Stereotactic radiosurgery versus whole-brain radiotherapy after intracranial metastasis resection: a systematic review and meta-analysis

Inclusion criteria consisted of any study that described adult patients (≥18 years of age) with a diagnosis of systemic cancer who underwent neurosurgery followed by whole-brain radiotherapy (WBRT) or by stereotactic radiosurgery (SRS) for treatment of up to three metastases to the brain. SRS was defined as a single or few fractions of high dose radiation to a small intracranial target that was encompassed by at least 50% of the prescribed dose [17]. Because Intraoperative Radiation Therapy (IORT) meets most of the salient features of SRS as described by the Radiation Oncology Therapy Group protocol, [18, 19] studies reporting IORT were included within the SRS treatment groups. Studies reporting on the use of Local Brain Radiotherapy (LBRT), however, were excluded for not meeting the criteria of SRS. Studies reporting outcomes on patients receiving prior WBRT and studies that did not compare post-operative WBRT to post-operative SRS were also excluded. After title and abstract screening, remaining articles were read full-text. Three authors (NL, IM, and LM) performed full-text screening and two (NL and LM) extracted the data. One senior author (MB) reviewed the included articles and the extracted data. Disagreements were solved by discussion.

The following data were extracted from each study whenever possible: author and year of published articles, number of patients in the study, patient characteristics (inclusion criteria, primary tumor type, age, type of intervention (whole-brain radiotherapy, stereotactic radiosurgery, radiation dose and regimen, time between resection and radiation), number of intracranial lesions at presentation, extent of extracranial disease), duration of follow-up, and primary and secondary outcomes of the study. When available, relative risks (RR) comparing SRS to WBRT were extracted for local recurrence risk, distant recurrence, leptomeningeal disease, and overall survival. Duration of overall survival (OS) and progression free survival (PFS) for both local and distant recurrence was also noted. Furthermore, data on the number of patients that died from neurological causes and on observed neurotoxicity were extracted. In one study reporting outcomes of multiple metastases per patient, the reported percent recurrence by number of cavities was considered as equivalent to the percent recurrence by number of patients, as the number of metastases per patient approximated 1:1 [20].

Study quality was assessed by the New-Castle Ottawa Scale, and the quality of our recommendation was assessed using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) criteria. Both assessments were made by two authors (NL and IM) independently [21, 22]. Discrepancies were solved by discussion.

To minimize heterogeneity amongst studies being compared, only cohort studies were included in the meta-analysis. Data analysis was performed using Comprehensive Meta-Analysis (CMA) Version 3 (Biostat, Inc., Englewood, NJ, USA). Overall fixed and random risk ratios were calculated for the same outcomes, depending on the available data in the articles. I-squared values were calculated to assess heterogeneity. For median overall survival, relative risks were calculated based on group size and median overall survival both for SRS and WBRT [23]. Meta-regression analysis on continent, age, and journal impact-factor was used when possible to assess sources of heterogeneity.

Data regarding local recurrence was available in five out of the eight studies included in our review [5, 18, 20, 26, 27]. The local recurrence incidence ranged from 0 to 60% in patients treated with adjuvant SRS and from 11 to 24% in patients treated with adjuvant WBRT. While no study reported a statistically significant difference in local recurrence rate between the two treatment groups, four of the five studies that included data on local recurrence did note differences in these rates. One study reported a near-significant hazard ratio (p = 0.09) that disfavored SRS over WBRT by a 68% higher rate of distant recurrence [18]. One Italian study from 1996 that did not specifically report a p-value comparing just these two treatment arms did find that 5 patients treated with adjuvant SRS were nearly four times as likely to experience local recurrence as compared to 14 patients treated with adjuvant WBRT [27]. Two other studies favored SRS for achieving local control, with percent differences in local recurrence ranging from 3 to 17% in favor of SRS [20, 26]. Only the study by Lee et al. demonstrated near-equal rates of local recurrence for both treatment groups, with a local recurrence of 11 and 12% for WBRT and SRS treatment groups, respectively [5].

