Impact of Dose Calculation Algorithm on Skin Dose for Breast Cancer Treated With Intensity Modulated Proton Therapy.

PURPOSE/OBJECTIVE(S) Historically, pencil beam algorithms (PBA), have been used for intensity-modulated proton therapy (IMPT) dose calculation. Recently, improvements in calculation speeds and a desire to generate more accurate plans have led to wider adoption of Monte Carlo (MC) algorithms for IMPT. Our goal was to ascertain the impact of both PBA and MC calculation algorithms on the skin dose in breast cancer treated with IMPT by means of measurements. MATERIALS/METHODS Twenty-one comprehensive chest wall patients were included in this study with treatment planning performed using a treatment planning system. All IMPT plans were generated with 1 or 2 en-face fields. Single- and multiple-field-optimization were used in 19 and 2 patients, respectively. The prescription doses were 50.4 Gy for 15 and 45 Gy for 6 patients, all at 1.8 Gy/fraction. The skin OAR was defined as a 3 mm thick rind anterior to the chest wall target volume. For 12 patients, the skin dose was optimized with the objectives of V95% < = 50%, V100% < = 5%, and D0.03cc < = 50.9 Gy. For the remaining 9 patients, the skin dose was optimized to control D0.03cc alone. The physical skin dose was measured with a parallel plate ionization chamber and a 2D ionization chamber array device at various depths in solid water ranging from 0.2 to 1.3 cm. The agreement of calculated to measured dose was evaluated by 2D gamma analysis (3%/3mm) and point dose comparison. RESULTS The average gamma passing rates for PBA and MC plans were 94% and 97%, respectively. The point dose comparison showed that the measured skin doses were systematically lower than the calculated dose in PBA plans by an average of 3.1% [range: -6.76% to 0.03%], whereas the measured doses were systematically higher than the calculated dose in MC plans by an average of 1.3% [range: -0.58% to 3.51%]. For the patients who received 50.4 Gy, the skin mean dose and D1% were 48.1 [range: 45.8 to 49.6] and 50.8 [range: 49.0 to 52.6] Gy. For the patients who received 45 Gy, the skin mean dose and D1% were 45.0 [range: 44.2 to 46.2] and 46.8 [range: 46.1 to 48.0] Gy. CONCLUSION This study demonstrated that comparing with the PBA algorithm generated plans, the MC generated plans yield a higher gamma passing rate (97% for MC vs. 94% for PBA), and a narrower 95% of confidence interval of skin dose (4.1 for MC vs. 6.8 for PBA). Measurements showed that MC computed plans systematically under-estimate the skin dose by an average of 1.3%, in contrast to PBA computed plans which systematically over-estimate the skin dose by an average of 3.1%. The clinical implications are that for individual patient, PBA could overestimate skin dose considerably more than MC. Therefore, cautions for skin dose should to be taken when PBA is replaced by MC in clinical practice, and it is recommended that appropriate optimization objectives be used during treatment planning to control skin dose.