| dc.description.abstract | We studied archival tissue samples of oral tongue squamous cell carcinoma (OTSCC) for the expression of genes that have been identified, by cDNA microarray studies and previous studies, to be highly discriminatory for lymph node spread. We performed immunohistochemistry on samples from 186 patients with squamous cell carcinoma using the following antibodies: CD44V9, COL5A1, DSG3, FAS, FTH1, IL22RA1, Ivl, KIF2A, Laminin 5, Llgl2, MMP8, MRP2, ODC1, P4HA1, P16, P53, S100a9, SRP19, and VEGF A.
VEGF expression, in the periphery of a tumour, predicted worse overall survival (60 vs 80% 5-year survival p=0.001). SRP19 expression in the centre (56% vs 78% 5-year survival p = 0.004) and periphery of tumour (60% vs 78% 5-year survival p = 0.015) predicted significantly worse overall survival.
VEGF and MMP8 positivity, in the periphery of a tumour, and P4HA1 negativity, in the centre of a tumour, were associated with lymph node metastases.
We constructed a logistic regression model which is far superior to the standard model of using tumour thickness above 5mm to decide on a neck dissection. It incorporates VEGF peripheral cores (+), P4HA1 central cores (-), MMP8 peripheral (+) and tumour thickness (Amen signature) to provide overall accuracy of prediction of 69.1% vs 58.1% for tumour thickness alone in the full sample. Combining our antibody predictors with tumour thickness increases accuracy to 69.1% from 65.7% using the three antibodies alone.
The superiority is particularly pronounced when comparing our signature’s accuracy of 73.0% in positive prediction of metastasis compared with tumour thickness of greater than 5mm with 53.2% accuracy in predicting metastasis. The use of the traditional tumour thickness predictor would result in an unnecessary neck dissection in 46.5% of patients, predicted to have nodal metastasis, vs 27.0% using the Amen signature. | |
