A Differentiation based immunohistochemical classifier that is prognostic for head and neck tumor patients. Poster at 2007 United States and Canadian Academy of Pathology (USCAP) Annual Meeting

DT Ross ¹ , BZ Ring ¹ , RS Seitz ² , RA Beck ² , S DeFoe ² , F Robert ⁴ , MT Schreeder ⁵ , CH Chung ⁶ , CS Kong ⁷ and QT Le ³ . ¹ Applied Genomics Inc., Burlingame, California, United States; ² Applied Genomics Inc., Huntsville, Alabama, United States; ³ Radiation Oncology, Stanford University, Stanford, California, United States; ⁴ Hematology and Oncology, University of Alabama, Birmingham, Alabama, United States; ⁵ Comprehensive Center of Huntsville, Huntsville, Alabama, United States; ⁶ Medicine, Vanderbilt University, Nashville, Tennessee and ⁷ Pathology, Stanford University, Stanford, California, United States.

Background: Gene expression profiling of tumors has revealed novel tumor subtypes within cancers derived from the same tissue and similarities between subtypes originating from different tissues. For example, the basal subtype of breast cancer expresses a signature in common with squamous cell carcinoma of the lung and a subtype of head and neck carcinoma (Chung et al., Cancer Cell, 2004). We have endeavored to translate gene expression signatures into IHC assays for distinguishing tumor subtypes.

Design: Tissue arrays were constructed from NSCLC tumor patients seen at the CCIH and the UAB, and head and neck patients seen at Stanford University. A five antibody classifier (TRIM29, CEACAM5, NCSTN, ABCG2, TLE3) trained on the CCIH cohort for distinguishing adeno from squamous carcinoma of the lung, was validated on the independent UAB cohort. Cox proportional hazard terms and classifier cut-offs were pre-specified using the CCIH NSCLC data and this model tested for classification of the head and neck patients.

Result: The classifier trained to separate adenocarcinoma from squamous carcinoma of the lung divided the H N patients into two distinct groups with statistically significant outcome differences. The sensitivity of the test for detecting patients at high risk of death due to disease was 83% and the specificity was 57%.

Conclusion: The clinically significant similarities defined by gene expression patterns or IHC staining patterns between carcinomas arising in different tissue origins suggests that there is a differentiation based classification of carcinoma that transcends tissue of origin. This classifier, as well as classifiers trained to directly predict outcome in head and neck tumors, may help select between therapeutic options for treating this notoriously aggressive tumor.

	(1.8 MB PDF)