Tucatinib–trastuzumab–capecitabine for treatment of leptomeningeal metastasis in women with HER2⁺ breast cancer: TBCRC049 phase 2 study results

Abstract

Treatments for leptomeningeal metastasis (LM) are limited and prognosis is poor. In this phase 2, nonrandomized, single-arm, multicenter study, we evaluated a tucatinib–trastuzumab–capecitabine regimen in patients with newly diagnosed LM and human epidermal growth factor receptor 2-positive (HER2⁺) breast cancer. The primary endpoint was overall survival; secondary endpoints included central nervous system progression-free survival, LM objective response, neurological symptom improvement, pharmacokinetics and safety. The trial met its prespecified interim efficacy threshold and exceeded the historical control of 4.4 months. Among 17 enrolled women, all had magnetic resonance imaging-confirmed LM, 15 (88%) were symptomatic and 8 (47%) had abnormal cerebrospinal fluid cytology. For a median follow-up of 18 months (range 9.0–26.7 months), 6 of 17 (41%) remained alive. Tucatinib reached therapeutic levels in the cerebrospinal fluid. The median overall survival was 10 months (95% confidence interval 4.1 months, not reached). The median time to central nervous system progression was 6.9 months (95% confidence interval 2.8, 13.8 months). Of 13 response-evaluable patients, 5 (38%) achieved composite LM objective response. Of 12 evaluable patients, 7 (58%) had improved neurological deficits. This prospective study suggests clinical benefit with a systemic regimen for HER2⁺ LM including objective responses, improved symptoms and extended survival. These data support systemic therapy as an approach in HER2⁺ breast cancer LM. ClinicalTrials.gov registration: NCT03501979.

Murthy and colleagues discussed recent research on the effect of a tucanitib–trastuzumab–capecitabine regimen in patients with HER2⁺ breast cancer and leptomeningeal metastasis.

Main

As new therapies extend the survival of cancer patients, leptomeningeal metastasis (LM) has become increasingly prevalent as a late-stage complication of systemic solid cancers¹, with breast cancer (BC) as the most common etiology². LM refers to the seeding of tumor cells to the leptomeninges and/or cerebrospinal fluid (CSF)³. Classically, LM presents with multifocal neurological deficits and progressive neurological decline and is diagnosed based on clinical findings, neuroaxis imaging findings and/or CSF cytology^1,2. Treatment options for patients with LM are limited and include radiation therapy, intrathecal therapy and, in select cases, systemic therapy⁴. The prognosis remains poor with an estimated median overall survival (OS) of 4–5 months^5–7.

Systemic therapies that better penetrate the brain are needed for LM and other central nervous system (CNS) metastases. Small-molecule HER2 kinase inhibitors may have the potential to penetrate the brain more effectively than antibody-based anti-HER2 agents, such as trastuzumab, pertuzumab and antibody–drug conjugates⁸. Recently, the antibody–drug conjugate trastuzumab–deruxtecan demonstrated encouraging activity in patients with LM disease (LMD) in two small retrospective studies.

Tucatinib is a small-molecule, HER2-targeted tyrosine kinase inhibitor (TKI) approved in Europe, the USA and several other countries for use in combination with trastuzumab and capecitabine in patients with both advanced unresectable and metastatic HER2⁺ breast cancer (BC), including patients with brain metastases, who have received one or more previous HER2-based regimens in the metastatic setting⁹. As a highly selective HER2-targeted TKI, tucatinib minimally inhibits other HER family receptors, such as epidermal growth factor receptor (EGFR or HER1), with lower potential for EGFR-related toxicities compared with the nonspecific HER family TKIs¹⁰. In the pivotal HER2CLIMB trial (NCT02614794) comparing tucatinib administered orally twice daily versus placebo, both in combination with trastuzumab and capecitabine, clinically meaningful and statistically significant improvements in progression-free survival (PFS), OS and confirmed objective response rate were observed in patients with HER2⁺ BC previously treated with trastuzumab, pertuzumab and T-DM1¹¹. Importantly, HER2CLIMB demonstrated a clinically meaningful and statistically significant improvement in PFS and a clinically important improvement in OS among patients with brain metastases¹¹. Furthermore, in an exploratory analysis of patients with brain metastases, patients treated with the tucatinib–trastuzumab–capecitabine regimen had a twofold increase in intracranial objective response rate, a two-thirds reduction in risk of intracranial progression or death and nearly half the risk of death¹². However, as is typical for trials in this setting, patients with LM were excluded from HER2CLIMB, leaving unknown the efficacy of this regimen in this important subgroup of patients who have limited treatment options and poor prognosis.

Tucatinib is predominantly metabolized by cytochrome P450 2C8 to the metabolite ONT-993 and to a lesser extent cleared via cytochrome P450 3A4 metabolism and biliary excretion. Assessing the bioavailability of tucatinib and ONT-993 into the CSF in a clinically relevant human patient population may contribute to the mechanistic understanding of tucatinib efficacy in patients with brain metastases.

We evaluated the safety, efficacy and pharmacokinetics of a tucatinib–trastuzumab–capecitabine regimen in patients with HER2⁺ BC and newly diagnosed LM.

