TRAUMATIC BRAIN INJURY (TBI) is an important cause of death worldwide and contributes to a considerable number of deaths and disability in the United States. The Centers for Disease Control and Prevention has estimated that about 50000 of the 1.4 million people who sustain a TBI in the United States will die as a direct consequence.1

Death from TBI can be either acute or distant from the traumatic event. Most deaths occur in a time frame from immediately to several months after injury.2,3 However, patients surviving the first 6 months postinjury have been posited to be at increased risk of premature death compared with non-head-injured controls since at least World War I.4 Causes of death could be related directly to the injury, such as status epilepticus, indirectly, such as falls, or not involve the central nervous system at all, such as cardiovascular disease. One distant and potentially life-threatening sequela, which has been the subject of several case reports,5,6 is intracranial neoplasm, presumably the result of growth of ectopic glial or meningeal cells that have been dislodged from their normal anatomical location by the trauma.

In 2007, in response to a request from the Department of Veterans Affairs, the Institute of Medicine convened a committee to undertake a review of the evidence surrounding long-term sequelae of TBI, particularly as it concerned veterans and active duty troops serving in Iraq and Afghanistan in Operation Iraqi Freedom and Operation Enduring Freedom, respectively. The report of the committee was published in 2008. It contained one chapter on "other" distant outcomes, specifically premature mortality, and intracranial neoplasms. In this article, we discuss the evidence that the committee reviewed regarding the possible association of TBI with premature mortality and intracranial neoplasms and its findings. Because of the uniquely military focus of the committee's review, we have separated discussions of TBI in military and civilian populations.

METHODS

We searched the published biomedical literature for studies that examined the association between the predictor variable traumatic brain injury and the outcomes of mortality and intracranial neoplasms that occurred at least 6 months postinjury. TBI had to be due to an external physical force rather than a degenerative or congenital condition. The methods have been described in detail elsewhere in this issue.7 Briefly, we searched multiple databases to identify potentially eligible articles, reviewed titles and abstracts, and abstracted data from those that met our inclusion criteria. We classified studies as primary studies if they had been published in a peer-reviewed journal, or undergone an equally rigorous review process, included details of their methods, included appropriate reference or unexposed groups, had sufficient statistical power to detect an effect, had sufficiently representative groups followed up to ensure external validity, and used reasonable methods to control for confounders and to minimize systematic error. Secondary studies were those that were less rigorous in their methods, but otherwise met inclusion criteria. We did not attempt to calculate summary statistics using meta-analytic methods but rather conducted a narrative systematic review. We examined the questions of premature mortality and posttraumatic intracranial neoplasms separately. See Tables 10.1 and 10.2 in Supplemental Digital Content 15 and 16, http://links.lww.com/jhtr/A31 and http://links.lww.com/jhtr/A32.

RESULTS

Premature mortality

We retrieved approximately 1900 titles in our initial search. Of these, 12 met our search criteria as primary studies and 11 as secondary studies and were reviewed. Mortality in TBI patients can be studied from the time of injury, from the time of discharge from inpatient acute-care hospitals, or from the time of admission into or discharge from inpatient rehabilitation. Rates in cohorts of patients at those different points of entry into a study will be different. For example, rates in patients from the time of injury will be greatest because they include early deaths. In contrast, survivors of the acute phase, who are studied during rehabilitation, are likely to have lower mortality. We begin with a review of military and civilian cohorts followed from the time of injury and then review cohorts recruited and followed from various stages of hospitalization and rehabilitation.

Three studies of TBI in military populations have examined the relationship between TBI and premature mortality in detail. These include studies of World War I veterans,8-10 World War II veterans,11 and Vietnam War veterans.12 The longest of these cohorts was established in the 1920s at the Hinverletztenheim, a hospital for brain injuries in Munich that had been designated a center for evaluation and treatment of more than 5000 brain-injured Bavarian veterans who had sustained penetrating and nonpenetrating head injuries in World War I.5 Follow-up studies of cohort members who had survived until age 35 and were followed until age 65 showed that the life expectancy was about 4 years shorter in those who had sustained a TBI compared to a noninjured veteran control group.6 Epilepsy was a significant risk factor for premature mortality, and cerebrovascular disease as a cause of death was significantly more common among TBI patients than among noninjured controls (17% vs 12%, P < .01).7 About 1000 records were randomly selected, and death certificates were sought from social welfare offices in Bavaria and West Germany. Vital statistics were obtained for 647 cases and matched with those of 616 uninjured Bavarian World War I decorated veterans who could be traced because they received pensions. Veterans were excluded if they were born before 1880 or died before the age of 35, if death dates were unknown, or if records were incomplete. Both penetrating TBI and nonpenetrating TBI were identified; posttraumatic epilepsy was also assessed.

