Introduction

Half of the adult Americans aged 18 years and over are considered to be regular drinkers [1]. Reportedly, 10 % of North Americans are excessive alcohol consumers, while 3 % of Americans self-report experiencing alcohol withdrawal [2]. Excess alcohol use contributes to 20 % of admissions to the intensive care unit (ICU) [3], and chronic alcoholism may affect as many as 50–60 % of trauma patients [4]. Many of these patients are at risk for developing early alcohol withdrawal syndrome (AWS) [5], particularly those with alcohol dependence (DSM criteria supplement 1). Withdrawal seizures may occur 12–24 h later, while few patients develop delirium tremens (DT), with symptoms including agitation, hallucinations, disorientation, tachycardia, hypertension, fever, agitation, and diaphoresis.

Despite the apparent high frequency among ICU patients, no guidelines for the recognition or management of AWS or DTs in the critically ill have been published. Developing clear evidence-based management directives remains challenging because of inconsistent approaches in individual studies including types of ICU, patient demographics, definitions of AWS risk, prophylaxis and treatment regimens, and outcomes reported.

The objectives of this study were to systematically review the ICU literature to identify AWS risk factors and tools validated for AWS detection, prevention strategies, treatment approaches, and appropriate outcomes among critically ill patients.

Methods

Search methodology

The authors developed an initial list of key words related to AWS in the ICU, and a professional librarian (Odette Hinse) expanded this list, developed corresponding medical subject heading terms, and searched relevant clinical databases (search details are in online supplement S2). Given the continuum and overlap between AWS, DTs, and alcohol withdrawal seizures and the limited high-quality data, all forms of alcohol withdrawal were labeled AWS and included in this project. The publications were reviewed focusing on diagnostic criteria (risk factors and screening tools), prevention and treatment protocols and outcomes. Articles could be considered in more than one category. Case reports and series, editorials, narrative reviews, systematic reviews, animal or in vitro studies and letters to the editor were all reviewed. Publications that contained original data were retained; all other publication types were excluded after careful content and reference review.

Quality of evidence was scored using OXFORD criteria (1–5) [6] and Grading of Recommendation Assessment, Development and Evaluation (GRADE) system (high, moderate, or low/very low) [7, 8]. Low and very low GRADE and four and five Oxford level studies were excluded. At least two authors independently performed the quality profile for each study, attaining perfect (100 %) concordance between reviewers for the OXFORD level of evidence. The GRADE criteria were also concordant between reviewers, but not as uniform, likely related to the paucity, inconsistency, and heterogeneity of data. Of 112 eligible articles, 26 were retained for grading based on content. Reviewing the references for all papers identified an additional eight articles. A total of 34 articles met our final search criteria (Fig. 1).

Fig. 1
figure 1

Description of search

Full size image

Results

Diagnostic criteria

AWS risk factors and incidence

The identification of explicit risk factors for AWS among ICU patients could not be completed because no study prospectively evaluated all risk factors, and because studies considered different ones. In addition, inclusion criteria were different across studies. Moreover, studies varied by type of hospital and ICU, and prospective prevention or treatment trials did not specify the number of overall ICU admissions to use as a denominator, again making AWS incidence and risk stratification challenging [4, 9–11]. The available incidence data and descriptors of AWS are summarized in Table 1.

Table 1 Described incidence of AWS across studies and populations
Full size table

Some studies based AWS risk on level of alcohol consumption alone, but with inconsistent thresholds [12–15]. Trauma patients who present with alcohol in their blood are often thought to be at risk for AWS, but when studied, they are not at greater risk for withdrawal; this feature does not correlate with chronic alcohol use [16].

