Purpose: Craniospinal radiation (CSRT) followed by a boost to the entire posterior fossa (PF) is standard postoperative therapy for patients with medulloblastoma. A large proportion of recurrences after treatment are local, with approximately 50-70% of recurrences occurring in the PF. It is unclear, however, whether these failures are occurring in the original tumor bed or outside the tumor bed, but still within the PF. With improved diagnostic imaging, better definition of tumor volumes, and the use of three-dimensional conformal therapy (3D CRT), we may be able to restrict the boost volume to the tumor bed plus a margin without compromising local control. This retrospective study analyzes the patterns of failure within the PF in a series of patients treated with radiation therapy (RT).
Methods: From July 1986 through February 1996, 114 patients >18 months and <18 years with medulloblastoma were treated at the University of Michigan and Children's Hospital of Philadelphia, with RT following surgical resection. Of 114, 27 (24%) were found to have a recurrence and form the basis for this study. RT consisted of CSRT followed by a boost to the entire posterior fossa. Some patients received adjuvant chemotherapy. Patient's preoperative magnetic resonance imaging (MRI) and/or computerized tomography (CT) studies were used to compare the original tumor volume with the specific region of local relapse. Failure was defined as MRI or CT evidence of recurrence or positive cerebrospinal fluid cytology. Relapse was scored as local, if it was within the original tumor bed, and regional if it was outside of the tumor bed but still within the PF.
Results: The median age of the 27 patients who relapsed was 8.6 years. Three patients were <3 years old. Of 27, 21 had disease localized to the PF. Of 26, 22 patients received chemotherapy during their treatment regimen; 1 patient did not have information on systemic treatment. The median dose of RT to the craniospinal axis was 32.5 Gy and to the PF was 55.2 Gy. The median time to recurrence was 19.5 months. Local failure within the tumor bed as any component of first failure occurred in 52% (14 of 27) of all failures, but as the solitary site of first failure in only 2 of 27 failures. Of 14 patients who failed in the tumor bed, 11 also failed in the spine, 8 of 14 also failed within the PF but outside the tumor bed, and 7 of 14 failed in all three locations. Local failure within the PF but outside the tumor bed as any component of first failure occurred in 41% (11 of 27) of all failures, but as the solitary site of first failure in only 1 of 27 failures. Of 11 patients who failed in the PF but outside the tumor bed, 9 also failed in the spine, 8 also failed within the tumor bed, and 7 failed in the all three locations. Of the failures outside the tumor bed but still within the PF, 7 of 11 failed in the leptomeninges, 1 in the brainstem parenchyma, and 3 in the PF parenchyma. Of 7 who failed in the PF leptomeninges, 6 also failed within the spine. Failure within the spine as any component of first failure occurred in 70% (19 of 27) of all failures and as the only site of first failure in 5 of 27 patients. Of 19 patients who failed in the spine, 11 also failed in the tumor bed, 9 also failed within the PF but outside the tumor bed, and 9 failed in the all three locations.
Conclusions: Leptomeningeal failure is a common component of failure and occurs in the leptomeninges of the PF, as well as the spine. Isolated tumor bed failure is a rarely observed event and occurred in only 2 of 27 failures described here. Similarly, parenchymal (nonleptomeningeal) failures in the PF but outside of the tumor bed were rare: 4 patients recurred in this manner, only 1 of whom was an isolated event without other sites of recurrence. Our data suggest that, when the entire PF is treated, very few failures develop in isolation in the PF outside the tumor bed. Further studies will be necessary to determine if RT to the tu