Does Glucocorticoid Administration Prevent Late Seizures after Head Injury?


Address correspondence and reprint requests to Dr. N.F. Watson at University of Washington Sleep Disorders Center, Box 359803, 325 Ninth Avenue, Seattle, WA 98104-2499, U.S.A. E-mail:


Summary: Purpose: Preventing posttraumatic epilepsy has been a difficult challenge. In this study we evaluated the association between glucocorticoid administration after traumatic brain injury (TBI) and posttraumatic seizures.

Methods: We examined a seizure-prevention trial database of 404 patients with severe TBI for exposure to glucocorticoids in the early (<1 week) posttraumatic period. After controlling for seizure risk, we compared the odds of developing first and second late posttraumatic seizures between those that received glucocorticoids and those that did not.

Results: Patients dosed with glucocorticoids within 1 day of their TBI were more likely to develop first late seizures than were those without [p = 0.04; hazard ratio = 1.74; 95% confidence interval (CI), 1.01–2.98]; whereas those receiving glucocorticoids ≥2 days after their injury had no similar association (p = 0.66; hazard ratio = 0.77; 95% CI, 0.23–2.56; p = 0.10 among the three groups). Receiving glucocorticoids within 1 day, or ≥2 days after TBI was not associated with second late seizure development.

Conclusions: Glucocorticoid treatment after TBI is not associated with decreased late posttraumatic seizures, and early treatment is associated with increased seizure activity.

In the United States, traumatic brain injury (TBI) affects nearly 300,000 people per year (1). TBI accounts for ∼4% of epilepsy and is a common cause of refractory epilepsy (2). Antiepileptic medications (AEDs), effective at preventing early (<1 week) posttraumatic seizures (PTSs), do not prevent late (>1 week) seizure development (3,4). Currently no effective therapies are available for preventing late PTSs.

Glucocorticoids were used in early posttraumatic periods in attempts to control life-threatening cerebral edema. These medications can mitigate posttraumatic ischemia (5), improve outcomes after traumatic spinal cord injuries, reduce posttraumatic epileptiform activity (6) in animal models, and treat infantile spasms (7). These factors make them attractive candidates to prevent late PTSs. This study sought to understand the impact of glucocorticoid medications on late PTS development.


Study design

This prospective cohort study used a dataset previously collected in a posttraumatic AED (phenytoin, PHT) seizure-prevention trial (3). Participants with a head injury admitted to Harborview Medical Center, a level one trauma center serving the Seattle metropolitan area and northwest Washington, were included if at least one of the following criteria was met: a cortical contusion visible on computed tomography scanning; a subdural, epidural, or intracerebral hematoma; a depressed skull fracture; a penetrating head wound; a seizure within 24 h of injury; or a score of ≤10 on the Glasgow Coma Scale (GCS) on admission.

Four-hundred four patients met inclusion criteria. We obtained consent or a waiver of consent for inclusion into the initial study, as approved by the institutional review board (IRB) of the University of Washington. The same IRB approved the current study. Patients were followed up for 2 years after injury for seizure development. Seizures were defined according to order of occurrence as first or second late seizures. Further details regarding the initial study have previously been published (3).

Clinical glucocorticoid use was not restricted by the initial study protocol. The study nurse recorded all medications administered to patients during the first week of hospitalization. We queried the database for patient exposure to glucocorticoid medications within a week of their TBI and considered any type, dose, or route of administration significant. The success of glucocorticoid treatment for CNS injury is time sensitive; therefore we performed a time-to-first-dose analysis of first and second late seizure development by dividing patients into those that received glucocorticoids acutely (0–1 days) and subacutely (2–7 days).

We analyzed the relation between total duration of glucocorticoid exposure and late seizure development. This analysis assumed that steroid exposure was a linear variable, with each 12 h of exposure having the same additional effect. Cumulative mortality as a function of steroid exposure also was examined.

Statistical analysis

Kaplan–Meier estimates for the percentage developing late seizures over time are presented and compared by using log rank tests. Hazard ratios (HRs) were used to describe the risk of developing first and second late PTSs. We controlled for seizure risk and injury severity with a multivariate Cox model. Control variables were selected in a stepwise fashion by using a proportional hazards model that considered only main effects. The variables considered were age, gender, race, number of nonreactive pupils on admission, treatment group (PHT vs. placebo), cortical contusions, depressed skull fracture, dural penetration, penetrating head wound, intracerebral hematoma, subdural hematoma, epidural hematoma, seizures within 24 h of injury, and GCS on admission. Five variables associated with the development of late PTSs were identified and controlled: nonreactive pupils on admission, cortical contusions, intracerebral hematoma, subdural hematoma, and early PTS.


