Ronald Eldridge

Ph.D., M.P.H.

Emory University

Postdoctoral Fellow

PLCO (Learn more about this study)

PLCO-338

Jan 31, 2018

Assessing mortality attributable to head and neck cancers caused by smoking, alcohol consumption, and human papillomavirus infection.

The primary causes of head and neck cancer (HNC) in the U.S. are smoking, alcohol consumption, and human papillomavirus (HPV) infection. There is strong suggestion that a patient’s risk of mortality differs by the primary cause of the tumor. However, this mortality difference has not been fully quantified. Due to the low incidence and mortality of HNC it is difficult to assess how smoking, alcohol consumption, and HPV infection may cause HNC-related mortality. Additionally, the five year survival rate of HNC is 67%, meaning many patients will likely survive their HNC diagnosis and die from other causes. For those patients who die from other causes, the effects of their HNC diagnosis on mortality may be more indirect – such as chemotherapy and radiotherapy affecting their long-term mortality. To accurately capture those indirect effects on mortality, novel methods are required.

Recent advances in causal mediation analysis have allowed researchers to decompose an estimated effect into causally identified direct and indirect effects. To date, this method has been mostly used to study biological mechanisms, but it can be applied in other ways. Provided accurate adjustment for confounding, the indirect effect can be viewed as the risk of the outcome caused by the mediator which in turn was caused by the exposure; for instance, the change in all-cause mortality due to HNCs caused by smoking, alcohol consumption, or HPV infection. Estimating that indirect effect would capture HNC cause-specific mortality, plus effects an HNC diagnosis might have on mortality when HNC is not the underlying cause of death.

PLCO is an ideal population for this study because it has pre-diagnosis information on smoking and alcohol consumption and potential confounders. In a nested case control subpopulation of the screening arm, it has pre-diagnosis serologic assays of HPV antibodies. Finally, the cohort has followed each subject for HNC incidence (including site: oral cavity, oropharynx, and larynx) and mortality.