Our series represents the largest in the published literature of RT for treatment of BM in EOC, with a total of 60 patients analyzed. We show that these patients can be effectively treated with RT, with or without resection of tumor, and that survival in these patients depends on KPS, number of BM, presence of LMD, and presence of uncontrolled primary tumor. We found no statistically significant effects of age, tumor histology, grade, initial disease stage, RT type, or use of surgical resection of metastases on OS after diagnosis of BM.
While CNS metastases are rare in EOC, they appear to be increasing in incidence as extracranial disease is better controlled with modern chemotherapy. In our series, patients with more advanced stage and grade comprise the majority of patients who develop BM. The most common presentation was a single BM in 47% of patients. Patients with single BM may be treated with surgery, SRS, PBRT, and/or WBRT. In our series, seven patients treated with SRS alone had a median OS of 60.2 months. While this patient subset is small, the long survival is noteworthy in a metastatic population, and likely indicates that these patients were highly selected with good control of extracranial disease and high KPS. Another smaller series has found longer survival in well-selected patients treated with SRS versus WBRT . Patients treated with only WBRT tend to have shorter median survival times , and this may be reflective of overall poorer KPS, stage, and number of BM.
The use of surgery and RT has been associated with longer survival after BM in multiple smaller studies [3–5, 8, 13, 23–25]; however, our study found no significant benefit to the use of surgery in 22 patients who received postoperative RT. Pothuri et al.  published a series from our institution of 14 patients treated with craniotomy and postoperative RT, and found median survival of 18 months and high rates of local control. Based on our current study and those preceding, we recommend that patients with a single BM, good KPS and limited extracranial disease be considered for SRS or both craniotomy and RT as clinically indicated. Ideally, single-modality SRS will be tested in more patients with EOC and single BM to come to a better understanding of the adequacy of SRS alone. In patients with multiple BM, some data suggest that resection of multiple metastases may improve outcomes [26, 27]; we could not verify this finding in our population, as only two patients fell into this category.
Our results add to the findings of the Radiation Therapy Oncology Group (RTOG) Recursive Partitioning Analysis (RPA) system  for BM in a variety of cancers, which identified KPS ≥70, age <65 years, controlled primary carcinoma, and no extracranial systemic metastases as being predictive of the longest survival after BM. We suspect that age was not significant in our patients because the majority of our patients fell under the 65-year-old cutoff used by the RPA.
A large but older series of EOC BM includes 72 patients from MD Anderson Cancer Center  treated heterogeneously since 1985, and none with upfront SRS or PBRT. In that series, 8 (11%) patients received steroids alone, 35 (51%) WBRT alone, 8 (11%) surgery alone, and only 12 (17%) had both surgery and WBRT. Twenty-five patients had a single BM, and 47 had multiple. The study does not provide information on the use of salvage therapies for CNS recurrence. Their patients had inferior outcomes to those in our study; median OS was just 6.9 months in patients with a single BM. On their univariate analysis of potential prognostic factors, they did not include KPS or primary tumor control, both of which are vitally important risk factors for outcomes after BM, and are important for determining treatment approach. Although comparing outcomes between retrospective studies is challenging, several possibilities may explain why our results are superior to this series. First, almost half of our population had a single BM, while two-thirds of the patients in the MD Anderson series had multiple metastases. Second, none of our patients were treated with steroids alone, a group that had a significantly worse hazard ratio of death and accounts for 11% of their population. Third, 37% of our patients received multimodality therapy that included surgery and RT, while only 17% of patients in their study were treated in a similar fashion and represented the group with the longest median survival. Surgical resection of a single BM has been shown to improve survival in a randomized clinical trial ; in this MD Anderson series, it does not appear that all patients with single BM received surgery, and none received SRS. Fourth, almost two-thirds of our patients were treated with chemotherapy following treatment of their BM, which may contribute to better outcomes (although we were unable to include it in our Cox regression analysis due to its vast heterogeneity). Fifth, it is not clear that patients with relapse received salvage therapy, while most of our patients received RT or chemotherapy at CNS relapse. Lastly, patients treated at our institution are followed closely and systematically for recurrence, and are treated with early salvage therapy at time of relapse; it is unclear what the standard follow-up entailed in other studies.
