ESTRO 2022 - Abstract Book

S252

Abstract book

ESTRO 2022

treatment. Plans were independently recalculated to assess dose calculation errors, delivery log files were used to evaluate delivery errors, and contemporaneously conducted beam output audits were used to evaluate machine output errors. The errors from each source were used to characterize the dose deviations that exist between the institution’s TPS dose and measured TLD dose, to determine how much of that deviation was accounted for by the error sources investigated. A dose difference metric, D, was used to describe the result, where a positive D value indicated the presence of true error, and a negative D value did not. Phantom results were evaluated in 2 groups, separated by a threshold total dose deviation (3.2%) (Kirby, 1992) to account for measurement uncertainty in phantom TLD doses. No delivery error was assessed for spine due to lack of log file data. Results Among all 3 phantoms, cases with absolute dose deviation within measurement uncertainty (< 3.2%), expectedly had a negative average D value, showing that not much, if any, of the error could be described by the categories investigated. Phantom results with substantial dose deviation (> 3.2%) had positive D values in all categories, indicating that some of the existing dose deviations were accounted for by the error modes we evaluated. Output error was the largest category of error for the lung phantom (16%), whereas dose calculation error was the major contributor for the H&N (49%) and spine (40%). The total magnitude of error quantified among these phantoms, i.e., the % of absolute dose deviation accounted for, was 20.5%, 68.3% and 70.9% for lung, H&N, and spine respectively (Fig. 1). These findings were further emphasized by the positive correlations found between absolute dose deviation and error type per phantom: lung = output (Pearson correlation; r = .54, p < .01), H&N = dose calculation (r = .72, p < .01) (Fig. 2), spine = dose calculation (r = .59, p < .01).

Conclusion Errors in radiotherapy remain prevalent and require attention and resolution in the community. For phantom results with non-negligible dose deviation, we identified 20.5%, 68.3% and 70.9 % of errors for the lung, H&N, and spine phantom respectively. Dose calculation and machine output errors both contributed substantially to the phantom error. As these are common sources of error, the radiation oncology community should pay particular attention to them.

Proffered Papers: Haematology

OC-0291 Patterns of failure in primary CNS lymphoma: the MSKCC experience from 2004-2018

K. Tringale 1 , C. Grommes 2 , B. Vachha 3 , H. Hubbeling 1 , A.N. Wijetunga 1 , C. Hajj 1 , J. Yahalom 1 , B. Imber 1

1 Memorial Sloan Kettering Cancer Center, Radiation Oncology, New York, USA; 2 Memorial Sloan Kettering Cancer Center, Neuro-Oncology, New York, USA; 3 Memorial Sloan Kettering Cancer Center, Radiology, New York, USA Purpose or Objective Treatment of primary CNS lymphoma (PCNSL) has rapidly evolved, emphasizing improved efficacy while reducing toxicity. Despite advances, many patients still relapse after first-line therapy. Clinical predictors and patterns of failure are critical to optimize and personalize induction and consolidation regimens and have been poorly studied in the modern era. Our objective was to evaluate PCNSL patterns of failure after contemporary treatment. Materials and Methods PCNSL patients treated at Memorial Sloan Kettering Cancer Center between 2004-2018 were evaluated. Initial site(s) of T1 post-contrast-enhancing disease on baseline MRI was characterized by laterality, uni- vs multifocality, supra- vs infratentorial, and deep nuclei, CSF, and/or radiographic meningeal involvement. For patients who received consolidation, disease status (radiographic evidence of contrast-enhancing disease [EOD] vs no evidence of disease [NED]) was evaluated