Tyrosine kinase inhibitors in HER2‐positive breast cancer brain metastases: A systematic review and meta‐analysis

Abstract Background Small tyrosine kinase inhibitors (TKIs) show activity against breast cancer brain metastases (BCBM) of the human epidermal growth factor receptor 2 (HER2)‐positive subtype. This meta‐analysis aimed to objectively explore the efficacy and safety of TKIs. Methods Electronic databases were searched for relevant clinical trials. We conducted a pairwise meta‐analysis, pooled analysis, and estimated summary survival curves to compare survival outcomes following TKIs therapy for BCBM patients using Stata version 16.0 or R x64 4.0.5. Results Thirteen clinical trials involving 987 HER2‐positive BCBM patients were analyzed. A trend of longer progression‐free survival (PFS) was observed in the TKI‐containing arm compared to the non‐TKI‐containing arm (hazard ratio = 0.64, 95% confidence interval [CI]: 0.35–1.15, p = 0.132), although the difference is not statistically significant. Summary survival curves reported the summary median PFS and overall survival were 7.9 months and 12.3 months. Subgroup analysis revealed that TKIs combined with capecitabine (TKI + Cap) regimens resulted in improved survival outcomes. Tucatinib may be more effective in BCBM patients. The main grade 3–5 adverse events (AEs) were diarrhea (22%, 95% CI: 14%–32%), neutropenia (11%, 95% CI: 5%–18%), hepatic toxicity (7%, 95% CI: 1%–16%), and sensory neuropathy (6%, 95% CI: 2%–12%). Conclusion TKIs therapy improved the survival outcomes of HER2‐positive BCBM patients, especially when combined with capecitabine and tolerable AEs. We also identified the clinical value of tucatinib, which appears to be the most favorable TKI drug for BCBM patients.


| BACKGROUND
Breast cancer (BC) is the most common cancer in women. 1 The incidence of BC brain metastasis (BCBM) has increased in recent years. BC is the second most common cause of brain metastases, accounts for 15%-20% of all brain metastases. 2,3 The human epidermal growth factor receptor 2 (HER2) amplified subtype is often considered to have a higher incidence of BM than other molecular subtypes. 4,5 The survival rate of BCBM patients is extremely low. The registHER study found that the median overall survival (OS) from the time of metastatic BC (MBC) diagnosis in patients with any central nervous system (CNS) event was significantly lower than that in patients with no CNS events (26.3 months vs. 44.6 months). 6 The treatment modalities for BCBM include whole-brain radiation therapy (WBRT), stereotactic radiosurgery (SRS), and surgery. Unfortunately, the time to intracranial progression after these therapies is approximately 6-12 months. [7][8][9] The European Association of Neuro-Oncology recommends that surgery be considered in patients with 1-3 intracranial metastases lesions, especially for lesions at least 3 cm in diameter. 10 Traditionally, WBRT has been recommended for patients with multiple BM; however, it provides disappointing local tumor control and poor clinical survival outcomes. Due to these limitations, SRS using a Gamma Knife, Linac, or Cyberknife has been proposed. 11 However, SRS may cause cognitive complications including memory impairment, seizures, cognitive disturbance, motor weakness, and ataxia. 12 Thus, systemic treatment is often combined with local treatment modalities that have poor efficacy.
The use of small-molecule tyrosine kinase inhibitors (TKIs), which inhibit both HER2 and EGFR targets, may improve patient outcomes. 13 TKIs bind to the intracellular domains of HER2, block tyrosine phosphorylation and downstream signaling events of ligand binding, and compete with ATP at the cytoplasmic catalytic kinase domain. They also have high blood-brain barrier (BBB) penetrability, which may better control intracranial lesions and produce little neurotoxicity. 14 Several TKIs, including lapatinib, neratinib, afatinib, pyrotinib, and tucatinib, have become available and are currently under investigation. In the pivotal single-group phase 2 study named LANDSCAPE, patients were included who had HER2positive BCBM and did not have previous WBRT, lapatinib, or capecitabine previously. The combination of lapatinib plus capecitabine improved CNS objective response rate (ORR) (65.9%, 95% CI: 50.1-79.5) compared with lapatinib alone. The results showed that lapatinib plus capecitabine could be recommended as a treatment for patients with CBM; however, it may be at the expense of relatively high toxicity. 15 New-generation TKIs, including neratinib, pyrotinib, and tucatinib, are promising agents for the treatment of HER2-positive BCBM. [16][17][18] Consequently, the role of TKI in the treatment of BM in HER2-positive BC patients has been gaining attention.
A debate is ongoing over the use of TKIs for treating HER2-positive BCBM owing to a lack of sufficient clinical evidence, mainly a preponderance of results from singlearm clinical trials and existing randomized controlled trials (RCTs) that reported mixed results. Questions have also been raised owing to differences in the baseline characteristics of patients in various clinical trials. In addition, there is no consensus on the optimal combination drug with TKIs and the TKI drug with better efficacy. A complete analysis of all relevant clinical studies is important to provide useful clinical references. Therefore, we conducted this metaanalysis using the latest data to provide a comprehensive overview of TKIs therapy for HER2-positive BCBM patients.

