Updates in the management of brain metastases

Abstract

The clinical management/understanding of brain metastases (BM) has changed substantially in the last 5 years, with key advances and clinical trials highlighted in this review. Several of these changes stem from improvements in systemic therapy, which have led to better systemic control and longer overall patient survival, associated with increased time at risk for developing BM. Development of systemic therapies capable of preventing BM and controlling both intracranial and extracranial disease once BM are diagnosed is paramount. The increase in use of stereotactic radiosurgery alone for many patients with multiple BM is an outgrowth of the desire to employ treatments focused on local control while minimizing cognitive effects associated with whole brain radiotherapy. Complications from BM and their treatment must be considered in comprehensive patient management, especially with greater awareness that the majority of patients do not die from their BM. Being aware of significant heterogeneity in prognosis and therapeutic options for patients with BM is crucial for appropriate management, with greater attention to developing individual patient treatment plans based on predicted outcomes; in this context, recent prognostic models of survival have been extensively revised to incorporate molecular markers unique to different primary cancers.

Brain metastases (BM) affect up to one-third of adults with cancer and are a significant cause of patient morbidity, anxiety, and mortality. With ∼200 000 patients affected each year in the United States,^1,2 a number that is somewhat speculative since the incidence of metastatic disease is not a reportable event, and a presumption of increasing incidence, BM represent an important public health burden that is 10 times more common than malignant primary brain tumors. Despite this, the BM population has received less attention, with historical exclusion from many clinical drug trials, frequent misperceptions about prognosis by physicians and patients, and the variable involvement of different care providers who may not follow the patient longitudinally. Management of BM has changed substantially in the past 5–10 years, with more varied treatments which have a greater focus on survivorship and mitigating the effects of treatment. This review provides a summary of key updates and considerations in the current management of BM for practicing clinicians.

Incidence

The incidence of BM is believed to be rising in the setting of improved systemic therapies that control systemic disease and prolong survival but cross the blood–brain barrier (BBB) too poorly to prevent the brain from becoming a sanctuary site for metastatic disease. Accurate estimates of population-level incidence of BM are difficult to ascertain, as national registries such as Surveillance, Epidemiology, and End Results capture information related to initial diagnosis and management of cancer but not events that typically occur later in the disease course, such as the development of BM. Prior reports suggest that ∼10%–35% of adult cancer patients develop BM.^1–4 In one metropolitan cancer registry from 1973 to 2001, the incidence proportion (IP) of BM across a variety of malignancies was 10%, defined as the percentage of patients who were diagnosed with BM at any point before death, with large differences according to primary tumor as well as age, gender, race, cancer stage, and age at cancer diagnosis.⁴ In that study, the IP of BM for the 4 most common cancers that metastasize to the brain were lung (20%), melanoma (7%), renal (7%), and breast (5%), when combining all initial stages together for a given primary tumor. When examined according to stage, patients with lung cancer had the highest IP for developing BM among all patients with a stage I cancer, at nearly 10%, whereas among patients diagnosed with a stage IV cancer, melanoma had the highest IP for BM at ∼37%. The notion that modern systemic therapies may alter the natural disease history and risk for BM in different cancers is perhaps most strikingly seen with human epidermal growth factor receptor 2 (HER2)–positive breast cancer,^5,6 melanoma,⁷ and anaplastic lymphoma kinase gene (ALK) rearranged non-small-cell lung cancer (NSCLC).^8–11 In the trastuzumab era, the incidence of BM among patients with advanced HER2-positive breast cancer is 40%–50%, although survival following BM has improved substantially.^5,12 In melanoma, where the brain is a frequent site of metastases, new immune checkpoint inhibitors and molecularly targeted agents provide survival benefit in systemic disease and may offer CNS control.⁷ In patients with ALK-rearranged NSCLC, the CNS is the first site of progression in 46% of patients treated with crizotinib.¹¹ Despite these remarkable subgroup findings, the only populations currently recommended by the National Comprehensive Cancer Network to undergo a screening brain MRI at diagnosis in the absence of neurologic symptoms are those with advanced melanoma, most patients with NSCLC, and all patients with small-cell lung cancer.¹³

Prognosis

While BM historically have been associated with a dismal prognosis and median survival of <6 months, a combination of earlier diagnosis in some cases, more effective treatments, and better understanding of which patients may survive for many months or years has led to improved prognostication tools. Estimating prognosis among patients with BM is clinically relevant, as it allows clinicians to recommend treatments that balance durability of intracranial tumor control with quality of life (QoL) and side effects of treatment. A landmark recursive partitioning analysis (RPA) of prognostic factors among 1200 pooled BM patients treated in 3 Radiation Therapy Oncology Group (RTOG) trials found that KPS, age, and extracranial disease status were key determinants of survival, with the longest-surviving group having a median survival of 7.1 months.¹⁴ While this RPA was validated and widely used, at least 5 additional prognostication models were published in the subsequent 10 years, in an effort to further refine survival estimates in an era of improving systemic therapies.¹⁵ The prognostic index that has now become the most prominent is the Graded Prognostic Assessment (GPA), which was originally developed from a database of nearly 2000 patients with newly diagnosed BM accrued to 4 RTOG protocols, and later serially refined to include patients treated across multiple institutions and to incorporate prognostic factors unique to different cancers, resulting in publication of the diagnosis-specific (DS) GPA in 2012.¹⁶ In the DS-GPA, the significant prognostic factors vary substantially according to the type of cancer and include factors such as age, KPS, extracranial disease status, number of BM, and in the case of breast cancer, the molecular subtype (Fig. 1). Importantly, the stratification of patients by DS-GPA score within each cancer reveals marked heterogeneity in outcome, showing the important differences in prognostication for different subgroups. For example, patients with small-cell lung cancer with the lowest DS-GPA score were found to have median survival of <3 months, whereas breast cancer patients with the highest DS-GPA score had median survival of >2 years. The DS-GPA is increasingly being used as a stratification tool for clinical trials in BM and continues to be refined with incorporation of tumor molecular markers for some cancers.¹⁷ The importance of molecular markers in estimating life expectancy was highlighted in a recent multi-institutional publication regarding prognostic factors among 90 patients with BM from ALK-rearranged NSCLC, which found that median overall survival (OS) after development of BM was over 4 years.¹⁸

Fig. 1.

Diagnosis-specific graded prognostic assessment worksheet to estimate survival from newly diagnosed brain metastases.* Abbreviations: GPA, Graded Prognostic Assessment; KPS, Karnofsky performance status; ECM, extracranial metastases; LumA, luminal A; LumB, luminal B; GI, gastrointestinal. Breast cancer subtype definitions: Basal: triple negative. LumA: estrogen receptor (ER)/progesterone receptor (PR) positive, (HER2) negative. LumB: triple positive. HER2: ER/PR negative, HER2 positive. *Figure reprinted with permission. ©2012 American Society of Clinical Oncology. All rights reserved. Sperduto PW et al: Summary report on the graded prognostic assessment: an accurate and facile diagnosis-specific tool to estimate survival for patients with brain metastases. Journal of Clinical Oncology, 30(4):419–425.

In addition to prognostication tools, recent literature describing outcomes among BM patients has highlighted the importance of considering systemic disease status when counseling patients and selecting treatments. A large Japanese prospective observational study by Yamamoto et al among 1194 patients with 1–10 BM under 3 cm in diameter and KPS ≥ 70 who were treated with stereotactic radiosurgery (SRS) alone found that among the 850 patients who had died as of last follow-up, 92% died from systemic disease progression, and not from their BM.¹⁹ This finding has been echoed in other recent studies^20,21 and runs counter to a common perception among physicians²² that the development of BM necessarily indicates a grim prognosis, with rapid fatality consequential to the BM for all patients. In reality, among patients with cancer, the status of their systemic disease and its responsiveness to systemic therapy are much more likely to drive mortality outcomes than the presence of BM in the current era and together are a key competing risk.²³ For the majority of patients with BM, the lack of effective systemic therapies implies that most succumb within a few months of BM diagnosis, but an increasing subset of patients are experiencing longer survival, especially in the context of improving systemic disease control. In the DS-GPA model cited above, however, only the presence or absence of systemic disease was included in the multivariate analysis used to derive the model, without information regarding whether systemic disease was under control versus progressing. One retrospective study reported that if this information was added to the model for breast cancer patients with BM, dramatic changes in median survival were found compared with those predicted by the DS-GPA model alone,²⁴ and another group has developed a nomogram incorporating systemic disease status which demonstrates significant heterogeneity within each RPA class or DS-GPA score group, allowing for more individualized survival estimation.²⁵ Therefore, prognostication tools continue to be refined and can increasingly help clinicians and patients best select treatments tailored to their situation.