Comparability of treatment groups

Study group characteristics are shown in Table 1. Demographic variables and cause of injury were similar between the glucocorticoid-treated and untreated groups. Injury characteristics were similar, with only dural penetration by surgery and nonreactive pupil(s) occurring more frequently in the glucocorticoid-treated subjects. Other than PHT, no other seizure-affecting medications were administered acutely to significant numbers of study participants.

Table 1. Study group characteristics
CharacteristicTiming of glucocorticoid exposure (days after traumatic brain injury)No glucocorticoid exposure (n = 279)
0–1 (n = 105)2–7 (n = 20)Total (n = 125)
  1. For marital status, unknown was considered “single.” All other unknown variables were considered “no” responses.

  2. ap < 0.05.

Demographic variables
 Age (SD)34 (17)29 (13)33 (17)35 (18)
 Sex (% male)80808075
 Marital status(% married)57655856
Cause of injury (number, %)
 Automobile35 (33)  6 (30)41 (33)91 (33)
 Motorcycle19 (18)  5 (25)24 (19)50 (18)
 Fall16 (15)  3 (15)19 (15)42 (15)
 Assault, fight, suicide attempt17 (16)  4 (20)21 (17)29 (10)
 Pedestrian7 (7)1 (5)8 (6)30 (11)
 Other/unknown11 (10)1 (5)12 (10)37 (13)
Injury characteristics (number, %)
 Dura penetration
     By surgerya97 (92)19 (95)116 (93) 237 (85) 
     By injury35 (33)  7 (35)42 (34)68 (24)
 Insertion of Camino intracranial pressure monitor94 (90)19 (95)113 (90) 229 (82) 
       Cortical contusions47 (45)14 (70)61 (49)137 (49) 
       Glasgow Coma Scale ≤1060 (57)11 (55)71 (57)160 (57) 
       Subdural hematoma38 (36)10 (50)48 (38)111 (40) 
       Nonreactive pupil(s)a39 (37)  9 (45)48 (38)78 (28)
Surgical treatment for
 Subdural hematoma23 (22)  7 (35)30 (24)67 (24)
 Epidural hematoma14 (13)  4 (20)18 (14)33 (12)
Depressed skull fracture12 (11)1 (5)13 (10)24 (9) 
 Intracerebral hematoma8 (8)  2 (10)10 (8) 21 (8) 
Seizures within 24 h of the injury14 (13)1 (5)15 (12)39 (14)
Subarachnoid hemorrhage11 (10)  2 (10)13 (10)19 (7) 
Penetrating injury10 (10)1 (5)11 (9) 18 (6) 
Removal of a bone flap6 (6)  3 (15)9 (7)15 (5) 
Shunt placement4 (4)  2 (10)6 (5)9 (3)

Glucocorticoid dose and timing

One hundred twenty-five of 404 patients in the seizure-prevention-trial database received glucocorticoids within 1 week of their TBI. One hundred five first received glucocorticoids within 1 day of their TBI (73, the day of injury; 32, the day after injury), and 20 first received glucocorticoids ≥2 days after their TBI. Of those given glucocorticoids, 122 (98%) received dexamethasone, three (2.4%) received prednisone, two (1.6%) received methylprednisolone, and two (1.6%) received hydrocortisone. Only three of the seven subjects that received prednisone, methylprednisolone, or hydrocortisone did not receive any concomitant doses of dexamethasone. Regarding total dexamethasone dosage, 82% of subjects received ≥20 mg, 70% received ≥40 mg, 57% received ≥80 mg, and 30% received ≥160 mg.

First late seizures

Figure 1 shows estimates of the proportion of patients with first late PTSs according to glucocorticoid exposure status. Patients dosed with glucocorticoids within 1 day of their TBI were more likely to develop first late seizures than were those without (p = 0.04). This early glucocorticoid exposure was associated with an increase of 74% in the risk of a first late seizure (95% CI, 1.01–2.98). Receiving glucocorticoids ≥2 days after a TBI was not associated with first late seizure development (p = 0.66; HR = 0.77; 95% CI, 0.23–2.56). Overall, 28 of 125 glucocorticoid-exposed patients experienced a first late seizure during observation for the study compared with 34 of 279 drug-free patients (p = 0.10, considering both time-to-first-dose groups controlling for the other identified risk factors).

Figure 1.

Proportion of patients with first late seizures as a function of glucocorticoid use and days from traumatic brain injury. A significant difference is observed among the three groups (Log rank, p = 0.02). Patients with early (0 to 1 day) glucocorticoid exposure (n = 105) were more likely to have first late seizures (24 seized, 34% estimated seizure rate by 2 years) than were the late (2–7 days, n = 20, four seized, 26%) or no-glucocorticoid group (n = 279, 34 seized, 20%). After controlling for seizure risk with the multivariate Cox model, patients dosed with glucocorticoids early continued to have a higher rate of first late seizure activity (p = 0.04).

