Background

Lung cancer is the leading cause of cancer-related death worldwide [1]. Non-small cell lung cancer (NSCLC) accounts for approximately 80% of all lung cancer cases, most of which are diagnosed at an advanced stage [2]. Many patients with NSCLC develop brain metastases (BrMs) during their disease course, which is reported in 40% of cases [3,4,5]. NSCLC patients with BrMs have a very poor prognosis and deteriorated quality of life [6, 7]; thus, the brain is an important metastatic site and has attracted particular attention.

Surgical resection serves as a localized therapeutic intervention for BrMs with distinct status and has been performed for selected patients [8, 9]. Common alternative localized treatment modalities encompass whole-brain radiotherapy (WBRT) and stereotactic radiosurgery (SRS). WBRT has been used for over 70 years and is currently prescribed for patients with multiple metastases or a relatively short life expectancy due to WBRT-induced delayed neurological injury. SRS has more recently emerged and is used for patients presenting with a limited number or volume of BrMs. Both methods rely on radiation, thereby enabling the targeting of multiple metastases, surgically unattainable deep lesions, or metastases in patients exhibiting relatively aggressive systemic conditions. As for surgical resection, it is predominantly recommended for patients with a solitary BrM in an accessible region and with controlled systemic disease [10]. Surgical resection offers a notable advantage in the histological diagnosis of targeted lesions; it has been reported that a diagnosis other than malignancy (such as radiation necrosis or abscess) is made in a certain proportion of cases [11]. Furthermore, it confers another important advantage of immediate improvement of the intracranial mass effect, with subsequent improvement of symptoms. Consequently, surgical resection has also been conducted for patients with severe progressive symptoms from BrMs and who require rapid improvement of symptoms. Some studies have shown that patients who undergo surgical resection of BrMs tend to show improvement in performance status (PS) [12,13,14]. However, little is known about the characteristics of patients whose PS could be expected to improve.

The efficacy of systemic therapy for NSCLC BrMs has been demonstrated in several studies [15]. Systemic treatments for NSCLC can be categorized as cytotoxic chemotherapy, tyrosine kinase inhibitors (TKIs) as targeted therapy for driver oncogene-addicted tumors, and immune-checkpoint inhibitors (ICIs). The treatment paradigm has undergone a transformation in recent decades, which has resulted from the clinical implementation of TKIs and ICIs. Each of these therapeutic modalities has shown clinical benefit for patients with BrMs to varying degrees [16, 17], but the optimal subsequent treatment for patients who have undergone surgical resection of BrMs remains uncertain, and the outcomes of each treatment have been poorly reported [18, 19].

This study focused on patients with NSCLC who had undergone surgical resection of BrMs. The aim was to clarify treatment characteristics and outcomes, with particular emphasis on PS changes caused by surgery, and subsequent systemic treatments.

Methods

Study population

This was a retrospective, cohort study including advanced NSCLC patients with BrMs who underwent surgical resection of BrMs of NSCLC between February 1, 2008, and December 31, 2021. Patients who (1) underwent surgical resection of BrMs only for diagnosis (not for treatment or palliation of symptoms) or (2) had missing clinical information before or after BrM resection were excluded. This study protocol was approved by the ethics committee of Kyoto University Hospital.

