Abstract

Sara Trivellato¹, Paolo Caricato^1,2, Valeria Faccenda^1,2, Gianluca Montanari¹, Valeria Tremolada¹, Raffaella Lucchini^3,4, Denis Panizza^1,4, Stefano Arcangeli^3,4, Elena De Ponti^1,4

¹ASST Monza, Medical Physics Department, Monza, Italy; ²University of Milan, Department of Physics, Milano, Italy; ³ASST Monza, Department of Radiation Oncology, Monza, Italy; ⁴University of Milan Bicocca, School of Medicine and Surgery, Monza, Italy

Nowadays, when dose constraints are fulfilled, medical physicists rely on their personal experience to evaluate treatment plan quality and no objective criteria are available to determine if the optimal plan has been obtained. Moderate hypofractionated prostate radiotherapy plans have been analysed to determine customized rectal and bladder dose constraints achievable for a given patient’s anatomy prior to optimization. 

Fifty treatment plans have been retrospectively selected. To quantitatively analyse volumetric relationships, an expansion-intersection volume (EIV) for rectum and bladder has been defined as the intersection volume between the organ at risk (OAR) isotropically expanded by 5 mm and the target (rectum EIV and bladder EIV). An empirical threshold line representing a feasible improvement in dosimetric parameters due to favorable patient anatomy has been traced for the rectum-V_46Gy and physicists were asked to improve those plans lying above it. Three plans over nine registered a feasible improvement. The remaining six plans were analyzed to justify their exclusion (sum rank test). The updated bladder and rectum dataset has been used to investigate linear regression as a function of EIVs and OAR volumes. Furthermore, data have been quantitatively analyzed performing a ROC analysis to estimate a bladder volume threshold above which a bladder constraints violation is not expected.

The excluded six plans showed a statistically significant lower bladder volume (p=0.0005) and bladder minus bladder EIV (p=0.0002). Regression analysis provided reasonable linear correlations: rectum-V_{37 Gy} and -V_{46 Gy} vs rectum EIV with R²=0.50 and R²=0.70, respectively (Figure 1); bladder-V_{41 Gy} and -V_{48 Gy} vs bladder volume with R²=0.43 and R²=0.41, respectively; bladder-V_{41 Gy} and -V_{48 Gy} vs bladder minus bladder EIV R²=0.49 and R²=0.47, respectively. As far as concerned the bladder evaluation, the ROC analysis showed an area under the curve of 0.70 [0.52- 0.88] and 0.80 [0.62- 0.97] for the bladder volume and the bladder minus bladder EIV, respectively (Figure 2). A bladder volume threshold of 112 cm³ showed a sensitivity of 79% and specificity of 50% (accuracy 75%) with a positive predictive value (PPV) of 91% and a negative predictive value (NPV) of 27%. A bladder minus bladder EIV of 63 cm³ showed a sensitivity of 95% and specificity of 50% (accuracy 89%) with a PPV of 92% and a NPV of 60%.

Highlighted linear relationships suggest institution-specific criteria to facilitate the plan optimization process with considerably lower constraints for small OAR-target intersection and the ROC analysis suggests a minimum bladder volume value to facilitate constraint fulfillment. This method helps to standardize institution-specific plan quality and consistency starting from OAR-treatment preparation and proximity to the target and it can be applied to different treatment schemes.