Lung Cancer Case Study

Identification of Non-Small Cell Lung Cancer Responders to Iressa® (Gefitinib)
 

Purpose: Determine if the multivariate integration of features derived from object-based morphometric image analysis of tissue histology and multiplexed immunofluorescence (IF) assays with clinical patient attributes could identify a quantitative phenotypic profile that predicts overall survival in patients with non-small cell lung cancer (NSCLC) treated with gefitinib.

Patients and Methods: We analyzed diagnostic tumor samples from 109 patients with advanced refractory non-small-cell lung cancer (NSCLC) who were ultimately treated with gefitinib. Tumor samples were subjected to a series of assays: EGFR DNA mutation analysis (AstraZeneca); EGFR immunohistochemistry (AstraZeneca); morphometric quantitation of H&E stained FFPE tissue (Aureon); Morphometric quantitation of 15 protein biomarkers including cytokeratin 18, Ki67, caspase-3 activated, CD34, EGFR, phosphorylated EGFR, phosphorylated ERK, phosphorylated AKT, PTEN, cyclin D1, phosphorylated mTOR, PI3K, VEGF, KDR (VEGFR2), and phosphorylated KDR - all of which were arranged in various multiplexed IF assays custom developed for this project (Aureon). A mathematical model employing support vector regression (Aureon DiscoveryPath SVRc) was utilized to integrate six clinical variables (gender, smoking history, age at diagnosis, tumor histology and ECOG performance status) with quantitative features derived from Aureon’s object-based morphometric imaging of tissue histology and multiplexed biomarker immunofluorescence assays (Aureon).

Results: Of the 109 patients, 87 had sufficient tumor samples for further analyses. 4 of 87 patients contained EGFR tyrosine kinase domain mutations and 51 patients had complete data from each of the domains (clinical, histological morphometry, and immunofluorescence). A model predicting overall survival was developed by utilizing Aureon’s multivariate algorithm. The model had a concordance index of 0.74, HR 5.26, P=.0002. Poor performance status, poorly differentiated histology by morphometry, and increased levels of activated caspase-3, phosphorylated KDR, and cyclin D1 were associated with reduced survival.

Conclusion: The integration of clinical, tissue morphometry, and biomarker data identified a set of features that were associated with a more aggressive disease phenotype in patients with NSCL treated with gefitinib and resulted in poor overall survival. The study evaluated only patients treated with gefitinib and hence cannot distinguish whether the features identified are associated with longer survival on gefitinib only or with longer survival regardless of treatment

(Click Image to Enlarge)

Table 1. The results of SVRc-FR (Support Vector Regression for Censored Data - Feature Reduction) on the 51-patient cohort. Performance status was the only clinical variable that passed the CI (concordance index) filter. Several models were tested, with the combination of clinical, IF and morphometric features resulting in the best concordance index, 0.74.

(Click Image to Enlarge)

Figure 1. Standard Hematoxylin and Eosin (H&E) digitized image of primary non-small cell lung cancer (adenocarcinoma; bronchoalveolar type) pre- and post-processing with image analysis software. A, Representative region demonstrating alveolar lumen (arrows) and compact tumor epithelial cells lining alveolar spaces. B, Same image post-processing to illustrate cellular components classified with respect to compartment: epithelial cytoplasm (light purple), epithelial nuclei (blue), stroma (pink), stromal nuclei including endothelial cells (green), alveolar lumen (white space; arrows). Magnification, 200X.

(Click Image to Enlarge)

Figure 2. Representative examples of antibody distribution within non-small cell lung cancer specimens. A, Caspase-3 activated and B, phosphorylated mTOR, (red; white arrows) focally in tumor epithelial cell cytoplasm and nuclei respectively. C, Cyclin D1 (red circles) in tumor epithelial nuclei. D, Phosphorylated KDR (red; white arrows) diffusely within cytoplasm of tumor epithelial cells. DAPI-stained nuclei are in blue. Magnification 200X.

(Click Image to Enlarge)

Figure 3. Survival curves for the 51 patients in the model, separated into 2 groups by model score above or below the cut-point of 39. A score of 39.5 or greater predicted a shortened overall survival time. HR = 5.26, p = .0002.