Risk and cause of death in post-traumatic epilepsy: a register-based retrospective cohort study

Objective Post-traumatic epilepsy (PTE) is common, but its impact on survival after traumatic brain injury (TBI) of different severity and in different demographic patient groups is unknown. We analyzed the risk of death associated with PTE with adjustment for TBI severity, causes of death, and the contribution of epilepsy as direct or contributing cause of death. Methods Register-based, retrospective cohort study. All individuals hospitalized in Sweden for a TBI between 2000 and 2010 without prior seizures were identified in the National Patient Register, with follow-up until 2017. Subsequent epilepsy was identified by ICD-10 codes. Time-dependent Cox proportional hazard ratio (HR) was used to assess hazard of death, with epilepsy as a time-updated covariate. Adjusted analyses for age, gender, injury severity and comorbidities were also performed. Causes of death were analyzed using the Cause of Death Register. Results Among 111 947 individuals with TBI, subsequent epilepsy diagnosis was associated with a crude HR of 2.3 (95% CI: 2.2–2.4) for death. Stratified analyses showed a HR of 7.8 (95% CI: 6.5–9.4) for death in younger individuals. Epilepsy was a more common underlying cause of death in younger individuals. Conclusion PTE is associated with a higher risk of death and epilepsy seems to contribute to a significant proportion of deaths, especially in younger age groups. Future studies on whether improved epilepsy treatment can reduce mortality are needed.


Introduction
Traumatic brain injury (TBI) is a common cause of epilepsy. Acquired epilepsy in general increases mortality to a median standardized mortality ratio of 2.3 [1]. In post-traumatic epilepsy (PTE), mortality varies in the literature. A US study found an almost three times increased mortality in PTE and an increased risk of death in younger age groups [2]. A Taiwanese study reported a twofold increased mortality in PTE, also after adjustment for age, sex, and comorbidities [3]. Similarly, a Finnish study found 1.75 times higher mortality in PTE after non-mild TBI, but whether PTE or initial TBI severity explained the excess mortality could not be addressed for sample size reasons [4]. This is a key question. If PTE is independently associated with a higher risk of death, studies are needed on possible mechanisms.
We used Swedish register data to investigate the risk of death in PTE, with adjustment for injury severity, in a population-based study including over 100 000 individuals hospitalized for TBI in Sweden between 2000 and 2010. We also analyzed to what extent epilepsy itself was stated as a direct or contributing cause of death.

Materials and methods
The study was conducted on data from comprehensive Swedish national registers, based on the personal identification number, unique to all inhabitants in Sweden. The National Patient Register (NPR), the Cause of Death Register (CDR) are managed by the National Board of Health and Welfare and the population register is managed by Statistics Sweden. Reporting to the NPR and CDR is mandatory for all health care providers. Information on all hospital inpatient care since 1987 and specialized outpatient care since 2001, with improvements in coverage until 2005 is available through the NPR. Data from the cause of death register identify an underlying cause of death and contributing causes of death (other diseases and injuries contributing to death).
All individuals aged ≥ 18 hospitalized for a traumatic brain injury (TBI) between 2000 and 2010 were included (Emergency room visits were not included). TBI was defined according to International Classification of Disease (ICD)-10 as a diagnostic code of S06, S02.0-02.1, S02.7, or S02.9 (n = 111,947). The inclusion criteria are described in detail in Fig. 1.

Statistical analyses
All analyses were performed in SPSS version 26 (IBM, NY, USA), except for the time-dependent multivariable Cox proportional regression modeling and Consecutive Cox models which were done in SAS, version 9.4 (Cary, NC, USA).
HR of death for the entire population and different subgroups was calculated by time-dependent Cox proportional regression modeling with admission date for TBI as start, death as event, and censoring at death or 2017-12-31, either with just PTE entered as a time-dependent covariate (univariate), or with multivariable adjustment for age, sex, type of injury, and time-updated CNS comorbidities. Consecutive Cox models were performed dividing the follow-up time into Significance analyses were performed using chi-squared test for Cox proportional regression model and z test for causes of death.

Sensitivity analyses
In one sensitivity analysis, individuals with stroke, CNStumors, and CNS-infection were excluded-leaving 88 584 individuals for analysis. In another sensitivity analysis, only individuals with a first epilepsy diagnosis within 2 years of index date (the typical latency of PTE) [5], were classified as PTE (n = 2060).

PTE vs non-PTE
The median age at the time of trauma was slightly higher in the PTE group than in the non-PTE group (58, range 18-99 vs 56, range 18-100). Male sex and brain comorbidities were more common in PTE. Individuals with PTE had in general a more severe TBI, although mild injury was most common in both groups ( Table 1).
The median time from injury to epilepsy was 2 years (range: 0-18 years) and median time from trauma to death was 6.31 years in the PTE group and 3.92 years in the non-PTE group.

Survival
During follow-up, 48.1% (95% CI: 46.6-49.6) in the PTE group died, compared to 37.6% (95% CI: 37.3-37.9) in the non-PTE group. The incidence proportion of death was highest during the first year after the epilepsy diagnosis in the PTE group, in which 27.1% (95% CI: 25.2-29.0) of the total deaths occurred.

All-cause mortality
All-cause mortality from the time of epilepsy was 77.3 deaths/1000 person-years in the PTE group. When stratified by injury severity, the all-cause mortality was 63.0 deaths/1000 person-years after focal cerebral injury, 64.4 deaths/1000 person-years after diffuse cerebral injury, 100.8 deaths/1000 person-years after extracerebral injury, 65.7 deaths/1000 person-years after fracture and 76.7 deaths per 1000 person-years after mild injury.

