Fax: (216) 445-1068
Salvage stereotactic radiosurgery effectively treats recurrences from whole-brain radiation therapy
Article first published online: 8 SEP 2008
Copyright © 2008 American Cancer Society
Volume 113, Issue 8, pages 2198–2204, 15 October 2008
How to Cite
Chao, S. T., Barnett, G. H., Vogelbaum, M. A., Angelov, L., Weil, R. J., Neyman, G., Reuther, A. M. and Suh, J. H. (2008), Salvage stereotactic radiosurgery effectively treats recurrences from whole-brain radiation therapy. Cancer, 113: 2198–2204. doi: 10.1002/cncr.23821
- Issue published online: 3 OCT 2008
- Article first published online: 8 SEP 2008
- Manuscript Accepted: 6 JUN 2008
- Manuscript Revised: 3 JUN 2008
- Manuscript Received: 12 FEB 2008
- brain metastases;
- whole-brain radiation therapy;
- stereotactic radiosurgery;
- gamma knife;
The purpose of the current study was to examine overall survival (OS) and time to local failure (LF) in patients who received salvage stereotactic radiosurgery (SRS) for recurrent brain metastases (BM) after initial management that included whole-brain radiation therapy (WBRT).
The records of 1789 BM patients from August 1989 to November 2004 were reviewed. Of these, 111 underwent WBRT as part of their initial management and SRS as salvage. Patients were stratified by Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis class, primary disease, dimension of the largest metastases and number of BM at initial diagnosis, and time to first brain recurrence after WBRT. Overall survival, survival after SRS, and time to local and distant failure were analyzed.
The median OS from the initial diagnosis of BM was 17.7 months. Median survival after salvage SRS for the entire cohort was 9.9 months. Median survival after salvage SRS was 12.3 months in patients who had their first recurrence >6 months after WBRT versus 6.8 months for those who developed disease recurrence ≤6 months after (P = .0061). Primary tumor site did not appear to affect survival after SRS. Twenty-eight patients (25%) developed local recurrence after their first SRS with a median time of 5.2 months. A dose <22 grays and lesion size >2 cm were found to be predictive of local failure.
In this study, patients who recurred after WBRT and were treated with salvage SRS were found to have good local control and survival after SRS. WBRT provided good initial control, as 45% of these patients failed >6 months after WBRT. Those with a longer time to failure after WBRT had significantly longer survival after SRS. Cancer 2008. © 2008 American Cancer Society.
Brain metastasis (BM) is the most common neurologic complication of cancer.1 The incidence of BM has been increasing because patients are living longer after the diagnosis of their primary disease. Typically, patients present with neurologic deficits and headaches that can be disabling. The diagnosis of BM usually carries a poor prognosis, with a median survival of approximately 4 months.2 Prognosis is influenced by multiple factors, which include histology, generalized performance status, age, control of primary disease, and lack of extracranial metastases. These factors constitute the Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) classification described by Gaspar et al.2
Prognosis is also influenced by therapy, which may include whole-brain radiation therapy (WBRT), surgical resection, or stereotactic radiosurgery (SRS). In a study by Patchell et al3 for single BM, the addition of surgery to WBRT improved overall survival and decreased local recurrence. The omission of WBRT after surgery, however, increases the risk of local and distant brain recurrence.4 Although there was no statistically significant difference in overall survival between the 2 groups, that study was not powered to detect the difference.
SRS is an alternative modality that may improve local control after WBRT. A prospective, randomized controlled trial was completed for patients with 1 to 3 BM comparing WBRT with WBRT with SRS boost. With the addition of the SRS boost, local control increased from 71% to 82% (P = .0132).5 Those undergoing SRS boost were more likely to have stable or improved Karnofsky performance status (KPS) at 6 months. There was also a modest improvement in survival, particularly in patients who were RPA class 1 (with a mean survival of 11.6 months) or possessed a favorable histology, defined as squamous or nonsmall cell histology, usually observed in lung cancer (with a mean survival of 5.6 months). Another study assessed the role of SRS in addition to WBRT for 2 to 4 BM and found that the rate of local failure at 1 year was 100% after WBRT alone, but only 8% for those patients who received an SRS boost.6 Survival did not differ between the 2 groups, but the study was not powered for survival. These 2 studies established the role of SRS boost.
