Brain metastasis (BM) is a common and grave complication in non-small cell lung cancer (NSCLC). Almost a third of NSCLC patients develop brain metastasis at some point during their disease course, with higher rates reported in patients with epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) mutations. BMs are associated with adverse neurocognitive function, poor quality of life, and dismal prognosis despite multidisciplinary treatment with surgery, radiation therapy (RT), and systemic agents.
Brain metastasis (BM) is a common and grave complication in non-small cell lung cancer (NSCLC). Almost a third of NSCLC patients develop brain metastasis at some point during their disease course, with higher rates reported in patients with epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) mutations . BMs are associated with adverse neurocognitive function, poor quality of life, and dismal prognosis despite multidisciplinary treatment with surgery, radiation therapy (RT), and systemic agents . There is, therefore, a critical need for more effective therapies for NSCLC patients with BMs.
Over the past few decades, understanding of tumor biology and the immune system has led to the development of immune checkpoint inhibitors (ICIs) that have revolutionized the treatment landscape for patients with advanced NSCLC. Blocking the programmed death protein-1 (PD-1), its ligand (PD-L1), and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) pathways has led to remarkable improvements in the outcomes of these patients.
There are, however, limited data on the central nervous system (CNS) efficacy of ICIs, as most of the pivotal trials on ICIs excluded or underrepresented patients with BMs . Some of the reasons for excluding these patients include concerns about the ability of monoclonal antibodies such as ICIs to penetrate the blood-brain barrier, diminished efficacy of ICIs due to concurrent steroid use, and hyperprogression of BMs . Despite these concerns, several retrospective studies and prospective trials point toward the activity and safety of ICIs in NSCLC patients with BMs.
In this review, we aim to summarize the current clinical evidence for the efficacy of ICIs in NSCLC patients with BMs, highlight the challenges of incorporating ICIs in treating these patients and identify areas for future research.
It is now known that the brain parenchyma is an immunologically active organ that initiates and regulates immune responses . Varying degrees of T cell infiltration or tumor-infiltrating lymphocytes (TILs) have been observed in BMs . Unlike cytotoxic chemotherapy and targeted therapies, because ICIs act by removing the inhibition of T cells by tumor cells, immune cell trafficking of peripherally activated T cells into the CNS is perhaps more critical than the penetration of the blood-brain barrier by the ICIs themselves . These observations, along with the recent discovery of the CNS lymphatic system and TME preconditions in the CNS that mimic extracranial metastases ] are challenging our long-held notions of immune privilege in the CNS and support the clinical development of ICI-based strategies in patients with BMs.
Currently, there are limited prospective data on the efficacy and safety of ICIs in NSCLC patients with BMs. Patients with BMs have historically been underrepresented in the clinical trials of ICIs in NSCLC. Besides, the majority of these trials do not report intracranial efficacy and other outcomes stratified by the presence or absence of BMs. As a result, our current understanding of the efficacy and safety of ICIs in BMs in NSCLC have primarily been derived from the single-arm phase 1–2 trials , expanded access programs , post hoc/pooled analyses of clinical trials , and retrospective series .
Given the concerns for increasing peritumor inflammation and vasogenic edema, the majority of ICI trials have excluded NSCLC patients with BMs that have not received local therapies such as RT. However, recent data presented below challenge this notion and provide evidence that ICI therapy by itself might be able to achieve intracranial response with acceptable safety in a select group of patients.
Goldberg et al. have recently reported an updated analysis of the NSCLC cohort of their phase 2 trial of pembrolizumab in patients with NSCLC or melanoma with untreated brain metastases . After a median follow-up of 8.3 months, 11 (29.7%, 95% CI, 15.9–47.0) of 37 patients in cohort 1 showed intracranial response with a CNS progression-free survival (PFS) of 2.3 months (95% CI, 1.9-not reached) with almost one-third of patients remaining progression-free in the CNS at 1 year. No responses were seen in the PD-L1 negative patients (5 patients in cohort 2). Treatment-related serious adverse events occurred in 6 (14%) of 42 patients and were comparable with adverse events reported in other ICI trials.
Another prospective trial that specifically evaluated NSCLC patients with untreated BMs was the Checkmate 012 . Patients in “arm M” of this phase 1 trial included 12 patients with at least one asymptomatic and untreated BM up to 30 mm in size. Patients were required to have at least one prior systemic therapy for NSCLC and could have up to four BMs. , no treatment-related neurologic adverse events were reported.
