Precision oncology requires a robust understanding of how genomic information relates to clinical outcomes and treatment response. To make this connection, we need to integrate multiple sources of data into single databases that have the power to inform new clinical insights.
This has historically been difficult because longitudinal clinical and genomic data have largely been aggregated in a handful of academic centers. Furthermore, “real world data” from the community – where 80% of clinical oncology care happens – is rarely incorporated. We need much more comprehensive datasets if we are to discover the genomic signatures in specific patient populations that will drive precision oncology forward. Doing so requires sharing data across organizations and engaging the local community settings where most patients receive their care, to better characterize cancer and develop new treatments.
A novel paradigm for generating a clinico-genomic database
At this year’s American Society of Clinical Oncology (ASCO) Annual Meeting, we are presenting a new way to rapidly generate and validate a large, research-grade, longitudinal clinico-genomic database. In partnership with Flatiron Health, we linked Foundation Medicine’s genomic data with clinical data from Flatiron’s electronic health records through a HIPAA-compliant, de-identified, IRB-approved process. This approach leverages important real-world insights from community practices, as well as academic centers, and allows us to explore the significance of molecular alterations in a more comprehensive fashion.
With this study, we validated this dataset’s ability to discover correlations and relationships between genomics and outcomes in molecular oncology. In a set of approximately 2,100 non-small cell lung cancer (NSCLC) patients, we confirmed existing knowledge of the molecular drivers of NSCLC and characterized predictive markers of response to immunotherapy. We found that the genomic characteristics of the NSCLC tumors in our database were consistent with prior studies in large populations, including The Cancer Genome Atlas. In addition, we recapitulated known relationships between certain driver mutations and overall survival, detecting predictors of response to targeted therapies.
This dataset also allowed us to demonstrate the utility of comprehensive genomic profiling, even for commonly tested mutations such as EGFR and ALK. By comparing prior (single marker) testing with our FoundationOne® assay, we found many cases where single marker testing had been negative for EGFR or ALK, but FoundationOne found an actionable mutation in that same gene. In fact, this happened in approximately 25% of all EGFR and ALK positive cases. As a result, many of these patients were able to be treated with the appropriate targeted therapy. This demonstrated the value of FoundationOne for findings such as EGFR and ALK that can be tested by single marker tests, and its ability to provide better accuracy for actionable findings that impact clinical care.
We were also able to characterize more recent findings in the literature, including the ability of tumor mutational burden (TMB) to help predict response to checkpoint inhibitor immunotherapy in several types of cancer, including NSCLC. In this dataset, we found that TMB predicted response to an FDA-approved checkpoint inhibitor, even in PD-L1 negative populations, demonstrating its potential to help oncologists prescribe more precise treatment.
Findings like these validate the use of this clinico-genomic database as a way to rapidly find answers to our questions about the genomics of cancer biology. With the results of this validation set, we are excited about what this clinico-genomic database can do to inform future research and speed advances in the field. Indeed, my colleague recently wrote about findings presented at the last American Association of Cancer Research (AACR) annual meeting which used this database to discover potentially novel biomarkers for cancer immunotherapies in NSCLC.
Extending the value of a clinico-genomic database
In the past year, this database has more than doubled in size to approximately 25,000 patients, and we expect continued growth over the next year. Over time we will have stronger and more powerful databases from which to extract important findings and advance the field of precision oncology.
We can also expand this methodology to disease areas such as melanoma, breast, colorectal, prostate and ovarian cancer, where comprehensive genomic profiling increasingly provides insights for treatment and drug development. As we link more genomic information to clinical outcomes, this tool can help direct novel target discovery through the identification of new driver mutations across a population. Clinical trial design may also be improved by using this information to enroll more specific patient populations that are most likely to respond to an investigational drug.
Through increased collaboration and data sharing across many different types of institutions --academic medical centers, community oncology providers, healthcare technology, molecular insights and pharmaceutical companies -- we can all drive the progress that leads to better cancer care.
Learn more about the clinico-genomic data at ASCO at our poster discussion on June 5th:
Development and validation of a real-world clinicogenomic database. (Abstract #2514, Poster Discussion Session: Developmental Therapeutics—Clinical Pharmacology and Experimental Therapeutics, 11:30AM-12:45PM)