Optimized planning target volume for intact cervical cancer

Purpose: To model interfraction clinical target volume (CTV) variation in patients with intact cervical cancer and design a planning target volume (PTV) that minimizes normal tissue dose while maximizing CTV coverage. Methods and Materials: We analyzed 50 patients undergoing external-beam radiotherapy for intact cervical cancer using daily online cone-beam computed tomography (CBCT). The CBCTs (n = 972) for each patient were rigidly registered to the planning CT. The CTV was delineated on the planning CT (CTV ₀) and the set of CBCTs ({CTV ₁-CTV ₂₅}). Manual (n = 98) and automated (n = 668) landmarks were placed over the surface of CTV ₀ with reference to defined anatomic structures. Normal vectors were extended from each landmark, and the minimum length required for a given probability of encompassing CTV ₁-CTV ₂₅ was computed. The resulting expansions were used to generate an optimized PTV. Results: The mean (SD; range) normal vector length to ensure 95% coverage was 4.3 mm (2.7 mm; 1-16 mm). The uniform expansion required to ensure 95% probability of CTV coverage was 13 mm. An anisotropic margin of 20 mm anteriorly and posteriorly and 10 mm superiorly, inferiorly, and laterally also would have ensured a 95% probability of CTV coverage. The volume of the 95% optimized PTV (1470 cm ³) was significantly lower than both the anisotropic PTV (2220 cm ³) and the uniformly expanded PTV (2110 cm ³) (p < 0.001). For a 95% probability of CTV coverage, normal lengths of 1-3 mm were found along the superior and lateral regions of CTV ₀, 5-10 mm along the interfaces of CTV ₀ with the bladder and rectum, and 10-14 mm along the anterior surface of CTV ₀ at the level of the uterus. Conclusion: Optimizing PTV definition according to surface landmarking resulted in a high probability of CTV coverage with reduced PTV volumes. Our results provide data justifying planning margins to use in practice and clinical trials.