Candidate Biomarkers for HPV-Negative Head and Neck Cancer Identified via Gene Expression Barcode Analysis. (September 2016)

• Record Type: Journal Article
• Authors: Yang, Shiayin F.; Bier-Laning, Carol M.; Adams, William; Zilliox, Michael J.
• Is Part Of: Otolaryngology--head and neck surgery. Volume 155:Number 3(2016:Sep.)
• Page Start: 416
• Page End: 422
• Publication Date: 2016-09
• Subjects: gene expression barcode -- microarray -- biomarkers -- HPV negative -- head and neck cancer; Head -- Surgery -- Periodicals; Neck -- Surgery -- Periodicals; Otolaryngology -- Periodicals; 617.51
• DOI: 10.1177/0194599816642436
• Languages: English
• ISSNs: 0194-5998

Abstract

Objectives: (1) Reanalyze publicly available genomic data for HPV-negative oral cavity squamous cell carcinoma to look for candidate biomarkers. (2) Evaluate the association of the identified biomarkers with survival. Study Design: Retrospective cohort study. Setting: Tertiary care. Subjects and Methods: Gene expression barcode analysis was applied to an existing publicly available data set of 54 HPV-negative oral cavity squamous cell carcinoma tumor samples to identify candidate genes associated with poor prognosis. Genes identified were evaluated for their association with survival on the basis of univariable and multivariable Cox proportional hazards models. Results: Three genes were found to be associated with poor prognosis. The most significant association was seen with spectrin expression. Subjects whose tumors expressed spectrin were 4.60 times more likely (hazard ratio; 95% confidence interval: 1.88-11.25) to die at any given time when compared with those without spectrin ( P = .001). On univariable analysis, subjects with late-stage cancer were 6.34 times more likely (hazard ratio; 95% confidence interval: 1.41-28.53; P = .02) to die at any given time, but interestingly, after controlling for spectrin, this effect was attenuated ( P = .07). Despite controlling for several possible confounding effects, the effect of spectrin remained hazardous throughout all multivariable models. This was true even after controlling for cancer stage and extracapsular extension ( P = .004). Conclusion: Our analysis of public genomic data shows promise in identifying biomarkers that may allow clinicians to make more accurate survival predictions. Spectrin is a strong candidate for further biomarker testing.