Frequency, etiologies, risk factors, and sequelae of falls among patients with brain metastases: A population- and institutional-level analysis

Abstract

Background

Falls in patients with cancer harbor potential for serious sequelae. Patients with brain metastases (BrM) may be especially susceptible to falls but supporting investigations are lacking. We assessed the frequency, etiologies, risk factors, and sequelae of falls in patients with BrM using 2 data sources.

Methods

We identified 42 648 and 111 patients with BrM utilizing Surveillance, Epidemiology, and End Results (SEER)-Medicare data (2008-2016) and Brigham and Women’s Hospital/Dana-Farber Cancer Institute (BWH/DFCI) institutional data (2015), respectively, and characterized falls in these populations.

Results

Among SEER-Medicare patients, 10 267 (24.1%) experienced a fall that prompted medical evaluation, with cumulative incidences at 3, 6, and 12 months of 18.0%, 24.3%, and 34.1%, respectively. On multivariable Fine/Gray’s regression, older age (≥81 or 76-80 vs 66-70 years, hazard ratio [HR] 1.18 [95% CI, 1.11-1.25], P < .001 and HR 1.10 [95% CI, 1.04-1.17], P < .001, respectively), Charlson comorbidity score of >2 vs 0-2 (HR 1.08 [95% CI, 1.03-1.13], P = .002) and urban residence (HR 1.08 [95% CI, 1.01-1.16], P = .03) were associated with falls. Married status (HR 0.94 [95% CI, 0.90-0.98], P = .004) and Asian vs white race (HR 0.90 [95% CI, 0.81-0.99], P = .03) were associated with reduced fall risk. Identified falls were more common among BWH/DFCI patients (N = 56, 50.4% of cohort), resulting in emergency department visits, hospitalizations, fractures, and intracranial hemorrhage in 33%, 23%, 11%, and 4% of patients, respectively.

Conclusions

Falls are common among patients with BrM, especially older/sicker patients, and can have deleterious consequences. Risk-reduction measures, such as home safety checks, physical therapy, and medication optimization, should be considered in this population.

Falls are common among patients with advanced cancer, with recent studies reporting fall rates over a 6- to 12-month period of 20%-30%.^1,2 Falls have the potential to result in serious injury and functional decline,³ which can be particularly deleterious for older patients who have additional comorbidities and a lower physiologic reserve.⁴ In addition, the fear of falling again and the loss of independence after a fall can adversely affect patient quality of life.^5–7 Among patients with advanced cancer, patients with brain metastases (BrM) may be at especially high risk of falling given the neurologic compromise often associated with the presence or treatment of intracranial disease.

BrM affect approximately 20%-40% of patients with solid malignancies and can be associated with significant symptomatology, including proprioceptive loss, visual compromise, dizziness, ataxia, and weakness, all of which have the potential to precipitate falls.^8,9 Although prior work has characterized the incidence of and risk factors for falls among patients with cancer,^1,2 large-scale studies focused on patients with BrM, a particularly susceptible patient population, are lacking. Moreover, a more robust understanding of the causes of falls in patients with BrM is needed in order to develop strategies for risk reduction.

Here, we performed population- and institutional-level analyses of older patients with BrM to characterize the incidence, risk factors, and sequelae of falls in this population. Utilization of the Surveillance, Epidemiology, and End Results (SEER)-Medicare database allowed us to assess this risk of clinically significant falls among a large, generalizable group of patients with BrM, and the use of institutional data from a tertiary cancer center allowed for a more granular assessment of patient- and disease-related risk factors.