Role of Whole Brain Radiotherapy

Whole brain radiotherapy (WBRT), including upfront surgery for a select minority of patients with bulky tumors, used to be the standard treatment for most patients with BM, given that WBRT could be initiated quickly, was widely available, treated both visible and occult BM, and provided symptom palliation.²⁶ During the past 2 decades, SRS has emerged as a common approach for patients with a limited number of BM, due to the conformal treatment targeting the metastasis alone, single-session delivery, minimal delay to continuing with systemic therapy, and importantly, less cognitive dysfunction. As recognition has grown that WBRT may cause prolonged fatigue and/or neurocognitive deficits among some patients with BM, there has been a desire to test whether WBRT can be safely omitted from the management of patients receiving SRS, and this has been the subject of 3 randomized trials from 2006 to 2011, highlighted below. A more recent, unpublished trial focusing on neurocognitive outcomes after SRS with or without WBRT, by the Alliance for Clinical Trials in Oncology (N0574), is described separately in the “Neurocognition and Quality of Life after WBRT” section below.²⁷ Collectively, these trials demonstrate consistent results for patients with a limited number (≤4) of BM: (i) there is no survival decrement by withholding WBRT; (ii) high rates of local disease control are achieved with SRS alone, but local control rates increase further with the addition of WBRT; (iii) higher rates of new, distant BM are observed among patients receiving SRS alone, necessitating more frequent salvage treatment, though at least one-quarter of patients who receive upfront WBRT still develop new BM after WBRT; and (iv) the risk of cognitive dysfunction is lowered, but not eliminated, when WBRT is withheld.

The Japanese Radiation Oncology Study Group (JROSG) randomized 132 patients with 1–4 BM to SRS alone versus WBRT (30 Gy in 10 fractions) plus SRS, and found no differences in median OS between arms (8.0 vs 7.5 mo, respectively) but found that WBRT significantly decreased both local recurrence (LR) from 27% to 11% and distant brain recurrence (DBR) from 60% to 35%.²⁸ Similarly, the European Organisation for Research and Treatment of Cancer (EORTC) enrolled 359 patients with 1–3 BM treated with either SRS or surgery and randomized them to observation versus WBRT (30 Gy in 10 fractions), with a primary endpoint of the maintenance of functional independence at 6 months.²⁹ Similar to the JROSG trial, the EORTC study found that median OS was not different between arms (10.9 vs 10.7 mo) and that WBRT reduced LR at 2 years from 31% to 19% and DBR from 48% to 33%, with no difference in functional independence. Of note, WBRT added no OS advantage after either SRS or surgery, echoing the findings of the classic “Patchell” trial of surgery with or without WBRT.³⁰ Finally, a smaller trial at the MD Anderson Cancer Center randomized 58 patients with 1–3 BM to SRS with or without WBRT (30 Gy in 12 fractions), with the primary endpoint of neurocognitive deterioration at 4 months, defined as a 5-point drop on Hopkins Verbal Learning Test–Revised (HVLT-R) total recall.³¹ Early stopping rules were met after results suggested higher neurocognitive decline in the SRS + WBRT arm (52% vs 24% for SRS alone). While the addition of WBRT was associated with lower LR (33% vs 0%) and DBR (73% vs 27%), similar to the JROSG and EORTC trials, surprisingly the SRS + WBRT arm was associated with shorter survival, with median and 1-year OS being 15.2 versus 5.7 months, and 63% versus 21% for SRS alone versus SRS + WBRT, respectively. This survival result has been scrutinized by some and attributed to imbalances between the treatment arms, particularly with regard to discordant systemic disease burden and rates of salvage therapy, despite randomization; for example, there were 16 deaths from systemic disease (64%) in the SRS + WBRT group compared with 10 deaths (50%) in the SRS alone group.

Taken together, these 3 trials established that the addition of WBRT to SRS for up to 4 BM categorically reduces LR in the original BM sites as well as providing prophylaxis against DBR, and all 3 trials found no OS advantage to WBRT. Based on these data, the American Society for Radiation Oncology (ASTRO) issued a recommendation in September 2014 discouraging the addition of WBRT to SRS for patients with limited BM.³² In 2015, investigators from these 3 trials published additional analyses using the original trial data, focusing on particular patient subgroups. First, in a meta-analysis using individual patient data from 364 patients with 1–4 BM who received SRS with or without WBRT from all 3 trials, Sahgal et al³³ found no significant OS difference between patients receiving SRS versus SRS + WBRT in the overall cohort, but the authors reported that patients aged ≤50 years at the time of treatment had longer OS when treated with SRS alone. Hazard ratios (HRs) for survival with SRS alone ranged from 0.46 to 0.64 for age subgroups <50 years, and became statistically insignificant among older patients. This provocative finding is hypothesis generating and has raised questions regarding the modeling used given the small numbers of young patients in these trials (<35 patients per arm), the lack of restaging to establish the extent of systemic disease prior to protocol entry, and how the competing risk of systemic disease was handled.³⁴ In contrast to these data, a secondary analysis by Aoyama et al³⁵ of the original JROSG trial poststratified according to DS-GPA found that among the subgroup of 88 patients with NSCLC and 1–4 BM, those with higher DS-GPA scores (2.5–4.0) had significantly longer median OS when treated with SRS + WBRT compared with SRS alone (16.7 vs 10.7 mo, P = .03), whereas there was no OS difference among NSCLC patients with lower DS-GPA scores. This observation implies that improved brain control with WBRT, which has been demonstrated in every SRS ± WBRT trial, may translate into an OS advantage specifically among high DS-GPA patients because they do not die as rapidly from extracranial progression. However, because of the small patient cohort, lack of stratification according to DS-GPA in the original trial, and treatment era of 1999–2003, prior to several systemic therapy advances that likely impact extracranial and intracranial disease control, this retrospective study remains provocative but only hypothesis generating.

In summary, there may be patient subgroups that derive differential benefits from SRS versus SRS + WBRT, but for now the clinical trial data, while underpowered, largely suggest no OS difference between these 2 approaches, and prospective trials are required to test hypotheses from the secondary studies cited above. Clinical controversy remains regarding optimal patient selection for receipt of SRS alone, SRS + WBRT, and WBRT alone.^36,37 In addition, it should be emphasized that the majority of patients treated in the landmark SRS, WBRT, and surgery trials for BM cited above were unselected NSCLC patients, and therefore caution should be applied when generalizing outcomes about local therapies from these trials to various subgroups, such as HER2+ breast cancer or ALK-rearranged NSCLC patients. A summary of several key recent studies examining survival and functional outcomes after SRS, with or without WBRT, is provided in Table 1.

Table 1.

Select studies of SRS ± WBRT for brain metastases

Studya	Year	Design	N	Primary Endpoint	OS (mo)	Functional Outcomes	Distant Brain Control
Alliance27	2015	Phase III RCT: SRS ± WBRT 1–3 BM	213	Any neurocognitive decline >1 SD at 3 mo.	SRS + WBRT: 7.5 SRS alone: 10.7 (P = NS)	Neurocognitive decline at 3 mo SRS + WBRT: 88% SRS alone: 62% (P = .002)	SRS + WBRT: 85% SRS alone: 51% (P < .001) [1 y]
EORTC29,67	2011	Phase III RCT: SRS or surgery ± WBRT 1–3 BM and stable systemic disease	359	PS ≤ 2 at 6 mo	SRS + WBRT: 10.9 SRS alone: 10.7 (P = NS)	Time to PS > 2 mo SRS + WBRT: 9.5 SRS alone: 10.0 (P = NS) Quality of life Inferior in WBRT arm at isolated timepoints, but transient	SRS + WBRT: 67% SRS alone: 52% (P = .023) [2 y]
MDACC31	2009	Phase III RCT: SRS ± WBRT 1–3 BM	58	5-point drop on HVLT-R total recall at 4 mo	SRS + WBRT: 5.7 SRS alone: 15.2 (P = .003)	Neurocognitive decline at 4 mo SRS + WBRT: 52% SRS alone: 24% (96% Bayesian posterior mean probability)	SRS + WBRT: 73% SRS alone: 45% (P = .02) [1 y]
JROSG28	2006	Phase III RCT: SRS ± WBRT 1–4 BM	132	OS	SRS + WBRT: 7.5 SRS alone: 8.0 (P = ns)	Neurologic decline at 12 mo SRS + WBRT: 28% SRS alone: 30% (P = .99)	SRS + WBRT: 53% SRS alone: 24% (P < .001) [1 y]
JLGK19	2014	Prospective observational: SRS alone 1–10 BM Stratified by 1 vs 2–4 vs 5–10 BM	1194	OS	1 BM: 13.9 2–4 BM: 10.8 5–10 BM: 10.8 (P = .78 for 2–4 vs 5 BM)	Neurologic decline at 12 mo 1 BM: 8% 2–4 BM: 9% 5–10 BM: 12% (P = .60)	1 BM: 63% 2–4 BM: 45% 5–10 BM: 36% (P < .0001 for 1 vs 2–4 BM; P = .067 for 2–4 vs 5–10 BM) [1 y]
EORTC/JROSG/MDACC33	2015	Pooled IPD meta-analysis of 3 RCTs: SRS ± WBRT for 1–4 BM analyzed according to patient age	364	OS	Overall SRS + WBRT: 8.2 SRS alone: 10.0 (P = NS) HR for SRS alone vs SRS + WBRT Age 35: 0.46 Age 40: 0.52 Age 45: 0.58 Age 50: 0.64 (all P < .05) Age >50 (P = ns)	NR	HR for SRS alone vs SRS + WBRT Age ≤ 50 (P = NS) Age 55: 1.67 Age 60: 1.95 Age 65: 2.27 Age 70: 2.65 Age 75: 3.09 Age 80: 3.60 (all P < .05)
JROSG35	2015	Secondary analysis of NSCLC pts in RCT: SRS ± WBRT for 1–4 BM Stratified by DS-GPA10 < vs ≥2.5	88	OS	DS-GPA 2.5–4.0 SRS + WBRT:16.7 SRS alone: 10.6 (P = .04) DS-GPA 0.5–2.0 SRS + WBRT: 4.8 SRS alone: 6.5 (P = NS)	MMSE score at last follow-up DS-GPA 2.5–4.0 SRS + WBRT: 26.5 SRS alone: 28.0 (P = .NS) DS-GPA 0.5–2.0 SRS + WBRT: 28.0 SRS alone: 27.5 (P = NS)	DS-GPA 2.5–4.0 SRS + WBRT: 91% SRS alone: 71% (P = NR) DS-GPA 0.5–2.0 SRS + WBRT: 88% SRS alone: 87% (P = NR) [timing NR]

Abbreviations: Alliance, Alliance for Clinical Trials in Oncology; MDACC, MD Anderson Cancer Center; JLGK, Japanese Leksell GammaKnife Society; RCT, randomized controlled trial; IPD, individual patient data; pts, patients; PS, performance status; ns, not significant; NR, not reported; MMSE, Mini-Mental State Examination.

