Together with prevention of secondary brain injury, this topic enjoys a large amount of attention from the CICM examiners. The exam candidate ignores this topic at their peril. Numerous past paper SAQs have been dedicated to it:
Much of the material gathered here comes from two chapters of Oh's Intensive Care manual:
Before going any further, it is important to discuss what we mean by raised intracranial pressure. The 2016 BTF guidelines have decided that the threshold value is 22 mmHg. This was on the basis of a study by Sorrentino et al (2012), which was a review of a patient database including 459 patients. More broadly, the survivors had an average ICP of around 18 mmHg, whereas those who died had an average ICP of 27 mmHg.
Much is made of head position. With good reason. It seems to position the patient at least 30° head up decreases the ICP but does not decrease the CPP (Feldman et al, 1992). In fact, at least in the pediatric population, the angle of the bed is directly related to intracranial pressure. Jiang et al did a meta-analysis of this in 2015, and concluded that while an elevation of 30-45° is optimal, anything is better than 0° and that the head-injured patient's head should be elevated to whatever height and angle is permitted by their concurrent spine or pelvic trauma.
A good study in 2002 (Mavrocordatos et al) found that the ICP increased significantly whenever the head was in a non-neutral position. Mind you, those were patients undergoing elective neurosurgical procedures, and "for ethical reasons" all patients with a raised ICP were excluded from the study. No matter; earlier studies threw ethics to the wind and performed their measurements in patients with real ICP problems (eg. Hulme and Cooper, 1976). The authors found astonishing increases of up to 50 mmHg in patients whose head was turned all the way to one side. Curiously, turning the head to the right seemed to create a greater increase in ICP, likely because in about 68% of people the right IJ is dominant (Lichtenstein et al, 2001)
A good study of intracranial pressure with and without the rigid collar found that one can decrease the intracranial pressure of a TBI patient by about 4-5 mmHg simply by removing the rigid collar and using something like sandbags to stabilise the neck.
Endotracheal tube ties represent a tight circumferential object wrapped around the head and neck of the patient. Everybody's entry into the topic of ICP control seems to mention the loosening of these ties to improve cerebral venous outflow (eg. Ngubane, 2011 and Raman et al, 2013) but nobody references anything in support of this. The only vaguely relevant study I could find is by Grady et al (1986), discussing the fact that jugular compression increases cerebral venous pressure enough to prevent air embolism during brain surgery.
Though Woda et al (1996) found that the mere presence of a right IJ line does not by itself cause an increase in intracranial pressure, people still avoid using them in TBI patients mainly because there is a nonzero risk of thrombosis, and the thrombus can certainly make things worse.
The effect of high PEEP is to reduce the venous return to the heart, and venous congestion of all organs results (the brain included). This only has a modest effect on ICP. McGuire et al (1997) found that in patients with normal ICP, a PEEP of 5 or so has absolutely no effect on intracranial pressure, but in patients with a raised ICP the increase of PEEP to 10 and 15 cm H2O resulted in a small rise of ICP (by about 1-2 mmHg).
The best single reference for this is the 2011 paper by Roberts et al ( "Sedation for critically ill adults with severe traumatic brain injury: A systematic review of randomized controlled trials" ). In summary, in 2011 there was no convincing evidence that one sedative agent is more efficacious than another.
People have been using propofol for ICP control ever since it first became available. The first studies of its effects on ICP seem to come from the late eighties (eg. Hartung et al, 1987), where its safety as an induction agent was investigated in TBI patients, using doses of around 1mg/kg. Vandesteen et al (1988) demonstrated that it has salutary effects on the most important parameters: "Cerebral blood flow decreased by 27.6% and cerebral vascular resistance by 51%.... Cerebral oxygen consumption decreased by 18.25%." This effect was recorded at a blood concentration of 4.06μg/ml, corresponding to fairly deep sedation (volunteers report waking from anaesthesia at a measured concentration of around 2 μg/ml; on induction of an un-premedicated elective case the dose tends to get into the 5-6 μg/ml range). Its effects on intraventricular ICP was first reported upon in 1993 by Bendriss et al, who were happy with its effects in elective VP shunt patients.