Clinical variables and outcomes

The investigation focused on the first surgical resection of BrMs in patients who underwent repeated resections. In all patients, Eastern Cooperative Oncology Group (ECOG) performance status (PS) was evaluated immediately before BrM resection. In addition, the ECOG PS after resection was evaluated as the best ECOG PS observed within 90 days after resection. The numbers of BrMs and complete resections were evaluated radiologically and clinically. The TNM classification was determined at the time of BrM resection, based on the 8th edition of the TNM staging system. For patients whose primary tumor had been previously surgically resected, the T and N categories were reported as X due to the unavailability of accurate staging information. Information on driver oncogenes was obtained through a review of electronic medical records; patients who underwent BrM resection in the early chronological period could not have been adequately profiled. Subsequent systemic therapy after BrM resection was defined as the systemic chemotherapy that the patients received after BrM resection and before disease progression or death, regardless of the site of progression. Subsequent systemic therapy was categorized as follows: (1) none; (2) cytotoxic agents; (3) TKIs; and (4) ICIs. Patients who received (2) cytotoxic chemotherapy plus (3) TKIs or (4) ICIs were categorized as (3) TKIs or (4) ICIs, respectively, considering the relatively stronger effect expected. Patients who received subsequent resection of BrM were defined as those who received any prior therapy for NSCLC, including surgical resection of the primary lesion, radiation therapy for the primary lesion or intracranial lesion, systemic chemotherapy, etc. Conversely, patients who received upfront resection of BrM included those who underwent BrM resection as the first treatment for NSCLC. The time between the date of BrM resection and that of death was calculated for each patient to evaluate overall survival (OS). The data cut-off was June 30, 2022.

Statistical analyses

Continuous variables are reported as means and ranges, and categorical variables as numbers and percentages. OS was calculated using Kaplan–Meier estimates and compared using the log-rank test. To investigate the potential factors associated with OS, univariate Cox proportional hazard models were constructed for all clinically relevant factors (age, sex, smoking status, ECOG PS before resection, previous anticancer therapy, complete resection, BrM histology, ECOG PS after resection, and subsequent therapy). The multivariate Cox proportional hazards model included factors that were clinically predicted to be associated with prognosis and factors we focused on (age, timing of resection, ECOG PS after resection, and subsequent therapy). In the comparison of characteristics, ECOG PS change was compared with the Wilcoxon signed-rank test, and the correlation between age and ECOG PS change was analyzed using the Mann–Whitney U test. Dichotomous variables were analyzed using Fisher’s exact test.

For the analyses, a two-tailed p-value of < 0.05 was considered significant. Statistical analyses were conducted using R (version 4.2.2) [20], data processing was conducted using the package dplyr (version 1.1.2) [21], description and analysis of characteristics were conducted using the package gtsummary (version 1.7.2) [22], survival analyses were conducted using the packages survival (version 3.5-7) [23] and survminer (version 0.4.9) [24], and figures were produced using the package ggplot2 (version 3.4.2) [25].

Results

Patients’ characteristics and overall survival

In total, 80 patients with NSCLC underwent surgical resection of BrMs between January 2008 and December 2021. Of them, 9 patients were excluded for the following reasons: 6 for BrM histology showing high-grade neuroendocrine carcinoma; 1 for BrM resection only for diagnosis; 1 for preceding BrM resection in another hospital; and 1 for missing information after BrM resection. Finally, 71 patients who were clinically indicated for BrM resection to alleviate or prevent symptoms were included. The patients’ characteristics are summarized in Table 1.

Table 1 Patient characteristics
Full size table

For all patients, the median OS was 18.3 (95% confidence interval [95% CI]: 8.7 - not reached) months, and the Kaplan-Meier curve is shown in Fig. 1.

Fig. 1
figure 1

Kaplan-Meier curve of overall survival

Full size image

Improvement of ECOG PS

The change in ECOG PS before and after BrM resection is shown in Fig. 2. In summary, 13 (18.3%) and 58 (81.7%) of 71 patients showed an ECOG PS of 0–1 and 2–4 before BrM resection, whereas 28 (39.4%) and 43 (60.6%) patients showed 0–1 and 2–4 after BrM resection, respectively, demonstrating significant improvement (p = 0.006). To further examine the characteristics of patients with improved ECOG PS after resection, patients with an ECOG PS of 2–4 before BrM resection were examined, and the baseline characteristics of patients whose ECOG PS improved to 0–1 after resection were compared to those whose ECOG PS did not improve (summarized in Table 2). The patients who had improved ECOG PS after resection were younger than those with no improvement in PS (median, 62 years versus 66 years; p = 0.041), and other clinically relevant factors showed no significant correlations.

