Management of Central Nervous System Metastases in Breast Cancer

Brain Metastases

In 2021 there are projected to be 284,200 new cases of breast cancer and 44,130 deaths from breast cancer in the United States alone. Presenting stage and current disease status (i.e., metastatic or not) are highly associated with the risk of brain metastases. Other risk factors include HER2-positive subtype, triple-negative subtype, young age, overall disease burden, and germline BRCA1/2 status.

Population-based studies within the Surveillance, Epidemiology, and End Results (SEER) database have demonstrated that less than 1% of all newly diagnosed breast cancer patients have brain metastases at initial presentation; however, 7.6% of patients presenting with de novo metastatic breast cancer have evidence of brain metastases at diagnosis, including 11.5% of patients with de novo HER2-positive metastatic breast cancer and 11.4% of patients with de novo triple-negative metastatic breast cancer. Among 15,204 patients initially presenting with stage I to III breast cancer in the National Comprehensive Cancer Network (NCCN) breast cancer outcomes database, those with HER2-positive or triple-negative subtype were more likely to recur initially in the central nervous system (CNS) compared with those with hormone receptor (HR)-positive, HER2-negative tumors (hazard ratio 3.97 [95% CI 2.36–6.72; P < 0.001] and hazard ratio 3.5 [2.10–5.85; P < 0.001], respectively), though the absolute risk of CNS as first site of relapse was low across subtypes. A similar study of 1434 patients with stage I/II breast cancer treated with breast conservation therapy reported a 5-year cumulative incidence of brain metastases of 1.7%; 0.1% for luminal A, 3.3% for luminal B, 3.2% for luminal-HER2, 3.7% for HER2, and 7.4% for triple-negative subtype. Other risk factors for brain metastases identified in this study were young age, high grade, and node-positivity. The risk of CNS relapse is increased in the setting of stage III disease, particularly in patients presenting with inflammatory breast cancer (IBC). In a study of 203 women presenting with stage III/IV IBC, the 2-year cumulative incidence of brain metastases was 18.7%. In a second study of 210 women presenting with IBC, the 3-year cumulative incidence of brain metastases was 15% in patients with stage III disease and 37% in those presenting with stage IV IBC.

Unfortunately, no adjuvant therapy has been proven to reduce the risk of CNS relapse. Across multiple adjuvant trastuzumab trials for early-stage, HER2-positive breast cancer, the combined risk of CNS as first site of relapse was 2.56% (95% CI 2.07%–3.01%) in those receiving trastuzumab compared with 1.94% (95% CI 1.54%–2.38%) in those who did not receive trastuzumab. In the ALLTO clinical trial testing the role of adjuvant lapatinib, the risk of CNS as first site of relapse was 2% in all treatment arms. The APHINITY trial similarly did not identify any signal of CNS prevention with adjuvant pertuzumab. In the KATHERINE trial, which enrolled patients with residual disease after neoadjuvant trastuzumab and chemotherapy, CNS as first site of relapse was noted in 4.3% of patients who received adjuvant trastuzumab versus 5.9% of those who switched to T-DM1. Notably, over half of the distant recurrences recorded in the T-DM1 arm were CNS relapses. In contrast, the ExteNET study tested the role of adjuvant neratinib and reported a potential signal for CNS prevention, albeit in a post hoc analysis of the subgroup of patients who were HR-positive/HER2-positive, less than 1 year post-trastuzumab, and had residual disease after neoadjuvant therapy (n = 295; incidence of CNS metastases 0.8% in neratinib arm vs. 3.6% in the placebo arm).

Once patients develop distant disease, there is a continuous risk of CNS involvement over time. In patients with HER2-positive, metastatic breast cancer, the lifetime risk of developing brain metastases is as high as 37% to 55%. Patients with triple-negative, metastatic disease face a lifetime risk of 25% to 46%. Median time from first distant recurrence to first CNS recurrence appears shorter in patients with triple-negative breast cancer, compared with other subtypes. Recent studies have also pointed to an increased risk in germline carriers of deleterious alterations in BRCA1 and BRCA2 . For example, in one study, 53% of BRCA1 and 50% of BRCA2 carriers developed parenchymal brain metastases and/or leptomeningeal disease (LMD), compared with 27% of 270 noncarriers.