To our knowledge, the optimal dose for SRS boost has not been determined prospectively. The RTOG standard of 24 grays (Gy) for tumors measuring ≤2 cm was based on toxicity, not efficacy, for patients in whom planned WBRT was not used.7 This was assessed retrospectively at the University of Kentucky.8 An SRS dose >20 Gy did not improve local control when compared with a dose of 20 Gy. There was a trend toward an increased risk of grade 3 or 4 neurotoxicity with doses >20 Gy versus 20 Gy (5.9% vs 1.9%; P = .078). The authors concluded that 20 Gy is the optimal dose for SRS boost for metastases measuring ≤2 cm when combined with WBRT.
In a Japanese study, patients were randomized to SRS with WBRT or SRS alone.9 There was no difference in survival noted between the 2 arms, but there was a 46.8% recurrence rate in SRS with WBRT arm versus 76.4% for SRS alone (P < .001). Local control and distant brain control were improved (P < .002 for local control and P < .003 for distant control). Salvage was less frequent in the SRS with WBRT treatment arm (P < .001). Despite the results, because the SRS alone arm did not result in neurologic death, SRS alone is an option. However, arguably the addition of WBRT from this study improves control and can be considered for all patients.
Despite good control rates from WBRT, especially with the addition of aggressive local therapy such as surgical resection or SRS, recurrences continue to occur locally and elsewhere in the brain. Repeat WBRT may cause neurotoxicity and is typically reserved for patients with poor prognoses.10 This leaves surgical resection and SRS as the remaining options. Of these 2 options, SRS may be an ideal option because it may be better than surgery at controlling the microscopic disease at the tumor margin that may be left behind after a resection.11 In addition, patients who have had BM are followed closely with serial imaging studies and therefore are more likely to have new metastases discovered when they are small and produce little or no mass effect. These lesions are ideal for SRS. Surgery, conversely, is often reserved for patients with neurologic deficits or symptoms and signs related to mass effect.11 In the current study, we retrospectively reviewed our series of patients who received WBRT as part of their initial management and who later developed recurrences that were treated with SRS.
MATERIALS AND METHODS
The records of 1789 patients with BM diagnosed from August 1989 to November 2004 were reviewed and entered into an Institutional Review Board-approved database. The presence of BM was confirmed by computed tomography (CT) and/or magnetic resonance imaging (MRI). Of these, 148 (8%) patients had WBRT as part of their initial management and subsequently developed a recurrence treated with SRS.
After excluding cases because of incomplete records, in which all the relevant treatment information and/or follow-up information were not attainable, 111 (75%) of the 148 patients comprised the study sample; the remaining 25% had inadequate information to be a part of this study.
These patients may have been treated with surgery and/or SRS as part of their initial management. Recurrence was either local or distant within the brain. We did not include patients treated for persistent disease. Patients had demonstrated progression in size or a new distant BM.
SRS was LINAC-based from 1989 to 1997, after which gamma knife (Elekta, Stockholm, Sweden) SRS was used. All but 1 patient received gamma knife SRS for salvage. SRS was typically dosed as per the RTOG 90-05 protocol.7 Per this protocol, doses were based on the maximal dimension (metastases measuring ≤2 cm received 24 Gy to the tumor margin, metastases measuring >2 cm but ≤3 cm received 18 Gy, and metastases measuring >3 cm received 15 Gy). In some cases, a lower dose was prescribed when the lesion was in the brainstem or next to a critical structure (as low as 9.6 Gy). Occasionally, a higher dose was given for smaller lesions at the discretion of the treating physician (as high as 25.4 Gy). Metastases measuring >4 cm were rarely treated because of concerns for radiation necrosis, but when they were treated a lower dose was given (approximately 12 Gy). No additional margins were used in treatment. Both CT and MRI data were used in planning. During planning, attempts were made to keep the conformality index and homogeneity index to ≤2.