Fig. 2
figure 2

Changes in ECOG performance status (PS) after resection of brain metastases

Full size image
Table 2 Comparison of characteristics for patients with ECOG PS of 2–4 before resection, stratified by ECOG PS after resection
Full size table

Subsequent therapy after BrM resection

As subsequent systemic therapy after BrM resection before progression, 21 patients (29.6%) received cytotoxic chemotherapy, 14 patients (19.7%) received TKIs, 3 patients (4.2%) received ICIs, and 34 patients (47.9%) received no subsequent systemic therapy. Kaplan-Meier curves are summarized in Fig. 3A; there was a trend for longer OS for patients receiving subsequent TKI or ICI therapies (p = 0.12).

Fig. 3
figure 3

Kaplan-Meier curves of overall survival stratified by subsequent therapy after brain metastasis (BrM) resection and before progression for all patients (A) and for patients with no previous anti-cancer therapy before brain metastasis resection (B)

Full size image

To clearly investigate the associations between subsequent systemic therapy and OS, patients who had received no anticancer treatment before BrM resection or who had received BrM resection as the first therapy for NSCLC after clinical diagnosis were selected. Kaplan-Meier curves are summarized in Fig. 3B; there was a trend for longer OS for patients receiving subsequent TKI or ICI therapies (p = 0.057).

As subsequent consolidative radiotherapy, 20 patients (28.2%) received whole brain radiotherapy and 8 patients (11.2%) received stereotactic radiosurgery therapy.

Univariate and multivariate analysis of OS

Univariate analysis for OS, including known clinical prognostic factors and the factors investigated above, was performed. The results are summarized in Table 3, and the ECOG PS after BrM resection, but not for ECOG PS before resection, was correlated with significantly longer OS (p = 0.007). Other clinically known prognostic factors showed no significant correlations with OS. The Kaplan-Meier curves of OS stratified by ECOG-PS after BrM resection are summarized in Fig. 4.

Table 3 Univariate and multivariate analyses of overall survival time
Full size table
Fig. 4
figure 4

Kaplan-Meier curves of overall survival stratified by ECOG performance status (PS) after brain metastasis (BrM) resection

Full size image

Furthermore, we conducted multivariate analysis; however, due to only 40 events occurred, the number of explanatory variables that could be considered was limited. We incorporated variables that were considered major prognostic factors (age, timing of resection [upfront or subsequent]) and factors of interest (ECOG PS after resection, subsequent therapy) into the model as explanatory variables. In this analysis, better ECOG PS after resection and subsequent TKI therapy were shown to be significantly associated with longer OS (Table 3).

Discussion

This investigation focused on NSCLC patients with surgically resected BrMs who showed an improvement in PS after BrM resection, which is more frequently observed in younger patients, and the associations between subsequent systemic therapy and prognosis.

In this study, the median OS of NSCLC patients with a surgically resected BrM was 18.3 months. This result aligns with historical controls [13, 26,27,28,29,30,31] from retrospective cohorts of NSCLC patients with resected BrM, which reported median OS ranging from 12 to 20 months. During earlier periods when the only possible intervention for BrM consisted of best supportive care, the prognosis of patients with BrM was 1–2 months [32]. With the introduction of WBRT in the 1950s, it improved to 4–6 months [10]. In recent years, the prognosis for NSCLC patients with BrMs has much improved, to 5–8 months [33]. Considering the inevitable bias of the present study in which patients in good general condition who could be considered for an invasive procedure, surgical resection of BrM, were selected, it is not surprising that patients who underwent surgical resection of BrMs showed a better prognosis.

