The authors acknowledge support from the Civilian Research Development Fund (CRDF), the National Institute on Alcohol Abuse and Alcoholism (R21-AA014543-01A1, KO5 AA 14906-01, I-P50 AA-12870-04), and the U.S. Department of Veterans Affairs via their Merit Review Program and Alcohol Research Center for their support of this study.
Reprint requests: John H. Krystal, MD, Alcohol Research Center (116-A), VA Connecticut Healthcare System, 950 Campbell Ave., West Haven, CT 06516; Fax: +203-937-3468; E-mail: firstname.lastname@example.org
Background: Benzodiazepines are the standard pharmacotherapies for ethanol detoxification, but concerns about their abuse potential and negative effects upon the transition to alcohol abstinence drive the search for new treatments. Glutamatergic activation and glutamate receptor up-regulation contribute to ethanol dependence and withdrawal. This study compared 3 antiglutamatergic strategies for ethanol detoxification with placebo and to the benzodiazepine, diazepam: the glutamate release inhibitor, lamotrigine; the N-methyl-d-aspartate glutamate receptor antagonist, memantine; and the AMPA/kainite receptor inhibitor, topiramate.
Methods: This placebo-controlled randomized single-blinded psychopharmacology trial studied male alcohol-dependent inpatients (n=127) with clinically significant alcohol withdrawal symptoms. Subjects were assigned to 1 of 5 treatments for 7 days: placebo, diazepam 10 mg TID, lamotrigine 25 mg QID, memantine 10 mg TID, or topiramate 25 mg QID. Additional diazepam was administered when the assigned medication failed to suppress withdrawal symptoms adequately.
Results: All active medications significantly reduced observer-rated and self-rated withdrawal severity, dysphoric mood, and supplementary diazepam administration compared with placebo. The active medications did not differ from diazepam.
Conclusions: This study provides the first systematic clinical evidence supporting the efficacy of a number of antiglutamatergic approaches for treating alcohol withdrawal symptoms. These data support the hypothesis that glutamatergic activation contributes to human alcohol withdrawal. Definitive studies of each of these medications are now needed to further evaluate their effectiveness in treating alcohol withdrawal.
Benzodiazepines are the medications most commonly prescribed to reduce the severity of ethanol withdrawal (Ntais et al., 2005). However, the identification of effective alternatives to benzodiazepine-assisted detoxification is a high priority because (1) there is concern about the abuse potential of benzodiazepines in patients with a propensity for addiction and (2) benzodiazepine prescription may increase the level of alcohol craving and relapse to alcohol use after the initiation of sobriety (Malcolm et al., 2002).
This study was approved by the institutional review boards at the Pavlov State Medical University (St. Petersburg, Russian Federation) and the Human Investigation Committee of the Yale University School of Medicine (New Haven, CT). All subjects gave written informed consent before their study participation.
A total of 174 patients were screened for this study, of which, 34 patients failed to meet inclusion criteria and 7 patients withdrew consent before randomization. Male inpatients [n=133; age: 43.0±9.7 (SD) years] meeting DSM-IV criteria for alcohol dependence on the basis of a structured diagnostic interview (First et al., 1996) were recruited from the 7-day Inpatient Detoxification Unit of the Leningrad Regional Center of Addictions (St. Petersburg, Russia). Study entry criteria included a history of most recent alcohol consumption between 8 and 48 hours before study entry and clinically significant alcohol withdrawal symptoms on the basis of the Clinical Institute Withdrawal Assessment—Alcohol, revised (CIWA-Ar score>10; Sullivan et al., 1989). Patients were excluded if they took psychoactive or anticonvulsant medications other than those prescribed in the study, met DSM-IV criteria for opiate dependence, needed urgent treatment for other symptoms, i.e., suicidal, homicidal, delirium tremens, or if they were at high risk for untoward side effects from study medications (i.e., liver function tests more than 3 times higher than the normal level, evidence of active renal, hepatic, or cardiac disease on the basis of history, EKG, or other laboratory findings).