^aArranged according to study design, with most recent Phase III trials listed first.

Stereotactic Radiosurgery Advances

Number or Volume of Metastases

While the landmark randomized trials from the last decade cited above enrolled patients with up to 4 BM, there has naturally been interest in the feasibility and outcomes of patients with >4 BM treated with SRS alone. There is nothing inherently limiting SRS only to patients with ≤4 BM, but the original randomized trials were conducted among patients with only limited intracranial disease, partly due to the belief that patients with many more lesions would be better served by WBRT or that SRS to so many lesions may be associated with excess toxicity or technical limitations. ASTRO published an evidence-based review on the role of SRS for BM in 2005 which recommended consideration of SRS for up to 4 BM that were ≤4 cm in size,³⁸ but this was prior to the publication of the seminal trials reviewed above, and the 2012 updated ASTRO guidelines³⁹ and 2011 American Association of Neurological Surgeons/Congress of Neurological Surgeons guidelines⁴⁰ do not make explicit recommendations for patients with more than 4 BM. Many centers have offered SRS to patients with a larger number of BM over the past 10 years based on retrospective studies suggesting safety and efficacy. A comprehensive prospective observational study in Japan among nearly 1200 patients from 23 institutions with up to 10 newly diagnosed BM treated with SRS alone was published in 2014¹⁹ and provided a helpful benchmark for understanding outcomes among this patient population. The largest single BM allowed per the inclusion criteria was <10 mL in volume and <3 cm in maximal diameter, and the total cumulative volume of all BM had to be ≤15 mL. The study included 208 patients with 5–10 BM, and this subgroup had a mean cumulative tumor volume of 3.54 mL (range, 0.02 to 13.90), 63% of whom had controlled systemic disease and 88% of whom had KPS ≥ 80. Baseline characteristics of the 5–10 BM group were not significantly different from the group with 2–4 BM, aside from slightly higher cumulative tumor volume (mean, 3.54 vs 3.07 mL), as expected. The principal finding of the paper was that median OS was similar between the 5–10 BM and the 2–4 BM groups (10.8 mo in both groups), which was statistically noninferior for the 5–10 BM group, based upon a noninferiority margin set at the upper 95% CI for an HR of 1.30. Secondary outcomes analysis demonstrated that patients with 2–4 BM as well as those with 5–10 BM had similar 2-year rates of neurologic death (5% vs 7%), new brain lesions (66% vs 72%), and salvage WBRT (10% vs 9%); the only parameter that was significantly higher among the 5–10 BM group was the risk of leptomeningeal dissemination (13% vs 22%). Toxicity was low and essentially identical between the 2–4 BM and 5–10 BM groups, with no significant differences in neurocognitive outcome at 4, 12, 24, or 36 months as assessed by the Mini-Mental State Examination. While not randomized, this large study provides some evidence of safety and efficacy of SRS alone for patients with up to 10 BM with a relatively low intracranial disease volume; of note, details regarding the frequency and costs of imaging and treatment were not reported. Additionally, investigators at the University of Pittsburgh have described their outcomes of SRS alone for ≥10 BM among a group of 61 patients who had a mean of 13 BM per patient, 62% of whom had received prior WBRT.⁴¹ They found that survival was superior among patients with 10–13 BM compared with ≥14 BM, but importantly, median survival in the entire cohort was only 4 months. The role of SRS for patients with ≥4 BM is currently under further study in randomized trials at UC San Francisco/North American Gamma Knife Consortium,⁴² and the MD Anderson Cancer Center.⁴³

Technical Advances for Multiple Targets

While the increased use of SRS for patients with higher numbers of BM is predominantly driven by a desire to avoid side effects associated with WBRT, it also stems from improved technology in the past 5–10 years. Modern SRS can be delivered with increasing speed while maintaining precision and accuracy, which improves procedural tolerability. Standard SRS delivery has historically entailed placing an isocenter target within each BM and then setting up and treating each BM sequentially, which can take several hours to complete for patients with 5–10 BM for some SRS treatment platforms. In contrast, there has been increasing literature and clinical experience over the past 5 years of using a single isocenter within the brain^44–47 and then using radiotherapy techniques such as helical tomotherapy and volumetric modulated arc therapy to deliver dose to multiple BM at the same time. While this technique relies on precise patient immobilization, it has the substantial advantage of being able to reduce treatment time for a patient with multiple BM down to minutes, without sacrificing accuracy or significantly increasing the integral radiation dose to the brain. For example, in one comparison of multitarget single-isocenter SRS versus standard sequential multiple-isocenter SRS using Gamma Knife (GK) among 28 patients with a mean of 3 BM, modeled median treatment time for standard GK was over 2 h compared with ∼20 min with the single-isocenter technique.⁴⁶ This expedited treatment time for single isocenter has also been published by other groups⁴⁴ and makes SRS treatment of patients with 5, 10, 15, or more BM vastly more feasible and tolerable for the patient. Similarly, several centers are now using volumetric modulated arc therapy to combine both WBRT and simultaneous hypofractionated radiation boosts to individual BM within the same integrated treatment session. This allows for dose escalation to individual BM, without requiring separate treatment sessions from the WBRT fractions.⁴⁸ Ultimately, the decision regarding whether to employ SRS, WBRT, or both for patients with multiple BM depends on clinician discretion, patient values, and logistical considerations. The absolute BM number cutoffs are becoming less crucial in the modern era, as intracranial BM recurrence and QoL concerns need to be balanced for individual patients on a case-by-case basis, especially in the setting of substantial increases in life expectancy from better systemic control for some subgroups of patients with BM, with a longer timeframe at risk for treatment sequelae.

Hypofractionation and Resection Cavity Treatment

Given that SRS is typically reserved for BM that are 3 cm or less in diameter in order to reduce the risk of radiation necrosis (RN) in surrounding normal tissues, there has been increasing study of using hypofractionated SRS for larger BM that are not surgically resected, or BM near sensitive structures such as the optic nerves and brainstem. Hypofractionated SRS, also referred to as hypofractionated stereotactic radiotherapy (SRT), typically entails delivery of 3–5 fractions on consecutive or alternating days, with the rationale that fractionating the treatment allows increased recovery of nearby normal tissues and may reduce toxicity. In one survey of CNS radiation oncologists in the United States, the 2 most common hypofractionation regimens used were 25 Gy in 5 fractions, and 21 Gy in 3 fractions.⁴⁹ Retrospective studies of hypofractionated SRS for BM have demonstrated a favorable toxicity profile, with <5% of patients experiencing acute grade 3 toxicity and ∼5% developing RN,^50,51 lower than might be anticipated if using single-fraction SRS for large lesions. However, the optimal dose/fractionation for hypofractionated SRS/SRT remains undefined, as the local control at 12 months ranged from 56% to 68% in 2 recent reports that predominantly utilized 25 Gy in 5 fractions,^50,51 suggesting that further dose intensification may be warranted when using hypofractionation, and prospective data are lacking.

Finally, there is a rapidly expanding recent body of literature on outcomes of single-fraction SRS or hypofractionated SRS targeting the resection cavity after surgical resection of BM⁵² as a means of avoiding adjuvant WBRT. An initial experience published by Stanford University in 2008 among 72 patients observed a 79% local control rate at 12 months, using a median dose at the margin of the target/cavity of 18.6 Gy.⁵³ Many retrospective studies have reported local control rates in the 85%–95% range,^54,55 and the addition of a 2-mm margin around the resection cavity has been shown to be associated with improved local control.⁵⁶ Some reports have suggested a higher risk of leptomeningeal dissemination among patients with breast cancer BM treated with cavity SRS,⁵⁷ though other studies have not observed this.⁵⁸ Hypofractionated SRS has also been used by many centers for larger surgical cavities measuring well over 3 cm, with local control found to be similar to single-fraction SRS.^59,60 However, a randomized trial has not yet been published examining how cavity SRS compares with adjuvant WBRT after surgical resection of BM, and 2 randomized trials are currently ongoing to answer this question with more statistical rigor. One is from the MD Anderson Cancer Center⁶¹ and is evaluating local control after either observation versus cavity SRS following resection of BM. The second is being conducted through the Alliance for Clinical Trials in Oncology (N107C)⁶² and examines OS and neurocognitive function 6 months following adjuvant WBRT versus cavity SRS for resected BM. In addition, there are emerging data on the use of preoperative SRS administered shortly before resection, as opposed to postoperative SRS, with the goals of improving local control due to improved targeting of an intact metastasis and reducing risks of leptomeningeal dissemination and RN. One multi-institutional retrospective study of 180 patients treated from 2005 to 2013 with SRS before or after resection, 37% of whom received preoperative SRS within 48 h before surgery, demonstrated similar rates of local control between those receiving pre- versus postoperative SRS but significantly higher rates of leptomeningeal disease (16.6% vs 3.2%, P = .01) and symptomatic RN (16.4% vs 4.9%, P = .01) at 2 years among those receiving postoperative SRS.⁶³ New trials are in development that will further compare pre- versus postoperative SRS.

