Director's Message: National Cancer Policy Forum: Computational Tools for Diagnosis and Decision-making

"Clinical Applications of Computational Methods in Precision Oncology" was the topic of the National Cancer Policy Forum at the National Academy of Medicine on October 29-30, 2018 [View workshop speakers and agenda]. I would like to summarize key observations, and provide you with an update on the critical issues and current barriers to implementation that were voiced during the workshop. These issues mirror those we at the Case Comprehensive Cancer Center face in collecting, recording, interpreting, and implementing decisions, which are based on computational analytics of imaging, digital pathology, genomics, and extraction of the EMR that are the focus of our clinical investigation. 

The current state of computational extraction and aggregation of these forms of data remains fragmented and in need of data standards, data reporting validation, use of independent validation data sets, and data set reporting to the FDA. Review and approval by the FDA of computational models for clinical decision-making, both in and beyond clinical trials, will require these same elements.

Presented examples of machine learning from these data sets include: improved genomic diagnosis (Dr. Mia Levy, Vanderbilt), risk stratification (Dr. Howard McLeod, Moffitt), decision-making algorithms (Dr. George "Holt" Oliver, Parkland Center for Clinical Innovation) and even in silico clinical trial design (Dr. Pratik Shah, MIT). Obvious utilities are the national clinical genomics trials - NCI MATCH and ASCO TAPUR.

Likewise, the FDA presented a perspective for reviewing IND and device approvals for computational tools, standardized analytics, risk stratification tools, and genomic calls that predict clinical trial utility. Validation tools will be essential, as noted by Dr. Giovanni Parmigiani (Dana Farber).

Apple, Google and Amazon are developing Application Programming Interfaces (APIs) that serve as the conduit for our patients to download My Chart and equivalent EMR data, which is then used by the patient. Patients can use this data as a reference for care received, for transfer of care, and transmission of their health record to researchers. Soon, these APIs will communicate genomic data for interpretation. A number of cancer centers are currently uploading such information, creating a mobile, digital genomics conduit between patient, clinician and investigator.

More from this workshop will be published in a year or so, but you now have a preview of the global issues related to computational tools for diagnosis and decision-making.