Predicting the outcome of oral potentially malignant disorders using a composite clinical, histopathological and molecular classifier

Abstract

Oral squamous cell carcinoma (OSCC) is associated with a high degree of morbidity and mortality. OSCCs are often preceded by oral potentially malignant disorders (OPMD) which have a higher propensity to undergo malignant transformation compared to clinically normal oral mucosa. However, traditional methods such as clinical and histopathological assessment of OPMD are unable to accurately predict clinical outcome. The search to identify the perfect biomarker for prognosticating malignant transformation in OPMD is still ongoing. This study was undertaken to construct a prognostic classifier for patients with OPMD by integrating clinical, histopathological and molecular factors as well as to discover a gene expression signature that characterises OPMD with a high risk of undergoing malignant transformation. The demographic and clinical features of a cohort of OPMD patients were studied in detail. Assessment and analysis of clinico-demographic features, histopathological features, differential gene expression, loss of heterozygosity (chromosomal regions 3p14.2, 9p21 and 17p13) and DNA ploidy status were performed on archived formalin-fixed paraffin-embedded tissue material from this cohort of patients. Gene expression studies revealed several genes that had statistically significant differential expression between cases that underwent malignant transformation and those that did not. A novel gene-signature for cases that had a propensity for undergoing malignant transformation was developed. Statistical model building was performed to construct a prognostic classifier for OPMD patients. A prognostic model composed of age at diagnosis, site of index OPMD, binary oral epithelial dysplasia grading and the novel gene-signature was found to be highly prognostic of clinical outcome for OPMD patients (concordance index: 0.85). Our findings show that a molecular biomarker driven prognostic classifier outperformed conventional methods for predicting clinical outcome in patients with OPMDs. The findings from this study have also reinforced that formalinfixed paraffin-embedded tissue can be used to generate a molecular classifier with clinical utility.