Impact of baseline brain metastases on survival and CNS progression in NSCLC patients treated with immune checkpoint inhibitors in real-world studies: a systematic review and meta-analysis

Introduction

Background

Brain metastases (BMs), the most common type of brain tumor in adults (1,2), are frequently associated with a dismal prognosis (3), and are characterized by low median survival rates and poor outcomes despite multimodal therapies (3). Lung cancer represents one of the three most common primary tumors that metastasize to the brain and is the tumor with the highest incidence of BMs, with an estimated prevalence of 20–56% (1,3,4) and an age-adjusted incidence of 4.60 per 100,000 in non-small cell lung cancer (NSCLC) and 0.87 per 100,000 in small cell lung cancer (5). Approximately 25–50% of patients with lung cancer develop BMs during the course of the disease (6).

Although recent advances towards the multidisciplinary management of BMs and brain tumors have substantially increased over the past decades, this disease still carries a high mortality, with reported 5-year survival rates as low as 4.7% across all tumor types (7). NSCLC, which accounts for approximately 85% of lung cancer histological subtypes (8) has an extremely poor prognosis in advanced metastatic stages, with 5-year overall survival (OS) rates of 10.7% in the immunotherapy era (9). Median OS has been reported at 15 months for the adenocarcinoma subtype and 9 months for non-adenocarcinoma subtypes (10,11).

The introduction of novel sequencing technologies (12) along with an improved understanding of cancer biology has led to the development of targeted therapies towards the interactions of malignant cells with the immune system. Of special importance is the discovery of immune checkpoint proteins such as the cytotoxic-T-lymphocyte associated protein 4 (CTLA-4) and programmed death-1/programmed death-ligand 1 (PD-1/PD-L1) (13). These proteins are surface receptors that modulate the immune system by balancing responses to antigenic threats while preserving self-tolerance, with some tumors exploiting these pathways to evade immune detection.

Rationale and knowledge gap

Although immune checkpoint inhibitors (ICIs) have demonstrated clinical benefit in patients with NSCLC and BMs, as evidenced by improved OS and progression-free survival (PFS) rates (14-19), data comparing BM status in NSCLC patients remains limited. Most studies do not stratify patients based on the presence of BMs at treatment initiation or include them only as part of subgroup analyses Most studies do not stratify patients based on the presence of BMs at treatment initiation or include them only based on their intracranial burden/symptomatic status (15-23). In major ICI trials, patients have not been stratified by baseline BMs (defined as the presence of BMs prior to ICI initiation), and very few studies have compared the intracranial efficacy between patients with and without BMs (21). It remains unclear whether the use of ICIs in patients without BMs at NSCLC diagnosis can reduce the incidence of, or delay, intracranial progression (IP) compared to those with BMs, and whether this translates into differences in survival outcomes.

Objective

Given the underrepresentation of patients with BMs in clinical trials, we sought to perform a systematic review and meta-analysis of real-world data studies to compare IP and survival outcomes in patients with NSCLC who were treated with ICIs, stratified by baseline BMs. Specifically, we aimed to assess the differences in IP, OS and PFS based on baseline BMs status at ICI initiation. We present this article in accordance with the PRISMA reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-750/rc).

Methods

This study was registered in the PROSPERO international systematic review (Registration ID 1014448).

Search strategy and inclusion criteria

A comprehensive literature search was conducted in PubMed, Embase and Cochrane library for articles published in English, from January 2015 to March 2025. The specific search terms used for each database are described in Table 1. Eligible studies included those that reported real-world cohorts of patients with NSCLC treated with ICIs and provided survival or IP outcomes comparing patients with and without BM at treatment initiation, either through stratified analyses or as a part of a multivariate analysis in which BM status was included as an independent variable. Systematic reviews/meta-analysis that directly compared or examined the effectiveness of ICIs stratified by BMs status in NSCLC were included for narrative discussion purposes only. We excluded clinical trials, case reports, case series, review articles and editorials.

Study selection and data extraction

Results were uploaded to Rayyan.ai, which is an online platform designed to facilitate duplicate detection and the screening process. Two researchers (W.J.S. and A.P.M.) independently screened titles, abstracts, and full texts. Discrepancies were resolved by consensus. Data were extracted into an excel spreadsheet and included: study type, study population, number of patients, type of ICI used, survival metrics, progression metrics, secondary outcomes, follow-up duration and main findings. For eligible studies, point-and-interval estimates of adjusted hazard ratios (HRs) for OS and PFS were extracted when available from univariate- and multivariate-analysis results, or calculated when necessary from the log-rank chi-square and number of events per arm. When raw IP data were available, odds ratios (ORs) were calculated using a 2×2 contingency table. The meta-analysis was performed using a random-effects model.

Outcomes and bias assessment

The primary outcomes analyzed in this study were OS, PFS and IP stratified by baseline BMs. Additional outcomes were extracted and tabulated into an excel spreadsheet for narrative purposes. Bias of real-world studies (RWs) was assessed using NIH Quality Assessment Tool for Observational Studies.

Statistical analysis

HRs for OS and PFS were extracted from each study, while ORs for IP were calculated using 2×2 contingency tables when raw IP data were available. All effect measures were pooled using a random-effects model. Heterogeneity was assessed using Cochran’s Q test and the I² statistic. Leave-One-Out sensitivity analyses were conducted for each outcome.

Sub-group analyses and meta-regressions were conducted to evaluate the impact of study-level characteristics, including systemic therapy regimen predominance (ICI only vs. ICI and other systemic therapies), prior local therapy and status of BMs (active/symptomatic vs. stable/asymptomatic), on pooled outcomes. A study was classified as “moderator predominant” if ≥50% of its population met the defining characteristic. For subgroup analyses, studies were categorized as moderator predominant, or moderator absent. For meta-regressions, the continuous proportion (%) of each moderator was used.

Publication bias was initially assessed using funnel plots and Egger’s test. For outcomes in which bias or small-study effects were suggested, additional analyses were conducted, including trim-and-fill, selection models, and influence diagnostics. Statistical significance was set at P<0.05. All analyses were conducted using RStudio (version 4.3.1) using the ‘meta’, and ‘metafor’ packages.

