Repository logo
 

Quantifying, Understanding and Predicting Differences Between Planned and Delivered Dose to Organs at Risk in Head & Neck Cancer Patients Undergoing Radical Radiotherapy to Promote Intelligently Targeted Adaptive Radiotherapy


Type

Thesis

Change log

Authors

Abstract

Introduction: Radical radiotherapy (RT) is an effective but toxic treatment for head and neck cancer (HNC). Contemporary radiotherapy techniques sculpt dose to target disease and avoid organs at risk (OARs), but anatomical change during treatment mean that the radiation dose delivered to the patient – delivered dose (DA), is different to that anticipated at planning – planned dose (DP). Modifying the RT plan during treatment – Adaptive Radiotherapy (ART) – could mitigate these risks by reducing dose to OARs. However, clinical data to guide patient selection for, and timing of ART, are for lacking.
Methods: 337 patients with HNC were recruited to the Cancer Research UK VoxTox study. Demographic, disease and treatment data were collated, and both DP and DA to organs at risk (OARs) were computed from daily megavoltage CT image guidance scans, using an open-source deformable image registration package (Elastix). Toxicity data were prospectively collected. Relationships between DP, DA and late toxicities were investigated with univariate, and logistic regression normal tissue complication probability (NTCP) modelling approaches. A sub-study of VoxTox recruited 18 patients who had MRI scans before RT fractions 1, 6, 16, and 26. Changes in salivary gland volumes and relative apparent diffusion coefficient (ADC) values were measured and related to toxicity events. Results: Spinal cord dose differences were small, and not predicted by weight loss or shape change. Mean DA to all other OARs was higher than DP; factors predicting higher DA included primary disease site, concomitant therapy, shape change and advanced neck disease. Nine patients (3.7%) saw DA>DP by 2Gy to more than half of the OARs assessed. These patients all had received bilateral neck RT for N-stage 2b oropharyngeal cancer. Strong uni- and multivariate relationships between OAR dose and toxicity were seen. Differences between DA and DP-based dose-toxicity models were minimal, and not statistically significant. On MRI, both parotid and submandibular glands shrank during treatment, whilst relative ADC rose. Relationships with toxicity were inconclusive.
Conclusions: Small differences between OAR DP and DA mean that DA-based toxicity prediction models confer negligible additional benefit at the population level. Factors such as primary disease sub-site, concomitant systemic therapy, staging and shape change may help to select the patients that do develop clinically significant dose differences, and would benefit most from ART for toxicity reduction.

Description

Date

2019-09-27

Advisors

Jena, Raj
Gallagher, Ferdia
Burnet, neil

Keywords

Head & Neck Cancer, Deformable Image Registration, Normal Tissue Complication Probability Modelling, Adaptive Radiotherapy

Qualification

Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge
Sponsorship
Addenbrooke's Charitable Trust (ACT) (24/15 A/Noble)
Cancer Research UK (C20/A20917)