Treating delirium tremens: Pharmaco*kinetic engineering with diazepam and phenobarbital (2024)

[PLEASE NOTE: For the most complete & updated material on alcohol withdrawal, please see the Internet Book of Critical Care Chapter on this topic here]
0
Introduction

Recently the New England Journal published a review article about delirium tremens which is somewhat misguided (see a scathing critique by The Poison Review). The article focused on traditional benzodiazepine therapy, overlooking recent evidence about phenobarbital. This post will explore how phenobarbital might fit into the treatment regimen for delirium tremens.

Neuropharmacology of alcohol withdrawal, phenobarbital and diazepam

Chronic alcoholism causes down-regulation of inhibitory GABA receptors and up-regulation of excitatory NMDA-subtype glutamate receptors. Thus, alcohol withdrawal is due to inadequate inhibitory GABA activity combined with excessive excitatory glutamate activity.

Phenobarbital and diazepam both bind to the GABA receptor in different locations. Benzodiazepines increase the frequency of GABA-receptor channel opening, whereas phenobarbital increases the duration of channel opening. Therefore, these drugs may act synergistically.

In addition to affecting the GABA receptor, phenobarbital also inhibits glutamate receptors. By simultaneously enhancing GABA activity and inhibiting glutamate activity, phenobarbital perfectly reverses the pathophysiology of alcohol withdrawal. This could help explain why some patients fail to respond to benzodiazepines but subsequently respond well to phenobarbital.

Pharmaco*kinetics

Alcohol withdrawal may be conceptualized as a dynamic sedative deficiency over time (image below). The challenge in treatment is to provide exactly enough sedatives to match this deficiency. Under-treatment exposes the patient to risks of seizure or agitation, whereas over-treatment increases the risk of aspiration and respiratory failure. A sedative regimen must be designed which is both flexible and powerful enough to precisely meet the patient’s sedative requirements. The architecture of this regimen requires careful attention to pharmaco*kinetics. For example, a common mistake when treating alcohol withdrawal is stacking doses of lorazepam: lorazepam may require 20 minutes to take effect, so repeating doses more frequently causes multiple doses to be administered before the first dose takes effect.

Intravenous phenobarbital takes effect over 20-30 minutes, and therefore can be given every thirty minutes without causing dose-stacking (Young 1987, Duby 2014, Gold 2010). Subsequently, it has an enormous half-life of about 4 days. This long half-life is helpful because it allows the drug to auto-taper, decreasing risk of rebound symptoms.

Although phenobarbital is more powerful, diazepam is more titratable. When given intravenously diazepam has maximal effect within 5 minutes, allowing frequent doses to up-titrate the diazepam level without a risk of stacking doses. Like phenobarbital, diazepam has a long terminal half-life allowing it to auto-taper.

IV diazepam’s rapid-onset and long half-life make it the benzodiazepine choice of champions (Goldfrank's 10th edition, Gold 2007). Lorazepam takes longer to up-titrate safely given slower duration of onset. Subsequently, lorazepam's intermediate half-life leads to problems with rebound withdrawal. Lorazepam is often used as a continuous infusion in severe withdrawal, which is fraught with hazard. Lorazepam's slow onset and intermediate half-life (14 hours) cause it to accumulate after up-titration of the infusion, risking over-sedation (it makes little pharmaco*kinetic sense to give a drug with a half-life of 14 hours as a continuous infusion). Lorazepam infusions also carry a high risk of propylene glycol intoxication (more on this below). The only possible role for lorazepam may be in patients with advanced cirrhosis, who have difficulty metabolizing diazepam (for more information see discussion by Bryan Hayes).

Is phenobarbital or diazepam better?

Benzodiazepines have conventionally been regarded as the drug of choice dating back to a study by Kaim 1969 on the prevention of delirium tremens which compared benzodiazepine (chlordiazepoxide), antipsychotics (chlorpromiazine), antihistamines (hydroxyzine), and thiamine. Of course, this amounts to comparing benzodiazepines to three straw men. There is little data comparing benzodiazepines to barbiturates. Some centers in the US and especially Europe have had success using phenobarbital instead of benzodiazepine.

