RCTs Comparing WBRT and SRS Boost versus WBRT Alone
The 2 RCTs comparing WBRT to WBRT and SRS boost5, 6 fundamentally ask the question of the role in optimizing local control by adding SRS as a boost when overall brain control is maximized by WBRT. Our meta-analysis was able to pool OS data for those treated with multiple brain metastases only, as the Kondziolka6 study restricted entry to patients with 2 to 4 brain metastases and the Andrews5 study reported data for those with single versus multiple brain (2 or 3) metastases separately. For multiple metastases, we conclude no OS advantage with the use of SRS as a boost to WBRT with an HR of 1.63 (95% CI 0.72-3.69; P = .24) (Fig. 1). However, the Andrews5 study reported a significant difference in OS for those patients with a single brain metastasis treated with WBRT and SRS boost compared with WBRT alone (median survival of 6.5 months vs 4.9 months, P = .0393). Importantly, this RCT was powered for OS and a priori for patients with single brain metastasis.5 Although OS for a treated single brain metastasis significantly favored WBRT and SRS boost (compared with WBRT alone) in the univariable analysis, the multivariable analysis yielded a P = .053 for treatment compared with P < .0001 for class 1 RPA status.5
Pooled results for local control from both RCTs5, 6 resulted in a statistically significant improvement in local control with an HR of 2.88 (95% CI 1.63-5.08; P = .0003) (Fig. 2) favoring WBRT and SRS boost. However, this result should be interpreted with caution as there are potential sources of bias. For example, the local failure rate with WBRT in the Kondziolka6 study was unexpectedly high at 100%, which has not been reproduced. Furthermore, this study was stopped early with only 14 patients randomized to WBRT alone and 13 to WBRT and SRS boost. Alternatively, the Andrews5 study completed with a much larger sample size with 164 patients randomized to WBRT and SRS boost and 167 patients randomized to WBRT alone. Although, they report local control favoring the addition of SRS boost with an 82% vs 71% local control rate at 1 year, missing data on local control was a confounding factor. The authors reported at the 3 month follow-up time point that only 60% of patients had imaging available for central review. Although, a local control benefit would not be unexpected given that additional dose to the tumor should indeed improve tumor control.
With respect to distant brain control, we could not pool the results from the 2 studies as data were not reported specifically for this endpoint. However, overall brain control (considering both local and distant brain failures) was described as the median time to any brain failure in the Kondziolka6 study. Failure occurred at 5 months in those treated with WBRT and 34 months for those treated with WBRT and SRS boost (P = .002).6 In contrast, Andrews et al5 report no significant differences between those treated with WBRT and SRS boost versus WBRT alone with respect to overall time to intracranial progression (P = .1278). Therefore, it is unclear from the 2 published trials whether there is a significant difference in the development of new brain metastases for patients treated with WBRT and SRS boost versus WBRT alone. However, intuitively as both arms are treated with WBRT, one would expect a similar risk of new distant brain metastases.
Therefore, we can conclude that SRS as a boost improves local control compared with WBRT alone, and for patients with single brain metastasis (as opposed to multiple) the local control benefit translates into a gain in OS. Hence, if local control is not optimized using SRS, then only is there potential for an OS advantage when the alternative is WBRT alone (and in patients with favorable prognostic factors, that is, RPA1 class 1 status).
SRS alone versus WBRT plus SRS boost
Fundamentally, those RCTs evaluating SRS alone to WBRT plus SRS boost7-9 were designed to answer the question of the need for WBRT when local control is optimized by using SRS in each treatment arm. The clinical aim of SRS alone was to avoid the potential side effects associated with WBRT, notably neurocognitive decline. We conclude based on our meta-analysis that the addition of WBRT to SRS significantly improves distant brain control with an HR of 2.15 (95% CI 1.55, 2.99; P < .00001) (Fig. 5), and local tumor control with an HR of 2.61 (95% CI 1.68-4.06; P < 0001) (Fig. 4). These results are biologically sound as subclinical microscopic disease in the brain remains untreated without WBRT, resulting in a greater risk of new brain metastases with time. This observation has been confirmed by RCTs evaluating prophylactic cranial irradiation (PCI) in lung cancer patients,13-15 as a lower incidence of brain metastases has been reported in those prophylactically radiated as opposed to those observed. The reported improvement in local tumor control is also expected given that the additional dose delivered by the WBRT serves to intensify the SRS dose given to the tumor.
However, despite these observed benefits of WBRT to SRS alone, we conclude no OS advantage for adjuvant WBRT based on the pooled data from the Chang8 and Aoyama7 study with a HR of 0.98 (95% CI 0.71-1.35, p = 0.88, Figure 3). This result confirms the Kocher9 study who reported no significant advantage for adjuvant WBRT in any of the subgroup arms, and that would include those randomized to SRS alone and WBRT plus SRS boost. As the OS curves were not reported for each subgroup independently, we could not pool their data for meta-analysis. The lack of OS advantage to WBRT can be explained by considering several factors. First, these studies were designed to optimize local control in both the experimental and control arms using SRS, as opposed to the Andrews5 and Kondziolka6 study where the control arm consisted of patients treated suboptimally for local control with WBRT alone. Second, the result likely reflects the observation that OS is determined according to those powerful prognostic factors such as primary cancer type,2 age,1 performance status,1 systemic disease status,1 and use of systemic chemotherapy,8 as opposed to overall brain control. Last, that OS survival is not compromised in patients treated with SRS alone as long as patients are followed closely and treated with salvage radiation (further SRS or WBRT) when appropriate at the time of intracranial relapse.7, 8, 9 Therefore we confirm, and provide updated robust HRs for the selected primary endpoints, those results from the initial Aoyama7 study such that WBRT plus SRS improves local control, distant brain control, and does not impact OS.