After obtaining written informed consent, subjects underwent a thorough medical evaluation including history, review of systems, physical examination, electrocardiogram, and a standard clinical laboratory assessment (CBC, liver function test, BUN, creatinine, urinanalysis, urine toxicology screen). The level of alcohol consumption for the 30 days before hospitalization was determined using the Timeline Follow-Back Method (Sobell and Sobell, 1992). At entry and throughout the study, the severity of alcohol withdrawal was assessed at least every 8 hours using a hospital-based clinical assessment based on the CIWA-Ar. Patients were immediately medicated with diazepam 10 mg, p.o. if symptom severity assessed by CIWA-Ar (score>10) or clinical assessment by the treating physician indicated that the level of alcohol withdrawal placed the patient at risk for a severe withdrawal-related medical complication. If this occurred, patients were then monitored intensively with the option of further diazepam administration up to every 4 hours until symptom severity was reduced to a moderate level. Each morning, raters blind to the treatment assignment administered an alcohol withdrawal severity scale (CIWA-Ar; Sullivan et al., 1989). Patients also completed a modified version of the Alcohol Withdrawal Symptoms Checklist (AWSC). This scale was developed by a group of investigators at the University of Pennsylvania (B. A. Flannery, personal communication) and susbstantially updated by the Pavlov Medical University research team to include minor symptoms of alcohol withdrawal (such as chill, depression, asthenia, sleep disorders, etc.) and make it more sensitive to the subtle changes in the withdrawal severity. As used, it consists of 17 self-report items rated between 0 and 4 in severity that reflect the severity of individual alcohol withdrawal symptoms. We recently completed a factor analysis of this scale (Pittman et al., 2007) that revealed that this scale is comprised of 5 factors consisting of the following items: factor 1, autonomic arousal: nervousness, sweating, tremor, sleep disturbance; factor 2, depression: depression, asthenia; factor 3, nausea and vomiting: nausea, vomiting; factor 4, physical tension: abdominal pain, muscle cramps, irritation/dysphoria; and factor 5, alcohol craving: chill, headache, craving for alcohol. Factors 1, 3, and 4 of the AWSC appear to be similar to factors assessed by the CIWA-Ar, while factor 2 (depression) and factor 5 (alcohol craving) are not included in the CIWA-Ar (Pittman et al., 2007). The level of dysphoric mood was also assessed daily using the Montgomery–Asberg Depression Rating Scale (MADRS; Montgomery and Asberg, 1979).
Block randomization was used with block sizes of 15, 20, and 25 randomly varied. Specifically, patients randomly chose from blocks of 15, 20, or 25 envelopes, each envelope containing 1 treatment assignment. Assessors were unaware of block sizes and were not involved with any aspect of the randomization. Patient treatment assignment was kept under lock and key throughout the study. Based on the homogeneity of the sample, we did not use stratification. The 133 study patients were assigned randomly to 1 of 6 groups for 7 treatment days: lamotrigine, memantine, topiramate, placebo, diazepam, and memantine with gradual dose titration. The last group was assigned to receiving an initial dose of memantine 10 mg on the first day, 20 mg on the second day, and 30 mg on the third day and thereafter. This treatment approach was discontinued after a small number of patients completed detoxification (n=6) to increase the feasibility of completing the other cells in a timely fashion. Retrospective analysis of these data suggested that patients receiving this graduated treatment did not differ from placebo, presumably because the effective dose of this medication was not achieved until withdrawal symptoms had spontaneously abated. Because this group size was not informative, it was dropped from all analyses. Thus, data are reported on 127 patients. All of the patients in this study smoked cigarettes and they had full access to an area that allowed smoking on the inpatient unit. Placebo treatment involved the administration of an inactive substance every 6 hours. Diazepam treatment was specified as 10 mg, p.o. every 8 hours for a total daily dose of 30 mg/d. Lamotrigine treatment consisted of 25 mg, p.o. every 6 hours for a total daily dose of 100 mg/d. Memantine treatment involved the administration of 10 mg, p.o. every 8 hours for a total of 30 mg/d. Topiramate treatment consisted of 25 mg, p.o. every 6 hours for a total of 100 mg/d. In all cases medication doses were selected as a compromise between expected efficacy and tolerabilitiy. With the exception of diazepam, we did not have clear guidance from the literature on dose selection. Therefore, we explored data collected in other types of studies. The dose of memantine was selected because we wished to avoid the maximum dose approved by the U.S. FDA (40 mg) and lower doses did not work in pilot testing, as noted above. The lamotrigine and topiramate doses were selected as a compromise between efficacy and anticipated side effects of dermatologic reactions (lamotrigine) and cognitive impairment (topiramate). The dose of topiramate selected for this study was equivalent to the daily dose of topiramate that had been shown to suppress alcohol withdrawal seizures in patients (Rustembegovic et al., 2002) and was twice the dose used to treat uncomplicated alcohol withdrawal (Choi et al., 2005).