To study which factors were prognostic for overall survival, survival after SRS, and time to local and distant failure, patients were stratified by RTOG RPA class, primary disease, dimension of the largest and number of BM at the time of initial diagnosis, and time to first brain recurrence after WBRT. Primary disease was categorized into breast cancer, nonsmall cell lung cancer, melanoma, renal cell carcinoma, and other. Dimension of the largest BM was categorized into <2 cm, >2 cm to 3 cm, >3 cm to 4 cm, and >4 cm. The number of BM was categorized into 1, 2, 3, or >3. Time to first recurrence was stratified as ≤3 months, >3 to 6 months, >6 months to 12 months, and >12 months.
Local failure was defined as any growth from previous MRI after SRS. If radiation necrosis was suspected, additional imaging in the form of positron emission tomography, single–photon emission CT, or serial follow-up MRI scans were used to establish a diagnosis. Ten patients had histologic confirmation of local disease recurrence. Distant failure was defined as new brain lesions that developed after SRS. Follow-up scans were usually performed at 2–month to 3-month intervals after SRS.
Time to diagnosis of BM and survival were calculated from the date of the first MRI or CT scan documenting BM. Survival was also measured from the date of salvage SRS. The time to distant brain recurrence was defined as the time to neuroimaging evidence of a new BM elsewhere in the brain. Time to local disease recurrence was defined as time to progression of a treated BM on neuroimaging.
The Kaplan-Meier method was used to test differences in overall survival, and time to local and distant recurrence after SRS. Analyses were performed using StatView software (version 5.0; SAS Institute Inc, Cary, NC) and S+.
The median follow-up after SRS was 7.1 months (range, 0.0-57.8 months). The median age of the patients was 56 years (range, 34-81 years). The primary diagnoses were as follows: nonsmall cell lung cancer in 48% (n = 53 patients), breast cancer in 23% (n = 25 patients), melanoma in 9% (n = 10 patients), renal cell in 8% (n = 9 patients), and other in 13% (n = 14 patients). Other primary diagnoses included endometrial, colorectal, sarcoma, ovarian, and unknown primary tumors. At diagnosis, patients presented with a median of 2 BM (range, 1-10 BM). The largest BM at the time of diagnosis ranged from 0.4 to 7 cm (median, 2 cm). Twenty-four patients were RPA class 1, 80 were RPA class 2, and 7 were RPA class 3 at diagnosis. All received WBRT after their diagnosis.
Total WBRT doses were available for 109 patients and doses per fraction were available for 108 patients. The median total WBRT dose was 37.5 Gy (range, 30-50 Gy). The median dose per fraction was 2.5 Gy (range, 1.8-3.0 Gy).
The median time to first recurrence of their BM was 5.7 months (range, 0.4-48 months). Thirty-one (28%) patients had their first recurrence ≤3 months after WBRT, 30 (27%) developed disease recurrence >3 to 6 months, 28 (25%) developed disease recurrence >6 to 12 months, and 22 (20%) developed disease recurrence >12 months after WBRT. Some patients recurred as little as 0.4 months after WBRT; these were confirmed to be true recurrences with growth of lesions noted on MRI. All patients received gamma knife for salvage, except 1 patient who received LINAC-based SRS. The median dose was 23.6 Gy (range, 9.6-25.4 Gy).
The median overall survival (OS) from the initial diagnosis of BM was 17.7 months. The median OS for the various stratifications are summarized in Table 1. RPA class (P = .16), primary disease (P = .28), size of the largest BM (P = .74), and number of BM (P = .92) did not appear to affect OS. The time to first disease recurrence after WBRT did significantly influence OS (P < .0001, univariate analysis).
|Stratification||Median OS, Mo||P|
|Class 1 (n=24)||24.6|
|Class 2 (n=80)||17.6|
|Class 3 (n=7)||10.2|
|Time to first recurrence. mo||<.0001|
Because none of the other variables were found to affect OS, survival from time of salvage SRS was calculated for time to recurrence after WBRT. The median survival after salvage SRS was 9.9 months in all patients. The median survivals after SRS when stratified by histology and time to first disease recurrence are summarized in Table 2. The median survival was 12.3 months in patients who had their first recurrence >6 months versus 6.8 months for those who developed disease recurrence ≤6 months after WBRT (P = .0061). The Kaplan-Meier survival curve is shown in Figure 1.
|Stratification||Median OS, Mo||P|
|Nonsmall cell lung (n=53)||8.2|
|Renal cell (n=9)||6.0|
|Time to first recurrence, mo||.05|
The location of the disease recurrence, whether it was a local recurrence or distant recurrence that occurred after WBRT, did not appear to affect survival after SRS. The Kaplan-Meier survival curve is shown in Figure 2. The median survival after SRS was 10.4 months for patients with local only recurrence, 12.9 months for distant only recurrence, and 7.3 months for those who fail locally and distantly after WBRT.
