Brain metastasis: Opportunities for novel therapies Authors: Zouina Sarfraz, Rupesh Kotecha, Arun Maharaj, Manmeet S Ahluwalia Published: 2026 ABSTRACT AbstractBrain metastases (BM) are a common and clinically significant manifestation of advanced cancer that increasingly influences treatment decisions across solid tumors. Management requires a tailored approach because intracranial disease may behave differently from extracranial disease, reflecting differences in drug penetration, blood-brain barrier biology, and the local immune milieu. This review summarizes the current biologic and clinical framework of BM, including molecular divergence from the primary tumor, the evolution of prognostic models, and the growing role of brain-specific risk stratification in treatment planning. It also reviews contemporary evidence for systemic therapies with intracranial activity across major tumor types, with particularly important advances in oncogene-driven non-small cell lung cancer, human epidermal growth factor receptor 2-positive breast cancer, melanoma, and renal cell carcinoma. In addition, the review examines the integration of systemic therapy with local approaches such as stereotactic radiosurgery, where safety appears to be more agent-specific than timing-specific. Emerging strategies, including focused ultrasound-mediated blood-brain barrier opening and tumor treating fields, may further broaden therapeutic possibilities. The management of BM increasingly depends on a multidisciplinary, biologically informed approach supported by more consistent intracranial endpoints and dedicated central nervous system-focused research.Keywords: Brain metastases, Blood-brain barrier, Targeted therapy, Immunotherapy, Antibody-drug conjugates, Stereotactic radiosurgery Introduction Brain metastases (BM) represent a significant therapeutic challenge of metastatic cancer care in the modern era. While improved systemic therapies have extended survival for patients with solid tumors, the central nervous system (CNS) remains a frequent site of relapse because intracranial disease control often lags behind extracranial response when drug exposure in the brain is inadequate. Many patients now survive long enough for brain metastases to become radiographically or clinically apparent, and those cells can persist behind the blood-brain barrier (BBB) and blood-tumor barrier (BTB). BM are increasingly common across a variety of tumor types and remain a dominant driver of morbidity, causing focal neurologic deficits, seizures, cognitive decline, and functional loss, driving corticosteroid use, reducing quality of life, and driving interruptions in otherwise effective systemic therapy. Approximately 10-40% of patients with solid tumors develop BM, corresponding to an estimated 70,000–400,000 cases annually in the United States, with incidence increasing over time alongside greater MRI utilization and improved systemic control of extracranial disease [1] . Risk is strongly subtype-driven and treatment-era dependent. For example, in patients with human epidermal growth factor receptor 2 (HER2)-amplified breast cancer, cumulative CNS involvement has been reported to be approximately 35–40% over the metastatic course in the trastuzumab era [[2], [3]]. Whereas, in anaplastic lymphoma kinase (ALK)-rearranged non-small cell lung cancer (NSCLC), the brain has historically been a frequent first site of progression on earlier-generation tyrosine kinase inhibitors [4]. These patterns make BM a distinct clinical priority that demands brain-specific thinking rather than extrapolation from systemic response. The objective of this review is to summarize contemporary approaches to risk stratification and patient categorization, describe evidence-based management strategies that integrate systemic and local therapies, and synthesize the available intracranial evidence to support informed clinical decision-making and optimize neurologic and overall outcomes in patients with BM. Brain Metastasis Ecosystem a Neurovascular and Immune Compartmentalization Neurovascular and Immune Compartmentalization Brain metastasis is biologically distinct, as tumor cells must enter a tightly regulated organ and adapt to neurovascular and therapy-driven constraints [5]. After hematogenous spread, cells arrest in the brain microvasculature, extravasate, and often grow along vascular basement membranes [6,7]. Many lesions initially expand via vessel co-option and remodeling rather than angiogenesis, with variable engagement by histology and stage [[8], [9], [10]]. The BBB is heterogeneously altered, forming a heterogeneous BTB with preserved efflux transport that limits and unevenly distributes drug exposure [[11], [12], [13]]. Early hypoxia and perivascular signaling further shape seeding before radiographic detection [6,14,15]. These features define a partially protected, spatially heterogeneous therapeutic compartment in which histology alone is insufficient to guide treatment selection. Systemic responses do not consistently predict intracranial control, reflecting pharmacologic and immunologic compartmentalization. Even after metastasis forms, BBB architecture and efflux transporters limit CNS drug delivery, producing lesion-level variability [[11], [12]]. Immune activity is similarly restricted: T-cell trafficking is limited, microglia and macrophages shape local responses, and corticosteroids suppress immunity [16]. Many lesions are therefore immunologically ‘cold,’ although radiotherapy can modulate the local immune microenvironment [[16], [17], [18]]. These factors help explain discordance between extracranial and intracranial outcomes, though CNS-penetrant therapies and antibody-drug conjugates suggest this gap may narrow in selected settings. bClinical Phenotypes and Sanctuary Dynamics Clinical Phenotypes and Sanctuary Dynamics BM span a spectrum from synchronous to metachronous disease, from limited to widespread burden, and distinct anatomic forms (parenchymal, dural, leptomeningeal). The oligometastatic paradigm reflects a lower disease burden with prognostic relevance [[19], [20], [21]]. However, adding brain lesions to extracranial oligometastases may shift patients to a polymetastatic state and worsen outcomes [22]. Certain primaries, such as melanoma and renal cell carcinoma, are prone to hemorrhagic metastases [23]. As systemic control improves, the CNS increasingly serves as a sanctuary site and locus of relapse due to biologic tropism and limited drug penetration [24,25]. cMolecular Divergence and Prognostic Integration Molecular Divergence and Prognostic Integration Given this heterogeneity, site-specific molecular assessment is increasingly important. BM often diverge from primary tumors through branched evolution, with gain or loss of actionable alterations and receptor discordance, particularly in breast cancer, affecting prognosis and treatment [26,27]. When feasible, intracranial tissue profiling is conducted. In leptomeningeal or CNS-only disease, CSF-derived ctDNA may better capture tumor biology, whereas plasma ctDNA can miss CNS-specific clones [28]. These considerations position molecular profiling as a practical treatment-selection tool and provide the rationale for mapping primary tumor distributions alongside key brain-tropic molecular alterations in Fig. I. Distribution of primary tumor origins and key molecular features associated with brain metastases.Incidence ranges are approximate and may vary by disease stage, molecular subtype, screening practices, and treatment era. Molecular features shown represent commonly reported or clinically actionable alterations associated with CNS tropism, therapeutic targeting, or prognostic relevance and are not exhaustive. Created in BioRender. Maharaj, A. (2026)https://BioRender.com/o4rkchl. Prognostication in BM has progressed from clinical-only schemas (i.e., recursive partitioning analysis [RPA], then diagnosis-specific graded prognostic assessment [DS-GPA]) to molecularly informed diagnosis-specific models that incorporate tumor biology, reflecting that survival is no longer explained by Karnofsky performance status (KPS) and lesion count alone [29,30]. In lung cancer, Lung-molGPA introduced actionable drivers (EGFR and ALK) alongside age, KPS, extracranial disease, and number of BM [31], and the most recent update further