Results

A total of 2,679 studies were screened. After removing 334 duplicates, 2,319 studies were excluded based on title and abstract review. Twenty-six full-text articles were assessed for eligibility. Of these, ten studies were excluded, seven of which analyzed outcomes in NSCLC patients with BMs patients treated with ICIs but lacked direct comparisons or subgroup analyses between BM+ and BM− patients. Ultimately, sixteen studies were included in the systematic review/meta-analysis: twelve were real-world observational studies, and four were systematic reviews or meta-analyses. The PRISMA flow diagram is presented in Figure 1.

Figure 1 PRISMA flow diagram depicting the selection process of this study. BM, brain metastasis; RCT, randomized controlled trial.

Baseline characteristics of the studies

Out of the 12 real-world data studies, 10 were retrospective and 2 had a prospective design. The 12 RWs included a total of 5,312 patients, of which 4,041 had no baseline BMs at treatment initiation and 1,271 had baseline BMs. Summary of RWs can be found in Table 2. Additionally, four systematic reviews/meta-analyses were retrieved and were summarized for narrative purposes (Table 3).

Summary of RWs

Out of the 12 observational studies, 11/12 studies reported no survival difference according to BMs status, contrasted to one study that reported BM as a poor survival prognostic factor. However, 6/12 studies reported increased IP in patients with BMs. Additionally, other variables were identified as independent prognostic factors for poor overall prognosis including performance status ≥2 (4/12) and use of steroids (3/12). Additionally, 3 studies reported intracranial activity in BM+ patients (Figure 2).

Figure 2 Bar plot representing the main findings in the observational studies included in this study. BM, brain metastasis; NSCLC, non-small cell lung cancer; IP, intracranial progression; PS, performance status; OS, overall survival; RT, radiotherapy.

Survival analysis

Among RWs (n=12), HRs for OS and PFS were extracted from 11 studies. These HRs were either directly reported with 95% confidence intervals (CIs) (23-27,30-34), or calculated from the log-rank chi-square and number of events per arm (28,29). However, in two studies (23,30), HR for OS or PFS, respectively, could not be extracted because the metrics were either not reported or the data were insufficient for calculation.

Compared to patients without baseline BMs, those with BMs at the initiation of ICI had a significant increase in the hazard of death. The pooled HR for OS was 1.15 (95% CI: 1.04–1.28), indicating a 15% higher risk of death in the BM+ group, with low heterogeneity (I²=0.0%) and a statistically significant overall effect (P=0.006) (Figure 3). Leave-one-out sensitivity analysis confirmed the robustness of the findings, showing consistent pooled estimates with no change in significance or heterogeneity across iterations.

Figure 3 Forest plot of hazard ratios for overall survival according to baseline BMs and leave-one-out analysis. BMs, brain metastases; CI, confidence interval; HR, hazard ratio; IV, inverse variance.

In addition, patients with baseline BMs had a significantly higher risk of disease progression compared to those without BMs, with a 19% higher hazard of progression (PFS: HR =1.19, 95% CI: 1.07–1.32; P=0.002), although interstudy heterogeneity was noticeable (I²=36.2%) (Figure 4). However, leave-one-out sensitivity analysis confirmed the stability of these findings, with pooled HRs ranging from 1.13 to 1.22 across iterations. Notably, heterogeneity decreased to 0% when excluding the study by Skribek et al., suggesting that this study may account for most of the variance between studies; however, the direction and the significance of the associations were not affected. Information on HR extraction for each study can be found in Tables S1,S2.

Figure 4 Forest plot of hazard ratios for progression-free survival according to baseline BMs and leave-one-out analysis. BMs, brain metastases; CI, confidence interval; HR, hazard ratio; IV, inverse variance.

Subgroup analyses

Due to the heterogeneous nature of RWs, we conducted subgroup analysis and meta-regressions analyzing three different moderators, in order to explore the differences in populations across studies. Outputs from meta-regression analyses are provided in the Tables S3,S4.

Systemic treatment regimens

Of 12 RWs, exact proportions of systemic regimens were available in 11 studies (Table S5). Among these, 8 were predominantly treated (≥50%) with ICI only regimens (either monotherapy or dual CTLA-4 + PD-1/PD-L1 inhibitors), whereas three studies were predominantly treated with ICI-based combination therapy. In 6/12 studies, both ICI monotherapy and combination regimens were used; the combination regimens most frequently included ICI + chemotherapy (24,28-30,33,34) and ICI + EGFR tyrosine kinase inhibitors (24).

In the subgroup analysis, the hazard of death for patients with baseline BMs compared to patients without was slightly higher in ICI-based combination-predominant studies than in ICI-only predominant studies; however, the association did not reach statistical significance in the combination-predominant subgroup (HR 1.23; 95% CI: 0.91–1.65). The formal test for subgroup differences was non-significant (P=0.60), suggesting that no evidence exists supporting that regimen predominance modified the overall effect (Figure 5). Consistently, meta-regression showed no significant association between the proportion of ICI-only regimens within studies and their effect sizes. The estimated increase in HR was only 0.06% for every 100% increase in the proportion of ICI-only regimens (P>0.99) (Table S3).

Figure 5 Subgroup analysis of systemic regimens across studies comparing (A) overall survival and (B) progression-free survival among NSCLC patients treated with immune checkpoint inhibitors. Studies were stratified into ICI-based combination therapy-predominant and ICI-only predominant cohorts. CI, confidence interval; HR, hazard ratio; ICI, immune checkpoint inhibitor; NSCLC, non-small cell lung cancer; SE, standard error.

Similarly, the overall PFS was not significantly different across subgroups (P=0.68) (Figure 5). Studies predominantly using ICI-only regimens showed a significantly higher hazard of death or progression at any point in patients with baseline BMs compared with those without (HR 1.12; 95% CI: 1.02–1.24) (Figure 5). Meta-regression demonstrated an estimated 4.25% increase of in the HR for every 100% increase in the proportion of ICI-only regimens, although this was not statistically significant (P=0.82) (Table S4).