Alcohol consumpion as a predictor for developing AWS is described with varying thresholds and varying classification schemes. A standard alcohol drink is typically defined as 12 grams of alcohol, which is equivalent to 355 ml (12 oz) of beer, 150 ml (5 oz) of wine, or 45 ml (1.5 oz) of 80-proof liquor. Using these definitions, the National Survey on Drug Use and Health 2010 reported that 52 % of Americans older than 12 years of age reported being current drinkers (at least one drink in the past 30 days), 23 % binge drank (five or more drinks on the same occasion, on at least 1 day in the past 30 days) and 6.7 % reported heavy drinking (≥5 drinks on the same occasion on 5 or more days in the past 30 days) [17]. These data contrast with primary care and hospitalized patients, in whom alcohol dependence has been reported to be as high as 20–42 % [18–20]. In an observational study in a group of alcohol-dependent patients (determined with DSM-IV positive and median daily alcohol consumption of >100 grams of alcohol), the incidence of withdrawal seizures and delirium was 17 % before prevention measures could be initiated [21].

Alcohol-dependent patients with a history of prior alcohol withdrawal or those consuming alcohol while being treated for an alcohol related disease constitute the greatest risk for withdrawal symptoms [13].

Screening tools for AWS

Several tools have been used to identify patients at risk for AWS including the CAGE questionnaire and the Short Michigan Alcohol Screening test (short-MAST) [22]. The CAGE questionnaire asks patients four questions: Have you ever felt you needed to Cut down on your drinking? Have people Annoyed you by criticizing your drinking? Have you ever felt Guilty about drinking? Have you ever felt you needed a drink first thing in the morning (Eye-opener) to steady your nerves or to get rid of a hangover? Abuse or dependence on alcohol is suggested by ≥2 “yes” answers, indicating further investigation is warranted [23, 24]. “Yes” responses to ≥3 questions and a daily consumption >60 g define alcohol-dependence or abuse according to DSM-III or IV criteria.

Despite its simplicity, CAGE may be limited in its use in the ICU because of its failure to predict severity of AWS or outcome [25]. Among 652 surgical oncology ICU patients, 24/26 (92 %) with CAGE scores ≥1 developed AWS; the two without AWS had CAGE scores of 1 and 3, and drank 4–6 drinks daily. Three patients with CAGE scores of 0 and alcohol intake of 2–8 drinks a day developed AWS [25]. A minority of authors propose considering patients with CAGE scores of 1 or 2 and alcohol consumption of 25–60 g/day social drinkers, combined with a biological marker (carbohydrate-deficient transferrin or CDT) to define alcohol abuse and justify AWS prophylaxis. Unfortunately, little data is available at this time to support the value of biomarkers to define AWS risk [26, 27].

The Short MAST is described in one study that evaluated critically ill patients with acute respiratory distress syndrome and multiple organ dysfunction [28]. Its value in critically ill patients, however, has not been psychometrically validated.

AWS assessment

Several withdrawal or agitation scales have been used to objectively rate symptom severity in patients experiencing AWS or DT. The revised Clinical Institute Withdrawal Assessment for alcohol scale (CIWA-Ar) and the Sedation Agitation Scale (SAS) are the most frequently described. The CIWA-Ar scale (Table 2) was revised for use in medical ICU patients [29] and used in five studies [4, 9, 10, 29, 30]. A score ≥20 reflects full-blown AWS in most studies [4, 10, 29], while prevention studies targeted varying CIWA scores (10 to <20) and different evaluation intervals (ranging from every 10 min to four times daily) [4, 10, 29, 30]. No study documented a link between frequency of assessments and outcomes, nor addressed the challenges of assessing CIWA-Ar in mechanically ventilated patients. Intubated patients were often excluded [14] or not mentioned at all [4, 11, 31]. Some studies allowed patients intubated after the onset of AWS to remain in the study during withdrawal treatment [10, 16, 32–34], and one study considered mechanical ventilation a complication of AWS treatment [35].

Table 2 Revised Clinical Institute Withdrawal Assessment for Alcohol (CIWA-Ar) scale
Full size table

The SAS was used to titrate pharmacologic therapy or as part of an AWS prevention protocol in two publications [14, 33]. A score ≥5 triggered pharmacologic intervention with a therapeutic goal score of 3–4. Titration was required at least every hour if agitation persisted in the treatment trial [33] and adjusted to maintain a score of 4 based on SAS measurements every 4 h in the prevention study [14].