Results

Patient characteristics

From 20 February 2019 to 22 June 2021, 17 women were enrolled and received at least one dose of tucatinib, capecitabine and trastuzumab at 4 sites in the USA (Fig. 1). The study was closed early (22 June 2021) before reaching the planned accrual target due to lack of enrollment after the US Food and Drug Administration (FDA) approval of tucatinib in the USA. Baseline disease characteristics of the 17 enrolled patients are shown in Table 1. All patients had magnetic resonance imaging (MRI) evidence of LM in the brain and eight (47%) had abnormal CSF cytology (five with malignant cells and three with atypical cells or equivocal). Fifteen patients (88%) had a history of brain metastases and eleven (71%) had received previous treatment for brain metastases. Three patients had newly diagnosed brain metastases concurrent with their LM disease diagnosis, for which they did not receive any therapy ahead of trial enrollment. The median patient age at study treatment initiation was 53 years (range 33–67 years). Five patients identified as black (29%) and 12 patients as white (71%). The mean and median lines of previous systemic therapy in the metastatic setting were 1. There was one patient who was heavily pretreated with eight previous lines of therapy in the metastatic setting and six had no previous lines of systemic therapy in the metastatic setting.

Fig. 1. Flowchart of study screening, enrollment, treatment and analysis populations.

Of 20 patients screened, 3 did not meet eligibility criteria and were excluded. Seventeen patients were enrolled and initiated protocol therapy with tucatinib, trastuzumab and capecitabine. Thirteen patients were evaluable for the LM response at the first restaging assessment. Four patients were not evaluable owing to early events, including symptomatic LM progression (n = 2), treatment-related toxicity (n = 1) and transition to a hospice (n = 1). Twelve patients discontinued treatment due to LM progression, seven of whom had concurrent brain metastasis progression. All 17 treated patients were included in the survival analyses. BM, brain metastasis.

Table 1.

Baseline disease characteristics

Baseline disease characteristics	Patients (N = 17) n (%)
Baseline CSF cytology
Positive	5 (29)
Negative	7 (41)
Equivocal	3 (18)
Unavailable	2 (12)a
Symptoms or signs attributable to LM
Yes	15 (88)
No	2 (12)
Estrogen receptor status
Positive	12 (71)
Negative	5 (29)
ECOG performance status
0	4 (24)
1	13 (76)
MRI evidence of LM
Brain only	11 (65)
Brain and spine	6 (35)
History of brain metastasis
Yes	15 (88)
Previous locoregional therapy	12 (65)
New or concurrent diagnosis (no previous treatment)	3 (18)
No	2 (12)
Previous treatment for parenchymal brain metastasis
Stereotactic radiosurgery	7b
Whole-brain radiation therapy	7b
No previous treatment	3
Prievious systemic HER2 regimens in metastatic setting
Trastuzumab containing	10 (59)
Pertuzumab containing	7 (41)
T-DM1 containing	3 (18)
Lapatinib containing	1 (6)
None	5 (29)
Number of previous lines in the metastatic setting
1	8
2	1
3	1
5	1
None	6
Extra-CNS disease
Yes	11 (65)
No	6 (35)

^aTechnical difficulty with sampling and insufficient fluid obtained for two patients.

^bTwo patients received both stereotactic radiosurgery and whole-brain radiation therapy.

Efficacy

The early clinical benefit endpoint for interim analysis was met, with 7 of the first 15 patients enrolled observed to have CNS PFS at 12 weeks. At data cutoff on 20 July 2021, 6 of 17 patients (35%) remained alive and 1 patient remained on study. Patients received a median of 5 treatment cycles (range 1–27), with median follow-up of 18 months (range 9–26.7 months), median OS 10 months (95% confidence interval (CI) 4.1 months, not reached (NR)) and median time to CNS progression 6.9 months (95% CI 2.3, 13.8 months) (Fig. 2a). Twelve patients discontinued treatment in the setting of progressive LM. More specific reasons for discontinuing treatment were CNS progression of both LM and intracranial brain metastases (six patients), CNS progression of LM only (five patients), both CNS and extracranial progression (one patient), extracranial progression only (two patients), toxicity (one patient) and patient decision to transition to a hospice (one patient), as illustrated in Fig. 2b. No patient had isolated progression of brain metastases without LM progression. The treatment duration, reason for discontinuation of treatment and survival status of each patient enrolled in the study are illustrated in the swimmer plot depicted in Fig. 2b.

Fig. 2. Analysis of survival and progression.

a, OS, PFS and time to CNS progression (TTP CNS) in 17 patients. b, Timeline of treatment, progression and survival outcomes in study patients. This swimmer plot illustrates the treatment duration, reason for discontinuation of treatment and survival status of individual patients in the study over time (measured in months). Each horizontal bar corresponds to a single patient. As indicated by the purple triangles, 12 patients discontinued due to progressive LM, of whom 7 also had progressive brain metastases. Three patients discontinued with systemic progression. Patient (Pt) 16 discontinued for toxicity and patient 5 discontinued by choice.

