers who are using current standard of care treatments for their patients and those who may not be complying with guidelines and policies. RWD, therefore, provide an opportunity for continuous quality improvement.
(FA): What challenges do we face in accessing and gaining insights from RWD?
DW/SL: While RWD can offer a lot of potential insights to help in the cancer fight, it has some shortcomings. Many real-world databases being used today in the oncology setting suffer from missing data and inherent biases. The Medicare database, for example, an enormously popular source for RWD analyses in the US, represents a patient group that is less racially diverse and more affluent than the actual US population. As we get better at merging RWD across multiple platforms, especially in the US’s fragmented healthcare ecosystem, the quality of RWD will continue to improve.
Many EMRs currently used in oncology practices allow important patient-specific, clinical data to remain buried in unstructured notes, lab reports, or other documents often scanned and stored leaving no capability of mining these data. RWD need to comprise discrete data points, not text, document scans, or the like, to generate potentially actionable insights. Abstracting useable information from unstructured data fields is challenging and resource intensive and most community oncology programs lack the resources to do it. Thus, a great deal of important data is being left on the table instead of being used to drive cancer treatment insights.
These RWD challenges point to another potential problem. When RWD have missing fields or possible biases it means that performing high-integrity investigations with these data requires a thoughtful and rigorous approach to biostatistical analyses. Again, in the community oncology setting, these trained resources are often not available. As the reliance on RWD increases, the research community must continue to uphold the highest standards for RWD-driven research and disseminate findings quickly and clearly to the entire oncology community. (FA): How can artificial intelligence (AI) facilitate the use of RWD to improve insight generation?
DW/SL: AI allows for faster and more accurate processing of the unstructured data we described above. Notes and scanned documents can be parsed using AI tools, recovering important RWD, including patient-reported outcomes and side effects, pathology findings, and complete molecular testing results.
As EMRs move toward capturing discrete data points instead of unstructured text, AI can facilitate data capture through appropriately placed alerts and the application of rules to “remind” clinicians to record pertinent data, such as diagnosis, stage, tumor markers, genetic and genomic results, medications, treatments, and so on. Properly architected clinical decision support platforms based on AI should be integrated into oncology EMR platforms to support the work of oncologists, helping them spend less time tracking information and more time with their patients.
AI can also improve the quality of cancer care through the application of machine learning. With RWD in place, AI systems can be trained to improve accuracy in diagnostics, treatment, supportive care, follow-up, rehabilitation, palliative, and end-of-life care. Recent advancements in molecular oncology, for example, include applications that can predict future cancer diagnoses based on patients’ vast gene data and can help determine the most appropriate treatments and supportive care to facilitate better patient outcomes. AI is a critical resource for decision support; indeed, the FDA has begun approving medical devices that incorporate AI. These devices are linked to studies that are rapidly improving the standard of cancer care3.
(FA): Randomized controlled trials will remain an important method in proving the safety and efficacy of new cancer treatments, but as RWD develop in breadth and quality, and as researchers become more adept at mining these data, RWD will have an increasingly positive impact on improving cancer care. Appropriate data capture is key to optimizing the potential that AI can deliver. References 1Nazha, B., Yang, J. C.-H., & Owonikoko, T. K. (2020, November 26). Benefits and limitations of real-world evidence: lessons from EGFR mutation-positive non-small-cell lung cancer. Fu- ture medicine. Retrieved June 14, 2022, from https://doi.org/10.2217/fon-2020-0951
2Rudrapatna, V. A., & Butte, A. J. (2020, February 3). Opportunities and challenges in using real-world data for Health Care. The Journal of Clinical Investigation. Retrieved June 14, 2022, from https://www.jci.org/articles/view/129197
3Shimizu, H., & Nakayama, K. I. (2020, March 4). Artificial Intelligence in oncology - Shimizu - Wiley Online Library. Wiley Online Library. Retrieved June 14, 2022, from https://onlineli- brary.wiley.com/doi/10.1111/cas.14377
Fredrick D. Ashbury, PhD
Chief Scientific Officer, VieCure Professor (Adj), Department of Oncology University of Calgary Professor (Adj), DLSPH, University of Toronto
Dawn Walter, MPH
Senior Data Analyst, VieCure
Shawn Lach, BSBA
Data Analyst, VieCure