Prevention

Four single center ICU alcohol withdrawal prevention studies were identified [9, 14, 15, 30]. Criteria for beginning AWS prophylaxis varied from a history of alcohol consumption to alcohol dependence, but used different criteria. Ethanol infusions were monitored in single arm studies [15, 30] or compared to treatment with benzodiazepines, clonidine, or antipsychotic medications [9, 14] applying widely varying administration protocols. Although the rate of ethanol elimination is highly unpredictable in alcohol-dependent patients [15], prophylactic ethanol infusion appeared moderately effective (13/32 patients developed alcohol withdrawal) [30]. An unblinded study compared ethanol to benzodiazepines in chronic alcoholics. The findings suggested similar efficacy since no patient in either arm developed withdrawal [14]. This may have reflected the low-risk of the selected cohort. The effect of four different preventive regimens: flunitrazepam–clonidine, chlomethiazole–haloperidol, flunitrazepam–haloperidol and ethanol, was compared in 197 alcohol dependent patients, with similar rates of withdrawal prevention and ICU length of stay [9]. Appendix 1 provides greater detail on the inclusion criteria and results of the four prevention studies described above.

Treatment

Ten moderate to high quality treatment studies for alcohol withdrawal in ICU patients were identified, including two prospective, controlled, randomized and blinded studies [4, 10], one prospective but lower quality effort [11], 3 pre-post reports [25, 31, 33], and four single-arm designs [25, 29, 31, 36]. Treatment was primarily benzodiazepine-based [4], [10], [25], [29], [31], [33], [36, 37], including lorazepam [25, 31, 33], flunitrazepam [4, 10, 37], midazolam [31], diazepam [33, 37], and chlordiazepoxide [33, 37]. Drug combinations in these studies included benzodiazepines, antipsychotics (haloperidol), chlomethiazole, phenobarbital, clonidine, propofol, carbamazepine, and valproate [4, 10, 25, 33, 35].

Protocol and symptoms driven treatment

Three groups evaluated the impact of using standardized protocols for AWS treatment [25, 31, 33]. Protocol-driven management was associated with less benzodiazepine use (p = 0.014), and lower complication rates (intubation, excessive sedation) but similar hospital and ICU length of stay in a study of 36 medical ICU patients [31]. Symptom-triggered therapy for resistant alcohol withdrawal in 96 ICU patients was associated with higher benzodiazepine doses but a significant reduction in mechanical ventilation (p = 0.008) [33] and transfers to the ICU for AWS-related causes were significantly decreased in patients with head and neck cancer treated with a protocol (p = 0.03) [25].

Several treatment studies linked specific clinical withdrawal symptoms with triggered medication administration [11, 25, 36]. Lansford grouped them into three distinct clusters that prompted different drug classes to treat AWS: central nervous system excitation (anxiety, restlessness, being bothered by bright lights or sounds) was treated with benzodiazepines; adrenergic hyperactivity (nausea or vomiting, tremor, sweating, hypertension, tachycardia, premature beats) was treated with clonidine; and delirium was treated with haloperidol [25]. A drug and symptom class-based protocol in surgical ICU patients treated with symptom-driven boluses of clonidine, haloperidol, and flunitrazepam suggested this protocolization decreased severity and duration of alcohol withdrawal symptoms (p ≤ 0.01), and led to shorter ICU and ventilation duration (p ≤ 0.01) [10].

A study of 159 trauma ICU patients compared the combinations of flunitrazepam-clonidine, chlomethiazole-haloperidol, or flunitrazepam–haloperidol [4]. The chlomethiazole–haloperidol group had significantly more pneumonia (p = 0.04) and longer mechanical ventilation duration (p = 0.03), while the flunitrazepam–clonidine group experienced significantly more cardiac complications (p = 0.005) [4, 10].

