Factors improving overall survival in breast cancer patients with leptomeningeal disease (LMD): A single institutional retrospective review

Background Breast cancer-related leptomeningeal disease (BC-LMD) is a dire diagnosis for 5–8% of patients with breast cancer (BC). We conducted a retrospective review of BC-LMD patients diagnosed at Moffitt Cancer Center (MCC) from 2011–2020, to determine the changing incidence of BC-LMD, which factors impact progression of BC CNS metastasis to BC-LMD, and which factors affect OS for patients with BC-LMD Methods Patients with BC and brain/spinal metastatic disease were identified. For those who eventually developed BC-LMD, we used Kaplan-Meier survival curve, log-rank test, univariable, and multivariate Cox proportional hazards regression model to identify factors affecting time from CNS metastasis to BC-LMD and OS. Results 128 cases of BC-LMD were identified. The proportion of BC-LMD to total BC patients was higher between 2016–2020 when compared to 2011–2015. Patients with HR + or HER2 + BC experienced longer times between CNS metastasis and LMD than patients with triple-negative breast cancer (TNBC). Systemic therapy and whole-brain radiation therapy (WBRT) prolonged progression to LMD in all patients. Hormone therapy in patients with HR + BC delayed BC-CNS metastasis to LMD progression. Lapatinib delayed progression to LMD in patients with HER2 + BC. Patients with TNBC-LMD had shorter OS compared to those with HR + and HER2 + BC-LMD. Systemic therapy, intrathecal (IT) therapy, and WBRT prolonged survival for all patients. Lapatinib and trastuzumab improved OS in patients with HER2 + BC-LMD. Conclusions Increasing rates of BC-LMD provide treatment challenges and opportunities for clinical trials. Trials testing lapatinib and/or similar tyrosine kinase inhibitors, IT therapies, and combination treatments are urgently needed.


Introduction
Leptomeningeal disease (LMD) is a dreadful complication occurring in approximately 5-8% of patients with breast cancer (BC) 1 . Median overall survival (OS) in untreated patients with LMD including from BC is approximately 1 month 1,2 . Aggressive treatment in breast cancer-related leptomeningeal disease (BC-LMD) can extend OS to 3-4 months, although this is BC subtype-speci c 3,4 . Beyond receptor subtype strati cation and examining the e cacy of HER2-targeted therapy, only a few studies have attempted to identify the clinical characteristics of BC-LMD [5][6][7] .
An earlier diagnosis of LMD is important to improve patient survival and enrollment in clinical trials.
However, Magnetic Resonance Imaging (MRI) has a wide variability to accurately diagnose LMD, 8 and the current gold standard of a positive CSF cytology has a very low sensitivity (< 50%) and positive predictive value. 9 To optimize the diagnostic yield beyond traditional CSF cytology, assays that detect circulating tumor cells and cell-free DNA have been developed over the past decade, but the clinical use of these new techniques to diagnose and treat LMD has not yet been de ned. 10 Breast cancer in the central nervous system (BC CNS) has been under-recognized, and the real epidemiological data remains under-reported, especially for those who develop BC-LMD. 11

Methods
This project was approved by the Scienti c Review Committee and Institutional Review Board at MCC (MCC #21524). A retrospective review of medical records was conducted to identify patients diagnosed and/or treated at MCC with BC and who also had a diagnosis of CNS metastases between January 1, 2011 and December 31, 2020. Only patients with con rmed diagnosis of BC and LMD were included. LMD diagnosis was con rmed with CSF cytology and/or MRI. Patients were excluded if the LMD diagnosis was based solely on clinical suspicion, if con rmatory MRI or CSF cytology was not available in the medical record, or if there was another malignancy which might seed the leptomeningeal space.

Data Collection
Demographics, BC receptor subtype, dates of BC/CNS metastasis/LMD diagnoses, and dates of censorship/death were collected. Method of diagnosis, treatments prior to and following CNS metastasis, and treatments following BC-LMD were also collected.

