FULL-LENGTH ORIGINAL RESEARCH

Risk of venous thromboembolism in people with epilepsy *Gabriel U. Martz, †Dulaney A. Wilson, †Angela M. Malek, and †Anbesaw W. Selassie Epilepsia, 55(11):1800–1807, 2014 doi: 10.1111/epi.12796

SUMMARY

Gabriel Martz is an epileptologist interested in improving quality of care and safety for patients.

Objective: Risk of venous thromboembolism (VTE) among people with epilepsy (PWEs) has not been previously reported. Standard VTE prevention methods may increase the risk of complications in this population. This statewide study assessed the risk of VTE in PWEs. Methods: Main risk categories were grouped into definite epilepsy (DE), probable epilepsy (PE), and migraine, a comparable neurologic condition. All inpatient, emergency department, and hospital-based outpatient encounters in South Carolina from January 1, 2000 through December 31, 2011, were evaluated for the primary outcome variable of VTE, defined as having a diagnosis of VTE at or after the diagnosis of epilepsy or migraine. Coagulopathies and common comorbidities of epilepsy were enumerated. Differences in VTE proportions were assessed using 95% confidence intervals (CIs). Association of VTE with epilepsy and migraine was evaluated with Cox proportional hazard modeling. Results: A total of 138,497 people with migraine (PWMs) and 67,900 PWEs (32,186 DE, 35,714 PE) were included. VTE occurred in 2.7% of PWEs (4.2% among DE), and 0.6% of PWMs. The hazard ratio for VTE in DE compared with PWMs was 3.08 (95% CI 2.76
3.42), adjusted for all covariables. Higher numbers of comorbidities were strongly associated with VTE. PWE had higher numbers of comorbidities (52% with 2+ comorbidities versus 23% of PWM), but the impact of comorbidities on VTE risk was larger in PWM. Significance: Higher VTE risk in PWE than PWM suggests risk factors associated with epilepsy, independent of chronic neurologic illness. VTE occurrence in PWE is comparable to published rates among people with cancer. KEY WORDS: Epilepsy, Epidemiology, Migraine, Venous thromboembolism, Comorbidity.

The spectrum of comorbidities associated with epilepsy has recently received significant attention.1,2 In addition to neurobehavioral issues such as depression and cognitive dysfunction,3 there is a high burden of somatic illnesses among people with epilepsy (PWEs)4; recent evidence suggests that PWEs have a higher prevalence of strokes and cardiovascular disease.2,4,5 However, venous thromboembolism (VTE) has not been previously evaluated in this Accepted August 8, 2014; Early View publication September 30, 2014. Departments of *Neurosciences and †Public Health Sciences, Medical University of South Carolina, Charleston, South Carolina, U.S.A. Address correspondence to Gabriel U. Martz, Department of Neurosciences, Medical University of South Carolina, 96 Jonathan Lucas St, MSC 616/ CSB 301, Charleston, SC 29425-6160, U.S.A. E-mail: gabriel.martz@norton healthcare.org Wiley Periodicals, Inc. © 2014 International League Against Epilepsy

population despite high rates of known risk factors such as obesity, reduced mobility, and frequent hospitalizations.6,7 Risk of VTE may be further magnified in the subgroup of patients with epilepsy who undergo elective admission for seizure monitoring. There are no established guidelines for prevention of VTE in this unique inpatient population.8 Commonly used guidelines for deep vein thrombosis (DVT) prophylaxis7 may not apply to this patient group because the majority of patients are presumed to be in the lowest risk categories. Furthermore, typical VTE preventive measures are often withheld due to potential for seizure-related injury. Many centers refrain from anticoagulation to avoid increasing the risk of hemorrhage during the acute seizures that are encouraged during the admission. Similarly, sequential compression devices may increase the risk of injury due to

1800

1801 VTE Risk in PWE confinement during convulsions, or a fall from the bed. Ambulation is often carefully restricted to prevent seizurerelated falls. These factors, combined with an average length of stay of >4 days,9 which is more than the >3 days of reduced mobility reported to amplify VTE risk,10 may result in an underrecognized high level of VTE risk for patients in the epilepsy monitoring unit (EMU). The current study was designed to determine whether PWEs have elevated VTE risk. Beyond the issues of chronic illness and comorbidity, PWEs may be at higher risk due to immobility, inflammatory processes,11 and autonomic dysregulation.12 VTE risk is highly relevant given its association with increased mortality in many conditions.13 The prevalence of VTE among PWEs was compared to that of people with migraine (PWMs) using a well-established, statewide database. This is a strong comparison group for isolating epilepsy-specific risk factors, because migraine has features in common with epilepsy, including chronicity with episodic flare-ups, similar medications, and a range of comorbid conditions.14 It was hypothesized that PWEs were more likely than PWMs to have VTE, after adjusting for measurable risk factors. To our knowledge, this is the first report of VTE risk in PWEs, and is an important first step toward ensuring best practice during both inpatient and outpatient care.