Methods

Patient Population and Study Design

SEER-Medicare cohort

The SEER registry contains demographic and clinical information for approximately 34.8% of cancer patients in the United States.¹⁰ For approximately 93% of Medicare patients in the SEER database, the SEER-Medicare program links Medicare claims data to SEER data.¹¹ 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 malignant neoplasm of the brain, cerebral meninges, and spinal cord, a methodology associated with a 97% sensitivity and 99% specificity in identifying patients with BrM via claims data.¹² Although in practice these diagnosis codes may also be utilized for patients with spinal cord metastases or leptomeningeal disease, validation efforts have demonstrated high specificity of these diagnosis codes in the identification of patients with intraparenchymal brain metastases, which was the population of interest for this study.^12,13 We utilized the date of the first BrM-associated claim as the BrM diagnosis date, an approach shown to be 92% sensitive for predicting the actual date of BrM diagnosis to within 30 days.¹³

Our initial cohort consisted of 68 207 patients with BrM aged 66 years or above at time of primary cancer and BrM diagnosis. To reliably identify patients who experienced a fall after a diagnosis of BrM, we excluded patients who entered hospice prior to BrM diagnosis (N = 1186), as well as those who did not have continuous Part A and B coverage or who had HMO enrollment from the year prior to BrM diagnosis through the date of censoring or death (N = 23 173). Given that comorbidity scores were calculated based on claims in the year prior to diagnosis, patients diagnosed with BrM at autopsy/death certificate were excluded (N = 413). Patients with missing zipcode-level information were excluded (N = 787). A total of 42 648 patients met the eligibility criteria (Supplementary Figure 1).

To determine whether a patient experienced a fall after BrM diagnosis, we searched for a diagnosis code for a fall (Table 1) any time after the BrM diagnosis date; to account for inaccuracies in dates linked to claims, we also included claims within 15 days preceding the BrM diagnosis date. We assessed for the following fall-related sequelae: (1) intracranial injury, including concussions, cerebral contusions, or laceration/intracranial bleeding and (2) fractures or dislocations (Table 1). In order for any sequelae to count as being fall-related, we mandated the dates of these claims be within 15 days of the date of the fall-related claim. Diagnosis codes for falls and their sequelae were captured from BrM diagnosis through the time of death or censoring.

Table 1.

Medicare Codes for Falls and Sequelae of Falls

ICD Diagnosis Codes
ICD-9-CM Falls (corresponding ICD-10-CM)	E880.X-E888.Xa, E804.0, E804.1, E804.2, E804.3, E804.8, E804.9, E987.0, E987.1, E987.2, E987.9, 7802 (W0XXXXX-W1XXXXX, V0XXXXX, V815XXA, R55, R296)
ICD-9-CM Intracranial injury	800.XX, 801.XX, 803.XX, 804.XX, 850.XX,851.XX, 852.XX, 853.XX, 854.XX, 430, 431, 432, 432.0, 432.1, 432.9
ICD-10-CM Intracranial injury	S02.0XXX, S02.1XXX, S02.8XXX, S02.9XXX, S06.0XXX, S06.1XXX, S06.2XXX, S06.3XXX, S06.4XXX, S06.5XXX, S06.6XXX, S06.8XXX, S06.9XXX, I60.XX, I61.XX, I62.XX
ICD-9-CM Fracture or dislocation	802.XX, 805.XX-839.XX
ICD-10-CM Fracture or dislocation	S03.XXXX, S12.XXXX, S13.XXXX, S22.XXXX, S23.XXXX, S32.XXXX, S33.XXXX, S42.XXXX, S43.XXXX, S52.XXXX, S53.XXXX, S62.XXXX, S63.XXXX, S72.XXXX, S73.XXXX, S82.XXXX, S83.XXXX, S92.XXXX, S93.XXXX

Abbreviations: ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10-CM, International Classification of Diseases, Tenth Revision, Clinical Modification.

^aExcludes E887 (fracture cause unspecified).

“X” refers to any potential numbers and serves as a placeholder to encompass any ICD codes in the fall category.