Chen et al.  used the RTOG RPA classification system in a population of 19 patients with EOC BM treated with various approaches, and found that surgery was associated with longer OS (33.7 months vs 7.4 months, p = 0.006). However, only 9 patients underwent surgery, and 8 of these also received adjuvant radiotherapy to the brain, making the patient population too small to draw definitive conclusions regarding the adequacy of surgery alone for brain metastases in this population. Their UVA found primary tumor control (p = 0.006) and number of BM (p = 0.005) to be associated with OS, but the number of events was too small to perform a true multivariate analysis. In our study, we also find that primary tumor control and number of BM are important. Our patient population differs in that we have three times as many patients, and all are treated with RT. In addition, our larger patient population likely includes patients that are less highly selected than in the Chen study. In fact, Chen et al. concede in their discussion section that the relatively long overall survival of their study (median, 16.3 months) may be attributed to patient factors that are pertinent to outcome but not accounted for in the study, such as systemic therapy and institutional preference for aggressive, multimodality therapy.
A recent multi-institutional retrospective study of 139 patients with brain metastases from various gynecologic malignancies identified a group of 56 patients with ovarian, fallopian tube, or peritoneal histology  (it is unclear how many of these patients had a true epithelial ovarian carcinoma). While comparing this “ovarian” subgroup to our patients is not entirely possible, given the heterogeneity of their subgroup, the study has several noteworthy findings. First, in the “ovarian” group, 80% received RT, about half of which also received surgery and/or chemotherapy for the BM. Second, median survival for the 56 patients was 12.5 months. Third, on a multivariate analysis across all gynecologic types, they found ovarian/tubal/peritoneal disease origin was associated with improved survival and recommended these patients be treated more aggressively when a BM is diagnosed.
In our series, 10 (20%) patients with BM also had LMD, and 9/10 (90%) had a synchronous or prior BM. LMD is an uncommon occurrence in advanced malignancy, estimated to occur in 4-15% of solid tumors . Previous series have reported an incidence of synchronous or preexisting CNS metastases in 28-75% of patients with LMD . Clarke et al.  reviewed a large series of LMD from our institution and reported that 70% of patients with LMD from solid tumors had previous or current brain disease. In that study, 59% of patients were treated with RT, 15% received supportive therapy only, and the remainder had some form of chemotherapy. Median OS after LMD was 2.3 months for patients with solid tumors; there were 2 cases of ovarian carcinoma. Another large series of 155 patients with LMD treated from 1980 to 2002 found a median OS in solid tumors of 2.8 months in non-breast solid tumors . Only one patient had ovarian cancer.
We found that median OS after LMD when present at initial BM diagnosis (n = 6) was 3.55 months (95% CI, 0.69-15.8). The patients in our series who developed LMD after a preceding BM diagnosis had no higher representation of one primary treatment modality, either surgery or RT, prior to developing LMD. Our patient population is different from these other large studies for several reasons. First, LMD was primarily treated with RT in our study, in contrast to other reports that frequently use intrathecal or systemic chemotherapy as the principal treatment for LMD [30, 31]. In one series of 31 patients treated with intrathecal chemotherapy, the response rate was 52% . In contrast, our LMD patients treated with RT who had follow-up imaging (n = 6) demonstrated 100% overall response rate (both complete and partial response) as seen on their first follow-up MRI. Second, all patients had LMD diagnosed by MRI, with or without CSF analysis. Lastly, our study represents a homogenous population of one disease, as opposed to the existing series that include one or two ovarian cancers mixed in with various other solid tumor types. Based on the favorable responses to RT and slightly longer survival in our series, we support administering RT for EOC LMD. EOC is known to be a biologically radiosensitive disease, and our results in the setting of LMD appear to confirm this fact.
There are some limitations of our study. First, we performed a retrospective analysis, which has its own well-known inherent drawbacks and biases, many of which have been outlined above. Second, not all patients in our study had pathologic confirmation of BM from EOC, which could theoretically have led us to include patients with primary CNS tumors or benign disease in this cohort. Third, because it was difficult to determine cause of death for many patients in this retrospective analysis, we could not determine BM-specific survival, which would be a relevant measure of outcome in this population. Lastly, while we were able to determine which patients received any systemic chemotherapy before and after RT for BM, we were unable to routinely quantify the number of cycles of chemotherapy they received and the specific agents with which they were treated. Prospectively collected data on chemotherapy is required to draw strong conclusions about the efficacy of systemic therapy in this patient population.