Neurocognition and Quality of Life after WBRT

Over the past 5 years, there has been accumulating evidence on deleterious effects patients may experience after WBRT in particular, as well as new approaches to potentially mitigate those harms. Most distressing to patients are the risks of neurocognitive decline after WBRT, which was investigated in the MD Anderson trial³¹ cited above, but was expanded in the much larger Alliance N0574 randomized trial²⁷ that was recently presented at the American Society for Clinical Oncology 2015 Annual Meeting. In the Alliance trial, 213 patients with up to 3 BM received SRS with or without WBRT, with primary endpoint being cognitive progression at 3 months, defined as a decline by >1 standard deviation in any of the 6 neurocognitive tests administered. At 3 months, cognitive decline was observed in 88.0% of patients in the combined arm versus 61.9% in the SRS-only arm (P = .002), including domains such as immediate recall, delayed recall, and verbal fluency. There were no statistically significant OS differences between arms, though intracranial tumor control was higher in the WBRT-containing arm, as anticipated. Another large recent study that informs our understanding of the risks of irradiation of the entire brain was a pooled secondary analysis of 2 randomized trials evaluating the role of prophylactic cranial irradiation (PCI) in non-small-cell or small-cell lung cancers, which included 410 patients who received PCI and 173 patients who did not.⁶⁴ PCI was associated with an approximately 3.5 times higher risk of decline in self-reported cognitive functioning at both 6 and 12 months compared with patients not receiving PCI (both P < .001). Objective declines in neurocognitive testing, as ascertained by the HVLT, were also associated in PCI at those same time points (both P = .002), but intriguingly, correlation between objective decline and self-reported decline was not significant (P = .86 at 12 mo). The clinical findings of cognitive decline following WBRT are echoed in a retrospective study of 68 patients who survived at least 1 year after treatment: 97% of patients who received WBRT developed white matter changes on MRI compared with only 3% of those who received SRS.⁶⁵

In addition to neurocognition specifically, QoL in general is an important consideration for patients and physicians in the selection of treatment for BM. A decline in neurocognition has been found to precede and predict a deterioration in QoL in this patient population,⁶⁶ and thus they are not completely distinct entities. The EORTC published health-related QoL results⁶⁷ in 2013 on the patients with 1–3 BM enrolled in the SRS or surgery ± WBRT trial²⁹ described above. Patient compliance with QoL recording had dropped to only 45% by 1 year of follow-up, and thus only the first year after treatment was analyzed. While QoL scores were generally lower among patients receiving WBRT in addition to SRS/surgery, the effects appeared to be relatively transient. QoL scores were statistically significantly lower for global QoL among patients who received WBRT only at 9 months, but not at 2, 3, 6, or 12 months, and similarly were lower for cognitive functioning at 12 months, fatigue at 2 months, and physical functioning at 2 months, as isolated timepoints. These data could be used to support the contention that QoL is lower among patients who have received WBRT, but more importantly show that most patients do not have chronically, consistently worse QoL globally or within specific domains after WBRT.

Approaches to Reduce Cognitive Side Effects from WBRT

In an effort to reduce the neurocognitive sequelae of WBRT, 3 recently reported trials have investigated different strategies to accompany WBRT. Firstly, RTOG 0614 was a Phase III trial investigating memantine, an N-methyl-d-aspartate receptor blocker shown to be effective in vascular dementia, and included 508 patients with BM who were randomized to either WBRT plus placebo versus WBRT plus daily memantine for 6 months.⁶⁸ The study narrowly missed meeting statistical significance on the primary endpoint of less decline in delayed recall at 6 months favoring the memantine arm (P = .059), likely a result of patient mortality with only 280 analyzable patients (55% of enrollees). The memantine arm did show significantly longer time to cognitive decline (HR 0.78, P = .01), with probability of cognitive decline at 6 months being 54% in the memantine arm and 65% in the placebo arm. Memantine is now being prescribed to patients receiving WBRT by many providers on the basis of these data, as both this trial and the Alliance trial of SRS with or without WBRT²⁷ again highlighted the cognitive side effects of standard WBRT. However, neurocognitive deterioration is still quite common among patients with BM, even among the superior-performing trial arms, with over half of the patients in the arm receiving memantine (RTOG 0614) and over half of the SRS-alone treated arm (Alliance) experiencing neurocognitive decline within 3–6 months.

A second strategy to ameliorate the neurocognitive effects of WBRT has been to use intensity-modulated radiotherapy (IMRT) treatment planning to develop conformal whole brain radiation plans that preferentially spare the bilateral hippocampi. This is based on the role of the hippocampus in developing and preserving memory. RTOG 0933 was a single-arm Phase II trial that enrolled 113 patients who received hippocampal-avoidance WBRT for BM and assessed neurocognitive function, including a primary endpoint of HVLT-R delayed recall at 4 months, compared with a historical control group.²¹ Only 42 patients were analyzable at 4 months, again highlighting the challenges to conducting trials among BM patients due to high mortality. Patients receiving hippocampal-avoidance WBRT had no measurable decline in QoL and a mean relative decline in delayed recall score of 7%, which was lower than the 30% decline seen in the historical control (P < .001). This study has been considered hypothesis generating and has not been adopted as standard of care, as it was only a single-arm study. NRG Oncology has initiated 2 trials to further explore whether hippocampal avoidance during WBRT decreases the incidence of neurocognitive sequelae. The first is NRG-CC001, a randomized Phase III trial examining memantine and WBRT with or without hippocampal avoidance, with a primary endpoint of time to neurocognitive decline.⁶⁹ The second trial will be NRG-CC003, a randomized Phase II/III trial of PCI with or without hippocampal avoidance among patients with limited or extensive stage small-cell lung cancer who achieve at least a partial remission to chemotherapy.

Finally, a recent multi-institutional trial has investigated the efficacy of the dementia medication donepezil, to evaluate whether this neurotransmitter modulator may improve cognitive function among patients who had previously received partial brain RT or WBRT. In the Phase III trial, 198 patients with primary (n = 130) or metastatic (n = 68) brain tumors who were at least 6 months out from brain-directed radiation received donepezil or placebo for 6 months.⁷⁰ Cognitive composite scores, the primary outcome, did not differ significantly between the treatment arms, yet significant differences in 2 domains of memory, as well as motor speed and dexterity domains, were statistically improved in the donepezil arm. In all, the recent trials of memantine, hippocampal avoidance, and donepezil all showed some improvements in neurocognition in patients with BM.

Diagnosis and Management of Radiation Necrosis after SRS

With the growing role of SRS in the management of BM, alongside advances in systemic therapy leading to longer overall patient survival, an increasingly common late side effect that physicians confront in clinic is the development of RN after SRS. After high-dose, conformal radiation to the brain such as with SRS, vascular endothelial cells may become damaged and lead to fibrinoid necrosis of small arteries, tissue hypoxia, and damage of neurons and glial cells, which is mediated by various cytokines as well as vascular endothelial growth factor (VEGF).^71,72 RN occurs in at least 10% of patients who receive SRS, typically between 6 and 18 months after treatment, and is often challenging to distinguish from recurrent/progressive BM based on conventional MRI given the often similar appearance of both contrast enhancement and edema.^73,74 In addition, radiographic evidence of SRS-induced breakdown of the BBB without development of an overt necrotic mass, as occurred in 10% of patients in the EORTC trial of SRS with or without WBRT,²⁹ is often referred to as “radiation treatment change” and may represent early stages of RN. Moderate doses of corticosteroids such as dexamethasone are often prescribed to control associated neurologic symptoms and/or potentially reduce local inflammation associated with RN. While the gold standard for diagnosis is surgical resection of the lesion, this is not always feasible or desirable, and may show a mix of RN and tumor. Several recent studies have investigated the role of various noninvasive neuroimaging approaches to help distinguish RN from tumor recurrence. These include MR perfusion⁷⁵ and dual-phase PET,⁷⁶ among others. While these tests can provide adjunctive information, no modality has yet shown adequate sensitivity and specificity to reliably differentiate these 2 phenomena noninvasively, as acknowledged in the recently published Response Assessment in Neuro-Oncology–Brain Metastases (RANO-BM) guidelines.⁷⁷ Most frequently, clinicians must rely on the pretest probability/clinical suspicion of RN, and serial MRI over time if the patient is asymptomatic, to guide management. While standard risk factors for RN include SRS dose, volume treated, prior irradiation, and chemotherapy,⁷³ there are growing reports that immunotherapies and targeted systemic therapies that cross the BBB may increase the risk of RN.^78–80