Institutional cohort

For the institutionally based analysis, we retrospectively identified 111 patients with newly diagnosed BrM at any age who received care at Brigham and Women’s Hospital/Dana-Farber Cancer Institute (BWH/DFCI) between June 1, 2015 and December 31, 2015. This start date was chosen based on a transition in the institutional electronic medical record system utilized at BWH/DFCI, which facilitated the collection of complete information relating to falls. In addition, we chose a total duration of 6 months in an attempt to choose a representative time period. The likely etiology of the fall was abstracted via chart review and included the following categories: (1) weakness, (2) ataxia, (3) syncope, (4) mechanical fall, (5) other/unknown cause. Seizures were not counted as falls; accordingly, if a patient had a seizure, lost consciousness, and subsequently experienced trauma, we did not consider such an event to constitute a fall.

The consequences of the fall, including fractures, intracranial bleeding, and/or subsequent escalation of care to the emergency department or hospital, were ascertained from the electronic medical record via manual chart review. Of note, as Massachusetts is not a SEER state, we expected minimal overlap between the BWH/DFCI and SEER-Medicare cohorts.

Statistical Methodology

Separate analyses were performed for the SEER-Medicare and institutional cohorts. Categorical baseline characteristics among patients who did vs did not fall were compared using the chi-square or, when feasible, the Fisher’s exact test (P-values were not provided for the SEER-Medicare cohort due to the large sample size). Normally and non-normally distributed continuous covariates were compared between groups using the unpaired t-test and Wilcoxon rank sum test, respectively.

In the SEER cohort, we used univariable and multivariable Fine and Gray’s competing risks regression to identify predictors of experiencing a fall using death from any cause as a competing risk; models were adjusted for age at BrM diagnosis, sex, race, Charlson comorbidity index (CCI, assessed by Deyo et al’s method),¹⁴ marital status, high school completion rate (zipcode level), median household income (zipcode level), residence type (non-urban/unknown vs urban), and primary tumor type. The proportional hazards assumption was assessed for all variables; for covariates with significant violations of the proportional hazards assumption, the interaction of the covariate with time was included in the regression model. Given the small number of patients in the institutional cohort, statistical modeling was not performed with this dataset; instead, descriptive statistics were used to describe the etiology and sequelae of falls. A 2-sided P-value <.05 was considered statistically significant. Analyses were performed using SAS version 9.4. This study was approved by our institutional review board; a waiver for informed consent was provided for institutional patients.

Results

SEER-Medicare Cohort

Among 42 648 Medicare patients with BrM, 1456 were alive at the time of last follow-up. Median follow-up among surviving patients without a fall was 20.4 months. Of the 42 648 Medicare patients with BrM, 10 267 (24.1%) experienced a fall that was captured by claims data in the period following BrM diagnosis. Baseline patient characteristics stratified by patients who experienced a fall vs not were similar overall (Table 2). Cumulative incidence of falls among the entire cohort at 3 months, 6 months, and 1 year were 18.0%, 24.3%, and 34.1%, respectively (Figure 1A). In the adjusted regression models assessing time-to-first fall, older age at BrM diagnosis (≥81 or 76-80 vs the reference of ≤70 years, hazard ratio [HR] 1.18 [95% CI, 1.11-1.25], P < .001, and HR 1.10 [95% CI, 1.04-1.17], P < .001, respectively), Charlson comorbidity score of >2 vs the reference of 0-2 (HR 1.08 [95% CI, 1.03-1.13], P = .002) and urban residence vs the reference of non-urban/unknown (HR 1.08 [95% CI, 1.01-1.16], P = .03) were associated with falls (Table 3). Married social status (HR 0.94 [95% CI, 0.90-0.98], P = .004) and Asian vs the reference of white race (HR 0.90 [95% CI, 0.81-0.99], P = .03) were associated with a decreased risk for falling (Table 3). As an exploratory analysis, we assessed whether particular comorbidities were more likely to be associated with falling in this patient population and found that peripheral vascular disease, cerebrovascular disease, dementia, and diabetes (with or without complications) were all predictive of falling (Supplementary Table 1). Cumulative incidence of falls stratified by significant covariates, including age group, marital status, Charlson comorbidity score, and race are displayed in Figure 1.

Table 2.

