Thesis
Development and validation of a head and neck cancer risk calculator
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2023
- Thesis identifier
- T16593
- Person Identifier (Local)
- 201788040
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- Background: Most new head and neck cancer (HNC) cases in the UK are diagnosed in an advanced disease stage. This is despite the availability of the 2-week wait (2ww), urgent suspected cancer referral pathway from primary to secondary care. Most HNCs are diagnosed from routes other than the 2ww, despite an increasing number of 2ww referrals. A symptom-based risk calculator had been previously designed to identify patients at high risk of HNC (AUC: 77%) but has not been widely adopted to date, having a lower AUC compared to other common cancer risk calculators (AUC >80%). Aim and Objectives: The aim of this study was to develop and validate a refinedversion of the HNC symptom-based calculator with the objective of increasing its prediction potential to be more in line with other cancer risk calculators. Design, Setting and Participants: The study was performed in two stages. The calculator development phase was based on a prospective cohort of new head and neck referrals to a secondary care centre in Glasgow (n=3,531, following sample size calculation). The validation phase was performed in a new prospective cohort of patients referred via the 2ww pathway in 41 secondary care centres across the UK (n=4,569) during the first wave of the COVID-19 pandemic. Main outcome measures: The main outcome measure was the area under the curve (AUC) and sensitivity and specificity combination of the final selected model at internal and external validation. Data collected included demographics, social history, presenting symptoms and signs and HNC diagnosis. Binary logistic regression analysisand random forest modelling with internal validation were performed to identify the best-performing model, followed by logistic regression external validation of the updated (HaNC-RC v.2) model. Results: The HaNC-RC v.2 had an improved AUC of 88.6% at internal validation. The model included age, gender, unintentional weight loss, smoking and alcohol history and a refined list of positive and negative symptoms of HNC. Two recommended referral thresholds were introduced based on sensitivity and specificitycombinations for a 2ww referral (cut-off: 7.1%; sensitivity: 85%, specificity: 78.3%) and urgent referral (cut-off: 2.2%; sensitivity: 97.1%; specificity: 52.9%). The AUC remained high at external validation (AUC: 83.96%; sensitivity:70%; specificity: 81%). The use of the HaNC-RC v.2 resulted in a reduction of the 2ww appointments by 70% during the first wave of the COVID-19 pandemic. Of the total of 256 cancers, 73.2% were seen in the high-risk group (2ww referral) and 16.5% in the moderate-risk group (urgent referral). These figures were much improved compared to those based on GP triaging using the national referral guidelines (59.9% and 25.4%, respectively) in the Glasgow region, without affecting the total numbers seen in each clinical setting. Conclusions: This study achieved its aim and objectives of developing and validating an updated version of a previously designed HNC risk calculator. The HaNC-RC v.2 has a much-improved AUC that remained high at external validation, and it could be used as a triaging aid for head and neck referrals in secondary or primary care pathways.
- Advisor / supervisor
- MacKenzie, Kenneth
- Kavanagh, Kimberley
- Lowit, Anja
- Resource Type
- DOI
Relations
Items
Thumbnail | Title | Date Uploaded | Visibility | Actions |
---|---|---|---|---|
|
PDF of thesis T16593 | 2023-06-05 | Public | Download |