Emerging treatments for RN that have published in the past 5 years include bevacizumab and laser interstitial thermal therapy (LITT; otherwise known as laser induced thermotherapy). Bevacizumab is an antibody against VEGF factor A and was tested in a small randomized trial of 14 patients who experienced symptomatic brain RN after fractionated radiotherapy for primary brain or head/neck tumors and were randomized to either placebo or bevacizumab every 3 weeks for 4 cycles.⁸¹ All patients who received bevacizumab had radiographic response with reduction in edema and contrast enhancement, and all had reduction in neurologic symptoms/signs, whereas no patients receiving placebo had an initial radiographic or symptomatic response, until the placebo group crossed over and received bevacizumab, and thereafter all experienced both radiographic and clinical response. Based on these data and additional institutional series, bevacizumab can be considered an option for BM patients without contraindications to its use, who have progressive symptoms from RN after SRS despite corticosteroids and are not felt to be good surgical candidates. However, to improve the evidence base for bevacizumab in the BM population, there is also an ongoing larger randomized Phase II study through the Alliance for Clinical Trials in Oncology evaluating the efficacy of corticosteroids with versus without bevacizumab for symptomatic RN specifically after SRS for BM, which will provide more robust data in an SRS-treated BM cohort.⁸² In contrast to bevacizumab, LITT is an emerging technology that uses thermocoagulation to destroy inflammatory cellular infiltrate in the region of RN, and may potentially be used for RN, progressive BM after SRS, or a combination of RN and progressive BM, with an evidence base that is small but expanding. Neurosurgeons who perform LITT typically drill a burr hole or may perform a small open craniotomy and insert a flexible LITT probe until reaching the region of interest, at which point the probe is heated to the desired range. The lesion may be biopsied through the same probe, prior to performing thermocoagulation. Single-institution reports have found excellent outcomes after LITT, with low risks among appropriately selected cases.⁸³ LITT represents another option for clinicians to consider for RN depending on availability, if corticosteroids have failed and full surgical resection of BM that previously received SRS is not feasible, especially if biopsy of the lesion is desired for diagnosis as part of the same procedure.

WBRT and Concurrent Systemic Therapy

Randomized trials to date have not demonstrated an OS benefit from adding systemic therapy to WBRT (Table 2).^84,85 With regard to other endpoints such as radiographic response rate, progression-free survival, and time to neurologic deterioration, results are mixed. However, toxicity is generally increased when adding systemic therapy.⁸⁴ While a recent Phase II study of concurrent WBRT with erlotinib among NSCLC patients showed low toxicity rates,⁸⁶ a larger, randomized Phase III trial (RTOG 0320) reported grades 3–5 toxicity rates of 41%–49% in the arms receiving either temozolomide (TMZ) or erlotinib concurrently with WBRT, compared with 11% receiving WBRT alone.⁸⁷ It has been speculated that the shorter survival observed among the 2 concurrent chemotherapy arms of the RTOG trial may be attributable to higher rates of toxicity, but the trial was underpowered to prove this assertion. Most studies of WBRT with concurrent systemic therapy have enrolled lung cancer patients, and further studies are needed to examine the role of adding systemic therapy to WBRT across a variety of malignancies. While TMZ is sometimes administered concurrently with WBRT for patients with BM from malignant melanoma, the limited prospective data on this subject suggest minimal benefit.⁸⁸ In addition, a Phase II randomized trial of concurrent TMZ with WBRT for patients with BM from breast cancer showed no improvement in objective response rate, progression-free survival, or OS.⁸⁹ Ongoing randomized studies include a study of WBRT with or without lapatinib for patients with BM from HER2-positive breast cancer.^90,91

Table 2.

Select trials of WBRT with concurrent systemic agents for brain metastases

Studya	Systemic Agent	Primary Cancer	Trial Design	N	WBRT Dose	Primary Endpoint	Overall Survival	CNS Response Rate (CR + PR) in Evaluable Patients	Other Outcomes
Guerrieri et al, 2004161	Carboplatin	NSCLC	Phase III randomized controlled trial (stopped prematurely due to poor accrual)	G1: WBRT (N = 21), G2: WBRT + carboplatin (N = 21)	20 Gy in 5 fractions	Overall survival	G1: 4.4 mo G2: 3.7 mo (P = NS)	G1: 10% G2: 29% (P = NS)
Ushio et al, 1991162	Chloroethylnitrosureas tegafur	Lung	Phase III randomized controlled trial	G1: WBRT (N = 25) G2: WBRT + chloroethylnitrosureas (N = 34) G3: WBRT + chloroethylnitrosureas + tegafur (N = 29)	40 Gy in 1.5–2 Gy fractions	Tumor control	G1: 27 wk G2: 29 wk G3: 30.5 wk (P = NS)	G1: 36% G2: 69% G3: 74% G1 vs G2 (P = NS) G2 vs G3 (P = NS) G1 vs G3 (P < .05)
Margolin et al, 200288	Temozolomide	Melanoma	Phase II single arm trial	N = 31	30 Gy in 3 Gy fractions	Objective response	6 mo	10%
Antonadou et al, 2002163	Temozolomide	Lung, breast, or unknown primary	Phase II randomized trial	G1: WBRT (N = 23) G2: WBRT + concurrent TMZ followed by 6 cycles of adjuvant TMZ (N = 25)	40 Gy in 2 Gy fractions	RR and neurologic symptom evaluation	G1: 7.0 mo G2: 8.6 mo (P = NS)	G1: OR 67% G2: OR 96% (P = .017)	Improvement in neurologic function in the TMZ group
Verger et al, 2005164	Temozolomide	Multiple (∼50% lung, 15% breast)	Phase II randomized trial (stopped prematurely due to poor accrual)	G1: WBRT (N = 41) G2: WBRT + concurrent TMZ followed by 2 cycles of adjuvant TMZ (N = 41)	30 Gy in 3 Gy fractions	Neurologic toxicity	G1: 3.1 mo G2: 4.5 mo (P = NS)	Response at 30 d G1: OR 32% G2: OR 32% (P = NS)	Freedom from intracranial progression at 90 d was improved in the TMZ group
Sperduto et al, 201287	Temozolomide or erlotinib	NSCLC	Phase III randomized trial	G1: WBRT + SRS (N = 44) G2: WBRT + SRS + TMZ (N = 40) G3: WBRT + SRS + erlotinib (N = 41)	WBRT: 37.5 Gy in 2.5 Gy fractions SRS dose was size dependent: lesions < 2 cm, 2.1 to 3.0 cm, and 3.1 to 4.0 cm received 24, 18, and 15 Gy, respectively	Overall survival	G1: 13.4 mo G2: 6.3 mo G3: 6.1 mo (P = NS for G1 vs G2 and G1 vs G3)	Not reported	No difference between groups in time to CNS progression Grade 3–5 toxicity rates were 11% in G1, 41% in G2, and 49% in G3; high toxicity rates may have led to inferior survival in the combination arms
Welsh et al, 201386	Erlotinib	NSCLC	Phase II single-arm trial	N = 40	35 Gy in 2.5 Gy fractions	Overall survival	11.8 mo	86%	No grade 4–5 toxicities Median survival was 19.1 vs 9.3 mo for patients with vs without an EGFR mutation
Knisely et al, 2008 (RTOG 0118)165	Thalidomide	Multiple (∼60% lung)	Phase III randomized trial	G1: WBRT (N = 93) G2: WBRT + thalidomide (N = 90)	37.5 Gy in 2.5 Gy fractions	Overall survival	G1: 3.9 mo G2: 3.9 mo	Not reported	No difference between arms in time to progression or rate of death due to brain metastases. 48% patients discontinued thalidomide due to side effects
Neuhaus et al, 2009166	Topotecan	Lung	Phase III randomized trial (stopped prematurely due to poor accrual	G1: WBRT (N = 49) G2: WBRT + topotecan (N = 47)	20 Gy in 2 Gy fractions	Overall survival	G1: 95 d G2: 87 d (P = NS)	Response 2 wk after treatment G1: 52% G2: 60%	No difference between groups in progression free survival

Abbreviations: CR, complete response; PR, partial response; OR, odds ratio. ^aArranged according to systemic agent and study design.

Surgery

The role of surgery in the management of BM has evolved over time and remains a key treatment modality to consider especially for BM over 3 cm in size or otherwise bulky lesions causing neurologic symptoms, and when tissue is necessary to establish a diagnosis. While the increasing availability of SRS may lead to fewer resections among BM that are borderline in size, improvements in systemic therapy leading to longer patient survival may be increasing the number of patients with BM overall, including those requiring surgery. As discussed previously, results from the EORTC trial of SRS or surgery with or without WBRT²⁹ echoed those from the classic trial by Patchell et al³⁰ of surgery with or without WBRT, in that neither study demonstrated an OS advantage with the addition of WBRT after surgery. While most clinicians still appropriately recommend radiotherapy (WBRT or cavity SRS) after resection of BM for improved local tumor control, it is important to recall that this has not been shown in randomized trials to improve survival, and there may be patient-specific scenarios where surgery alone followed by surveillance is appropriate.

