Prevalence of benzodiazepine use disorder during hospitalization for alcohol detoxification

Benzodiazepines (BZDs) are the first‐line treatment of alcohol withdrawal. Comorbidity between benzodiazepine use disorder (BUD) and alcohol use disorders (AUD) is common. However, the risk factors are poorly characterized due to the paucity of available BUD screening tools. The present study aimed to rectify this by conducting an observational screening investigation for BUD in patients hospitalized for alcohol detoxification in a specialized unit. During a face‐to‐face interview, a short BUD screening tool, Echelle Cognitive d'Attachement aux benzodiazépines (ECAB), was administered to record recent patterns of BZD use, thereby allowing categorization of AUD patients as follows: non‐BZD users, BZD users without BUD, and BUD (ECAB ≥6). Clinical and sociodemographic risk factors were identified and recorded during clinical assessment and were analyzed using nonparametric bivariate tests and multinomial regression for association with BUD, with p < 0.05 for significance. Of the 150 AUD patients, 23 (15%) had comorbid BUD. Several variables were associated with ECAB score, with their independence being verified using multinomial regression, with lower risk of BUD versus BZD use, when the initial prescriber was an addiction specialist compared with a psychiatrist or a general practitioner [odds ratio (OR) = 0.12, 95% confidence interval (CI) = 0.14–0.75]. A higher risk of BZD use versus no use was evident when comorbid psychiatric disorders were present (OR = 9.2, 95%CI = 1.3–65). Our findings raise clinicians' awareness that in patients hospitalized for alcohol detoxification, BUD is highly prevalent but not specifically related to psychiatric disorders. BUD can be effectively screened by utilization of the ECAB.


| INTRODUCTION
Benzodiazepine medication is prescribed for anxiety, sleep disorders as well as alcohol withdrawal prevention or treatment. The data regarding the misuse and the adverse events of this class of drugs have mostly been gathered in the geriatric population, including the risk of falls, delirium [1], and possible increased incidence of neurodegenerative disorders in chronic users [2]. Data also indicate an increase in all-cause mortality in chronic benzodiazepine users from clinical samples with substance use disorders [3], especially in cases of concurrent use of other sedatives [4] or alcohol [5]. Despite this wealth of adverse health consequences, studies assessing benzodiazepine use disorder (BUD) as an addiction, indicated by the loss of control over such drug, are rare [6]. BUD is evident in approximately 9% of regular benzodiazepine users [7] and 1.5% of ever-users [6]. BUD is a relatively recent, but growing, concern with up to 17% of ever-users who show signs of misuse [6]. The French are among the highest benzodiazepine users in the world [5,8,9]. However, investigation of BUD in a French sample is rare.
Alcohol use disorder (AUD), characterized by a loss of control over alcohol intake, is a major complication of repeated alcohol use. A study published in the Lancet in 2010 [10] highlighted AUD as the most damaging of the addictions when combining damage to self and others, in comparison with heroin and cocaine addiction. AUD is often comorbid with other addictions, but its relationship with BUD has been relatively neglected.
The overlaps and interactions of AUD and BUD are manifold. Firstly, benzodiazepine and alcohol are two allosteric modulators of the gamma-aminobutyric acid (GABA)-A receptor [11], with benzodiazepine and alcohol potentiating each other's effects. Secondly, benzodiazepine medications are the first-line curative and preventive treatment of alcohol withdrawal [12], making AUD patients more likely to have benzodiazepine exposure. Thirdly, both AUD and BUD are highly comorbid with other psychiatric disorders (including mood, psychotic, anxiety, and personality disorders), where benzodiazepines are still not uncommonly prescribed on a long-term basis, despite increased risk for adverse events and doubtful efficacy [13,14]. However, in comparison with opioid use disorders [15][16][17][18], there has been relatively little investigation of the characteristics of BUD in people with AUD. One notable exception is a French study of 1005 in/out-patients seeking treatment for AUD, wherein 413 benzodiazepine/Z-drug users (41%) were identified and 217 (53%) of whom fulfilled the DSM-IV criteria for BUD [19]. These alarming findings call for replication and extension. Therefore, we undertook an observational study in patients hospitalized for alcohol detoxification in a specialized unit from a University Hospital Centre in Paris, France. Our primary aim was to evaluate the prevalence of BUD in AUD inpatients. The secondary objective was to identify the risk factors for comorbid BUD in AUD inpatients, eventually identifying profiles of AUD patients with the greatest risk of comorbid BUD.