As noted above, this study was a single-blind study with randomized assignment. Because study medications were not encapsulated, there was a potential for subjects to learn their group assignment by studying their medications if they were aware of distinctive markings associated with each study medication. We suspect that the blind was largely intact in this patient group based on informal clinical interactions with patients, although the integrity of the blind was not formally assessed. To promote the integrity of the blind, no medication could be identified by its administration schedule. Placebo, lamotrigine, and topiramate were administered 4 times daily, while diazepam and memantine were administered 3 times daily.
Data for each outcome (CIWA-Ar, AWSC, MADRS) were approximately normally distributed as determined using Kolmogorov–Smirnov test statistics and normal probability plots, and further from residual plots between predicted and residual values from the models. Initially, we determined whether each drug tested differed significantly from placebo. Linear mixed models were used to assess differences between drugs for each continuous outcome measured over time. In each model, rating scale data constitute the dependent outcome, while drug (placebo, diazepam, lamotrigine, memantine, and topiramate) was included as a fixed effect, and time (treatment days 1–3) was included as a within-subject explanatory factor. The interaction between drug and time was also fitted. The best-fitting variance–covariance structure was assessed using information criteria. Dunnett's multiple comparisons procedure (Dunnett, 1955) was performed to compare each drug versus placebo averaged over time. If the drug by time interaction was significant, each drug was further compared with placebo at each time point using specified contrasts, followed by the Bonferroni method to adjust elevated Type I error. As each test drug was significantly better compared with placebo, we refit the data excluding placebo. In these analyses, the same models described above were used with Tukey's multiple comparisons procedure (Miller, 1981) used in place of Dunnett's procedure. The 5 AWSC factors were not normally distributed and thus were analyzed by the nonparametric method by Brunner (Brunner et al., 2002), which generates an ANOVA-type statistic (ATS). In these models, the same factors described above were included in the models and post hoc comparisons were adjusted using Bonferroni corrections. None of the following covariates influenced the outcomes: age, years of alcohol dependence, binge duration (days), hours since last drink, average daily alcohol intake (g), maximum daily tolerance, and average daily tolerance. Therefore, their effects are not reported in the text. Administration of benzodiazepine to compensate for lack of efficacy of the research medication was assessed on a daily basis using contingency tables and analyzed with Fisher's exact test. All analyses were performed on an intent-to-treat basis.
As presented in Table 1, the medication groups did not differ significantly in their baseline characteristics. Also, there were no group differences in day 1 withdrawal severity as reflected by the score on CIWA-Ar [F(4, 122)=0.15, p=0.96] or total AWSC [F(4, 122)=0.74, p=0.57].
Table 1. Demographics and Clinical Characteristics
Number of subjects
Age (y; M ± SE)
AWS presence (y; M ± SE)
Binge duration (d; M ± SE)
Period since last drink (h; M ± SE)
Average daily ethanol: g/standard drinksa (M ± SE)
Comparison With Placebo. The effects of drug [F(4, 122)=3.85, p=0.006], time [F(2, 122)=7.07, p<0.0001], and drug by time [F(8, 122)=5.04, p<0.0001] were significant. When averaged over time, only diazepam [t(122)=−2.59, Dunnett-adjusted p=0.04], and lamotrigine [t(122)=−3.81, Dunnett-adjusted p=0.0008] were significantly superior to placebo in reducing CIWA-Ar scores (see Table 2 for 95% confidence intervals). Except for a marginal memantine effect at test day 2 (Bonferroni-adjusted p=0.07), all medications were significantly superior to placebo when compared at test days 2 and 3 (Bonferroni-adjusted p<0.01) (Fig. 1).
Table 2. Estimated Drug Effects and Adjusted 95% CI
Mean difference from placebo average over test days
Comparison Among the Active Medications. Excluding placebo, there was a significant drug by time interaction [F(6, 98)=2.27, p=0.04]. Averaged over time, there were no significant differences between the drugs (all Tukey-adjusted p>0.1). Further, no group differences were seen on test day 1. However, on both test days 2 and 3, lamotrigine-treated subjects had a significantly lower CIWA-Ar score than subjects treated with memantine and topiramate (Bonferroni-adjusted p<0.05). No medications differed significantly in efficacy from diazepam.
Self-Rated Withdrawal Assessment (AWSC)
AWSC Total Score
Comparison With Placebo. As shown in Fig. 2, all of the treatments were statistically superior to placebo. The effects of drug [F(4, 122)=8.93, p<0.0001], time [F(2, 122)=11.56, p<0.0001], and the interaction of drug and time [F(8, 122)=7.76, p<0.0001] were highly significant. Averaged over time, each drug showed at least marginal superiority to placebo (Dunnett-adjusted p<0.06) (Table 2). While there were no differences between each drug and placebo measured at test day 1, all drugs were significantly different from placebo on the second and third test days (Bonferroni-adjusted p<0.04).