Baseline Characteristics of SEER-Medicare Patients With Brain Metastases

	Did Not Fall (N = 32 381)	Did Fall (N = 10 267)
Sex, N (%)
Male	15 877 (49)	4790 (47)
Female	16 504 (51)	5477 (53)
Race, N (%)
White	26 943 (83)	8592 (84)
African American	2490 (8)	780 (8)
Hispanic	1374 (4)	430 (4)
Asian/Pacific Islander	1456 (5)	413 (4)
Other/Unknown	118 (<1)	52 (1)
Marital status, N (%)
Married/domestic partnership	17 578 (54)	5307 (52)
Unmarried/single	13 305 (41)	4383 (43)
Unknown	1498 (5)	577 (6)
Type of residence, N (%)
Urban	28 637 (88)	9231 (90)
Non-urban	3716 (11)	1036 (10)
Unknown	28 (<1)
Percentage graduated from high school, median (IQR)a	87 (79-92)	87 (80-93)
Household income (per 10K USD), median (IQR)	5.3 (3.9-7.1)	5.4 (4.0-7.3)
Age at diagnosis of brain metastases, years, mean (SD)	76 (6)	76 (6)
Charlson comorbidity index, N (%)b
0-2	25 564 (79)	7985 (78)
>2	6356 (20)	2143 (21)
Unknown	461 (1)	139 (1)
Primary tumor type, N (%)
NSCLC	17 044 (53)	5064 (49)
Breast	3007 (9)	1169 (11)
Melanoma	2125 (7)	720 (7)
Other	10 205 (32)	3314 (32)

Abbreviations: N, number; NSCLC, non-small cell lung cancer; SD, standard deviation; USD, United States Dollars.

Note: (1) Percentages may not add up to 100 due to rounding; (2) Categories for certain variables were grouped together so as to comply with NCI data policy of not displaying any cells with values <11.

^aZipcode level.

^bExcluded diagnosis of metastatic cancer so as not to inflate all scores by 6 points.

Figure 1.

Cumulative incidence of falls among SEER-Medicare patients after a diagnosis of brain metastases. Cumulative incidence of falls among SEER-Medicare patients after brain metastasis diagnosis for the entire cohort (A) and via stratification by age (B) marital status (C) Charlson comorbidity index (D), and race (Asian vs non-Asian) (E). The number of patients at risk at each time point are displayed under each panel. Note that patients with unknown marital status, Charlson comorbidity index, and race were removed in order to comply with the NCI policy of not displaying data for groups with N < 11. Of note, for patients with falls in the 15 days leading up to a diagnosis of brain metastases, time-to-fall was considered “zero” to avoid a negative time to event. Abbreviations: BrM, brain metastases; NCI, National Cancer Institute; SEER, Surveillance, Epidemiology, and End Results.

Table 3.

Univariable and Multivariable Fine and Gray Competing Risks Regression for Time to Fall Among Medicare Patients With Brain Metastases

