Fever in the neurosurgical patient has been explored multiple times by the CICM examiners, mostly taking the form of the "Does This CSF Look Infected To You" sort of data interpretation question, and not until Question 1 from the second paper of 2023 had they asked for anything detailed or nuanced. In fact "detailed and nuanced" were key words from the marking rubric, and to achive a maximum score the candidate would have had to demonstrate "clinical perspective" and that they were "able to appreciate the challenges of making the diagnosis of ventriculitis".
In summary:
Considering how common the situation, it is remarkable that it is so hard to find a quality resource on this topic, with many falling short in either the "nuance" or the "detail" category of the college rubric. For example, Laws & Jallo, whose 2010 article comes up in the top three search results for "fever in neurosurgical patient", include a list of non-infectious causes of fever that includes "Neoplastic Disease: Primary CNS Tumors and Malignancies, Leukemias and Lymphomas, Metastatic Disease, Oncologic Diseases", which is sufficiently generic and repetitive to earn no marks from the CICM examiners. Of the papers that were useful in compiling the summary below, the best came from Chavali et al (2019). For the management of ventriculitis specifically, one cannot go past the 2017 IDSA clinical practice guidelines
If one were pushed against the wall, one might produce a list like this, classifying the causes as "infectious" and "non-infectious":
The candidate who wishes to absolutely excel at this will make a list of these either according to the "clinical perspective", i.e. from most to least likely, or will regurgitate a series of possibilities using some kind of cognitive aid like VINDICATE or VITAMINC to help them remember different categories. Sources of sepsis can also be generated by thinking of patients in head-to-toe terms, or by using the crude mnemonic device below:
However this is unlikely to ever become the source of a large number of marks, as end-stage ICU trainees are all expected to be able to effortlessly generate long lists of differentials like this, and a question like this would not discriminate very well between them. "Nuance" and "detail" lurk in the neurological causes of fever and the distribution of the incidence among different diagnostic categories of neurocritical patients.
How would you add nuance and detail into an answer about the possible causes of fever in a neurosrugical patient? One might try to make oneself appear more clever by gesturing in the direction of pre-test probabilities and patient characteristics. Some patients in the neurotrauma ICU are going to be more likely to develop fevers than others, and the most prevalent causes of fever among them are going to be different. Commichau et al (2003) reported a 25% incidence of fever in their prospective cohort, with about half of the temperatures beign explained by some infectious aetiology, and 28% remaining unexplained following whatever the authors considered a "complete diagnostic evaluation". For others, the incidence was lower (16%); others yet found it to be higher (40-60%), but most would agree that the patients at greatest risk of developing a fever in the ICU are the subarachnoid haemorrhage patients. In that group the incidence was 65% for Commichau et al and 48% for Rabinstein & Sandhu (2007), which was the highest incidence followed by patients with traumatic brain injury and those having refractory seizures. Of these SAH patients the vast majority had negative cultures and were described as having a fever of "central" cause.
Occasionally, all the cultures are negative, and the concept of "central" or "neurogenic" fever is trotted out to explain the high temperature, which defies logic as by definition all fevers are central and neurogenic in origin, thermoregulation being a hypothalamic function. This term is loosely used to describe temperature fluctuations which are due to some intrisncially "neurological" factor mostly related to the original brain pathology with which the patient was admitted to ICU, and attributed to various pyrogenic chemicals released in the process of brain injury or haemorrhage. "Central fever" is therefore the incantation used to ward away the use of antibiotics in this group.
How does this work? The mechanism is supposed to be the action of locally generated inflammatory cytokines directly on the organum vasculosum lamina terminalis and the preoptic area of the hypothalamus, making these sensors believe that the patient has a systemic infection (a situation akin to holding a lighter flame to the temperature sensor of the sprinkler system). These pyrogens originate from broken brain tissue and blood, or so the folklore says. Goyal et al (2020) list some generic cytokines and inflammatory mediators which are implicated, but a careful review of their references suggests that these are merely cytokines that are implicated in producing fever more broadly. Still, it appears these are present in the CSF of human neurosurgical patients. Animal data suggests that the presence of blood in the ventricles is essential to this process, and specifically blood in the third ventricle (Migliorino et al, 2023). Prostaglandins do not seem to be involved, which suggests that drugs acting on those pathways (eg. NSAIDs) should not be effective to ameliorate the temperature spikes.
So, confronted with a febrile neurosurgical patient, how does one convince oneself that this fever is "central" and that antibiotics are unnecessary? Unfortunately no guidelines or diagnostic criteria are available to help, even to discriminate infectious from noninfectious causes of fever in this group, let alone to make the diagnosis of "central" fever. There is usually neither the time nor the interest to do CSF cytokine levels, nor would this really be relevant anyway. Most reasonable people settle on this diagnosis when they are satisfied that:
Of the latter:
It does not require a wild leap of the imagination to consider that the mere presence of a large unwanted plastic object in your brain is enough to attract the attention of the immune system and to create a "pleocytosis", i.e. an increased white cell count. Under normal circumstances about 1 in 500-1500 cells in the CSF is allowed to be a leukocyte, but these are values derived from meningitis in the community, where patients are having "traumatic taps" during their LP with tiny flecks of blood being aspirated into their fresh lumbar cistern sample. What a "normal" white cell count should be in the cerebral CSF of patients with massive amounts of intraventricular blood is not well described. Moreover, and much more disturbingly, the WCC of the EVD-derived CSF appears to change from sample to sample within the same patient, even without various change in patient care (eg. repositioning), such that even serially drawn CSF samples from the same patient in the same position yielded wildly different results. Lastly, patients with normal CSF cell counts are still occasionally found to have culture-proven ventriculitis, as in Mayhall et al (1984). Still, brave souls have attempted to derive some meaning from these values, and found that occasionally (eg. in combination with other variables) they have some predictive utility. For instance, Willer‐Hansen et al (2022) and Brooks et al (2022) were able to find enough ventriculitis patients to publish their WCC/RBC ratios with a straight face; the Australian data from the latter (assuming ventriculitis has some kind of national character) gives a figure of 1:106 for a fair and reasonable AUROC of 0.706. In other words, it appears that you are permitted about five times as much WCCs in your CSF if the sample is collected from an indwelling EVD.
