Head and Neck Cancer Risk Factors and Risk Prediction Model Validation

NCI R21 funded for "Development, Application, and Evaluation of Prediction Models for Cancer Risk and Prognosis" (PA-10-026). Project start date: March 1, 2012

Head and neck cancer (HNC), including malignancies of the oral cavity, oropharynx, hypopharynx, and larynx, is the sixth most common cancer worldwide, accounting for 6% of all cancers (Parkin 2005). Worldwide, more than half a million HNC cases and 300,000 deaths due to HNC are estimated to occur each year (Ferlay 2004). In the US, 35,530 men and 13,730 women were diagnosed with HNC in 2010 (Cancer Facts & Figures). Tobacco and alcohol account for a large proportion of HNC cases in Latin America (82.9%), and Europe (84.3%; Hashibe 2009). However, in the US, tobacco and alcohol only account for 50.5% of HNC cases (Hashibe 2009).

Although the HNC incidence rates have been decreasing with the decreasing prevalence of tobacco smoking in the US over the last few decades, the incidence rates for tonsil and tongue cancers overall (Ryerson 2008) and for oral cavity and pharyngeal cancer among young women (Bleyer 2009) have been increasing in the US. The alarming trend for oropharyngeal cancer might be due to human papillomavirus (HPV) infection, a recognized cause of oropharyngeal cancer (IARC Monograph, 2007). In North America, 40-80% of oropharyngeal cancer cases are HPV positive (Marur 2010).

Additional risk factors are family history of head and neck cancer (Negri 2009), asbestos and inorganic acid mists (occupational) for laryngeal cancer (Straif 2009; Baan 2009), and genetic variants in the alcohol metabolism genes ADH1B and ADH7 (Hashibe 2008). Individuals who have oral precancers such as oral leukoplakia have a higher likelihood of developing oral cavity cancers (Amagasa 2011). Other suspected HNC risk factors include socioeconomic status (Conway 2008), involuntary smoking (Lee 2008), marijuana use (Berthiller 2009), body mass index (Gaudet 2010), coffee intake (Galeone 2010), oral hygiene (oral cavity cancer; Guha 2007, La Vecchia 2009), physical activity (Leitzmann 2008), and sexual behaviors (Heck 2010).

Cancer risk prediction models on cancers of the breast, colorectum, prostate, lung, and ovary have been published (Freedman2005). To date, a risk prediction model for HNC has not been developed, possibly due to the limited sample size in individual studies for risk model development and validation.

We plan to develop a HNC risk prediction model within the International Head and Neck Cancer Epidemiology (INHANCE) Consortium (http://inhance.iarc.fr). The INHANCE consortium was established in 2004 to study the etiology of head and neck cancer across various international epidemiologic studies, chiefly of case-control design. INHANCE is coordinated through the University of Utah (Hashibe) and Mt. Sinai School of Medicine (Boffetta). Presently, INHANCE includes over 35 studies, which when taken collectively provide questionnaire data on over 25,000 cases and over 35,000 controls, and biological samples from a majority of the study populations. We have thus far published 13 papers based on pooled analysis of different aspects of head and neck cancer epidemiology (Hashibe 2007, Lee 2008, Negri 2009, Purdue 2009, Hashibe 2009, Berthiller, Conway 2009, Lubin 2009, Marron 2010, Heck 2010, Gaudet 2010, Galeone2010, Sturgis 2011).

While the INHANCE pooled data is a powerful research tool, a limitation is that the INHANCE estimates of attributable fractions are not ideal because the majority of the case-control studies are hospital-based and thus the controls may not be representative of the general population. Additionally, we are unable to circumvent the case-control study design limitation of recall bias, since most of the risk factors we have investigated have relied on self reports after the patients were diagnosed with cancer. Thus, we believe that a collaboration between the INHANCE consortium and the Cohort Consortium will be an opportunity to greatly advance epidemiologic research of head and neck cancer.

Our goals are to estimate risk ratios and attributable fractions (AF) for established and suspected HNC risk factors, and in the cohort data. We are currently developing the first HNC risk prediction model for HNC overall and for HNC subsites (oral cavity, oropharynx, hypopharynx, larynx) in the INHANCE consortium. The Cohort Consortium would be the ideal setting to validate the INHANCE head and neck cancer risk prediction model.

Aim 1: To estimate the proportion of HNC cases attributed to factors other than tobacco and alcohol in population-based prospective data. We will estimate risk ratios (RR) and attributable fractions (AF) for established and suspected HNC risk factors, and compare the estimates between the Cohort Consortium and INHANCE consortium.

Aim 2: To validate the INHANCE HNC risk prediction model with data from the Cohort Consortium. We will assess the performance of the INHANCE HNC risk prediction model for HNC overall and for HNC subsites separately (oral cavity, oropharynx, hypopharynx, larynx).

We propose to conduct a pooled analysis of cohort studies focusing on HNC patients. RRs and AFs for established and suspected HNC risk factors will be estimated using Cox proportional hazard models. The suspected HNC risk factors include socioeconomic status, involuntary smoking, marijuana use, body mass index, coffee/tea intake, oral hygiene, physical activity, sexual behaviors, and reproductive factors for women. We will assess the performance of the INHANCE HNC risk prediction model in discriminating cases and controls.

A cohort consortium approach is necessary to understand the role of several suspected HNC risk factors without the limitations inherent in case-control study design, and obtain population-based attributable fraction estimates. Since not all of the cohorts are population-based, we will stratify on population-based status. The validation of the INHANCE HNC risk prediction model requires prospective data that is independent of the INHANCE Consortium studies.

50 head and neck cancer cases

Head and neck cancer overall and by subsite (oral cavity, oropharynx, hypopharynx, larynx), with histology.
ICD codes: C00 Lip, C01 Base of the tongue, C02 Other and unspecified parts of the tongue, C03 Gum, C04 Floor of mouth, C05 Palate, C06 Other and unspecified parts of the mouth, C09 Tonsil, C10 Oropharynx, C12 Pyriform sinus

Minimum data: 1) Socioeconomic status (SES):Education,income, 2) Tobacco smoking, 3) Alcohol drinking
For other exposure data, we would like to request it only if it is available: 1) HPV 16/18 infection, 2) Body Mass Index, 3) Chewing habits, 4) Involuntary smoking, 5) Marijuana use, 6) Oral precancers, 7) Occupation (asbestos), 8) Indoor air pollution exposure, 9) Family history of cancer, 10) Dietary factors (fruits/vegetables/coffee/tea/caffeine), 11) SNPs: ADHs, 12) Sexual behaviors, 13) Oral hygiene, 14) Reproductive factors, 15) Physical activity, 16) Previous medical conditions.

Age, sex, race/ethnicity, study center

No