Show simple item record

dc.rights.licenseCC-BY-NC-ND
dc.contributor.advisorSchuit, E.
dc.contributor.authorHeilmaier, Christina
dc.date.accessioned2022-06-04T00:00:40Z
dc.date.available2022-06-04T00:00:40Z
dc.date.issued2022
dc.identifier.urihttps://studenttheses.uu.nl/handle/20.500.12932/41607
dc.description.abstractBackground: Photon-based radiation therapy is important in the treatment of head-and-neck cancer (HNC). However, potential post-radiogenic complications may have a long-lasting negative impact on the patient's quality of life. The current study focused on the presence of laryngeal edema 6 months after irradiation of HNC and had three main objectives: (I) external validation of the normal tissue complication probability (NTCP)-model published by Rancati based on the equivalent uniform dose (EUD), (II) development of new NTCP-models using a large training cohort, (III) external validation of the new models with an independent validation data set. Methods: The training cohort consisted of n=750 and the validation cohort of n=395 patients. For both cohorts, a sub-cohort including patients with non-laryngeal HNC was formed (n=416 and n=227 participants, respectively). Multiple imputation was executed to compensate for missing values. Rancati’s model performance was assessed with both the training and the validation cohort, determining the discrimination (area-under-the-curve, AUC) and calibration parameters (calibration intercept and slope). Following, candidate predictors for new models were selected, their correlation and multicollinearity was checked and the required sample size was calculated. New models were set up with stepwise regression (p<0.20), considering different (sub-)cohorts and candidate predictors to comprehensively answer the research question. The performance of the new models was analyzed and cross-validation (cv) was executed to account for optimism. Finally, external validation of the new models was performed. Results: External validation of Rancati’s model revealed a poor calibration and moderate model performance with the training data set (AUC=0.66; 95%-confidence interval (CI): 0.60-0.71). Analysis of 11 different NTCP-models showed that the model consisting of dose values (= mean dose to the glottic and supraglottic area = GLOTTIC DMEAN and SUPRAGLOTTIC DMEAN), the treatment regimen (SEQUENCE), and clinical parameters (SMOKING and AGE) yielded the best result in estimating laryngeal edema in non-laryngeal HNC patients (cvAUC: 0.86 (0.74-0.97), calibration intercept: 0.06 (0.04-0.08), slope: 1.01 (1.0-1.01)). In contrast, only the dose values (GLOTTIC DMEAN and SUPRAGLOTTIC DMEAN) turned out to be relevant in the sub-cohort of laryngeal HNC patients (cvAUC: 0.66 (0.58-0.74), intercept: 0.24 (-0.19-0.46), slope: 1.17 (1.02-1.27)) and the total cohort (cvAUC: 0.75 (0.74-0.77), intercept: 0.04 (0.03-0.05), slope: 1.02 (1.02-1.02)), which could be attributed to differences in the clinical characteristics of the respective patients. Regardless of the patient cohort, the combination of GLOTTIC DMEAN and SUPRAGLOTTIC DMEAN was superior to GLOTTIC DMAX and SUPRAGLOTTIC DMAX (= maximum dose to the glottic and supraglottic area) and the laryngeal EUD. Moderate to high correlation between the dose values was evident (range of correlation coefficients=0.69-0.92) but without significant multicollinearity as indicated by variance-inflation-factors of 1.01-2.16. Comparison of the external validation of Rancati’s and the new NTCP-models with the validation data set revealed better performance parameters of the new models (e.g., AUC: 0.69-0.97 versus 0.51 for Rancati). Conclusions: New NTCP-models relying on dosimetric and non-dosimetric variables achieve a better prediction of the risk of post-radiogenic laryngeal edema in HNC patients compared to recent models based entirely on dose data. This is an essential step towards individualized patient treatment.
dc.description.sponsorshipUtrecht University
dc.language.isoEN
dc.subjectComprehensive development of various statistical models to predict the development of post-radiogenic laryngeal edema after irradiation of head and neck tumors.
dc.titleDevelopment of NTCP-models to predict the presence of laryngeal edema six months after radiotherapy of head-and-neck cancer
dc.type.contentPostgraduate Master Thesis
dc.rights.accessrightsOpen Access
dc.subject.keywordsNTCP-model, larynx edema, post-radiogenic, head-and-neck. cancer
dc.subject.courseuuEpidemiology Postgraduate
dc.thesis.id4218


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record