Several other ongoing single-arm phase 2 trials are evaluating the role of ICI in patients with untreated BMs (Clinicaltrials.gov NCT02681549, NCT02886585, NCT03526900) (Table 2). The intracranial efficacy will be measured by modified RECIST in the first study, while Response Assessment in Neuro-Oncology Brain Metastases (RANO-BM) criteria will be used for the other two studies .
In addition to these prospective trials, several other retrospective studies have suggested the potential efficacy of ICI alone in untreated BMs and are summarized in Table 1. However, the patients in the above studies were highly selected and only included patients that had small BMs and were asymptomatic. Therefore, further studies are needed to better clarify the efficacy and safety of ICI alone for untreated BMs that are larger or are symptomatic.
|Author, Year||Trial||Phase||LOT||N||Histology||PD-L1||CNS Disease||ICI Arm||Comparator Arm, If Present||F/u||CNS ORR||Median CNS PFS||Extracranial ORR||DOR||Median PFS (mo) or PFS HR||Median OS (mo) or OS HR||Notes|
|Mansfield, 2019 ||Pooled Analysis of KEYNOTE-001, −010, −024, and −042||Ib to III||≥1||293/3170||Squamous + non-squamous||PD-L1 TPS ≥1%||Treated and stable||Pembro||Chemo||18.4||-||-||26.1% (20.2–32.8) vs. 25.8% (23.7–27.9)||NR (IQR 3.3 to 46.2+) vs. 30.4 (IQR 1.4+ to 49.3+)||0.96 (0.73–1.25) vs. 0.91 (0.84–0.99)||13.4 vs. 10.3; 0.83 (0.62–1.10) [vs. 14.8 vs. 11.3; 0.78 (0.71–0.85)]||TRAEs occurred similarly with pembro vs. chemo both in pts with BM (66% vs. 84%) and without (67% vs. 88%)|
|Goldberg, 2020 ||NCT02085070||II||≥1||42||Squamous + non-squamous||PD-L1 TPS ≥1% (n = 37) or 0% (n = 5)||Untreated and asymptomatic (5 mm to 20 mm)||Pembro||-||8.3||29.7% (15.9–47.0)||2.3 (1.9-not reached)||-||6.9 (IQR 3.7–22.4)||1.9 (1.8–3.7)||9.9 (7.5–29.8)||6/27 patients had discordant response.|
|Goldman, 2016 ||Pooled Analysis of Checkmate 063, 017 and 057||II to III||≥2||46||Squamous + non-squamous||NA||Treated and stable||Nivo||Docetaxel||8.4||-||-||-||-||-||Checkmate 017: 4.99 vs. 3.86 (nivo vs. docetaxel) (HR not reported); Checkmate 057: 7.61 vs. 7.33; 1.04 (0.62–1.76)||CNS TRAEs occurred in 5 pts (11%) and were all gr 1–2 (paresthesia, n = 2; dizziness, somnolence, and tremor, n = 1 each)|
|Hellman, 2017 ||Checkmate 012, Arm M||I||≥2||12||-||-||Untreated and asymptomatic (≤3 cm and ≤4 in number)||Nivo||-||-||16.7 (2.1–48.4)||-||-||-||1.6 (0.92–2.50)||8.0 (1.38–15.50)||2 out of 12 patients achieved intracranial responses, including a patient with leptomeningeal disease|
|Lukas, 2017 ||Pooled analysis of PCD4989g, BIRCH, FIR, POPLAR, and OAK||I to III||≥2||79/1452||Squamous + non-squamous||Unselected||Treated and stable||Atezo||Chemo||-||-||-||-||-||-||20.1 vs. 11.9; 0.54 (0.31–0.94) vs. 13.0 vs. 9.4; 0.75 (0.63–0.89)||Incidence of all AEs and SAEs was similar in pts with or without BMs. The most common treatment + R8-related neurological AE was headache in 6 (8%) pts with and 42 (3%) pts without BM.|
|Gagdeel, 2018 ||OAK||III||≥2||123/850||Squamous + non-squamous||Unselected||Treated and stable||Atezo||Docetaxel||28||-||Time to radiographic identification of new symptomatic BM: NR vs. 9.5; 0.38 (0.16–0.91)||-||-||-||16.0 vs. 11.9; 0.74 (0.49–1.13) [vs. 13.2 vs. 9.3; 0.74 (0.63–0.88)]||No treatment-related grade 4–5 neurologic AEs or SAEs were observed in patients with a history of asymptomatic, treated BM, and there was a low incidence of treatment-related grade 3 neurologic AEs (5%).|