The pooled risk ratio comparing local recurrence incidence between SRS vs. WBRT groups was 0.79 (95% CI: 0.48–1.29; fixed effects model) and demonstrated no significant difference in rates of local recurrence (Fig. 2). Results from the random effects model were similar (RR: 0.88; 95%-CI = 0.40–1.92) and moderate heterogeneity was observed (I² = 49.44%, p-heterogeneity = 0.10, 3 studies). This analysis was based on 163 patients receiving post-operative WBRT and 138 patients receiving postoperative SRS.

Univariate meta-regression on age (p = 0.02) and continent (p = 0.04) were identified as a source of heterogeneity. With increasing age, the RR for local recurrence was in favor of SRS as compared to WBRT. Furthermore, Europe, in comparison to North America was associated with a RR for local recurrence in favor of WBRT. Journal impact factor was not identified as a source of heterogeneity (p = 0.97).

Data regarding distant progression was available in three out of the eight studies included in our study (Fig. 3) [5, 20, 26]. The distant progression incidence ranged from 6 to 50% in patients treated with adjuvant SRS and from 17 to 44% in patients treated with adjuvant WBRT [5, 20, 26]. Three of the studies, while not reaching statistical significance, also reported distant progression rates that were 6 to 11% higher in SRS patients as compared to WBRT patients [20, 26]. One study reported better distant control in 17 patients treated with SRS as compared to 109 treated with WBRT [5]. One additional study reported a statistically significant distant progression hazard ratio of 2.17 for patients treated with SRS (single dose range: 14 to 24 Gy) after resection compared to those treated with WBRT (dose: 30 Gy in 10 fractions or 37.5 Gy in 15 fractions) [18]. The pooled risk ratio comparing distant recurrence rates between SRS versus WBRT was 1.09 (95%-CI: 0.74–1.60; fixed effects model) and demonstrated no significant difference in rates of distant recurrence (Fig. 3). Results from the random effects model were similar (RR: 0.94; 95%-CI: 0.39–2.26), and moderate heterogeneity was observed (I² = 50.5%, p-heterogeneity = 0.28, 3 studies, Fig. 3). Continent (p = 0.52), age (p = 0.12), and journal impact factor (p = 0.52) were not identified as sources of heterogeneity. This analysis was based on 163 patients receiving postoperative WBRT and 138 patients receiving postoperative SRS.

Two studies reported on the development of leptomeningeal disease (LMD) following radiotherapy [18, 20]. Retrospectively, SRS treatment relative to WBRT was associated with a higher risk of LMD occurrence as per Hsieh et al. [18] (HR:2.44; 95%-CI:1.15–5.18) and Patel et al. [24] (HR:5.67; 95%-CI: 1.50–21.51).

The pooled relative risk for the development of LMD following postoperative SRS to postoperative WBRT was 2.99 (95% CI 1.55–5.76), indicating that resection followed by SRS increases the risk for developing LMD as compared to resection followed by WBRT (Fig. 4). There was low heterogeneity (I²: 14.4% p-heterogeneity: 0.28, 2 studies) between the included studies. This remained significant in the random effects model (RR: 3.09, 95% CI: 1.47–6.48). The analysis is based on 192 patients receiving resection followed by WBRT and 152 patients receiving resection followed by SRS.

Only one study out of the 8 included in this review found a significant difference in survival between patients treated with SRS following resection as opposed to WBRT [24]. In their 2011 institutional study, Elaimy et al. retrospectively compared six different treatment regimens in patients with metastatic lesions in the brain [24]. Using SRS only as a reference, they found that 15 selected patients treated with SRS following resection had a 68% lower risk of death as compared to 65 patients treated with SRS alone (p = 0.02). In contrast, they reported a slightly higher risk of death (HR = 1.04) in 11 patients treated with WBRT after resection as compared to SRS alone [24]. While the other eight studies did not report significant differences in survival between the two treatment groups, two studies were slightly in favor of SRS, [18, 26] while three [5, 27, 28] indicated slightly better outcomes after WBRT, with one even reaching a near-significant level in favor of WBRT (p = 0.06) [5].