A similar excess of premature mortality was found among World War II veterans who had sustained penetrating head injuries compared with a matched control group that had peripheral nerve injuries (28% vs 17%, P = 0.03).8 This excess was concentrated almost exclusively in those who had posttraumatic epilepsy. Among male Vietnam veterans who also had sustained penetrating head injuries, excess mortality was also found but seemed limited to the first 3 years postinjury compared with actuarial projections for North American men.9

In civilian populations, a retrospective cohort study of patients with TBI who sought medical care in Olmstead County, Minnesota, identified 1448 cases of TBI from 1985 to 2000, 164 (11%) of which were moderate to severe and 1284 (80%) were mild.13 Excluding deaths that occurred in the first 6 months postinjury, there were 14 deaths in those with moderate to severe TBI and 69 in those with mild TBI. Neither of these differed from what was expected, using mortality tables from the 1990 Minnesota white population (for moderate-severe, 12.7 deaths expected, relative risk [RR] 1.10, 95% confidence interval [CI] 0.60-1.85; for mild, 58.6 expected, RR 1.18, 95% CI 0.92-1.49).

The only other study to examine premature mortality in a general population comes from the sports literature. To study the influence of intense physical training on life expectancy, Bianco and colleagues examined male athletes born during 1860-1930 who had been inducted into various halls of fame: baseball (154), ice hockey (130), tennis (83), football (81), boxing (81), track and field (59), basketball (58), swimming (37), and wrestling (32).14 The study, however, also allowed for mortality comparison between boxers and other athletes. The overall median life expectancy was 76.0 years; boxers had the lowest median life expectancy (73.0 years), but no differences in mortality were found by number of rounds or bouts, presumably a proxy for numbers of TBIs.

Other studies in civilian populations have followed patients from acute-care discharge, admission to rehabilitation centers, and discharge from rehabilitation centers. For example, a population-based study in South Carolina found that 8.4% of persons who survived acute-care hospitalization died within 15 months of discharge; 17% of these deaths were classified as TBI-related, and almost two-thirds of these deaths occurred within 3 months of discharge.15 Younger males, older males, those who had sustained more severe TBI, and those treated in hospitals without trauma centers were at the greatest risk of early mortality. An Australian study of persons admitted to a rehabilitation hospital found an excess mortality rate of 3.8 over the 5 years postdischarge compared with the rate in the general population with increased risk of death marginally associated with male sex, greater age, and a premorbid psychiatric history.16 Similarly, a review of patients with moderate to severe TBI admitted to Traumatic Brain Injury Model Systems' rehabilitation centers in the United States found that individuals with TBI were at twice greater risk of death than those in the general population (95% CI 1.69-2.31) and that life expectancy was decreased by an average of 7 years.17 Others have identified similar degrees of excess mortality in TBI patients discharged from rehabilitation in Pennsylvania (standardized mortality ratio [SMR] 2.78)18 and among those with chronic residual disabilities in California (SMR 2.5 among ambulatory patients).19 Several of these indicate that better functional status at the time of discharge from rehabilitation is a leading indicator of longer survival time.10,15,16

A series of other studies have reported on mortality in people who have TBI, primarily using cohort designs but without external comparison groups; we classified these as secondary studies. The best of the studies have followed patients from the time of TBI. Others have examined subsets of the universe of patients with TBI, such as those admitted into rehabilitation hospitals, in which case the more selective nature of the population (survivors of the acute period who enter rehabilitation) may lead to death rates different from that in the whole universe of TBI patients who survive the acute period. Six population-based studies found consistent results regarding TBI and increased mortality20-25 with mortality increasing with severity of TBI and age. Similar associations were found in consecutive series of patients admitted to hospitals with TBI.26-30

Posttraumatic intracranial neoplasms

Fourteen studies met our search criteria for primary studies and 2 as secondary studies and were reviewed. Studies that have examined the association between TBI and subsequent development of intracranial neoplasms have been exclusively retrospective and have included both cohort studies and case-control studies. Tumor types that have been studied include gliomas, including astrocytomas, ependymomas, and oligodendrogliomas; meningiomas; neurilemmomas; medulloblastomas; and vascular tumors.