Leptomeningeal Disease

LMD is less common than parenchymal brain metastases, and occurs in less than 5% of patients with metastatic disease. Breast cancer is the second most frequent solid tumor associated with LMD. The incidence of LMD is also relatively enriched in patients with lobular histology, and in those with estrogen receptor (ER)-negative tumors.

Epidural Spinal Cord Compression (ESCC)

A study of 35,197 patients with breast cancer from Ontario, Canada reported a 5-year cumulative incidence of metastatic ESCC of 5.52%. ESCC is rarely the initial presentation of breast cancer (0.11%). Lu and colleagues identified four important risk factors to stratify the risk for confirmed ESCC among patients presenting with suspected ESCC, including known bone metastases for >2 years, metastatic disease at initial diagnosis, objective weakness, and vertebral compression fracture on spine radiograph. The risk of confirmed ESCC ranged from 12% in those with no risk factors to 85% in patients with >3 risk factors.

Brain Metastases

Because routine surveillance imaging of the brain is not currently standard practice for asymptomatic patients without a history of CNS involvement, a low threshold for brain MRI with and without contrast should be considered in patients presenting with suggestive symptoms, particularly for those in high-risk groups. Signs and symptoms from brain metastases are varied and caused by the displacement or destruction of structures within the brain either directly by tumor and/or associated peritumoral edema. Symptoms can include headache, focal neurologic dysfunction (e.g., weakness, visual field deficits, imbalance/incoordination), and seizures. Half of patients experience headaches, while associated nausea or vomiting is noted in 40% of patients. An abnormal neurologic examination, or a significant change in prior headache pattern, also suggest that the headache may be caused by a tumor. Between 15% and 18% of patients present with seizures ; among seizure-naïve patients at initial presentation of intracranial disease, approximately 10% to 11% of patients will later develop seizures. Encephalopathy is a less common presentation, and more commonly due to toxic-metabolic factors, rather than occult brain metastases. Although hemorrhage can be seen, anticoagulation is safe so long as patients have not had active bleeding of their brain metastases.

Leptomeningeal Disease

The clinical hallmark of LMD is a constellation of signs and symptoms caused by multifocal involvement of the neuraxis (i.e., cerebrum, cranial nerves, and spinal roots). Frequent manifestations of LMD include headache, nausea, and cranial nerve deficits, particularly to the nerves that control extraocular muscles. In a study of 102 patients with LMD, 33%, 39%, and 13% of patients had signs and symptoms involving one, two, or all three levels of the neuro-axis, respectively. Careful neurologic examination may reveal ocular muscle paresis, facial weakness, diminished hearing and/or vision, trigeminal neuropathy, hypoglossal neuropathy, facial numbness, and diminished gag reflex. Other findings include seizures, hemiparesis, and cerebellar dysfunction. Invasion of the spinal roots can be associated with weakness, paresthesias, back and neck pain, radicular pain, and/or bowel and bladder dysfunction. Meningeal signs of nuchal rigidity and pain on straight leg raises are common. Symptoms of increased intracranial pressure occur in approximately half of patients with LMD and include headache, nausea, vomiting, dizziness, and altered vision.

Epidural Spinal Cord Compression

ESCC is a medical emergency. Prompt recognition and treatment is critical to reduce morbidity and preserve neurologic function. The most common presenting symptoms are back pain, muscle weakness, sensory loss, and sphincter dysfunction. In a series of 100 patients presenting with radiculopathy or myelopathy, epidural metastases were diagnosed in 54% of cases, suggesting a very high pretest probability and strong recommendation for rapid assessment in patients with these signs/symptoms.

Up to 85% of patients have partial weakness or paresis, which is generally symmetrical. Sensory deficits, including ascending paresthesias, are also common, and correspond to the dermatome level. Radiculopathy can occur with lateralized epidural lesions. “Saddle anesthesia” is frequent in patients who have cauda equina syndrome. Urinary hesitancy, retention, and incontinence usually occur as a late manifestation of myelopathy and are relatively uncommon at initial presentation.