incorporates programmed death-ligand 1 (PD-L1) status to align prognosis with modern immunotherapy-era outcomes [32]. In breast cancer, Breast-GPA and its contemporary update integrate biologic subtype (receptor status) together with age, KPS, extracranial metastases, and intracranial burden, acknowledging differential trajectories across HER2-positive, hormone receptor-positive, and triple-negative disease [33,34]. In melanoma, the melanoma-molGPA incorporates genomic markers (e.g., BRAF, NRAS, c-KIT) with core clinical variables, and the 2025 update adds systemic-therapy context (including immunotherapy) and lactate dehydrogenase (LDH) to reflect current treatment-era survivorship patterns [35,36]. In RCC, the updated Renal GPA expands beyond KPS and BM number to include extracranial metastases and hemoglobin, consistent with broader risk biology in the targeted-therapy era [37,38]. Molecularly informed GPA frameworks integrate intracranial burden with tumor biology and treatment context, improving prognostic accuracy beyond anatomic metrics. They support counseling, trial stratification, and treatment selection by aligning survival estimates with CNS-specific therapeutic and disease factors. Tumor-Specific Novel Systemic Therapies with Intracranial Data Interpretation of intracranial efficacy for tumor-specific systemic therapies remains limited by heterogeneous trial designs. Studies often combine treated and untreated BM without stratified reporting, incompletely document prior radiotherapy, and use inconsistent eligibility definitions for “active,” “stable,” or “progressive” disease, limiting attribution of response to systemic therapy alone [39,40]. CNS endpoints are frequently secondary or post hoc, particularly in lung cancer, and response assessment varies across RECIST, RANO-BM, CT-based follow-up, and modified lesion thresholds [[41], [42], [43]]. Outcomes restricted to measurable lesions may overestimate intracranial objective response rate (iORR) relative to broader CNS cohorts [44,45], while symptomatic patients and corticosteroid-dependent patients remain underrepresented [46]. Additional variability in lesion size, treatment line, and molecular evolution after prior therapy further limits cross-trial comparisons [47]. Thus, intracranial response rates should be interpreted cautiously when used to support deferral of upfront radiation therapy. The genomic distribution of actionable alterations in advanced non–small-cell lung adenocarcinoma is shown in Fig. II. Intracranial activity varies substantially across systemic therapy classes, as summarized in Fig. III. Incidence of oncogenic driver alterations in advanced non–small-cell lung adenocarcinoma. Percentages reproduce published estimates and may not sum to 100%. Adapted from Friedlaender et al. [48] Class-level intracranial activity of selected CNS-active systemic therapies. Representative iORR and median iPFS estimates from selected studies discussed in this review. Bubble size reflects the intracranial response-evaluable population contributing to iORR. Values are not intended for formal cross-trial comparison. Non-small cell lung cancer i ICI and Triplet Therapy for Driver-Negative NSCLC With Brain Metastases ICI and Triplet Therapy for Driver-Negative NSCLC With Brain Metastases Evidence supporting immune checkpoint inhibitor (ICI)-based therapy in driver-negative NSCLC with BM has expanded, although intracranial control remains less durable than in oncogene-driven disease. Evidence of CNS activity with PD-1 blockade alone was demonstrated in a phase II prospective pembrolizumab cohort restricted to asymptomatic, steroid-independent BM. Among patients with PD-L1-positive tumors, the iORR was 29.7%, with no responses in PD-L1-negative disease and a median intracranial progression-free survival (iPFS) of 2.3 months (95% CI, 1.9-not reached), establishing that single-agent PD-1 inhibition can induce CNS responses but in a highly-selected population with limited durability [49]. Subsequent phase II studies incorporating chemotherapy have reported higher and more consistent intracranial activity in treatment-naïve disease. In ATEZO-BRAIN, first-line atezolizumab plus carboplatin-pemetrexed achieved an iORR of 42.7% and median iPFS of 6.9 months, despite inclusion of patients requiring baseline corticosteroids and without restriction to PD-L1-high tumors [50]. The CAP-BRAIN trial evaluating camrelizumab with platinum-pemetrexed reported an iORR of 57.5% with a median iPFS of 8.4 months (95% CI, 6.2-not reached), with intracranial and extracranial responses largely concordant [51]. Similarly, a multicenter phase II study of tislelizumab combined with chemotherapy that permitted symptomatic or recently irradiated BM demonstrated an iORR of 46.7% and a 12-month iPFS rate of 55.8%, reinforcing reproducible intracranial activity across chemo-ICI platforms in clinically heterogeneous populations [52]. Triplet strategies incorporating anti-angiogenic therapy provide additional signals of intracranial benefit. In the phase III IMpower150 trial, atezolizumab plus bevacizumab and chemotherapy improved survival compared with bevacizumab-chemotherapy in the overall study population [53]. In an exploratory CNS analysis, atezolizumab, bevacizumab, carboplatin, and paclitaxel (ABCP) was associated with a numerically longer time to development of new brain metastases than bevacizumab, carboplatin, and paclitaxel (BCP) (HR, 0.68; 95% CI, 0.39–1.19), although the confidence interval crossed unity [54]. Dual-checkpoint blockade combined with chemotherapy also demonstrates CNS activity. In the randomized CheckMate 9LA trial, among patients with previously locally treated baseline BM, nivolumab plus ipilimumab with two cycles of platinum-doublet chemotherapy improved intracranial PFS versus chemotherapy alone (HR 0.40; 95% CI, 0.24–0.65) [55]. CNS endpoints were exploratory but provide randomized evidence that immunotherapy-based triplets can achieve meaningful intracranial disease control in metastatic NSCLC without sensitizing EGFR or ALK alterations [55]. Across prospective studies, single-agent PD-1/PD-L1 blockade produces intracranial responses in approximately one-third of selected PD-L1-positive patients, whereas chemo-ICI doublets yield response rates in the 40–60% range and median iPFS of approximately 7–8 months; randomized triplet regimens have shown improved intracranial progression outcomes, although dedicated intracranial response rates were not reported. While the magnitude and durability of intracranial benefit remain inferior to targeted therapies in patients with oncogene-driven disease, these regimens represent the principal systemic platform for patients with driver-negative NSCLC and BM, particularly when integrated with chemotherapy and, in selected cases, anti-angiogenic therapy or radiotherapy. iiEGFR-Mutated NSCLC With BM EGFR-Mutated NSCLC With BM aOsimertinib Osimertinib Across literature specifically reporting outcomes in patients with baseline BM, osimertinib has demonstrated reproducible intracranial activity that largely parallels systemic disease control, although the magnitude and durability of benefit vary by treatment line and cohort enrichment. In the randomized phase III FLAURA trial, 61 patients with baseline BM treated with first-line osimertinib achieved an iORR of 66% (40/61). CNS progression occurred in 12/61 (20%), and new intracranial lesions developed in 7/61 (12%). Median iPFS was not reached, with a lower bound of 16.5 months on the 95% confidence interval, reflecting durable intracranial control in a population limited to stable or asymptomatic disease at baseline [45]. The FLAURA trial's results led to the U.S. FDA's expanded first-line approval of osimertinib on April 18, 2018, for patients with EGFR-mutated metastatic NSCLC, which represents a major advancement in the integration of CNS-active targeted therapy for this population [56]. In the second-line, T790M-positive setting evaluated in AURA3, intracranial efficacy remained evident but more modest, with an iORR of 40% (30/75) and median iPFS of 11.7 months (95% CI, 10.0-not calculable). CNS progression was observed in 12/75 (16%) and new lesions in 8/75 (11%), suggesting preserved but shorter-duration intracranial control in a more heavily pretreated population [44]. Prospective