ESTRO GUIDE 2017

Quantitative Methods in Radiation Oncology: Models, Trials and Clinical Outcomes 8-11 October 2017 Maastricht, The Netherlands

TARGET GROUP The course is aimed at physicians, medical physicists, biologists and radiation therapists (RTTs).

I, II, III, and IV trial designs, meta-analysis, clinical endpoints, survival statistics and the Cox Proportional Hazards Model • Statistical modelling and exploratory data analysis, simple mechanistic models, external and internal va- lidity of models, bootstrap andMonte Carlomethods, goodness of fit • Dose-response models, normal tissue complication probability (NTCP) and tumour control probability (TCP) models, modelling combinedmodality therapy, patient-to-patient variability in response, the line- ar-quadraticmodel and beyond, generalised equivalent uniform dose, use of models in treatment planning • Predictive assays, ROC curves and AUC, sensitivity, specificity, positive and negative predictive value • High dimensionality data sets, machine learning, data mining, over-fitting, training and validation sets, sample splitting, K-fold validation. PREREQUISITES No specific requirements are needed for attending this course although a broad familiarity with the principles of cancer medicine and radiation oncology is expected. TEACHINGMETHODS The four-day course consists of: • 27 didactic 45-minute lectures • 4 half-hour interactive discussion sessions • A practical exercise (1.25 h) • An interactive data analysis session (1.25 h) • A Meet-the-professor session where you can bring- your-own data analysis project and discuss one-on-one with faculty members (10-minute time slots, 1.25 h total time). METHODS OF ASSESSMENT • Course evaluation form • Self-assessment tools are integrated in some of the discussion sessions.

COURSE AIM The aimof this course is tomake the attendees better at makingmodel-supported decisions. Radiation oncology probably has the most solid quantitative foundation amongmedical specialties. As in other specialties, results of randomised controlled trials form evidence-based treatment guidelines; but in addition, prognostic and predictive models provide clinical decision support for individualisedmanagement of cases. Radiation bioeffect models of Normal Tissue Complication Probability (NTCP) and Tumour Control Probability (TCP) have become much more refined and are increasingly being validated in independent datasets. While integration of quantitative estimates of various treatment outcomes is likely to improve patient care, it is also important to understand the limitations of model estimates and to be able to assess the validity or quality of a statistical data analysis or a mathematical model. Uncritical reliance on model results may compromise patient safety or treatment outcome. LEARNINGOUTCOMES By the end of this course participants should be able to: • Broadly describe themost commonly used quantitative methods in radiation oncology and radiation biology and the assumptions behind these • Identify appropriate quantitative methods of analysis for a given data set • Critically evaluate modelling results especially with respect to proper validation and estimates of uncer- tainties. COURSE CONTENT • Models and modelling, hypothesis testing and pa- rameter estimation, type I and II uncertainties • Clinical trials and evidence-basedmedicine, phase 0,

82

ESTRO SCHOOL