Seizures Among Patients With Brain Metastases

Abstract

Objective

To test the hypothesis that subets of patients with brain metastases (BrM) without seizures at intracranial presentation are at increased risk for developing seizures, we characterized the incidence and risk factors for seizure development among seizure-naive patients with BrMs.

Methods

We identified 15,863 and 1,453 patients with BrM utilizing Surveillance, Epidemiology, and End Results (SEER)–Medicare data (2008–2016) and Brigham and Women's Hospital/Dana Farber Cancer Institute (2000–2015) institutional data, respectively. Cumulative incidence curves and Fine/Gray competing risks regression were used to characterize seizure incidence and risk factors, respectively.

Results

Among SEER-Medicare and institutional patients, 1,588 (10.0%) and 169 (11.6%) developed seizures, respectively. On multivariable regression of the SEER-Medicare cohort, Black vs White race (hazard ratio [HR] 1.45 [95% confidence interval (CI), 1.22–1.73], p < 0.001), urban vs nonurban residence (HR 1.41 [95% CI, 1.17–1.70], p < 0.001), melanoma vs non-small cell lung cancer (NSCLC) as primary tumor type (HR 1.44 [95% CI, 1.20–1.73], p < 0.001), and receipt of brain-directed stereotactic radiation (HR 1.67 [95% CI, 1.44–1.94], p < 0.001) were associated with greater seizure risk. On multivariable regression of the institutional cohort, melanoma vs NSCLC (HR 1.70 [95% CI, 1.09–2.64], p = 0.02), >4 BrM at diagnosis (HR 1.60 [95% CI, 1.12–2.29], p = 0.01), presence of BrM in a high-risk location (HR 3.62 [95% CI, 1.60–8.18], p = 0.002), and lack of local brain-directed therapy (HR 3.08 [95% CI, 1.45–6.52], p = 0.003) were associated with greater risk of seizure development.

Conclusions

The role of antiseizure medications among select patients with BrM should be re-explored, particularly for those with melanoma, a greater intracranial disease burden, or BrM in high-risk locations.

Brain metastases (BrM) affect between 20% and 40% of patients with solid malignancies and are associated with significant clinical sequelae,^1,2 including development of seizures. Seizures appear to be a presenting symptom in approximately 10%–20% of patients diagnosed with BrM, although estimates based on prior studies have varied with regard to this estimate.^3–5 In patients presenting with seizures, initiation of an antiseizure medication (ASM) to minimize development of further seizures is indicated.⁶ Among patients who do not present with seizures at BrM diagnosis, the incidence of subsequent seizures remains poorly described. Although current guidelines do not recommend initiation of an ASM following a diagnosis of BrM in seizure-naive patients,^7,8 such guidelines are primarily based on a limited number of older, single-institution studies involving small, heterogeneous cohorts that did not demonstrate a significant decrease in seizure risk among patients receiving ASMs vs not.^9–11 In addition, the increasing use of therapies associated with late inflammation or necrosis, such as stereotactic radiation and immunotherapy, may increase seizure risk in more contemporary cohorts.^12–14 The development of seizures has the potential for significant detriment on patient quality of life as patients with seizures experience restrictions on activities such as driving, operating heavy machinery, recreational activities such as swimming alone, and certain job functions; moreover, the sudden onset of a seizure during such activities also places patients at greater risk for injuries, such as burns and fractures, or death. It follows that seizures often prompt emergency department visits that could potentially have been avoided if preventative ASMs had been utilized, which has implications for both patients and health care systems. Hence, renewed consideration of upfront utilization of ASMs among select patients with BrM treated in the contemporary era is warranted.

We sought to characterize the incidence of seizure development among patients with BrM who were seizure-naive at intracranial presentation on both a population and institutional level and to identify subsets of patients at increased risk of seizures who may potentially benefit from utilization of preventative ASMs if our results are validated prospectively. The utilization of the Surveillance, Epidemiology, and End Results (SEER)–Medicare database allowed us to assess this risk among a larger and more generalizable group of patients with BrM, while the use of institutional data allowed for a more granular assessment of patient- and disease-related risk factors.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

This research did not involve experiments on humans. A waiver of informed consent was obtained given the retrospective nature of the study and given that many patients were deceased or not immediately reachable. No patients are recognizable from the work published here and the data presented in the manuscript are not associated with a clinical trial. This study was approved by our institutional review board.

