[PLEASE NOTE: For the most complete & updated material on alcohol withdrawal, please see the Internet Book of Critical Care Chapter on this topic here]
Over the past year, the five papers below were published regarding the use of phenobarbital in alcohol withdrawal. These studies aren’t massive, glittering multicenter RCTs (and, realistically, it’s dubious whether such a study will happen). However, these papers bring some unique and interesting perspectives to the table. This post will summarize them briefly – a bit like a sampler flight of beers (figure above).
I. Pre-post designs
The following studies involve changes in protocol over time, leading to different cohorts of patients being treated differently.
Tidwell WP et al. Treatment of alcohol withdrawal syndrome: Phenobarbital vs. CIWA-AR protocol.
This is a retrospective cohort study at Saint Thomas West Hospital in Nashville comparing patients treated with CIWA-triggered benzodiazepines (pre-intervention) versus patients treated after implementation of a phenobarbital protocol. The phenobarbital protocol involved fixed doses of phenobarbital supplemented with small doses of PRN lorazepam if needed:
In contrast, patients treated with benzodiazepines alone received the following treatment:
The baseline demographics of the two groups were well matched:
Outcomes were substantially better in the phenobarbital group (table below). In particular, only one patient in the phenobarbital group required intubation (and this intubation procedure was actually performed prior to the patient's receiving any phenobarbital). Patients in the phenobarbital group were less likely to require other agents (e.g. haloperidol or dexmedetomidine).
The doses of phenobarbital used here are on the low end compared to phenobarbital-monotherapy protocols. The efficacy of such low doses might be explained by synergy between phenobarbital and lorazepam.
It is often feared that barbiturates may have a greater risk of respiratory depression compared to benzodiazepines. This study argues for the opposite view: addition of phenobarbital (with a concomitant reduction in benzodiazepine doses) correlated with a reduced rate of intubation. This is consistent with my experience: benzodiazepines (and especially benzodiazepine-induced paradoxical reactions and benzodiazepine-induced delirium) may be responsible for precipitating the need for intubation.
The discussion section points out that oral phenobarbital is considerably less expensive than intravenous phenobarbital. Although intravenous phenobarbital does have pharmacokinetic advantages for the sickest patients, oral phenobarbital seems fine for most patients (especially those along a more mild-moderate spectrum of withdrawal).
Nelson AC et al. Benzodiazepines vs barbiturates for alcohol withdrawal: Analysis of three different treatment protocols.
This is a retrospective study at the University of Colorado performed during a benzodiazepine shortage. Three alcohol withdrawal protocols used during three different time periods were compared: one using only IV diazepam, one using both IV lorazepam and phenobarbital, and one only utilizing IV phenobarbital:
Initially there were 299-500 patients in each group. Subsequently, 100 patients were selected at random from within each group. Although this may have been necessary to allow performance of the study, it did limit statistical power. Patients were fairly well-matched at baseline, although there were fewer patients with severe withdrawal in the diazepam group:
Outcomes were generally similar (table below). Patients in the diazepam group were less likely to be admitted. Protocol violations occurred in 58% of patients in the phenobarbital group (explaining the use of benzodiazepine in these patients and relatively low phenobarbital doses).
The authors concluded that phenobarbital is safe and effective, with similar results compared to benzodiazepines.
II. Intercurrent protocol comparisons
The following studies involve a group of patients treated with various protocols at the same time. Choice of different protocols was left to practitioner discretion.
Nisavic M et al. Use of phenobarbital in alcohol withdrawal management – A retrospective comparison study of phenobarbital and benzodiazepines for acute alcohol withdrawal management in general medical patients.
This is a retrospective study comparing patients treated with a benzodiazepine protocol versus a phenobarbital-monotherapy protocol at Massachusetts General Hospital between 2007-2011. The management strategy was chosen by treating clinicians. Most patients were treated with benzodiazepines (419 versus 143). Patients treated with phenobarbital were considerably higher risk (with a higher rate of current seizure, prior seizure, and delirium tremens):
Despite being higher risk, patients treated with phenobarbital had very similar outcomes (with some positive trends towards less delirium and lower rates of leaving against medical advice):
Among patients initially treated with benzodiazepines, 16 were benzodiazepine-refractory and required treatment with the phenobarbital protocol. These patients responded to phenobarbital.
Overall, phenobarbital monotherapy performed well in this study. Patients treated with phenobarbital were sicker, yet their outcomes were equivalent (perhaps a wee bit better). Phenobarbital monotherapy appears uniformly effective, whereas some patients failed benzodiazepine therapy and required phenobarbital.
The protocol these authors used is interesting (figure above). Patients were initially loaded with 6-15 mg/kg in divided doses over several hours. The loading dose was determined based on an assessment of risk of alcohol withdrawal versus the risk of respiratory compromise. Subsequently, patients received small maintenance doses for up to a week. The maintenance dose was initially calculated to maintain a steady drug level, with tapering off over time. Thus, the doses of phenobarbital were largely fixed. Residual symptoms seem to have been treated with other drugs if necessary (e.g. haloperidol).
This study thus validates the concept that after giving a large dose of phenobarbital, it’s safe to treat residual symptoms with non-GABA-ergic medications (e.g. haloperidol). Limiting phenobarbital doses and transitioning to non-GABA agents after a certain point is an important component of many phenobarbital protocols.