Methods Data sources This study represents the civilian population of South Carolina (SC) utilizing mandated hospital discharge and emergency department (ED) visit datasets, which include hospitalbased outpatient department (OPD) visits reported via the uniform billing system (UB-04).15 Variables used for analysis are shown . The Medical University of South Carolina Institutional Review Board exempted this study. Study design, setting, and population This study was designed as a retrospective cohort study. SC residents hospitalized or treated and released from the ED and OPD from January 1, 2000 through December 31, 2011 were eligible. Case and control cohorts were ascertained using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes for epilepsy (345.x), seizure Not Otherwise Specified (780.39) and migraine (346.x).15 Discharge diagnosis narratives are coded by certified health information specialists who are operating under the standard coding guideline of National Center for Health Statistics and the Center for Medicare and Medicaid Services. The accuracy of the diagnoses, which refers to the extent to which the ICD-9-CM codes in the billing dataset accurately reflect the attending physicians’ discharge diagnosis narratives, was 85.2%.16

Definitions Diagnoses of epilepsy and migraine were established by the treating clinician at the time of the inpatient, ED, or OPD visit and utilized as the main exposure of interest. Case ascertainment was based on an algorithm developed previously16; briefly, cases of definite epilepsy (DE) required (1) two or more encounters with diagnosis codes of 345.x or (2) one encounter with diagnosis of 345.x preceded by an encounter with diagnosis of ‘seizure not otherwise specified’ (780.39). Cases of probable epilepsy (PE) had only a single encounter during the study period with a diagnosis code of epilepsy (345.xx). Individuals with diagnostic codes for both epilepsy and migraine were excluded. The outcome of interest was the presence or absence of VTE ascertained with diagnosis codes of 415.1, 451.1x, 451.2, 451.81, 451.9, 453.1, 453.2, 453.4, 453.8, 453.9, and V12.51, at the time of, or subsequent to, an encounter with diagnosis of epilepsy or migraine. Race was categorized into white, black, and all others. Place of residence was defined as rural or urban, according to the county codes of rural-urban designation by the Office of Management and Budget. Mortality status from any cause was determined for all subjects, with the SC Multiple Causes of Death Data file through December 31, 2011. Coagulopathy disorders were identified via secondary diagnosis codes of 289.81, 289.82, 289.9, 790.92, and 790.95 present in any encounter during the study period. Months of follow-up after index diagnosis and number of encounters were recorded. The co-occurrence of 35 common comorbidities of epilepsy noted in the literature2,17,18 were identified by their diagnosis codes (Table S1). Presence of any of these 35 conditions over the course of the study period was noted, and subjects were grouped by number of comorbidities registered: 0–1, 2–4, 5–6, and ≥7. The remaining variables were classified as shown in Table 1. It was hypothesized that the portion of the population with increased age and comorbidities may be overrepresented in the VTE group, although they may be less commonly admitted to EMUs. To better characterize this potentially lower risk population, the analysis was re-performed on a subgroup defined by age 12–64 years with insurance coverage (any type), fewer than two comorbid conditions, and without coagulopathy. As earlier, proportions with VTE among PWEs and PWMs were determined, as well as the hazard ratio. Statistical analysis Statistical analysis was performed using SAS software statistical package, V9.3.1 (SAS Institute Inc., Cary, NC, U.S.A.). Proportions corresponding to demographic and clinical characteristics were compared among the three groups by constructing 95% confidence intervals (CIs) under the assumption of independence and normal approximation (Table 1). Overlapping CIs suggest no significant Epilepsia, 55(11):1800–1807, 2014 doi: 10.1111/epi.12796

1802 G. U. Martz et al. Table 1. Demographic and clinical characteristics of case and control groups

Race

Sex Age group

Payer status

MSA

Mortality Number of comorbidities

Coagulopathy Follow-up (months) Number of visits per annum

Probable epilepsy (n = 35,714)

Migraine (n = 138,497)

Total (N = 206,397)

n

%

95% CI

n

%

95% CI

n

%

95% CI

Yes No White Black Other Male Female 65+ 35 to 64 19 to 34 13 to 18 6 to 12 0 to 5 Uninsured Indigent care Medicaid Medicare Commercial Rural Urban Out of state Yes No 7+ 5 to 6 2 to 4 0 to 1 Yes No Median (IQR)

2,647 203,750 139,500 58,172 8,725 61,744 144,653 24,081 103,075 59,909 10,792 5,341 3,199 33,504 10,719 33,787 40,289 88,098 74,262 121,861 10,267 17,587 188,810 2,470 8,071 56,059 139,797 1,051 205,346 206,397

1,340 30,846 18,788 12,415 983 15,559 16,627 7,162 15,263 6,054 1,247 1,339 1,121 4,062 1,276 7,305 11,968 7,575 12,073 19,532 581 5,597 26,589 1,872 4,706 15,308 10,300 463 31,723 32,186

