A Multiparametric Gene-Based Diagnosis Model for HNSCC

• Published in: Otolaryngology-head and neck surgery 2008-08, Vol.139 (2), p.P88-P89
• Main Authors: Worsham, Maria J., Lu, Mei, Sethi, Seema, Kapke, Alissa, Chen, Kang-Mei, Havard, Shaleta E., Benninger, Michael S.
• Format: Article
• Language: English
• ISSN: 0194-5998; 1097-6817
• DOI: 10.1016/j.otohns.2008.05.490

Problem A current shortcoming in the diagnosis, prognosis and treatment of HNSCC is a lack of methods that adequately addresses the complexity and diversity of the disease. Diagnostic and prognostic marker systems based on single parameters have generally proven inadequate. Multiparametric methods, which rely on many pieces of information, are ideally suited to the grouping of tumor subtypes, identification of specific patterns of disease progression, and in predicting clinical outcomes. Methods In a retrospective multi-ethnic primary HNSCC cohort drawn from a primary healthcare setting, and constructed through re-review of the primary biopsy, gene alterations (104 genes) and clinical variables (9 histopathology and 3 demographic variables) were evaluated as predictors of stage (TNM, early versus late). Statistical analysis compared logistic regression and Classification and Regression Tree (CART®) analyses to derive the most predictive model, assessed using receiver operating characteristic (ROC) curve analysis. Results Considering all clinical and gene variables for the 360 primary HNSCC study cohort, the multivariate logistic regression model retained only tumor grade, sample type (radical dissection) and their interaction with ROC as 64%. CART® generated a multivariable model with 12 variables: clinical variables of age, pattern of invasion, and tumor grade and gene variables of TP53, F3, TFF1, CDKN2A, KIAA0170, HS222808, TANK, MYC, and UTY1, with an ROC of 0.82%. Conclusion A group of clinical variables in multiparametric combinations with molecular alterations discriminated early and late stage HNSCC. CART® improved the model's performance. Significance Validation of this initial multiparametric strategy for predicting late stage HNSCC comprising several genes and clinical factors, currently underway, should yield a multiparametric, comprehensive genome-wide molecular algorithm integrated with clinical risk factors in order to refine HNSCC diagnosis and prognosis associated with clinical and pathological staging to aid in the clinical management of patients at the earliest stages. Support NIH R01 DE15990.