An updated review on systemic therapy for brain metastases in non-small cell lung cancer.

Introduction
Brain metastases (BrM) are common, found in 15–20% of non-small cell lung cancer (NSCLC) patients at initial diagnosis and 30–40% later in the disease course (1,2). Historically, treatment strategies for BrM included surgery and radiation therapy (RT), with limited utilization of systemic chemotherapies, which were thought to be too large or hydrophilic to cross the blood-brain barrier. Stereotactic radiosurgery (SRS), the mainstay of current RT protocols, produces local control rates up to almost 90% for NSCLC BrM (3). But, neurotoxic sequelae, such as radiation necrosis and neurocognitive changes, can still occur (3). Over the past 20 years, as tyrosine kinase inhibitors (TKIs) replaced chemotherapy as first-line therapy for patients with epidermal growth factor receptor mutated (EGFRm) or anaplastic lymphoma kinase fusion positive (ALK+) NSCLC, there has been a shift in strategy. In some cases, the central nervous system (CNS) activity of these drugs may allow clinicians and patients to defer surgery or radiation and their associated toxicities.
Since Riess et al. 2013 (4), several newer CNS penetrant therapies have emerged for the treatment of NSCLC BrM including TKIs, antibody-drug conjugates (ADCs), and immune-checkpoint inhibitors (ICIs) (1). Despite the increased availability of new targeted treatments, data for CNS-specific outcomes in NSCLC clinical trials is limited. Encouragingly, over the past decade, an increasing number of studies allow patients with CNS disease to enroll, but lack of uniform entry criteria, imaging protocols, and outcome measures still hinders direct drug-to-drug and drug to local therapy comparisons (1).
This article summarizes the development of selected CNS active therapies in NSCLC since our last review. We have selected several trials assessing EGFR, ALK, RET, ROS1, NTRK TKIs, ADCs and anti-PD1 inhibitors. While exciting progress has been made, more work is needed to define BrM response, and clearly integrate new therapies into treatment plans that include a combination of systemic therapy, RT, and surgery.