Included in the primary studies were 3 large population-based retrospective cohorts, conducted in Olmsted County, Minnesota,31 Denmark,32 and Sweden33. These studies used medical records to ascertain TBI and then looked for incident intracranial neoplasms. Together these studies followed more than 500000 patients from the time of an initial TBI and identified 665 tumors (Table 1). In both the Swedish and Danish studies, an excess of tumors was noted in the first year postinjury; this excess is believed to be due to incidental diagnosis of brain tumors found during the work-up of TBI. After 1 year, in none of these studies was an excess of tumors found among TBI patients compared with what was expected among non-TBI patients (Minnesota, RR 1.0; 95% CI, 0.3-2.6; in Denmark, standardized incidence ratio [SIR] 31.15; 95% CI, 0.99-1.32; in Sweden, SIR 1.0, 95% CI, 0.9-1.2), and in no study was there an excess of a specific type of tumor. Thus, these 3 studies found that the incidence of brain tumors after severe head trauma was no different from the incidence in the general population if cases diagnosed in the first year after trauma were excluded.

TABLE 1. Incidence of intracranial neoplasms following TBI, population-based studies

Eleven other studies used self-reports or self-reports of physicians' diagnoses to ascertain exposure, which creates some methodological limitations because of the risk of recall bias and consequent overestimation of risk. These studies had mixed results. Studies from southern Ontario,34 Tennessee,35 and multiple centers in the United States, Europe,36 and Germany37 found no relationship using case-control designs. In addition, the 2 secondary studies identified from Minnesota38 and Italy39 had negative findings. However, 7 other studies conducted in the northeastern United States,40 northeastern China,41 Brazil,42 western Washington,43 and Los Angeles44-46 found a relationship. In particular, the results of the 4 Preston-Martin studies33,41-43 suggest that the odds of meningioma are increased in people who have had a TBI, especially those with relatively remote histories (15 years or more before); this was also found in studies by Phillips and colleagues40 and in a study with more heterogeneous histologic subtypes by Monteiro and colleagues.39

The series of studies by Preston-Martin and colleagues is of particular interest. In 1980 she and her colleagues conducted a population-based case-control study of intracranial meningioma in women in Los Angeles41; a study using similar methods was conducted in men in 1983.42 All microscopically confirmed cases of meningioma were identified in the Los Angeles County Cancer Surveillance Program, and each subject was matched to a control residing in the same neighborhood by sex, race, and year of birth. In the first study, women who received a diagnosis in 1972-1975 completed questionnaires regarding risk factors; questionnaires were returned by 189 subjects and 185 controls. A history of medically treated TBI was found to be a risk factor for meningioma (odds ratio [OR], 2.0; 95% CI, 1.2-3.5) and was independent of having had head or neck radiography. In the second study of men with meningioma, 120 subjects who received a diagnosis in 1972-1979 were matched with 105 neighborhood controls. TBI was associated with increased odds of meningioma if the person had ever participated in boxing as a sport (OR, 2.0; 95% CI, 1.1-3.2) or had a serious TBI that resulted in loss of consciousness (LOC) or a permanent scar (OR, 1.9; 95% CI, 1.1-3.2). The authors noted that many of the men with TBI had not received medical treatment.

The investigators then went on to study the association between serious TBI-defined as resulting in a medical visit, LOC, or dizziness-and brain tumor in men using similar methods.43 In this study, however, they focused on 2 tumor types-glioma and meningioma-first diagnosed in 1980-1984 in Los Angeles County and identified in the Cancer Surveillance Program. The 272 subjects (202 with glioma and 70 with meningioma) were interviewed and were compared with 272 neighbor controls. To be included, serious TBI had to have occurred 2 years or more before the brain cancer diagnosis. The OR for having had a serious TBI 20 years or more before was 0.8 (95% CI, 0.5-1.3) in those with glioma but 2.3 (95% CI, 1.1-5.4) in those with meningioma, suggesting that recall bias did not underlie the finding of increased odds associated with meningioma. Furthermore, the odds of meningioma, but not glioma, increased with the number of serious TBIs: with 1 TBI, 1.3 (95% CI, 0.6-2.9), with 2 TBIs, 2.1 (95% CI, 0.8-5.9), and with 3 or more TBIs, 6.2 (95% CI, 1.2-31.7).

Expanding the case-control approach internationally to access greater numbers of cases, the investigators studied men and women with gliomas and meningiomas in 6 countries (2 centers in Australia, 1 in France, 1 in Germany, 2 in Canada, 1 in Sweden, and 1 in the United States).33 Glioma and meningioma cases were in 729 men and 779 women who received diagnoses of either glioma or meningioma in 1984-1992 and were matched to controls by sex, age, and education. Participants were asked about medically treated TBIs, which were classified as serious if they caused LOC or amnesia or required hospitalization. Injuries were also classified by time before the brain cancer diagnosis. With these larger numbers, the investigators found that odds ratios for glioma or meningioma were not significantly increased in men who had had any TBI 5 years or more before diagnosis (OR, 1.18; 95% CI, 0.94-1.48, and OR, 1.49; 95% CI, 0.86-2.57, respectively) or who had had a severe TBI 5 years or more before (OR, 1.13; 95% CI, 0.87-1.48, and OR, 1.15; 95% CI, 0.57-2.34, respectively). There was no significant increase in the risk of either brain tumor in women regardless of TBI severity. Men had a slightly increased OR for glioma if they had sustained more than 1 TBI 5 years or more before (OR, 1.52; 95% CI, 1.00-2.32) but not if they had more than 1 TBI regardless of timing (OR, 1.67; 95% CI, 0.56-4.98). In men who had sustained their TBI 15 to 24 years before diagnosis, there was a statistically significant increase in the risk of meningioma (OR, 5.35; 95% CI, 1.72-16.62), but the increase was not seen in connection with other latent periods or in women.

