Phenobarbital versus benzodiazepines in alcohol withdrawal syndrome

Abstract

Aim

Phenobarbital, a long‐acting barbiturate, presents an alternative to conventional benzodiazepine treatment for alcohol withdrawal syndrome (AWS). Currently, existing research offers only modest guidance on the safety and effectiveness of phenobarbital in managing AWS in hospital settings. The study objective was to assess if a phenobarbital protocol for the treatment of AWS reduces respiratory complications when compared to a more traditionally used benzodiazepine protocol.

Methods

A retrospective cohort study analyzing adults who received either phenobarbital or benzodiazepine‐based treatment for AWS over a 4‐year period, 2015–2019, in a community teaching hospital in a large academic medical system.

Results

A total of 147 patient encounters were included (76 phenobarbital and 71 benzodiazepine). Phenobarbital was associated with a significantly decreased risk of respiratory complications, defined by the occurrence of intubation (15/76 phenobarbital [20%] vs. 36/71 benzodiazepine [51%]) and decreased incidence of the requirement of six or greater liters of oxygen when compared with benzodiazepines (10/76 [13%] vs. 28/71 [39%]). There was a significantly higher incidence of pneumonia in benzodiazepine patients (15/76 [20%] vs. 33/71 [47%]). Mode Richmond Agitation Sedation Scale (RASS) scores were more frequently at goal (0 to −1) between 9 and 48 h after the loading dose of study medication for phenobarbital patients. Median hospital and ICU length of stay were significantly shorter for phenobarbital patients when compared with benzodiazepine patients (5 vs. 10 days and 2 vs. 4 days, respectively).

Conclusion

Parenteral phenobarbital loading doses with an oral phenobarbital tapered protocol for AWS resulted in decreased risk of respiratory complications when compared to standard treatment with benzodiazepines.

Phenobarbital, a long‐acting barbiturate, presents as an alternative to conventional benzodiazepine treatment for alcohol withdrawal syndrome (AWS). This is a retrospective cohort study analyzing adults who received either phenobarbital or benzodiazepine‐based treatment for AWS over a 4‐year period, 2015 to 2019, in a community teaching hospital in a large academic medical system. Parenteral phenobarbital loading doses with an oral phenobarbital tapered protocol for AWS resulted in decreased risk of respiratory complications when compared to standard treatment with benzodiazepines.

1. INTRODUCTION

Alcohol is one of the oldest and most widely used substances in the world. ¹ Alcohol is a central nervous system (CNS) depressant that can promote changes in neuronal pathways leading to behavioral alterations; its exact mechanism is complex and can vary from person to person. ² , ³ , ⁴ Alcohol enhances the inhibitory effects of gamma‐aminobutyric acid (GABA) receptors in the CNS and decreases excitatory glutamatergic activity by decreasing the number of ionotropic glutamatergic receptors. ³ , ⁵ It exerts effects on the dopaminergic reward system through the opioid and dopaminergic systems ² , increases serotonin transmission acutely and reduces it during alcohol withdrawal ⁶ , and activates the G protein‐coupled inwardly rectifying K (GIRK) channels. ³ Liver alcohol dehydrogenase is the major enzyme system for metabolizing alcohol. ² Alcohol use is associated with 87 798 deaths annually in the United States, many of these due to alcohol use disorder (AUD). ⁷ In one study, nearly 8% of all hospital admissions, 16% of postsurgical patients, and 31% of trauma patients suffered from AUD and eventually developed alcohol withdrawal syndrome (AWS). ⁸

The National Institute on Alcohol Abuse and Alcoholism (NIAAA) ⁹ reports that approximately 16 million people in the United States have AUD and are therefore at increased risk of developing AWS. AWS can occur as early as 6 h and may persist for up to 7 days after alcohol cessation. AWS can include seizures, hallucinations, delirium, and/or autonomic instability. The incidence of AWS is higher in critically ill patients; AWS is harder to detect in this setting, as it shares many similarities with physiologic responses observed in critical illness. ¹⁰ , ¹¹ , ¹² Therefore, prompt identification and management of AWS in hospitals remain a challenge.