Patient Population and Study Design

SEER-Medicare Cohort

The SEER registry contains demographic and clinical information for approximately 34.8% of patients with cancer in the United States.¹⁵ For approximately 93% of Medicare patients in the SEER database, the SEER-Medicare program links Medicare claims data to SEER data.¹⁶ We utilized the SEER-Medicare database to identify patients aged 66 and older diagnosed with BrM between 2008 and 2016. In order to identify patients with BrM, we required patients to have ≥3 claims associated with an ICD-9-CM (198.3) or ICD-10-CM (C79.31, 79.32) diagnosis code for secondary neoplasm of the brain, cerebral meninges, and spinal cord, a methodology associated with 97% sensitivity and 99% specificity in identifying patients with BrM via claims data.¹⁷ We utilized the date of the first BrM-associated claim as the BrM diagnosis date, an approach demonstrated to be 92% sensitive for predicting the actual date of BrM diagnosis to within 30 days.¹⁸

To reliably identify patients with a diagnosis code of seizures or with a prescription for ASMs, we mandated continuous Part A, B, and D coverage with no health maintenance organization enrollment from the year prior to BrM diagnosis through the date of censoring or death (n = 19,374). Patients diagnosed at autopsy/death certificate were excluded from the cohort (n = 170).

To restrict the cohort to patients who were seizure-naive, we eliminated patients with an ICD-9-CM (345.XX, 780.3X) or ICD-10-CM (G40.XX, R56.XX) diagnosis code associated with epilepsy or seizures in the year prior to and the 14 days after BrM diagnosis date, as well as any patients who received an ASM (table 1) in the same period (n = 3,341). Such methodology also facilitated exclusion of patients on ASMs for non–seizure-related indications. The final cohort consisted of 15,863 patients.

Table 1.

Seizure Medication List

To estimate the incidence of seizures in this cohort, we searched for a diagnosis code for epilepsy/seizures or initiation of a long-term ASM at any time 15 days or more after the BrM diagnosis date. Long-term ASM use was defined as prescription of an ASM for ≥30 days outside the context of brain-directed stereotactic radiation or neurosurgery, interventions for which ASM prophylaxis is sometimes utilized.^19,20 Seizure-related diagnosis codes and ASMs were captured from BrM diagnosis through death or censoring.

Institutional Cohort

For the institutionally based analysis, we retrospectively identified 1,453 patients with newly diagnosed BrM managed at Brigham and Women's Hospital/Dana-Farber Cancer Institute between 2000 and 2015 without seizures at intracranial presentation. Development of seizures after initial diagnosis was based on manual chart review performed by 2 radiation oncologists specializing in tumors of the CNS. Seizures predating the imaging study used to diagnose brain metastases (e.g., MRI of the brain, CT of the head) were considered to have developed at intracranial presentation, and such patients were not included in our study. Seizures developing anytime thereafter were considered to have developed after initial diagnosis of brain metastases. Given that Massachusetts is not a SEER state, we expected minimal overlap between the SEER-Medicare and institutional cohorts.

Statistical Methodology

Our goal was to utilize 2 distinct data sources to determine the incidence of seizures among patients with BrM who were seizure-naive at intracranial presentation and also to assess demographic and clinical predictors of seizure risk in these patients.

Separate analyses were performed for the SEER-Medicare vs institutional cohorts. Categorical baseline characteristics among patients who did vs did not develop seizures were compared using the χ² test. Normally and non-normally distributed continuous covariates were compared between groups using the t test and Wilcoxon rank sum test, respectively. Cumulative seizure incidence over time was calculated and displayed graphically; comparisons were made with the Gray test.