| METHODS
Our study was registered on INPLASY (INPLASY202170064) and the details are available at inpla sy.com (https://doi. org/10.37766/ inpla sy2021.7.0064). The protocol for this systematic review and meta-analysis has not been previously published in peer-reviewed journals.

| Selection criteria
We developed the eligibility and exclusion criteria for this study in advance. All clinical trials (RCTs or single-arm trials) investigated TKI regimens for the treatment of HER2positive BC patients with brain metastases. A subgroup analysis of patients with brain metastases within relevant trials was also performed in our study. Survival or tumor response data were available for all patients. The studies were grouped based on treatment regimens, regardless of active or inactive brain metastases, treatment lines, or prior treatment. Phase  were included in the analysis. Case reports, systematic reviews, and retrospective studies were excluded from the analysis. In case of multiple publications from the same trial, only the most recent publication was included.

| Search strategy
This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Searches were performed on PubMed, Embase, and Cochrane Central Register of Clinical Trials, utilizing the following keywords: "(breast OR mammary) AND (cancer OR carcinoma OR malignant OR neoplasm OR tumor) AND (HER-2 OR HER2 OR HER2/neu OR ERBB2 OR human epidermal growth factor receptor 2) AND (positive OR +) AND (brain OR central nervous system) AND (metastasis OR metastases OR metastatic) AND (smallmolecule tyrosine kinase inhibitors OR lapatinib OR neratinib OR afatinib OR tucatinib OR pyrotinib)", only in English, without any limitations on factors such as time or race. We also assessed the ASCO and ESMO annual meetings and SABCS online articles, and included articles published before January 1, 2023. The references of all studies that met the eligibility criteria were examined for other relevant studies, especially for relevant meta-analyses.

| Data extraction
Two investigators (YSY and JZ) independently extracted the data, and all included items required consensus. The first author, year of publication, trial name, locations, trial phase, number of analyzed participants, therapy regimens, therapy details, Eastern Cooperative Oncology Group Performance Status, and the efficacy endpoints of each eligible trial were recorded. The primary outcome was progression-free survival (PFS). The secondary outcomes were CNS ORR, median PFS, median OS, and grade 3-5 adverse events (AEs). All data were recorded from the original articles or using the GetData Graph Digitizer if the original statistical curve pictures were acquired from the articles.

| Definition of outcomes
Outcome analyses, including PFS, CNS ORR, median PFS, and median OS, were conducted based on the investigator-assessed response according to RECIST 1.1. The data have also been corrected and are indicated in the Characteristics of the Trials. Neutropenia, vomiting/nausea, diarrhea, stomatitis mucositis, skin and subcutaneous tissue disorders, sensory neuropathy, and hepatic toxicity were considered as the most crucial side effects. Grade 3-5 AEs were calculated using the National Cancer Institute Common Terminology Criteria version 4.0.