Benzodiazepines, traditionally used for AWS treatment, work on the GABA_A receptor by increasing its opening frequency. ¹³ However, some patients may not respond adequately to benzodiazepines as gradual changes to GABA conformation(s) occur with chronic alcohol use and may result in cross‐tolerance to benzodiazepines. ¹³ , ¹⁴ Additionally, benzodiazepines require endogenous GABA to bind to the GABA channel for them to properly exert their effect. Patients with AUD often have low levels of endogenous GABA, meaning benzodiazepines may be unable to exert the desired effect. These patients often develop benzodiazepine‐resistant alcohol withdrawal. ¹⁵ Benzodiazepines have a well‐known abuse and dependence liability. ¹⁶ Finally, benzodiazepines pose challenges in hospitalized patients by increasing risk for over‐sedation, memory deficits, and respiratory depression in a compromised population. ¹³ , ¹⁷ , ¹⁸ AUD predisposes patients to pulmonary infections through various mechanisms including the suppression of cough and gag reflexes, decreased mucociliary clearance, decreased alveolar macrophage function, and overall dysfunction of cell‐mediated immunity. ¹ , ¹⁹

Phenobarbital, a long‐acting barbiturate, presents an alternative to conventional benzodiazepine treatment for AWS. ¹³ It is a CNS depressant that promotes binding to GABA_A receptors, modulates chloride currents through receptor channels, and inhibits glutamate‐induced depolarizations. Phenobarbital may reduce cross‐tolerance between benzodiazepines and alcohol through decreasing activity at glutamate AMPA and kainate receptors and acting at GABA_A receptors to increase channel open time. ¹³ , ²⁰ , ²¹ Phenobarbital is metabolized in the liver by CYP2C9 with minor metabolism by CYP2C19 and CYP2E1. A quarter of the dose of phenobarbital is excreted unchanged in the urine. In adults, the half‐life of phenobarbital is aproximately 100 h. ²² Phenobarbital's rapid onset and long duration of action allow the drug to treat acute AWS symptoms and to prevent its recurrence. ²³ However, concerns remain regarding the use of phenobarbital for treatment of AWS due to its lack of thorough evaluation, narrow therapeutic range, and concern that it may interact with other drugs. ²⁴

Due to the known effects of AUD on the respiratory system, it is important to asses if a phenobarbital protocol for treatment of AWS reduces respiratory complications when compared to a benzodiazepine protocol to provide a safe and efficacious alternate treatment option. Because of the prevalence of AUD and, ultimately, AWS among hospitalized patients; the known complications of benzodiazepine treatment, and the potential advantages of using phenobarbital to treat AWS; it was hypothesized that a tapered phenobarbital protocol for AWS would be associated with fewer respiratory complications when compared to a benzodiazepine protocol. Respiratory complications were defined as incidence of intubation and requirment of at least 6 liters of oxygen.

2. MATERIALS AND METHODS

2.1. Study design, setting, and patient population

The Duke University Health System Institutional Review Board approved this single‐center, retrospective cohort study conducted in the emergency room and a 22‐bed Intensive Care Unit (ICU) in Duke Regional Hospital (DRH), a 369‐bed community teaching hospital. All patients included in the study began treatment in the Emergency Department or ICU and ultimately completed the course in the ICU. The phenobarbital order panel in the electronic health record (EHR) was released in December 2017 due to a need for an alternative to the traditional benzodiazepine treatment regimen. The purpose of this study was to assess the safety and efficacy of parenteral phenobarbital followed by an oral phenobarbital taper for the treatment of patients with or at risk for moderate to severe AWS with a hospital admission between 2015 and 2019. Moderate to severe AWS was defined as Prediction of Alcohol Withdrawal Severity Scale (PAWSS) ≥ 4, persistent Clinical Institute Withdrawal Assessment for Alcohol (CIWA) score > 15, continued objective signs of withdrawal appropriate for treatment per CIWA, prior hospitalization for AWS, history of alcohol withdrawal seizures, admission blood alcohol level (BAL) ≥ 200 mg/dL, or clinical judgment by the health care provider. ²⁵ Adult patients were included if they were initiated on either the phenobarbital protocol or received parenteral benzodiazepine‐based treatment for alcohol withdrawal. Patients were excluded if they were pregnant, listed an allergy to any study medication, or had acute intermittent porphyria, fulminant liver disease, or epilepsy.

2.2. Study treatment

The first study cohort included patients who initiated the phenobarbital protocol from December 2017 to January 2019. They received a 4‐day tapered protocol with a parenteral loading dose. The phenobarbital protocol represents adaptations of protocols used successfully at other institutions. ¹³ , ²⁶ The 4‐day treatment included a loading dose of either 6 or 10 mg/kg based on ideal body weight (IBW) divided into three intramuscular doses, 3 h apart on day 1, followed by a tablet maintenance dose taper for days 2 through 4. If total body weight (TBW) was less than IBW, then TBW was utilized for the dosing. Per protocol, nurses were authorized to administer a breakthrough parenteral dose of phenobarbital 65 mg every 6 h as needed for withdrawal symptoms (Figure 1).

FIGURE 1.