DISCUSSION

Premature mortality

We found clear evidence of increased mortality in the acute phase after TBI, and for some time afterward in both military and civilian populations with moderate to severe TBI. In the military literature, posttraumatic epilepsy in patients who initially survive penetrating head injuries is associated with an increased risk of death and about a 5-year decrease in life expectancy. In the civilian literature, a large population-based study in Minnesota suggested that although there is clear evidence of increased mortality in the first 6 months after injury, there is no evidence of increased mortality in patients with TBI beyond 6 months, regardless of severity.10 Studies of the subset of more severely injured patients who survive initial hospitalization and require inpatient rehabilitation demonstrate a worse prognosis, consistent with the greater degree of residual compromise: mortality some 2 to 7 times higher than in age- and sex-matched comparison populations (Box 1).

Box 1. Committee findings regarding the association between premature mortality and TBI

Posttraumatic intracranial neoplasms

We found the literature regarding the risk of posttraumatic intracranial neoplasms more opaque. There is a clear relationship between TBI and the diagnosis of intracranial neoplasms in the first year postinjury, presumably the result of the incidental finding of brain tumors discovered during evaluation of the TBI. Whether there is a longer-term risk of intracranial neoplasms pathophysiologically associated with the injury is substantially less clear. While the bulk of the evidence suggests no association, there is some countervailing evidence. The 3 large population-based registry studies in Minnesota,28 Denmark,29 and Sweden30 found no association between TBI and risk of brain tumors. These were very large and compelling studies, primarily because they were able to limit overascertainment of exposure among cases due to self-reporting of TBI; the Minnesota study was able to ascertain exposure up to 44 years earlier, and there were more than 3.2 million person-years of follow-up in the Swedish study. Nonetheless, the Danish study almost reached statistical significance (95% lower confidence bound of the SIR was 0.99). In addition, well-conducted case-control studies found a relatively specific association between TBI and risk of later meningioma as opposed to other tumor types and a latent period of 10 years or more.37-43 This suggests the potential for a weak but significant association between TBI and meningioma after a latency period of 10 years or more after TBI, and points to the need for longer-term follow-up, especially in large registry-based studies, to understand if there is measurable risk and, if it is increased, when and with what types of tumors it is most likely to be observable (Box 2).

Box 2. Committee findings regarding the association between intracranial neoplasms and TBI

As with any study based on a review of the published literature, we were limited by what studies had been done. Our review was further circumscribed to focus on the epidemiologic and clinical literature in humans exposed to TBI and did not include animal studies or therapeutic studies, which might have been useful in further elucidating the risks of these longer-term outcomes. Moreover, the preponderance of traumatic brain injuries in military populations serving in Iraq and Afghanistan have some component of blast injuries, often complicated by other neurological and somatic trauma.47,48 This is likely a different mechanism of injury and pathophysiology than that in the vast majority of studies we reviewed,49 which typically involved blunt head trauma. Even in military studies, the mechanism was penetrating trauma with loss of brain tissue with the possible exception of the few studies from World War I where blast waves may have been focused and concentrated by the architecture of trenches. How much we can generalize from these other forms of TBI is not known, and the risk of long-term sequelae may be different. This problem is compounded by the mixing of mild TBI and more severe forms in several articles.

Specifically in our examination of premature mortality, care has changed significantly since many of the studies we examined were published. What was the standard of care for neurorehabilitation in 1985, when the large Minnesota study began accruing patients, has changed, and patients who may have died in past years are surviving with better functional outcomes.50 In addition, none of the studies we reviewed that examined the risk of intracerebral neoplasms included military populations, who may differ substantially from the civilian populations these studies observed, for instance in terms of age and types of exposures.

Nonetheless, we feel that we have amassed a substantial body of evidence regarding these uncommon outcomes. As cohort studies follow the long-term outcomes of service members returning from these conflicts, we hopefully will be able to better estimate the risks of premature mortality and intracerebral neoplasms in these populations.

REFERENCES