Abstract

Cell-based Point-of-Care Oncology Tool (POCOT) For Precision Medicine

John McDevitt, Chair, Department of Biomaterials, New York University College of Dentistry Bioengineering Institute

The U.S. health care spending in 2016 reached $3.3 trillion or $10,348 per person. Health spending now accounts for 17.9% the US Gross Domestic Product. Chronic diseases like cardiac and cancer account for the majority of the costs in this area. Further, contributing to the $2.8B total costs for each new pharmaceutical is the cumbersome decade long approval process. These somber statistics seem to be offset by the promise of cutting edge biomarker developments with 157,000 biomarker papers published in last decade, yet only ~1 biomarker per year gets approved by the FDA. Likewise, the true potential for precision medicine is tampered by an infrastructure and incentive structure that slows the arrival of the benefits of precision medicine. To overcome these limitations, transformative new tools are desperately needed to enable a new era of medicine through research, technology, and policies that empower patients, researchers, and providers to work together toward development of individualized treatment. This talk features the development, optimization and validation of the first cell-based point-of-care oncology tool (POCOT) for precision medicine. Using single-cell data collected non-invasively from cytology samples of prospectively recruited patients with gold-standard-confirmed diagnoses, a series of predictive models were developed and validated resulting in a “continuous numerical risk score”. Model development consisted of: (1) training binary classification models for each diagnostic class pair, (2) pairwise coupling to obtain diagnostic class probabilities, and (3) a weighted aggregation to obtain a final risk score on a continuous scale.  Diagnostic accuracy based on optimized cutpoints for the validation dataset ranged from 76% for Benign lesions, to 82.4% for Dysplastic, 89.6% for Malignant, and 97.6% for Normal controls. The weighted aggregation of pairwise probabilities into a continuous variable was associated with a minor decrease of 6.4% for overall accuracy and demonstrated a strong positive relationship with diagnostic severity (Pearson’s coefficient = 0.805 for validation dataset). Frequencies of 5 cellular phenotypes recorded for 3 different ranges of the numeric index score agreed with common trends observed in conventional cytopathology. Finally, a simulation demonstrated that the numeric index can respond to changes in as few as 1% of the cells within a cytology sample, which is a necessary requirement for future evaluations of its utility for lesion monitoring.  The first cell-based point-of-care oncology tool for precision medicine is demonstared in the context of a continuous numeric index for potentially malignant oral disorders. These efforts result in a sensitive, accurate, and non-invasive method for enabling monitoring of these lesions. Continuous quantitative severity indices devoid of human subjectivity and categorization bias could have significant implications in a variety of medical decision making situations where microscopic evaluations and grading of biopsy samples by pathologists currently determine diagnostic, prognostic and treatment decisions. As such, this numeric index has the potential to monitor disease progression, recurrence, and the need for therapeutic intervention.