ESTRO 37 Abstract book

S342

ESTRO 37

PET image biomarkers (intensity and textural) improved the prediction of xerostomia at 12 months over the reference model (contralateral parotid mean dose and baseline xerostomy). Genomics and proteomics are mostly used for TCP prediction improvement, but can even be used for refinement of NTCP prediction. Identification of distinctive and significant factors in patients population allows more and more accurate risk stratification towards individualized TCP and NTCP calculation and customized radiation treatment plans. Biotechnological and computational tools represent an intellectual challenge for radiation oncologists that historically are the physicians more prone to face physical and mathematical problems. The necessarily tight cooperation with our physicist companions is today requested more than ever. SP-0651 Clinical potential of human induced- pluripotent stem cells in the management of normal tissue radiation damage M. Benderitter 1 1 Institut de Radioprotection et de Sûreté Nucléaire, Laboratory of Radiobiology & Experimental Radiooncology, Fontenay-aux-Roses- Paris, France Cell therapy was demonstrated of main importance in the management of normal tissue radiation damage. Preclinical and clinical trial data suggest that mesenchymal stem cells (MSCs) are a practical and safe source of cells for stem cell-based therapies of severe tissue damage consecutive to radiation overexposure. MSCs were shown to migrate to damaged tissues supporting wound healing through a “cell drug” mode of action restoring skin and gut functions after irradiation. However, technical limits associated with large-scale ex vivo expansion indicate that alternative source is required to obtain sufficient cell numbers of the appropriate lineage to treat patients with severe disease. Based on their pluripotency and unlimited expansion potential, induced pluripotent stem cells (iPSCs) are considered a promising resource for regenerative medicine. Like naturally occurring stem cells, these artificially induced cells can self-renew and develop into almost any cell in the body (pluripotency). Clinical iPSC banks of selected universal donors should allow their use for large scale allogeneic grafts. The hiPSC-derived MSCs possess important advantages, including the capacity to generate a virtually unlimited supply of therapeutic cells and control differentiation to ensure optimum safety and potency before transplantation, which could in turn overcome the drawbacks of current MSC therapy. We demonstrate that 1) hiPSC-derived MSCs could be expanded using GMP- grade culture conditions, 2) by using a mouse model of radio-induced skin injury, hiPSC-derived MSCs have a therapeutic benefit for skin wound healing similar to adipose-derived MSCs. Moreover, our group describes a GMP-grade system to produce hiPSCs, a cell population capable of reconstituting human hematopoiesis. We demonstrate that i) hiPSC-derived hematopoietic stem cells (HSCs) from healthy donor are capable of reconstituting a functional human hematopoiesis in a radio-induced aplasia preclinical model, ii) hiPSC-derived HSCs from aplastic anemia patients or acute leukemia affected patients retain this ability. Our study prepares a new approach of autologous graft (from the cells of the patient) of cells for healthy tissue damage after radiation exposure. It could potentially pave the way to the constitution of universal banks of stem cells, which would radically increase the capacity of support and treatment of tissue exposed to high doses of ionizing radiation and in the management of chronic late radiotherapy side effects.

tissue complications are the clinical endpoints considered by a radiation oncologist. In a judgement of a treatment plan designed for a given patient, the trade-off between tumor control probability (TCP) and normal tissue complication probability (NTCP) is the kingmaker of the decisional process. Modern clinical radiobiology started with the use of 3D treatment planning implying a different geometry of the beam entry in the body compared to the conventional parallel opposed technique. Emami et al. charted in the seminal paper of 1991 normal tissue tolerance in the case of partial volume and/or inhomogeneous organ irradiation. The few available toxicity records reported in the scientific literature were very heterogeneous and qualitatively poor. However, with all these limitations, the information gathered in that article enriched by the personal experience of expert radiation oncologists, resulted in proposed dose/volume effect for 1/3, 2/3 and whole volume of 28 critical sites. The data were fundamental for the birth of the Lyman-Kutcher-Burman (LKB) and the relative seriality models, mathematical radiobiological tools largely used by the radiation oncology community for plan evaluation. These predictive models centered on the relation between dosimetric information, represented by the graphical simplification of dose-volume histograms (DVH), and clinical endpoints considered (i.e. NTCP). The weak points of these useful tools consisted in a large approximation of the prediction due to the low quality of the toxicity information, and the fact that the relation was made just with a single virtual uniform dose created by an algorithm to translate heterogeneous distribution of a DVH to uniform dose to a part or a whole of an organ at risk (OAR). In 2010 the quantitative analysis of normal tissue effects in the clinic (QUANTEC), updated the Emami’s information with almost twenty years more of clinical experience with 3D-conformal and intensity modulated radiation therapy. Moreover, the report gave some practical guidance in classifying toxicity risk and suggested studies design to improve the knowledge about the issue. Meanwhile several researchers involved in the radiation oncology discipline proposed predictive models taking into account more than one dosimetric parameter, and patient specific, treatment, topographic, or imaging information in the determinism of a given toxicity. An important contribution to these more complex prediction models came from the work of El Naqa et al. in 2006 that elegantly described alternative methods for building multivariable dose–response models. Behind the tools for data collection and analysis (i.e. CERR, Computational Environment for Radiotherapy Research) and the sophisticated statistical methodology, their work represents a very good source of information on the modalities to report and collect data about toxicity to build NTCP models. In this framework, our group studied homogeneously treated Hodgkin’s lymphoma patient populations testing for relation among a single clinical endpoint and multiple clinical and dosimetric factors. For hypothyroidism, we showed that the absolute volume of thyroid gland exceeding 30 Gy in combination with thyroid gland volume and gender obtained the best prediction power. For hearth valve defects, we found a consistent association with lung volume. Interestingly, a model for lung fibrosis prediction included mass of hearth receiving dose > 30 Gy. The technology progress and the advanced computational systems generated the “omics” science, with the possibility to relate clinical endpoints even with genomic, proteomic and radiomic factors. We topographically identified by a voxel-based approach (2017) the cervical portion of the esophagus and the cricopharyngeus muscle as regions with significant differences in dose distribution between head and neck cancer patients that developed or not radiation induced acute disphagia. van Dijk et al. (2017) found that 18 FDG-