Phenobarbital protocol. The 4‐day treatment included a loading dose of either 6 or 10 mg/kg based on ideal body weight (IBW) divided into three intramuscular doses; 3 h apart on day 1; followed by a tablet maintenance dose taper for days 2 through 4. If total body weight (TBW) was less than IBW; then TBW was utilized for the dosing. Per protocol; nurses were authorized to administer a breakthrough parenteral dose of 65 mg every 6 h as needed for withdrawal symptoms. The recommended loading dose of either phenobarbital 6 mg/kg or 10 mg/kg was based on risk of sedation and respiratory compromise. Risk factors for sedation included age of 65 or older; hepatic dysfunction; or cirrhosis. Risk factors for respiratory compromise included pneumonia; chronic obstructive pulmonary disease (COPD); asthma; interstitial lung disease; or pulmonary fibrosis. If patients had one of the above risk factors; then the protocol would recommend the lower loading dose of phenobarbital; 6 mg/kg.

The recommended loading dose of either phenobarbital 6 mg/kg or 10 mg/kg was based on the risk of sedation and respiratory compromise. Risk factors for sedation included age 65 or older, hepatic dysfunction, or cirrhosis. Risk factors for respiratory compromise included pneumonia, chronic obstructive pulmonary disease (COPD), asthma, interstitial lung disease, or pulmonary fibrosis. If patients had one of the above risk factors, the protocol would recommend the lower loading dose of phenobarbital, 6 mg/kg.

The second study cohort analyzed included patients initiated on a benzodiazepine‐based treatment regimen with ICU admission between January 2015 and January 2019. All benzodiazepine patients received the same nursing‐driven protocol, the Clinical Institute Withdrawal Assessment for Alcohol‐ Revised (CIWA‐Ar) protocol. The CIWA‐Ar scale is a 10‐item tool originally intended for voluntary detoxification centers to detect and monitor AWS. Its components include nausea/vomiting, tremor, diaphoresis, anxiety, agitation, tactile disturbances, auditory disturbances, visual disturbances, headache, and orientation. Scores for each item range from 0 to 7. Cumulative scores of 8–10 indicate mild withdrawal. Scores of 8–15 indicate moderate withdrawal, and scores of 15 or greater indicate severe withdrawal. ²⁵

2.3. Study outcomes

Baseline characteristics were evaluated through a retrospective chart review of the EHR. The primary objective was to assess if a parenteral phenobarbital loading dose followed by an oral tapered phenobarbital protocol was associated with decreased respiratory complications, defined by decreased rates of intubation and less frequent requirements for at least 6 L of oxygen when compared to benzodiazepine treatment in patients with AWS. Secondary outcomes included hospital and ICU length of stay (LOS), the incidence of seizure, or pneumonia, the total number of additional medications used as adjunct treatments within the first 2 weeks of hospitalization, Richmond Agitation Sedation Scale (RASS) mode scores, and type of discharge. RASS mode scores were reviewed for specific intervals post‐initiation of study medication (0–9 h, 9–24 h, 24–48 h, and 48–96 h). Type of discharge was categorized into one of the following: (1) against medical advice (AMA); (2) planned discharge (home, skilled nursing facility (SNF), rehabilitation center, or another hospital); or (3) death from any cause or hospice.

Medications for adjunct treatment during hospitalization were ascertained, including dexmedetomidine, haloperidol, and quetiapine. Dexmedetomidine is an α ₂‐adrenoceptor agonist extensively metabolized through glucuronidation and cytochrome P450 metabolism in the liver. ²⁷ , ²⁸ The volume of distribution at a steady state of dexmedetomidine is approximately 118 L, and the mean elimination half‐life is 1.5–3 h following IV administration. ²⁹ Haloperidol works via dopamine receptor antagonism in the CNS. Haloperidol is extensively metabolized by hepatic enzymes (partially by members of the CYP450 family). Only 1% is excreted unchanged in the urine. ³⁰ Quetiapine is an antagonist at D₂ and 5‐HT_2A receptors and downregulates 5‐HT_2A receptors. ³¹ , ³² Quetiapine is rapidly absorbed in the body, reaching maximum observed plasma concentration within 1–2 h. The drug is 83% bound to serum proteins, and has a mean terminal half‐life of 7 h. The drug exhibits linear pharmacokinetics. ³²

2.4. Statistical analysis

All analyses were performed using IBM SPSS v.28 (City is Armonk, NY). The primary outcome was evaluated utilizing Fisher's exact tests. Mann–Whitney U and chi‐square tests were conducted to compare secondary clinical endpoints. Baseline demographics and clinical characteristics were compared using independent samples t‐tests, Mann–Whitney U, and chi‐square tests. The type of discharge was analyzed using a chi‐square test. Descriptive statistics were performed for the total amount of study medication administered. Power was set to 0.8 (80%), and alpha was set to 0.05, with an estimated sample size of 147. ³³

Multivariable regression was completed to control for confounding covariates. To determine if treatment exposure was associated with clinical outcomes, accounting for discharge type and blood alcohol level (BAL), a series of hierarchical linear regressions were conducted on significant primary and secondary continuous outcomes. Ordinal logistic regression was performed to determine if treatment exposure was associated with the number of additional medications to treat delirium when accounting for discharge type. Logistic regression models were conducted to examine binary outcomes.