The rapidity of withdrawal symptom manifestation and the speed of progression to full-blown withdrawal syndrome, if left untreated, were emphasized in several studies [31, 33, 36]. Aggressive treatment within the first 8–24 h appears crucial to ensure rapid symptom control, with no trial addressing optimal timing and frequency of assessments.

Appendix 2 summarizes patient characteristics and AWS treatments considered in the treatment portion of this review.

Outcomes

Studies describing alcohol withdrawal and delirium tremens necessarily include alcohol dependent patients. No study, however, compared alcohol consumers to alcohol dependent patients, or contrasted these two groups with occasional drinkers or abstainers, in order to stratify complication risk categories. Alcohol dependent patients are reported to have higher infection, sepsis and septic shock rates [38], are more likely to get admitted to the ICU and die in the hospital [39], and cost more in hospital resource dollars than patients not admitted to ICU for alcohol-related problems [32]. The few studies that addressed co-variates emphasized a high rate of ICU admission attributable to alcohol-related diagnoses, but once admitted, their length of stay and costs were no different from non-alcoholic medical [3] or trauma [40] patients.

Financial estimates varied greatly in methods, but an episode of alcohol withdrawal cost $7,462 per patient for benzodiazepines and ICU monitoring in one study [36], while among patients admitted for alcohol-related problems, the cost of an ICU stay was significantly higher in the alcoholic group ($52,527) compared to non-alcoholics ($43,136) [41]. Implementing AWS management guidelines is associated with a reduction in benzodiazepine acquisition costs and ICU length of stay [37].

Discrepancies in reported outcomes for various drinking intensity categories may relate to different definitions of risk, with some studies finding no differences between at risk and non-at-risk drinkers in ICU morbidity or mortality [39, 42], while patients with acute respiratory distress syndrome had worse outcomes if they regularly consumed alcohol [28]. Patients admitted to ICU with alcohol related complications (cirrhosis, GI bleed, intoxication, withdrawal) had a longer length of stay and higher mortality if more than one alcohol-related clinical feature contributed to their ICU admission [43, 44] or if chronic illness or delirium were present or mechanical ventilation required [43, 45].

Not surprisingly, critically ill trauma patients developing AWS have a longer duration of mechanical ventilation and ICU stay, more frequent pneumonia, urinary tract infections, sepsis, and septic shock, and higher mortality [46]. These adverse events may be associated with alcohol use, but immune down-regulation has also been associated with the pharmacological effects of morphine, propofol and benzodiazepines [47]. Patients experiencing AWS require more frequent tracheostomy and PEG feeding tubes, and require higher doses of sedation [48, 49], which has been associated with prolonged mechanical ventilation and length of stay [50–52].

Long term outcome data are sparse, but patients admitted to the ICU for DT are often seen again in the emergency room within two years related to AWS or alcohol related complications [53]. Appendix 3 presents the outcomes extracted from included studies.

Discussion

Despite an estimated prevalence of chronic alcoholism affecting up to 20–40 % of hospitalized patients and 50–60 % of trauma patients, little high-quality data for how best to prevent, diagnose, and treat AWS in the ICU is available and ICU-specific guidelines have not been published.

Methods to identify ICU patients at risk for AWS include alcohol related questionnaires (CAGE), alcohol consumption documentation, the Short MAST, and a history of prior AWS or seizures, while the reliability of biochemical markers such as CDT has yet to be determined. The relative validity of these variables has not been compared. A prior history of alcohol withdrawal and seizures should be considered a significant risk for AWS, though this information may not be obtainable in all patients. Asking patients or next of kin about alcohol consumption and withdrawal may stratify patients at risk for AWS or DT, and facilitate recognition of the need for prompt titrated pharmacological management.

Risk thresholds in prevention studies vary widely, and best prevention pharmacotherapy has not been defined. The threshold between prevention studies for patients at risk or with minor symptoms and treatment studies with more significant AWS components is often been blurred, and clearer definitions are needed.

