Margriet Kwint

Prognostic value of volumetric changes during concurrent chemoradiation 87 5 Introduction: Patients with locally advanced non–small cell lung cancer (LA-NSCLC) are preferably treated with concurrent chemoradiation (CCRT). Although this treatment has a curative intent, overall survival (OS) is poor with 44-59% surviving at 2 years (1- 3). In order to distinguish patients with better or worse OS, it is important to build accurate prediction models for individual treatment outcomes. Current prediction models mainly use baseline characteristics to predict treatment outcomes (4-11). An important step to improve these prediction models might be to incorporate longitudinal data of tumor volume change derived during treatment. The predictive value of tumor volume changes during treatment has been studied previously (12-15). Although these studies hinted towards a predictive potential, the observed associations were inconsistent and the performed studies were of limited quality due to small sample sizes. Furthermore, the performed analyses could have been too simplistic. For example, observed GTV-changes during treatment were dichotomized below or above the median (13-15). The assumption that 2 groups (above/below median) represent the treatment response of all NSCLC patients might be a misconception (16). In order to identify various subgroups of patients with distinct treatment responses, more advanced statistical techniques, such as latent class mixed modelling, could be useful (17-20). The use of image guided radiotherapy (IGRT) (e.g. Cone Beam-CT(CBCT)), resulted in an increase of readily available imaging data of tumor volume during radiation treatment(21). Consequently, this data could be used as a longitudinal parameter for treatment outcomes. Subgroups with particular tumor volume changes could have worse/better outcomes compared to others. The addition of this more dynamic information might improve the prognostic accuracy of the current baseline prediction models (4-10). Therefore, the aim of this study is to identify subgroups of LA-NSCLC patients showing distinct treatment responses during CCRT and to investigate whether the identified subgroups are associated with treatment outcomes.