Pretreated status

Local pre-treatment status prior to ICI initiation was reported in all 12 RWs. In 5 of the 12 studies, the population was classified as predominantly pre-treated with local therapies before ICI initiation, whereas in another 5 studies, most patients were not pre-treated before starting immunotherapy. In one study (Zhang et al.) (27), the proportion of pretreated and non-pretreated patients was equal (50%).

Subgroup analysis for OS showed that patients with baseline BMs in studies classified as non-pretreated dominant had a significantly higher hazard of death compared with those without baseline BMs (HR 1.24; 95% 1.05–1.47). In contrast, in pretreated-dominant studies, the hazard of death was higher but not statistically significant (HR 1.09; 95% CI: 0.95–1.24). Subgroup differences were also not statistically significant (P=0.22) (Figure 6).

Figure 6 Subgroup analysis of pretreated status across studies comparing (A) overall survival and (B) progression-free survival among NSCLC patients treated with immune checkpoint inhibitors. Studies were stratified into pretreated predominant and non-pretreated predominant cohorts. CI, confidence interval; HR, hazard ratio; NSCLC, non-small cell lung cancer; SE, standard error.

Regarding PFS, patients with baseline BMs in studies with a non-pretreated dominant population had a 29% higher hazard of death or progression at any point in time compared with patients without baseline BMs. In studies classified as pre-treated dominant, this effect was attenuated to a non-significant difference between those with and without BMs at ICI initiation (Figure 6). Meta-regression showed minimal variability explained by pre-treatment status (t²=0.0116), with a moderate amount of residual heterogeneity (I²=33%) not being accounted by the moderator. Additionally, for every 100% increase in the proportion of pretreated patients within studies, the HR decreased by an estimated 27%, although this was not statistically increase in the HR (P=0.27) (Table S4).

BMs lesions activity status

Active BMs lesions were defined in most studies as newly diagnosed, progressing and not-previously treated lesions. In studies where explicit definition of active BMs was not provided, neurologic symptomatic status was used as a surrogate for active disease. Overall, 9 of the 12 RWs reported active/symptomatic status. Among these, 6 studies, (23,24,26-28,30) included patients with predominantly asymptomatic/stable lesions, while 3 studies (29,32,33) included populations mostly composed of active/symptomatic lesions. Three studies did not report this information (25,31,34).

Subgroup analysis for OS demonstrated a non-significant higher hazard of death among patients with baseline BMs in both asymptomatic/stable-predominant studies and active/symptomatic-predominant studies. The magnitude of effect was greater in the active/symptomatic subgroup (HR 1.19; 95% CI: 0.98–1.45 and HR 1.37; 95% CI: 1.00–1.88, respectively) (Figure 7). Meta-regression showed a non-significant 14% increase in the HR for every 100% increase in the proportion of patients with symptomatic lesions within studies (P=0.72) (Table S3).

Figure 7 Subgroup analysis of brain metastasis burden across studies, comparing (A) overall survival and (B) progression-free survival among NSCLC patients treated with immune checkpoint inhibitors. Studies were stratified into stable/asymptomatic lesion-predominant versus active/symptomatic lesion-predominant cohorts. CI, confidence interval; HR, hazard ratio; NSCLC, non-small cell lung cancer; SE, standard error.

Similarly, patients with active/symptomatic brain lesions had a 49% higher hazard of progression or death at any time point, although this did not reach statistical significance (HR 1.49; 95% CI: 0.96–2.33) (Figure 7). Subgroup differences for both OS and PFS were not statistically significant. Importantly, meta-regression showed a that a higher proportion of active/symptomatic BMs was associated with a 68% increase in the HR (HR 1.68 per 100% increase in symptomatic lesions within studies), although this was not statistically significant (P=0.12). This moderator explained 45% of between-study heterogeneity, while residual heterogeneity remained moderate (I²=36%). These findings suggest that the proportion of active versus stable lesions may be regarded as a meaningful contributor to variability across studies (Figure 7).

IP

Baseline BMs status was significantly associated with IP. Of the 12 RWs, six reported raw IP data (24,26-30), from which crude OR were calculated and included in the pooled analysis. In one of these six studies (Swart et al.), cumulative incidence was reported instead of raw progression events (29); therefore, we derived approximate event counts, introducing minor imprecision but allowing the inclusion of relevant data. Additionally, among these 6 studies, three used the Response Evaluation Criteria in Solid Tumors 1.1 (RECIST) to define IP (24,27,28), one used modified RECIST criteria (30), one relied on local institutional criteria (26), and one did not report the criteria used (29) (Table S6). Imaging intervals ranged from 6–12 weeks, with most of the studies using MRI. The pooled analysis showed that patients without baseline BMs had significantly lower odds of experiencing IP during ICI therapy compared to those with BMs (pooled OR =0.10, 95% CI: 0.05–0.20) (Figure 8). Nonetheless, heterogeneity was high (I²=77.7%, P=0.0004), indicating substantial variability in effect estimates across studies. Leave-one-out analysis identified Hendriks et al. as the study with the greatest influence on the overall effect; removing this study substantially reduced heterogeneity (I²=41%) substantially, likely due as this study was the one with the smallest standard error (Figure 8).

Figure 8 Forest plot of odds ratio for intracranial progression according to baseline BMs and leave-one-out analysis. BMs, brain metastases; CI, confidence interval; OR, odds ratio.

Publication bias

The risk of bias assessment for the included studies is presented in Figures 9,10. Most studies demonstrated a low risk of bias and were methodologically rigorous. All studies clearly defined their populations and defined the outcomes of interest. Only three of the studies had an eligible patient participation rate below 50%, and only two failed to adjust for potential confounding factors. In the two prospective studies, included loss to follow-up was not reported. Funnel plot analysis suggested mild asymmetry for studies reporting OS, and Egger’s test was statistically significant (P=0.01), suggesting potential small-study effects or publication bias (Figure S1). Given this suggestion, we performed a trim-and-fill analysis, which imputed four additional studies (Figure S2). The adjusted pooled HR was 1.10 (95% CI: 1.00–1.21); which is directionally consistent with the original estimate, showing minimal attenuation of the overall effect.

