Statistical and Epidemiological Challenges in Using the NHANES Assessment of Oral Human Papillomavirus Infection to Study Risk of Infection and of Oropharyngeal Cancer in the US

• Activity Number: 358 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Barry Graubard* and Anil Chaturvedi and Joseph Tota and Hormuzd Katki and Maura Gillison
• Companies: National Cancer Institute and National Cancer Institute and National Cancer Institute and US National Cancer Institute and MD Anderson
• Keywords: NHANES; Risk of oral HPV; Complex survey data; quasi-score tests; propensity weighting; area under ROC

Abstract

NHANES 2009-16 collected saliva in a cross-sectional stratified multistage cluster probability survey of the US to study risk factors for oral human papillomavirus (HPV), which is strongly associated with incidence of oropharyngeal cancer (OPC). Three uses of this data and related statistical issues are: 1) identify risk factors for the prevalence of oral HPV infection; 2) examine effectiveness of prophylactic vaccination for HPV infection; and 3) develop/validate an absolute risk model of 1-year incident OPC in US with NHANES as a control series and non-representative clinical samples (i.e., nonprobability samples) of OPC as a case series. The statistical methods that address these data uses while accounting for the complex sample design of the NHANES are presented. The methods include predictive margins to estimate covariate/risk factor adjusted prevalence differences, cubic regression splines to assess nonlinear relationships between prevalence of oral HPV and age, quasi-score tests to test the impact of HPV vaccination on small HPV infection rates, and propensity weighting, cross-validation, and estimation of area under the ROC curve to evaluate the absolute risk models.