Source data

LM response assessment

Of 17 patients, 13 were evaluable for composite LM response assessment. Of the 13 patients, 5 (38%) evaluable for composite LM objective response met the criteria for response (Supplementary Table 1) per the composite assessment criteria incorporating a neurological clinical evaluation, neuroaxis imaging and CSF cytopathology, as depicted in Supplementary Table 2. For all five patients, LM response was observed at the first response assessment timepoint. All 13 (100%) evaluable patients met the criteria for clinical benefit, defined as stable disease (SD) or better on composite LM assessment. Improvement of target deficits in the clinical evaluation component of the assessment was observed in 7 (58%) of 12 eligible patients with at least 1 target deficit at baseline, as shown in Fig. 3. Six of the seven patients with improved target deficits improved by first restaging timepoint and none received an increased dose of corticosteroids during this interval. The remaining five patients achieved best response of stable target deficits. LM radiographic responses were observed in five patients, with a best response of complete response (CR) in one patient and a best response of partial response (PR) in four patients, as depicted in Supplementary Table 1. Eight patients had a best radiographic response of SD. CSF cytological CRs were observed in two of five patients with positive CSF cytology at baseline. Among the four enrolled patients nonevaluable for response, two discontinued treatment due to symptoms suggesting progression of LM, one due to liver toxicity and one opted to transition to a hospice, all before first restaging.

Fig. 3. Target deficits identified at baseline and attributed to LM, with best response to treatment.

Of the 12 evaluable patients with at least one target neurological deficit identified at baseline (up to four for each patient) and attributed to LM, most experienced improvement in deficits, as indicated by the green boxes on this heat map; yellow boxes indicate the best response of SD. BLE, bilateral lower extremity; LLE, left lower extremity. aAt first restaging (before cycle 3), with exception of patient 17 (before cycle 6).

Safety

All patients were evaluable for safety. Treatment-emergent adverse events (TEAEs) are shown in Supplementary Table 3. Most TEAEs were grades 1 and 2 and included diarrhea, nausea, vomiting, hand–foot syndrome and liver function test elevation. Grade 3 TEAEs reported were nausea and vomiting in one patient, hand–foot syndrome in one patient and liver function test elevation in three patients. A grade 4 TEAE of alanine aminotransferase elevation in one patient after one cycle led to study drug discontinuation and resolved after 1 month. Most TEAEs improved or resolved with appropriate supportive care and dose modifications (including treatment interruptions or dose reductions). No treatment-emergent neurotoxicities were observed.

Pharmacokinetics

Cycle 2 pharmacokinetic data for two patients were excluded (UAB-010 due to suspected incorrect dosing as assessed by the lead investigator and UAB-005 due to withdrawn consent). CSF concentrations were not available for patients UCF-006 and UCF-012.

Tucatinib and ONT-993 plasma concentrations after tucatinib 300 mg twice daily on day 1 of cycles 1 and 2 (Fig. 4a–c) were within the expected range and exhibited high interindividual variability. Median (range) steady-state (cycle 2) tucatinib and ONT-993 plasma concentrations at 2–3 h (expected maximum concentration (C_max)) were 510 ng ml⁻¹ (260–730 ng ml⁻¹) and 61 ng ml⁻¹ (25–140 ng ml⁻¹), respectively, consistent with population pharmacokinetic model-predicted C_max values for tucatinib of 747 ng ml⁻¹ in patients with metastatic BC (mBC)¹³. Accordingly, unbound tucatinib plasma (plasma_ub) concentrations were 15 ng ml⁻¹ (7.4–21 ng ml⁻¹) at steady state, 2–3 h post-dose.

Fig. 4. Tucatinib and ONT-993 concentrations in plasma and CSF after administration of tucatinib 300 mg twice daily.

a–c, Arithmetic means with s.d. and individual values for tucatinib and ONT-993 plasma concentration versus time profiles on day 1 of cycles 1 and 2 after 300 mg tucatinib twice daily. a, Tucatinib and ONT-993 plasma concentrations versus time (cycles 1 and 2, arithmetic mean (s.d.). b, Tucatinib cycle 1 plasma concentrations versus time (individual values). c, Tucatinib cycle 2 plasma concentrations versus time (individual values). d–f, Arithmetic means and individual values for tucatinib and ONT-993 CSF concentration versus time profiles on day 1 of cycles 1 and 2 after 300 mg tucatinib twice daily. d, Tucatinib and ONT-993 CSF concentrations versus time (cycles 1 and 2, arithmetic mean (s.d.). e, Tucatinib cycle 1 CSF concentrations versus time (individual values). f, Tucatinib cycle 2 CSF concentrations versus time (individual values).

Source data

Tucatinib and ONT-993 were detectable in the CSF after 300 mg tucatinib twice daily with 2 h post-first tucatinib dose (Fig. 4d–f). Individual total CSF concentrations across all timepoints in cycle 1, day 1 ranged from 0.57 ng ml⁻¹ to 12 ng ml⁻¹ for tucatinib and from 0.21 ng ml⁻¹ to 2.0 ng ml⁻¹ for ONT-993. Individual total CSF concentrations across all timepoints in cycle 2, day 1 (steady-state) ranged from 0.94 ng ml⁻¹ to 25 ng ml⁻¹ for tucatinib and from 0.31 ng ml⁻¹ to 4.7 ng ml⁻¹ for ONT-993.