Statistical Analysis
Descriptive statistics including frequency, percentage, median, and range were calculated for patients' demographics and clinical characteristics. Differences in continuous variables between patient groups were statistically tested using the Kruskal-Wallis tests. The associations between categorical variables and endpoints were evaluated using Chi-square test or Fisher's exact test.
Median time between CNS metastatic disease and LMD, and OS post-LMD were estimated using the Kaplan-Meier method. Univariable Cox proportional hazards regression analysis was used to estimate hazard ratios and their 95% con dence interval (CI). Signi cant variables at univariable analysis were subsequently tested in multivariable Cox regression analysis. All reported p values were two-sided, and signi cance level was 0.05 (p < 0.05). Analyses were performed using R version 4.1.0.

Cases and Demographics
One-hundred-twenty-eight patients were identi ed who met radiographic and/or CSF cytology criteria for BC-LMD diagnosis. Forty patients were identi ed between 2011-2015, whereas 88 patients were identi ed between 2016-2020 (Fig. 1A). The proportion of BC-LMD patients to total BC patients was signi cantly higher between 2016-2020 when compared to 2011-2015 (p = 0.0168, Fig. 1B Out of all the patients identi ed, 100 (78%) had systemic metastasis prior to the diagnosis of LMD, and 114 (89%) had CNS metastasis prior to the diagnosis of LMD. Patients with HR + and HER2 + BC were more likely to develop systemic metastasis prior to the diagnosis of LMD when compared to patients with TNBC (i.e., 86% and 79%, respectively, versus 57%; p = 0.007). All three patient cohorts were equally likely to develop CNS metastatic disease prior to the diagnosis of LMD. Patients' demographic data is summarized in Table 1.

Factors Affecting Time between CNS Metastasis Diagnosis and LMD Diagnosis
Only patients who had at least a one-month gap between CNS metastasis and LMD diagnoses were included in the analysis. Of the 48 BC-LMD patients that met this criterion, 19 had HR + BC (40%), 19 HER2 + BC (40%), and 10 TNBC (20%). Demographics for these patients is summarized in Table 2, and post-CNS metastasis/pre-LMD treatment data is summarized in Supplementary Table 1.
Patients with HR + BC who received any hormone therapy had signi cantly longer times between their CNS metastasis diagnosis and LMD diagnosis ( Supplementary Fig. 1A, p = 0.0005). More speci cally, HR + patients that were treated with letrozole had a longer time to LMD diagnosis from CNS metastasis diagnosis ( Supplementary Fig. 1B, p = 0.0007). Similar trends were observed in HR + BC patients who were treated with the aromatase inhibitor exemestane ( Supplementary Fig. 1C, p = 0.0015) and CDK4/6 inhibitor palbociclib ( Supplementary Fig. 1D, p = 0.023). Finally, the mTOR inhibitor everolimus also showed prolonged time between CNS metastasis and LMD diagnosis in HR + BC patients ( Supplementary  Fig. 1E, p = .0013). Of all these treatments, exemestane and palbociclib also showed signi cance in multivariate analysis ( Supplementary Fig. 1F).
Anti-HER2 therapy lapatinib was associated with longer times between CNS metastasis and BC-LMD ( Supplementary Fig. 2, p = 0.026) for HER2 + BC-LMD patients when compared to patients of any subtype who did not receive this treatment. Unfortunately, no systemic therapy was found to be associated with signi cantly longer time between CNS metastasis diagnosis to LMD diagnosis in TNBC patients.
Finally, the effects of brain tumor excision and radiosurgery were also analyzed, but patients receiving these treatments experienced no differences in the time between CNS metastasis and LMD when compared to patients who did not receive such treatments ( Supplementary Fig. 3).