CNS-enriched cohorts offer additional granularity regarding response durability and patterns of progression. In APOLLO, which required measurable CNS disease and permitted prior local therapy, the iORR reached 68.8% (22/32 evaluable), with CNS progression in 11/38 (29%); notably, CNS-only progression accounted for 7/38 (18%), highlighting that intracranial progression remains a clinically relevant event even with high response rates. iPFS was not separately reported, limiting cross-study comparison of durability despite strong response rates [57]. The radiotherapy-naïve OCEAN T790M cohort reported a BM response rate of 66.7% (26/39; 90% CI, 54.3–79.1) and a median BM-related PFS of 25.2 months (95% CI, 7.0–34.5), suggesting prolonged intracranial control in the absence of prior CNS radiation, although CNS progression counts were not separately reported [58]. The subsequent first-line OCEAN cohort demonstrated an iORR of 76.9% (20/26) and median iPFS of 22.0 months (95% CI, 9.7-not evaluable). CNS progression occurred in 7/26 (26.9%) by BM-PFS criteria; however, new-lesion incidence was not reported, and the use of BM-PFS events as a surrogate for intracranial progression should be interpreted cautiously when comparing with randomized datasets [59]. Randomized data incorporating combination therapy further establish the relationship between systemic intensification and intracranial control. In FLAURA2, among 222 patients (CNS full analysis set) with baseline BM, intracranial response rates were 69% (72/104) with osimertinib alone and 73% (86/118) with osimertinib plus platinum–pemetrexed [60,61]. Median iPFS was 27.6 months (95% CI, 22.1-not calculable) with monotherapy and 30.2 months (95% CI, 28.4-not calculable) with combination therapy. CNS progression occurred in 31/104 (30%) and 28/118 (24%), respectively, with new CNS lesions reported in 12/104 (12%) and 9/118 (8%). While combination therapy numerically prolonged iPFS and reduced CNS progression rates, the incremental intracranial benefit appeared modest relative to monotherapy, and both arms demonstrated prolonged CNS control exceeding two years in many patients [60,61]. Beyond the pivotal trials, prospective reports add context to the intracranial activity of osimertinib and the constraints of response-evaluable disease and imaging schedules. In a nonrandomized phase II study enrolling only patients with asymptomatic BM (NCT02736513), iORR was 84.2% (16/19) in treatment-naïve patients, 66.7% (12/18) in T790M-positive pretreated patients, and 50% (5/10) in T790M-negative pretreated patients [62]; corresponding intracranial disease control rates were 94.7%, 94.4%, and 80%, respectively. Median iPFS was 11.8, 7.6, and 6.3 months across these strata, and dose escalation to 160 mg for isolated intracranial progression produced an intracranial response rate of 54%, disease control rate of 72.7%, and median iPFS of 4.3 months, supporting a CNS-salvage strategy when local therapy can be deferred [62]. In a cohort enriched for prior whole-brain radiotherapy exposure, intracranial ORR was 9.5% and intracranial disease control rate 81.0%, illustrating how heavy pretreatment and limited measurable targets can suppress RECIST-defined intracranial response despite clinical CNS benefit [63]. Strategy-based prospective data from the APPLE program further indicate that up-front osimertinib reduces the risk of intracranial progression compared with sequential EGFR-TKI strategies, with longer median brain PFS observed with up-front osimertinib (34.3 months) compared with sequential approaches (22.3 months), and a hazard ratio of 0.54 (90% CI, 0.34–0.86) for brain progression [64]. bAfatinib Afatinib Evidence for afatinib in patients with EGFR-mutated NSCLC and BM derives largely from early randomized trials that permitted baseline BM and subsequent retrospective cohorts reporting CNS outcomes. In the LUX-Lung program (LUX-Lung 2, 3, and 7), patients with treated or stable BM were included, and afatinib demonstrated systemic activity comparable to that observed in patients without BM [[65], [66], [67]]. In a pooled analysis of patients with baseline BM from LUX-Lung 3 and 6, median PFS was 8.2 months with afatinib versus 5.4 months with platinum-based chemotherapy, with ORR of 72.9% and 24.2%, respectively [68]. Retrospective series suggest moderate intracranial activity. Reported iORRs range from nearly 49% to 81%, with median iPFS around 12 months in selected cohorts, although CNS progression remains common and frequently represents the first site of failure [[69], [70], [71]]. Afatinib received U.S. FDA approval in July 2013 for first-line metastatic EGFR-mutant NSCLC [[72], [73]]. However, with the emergence of third-generation EGFR inhibitors with improved CNS penetration, its role in contemporary practice is limited and generally considered secondary. cAmivantamab plus Lazertinib Amivantamab plus Lazertinib The CNS signal with amivantamab plus lazertinib must be interpreted in the context of who was enrolled and when in the disease trajectory the regimen was deployed. In CHRYSALIS-2 Cohort A, patients were heavily pretreated, post-osimertinib and post-platinum, with 66/162 harboring baseline BM. Among the 66 patients with brain metastases identified on baseline imaging, 29 had untreated BM and 37 had treated BM. Of the 60 patients with at least one postbaseline brain scan, 17 (28%) demonstrated an intracranial response; among the 26 evaluable patients with untreated BM, 7 (27%) demonstrated an intracranial response [74]. iORR was 28% (17/60 evaluable), with a similar signal in untreated BM (27%), which is a meaningful observation in a refractory population where systemic median PFS was 4.5 months. The activity of dual EGFR-MET extracellular blockade layered onto a CNS-penetrant TKI is modest but biologically credible. This is salvage CNS control, not dominance. The dedicated phase II CNS study (NCT04965090) deliberately enrolled patients with active or progressive CNS disease, including measurable BM (n = 20); iORR reached 40% with median iPFS of 6.0 months (95% CI 3.8-NR) [75]. These were not stable screening lesions, but rather they were biologically progressing CNS tumors. This study establishes proof of intracranial activity beyond radiographic stability and confirms feasibility of prospective CNS-specific assessment in this molecular subset. The first-line setting changes the narrative. In MARIPOSA, asymptomatic or treated/stable BM were permitted [76,77]. With longer follow-up, PFS favored amivantamab plus lazertinib both in patients with and without baseline BM, with hazard ratio 0.69 in the BM subgroup. Median PFS among patients with baseline BM was 18.3 months with amivantamab plus lazertinib versus 13.0 months with osimertinib. Participants treated with amivantamab–lazertinib showed early evidence of a reduced risk of CNS progression and sustained CNS control compared with osimertinib. However, MARIPOSA enrolled predominantly stable or asymptomatic BM, and intracranial response rates were high in both arms in measurable disease (as reflected in the CNS analysis). The distinction is durability and delayed CNS progression rather than dramatic superiority in iORR. In MARIPOSA-2, post-osimertinib, amivantamab-containing regimens improved median iPFS versus chemotherapy (12.5–12.8 vs 8.3 months) [78]. This confirms CNS contribution of EGFR-MET targeting even after third-generation TKI resistance, though attribution between antibody effect and chemotherapy synergy remains complex. A consistent theme emerges. In refractory CNS disease, amivantamab plus lazertinib produces iORR up to 40% with finite durability. In first-line disease, the combination reduces risk of CNS progression and prolongs intracranial control relative to osimertinib. In post-osimertinib progression, amivantamab-containing regimens extend iPFS versus chemotherapy. On August 19, 2024, the U.S. FDA approved lazertinib in combination with amivantamab for the first-line treatment of locally advanced or metastatic EGFR exon 19 deletion or exon 21 L858R-mutant NSCLC, based on the MARIPOSA phase III trial demonstrating significantly improved PFS over osimertinib [79]. What distinguishes amivantamab plus lazertinib from osimertinib is not simply penetration – it is mechanistic breadth. The bispecific antibody addresses extracellular