NSCLC patients who underwent surgical resection of BrMs demonstrated improvements in PS, which is comparable to previous studies [12,13,14]. From the further analysis of patients who had poor ECOG PS prior to BrM resection, those who experienced PS improvement postoperatively had a younger age than those who showed no PS improvement. To the best of our knowledge, no previous study investigated the baseline characteristics of patients who tend to show PS improvement after BrM resection. It is plausible, given that a younger age has been identified as a factor associated with a good prognosis for patients who underwent surgical resection of BrMs [30, 34]. The cause is unknown, but it could result from superior baseline status or preserved functional capacity prior to the appearance of symptoms, encompassing comorbidities unrelated to NSCLC. On univariate analyses for OS, ECOG PS after BrM resection, rather than preoperative values, was associated with a better prognosis. Previous studies focused mainly on ECOG PS before BrM resection and posited it as a potential factor associated with a good prognosis [28, 30, 35, 36]. The reason for this discrepancy is unknown, but it could be influenced by the diverse characteristics of the participating patients. Specifically, the selection criteria for surgical resection could largely rely on medical accessibility or the collective agreement of medical practitioners within various countries and regions. In summary, younger patients with symptomatic and operable BrMs could be deemed suitable candidates for surgical resection to alleviate symptoms and improve their prognoses, although further investigations are needed.

Advances in systemic therapy could affect the outcomes of NSCLC patients with resected BrMs. In previous studies, TKIs or ICIs were reported to be associated with superior outcomes in selected NSCLC patients with BrMs which were not surgically resected [18, 19]. The results of the present study aligned with these studies, and the present one specifically focused on the prognosis of NSCLC patients with surgically resected BrMs for the first time. However, the present cohort had an inevitable bias, as those who underwent BrM resection with or without preceding systemic treatment were included. To circumvent this bias, patients who received BrM resection as the initial therapy for NSCLC were examined, and more pronounced associations between subsequent systemic treatments and survival outcomes were found. Of course, the results could not be interpreted simply as causal connections between subsequent systemic treatment and survival outcomes because of the inevitable selection bias; patients who could receive only cytotoxic chemotherapy or no systemic therapy must have some background characteristics precluding access to other treatments. Moreover, patients who had undergone BrM resection in previous decades lacked comprehensive genetic profiling of their tumors and no chance to receive TKI or ICI treatment. The effects of TKIs or ICIs for BrMs have been widely studied in preclinical and clinical retrospective studies, but they are seldom investigated in prospective clinical trials, since patients with BrMs are excluded from the majority of trials. Consequently, real-world studies are still needed to investigate and optimize therapy for NSCLC patients with BrMs.

The present study had several limitations. First, it was a retrospective study and included only patients who underwent surgical resection of BrMs. The background characteristics of patients undergoing BrM resection were heterogeneous, which could affect the results and their interpretations. Furthermore, we acknowledge that patients selected for surgical resection of brain metastases likely had better overall health and prognosis compared to the general population of NSCLC patients with BrMs. This introduces a potential selection bias, which may limit the generalizability of our findings to all NSCLC patients with BrMs. Second, given that patients who underwent BrM resection over an extended period (2009–2022) were included, genomic profiling of tumors was not conducted uniformly. Additionally, during this extended study period, treatment protocols and diagnostic methods underwent significant advancements, which could have potentially affected the observed outcomes. Third, the present study included a limited number of participants. Consequently, the statistical power could be limited. Fourth, our study design did not include a non-surgical control group, which limits our ability to directly compare the effectiveness of surgical resection against other treatment modalities for NSCLC brain metastases. This was intentional, as our primary aim was to describe outcomes in surgically treated patients, with a focus on post-operative performance status changes and subsequent treatments. Fifth, due to the small number of patients in each subgroup receiving subsequent systemic therapies, we were unable to provide detailed statistical comparisons for the survival outcomes across these therapies. This limitation in sample size for subgroup analysis restricts our ability to draw definitive conclusions about the relative efficacy of various systemic treatments following BrM resection. Future studies with larger cohorts would be valuable to further elucidate the impact of specific systemic therapies on survival outcomes in this patient population.

Conclusions

In this study, the prognosis and characteristics of NSCLC patients with BrMs who underwent surgical resection of BrMs were investigated. The results suggested that PS improvement after BrM resection could be expected, particularly in younger patients. Recent advances in systemic treatment such as TKIs and ICIs could potentially correlate with survival outcomes.