3. RESULTS

1259 patients were admitted to the ICU between January 2015 and January 2019 who met the study criteria for alcohol withdrawal. 147 patients were included in the study: 76 patients were treated with phenobarbital, and 71 patients were treated with a benzodiazepine‐based regimen. The primary reason for exclusion was no administration of AWS treatment medication.

Baseline demographics and clinical characteristics (Table 1) were similar between cohorts. The mean Charlson Comorbidity Index score was 2.3 for phenobarbital patients and 2.6 for benzodiazepine patients. The mean age was 52 years for phenobarbital patients and 51 years for benzodiazepine patients, with both treatments being predominantly male at 70% and 80%, respectively. There were two exceptions to the similarity of baseline characteristics. First, there were significant differences in collecting a BAL at admission. The phenobarbital patients had a significantly lower rate of BAL collection compared to the benzodiazepine patients (65% vs. 86%; p = 0.004). Second, the laboratory result of the median BAL was significantly higher for the phenobarbital patients 43 mg/dL (min <5 mg/dL, max 562 mg/dL) vs. 5 mg/dL (min <5 mg/dL, max 336 mg/dL; p < 0.001). A BAL of <5 mg/dL is reflective of a not detected level. Previous alcohol‐related healthcare visits occurred more frequently in phenobarbital patients, although this was not a statistically significant outcome (46/76 [61%] vs. 32/71 [45%] patient encounters; p = 0.07).

TABLE 1.

Baseline demographic and clinical characteristics.

Characteristic	Phenobarbital, n = 76	Benzodiazepine, n = 71	p‐Value
Mean age, years (SD)	52 (11)	51 (11)	0.454
Male, n (%)	53 (70)	57 (80)	0.183
Race, n (%)
Caucasian/White	46 (61)	42 (59)	0.729
African American/Black	25 (33)	24 (34)
Asian	3 (4)	1 (1)
Hispanic	1 (1)	3 (4)
Two or more races	1 (1)	1 (1)
BMI, n (%)
Underweight (<18.5)	6 (8)	4 (6)	0.599
Normal (18–24.9)	31 (41)	31 (44)
Overweight (25–29.9)	25 (33)	18 (25)
Obese (≥30)	14 (18)	18 (25)
Admission BAL collected, n (%)	49 (65)	61 (86)	0.004
Median BAL mg/dL (min, max) a	43 (<5, 562)	5 (<5, 336)	<0.001
Respiratory comorbidities, n (%)
Asthma	4 (5)	4 (6)	0.795
COPD	8 (11)	10 (14)
Mean CCI (SD)	2.3 (2)	2.6 (2)	0.607
Prior alcohol‐related visit, n (%)	46 (61)	32 (45)	0.07
Initial RASS (min, max)	0 (−4, 4)	0 (−5, 4)	0.868

Note: The baseline demographics and clinical characteristics were similar between the two groups, with two exceptions the phenobarbital patients had a significantly lower rate of BAL collection compared to the benzodiazepine patients (65% vs. 86%; p = 0.004). The laboratory result of the median BAL was significantly higher for the phenobarbital patients (43 mg/dL (min <5 mg/dL, max 562 mg/dL) vs. 5 mg/dL (min <5 mg/dL, max 336 mg/dL); p < 0.001). This means the phenobarbital‐treated patients had higher blood alcohol levels on arrival.

Abbreviations: BAL, blood alcohol level; BMI, body mass index; CCI, Charlson Comorbidity Index; COPD, chronic obstructive pulmonary disease; RASS, Richmond Agitation Sedation Scale; SD, standard deviation.

^a n = 49 phenobarbital and n = 61 benzodiazepine.

Phenobarbital treatment was associated with a statistically significant decreased occurrence of respiratory complications, including intubation (15/76 [20%] vs. 36/71 [51%] patients; p < 0.001). Phenobarbital treatment was also associated with a statistically significant lower occurrence of 6 or greater liters of oxygen (10/76 [13%] vs. 28/71 [39%] p < 0.001) (Table 2). Of note, intubation refers to any time after the initiation of treatment for AWS with either phenobarbital or benzodiazepines.