We used univariable and multivariable Fine and Gray competing risks regression to identify predictors of seizure development, the primary outcome of the study, using death from any cause as a competing risk; models were adjusted for age at BrM diagnosis, sex, race, Charlson comorbidity index (assessed by the Deyo et al.²¹ method), primary tumor type, and initial BrM treatment strategy. The SEER-Medicare model also included marital status, high school completion rate (zip code–level), median household income (zip code–level), residence type (nonurban/unknown vs urban), and type of managing hospital (medical school–associated hospital vs not). The institutional models also included smoking history, Karnofsky performance status, and the following covariates abstracted at initial presentation of intracranial disease: presence of a BrM greater than 3 cm in maximal unidimensional size vs not, presence of more than 4 BrM vs not, presence of at least 1 BrM in a high-risk intracranial location for seizures (supratentorial frontal, parietal, temporal, or occipital regions) vs no BrM in high-risk locations (cerebellum, brainstem, basal ganglia, or thalamus), presence of hemorrhagic metastases vs not, and presence of leptomeningeal disease vs not. Race was based on self-report for both cohorts of patients; sex was self-reported for patients at Brigham and Women's Hospital/Dana Farber Cancer Institute and derived per SEER coding guidelines for the SEER-Medicare cohort.¹⁵ For delineation of the initial BrM treatment strategy using SEER-Medicare data, brain-directed stereotactic radiation could be readily identified via claims. Among patients who received brain-directed radiation in a nonstereotactic manner, we anticipate that the vast majority received whole brain radiation. Typically in a small percentage of patients, some centers in the United States administer partial brain radiation in a nonstereotactic manner, and both whole brain radiation and nonstereotactic partial brain radiation are captured among claims as brain-directed nonstereotactic radiation. For this reason, unlike our institutional patients whom we could subset based on receipt of stereotactic or whole brain radiation therapy (as we do not utilize nonstereotactic partial brain radiation at our institution), we were only able to divide SEER-Medicare patients receiving intracranial radiation into those receiving stereotactic vs nonstereotactic brain-directed radiation. For covariates with significant violations of the proportional hazards assumption, the interaction of the covariate with time was included in the model. A 2-sided p value <0.05 was considered statistically significant. Analyses were performed using SAS version 9.4.

Data Availability

SEER-Medicare data cannot be made publicly available in compliance with the National Cancer Institute guidelines. De-identified institutional data will be shared on reasonable request to the corresponding author pending institutional approval.

Results

SEER-Medicare Cohort

Among 15,863 Medicare patients without seizures at diagnosis of BrM, 1,588 (10.0%) subsequently developed seizures. Baseline patient characteristics stratified by development of seizures vs not are depicted in table 2.

Table 2.

Baseline Characteristics of Surveillance, Epidemiology, and End Results (SEER)–Medicare Patients With Brain Metastases

Cumulative incidence of seizures for the entire cohort at 1 year was 8.6% (figure 1A). When stratified by nonmelanoma vs melanoma, 1-year rates of seizure development were 8.3% vs 13.0%, respectively (p < 0.001; figure 1B). When stratified by non–Black vs Black patients, 1-year rates of seizure development were 8.4% vs 10.9%, respectively (p = 0.005; figure 1C). Assuming a 50% relative reduction in seizure incidence with the use of ASMs, the number needed to treat to prevent 1 seizure would be 25 for the general population.

Figure 1. Cumulative Incidence of Seizures Among Surveillance, Epidemiology, and End Results (SEER)–Medicare Patients Without Seizures at Brain Metastasis Diagnosis.

Cumulative incidence of seizures among SEER-Medicare patients without seizures at brain metastasis diagnosis for the entire cohort (A) and via stratification by primary tumor type (melanoma vs nonmelanoma) (B) and race (Black vs non-Black) (C). The number of patients at risk at each time point are displayed under each panel. Note that patients with unknown race were removed from (C) in order to comply with the National Cancer Institute policy of not displaying data for groups with n < 11. For the same reasons, exact number at risk below each curve is suppressed when displaying exact numbers that would allow an n < 11 to be directly derived.