Twenty-eight (25%) of the patients developed a local failure (LF) at a site of their first salvage SRS, at a median of 5.2 months. The local control rate at 1 year was 68% and that at 2 years was 59%. Table 3 compares median time to LF stratified by histology and time to first disease recurrence (which includes censoring for patients who did not fail). Histology (P = .49) or time to first disease recurrence after WBRT (P = .14) did not affect time to LF. The size of the metastasis and SRS dose were examined for each lesion. A total of 243 metastases were treated, of which 31 failed. With regard to size, 91% of the metastatic lesions measuring ≤2 cm were controlled at 12 months versus 62% of the lesions measuring >2 cm (P < .0001) (Fig. 3). As for dose, 92% of the metastatic lesions receiving ≥22 Gy were controlled at 12 months versus 72% for those receiving <22 Gy (P = .0013) (Fig. 4). This dose was chosen as a cutoff based on previous data from our institution.12 It is important to note that lesion size and SRS dose are highly correlated factors, because larger lesions are typically treated with a lower dose to control toxicity.
|Stratification||Median Time to LF, Mo||P|
|Nonsmall cell lung (e=12)||15.3|
|Renal cell (e=2)||NR|
|Time to first recurrence, mo||.14|
Thirty-four patients (31%) had distant failure (DF) after SRS, with a median of 14.5 months. The distant control rate was 86% at 1 year, 51% at 2 years, and 30% at 3 years. Table 4 compares median time to DF stratified by histology and time to first recurrence. Histology (P = .57) or time to first recurrence after WBRT (P = .74) did not affect time to DF.
|Stratification||Median Time to DF, Mo||P|
|Nonsmall cell lung (e=12)||27.3|
|Renal cell (e=2)||35.9|
|Time to first recurrence, mo||.74|
Of the patients who developed a local recurrence after SRS, 10 underwent resection with pathology confirming recurrent disease; of these 10, 3 patients received carmustine and 1 received adjuvant temozolomide. One patient was treated with repeat SRS and temozolomide. Four patients were treated with temozolomide and 1 patient underwent treatment with intra–arterial carboplatin and temozolomide. Ten patients were followed with no further therapy.
Two patients were confirmed to have radiation necrosis after SRS. One patient underwent 18-fluorodeoxyglucose positron emission tomography confirming necrosis 1.5 years after SRS. The other patient underwent resection 5.5 months after SRS with pathology confirming radiation necrosis.
Other toxicities included 1 patient who had a seizure during SRS. It is unclear whether her seizures were aggravated by SRS treatment because she had minor seizures before treatment. Another reported severe fatigue after SRS and slept >12 hours per day. No other major toxicities were noted.
SRS may be used in the upfront setting to improve local control and survival or in a salvage setting after failure from other therapy, including WBRT and surgery. In the salvage setting when WBRT was given upfront, the remaining options are often SRS or surgery. Surgery may be selected in cases in which there is only 1 recurrent lesion or the lesion is large or symptomatic, and in patients with good functional status and prognosis. Alternatively, SRS has the benefit of being a minimally invasive, outpatient procedure without the use of general anesthesia. Other advantages include minimal recovery time, less risk for bleeding or infection, and decreased costs.13 SRS can also treat surgically inaccessible lesions.
Several retrospective studies have been performed regarding the use of salvage SRS. One study reported in 1995 involved 182 patients (295 lesions) with 69% having recurrent BM.14 SRS was LINAC-based. The median OS from radiosurgery was 9.4 months. Increased survival was associated with the absence of systemic disease, age <60 years, fewer than 3 lesions, and female sex. The 2-year local control rate was 65%. The authors concluded that SRS had similar control rates as surgery with good survival. However, unlike our study, Alexander et al14 included patients treated with SRS as boost.
