Diagnosis early in the natural history of disease promises to lead to improved health management. However, over-diagnosis needs to be avoided. Valuable lessons learned from cancer will be instructive for other chronic diseases. The integration of Information from multiple disparate biological compartments to generate informative probabilistic models will be paramount. How the healthcare system determines applicable evidence level and rewards value with reimbursement will be critical.

Many recent clinical diagnostic tools have been developed using machine learning and must be evaluated using a 3-step process of deriving, validating, and establishing the clinical effectiveness of the tool. Several key considerations are appropriate machine learning methods used for the type and size of data, proper tuning of prespecified settings (called hyperparameters) on a dataset independent of the final validation set, and a rigorous reference standard.

The U.S. has the highest per capita healthcare costs, with substantial financial and clinical variability due to over- and under-use of mispriced technologies (“moral hazard”). Physician-led professional societies are creating/owning quality metrics for assessing performance. Under MACRA, performance also is assessed based on costs for an episode of care. To succeed under these new incentives, physicians will need risk-based algorithms assessing technology quality and costs; representative examples to be presented.

Medicare has historically been far behind the curve in providing coverage for preventive services, especially novel diagnostic tests. Medicare’s approach is so slow, it risks actually discouraging investing because of the long and slow policy pipeline. Medicare can adopt new preventive benefits for diagnostics if they are directly legislated by Congress. In addition, Medicare can adopt new diagnostics through the year-long National Coverage Determination (NCD) process, if they have been previously endorsed by the U.S. Preventive Services Task Force. These serial steps can add many years to the adoption of new technology and depend on irregular multi-year timelines in the case of USPSTF. We will discuss the pros and cons of the status quo, best practices, and case studies for coping with the status quo, and the outlook for new policy that would better encourage much-needed innovations.

For many diseases, population screening is essential to improving patient survival. Despite the promise of blood-based “liquid biopsies," progress has been limited necessitating new multi-omics approaches incorporating multiple technologies. We discuss new approaches to biomarker discovery, algorithm development, and clinical validation using mechanistic, statistical, and machine learning approaches for handling of multi-omics data. We further discuss the need to improve clinical decision support systems to facilitate clinical decision making.

GRAIL is a healthcare company whose mission is to detect cancer early, when it can be cured. GRAIL is focused on alleviating the global burden of cancer by developing pioneering technology to detect and identify multiple deadly cancer types early. The company is using the power of next-generation sequencing, population-scale clinical studies, and state-of-the-art computer science and data science to enhance the scientific understanding of cancer biology, and to develop its multi-cancer early detection blood test.

Non-alcoholic fatty liver disease (NAFLD) affects approximately 25% of adults and 10% of children in the U.S. and is associated with obesity, type 2 diabetes mellitus, dyslipidemia and hypertension. Severe clinical outcomes include decompensated cirrhosis and hepatocellular carcinoma. The current diagnostic standard for non-alcoholic steatohepatitis (NASH), the more advanced form of NAFLD, is based on liver biopsies. Early diagnosis of the higher risk individuals is paramount. Representation and active engagement of scientific experts from all stakeholder groups in a non-competitive environment increases clarity and standardization while decreasing uncertainty. Lessons learned will be summarized.

Breast cancer screening by mammography, which began in the U.S. about 40 years ago, has led to significant increases in the incidence of early-stage breast cancers, including ductal carcinoma in situ (DCIS), also called precancer or stage 0 disease. However, the expected reciprocal decrease in subsequent late-stage breast cancers was not found. It is now clear that some screen-detected breast “cancers” are indolent lesions without significant malignant potential.

Saliva is a bodily fluid that we produce ~600ml per day and harbors multiple omics constituents that can be harnessed non-invasively and painlessly for personalized and precision medicine, making it ideal for liquid biopsy applications. Yet, PCR-based technologies cannot detect ctDNA in saliva samples, whereas an emerging liquid biopsy platform “Electric Field Induced Release and Measurement (EFIRM)” can consistently detect ctDNA from NSCLC patients with actionable mutations in plasma and saliva with concordance of 95%+ with tissue/biopsy-based genotyping, including early stage lesions.

Early detection of cancer has the potential to significantly reduce cancer deaths. Liquid biopsies provide an opportunity to develop tests for the detection of multiple-types of cancer in a single test. We will discuss the opportunities and the challenges of developing and utilizing such test.

Exosomes reflect promising cancer biomarkers due to their abundance in biofluids, protein, and RNA contents reflecting parental cells, and stability in circulation. Pancreatic ductal adenocarcinomas (PDAC) represent highly lethal cancers often presenting at advanced stages. Integrating exosomes into monitoring programs for PDAC could improve clinical outcomes. This talk will discuss novel nanosensors we developed to analyze exosomes in blood and recent translational strategies to position them into clinical workflows.