Article Text
Statistics from Altmetric.com
I read with great interest the recent article by Townend et al1 in which the authors studied the predictive value of protein S-100B in patients with head injury upon performance in the extended Glasgow outcome scale (GOSE). One important criticism is that the study was performed in patients with head injury defined as “any blow to the head causing a clinical diagnosis of head injury to be made, even if insufficient to cause definite loss of consciousness” and not only in patients with traumatic brain injury, which is defined at least through loss of consciousness, amnesia, or postconcussion syndrome. Consequently, relevant abnormality of the brain even in minor traumatic brain injury was only detected in a few patients.
In addition, cerebral computed tomography (CT) was only performed in 15 of 148 patients. The extent of possible traumatic brain injury in the patients in the study by Townend et al1 cannot be estimated. Patients with frontal contusion lesions in CCT and/or diffuse axonal injury were not separately identified in this study. Those patients are at high risk of having neuropsychological deficits and also frequently suffer from loss of insight. This may falsify the outcome measured by the extended Glasgow coma scale that was obtained by telephone interview only. Assessment by phone has limitations and cannot substitute a detailed neurological and neuropsychological examination that would reveal the above mentioned deficits.
In literature, CT controlled studies by Romner et al2 (RIA), Ingebrigtsen et al3 (RIA) and Biberthaler et al4 (LIA-mat) calculated that an undetectable protein S-100B or protein S-100B below a cut off point at 0.1 ng/ml predict normal intracranial findings in CT. Herrmann et al5 (LIA-mat) showed that an initial S-100B value above 0.14 ng/ml has a high predictive value for short-term and long term neuropsychological deficits in traumatic brain injury. A prospective study has not been performed yet.
Because in the study by Townend et al1 measurements of protein S-100B were performed retrospectively without CT control or short-term or long term clinical monitoring, the study is of no clinical value.
Before implementation of a much needed neurobiochemical marker of brain damage in traumatic brain injury, there is a need for a prospective study of protein S-100B as a neurobiochemical marker of brain damage. This would include hospital monitoring of the patients with an initial cranial CT and MRI control as well as a short-term and long term neuropsychological follow up.
References
Authors’ reply
We thank Wunderlich for his thought provoking letter. The chief reservations expressed regarding the potential applicability of our findings seem to be that the entry criteria were too broad, that the measurements of protein S-100B were made retrospectively, that no CT control was performed, and that follow up might be biased. We will consider these points in turn.
The entry criteria were kept as broad as possible to enable the full cross section of patients with head injury to be evaluated. We are aware that traumatic brain injury is delineated more precisely according to the presence of a period of altered consciousness, particularly for research purposes. However, we were unable to find evidence in published literature that disability after head injury is confined to this group, although the rate would be expected to be higher than in those without such an alteration of conscious level. Also, there are no published data that we are aware of that demonstrate S-100B levels in those without altered consciousness after head trauma. We thought this would be of interest as clearly if there was a large proportion of this group with raised S-100B level and a uniformly good outcome, then its use as a prognostic marker would be limited by this false positive rate. For these reasons we consider our entry criteria apposite. By keeping patient selection as simple as possible, we anticipate that the least experienced practitioner would be competent to identify a patient in their practice that would be represented by our data. Our study, therefore, passes the test of applicability.
Wunderlich states that our S-100B measurements were made retrospectively. Our cohort was recruited prospectively. The blood samples taken for S-100B level estimation taken at initial assessment were analysed once outcome had been assessed. Blinding the outcome assessor to the S-100B level in this way was intended to reduce the risk of bias. Ours, therefore, is a prospective study. If the aim of Wunderlich’s comment was to reflect the need for prospective validation of the cut off points we derived, then we agree, and are collecting data to that end.
The role of CT in the prediction of head injury outcome, or the relation between S-100B level and CT findings were not the aims of this study. CT data were included to demonstrate the infrequency of the use of this imaging modality in current UK practice, and thereby emphasise the role a serum marker might have. The purpose of CT in the emergent care of the patients with head injury is to identify lesions amenable to surgical intervention. Patients disabled after so called mild head injury often have normal CT scans, indeed our data suggest that currently in the UK many such patients will not even undergo such an investigation. There is also evidence that serum S-100B is a better predictor of outcome than Marshall CT classification after severe head injury.1 We therefore foresaw little benefit in this study in correlating S-100B with CT abnormality. CT is likely to remain a poor surrogate for the entity we specifically sought to assess, namely neurological disability, which we scored directly using a validated tool. Routine CT, therefore, was not necessary in this study. Clearly a serum marker that, if “negative”, could exclude a lesion requiring surgical intervention would be of immense value, but that was not the purpose of this study.
The possibility of the misrepresentation of outcome by patients with undiagnosed frontal contusions because of lack of insight was not considered when designing our study. That this effect might be exacerbated by telephone follow up is conceded. However, despite their limitations, we believe our arrangements ensured a reasonable follow up rate. This is not routinely the case in head injury studies. We also believe that the validity of our outcome measure has been demonstrated in published literature, and is also clinically relevant. The purpose of attempting to predict outcome as we have done is to identify those patients with head injury likely to benefit from intervention. If the assessment of that need is based on their inability to return to their previous life, rather than important but not so obviously relevant neuropsychological impairments, then a more compelling case can be made for such a programme to be resourced.