Second late seizures

Figure 2 shows estimates of the proportion of patients with second late PTSs according to glucocorticoid-exposure status. Receiving glucocorticoids within 1 day (p = 0.28; HR = 1.41; 95% CI, 0.75–2.63), or ≥2 days (p = 0.54; HR = 0.63; 95% CI, 0.15–2.74) after TBI was not associated with second late PTS development. Overall, 20 of 125 glucocorticoid-exposed individuals had a second late seizure during the period of observation compared with 28 of 279 drug-free patients (p = 0.41, considering both time-to-first-dose groups, controlling for the other identified risk factors).

Figure 2.

Proportion of patients with second late seizures as a function of glucocorticoid use and days from traumatic brain injury. No difference is found among all three groups regarding second late seizure development (Log rank, p = 0.21). Controlling for seizure risk with the multivariate Cox model further supports this conclusion (p = 0.28 for the 0- to 1-day group; p = 0.54 for the 2- to 7-day group).

Seizures and length of steroid exposure

We used previously established predictors from our multivariate Cox model to assess time to first late seizure development as a function of glucocorticoid-exposure duration. No cumulative effect was found of glucocorticoid exposure on late seizure development (p = 0.63; HR = 1.16; 95% CI, 0.63–2.16).


No difference was noted in the cumulative mortality between the glucocorticoid-treated and -untreated groups (Kaplan–Meier survival analysis, Log rank, p = 0.57). Sixty-one of 279 patients that did not receive glucocorticoids and 31 of 125 that did, died during the follow-up period: 15% in each group by day 8; 19% in each group by 1 month; 22% and 23%, respectively, by year 1; and 22 and 25%, respectively, by the end of year 2.


In this prospective cohort study, glucocorticoids were not associated with decreased late PTSs. When administered within 1 day of a TBI, glucocorticoids were significantly associated with an increased likelihood of developing a first late seizure. Duration of glucocorticoid exposure did not independently affect late seizure development. We hypothesized that glucocorticoids would prevent late PTSs. Our results do not support this for many potential reasons.

Animal models show that glucocorticoids can worsen excitotoxicity and oxidative stress. Oxidative cascades are implicated in both TBI and animal-based models of epileptogenesis (8–10). Glucocorticoid treatment during kindling epileptogenesis results in more severe tonic–clonic seizures, higher-amplitude CA1 spikes in the hippocampus, and increased amplitude of voltage-gated calcium currents compared with placebo (11). Glucocorticoids have been shown to induce reversible dendritic atrophies in the hippocampal CA3 region and to decrease proliferation and neurogenesis in the dentate gyrus (12). These findings may explain not only why glucocorticoids failed to prevent late PTSs, but also why we observed increased seizure activity after early posttraumatic glucocorticoid treatment.

When our initial seizure-prevention trial was under way, patients with TBI were often treated with glucocorticoids in the hope of reducing cerebral edema. A body of research since the end of that trial does not support this practice. Glucocorticoid medications attenuate vasogenic but not cytotoxic edema, and thus do not effectively reduce intracranial pressure in the critical early stages of acute TBI (13). TBI outcome studies show that glucocorticoids do not benefit patients with moderate and severe head injury (14). Thus glucocorticoid use after TBI is not the current standard of care.

This prospective cohort study was limited by the nonrandomized administration of glucocorticoids. Indeed, these medications may have been given preferentially to high-risk patients. Although we controlled for seizure risk and injury severity with our backward stepwise selection model, this adjustment could have missed other, undocumented variables associated with high-risk patients. However, the lack of difference in mortality and similar injury characteristics points to comparable injury severity for the two groups. For these reasons, we believe that seizure risk was adequately controlled.

Although most patients in this study received significant amounts of dexamethasone, glucocorticoid type, dose, and exact times of administration were not standardized. Furthermore, glucocorticoid exposure was documented for only the first week after TBI. This could have attenuated treatment effects, particularly those seen at higher dexamethasone doses administered over longer periods. These issues could be resolved with a randomized, double-blind, placebo-controlled trial. Unfortunately, the lack of clinical equipoise regarding posttraumatic glucocorticoid use makes such a trial unfeasible.

Our study revealed no benefit of glucocorticoid medications in preventing late PTSs, and first late seizures were more common in those receiving glucocorticoids within 1 day of their TBI. This study provides further evidence to support the position paper of The Brain Trauma Foundation of the American Association of Neurological Surgeons Joint Section on Neurotrauma and Critical Care advising against the routine use of glucocorticoids in TBI (15).