Figure 9 Traffic-light plot for risk of bias based on the NIH Quality Assessment Tool for observational studies.

Figure 10 Summary plot for risk of bias based on the NIH Quality Assessment tool for observational studies.

A selection model analysis was also conducted; however, the model was unstable. This occurred because one of the P value intervals contained no studies, causing the estimated selection weights to reach boundary values (~100%) and generating extremely large standard errors. Additionally, the very low between-study heterogeneity (tau² =0.054) limited the model’s ability to reliably estimate the selection parameters.

To further assess robustness, influence diagnostics were performed. This analysis identified the study by Hendriks et al. as having the highest influence across multiple metrics (although none exceeded conventional thresholds) (Figure S3). Removal of the study did not meaningfully change the pooled effect size or heterogeneity of the study demonstrating that despite its relatively large weight, it does not alter the overall findings. Overall, these results suggest no meaningful evidence of publication bias. The direction of the pooled effect remained stable after trim-and-fill and no single study demonstrated significant influence on the pooled results.

On the other hand, both the funnel plot and Egger´s test for PFS indicated no evidence of publication bias (P=0.22), thus, no additional publication bias analysis were required (Figure S4). Nonetheless, trim-and-fill analysis imputed 2 studies, with an adjusted pooled HR consistent with the direction of the original estimate (HR 1.13; 95% CI: 0.95–1.32) (Figure S5).

Discussion

Key findings

In this systematic review and pooled analysis, we investigated real-world data studies to evaluate at the impact of ICIs on IP and survival outcomes in patients with NSCLC, stratified by the presence of BMs at treatment initiation. Overall, our findings suggest that ICI provide a greater reduction in the risk of IP in patients without baseline BMs compared with those with BMs (Figure 5). Patients without BMs also demonstrated better pooled OS and PFS than those with central nervous system (CNS) disease (Figures 3,4). Additionally, subgroup analysis showed that, in studies with predominantly non-pretreated lesions, patients with baseline BMs had significantly worse OS and numerically worse PFS compared with patients without baseline BMs.

Comparison with similar research

Although some RCTs and prospective trials have included patients with BMs, overall data remain limited because many trials still exclude these patients or use strict eligibility criteria, and because of the high mortality associated with this population (39-48). Nevertheless, several prospective trials and RCTs have indeed explored intracranial and survival outcomes in patients with NSCLC BMs treated with ICIs. For example, a post hoc pooled analysis of three RCTs (KEYNOTE-001, KEYNOTE-010, and KEYNOTE-042) evaluated 3,180 patients stratified by BMs status. Among these, 293 patients had BMs and 2,877 did not. The pooled analysis showed significantly improved OS and PFS in patients without BMs, especially among those with PD-L1 tumor proportion scores ≥50%. While survival improvements were also seen in patients with BMs, the HRs were not statistically significant. Additionally, both the objective response rate and duration of response were longer in patients receiving pembrolizumab compared to chemotherapy, regardless of BM status (49). Similarly, a post-hoc pooled analysis of 1,298 patients with advanced NSCLC (171 patients with BMs and 1,127 without BMs) showed significantly longer OS and PFS in both BM+ and BM− groups treated with pembrolizumab plus chemotherapy versus chemotherapy alone (19).

Further supporting this evidence, we identified five systematic reviews/meta-analyses pooling survival data from RCTs (n=3) and RWs (n=2). These meta-analyses either directly compared or analyzed the outcomes of patients with NSCLC treated with ICIs, stratifying by BMs status (35-38,50) (Table 2). Overall, RCT-based systematic reviews and meta-analysis reported improved survival with ICI treatment compared to chemotherapy in both BM+ and BM− patients (35). One study (37) found better pooled OS in patients with BMs treated with first-line ICIs but lower PFS compared to patients without BMs, while another study (36) showed improved survival in both groups treated with ICIs, with a greater benefit in the BM− subgroup.

On the other hand, one RWs-based systematic review and meta-analysis reported no significant differences in the 1- and 2-year pooled OS rates between BM+ and BM− patients treated with ICIs (38). Importantly, none of the RWs included in our analysis overlapped with those reported in Juarez-Garcia et al.

Main findings

We identified multiple observational studies that reported both systemic and intracranial ICI efficacy by comparing baseline BM status (23-34) (Table 1). Eleven studies reported no statistically significant difference in terms of OS or PFS. Despite only one study (32) finding significantly worse survival in BM+ patients, BM status was not consistently identified as an independent predictor of worse survival across studies. Other clinical factors that were reported as independent predictors of worse outcomes included performance status and steroid use (Figure 2). Additionally, in 6/12 studies that reported IP data, patients with BM+ had a higher rate of IP than patients without, posing the question on the effectiveness of ICIs for slowing IP rates in patients without established CNS disease. While the majority of RWs included in this meta-analysis did not report significant differences in survival outcomes between patients with and without baseline BMs, our pooled analysis revealed a statistically significant differences in survival and IP, with worse outcomes in the BM+ group. This discrepancy likely reflects the increased statistical power provided by meta-analytic methods, detecting modest but consistent effects that individual RWs may miss due to smaller sample sizes or heterogeneous endpoints.

On the other hand, subgroup analyses showed that patients with baseline BMs at ICI initiation experienced worse outcomes when predominantly treated with ICI-only regimens, had non-pretreated lesions, or had active/symptomatic disease (although these findings were not statistically significant). Although none of these moderators reached statistical significance in meta-regression, it is notable that neurological burden accounted for nearly half (45%) of the between-study variability in PFS, suggesting that lesion activity may be an important contributor to outcome differences across RWs.

The studies that contributed to the pooled estimate for OS displayed almost no heterogeneity in HR estimates, and leave-out-analysis confirmed the robustness of our findings, with no variation in heterogeneity or effect size across iterations (Figure 3). Although the funnel plot and Egger’s test suggested the possibility of publication bias, trim-and-fill and influence diagnostics analysis demonstrated a consistent direction of the pooled effect and showed no evidence of a significant influence on the pooled results by a single study. In contrast, the studies that contributed to the pooled estimate for PFS showed minimal to moderate effect-size heterogeneity, with low variability in the leave-one-out sensitivity analysis and no evidence of publication bias, as indicated by a non-significant Egger’s test and a symmetrical funnel plot. The pooled estimate for IP was accompanied by high effect-size heterogeneity among studies, which may be attributable to variability in the reporting and the use of raw data to calculate unadjusted ORs.