| Statistical methods
We conducted a pairwise meta-analysis to evaluate PFS as hazard ratios (HRs) with corresponding 95% confidence intervals (95% CI) between TKI-containing and non-TKI-containing regimens using available data. Owing to limited data, single-arm trials were also included in our study to better describe the efficacy of TKIs in BCBM patients. Because no HRs or risk ratios could be produced with these single-arm studies, we conducted them in three steps to better demonstrate their effectiveness. First, we attempted to estimate tumor response and survival results using a pooled analysis from a quantitative perspective. Continuous variables, including median PFS and median OS, are presented as pooled mean values. Dichotomous variables, including CNS ORR and grade 3-5 AEs incidence rates, were presented as pooled effect sizes with 95% CIs. Second, to indicate the analysis of time survived to an outcome event that cannot be reflected in pooled analysis, a R package entitled MetaSurv was used to estimate "summary survival curves" from single-arm studies' K-M curves by the product-limit estimator method, then the summary median survival can be calculated based on the summary survival curve. 19 All data were extracted using the GetData Graph Digitizer if the original statistical curve pictures were acquired from the articles or supplementary materials. Thirdly, subgroup analysis was performed using pooled analysis or estimating "summary survival curves." We planned a subgroup analysis for the following subgroups 1 : to explore the treatment effect of TKIs with different combinations of drugs across subgroups, subgroup analysis was performed using the following classification variables: TKIs were used alone (monotherapy), TKIs were combined with capecitabine (TKI + Cap), TKIs were combined with other drugs except capecitabine (TKI + Others) 2 ; to explore the treatment effect of different types of TKI drugs including lapatinib, tucatinib, neratinib, and pyrotinib.
Cochran's Q chi-square test was used to examine the heterogeneity across studies, and inconsistencies were tested using the I 2 statistic. We used a fixed-effects model for the analysis if the pooled results showed low heterogeneity (I 2 ≤ 50%). Conversely, a random-effects model was used for significant heterogeneity (I 2 > 50%). Begg's and Egger's tests were used to examine the potential publication bias of the included clinical trials. All statistical analyses were performed using Stata version 16.0 or R x64 4.0.5, and statistical significance was set at p < 0.05.

| Quality assessment
The study quality of the five included RCTs was assessed using the Cochrane risk of bias tool. Eight single-arm studies were assessed using the MINORS index scored. 20 We included clinical trials that could ensure their high quality and integrity owing to the high heterogeneity that may appear in single-arm meta-analyses, and a randomized effect model was applied to minimize the bias.

| Patient and public involvement
The patients and the public were not involved in the design, conduct, reporting, or dissemination plans of our research.

| Overview of literature search and study characteristics
The initial search revealed 670 potentially relevant manuscripts and one additional abstract. After a detailed review of the titles and abstracts, 548 manuscripts were excluded. The full texts of the remaining 123 articles were read, and a total of 14 articles from 13 trials-including five RCTs and eight single-arm clinical trials that met the requirements-were obtained according to the inclusion and exclusion criteria ( Figure 1). [15][16][17][21][22][23][24][25][26][27][28][29][30][31] All details of the included trials are described in Table S1. The corresponding PRISMA checklist is shown in Appendix S1. Although the study by Krop et al. was a retrospective study, the results were from a high-quality multicenter, randomized, open-label, phase III trial called EMILIA. 21 Because of the high quality of the study, it was included in our analysis. Among the subgroups, four articles incorporated the results of TKIs for HER2-positive BCBM, and one article met the eligibility criteria for this metaanalysis from the abstracts published in ASCO. All eligible studies, which included 987 patients, were published between 2008 and 2022.

| Quality assessment
The chart of bias assessment showed a low-risk bias in all included trials (Appendix S2). Two of the five randomized controlled studies generated detailed random sequences. Because most of the studies (3/5) had openlabel designs, performance bias might have existed, but did not affect the final outcome. All the RCTs provided complete outcome data without bias. However, no allocation concealment or blinding methods have been proposed. The MINORS index score was between 14 and 15 points in eight single studies, which was appropriate for the current study.