The steady-state tucatinib CSF concentrations and total CSF-to-calculated plasma_ub ratios were generally consistent over time (Fig. 5a,b). Tucatinib concentrations in the CSF per timepoint were generally in a similar range to corresponding plasma_ub tucatinib concentrations. Median tucatinib total CSF-to-plasma_ub ratios per timepoint ranged from 0.3 to 1.0 for cycle 1 and from 0.59 to 1.1 in cycle 2 (steady state, Fig. 5b). For ONT-993, median ratios per timepoint ranged from 0.32 to 2.4 for cycle 1 and 0.81 to 2.1 in cycle 2 (steady state, Fig. 5d).

Fig. 5. Tucatinib and ONT-993 CSF-to-calculated plasma_ub ratios after administration of tucatinib 300 mg twice daily.

a, Tucatinib concentrations in CSF at 0, 2, 5 and 24 h after tucatinib administration on day 1 of cycles 1 and 2. b, Tucatinib CSF-to-plasma_ub ratios at 0, 2, 5 and 24 h after tucatinib administration on day 1 of cycles 1 and 2. c, ONT-993 concentrations in CSF at 0, 2, 5 and 24 h after tucatinib administration on day 1 of cycles 1 and 2. d, ONT-993 CSF-to-plasma_ub ratios at 0, 2, 5 and 24 h after tucatinib administration on day 1 of cycles 1 and 2. The lines (both dashed and solid) represent the median value for each timepoint. C, Cycle. PK sampling ranges of 2–3 h and 5–7 h were simplified in the figure to 2 and 5 h, respectively.

Source data

Patient-reported outcomes, symptom burden and QoL assessments

Among the 17 participants, the mean Eastern Cooperative Oncology Group performance status score at cycle 1 or baseline was 1 (range, 0–1). All 17 participants completed the majority of both the Linear Analogue Scale Assessment of Quality of Life (LASA) and the MD Anderson Symptom Inventory Brain Tumor (MDASI-BT) questionnaires at protocol-specified timepoints (Fig. 6). The mean LASA score at cycle 1 or baseline was 31 (range 11–47), with higher scores indicating better quality of life (QoL). The mean improvement in the LASA score over the course of the study was 13.5 (Fig. 6a). Five patients experienced an improvement of 10+ points and one experienced an improvement of 20+ points. The mean MDASI-BT score at cycle 1 or baseline was 61 (range 0–213), with higher scores indicating increased symptom burden (Fig. 6b). The mean improvement in the MDASI-BT score over the course of the study was 32.

Fig. 6. Treatment-associated improvements in QoL and symptom burden.

All participants completed the majority of the LASA and MDASI-BT questionnaires across timepoints. Mean LASA and mean MDASI-BT overall scores improved compared to baseline. Graphs depict best improvement in overall scores from baseline assessment in patients with baseline and at least one post-baseline score. a, Change in the LASA overall score. The LASA score reflects QoL, with higher scores indicating better QoL. The x axis represents the change in score from baseline. Positive values indicate improved QoL and negative values worsened QoL. b, Change in the MDASI-BT overall score. The MDASI-BT score reflects symptom burden, with lower scores indicating fewer symptoms. The x axis represents the change in score. Negative values indicate improved symptom burden and positive values worsened symptom burden.

Source data

Discussion

This study provides prospective evidence suggesting clinical benefit with a systemic regimen for HER2⁺ LM. We observed longer survival in patients with LM from HER2⁺ mBC who were treated with tucatinib, trastuzumab and capecitabine, relative to historical controls. We also observed an improvement in neurological signs and symptoms in most patients and objective LM responses in over a third of patients, using a composite criterion. Furthermore, the regimen was well tolerated. The safety data were consistent with data reported in previous studies of tucatinib in combination with trastuzumab and capecitabine^9,11. These data expand the scope of the tucatinib triplet regimen in the CNS to LM from HER2⁺ BC.

Diagnosis and response assessment in LM is challenging, in part due to the poor sensitivity of CSF cytology, the poor specificity of imaging and neurological symptoms that can be difficult to interpret and attribute. Although most studies utilize a composite assessment of CSF cytopathology, neuroaxis imaging and neurological clinical evaluation, no validated algorithm for evaluation exists. In this study, response of LM to treatment was determined using an adaptation of the criteria proposed by the Response Assessment in Neuro-Oncology–Leptomeningeal Metastases (RANO-LM) working group¹⁴. According to these criteria, approximately one-third of patients achieved a composite response. In addition, our study is encouraging in that most patients achieved neurological clinical benefit from the regimen. Most experienced improvement of neurological deficits, which is not commonly reported with LM-directed drug therapies, for which the goal historically has been to delay further neurological deterioration.

In this study, tucatinib and ONT-993 were detectable in the CSF of all 13 patients for whom CSF data were available after the first tucatinib dose and were comparable to steady-state unbound plasma concentrations. This study documented evidence of tucatinib distribution in the CSF in patients with HER2⁺ mBC and supported the previously observed activity in patients with brain metastases. The tucatinib unbound plasma fractions in actual individual patients in this study were not estimated and may affect the estimation of the CSF-to-plasma_ub ratios, although it is not anticipated to greatly impact the overall interpretations. Although detected in the CSF, ONT-993 is not expected to contribute to overall pharmacological activity based on the concentrations and potency. Circulating tumor (ct)DNA analysis with matched CSF and blood correlative samples is under way and will be reported separately.