Once withdrawal occurs, early and frequent assessment of withdrawal symptoms is essential, particularly in the first 24–48 h. The CIWA-Ar relies on patient communication for information regarding nausea/vomiting, anxiety, tactile and auditory disturbances, and headache. This tool may not be applicable or reliable in critically ill patients, particularly in mechanically ventilated patients. Titrating treatment or prophylaxis to agitation symptoms such identified with the SAS in general ICU patients, and with CIWA-Ar whenever feasible, appears the best approach to match drug dosing and symptom severity and improve outcomes.

Benzodiazepines are a mainstay of AWS treatment, despite uncertainty about their effectiveness and safety. Barbiturates and propofol appear safe and effective GABA alternatives for AWS. Clonidine as a combination regimen is efficacious in reducing the adrenergic symptoms of AWS, and combination therapy with benzodiazepines, alpha-2 agonists and antipsychotics was associated with good outcomes in multiple studies [10, 25]. Dexmedetomidine is structurally similar to clonidine; however, AWS management with this drug is limited to case reports [54]. Its usefulness as adjunctive therapy to benzodiazepines has been reported [55]. Although best practices in prevention and treatment strategy cannot be established given the limited evidence and inconsistent designs, combination pharmacotherapy, titrated to symptom severity and linked to symptom clusters appears promising.

Published ICU studies of patients with AWS have reported inconsistent outcomes since they have applied highly variable definitions. The standardization of definitions for AWS will facilitate future comparisons and systematic reviews. In addition to standard patient demographics and history of alcohol ingestion, additional beneficial data should include whether an ICU admission was for AWS or other critical illness, whether the study was intended for prevention or treatment, what monitoring tools were applied, and what the thresholds were for each pharmacologic intervention and what triggered their administration. The ICU-related complications such as sepsis, pneumonia, mechanical ventilation, ARDS, duration of ICU and ventilator therapy, and discharge location should also be sought.

Several limitations of our manuscript deserve comment. Because of limited high quality data, we combined research addressing various forms of AWS from different types of ICUs using different study populations, designs, thresholds for treatment, AWS prevention or treatment protocols, and patient outcomes. This variability highlights the need for more uniform research approaches to this complex area before treatment guidelines based on evidence can be proposed. Delirium, a common and morbid complication of critical care admission, is both a risk factor and a potential confounder for alcohol withdrawal symptoms which we were not able to address [45]. In addition, our filtering process was qualitative by necessity, but our consensus between authors is likely to have reduced potential bias.

Significant gaps remain in the current literature, and these should stimulate future studies. The detection of withdrawal or risk for AWS in mechanically ventilated patients, particularly when no history is available from the patient, has not been studied. Patients presenting with psychomotor slowing rather than agitation have largely been ignored in the critical care setting [56]. Additional gaps in knowledge that require study include identification of the optimal methods to stratify risk for alcohol withdrawal and delirium tremens in various populations, establishing whether prophylaxis in high-risk patients is beneficial and safe, determining whether prophylactic ethanol administration best serves this purpose, comparative trials of alcohol withdrawal treatment, and outcome studies that consider the many confounders (alcohol consumption among them) likely to blur the link between alcohol withdrawal and outcomes. Finally, whether adverse events are related to alcohol consumption, dependence, withdrawal, patient co-morbidities, or the treatment for AWS remains unclear.

Conclusion

Early detection of alcoholic patients at risk for AWS should be routine for ICU patients, and is likely best performed based on identifying a history of heavy alcohol consumption and prior withdrawal events. Early and aggressive treatment with combination therapy regardless of pharmacologic agents such as benzodiazepines, alpha-2 agonists, and anti-psychotics should be titrated to specific withdrawal symptoms. Outcomes should be stratified by alcohol use, alcohol withdrawal, and concomitant co-morbidities, and should provide extended follow-up and ongoing efforts towards alcohol detoxification and abstinence. Large prospective trials in critically ill patients, particularly those who are intubated and mechanically ventilated, are needed to evaluate the best tools to assess the presence and the severity of acute withdrawal syndrome and the optimal pharmacologic approaches for prevention and treatment.