Hierarchical clustering analysis of tissue microarray immunostaining data identifies prognostically significant subsets of stage IIIA N2 non-small cell lung cancer patients. An Alpe-Adria Thoracic Oncology Multidisciplinary group study (ATOM 014)

Abstract

10085 Background: Based on gene expression profiling, prognostically relevant cluster groups of patients have been recently identified for breast cancer, lung cancer, and lymphoma. The development of tissue microarray (TMA) technology provides methodology for the concomitant analysis of multiple proteins on large numbers of tumor samples. Methods: Here, we report the hierarchical clustering analysis of TMA immunostaining data from 87 patients with stage IIIA pN2 non-small cell lung cancer (NSCLC), treated with surgery alone between 1985 and 1997. A panel of 10 markers was tested by immunohistochemistry: EGFR, ErbB-2, c-kit, COX-2, survivin, bcl-2, cyclin D1, cyclin B1, MMP-2 and MMP-9. Univariate, multivariate analyses and unsupervised hierarchical clustering analysis by using these 10 markers were performed. Results: Bcl-2 (p<0.0001) and cyclin D1 (p=0.0036) are more expressed in squamous cell carcinoma (SCC), while MMP-2 (p=0.0115), MMP-9 (p=0.0075) and survivin (p=0.02) show increased expression levels in other histological subtypes. In univariate analysis, only squamous cell histology, bcl-2 and cyclin D1 expression were favorable prognostic factors (p=0.0149, p=0.0013, p<0.0001, respectively), while MMP-2 expression showed a significant correlation with poor outcome (p=0.013). In multivariate analysis, cyclin D1 and MMP-2 were the only positive and negative prognostic factors, respectively (p<0.0001, p=0.06). Unsupervised hierarchical clustering analysis produced 5 distinct cluster groups and the deduced dendrogram identified 2 prognostically significant subsets of patients, with a median survival of 51 (groups 1–2) and 10 (groups 3–4-5) months, respectively (p<0.0001). Notably, groups 1–2 were mostly composed of SCC (80%). Conclusions: This study suggests that hierarchical clustering of TMA immunostaining data by using a limited set of markers might provide a useful tool for the identification of radically resected NSCLC patients at high risk of recurrence, likely to benefit from more aggressive treatment. No significant financial relationships to disclose.