Investigators have developed prognostic models that discriminate between patients with and without posttraumatic seizures (PTS) at year 1 and year 2 after traumatic brain injury (TBI), but perform little better than chance at predicting PTS, according to research published in the September issue of Epilepsia. The models do, however, identify potentially important predictors that may help to identify populations at risk for PTS.
“Individuals with characteristics identified in prognostic models as predictors of PTS represent subpopulations that may benefit from tailored seizure prophylaxis guidelines addressing unique premorbid characteristics, pathologies, and procedures,” said Anne Ritter, MPH, PhD, of Uniformed Services University of the Health Sciences in Bethesda, Maryland.
PTS is commonly recognized as a complication of TBI that may be acute or chronic. Although risk factors for PTS have been identified, predicting who will develop PTS remains difficult. Current PTS prognostic models are not widely accepted for clinical use and do not reflect current trends in injury, diagnosis, or care.
“Accurate PTS risk prediction could help define high-risk populations in support of clinical intervention trials. Predictive models could also inform clinical algorithms to identify individuals likely to benefit from tailored seizure prophylaxis or treatment,” said Dr. Ritter.
Dr. Ritter and colleagues conducted a study to develop and internally validate preliminary prognostic regression models that predict PTS during acute care hospitalization and at year 1 and year 2 post injury.
Study Population
Eligible participants had moderate or severe TBI, were admitted to a participating hospital emergency department within 72 hours of injury, were age 16 or older, and received acute care and inpatient rehabilitation within a TBI Model System (TBIMS) designated hospital system. Moderate or severe TBI was defined as posttraumatic amnesia lasting longer than 24 h, loss of consciousness lasting longer than 30 minutes, an emergency department Glasgow Coma Scale score less than 13, or positive neuroimaging findings.
People injured between October 1, 2011, and August 31, 2014, were included in the study. Patients injured during this time period, but not eligible for year 1 follow-up were excluded, however. Data for all participants were selected from the TBIMS National Database.
Researchers limited data to those collected at enrollment, year 1, or year 2 post injury. Enrollment data were collected using chart review and interview and included demographic, social, and injury characteristics, as well as preinjury personal and medical history and acute hospitalization outcome. PTS status, defined as the presence or absence of seizure activity, was the main outcome. It was determined during acute hospitalization, at year 1, and at year 2. In addition, investigators used multivariable logistic regression to generate prognostic models for PTS during acute hospitalization, at year 1, and at year 2. They internally validated models with resampling.
PTS Predictors of Interest
Of 2,136 participants, 2,042 had data available on all predictors identified in simple logistic regression for seizure during acute hospitalization. The sample’s demographic and clinical variables were similar in this investigation to those in previous TBI studies.
The final year 1 prognostic model identified injury severity, subdural hematoma, contusion load, craniotomy, craniectomy, seizure during acute hospitalization, preinjury condition limiting physical activity, preinjury mental health treatment or psychiatric hospitalization, and incarceration as risk factors for PTS. Craniectomy was the most statistically significant predictor in the final model.
At year 2, following validation, predictor variables included subdural hematoma, intraparenchymal fragment, craniotomy, craniectomy, seizure during acute hospitalization, and preinjury incarceration. Acute hospitalization seizure and craniectomy were the most statistically significant predictors of PTS at year 2.
Overall, the models displayed poor discrimination ability for PTS; however, these models may have added benefit, compared with prior models that were not being used clinically, said Dr. Ritter.
“These models must be examined in independent study populations to determine discriminability and validity outside the TBIMS population,” she added.
—Erica Tricarico
Suggested Reading
Ritter AC, Wagner AK, Szaflarski, et al. Prognostic models for predicting posttraumatic seizures during acute hospitalization, and 1 and 2 years following traumatic brain injury. Epilepsia. 2016;57(9):1503-1514.