The pooled relative risk for median overall survival comparing patients treated with SRS after resection to those treated with WBRT was 0.51 (95%-CI: 0.44–0.54; I²: 86.6%, p-heterogeneity < 0.01; 4 studies), suggesting a favorable outcome for WBRT with considerable heterogeneity (Fig. 5) [5, 24, 26, 27]. However, the overall survival was non-significant using the random-effect model (RR: 0.78; 95%-CI: 0.40–1.00) [5, 24, 26, 27]. The analysis was based on 174 patients receiving post-operative WBRT and 153 patients receiving post-operative SRS. Univariate meta-regression on continent (p < 0.01) and journal impact factor (p < 0.01) were identified as sources of heterogeneity. Europe and Asia, in comparison to North America, were associated with a RR for overall survival in favor of SRS. A lower impact factor journal showed a RR in favor of WBRT as compared to SRS. Age was not identified as a source of heterogeneity (p = 0.14).

Radiation necrosis and leukoencephalopathy were each reported in one study. Patel et al. found that symptomatic radiation necrosis, requiring the use of steroid or other interventions, was 27% in the SRS group and 0% in the WBRT group (p < 0.01). After 1 year, the rate of radiographic leukoencephalopathy was 7% following SRS vs. 47% following WBRT (p < 0.01). The relative risk of leukoencephalopathy for SRS compared to WBRT in this study was 0.15 (95% CI: 0.11–0.21).

A major limitation of this review and meta-analysis was the heterogeneity among the included studies. Amongst the nine studies included, one included only colon cancer patients, one included melanoma patients only, and the remaining six consisted of patient populations with varying proportions of melanoma, colon, renal, breast, and unknown histologies [5, 18, 20, 24, 26, 28]. Additionally, the number of brain metastases was not reported homogenously between studies, and reported outcomes were not specific to the number of brain metastases and treatment modalities. Therefore, it was not possible to perform a subgroup analysis by histology type or number of metastases, or address potential sources of heterogeneity by these covariates. This is most likely of importance, because certain tumors, such as melanoma, colon, and renal carcinomas, are considered radioresistant for WBRT [10, 39, 40] and it is possible that outcomes could differ in patients with only a single metastasis [41, 42]. Of note, one study included in the analysis reported patients with 2–4 metastases in a single group, but noted that the number of metastases did not have an effect on the outcomes [24]. A further limitation of this analysis was that some studies reported death as a hazard ratio, while other used the Kaplan Meier method. Reporting a hazard ratio is a more accurate method and the variation in reporting limits the strength of the conclusions. Lastly, the studies included in the analysis were all retrospective in nature. Therefore, the different treatment-strategies implemented for each set of patients were the result of the tumor boards’ selection of patients whom they felt would benefit from that regimen, resulting in a selection bias. This is further complicated by the different forms of radiation administered in each study, as some studies described outcomes of IORT instead of SRS [18]. No studies were identified that compared WBRT with hypofractionated SRS, which has become a viable treatment strategy for lesions not treatable with a single fraction SRS [43, 44].

Despite these limitations, this meta-analysis based on cohort studies evaluated several different outcomes after extensive review of the literature. To our knowledge, this is the first meta-analysis on this topic. The use of both random- and fixed-effect models together with heterogeneity analysis resulted in a critical evaluation of our outcomes. Furthermore, an increased risk for the development of LMD after SRS compared to WBRT following surgical resection was identified.

There is a clear necessity for the improvement of care for brain metastasis patients as they tend to live longer. Currently, several treatment strategies exist for brain metastasis, but data on the most effective strategy is lacking. Factors such as radiation dose, treatment timing, chemotherapy combinations among others should be evaluated in future studies in order to evaluate the most effective treatment modalities. The wide range of treatment modalities also provides a potential to strive towards more individualized patient care. In addition, evaluated outcomes should not be limited to survival or progression and should include quality of life. Indeed, currently a phase III randomized trial is being conducted to evaluate postoperative SRS versus WBRT in brain metastasis patients (NCT01372774) [45]. Primary endpoints will be evaluation of survival and neurocognitive progression and secondary endpoints will be QOL, adverse events, and functional independence among others. Last, the use of “big data” may provide a method to evaluate these outcomes more thoroughly.