4.2 95.8 58.4 38.6 3.1 48.3 51.7 22.3 47.4 18.8 3.9 4.2 3.5 12.6 4.0 22.7 37.2 23.5 37.5 60.7 1.8 17.4 82.6 5.8 14.6 47.6 32.0 1.4 98.6 13

3.9
4.4 95.6
96.1 57.8
58.9 38.0
39.1 2.9
3.2 47.8
48.9 51.1
52.2 21.8
22.7 46.9
48.0 18.4
19.2 3.7
4.1 3.9
4.4 3.3
3.7 12.3
13.0 3.8
4.2 22.2
23.2 36.7
37.7 23.1
24.0 37.0
38.0 60.1
61.2 1.7
2.0 17.0
17.8 82.2
83.0 5.6
6.1 14.2
15.0 47.0
48.1 31.5
32.5 1.3
1.6 98.4
98.7 4
29

454 35,260 23,147 11,062 1,505 17,367 18,347 8,016 15,867 6,613 1,739 1,618 1,861 4,925 1,773 6,349 11,744 10,923 12,211 20,825 2,676 6,638 29,076 36 789 12,667 22,222 134 35,580 35,714

1.3 98.7 64.8 31.0 4.2 48.6 51.4 22.4 44.4 18.5 4.9 4.5 5.2 13.8 5.0 17.8 32.9 30.6 34.2 58.3 7.5 18.6 81.4 0.1 2.2 35.5 62.2 0.4 99.6 24

1.2
1.4 98.6
98.8 64.3
65.3 30.5
31.5 4.0
4.4 48.1
49.1 50.9
51.9 22.0
22.9 43.9
44.9 18.1
18.9 4.6
5.1 4.3
4.8 5.0
5.4 13.4
14.2 4.7
5.2 17.4
18.2 32.4
33.4 30.1
31.1 33.7
34.7 57.8
58.8 7.2
7.8 18.2
19.0 81.0
81.8 0.1
0.1 2.1
2.4 35.0
36.0 61.7
62.7 0.3
0.4 99.6
99.7 8
45

853 137,644 97,565 34,695 6,237 28,818 109,679 8,903 71,945 47,242 7,806 2,384 217 24,517 7,670 20,133 16,577 69,600 49,978 81,504 7,010 5,352 133,145 562 2,576 28,084 107,275 454 138,043 138,497

0.6 99.4 70.4 25.1 4.5 20.8 79.2 6.4 51.9 34.1 5.6 1.7 0.2 17.7 5.5 14.5 12.0 50.3 36.1 58.8 5.1 3.9 96.1 0.4 1.9 20.3 77.5 0.3 99.7 47

0.6
0.7 99.3
99.4 70.2
70.7 24.8
25.3 4.4
4.6 20.6
21.0 79.0
79.4 6.3
6.6 51.7
52.2 33.9
34.4 5.5
5.8 1.7
1.8 0.1
0.2 17.5
17.9 5.4
5.7 14.4
14.7 11.8
12.1 50.0
50.5 35.8
36.3 58.6
59.1 4.9
5.2 3.8
4.0 96.0
96.2 0.4
0.4 1.8
1.9 20.1
20.5 77.2
77.7 0.3
0.4 99.6
99.7 18–85

Median (IQR)

206,397

32,186

3

35,714

1

138,497

1

Characteristics VTE

Definite epilepsy (n = 32,186)

2
5

1
1

1–2

VTE, venous thromboembolism; MSA, metropolitan statistical area; IQR, interquartile range (25th and 75th percentile).

difference in proportions. Similarly, p-values were provided to demonstrate statistical significance across risk groups. Univariate (crude) associations between dependent and independent variables were evaluated. Potential confounders included in the multivariable analyses were variables that were associated with the main exposure and outcome variables. Based on these evaluations, demographic characteristics, insurance type, the number of comorbid conditions, and coagulopathy disorders met the database definition of confounders, and there was no evidence of interactions. The relationship between the dependent variable (VTE) and independent variables was examined as a function of time using Cox proportional hazards regression averaged over the entire time of follow-up. Survival time (T) was defined by the number of months from date of the initial diagnosis (epilepsy or migraine) to the date of the earliest subsequent clinical encounter including VTE in the diagnoses listing. The model assumes the hazard ratio (HR) Epilepsia, 55(11):1800–1807, 2014 doi: 10.1111/epi.12796

comparing any two specifications of predictors is constant over time. This assumption was tested with the goodnessof-fit chi-square test, which compares the observed and expected VTE probabilities, and by graphical means using the log-log Kaplan-Meier (KM) curves.19 Variables with bivariate association p-values ≤0.20 were included in the multivariable model. Multicollinearity among covariables was evaluated by assessing deviations of regression coefficients and their standard errors in the fitted univariate and multivariable models.20 Covariables were entered simultaneously into the model. A log-rank test validated the homogeneity of VTE probability curves by type of disease strata.21 p-values