TABLE 2.

Primary outcome and secondary clinical outcomes.

Clinical Outcomes	Phenobarbital; n = 76	Benzodiazepine; n = 71	p‐Value
Primary outcome
Intubation (after initiation of AWS protocol); n (%)	15 (20)	36 (51)	<0.001
≥6 L of oxygen; n (%)	10 (13)	28 (39)	<0.001
Secondary outcomes
Pneumonia incidence; n (%)	15 (20)	33 (47)	<0.001
Seizure incidence; n (%)	5 (7)	6 (9)	0.759
Median seizures; (min; max)	0 (0; 3)	0 (0; 6)	0.677
Median LOS in days (min; max)
ICU	2 (0; 11.2)	4.2 (0.7; 30)	<0.001
Hospital	5.3 (0.3; 20.7)	9.9 (1.4; 63.7)	<0.001
Median additional medications for delirium; (min; max)	0 (0; 3)	1 (0; 3)	<0.001
Dexmedetomidine	12 (16)	40 (56)	<0.001
Haloperidol	16 (21)	28 (39)	0.019
Quetiapine	8 (11)	18 (25)	0.029
Mode RASS
0–9 h (min; max)	0 (−4; 3)	‐1 (−5; 4)	0.525
9–24 h (min; max)	0 (−5; 3)	−2 (−5; 4)	0.002
24–48 h (min; max)	0 (−5; 1)	−1 (−5; 3)	<0.001
48–96 h (min; max)	0 (−5; 1)	0 (−5; 3)	0.357
Discharge type; n (%)
Against medical advice	16 (21)	3 (4)	0.01
Planned	55 (72)	62 (87)
Death from any cause/hospice	5 (7)	6 (9)

Note: Phenobarbital treatment was associated with a statistically significant decreased occurrence of respiratory complications including intubation (15/76 [20%] vs. 36/71 [51%] patients; p < 0.001). Phenobarbital treatment was also associated with a statistically significant lower occurrence of 6 or greater liters of oxygen (10/76 [13%] vs. 28/71 [39%] p < 0.001). Of note; intubation refers to any time after the initiation of treatment for AWS with either phenobarbital or benzodiazepines.

Abbreviations: ICU, intensive care unit; LOS, length of stay; Max, maximum; Min, minimum; Planned: home, skilled nursing facility, rehabilitation, other hospitals; RASS, Richmond Agitation Sedation Scale.

Phenobarbital treatment was associated with a shorter median ICU LOS (median, [minimum, maximum]) (2 [0, 11] vs. 4 [0.7, 30] days), and hospital LOS (5 [0.3, 21] vs. 10 [1, 64] days). Phenobarbital treatment was associated with significantly fewer total additional medications for the treatment of delirium within 2 weeks when compared to benzodiazepine‐treated patients (0 vs. 1 additional medication) including dexmedetomidine (12/76 [16%] vs. 40/71 [56%] patients), haloperidol (16/76 [21%] vs. 28/71 [39%]), and quetiapine (8/76 [11%] vs. 18/71 [25%]). Benzodiazepine‐treated patients had significantly more pneumonia diagnoses (15/76 [20%] vs. 33/71 [47%]). RASS mode score comparisons showed phenobarbital‐treated patients were significantly less sedated with a smaller range of scores at the 9 to 24‐hour period (0 [−5, 3] vs. ‐2 [−5, 4]) and 24 to 48‐hour period (0 [−5, 1] vs. −1 [−5, 3]), relative to the benzodiazepine‐treated patients. The mean total phenobarbital dose during hospitalization, including as‐needed doses, was 968 mg (standard deviation (SD) 580 mg) and a median dose of 803 mg. The mean total dose of benzodiazepines in lorazepam equivalents was 86 mg (SD 106 mg) and median dose of 40 mg.

3.1. Sensitivity analyses

Sensitivity analyses were completed to control for possible covariates of type of discharge (AMA vs. planned) and BAL. Two separate hierarchical linear regressions were conducted evaluating ICU LOS, hospital LOS, RASS mode score at 9–24 h, and RASS mode score at 24–48 h to determine if treatment was predictive of these clinical outcomes. Type of discharge was a statistically significant outcome, with a higher percentage of phenobarbital patients leaving AMA. The phenobarbital patients had a significantly higher BAL at admission. Notably, even though the admission BAL was higher in the phenobarbital cohort, the initial RASS score was not statistically different between cohorts. The initial RASS score was defined as the first RASS score recorded in the patient's EHR immediately following the first dose of study medication administration, which was determined through a retrospective chart review.