|Hendriks, 2019 ||Retrospective study that included patients on routine clinical care, EAPs, compassionate use programs, and clinical trials||1||255/1025||Squamous + non-squamous||Unselected||Untreated and asymptomatic or treated and stable (stable or decreasing symptoms allowed)||Anti-PD-1 or anti-PD-L1 monotherapy||-||15.8||27.3 (PD-L1 positive patients (n = 14): 35.7%; PD-L1 negative (n = 9): 11.1%)||-||20.6% vs. 22.7%||-||1.7 (1.5–2.1) vs. 2.1 (1.9–2.5) (with and without BM)||8.6 (6.8–12.0) vs. 11.0 (8.6–13.8)||Multivariable analysis showed that steroid use (HR, 2.37) was associated with poorer OS, whereas stable BMs (HR, 0.62) and higher ds-GPA classification (HR, 0.48–0.52) were associated with improved OS.|
|Crino, 2017 ||Retrospective (Italian EAP)||≥2||409/1588||Non-squamous||-||Asymptomatic||Nivo||-||6.1||-||-||-||-||-||-||-|
|Molinier, 2017 ||Retrospective (French EAP)||≥2||130/600||Squamous + non-squamous||Unselected||NR||Nivo||12% partial response||6.6||7 patients had all-grade neurological symptoms, 1 (0.1%) grade 3, not specified whether it was BM patient or not.|
|Gauvain, 2018 ||Retrospective||≥2||43/191||Squamous + non-squamous||unselected||Included all patients whether treated or not, symptomatic or asymptomatic||Nivo||-||5.8||9.0 (3.0–23.0)||3.9 (2.8–11.1)||11% (4–26)||-||-||-||Five neurological events occurred, including 1 grade-4 transient ischemic attack of uncertain imputability and 1 grade-3 neurological deficit; neither required nivo discontinuation.|
|Cortinovis, 2017 ||Retrospective (EAP Italy)||≥2||38/372||Squamous||unselected||Treated and stable||Nivo||-||4.5||-||-||-||-||5.5||6.5||Disease control rate was 47.3%, including 1 complete response, 6 partial responses, and 11 stable diseases. Out of the 38 patients included, only 1 discontinued treatment due to AE (2.6%), whereas 21 pts (55.3%) discontinued treatment for non-toxicity related reasons.|
|Watanabe, 2017 ||Retrospective||≥2||4 out of 48||-||-||Untreated||Nivo||-||-||-||-||-||-||1.8||-||None of the BM patients treated with nivolumab achieved intracranial response.|
|Dudnik, 2016 ||Retrospective||≥2||5||Squamous + non-squamous||NA||Untreated but asymptomatic||Nivo||-||-||-||-||-||-||-||-||Two intracranial responses were observed, including one complete response of parenchymal brain metastases and one partial response of leptomeningeal carcinomatosis. All of the responses were rapid and durable. Importantly, no grade 3/4 adverse events were seen. Systemic responses and intracranial responses were largely concordant|
|Bjørnhart, 2019 ||Retrospective||-||21||-||-||-||Nivo or pembro||-||-||4.8||-||-||-||4.2 (2.5–5.9)||8.2 (1.0–15.5)||-|
|Dumenil, 2018 ||Retrospective||-||10||-||-||Nivo||-||-||-||-||-||-||-||3.1||-|
|Garde-Noguera, 2018 ||Retrospective||-||38||-||-||Nivo||-||-||-||-||17.2||-||1.6||3.1||-|
|Sun, 2020 ||Retrospective||≥1||66||Squamous + non-squamous||Treated and stable or received RT within 30 days of starting pembro or untreated||-||-||15||-||-||-||-||9.0 vs. 7.9 (with or without BM)||18.0 vs. 21.0 (with or without BMs)||13 treated with pembro alone, intracranial responses included 2 CR, 2 PR, 3 SD, and 4 PD. On multivariable analysis, female sex, ECOG 0–1, adenocarcinoma histology, and P as first line therapy were associated with improved PFS and OS. Presence of BM, baseline steroid use, and timing of local RT (before vs. after P) were not associated with inferior survival|
The addition of RT to ICI has been investigated as a means to create synergy between the two treatment modalities by priming the immune response and, possibly, an abscopal response (tumor regression at a site distant from the primary site of radiotherapy) . Although RT can lead to leukopenia and cause immunosuppression, it can also stimulate the innate and adaptive immune system through the release of tumor cell antigens and activation of critical molecular pathways, including increased PD-L1 expression .
