uccesses in understanding disease biology lead to the discovery and use of complex biomarkers in drug development, characterizing patient disease and making patient care more effective. High-technology methods help address the opportunities and challenges of using complex biomarkers, most often in clinical cancer research geared toward precision medicine. These technologies are less invasive and/or provide more data than standard tissue biopsy and imaging.
Technologies for Characterizing Complex Biomarkers
Tumor heterogeneity is a challenge that is not fully addressed by LBx, but technologies such as molecular imaging and digital pathology are pathways for biomarker development that provide the spatial molecular mapping of tumors and the TME. Molecular imaging probes are available for many common cancer mutations. Like LBx, molecular imaging is noninvasive, but its use has been limited by high cost, requirements for multiple scans, and limited access.
Digital pathology provides advantages over standard tissue biopsy in that the tissue may be turned into virtual slides allowing for multiple analyses, sharing access to the slides among collaborating researchers, and improving diagnoses and outcomes by providing capabilities for quantitative evaluation of tissue parameters. Digital pathology has been called “tissue-omics,” recognizing its similarity to the other ’omics in terms of the type and amount of data generated.
Collection and analysis of real-world data, i.e., patient data from sources other than specialized CRFs designed for capturing data from a clinical trial, is of high interest to researchers and regulators. Development and implementation of these data, including that from electronic medical records and digital health technologies (e.g., wearables), may enrich the understanding of adverse events, efficacy, and other indicators of patient status and provide data to complement other assays and biomarkers.
All the technologies described produce “Big Data” and increase the need for data sciences in clinical cancer drug development and care. Data science (e.g., AI approaches) may be applied to synthesize multiple data sources to help with therapy and clinical trial optimization, developing biomarker-based innovative endpoints, and more. For example, the benefits of genomic sequencing can be realized by using data science to create high-dimensional, complex molecular signatures. Digitization of histopathology specimens using algorithms that segment specimens to get patient subtypes is another area where data science might be used to generate biomarkers.
Carefully curated resources are needed to generate relevant data for researchers. Examples of available resources include the American Association for Cancer Research’s Project GENIE, the American Society of Clinical Oncology’s CancerLinQ, and the National Cancer Institute’s Genomic Data Commons (GDC). More specifically, the GDC houses the Genome Cancer Atlas (TCGA) and its follow-on Human Tumor Atlas Network (HTAN). Beyond genomic data, HTAN provides different analysis methods including spatial ’omics measurements and single-cell RNA sequencing.
Despite the opportunities, barriers exist to using Big Data in clinical research. These include reluctance to share patient data, incompatible workflows, regulatory issues, and the need for well-governed partnerships. The FDA has shown increasing interest in complex biomarkers and has provided criteria for regulatory approval. These include definition of the biomarker and intended use (which determines the regulatory pathway), assay sensitivity and specificity, reliability of results, and factors in detectability (different assays may pick different populations).
Standards Are Needed
- The FDA has recommended the development of quality-control materials (QCM) for LBx that demonstrate functional comparability to clinical specimens.
- The Foundation for the National Institutes of Health (FNIH) is evaluating QCM with the intent that they can be used to improve the interoperability of LBx assay data.
- The BLOODPAC consortium has also evaluated assay reference materials.
- Friends of Cancer Research is conducting the project ctMoniTR to explore methodology and to set standards for the collection and analysis of ctDNA data.
Standardization of LBx (e.g., assay protocols, specimen collection, and data collection and analysis) is generally challenging and is the subject matter of several public-private collaborations, one of which is the FNIH International Liquid Biopsy Standardization Alliance (ILSA), designated by the FDA as a Collaborative Community, and including BLOODPAC, ctMoniTR, the FNIH QCM project, and several international associations dedicated to the development of LBx. Also, the Digital Pathology Association has a Regulatory and Standards Task Force charged with promoting standards in the research community. The Medical Device Innovation Consortium (MDIC) focuses on strategies and standards to integrate these new technologies into clinical use.
The standardization of real-world data may be the most challenging of all. Electronic medical records often have missing data and text fields with variable syntax. It is difficult to engage healthcare professionals who create and manage these records in applying standard terminology or other aids to improve interoperability of these data. Again, several initiatives are tackling this issue. One of the most prominent is the application of Minimal Common Oncology Data Elements (mCODE). mCODE provides minimum recommended standards for the structure and content of health record information.
Application of Complex Biomarkers―Early Detection and Prevention
Challenges in early disease include the high possibility of over- or underdiagnosis (from the small amounts of analyte detectable before cancer develops); the current high cost to and time burden for patients who might need frequent retesting; and requirements for counseling patient anxiety. The biggest challenge may be the analytical and clinical validation of assays, since large amounts of quality data are needed (well-defined cohorts, sufficient follow-up time in screened patients, and standardized protocols and assessments).
Nonetheless, in oncology, Multicancer Early Detection (MCED) is emerging as an exciting new technology for evaluating early disease. MCED tests are LBx that are used to evaluate DNA or proteins and other factors from cancerous or precancerous cells. The test results are analyzed by AI methods for patterns predictive of cancer across multiple targets (most often methylation patterns) and identifying the cancer’s origin. Illumina’s Galleri and the Exact Sciences tests may be the best developed. For example, the performance of Exact Sciences’ test has been evaluated in a large-scale prospective interventional study, showing that cancers in multiple organs could be detected, the majority of which have no early detection options. The Exact Sciences prototype MCED assay uses three complex biomarkers (aneuploidy, methylation, protein biomarkers) and four DNA mutation biomarkers; the test showed sensitivity and strong specificity in early-stage cancers. Additional evaluation is needed, and Exact Sciences is planning a pivotal study aimed at obtaining FDA approval. It is noteworthy that the National Cancer Institute is looking to build a clinical network for evaluating MCED tests and planning studies to evaluate their clinical utility.