Hazard ratio for death
The HR of death was doubled in patients with PTE, and only slightly affected by adjustment for age, sex, type of injury, and comorbidities. Stratified analyses showed that the association between PTE and death was significantly higher in younger age groups, with an eightfold hazard of death in individuals 18-39 years of age (Table 2). There was no significant difference between the adjusted HR of death in men compared to women, or between different types of structural TBI. When adjusting for age, sex, type of injury, and comorbidities, the HR was still increased in patients with PTE in all patient strata.

Cause of death
We next compared registered causes of death between individuals with PTE and controls. Cerebrovascular disorders, disorders of the nervous system, disorders of the digestive system, and external causes (including trauma) were more common causes of death in individuals with PTE (Table 3).

Death associated with epilepsy
Epilepsy, including seizure and status epilepticus, was an underlying or contributing cause of death in 18.4% (95% CI: 16.8-20.1) and the underlying cause in 2.3% (95% CI: 1.7-3.0%), of the individuals with PTE. The proportion of deaths with epilepsy as underlying or contributing cause was stable over the first five years from the diagnosis. (Fig. 3). In the PTE group, epilepsy or status epilepticus was a significantly more common cause of death in individuals dying < 60 years of age, compared to the two older age groups.

Sensitivity analyses
Since PTE typically develops within 2 years of TBI [5], we performed a sensitivity analysis including only individuals with epilepsy diagnosed within 2 years of their injury. In this group, the all-cause mortality from the time of epilepsy diagnosis was 69.6 per 1000 patient years in this group and the HR for death was 1.9 (95%CI: 1.7-2.0). Another sensitivity analysis assessing the HR for death excluding individuals with a diagnosis for stroke, CNS-infection, and CNS-tumors was performed, showing no significant difference from the main analysis-2.2 (95%CI: 2.1-2.4).

Discussion
We found a twofold increased risk of death after PTE, and the risk remained elevated with adjustment for, sex, age, type of injury, and comorbidities. To our knowledge, this is the largest study so far evaluating risk of death in PTE and one of the first with a data set allowing adjustment for TBI severity. The elevated HR associated with PTE was highest in younger individuals. The fact that PTE in itself was associated with increased risk of death, and not merely a representation of TBI severity, highlighting a need for more studies on mechanisms and preventive measures. The findings are probably Fig. 2 Causes of death in exposed group compared to unexposed, stratified by age groups at time of death. IHD ischemic heart disease, CVD cerebrovascular disease, OHD other heart disease, ID infectious diseases, ENMD endocrine nutritional and metabolic disease, MBD mental and behavioural diseases, DNS disorder of nervous system, DRS disorder of respiratory system, DDS disorder of digestive system, ECM external cause of mortality generalizable; the rate of death in our material was similar to that seen in a recent study on PTE [3], which is in turn substantially higher than after TBI in general [7]. Epilepsy, seizure, or status epilepticus was stated as an underlying or contributing cause of death in 18,4% of patients with PTE, with epilepsy being a more common cause of death in younger individuals.
That the association between death and PTE was highest in younger age groups is a cause for concern. External causes of death were significantly more common in PTE, iterating previous reports that accidents are overrepresented in adults with epilepsy [8]. Sudden unexpected death in epilepsy is another common cause of death among younger adults with epilepsy [9], suicide is also more common among individuals with epilepsy and those after TBI [10,11]. Our findings indicate that the number of life years lost to epilepsy-related deaths in PTE could be substantial. It is probably premature to conclude that PTE-related deaths are more common in young individuals. Older individuals are at higher risk of dying in general, which could have attenuated the effect of PTE-associated mortality in our regression models.
PTE carried an increased adjusted HR for death to a similar degree in both males and females, This is in line with previous findings in PTE [3] and suggests that although males have a higher standardized mortality ratio among epilepsy patients in general [1], this may not extend to PTE. Furthermore, the adjusted increased HR for death associated with PTE was similar in different trauma severity groups, indicating that the effect of PTE is proportional to the trauma severity.
Our study was register-based, with associated strengths and weaknesses. One strength is that NPR-registration in Sweden is mandatory in specialized health care, and an ICDcode for epilepsy has > 90 percent accuracy and the register in general an accuracy of 85-95% [12,13]. Another strength is the cohort size and population-wide design, which increases generalizability, at least to high-income countries with a similar panorama of injury mechanisms for TBI. The levels of PTE and death closely match those described in the literature [3]. There are also limitations. Since the study was register based, we are unable to assess clinical signs related to the trauma which can affect epileptogenesis, such as location of trauma lesion, length of unconsciousness or amnesia [14]. Another limitation is that individuals classified as PTE may have epilepsy for other reasons than TBI, to address this issue a sensitivity analyses was performed, including only patients that received the epilepsy diagnosis within two years after TBI-the typical latency of PTE [5]showing only a slightly lower HR, and a similar all-cause mortality ( Table 2). Another analysis assessing possible contribution of other comorbidities as a cause of the epilepsy, we excluded individuals with stroke, CNS-tumor, and CNS-infection, which did not alter our main results. Since epilepsy is generally under-recognized as cause of death, findings on causes of death should be interpreted with some caution. Our findings raise important questions. Can improved epilepsy management reduce mortality? Better information on seizure-related risks could perhaps prevent or mitigate accidents. Since many patients with PTE were older and cerebrovascular disease was an important cause of death, another question is whether use of enzyme-inducing antiseizure medications like phenytoin or carbamazepine may have interfered with cardiovascular primary or secondary prophylaxis.
In conclusion, we found an increased HR of death in individuals with PTE, also with adjustment for TBI severity. The HR was mainly increased in younger individuals. More in-depth study of the healthcare provided to PTE patients are needed to elucidate modifiable mechanisms underlying the higher risk of death.