RISK OF NODAL DISEASE IN PATIENTS WITH ORAL SQAMOUS CELL CARCINOMA IDENTIFIED BY IMMUNE-GENE EXPRESSION PROFILING

Presentation: S003
Topic: Oral Cavity
Type: Oral
Date: Wednesday, April 26, 2017
Session: 10:15 AM - 11:00 AM Oral Cavity
Authors: Kelly Y Liu, Catherine F Poh, DDS, PhD, FRCPC
Institution(s): The University of British Columbia

Importance: Nodal disease (N+) is the most significant prognostic factor in oral squamous cell carcinoma (OSCC). Preventive neck dissection (ND) may improve outcome; however, clinico-pathological features, including depth of invasion (DOI), are poor indicators and often result in unnecessary or delayed ND leading to morbidity and mortality. There is an urgent need to identify new markers(s) to stratify patients for appropriate treatment.

Objectives: To explore differential gene expression (DGE) on cancer immunology-related genes of OSCC with known nodal status.

Design: A pilot case control study on OSCC patients who were treated with curative surgery with at least 3-year postoperative follow-up.

Methods: Patients with available fresh-frozen tissues of the primary tumors were identified from a pan-Canadian COOLS trial. Tissue containing >80% tumor cells were processed for total RNA purification (AllPrep® DNA/RNA/miRNA Universal Kit,  Quagene©). Commercially available nCounter® GX assay with PanCancer Immune Profiling Panel (Nanostring Technologies, Seattle, WA) was used to quantify expression of 730 genes specific to immune cells, tumor cells, or immunological pathways. Statistical analysis was performed using R base functions or packages (v3.3.2), with p<0.05 (2-sided) considered significant. Student’s t-test or chi-square test was used for comparing patient or clinico-pathological variables between N0/N+ groups; and general logistic regression for examining odds ratio of these variables with N+. In DGE analysis, difference in gene expression was described in log2 transformed fold-change (FC); association of expression with N+/N0 was analyzed with multivariate linear regression model followed by p-value correction using Benjamini-Hochberg controlling false discovery rate (FDR); and unsupervised hierarchical clustering was used to examine the predictive value (sensitivity, SE and specificity, SP) on N+/N0.

Results: The study cohort consisted of 87 patients of whom 43(49%) were N+ either at time of surgery (51%) or within 2 years of follow-up (49%); whereas 44(51%) patients remained N0 for at least 3 years. There was no difference between the groups (in age, gender, smoking history, or tumor site) except for tumor Grade III and DOI ≥4mm. Using multivariate logistic regression analysis, Grade III was the only factor significantly associated with N+ status (OR, 25; 95% CI, 5.5-157.3; p<0.001) but not DOI ≥4mm (OR 2.0; 95% CI, 0.7-6.0; p=0.2). Combination of Grade III and ≥4mm showed 39.5% SE and 95.5% SP. For patients who were neither Grade III or ≥4mm, we identified a subset of 21 genes (FDR<0.001 and ≥1 FC) which had SE of 61.5% and SP of 78.6%; outperforming conventional risk marker using ≥4mm DOI alone (SE, 65%; SP, 43%). With the combination of clinico-pathological and genomic markers, we would perform 56 NDs (10 salvage and 9 unnecessary NDs) vs using clinic-pathological alone (69 NDs; 9 salvage and 24 unnecessary NDs).

Conclusion: This preliminary data identified a panel of genomic markers that can potentially be used for stratifiying patient's treatment according ot their risk of nodal disease; ultimately, reduce cost and improve quality of life.  Further validation study is ongoing and warranted.