Statistical Methods for Biomarker Analysis for Head and Neck Carcinogenesis and Prevention

This chapter focuses on various statistical methods used for biomarker analysis for head and neck carcinogenesis and prevention. Various statistical methods are illustrated by using data obtained from a retinoid chemoprevention trial in patients with oral premalignant lesions (OPLs). Because of patient refusal, loss to follow-up, and limited amount of tissues, biomarkers in the chapter have not been analyzed on all participants, but only on the available tissue samples. Knowing the type of biomarker is essential for choosing the proper statistical method for biomarker analysis. Biomarkers can be measured qualitatively, quantitatively, or semiquantitatively. Qualitative biomarker measures can also produce categorical variables. The Cox proportional hazards model, a general semiparametric method, is used for analyzing time-to-event data. The application of an interval event chart provides a graphic assessment of the combined effect of histology and prior cancer history on time to cancer development. Event charts are very useful in revealing the association of multiple time-to-event data at the individual level.