Brain Metastases

The differential diagnosis for a brain lesion includes primary brain tumor, infectious abscess, progressive multifocal leukoencephalopathy, demyelination, vascular anomaly, cerebral infarction or bleeding, and radiation necrosis (if the patient has had prior radiation therapy [RT] to the brain).

For diagnostic evaluation, gadolinium-enhanced MRI is strongly preferred over contrast-enhanced computerized tomography (CT) or 18-fluorodeoxyglucose positron emission tomography (FDG-PET), due to increased sensitivity for posterior fossa and cortical lesions, and small lesions, especially when thin slices (1 mm) are utilized on T1 postcontrast sequences. In a study of 50 patients with suspected brain metastasis, 70 cerebral metastases were identified in 20 patients by MRI, but the sensitivity of FDG-PET fused with CT imaging for detection of all 70 lesions was only 20%. Currently, MRI spectroscopy, MRI perfusion, and FDG-PET are used primarily to help differentiate between tumor recurrence and radiation necrosis.

For patients presenting with a single CNS lesion, surgical resection for both diagnostic and therapeutic purposes should be strongly considered. In a randomized trial of surgical resection for patients with solid tumors presenting with a single suspected brain metastasis, 11% of patients were found to have alternative diagnoses, including primary tumors, abscess, and inflammatory reaction. If surgical resection is not indicated, stereotactic biopsy can be considered.

If brain metastasis tissue is available, ER, progesterone receptor (PR), and HER2 testing should be routinely performed, as changes in tumor subtype compared with the primary tumor and other distant metastatic sites can occur. In a study of 219 patients with resected breast cancer brain metastases, 22.8% of patients had a tumor subtype switch, including 37.5% of patients who had primary HR-positive/HER2-negative disease. Additionally, 14.8% of initially HER2-negative patients gained HER2 in their brain metastasis. Loss of receptor expression was generally associated with worse survival (hazard ratio 1.80, P = 0.03), whereas gain of HER2 was associated with a nonsignificant trend to improved survival (hazard ratio 0.64, P = 0.17). A similar study including 316 patients reported a subtype discordance rate of 32%, with 13% of patients experiencing HER2 gain in their brain metastasis.

Leptomeningeal Disease

The differential diagnosis for LMD includes infectious etiologies (i.e., bacterial meningitis, Lyme disease), autoimmune disorders (i.e., vasculitis, sarcoidosis), and artifact (i.e., post-RT changes, postlumbar puncture changes). It is also important to distinguish classical LMD from postsurgical pachymeningeal seeding in which tumor cells dispersed at the time of surgery form nodular masses on the pachymeninges, as the natural history and management of classical LMD versus pachymeningeal seeding differs significantly. The diagnosis of LMD is based on obtaining a detailed history and identifying neurologic signs on physical examination that suggest multifocal involvement of the neuraxis.

MRI including brain and spine should be performed in patients with suspected LMD. The sensitivity and specificity of MRI for LMD is 76% to 77%. By contrast, the sensitivity of CT is low (~40%). Typical MRI findings in the brain include thin, diffuse leptomeningeal enhancement, multiple nodular or plaque-like deposits adherent to the dura, and tumor masses with or without hydrocephalus ( Fig. 65.1 ). On MRI, the entirety of the spinal cord, as well as the cauda equina, can exhibit linear enhancement. Nodular deposits may be seen on spinal nerve roots and on the cauda equina.

The presence of tumor cells in the CSF is considered diagnostic for LMD; however, the sensitivity is approximately 75% when a single sample is analyzed. The sensitivity of CSF cytology increases with the amount of CSF collected and the number of collections, with up to 98% sensitivity when four serial CSF samples are analyzed. Typical CSF findings include elevated opening pressure, elevated protein level, and low glucose concentration. The combination of MRI and CSF cytology increases the diagnostic accuracy for LMD. The EANO-ESMO classifies type I LMD by positive CSF cytology, whereas type II LM is defined by typical clinical and MRI signs.