	Univariable HR (95% CI)	Univariable P	Multivariable HR (95% CI)	Multivariable P
Age at BrM diagnosis, years
66-70	Ref		Ref
71-75	1.04 (0.99-1.10)	.13	1.03 (0.97-1.09)	.32
76-80	1.13 (1.07-1.20)	<.001	1.10 (1.04-1.17)	<.001
≥81	1.23 (1.16-1.30)	<.001	1.18 (1.11-1.25)	<.001
Sex		<.001		.13
Male	Ref		Ref
Female	1.08 (1.04-1.12)		1.03 (0.99-1.08)
Race
White	Ref		Ref
African American	0.99 (0.92-1.06)	.74	1.00 (0.93-1.08)	.96
Hispanic	0.98 (0.89-1.08)	.74	0.98 (0.89-1.09)	.75
Asian/Pacific Islander	0.89 (0.81-0.98)	.02	0.90 (0.81-0.99)	.03
Other/Unknown	1.33 (1.02-1.74)	.04	1.31 (1.00-1.71)	.05
Marital status at diagnosis
Unmarried/single	Ref		Ref
Married/partnered	0.92 (0.88-0.96)	<.001	0.94 (0.90-0.98)	.004
Unknown	1.15 (1.05-1.25)	.002	1.14 (1.04-1.24)	.004
Graduated from high school (per % increase)	1.00 (1.00-1.01)	<.001	1.00 (1.00-1.00)	.81
Household income (per 10K USD increase)	1.02 (1.01-1.03)	<.001	1.02 (1.01-1.03)	<.001
Household income (per 10K USD increase) * log(time)a	1.01 (1.01-1.01)	<.001	1.01 (1.01-1.01)	<.001
Residence		<.001		.03
Non-urban/Unknownb	Ref		Ref
Urban	1.14 (1.07-1.21)		1.08 (1.01-1.16)
Charlson comorbidity index
0-2	Ref		Ref
>2	1.09 (1.04-1.14)	<.001	1.08 (1.03-1.13)	.002
Unknown	0.99 (0.84-1.17)	.91	1.02 (0.87-1.21)	.78
Primary tumor type
NSCLC	Ref		Ref
Breast	1.26 (1.18-1.34)	<.001	1.20 (1.12-1.28)	<.001
Breast * log(time)a	1.07 (1.04-1.09)	<.001	1.07 (1.04-1.09)	<.001
Melanoma	1.12 (1.04-1.21)	.004	1.07 (0.99-1.16)	.09
Other	1.08 (1.03-1.13)	<.001	1.07 (1.02-1.12)	.003

Abbreviations: BrM, brain metastases; CI, confidence interval; HR, hazard ratio; N, number; NSCLC, non-small cell lung cancer; USD, United States Dollars.

^aFor covariates with significant violations of the proportional hazards assumption, the interaction of the covariate with time (denoted by *) was included in the model. Time reflects the period between BrM diagnosis and a fall event, a competing event, or censoring. A P-value <.05 of a covariate * log(time) indicates that the HR of the covariate varies with time.

^b“Non-urban” and “unknown” categories were grouped in this manner to allow for model convergence.

Among the 10 267 patients who experienced a fall, 1633 (15.9%) experienced a fracture/dislocation, while 941 (9.2%) patients experienced an intracranial injury/bleed.

Institutional Cohort

Among 111 patients diagnosed with BrM at BWH/DFCI between June 1, 2015 and December 31, 2015, 53 were alive at the time of last follow-up. Median follow-up among surviving patients without a fall was 13.5 months. Of the 111 BWH/DFCI patients with BrM, 56 (50.4%) experienced a fall after BrM diagnosis. Baseline characteristics for patients from BWH/DFCI with BrM who experienced a fall after BrM diagnosis vs not are presented in Table 4.

Table 4.

Baseline Characteristics of Single-Institution Patients With Brain Metastases Who Experienced a Fall

	Did Not Fall (N = 55)	Did Fall (N = 56)	P
Age at diagnosis of BrM, years, mean (SD)	61 (12)	63 (11)	.55
Sex, N (%)			1.00
Male	22 (40)	23 (41)
Female	33 (60)	33 (59)
Race, N (%)			.26
White	45 (82)	52 (93)
African American	3 (5)	3 (5)
Hispanic	1 (2)	0 (0)
Asian/Pacific Islander	3 (5)	1 (2)
Other/Unknown	3 (5)	0 (0)
Karnofsky performance status, N (%)			.71
30-80	28 (51)	26 (46)
90-100	27 (49)	30 (54)
Charlson comorbidity index, N (%)			.09
0-1	53 (96)	48 (86)
>1	2 (4)	8 (14)
Numberof BrM at diagnosis of intracranial involvement, median (IQR)	2 (1-4)	3 (1-6)	.11
Largest BrM in mm at diagnosis of intracranial involvement, median (IQR)	16 (9-27)	19 (11-29)	.44
Primary tumor type, N (%)			.97
NSCLC	23 (42)	22 (39)
Breast	9 (16)	11 (20)
Melanoma	5 (9)	6 (11)
Othera	18 (33)	17 (30)
Initial brain-directed treatment strategy, N (%)			.50
WBRT, without SRS/SRT or resection	13 (24)	16 (29)
SRS/SRT, without resection	30 (55)	24 (43)
Any resection with or without radiation	12 (22)	15 (27)
No local therapy	0 (0)	1 (2)

Abbreviations: BrM, brain metastases; IQR, interquartile range; mm, millimeters; N, number; NSCLC, non-small cell lung cancer; SD, standard deviation; SRS/SRT, stereotactic radiosurgery/radiation therapy; WBRT, whole-brain radiotherapy.