How good is it compared to other agents? This is uncertain. Much of what we know about its use in TBI has been extrapolated from non-TBI studies. For example: Kikuta et al (2006) found that it decreased intracranial pressure when compared to an alternative agent. However, that was a study of elective patients with Moyamoya disease. And the "alternative agent" was sevoflurane, a notorious cerebral vasodilator. Something similar was done by Petersen et al in 2003, among patients with brain tumours. Conclusion? Pointless, as far as determining the "bestest-ever sedative for TBI". But, those people did confirmed that there is a dose-related decrease in CMRO2 and ICP. Prior to this, we had only small case series to support its use in TBI (eg. Herregods et al, 1988). At least its better than the cat studies we used to determine our ideal MAP target.
The objectives of using opiate analgesia in traumatic brain injury are:
It is believed that some analgesia is mandatory. How much of it? That is still a topic of debate. Some things are becoming clear, however.
For one, continuous infusions are better than boluses. An excellent recent study by Welch et al (2016) found that among the paediatric population boluses of fentanyl (and surprisingly midazolam) not only fail to reduce the ICP, but are actually associated with increases in intracranial pressure. We are talking about big boluses, 15-20mg morphine or 150-200μcg fentanyl. Opiate boluses also seem to decrease the cerebral response to hypertension by impairing cerebral autoregulation. In 2000 de Nadal et al performed a study comparing them to one another among a group of severe TBI patients. The drugs were given as boluses, 0.2mg/kg of morphine and 2 μcg/kg of fentanyl (so, 10-15 mg of morphine and say 150-200 μcg of fentanyl). Increased intracranial pressure was observed with both.
Locally, morphine and fentanyl are available as options for TBI patients. And among the locals there exists a range of (possibly) informed expert opinions regarding the supremacy of one opiate over another when it comes to it's effect on the ICP. There is actually surprisingly little evidence to support this difference in opinion. For instance, Lauer et al (1997) randomised a small (n=15) group of severe TBI patients to either receive fentanyl sufentanil or morphine as a four-hour infusion, during which vital signs were monitored .In case one is wondering, the doses they used were around 1.6 μcg/kg /hr of fentanyl, 0.03mg/kg/hr of morphine and 0.33 μcg/kg/hr of sufentanil. To round these doses up into recognisable infusion rates, that corresponds to about 100mcg/hr of fentanyl, 2mg/hr of morphine and 25mcg/hr of sufentanil.
The results? ICP did not change from baseline in any group. The fentanyl patients had a slightly lower ICP than the others during a few points along the sampling timeframe, but otherwise there was no clear eyebrow-raising advantage associated with any of the drugs used here. Of course, with such a small sample, the signal may have been lost in the noise (see the above-indented graph I stole from Lauer et al).
In summary, opiate infusions and boluses probably do nothing for ICP, unless the patient is actually experiencing opiate-responsive pain; and large boluses of opiates might actually increase the ICP. Plus, all opiates are more or less the same as far their effect on ICP is concerned.
The use of something like midazolam as a second agent has been routine in brain injury. People tend to add it to propofol as propofol-sparing agent, when the ICP is not very well controlled and the patient becomes haemodynamically unstable from the constant propofol boluses. An excellent review of sedation in TBI (Roberts et al, 2011) could not examine this practice directly, but was able to turn up three trials comparing midazolam to propofol. There was no mortality difference (as one would expect - it depends on other things). More surprisingly, midazolam had no impact on the duration of ICU stay. The review concluded that there was insufficient evidence to recommend midazolam over propofol or vice versa.