It is clear that either benzodiazepines or phenobarbital may be successful in treating most patients with alcohol withdrawal. However, phenobarbital appears to be more potent, given that there is a subset of patients who fail to respond to benzodiazepines yet respond well to phenobarbital. As discussed above, this may reflect that phenobarbital more fully treats the underlying neurotransmitter abnormalities in alcohol withdrawal. Therefore, if you were stranded on a desert island and needed one drug to treat alcohol withdrawal, IV phenobarbital might be better.

It is possible that a carefully designed combinationof these drugs could be the best approach. As discussed above, the two agents may have synergistic activity on GABA receptors. Phenobarbital is more powerful, while diazepam is more easily titratable: perhaps a combination of the two drugs could take advantage of both properties.

What is the best treatment target?

When designing any treatment algorithm, it is essential to chose the goal of therapy wisely. The success of recent investigations by Gold 2007 and Duby 2014 has been attributed to their utilization of phenobarbital. However, its entirely possible that clever selection of treatment goal was also instrumental to their success. These investigators used standard ICU sedation scales, targeting either a score of 0 to -2 on the Richmond Agitation Sedation Score (Gold 2007) or a score of 3-4 on the Riker Sedation Analgesia Scale (Duby 2014; see tables below). Either way, this amounts to stopping medication when the patient has reached an alert and calm state. Although traditional treatment has targeted drowsiness or sleep, stopping medication as soon as the patient is asymptomatic provides a greater margin of safety to avoid oversedation. By targeting a calm/awake state, you would have to really overshoot your goal before any serious consequences occurred (e.g., aspiration, loss of airway). More complicated alcohol withdrawal scales (e.g. the CIWA scale) are labor-intensive, excessively complicated, and confounded by many factors (Duby 2014).

Previous literature has recommended targeting a normal heart rate as well. However, there are many reasons for a patient to be tachycardic (e.g., hypovolemia, sepsis, heart failure, etc.). Assuming that all tachycardia is due to alcohol withdrawal is a dangerous assumption. For example, in one case at Genius General a patient with tachycardia due to unrecognized alcoholic cardiomyopathy was over-treated due to persistent tachycardia leading to intubation.

Treatment algorithm for delirium tremens

(a) Load the patient with 10 mg/kg phenobarbital IV over 30 minutes.

Rosanson 2012randomized 102 patients withalcohol withdrawal to receive either placebo or 10 mg/kg phenobarbital intravenously followed by usual care with PRN lorazepam. Patients treated with phenobarbital were less likely to require ICU admission (8% vs. 25%), continuous lorazepam infusion (4% vs. 31%), or restraints (29% vs. 45%). No complications from phenobarbital were noted. The reason that they chose a 10 mg/kg dose was simply that this was the highest dose that could be approved by their institutional IRB.

10 mg/kg phenobarbital may seem like a high dose, but it is actually a moderate and very safe dose when given up-front before the patient has benzodiazepine on board. There is a predictable linear relationship between phenobarbital dose and serum concentration (figure below). Tangmose 2010 and Young 1987 studied phenobarbital levels in patients undergoing withdrawal, and both papers agree that a dose of 10 mg/kg will yield a plasma level near 16 ug/ml, near the lower end of the therapeutic range (15-40 ug/ml).

Relationship between cumulative phenobarbital dose and plasma phenobarbital concentration among patients treated for alcohol withdrawal (Tangmose 2010).We have addedgreen linesindicating the plasma therapeutic range for phenobarbital (64-172 micromol/L = 15-40 ug/ml), anorange lineindicating the level at which mild signs of toxicity are usually noted such as ataxia and nystagmus (225 micromol/L = 50 ug/ml), and ared lineindicating the lowest level which has been associated with stupor or coma (>280 micromol/L = 65 ug/ml)(Lee 2013).