One study that investigated the use of SRS in a salvage setting was reported in 1990 and involved 18 patients (21 lesions) with recurrent or persistent BM treated with LINAC SRS.15 All but 1 received prior WBRT. The most common histology in this study was lung followed by breast cancer and melanoma. The median follow-up was 9 months. All tumors were reported to be controlled in the radiosurgery field, although 2 patients failed at the margin of the treated volume, requiring surgical resection and 125I brachytherapy. Adverse events were limited and transient. There were no cases of symptomatic radiation necrosis reported.
A French study reported in 2001 focused specifically on patients treated with WBRT upfront followed by SRS in a salvage setting.16 This involved 54 consecutive patients (97 lesions) treated with LINAC-based SRS for recurrent disease after WBRT. The median follow-up was 9 months. The median survival was 7.8 months. The 1-year and 2-year local control rates were 91.3% and 84%, respectively. The 1- and 2-year brain new event-free survival rates were 65% and 57%, respectively. On univariate analysis, KPS, RPA class, score index for radiosurgery (SIR), and interval between WBRT and radiosurgery were prognostic for overall survival and brain disease-free survival. On multivariate analysis, only RPA and interval between WBRT and radiosurgery predicted for overall survival and brain-disease free survival. For local control, no prognostic factors were found.
We reviewed our series of patients who received WBRT in an upfront setting and developed recurrences and were treated with salvage SRS. Because this group of patients had received radiation in the form of WBRT, we evaluated the benefit of SRS with respect to control rates and overall survival. We also wanted to study the prognostic factors associated with better survival in this group of patients by evaluating a large number of patients treated at a single institution. To our knowledge this is the largest series to date studying specifically salvage SRS after WBRT. In our group of 111 patients, the median OS was 17.7 months from the diagnosis of the first BM and the survival was 9.9 months from the time of salvage SRS. These results suggest that through the use of SRS, prolonged survival was achievable and are comparable to the previous studies discussed above. Our results demonstrate longer survival when compared with some studies using SRS upfront because these patients were self-selected to do well by virtue of having time to develop recurrent BM and not dying of their systemic disease.
Unlike the Noel et al16 article, RPA did not predict for improved survival in this group of patients in this study. Other factors such as primary disease, size of the largest BM, or number of lesions at initial diagnosis also were not found to predict for survival. Conversely, longer time to diagnosis from WBRT to SRS was prognostic for longer OS. This also translated into longer survival after SRS. This is consistent with the Noel et al study mentioned above. Table 5 summarizes the data from previous studies and the current one.
|Series||No. of Patients||Median Survival After SRS, Mo||Prognostic Factors for Survival||Inclusion|
|Alexander 199514||182||9.4||No active systemic disease, age <60 y, <3 lesions, female||Upfront and salvage (69% of patients)|
|Noel 200116||54||7.8||RPA and interval between WBRT and SRS||Salvage|
|Current study||111||9.9||Interval between WBRT and SRS||Salvage|
Also investigated was the local recurrence rate after initial salvage SRS. Primary disease and time from WBRT to the development of recurrent disease were not found to be predictive of local recurrence. Only 25% of the patients failed at the site of first salvage SRS. Dose of SRS and size of the metastasis did predict for LF of each metastasis, which is consistent with previous data from our institution.12 Although a lower dose of 20 Gy for metastases measuring ≤2 cm may be ideal when given as a boost with WBRT,8 in this setting as salvage therapy, SRS was given alone and a higher dose may be needed. In total, these results are consistent with previous studies.14-16 No factors predicted for distant recurrence.
This study adds to the existing data regarding the role of salvage SRS. This study is limited by the finding that this is retrospective, which may lead to bias, although this was minimized by having a single experienced clinician (S.T.C.) re-review all the records in the database. A prospective study would be ideal and provide more accurate data for prognostic factors for salvage SRS. Currently, we are analyzing our patients treated with upfront SRS for comparison.
SRS is effective in treating recurrences after WBRT. Time to recurrence is a factor that appears to result in improved survival after SRS. Survival was 9.9 months after SRS in all patients. Smaller metastases (defined as ≤2 cm) and an SRS dose of ≥22 Gy appear to result in better local control. Based on these results, we continue to offer this treatment approach for patients with recurrent BM after WBRT.