EGFR and MET-driven escape while lazertinib maintains catalytic inhibition with CNS penetration. The CNS efficacy appears to track with this dual targeting, particularly in MET-enriched resistance states that are disproportionately represented in BM. The question is not whether amivantamab plus lazertinib is “CNS active.” It is whether it meaningfully alters intracranial natural history across lines of therapy. The data suggest it does; most convincingly in delaying CNS progression in the first line setting and maintaining measurable intracranial activity after osimertinib failure. iiiALK-Rearranged NSCLC with BM ALK-Rearranged NSCLC with BM BM are present in approximately one-quarter to one-third of patients with ALK-rearranged NSCLC at diagnosis and remain a dominant site of therapeutic failure [80,81]. CNS control has therefore become a defining feature of drug selection in this population. Crizotinib (first-generation) demonstrated intracranial activity but incomplete CNS protection. In PROFILE 1014 (NCT01154140), among patients with treated, stable baseline BM, intracranial time to tumor progression was 15.7 months with crizotinib versus 12.5 months with chemotherapy (HR 0.45; 95% CI 0.19–1.07), with intracranial progression events in 9/39 (23%) with crizotinib versus 12/40 (30%) with chemotherapy [82]. Pooled analyses of PROFILE 1005/1007 showed modest CNS activity, with iORR of 4/22 (18%) in untreated BM and 6/18 (33%) in previously irradiated BM [83]. These findings established proof of intracranial activity highlighting the CNS as a persistent sanctuary site under first-generation inhibition. Second-generation ALK inhibitors redefined intracranial efficacy in the first-line setting. In ALEX (NCT02075840), alectinib prolonged PFS in patients with baseline BM (25.4 months vs 7.4 months with crizotinib), reflecting durable CNS control [84]. In ALESIA (NCT02838420), iORR among patients with baseline BM was 32/44 (73%) with alectinib versus 5/23 (22%) with crizotinib, and CNS progression without prior systemic progression occurred in 12/125 (10%) with alectinib versus 22/62 (36%) with crizotinib [85]. Brigatinib demonstrated similar superiority in ALTA-1L (NCT02737501), with iORR by BIRC favoring brigatinib across subgroups (Asian: 13/21 [62%] vs 6/18 [33%]; non-Asian: 18/26 [69%] vs 1/31 [3%]) [86]. Ensartinib in eXalt3 (NCT02767804) improved iORR in measurable BM (7/11 [63.6%] vs 4/19 [21.1%]) [87]. Ceritinib also demonstrated CNS activity. In ASCEND-4 (NCT01828099), iORR among measurable baseline BM was 16/22 (72.7%) with ceritinib and 6/22 (27.3%) with chemotherapy [88], and in ASCEND-5 (post-crizotinib), iORR was 6/17 (35%) with ceritinib and 1/20 (5%) with chemotherapy [89]. Second-generation inhibitors shifted intracranial control from partial activity to consistent, durable CNS disease management. Lorlatinib (third generation) was engineered for enhanced CNS penetration and demonstrated marked intracranial activity in CROWN (NCT03052608). In the 5-year analysis, among patients with baseline BM, median PFS was not reached with lorlatinib versus 6.0 months with crizotinib (HR, 0.08; 95% CI, 0.04–0.19), with a 5-year PFS rate of 53% with lorlatinib. Median time to intracranial progression was not reached versus 7.2 months, respectively (HR, 0.03; 95% CI, 0.01–0.13). In the overall population, median time to intracranial progression was not reached with lorlatinib versus 16.4 months with crizotinib (HR, 0.06; 95% CI, 0.03–0.12), with 5-year freedom from intracranial progression of 92% versus 21% [90]. In previously treated disease (NCT03909971), lorlatinib retained intracranial activity but responses were attenuated after prior next-generation ALK TKIs (iORR 29/36 [80.6%] post-crizotinib vs 10/21 [47.6%] after ≥1 non-crizotinib ALK TKI; iPFS 4.8 months in the latter cohort) [91]. Combination strategies remain exploratory. Alectinib plus bevacizumab produced high intracranial response in small cohorts: iORR 7/9 (77.8%) in a phase I/II study (NCT02521051) [92] and 6/7 (85.7%) in ALEK-B (NCT03779191) [93], with a reported 36-month iPFS rate of 87.8%. These data are hypothesis-generating given limited sample size and single-arm design. Across successive generations of ALK tyrosine kinase inhibitors, intracranial control has progressed from modest activity with first-generation agents to durable suppression with second-generation therapies and marked reduction in intracranial progression risk with third-generation inhibition. The CNS is no longer a predictable early site of failure when contemporary ALK TKIs are used upfront. ivKRAS G12C Inhibition KRAS G12C Inhibition BM are frequent in KRAS G12C-mutant NSCLC and have historically contributed to early systemic and intracranial progression. The available data indicate that KRAS G12C inhibitors possess clinically meaningful CNS activity, although the depth and durability of control vary by disease state at treatment initiation and by agent. In randomized settings restricted to treated, stable BM, both adagrasib and sotorasib demonstrate intracranial benefit relative to docetaxel. In the CNS analysis of CodeBreaK 200, sotorasib extended median iPFS to 9.6 months compared with 4.5 months with chemotherapy (HR: 0.43) [94]. In KRYSTAL-12, intracranial responses were observed with adagrasib in patients with baseline BM (24% compared with 11% on docetaxel) [95]. These findings establish activity beyond cytotoxic therapy, primarily in cohorts with locally controlled CNS disease. Prospective evidence in untreated BM is more limited. In the untreated CNS cohort of KRYSTAL-1, adagrasib achieved an intracranial response rate of 42% with a median iPFS of 5.4 months [96]. Pharmacokinetic data demonstrate measurable CSF penetration, supporting observed radiographic regression [97]. Comparable untreated-CNS datasets for sotorasib remain less defined [98], and additional prospective data with IBI351 excluded active lesions [99]. Real-world outcomes further contextualize these findings. In the IFCT-2102 Lung KG12Ci cohort, which included patients with progressing BM at initiation, cerebral response was 20.1% and median real-world PFS was 3.5 months [100]. KRAS G12C inhibitors demonstrate more durable control in treated, stable metastases, with more heterogeneous outcomes in untreated disease. Intracranial responses may be achievable, and progression can be delayed relative to chemotherapy; however, the duration of CNS control remains intermediate when viewed against established CNS-penetrant therapies in other oncogene-driven NSCLC subsets. Emerging combination strategies aim to overcome adaptive resistance to KRAS G12C inhibition, which commonly occurs through reactivation of upstream receptor tyrosine kinase signaling and MAPK pathway output. One approach combines KRAS inhibition with EGFR blockade; in the phase II KROCUS study evaluating fulzerasib plus cetuximab, 40.7% of enrolled patients had baseline BM, although intracranial-specific outcomes have not yet been reported [101]. Additional investigational approaches include SOS1 inhibition to suppress upstream KRAS pathway reactivation [102]. vRET-, ROS1-, NTRK-, and MET-Directed Therapies RET-, ROS1-, NTRK-, and MET-Directed Therapies In RET-rearranged NSCLC with BM, selective RET inhibition shows consistent intracranial activity. In LIBRETTO-001, intracranial responses occurred in 91% (10/11) of patients with measurable CNS disease (in the initial NEJM report); dedicated analyses reported iORR 82% (n=22), and in the final LIBRETTO-001 analysis iORR 85% among measurable lesions [[103], [104], [105]]. LIBRETTO-431 compared first-line selpercatinib with chemo ± pembrolizumab, allowed treated or stable BM, and showed reduced CNS progression risk (HR 0.28) [106]. In LIBRETTO-321, intracranial responses occurred in 4/5 patients with measurable CNS lesions [107]. Pralsetinib in ARROW demonstrated intracranial responses in 7/10 patients with measurable BM [108]. In ROS1-rearranged NSCLC, BM are a frequent site of progression, and prospective datasets now delineate the intracranial activity of contemporary ROS1 TKIs. In TRUST-I/TRUST-II, taletrectinib produced iORR 76.5% (13/17) in TKI-naïve and 65.6% (21/32) in pretreated patients with measurable BM [109]. In TRIDENT-1, repotrectinib achieved intracranial responses in 8/9 (89%) TKI-naïve and 5/13 (38%) after one prior ROS1 TKI without chemotherapy [110]. Earlier datasets contextualize CNS risk. In EUCROSS, three patients had measurable CNS target lesions, all of whom achieved stable disease as their best intracranial response. Among 17 patients undergoing serial brain MRI assessment, median iPFS was 32.2 months, with isolated CNS progression in 24% of patients [111]. In BAROSSA, brigatinib showed iORR in 0% (0/3) among TKI-naïve patients and 4/6 (66.7%) after crizotinib [112]. Comparable intracranial responses have also been reported with other ROS1 inhibitors. Entrectinib (ALKA-372-001/STARTRK-1/2) had intracranial ORR of 79.2% in measurable CNS disease [113]; lorlatinib had an iORR of 64% (TKI-naïve) and 50% (post-crizotinib) [114], with a dedicated post-crizotinib CNS-only study reporting iORR in 87% [115]; and ceritinib intracranial disease control in 5/8 (63%) patients with BM [116]. In an integrated STARTRK analysis of NTRK fusion-positive NSCLC, entrectinib achieved an iORR of 9/14 (64.3%) with median iPFS 32.7 months in patients with baseline BM [117]. In METex14-altered disease, gumarontinib produced investigator-assessed intracranial partial responses in 5/5 target CNS lesions [118]. Capmatinib in GEOMETRY mono-1 demonstrated intracranial responses in 16/28 (57%) patients with baseline BM [119]. Tepotinib across VISION BM analyses showed iORR 10/15 (66.7%) in RANO-BM-evaluable disease and 12/15 (80%) in a pooled evaluable cohort, with measurable-lesion responses in 5/7 (71.4%) [[120], [121], [122]]. Across RET-, ROS1-, NTRK-, and MET-altered NSCLC, selective TKIs with established CNS penetration (selpercatinib and pralsetinib; repotrectinib and taletrectinib; entrectinib; capmatinib, tepotinib, and gumarontinib) demonstrate high intracranial response rates in measurable BM, with most reported estimates falling between approximately 60% and 90%. These estimates derive from small cohorts and vary widely, and responses are generally attenuated after prior TKI exposure. In these rare molecular subsets, effective systemic targeting translates into reproducible intracranial responses, although durability and resistance remain incompletely defined in small BM cohorts. From a clinical perspective, NSCLC has mature intracranial systemic therapy data, particularly in molecularly defined disease. CNS-active targeted therapies can provide meaningful intracranial control in selected oncogene-driven subsets, whereas chemo-immunotherapy remains a key systemic approach for driver-negative disease, with treatment decisions still shaped by symptoms, steroid requirement, prior radiotherapy, and urgency of local control. HER2-positive breast cancer Tyrosine kinase inhibitors Early HER2 TKIs showed intracranial activity but with limited durability. In LANDSCAPE, lapatinib plus capecitabine achieved an iORR of 65.9% with median iPFS 5.5 months [123], although pooled analyses across studies reported a lower CNS-ORR of about 29% [124]. Neratinib-based regimens demonstrated similar activity. In TBCRC 022, neratinib plus capecitabine produced a CNS-ORR of 49% in TKI-naïve patients (33% after prior lapatinib), with median PFS of 5.5 months [125]. In NEfERT-T, neratinib reduced CNS recurrences (8.3% vs 17.3%) [126], and in NALA, neratinib plus capecitabine lowered the need for CNS intervention compared with lapatinib (22.8% vs 29.2%) [127]. Overall, these studies established proof of intracranial activity for earlier HER2 TKIs but with modest durability compared with newer agents. The intracranial activity of tucatinib was established in the pivotal HER2CLIMB trial, which enrolled patients with previously treated HER2-positive metastatic breast cancer and included a prespecified cohort with BM (291/612 patients). Patients were randomized to tucatinib + trastuzumab + capecitabine or placebo plus trastuzumab and capecitabine, and both active (untreated or progressing) and stable BM were permitted [[128], [129]]. Among patients with BM, median iPFS was 9.9 months (95% CI 8.4–11.7) with tucatinib versus 4.2 months (95% CI 3.6–5.7) with control therapy. In patients with active measurable BM, confirmed iORR was 47.3% (95% CI 33.7–61.2) with tucatinib compared with 20.0% (95% CI 5.7–43.7) in the control arm. Updated analyses demonstrated consistent benefit across subgroups, including active BMs (iPFS 9.6 vs 4.0 months) and stable BMs (13.9 vs 5.6 months), with longer intracranial duration of response (8.6 vs 3.0 months). Tucatinib was also associated with a reduced risk of developing new brain lesions as the first site of progression (HR 0.55) [[129], [130]]. Additional data with pyrotinib provide supportive evidence. In the phase II PERMEATE trial, 78 patients with HER2-positive metastatic breast cancer and active BM received pyrotinib plus capecitabine [131]. Patients were stratified as RT-naïve (n = 59) or progressing after radiotherapy (n = 19). Intracranial ORR was 74.6% (44/59) in RT-naïve patients and 42.1% (8/19) after prior radiotherapy, with median iPFS of 13.6 months (95% CI 9.0–15.8) and 5.7 months (95% CI 3.4–11.5), respectively [131]. In the randomized phase III PHENIX trial,[132] 31 of 279 enrolled patients had baseline BM. Among the subgroup with untreated baseline BM, progressive BM occurred in 73.3% receiving pyrotinib versus 87.5% with placebo, with median time to brain progression of 168 vs 127 days, respectively. Antibody-drug conjugates In the phase IIIb KAMILLA study of ado-trastuzumab emtansine (T-DM1), 398 patients had baseline BM. Among 126 patients with measurable BM, ≥30% intracranial tumor shrinkage occurred in 42.9% (54/126), including 49.3% (33/67) of patients without prior brain radiotherapy [133]. Following the experience with T-DM1, more contemporary HER2-directed ADC data in breast cancer BM are largely derived from trastuzumab deruxtecan (T-DXd). In the DESTINY-Breast01 subgroup analysis of patients with previously treated, stable BM, intracranial responses were observed in 47.1% of patients with measurable disease, with intracranial progression reported in 8.3%, supporting activity in a heavily pretreated population with prior local therapy [134]. Prospective and randomized data further characterize this signal. In DESTINY-Breast03, among patients with clinically inactive/asymptomatic baseline BM, intracranial ORR was 65.7% with T-DXd versus 34.3% with T-DM1; intracranial progressive disease as best response occurred in 0/35 versus 7/35 (20.0%), respectively. [135] In the phase 3b/4 DESTINY-Breast12 study [136], which included both stable and active BM, the 12-month iPFS was 58.9%, and intracranial responses were observed across stable and active cohorts, including untreated lesions. Single-arm prospective studies complement these findings. In TUXEDO-1 [137], conducted in patients with active BM without immediate indication for local therapy, intracranial response was 73.3%, including complete responses. In DEBBRAH, intracranial responses were reported in both untreated and post-local-therapy active metastases, with an overall iORR of 41.7% across HER2-low active cohorts [138]. Additional HER2-directed agents have emerging or limited data relevant to patients with BM, including disitamab vedotin (RC48), for which case-level evidence has been reported [139], and ARX788, which has prospective BM-specific data [140]. Trastuzumab duocarmazine [141] and HER2-bispecific antibodies such as zanidatamab [142] have demonstrated activity in advanced HER2-positive or HER2-expressing disease, but dedicated intracranial efficacy data remain limited. In HER2-positive breast cancer, tucatinib-based therapy and T-DXd have meaningfully broadened treatment options for patients with BM, including selected patients with active intracranial disease. In practice, these data support multidisciplinary sequencing rather than uniform deferral of local therapy. Melanoma The randomized phase 2 Anti-PD-1 Brain Collaboration (ABC; NCT02374242) trial established the intracranial activity of combination nivolumab plus ipilimumab in patients with active, immunotherapy-naïve melanoma BM (MBM). Eligible patients had at least one measurable intracranial lesion 5–40 mm on gadolinium-enhanced MRI. Among patients with asymptomatic BM and no prior brain-directed therapy, iORR was 16/35 (46%) in the initial 2018 analysis and 18/35 (51%) with extended