In the adjusted regression models assessing time to first seizure, Black race vs White race (hazard ratio [HR] 1.45 [95% confidence interval (CI), 1.22–1.73], p < 0.001), higher median zip code–level household income (HR 1.04 [95% CI, 1.02–1.07], p = 0.001), urban vs nonurban/unknown residency (HR 1.41 [95% CI, 1.17–1.70], p < 0.001), melanoma as primary tumor type vs the reference of non-small cell lung cancer (NSCLC) (HR 1.44 [95% CI, 1.20–1.73], p < 0.001), and receipt of brain-directed stereotactic radiation vs the reference of nonstereotactic brain-directed radiation (HR 1.67 [95% CI, 1.44–1.94], p < 0.001) were associated with a greater risk for seizure development (table 3). Esophagus as the primary tumor type vs the reference of NSCLC (HR 0.51 [95% CI, 0.29–0.91], p = 0.02), type of managing hospital (non–medical school–associated vs medical school–associated, HR 0.87 [95% CI, 0.78–0.98], p = 0.02), and lack of brain-directed therapy (radiation or surgery) vs the reference of local nonstereotactic brain radiation (HR 0.68 [95% CI, 0.60–0.78], p < 0.001) were all associated with a decreased risk for seizure development (table 3). Upon removal of patients whose urban vs nonurban residential status was unknown, urban residency remained a significant risk factor for seizure development when compared to the reference of nonurban residence (HR 1.42 [95% CI, 1.18–1.91], p < 0.001).

Table 3.

Univariable and Multivariable Fine and Gray Competing Risks Regression for Seizure Development Among Medicare Patients With Brain Metastases (BrM)

Institutional Cohort

Among 1,453 patients without seizures at diagnosis of BrM, 169 (11.6%) subsequently developed seizures. Of these 169 seizures, 54 (32.0%) were focal aware, 31 (18.3%) were focal unaware (or focal impaired awareness), 65 (38.5%) were bilateral tonic-clonic or other convulsive type, and 19 (13.7%) were of indeterminate type. Baseline patient characteristics based on development of seizures vs not are depicted in table 4.

Table 4.

Baseline Characteristics of Single-Institution Patients With Brain Metastases (BrM)

Among institutional patients, cumulative incidence of seizures at 1 year was 7.3% (figure 2A). When stratified by nonmelanoma vs melanoma, 1-year rates of seizure development were 5.9% vs 15.0%, respectively (p < 0.001; figure 2B). When comparing patients with ≤4 BrM to those with >4 BrM, 1-year seizure rates were 6.9% vs 8.4%, respectively (p = 0.05; figure 2C). Finally, 1-year incidence of seizures for patients with BrM in low-risk vs high-risk locations were 2.0% vs 9.8%, respectively (p < 0.001; figure 2D). Assuming a 50% relative reduction in seizure incidence with the use of ASMs, the number needed to treat to prevent 1 seizure would be 25 for the general population.

Figure 2. Cumulative Incidence of Seizures Among Brigham and Women's Hospital/Dana Farber Cancer Institute (BWH/DFCI) Patients Without Seizures at Brain Metastasis (BrM) Diagnosis.

Cumulative incidence of seizures among BWH/DFCI patients without seizures at BrM diagnosis for the entire cohort (A) and via stratification by primary tumor type (melanoma vs nonmelanoma) (B), number of BrM present at diagnosis of intracranial disease (>4 vs ≤4) (C), and location of BrM at diagnosis of intracranial disease (any high-risk/epileptogenic locations vs entirely low-risk/nonepileptogenic locations) (D). The number of patients at risk at each time point is displayed under each panel.

In the adjusted regression models assessing time to seizure, melanoma as primary tumor type vs NSCLC (HR 1.70 [95% CI, 1.09–2.64], p = 0.02), harboring >4 BrM vs ≤4 BrM at intracranial presentation (HR 1.60 [95% CI, 1.12–2.29], p = 0.01), the presence of BrM in a high-risk vs low-risk intracranial location at intracranial presentation (HR 3.62 [95% CI, 1.60–8.18], p = 0.002), and lack of local, brain-directed therapy vs whole-brain radiotherapy (HR 3.08 [95% CI, 1.45–6.52], p = 0.003) were associated with a greater risk for seizures (table 5). When assessing the association between the location of supratentorial disease and seizure risk, we found that the presence of frontal metastases vs not (HR 2.25 [95% CI, 1.66–3.07, p < 0.001) and parietal metastases vs not (HR 1.72 [95% CI, 1.24–2.40], p = 0.001) were associated with seizure development, while no significant association was seen for temporal (HR 1.35 [95% CI 0.94–1.93], p = 0.10) and occipital (HR 1.29 [95% CI, 0.89–1.85], p = 0.17) disease.