Question 1 from the second paper of 2023 invited the candidates to discuss the challenges associated with making the diagnosis of ventriculitis, revealing that the examiners find the exercise challenging. The main reasons for this, apart from the difficulties with the CSF cell count, is that ventriculitis mainly has meningitis-like symptoms, and the comatose TBI patient is unlikely to be able to report any of these. Nor is their neck stiffness easily elicited, nor can anyone ever remember what Kernig's sign and Brudzinski's sign actually are, nor are these even remotely reliable in the diagnosis of meningitis among conscious and cooperative patients, let alone unconscious intubated ones. In fact, there are enough challenges that to list them requires a structure.
There are many, and it will never be clear to the observer which definition each individual neurosurgeon will adhere to, or how tightly. Ramanan et al (2021) list five competing definitions, as an example. To reduce the cognitive load for the CICM exam candidate, the CDC definition will be reproduced here, for no reason other than the fact that it is the longest, and therefore the most attractive to the question-writer.
CDC criteria for ventriculitis
The exam candidate is reminded that the SAQ asked for challenges rather than to regurgitate a definition, which means the candidate mainly needed to complain about a lack of an accepted definition in order to pass.
In general, SAQs which ask the candidate to describe their assessment are usually expected to be answered in terms of history, examination, investigations and imaging, which means the following structure:
Bacteria do grow quickly, but CSF is usually sterile to begin with, which means there should be some lag time between the insertion of the contaminated EVD and the development of ventriculitis. Ergo, fevers in the first 48-72 hours following EVD insertion are highly unlikely to be related to intracranial infection. This is probably the only one of these features that could be determined from basic principles. Other historical features would have to come from the empirical findings of audits and prospective cohort studies, such as Savin et al (2018), Kim et al (2012) and Dorrenstejn (2019):
Patient characteristics:
Intraventrciular blood
Systemic infection
Duration of catheterization (11 days seems to be the cutoff for increasing risk of infection)
Immune compromise
Features of surgery:
Craniotomy (vs burrhole)
Superficial surgical site infections
CSF leakage
Catheter management
EVD, rather than a parenchymal pressure transducer
Frequent sampling
EVD irrigation
Non-tunnelled vs. tunnelled EVD
Symptoms, in case the patient has the wherewithall to report them, might include:
Headache
Photophobia
Neck pain
For the intubated patient,
For the conscious patient, also:
A gold standard is obviously a positve culture, but other features are also available:
Let's say you are convinced the CSF is the source of the infection. What will you do about it? The 2017 IDSA guidelines are probably the most comprehensive to answer that question. In summary:
The role of this strategy is mostly reserved for those situations where the patient has failed to respond to more conventional therapy. A failure of response, of course, is in the eye of the beholder, but most reasonable people would describe it as the persistence of positive CSF cultures or non-resolving clinical features of ventriculitis. Other possible reasons include:
The practical aspects that one would need to consider include:
Is this... safe? Or rather, to put it in a less idiotic wording, is the presence of highly concentrated vancomycin in your brain somehow distracting? Will the patient notice this, will they feel pain, or will the antibiotic have some kind of CNS effects at this ridiculious concentration? Well. It would be logical to expect the CNS toxicity of drugs already known to have neurotoxic effects (eg. gentamicin) to be enhanced by intrathecal administration. However, for the others, the toxicity is remarkably mild; or at least the patients are too obtunded to report any disturbance to their higher level function. Llave et al (2021) performed a thorough review and found very little evidence of CNS toxicity. In short, almost everyone would agree that the benefits outweigh the risk.
On the other hand, let's say everyone has convinced each other that the patient is febrile because of those mysterious cytokine-mediated central processes. What would you do about it? Anything? Is this fever even a problem?
In short, the answer is yes, the fever is a problem and you would absolutely make every effort to do something about it. Each category of neurocritical ICU patient will be somehow disadvantaged by a high temperature:
Goyal et al recommend the following options, with the caveat that they are probably unscientific, and that they are mostly known from case reports:
This is unknown. Studies looking at the aforementioned "cumulative fever burden" looked at the first thirteen days of admission, so we know those two weeks are important. However what the effect is of fevers in the distant future is more difficult to determine. Lavinio et al, in a 2023 consensus statement on therapeutic temperature management in neuroICU patients, only offered that "TTM should be maintained for as long as there is potential for secondary brain damage", which leaves the interpretation very open, as one might make the argument that every male aged 18-24 with a driving licence is at risk of secondary brain injury. This lack of consensus regarding the optimal duration of active fever management in neuroICU patients overlaps with the lack of consensus regarding the optimal strategies for temperature management more broadly, and is therefore unlikely to ever appear as an SAQ, unless it is to help trainees demonstrate that they understand the controversy (eg. as a "discuss" question)
Chavali, Siddharth, et al. "Fever in the neurointensive care unit." Journal of Neuroanaesthesiology and Critical Care 6.03 (2019): 275-283.
