Verification Using Osld In Vivo Dosimetry of the Predicted Skin Surface Dose in Patients Receiving Post-Mastectomy Radiotherapy

Abstract

In patients receiving post-mastectomy radiotherapy (PMRT), the skin overlying the chest wall is considered at-risk of tumour recurrence, often necessitating the use of bolus material. The accuracy with which three-dimensional treatment planning systems predict radiation dose delivered to the skin surface is poorly characterised. The purpose of this study was to evaluate whether dose to the skin surface, underneath bolus, was accurately predicted by a commercially available 3D-treatment planning system in the PMRT clinical setting. Optically stimulated luminescent dosimeters (OSLDs) are a widely used method of in vivo dosimetry, with their use at the skin surface having been examined previously. However, their use in the evaluation of skin surface dose in the setting of PMRT has not previously been documented. In vivo dosimetry using OSLDs was performed prospectively in 20 patients receiving PMRT following ethics approval. An array of 9 OSLDs were applied in a 3x3 grid arrangement, over a 10x10cm field on the ipsilateral chest wall or neobreast (non-autologous reconstruction post-mastectomy), referenced to the midline tattoo. Dosimetry data were recorded on three separate days during the first half of the PMRT course (50 Gy/25#) for each patient. The 9 in vivo dosimetry points corresponded to the same anatomical location on the 3D treatment planning DICOM image, 0.45mm below the skin surface, from which the predicted dose was calculated using the departmental planning algorithm. Estimated bias was calculated using the mean difference between the planned and measured dose, and a p-value of a t-test for systematic bias given for each point. Bland-Altman limits of agreement are stated, which are the values between which 95% of within subject differences in measurement are expected to fall. The coefficient of variation (CV) of the 3 OSLD readings at each of the 9 points was low for 8 points (≤ 4.4%) demonstrating comparable dose-received per fraction at each point; only the inferior-medial point CV was higher at 7.6%. The mean difference for each of the 9 points ranged between 0.47-1.30 Gy (0.94%-2.6% of prescribed dose), representing less dose received compared to the planned dose. For 7 of 9 points, the mean estimated bias was < 1 Gy of predicted radiation dose. Bland-Altman limits of agreement were widest at the superior medial position ( -5.98Gy to 7.70Gy). T-test for systematic bias was significant (p=0.03) in only one of the 9 points (inferior-lateral point). Our data indicate that the skin surface dose underneath bolus is predicted with a reasonable level of accuracy utilising a commercially available 3D treatment-planning system. Bland-Altman limits of agreement were wide, indicating a limitation of the study being the limited number of fractions (3 of 25) from which our data could be extrapolated.