Table 5.

Univariable and Multivariable Fine and Gray Competing Risks Regression for Seizure Development Among Single-Institution Patients With Brain Metastases (BrM)

Discussion

In this study of over 17,000 patients across 2 distinct cohorts, we described the incidence and risk factors for development of seizures among patients with BrM using data from both a population-based registry and a comprehensive cancer center. Our data suggest that for a typical patient without seizures at initial intracranial presentation, a lifetime seizure risk of 10%–11% can be expected. However, incidence is higher among a subset of patients, including those with melanoma as their primary tumor type, patients of Black race, those with a significant intracranial disease burden, or those with BrM in epileptogenic intracranial locations.

Current guidelines suggest that ASMs should only be prescribed to patients with BrM who have already had a seizure; upfront initiation of ASMs is not recommended in seizure-naive patients with BrM as primary prevention.^7–9 These guidelines are supported by a small number of randomized controlled trials (RCTs) and retrospective studies in seizure-naive patients that failed to demonstrate a difference in seizure incidence among patients prescribed ASMs vs not. In one study by Glantz et al.,¹¹ 74 patients with newly diagnosed brain tumors (59 with BrM) and no prior seizure history were randomized to receive divalproex sodium or placebo. No significant difference in seizures was noted between groups (odds ratio for seizure development comparing divalproex sodium to placebo group 1.7 [95% CI, 0.6–4.6; p = 0.3]), although given the considerably wide CI, the overall utility of the data is limited. Similar concerns apply to an RCT by Forsyth et al.,¹⁰ in which 100 seizure-naive patients with brain tumors, of which 60 had BrM, were randomized to receive phenytoin or phenobarbital vs no anticonvulsant. Although no significant difference in seizure rates between the 2 groups was observed at 3 months, given the study's small sample size and guarded power, no definitive conclusions could be drawn. Collectively, these studies have provided the rationale for withholding ASMs in patients with BrM and no prior seizure history.⁹

Among seizure-naive patients status postcraniotomy for resection of a supratentorial brain metastasis, data on whether to administer ASMs are conflicting. A retrospective study,²² as well as 2 RCTs including patients with BrM,^23,24 demonstrated no significant difference in seizure rate among patients who did vs did not receive ASMs after neurosurgical resection. However, other studies have suggested benefit to routine utilization of ASMs in this population,^25,26 and some consensus guidelines generally support temporary use of ASMs in such patients.⁷ However, future studies are needed to further elucidate specific indications for and the duration of ASMs in the postcraniotomy setting among patients with resected BrM.

Although the aforementioned studies by Glantz et al.¹¹ and Forsyth et al.¹⁰ did not demonstrate a role for routine use of ASMs in the BrM population at large, these studies also included patients with primary brain tumors, limiting the power of BrM-specific subgroup analyses. Moreover, these RCTs utilized older ASMs that are not as commonly prescribed today, limiting their applicability to the modern neuro-oncologic landscape. In addition, potentially more epileptogenic therapies such as immunotherapy and stereotactic radiation, both of which can contribute to necrosis that secondarily increases seizure risk,^12–14 were not commonly used in such eras. Therefore, although prior RCTs did not demonstrate preventative utility of ASMs in the BrM population at large, given the small and heterogeneous patient populations studied, they may have been unable to identify effect modification suggesting a benefit in higher-risk subgroups. Here, we assessed seizure incidence at a national and institutional level in contemporary cohorts of patients with BrM and detected a seizure rate between 10% and 11%, relatively consistent with prior reports in the literature.³ However, given the size of our cohorts, relative to most prior work, we were able to identify particular subgroups at especially high risk of seizures who might benefit from ASM therapy. Even among these higher risk cohorts, given the side effects associated with ASMs, the ultimate decision regarding whether or not to initiate an ASM should be made on an individual level after careful patient–provider discussion, and based in part on a patient's performance status, how motivated he or she is to minimize seizure risk given his or her daily activities, and how well-tolerated an initial trial of ASM therapy is by a particular patient.