Comparison Among the Active Drugs. In the analysis excluding placebo, the effects of drug [F(3, 98)=3.87, p<0.01], time [F(2, 98)=9.40, p<0.0001], and the interaction of drug and time [F(6, 98)=2.96, p<0.01] were significant. Averaged over time, lamotrigine-treated subjects had significantly lower AWSC scores than subjects receiving memantine [t(98)=−3.05, Tukey-adjusted p=0.02]. However, none of the medication groups differed significantly from diazepam. Further post hoc testing revealed no significant differences between the medication effects on day 1. However, on days 2 and 3, lamotrigine tended to have lower AWSC scores relative to memantine and topiramate (Bonferroni-adjusted p<0.07). These effects remained present after covariate adjustment.
AWSC Factor Scores
Autonomic Arousal. All of the active medications were efficacious relative to placebo as reflected by the significant interaction of drug by time effects [ATS(7.38)=12.0, p<0.0005]. Comparison among the active medications also revealed differences [drug by time interaction: ATS(5.59)=3.0, p=0.008]. Averaged over time, lamotrigine was superior to both memantine [ATS(1)=12.7, adjusted p=0.002] and topiramate [ATS(1)=14.8, adjusted p=0.0006] while diazepam was superior to topiramate [ATS(1)=6.8, adjusted p=0.05].
Depression. All of the active medications were also significantly superior to placebo in reducing withdrawal-related depressed mood [drug by time interaction: ATS(7.3)=8.43, p<0.0005]. Averaged over time, each drug was superior to placebo on this outcome measure (Bonferroni-adjusted p<0.01). There were no significant differences among the active medications in their efficacy on this outcome measure after adjusting for multiple comparisons.
GI Discomfort. There was a significant reduction in nausea and vomiting by all of the active medications [drug by time interaction: ATS(7.04)=2.84, p=0.03]. There were no significant differences among the active medications on this outcome measure.
Physical Tension. There was a significant drug by time interaction for this outcome measure [ATS(6.88)=2.89, p=0.03]. However, none of the active medications were significantly superior to placebo on this outcome measure.
Alcohol Craving. All of the active medications reduced alcohol craving significantly compared with placebo as reflected by a significant drug by time interaction effect [ATS(6.76)=4.89, p<0.0005] and by significant post hoc comparisons of the active agents with placebo when data were averaged over time (p<0.005 adjusted for multiple comparisons). Comparisons among the active medications did not reveal significant differences after adjusting for multiple comparisons.
Level of Dysphoric Mood (MADRS)
Comparison With Placebo. The effects of drug [F(4, 114)=9.28, p<0.0001], time [F(2, 228)=7.48, p<0.0001], and the interaction of drug and time [F(8,228)=11.3, p<0.0001] were significant (Fig. 3). All drugs were superior in reducing MADRS scores compared with placebo when data were averaged over time (Dunnett-adjusted p<0.003), on day 2 (Bonferroni-adjusted p<0.0005), and day 3 (Bonferroni-adjusted p<0.0005).
Comparison Among Active Medications. Excluding placebo, the effect of drug trended toward significance [F(3, 92)=2.69, p=0.05]; however, the drug by time interaction was not significant [F(6, 92)=1.39, p=0.2]. Again, inclusion of covariates did not affect the results.
Benzodiazepine Administration Due to Lack of Treatment Efficacy
Comparison With Placebo. A substantially higher percentage of patients treated with placebo received diazepam due to lack of treatment efficacy compared with the active medications (placebo: 88%, diazepam: 12%, lamotrigine: 20%, memantine: 27%, topiramate: 38%; Fisher's exact test, p<0.0001).
Comparison Among Active Drugs. When placebo was excluded from the analysis, there were no significant differences in necessity to administer benzodiazepines due to lack of treatment efficacy (Fisher's exact test, p=0.40). Similar results were found when comparing patterns from test day 1 with test day 3.
Medication Side Effects. No serious adverse events occurred during the course of this study. In addition, information on 10 specific side effects was obtained daily during the study based on side effects generally associated with these classes of medications (rash, asthenia, headache, nausea, dizziness, pain, itching, depression, sedation, and thought/concentration disturbance). Diazepam was the only medication associated with these side effects in this study. On the second study day, 5 (20%) subjects receiving diazepam reported at least 1 side effect [asthenia (n=1), dizziness (n=2), sedation (n=3), thought/concentration disturbance (n=1)]. On the third study day, 4 (16%) of diazepam-treated subjects reported at least 1 side effect [asthenia (n=1), dizziness (n=3), sedation (n=2), thought/concentration disturbance (n=1)].