The first hierarchical linear regression accounted for discharge type (AMA vs. planned), and the second accounted for BAL. Discharge types (AMA vs. planned) or BAL were entered in model 1, and treatment (phenobarbital vs. benzodiazepines) was entered in model 2. Change in r‐square between model 1 and model 2 was tested for significance. For ICU LOS, the addition of treatment to the model added an additional 7.1% of variance explained for the discharge type (beta = 0.278, p < 0.001) and an additional 7.8% of variance explained for BAL (beta = 0.287, p = 0.003). For discharge type, both discharged AMA (beta = −0.271, p = 0.26) and treatment (beta = 0.278, p < 0.001) were significant predictors of ICU LOS. For BAL, treatment was a significant predictor, but BAL (beta = −0.047, p = 0.618) was not a significant predictor. For hospital LOS, treatment was the only significant predictor for both the discharge type and BAL. The addition of treatment to the model added an additional 9.7% (beta = 0.321, p < 0.001) and 10.6% (beta = 0.336, p < 0.001) of variance explained when evaluating the discharge type and BAL, respectively. For RASS mode score between 9 and 24 h, the addition of treatment to the model added an additional 5.6% (beta = −0.244, p = 0.004) and 9.6% (beta = −0.316, p < 0.001) of the variance explained for discharge type and BAL, respectively. When evaluating the discharge type, both planned discharge (beta = 0.294, p = 0.017) and treatment were significant predictors of RASS mode score between 9 and 24 h. Treatment was a significant predictor of RASS mode score between 9 and 24 h, but BAL was not (beta = 0.016, p = 0.864). For RASS mode score between 24 and 48 h, the addition of treatment to the model added an additional 7.4% (beta = −0.2794, p = 0.001) and 11.5% (beta = −0.346, p < 0.001) of the variance explained, respectively. For the final model, discharged AMA (beta = 0.306, p = 0.011), planned discharge (beta = 0.352, p = 0.003), and treatment (beta = −0.2794, p = 0.001) were significant predictors of RASS mode score between 24 and 48 h. BAL was not a significant predictor (beta = 0.064, p = 0.495).

Ordinal logistic regressions examining total additional medications for the treatment of delirium, accounting for discharge and BAL, were both significant (X² (3) = 33.261, p < 0.001 with a Nagelkerke R ² = 0.223 and X² (3) = 20.290, p < 0.001 with a Nagelkerke R ² = 0.185, respectively). Only treatment was a significant predictor for total additional medications for treating delirium (type of discharge model: estimate = −1.52, p < 0.001 and BAL model: estimate = −1.61, p < 0.001). Discharged AMA (estimate = 0.569, p = 0.511), planned discharge (estimate = −0.814, p = 0.204), and BAL (estimate = −0.001, p = 0.51) were not significant predictors for total number of additional medications for the treatment of delirium.

Odds ratios (95% CI) were evaluated through logistic regression models to examine binary outcomes of incidence of pneumonia, use of dexmedetomidine, haloperidol, intubation at any point, and 6+ liters of oxygen. A statistically significant model p‐value was defined as p < 0.05, and marginally significant was p < 0.1. Due to a lack of variation in data, the model for quetiapine within 2 weeks did not converge. Treatment was a statistically significant predictor of the outcomes in all models except for prediction of haloperidol, which was marginally significant (p = 0.08) when considering the type of discharge, and treatment was a statistically significant predictor of the outcomes in all models when taking into account BAL.

4. DISCUSSION

This retrospective cohort study adds to a growing body of literature demonstrating that phenobarbital is a relatively safe and efficacious alternative to the traditional therapy of benzodiazepines for AWS. Compared to patients who received benzodiazepines, those who received phenobarbital developed fewer respiratory complications as defined by a lower incidence of both of the following primary outcomes: need for mechanical ventilation and need for more than 6 L/min of supplementary oxygen.

Benzodiazepines have become the drugs of choice for AWS; however, they are not ideal agents in treating most of the symptoms of AWS because of their short duration of action and accumulation in fat‐soluble tissues. For example, diazepam maintains its anticonvulsant effect for only about 20 min, and oral diazepam accumulates during chronic therapy when used for its sedative effect, especially in patients with liver disease. ³⁴ , ³⁵ , ³⁶ In fact, because of risks posed by benzodiazepines, including increased ICU LOS, delirium, anxiety, depression, cognitive dysfunction, and length of mechanical ventilation, they are no longer recommended for sedation in critically ill adults. ³⁷ , ³⁸