follow-up, compared with 5/25 (20%) with nivolumab alone. In contrast, among patients with prior local brain therapy failure, neurological symptoms, and/or leptomeningeal disease, intracranial response was 1/16 (6%) with nivolumab monotherapy. Median intracranial progression-free survival was not reached (95% CI: 2.9-NR) with the combination in untreated asymptomatic disease, versus 2.5 months (1.7–2.8) with nivolumab alone and 2.3 months (1.4–4.3) in previously treated or symptomatic patients. With 7-year follow-up, intracranial progression-free survival rates were 42% with combination therapy, 15% with nivolumab alone, and 6% in the symptomatic/previously treated population [143,144]. Targeted therapy has also demonstrated intracranial activity in BRAFV600-mutant melanoma. In the phase II COMBI-MB trial, dabrafenib plus trametinib achieved intracranial responses in 44/76 patients (58%) with untreated asymptomatic BM; however, durability was limited, with a median iPFS of 5.6 months and median overall survival of 10.8 months, contrasting with the more sustained intracranial control observed with immune checkpoint blockade [145]. CheckMate 204 (NCT02320058) prospectively evaluated nivolumab plus ipilimumab in 119 patients with measurable MBM 0.5–3.0 cm, requiring at least one non-irradiated lesion [46]. Among 101 patients who were asymptomatic and not receiving corticosteroids at baseline, iORR was 53.5%, with a 36-month iPFS of 54.1%; median iPFS was not reached. Among 18 patients with neurological symptoms, who were permitted low-dose dexamethasone, iORR was 16.7%, median iPFS was 1.2 months, and the 36-month iPFS rate was 18.9%. Prior stereotactic radiotherapy was permitted provided at least one unirradiated lesion remained (9/101 [8.9%] in asymptomatic patients; 3/18 [16.7%] in symptomatic patients) [46]. These results closely parallel the ABC experience, demonstrating durable intracranial control with dual checkpoint blockade in untreated, asymptomatic disease and substantially reduced activity in symptomatic or steroid-requiring patients. Additional randomized data from NIBIT-M2 (NCT02460068) in untreated, asymptomatic MBM (5–20 mm) showed an iORR of 12/27 (44.4%) with nivolumab plus ipilimumab, median iPFS of 8.7 months, and intracranial progression in 17/27 (63%) [146]. In contrast, the phase 2 TRICOTEL study (NCT03625141) evaluated atezolizumab plus vemurafenib and cobimetinib in 65 patients with BRAFV600-mutant MBM ≥5 mm, including 26/65 (40%) with symptomatic CNS disease. Intracranial objective response by independent review was 42% (95% CI: 29–54), median iPFS was 5.3 months (95% CI: 3.8–7.2), and intracranial progressive disease or death occurred in 51/65 (78%) [147]. Overall, these data show that upfront nivolumab plus ipilimumab provides durable intracranial benefit in untreated, asymptomatic MBM, whereas outcomes are markedly attenuated in symptomatic or steroid-exposed populations; triplet targeted-immunotherapy regimens demonstrate meaningful intracranial responses in BRAFV600-mutant disease, including symptomatic patients, but with shorter median iPFS. In melanoma, dual immune checkpoint blockade provides clinically meaningful intracranial activity, especially in asymptomatic patients not requiring corticosteroids. BRAFV600-targeted therapy remains important when rapid disease control is needed, although durability may be more limited. Renal cell carcinoma Literature on dual checkpoint blockade in renal cell carcinoma brain metastases remains limited but instructive. In the BM cohort of CheckMate 920, nivolumab plus ipilimumab in previously untreated advanced RCC with asymptomatic, untreated BM (N = 28) achieved a systemic ORR of 32%, median PFS 9.0 months, and intracranial progression in 25% of patients, with no new safety signals [148]. By contrast, the dedicated phase II GETUG-AFU 26 NIVOREN study of nivolumab monotherapy after VEGFR therapy demonstrated limited intracranial activity in untreated BM (N = 39), with an iORR of 12% and median iPFS 2.7 months; 77% experienced intracranial progression, nearly half with new lesions [149]. These data suggest that PD-1 monotherapy has limited intrinsic activity in untreated CNS disease, reinforcing the importance of timely and effective local control in this population. Real-world experience with ipilimumab plus nivolumab further supports this integration paradigm. In a multicenter retrospective cohort (N = 19), where 95% received prior CNS-directed therapy, median iPFS reached 14.7 months (95% CI: 7.2-NR) [150]. The signal is consistent; dual ICI appears clinically active in selected, largely pretreated or locally controlled RCC-BM, whereas untreated intracranial disease remains a high-risk subset in which focal therapy should be prioritized before or alongside systemic immunotherapy. Cabozantinib has shown clinically meaningful activity in RCC BM across prospective and real-world datasets. In the phase II CABRAMET trial, which enrolled patients with at least one non-locally pretreated measurable BM, brain partial responses occurred in 61.5% (16/26), the 6-month BM progression-free rate was 56%, and median BM-PFS was 10.7 months (95% CI, 5.4–NR), with median OS of 15.0 months (95% CI, 9.3–35.0) [151]. In the largest retrospective international cohort (n = 88), iORR was 55% (95% CI: 36–73) in patients with progressing BM without recent local therapy and 47% (95% CI: 33–61) in those receiving concomitant or prior brain-directed treatment, with 6-month intracranial progression rates of 25% and 15%, respectively, and no treatment-related neurologic adverse events reported [152]. Real-world data from the Italian managed access program (n = 12, with 9 evaluable) similarly showed a brain-specific response rate of [5/9] 55% (CR [2/9] 22%, PR [3/9] 33%), median PFS 5.8 months, median OS 8.8 months, and no signal of neurologic toxicity [153]. Across these datasets, cabozantinib has produced intracranial response rates of approximately 47–55% in retrospective and real-world cohorts and a 61.5% brain partial response rate in prospective evaluation, with a median brain-specific PFS of approximately 11 months and no treatment-related neurologic adverse events reported in these cohorts, including in patients with measurable, non-locally pretreated RCC-BM. In RCC-BM, cabozantinib and selected immunotherapy-based approaches show clinically relevant CNS activity, but the evidence base remains smaller than in NSCLC or breast cancer. For patients with symptomatic, untreated, or progressing BM, focal therapy remains an important component of care. Integration of Local Therapies with Modern Systemic Therapies aStereotactic Radiosurgery timing with ICI/TKI/ADCs Stereotactic Radiosurgery timing with ICI/TKI/ADCs Concurrent or closely sequenced ICI with Stereotactic Radiosurgery/Stereotactic Radiotherapy (SRS/SRT) may be feasible with emerging evidence of generally low absolute radionecrosis (RN) in prospective datasets, although retrospective cohorts show variable risk signals. Early prospective studies placing ICI within days to nearly 2 weeks of SRS report high local control and limited RN, including pembrolizumab delivered 2–3 days before SRS with favorable safety and control signals [154], nivolumab within 14 days of SRS with protocolized central review and intracranial efficacy endpoints [155], nivolumab one week before SRS with no RN observed in long-term follow-up [156], ipilimumab initiated two days after SRS in a phase I melanoma trial [157], and dual-ICI within ≤7 days of SRS in NSCLC with safety endpoints reported [158]. Larger retrospective series using 2-4-week concurrency definitions show low absolute RN rates but mixed directional signals. In a 657-patient, 4182-metastasis cohort, any RN occurred in 10% and symptomatic RN in 6.8% with no timing effect, while risk was tracked with V12 [159]; another cohort (97 patients, 580 lesions) reported higher 1-year adverse radiation effect with concurrent therapy (6.4%) vs non-concurrent therapy (2.0%; HR: 4.47, p = 0.005) but improved OS [160]; a 260-patient retrospective cohort found pathologic RN in 3% with no difference by timing and improved intracranial control with concurrent therapy [161]; and a half-life-anchored analysis reported 12-month radiographic RN of 3.5% with low symptomatic RN counts among 150 patients/1003 BM [162]. These data