Pharmacological Treatment of Alcohol Dependence:  A Review of the Evidence FREE

Alcohol dependence, or alcoholism, is often a progressive chronic disorder that both the American Medical Association and the American Psychiatric Association recognize as a disease.¹ It is a common disorder posing a heavy burden on patients, their families, and society. Epidemiological studies indicate that alcoholism will affect approximately 10% of Americans at some time in their lives, with men affected more than women.^2- 5 This is a high prevalence rate compared with many other diseases and highlights the public health significance of alcohol dependence. In addition to the estimated 100,000 US citizens who die each year because of alcohol-related causes, including traffic collisions and cirrhosis of the liver,⁶ alcoholism costs the nation an estimated $166 billion annually in direct and indirect health and social costs.⁷

Treating alcoholism involves several stages. The initial stage deals with acute withdrawal and detoxification.⁸ Later stages attempt to maintain patients in remission and develop a lifestyle compatible with long-term abstinence (ie, avoidance of any alcohol consumption). This latter aspect of patient management has traditionally involved psychosocial interventions (including Alcoholics Anonymous and various counseling approaches), pharmacological modalities (primarily disulfiram), and combinations of both. Although many individuals do achieve long-term sobriety with treatment, others continue to relapse and deteriorate despite multiple courses of treatment. Pharmacological treatment is emerging as another means to enhance abstinence and prevent relapse, complementing psychosocial interventions that have been in use for many years. The findings we report here are part of a rigorous systematic review of the relevant scientific literature on the pharmacological treatment of alcoholism conducted by the Research Triangle Institute–University of North Carolina at Chapel Hill Evidence-based Practice Center and sponsored by the Agency for Health Care Policy and Research (AHCPR).⁹

We conducted a detailed search of the scientific literature on 5 classes of drugs.⁹ We focused on early drugs used for alcoholism (disulfiram and lithium) and pharmacotherapies that have been used more recently (the opiate antagonists naltrexone and nalmefene, acamprosate, and various serotonergic agents) (Table 1). These classes of medications represent the major agents used or recently tested for the treatment of alcohol dependence after alcohol withdrawal.

With the advice of a technical expert advisory group (see "Acknowledgment" section), we developed study questions and methods. We reviewed the efficacy of these agents for reduction in the number of standard drinks of alcohol, reduction in the number of drinking days (or increase in the number of nondrinking days), overall resumption of drinking, number of episodes of heavy drinking, time to first drink, reduction in the rate of relapse determined by a priori criteria, and craving. Our results are drawn largely from intention-to-treat analyses; the full evidence report⁹ provides information on subgroup and follow-up analyses, occurrence rates of adverse effects, and patient compliance.

Literature SearchedTable 2 shows the inclusion and exclusion criteria for articles selected for review. We used the following databases: MEDLINE, HealthSTAR, the American Society of Health-System Pharmacists International Pharmaceutical Abstracts database, EMBASE, Alcohol and Alcohol Problems Database, and PsycINFO. Search strategies included all medications of interest and related classes of medications, alcoholism, alcohol drinking, and study and design characteristics. We also reviewed materials available from the Cochrane Collaboration, published abstracts, and US Food and Drug Administration documents.

To obtain data for our evidence tables, we developed a data extraction form, a follow-up results form, a comorbid study results form, an adverse-effect profile, and an extraction guide. We used dual abstraction procedures (1 content reviewer and 1 methods reviewer); the principal investigators (S.L.W. and J.C.G.) completed a third conflict-resolution review.

To assess the quality of individual articles, reviewers completed a quality rating form covering study design, diagnostic and outcome measures, statistical analyses, and discussion. The form used standard methodological items, such as loss to follow-up, blinding, and comparability between treatment groups. It also included items specific to alcohol dependence, such as authors' alcoholism criteria and method of assessment. The quality score was independent of the article's findings on efficacy of the pharmacological intervention. Individual scores summed to 40 points and were rescaled to 100 points (scale available from the authors on request). We judged the overall strength of the evidence on efficacy based on the consistency across studies, quality of studies, sample size, and magnitude of effects (Table 3).

After applying our inclusion and exclusion criteria and reviewing the titles and/or abstracts of 375 articles, we fully abstracted and analyzed data for 41 studies and 11 follow-up or subgroup evaluations, including 11 disulfiram studies, with 1 subgroup publication; 1 nalmefene article; 3 naltrexone studies, with 5 subgroup publications; 9 acamprosate articles; 9 serotonergic agent studies not restricted to comorbid populations, with 1 subgroup study; 3 serotonergic agent studies restricted to persons with coexisting depression or anxiety, with 1 subgroup analysis; and 6 lithium articles, with 3 subgroup analyses. One article addressed both fluoxetine and acamprosate. We did not exclude any articles from analysis on the basis of poor quality ratings; this judgment is reflected in the discussions and rating of the strength of evidence.

Table 4 shows the primary findings with evidence graded on the A-B-C-I scale defined in Table 3. When multiple citations refer to the same trial, only the original is included in Table 4. We included in Table 4 only those studies that provided information on standard alcohol outcomes, including drinking days, return to drinking, time to first drink, alcohol consumed per unit of time, episodes of heavy drinking, craving, or relapse.

