PARENT SESSION
1:30 PM to 3:30 PM
Tuesday, April 23, 2002
Poster Session 20 Predictive Assays
Room: Nevada Exhibition Center
(P25-265) Gene expression profiling as a predictor of tumor response to radiation.
Burrows, Thomas*,1, Weil, Michael1, Stivers, David4, El-Naggar, Adel3, Ang, Kian2, Story, Michael1, 1 Department of Experimental Radiation Oncology, Houston, TX4 Department of Biostatistics, Houston, TX3 Department of Pathology, Houston, TX2 Department of Radiation Oncology, Houston, TX
ABSTRACT-
In a previous study, surgical pathologic criteria (SPC) were used to stratify patients according to their risk for local-regional recurrence and morbidity in a population of over 200 patients treated at MDACC for head and neck squamous cell carcinoma (HNSCC) by surgery and radiation (Ang et al., IJROBP, 2001) consistent with a single treatment protocol developed here. Unfortunately, SPC does not predict relapse rate nor the pattern of relapse (local-regional vs. metastasis). Over 400 tumor specimens with SPC and long-term patient response data have now been collected. We proposed that gene expression profiling could distinguish risk categories in this population and furthermore, it could predict the risk for and pattern of relapse in patients at high risk for recurrence and poor survival. Total RNA from a number of tumor samples was collected and gene expression patterns were developed using Research Genetics GF200 and GF211 microarrays. In our first analysis of gene expression, 6 HNSCC samples from our population were compared against 6 variant carcinomas seen in H&N cancer. Cluster analysis successfully segregated the 6 HNSCC samples into a single branch while variants clustered together in another branch. Having established that differences in gene expression can discern types of carcinomas we examined gene expression in tumor cells within our patient population. From the high risk group, 4 tumors were chosen, 3 from patients where there was no recurrence (P57, P61, P63) and one from a patient who had a metastasis (P12). 4000 known genes were examined (GF211 array). In blinded analysis of gene expression sample P12 was identified as having an expression pattern divergent from the other 3 samples. Multi-dimensional scaling also segregated P12 from the other 3 tumor samples. Using the 97th percentile of studentized residual as an arbitrary cutoff for significance, approximately 100 genes were identified. A number of these differentially regulated genes are associated with metastasis and angiogenesis and include the integrins, fibronectin 1, CD59, CD37, myosin, tubulin, plasminogen activator, thrombospondin, and laminin, as examples. Other tumor samples are being added to increase the statistical validity of our analysis and to determine if patients can be stratified into risk groups by gene expression analysis.
KEYWORDS: gene expression profiling, tumor response, radiation, risk assessment