A recent report from the Centers for Disease Control and Prevention (CDC) estimates 2.8 million traumatic brain injury (TBI) deaths, hospitalizations and emergency department visits occurred in the United States in 2013. That number includes 56,000 deaths, 282,000 hospitalizations, and 2.5 million emergency department (ED) visits related to TBI. The latter increased more than 50% between 2007 and 2013 (Taylor et al., 2017).

Among the potential consequences of TBI are seizures and epilepsy. The latency of seizure occurrence after TBI is commonly categorized according to three intervals. Immediate seizures occur less than 24 h after TBI. Early seizures occur less than one week after TBI, attributed to acute, but not necessarily irreversible, pathophysiologic changes in cerebral function (Englander et al., 2003). Late seizures occur more than a week after TBI, and if recurrent, constitute the diagnosis of post traumatic epilepsy (PTE) (Lowenstein, 2009). TBI can result in several potentially epileptogenic alterations, including neuronal, axonal, and vascular damage, as well as parenchymal and subarachnoid hemorrhage. TBI initiates cascades of molecular and cellular changes including excitotoxicity, gliosis, and neuroinflammation, as well as later toxicity caused by iron-rich hemoglobin breakdown products. (Glushakov et al., 2016; McNamara et al., 2006; Pitkänen et al., 2016, 2014)

The overall incidence of PTE in hospitalized populations, comprising a range of TBI severity mainly from closed head injuries, is about 3–5% (Chen et al., 2009; Annegers et al., 1998). In a study identifying 5984 episodes of TBI in Olmsted County, Minnesota from 1935 to 1984, the probability of PTE ranged from 0.7% to 10.0% in five years follow-up and 2.1% to 16.7% in 30 years follow-up, correlating with the severity of TBI (Annegers et al., 1998). Penetrating head injuries as seen in military veterans show the highest incidence of PTE, with estimates ranging from a 5-year cumulative incidence of 28% to a 15-year cumulative incidence of 53% (Salazar et al., 1985; Salazar and Grafman, 2014; Raymont et al., 2010; Caveness et al., 1979). There has been no study of PTE representing the entire population of the US.

Identified risk factors of PTE include chronic alcoholism, age of 65 years or older, penetrating injuries, traumatic intracranial hemorrhage, severity of injury, posttraumatic amnesia or loss of consciousness for more than one day, trauma-related focal neurologic deficits, depressed skull fractures, cerebral contusions, and retained bone and metal fragments (Englander et al., 2003; Annegers et al., 1998).

An important goal in the acute and long-term management of TBI is the prevention of PTE. Several randomized clinical trials have shown the effectiveness of anti-epileptic drugs (AEDs) in the management of early seizures after TBI (Kirmani et al., 2016), including prophylactic anti-epileptics (Schierhout and Roberts, 2012; Temkin et al., 1990). Longer, limited-term prophylactic use of three older AEDs: phenytoin (PHT), carbamazepine (CBZ), and valproate (VPA) have been evaluated in clinical trials, but have shown no benefit in the prevention of late seizures or epilepsy (Kirmani et al., 2016; Temkin et al., 1999).

This study addresses the incidence of PTE and its risk factors—including TBI severity and early seizures—among persons with medically attended TBI who are enrolled in a large database broadly representing the U.S. insured population. We also explore the possible effectiveness of prophylactic AED use after TBI to prevent the development of PTE, as this has not been evaluated for many AEDs.

From 2006 to 2012, we identified 2,345,758 total TBI patients (1,851,368 patients from commercial and Medicare databases and 494,390 patients from Medicaid databases, Fig. 1) among whom there were 77,014 (3.3%) pre-existing epilepsy patients and 263,149 (11.2%) patients who were taking AEDs prior to TBI. We excluded those patients with preexisting epilepsy or prior AED use from the TBI cases that we analyzed. The study population included 2,059,870 TBI patients and 2,198,173 non-TBI matched controls (Fig. 1).

Overall, the 9-year cumulative incidence of epilepsy among TBI patients in our study was more than two-fold that among non-TBI controls. This is consistent with other population-based studies demonstrating an increased risk of PTE across the spectrum of TBI severity, although the effect is lower in magnitude than in those studies. (Annegers et al., 1998; Ferguson et al., 2010) This is because our study population includes milder TBIs seen in emergency departments or outpatient clinics, whereas some other studies only included more serious injuries admitted to hospital. We may also note that the one-year incidence of epilepsy (0.19%) in our control population was also substantially higher than annual incidence rates commonly cited for general populations. This may be explainable in part by the age distribution of our controls, given that these are matched to a TBI population. TBI populations, including ours, over-represents children under 15 (and especially children under 5) years of age, where incidence rates of TBI are much higher, as are incidence rates of epilepsy.

The incidence of early seizures after TBI was 0.5% in our study, lower than the incidence reported in other studies, which have established early seizures as a known risk factor for PTE. (Szaflarski et al., 2014) The reason is that our study population is a somewhat different from some other studies. Estimates from other studies have ranged from 2.1% (in a population-based Olmsted County, Minnesota study) to 16.3% (in a TBI rehabilitation center) (Englander et al., 2003; Annegers et al., 1998, 1980; Asikainen et al., 1999). The differences may be attributed to variations in definitions, study designs, and study populations, some of which (e.g., TBI rehabilitation centers) have more severely injured subjects. We included all visits to EDs, which means that more mild cases were allowed. Our study is the first to report early seizure occurrence in a broad spectrum of privately insured, Medicaid and Medicare TBI patients in the U.S. population.