Additional RWs studies (51-58) have provided insights into the CNS efficacy and survival benefit of ICIs in patients with BMs, although no direct comparison between BM+ and BM− patients were made in these studies. For example, Kemmotsu et al. found a significantly lower rate of IP in ICI-treated patients compared to those treated with cisplatin-based chemotherapy (55). Similarly, a multicentric cohort involving 322 patients from five different European countries reported that patients with baseline BMs had a significantly longer median time to IP (12 months, 95% CI: 9.1–36.0) than those without baseline BMs (5 months, 95% CI: 3.0–NA; P=0.002) (53). Other retrospective studies also demonstrated survival benefits of ICIs in patients with BMs (51,52,58). However, some RWs such as the one by Tozuka et al. did not did not find a survival benefit of ICIs in this population (57).

Clinical implications

In our pooled results, patients without baseline BMs had a lower risk of death and IP compared to those with baseline BMs during ICI treatment. Importantly, the variability in effect estimates across studies may be partly attributable to differences in systemic therapy regimens, as well as baseline prognostic factors such as pretreatment status and the neurological burden of brain lesions. These differences in survival and intracranial efficacy may be partially explained by the limited permeability of the blood-brain barrier (BBB), as the brain is considered an immune privileged organ (59).

A study by Kim et al., compared intracranial and extracranial responses to ICIs in patients with lung cancer and BMs through immunohistochemical analysis of resected primary lung and secondary brain tumor specimens. Their findings revealed a significantly reduced immune response in BMs, with a lower proportion of PD-1+ tumor invading lymphocytes in the brain specimens compared to the primary lung tumors (60). This immune attenuation may contribute to the suboptimal efficacy of ICIs in the CNS. Nevertheless, despite this reduced intracranial immune activation, ICIs still offer therapeutic benefit in patients with BMs. This is likely due to the systemic activation of peripheral T cells, which may exert anti-tumor effects even in the absence of direct BBB penetration (61,62). This systemic mechanism could explain the observed survival benefit in patients with active BMs (32).

Our findings support the early initiation of ICI therapy in patients who have not yet developed CNS disease involvement. Doing so may not only reduce the risk of IP, but also improve survival outcomes and quality of life, as suggested by Bjørnhart et al. (33).

Finally, it is worth noting that none of the screened studies stratified survival or IP outcomes based on the timing of BM development, such as synchronous (at or <3 months from NSCLC diagnosis) or metachronous (>3 months after NSCLC diagnosis) BMs. This represents a critical gap in the literature, as observational data suggest that the chronicity of BM may independently influence prognosis (56,63-66). However, the direction of this effect is still debated: some studies suggest worse outcomes in synchronous BMs (63,64) while others report worse OS in metachronous BMs (65).

Future framework

Looking ahead, there remains a need for standardized definitions and reporting infrastructure for identifying baseline BMs, the degree of IP as identified on surveillance imaging, CNS-specific response criteria—such as the response assessment in neuro-oncology brain metastases (RANO-BM) criteria and diagnosis-specific graded prognostic assessment (DS-GPA) score—and specific timepoints for follow-up of surveillance imaging. Furthermore, it is of crucial importance that novel biomarkers be identified and incorporated into subsequent investigations to both predict and quantify ICI-based therapeutic efficacy for BM patients. One promising biomarker is PD-L1, with preliminary data suggesting its association with clinical outcomes in BM patients treated with ICIs (49,67).

Further research is also needed to investigate underrepresented subgroups where current data are limited. This includes examining the prognostic relevance of BM chronicity, as well as the influence of other sociodemographic factors such as race, ethnicity, socioeconomic status and real-world access to ICIs on clinical outcomes. Finally, the role of ICIs in high-mortality populations, such as patients with leptomeningeal disease or active progression of CNS lesions, remains poorly understood and warrants focused investigation with these targeted therapies.

Limitations

This meta-analysis was based on real-world data, where study designs are heterogenous (e.g., differences in baseline BMs definitions), patient population is heterogeneous, potential selection bias might exist (e.g., patients with better clinical status being more likely to receive treatment), patients’ characteristics are not controlled/adjusted for, variation in ICI regimens and concurrent treatments occurs, and other confounding factors are difficult to control. Despite the fact that (I) some of the retrospective studies included in this review demonstrated associations, and (II) pooled analyses of survival metrics improved the power of the overall associations, these studies only provide a certain degree of correlation, without providing evidence of causality. Additionally, the OR’s extracted from these studies were unadjusted for confounding factors, thus limiting the potential association of this measure of association, and most of the studies lacked consistent intracranial-specific outcome reporting data which limited the amount of information that could be retrieved. Importantly, choosing IP as an endpoint in RWs is prone to bias due to heterogeneity in the definitions of IP, the types of modalities used and the follow-up imaging interval across studies.