Historically, AWS has been managed by administering benzodiazepines, supportive care, and continuous evaluation using a validated clinical scale, such as the CIWA scale. ²⁵  However, the CIWA protocol places a heavy burden on the nursing staff, requiring frequent reassessment of the patient beyond traditional clinical monitoring. As discussed previously, patients receiving large cumulative doses of lorazepam are at risk for adverse effects, including the development of both delirium and respiratory depression. ³⁹ Other pharmacological agents, such as dexmedetomidine and propofol, can be used as adjuncts to reduce the total amount of benzodiazepines given or to control refractory symptoms. This approach, however, may increase cost (in the case of dexmedetomidine) or require mechanical ventilation (in the case of propofol). ⁴⁰ Little has been published supporting the use of alternative strategies for treating AWS. Thus, clinicians are left with few options managing these complex patients. ⁴¹

Phenobarbital is a candidate for the treatment of AWS for many reasons. There is cross‐tolerance between barbiturates and ethanol. The medication has a wide margin of safety when used for sedative‐hypnotic withdrawal. The doses that are efficacious for treating withdrawal symptoms do not produce significant CNS depression. Phenobarbital has a rapid onset of action, so monitoring the clinical effects of phenobarbital loading doses is practical. At therapeutic serum concentrations, phenobarbital has an anticonvulsant efficacy like diazepam or phenytoin. ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ Finally, because of its long duration of action (half‐life of approximately 90–100 h), serum phenobarbital concentrations remain stable for long periods after single loading doses. ²² The long half‐life of phenobarbital eases the administration burden compared with benzodiazepines, which may need to be given more than once per hour. Phenobarbital's long half‐life allows for a gradual transition off therapy after the last dose is provided.

This current work further supports the use of phenobarbital in treating AWS. In this study, patients who received phenobarbital versus those receiving benzodiazepines developed fewer respiratory complications as defined by a lower incidence of both of the following primary outcomes: need for mechanical ventilation and need for more than 6 L/min of supplementary oxygen. The above findings might be explained by baseline differences in cohorts. However, there was no differences in the mean Charlson Comorbidity Index between cohorts. There were also no differences in age, gender, race, BMI, and respiratory comorbidities, including COPD and asthma, between cohorts. The benzodiazepine cohort did have a significantly greater percentage of BAL samples collected on admission; however, fewer of these samples met the positive cutoff value for the hospital laboratory versus the phenobarbital‐treated cohort. This means the phenobarbital‐treated patients had higher blood alcohol levels on arrival. This could be due to one of the following: (a) patients who ended up getting phenobarbital may have been more intoxicated on arrival to the hospital than patients who received benzodiazepines; (b) patients who received phenobarbital presented closer to their peak alcohol ingestion time; and/or (c) patients who received phenobarbital were more physiologically tolerant of alcohol than the ones receiving benzodiazepines. Notably, all three explanations suggest that the patients who received phenobarbital presented at a possibly higher risk for complicated withdrawal than those who received benzodiazepines.

Pneumonia incidence, length of ICU stay, and length of hospital stay were greater in the benzodiazepine‐treated cohort. These findings are consistent with prior reports that alcohol consumption can contribute to developing pneumonia. ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ It is known that AUD predisposes patients to pulmonary infections through various mechanisms. ¹ , ¹⁹ Previous studies have shown that exposure to benzodiazepines can increase the risk of pneumonia, particularly in patients with other risk factors, including AUD and AWS. ⁵¹ , ⁵² The increased risk may occur at the cellular level through benzodiazepine actions on GABA signaling. One potential explanation is the lack of downregulation of GABA_A receptors in patients with pneumonia exposed to benzodiazepines leading to acidosis in the intracellular macrophage, affecting its ability to serve as a phagocyte and kill. ¹ , ⁵³ This may explain the increased pneumonia incidence and increased hospital and ICU stay in benzodiazepine‐treated patients, as they require additional recovery time due to dysfunctional cellular immunity. Overall, this suggests that a phenobarbital protocol may positively impact a patient's ability to recover from illness associated with AWS.

Our findings also suggest phenobarbital‐treated patients were less sedated than benzodiazepine‐treated patients and had significantly fewer adjunct medications administered within the first 2 weeks of hospitalization than benzodiazepine‐treated patients. RASS scores were more frequently at goal (0 to −1) between 9 and 48 h after the loading dose of study medication for phenobarbital patients indicating these patients were significantly less sedated. Adjunct medications administered included haloperidol, dexmedetomidine, and quetiapine. Fewer adjunct medications administered within the first 2 weeks of hospitalization in phenobarbital‐treated patients can mean multiple things. For example, phenobarbital patients may have been less sedated due to the need for fewer adjunct medications. Additionally, prior studies have shown that benzodiazepine administration increases the risk of delirium in critically ill adults, especially older adults. ⁵⁴ , ⁵⁵ , ⁵⁶ Delirium is an acute confusioned state marked by severely disorganized cognition, fluctuating course, additional deficits, and disturbed awareness. ⁵⁷ It is possible that additional medications were administered more frequently to treat delirium that may have been attributed to benzodiazepine administration, but further work is necessary to clarify this. Overall, this suggests that a phenobarbital protocol may provide a viable alternative to benzodiazepine treatment for AWS and limit side effects, including over‐sedation caused by benzodiazepine treatment.