OS with BC-LMD
Treatment data for patients post-LMD diagnosis is summarized in Supplementary Table 2 Furthermore, 55 patients received IT therapy post-LMD diagnosis, and they had a median OS of 11.8 months [8.2, 14.5]. This was signi cantly greater than the 1.9 month [1.4, 2.7] month median OS of patients that did not receive IT therapy (Fig. 3C, p < 0.0001). Speci cally, patients receiving IT methotrexate had a 5.5 month longer median OS than patients who did not (Fig. 3D, p = 0.0096), and patients receiving IT thiotepa had a median OS that was 8.4 months longer than those who did not (Fig. 3E, p = 0.0354). Improved OS was also observed in the 8 patients who received only IT therapy for treatment post LMD diagnosis. Their median OS of 11.24 [0. 21,19.27] months was signi cantly greater than the OS of the 22 patients that did not receive any therapy post BC-LMD diagnosis (Fig. 4, p = 0.041).
Finally, 61 patients receiving WBRT survived a median of 6.5 [4.7, 9.8] months while 67 patients who did not receive WBRT survived a median of 2.7 months [1.7, 5.3] (Fig. 3F, p = 0.0274). However, The OS for patients that received only WBRT post-LMD diagnosis was 1.86 months [0.39, 10.11], which was not signi cantly greater than the OS for patients that received no therapy (Fig. 4).  Fig. 5A-C, p = 0.015, 0.0261 and 0.0352, respectively). HER2 + BC-LMD patients receiving IT therapy had an improved median OS by 11.1 months when compared to HER2 + BC-LMD patients that did not receive IT therapy ( Supplementary Fig. 5D, p < 0.0001). Importantly, all HER2 + patients that received IT therapy received IT trastuzumab as a part of their treatment regimen.
Although treatment with systemic therapy collectively enhanced survival in patients with TNBC when compared to those patients who did not receive systemic therapy (3.4 months versus 1.2 months, Supplementary Fig. 6A, p = 0.0334), no individual therapy was found to be associated with prolonged survival in this subset of patients. However, IT therapy was found to increase OS in TNBC patients compared those that did not receive the therapy (8.9 months versus 1.1 months, Supplementary Fig. 6B, p = 0.001).