suggest that when excess toxicity is observed, it more consistently aligns with dose-volume parameters rather than the calendar interval alone. Concurrent or closely sequenced TKI with SRS appears feasible in modern practice, although risk signals vary by TKI generation and radiosensitization potential. An early randomized dataset evaluating concurrent erlotinib with combined WBRT and SRS (RTOG 0320) demonstrated substantially higher grade 3–5 toxicity (49% vs 11%) and reported necrosis events without survival benefit, suggesting that excess toxicity was likely driven by the three-modality combination and supporting caution with radiosensitizing regimens [163]. More contemporary studies support the feasibility of combining CNS-penetrant EGFR TKIs with radiotherapy, although optimal sequencing remains uncertain. A retrospective cohort defined concurrent EGFR-TKI and radiotherapy as administration within 28 days and reported survival outcomes across treatment strategies [164]. In the prospective randomized STARLET joint analysis, grade 3/4 radionecrosis occurred in 5% of patients treated with SRS plus osimertinib, supporting feasibility of the combination without establishing an optimal treatment-interruption interval [165]. HER2-directed TKIs demonstrate heterogeneous safety signals. Lapatinib-based radiotherapy has been associated with a higher severe-adverse-event signal, although a consistent increase in radionecrosis has not been established [166]. Prospective evaluation of HER2-TKI-based regimens delivered alongside SRS is ongoing [167]. Concurrent or closely sequenced ADC therapy with stereotactic RT appears feasible overall, but RN risk seems to vary by agent rather than across the class. The most consistent signal is seen with T-DM1, where receipt of ≥1 cycle after SRS was associated with higher RN risk on multivariable analysis (HR: 2.5) in a dedicated stereotactic cohort, although exposure windows were defined broadly (post-SRS within months rather than day-level intervals) [168]. In contrast, data for newer ADCs suggest lower absolute risk. For T-DXd, multi-institutional SRS series report low symptomatic RN rates (1% per lesion) when delivered during treatment or within 1 month of infusion [169], while prospective registry data and large real-world cohorts, defining concurrency from same-day delivery up to 10 days, show only rare RN events without a clear increase in severe toxicity [[170], [171]]. For sacituzumab govitecan (SG), available multi-institutional SRS cohorts report no symptomatic RN when stereotactic RT is delivered during treatment, supporting feasibility of continuation through SRS in most cases [172]. Pooled analyses show an overall increase in RN risk with concurrent ADC-RT exposure (RR: 2.7), but this appears largely driven by T-DM1-containing datasets rather than newer ADCs [173]. T-DM1 demonstrates the clearest RN risk signal near SRS, whereas T-DXd and SG generally show acceptable safety with low absolute symptomatic RN rates in current prospective and multicenter cohorts. These data support an agent-specific approach to SRS sequencing. Across available studies, reported radionecrosis signals appear more dependent on the systemic agent than on the exact treatment interval, although the evidence remains heterogeneous (Fig. 4). bFUS-Mediated Blood-Brain Barrier Opening FUS-Mediated Blood-Brain Barrier Opening Agent-specific radionecrosis and toxicity signals with stereotactic radiosurgery. Evidence for ICI, TKI, and ADC therapies delivered concurrently with or near SRS. Categories reflect the direction and consistency of available evidence rather than absolute risk or a recommended treatment interval. Note: ICI is placed in the middle column because timing data are inconsistent across cohorts, not because the overall radionecrosis signal is increased. ADC, antibody-drug conjugate; ICI, immune checkpoint inhibitor; RN, radionecrosis; SRS, stereotactic radiosurgery; T-DM1, ado-trastuzumab emtansine; T-DXd, trastuzumab deruxtecan; TKI, tyrosine kinase inhibitor; WBRT, whole-brain radiotherapy. [[154], [155], [156], [157], [158], [159], [160], [161], [162], [163], [164], [165], [166], [167], [168], [169], [170], [171], [172], [173]]. The BBB and BTB remain major determinants of intracranial drug exposure in patients with BM, particularly for large molecules such as monoclonal antibodies and ADCs. Low-intensity focused ultrasound (FUS), delivered transcranially in conjunction with intravenous microbubbles, enables transient, spatially targeted, and reversible BBB opening [174]. Oscillation of circulating microbubbles within the ultrasound field produces mechanical effects on the cerebrovascular endothelium, resulting in temporary tight junction modulation and increased vascular permeability, a phenomenon demonstrated in preclinical brain metastasis models using MR-guided FUS with microbubbles [[175], [176], [177]]. BBB integrity typically restores within hours, and contrast-enhanced MRI can confirm the focal region of opening [175]. Preclinical studies in BM models support the mechanistic rationale for this approach. For example, FUS-mediated BBB disruption has enhanced delivery and therapeutic activity of HER2-targeted antibodies in breast cancer BM models [178], increased tumor accumulation of chemotherapy in triple-negative breast cancer BM [179], and improved delivery of HER2-targeted NK-92 cells to intracranial metastases [180]. The first clinical proof-of-concept for antibody delivery enhancement in BM was reported by Meng and colleagues in 2021 (NCT03714243) in HER2-positive breast cancer BM [181]. In this first-in-human, single-arm study (4 patients; 20 MR-guided FUS sessions), patients received systemic trastuzumab-based therapy with radiolabeled trastuzumab SPECT imaging to quantify intracranial delivery. MRgFUS-treated lesions demonstrated a mean increase of 101 ± 71% in standardized uptake value ratio (SUVR) compared with −18 ± 26% in untreated control lesions, and 87 ± 17% of sonicated voxels showed >20% increases in SUVR (voxel-wise increases up to 450%). No treatment-related serious adverse events were observed, and mean tumor measurements decreased by 19 ± 12% over the treatment course. These findings provided the first direct human evidence that transcranial FUS can safely improve monoclonal antibody delivery to metastatic brain lesions [181]. Beyond trastuzumab, FUS-mediated BBB opening is now being prospectively evaluated in combination with immunotherapy. The ongoing LIMITLESS randomized controlled trial (NCT05317858) enrolls patients with EGFR- and ALK-negative NSCLC-BM (≤3 lesions, ≥1 measurable per RANO-BM) and randomizes patients 2:1 to low-intensity focused ultrasound (LIFU) plus pembrolizumab versus pembrolizumab alone [182]. LIFU is performed before each pembrolizumab cycle. The primary endpoint is 6-month overall objective response rate per RANO-BM criteria, with exploratory endpoints including intracranial progression-free survival and overall survival. Enrollment began in 2022, and mature efficacy data are pending [182]. Multiple focused ultrasound platforms are being applied in neuro-oncology to achieve transient BBB opening, with differences primarily related to invasiveness and targeting architecture. Implantable devices such as SonoCloud are surgically placed and permit repeated, localized BBB disruption without skull-related acoustic distortion [183,184], whereas transcranial MR-guided phased-array systems such as Exablate deliver noninvasive, electronically steered sonication under real-time MRI guidance with acoustic emissions monitoring and contrast-enhanced imaging to verify BBB opening and subsequent barrier restoration [185]. Frameless neuronavigation-based approaches aim to streamline outpatient workflows but depend on accurate registration and head stabilization. Among noninvasive strategies, transcranial MR-guided systems are currently the most clinically mature. While current literature establishes biological plausibility, prospective and randomized clinical validation in metastatic populations remains essential. cTTFields after SRS TTFields after SRS Tumor Treating Fields (TTFields) are a notable device-based therapy evaluated in a randomized phase III framework following SRS for BM. The METIS trial evaluated this approach in patients with 1–10 newly