Apart from the findings of these small scale trials, midazolam has several known disadvantages. It is relatively long acting, and has little effect on CMRO2 (certainly less than propofol). Nugent et al (1982) were the early pioneers of dog-based midazolam research. This group found that at low (safe) doses of 0.2mg/kg the midazolam infusion decreased cerebral blood flow but left cerebral oxygen consumption unaffected. For the normal 70 kg adult, that comes to 14mg (or 20-28mg for the normal Western Sydney resident). Not satisfied with this, Nugent's team then went on to use truly ridiculous doses, ranging all the way up to 10 mg/kg. That would be a whole-human dose of 700mg of midazolam. This absurd dose did result in a decrease of cerebral metabolism, but only by 45% (making one think that it might have been the effect of some excipient rather than of midazolam itself). Judging from these dog studies, one must administer midazolam in such doses that would give rise to hilarious disciplinary action against the prescriber. In 1993, Papazian et al confirmed that it has only a modest effect on raised ICP among a group of severe head injury patients (in doses of 0.15mg/kg, corresponding to 10-15mg boluses).
The primary exam candidates from years gone by will recall the conventional wisdom, that ketamine should never be used for induction of patients with closed head injury because it can increase intracranial pressure. This folklore actually comes from the distant dawn age of intensive care. In 1972, Shapiro et al published a paper describing an increase in ICP in response to ketamine boluses. The graphed time trend of ICP is reproduced below, without any permission whatsoever:
Five patients with various causes of raised ICP (mainly hydrocephalus) were anaethetised with 2mg/kg of ketamine. The mean ICP increase was actually quite massive: the authors report an increase from a mean of 13 mmHg to a mean of 63mmHg. The nightmarish ICP spike came right ater the ketamine bolus, and required thiopentone to smoothe out. This article (and several others like it) generated a mass migration of critical care specialsists away from ketamine. Whole generations grew up thinking that ketamine is an evil brain-swelling death drug, and should be avoided in head injury patients.
This myth has taken decades to dismiss. Good examples of thorough ketamine studies debunking its bad reputation can be found in the nineties and noughties. For instance, Albanese et al (1997) studied eight TBI patients undergoing ICP monitoring with a propofol co-infusion (3mg/kg/hr, or about 20ml/hr as is the standard for TBI locally). Three different bolus doses of ketamine were explored: 1.5mg/kg, 3mg/kg and a whopping 5mg/kg (350mg of ketamine for the 70kg adult). All doses studied resulted in a significant decrease in ICP; EEG demonstrated burst suppression with the higher doses. The ICP decrease was modest (it decreased by 5mmHg, from a mean ICP of 14 mmHg down to a mean of 9 mmHg) but this had served to demonstrated an important point.
More recently, Zeiler et al (2014) performed a systematic review of the uses of ketamine in TBI. In total,m they were able to scrape together 101 adult patients and 55 paediatric patients. In none of the studies was there any ICP increase associated with the use of ketamine. No adverse events were seen, and some of the authors reported substantial increases in MAP and CPP, with the resulting decrease in vasopressor requirements. Zeiler et al did not look at outcomes, because they were generally very poorly documented. The 2015 review by Cohen et al looked at a larger selection of studies, and found that ketamine had no effect on neurological outcomes, length of ICU stay or mortality. No sustained effect on ICP was observed (of the ten studies, two reported an increase and two reported a decrease).
In summary, ketamine is no longer contraindicated in severe traumatic brain injury (level 2b evidence), but it is also unlikely to become a miraculous rescue therapy.
In short, there are no guidelines offered for this issue. Brain Trauma Foundation has nothing to say about this topic. Neuromuscular blockade is used routinely in the ICU to control intracranial pressure, and it probably helps to some extent for patients who are cooled (and shivering) or patients who are under-sedated (and coughing). It is supposed to decrease the spikes of ICP experienced with regular tracheal suctioning.
However, nobody seems to agree whether continuous infusions of NMJ blockers actually improve outcomes. It certainly decreases whole-body metabolic demand, but that really doesn't answer the question. Certainly, prolonged used of paralysing toxins tends to lead to a whole host of extracranial complications, and there is some evidence that routine use in all TBI patients does not influence outcome.
And of course it raises the question: if you have enough opiates on board that the patient is plainly unresponsive to all forms of stimuli, then do you really need a neuromuscular junction blocker? Kerr et al (1998) discovered that yes, in fact you do. In their study, three groups of patients were randomised to opiates, no opiates, and opiates plus vecuronium. Then, aggressive suctioning was performed by cackling villains. The ICP spikes of the vecuronium group were significantly smaller as compared to the opiates-only group.