EGFR-TKIs
EGFR-TKIs have become the standard of care for newly diagnosed EGFRm NSCLC, replacing up-front platinum-based chemotherapy. First-generation TKIs erlotinib and gefitinib showed activity against both systemic and CNS disease, but prospective trials of these drugs did not routinely include BrM patients, let alone previously untreated BrM. Some literature suggests there is no significant difference in overall survival (OS) between NSCLC patients without versus with BrM on 1st generation drugs [median 23.7 vs. 22.3 months, P=0.82 (5)]. However, subsequent generations of TKIs were developed with both better systemic and CNS efficacy, and with more routine inclusion of BrM in more recent trials (6-8), we have a better understanding of CNS drug activity in NSCLC BrM patients.
The LUX-LUNG trials (8) compared second-generation TKI afatanib against chemotherapy (cisplatin-pemetrexed in LUX-LUNG 3 and cisplatin-gemcitabine in LUX-Lung 6). Stable asymptomatic BrM were included; 12.2% and 13.5% of patients in LUX-LUNG 3 and LUX-Lung 6 respectively. Median progression-free survival (mPFS) was improved in the afatanib arms compared to chemotherapy in BrM patients, but not statistically significant, limited by small sample sizes in both trials (Table 1). Translating these data to clinical practice is challenging, as patients with symptomatic BrM and leptomeningeal disease were excluded. Inclusion of BrM patients pretreated with RT also clouds the interpretation of any reported response rates.
Osimertinib, a third-generation EGFR TKI, was compared against first generation TKIs (erlotinib or gefitinib) in the pivotal FLAURA trial (6) for newly diagnosed EGFRm NSCLC patients. mPFS was improved for patients with BrM (15.2 vs. 9.6 months; Table 1). Secondary analyses for CNS-specific outcomes showed CNS-mPFS was not reached with osimertinib. Furthermore, CNS progression was reduced in the osimertinib arm compared to the first-generation TKI arm (20% vs. 39%, Table 1) (9). The inclusion of stable BrM post-RT in FLAURA makes interpreting CNS endpoints challenging. To address this, the phase II single arm OCEAN study (10) treated RT-naïve EGFRm NSCLC BrM patients with osimertinib monotherapy. Patients were divided into a T790M vs. TKI-naïve cohorts, and both multiple and symptomatic BrM were allowed. Response rates (RR) to osimertinib in the BrM cohort were 66.7% and 70% using PAREXEL and RECIST criteria, respectively, with a mPFS of 25.2 months across 39 BrM patients (Table 1).
The FLAURA 2 trial (7) compared upfront osimertinib monotherapy against osimertinib with pemetrexed and platinum chemotherapy. mPFS favored the osimertinib-chemo combination group (24.9 vs. 13.8 months, Table 1). In a secondary CNS-specific analysis, CNS-PFS was prolonged in the combination arm compared to monotherapy (20.1 vs. 13.9 for all BrM, 30.2 vs. 27.6 months for measurable BrM) (11). CNS-ORR was similar between both arms (73% vs. 69% for all BrM, 88% vs. 87% for measurable BrM), although a larger difference was seen for the CNS complete response (CR) rate (59% vs. 43% for all BrM, 48% vs. 16% for measurable BrM) (11). Notably, combination osimertinib-chemotherapy led to a higher rate of grade 3 (G3) adverse events (AEs) compared to monotherapy (49% vs. 13% for any possibly causally related AE), and more frequent discontinuation compared to osimertinib alone (45% vs. 6%) (10). Therefore, in clinical practice, prolonged CNS-PFS must be weighed against the higher risk of toxicity with combination osimertinib-chemotherapy.
Resistance to osimertinib is often mediated by MET amplification. Amivantamab, a bispecific antibody targeting MET and EGFR, has been studied as monotherapy and in combination with third-generation EGFR-TKI Lazertinib (ami/laz). The MARIPOSA trial (12) demonstrated improved PFS in BrM patients treated with ami/laz vs. osimertinib (18.3 vs. 13 months) (Table 1). The follow-up MARIPOSA-2 study (13) compared ami/laz/chemotherapy vs. ami/chemotherapy vs. chemotherapy for patients who progressed on osimertinib. BrM patients comprised 41–46% of the cohorts. mPFS was 8.3, 6.3, and 4.2 months in the ami/laz/chemotherapy, ami/chemo, and chemotherapy groups, respectively (Table 1). G3 AE’s occurred at a higher rate with ami/laz, including mucositis, nausea, fatigue, infusion reactions, transaminitis and rash (14). In clinical practice, these toxicities often limit use of ami/laz.
As more EGFR-TKIs are developed, clinicians are faced with increasingly complex treatment decisions, including how to incorporate targeted therapies into the greater treatment schema of surgery and RT. Optimal treatment timing and effective combinatorial strategies with locoregional therapy are not clearly defined with a paucity of randomized prospective trials directly comparing systemic therapy to RT. TURBO (15) is the latest large, retrospective study to evaluate upfront targeted therapy compared to combination targeted therapy and radiotherapy for NSCLC BrM patients. This study included treatment-naïve EGFRm or ALK+ NSCLC patients who received a first-line TKI alone vs. TKI + SRS. In patients with BrM ≥1 cm, mPFS was increased in the TKI + SRS group (Table 1). CNS PFS was increased in TKI + SRS vs. TKI alone (26 vs. 20 months), though not statistically significant. No OS benefit was seen (41 vs. 40 months). These results suggest that TKI + SRS may improve local control, but not necessarily survival for patients with larger symptomatic BrM. In contrast, TKI alone may be appropriate for patients with asymptomatic CNS disease. Several current prospective studies investigating upfront TKI vs. SRS will help better inform optimal treatment sequencing, including NCT 03769103, 03497767 and 05987644.
To summarize, osimertinib is the current drug of choice in newly diagnosed NSCLC with asymptomatic or minimally symptomatic BrM for whom radiosurgery may be deferred until time of progression. Osimertinib with platinum-pemetrexed per FLAURA2 is usually only considered as first-line therapy for patients with a heavier burden of CNS disease given the side effect profile. Patients who develop asymptomatic or minimally symptomatic CNS progression on early generation TKIs may be switched to osimertinib monotherapy. If BrM progress on osimertinib, double or pulse dose osimertinib can be considered. Alternatively, pemetrexed/bevacizumab or amivantamab/lazertinib can be added per MARIPOSA, and/or SRS. As the latter treatment regimens are associated with a higher degree of toxicities, they are typically reserved for patients with substantial burden of disease or for the next line setting. Large/or severely symptomatic lesions that need to be addressed with surgery and post-operative SRS are usually treated with TKI thereafter. With increasing treatment combinations, multidisciplinary discussions amongst neuro-oncology, medical oncology, neurosurgery, and radiation oncology are crucial to determine an individualized treatment approach for each patient.