Discussion
Data from MCC supports a signi cant rise in the number of BC-LMD cases between 2011-2020. Greater institutional efforts to identify and treat LMD patients were made in the latter half of the study period, which may have driven the increased in BC-LMD diagnosis. A slight trend towards increased proportions of patients with TNBC were seen in the current review, but all cases increased over the study period. One of the rst to report this, showed that HR status in uences the risk of developing LMD. 12 This is supported by recent data suggesting HR + BC-LMD likely comprises the majority of all cases (range 48-66%), and rates of HER2 + and TNBC-LMD are more variable. 12-16 HER2 + BC-LMD represents 14-47.4% of all cases, and TNBC-LMD varies between 13.1-40% across studies. [12][13][14][15][16] The proportion of HR + patients was similar in our study at 51% compared to HER2 + BC-LMD (27%) and TNBC (22%). BC subtype affected time between CNS metastasis and BC-LMD diagnoses. TNBC had the fastest progression to BC-LMD compared to HR + and HER2 + BC-LMD. Even though HER2 + cancer may have a predisposition for CNS invasion, a tendency to invade the leptomeninges has not been clinically revealed. 1,17 It has been shown that increased survival in patients with BC generally correlates with increased incidence of CNS metastases. 18 Prolonged survival and risk of developing metastases, may confer a greater risk of developing BC-LMD. 14 Historically, HER2 + and TNBC have carried an increased risk of developing CNS metastasis, and an associated decreased survival relative to HR + disease. 19,20 Earlier work suggested no difference among BC-LMD based on the molecular subtype. 21 However, our ndings support more recent studies demonstrating an increased risk of developing LMD for TNBC. 12,13,15,16,22 This study also showed the use of any systemic treatment and/or WBRT post-CNS metastasis delays progression from to BC-LMD. HR + patients receiving hormone therapies and/or kinase inhibitors may experience prolonged times between CNS metastasis and LMD diagnosis. For HER2 + patients, treatment with lapatinib demonstrated similar results. However, the studies stringent criterion for stratifying patients by at least a one-month period between CNS metastasis and LMD severely reduced the total patient sample from 128 to 48 patients. Follow-up studies analyzing signi cantly larger patient samples is required to de nitively determine factors that prolong time between CNS metastasis and BC-LMD.
Survival did not change over the study period. Five cases identi ed in the rst two years of the study survived longer than two years, but survival over the decade remained static even accounting for these outliers. Median survival after the diagnosis of BC-LMD was 4.7 months in this study. TNBC LMD had the shortest median survival of 2 months, followed by 5.3 months in HR + and 8.4 months in HER2+. These ndings concur with previous studiesand BC-LMD survival depending on BC subtype. 22,23 While longer survival for patients with HER2 + BC-LMD likely stems from HER2 targeted systemic and IT chemotherapy. 1,17,22,24 differences in survival in HR + or TNBC may be driven by other factors. Patients with HR + BC-LMD were ve times as likely to receive systemic therapy and four times as likely to receive IT chemotherapy following BC-LMD diagnosis compared to those with TNBC. It is tempting to say, therefore, that intention to treat is the primary driver of prolonged survival, but other factors including poor performance status and extent of extracranial metastatic disease may be confounders in this data. 23 Even so, we were unable to identify any other single factor accounting for differences in survival among patients with HR + or TNBC.
The use of systemic therapy and IT chemotherapy appeared to improve survival following the diagnosis of LMD regardless of HR subtype. IT trastuzumab showed prolonged survival in HER2 + LMD. 22,25,26 In 2018, the rst phase 1 study of IT trastuzumab showed 150 mg weekly dosing achieved steady-state levels after 1 week and was well tolerated. 27 Prior to this publicationy, patients with HER2 + LMD treated at our institution received weekly IT doses of trastuzumab of less than 150 mg. Our study demonstrates that IT trastuzumab dosing (i.e. physiologic dosing) resulted in improved survival for patients with HER2 + disease.
To maximize IT HER2 targeted therapy, our institution opened a phase I/II study of radiotherapy followed by IT trastuzumab and pertuzumab in patients with HER2 + BC-LMD to evaluate safety and treatment outcomes (NCT04588545). 24 Radiotherapy can eliminate tumor blockages within the leptomeninges, allowing for IT therapy to properly ow along the CSF. 28 The current study, however, found no statistical difference in the OS between patients who received only IT therapy versus WBRT and IT therapy. However, the latter cohort included only four patients. Further investigation is warranted.
There is an increasing interest in HER2-targeting therapies in BC-LMD. Prior studies showed that pertuzumab-based therapies improved progression-free survival (PFS) when used with trastuzumab and taxanes, and when given as a rst or second-line chemotherapy 29, 30 y 31 .