This work corroborates findings in other studies that the use of a phenobarbital protocol for AWS was associated with a significant reduction in ICU LOS; a decreased use of adjunct medications, including quetiapine, haloperidol, and dexmedetomidine; and a decreased need for invasive mechanical ventilation. ⁴¹ , ⁵⁸ , ⁵⁹ While prior studies have shown phenobarbital's safety and efficacy for AWS, most of these studies have reviewed the use of intravenous phenobarbital compared with a benzodiazepine treatment regimen. ¹³ , ²⁶ , ⁴¹ , ⁶⁰ , ⁶¹ Previous studies on using phenobarbital to treat AWS have relied on weight‐based dosing (10 mg/kg) or escalating or de‐escalating doses, beginning at 60 mg or 260 mg, respectively. ²³ , ⁵⁹ , ⁶¹ , ⁶² A simple and practical intramuscular phenobarbital protocol has yet to be established. ⁴¹

This study provides further evidence for lower doses of intramuscular phenobarbital and a shortened oral taper to treat AWS. A single intramuscular injection can deliver 390 mg of phenobarbital into the body while permitting a slow drug release into the bloodstream. In contrast, intravenous administration immediately releases all administered phenobarbital into the bloodstream. The protocol used in this study could easily be implemented on hospital wards, whereas intravenous loading typically requires an ICU or ED setting.

Potential limitations of this study include the retrospective design, the concurrent dexmedetomidine shortage, and medication crossover. First, the retrospective study design means that the cohorts could have been impacted by unmeasured confounders. Second, the dexmedetomidine shortage began close to the same time as the initiation of the phenobarbital protocol and continued throughout the study period. This may have increased the benzodiazepine cohort's relative access to dexmedetomidine. However, since dexmedetomidine is a useful and safe adjunct for treating AWS, the relative absence of dexmedetomidine in the phenobarbital group further supports the safety and efficacy of phenobarbital as a primary therapy for AWS. Third, the study's retrospective nature created the potential for medication crossover, i.e., it was possible for phenobarbital‐treated patients to receive benzodiazepines. Approximately 10% of phenobarbital patients did receive at least one dose of benzodiazepine either concomitantly or after completing the phenobarbital protocol.

Based on the findings from the current study and other published papers, our health system has expanded this phenobarbital protocol across all inpatient locations. Future studies will measure respiratory complications, hospital, and ICU LOS, the incidence of pneumonia, the total number of additional adjunct medications for the treatment of AWS, and RASS mode scores in patient populations treated with benzodiazepines versus phenobarbital.

Overall, this study demonstrates that phenobarbital may serve as a promising alternative to benzodiazepine‐based treatment for AWS. Despite the phenobarbital cohort including more patients at greater risk for withdrawal (i.e., as manifested by a higher admission BAL), these patients generally experienced better outcomes than those who received benzodiazepines. Compared to benzodiazepine administration for AWS, phenobarbital may help to reduce respiratory complications. Further, the intramuscular route and relatively low doses of phenobarbital used in this study provide a pathway for usage beyond the ICU setting in hospital wards. However, additional studies are needed to test the safety of this protocol in the hospital ward setting.

AUTHOR CONTRIBUTIONS

The authors confirm contribution to the paper as follows: study conception and design: D. Malone; D. MacElroy; M. Al‐Hegelan; and Y. Bronshteyn; data collection: D. Malone; J. Thompson; interpretation of results: D. Malone; B. Costin; J. Thompson; and Y. Bronshteyn; draft manuscript preparation: D. Malone; B. Costin; and Y. Bronshteyn. All authors reviewed the results and approved the final version of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICAL APPROVAL

Approval of the research protocol by an Institutional Reviewer Board: Approval of the research protocol by an Institutional Reviewer Board was obtained.

Informed Consent: N/A.

Registry and the Registration No. of the study/trial: N/A.

Animal Studies: N/A.

Supporting information

Appendix S1

Malone D, Costin BN, MacElroy D, Al‐Hegelan M, Thompson J, Bronshteyn Y. Phenobarbital versus benzodiazepines in alcohol withdrawal syndrome. Neuropsychopharmacol Rep. 2023;43:532–541. 10.1002/npr2.12347

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available as an Appendix S1.