THE article by Dr Terrio and colleagues1 is the first report that describes the complete postdeployment screening process for traumatic brain injury (TBI) that US soldiers are undergoing on return from combat. The authors found a prevalence rate of 22.8% for "confirmed" TBI in their study sample, which was similar to that found in previous reports using cross-sectional self-report methods.2,3 Although a commendable effort, the article has serious methodological and scientific problems that render the findings and conclusions invalid.
The authors used the term "TBI" throughout the article and failed to provide information on what proportion of cases were in the mild (concussion) category. Presumably most, or all, of these cases were concussions, but the failure to characterize them as such makes it impossible to place this work within the broader concussion literature.
The authors concluded that, on the basis of their clinical expertise, they were able to both "confirm" that TBI occurred sometime during deployment (months earlier) and establish that postdeployment symptoms-such as irritability, headaches, dizziness, or memory problems-were in fact caused by these TBIs. Such causal inferences cannot be made on the basis of retrospective cross-sectional clinical data or through the statistical procedures utilized. There is no clinically validated method of determining whether altered consciousness was caused by TBI, acute stress, pain, battlefield disorientation, or other factors. There is also no validated way of determining the etiology of postdeployment symptoms months after injury events. Several well-controlled studies using independent measures have demonstrated that nonhead injuries are just as likely to produce "postconcussion" symptoms as are concussions.3-7
The authors themselves acknowledge at the end of their article the reality that since many of the TBIs were sustained during combat, this "[horizontal ellipsis] begs the question of whether symptoms recalled were wholly or in part related to other deployment-related medical or psychiatric conditions including posttraumatic stress." The authors' own data raise questions as to the specificity of symptoms, when they observed that memory problems often had an onset sometime after the acute period, in contrast to the extensive concussion literature, which shows that cognitive problems are most prominent only within the first few days after concussion.8
The most serious error was to report odds ratios of the association of postdeployment symptoms with TBI as evidence for a causal association. These statistical analyses were invalid for 2 reasons. First, the 5 symptom questions were not independent from the TBI questions in the postdeployment questionnaire, and thus the odds ratios violated basic statistical principles. Second, clinicians who were specially trained to focus on TBI were allowed to change the soldiers' answers in the questionnaires, thus biasing any comparisons between TBI and other injuries, again violating scientific principles.
In conclusion, the article offers no new information on the accuracy, validity, or effectiveness of postdeployment screening for mild TBI (concussion). The methodology employed was unscientific and seriously biased in the direction of demonstrating causal associations simply as an artifact of the screening procedures.
Charles W. Hoge, MD
Walter Reed Army Institute of Research, 503 Robert Grant Avenue, Silver Spring, Maryland
REFERENCES