| MATERIAL AND METHODS
This was a cross-sectional study using medical charts and single, face-to-face interview for collecting clinical and sociodemographic data.

| Participants
We recruited all patients consecutively hospitalized for alcohol detoxification in an addiction inpatient unit, between May 6 and September 16, 2019. Two patients could not be assessed for eligibility due to scheduling demands.
This unit is located in a Paris university hospital and specializes in complex detoxification procedures. The only inclusion criterion was being hospitalized for alcohol detoxification, regardless of possible other indications and other substances involved at the time of admission. Exclusion criteria were as follows: patients who did not speak French or having very severe cognitive impairment. The study assessment was included in the "care-as-usual" procedures. Thus, participants were not required to provide written informed consent, but the investigator recorded that they had no objection to have their clinical data used for the study, in compliance with the French law on bioethics in clinical research. As such, the study was approved under the decision no. 1-20-090 ID 9604 from the Comité de Protection des Personnes Sud-Ouest et Outre-Mer 1.

| Assessments
Benzodiazepine use was assessed in two steps. Firstly, we asked whether the patient had used any benzodiazepine in the 6 months prior to hospitalization. Following a positive response, the patterns and the trajectory of benzodiazepines use were investigated, including the following: indication of use during alcohol detoxification (yes/no), name of the prescribed benzodiazepine (further categorized into long vs. intermediate + short half-life [20]), maximum dosage ever taken, age at first use, and initial prescriber (categorized as general practitioner, psychiatrist, addiction specialist, or other specialist). Importantly, we did not consider baclofen as a benzodiazepine, although Z-drugs were included as within BUD. Next, as an assessment of BUD in benzodiazepine users, Echelle Cognitive d'Attachement aux benzodiazépines (cognitive benzodiazepine attachment scale, ECAB) [21] was applied. The ECAB consists of 10 binary items rated 1 or 0. A total score of 6 or more discriminates between the presence versus absence of BUD patients (sensitivity of 94%; specificity of 81%). Although this tool has not been formally validated, it is the only recommended assessment by the French Health Authorities (HAS), as stated in their 2007 recommendations [22]. The ECAB is an easy-to-use self-rated scale with high reproducibility [21]. These characteristics make it suitable for an inpatient AUD population where "ease of use" is a highly desirable characteristic for both recruited patients and clinicians, who may use it in the future. There is a paucity of questionnaires investigating BUD, with few validation studies. One is much lengthier than the ECAB [23], while the other focuses on DSM criteria [24], which may lack sensitivity for benzodiazepine users. Neither of these two other assessment tools were available in French at the time of this study, making ECAB the only practicable option.
Age, gender, and body mass index (BMI) were recorded as were data on medication regimens at both admission and discharge, as possible risk factors for BUD. These medications were categorized as follows: (i) non-BZD psychotropic medication, further categorized into antidepressants, mood-stabilizers, antipsychotics (both first and second generation), medication for opioid use disorder (MOUD), and AUD medications not aimed at detoxification and available in France at the time of the study (i.e., acamprosate, baclofene, disulfirame, nalmefene, and naltrexone) [25] and (ii) nonpsychotropic drugs. Psychotropics all belonged to the category "N" from the Anatomical Therapeutic Chemical (ATC) Classification, limited to medications used in psychiatric disorders, as summarized in various recommendations from the HAS and reported in previous studies from our group [26]. Antipsychotics of the first and second generation were mixed altogether, thus, this category also included neuroleptics, which are mostly used for symptomatic treatment. Any history of substance use and psychiatric disorders were also recorded, based on patients' medical charts and/or recent specialized evaluation. The presence of lifetime psychiatric diagnosis was categorized as follows: "no psychiatric diagnosis", when no such disorder was ever evident in medical charts or during the current inpatient stay or "possible psychiatric diagnosis", if such disorder was suspected but refuted by a psychiatrist or when a psychiatric disorder was suspected due to psychiatric symptoms or psychotropic medications, yet without psychiatric assessment, and "certain psychiatric diagnosis" when a psychiatrist entered a psychiatric diagnosis. Finally, we identified cognitive impairment using medical charts in cases were an ICD10 (2008 version) F00-F09 diagnosis was present or based on current assessment by the Montreal Cognitive Assessment (MOCA) [27]. In our clinical unit, the MOCA is systematically administered to all patients admitted for alcohol detoxification, with typical administration occurring between day 8 and 15, when benzodiazepines have been either withdrawn or are below 10% of the initial dose. We applied the commonly accepted MOCA cutoff <26 to define current cognitive impairment [27].

