Abstract
To develop and validate a lab-on-a-chip method to analyze exfoliated cells collected from a non-invasive brush biopsy in a NIDCR-sponsored 950-patient multi-site international clinical trial.
Three groups of subjects (PMOD, OSCC, and normal controls) were enrolled at three American and one European sites in this prospective study. Matched brush and scalpel biopsies were collected from all non-control subjects. Three-color immunofluorescence analysis of cellular samples (an average of ∼2000 cells/sample) for up to 25 cytomorphometric parameters and biomarker expression (of seven candidate biomarkers) were generated in an automated manner and processed using sophisticated image analysis algorithms based on pattern recognition techniques and advanced statistical methods.
At the time of this writing, 280 samples from the training phase have been processed, and preliminary data for single parameters exhibit high discriminating ability when compared to gold standard histopathology, to differentiate between each category of normal, benign, dysplastic, and OSCC samples in a dichotomized way. In this talk, we expect to report final results from the full 950-patient trial, with completion anticipated in March 2013, with samples from 2/3 of the patients to serve as training data, and 1/3 of the data analyzed blindly in the validation stage. The performance of single parameters as well as various multivariate models combining morphological and molecular biomarkers, measured as AUCs relating to the ability to discriminate between each disease category will be presented, as well as in low versus high risk models for both the training and validation sets.
Based on very promising preliminary results, we expect the outcome of this study to produce a tool based on LOC analysis of exfoliated cells collected non invasively and capable of discriminating between benign and dysplastic or OSCC lesions with high sensitivity and specificity with multivariate panels of molecular and cytomorphometric parameters.