ALK-TKIs
ALK fusions occur in around 5% of NSCLC patients and are associated with a higher incidence of BrM. Crizotinib was the first ALK-TKI to receive approval, and though some CNS efficacy was reported, the CNS was often the site of first recurrences. The second-generation ALK-TKI alectinib was studied against crizotinib in the J-ALEX (16) and ALEX (17) trials, which improved mPFS in ALK+ NSCLC patients. Patients with stable BrM were allowed in both trials and had higher CNS-response rate (CNS-RR) in the alectinib arm of ALEX compared to crizotinib (81% vs. 50% for measurable baseline BrM, 59% vs. 26% for all BrM) and longer time to CNS progression in J-ALEX (Table 2). These results were reproduced in the ALESIA trial (18), which confirmed superior PFS for BrM patients with alectinib vs. crizotinib (mPFS not estimable vs. 10.7 months respectively; Table 2) in an Asian population. Brigatinib, another 2nd generation ALK-TKI, was studied in ALTA-1L (19) for advanced NSCLC patients. 12-month survival without intracranial disease progression was improved in the BrM patients treated with brigatinib vs. crizotinib (67%, vs. 21%).
Lorlatinib is a 3rd generation ALK-TKI that can overcome resistance to earlier generation ALK-TKIs. The CROWN trial (20) compared lorlatinib to crizotinib in ALK+ NSCLC patients and allowed asymptomatic treated or untreated BrM. In patients with measurable CNS disease, mPFS was not reached in lorlatinib arm vs. 6 months with crizotinib (Table 2). Lorlatinib also led to a higher intracranial overall response rate (IC-ORR) of 66% vs. 20% for measurable disease, and a longer intracranial duration of response (DOR) (NR vs. 10.2 months). Notably, 77% of patients treated with lorlatinib experienced G3/4 AEs versus 57% of patients in the crizotinib arm. The majority of side effects from lorlatinib included hypertriglyceridemia and hypertension; dose reduction occurred in 23% of patients and temporary discontinuation in 62% of patients. Perhaps unsurprisingly, increased CNS penetration from lorlatinib also produces a higher rate of neurologic side effects affecting mood, cognition, and/or speech (21). Given this tradeoff between increased intracranial efficacy and CNS toxicities, lorlatinib is used with close attention to side effects and a low threshold for dose reduction; in select patients, a second-generation ALK TKI may be used prior to lorlatinib.
In clinical practice, newly diagnosed ALK+ NSCLC BrM patients who have asymptomatic or minimally symptomatic disease may be treated upfront with a 2nd or 3rd generation ALK-TKI, in which case SRS may be deferred until evidence of progression. If patients progress on a 2nd generation ALK-TKI without significant worsening of BrM symptoms, lorlatinib should be considered with careful monitoring of neurologic side effects that may develop. Large and/or significantly symptomatic CNS lesions usually will require discussion of locoregional therapy that can be followed by TKI.

RET TKIs
RET fusions (RET+) occur in 1–2% of patients with NSCLC. RET TKIs are effective in these patients, including those with BrM. The randomized Phase 3 trial LIBRETTO-431 (22) examined selpercatinib vs. platinum-based chemotherapy with or without pembrolizumab in patients with advanced RET+ NSCLC (24.8 vs. 11.2 months). Within the subgroup of 29 patients with measurable BrM, the selpercatinib group had higher IC-RR than control (82% vs. 58%). Pralsetinib was studied in the ARROW trial (23) in patients with RET+ NSCLC. Of the ten patients with evaluable BrM, 70% experienced an intracranial response, including three patients with CR. Median DOR was 10.5 months, with 71% of responses still ongoing at 6 months and 36% at 12 months.

ROS1 TKIs
Several TKIs with activity against ROS1 rearrangements have been developed in recent years. Despite the systemic efficacy of crizotinib, a first generation ROS1 TKI, activity against BrM is much more limited (24). Subsequent ROS1 TKIs have since showed improved CNS activity. A combined analysis of 3 clinical trials evaluating entrectinib (ALKA-372-001, STARTRK-1 and STARTRK-2) incorporated 25 metastatic ROS1+ NSCLC patients with measurable baseline BrM and reported an IC-ORR of 80% with median IC-DOR of 12.9 months (25). Modest response was also seen in patients previously treated with crizotinib who had CNS progression (25). Repotrectinib, a separate next-generation ROS1/NTRK TKI with better CNS penetration, also yielded intracranial responses in the TRIDENT-1 trial (26). ROS1+ NSCLC patients with asymptomatic BrM were allowed, and 9 patients with measurable BrM were included in the analysis. IC-RR was 89% in ROS1-TKI naive patients and 38% in pretreated patients.