Note: (1) Percentages may not add up to 100 due to rounding.

^aSites included lung, liver, bone, adrenal gland, distant lymph nodes, and soft tissue.

Among the 56 patients who fell, there were a total of 171 falls, with 50% of patients (n = 28) experiencing more than fall. The most common etiologies of falls were: mechanical, (28% of falls), followed by weakness (16% of falls) and ataxia (16% of falls; Table 5). Of the 171 falls among the institutional cohort, 57 (33%) resulted in an emergency department visit, 39 (23%) resulted in an inpatient hospitalization, 19 (11%) were associated with a subsequent fracture, and 6 (4%) were associated with intracranial hemorrhage (Supplementary Figure 2).

Table 5.

Etiologies and Sequelae of Falls Among Single-Institution Patients With Brain Metastases Who Experienced a Fall

Etiology of Fall	Number of Patientsa (%)	Number of Fallsb (%)
Weakness	14 (25)	27 (16)
Ataxia	14 (25)	27 (16)
Syncope	6 (11)	8 (5)
Mechanical	21 (38)	48 (28)
Other	3 (5)	3 (2)
Sequelae of Falls
Emergency department visit	29 (52)	57 (33)
Inpatient hospitalization	25 (45)	39 (23)
Intracranial hemorrhage	6 (11)	6 (4)
Fracture	9 (16)	19 (11)

^aPertaining to the 56 patients who fell.

^bOf a total of 171 falls.

Discussion

In this study, we described the incidence, risk factors, causes, and sequelae of falls in patients with BrM using both a population-based registry as well as data from a tertiary cancer center. At a population level, we found that 24.1% of patients experienced a fall after a diagnosis of brain metastases, while at the institutional level, we found that this rate was significantly higher (over 50%). Among the SEER-Medicare cohort, we found that especially old patients were at higher risk of falls, as were patients harboring greater comorbidity, while married individuals were less at risk for falling. Although the HRs for the above predictors of falls were modest in magnitude, the absolute increase in fall risk linked to these predictors is meaningful given that the absolute incidence of falls we observed in both cohorts was substantial. Among patients who fell, consistent across both the SEER-Medicare and institutional datasets, we found that a high rate of serious sequelae ensued, including fractures in 15.9% of SEER-Medicare patients and 16% of BWH/DFCI patients, and intracranial injury/bleeding in 9.1% of SEER-Medicare patients and 11% of BWH/DFCI patients. In addition, among BWH/DFCI patients, we found that 33% of falls led to an emergency department visit, and 23% led to a hospitalization. These data suggest that falls are common among older patients with BrM and may have serious consequences on their functional status and well-being.

The utilization of both population- and institutional-level datasets allowed us to explore different aspects of our study question by harnessing the strengths of each dataset; although the SEER-Medicare data captured only falls that resulted in a management claim (ie, likely more severe falls), the large number of patients in the SEER cohort facilitated statistical analyses assessing risk factors for falls, and the generalizable nature of SEER data allowed us to feel confident that the results obtained would apply to most patients on a national scale. In contrast, while the total number of patients in our institutional cohort was small, all data were collected via manual chart review and therefore not dependent on billing codes to identify whether a patient experienced a fall or not. We were also able to explore the causes of falls using manual chart review. Collectively, the 2 datasets allowed us to characterize the burden of falls among patients with BrM comprehensively.