diagnosed, mutation-negative NSCLC-BM, using time to intracranial progression, assessed by blinded independent central review per RANO-BM, as the primary endpoint [186]. Secondary endpoints included OS, neurocognitive outcomes, quality of life (QoL), safety, and patterns of intracranial failure. TTFields significantly delayed time to intracranial progression (HR 0.72; 95% CI, 0.53–0.98; Fine-Gray P=.044), without worsening QoL or cognitive function [186]. No statistically significant OS difference was observed. Device-related adverse events were mainly grade ≤2 skin events. In exploratory analyses, patients receiving ICIs demonstrated a more pronounced intracranial benefit signal [186,187], suggesting potentially greater benefit in this subgroup. Beyond METIS, evidence remains limited. The RAD 1704 pilot study (NCT03488472) in small-cell lung cancer evaluated SRS followed by continuous TTFields (NovoTTF-200A) but enrolled only one patient and terminated early, yielding no interpretable efficacy data [188]. Similarly, a phase I melanoma study (NCT05341349) was designed to evaluate TTFields in combination with SRS plus pembrolizumab or nivolumab/ipilimumab, with grade 3 CNS toxicity as the primary endpoint and secondary measures including intracranial control, progression-free survival, overall survival, radionecrosis, and biomarker correlates [189]. The study was subsequently terminated because of slow accrual after enrolling one patient, with no results reported. Outside NSCLC, clinical evidence for TTFields following SRS therefore remains extremely limited. TTFields are U.S. FDA-approved for glioblastoma and for metastatic NSCLC that has progressed on or after platinum-based therapy, in combination with a PD-1/PD-L1 inhibitor or docetaxel, and are authorized under a Humanitarian Device Exemption for unresectable locally advanced or metastatic malignant pleural mesothelioma in combination with chemotherapy; however, neither authorization extends to the post-SRS brain metastasis indication evaluated in METIS [[190], [191], [192]]. dComprehensive Decision Framework Comprehensive Decision Framework The evolving intracranial efficacy of contemporary systemic therapies, the agent-specific safety considerations surrounding SRS, and emerging adjunctive approaches such as FUS and TTFields merit the need for an integrated decision framework. Initial management of BM increasingly depends on tumor biology, steroid requirement, intracranial disease burden, and the magnitude of reported intracranial activity of available systemic therapies. A conceptual framework integrating these elements is presented in Fig. V. Proposed Decision Framework for Initial Management of Brain Metastases.Intracranial activity categories are based on reported intracranial objective response rates from dedicated or prespecified CNS analyses. Thresholds are conceptual and intended to reflect magnitude of intracranial efficacy rather than strict numerical cutoffs. Final management should be determined through multidisciplinary evaluation.Abbreviations:CNS, central nervous system; iORR, intracranial objective response rate; SRS, stereotactic radiosurgery. Future Directions Emerging strategies for BM extend beyond refining systemic cytotoxic potency to overcoming BBB/BTB through engineered delivery mechanisms. Bispecific ADC constructs are highlighted as a promising class that can engage multiple tumor antigens simultaneously, improving internalization and potency while potentially mitigating resistance mechanisms seen with conventional ADCs [193]. Additionally, BBB shuttle technologies, such as engineered peptide or receptor-mediated transport platforms (e.g., “Brainshuttle”), demonstrate that large biologics can be ferried into the CNS more effectively by exploiting endogenous transcytosis pathways, improving central exposure without disruptive barrier damage [194]. These approaches mirror longstanding challenges in CNS drug delivery and offer a rational pathway for designing ADCs and bispecifics with enhanced intracranial penetration. Complementary to molecular engineering, FUS-mediated BBB opening continues to mature as a clinical adjunct that transiently and reversibly enhances local CNS permeability to systemic therapies. FUS platforms leveraging microbubbles have been shown to facilitate targeted intratumoral delivery and are being evaluated in early human studies aimed at enhancing drug uptake in brain tumors [181]. Integrating stereotactic radiosurgery with FUS-enhanced delivery and CNS-optimized ADCs could create a tripartite paradigm that couples local cytoreduction, spatially targeted barrier modulation, and potent systemic payload delivery. Such multimodal strategies, grounded in both biologic design and delivery science, are poised to redefine intracranial drug efficacy in BM. Conclusion BM represent a biologically and therapeutically distinct manifestation of systemic cancer in which outcomes are shaped by tumor biology, neurovascular barriers, and compartmentalized immune dynamics. Advances in targeted therapies, immune checkpoint blockade, and ADCs have demonstrated that meaningful intracranial responses are achievable across several tumor types, particularly when agents possess intrinsic CNS activity. In oncogene-driven NSCLC, modern TKIs have transformed intracranial control, while immunotherapy combinations and next-generation HER2-directed therapies have improved CNS outcomes in melanoma and breast cancer. Despite these advances, intracranial outcomes often remain less durable than systemic responses, and interpretation of CNS activity is frequently limited by heterogeneous trial design, inconsistent endpoints, and underrepresentation of patients with symptomatic or steroid-dependent disease. Optimal management therefore requires integration of systemic therapy with focal approaches such as stereotactic radiosurgery, guided by tumor biology, disease tempo, and demonstrated CNS drug activity. Continued progress will depend on prospective trials that consistently define and report intracranial response, intracranial progression, new CNS lesion development, prior radiotherapy exposure, steroid use, and symptomatic status. Greater standardization of CNS endpoints, ideally using brain metastasis-specific criteria such as RANO-BM when feasible, will be essential to improve cross-trial interpretation and guide multidisciplinary treatment decisions. Author contributions Zouina Sarfraz: Literature review, data synthesis, manuscript drafting, and manuscript revision. Rupesh Kotecha: Conceptualization, intellectual input, critical manuscript revision, and supervision. Arun Maharaj: Figure preparation, manuscript revision, and editorial input. Manmeet S. Ahluwalia: Conceptualization, overall supervision, strategic direction, critical manuscript revision, and final approval of the submitted version. All authors approved the final manuscript and accept responsibility for the work. Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: ZS and AM report no conflicts of interest. RK reports institutional research funding from Medtronic Inc., Blue Earth Diagnostics, NovoCure Inc., GT Medical Technologies, AstraZeneca, Exelixis, Brainlab, Cantex Pharmaceuticals, Ion Beam Applications, Kazia Therapeutics, and Black Diamond Therapeutics; consulting fees from Kazia Therapeutics, Elekta AB, NovoCure Inc., Telix Pharmaceuticals, GT Medical Technologies, Neosoma, and Radiopharm Theranostics Limited; honoraria from Elekta AB, Accuray Inc., NovoCure Inc., Brainlab, and Ion Beam Applications; support for travel or meeting attendance from Elekta AB, NovoCure Inc., Brainlab, GT Medical Technologies, Zap Surgical Systems, and Ion Beam Applications; participation on a Data Safety Monitoring Board or advisory board for GT Medical Technologies, Insightec Ltd., Plus Therapeutics, Inc., and Servier; and stock options in Neosoma Inc. MSA has grants from Pfizer and Incyte; is a consultant for Bayer, Sumitomo Pharma Oncology, Autem Therapeutics, GT Medical Technologies, Modifi Biosciences, Bugworks, Allovir, EquilliumBio, VBI Vaccines, Servier Pharmaceuticals, Incyte, Recordati, Novocure, AbbVie, and Genmab; and has stock options in Bugworks, Mimivax, Trisalus Lifesciences, LiveAI, and Ryght. Acknowledgments None. Footnotes This article is part of a special issue on published in Neurotherapeutics References