We found the cumulative incidence of epilepsy over nine years approached 50% among TBI patients with early seizures, and the risk of epilepsy among TBI patients with early seizures was 37 times higher than that for those without early seizures. The apparent risk of epilepsy among patients with severe TBI was nearly 14 times higher than that for those with mild TBI. This appears consistent with other population-based studies of U.S. localities, although the actual incidence of PTE in relation to these factors was quite different (Annegers et al., 1998; Ferguson et al., 2010). The study by Annegers (Annegers et al., 1998) found a much lower incidence of PTE in the early seizure group. It is not clear how to explain that difference, but the definition of PTE may have been different. The Ferguson study (Ferguson et al., 2010) only looked at inpatients and found similar tendencies, but at a higher incidence, which would be expected, since those subjects may have been more severe TBI patients. Both studies had a fairly small number of PTE patients.

Recent guidelines for the management of severe TBI from the Brain Trauma Foundation and the American Association of Neurological Surgeons suggest patients with severe TBI may be placed on an AED soon after trauma to prevent early seizures (Carney et al., 2016). Our findings indicate that in practice, 47% of patients with early seizures and 80% of patients with severe TBI had no recorded AED use prior to the diagnosis of PTE. However, it is possible that some received AEDs only while hospitalized, which is not recorded in the database. An older survey of 127 neurosurgery clinics on antiepileptic prophylaxis in patients with TBI indicates a variety of attitudes towards prophylaxis for seizures: in 12% of the responding institutions, antiepileptic prophylaxis is given to every patient, in 36%, no prophylaxis is carried out, and in 52% some patients receive prophylaxis while others do not (Dauch et al., 1996). Penetrating injuries, intracranial hemorrhages and electroencephalographic abnormalities were the most frequent reasons for the prophylaxis.

The prophylactic use of any standard AED we examined appeared to confer no reduction in risk of PTE. Indeed that risk appeared elevated for nearly all drugs, a result we suspect may be attributable to un-detected cases of preexisting epilepsy in our study population, as well as confounding risk factors such as type of injury and nuances of severity that we were unable to take account of. Our findings are consistent with other studies that have examined the potential efficacy of specific AEDs—phenytoin, carbamazepine, valproate, and levetiracetam—finding no benefit, as described in a recent systematic review (Thompson et al., 2015).

One possible exception of interest was acetazolamide, which we found was associated with a decreased risk of PTE. This suggestive finding, of course, needs replication through more rigorous studies. Most likely it was not used as an anti-epileptic drug, but as a treatment of brain edema following TBI (Sturdivant et al., 2016). If verified, possible explanations may involve acetazolamide’s unique action among AEDs as a selective inhibitor of carbonic anhydrase. Acetazolamide increases local cerebral blood flow (Bickler et al., 1988), an effect confirmed in an animal model of brain injury, which finds a normalization of the ratio between local cerebral metabolic rate for glucose and local cerebral blood flow and a consequent reduction in axonal injury (Harris et al., 2012). Apart from these effects, in animal models of epilepsy, acetazolamide also generally decreases the interval of occurrence of ictal and interictal discharges (Hamidi and Avoli, 2015).

Our definitions of TBI and epilepsy, using coded administrative data, are consistent with recent international recommendations. The large number of TBI cases and PTE cases is statistically robust, providing credible estimates of the incidence of early seizures, PTE, and the association between AEDs and PTE. The sample draws subjects across the spectrum of epilepsy from multiple sectors throughout the insured population of the United States, thereby being more broadly representative than either local community-based studies or clinic-based studies. And our 11 years of data can provide a longitudinal view of the development of PTE following TBI and prophylaxis.

There are, however, several limitations to our analysis. The dataset we used in this study is claim-based and susceptible to miscoding and missing information. ICD-9-CM-based coding can result in misclassifications of the occurrence of both TBI and epilepsy. The accuracy of ICD-9-CM coding for our indicators of TBI severity have not been validated. AED use may be misclassified in some patients. As noted above, AED received during hospitalization are not recorded in the database. Also, claimed-based data represent prescriptions filled only; it cannot be determine whether patients adhered to medication used the AEDs as prescribed.

Misclassification of some preexisting epilepsy cases as new PTE cases may occur also. Because the minimum 2-year baseline is not always long enough to detect prevalent epilepsy cases (International League Against Epilepsy et al., 1993; Thurman et al., 2011), we may have included pre-existing epilepsy patients who have infrequent or mild seizures for which they rarely obtain medical care or do not take medication, or who receive care in part from other sources not covered by insurance plans included in our data set.

This study uses linked inpatients and outpatient data that describe all medical encounters for all individuals in a population. When individuals have multiple medical encounters described by epilepsy or seizure codes, the likelihood of identifying true cases of epilepsy is high; however, the codes per se do not distinguish new-onset from long-established cases (Thurman et al., 2011).

The claims data that we analyzed do not represent persons lacking health insurance (a status more frequent among younger adults) and may not fully represent the insured population of the United States. The commercial and Medicare supplemental datasets are contributed by over 150 large employers and about 20 health plans. The population submitting Medicare Supplemental claims may differ from the Medicare population without supplemental private insurance, which was not included in our study.

All these factors may substantially reduce confidence in any extrapolation to findings for the U.S. population as a whole.

Finally, information identifying race was not provided in Medicare and Commercial Claims data, although incomplete information was contained in Medicaid data. We combined Medicaid, Medicare and Commercial claims data together, and thus could not use race as a variable.

We identified advancing age, early seizures, and increased severity TBI as risk factors of PTE. In caring for and following patients with these factors, counseling and increased vigilance to provide early treatment of emerging PTE may be indicated.

Further studies are needed to test the epilepsy prevention efficacy of acetazolamide prophylaxis, as are studies of prognostic indicators, including biomarkers, to identify which patients will develop epilepsy and which might benefit from preventive treatment. Finally, more research needs to be done on the effectiveness of AEDs in the first seven days or more after TBI.