This is a mindlessly simple manoeuvre which accomplishes much good, with little harm to follow. To extract some CSF from the skull will decrease intracranial pressure by exploiting the Monro-Kellie doctrine. Less material inside the skull means less pressure inside the skull. The act of trying to drain the EVD will yield useful information. If it fails to drain, obviously its kinked or blocked, and this could be the problem. If it drains and fresh blood comes forth through it, perhaps another CT brain is in order.
The best answer is "some". If you open the EVD and no CSF comes out, the EVD is in trouble and you should get on the phone to the neurosurgeons, yelling angrily about anterior horns and such. The EVD is not draining and this is a serious problem.
Of course, if it is draining, there is hope. The daily rate of CSF production varies, but most authorities will say that the daily volume is about 550ml. The rate is not influenced by the intracranial pressure. Thus, you should expect 20-24ml to drain per hour.
Additionally, the brain-injured person will have greatly reduced intracranial compliance, which means that even a few ml of CSF will result in a profound decrease in ICP. Thus, you should not expect a gush of fluid.
So. 20-24ml per hour is being produced. Let us say that during that hour, you intermittently open the EVD to drain some CSF. The closer you approach that 20-24ml volume, the more it means the CSF is not being reabsorbed by conventional means; i.e. there is either some barrier to its normal circulation, or to its normal resorption. In short, anything approaching 20ml/hr of CSF drainage through the EVD is cause for concern, and typically triggers a repeat CT brain.
There is a surprising amount of variation in the local practice. Some centres will have very little EVD opening, whereas others will let the CSF out at regular intervals, or whenever the ICP rises beyond a certain threshold. There is some risk involved. The EVD is in one lateral ventricle, and if too much CSF is drained from it the midline shift may get worse. In spite of such concerns, the 2016 BTF Guidelines recommend continuous drainage, if you are going to drain it (instead of intermittently opening the EVD). This seems to be based on a 2014 study by Nwachuku et al where patients had oth an EVD and a Codmans parenchymal monitor. The authors found that the continuously open EVD group did not have as many pressure spikes (which makes sense). Neurological outcome or mortality were not explored by this group.
This topic is large enough to merit its own page. Indeed, there are all sorts of words written about which hypertonic glop to infuse, how fast, when, and for how long.
Locally, hypertonic saline infusions are commonly seen as a means of maintaining a supra-physiological serum sodium level, because an impression has formed among the critical care public that being hypernatremic is somehow good for the brain. This impression probably formed on the basis of studies which have associated hyponatremia with increased mortality among TBI patients. Low sodium bad, therefore high sodium good?
Where did this association of hyponatremia and increased TBI mortality come from? Usually, a number like "7-60% increase in mortality" is quoted. This figure is propagated through such review articles as Diringer et al (2006) and Tisdall et al (2006). Those authors in turn get their mortality statistics from a 1997 review article by Fried and Palevsky. They, in turn, quote Andersen et al (1985) and Tierney et al (1986). Chasing these references has revealed that both were observational cohorts of nonspecific hospital inpatients - not ICU, and not brain trauma. Andersen et al found a 60-fold increase in mortality, not a 60 percent increase (that's an important mistake to note, attributable to Chinese whispers). Tierney et al reported on retrospective data arranged into case-control pairs, and found that after confounders were removed the increase in mortality was only seven-fold. This seems to be the origin of those numbers.
To confuse the issue further, hypernatremia has also been associated with increased mortality among TBI patients. For example, Maggiore et al in 2009 found that hypernatremia in a mean range of 148-152 mmol/L was associated with an increased risk of death by a hazard ratio of 3.00, although some of that group had diabetes insipidus (that study is discussed in greater depth in the chapter on polyuria following traumatic brain injury). Other authors have found that only a sodium consistently greater than 160 mmol/L was associated with increased mortality (Aiyagari et al, 2006). This should not give one the impression that a sodium of 159 mmol/L was relatively safe. Hypernatremic patients in that retrospective cohort also spent longer in ICU, in hospital, on a ventilator and had a lower median admission GCS (8 vs. 14).