Other emerging updates with regard to the surgical management of BM have been cited above, including the increasing prevalence of postoperative SRS to the surgical cavity, which avoids WBRT, and investigation of preoperative SRS, which has been shown in at least one study to have similar local control as cavity SRS but with higher rates of both leptomeningeal disease and RN.⁶³ In addition, LITT is a new tool for neurosurgeons in the management of RN after SRS for BM as highlighted above, though its evidence base is thin and requires more study to better establish efficacy and optimal indications.

Systemic Therapy

Challenges in Treating Brain Metastases with Systemic Therapy

Because active BM often coexist with active systemic disease, antitumor agents that can control both intracranial and extracranial disease are needed. Unfortunately, few clinical trials of systemic agents have been conducted to date in patients with BM, and this population has frequently been excluded from clinical trials of emerging investigational drugs.⁹² Comparison across BM trials has also been hampered by clinical trial design including enrollment of heterogeneous populations and varying definitions of criteria to assess response and progression. Response Evaluation Criteria in Solid Tumors (RECIST) for response assessment sums representative target lesions across all organ sites, including CNS and non-CNS sites. For therapies with differential responses in CNS and non-CNS locations, RECIST may not adequately describe CNS progression or response. Strong consideration should be given to assessing CNS and non-CNS for progression as separate compartments. Indeed, RANO-BM is an international collaboration attempting to address these issues in clinical trial design and radiographic assessment by developing uniform radiographic response criteria for use in BM trials.^77,93,94 RANO-BM recommends assessing the CNS and non-CNS sites according to RANO-BM criteria and RECIST, respectively. These radiographic response criteria proposed by RANO-BM are now being adopted in clinical trials of patients with BM where the CNS outcomes are important primary objectives. In addition, the Jumpstarting Brain Tumor Drug Development Coalition Imaging Standardization Steering Committee (an international collaboration which includes members of RANO-BM) recently published consensus recommendations for a standardized brain tumor imaging protocol in clinical trials.⁹⁵

BM differ from metastases to other organs from a pathophysiologic and clinical perspective.⁹⁶ The brain has a unique microenvironment and an immune system distinct from other organs.⁹⁶ For example, Zhang et al⁹⁷ showed that tumor cells with normal expression of phosphatase and tensin homolog (PTEN) lose PTEN expression after dissemination to the brain (but not to other organs) and regain PTEN expression after leaving the brain microenvironment in a process that is epigenetically regulated by microRNAs from brain astrocytes. BM patients may also develop neurologic deficits and seizures, which could require the use of medications (ie, enzyme-inducing anticonvulsants, steroids) that can interfere with the metabolism of antitumor agents. These issues in part lead to the perception that BM patients are less than ideal candidates for early-phase clinical trials. However, in a review of Phase I clinical trials at MD Anderson Cancer Center comparing patients with treated, stable BM versus those without BM, the presence of BM was not an independent factor for survival, and the time to treatment failure was not different between groups.⁹² In addition, the incidence of serious adverse events including neurologic toxicity was not increased in the BM population. These data suggest that patients with treated and stable BM should indeed be included in early-phase clinical trials.

Another way that BM differ from metastases to other organs is the presence of the BBB, which is a specialized cellular barrier that prevents macromolecules from entering the brain and can efficiently pump out unwanted compounds through efflux transporters such as P-glycoprotein and the breast cancer resistance protein (ATP-binding cassette subfamily G member 2).⁹⁸ This is best demonstrated by targeted agents associated with good systemic control but insufficient CNS control, such as trastuzumab in HER2-positive breast cancer⁹⁹ and crizotinib in NSCLC with ALK-gene rearrangement.^100,101 Intrathecal chemotherapy can deliver drug directly to the CSF but cannot reliably treat parenchymal disease or even bulky leptomeningeal disease, as diffusion of drug into tumor deposits thicker than 1 mm, along nerve root sleeves, and into the Virchow–Robin spaces is limited.^102,103

The BBB in BM is somewhat disrupted, as demonstrated by leakage of gadolinium enhancement into the tumor on brain MRI. However, there is debate about how disrupted the BBB is and whether we can rely on this disruption (however minimal) to permit sufficient drug penetration. For example, mouse models of BM suggest that penetration of lapatinib, a HER2-directed tyrosine kinase inhibitor (TKI), is variable between and within metastases, with some lesions demonstrating very high levels of lapatinib (17% freely permeable) and others not statistically distinct from normal brain.¹⁰⁴ The degree of drug penetration was not well correlated with lesion size. Similar findings were obtained in a prospective clinical study of capecitabine and lapatinib uptake in surgically resected BM from 12 breast cancer patients.¹⁰⁵ Detectable levels of capecitabine or lapatinib were found in resected tumor tissue, but drug distribution was widely variable among patients. In another experimental model of breast cancer, BM uptake of ¹⁴C-paclitaxel and ¹⁴C-doxorubicin was generally greater than normal brain but <15% of that of other tissues or peripheral metastases and reached cytotoxic concentrations in only a small subset (∼10%) of the most permeable metastases.¹⁰⁴

Attempts to improve drug delivery across the BBB include inhibition of drug efflux transporters, pulsatile dosing regimens, and BBB disruption with ultrasound, although these strategies remain in clinical development.¹⁰⁶ ANG1005 combines paclitaxel with an amino acid peptide that targets the lipoprotein receptor-related protein 1 receptor (which mediates transcytosis on endothelial cells of the BBB and mediates endocytosis on multiple tumor cells). A Phase II study of ANG1005 in breast cancer patients with BM demonstrated antitumor activity in the CNS with radiographic response rates of 22%–40%.¹⁰⁷ Escalating dosing of the epidermal growth factor receptor (EGFR) TKIs gefitinib and erlotinib may increase CNS penetration.¹⁰⁸ Pulsatile dosing of erlotinib at 1500 mg once weekly is associated with a CSF concentration of 130 nM, sufficient to inhibit 50% growth in cancer cell lines.¹⁰⁹ A retrospective analysis suggested that pulsatile dosing is well tolerated and effective in patients with EGFR-mutant lung cancers with CNS metastases (parenchymal and leptomeningeal) even after progression on standard dosing erlotinib.¹¹⁰

Some antitumor agents are not dependent on whether they can cross an intact or even variably disrupted BBB—for instance, immune checkpoint blockade therapies such as the cytotoxic T lymphocyte antigen (CTLA)–4 antibody ipilimumab, which relies on activated T cells in the periphery to migrate into the CNS, as well as angiogenesis inhibitors such as bevacizumab, which need reach only the luminal side of the vascular wall. Both of these agents have demonstrated activity in BM and, as detailed above in the case of bevacizumab, against RN resulting from prior SRS. Of note, both ipilimumab and bevacizumab are monoclonal antibodies and are generally considered too large to cross an intact BBB. Nonetheless, ipilimumab has activity in BM from melanoma.⁷ In a Phase II trial of ipilimumab in metastatic melanoma patients who were either neurologically asymptomatic (n = 51) or symptomatic and on a stable dose of corticosteroids (n = 21), the median OS was 7 months and 3.4 months, respectively.¹¹¹ The FDA's expanded access program reported 1-year survival rates of 20% in melanoma patients with stable, asymptomatic BM treated with ipilimumab 10 mg/kg.¹¹² In Italy, an ongoing randomized Phase III trial is examining the benefit of adding fotemustine or the anti–programmed cell death protein 1 (PD1) agent nivolumab to ipilimumab in patients with BM from metastatic melanoma,¹¹³ and 2 other trials, based in the US and Australia, are also testing combination nivolumab plus ipilimumab in melanoma metastatic to the brain.^113,114

Due to concerns for CNS hemorrhage, BM patients were excluded from early trials of bevacizumab.¹¹⁵ However, analysis of datasets from bevacizumab clinical trials demonstrated low rates of CNS hemorrhage in patients with CNS metastases.¹¹⁵ In patients with occult CNS metastases enrolled in randomized controlled Phase II/III trials, 3 of 91 bevacizumab-treated patients (3.3%) developed grade 4 cerebral hemorrhage, while 1 of 96 control patients (1.0%) developed a grade 5 cerebral hemorrhage. The inclusion of bevacizumab with standard cytotoxic agents has since been associated with durable response rates in BM from breast cancer¹¹⁶ and in asymptomatic, untreated BM from nonsquamous NSCLC.¹¹⁷ In the latter patient population, bevacizumab plus carboplatin and paclitaxel demonstrated an intracranial response rate of 61.2% and an extracranial response rate of 64.2% with an acceptable safety profile with only one grade 1 intracranial hemorrhage.¹¹⁷ There is an ongoing randomized Phase III trial examining the addition of bevacizumab to cisplatin and pemetrexed as first-line therapy for patients with asymptomatic BM from NSCLC.¹¹⁸

Finally, targeted treatments of BM are often based on the molecular profiles of the primary tumor, as opposed to the BM itself. However, recent studies comparing molecular profiles of matched BM and primary tumors demonstrate that somatic alterations in the BM are frequently discordant with those in the primary tumor.^119–121 Brastianos et al¹²⁰ performed whole-exome sequencing of 86 matched BM and primary tumors (mainly breast, lung, and renal cell cancers) and found that 53% of BM harbored a potentially clinically actionable somatic alteration not detected in the matched primary tumor. The brain microenvironment may play an important role in driving some of these alterations. These data suggest that for clinical trials of targeted agents in BM, biopsy or surgical resection of BM may be indicated for molecular profiling.

