Abstract Book

S331

ESTRO 37

There was a trend for higher bacterial diversity in patients with no symptoms (p=0.09). Symptom group by timepoint was significant when modelling the evolution of this diversity (p=0.05), with the intercept also significant (p<0.0001), adding evidence to the importance of microbial diversity in predicting risk of radiotherapy- induced gastrointestinal side-effects (figure 1). There was a trend for significance of symptom group when modelling the evolution of Roseburia (p=0.08) with the intercept different between groups (p=0.007). Although not statistically significant, evolution of proportions of Sutterella appear different between symptom groups, with the relatively low proportions of this genus making comparisons difficult without very large patient numbers.

and slope. Furthermore, the Bayesian (BIC) was used to assess improvement of fit of the model after the addition of a new class. We tested whether the distribution of patient-reported toxicities were significantly different between classes: 1) at baseline, and 2) during follow-up (Chi-square). Results In the PF domain four distinct classes (subgroups) were identified (Figure). The majority (63%) showed consistent high (class 1), middle (3) or low (4) PF, 12% experienced a clear decrease in PF (class 2). Toxicity patterns were different between classes for most GI and GU toxicities during follow-up, and for most GU toxicities at baseline (Table). Comparing class 2 (QoL deterioration) with class 1 (with comparable baseline PF), class 2 developed significantly more GI complaints during follow-up (pain/cramps p=0.01, diarrhea p=0.006, mucus p=0.02), whereas complaints were comparable at baseline. With respect to GU, class 2 had considerable higher rates at both baseline and follow-up (p<0.05). Conclusion With this study, we identified four distinct trajectories of physical functioning during follow-up of prostate cancer patients treated with radiotherapy. Toxicity profiles of identified classes were significantly different, at baseline and during follow-up. Being at risk for physical functioning deterioration seems not only associated with the development of radiation-induced toxicities but also with baseline symptoms. This suggests a complex relationship between physical functioning and toxicity. Such knowledge can be valuable for patient care with a more personalized approach.

Figure 1: Difference of Chao1-diversity at baseline (A) and dynamics of Chao1 α-diversity (B), Roseburia (B), and Sutterella (C) over time. IBDQ class: 0 = no symptoms, 1 = mild symptoms, 2 = persistent symptoms. IBDQb = IBDQ- bowel subset. Conclusion Gut microbiota patterns differ between patients with and without radiotherapy-induced gastrointestinal side- effects. Larger cohorts are needed to confirm these exploratory observations. Assessment of microbiota-host functionality will shed light on disease mechanisms. PV-0625 Quality of Life trajectories and correlation with toxicity after radiotherapy for prostate cancer K. De Vries 1 , I. Walraven 2 , L. Incrocci 1 , W. Heemsbergen 1 1 Erasmus MC Cancer Institute, Radiation Oncology, Rotterdam, The Netherlands 2 Netherlands Cancer Institute, Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective We previously observed a significant post-treatment deterioration in the physical functioning (PF) domain as measured on the quality of life (QoL) questionnaire SF36, in prostate cancer patients treated with radiotherapy. However, this was based on group averages which do not provide insights in the individual trajectories of QoL. Neither does it identify the subgroup of patients who experience a decrease in PF, and the toxicities these patients suffer from. Therefore the aim of this study was to identify these longitudinal trajectories by advanced statistical modeling, and to correlate these with toxicity. Material and Methods One hundred and sixty seven (n=167) patients were selected from a previous dose escalation study for localized prostate cancer (68 Gy or 78 Gy). Patients with a clinical relapse < 3y were excluded due to possible influence on QoL. Selected patients reported PF on the SF-36 and patient-reported toxicity on symptom questionnaires (scheduled at baseline, 6 months, y1-y3). The PF domain consists of 10 items about physical activities (e.g. walking distances). To identify classes (subgroups) of patients with distinct trajectories latent class growth mixture modeling was used. Hereby, each identified class is characterized by a distinct intercept