Although we could not pool the data for meta-analysis for the secondary endpoints proposed in this study, it is the interpretation of those secondary endpoints that provide insight as to why some have been such strong advocates of SRS alone. First, with respect to neurocognitive outcomes, the Chang study8 used a validated neurocognitive instrument to determine the impact of WBRT on cognition. This study was uniquely powered for this intent as the primary endpoint was based on the HVLT. Chang8 reported a significant improvement in learning and memory function at 4 months in those treated with SRS alone, and the benefit persisted at 6 months. With respect to longer term neurocognitive adverse sequelae of WBRT, a negative impact on learning and memory was recently reported even up to 1 year after WBRT after PCI again based on HVLT assessments.16 This PCI RCT16 on patients with lung cancer is highly relevant as neither arm had brain metastases at baseline, and worse learning and memory neurocognitive function was observed despite the lower incidence of metastases in the PCI arm compared with the controls. Therefore, one can conclude that it is the treatment effect of WBRT that affects neurocognition as opposed to the conjecture that brain metastases recurrence is the major source of neurocognitive decline.
Although the Aoyama study7 did evaluate neurocognition using the MMSE tool, this was not an a priori endpoint and optionally assessed. In 28 of the 44 patients who lived at least 1 year after treatment, MMSE scores were available at least once at the median follow-up times of 30.5 months in the WBRT plus SRS boost arm (16 patients) and 20.7 months in the SRS alone arm (12 patients).7 Aoyama reports no significant difference in MMSE scores between the 2 arms.7 However, the MMSE is a dementia screening tool and is considered at best a poor measure of neurocognition in brain metastases patients.12 Therefore, we could not pool the MMSE data with the formal neurocognitive HVLT results from the Chang8 study, and the latter remains the only study with outcomes based on a validated instrument to test neurocognition.
With respect to maintaining good performance status over time, as defined by a WHO score of at least 2, Kocher et al9 reported no significant difference among those treated with WBRT plus SRS boost compared with SRS alone. This study was specifically powered for this endpoint. A similar result was observed in the Aoyama et al7 RCT, where systemic functional preservation rates as measured by a KPS of ≥70 were not significantly different in those randomized to WBRT plus SRS boost and those to SRS alone (P = .53). Furthermore, in those studies reporting the risk of neurologic death according to the treatment arm, there has been no significant difference favoring 1 treatment approach over another. The relevant data from the RCTs are summarized in Table 1. Therefore, one can conclude that despite a greater risk of new brain metastases and risk of local tumor progression when treating with SRS alone, there is no adverse impact for patients in preserving their functional independence or in their cause of death. We postulate that this observation is associated with the timely use of salvage brain treatments before symptomatic manifestation of the recurrent tumors, which is possible as patients are followed with frequent imaging and clinical follow-up on study. However, we cannot validate this hypothesis within the reported RCT data.
With respect to QOL, the lack of reported data in the RCTs makes any conclusion of potential benefits of 1 treatment approach over another impossible. Although the Chang8 study did report better QOL scores for the SRS alone cohort, the wide CI in the data made the result inconclusive. With respect to toxicity, serious adverse events based on the individual RCTs may favor use of SRS alone. In particular, an increased risk of radiation necrosis with WBRT plus SRS boost has been reported compared with SRS alone by Kocher et al.9 Furthermore, leukoencephalopathy as a late effect was observed more in patients treated with WBRT as reported by Aoyama et al.7 Last, we learned from the Chang8 study that patients treated with SRS alone were treated sooner with systemic therapy and with more cycles of systemic therapy, and this may be a potential benefit for patients treated with SRS alone to consider. This imbalance in the treatment arms may have been a factor in the reported longer OS observed in the SRS alone arm compared with the WBRT and SRS boost arm in that study; however, based on meta-analysis and the Kocher study9 we observe no OS difference according to treatment arm.
For selected patients with up to 4 brain metastases eligible for SRS, our meta-analysis concludes no OS benefit for WBRT plus SRS boost compared with SRS alone despite significant gains in both local and distant brain tumor control with WBRT. SRS alone may allow patients to optimally retain their neurocognitive function, experience fewer serious late side effects, and are not at adverse risk with respect to maintaining performance status. Therefore, we conclude that SRS alone with frequent magnetic resonance imaging (MRI)-based follow-ups in order to salvage recurrent brain metastases before symptomatic manifestations, should be routinely offered to selected patients as a treatment option to consider.