None of this gibberish answers the question, "What should my patient's sodium be?". There have been several studies to answer this question, well summarised in the "discussion" section of the most recent paper by Wells et al (2012). The authors studied their local population of head injury patients, in an institution where sodium levels of 145-155mmol/L were targeted as a part of their routine practice, using hypertonic saline. Wells and colleagues challenged the hypothesis that the pursuit of abnormally high serum sodium is somehow beneficial, even in terms of intracranial pressure. Eighty-one TBI patients were observed. Their serum sodium values did not correlate with their ICP behaviour: having a high sodium did not protect them from ICP spikes, nor did it decrease the maximum ICP during those spikes, nor was there any "sparing" effect on the number of ICP-controlling interventions.
On balance, the data seem to point to a lack of effect. "Two studies suggested that increasing the serum sodium improved ICP control (total n = 58) and two did not (total n = 113). The fifth study suggested that continuous infusion of HTS may be detrimental." Most recently, Griesdale et al (2016) were at least able to demonstrate safety: among their severely brain-injured cohort of 231 retrospective cases, there was no mortality increase asosciated with a mildly raised sodium (mean around 146 mmol/L). The patients receiving hypertonic saline appeared to have decreased intracranial pressure, but none of the other ICP management interventions were reported or corrected for, and the ICP measurements used by the investigators were once daily measurements. An excellent takedown of targeted hypernatremia comes from the same issue of the Canadian Journal of Anaesthesia, in the editorial by Bonaventure et al.
So: probably best to keep it around 145mmol/L, where it's still normal-ish.
So, the patient is properly positioned, full of sedatives and completely paralysed, with the EVD getting drained every fifteen minutes, and the ICP is still crap. What's your move?
Barbiturates used to be the standard choice for sedation in the dark ages of critical care. In 1939, Browder and Meyers suggested that barbiturate sedation be considered alongside chloral hydrate bromides and paraldehyde as one of the "palliative, supportive and hygienic measures" used once a surgically amenable lesion has been excluded.
The 2007 iteration of the Brain Trauma Foundation recommends titrating one's barbiturate therapy to burst suppression on EEG. Its place is in a situation when all surgical and medical interventions have failed to control the ICP, i.e. where the term "super-refractory raised ICP" is appropriate. How often is this required in the modern world? Majdan et al (2013) reviewed the practice of five European countries and found that fewer than 20% of TBI patients required barbiturate coma, and in only 6% did the intensivist resort to a high dose infusion. It is seen as an effective strategy: in 69% of cases the ICP decreased to a satisfactory degree.
Of course, no outcome difference was found by Majdan's team (whether functionally or in terms of survival). Barbiturate coma only seems to improve the chances of an "acceptable outcome" in the paediatric population (Mellion et al, 2013). A Cochrane review of adult data (Roberts et al, 2012) found no mortality benefit, and no difference in neurological performance in the long term. Sure, the ICP was well controlled, but one in four patients developed clinically significant hypotension.
Another disadvantage of barbiturate therapy for raised ICP is the well-known effect of barbiturates on the immune system. Bronchard et al (2004) found that barbiturate coma was associated with an increased risk of VAP (a rate of 42%, vs 12% of non-barbiturated patients). This finding was reproduced by post-hoc analysis of the Corti-TC and SPIRIT trials (Lasocki et al, 2015) - the OR was around 5.6.
In summary, barbiturates will control your ICP, but at the cost of infectious complications and without much of a long-term functional benefit or mortality improvement.