Chemoprevention of Brain Metastases

In cancer patients with a high risk for CNS relapse, interest in pharmacologic prevention of BM (using the same classes of agents with proven activity either systemically for that cancer or in primary brain tumors) is growing.¹²² Such a strategy may prevent brain recurrence while preserving neurocognition and QoL.¹²³ In preclinical mouse models of brain-tropic breast cancer, TMZ administered 3 days after inoculation prevented the formation of experimental BM from cells negative for O⁶-methylguanine-DNA methyltransferase.¹²⁴ In a retrospective study of patients with EGFR-mutant advanced NSCLC, the risk of CNS progression initially treated with gefitinib or erlotinib is lower compared with upfront chemotherapy.¹²⁵ These data suggest a role for chemoprevention of CNS metastases in specific patient populations.

Chemoprevention strategies have been preliminarily tested in clinical trials. While these studies have not met their primary endpoints of preventing BM, they are informative for future study design. The CEREBEL study was a Phase III, randomized trial of lapatinib + capecitabine versus trastuzumab + capecitabine for prevention of CNS metastases.¹²⁶ Unfortunately, the study was closed early and was inconclusive for its primary endpoint. Twenty percent of patients failed screening due to asymptomatic BM on brain MRI at study entry. The 3% and 5% rates of CNS as first site of relapse with lapatinib + capecitabine and trastuzumab + capecitabine, respectively, were far lower than the expected rates of 12% and 20%, respectively; these expected rates were based on historical trials in which baseline CNS imaging was not required (and hence likely included a substantial number of patients with occult BM who then had a higher likelihood of progression in the brain during the course of the trial). Therefore, the study was underpowered for its primary endpoint. A Phase II randomized study of maintenance TMZ versus observation for prevention of BM in high-risk NSCLC was also closed early with 53 patients enrolled.¹²³ There was no difference in the incidence of BM at one year (18% in the TMZ arm and 13% in the observation arm, P = .6995). Again, the incidence of BM was lower than expected (13% in the observation arm vs the 40% predicted based on historical data). Furthermore, TMZ prophylaxis is unlikely to be effective in view of its overall very low antitumor activity against NSCLC in general. Future chemoprevention efforts will need to be redesigned based on improved agents, a better understanding of the populations at risk for CNS relapse, pharmacokinetics and pharmacodynamics, the emergence of improved imaging methods, and a better understanding of the safety of combining systemic therapy with radiotherapy. For the present and the foreseeable future, the best form of prophylaxis against BM will continue to be the discovery of agents and strategies with increasing overall activity against each type and subset of cancer.

Tumor-Specific Systemic Therapies for Brain Metastases

Guidelines from the National Comprehensive Cancer Network for management of BM recommend local radiation or surgery as upfront treatment.¹²⁷ Indeed, no chemotherapy is approved by the US FDA for management of BM from breast cancer, lung cancer, melanoma, or other solid tumors, and none of the available systemic therapies are generally considered first-line treatment for BM.¹²⁸ Nonetheless, in a few select patients with asymptomatic or minimally symptomatic BM, treatments with established efficacy in BM may be reasonable to consider as initial therapy.

In breast cancer patients, the risk of developing BM and survival after BM diagnosis vary according to tumor subtype.^99,129 In patients with triple negative breast cancer, the risk of CNS relapse may be as high as 46%, and CNS involvement often occurs in the setting of active systemic disease.⁹⁹ This high rate of CNS involvement is unlikely to be due to a sanctuary effect, but rather to the lack of effective therapies in general for this aggressive subtype of breast cancer.²³ In breast cancer patients with HER2-positive disease, up to 40%–50% will develop BM,⁵ often despite controlled systemic disease.⁹⁹ Some studies suggest that up to half of HER2-positive patients die of BM-related causes,¹³⁰ emphasizing the importance of CNS control in this patient population. For the HER2-positive patient population, the American Society of Clinical Oncology offers recommendations on the management of BM.⁶ Clinical trials of lapatinib in combination with capecitabine have demonstrated activity with a CNS objective response rate of 20% in patients with previously treated BM¹³¹ and 66% (using a volumetric response criteria) in patients with previously untreated BM¹³² (Table 3). More recent trials in the HER2-positive population with BM have examined the potential benefit of other HER2-directed TKIs. Afatinib (alone or in combination with vinorelbine)¹³³ did not demonstrate efficacy over investigator's choice for management of BM in HER2-positive patients and was not as well tolerated. Single-agent neratinib had low efficacy in previously treated BM with a CNS objective response rate of only 8%,¹³⁴ but neratinib in combination with chemotherapy may prove more efficacious. In the NEfERT-T trial for untreated metastatic HER2-positive breast cancer (including women with asymptomatic CNS metastases), the neratinib + paclitaxel arm had a lower incidence of CNS recurrences and delayed time to CNS metastases compared with trastuzumab + paclitaxel.¹³⁵ However, these results require validation in a larger study. Other promising agents in clinical trials for breast cancer patients with BM include trastuzumab + emtansine (for HER2+ disease),¹³⁶ pathway inhibitors of phosphatidylinositol 3-kinase–mammalian target of rapamycin, CDK4 inhibitors, and poly(ADP-ribose) polymerase inhibitors.¹³⁷

Table 3.

Select trials of systemic agents for brain metastases according to primary cancer

Study	Systemic Agent	Mechanism of Action	Population	Prior Treatment for BM?	Trial Design	N	Primary endpoint	CNS Response Rate (CR + PR)	Overall Survival
Melanoma
Margolin et al, Lancet Oncol 2012111	Ipilimumab	CTCL-4 inhibitor	Melanoma metastatic to brain	41%–48% had received prior brain irradiation (WBRT or SRS)	Phase II	72 total: 51 in cohort A: neurologically asymptomatic (N = 51) Cohort B: symptomatic (N = 21)	Disease control (CR, PR, SD) at 12 wk = 18% in A and 5% in B	16% cohort A, 5% cohort B	7.0 mo in cohort A and 3.7 mo in cohort B
Long et al, Lancet Oncol 2012138	Dabrafenib	BRAF inhibitor	Val600Glu or Val600Lys BRAF mutant melanoma metastatic to brain	Cohort A previously untreated BM. Cohort B prior local therapy (surgery, WBRT, SRS)	Phase II	172 total: 89 in cohort A and 83 in cohort B	Intracranial response (CR + PR) Val600Glu 39.2% cohort A, 30.8% cohort B. Val600 Lys 6.7% cohort A, 22.2% cohort B	Val600Glu 39.2% cohort A, 30.8% cohort B. Val600 Lys 6.7% cohort A, 22.2% cohort B	Val600Glu 33.1 mo cohort A, 31.4 mo Cohort B Val600Lys 16.3 mo cohort A, 21.9 mo cohort B
Breast Cancer
Lin et al, CCR 2009131	Lapatinib	HER2 inhibitor	HER2+ breast cancer metastatic to brain	All previously treated with brain irradiation (WBRT or SRS)	Phase II	242: 95 in cohort A (ECOG 0–1 and 1–2 prior trastuzumab regimens) and 147 in cohort B (ECOG 2 and/or >2 prior trastuzumab regimens)	CNS objective response (50% or greater volumetric reduction)	6% in cohort A and cohort B	6.4 mo (cohort A 9.6 mo, cohort B 5.5 mo)
	Lapatinib + capecitabine	HER2 inhibitor (lapatinib) and prodrug that converts to 5-fluorouracil (capecitabine)	HER2+ breast cancer metastatic to brain	Extension phase for patients on study who progressed on lapatinib	Phase II	Subset of 50 patients in above study	CNS objective response (50% or greater volumetric reduction)	20%	Not reported
Bachelot Lancet Oncol 2013 LANDSCAPE132	Lapatinib + capecitabine	HER2 inhibitor (lapatinib) and prodrug that converts to 5-fluorouracil (capecitabine)	HER2+ breast cancer metastatic to brain	Untreated (no prior WBRT, SRS, capecitabine, or lapatinib)	Phase II	45	CNS objective response rate (CR + PR)	49%	17 mo
Pivot et al, J Clin Onc 2015 CEREBEL126	Lapatinib + capecitabine (lap + cap) vs trastuzumab + capecitabine (tras + cap)	HER2 inhibitors (lap, tras), prodrug that converts to 5-fluorouracil (cap)	HER2+ metastatic breast cancer without CNS metastases at baseline	No brain metastases at baseline	Phase III, randomized study, preventative trial. Study terminated early	540 (271 lap + cap and 269 tras + cap)	Incidence of CNS mets as first site of relapse: 3% in the lap + cap and 5% in tras + cap, difference not statistically significant	n/a	22.7 mo for lap + cap and 27.3 mo for tras + cap
NSCLC
Bearz et al, Lung Cancer 2010146	Pemetrexed	Folate anti-metabolite	NSCLC metastatic to brain	71.8% received prior brain irradiation	Retrospective review	39	Not specified	38.4%	10 mo
Ceresoli et al, Ann Oncol 2004147	Gefitinib	EGFR inhibitor	Unselected patients with NSCLC metastatic to brain	2 groups: prior WBRT (44%) and no prior WBRT (56%)	Phase II	41	Disease control rate (CR + PR + SD)	10%	5 mo
Wu et al, Ann Oncol 2013 (CTONG-0803)148	Erlotinib	EGFR inhibitor	East Asian patients with adenocarcinoma or EGFR mutant NSCLC, asymptomatic brain metastases (total 16.7% mutation positive)	35.4% received prior WBRT	Phase II	48	PFS	58.3% systemic +intracranial RR (intracranial RR alone not reported)	18.9 mo [patients with EGFR mutant disease 37.5 mo vs EGFR wild-type 18.4 mo vs unknown EGFR status 19.4 mo, P-value .14]
Besse et al, Clin Ca Res 2015117	Bevacziumab + carboplatin and paclitaxel	VEGF inhibitor (bevacziumab), alkylating agent (carboplatin), taxane (paclitaxel)	Unselected NSCLC patients with asymptomatic, prevously untreated brain metastases	No prior treatment	Phase II (noncomparative study with 2 arms atlhough only the bevacizuamb + carboplatin and paclitaxel arm results reported)	67 in the B + CP arm	Investigator-assessed 6-month PFS rate	61.2%	16 mo
Ou et al, J Clin Oncol 2016155	Alectinib	ALK inhibitor	ALK-rearranged NSCLC resistant to crizotinib, patients with stable treated or asymptomatic untreated brain and leptomeningeal metastases allowed	61 (73%) of 84 patients with CNS metastases at study entry had prior brain irradiation	Phase II	84 (61%) of the 138 patients enrolled on study had CNS metastases at baseline	Objective response rate (CNS objective response rate was a secondary objective)	57% in 35 patients with measurable CNS lesions	Not reported
Boggs et al, Lung Ca 2014123	Temozolomide (TMZ)	Alkylating agent	NSCLC stage IIIA, IIIB, or IV (only pleural or pericardia effusion), no brain metastases at baseline	No brain metastases at baseline	Phase II randomized trial – preventative trial Study terminated early	53 (26 TMZ, 27 observation)	Incidence of brain metastases at 12 mo: 18% in TMZ, 13% in observation	n/a	27.1 mo in TMZ vs 22.5 mo in observation (P = .7)

