Epidemiology and risk prediction modelling of head and neck cancer

Smith, Craig D. L. (2024) Epidemiology and risk prediction modelling of head and neck cancer. PhD thesis, University of Glasgow.

Abstract

Head and neck cancer (HNC), typically defined as squamous cell carcinomas of the oral cavity, pharynx and larynx, is the 7 th most common cancer type globally. In the UK there over 12,000 new cases each year, with some of the highest incidence rates in Scotland. Major risk factors of HNC include smoking and alcohol consumption, both acting with independently and synergistically on HNC risk. Other key factors include socio-demographics (age, sex and socioeconomic deprivation). Human Papillomavirus (HPV) is another major risk factor for cancers of the oropharynx. With HNC incidence continuing to increase, primarily driven by the growing burden of HPV-associated oropharynx cancer (OPC) cases, there is a need for an increased focus on further understanding epidemiological changes and developing prevention strategies.

Chapter 1 summarises the literature focusing on definitions, epidemiology and risk factors for HNC. The chapter also goes on to describe the literature on approaches to the prevention of HNC.

Chapter 2 details the overarching aims and objective of this thesis and of each of the research studies.

Chapter 3 is presented as a journal publication. In this chapter, a rapid review of the literature using a systematic search of multiple databases was conducted on existing HNC risk models. In this review, 14 studies were identified according to pre-planned search criteria. The papers were assessed using the Prediction model Risk Of Bias ASsessment Tool (PROBAST) framework, in addition to an independent quality assessment. Six of the 14 models were classed as high quality, and of these, three were also high performing (AUC > 0.8, 0.87–0.89). A narrative synthesis of these models was performed and found that high-quality models had tailored the selection of predictors to the target populations. The synthesis identified models should include behavioural and sociodemographic predictors, in addition to clinical variables or biomarkers, where practical. The review found that some existing models have potential to predict HNC risk, but there is scope for improvement, with many of the models lacking external validation.

Chapter 4 is presented as a journal publication. In this chapter, data from the Scottish Cancer Registry were analysed. HNC incidence trends were evaluated by calculating age-standardised rates and using Poisson regression with interaction tests to assess changes in HNC incidence over time (2001-2020). This analysis revealed that the burden of HNC is changing (RR = 1.05, 95% CI = 1.01-1.09), primarily driven by increasing OPC rates (RR = 1.78, 95% CI = 1.65-1.93), accompanied by a stabilisation of OCC rates and a decline in larynx cancer rates (RR = 0.73, 95% CI = 0.68-0.79). However, the socio-demographics of people with HNC in Scotland have remained largely unchanged. Incidence over time was consistently higher among males (RR = 2.83, 95% CI = 2.74 - 2.91) and among those aged 60-64 and 65-69 years. Large area-based socioeconomic inequalities also existed and remained wide over the study period in all subsites, with the highest rates observed among those from the most socioeconomically deprived areas when compared with those from the least (RR = 2.87, 95% CI = 2.73-3.00). The Cancer Registry analysis assessing the HNC sociodemographic profile over time largely supported the subsequent use of a case-control study (ARCAGE) conducted in 2002-2004 for the HNC risk prediction model development and subsequent validation in the UK Biobank cohort study, which is still ongoing (2006 – present).

Chapter 5 is presented in a traditional thesis chapter format. It aimed, to investigate the associations between individual socioeconomic status (SES) and HPV-positive and HPV-negative OPC in a European multicentre case-control study (Alcohol Related Cancers And Genetic susceptibility in Europe – ARCAGE, 2002- 2004). Logistic regression analysis found no conclusive evidence of an association with a higher socioeconomic status and HPV-positive OPC (University education vs primary education; OR = 3.07, 95% CI = 0.92 - 10.30). However, this analysis was constrained by limited numbers of HPV-positive OPC cases (n = 74). This work lays the foundations for future international pooled analyses of multiple case-control and cohort studies within the HEADSpAcE (Head and Neck Cancers in South America and Europe) Consortium.

Chapter 6 is presented as a journal publication, detailing the process of developing and validating a HNC risk model. Following the findings of the review and supported by the findings of the registry analysis, a clinical risk prediction model was developed with data from the ARCAGE case-control study (1926 HNC cases and 2043 controls). Using established predictors from the literature, a clinical risk prediction model was devised using predictors chosen with ease of use in primary dental care at the forefront. The model exhibited fair performance in the developmental dataset (AUC = 0.75, 95% CI = 0.74–0.77) and had acceptable, but more limited, performance in the validation dataset, the UK Biobank cohort (384,616 participants, 1177 HNC cases; AUC = 0.62, 95% CI, 0.61 - 0.64). Such a model has potential to be developed into a tool and feasibility tested in primary dental care for prompting preventive interventions (e.g. smoking cessation, alcohol counselling) and recall intervals. The model could also potentially be improved with the use of biomarkers (e.g. HPV).

Chapter 7 is presented in a traditional thesis chapter format. It aimed to describe the existing HPV data within the UK Biobank. The viability of using multiple imputation to project HPV serostatus was assessed. Analysis of available HPV data within the UK Biobank revealed some associations with HPV-serostatus, mostly sexual behaviours, but the substantial volume of missing HPV data (>98% missing) made any potential imputations untenable. The utility of exploring HPV mediated HNCs in the UK Biobank at the time of writing, was limited by the small, random sample of HPV-serology data available.

Chapter 8 is presented as a traditional thesis discussion. It describes the key findings of this thesis, including the under-utilisation of HNC risk models, a lack of validation and feasibility testing, the unchanging socio-demographics of HNC and the performance of the ARCAGE model with recommendations for future feasibility testing and areas for improvement. This chapter also compares the findings of this thesis with the wider literature and details other advancements published over the time-period of this thesis. The strengths and limitations of each study are also detailed, followed by recommendations for practice, policy and future research.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Supported by funding from TRACC (to Train and Retain Academic Cancer Clinicians) programme and Cancer Research UK.
• Subjects: R Medicine > R Medicine (General); R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
• Colleges/Schools: College of Medical Veterinary and Life Sciences > School of Cancer Sciences
• Supervisor's Name: Conway, Professor David, McMahon, Dr. Alex and Inman, Professor Gareth
• Date of Award: 2024
• Depositing User: Theses Team
• Unique ID: glathesis:2024-84508
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 28 Aug 2024 13:40
• Last Modified: 30 Aug 2024 10:28
• Thesis DOI: 10.5525/gla.thesis.84508
• URI: https://theses.gla.ac.uk/id/eprint/84508

Related URLs

• Article DOI
• Article DOI