| Pairwise meta-analysis for PFS
Five RCTs compared the HRs of PFS compared TKIcontaining regimens with non-TKI-containing regimens. The pooled analysis results are provided in Figure 2, which revealed a PFS trend favoring TKIcontaining regimens was observed in BCBM patients but without statistical significance (HR = 0.64, 95% CI: 0.35-1.15; p = 0.132) as calculated by the random-effects model.

| "Summary survival curves" in singlearm studies
Five arms from four trials reported the PFS K-M curves, and the summary PFS and 95% CI curves are shown in Figure 4A with the fixed-effects model. We also found favorable efficacy of TKI-containing regimens and TKI + Cap-containing regimen subgroups, based on the analysis of survival to an outcome event. The summarized PFS curves of patients treated with TKIs have showed the summary median PFS time was 7.9 months (95% CI: 6.2-10.0; I 2 = 0.00%). While TKI + Cap-containing regimens subgroup showed an obviously higher summary median PFS time of 9.1 months (95% CI: 6.8-10.6; I 2 = 0.00%) ( Figure 4C). The OS K-M curves were presented in five trials; the summary median OS time using the fixed-effects model was 12.3 months (95% CI: 10.0-14.4; I 2 = 0.00%) with TKI-containing regimens treatment ( Figure 4B). While four studies presented the OS K-M curves, the F I G U R E 1 Flowchart diagram of selected studies included in the metaanalysis.