NTRK TKIs
NTRK gene fusions, targeted by TKIs such as entrectinib and larotrectinib, occur in <1% of NSCLC patients (27). Entrectinib, previously mentioned as a ROS1+ TKI, also targets NTRK and ALK. An integrated analysis of the ALKA-372-001, STARTRK-1 and STARTRK-2 studies included 22 patients with advanced NTRK+ NSCLC and patients with other histologies (27). Overall IC-ORR was 63.6% for the 11 patients with measurable baseline BrM from all histologies, and specifically 80% for the five NSCLC patients. CNS efficacy of larotrectinib has been demonstrated in case reports, even achieving complete intracranial response in one patient with NTRK+ NSCLC and multiple BrM (28). A smaller study also confirmed intracranial response in 4 out of 5 patients with NTRK+ solid cancers and BrM (29).

ADCs
ADCs represent a novel class of compounds that have substantially transformed the treatment landscape for various cancers, including NSCLC. ADCs are composed of a monoclonal antibody chemically linked to a potent cytotoxic agent with molecular weights in the hundreds of thousands of kDa. ADCs were presumed not to have CNS activity based on their size. However, trastuzumab deruxtecan (T-DXd), an ADC consisting of a HER2-directed monoclonal antibody linked to a topoisomerase I inhibitor, demonstrated clear and durable activity in HER2+ breast cancer BrM patients. In the DESTINY-Breast03 trial, metastatic HER2+ breast cancer patients with stable and pretreated BrM had superior mPFS (15 vs. 3 months) and OS (not reached vs. 25.1 months) when treated with T-DXd vs. ado-trastuzumab emtansine (T-DM1). IC-ORR was almost doubled with T-DXd (65.7% vs. 34.3%) (30). T-DXd also proved beneficial for patients with active BrM per the phase 2 DEBBRAH and TUXEDO-1 trials (IC-ORR rates of 44–73%), and the phase 3/4 DESTINY-Breast12 trial (CNS-ORR for active BrM =62.3%) (30-34).
T-DXd has also been studied in HER2-altered NSCLC. Although most studies included BrM, they were not intentionally designed to evaluate intracranial endpoints; as a result, reported outcomes are retrospective and less rigorous than in breast cancer trials. In DESTINY-Lung01, T-DXd led to durable responses in patients with refractory metastatic HER2-mutant (HER2m) and HER2-overexpressing NSCLC. 36% of HER2m patients had stable asymptomatic BrM, over half of whom had not received radiotherapy. ORR and mPFS in the BrM patients were relatively similar compared to the overall cohort (54.5% vs. 55%, 7.1 vs. 8.2 months) (35). Results were encouraging in the HER2-overexpressing BrM cohort as well, with ORR 50% and mOS not reached in the 5.4 mg/kg dosing group (36). The subsequent DESTINY-Lung02 confirmed efficacy of T-DXd in HER2m NSCLC patients with stable BrM, with ORR 60% and 45.5% in the 5.4 and 6.4 mg/kg cohort, respectively (37). As neither of these trials reported CNS-specific outcomes and incorporated BrM patients with variable pre-enrollment RT histories, it is difficult to draw definitive conclusions about CNS activity. The upcoming DESTINY-Lung04 trial will evaluate both systemic and intracranial efficacy of first-line T-DXd in advanced HERm NSCLC patients compared against standard-of-care.
HER3 is overexpressed in NSCLC and portends worse outcomes (38). Patritumab-deruxtecan is an ADC comprised of an anti-HER3 antibody combined with a topoisomerase I inhibitor. In the phase 2 HERTHENA-Lung01 trial, HER3-DXd was studied in advanced EGFRm NSCLC patients pretreated with EGFR-TKI and platinum-based chemotherapy. 30 asymptomatic BrM patients without prior RT were included. CNS-ORR was 33.3%, suggestive of some degree of intracranial activity (39). The subsequent HERTHENA-Lung02 study compared HER3-DXd against platinum-based chemotherapy in advanced pre-treated NSCLC patients, with stable BrM allowed. Although there was a small yet nonsignificant trend towards prolonged IC-PFS in the HER3-DXd arm (5.4 vs. 4.2 months), the trial did not reach its OS endpoint on the most recent follow-up (mOS 16 vs. 15.9 months) (40). Efficacy of HER3-DXd for leptomeningeal disease also appears to be modest based on the TUXEDO-3 trial, with IC-ORR reported as 11.1% (41). As none of these trials specifically selected for patients with HER3 alterations, activity of HER3-DXd for HER3-altered tumors with BrM is yet to be determined.
Sacituzumab-govitecan (SG) and datopotamab deruxtecan (Dato-DXd) are both Trop-2 directed ADC’s containing topoisomerase I inhibitor payloads. Initially FDA-approved for metastatic triple negative breast cancer, SG was then studied in the EVOKE-01 trial, which did not demonstrate significant OS benefit for metastatic refractory NSCLC patients. Despite including a small portion of stable BrM, no CNS outcomes were described (42). Dato-DXd, which has a more potent payload and longer half-life than SG (43), has yielded more encouraging results for BrM patients. The TROPION-Lung01 study demonstrated improved PFS for Dato-DXd when compared to docetaxel in advanced NSCLC. Subgroup analysis of the 170 patients with previously treated or inactive BrM demonstrated continued PFS benefit in the Dato-DXd arm (HR =0.64), although other CNS-specific outcomes were not reported (44).
Dato-DXd remains effective even after treatment with targeted therapy, making it an important next-line agent to consider for patients who have progressed on EGFR or ALK TKI. In the TROPION-Lung05 trial, advanced NSCLC patients with driver mutations previously treated with targeted therapy and platinum chemotherapy were treated with Dato-DXd. Notably over half of the patients had baseline stable BrM, 28% of which were untreated. Exploratory analysis showed promising CNS efficacy with IC-ORR 22% and disease control rate 72% (45).
The emergence of ADCs represents an exciting novel therapeutic class for NSCLC BrM patients. Newer ADCs, such as HER3-DXd and Dato-DXd, may become important next-line considerations for patients who progress after TKI therapy.

