Validation of prognostic molecular markers in malignant pleural mesothelioma
Division of Thoracic Surgery, Brigham and Womenâs Hospital and Harvard Medical School, MA, Boston
AIMS: We have developed a simplified technique to predict prognosis in thoracic cancers using ratios of gene expression levels (Gordon et al., 2003, Cancer Epidemiol Biomarkers Prev, 12: 905-910). Using profiling with microarrays, we previously identified new prognostic markers and applied the gene ratio method to develop a prognostic test for patients with mesothelioma undergoing trimodality therapy (Gordon et al. 2003, JNCI, 95(8): 598-605). This test uses 4 genes in 3 gene pair ratios: KIAA0977/GDIA1, L6/CTHBP, and L6/GDIA1. Here we have tested the efficacy of the new prognostic markers and the specific prognostic test to a different cohort of patients with mesothelioma. METHODS: We used microarrays containing approximately 22,000 genes to analyze gene expression in 39 mesothelioma tumors with linked clinical data. We first utilized the previously identified top 20 prognostic genes to perform supervised cluster analysis and then use Kaplan-Meier analysis to compare survival in the resulting cohorts. Next we applied the 4 gene prognostic test to these 39 samples using real time quantitative RT-PCR to assign specimens to poor and good outcome group. These resulting groups were compared for survival as above. RESULTS: Using Kaplan-Meier survival analysis and PAM clustering of the 39 MPM samples of the present study, we found that we could identify two patient subsets with significantly different clinical outcome (P=0.00102). A random selection of genes demonstrated that there was a low likelihood (approximately 0.1%, i.e., 14/10,000 iterations) of observing these results by chance alone. These results validate the importance of the selected prognostic markers in mesothelioma. We next examined the accuracy of the proposed ratio based prognostic test. Using quantitative RT-PCR, we calculated the combined score for these 3 ratios in all 39 samples and assigned each to an outcome group, i.e. combined scores >1 and <1 were predicted to be associated with good and poor outcome, respectively. We found that this test was able to significantly (P=0.037) predict clinical outcome in these 39 samples. The estimated median survival (33 months) of the good outcome subset was almost 3-fold higher than the estimated median survival of the poor outcome subset (12 months). CONCLUSIONS: We propose a new expression ratio-based test for predicting outcome for patients with mesothelioma. This test improves on current techniques and may be ultimately used to assign treatment strategies.
Paper presented at the International Symposium on Predictive Oncology and Intervention Strategies; Nice, France; February 7 - 10, 2004; in oral session 893 (Molecular pathology - Part I).