Please note that the “therapeutic range” of the phenobarbital level is defined with regards to use of phenobarbital for seizure. When used for delirium tremens, the phenobarbital is titrated to clinical effect, and a given patient could require doses above this range. However, the concept of therapeutic range is useful because it is established that phenobarbital levels in this range are safe and rarely cause respiratory depression by itself (Hayner 2009). Furthermore, due to cross-tolerance between alcohol and phenobarbital, an alcoholic suffering from withdrawal should be resistant to phenobarbital intoxication. This explains why it is safe to give a 10 mg/kg bolus of phenobarbital at the beginning of therapy for alcohol withdrawal. Indeed, a loading dose of 15 mg/kg was used by Ives 1991, but there isn't yet evidence to support the safety of giving this dose intravenously.

To provide additional context regarding necessary doses of phenobarbital, Tangmose 2010 described administration of up to 3,000 mg of phenobarbital alone to 348 patients without a single episode of respiratory suppression (figure above). Michaelsen 2010reported the following doses when phenobarbital alone was used to treat alcohol withdrawal:

One advantage of loading with a fair amount of phenobarbital initially is that this will last the duration of the patient’s alcohol withdrawal symptoms, and should reduce the sedative requirement throughout the patient’s entire course (image below). The safety of using 10 mg/kg phenobarbital by itself has been established. However, in the presence of other sedatives such as benzodiazepines, the dose-response relationship for phenobarbital is no longer reliable. Therefore, once the patient has been loaded with benzodiazepines, the opportunity to safely give 10 mg/kg phenobarbital is lost. Thus it makes sense to take advantage of this one-time opportunity by immediately loading up-front with phenobarbital.

(b) Titrate IV diazepam to effect

The next step is to up-titrate IV diazepam. The target dose of diazepam to use at this point is not entirely clear. Since diazepam may synergize with phenobarbital, diazepam may be about twice as potent here as it would be alone (Rosanson 2012). Therefore, it may be safer to ramp up the dose of diazepam a bit slower than usual. The great majority of patients will likely respond at this point.

(c) Additional phenobarbital

There is a rare sub-population of patients who are refractory to large doses of benzodiazepines, but subsequently respond well to phenobarbital. For example, Hjermo 2010 described eight patients who failed to respond to large doses of diazepam (410-2,540 mg) who all responded well to administration of phenobarbital (906 +/- 430 mg). As discussed above, 10 mg/kg is actually not a lot of phenobarbital. Some patients require more.

Barbiturates are arguably the most potent neurosupressive medications available. Barbiturates can terminate nearly any seizure, even seizures refractory to enormous doses of propofol, benzodiazepines, and numerous other antiepileptics. Barbiturate overdose can electrically silence the brain and convincingly mimic brain death. Therefore, we question the concept that its possible for delirium tremens to “fail” to respond to phenobarbital. More likely, we are failing to use an adequate dosage of phenobarbital. As shown in figures earlier, some patients may require thousands of milligrams. Using a conventional approach of diazepam up-titration followed by phenobarbital (total phenobarbital dose 25%-75% quartile = 130-1430 mg), the primary reason for intubation was failure to control agitation (8/9 patients) rather than over-sedation (Gold 2007). This suggests that even with fairly aggressive algorithms involving diazepam followed by phenobarbital, the primary problem continues to be under-treatment rather than over-treatment.

Maintenance

The approach above describes an initial strategy for gaining control of a patient with acute delirium tremens and agitation. Medications used (diazepam and phenobarbital) have a long half-life allowing them to auto-taper without requiring scheduled re-dosing. Maintence therapy therefore consists of intermittent as-needed use of medications based on the agent and dose which were successful initially (i.e., PRN diazepam or phenobarbital).

Avoidance of propylene glycol toxicity

The New England Journal review article recommends high-dose benzodiazepines without mentioning that diazepam and lorazepam are formulated in propylene glycol, leading to a risk of propylene glycol toxicity at high doses. Propylene glycol toxicity may cause lactic acidosis, renal failure, and multiorgan failure mimicking sepsis. IV Lorazepam contains 80% propylene glycol, more propylene glycol than any other commonly used drug (Zar 2007). There is a linear relationship between the rate of infusion of lorazepam and the propylene glycol level, with potentially toxic levels often occurring at infusion rates above 10 mg/hr (figure below). Propylene glycol intoxication is less common with diazepam, although it has been reported in patients receiving >2,000 mg (Wilson 2005). Zar 2007 recommended avoiding more than 166 mg/day lorazepam (7 mg/hour infusion) or 832 mg/day diazepam. Adjunctive phenobarbital reduces the requirement for benzodiazepine, rendering propylene glycol intoxication less likely. Although phenobarbital is also formulated in propylene glycol, the quantity of propylene glycol administered with phenobarbital is minimal (Hayner 2009). Any treatment algorithm involving reasonable doses of diazepam and phenobarbital should avoid propylene glycol intoxication.