Prior work has demonstrated that falls are a common problem among patients with cancer, with most studies reporting rates between approximately 20%-30% over 3-12 months,² and a subset of studies identifying even higher rates.^1,15 For example, in a population-based study of older cancer survivors, the prevalence of falls among patients with prostate and lung cancer in the 1-2 years prior to cancer diagnosis were 12% and 17%, respectively, while rates in the 1-2 years after cancer diagnosis rose to 17%-20% and 28%, respectively.¹⁵ In another prospective study of 185 patients with metastatic or locoregionally advanced cancer, the authors found that the presence of intracranial malignancy (primary or metastatic) was an independent risk factor for falling, although the absolute number of patients with primary/secondary intracranial disease was reasonably small (N = 24) and particular risk factors for and sequelae related to falling within this population were not reported on.¹ Although these studies have characterized the serious nature of falls among patients with advanced cancer, none have focused exclusively on patients with brain metastases, a group that appears to be at particularly high risk of experiencing a fall.

When assessing risk factors for falls within the SEER-Medicare cohort, we found that older patients and those with greater comorbidities were most at-risk. The increased risk of falling among older and sicker patients with BrM has serious implications, as such patients have a lower physiologic reserve,⁴ suffer from higher rates of osteopenia and osteoporosis,¹⁶ have impaired tissue regeneration and immunologic function,⁶ and are more likely to be on blood-thinning medications,¹⁷ all factors that increase the chance of significant injury after a fall. Prior studies have shown that falls in older patients commonly result in serious injuries and can have devastating consequences on functional ability.^18,19 Falls are the most common cause of traumatic brain injury and hip fractures among older patients, and although rare, approximately 25% of older patients who suffer a hip fracture will die within 1 year.³ In this study, among both cohorts, we identified high rates of serious injuries, including intracranial injuries, fractures, or dislocations, that were associated with a fall. If afflicted by such sequelae, patients may experience a decline in their ability to perform activities of daily living, require increased assistance with mobility, and find themselves increasingly dependent on caretakers.

In addition to the physical consequences, falling can be a source of significant psychological stress for patients. The experience of one fall causes many patients to live in fear of experiencing a subsequent fall in the future,⁵ and it is common for such patients to therefore limit their activity and socially isolate, which can lead to anxiety, depression, and overall diminished quality of life.^6,7 The consequences of a fall can be especially problematic for patients with BrM, who already may suffer from a high burden of neurologic symptoms. Another important consequence of fall-related injuries, for both patients and health systems, relates to the increased need for hospital-level care after a fall. Among BWH/DFCI patients, we found that 33% and 23% of falls led to emergency department visits and inpatient hospitalizations, respectively. Requiring hospital-level care can be psychologically challenging for patients, and prior work has shown that patients with advanced cancer report lower quality of life when hospitalized.^20,21 In addition, patients with BrM are often receiving systemic treatments or steroids, both of which can dampen the immune system. Exposing such patients to the emergency department and inpatient settings increases their risk for nosocomial infections that can yield added risk for deleterious outcomes.^22,23

In addition to characterizing the incidence and risk factors for falling, we also sought to identify the causes of falls in this patient population. Among patients at BWH/DFCI, the most common etiologies of falling were: mechanical (28%), weakness (16%), ataxia (16%), and syncope (5%). This is consistent with other studies among patients with cancer that have identified weakness, dizziness, and difficulties with balance as the most common precipitants of falls.²⁴ Such neurologic symptomatology is particularly common among patients with BrM, highlighting the increased susceptibility of this population.