Conclusions

In conclusion, this study highlights real-world evidence supporting the efficacy of ICIs for patients with and without BMs, with a modestly greater benefit in survival and intracranial disease control among patients without BMs. There is also a suggested improvement in survival among patients with pretreated, stable/asymptomatic lesions and those receive ICI-combination regimens. These findings also underscore the limited number of studies that directly compare both populations using consistent and standardized endpoints, such as IP rates or the chronicity of secondary BMs lesions. Importantly, the inclusion of RWs is particularly valuable, given their broader and more demographically representative populations compared to RCTs. This enables comparisons between BM+ and BM– patients that are often not feasible in controlled trials settings. Further prospective studies and clinical trials are needed to include more diverse BMs populations with BMs and to address current gaps in the literature. Expanding therapeutic strategies for these patients may help improve both survival outcomes and quality of life.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the PRISMA reporting checklist. Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-750/rc

Peer Review File: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-750/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-750/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sacks P, Rahman M. Epidemiology of Brain Metastases. Neurosurg Clin N Am 2020;31:481-8. [Crossref] [PubMed]
2. Lu Y, Huang Y, Zhu C, et al. Cancer brain metastasis: molecular mechanisms and therapeutic strategies. Mol Biomed 2025;6:12. [Crossref] [PubMed]
3. Achrol AS, Rennert RC, Anders C, et al. Brain metastases. Nat Rev Dis Primers 2019;5:5. [Crossref] [PubMed]
4. Berghoff AS, Schur S, Füreder LM, et al. Descriptive statistical analysis of a real life cohort of 2419 patients with brain metastases of solid cancers. ESMO Open 2016;1:e000024. [Crossref] [PubMed]
5. Zhang C, Zhang H, Jia W. Trends of Incidence and Survival of Brain Metastases from Lung Cancer. World Neurosurg 2025; Epub ahead of print. [Crossref]
6. Wang S, Uriel M, Cheng H. Lung Cancer with Brain Metastasis-Treatment Strategies and Molecular Characteristics. J Clin Med 2024;13:7371. [Crossref] [PubMed]
7. Yuan J, Cheng Z, Feng J, et al. Prognosis of lung cancer with simple brain metastasis patients and establishment of survival prediction models: a study based on real events. BMC Pulm Med 2022;22:162. [Crossref] [PubMed]
8. Gillespie CS, Mustafa MA, Richardson GE, et al. Genomic Alterations and the Incidence of Brain Metastases in Advanced and Metastatic NSCLC: A Systematic Review and Meta-Analysis. J Thorac Oncol 2023;18:1703-13. [Crossref] [PubMed]
9. Wang Y, Kondrat K, Adhikari J, et al. Survival trends among patients with metastatic non-small cell lung cancer before and after the approval of immunotherapy in the United States: A Surveillance, Epidemiology, and End Results database-based study. Cancer 2025;131:e35476. [Crossref] [PubMed]
10. Sperduto PW, Mesko S, Li J, et al. Survival in Patients With Brain Metastases: Summary Report on the Updated Diagnosis-Specific Graded Prognostic Assessment and Definition of the Eligibility Quotient. J Clin Oncol 2020;38:3773-84. [Crossref] [PubMed]
11. Önal Ö, Koçer M, Eroğlu HN, et al. Survival analysis and factors affecting survival in patients who presented to the medical oncology unit with non-small cell lung cancer. Turk J Med Sci 2020;50:1838-50. [Crossref] [PubMed]
12. Wang D, Liu B, Zhang Z. Accelerating the understanding of cancer biology through the lens of genomics. Cell 2023;186:1755-71. [Crossref] [PubMed]
13. Arafat Hossain M. A comprehensive review of immune checkpoint inhibitors for cancer treatment. Int Immunopharmacol 2024;143:113365. [Crossref] [PubMed]
14. Rios-Hoyo A, Arriola E. Immunotherapy and brain metastasis in lung cancer: connecting bench side science to the clinic. Front Immunol 2023;14:1221097. [Crossref] [PubMed]
15. Goldberg SB, Schalper KA, Gettinger SN, et al. Pembrolizumab for management of patients with NSCLC and brain metastases: long-term results and biomarker analysis from a non-randomised, open-label, phase 2 trial. Lancet Oncol 2020;21:655-63. [Crossref] [PubMed]
16. Reck M, Ciuleanu TE, Lee JS, et al. Systemic and Intracranial Outcomes With First-Line Nivolumab Plus Ipilimumab in Patients With Metastatic NSCLC and Baseline Brain Metastases From CheckMate 227 Part 1. J Thorac Oncol 2023;18:1055-69. [Crossref] [PubMed]
17. Gadgeel SM, Lukas RV, Goldschmidt J, et al. Atezolizumab in patients with advanced non-small cell lung cancer and history of asymptomatic, treated brain metastases: Exploratory analyses of the phase III OAK study. Lung Cancer 2019;128:105-12. [Crossref] [PubMed]
18. Paz-Ares LG, Ciuleanu TE, Cobo M, et al. First-Line Nivolumab Plus Ipilimumab With Chemotherapy Versus Chemotherapy Alone for Metastatic NSCLC in CheckMate 9LA: 3-Year Clinical Update and Outcomes in Patients With Brain Metastases or Select Somatic Mutations. J Thorac Oncol 2023;18:204-22.
19. Powell SF, Rodríguez-Abreu D, Langer CJ, et al. Outcomes With Pembrolizumab Plus Platinum-Based Chemotherapy for Patients With NSCLC and Stable Brain Metastases: Pooled Analysis of KEYNOTE-021, -189, and -407. J Thorac Oncol 2021;16:1883-92. [Crossref] [PubMed]