This study provided the first systematic evidence for the clinical efficacy of antiglutamatergic pharmacologic supports for ethanol detoxification in moderately severely alcohol-dependent male patients who exhibited clinically significant ethanol withdrawal symptoms at the time of study entry. The principal findings from this study are that lamotrigine, memantine, and topiramate, drugs that, respectively, reduce glutamate release, block NMDA glutamate receptors, and reduce the function of AMPA/kainite receptors, all show efficacy for ethanol detoxification that is not different from diazepam, a drug that facilitates GABAA receptor function, and that is superior to placebo. This pattern was observed with observer-rated assessment (CIWA-Ar), self-rated assessment (AWSC), need for additional diazepam due to lack of efficacy, and dysphoric mood (MADRS). Further, the increased likelihood that an individual receiving placebo would receive supplemental diazepam relative to other groups may have led this study to underestimate the true differential efficacy of the active medications relative to placebo. This study provided preliminary evidence that diazepam was the study medication most frequently associated with side effects. Future studies will be needed to determine whether the antiglutamatergic medications provide superior tolerability than diazepam at doses that provide similarly effective suppression of alcohol withdrawal symptoms. Also, it will be important to determine whether these findings apply to alcohol-dependent women.
This study also provided suggestive evidence of subtle advantages of lamotrigine over memantine and topiramate in reducing observer-rated (CIWA-Ar) and self-rated (AWSC) withdrawal severity. This result might reflect an advantage of glutamate release inhibition over pharmacologic strategies that target individual glutamate receptor subpopulations. However, it is not clear that this inference can be drawn from this study. This study compared single doses of each drug. These doses were derived from our clinical experience with these medications and our interpretation of the published literature related to the safety and efficacy of these medications when prescribed for other indications. Further, the topiramate dose selected for this study had previously been shown to suppress alcohol withdrawal seizures (Rustembegovic et al., 2002) and it was twice the dose used to treat uncomplicated alcohol withdrawal symptoms (Choi et al., 2005). However, it is not clear that the doses of each medication in this study were the optimal doses for ethanol detoxification. Lamotrigine, memantine, and topiramate were prescribed at doses below their maximum tolerated doses in the general population. In this study, this step was taken to improve the expected tolerability of each drug. However, it is possible that optimizing the dose of each medication would alter the relative efficacy of the antiglutamatergic medications in this study.
In addition, it is possible that the mechanisms evaluated in this study could have been probed with alternative medications that might have produced different findings. In glutamate release inhibition, alternative medications include the cation channel antagonist, riluzole, group II metabotropic glutamate receptor (mGluR) agonists, or cannabinoid receptor agonists. Alternative NMDA receptor antagonists included memantine-like drugs (amantadine, neramexane), low-potency NMDA receptor antagonists (dextromethorphan), NMDA receptor subtype selective antagonists, competitive NMDA receptor antagonists, and the psychotigenic uncompetitive NMDA receptor antagonists (ketamine). Selective AMPA and kainite glutamate receptor antagonists might also be studied, when available for human administration. However, this study did not explore all conceivable glutamate receptor-related targets, only those classes with representative medications currently available to American or Russian physicians. For this reason, this study did not explore group I mGluR antagonists or potent and selective glycineB receptor antagonists.
All of the active medications in this study showed evidence of safety and efficacy. No patients were withdrawn from the study due to untoward medication effects. It remains to be seen whether antiglutamatergic medications will be superior to diazepam in impact on relapse to drinking following detoxification or in preventing withdrawal-related neuroplasticity (Malcolm et al., 2002). Also, it is not clear whether antiglutamatergic detoxification strategies are superior to other anticonvulsant detoxification strategies (Johnson et al., 2004). However, we now have a growing number of potential alternative approaches to benzodiazepine-assisted ethanol detoxification that appear to show comparable efficacy. Thus, there are a growing number of alternative approaches to alcohol detoxification that await definitive evaluation of their efficacy, safety, and cost effectiveness.
The authors thank the Clinical Staff of the St. Petersburg Regional Center of Addictions and Psychopharmacology for their expert contributions to the success of this study. Dr. Krystal is a paid consultant to Merz Pharmaceutical and Forest Laboratories, makers of memantine; Glaxo SmithKline, the maker of lamotrigine; and Johnson and Johnson Pharmaceutical, the maker of topiramate. These companies provided no financial support or drug for this study. This study is registered with ClinicalTrials.gov under the following ID: RCT-RCT_LOND_GAT_AWS_2003.