|Clinicaltrials.Gov Identifier||Phase||Disease||Major Inclusion Criteria||Steroids||Intervention||Estimated Enrollment|
|NCT02681549||II||NSCLC and melanoma||At least one untreated BM 5–20 mm, asymptomatic, and not requiring steroids, PD-L1 positive||Steroids not permitted||Pembrolizumab plus bevacizumab||53|
|NCT02886585||II||NSCLC and melanoma||Untreated asymptomatic BM or progressive asymptomatic BM measuring ≥10 mm or cytology positive neoplastic meningitis||Stable dose of dexamethasone 2 mg/day or less for 7 days prior to initiation of treatment||Pembrolizumab||102|
|NCT03526900||II||NSCLC||Untreated BM, asymptomatic, and ≤4 mg dexamethasone/day||Up to ≤4 mg dexamethasone/day allowed as long as patients are asymptomatic or minimally symptomatic||Atezolizumab plus carboplatin plus pemetrexed, followed by maintenance atezolizumab plus pemetrexed||40|
Several studies, including a meta-analysis of retrospective data on stereotactic radiosurgery (SRS) and ICI, have suggested better OS with concurrent rather than sequential ICI and SRS . reported retrospective data on NSCLC BM patients who received SRS and anti-PD-1/PD-L1 therapy . Patients who received RT during or before ICI therapy achieved a better distant brain control rate at 6 months compared with patients who received ICI before RT. The rate of acute neurotoxicity was similar among patients who received SRS alone or with ICI
Discussion on the optimal timing and dosing of RT with ICI was outside the scope of this review, and the results of ongoing clinical trials of RT in patients with BMs are expected to give more insight into this important clinical question (such as NCT02696993, NCT02858869, NCT02710253).
Several post hoc analyses of ICIs in NSCLC patients with pretreated, asymptomatic, and stable BMs have been reported in recent years . In a pooled analysis of NSCLC patients with BMs enrolled in three trials with nivolumab (CheckMate 063, 017, and 057), 46 patients who received nivolumab as second-line treatment displayed acceptable safety and promising efficacy when compared to docetaxel . Similarly, in another pooled analysis of pembrolizumab monotherapy trials (KEYNOTE-001, 010, 024, and 042), pembrolizumab showed improved survival with pembrolizumab compared with chemotherapy, irrespective of BM at baseline .
The combination of ICIs with chemotherapy represents another recent advance in the treatment of advanced NSCLC patients, with multiple front-line trials showing the combination to be superior to chemotherapy alone . A pooled analysis of pembrolizumab plus chemotherapy trials (KEYNOTE-021, 189, and 407) has shown the combination to improve survival irrespective of the presence or absence of BM at baseline (Table 3) .
|Author, Year||Trial||Phase||LOT||N||Histology||PD-L1||CNS Disease||ICI Arm||Comparator Arm, If Present||F/u||Extracranial ORR, %||DOR, mo||Median PFS (mo) or HR for PFS||Median OS (mo) or HR for OS||Notes|
|Powell, 2019 ||Pooled analysis of KEYNOTE-021, 189, and 407||II, III||1||171/1298||Squamous + non-squamous||Unselected||Treated and stable||Pembro + chemo||Chemo||10.9||39 vs. 19.7 [vs. 54.6 vs. 31.8]||11.3 vs. 6.8 [vs. 12.2 vs. 6.0]||6.9 vs. 4.1; 0.44 (0.31–0.62) [vs. 8.8 vs. 5.3; 0.55 (0.48–0.63)]||18.8 vs. 7.6; 0.48 (0.32–0.70) [vs. 22.5 vs. 13.5; 0.63 (0.53–0.75)]||All-cause grade 3–5 AEs with Pembro + chemo vs. chemo alone occurred in 81.4% vs. 70.3% of pts with BM and 68.3% vs. 65.6% without BM.|
|Afzal, 2018 ||Retrospective||≥1||18/54||Non-squamous||Unselected||Treated and stable||Pembro + chemo||30||80||6.5||13.7||-|
|Borghaei, 2020 ||Checkmate 227||III||1||135/1739||Squamous + non-squamous||Unselected||Treated and stable||Ipi + nivo||Chemo||29.3 (minimum follow-up)||33 vs. 26 [vs. 33 vs. 28]||24.9 (11.3–NR) vs. 8.4 (4.2–13.9) [vs. 19.6 (15.5–28.6) vs. 5.8 (4.8–6.9)]||5.4 vs. 5.8; 0.79 (0.52–1.19) 4.9 [vs. 5.4; 0.81 (0.70–0.93)]||18.8 vs. 13.7; 0.57 (0.38–0.85) [vs. 17.1 vs. 13.9; 0.76 (0.66–0.88)]||Any-grade nervous system adverse events were reported in 46% of pts with BM treated with ipi + nivo and 42% of those treated with chemo, most were grade 1–2.|
Borghaei et al. recently presented a post hoc analysis of the BM-positive cohort from the Checkmate 227 trial that randomized advanced NSCLC patients into first-line ipilimumab plus nivolumab versus chemotherapy . The data suggested similar efficacy and safety of dual-ICI therapy for NSCLC patients irrespective of the presence or absence of BMs at baseline  (Table 3).