Sullivan SM et al. Comparison of phenobarbital-adjunct versus benzodiazepine-only approach for alcohol withdrawal syndrome in the emergency department.
This is a retrospective study performed in two academic emergency departments associated with the University of Arizona. Patients were categorized into groups depending on whether or not they had received intravenous phenobarbital (with or without benzodiazepines) in the emergency department. No specific protocol had been in place to guide phenobarbital use.
Initially, 233 patients were identified who had received phenobarbital versus 3553 who had not. Similar to the Nelson et al study above, these groups were randomly reduced in size to ~100 patients each (to facilitate feasible performance of the study). Baseline variables are shown below. As in the Nissavic study above, phenobarbital was preferentially used in patients presenting with seizure history or delirium tremens.
The median cumulative dose of phenobarbital in the emergency department was 260 mg (with an interquartile range of 130-500 mg). 81% of patients in the phenobarbital group also received benzodiazepines.
Patient outcomes were generally similar (table below). Patients in the phenobarbital level had less severe symptoms after 24 hours and they had a shorter hospital length of stay:
The authors of this study concluded that phenobarbital didn’t work (see their abstract reproduced above). There are two main problems with this conclusion:
- The study wasn’t randomized: phenobarbital was administered in a cohort of sicker patients. Similar to the Nissavic study, phenobarbital was applied to a sicker group of patients and achieved somewhat better overall outcomes. In total, this may indirectly imply that the phenobarbital is working.
- The dose of phenobarbital used in the emergency department was small (median of 260 mg). Patients in both groups were treated mostly with benzodiazepines, so it’s unrealistic to expect a huge difference between groups (this is a problem with many of the above studies).
III. Other study designs
Waldee E and Phan SV. Evaluating the use of phenobarbital for the management of alcohol withdrawal in psychiatric inpatients.
This is a retrospective, observational study describing the use of phenobarbital monotherapy to prevent or treat alcohol withdrawal among 122 psychiatric inpatients at the University of Georgia.
Baseline properties of the patients are shown above. Most patients (70%) had symptoms of alcohol withdrawal. These are severe alcoholics who were often inebriated initially. Their history of alcoholism was impressive, with an average of 17 drinks per day and 25 years of drinking history. Before going further, pause and imagine what would happen if this cohort of patients were admitted to a medicine ward and treated with benzodiazepines. At least some would probably fail treatment and require ICU transfer.
Patients were treated with oral +/- intramuscular phenobarbital. The average cumulative dose was 421 mg +/- 288 mg. Ten patients received additional treatments (including clonidine, diazepam, or lorazepam).
The outcomes are astonishingly good. No patients required transfer out of the psychiatric unit (e.g. to the medicine unit or ICU). Among patients with symptoms of alcohol withdrawal, nearly all resolved. Among patients treated prophylactically to prevent alcohol withdrawal, 34/36 developed no symptoms of withdrawal. 15 patients did experience some sedation which was self-limiting.
The primary endpoint was safety. The authors demonstrated that phenobarbital had no appreciable effect on respiratory rate (table above). The demonstration that phenobarbital can be safely administered in a psychiatric unit implies that this intervention should also be quite safe in other settings as well (e.g. hospital ward or ICU). However, this shouldn’t be overly surprising, as the use of phenobarbital in psychiatric units has already been demonstrated in prior studies (20682131).
Comparing this study to the other studies discussed herein provides some interesting contrasts. Why did no patients require intubation in this study? It’s probably because they were pre-screened before admission to psychiatry, thereby eliminating other active medical or surgical problems (e.g. trauma, pneumonia). This emphasizes that when we read some studies that involve a 5-10% intubation rate, these intubations are potentially due to other problems (e.g. pneumonia) – not necessarily alcohol withdrawal.
- Phenobarbital use appeared to produce similar or perhaps better results, compared to benzodiazepine therapy alone.
- Phenobarbital’s efficacy may be limited by suboptimal dosing. Despite many studies being motivated by Rosenson’s RCT (which used front-loaded 10 mg/kg phenobarbital), in practice most studies used considerably lower doses (22999778).
- These studies refute the concept that phenobarbital increases the risk of intubation compared to benzodiazepines. In studies comparing phenobarbital versus benzodiazepines, rates of intubation were similar or lower with phenobarbital. One study utilizing phenobarbital monotherapy achieved an intubation rate of zero among 122 patients.
- Overall, studies utilizing explicit guidelines to administer phenobarbital seemed to have greater success (Tidwell et al. and Nisavic et al.). This could be explained by guidelines' facilitating clinician comfort with the administration of higher cumulative doses of phenobarbital.
- Phenobarbital for EtOH w/d #1 (12/14)
- Phenobarbital for EtOH #2 (monotherapy, 10/15)
- Phenobarbital for EtOH #3 (monotherapy reloaded, 12/17)
- Phenobarbital for EtOH #4 (one order to sedate them all, 10/18)
- Coming in three days: IBCC chapter on EtOH withdrawal.
- [PLEASE NOTE: For the most complete & updated material on alcohol withdrawal, please see the Internet Book of Critical Care Chapter on this topic here]
PubMed is down currently (again), so I will need to add in the references when PubMed is back up – hopefully later today. My apologies.