| Statistical analysis
Our assessment of benzodiazepine use yielded three groups of interest: 1. "no or very light benzodiazepine users", which included participants without recent benzodiazepine use OR current benzodiazepine use but with an ECAB score of 0; 2. "benzodiazepine users", which included participants with recent use but without evidence of BUD, having an ECAB score between 1 and 6; 3. "BUD", with ECAB ≥6 (best estimate for BUD).
We used Kruskal-Wallis, Chi 2 , and Fisher's exact test to investigate statistical associations between ECAB category and continuous and categorical variables, respectively. The significance threshold was set at 0.05 for the analyses in all three groups and to correct for multiple testing, at p = 0.05/3 = 0.017 for post hoc pairwise analyses, which were performed when the global test across the three groups was significant.
Variables significantly associated with our predefined BZD users' groups were entered as independent variables in a multinomial regression model, which was run twice: once with "no or very light benzodiazepine users" as the reference category and once with "benzodiazepine users", in order to yield an additional odds ratio for "BUD" versus "benzodiazepine users". We also report the model overall explanatory properties (McFadden R 2 ) and multicollinearity (variance inflation factor). Should our sample prove be too small to perform all subgroup analysis in multinomial regression, we also ran simpler models after removing some of the independent variables, until the relevant investigations could be performed entirely. These analyses were deemed exploratory to enhance interpretation of the findings.
The risk of comorbid BUD was significantly different (p = 0.012), varying as a function of first prescriber, T A B L E 1 Comparison of benzodiazepine users in alcohol use disorder inpatients hospitalized for alcohol detoxification.
Level of benzodiazepine use/use disorder at the ECAB increasing from addiction specialist (13%) to psychiatrist (35%) to general practitioner (43%), with one pairwise comparison significant between the BUD versus BZD users' groups ( p = 0.028). Neither the duration of intake nor the duration of action of the particular benzodiazepine was associated with BUD risk ( p = 0.991 and p = 0.125, respectively). Several variables related to psychiatric disorders were significantly associated with BUD, namely, more frequent psychiatric diagnosis ( p < 0.001), with a significantly higher frequency of BUD in the "confirmed" group compared with the others (96% and 74% vs. cumulative 4% and 26%, p < 0.001); a higher number of nonbenzodiazepine psychotropic drugs in both BZD users groups (median = 2 for both groups compared with 1 for nonusers, p = 0.006); and more frequent current antidepressant prescription (48% and 43% in the BUD and the BDZ users' groups, respectively, vs. 23% in the nonuser group; p = 0.01). Other psychotropic treatments were not associated with the risk of BUD (mood-stabilizers, p = 0.052; neuroleptics, p = 0.113; MOUD, p = 0.063 and medication for AUD, p = 0.387). Age and female gender were not associated with the risk of comorbid BUD ( p = 0.491 and 0.751, respectively). In comparison of BZD use versus nonuser groups, emergency hospitalization was not significantly different ( p = 0.174) but was significantly different for "reason for hospitalization" (pairwise p < 0.001), as well as across all groups ( p = 0.001). In the BZD use versus nonuser groups, the proportions of admission for "alcohol detoxification only" (82% vs. 63%), "alcohol + any other substance" (12% vs. 3%), and "alcohol detoxification + complication" (6% vs. 34%) seemed different. Variables related to  F I G U R E 1 Plot of multinomial logistic regression with benzodiazepine (BZD) use /dependence/no use as the dependent variable, as estimated by the questionnaire of attachment to benzodiazepines (ECAB). Some associations could not be computed due to insufficient subgroup size (all "possible psychiatric diagnosis" values and "confirmed psychiatric diagnosis" for groups other than "BZD use vs. no use").
AUD (duration of the disorder, cirrhosis, cognitive impairment) (smallest p-value = 0.845) and the number of cigarettes smoked per day (p = 0.762) were not associated with BUD. About 97% of all participants were current tobacco smokers. Multinomial regression (Figure 1) was performed with BZD users' groups as the dependent variable and age, gender, psychiatric diagnosis, duration of BZD use, and initial prescriber as the independent factors (the two latter were naturally missing for nonusers of BZD). Significant associations remained between initial prescriber and BZD users' group, where having an addiction specialist as the initial BZD prescriber was associated with a lower risk of BUD versus BZD use [OR = 0.12, 95% confidence interval (CI) = 0.14-0.75, p = 0.0216] and having confirmed psychiatric disorder was associated with increased risk of BZD use versus no use (OR = 2.6, 95% CI = 1.6-6.8, p = 0.0445). In order to investigate all psychiatric diagnosis groups, we had to remove the initial BZD prescriber and BZD duration. By doing so, we show that confirmed psychiatric disorder was independently associated with both BZD use and BUD compared with no use (ORs = 7.3 and 34, p = 9.9 Â 10 À6 and 0.0009, respectively), but not with BUD compared with BZD use (p = 0.158) (data not shown, available on request).