Despite concerns for increased risk of radiation necrosis, there is sparse data on the optimal timing and sequencing of RT and ICI in patients who require RT for symptomatic BMs . Most retrospective studies have shown a manageable short-term safety profile for patients receiving intracranial RT concurrently for BMs while on ICI for extracranial disease .
Radiation necrosis or treatment-induced brain tissue necrosis is a critical delayed complication of radiation therapy that usually develops 6 months to 2 years after radiation . Radiation necrosis is thought to be more common with higher doses per fractionation and with concurrent chemotherapy or radiosensitizers,  and is thought to be more frequent with SRS, especially in the setting of concurrent administration of ICIs . Radiation necrosis is difficult to distinguish from tumor recurrence radiographically and often requires biopsy or serial imaging, as radiation necrosis tends to regress spontaneously after an initial period of growth.
Development of radiation necrosis in patients receiving ICI can be challenging as these patients often require moderate to high doses of steroids that can potentially lower both intracranial and extracranial efficacy of ICIs . In these scenarios, bevacizumab and surgery can be used selectively to control symptoms and facilitate steroid taper .
Pseudoprogression involves a transient enlargement of existing lesions or the appearance of new lesions mimicking tumor progression, which resolves on longitudinal imaging . ICIs (particularly anti-CTLA-4 agents) have been known to result in pseudoprogression when used to treat BMs . For minimally symptomatic lesions, close follow-up with serial imaging can avoid unnecessary tumor-directed therapies. Sometimes, a biopsy is needed to distinguish treatment-related changes from progressive tumors and to guide further therapy .
Evaluating the actual clinical impact of neurologic adverse events in BM patients treated with ICIs is difficult given issues with assessing the adverse events as being tumor-associated inflammatory response, paraneoplastic, or truly autoimmune events, and variable reporting of neurotoxicity across trials . The majority of data for ICI-related neurotoxicity in BM patients come from trials conducted in melanoma patients. CNS autoimmune toxicities due to ICIs are rare but can include myasthenia gravis, encephalitis, aseptic meningitis, and rarely, multiple sclerosis. Available studies on the use of ICIs in NSCLC BMs so far have reported a very low incidence of autoimmune CNS toxicities in these patients .
In the OAK trial of subsequent line atezolizumab versus docetaxel, the risk of identifying new symptomatic brain lesions (patients were not required to undergo regularly scheduled follow-up scans and were instead symptom-driven) in patients with a history of asymptomatic, treated BM appeared to be lower than with docetaxel. In patients without BM at baseline, a larger sample size and a longer follow-up period are needed to generate enough data to draw any meaningful conclusion regarding the future risk of BMs in patients treated with atezolizumab . Similar results were seen in the pooled analysis of five studies (PCD4989g, BIRCH, FIR, POPLAR, and OAK) that evaluated subsequent-line atezolizumab versus docetaxel, with a lower risk of developing a new CNS lesion with atezolizumab (median time to develop a new CNS lesion, not reached versus 9.5 months; HR 0.42, 95% CI, 0.15–1.18) .
Post hoc analysis of data from the IMpower 150 trial showed that the combination of atezolizumab plus bevacizumab plus carboplatin plus paclitaxel (ABCP) might delay the time to new BM development compared with atezolizumab plus carboplatin plus paclitaxel (ACP) . Authors reported that with a minimum follow-up of 32.4 months (with the development of BM in 100 patients), the ABCP regimen was associated with a lower rate of new BM development (7.0%) compared with ACP (11.9%) and BCB (6.0%) regimens. Median time to develop new BMs was not reached in any arms; however, a trend toward delayed time to new BM development was seen in the ABCP arm versus the BCP arm (HR 0.68; 95% CI, 0.39–1.19). The rates of grade 3–4 treatment-related adverse events were similar between the patients with and without BMs but were slightly higher in the ABCP arm than the ACP and BCP arms .