Factors Affecting Setup Errors In Three-Dimensional Conformal Radiotherapy Of Thoracic Tumor

BACKGROUND AND PURPOSE: Margins between the clinical target volume and planning target volume are used in radiotherapy to compensate setup errors, which may influence the radiobiological effects of the treatment. This article evaluates patient-related factors affecting setup errors in three-dimensional conformal radiotherapy (3D-CRT) of thoracic tumor using an electronic portal imaging device (EPID) and calculates the individualized optimal setup margins for target in the mediastinum.METHODS AND MATERIALS: This study analyzed 552 records of 136 patients with thoracic tumor (major target in the mediastinum)were consecutively treated with 3D-CRT from May 2008 to Mar 2009. Digitally reconstructed radiographs (DRRs) were created for the orthogonal portals taken at the isocenter. Treatment verification was done using an amorphous silicon detector portal imaging device for using orthogonal portals and the DRR was used as a reference image. An image matching software was used to match the anatomical landmarks in the DRR and the portal imaging and the displacement of the portals in x, y axis and rotation were noted in the anteroposterior (AP) and lateral images. Electronic portal imaging was repeated weekly and an median of 5 images per patient was recorded. The means and standard deviations of positioning offset in all the directions were calculated as individual systematic and random errors for the each patient respectively. Population systematic (Sigma) and random errors (sigma) come from the standard deviations of individual systematic and random errors respectively. Multiple linear regression was applied for multivariate analysis by using individual systematic and random errors in all the directions as independent variable and using gender, entity, location of lesion and total lung volume as the dependent variables simultaneously. Stepwise regression were applied for screening of independent variables. Entry and removal probilities of F are 0.10 and 0.15 respectively. The uncertainty margins were calculated using a dose-probability-based margin recipe.RESULTS: The medians of translational setup errors from all data were 0cm (-1.44～1.34cm), 0.1cm (-1.55～1.80cm), and 0cm (-1.19～1.13cm) in the left-right(LR), superior-inferior (SI) , and anterior-posterior (AP)directions, respectively. The medians of rotations around the AP and LR axes were 0 degree (-3.50～1.81 degree) and 0 degree (-6.21～1.11 degree), respectively. Margins for the LR, SI and AP directions were 0.78cm (Sigma =0.26, sigma = 0.18), 1.03cm (Sigma =0.36, sigma = 0.20), and 0.61 cm (Sigma = 0.20, sigma = 0.16) , respectively. Statistically significant correlations were observed in location of lesion (R2=0.030, P=0.042), gender (R2=0.029, P=0.048) and total lung volume (TLV) (R2=0.032, P=0.036) for individual random errors in SI directions (1.03cm and1.02cm for leisions above and below hilum of lung respectively), individual systemetic errors in AP (0.49cm and 0.64cm for male and female respectively) and individual random errors in AP directions (0.60cm and 0.62cm for TLVs above and below 3 500cc respectively), respectively.CONCLUSIONS: Anisotropic margins for thoracic tumor patients received 3D-CRT in the LR and SI directions were 0.78cm and 1.03cm, respectively, at least. Gender were independent factor influencing setup errors for these patients. Gender-specific safety margins should be 0.49cm and 1.64cm in the AP direction, respectively, at least.