The most appropriate endpoint for LM trials is the subject of ongoing discourse¹⁵. As response assessment in LM can be difficult to determine and prognosis in this patient population is often determined by LM, we chose OS as our primary endpoint. Nevertheless, identifying an appropriate historical control in LM is difficult. Until recently, the body of the LM literature comprised primarily retrospective studies in addition to six randomized studies^16–21. These studies—most published one or more decades ago—tended to be small and include heterogeneous populations of patients, with variable endpoints and criteria for LM diagnosis and response assessment. In clinical practice, management of LM varies widely and is primarily guided by expert opinion.

Since the development of our study, a randomized study of intrathecal cytarabine in LM has been reported, as well as two prospective studies on intrathecal trastuzumab in the HER2⁺ LM population, a randomized study evaluating proton craniospinal radiation in LM and small retrospective studies of trastuzumab–deruxtecan, which included patients with LM^20,22–25. Although these contemporary studies represent important advances, the literature still suffers from a heterogeneous patient population and a lack of harmonization of study criteria and endpoints, relative rarity, complicated clinical courses and the absence of validated response assessment criteria.

Oberkampf et al. reported in a phase 2 study of intrathecal trastuzumab that, among 19 patients with HER2⁺ BC LM, the median LM-related PFS was 5.9 months and the median OS was 7.9 months. The clinical neurological PFS after 8 weeks of treatment was the primary endpoint of the study: 14 patients (74%) were free of neurological progression at 8 weeks (ref. ²²). In a phase 1 and 2 study, Kumthekar et al. reported a median overall survival of 10.5 months for the phase 2 dose-treated subset of 23 patients with HER2⁺ BC and a median PFS of 2.8 months (ref. ²³). The primary endpoint of the study was disease response, utilizing a scale integrating clinical, radiographic and cytological response; the response rate was 17.9% for the HER2⁺ BC cohort²³. Intrathecal trastuzumab was well tolerated in both studies. Our median OS is consistent with these two studies and all patients fared better than the median OS of 4–5 months reported in historical controls. Of further interest, in a randomized phase 2 study, Yang et al. found that proton craniospinal radiation was well tolerated and notably improved both CNS PFS (median 7.5 months versus 2.3 months; the primary endpoint) and OS (9. months versus 6 months) over photon involved-field radiation in patients with LM and nonsmall-cell lung cancer and BC²⁶. In addition, a recent study provides the first randomized evidence of clinical benefit of intrathecal chemotherapy in LM. In the intent-to-treat group, median LM PFS (the primary endpoint) was 3.8 months in the experimental group (intrathecal cytarabine plus physician’s choice of systemic therapy) versus 2.2 months in the control group (systemic therapy alone) (P = 0.04); median OS was 7.3 months in the experimental group versus 4 months in the control group (P = 0.51)²⁰. Their new agents are under investigation for the treatment of LM^24,25,27. Recently, the antibody–drug conjugate trastuzumab–deruxtecan demonstrated activity in two small retrospective studies in patients with HER2⁺ LM. ROSET-BM (n = 19 with LM) showed 12-month PFS and OS rates of 60.7% and 87.1%, respectively²⁵. In a case series of eight patients with HER2⁺ LM, all patients derived clinical benefit and four (50%) achieved a partial radiographic response²⁴.

Our study provides an important addition to the contemporary literature about the management of LM beyond the favorable efficacy and safety data described above. The study only included patients with HER2⁺ mBC and newly diagnosed, untreated LM, as opposed to earlier trials which often included more heterogenous populations. Still, many of our patients had previously diagnosed parenchymal brain metastases treated with modalities to include radiation and therefore represent a recurrent CNS metastases population. Furthermore, the study entry criteria required CSF confirmation of LM or MRI evidence (MRI reviewed by a radiologist) plus neurological signs and/or symptoms attributed to LM, increasing the confidence that enrolled patients in fact had a diagnosis of LM. A particular strength of our study was the prospective evaluation of patient-reported outcomes. Notably, many patients experienced improvement in patient-reported QoL and symptom burden, with both the mean LASA and the mean MDASI-BT overall scores improved compared to baseline. Improvements in QoL have not been consistently reported with intrathecal chemotherapies and radiation therapy, including proton craniospinal Irradiation, may come with a cost to QoL and short-term and long-term cognition. Importantly, this study provides evidence of the feasibility of enrolling LM patients to a systemic therapy trial, even on the background of the COVID pandemic.

Important questions are now raised about sequencing available therapies in LM, particularly in patients with HER2⁺ mBC. The availability of a systemic regimen for treating LM may represent an opportunity to defer intrathecal therapy, which is a more invasive treatment administered often weekly or even twice weekly that may negatively impact QoL for patients. Furthermore, although intrathecal chemotherapy is most suitable for treating microscopic and linear macroscopic LM due to its limited penetrance beyond the subarachnoid space²⁸, tucatinib in combination with trastuzumab and capecitabine affords an opportunity to treat nodular LM along with parenchymal brain metastasis and systemic disease. This regimen may also represent an opportunity to defer radiation to the neuroaxis, particularly in patients who are asymptomatic or minimally symptomatic.

