Clinical studies of the role of inflammation in brain tissue rely on determinations of biomarkers from clinically accessible fluid compartments. Plasma is clearly the most accessible and can be measured repeatedly to determine changes over time. We have shown that the concentrations of many of the commonly measured markers were similar in plasma and CSF compartments and in some instances plasma markers were higher. Direct comparisons with studies reported previously are problematical, since the assays differ and virtually none make reference to recognised standards. CSF IL-1ß has been reported as undetectable, or low and stable, in some studies of SAH [3, 13], while another study reported it as increased, but in the low picogram range , and two others recorded tens to hundreds of picograms per millilitre [5, 6]. TNF-α has similarly been found to be low or undetectable in some studies [4, 13], but increased in others [6, 14]. As in our study, IL-1Ra and TNFR1 concentrations have been reported as similar in CSF and plasma or serum, but with an increase after SAH in at least some patients [14, 15].
In contrast, CSF concentrations of IL-6 and IL-8 were several orders of magnitude greater than those in plasma after SAH. Other studies have consistently found that CSF concentrations of IL-6 and IL-8 are increased one or two orders of magnitude above those in normal CSF (taken preoperatively for spinal anaesthesia, during elective aneurysm clipping or for diagnostic purposes that subsequently found no inflammatory cause) or plasma values [3, 4, 6, 13, 16, 17]. There was a superficially steady relationship between plasma and CSF concentrations for individual patients, particularly for IL-8. However, the relationship between these compartments in different individuals varies considerably, as indicated by the poor correlation between compartment values for the cohort. This could reflect variability between individuals in the transport of cytokines from CSF to plasma, or simply be due to production occurring discretely in each compartment. The probability of the latter explanation being correct is suggested by the dissociation between plasma and CSF IL-6 following infection.
The absence of a gradient concentration between CSF and plasma for most markers, and the lack of a clear relationship between IL-6 or IL-8 in CSF and plasma, suggest that plasma concentrations of inflammatory markers cannot provide useful information about inflammation in the brain. Several neurosurgical studies have used microdialysis to address this issue, including a study in patients with SAH [18, 19]. However, apart from the difficulty of placing the dialysis probes such that they provide a representative sample of events taking place in damaged tissue, collected data may reflect tissue damage caused by inserting the probes. It is notable that in a study where a significant change was seen over the collection period (for IL-6, IL-8, macrophage inflammatory protein 1ß and fibroblast growth factor 2) the highest values were observed in the first 6 h following probe insertion and these values fell over the next 24 h .
It may be that systemic inflammation and plasma markers are of greater significance than markers detected in the CNS, perhaps because this reflects neurovascular involvement. This is borne out to some extent by the reported associations between systemic inflammation or peripheral inflammatory markers and clinical outcome in situations following brain damage [7, 20–23].
Our observations additionally raised questions as to what might be causing some of the changes in cytokine concentration within compartments and altering the relationship between central and peripheral concentrations, particularly for IL-6. After reviewing the clinical report forms and patient notes the most consistent other event occurring post-EVD insertion was infection. It is noteworthy that rather little mention is made of infection in study reports or commentaries on the association between cytokines or inflammation following SAH. Infection is a classic activator of inflammation and immunity, and bacterial products are widely used to activate inflammatory responses in experimental settings [24–27]. Ignoring the impact that infection may have on measures of inflammation seems surprising in any study of inflammatory markers, but perhaps particularly in conditions where infection is prevalent. Quite apart from its potential to induce production of cytokines, infection could modify cytokine production occurring as a result of the underlying pathology or might independently have an important influence on clinical outcome. Infection therefore has the potential to obscure relationships between a clinical condition and inflammation, or to suggest associations where both outcome and inflammatory markers are independently influenced by infection.
Depending upon the assessment measures used, infection could potentially alter clinical outcome in several ways. It may influence outcome by directly affecting body systems not threatened by the underlying clinical condition of interest, or modify metabolic functions, including vascular or renal function, such that the primary pathology progresses more rapidly. Alternatively, the infectious stimulus may exacerbate inflammatory responses associated with the condition being investigated. In this respect, several studies have shown that infectious stimuli increase the severity of CNS ischaemic pathology attributable to inflammation [28, 29].
Since each of the measured markers has been associated with inflammation it might be expected that most would be similarly affected by infection. Considering the remarkably similar production pattern of IL-6 and IL-8 in these SAH patients, it is perhaps particularly surprising that IL-8 production was quite resistant to modification by infection. Induction of IL-6 seemed particularly sensitive to infection in the patients studied and these data suggest that monitoring IL-6 might be used as a marker of infection.
The association between IL-6 and peripheral infection has been well described ([25–27, 30, 31] and has been indicated as of value in diagnosis [32–34]. It is important to note that ours was a post-hoc analysis, driven by observed data, and therefore likely to overstate the association. Nevertheless, the strength of association is striking and merits further confirmatory studies. Elsewhere, CSF cytokines have been evaluated to diagnose infection in a variety of neurosurgical conditions and IL-1 and TNF were suggested to be of use, whereas IL-6 and IL-8 were not . This contrasts with our study, where IL-1 or TNF were barely detectable and only IL-6 production was consistently modified by infection.
At least three other studies have considered the value of monitoring CSF IL-6 to aid the diagnosis of VRI. Of these, measurement of CSF in four pre-term infants with hydrocephalus suggested that it may be useful . Analysis of IL-6 in 21 adult patients with EVDs concluded that IL-6 was not of value . However, a subsequent study in 75 patients with EVDs concluded that IL-6 had significant value for predicting VRI 1 day before other diagnostic markers .
The apparent lack of consensus on how VRI is diagnosed highlights the need for an improved method for discrimination. Even positive culture is not definitive and may be identified as due to catheter colonisation or contamination in the majority of cases . Similarly, lack of consensus on use of prophylactic antibiotic cover during use of intracranial monitors and EVDs , combined with current concern over inappropriate use of antibiotics, identifies the need for improved, early identification of infection. This study suggests that there would be value in clarifying whether rapid IL-6 measurement may provide a sufficiently sensitive and specific marker to identify VRI early and make a better informed decision on antibiotic use.