Several newer techniques are being studied to increase the sensitivity for the diagnosis of LMD. In one study of 78 solid tumor patients who presented with clinical data suggestive of LMD, the sensitivity of flow cytometry immunophenotyping was higher than with conventional cytology (75.5% vs. 65.3%), with similar specificity (96.1% vs. 100%). In a study of 95 patients with epithelial tumors (n = 36 with breast cancer), the CellSearch circulating tumor cell (CTC) assay achieved 93% sensitivity and 95% specificity for the diagnosis of LMD, using a >1 CSF-CTC/mL cutoff. Studies reporting the use of technologies to detect and characterize cell-free DNA in CSF have been generally quite small, and have focused primarily on the ability of the assays to detect specific genetic alterations (e.g., EGFR mutations, KRAS mutations, ERBB2 amplification).

Epidural Spinal Cord Compression

Patients with known breast carcinoma who develop persistent back, neck, or radicular pain should have an MRI scan. The differential diagnosis when considering ESCC includes benign musculoskeletal disease, spinal abscess, and radiation myelopathy. To diagnose ESCC, the metastatic tumor must extrinsically compress the thecal sac ( Fig. 65.2 ). Plain radiographs are insufficient to observe this compression, so MRI of the spine with and without contrast or myelography are the diagnostic tools of choice. CT scanning is more sensitive than plain films but is not as sensitive or as specific as MRI. In general, MRI with and without contrast is preferred over MRI myelography because the latter requires a lumbar puncture, is contraindicated if the subarachnoid space at the level of the spinal cord is obstructed, and does not provide soft tissue resolution.

Brain Metastases

Historically, patients with brain metastases have had a poor prognosis, with median survival between 4 and 6 months. However, survival has improved, most notably in patients with HER2-positive breast cancer.

Sperduto and colleagues reported on results of a multiinstitutional, multinational database of 2473 patients with breast cancer who experienced a new diagnosis of brain metastases between 2006 and 2017. Prognostic factors were tumor subtype, performance status, age, number of brain metastases, and presence of extracranial metastases, and these factors were incorporated into a Graded Prognostic Assessment (GPA) score. Median overall survival (OS) from diagnosis of brain metastases ranged from 6 months in the GPA 0 group to 36 months in the GPA 3.5–4.0 group ( Table 65.1 ). Patients with HER2-positive tumors experienced the longest survival; those with triple-negative tumors fared the worst.

The importance of tumor subtype in predicting outcomes has been consistently reported. Within the population-based SEER registry, median survival after a brain metastasis diagnosis ranged from 6.0 months in patients with triple-negative disease to 21.0 months in patients with HR-positive, HER2-positive breast cancer. In the HER2CLIMB clinical trial, which enrolled patients with HER2-positive metastatic breast cancer pretreated with taxane, trastuzumab, pertuzumab, and ado-trastuzumab emtansine, median OS from randomization in the triplet arm (tucatinib, capecitabine, trastuzumab) in the subset of patients with active brain metastases at study entry was 20.7 months. In contrast, most studies report median OS <6 months in patients with triple-negative breast cancer and brain metastases, and triple-negative status confers poor prognosis even in patients presenting with limited (one to three brain metastases) CNS involvement, compared with other breast cancer subtypes.

Leptomeningeal Disease

Unfortunately, survival after a diagnosis of LMD remains very limited, particularly among patients who present with poor performance status. Other prognostic factors have been defined in the Epidemiological Strategy and Medical Economics (ESME) database, which included 22,266 patients with metastatic breast cancer diagnosed between 2008 and 2016, of which 312 patients received intrathecal (IT) therapy for a diagnosis of LMD. Poor prognostic factors included: triple-negative subtype (hazard ratio 1.81, 95% CI 1.32–2.47 vs. HR-positive/HER2-negative subtype as reference), > treatment lines (hazard ratio 1.88, 95% CI 1.30–2.73), and >3 other metastatic sites (hazard ratio 1.33, 95% CI 1.01–1.74). Concomitant systemic therapy was associated with longer OS (hazard ratio 0.47, 95% CI 0.35–0.62). Median OS after LMD diagnosis was 5.1 months (95% CI 4.1–7.3) for HR-positive/HER2-negative, 5.6 months (95% CI 2.9–11.6) for HER2-positive, and 3.0 months (95% CI 1.7–5.1) for triple-negative subtype, respectively ( P < 0.0001).