20. Wei Y, Xu Y, Wang M. Immune checkpoint inhibitors for the treatment of non-small cell lung cancer brain metastases. Chin Med J (Engl) 2023;136:1523-31. [Crossref] [PubMed]
21. Jablonska PA, Das A. Management of brain metastases in non-small cell lung cancer without actionable driver mutations—the need to dive deeper in the right ‘pool’. Transl Lung Cancer Res 2023;12:1966-71. [Crossref] [PubMed]
22. Lin NU, Prowell T, Tan AR, et al. Modernizing Clinical Trial Eligibility Criteria: Recommendations of the American Society of Clinical Oncology-Friends of Cancer Research Brain Metastases Working Group. J Clin Oncol 2017;35:3760-73. [Crossref] [PubMed]
23. Metro G, Banna GL, Signorelli D, et al. Efficacy of Pembrolizumab Monotherapy in Patients With or Without Brain Metastases From Advanced Non-Small Cell Lung Cancer With a PD-L1 Expression ≥50%. J Immunother 2020;43:299-306. [Crossref] [PubMed]
24. Guo T, Chu L, Chu X, et al. Brain metastases, patterns of intracranial progression, and the clinical value of upfront cranial radiotherapy in patients with metastatic non-small cell lung cancer treated with PD-1/PD-L1 inhibitors. Transl Lung Cancer Res 2022;11:173-87. [Crossref] [PubMed]
25. Descourt R, Greillier L, Perol M, et al. First-line single-agent pembrolizumab for PD-L1-positive (tumor proportion score ≥ 50%) advanced non-small cell lung cancer in the real world: impact in brain metastasis: a national French multicentric cohort (ESCKEYP GFPC study). Cancer Immunol Immunother 2023;72:91-9. [Crossref] [PubMed]
26. Hendriks LEL, Henon C, Auclin E, et al. Outcome of Patients with Non-Small Cell Lung Cancer and Brain Metastases Treated with Checkpoint Inhibitors. J Thorac Oncol 2019;14:1244-54. [Crossref] [PubMed]
27. Zhang G, Cheng R, Wang H, et al. Comparable outcomes of nivolumab in patients with advanced NSCLC presenting with or without brain metastases: a retrospective cohort study. Cancer Immunol Immunother 2020;69:399-405. [Crossref] [PubMed]
28. Masuda T, Tsubata Y, Hata K, et al. Efficacy of immune checkpoint inhibitors according to programmed cell death-ligand 1 expression in patients with non-small cell lung cancer and brain metastasis: A real-world prospective observational study. Thorac Cancer 2024;15:2408-17. [Crossref] [PubMed]
29. Swart EM, Noordhof AL, Damhuis RAM, et al. Survival of patients with KRAS G12C mutated stage IV non-small cell lung cancer with and without brain metastases treated with immune checkpoint inhibitors. Lung Cancer 2023;182:107290. [Crossref] [PubMed]
30. Zhou J, Wu Y, Xie M, et al. The clinical outcome and risk factors analysis of immune checkpoint inhibitor-based treatment in lung adenocarcinoma patients with brain metastases. Transl Lung Cancer Res 2022;11:656-69. [Crossref] [PubMed]
31. Cortellini A, Tiseo M, Banna GL, et al. Clinicopathologic correlates of first-line pembrolizumab effectiveness in patients with advanced NSCLC and a PD-L1 expression of ≥ 50. Cancer Immunol Immunother 2020;69:2209-21. [Crossref] [PubMed]
32. Skribek M, Rounis K, Makrakis D, et al. Outcome of Patients with NSCLC and Brain Metastases Treated with Immune Checkpoint Inhibitors in a ‘Real-Life’ Setting. Cancers (Basel) 2020;12:3707. [Crossref] [PubMed]
33. Bjørnhart B, Hansen KH, Asmussen JT, et al. Effect and Tolerability of Immunotherapy in Patients with NSCLC with or without Brain Metastasis. Cancers (Basel) 2022;14:1682. [Crossref] [PubMed]
34. Sun L, Davis CW, Hwang WT, et al. Outcomes in Patients With Non-small-cell Lung Cancer With Brain Metastases Treated With Pembrolizumab-based Therapy. Clin Lung Cancer 2021;22:58-66.e3. [Crossref] [PubMed]
35. Li S, Zhang H, Liu T, et al. The Effect of Asymptomatic and/or Treated Brain Metastases on Efficacy of Immune Checkpoint Inhibitors in Metastatic Non-Small Cell Lung Cancer: A Meta-Analysis. Front Oncol 2021;11:702924. [Crossref] [PubMed]
36. Wang Y, Zhang Q, Chen C, et al. Association of Brain Metastases With Immune Checkpoint Inhibitors Efficacy in Advanced Lung Cancer: A Systematic Review and Meta-Analysis. Front Oncol 2021;11:721760. [Crossref] [PubMed]
37. Siciliano MA, Caridà G, Ciliberto D, et al. Efficacy and safety of first-line checkpoint inhibitors-based treatments for non-oncogene-addicted non-small-cell lung cancer: a systematic review and meta-analysis. ESMO Open 2022;7:100465. [Crossref] [PubMed]
38. Juarez-Garcia A, Sharma R, Hunger M, et al. Real-world effectiveness of immunotherapies in pre-treated, advanced non-small cell lung cancer Patients: A systematic literature review. Lung Cancer 2022;166:205-20. [Crossref] [PubMed]
39. Reck M, Rodríguez-Abreu D, Robinson AG, et al. Updated Analysis of KEYNOTE-024: Pembrolizumab Versus Platinum-Based Chemotherapy for Advanced Non-Small-Cell Lung Cancer With PD-L1 Tumor Proportion Score of 50% or Greater. J Clin Oncol 2019;37:537-46. [Crossref] [PubMed]
40. Reck M, Luft A, Szczesna A, et al. Phase III Randomized Trial of Ipilimumab Plus Etoposide and Platinum Versus Placebo Plus Etoposide and Platinum in Extensive-Stage Small-Cell Lung Cancer. J Clin Oncol 2016;34:3740-8. [Crossref] [PubMed]
41. Borghaei H, Paz-Ares L, Horn L, et al. Nivolumab versus Docetaxel in Advanced Nonsquamous Non-Small-Cell Lung Cancer. N Engl J Med 2015;373:1627-39. [Crossref] [PubMed]
42. Rudin CM, Awad MM, Navarro A, et al. Pembrolizumab or Placebo Plus Etoposide and Platinum as First-Line Therapy for Extensive-Stage Small-Cell Lung Cancer: Randomized, Double-Blind, Phase III KEYNOTE-604 Study. J Clin Oncol 2020;38:2369-79. [Crossref] [PubMed]