Propylene glycol concentration as a function of lorazepam infusion rate (Horinek 2009).We have added a line indicating 100 mg/dL propylene glycol, the level at which propylene glycol may potentially cause clinical deterioration (Zar 2007).

The New England Journal review article recommends a lorazepam infusion at 10-30 mg/hour. To support this recommendation it cited a protocol by DeCarolis 2007. However, DeCarolis included within their protocol structured down-titration of lorazepam infusion as well as daily monitoring of serum osmolality for early detection of propylene glycol intoxication. In the abstract of their paper, they cautioned that “the use of the protocol is effective but requires close monitoring to ensure protocol compliance and to avoid potential propylene glycol toxicity.” By omitting safeguards within the DeCarolis protocol, the New England Journal review article makes potentially dangerous recommendations.

Dexmedetomidine?

Dexmedetomidine's ability to provide titratable sedation without affecting airway reflexes could make it useful. However, dexmedetomidine doesn't address the underlying neurobiology of alcohol withdrawal, leading to a potential danger of masking symptoms without truly treating the disease. Unlike diazepam and phenobarbital, which are both supported by decades of experience, there is little evidence regarding dexmedetomidine. We await further studies prior to reaching any conclusion about dexmedetomidine.

Too early for phenobarbital-first approach? Or too late?

Some may argue that it is too early to move to a phenobarbital-first approach, given only one recent study on this by Rosanson 2012. However, there is another study regarding up-front loading with intramuscular phenobarbital 15 mg/kg by Ives 1991, describing a protocol which had been in use for 25 years at that point in time. Literally decades of experience exist regarding the use of phenobarbital monotherapy for alcohol withdrawal. Success combining phenobarbital and benzodiazepines has been reported in numerous studies, and is currently well accepted. The pharmacology of phenobarbital loading and the safety of various phenobarbital levels has been worked out in epilepsy over many decades. Therefore, it is our opinion that at this point there is sufficient clinical and physiological evidence to consider moving to a phenobarbital-first approach. According to Stephen King, “sooner or later, everything old is new again.” At this point, phenobarbital is arguably more old than new.

Take-home considerations
0

• Consider initiating treatment for alcohol withdrawal with 10 mg/kg phenobarbital infused over 30 minutes. This is proven to be safe. It may provide a nice baseline level of sedation, blunting the severity of the disease.
• Although phenobarbital is extremely effective, some patients may require very high doses. Rather than a patient “failing” to respond to phenobarbital, it is more likely that we are failing to provide an adequate dose of phenobarbital.
• Lorazepam infusions may cause oversedation (due to accumulation of lorazepam) or propylene glycol toxicity. IV bolus diazepam is probably the best benzodiazepine for management of alcohol withdrawal.
• Consider targeting treatment to achieve an awake and calm state. CIWA scales and vital signs may be confounded by a variety of other factors.

[PLEASE NOTE: This post has been updated with anew postdescribing our current approach to alcohol withdrawal. The material here is still correct, but it does not represent our current practice. Ifyou have time we would recommend reading this post first, and then reading our newer post.]

Coauthored with neurointensivist colleague and drinking buddy Ryan Clouser (@NeuroCritGuy).

Image credits:

–http://www.freeimages.com/photo/engine-vrooom-1-1546825

– RASS scale from http://www.resus.com.au/uncategorized/how-deeply-should-we-sedate-our-patients-post-intubation/

– SAS scale from http://ajcc.aacnjournals.org/content/12/4/343/T1.expansion