The current study has several limitations. First, although actual accrual initially outpaced projected accrual rates, the study was terminated early due to lack of accrual after the FDA approval of tucatinib in combination with trastuzumab and capecitabine for patients with HER2⁺ mBC with and without brain metastasis. Second, given that LM from HER2⁺ mBC is a rare disease, and considering the limited LM literature, the study was designed as a single arm utilizing a historical control based on retrospective data. Historically, patients with LM largely had progressive systemic disease and poor functional status at the time of diagnosis. Today, many patients present with stable or controlled systemic disease and good or excellent functional status, mirroring our enrolled patients. As is typically the case in LM studies, CSF confirmation of LM was not a requirement for enrollment in our study. However, patients without positive CSF cytology did have neuroaxis imaging consistent with LM, as well as clinical findings (neurological signs and/or symptoms). Although our clinical trial conditions may not be entirely generalizable to the broader LM population, the study did allow enrollment of patients with a relatively low functional status and, given the tolerability and ease of use and convenient schedule compared to intrathecal therapy, it may be an option available to a broad population of patients with HER2⁺ LM.

The combination of tucatinib, capecitabine and trastuzumab achieved a clinically meaningful improvement in OS in relation to historical controls for patients with LM from HER2⁺ mBC, as well as improvement in neurological symptoms and signs in most participants. The combination was well tolerated with a manageable safety profile. This study provides the prospective data of a HER2⁺-directed systemic therapy providing clinically meaningful benefit for patients with LM from HER2⁺ BC. Although this is an importance advance, further studies evaluating CNS-penetrant drugs, as well as the best combinations and sequence of therapies in this patient population, are needed. Unanswered questions include whether there may be an additive or synergistic benefit of combining the tucatinib regimen with intrathecal therapies, such as intrathecal trastuzumab. Planned future analysis for this study include evaluating blood and CSF samples to assess circulating tumor (ct)DNA concentrations and genomic sequencing to identify candidate biomarkers that may predict response to therapy and provide better understanding of underlying mechanisms.

Methods

Study design

The study was conducted in accordance with the International Conference on Harmonisation Good Clinical Practice guidelines and the Declaration of Helsinki. The study protocol was reviewed and approved by the institutional review board at each participating institution and written informed consent was obtained from all participants before receiving study treatment.

TBCRC049 (NCT03501979; registered at https://clinicaltrials.gov/study/NCT03501979 on 20 February 2019) is an investigator-initiated, phase 2, single-arm, nonrandomized, multicenter study evaluating the combination of tucatinib, trastuzumab and capecitabine for treatment of patients with HER2⁺ BC and newly diagnosed LM.

The study enrolled at four sites in the USA—University of Texas MD Anderson Cancer Center; University of Alabama at Birmingham O’Neal Comprehensive Cancer Center; University of California San Francisco; University of Michigan—through the Translational Breast Cancer Research Consortium (TBCRC). Eligible patients were adults (men and women, based on self-report), aged ≥18 years at the time of consent with histologically proven HER2⁺ mBC by immunohistochemistry 3⁺ and/or FISH ratio >2.0 or average HER2 copy number >6.0 signals per cell or per current American Society of Clinical Oncology—College of American Pathologists or National Comprehensive Cancer Network guidelines, and newly diagnosed LM, defined as positive CSF cytology and/or radiographic evidence of LM plus neurological signs and/or symptoms attributed to LM and Karnofsky Performance Status ≥50. Patients were allowed to enroll while on a stable dose of corticosteroids, when clinically indicated. No previous LM-directed therapy was allowed, including radiation, intravenous and intrathecal therapy. Patients with treated or concurrent or new parenchymal brain metastases were eligible. Patients must not have received whole-brain radiotherapy for parenchymal metastases within 2 weeks or focal CNS radiotherapy within 1 week before the first dose of the study drug. Patients were excluded if they did not have HER2⁺ BC, lacked newly diagnosed LM, had previous LM-directed therapy or had received whole-brain radiotherapy.

Patients were treated with 21-day cycles of oral tucatinib 300 mg twice daily, oral capecitabine 1,000 mg m⁻² twice daily on days 1–14 and intravenous trastuzumab 6 mg kg⁻¹ on day 21 (a loading dose of 8 mg kg⁻¹ was administered on cycle 1, day 1 if maintenance therapy was not received within the previous 30 d). To replicate the findings of this phase 2 study, a clinical trial should enroll patients with HER2⁺ BC and newly diagnosed LM who meet eligibility criteria, including MRI-confirmed LM and either abnormal CSF cytology or neurological symptoms. Patients should receive tucatinib (300 mg twice daily), capecitabine (1,000 mg m⁻² twice daily, on days 1–14 of a 21-d cycle) and trastuzumab (6 mg kg⁻¹ intravenously every 21 d, with an 8 mg kg⁻¹ loading dose if needed). The primary endpoint is OS, with secondary endpoints assessing CNS PFS, composite LM response (via CSF cytology, MRI and neurological function) and patient-reported QoL. Safety should be monitored using Common Terminology Criteria for Adverse Event (CTCAE) criteria, tracking common adverse events such as diarrhea, nausea and liver function changes. Pharmacokinetic analysis should measure tucatinib and ONT-993 levels in CSF and plasma, alongside ctDNA analysis to explore biomarkers. The study follows ethical guidelines, uses historical controls for OS comparison and aims for ~30 patients in a Gehan-like design. The study was a single-arm, nonrandomized, phase 2 trial due to the rarity and aggressive nature of LMs from HER2⁺ BC, which makes large-scale randomization challenging. Ethical concerns also played a role in not designing this as a randomized study, because withholding effective systemic therapy from a control group would be inappropriate given the historically poor prognosis. Instead, the study used historical controls to evaluate efficacy. A single-arm design allowed for faster enrollment and treatment initiation, crucial for this rapidly progressive disease. As there was no randomized control arm, blinding was unnecessary to prevent bias in treatment allocation.