Because The DECRA trial has not demonstrated any improvement in outcome (in fact, long term outcome turned out to be worse) this measure is controversial. The issue is discussed at greater length in other chapters:
In short, decompressive craniectomy does not influence mortality in TBI, and the survivors still tend to be severely disabled (there is some thought that axonal stretching and deformation of the brain architecture is responsible). Many neurosurgeons are reluctant to proceed with it, because of the impression that they would be rescuing a severely disabled person, and that this may not benefit either that person or society in a broader sense. Weirdly, an economic evaluation of decompressive craniectomy in the US (Alali et al, 2014) has revealed that it is in fact cheaper (at least among young patients) than barbiturate coma.
In summary, the present trend is to steer away from decompressive craniectomy. However, it must be pointed out that trials such as DECRA involved "primary" decompression, that is to say decompression early in the course of therapy. More data is awaited with the publication of the RESCUEicp trial, where decompressive craniectomy is treated as a last-ditch measure, and is compared to maximal medical therapy as represented by barbiturate coma.
Like barbiturates, hypothermia was in vogue in the distant past. However, a 2009 Cochrane review of hypothermia in traumatic brain injury recommended nothing. It turned out that hypothermia looked like a better treatment only in those trials which were poorly designed. Everyone agrees that fever is bad, and you should at least maintain normothermia, but hypothermia has not been the neuroprotective magic bullet everybody thought it was.
The recent Eurotherm 3235 trial (Andrews et al, 2015) found no mortality or functional outcome improvement with indiscriminate use of hypothermia. However, some might argue that the "second tier" is not where this therapy belongs. In a critical response, many pointed out that this trial provided evidence against therapeutic hypothermia as an early measure to lower intracranial pressure, and not against its use for refractory raised ICP. It certainly was effective in decreasing the ICP in Eurotherm, and as a last-ditch rescue therapy it may still have a role to play, so their argument went. The authors replied:
"We see no plausible explanation why waiting until all other therapies have failed to control intracranial pressure would result in benefit from therapeutic hypothermia, when the pathophysiology of brain swelling is similar."
In summary, the use of hypothermia will decrease your ICP, but there may be little improvement in the outcome, and there will certainly be complications (depending on how low you go).
Increased intracranial pressure associated with cerebral trauma responds to THAM (it is at least as effective as 20% mannitol). However, the effect - apart for lowering CSF CO2 - is likely the same as for mannitol, i.e. osmotic diuresis. This therapy was very hot in the 1970s. Specific unique indications may exist outside of the realm of traumatic brain injury. For example, Akioka et al (1976) found that it reverses the increase in ICP due to hypercapnea. Promising early trials (eg. Wolf et al, 1992) likely do not represent true effect because they were performed in an era which pre-dates modern TBI management, a time when patients would be routinely hyperventilated for prolonged periods. The most recent entry into this topic was made by Zeiler et al, who performed a systematic review THAM in 2014 and gave it a lukewarm recommendation ("may be useful", they said).
Routinely used to treat migraines, this ergot alkaloid is attractive by its effect on the ceberal circulation. Asgeirsson et al (1995) performed a series of clever cerebral blood flow studies and concluded that its effect on reducing ICP was mainly due to the constriction of large cerebral veins. Unfortunately, this vasoconstrictor substance is also prone to causing renal failure and gangrene (Gupta et al, 1996) and so its popularity has waned somewhat.
Indomethacin is a cerebral vasoconstrictor, among other things. Slavik et al (1999) found it reduced the cerebral blood flow of healthy volunteers by 26-40%. Boluses of 30-50mg reduced ICP by 37%-52%, and increased CPP by 14%. Sadler et al performed a systematic review of this practice in 2014. In a total population of 177 patients, "all but one study documented a decrease in ICP with indomethacin administration, with both bolus and continuous infusions". Most of the studies failed to focus on patient outcome, and on recorded a troubling reduction in SjvO2, suggesting that reducing cerebral blood flow may not be such a good thing (surprise). "This compound should be considered experimental", the authors concluded.
The CRASH trial investigators have demonstrated that steroids in brain injury decrease both short-term and long-term survival. There is no role for steroids here. In fact, this is the only BTF recommendation which is supported by Class I evidence. So, forget steroids. They have an established role to play in managing the raised ICP in the context of malignancy, and are mentioned in this chapter for that reason.