Author of

• 15382

  +

  P3.04 - Poster Session/ Biology, Pathology, and Molecular Testing (ID 235)

  • Event: WCLC 2015
  • Type: Poster
  • Track: Biology, Pathology, and Molecular Testing
  • Presentations: 1
  • Moderators:
  • Coordinates: 9/09/2015, 09:30 - 17:00, Exhibit Hall (Hall B+C)

  • 15458

    +

    P3.04-076 - The Crux of Molecular Prognostications in NSCLC: An Optimized Biomarker Panel Fails to Outperform Clinical Parameters (ID 2586)

    09:30 - 09:30  |  Author(s): H. Koyi
    • Abstract

    Background:
    The best known prognostic factors for non-small cell lung cancer (NSCLC) patients are age, tumor stage and performance status. Numerous proteins have been analyzed to improve the traditional prognostication. Even though some proteins have shown prognostic value, the performance is not sufficient to be introduced in the clinical routine. The aim of this study was to generate a prognostic classifier based on proteins that previously have shown reproducible prognostic value and represent different aspects of tumorigenesis.
    
    Methods:
    The selection of proteins was based on literature search, meta-analysis of gene expression data sets and availability of reliable antibodies towards these proteins. Finally, five proteins (Ki67, EZH2, SLC2A1, TTF1 and CADM1) were chosen and analyzed by immunohistochemistry on tissue microarrays comprising NSCLC tissue patients (n=673), divided into a training and a validation cohort. For each patient, one score was obtained for each of the five antibodies, integrating the staining intensity and the fraction of stained tumor cells. Analyses were performed using all possible combinations of proteins and tested with or without clinical parameters. The C-index was used to develop the best prediction model on a training cohort (n=326) and the model was subsequently validated in the validation cohort (n=347).
    
    Results:
    All five proteins showed a significant prognostic impact in the univariate and the multivariate Cox analyses. Using a combination of the protein scores, the model was then fitted to provide the best prognostic performance (C-index=0.60). This did, however, not outperform the use of clinical parameters alone (C-index=0.62). The same was true when the analyses were performed separately for the adenocarcinoma (C-index=0.60) and the squamous cell carcinoma subgroup, respectively (C-index=0.60). More importantly, the addition of protein data to the clinical information (C-index=0.62) did not improve the prognostic value of the clinical parameters alone (C-index=0.60). To substantiate the results of our test cohort, we transferred the best prognostic model for all NSCLC, only adenocarcinomas and only squamous cell carcinomas respectively to a validation cohort. Again, all proteins showed prognostic relevance in the univariate analysis but did not perform better, alone or in combination, than the clinical parameters.
    
    Conclusion:
    Here we have performed a comprehensive analysis in order to obtain the best survival prediction model by using clinical parameters and the expression of five proteins. Although we chose strict criteria for protein marker selection, the prognostic power of these proteins was inferior to the traditional clinical parameters. Our findings question the general concept of using protein markers for prognostication in NSCLC but stress the value of careful assessment of traditional parameters in clinical practice.