Abbreviations: CR, complete response; PR, partial response; PFS, progression-free survival; RR, response rate; mets, metastases.

For melanoma patients, up to 60% will develop BM during the course of their illness.⁷ Ipilimumab¹¹¹ (as previously discussed) or BRAF inhibitors such as dabrafenib¹³⁸ (for patients with BRAF^V600 mutations) can be considered for management of progressive BM (Table 3). A Phase II study of dabrafenib in patients with Val600Glu or Val600Lys BRAF-mutant melanoma and previously untreated BM (n = 89) or treated BM (n = 83) demonstrated remarkable median OS of 33 weeks and 31 weeks, respectively.¹³⁸ In a Phase III randomized study of 704 patients with BRAF-mutant metastatic melanoma that allowed patients with stable, treated BM, the combination of a BRAF inhibitor and an inhibitor of mitogen/extracellular signal-regulated kinase (MEK) demonstrated superior progression-free survival and OS compared with a BRAF inhibitor alone.¹³⁹ The efficacy of combination therapy with a BRAF inhibitor and a MEK inhibitor for treatment of BRAF-mutant melanoma BM is now being evaluated in 2 Phase II studies.^140,141 Anti-PD1 checkpoint inhibitors such as nivolumab and pembrolizumab are also under investigation for BM from metastatic melanoma. Preliminary evidence from a small prospective series suggests that pembrolizumab may be active against melanoma metastatic to the brain and may be less toxic than ipilimumab.^142,143 The combination of ipilimumab and nivolumab has shown enhanced activity in systemic metastases of melanoma, albeit at the cost of increased immune-related toxicity.^144,145 Nonetheless, this strategy of combining anti–CTLA-4 and anti-PD1 agents is also undergoing evaluation in several prospective studies for melanoma patients with BM.^113,114

The options for NSCLC patients with BM may include pemetrexed¹⁴⁶ or an EGFR TKI such as gefitinib¹⁴⁷ or erlotinib¹⁴⁸ (for patients with activating EGFR mutations) (Table 3). Response rates of BM to EGFR TKIs at conventional doses in patients with EGFR-mutant NSCLC may be as high as 60%–80%.¹⁴⁹ As previously mentioned, pulsatile dosing of erlotinib administered once weekly can be an effective strategy for management of BM even in patients who have progressed on daily erlotinib.¹¹⁰ For the 2%–7% of NSCLC patients who harbor ALK-gene rearrangements, CNS metastases are a relatively common complication and represent a major barrier to achieving long-lasting disease control.^149,150 In a Phase III randomized trial of the ALK inhibitor crizotinib, ∼35% of patients with ALK-rearranged NSCLC had BM at the time of study entry.¹⁵¹ Based on pooled analysis of 2 large clinical trials of crizotinib in patients with ALK-rearranged advanced NSCLC, 20% of patients without BM at baseline and 70% of patients with known BM developed CNS progression while on crizotinib.⁸ Second-generation ALK inhibitors with increased potency such as alectinib and ceritinib may have better CNS penetration. Preclinical pharmacokinetic studies of alectinib demonstrate improved CNS penetration over crizotinib, with CNS concentrations 63%–94% of serum concentrations.¹⁵² In Phase I studies of alectinib or ceritinib that allowed patients with CNS metastases and no prior brain radiotherapy at study entry, 3 of 4 patients with measurable CNS disease in the alectinib study and 6 of 11 patients with measurable CNS disease in the ceritinib study achieved a complete or partial CNS response.^153,154 In a separate study of alectinib for crizotinib-resistant ALK-rearranged NSCLC that allowed stable, treated, or asymptomatic untreated BM at baseline, the CNS response rate in the 35 patients with measurable CNS disease was 57%.¹⁵⁵ At 12 months, the cumulative CNS progression rate (24.8%) was lower than the cumulative non-CNS progression rate (33.2%) for all patients. Ongoing trials of second-generation ALK inhibitors for BM in ALK-rearranged NSCLC include a Phase II study of ceritinib for ALK-rearranged BM and leptomeningeal disease.¹⁵⁶ The randomized Phase III ALEX trial comparing alectinib versus crizotinib in treatment-naive patients with ALK-rearranged NSCLC allows enrollment of patients with asymptomatic BM and leptomeningeal disease and will measure time to CNS progression.¹⁵⁷ Given the recent approval of anti-PD1 antibodies as second-line therapy of both squamous and nonsquamous NSCLC, further study of these checkpoint inhibitors—alone^158,159 or in combinations with other agents¹⁶⁰—in BM of NSCLC will also be pursued.

Conclusions and Future Directions

The clinical management and understanding of BM has changed substantially in the last 5 years. Several of these changes stem from improvements in systemic therapy, which have led to better systemic control and longer patient OS, associated with increased time at risk for developing BM. Being aware of significant heterogeneity in prognosis and therapeutic options for patients with BM is crucial for appropriate management, with greater attention to developing individual patient treatment plans based on predicted outcomes. The increase in use of SRS alone for many patients with multiple BM is an outgrowth of the desire to employ treatments focused on local control while minimizing cognitive effects associated with WBRT. Complications from BM and their treatment must be considered in comprehensive patient management, especially with greater awareness that the majority of patients do not die from their BM. Development of systemic therapies capable of preventing BM and controlling both intracranial and extracranial disease once BM are diagnosed is paramount. Many clinical trials are under way to establish the efficacy of systemic agents in BM, particularly for breast cancer, lung cancer, and melanoma. Standardization of radiographic response criteria and clinical trial design by the brain metastases working group of RANO will hopefully provide more uniformity in assessment of brain metastases and improve cross-trial comparison. Future directions and challenges also include development of drugs with better CNS penetration, earlier inclusion of BM patients in drug development, as well as increased understanding of the molecular basis of BM and the role of chemoprevention for BM.

Funding

There was no specific funding source for the information presented in this review article. M.P.M.: research funding from Novocure and Cellectar. N.U.L.: research funding from Genentech, GlaxoSmithKline, Array Biopharma, Oncothyreon, Kadmon, Novartis, and Puma.

Conflict of interest statement. E.Q.L.: consultant for Genentech; honoraria from UpToDate. M.P.M.: board of directors for Pharmacyclics; stock in Pharmacyclics; honoraria from Cavion; research funding from Novocure and Cellectar; consultant/advisory role for Elekta, Novartis, and Novocure. B.M.A.: consultant for Abbvie and Precision Health Economics. N.U.L.: research funding from Genentech, GSK, Array Biopharma, Oncothyreon, Kadmon, Novartis, and Puma. P.Y.W.: advisory boards of Cavion, Cortice Biosciences, Foundation Medicine, Genentech, Monteris, Novartis, Novocure, Regeneron, and Vascular Biogenic; speaker bureau for Merck.