| DISCUSSION
In this monocentric, cross-sectional study of benzodiazepine use and use disorder measured by the ECAB questionnaire in 150 patients hospitalized for alcohol detoxification, we found a 15% prevalence of BUD. The risk factors of BUD were related to initial prescriber and lifetime psychiatric disorders in bivariate analysis. Refining these findings using multinomial regression revealed that BUD versus BZD use was less likely when the initial prescriber was an addiction specialist and that psychiatric disorders were associated with BZD use and BUD compared with no use but not with BUD compared with BZD use.
The 15% prevalence of BUD in the current study is consistent with previous studies in similar populations [19] and with the 19% past-month nonmedical benzodiazepine users in a US AUD sample [28]. One study of current benzodiazepine users (ECAB >0 and <6) showed a prevalence of 26% higher than in the general population [7]. This prevalence was close to the 31% misuse prevalence reported in a French psychiatric population [13] but twice as low as the 53% reported by Morel et al. in their subsample of current benzodiazepine users. This discrepancy could reflect true sample differences, although exposure to benzodiazepines and to social adversity, which would tend to increase BUD risk, were higher in our sample. Other explanations include the following: the possibly high sensitivity of the use of a nonvalidated self-rating questionnaire and/or the fact that our sample included patients, who were mostly in treatment for months/years, as compared with Morel et al.'s, which contained a substantial proportion of first-episode AUD patients. This latter hypothesis is supported by our finding that benzodiazepines initiated by addiction specialists were less likely than those initiated by other practitioners to be associated with BUD.
Psychiatric disorders were differentially associated with BZD use versus no use but not with BUD versus BZD use. This finding is novel and related to a recent French publication reporting that benzodiazepine misuse was common in psychiatric inpatients compared with the general population [13]. This suggests that, in our sample with severe AUD, individuals with psychiatric disorders were more likely to be treated with BZD, with possibly shared risk factors between psychiatric disorders and BUD, although this did not lead to an increase in BUD risk. This interpretation is subject to caution since our regression model showed limited reliability regarding associations between psychiatric disorders and BZD users' groups (some ORs could not be computed). Multinomial regression modeling following removal of initial BZD prescriber and BZD duration reinforced our original finding that psychiatric disorders increased the general risk for BZD use in our sample, possibly for long durations or at high dosage, but not the specific risk of BUD.
The occupational title of the first-time prescriber was significantly associated with the presence of BUD, increasing from addiction specialist to psychiatrist to general practitioner. We hypothesize that addiction specialists are more aware and better prepared to manage the risk of BUD than other practitioners, either by a deeper assessment of previous BUD history or at-risk BZD use patterns or by using more efficient strategies to prevent BUD in current users. Psychiatrists probably face particularly complex situations of comorbid AUD and other psychiatric disorders and general practitioners may encounter difficulties tapering/withdrawing benzodiazepine using in AUD populations.
Gender was not significantly associated with BUD in our specific sample of patients with severe AUD patients undergoing inpatient alcohol detoxification, suggestive an equal risk in this high-risk population, in contrast to the observed sex-related risk observed in the general population. The p-value of 0.419 for gender x BUD interaction indicates no significant association in the AUD population. This is a novel and relevant finding as and should encourage clinicians to monitor the risk of BUD in both genders equally.
Short-versus long-acting BZD subtypes did not convey any differential BUD risk in AUD. Short-acting compounds, across different substances, are typically regarded as more addictive, including benzodiazepines [29]. Our findings suggest that this is not the case in the high-risk group of patients with severe AUD. Likewise, the duration of benzodiazepine use, the duration of AUD, nor age at interview were associated with BUD. A trend is evident for comorbid opioid use, with 22% of people taking MOUD having an ECAB > 6, compared with 5% and 11% in the other groups ( p = 0.063). The probability indicates a possible lack of power in this analysis, which would be parsimonious with previous evidence indicating that opioid-dependent individuals (whether treated with medications or not) have a high BUD risk [17,18]. Finally, the overall nonpsychiatric severity profile-as estimated by assessing cognition, cirrhosis, emergency admission, nonpsychotropic medications-was not associated with BUD.