Of particular interest in the current study, ten HER2 + BC-LMD patients exhibited a signi cant increase in OS in association with systemic treatment with lapatinib, a reversible tyrosine kinase inhibitor (TKI). 32,33 Lapatinib with capecitabine is well tolerated 34 and improved PFS to 8.4 months versus 4.1 months using capecitabine alone. 35 Lapatinib + capecitabine BC brain metastases showed overall response rates (ORR) of 59.1% among treatment naïve patients 36 and 21% among patients who may or may not have been exposed to either agent previously. 37 However, the CEREBEL trial demonstrated no difference in PFS between capecitabine plus either lapatinib or trastuzumab 38 .The EMILIA trial showed that lapatinib plus capecitabine was both less tolerable and less e cacious than trastuzumab emtansine in prolonging PFS. 39 The LANTERN trial, a phase II trial comparing lapatinib-capecitabine versus trastuzumabcapecitabine therapy in HER2 + BC with CNS metastasis showed no signi cant difference in PFS but a trend favoring trastuzumab-capecitabine. 40 Aside from two case studies, there is no prospective data describing the e cacy of lapatinib in the treatment of BC-LMD. 41,42 Neratinib and pyrotinib are similar to lapatinib except that they irreversibly bind to the HER intracellular phosphorylase domain and have e cacy in BC brain metastases. 33 The NALA trial compared lapatinib and neratinib and both with capecitabine in BC brain metastases and showed similar ORR (26.7% versus 32.8%, respectively), but signi cantly longer duration of response for neratinib rather than lapatinib (8.5 versus 5.6 months, respectively). The bene ts of these TKIs were overshadowed by the combination of T-DM1 and tucatinib, as tolerability was better and e cacy was at least as good for these agents. 16,33 Neratinib + capecitabine enhanced OS to 10 months and improved neurological symptoms in 60% of patients with HER2 + BC LMD. 43 Tucatinib is a newer TKIthat has shown activity in HER2 + BC brain metastases, when combined with trastuzumab and capecitabine. [44][45][46] CSF pharmacokinetic analysis revealed detectable levels of tucatinib within 2 hours of administration (NCT03501979). 47 Further studies using TKIs in BC-LMD are needed. Other studies evaluated these agents in CNS metastases and the ORR in BC-LMD was not established.
At the present time, there is a lack of effective treatments in TNBC-LMD. The current study showed an association between IT therapy and OS among BC-LMD patients, supporting prior studies. 22 Similarly, we found that patients with TNBC-LMD were less likely to receive treatment after their LMD diagnosis. Prospective trials speci cally targeting TNBC-related LMD are critically needed.
A general observation based on the current data conforms with previous work showing that treatment of any type following diagnosis of BC-LMD improved survival to 6.54 months. 3,22,23 It might be surmised that greater intention to treat would improve survival, as this study found that patients that did not receive any treatment survived a median of 1.07 months. Survival has also been shown to vary when considering diagnostic modality: cytology versus MRI alone. 2 However, we found no difference in OS based on the diagnostic modality. This contradiction may relate to our exclusion of cases of BC-LMD, which were treated on clinical suspicion of LMD but for which no CSF cytology or MRI evidence of disease was found. 17 There may also be improving awareness and recognition of typical radiographic LMD features, which increases the sensitivity of MRI in our institution.
To expand systemic treatment affecting CNS and LMD-related cancer, various novel immunotherapy approaches are being assessed in the management of BC-LMD. The use of systemic pembrolizumab 48 or systemic ipilimumab and nivolumab 49 showed promise for a variety of LMD patients (most of whom had LMD from BC), but the median survival was only 3.6 months and 2.9 months, respectively. Only 11 patients at our institution received systemic immunotherapy following diagnosis with BC-LMD, and no associated bene t was found. An alternative approach in trials now includes IT bispeci c antibody-armed T-cells that can be directed against HER2 (NCT03661424). A similar trial with IT HER2-directed chimeric antigen receptor (CAR) T cells (NCT03696030) is recruiting currently. A phase 3 study is planned to use the systemic administration of blood-CSF penetrant drug ANG1005 (a paclitaxel-peptide conjugate that crosses the blood CSF barrier via a low-density lipoprotein receptor-related protein-1 (LRP-1) mediated transcytosis). A phase 2 study with the same agent showed activity and an average survival of 8 months in BC patients with LMD. 50 At the time of this report, there are only 23 recruiting or active clinical trials in the US targeting LMD generally and even fewer targeting BC-LMD. Even with improvements in outcomes for HER2 + BC-LMD, the need to nd new therapies which improve OS in BC-LMD is dire.

Declarations
Ethical Approval: This study was approved by the Ethics Committee and Internal Review Board. This is a retrospective study and does not require informed consent to participate or publish the study.    8 months). C) Patients receiving intrathecal therapy post BC-LMD diagnosis had a signi cantly higher median survival time (11.8 months) when compared to BC-LMD patients that did not receive intrathecal therapy (1.9 months). D-E) More speci cally, overall survival median times were higher in patients that received intrathecal methotrexate (8.4 months) and/or intrathecal thiotepa (12 months) versus those that did not (2.9 months; 3.6 months, respectively). F) patients receiving WBRT post BC-LMD diagnosis had a higher overall survival median time (6.5 months) than those who did not receive WBRT (2.7 months).