Endpoints and assessments

The primary study objective was to determine OS. Secondary objectives included and reported herein included: safety; time to CNS progression, LM composite objective response assessment and clinical benefit, pharmacokinetics and patient-reported outcomes and QoL assessments.

All patients who initiated treatment on protocol were considered evaluable in the modified intent-to-treat analysis for the primary endpoint (OS). Patients who received at least one cycle of therapy and had their disease re-evaluated by the protocol-dictated CNS response assessment parameters (CSF cytopathology, neuroaxis imaging, neurological clinical evaluation) were considered evaluable for LM objective response. At each cycle visit, evaluation of symptoms was performed using the MDASI-BT patient questionnaire²⁹. A neurological exam recorded using the Neurological Assessment in Neuro-Oncology (NANO) group scoring instrument was performed at each cycle visit, as was CSF analysis via Ommaya reservoir or lumbar puncture³⁰. MR neuroaxis imaging was performed every other cycle, with findings recorded by the radiologist in a provided scorecard. Extracranial imaging by computed tomography scan or positron emission tomography–computed tomography was performed at screening and every four cycles (or sooner if clinically indicated) for evaluation of extra-CNS disease while patients were on study. Patients were followed by the treating medical oncologist and neuro-oncologist.

Response assessment

The LM objective response evaluation used a composite assessment derived from the RANO-LM working group¹⁴. The composite assessment incorporates CSF cytopathology, neuroaxis imaging and a neurological clinical evaluation, which included: (1) a neurological examination recorded using the NANO scoring instrument; and (2) an evaluation of symptoms not captured on physical examination, such as headache and seizure, using the MDASI-BT patient questionnaire. Up to four target deficits (neurological signs or symptoms) attributable to LM were identified in the baseline neurological clinical evaluation and monitored longitudinally. The following were required for response per the composite LM response criteria: (1) negative CSF cytopathology (confirmed on subsequent CSF evaluation) when CSF cytopathology was positive for malignant cells at baseline; (2) definitive improvement on neuroaxis imaging defined as radiological CR or PR; and (3) Improved or stable target deficits on neurological clinical evaluation when target deficits were present at baseline. The objective composite LM response was defined as a CR or PR per the criteria. Clinical benefit was defined as CR, PR or SD. In addition to the standardized NANO scorecard used for clinical evaluation, a standardized imaging scorecard adapted from the proposed RANO imaging scorecard was created by the investigators and used for recording the imaging evaluation. The neuroaxis imaging evaluations were completed locally by radiologists at each site.

Patient-reported outcomes and QoL assessments

Longitudinal symptom burden, patient-reported outcomes and QoL assessments were conducted. The MDASI-BT self-assessment was performed every cycle to assess patient-reported outcomes and symptoms. The LASA QoL self-assessment was administered every other cycle. MDASI-BT and LASA data was summarized at each timepoint using descriptive statistics and evaluated the change in scores over time.

Safety

Safety was evaluated in all patients who received at least one dose of tucatinib, trastuzumab and capecitabine. Adverse events were defined according to the Medical Dictionary for Regulatory Activities and severity was graded according to the National Cancer Institute CTCAE, v4.

Pharmacokinetics and correlative studies

Pharmacokinetic assessments in plasma and CSF

To determine the bioavailability of tucatinib in the CSF of patients with LMs, we conducted a parallel assessment of pharmacokinetics in the CSF and the plasma, for the first 15 patients enrolled in the study. Plasma was collected via intravenous peripheral blood draw and CSF was collected via Ommaya reservoir tap on day 1 of cycles 1 and 2 at 0 h (baseline), 2–3 h (T_max), 5–7 h and 24 h (optional) after initiation of tucatinib 300 mg twice daily. Tucatinib and ONT-993 concentrations were quantified in plasma (n = 14) and CSF (n = 13; CSF data unavailable for two patients) using validated liquid chromatography–tandem mass spectrometry method³¹.

Correlative assessments in plasma and CSF

Biospecimens (blood and CSF) were collected at baseline and each cycle for ctDNA correlative analyses, which is ongoing and will be reported in a separate publication.

Pharmacokinetic analyses

Pharmacokinetic analyses were conducted using WinNonlin v8.1 (Certara L.P.) and plots were constructed using GraphPad Prism v8.1.0 for Windows (GraphPad Software).

Plasma_ub concentrations for tucatinib and its metabolite, ONT-993, were calculated using an average unbound fraction of 0.029 (value from US Prescribing Information; true individual plasma binding may vary). All CSF concentrations reported represent total tucatinib in the CSF.