Anti-PD1
ICIs have demonstrated modest efficacy in NSCLC BrM, although reporting of intracranial outcomes remains limited. Most studies report sustained systemic PFS and OS benefit for BrM patients treated with ICI, with few comments on actual intracranial outcome measures, making it difficult to weigh systemic therapy alone against local CNS treatments. A small study of pembrolizumab in non-PD-L1 expression-selected patients, demonstrated a CNS RR of 33% (46). Combination therapies have also shown some benefit. First-line nivolumab (nivo) and ipilimumab (ipi) improved survival when compared to platinum-doublet chemotherapy for NSCLC patients without driver mutations per the CheckMate 227 Part 1 trial, with asymptomatic and pretreated BrM allowed (47). Post-hoc exploratory analysis at 5 years demonstrated some OS benefit for the patients with baseline BrM (17.4 vs. 13.7 months, HR =0.63), mildly improved 5-year IC-PFS in the combination ipi/nivo arm (16% vs. 6%), as well as a reduction in the development of new BrM when compared to chemotherapy (4% vs. 20%). Treatment-related AEs occurred in 77% of the patients in the ipi/nivo arm vs. 76% in the chemotherapy arm, raising the concern that drug toxicities should be heavily considered in the treatment calculus for our NSCLC BrM patients (47). With an intracranial response rate limited to around 33% (46), ICIs appear to be less effective than upfront RT, which achieves much higher local control rates of around 90% (3). Furthermore, whereas SRS results in similar local control for the treatment of BrM and lung metastases (3,48), ICI results in an inferior overall response rate in brain compared to lung metastases (49). Therefore, combination ICI and RT remains the most common approach for patients with non-driver mutation NSCLC, particularly those with elevated expression of PD-L1. The optimal timing of ICI plus SRS, and head-to-head comparisons of ICI vs. SRS vs. ICI + SRS, have yet to be prospectively studied. One retrospective series that evaluated SRS followed by ICI within 90 days in NSCLC BrM patients reported superior OS and intracranial control in patients with higher PD-L1 expression (50). Notably, elevated PD-L1 expression was also associated with a higher rate of symptomatic radiation necrosis, which should be factored into treatment decisions.

Conclusions
Since our prior review of CNS active drugs for NSCLC BrM, the treatment landscape has greatly evolved. Patients and providers have access to new, more effective systemic medications that can help delay or defer local CNS treatments such as surgery and radiation. CNS active drugs can also prevent new BrM, making it even more important for patients and clinicians to understand the growing number of treatment options. While progress has been made, studies that compare drug-to-drug or drug-to-local therapy efficacy specifically in BrM, are rare. Well-designed prospective randomized trials that enroll BrM patients, including those with active and untreated BrM, and intentionally assess CNS-specific outcomes are urgently needed.

Supplementary
The article’s supplementary files as