Several interventions have been proposed to reduce the risk of falls among patients with cancer. A recent systemic review evaluated 11 studies examining the effect of exercise on fall incidence among patients with cancer.²⁵ Of the 11 studies, only one study directly measured fall rates and found no difference in the incidence of falls among patients randomized to an exercise-based intervention compared to patients receiving usual care; however, across other studies included in the review, exercise programs were associated with improvements in strength and balance. Although these additional studies did not evaluate whether such improvements in strength and balance translated to a lower number of falls, such an effect may be plausible and should be explored further in future studies.²⁵ In addition to exercise and physical therapy, another potential intervention that holds promise in reducing fall rates is pharmacologic optimization; patients with BrM are commonly on several classes of medications for symptom management, including analgesics and benzodiazepines,²⁶ both of which are associated with an increased risk for falling.²⁷ In addition, existing medications, such as anti-hypertensives or diuretics, harbor potential to increase fall risk²⁸ after patients are diagnosed with BrM, especially given the poor oral intake/hydration and potential subsequent hypotension that can ensue in such patients. Steroids are also often utilized in the management of patients with brain metastases, primarily among patients with neurologic symptomatology. While helpful in controlling the effects of intracranial edema and alleviating a variety of symptoms that patients with brain metastases can experience, prolonged use of steroids at higher daily doses can contribute to myopathy with proximal paresis of the legs, adding to fall risk.²⁹ Consequently, steroids should be utilized at the lowest effective dose for the shortest amount of time possible. Medication reconciliation and optimization, including removal of unnecessary medications and careful consideration of doses, seems especially warranted for patients with brain metastases. Physical therapy also offers the potential recuperation of existing neurologic deficits and may help mitigate fall risk. Finally, environmental factors such as the addition of handrails to homes, ensuring patients have appropriate footwear, and utilizing mobility assistance devices such as canes, walkers, and/or wheelchairs, when indicated, may prove helpful in this population.¹⁹

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.³⁰ However, in contrast to other metastatic sites, BrM are commonly treated with local therapies for which diagnostic/billing codes exist, and claims data can therefore be used to reliably identify BrM, a methodology that has previously been validated and shown to have a high sensitivity (>97%) and specificity (99%).^12,13 Second, we acknowledge that multivariable modeling engenders multiple testing and that a P-value threshold to assess significance of .05 carries the risk of a false-positive result. Consequently, we would suggest that our results be considered hypothesis-generating rather than definitive, which also suits the retrospective nature of the work. Third, claims data do not allow us to determine the underlying etiology for a fall or definitively link the effects of oncologic treatment to the risk of falls. Fourth, the estimate of fall incidence was significantly lower in the SEER-Medicare cohort compared to the institutional cohort, and this is likely secondary to the limitations of claims data in comprehensively identifying falls, particularly ones that do not prompt medical evaluation. Hence, the incidence of falls and fall-related injuries stemming from the population-based data presented here are likely significantly underestimated given the reliance on billing claims for identification. However, we attempted to overcome this limitation of the SEER-Medicare data by complementing this dataset with use of our institutional dataset, for which complete information could be manually collected via chart review, and which is likely more reflective of the proportion of patients with BrM who experience a fall after the development of intracranial disease. Finally, although we identified a significant number of falls among patients with BrM, given the retrospective nature of our work, our data should not be interpreted as demonstrating direct causality between BrM diagnosis and the risk of falling. Similarly, we cannot be certain that the fractures and intracranial injuries that we identified via claims data are truly secondary to falls and not some other cause. However, the relatively consistent rate of such injuries identified after falls among the institutional cohort offer support to the rates identified via the SEER-Medicare data. An important aim of future studies would be to characterize rates of falls, fractures, and intracranial injuries in patients with brain metastases vs not.

Conclusions

In this large study of patients with BrM, we used 2 complementary datasets to describe the incidence, risk factors, etiologies, and sequelae of falls among this highly susceptible patient population. The results of our study collectively suggest that falls are a serious problem among patients with BrM. Given the consequences on physical functioning, emotional well-being, and overall quality of life associated with falls, clinicians caring for patients with BrM should consider proactive risk-reduction strategies when evaluating such patients in clinic, including exercise programs, more careful and frequent medication reconciliation, and implementation of household/environmental safety measures.

Funding

No funding was required for this study.

Conflict of interest statement. Dr A.A.A. reports research funding from Varian Medical Systems and NH TherAguix and consulting fees from Novartis and Seagen. The remaining authors declare no conflicts of interest.