| DISCUSSION
Survival outcomes of patients with HER2-positive BC have greatly improved with the availability of additional treatment approaches. However, an increase in the incidence of CBM has been reported over the last few years. This may be due to the combination of poor penetrability of the BBB to chemotherapy drugs, monoclonal antibodies, and antibody-dependent cell-mediated cytotoxicity (ADCC) drugs, allowing the CNS to become a "shelter" for cancer cells, and advanced medical imaging technology becoming better at detecting brain metastatic lesions. 32,33 The presupposition for disease control and overall efficacy in treating CBM is adequate target engagement and drug penetration into the brain and tumors. 34 Among the trials included in this meta-analysis, trastuzumab was included in most previous lines of therapy, and the majority of patients had previously received brain radiation therapy, including WBRT or SRS. Our meta-analysis showed that TKI-containing regimens are an effective treatment protocol for HER2-positive BCBM patients, with favorable summary median PFS and OS (7.9 months, 12.3 months; respectively). Small-molecule TKIs have a lower molecular weight, which allows them to effectively penetrate the BBB. 35 In addition, TKI drugs inhibit both constitutive and ligand-induced ErbB signaling compared to monoclonal antibodies. 35 The unique physicochemical properties mentioned above have produced advantageous results. The efficacy of TKIs has been confirmed in animal models of preclinical HER2-positive BCBM animal models. Gril et al. found that animal models treated with lapatinib had fewer large metastatic sites than those of patients treated with vehicle. 36 Furthermore, clinical trials have F I G U R E 5 Pooled grade 3-5 adverse events in HER2-positive breast cancer brain metastases patients included.
revealed the efficacy and toxicity of TKI-containing regimens in HER2-positive BCBM patients. 16,17,22,37 Subgroup analysis showed that the TKI + Cap regimens had a more substantial benefit with a favorable CNS ORR of 43% compared with that of TKI (monotherapy) (5%) and TKI + Others (11%) and produced longer median PFS and OS. These results are consistent with those of a phase II trial that assessed lapatinib efficacy and reported a CNS ORR of 6%. Of the 50 patients who entered the capecitabine extension phase, 20% achieved an objective CNS response. 29 Chefrour et al. attempted to elucidate the reason for the improved efficacy of this combination through molecular determinants of the response. They indicated that lapatinib and capecitabine could regulate each other's molecular determinants of response, and this synergistic interaction was also observed in subcutaneous BC mouse models. 38 All of these results suggest that of all TKI-containing strategies (TKI Monotherapy or combined with other drugs), the combination of TKI and capecitabine is the most effective treatment method for HER2positive BCBM patients.
The TKIs available for clinical practice or under investigation mainly include lapatinib, neratinib, afatinib, pyrotinib, and tucatinib. However, the optimal treatment for HER2-positive BCBM remains controversial. We performed a subgroup analysis to explore the efficacy of the different TKIs. Owing to limited data, we could not perform any summary analysis apart from Lapatinibcontaining regimens. However, the HER2CLIMB showed significantly better outcomes of a tucatinib-containing regimen of 9.9 months for PFS and 18.1 months for OS compared with those of lapatinib-containing regimens of 5.90 months for PFS and 12.99 months for OS. 39 Jing et al. studied the advantages of tucatinib using mechanistic modeling of CNS pharmacokinetics and target engagement of TKIs, including lapatinib, neratinib, and tucatinib. As revealed by the physiologically based pharmacokinetic model approach, tucatinib can induce sufficient HER2 inhibition in both disrupted and largely intact BBB; conversely, lapatinib and neratinib may not achieve adequate target inhibition. 34 These mechanistic models are in line with the available efficacy data, and our results support the pharmacological value of tucatinib for the treatment of HER2-positive BCBM patients. Therefore, based on the current data, tucatinib may be the most effective TKI.
HER2-targeting antibody therapies, including trastuzumab, trastuzumab emtansine (T-DM1), trastuzumab deruxtecan (T-DXd), and pertuzumab, were discovered in a medical history. Generally, they have limited roles in treating BM as first-line therapy because of the BBB. However, recent studies have suggested the opposite view, as some clinical trials have shown that large-molecule drugs can be effective treatments for BCBM. [40][41][42][43][44] The results of preclinical trials may explain these findings. The BBB can be disturbed during brain metastasis and replaced by the blood-tumor barrier, which has better drug permeability. 45,46 Dijkers et al. found that 89 Zr-trastuzumab was detected in brain lesions at an 18-fold higher concentration than that in normal brain tissue, which supports the feasibility of HER2-targeting therapy in patients with BCBM. 47 Clinical trial results also support this phenomenon, showing a CNS response to ADCC drugs. KAMILLA, a single-arm phase IIIb trial, evaluated the efficacy of T-DM1 in patients with metastatic BC, and a subgroup analysis demonstrated that T-DM1 extended PFS and OS in patients with baseline BM. 42 T-DM1 is no longer the newest HER2-targeted ADCC drug, T-DXd. T-DXd also appeared to affect BCBM. In the randomized DESTINY-Breast03 study, T-DXd was found to significantly increase PFS in patients with BM. 43,44 However, these data were compared in patients with stable brain metastases at baseline. Therefore, a single-arm phase II trial called TUXEDO-1 enrolled patients with HER2-positive BC with active BM, and the results showed that T-DXd still had a remarkable CNS ORR (73.3%). 43 However, direct head-to-head comparisons of TKI and ADCC drugs are lacking. Currently, TKIs play a leading role in the treatment of BCBM patients.
To our knowledge, this study is the first to assess the efficacy and safety of TKI-containing regimens for the treatment of HER2-positive BCBM. Our meta-analysis included five RCTs and eight single-arm clinical trials of high quality and integrity. A pairwise meta-analysis was performed to compare the efficacy of TKI-containing regimens and non-TKI regimens. Pooled analyses and summary survival curves obtained from single-arm studies provided quantitative and intuitive results. However, trial design discrepancies between different studies can account for the high heterogeneity of pooled analyses in single-arm studies. Subgroup analysis was used to reduce heterogeneity and yielded reliable results.

| CONCLUSION
TKI therapy improves the survival outcomes of HER2positive BCBM patients, especially when combined with capecitabine and tolerable AEs. We also paid attention to the clinical value of tucatinib, which appears to be the most favorable TKI drug for BCBM patients.

FUNDING INFORMATION
This study was not supported by any financial or personal relationships with other people or organizations.

CONFLICT OF INTEREST STATEMENT
The authors have declared that no competing interests exist.

DATA AVAILABILITY STATEMENT
All data generated or analyzed in this study are included in the published article [and its supplementary information files].

PATIENT CONSENT
This article does not include studies with human participants conducted by any author.

INFORMED CONSENT
As this study contained data released from the published literature, informed consent was not required.