Statistical analysis

The trial was planned for total enrollment of 30 patients in a Gehan-like trial design with an interim futility analysis and intent to estimate OS³². This sample size is similar to those reported in previous prospective studies^22,23 of LM in patients with HER2⁺ BC. This design used an early clinical benefit endpoint for interim analysis after the first 15 patients enrolled. ‘Success’ was defined as PFS at 12 weeks in the CNS. Enrollment was to stop if there were <2 successes after the first 15 patients. An event was considered to be either CNS progression before 12 weeks or death from any cause. This early clinical benefit endpoint was met and the trial proceeded to enroll with the ultimate goal of estimating the OS curve. The combination treatment would be declared worthy of further study if the median OS matched or was higher than the historical control OS of 4.4 months for patients with HER2⁺ BC and LM, based on the most robust dataset available at the time of protocol development, which was a retrospective analysis of 56 patients from a single tertiary care institution⁵. If the true median OS were 6.0 months, we would expect to observe a median OS > 4.4 months with a probability of 88%. However, the study was closed prematurely with 17 patients enrolled due to lack of accrual after the FDA approval of tucatinib in April 2020 and its subsequent commercial availability. Data collection and analysis were not performed blind to the conditions of the experiments. Data points were not excluded from the analyses for any reason. Data distribution was assumed to be normal but this was not formally tested.

All statistical analyses were performed using JMP 18.0.1. The upper and lower 95% confidence limits of the median times to the event used the Brookmeyer–Crowley method³³.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

R.K.M., B.J.O., D.A.B., A.S.-R., M.G.M., J.J., J.W., J.S.-G., A.T.-E., M.L., K.R., M.M., A.M., S.D.F., J.F.d.G., I.E.K., V.V., M.F.R., A.C.W., D.T., N.U.L. and E.M.S.-R. participated in study design and conduct, in data acquisition, analysis and interpretation and in drafting and revising the paper. D.A.B. performed statistical analyses. A.T.-E. verified the pharmacokinetic data. All authors approved the final version before submission.

Peer review

Peer review information

Nature Cancer thanks Susan Geyer and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

Deidentified data underlying the findings described in this article may be obtained in accordance with the TBCRC data-sharing policy. Requests for TBCRC data or tissue for the purpose of carrying out ancillary analyses that are not part of the objectives of the parent TBCRC study must be reviewed by the TBCRC Correlative Science Review Committee. Source data are provided with this paper.

Competing interests

R.K.M. has received research funding (institutional) from Zanrong Pharma, Seattle Genetics, Pfizer, Daiichi Sankyo, Astra Zeneca and EMD-Serono and honorarium and/or consulting fees from Sanofi, Novartis, Astra Zeneca, Pfizer, Genentech/Roche, Seattle Genetics and Puma. B.J.O. has received research funding (institutional) from EMD-Serono, Seagen, consulting fees from Seagen and scientific and/or advisory committee fees from PlusTherapeutics. D.A.B. is co-owner of Berry Consultants, LLC, a statistical consulting company. A.T.-E. has Seagen employment and/or stocks. A.M. has received research funding (institutional) from Lilly, Merrimack, Novartis, Genentech and/or Roche, Millenium Pharmaceuticals, Eisai, Seagen, Pfizer, Tempus, Molecular Templates, Dantari, Pharmaceuticals, Suzhou Zanrong Pharma, Merck Sharp & Dohme Corp and F. Hoffman LaRoche, consulting and/or advisory fees from Eisai, Lilly, Seattle Genetics and other from Taiho Pharmaceutical. J.F.d.G. has received grant support from Carthera and HaiHe pharmaceuticals, consulting fees from MundiPharma, Insightec, Bioasis Technologies, Kintara, Kazia, Monteris, Karyopharm, Sumitomo Dainippon Pharma Oncology, Merck, Sapience Therapeutics, VBI Vaccines, Chimerix, Aucentra, Midatech, Telix Pharma, Alpha Pharmaceuticals and Servier and other from Alaunos and Brii Biosciences. M.F.R. has received research grants from Pfizer and Genentech and consulting honoraria from Macrogenics, Seagen, Novartis and AstraZeneca. N.U.L. has received contracts and/or grant funding (institutional) from Genentech, Merck, Pfizer, Seattle Genetics, AstraZeneca, Zion Pharmaceuticals and Olema Pharmaceuticals, royalties from Up to Date and consulting fees from Puma, Seattle Genetics, Daiichi Sankyo, AstraZeneca, Prelude Therapeutics, Denali Therapeutics, Olema Pharmaceuticals, Aleta BioPharma, Affinia Therapeutics, Voyager Therapeutics, Janssen, BluePrint Therapeutics and Artera, Inc. and Stemline/Menarini. E.M.S.-R. has received research support from S. G. Komen, Vfoundation, Breast Cancer Research Foundation of Alabama and NIH, clinical research (institutional) funding from Pfizer, Seagen, Tesaro, Lilly, Merck, Genetech, Calithera and Corcept Therapeutics, Inc., and consulting and/or advisory fees from Lilly, Mylan, Novartis, Immunomedics, AstraZeneca, Seagen and Merck. A.C.W., A.S.-R., D.T., I.E.K., J.J., J.S.-G., J.W., K.R., M.G.M., M.L., M.M., S.D.F. and V.V. declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43018-026-01120-7.