Introduction
Precision medicine is the commitment to deliver the right therapy to the right patient at the right time. This is a challenging goal, especially in oncology, given the wide diversity of diseases and patients. To address the challenges, drug developers are increasingly analyzing real world data (RWD) to generate real world evidence (RWE) to complement what can be learned from traditional randomized, controlled clinical trials (RCTs). With an esteemed panel of experts—Siraj Ali (Lunit), Keith Flaherty (MGH), Eleonora Iob (PHARMO), and Saurabh Sewak (Foghorn)—Dennis Chang (Lumanity) led a whirlwind tour of a variety of RWD/RWE types and uses in precision oncology.
Clinicogenomic RWD
A tremendous amount of data is already being generated from the integration of molecular testing data (including next-generation sequencing [NGS] data) with clinical outcomes data found in electronic medical records (EMRs).
Drug developers can leverage such clinicogenomic data to:
- Gain insights for clinical trial design and development strategy: e.g., anticipating challenges in enrollment by understanding how rare a biomarker truly is in the target patient population. MSI-H (microsatellite instability-high) status in colorectal cancer (CRC) is a good example, where the rate in metastatic CRC is far lower than in non-metastatic CRC.
- Gain insights for commercial assessment and strategy: e.g., developing a clearer picture of the unmet need—and thereby a clearer business case—by seeing how a biomarker-defined population responds to standard of care treatment. For instance, the responsiveness of KRAS-mutant NSCLC to immune checkpoint inhibitors (ICIs) shows that the bar to beat in first-line treatment is much higher than for ICI-resistant subpopulations of NSCLC.
- Support regulatory approval in conjunction with clinical trial data: For example, indication expansion by showing real-world efficacy in a population not previously included in the pivotal trial. An example would be the approval of palbociclib for male breast cancer.
Patient-reported Outcomes as RWD
Although not yet in widespread use, there are mobile apps that enable real-time communication from patients to clinicians on their experience on therapy. With these apps, patients can report symptoms and treatment-emergent adverse events with greater resolution and reliability than in office/clinic visits.
Drug developers can harness such patient-reported RWD to:
- Gain insights on variations in patient experience and outcomes: to enable optimization of dosing and AE management
- Generate evidence on the impact of therapy on quality of life: to complement clinical trial data
For these applications, the patient-reported data would also need to link to other data—at minimum which therapies are being used and ideally to other clinical information (comorbidities, etc.).
Importantly, clinicians can also harness the data to improve patient care. In fact, RCTs have shown that the use of such apps can produce significant, meaningful improvements in survival. A key to real-world implementation is that the apps have to be simple and not burdensome to use: a small number of questions answered every day is far more practical than an extensive questionnaire that will not reliably be completed.
Digital Pathology and Radiomic RWD
Drug developers can use artificial intelligence/machine learning (AI/ML) to analyze digital images of pathology slides and/or radiology scans to:
- Gain insight into correlates of treatment response and outcomes: to generate hypotheses for future drug and biomarker development
- Discover novel predictive biomarkers that can be used to select patients for trials or treatment.
Published studies show proof of principle that a standard H&E stained pathology slide contains invaluable information that AI/ML can use to predict response or non-response to immunotherapy. There are studies showing that routine CT scans can be mined to make similar predictions. Moreover, the combination of these and other modalities may have even greater predictive power.
Conclusions
We have reason to be excited. The many use cases listed above, for a variety of RWD types, are already compelling, and the trend is toward even greater abundance of data, greater integration of multiple data types, and greater analytic power in the future. However, we must also acknowledge that there are challenges to overcome. For example, we must improve data quality and completeness, reduce biases and increase standardization of data collection and analysis, manage the complexity of data, and improve data access—as many of the most powerful datasets are behind costly paywalls. Public investment and public-private partnerships are likely critical to realize the full potential of RWE in precision oncology.