| Limitations of the study
This was a monocentric, retrospective study, carried out in Paris in a single center, conducted with inpatients suffering from severe addictive disorders, which can limit the generalizability of our findings. The ECAB scale appeared to be an effective standardized means for the early identification of BUD, given the paucity of relevant available tools. However, the interpretation of our findings may be limited by the lack of proper psychometric validation of the ECAB in nongeriatric populations. Due to stigma and/or recall issues, patients may further underreport their history of addiction and/or other psychiatric diagnoses. AUD medication was collected on a yes/no basis, preventing us from conducting possibly relevant investigations between the type of AUD medication and ECAB score. Finally, there was no control group to check whether our results are specific to our population.
The current study has several strengths. It was one of the first studies carried out in France in patients with severe AUD using a validated questionnaire to assess BUD. We were systematic and only had two patients not included in the study, indicative of no selection bias. Our study, although exploratory, was carried out using a combination of standardized tools and basic clinical questions in a consecutively recruited sample of patients with severe addiction profiles. The ECAB standardized tool was used to identify patients with benzodiazepine-related disorders which, despite their prevalence in this population, remained understudied to date-as has been noted by other groups [4]. The ECAB scale appears to be an effective standardized means for the early identification of BUD. It will be interesting to compare the ECAB scale with DSM V criteria, especially in regard to clinical utility.
The data also indicate that no other clinical or sociodemographic characteristics of patients with AUD (patterns of benzodiazepine use, AUD history and complications, age, gender) were associated with the risk of BUD, which should encourage clinicians to closely monitor BUD risk in all AUD patients.

| CONCLUSION
Very few epidemiological studies in France have investigated BUD and alcohol/benzodiazepine co-dependence, especially using standardized instruments. In this monocentric cross-sectional study of benzodiazepine use and dependence measured by the ECAB questionnaire in 150 patients hospitalized for alcohol detoxification, we observed a prevalence of BUD, estimated with the ECAB score of 15%. Using a threegroup variable for categorizing BZD use versus BUD versus no use, we identified a relevant and novel association between BZD use and psychiatric disorders, as compared with no use, in a population with severe AUD.

| PERSPECTIVE
Longitudinal studies would better investigate the relationships between initial prescriber and BUD and between psychiatric disorders and BZD use. Optimizing psychiatric care for patients using BZD may eventually reduce BUD prevalence. It will be interesting to assess whether patients with both AUD and BUD have particular trajectory, such as more frequent adverse events, more severe psychiatric disorders, and/or a lower life expectancy compared with those with only AUD or BUD. The development of the ECAB should aid the investigation and treatment of this common clinical comorbid presentation.
Institute for Research (INSERM), and the NARSAD. No other author has conflict of interest to declare.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The study was approved under the decision no. 1-20-090 ID 9604 from the Comité de Protection des Personnes Sud-Ouest et Outre-Mer 1 and fulfilled ethical standards of the Helsinki declaration 1989.

CLINICAL TRIAL REGISTRATION
Not applicable.

PATIENT CONSENT
The study assessment was included in the "care-asusual" procedures. Thus, participants were not required to provide written informed consent, but the investigator recorded that they had no objection to have their clinical data used for the study, in compliance with the French law on bioethics in clinical research.

PERMISSION TO REPRODUCE MATERIAL FROM OTHER SOURCE
Not applicable.