In both datasets, we found that patients with melanoma were more likely to develop seizures as compared to the reference of patients with NSCLC. This is consistent with prior work that has suggested that patients with melanoma-associated BrM may be at higher risk for seizures, possibly due to the hemorrhagic nature of such lesions, degree of surrounding edema, or the routine use of immunotherapy in this condition.^27–29

We also found a significantly greater incidence of seizures among Black patients as compared to White patients in the SEER-Medicare cohort (HR 1.45, p < 0.001), although a significant association among institutional patients was not seen (HR 1.76, p = 0.09). There are several reasons that could potentially account for the differential rates we observed across races. First, widespread disparities in access to care and degree of follow-up exist among races, possibly contributing to later detection and worse outcomes among Black populations with BrM.^30–32 Given that Black patients may have fewer physician encounters and lower rates of surveillance imaging compared to White patients following a diagnosis of BrM, undetected tumor growth and subsequently delayed interventions related to such growth could predispose Black patients to developing seizures. In addition, it has previously been shown that among patients with metastatic cancer, Black patients are significantly less likely to receive supportive medications for common symptoms compared to their White counterparts.^33,34 In the BrM population specifically, Black patients have been shown to be less likely to receive steroids than White patients, and the link between uncontrolled edema (given mitigation of edema with steroids) and seizures may explain some of the differences in seizure rates that we observed.^34,35 Moreover, compared to White patients, Black patients with cancer have previously been shown to experience more cancer-related financial problems and to more frequently forego prescription medications due to cost.^36,37 Finally, numerous studies have documented higher rates of medical mistrust among Black patients, which has been linked to underutilization of care.^38,39 Hence, the higher incidence of seizures we observed among Black patients could be reflective of more limited follow-up care, poorer access to ASMs, more limited use of other supportive medications such as steroids, implications of out-of-pocket cost, or disengagement with the medical system.

Interestingly, among the SEER-Medicare cohort, we also found an increased incidence of seizures among patients living in an urban residence compared to those living in a nonurban residence. Although future studies will be needed to fully explore this association, it is possible that patients living in an urban environment are more frequently able to seek medical care following a seizure due to their proximity to hospitals compared to those in nonurban environments and that this increased risk represents a detection bias. Further study of this association is warranted.

In addition, while not possible with SEER-Medicare data, our institutional data afforded the opportunity to associate seizure risk with the location and burden of intracranial disease. We found that patients harboring metastases located in regions traditionally associated with higher seizure risk^40,41 (supratentorial frontal, parietal, temporal, or occipital lobes) vs lower seizure risk (cerebellum, brainstem, basal ganglia, or thalamus) displayed greater seizure incidence. We also found that patients with >4 BrM were substantially more likely to develop seizures as compared to patients with ≤4 BrM, consistent with prior work that has demonstrated that the number of intracranial lesions correlates with seizure risk.⁴² Although the mechanism by which BrM cause seizures is not well-understood, it is thought that by inducing inflammation and hypoxia in the nearby environment, BrM lead to metabolic imbalances that may precipitate seizure activity.⁴³ It follows then that the intracranial environment among patients with a higher number of BrM is subject to a greater degree of dysregulation and subsequently an increased risk for seizures.

