A prototype tobacco-associated oral squamous cell carcinoma classifier using RNA from brush cytology

Oral cancer in the form of squamous cell carcinoma (OSCC) is typically detected in advanced stages when treatment is complex and may not be curative. The need for surgical biopsy may contribute to delays in diagnosis and impede early detection. Multiple studies of RNA from surgically obtained tumor samples have revealed many genes differentially expressed with this disease. We sought to determine whether the identified mRNAs could be used as markers by a non-invasive detection system for OSCC using RNA from brush cytology.

Levels of mRNAs from 21 genes known to be differentially expressed in head and neck squamous cell carcinoma surgical samples, compared with controls, were shown to be quantifiable in oral brush cytology samples. These mRNAs were quantified in a training set of 14 tumor and 20 non-malignant brush cytology samples from tobacco/betel nut users. With the measurement of two additional mRNAs and analysis using support vector machines algorithm for class prediction of these cancers was produced.

This OSCC classifier based on the levels of 5 mRNAs in RNA from brush cytology initially showed 0.93 sensitivity and 0.91 specificity in differentiating OSCC from benign oral mucosal lesions based on leave-one-out cross-validation. When used on a test set of 19 samples from 6 OSCCs and 13 non-malignant oral lesions, we found misclassification of only one OSCC and one benign lesion.

This shows the promise of using RNA from brush cytology for early OSCC detection and the potential for clinical usage of this non-invasive classifier.