43. Paz-Ares L, Dvorkin M, Chen Y, et al. Durvalumab plus platinum-etoposide versus platinum-etoposide in first-line treatment of extensive-stage small-cell lung cancer (CASPIAN): a randomised, controlled, open-label, phase 3 trial. Lancet 2019;394:1929-39. [Crossref] [PubMed]
44. Brahmer J, Reckamp KL, Baas P, et al. Nivolumab versus Docetaxel in Advanced Squamous-Cell Non-Small-Cell Lung Cancer. N Engl J Med 2015;373:123-35. [Crossref] [PubMed]
45. Wu YL, Lu S, Cheng Y, et al. Nivolumab Versus Docetaxel in a Predominantly Chinese Patient Population With Previously Treated Advanced NSCLC: CheckMate 078 Randomized Phase III Clinical Trial. J Thorac Oncol 2019;14:867-75. [Crossref] [PubMed]
46. Hellmann MD, Paz-Ares L, Bernabe Caro R, et al. Nivolumab plus Ipilimumab in Advanced Non-Small-Cell Lung Cancer. N Engl J Med 2019;381:2020-31. [Crossref] [PubMed]
47. Horn L, Mansfield AS, Szczęsna A, et al. First-Line Atezolizumab plus Chemotherapy in Extensive-Stage Small-Cell Lung Cancer. N Engl J Med 2018;379:2220-9. [Crossref] [PubMed]
48. Gandhi L, Rodríguez-Abreu D, Gadgeel S, et al. Pembrolizumab plus Chemotherapy in Metastatic Non-Small-Cell Lung Cancer. N Engl J Med 2018;378:2078-92. [Crossref] [PubMed]
49. Mansfield AS, Herbst RS, de Castro G Jr, et al. Outcomes With Pembrolizumab Monotherapy in Patients With Programmed Death-Ligand 1-Positive NSCLC With Brain Metastases: Pooled Analysis of KEYNOTE-001, 010, 024, and 042. JTO Clin Res Rep 2021;2:100205. [Crossref] [PubMed]
50. Chen M, Wei L, Wang Q, et al. Efficacy of different therapies for brain metastases of non-small cell lung cancer: a systematic review and meta-analysis. Transl Lung Cancer Res 2023;12:689-706. [Crossref] [PubMed]
51. Sun C, Zhou F, Li X, et al. PD-1/PD-L1 Inhibitor Combined with Chemotherapy Can Improve the Survival of Non-Small Cell Lung Cancer Patients with Brain Metastases. Onco Targets Ther 2020;13:12777-86. [Crossref] [PubMed]
52. Takamori S, Komiya T, Powell E. Survival benefit from immunocheckpoint inhibitors in stage IV non-small cell lung cancer patients with brain metastases: A National Cancer Database propensity-matched analysis. Cancer Med 2021;10:923-32. [Crossref] [PubMed]
53. Slowley A, Phiri K, Multani JK, et al. Real-world treatment patterns and clinical outcomes after introduction of immune checkpoint inhibitors: Results from a retrospective chart review of patients with advanced/metastatic non-small cell lung cancer in the EU5. Thorac Cancer 2023;14:2846-58. [Crossref] [PubMed]
54. Decroisette C, Greillier L, Curcio H, et al. Three-Year Overall Survival of Patients With Advanced Non-Small-Cell Lung Cancers With ≥50% PD-L1 Expression Treated With First-Line Pembrolizumab Monotherapy in a Real-World Setting (ESCKEYP GFPC Study). J Immunother 2023; Epub ahead of print. [Crossref]
55. Kemmotsu N, Ninomiya K, Kunimasa K, et al. Low frequency of intracranial progression in advanced NSCLC patients treated with cancer immunotherapies. Int J Cancer 2024;154:169-79. [Crossref] [PubMed]
56. Cantor DJ, Davis C, Ciunci C, et al. Brief Report: Long-Term Follow-Up of Adjuvant Pembrolizumab After Locally Ablative Therapy for Oligometastatic NSCLC. JTO Clin Res Rep 2024;5:100667. [Crossref] [PubMed]
57. Tozuka T, Kitazono S, Sakamoto H, et al. Poor efficacy of anti-programmed cell death-1/ligand 1 monotherapy for non-small cell lung cancer patients with active brain metastases. Thorac Cancer 2020;11:2465-72. [Crossref] [PubMed]
58. Crinò L, Bronte G, Bidoli P, et al. Nivolumab and brain metastases in patients with advanced non-squamous non-small cell lung cancer. Lung Cancer 2019;129:35-40. [Crossref] [PubMed]
59. Yoshida TM, Wang A, Hafler DA. Basic principles of neuroimmunology. Semin Immunopathol 2022;44:685-95. [Crossref] [PubMed]
60. Kim R, Keam B, Kim S, et al. Differences in tumor microenvironments between primary lung tumors and brain metastases in lung cancer patients: therapeutic implications for immune checkpoint inhibitors. BMC Cancer 2019;19:19. [Crossref] [PubMed]
61. Parakh S, Park JJ, Mendis S, et al. Efficacy of anti-PD-1 therapy in patients with melanoma brain metastases. Br J Cancer 2017;116:1558-63. [Crossref] [PubMed]
62. Yan X, Qu F, Zhou Y. Progress of immune checkpoint inhibitors therapy for non-small cell lung cancer with brain metastases. Lung Cancer 2023;184:107322. [Crossref] [PubMed]
63. Li J, Zhang X, Wang Y, et al. Clinicopathological characteristics and prognosis of synchronous brain metastases from non-small cell lung cancer compared with metachronous brain metastases. Front Oncol 2024;14:1400792. [Crossref] [PubMed]
64. Chen K, Zhang F, Fan Y, et al. Lung-molGPA Index Predicts Survival Outcomes of Non-Small-Cell Lung Cancer Patients with Synchronous or Metachronous Brain Metastases. Onco Targets Ther 2020;13:8837-44. [Crossref] [PubMed]
65. Ouyang W, Yu J, Zhou Y, et al. Metachronous Brain Metastasis in patients with EGFR-mutant NSCLC indicates a worse prognosis. J Cancer 2020;11:7283-90. [Crossref] [PubMed]
66. Shibahara I, Kanamori M, Watanabe T, et al. Clinical Features of Precocious, Synchronous, and Metachronous Brain Metastases and the Role of Tumor Resection. World Neurosurg 2018;113:e1-9. [Crossref] [PubMed]
67. Spigel DR, Chaft JE, Gettinger S, et al. FIR: Efficacy, Safety, and Biomarker Analysis of a Phase II Open-Label Study of Atezolizumab in PD-L1-Selected Patients With NSCLC. J Thorac Oncol 2018;13:1733-42. [Crossref] [PubMed]