Taken together, the above results suggest that certain populations with BrM, such as patients with melanoma, with disease in sensitive intracranial locations, or with a greater intracranial disease burden, may be at higher risk of developing seizures after initial BrM diagnosis. While ASMs are currently not routinely prescribed to patients with BrM who do not present with a seizure, our study suggests that there may be a role for ASMs in higher-risk subpopulations. This suggestion is reinforced by the high percentage of seizures we identified that impaired cognition or awareness (either focal unaware or those that progressed to convulsions), which could have marked consequences if patients are operating a motor vehicle, walking in a perilous location, or caring for loved ones at the time of seizure. Given that newer ASMs (e.g., levetiracetam, lacosamide, topiramate, and oxcarbazepine) have proven efficacy and better side effect profiles compared to older drugs,⁴⁴ re-evaluation of the role of ASMs in high-risk, seizure-naive patients with BrM seems reasonable. Although prospective studies are needed to determine the true effect of ASM utilization in these subgroups, our data suggest that for select patients who are motivated to minimize seizure risk and who demonstrate reasonable medication tolerance following an initial trial of ASMs, renewed consideration of longer-term ASM utilization among the oncology community may be warranted.

Finally, although we also detected a higher rate of seizures among Black patients, it should be noted that unlike primary tumor site and intracranial disease burden, which have a biological basis that could explain their association with an increased risk for seizures, as discussed above, the differences in seizure rates we observed among races are more likely reflective of socioeconomic disparities, access to care, and other nonbiologic factors. Future work is needed to identify and overcome the structural barriers that could be underlying such race-based differences in symptomatology and oncologic outcomes in this population.

Our work should be considered in the context of its limitations. First, the National Cancer Institute advises caution when utilizing claims to identify metastases after primary cancer diagnosis.⁴⁵ However, in contrast to other metastatic sites, BrM are commonly treated with local therapies for which diagnostic/billing codes exist; claims data can therefore be used to reliably identify BrM, a methodology that has previously been validated with manual chart review and shown to have a high sensitivity (>97%) and specificity (99%) for identification of BrM.^17,18 Secondly, for the SEER-Medicare analysis, given that prescription claims data for patients on hospice may be incomplete, seizure incidence may have been underestimated in this cohort due to incomplete capture of ASMs following entry into hospice. Moreover, we acknowledge that the risk for seizure development may increase or decrease throughout a patient's disease course based on tumor-related changes, oncologic intracranial interventions such as radiation or surgery, and other factors; the focus of this work was to characterize seizure risk secondary to characteristics at the time of intracranial disease diagnosis. In addition, while beyond the scope of the present work, future studies should study the rates of emergency department visits or inpatient hospitalizations secondary to seizures in the BrM population and analyze the cost associated with such potentially preventable visits. Despite these limitations, the utilization of data from both a population-based registry, as well as a single institution, was a strength of our study and allowed us to both independently examine and validate the incidence and risk factors for seizure development in patients with BrM. Ultimately, however, these hypothesis-generating data need to be confirmed in the setting of an RCT before warranting changes to existing clinical guidelines.

In this study of over 17,000 seizure-naive patients with newly diagnosed BrM, we describe the incidence and risk factors for subsequent seizure development at both a population and single-institution level. The results of our study suggest that a significant proportion of patients with BrM will develop seizures and that certain subgroups may be at particularly high risk for seizure development. Given that seizures have the potential to cause serious harm and also profoundly decrease quality of life for patients with BrM, these findings may be useful in counseling patients about individual seizure risk and may inform future studies evaluating the role of upfront initiation of ASMs for particularly vulnerable subgroups.

Glossary

ASM

antiseizure medication

BrM

brain metastases

CI

confidence interval

HR

hazard ratio

ICD-9-CM

International Classification of Diseases, 9th revision, Clinical Modification

ICD-10-CM

International Classification of Diseases–10, Clinical Modification

NSCLC

non-small cell lung cancer

SEER

Surveillance, Epidemiology, and End Results

Appendix. Authors

Study Funding

No targeted funding reported.

Disclosures

Dr. Lamba, Dr. Catalano, Dr. Cagney, Dr. Haas-Kogan, Dr. Bubrick, and Dr. Wen report no disclosures. Dr. Aizer reports research funding from Varian Medical Systems and consulting fees from Novartis. Go to Neurology.org/Nhttps://n.neurology.org/lookup/doi/10.1212/WNL.0000000000011459 for full disclosures.