Disulfiram. Although the literature is substantial on use of this drug to treat alcoholism (135 studies), the number of controlled clinical trials is limited; of the 11 trials we reviewed, 5 concerned oral disulfiram^10- 15 and the remainder studied disulfiram implants.^16- 21

Oral Disulfiram (Maintenance Dosage, 200-250 mg/d). The 5 oral disulfiram studies involved 1207 subjects; 3 were placebo controlled and used double-blind random assignment.^10- 11,13 Outcome measures and results varied; they provide modest evidence that disulfiram reduces drinking frequencies without significantly enhancing abstinence rates (Table 4). The only group that studied supervised administration of the drug found that subjects taking disulfiram had significantly improved outcomes in drinking frequency and amount consumed¹⁴; however, the trial did not maintain double-blind standards (subjects were told after randomization if they were receiving disulfiram).

Disulfiram Implants (1-Time Dose, 800-1000 mg). Although implants of disulfiram appear to be an attractive method to ensure compliance, the bioavailability of disulfiram following implantation has not been demonstrated in the therapeutic trials. Six studies (combined enrollment, n = 282) used disulfiram implants; 5 used a placebo, most commonly a saline injection.^17- 21 The outcome measures reflect diverse results (Table 4).

We judged the efficacy evidence for disulfiram (modalities combined) as warranting a grade of B—inconsistent but positive evidence that disulfiram reduces the number of drinking days, and inconsistent but more often negative evidence about other outcome measures.

Opiate Antagonists. Of the 77 research articles identified, 3 controlled studies evaluated the efficacy of naltrexone in alcohol-dependent patients^22- 24 and 1 pilot study evaluated nalmefene.²⁵ We also reviewed substudies of these trials.^26- 30

Naltrexone (50 mg/d). Although the overall sample size for the 3 naltrexone studies is modest (n = 271 randomized patients), the trials are recent and of good quality, with consistent results on relapse rates and drinking frequencies (Table 4). One study investigated the effect of both naltrexone vs placebo and supportive therapy vs coping skills therapy, thus providing the only work to address the interaction between pharmacological and psychosocial interventions.²² Naltrexone had a positive effect (on abstinence) only when combined with supportive psychotherapy; patients receiving naltrexone and coping skills therapy showed no such effect. Other outcomes show a more consistent pattern (Table 4). Among all patients in the first 2 trials,^22- 23 relapse rates at the end of the trials were higher for the placebo groups (54%²³ and approximately 80%²²) than for the naltrexone groups (23%²³ and approximately 40%²²). In a third trial,²⁴ end-of-study relapse rates for all subjects were 53% and 35% for placebo and naltrexone patients, respectively; however, for compliant patients, the figures were 52% and 14%. Of 2 studies that measured a biological marker for alcohol consumption (γ-glutamyl transpeptidase [GGTP]), one found that naltrexone recipients had significantly lower GGTP levels at the end of the trial²⁴ and the other earlier trial had a nonsignificant trend in that direction.²³ O'Malley et al²⁸ studied patients for an additional 6 months after trial completion and found that compared with subjects receiving placebo, those receiving naltrexone had lower rates of heavy drinking. Reports on subsets of patients who sampled alcohol found that naltrexone diminished the subjective "high" after alcohol consumption.^24,26,29

Nalmefene (10-40 mg/d). The nalmefene trial randomized 21 subjects to placebo and 2 nalmefene dosage groups; 8 subjects completed the study.²⁵ The significant finding was a lower relapse rate in the high-dosage nalmefene group (data not shown).

The high quality of these trials coupled with the consistent finding for an effect of naltrexone in reducing relapse rates and drinking frequencies warranted a grade of A; we gave nalmefene a grade of I.

Acamprosate (1300-2000 mg/d). Of the 37 acamprosate studies identified in our literature search, we abstracted data for 9.^31- 39 A total of 2170 patients were enrolled; all trials used a double-blind method and 1 used a crossover design.³¹

The most consistent finding in the acamprosate trials has been its effect on drinking frequency (Table 4); nondrinking days were typically increased by 30% to 50% (eg, from 162 to 225 nondrinking days³⁵). Several studies also found that acamprosate approximately doubled abstinence rates, although a majority of patients returned to drinking while taking acamprosate. The acamprosate trials did not assess relapse rates. Of 4 trials assessing GGTP, 2 reported improvements among acamprosate recipients,^33,39 1 indicated a trend toward improvement,³⁴ and 1 reported no effect.³⁵

Among subjects followed up for 48 weeks after trial completion, patients taking acamprosate showed a greater number of cumulative nondrinking days than did those receiving placebo since entry into the trial (387 vs 250 days) and a higher abstinence rate (40% vs 17%).³⁵ Among subjects followed up for 12 months after trial completion, acamprosate continued to exert a positive effect on abstinence rates but not number of nondrinking days.³⁶ Acamprosate recipients had higher rates of abstinence at 6 months after treatment than controls (43% vs 30%).³⁹ Ladewig et al³² did not find a carryover effect for acamprosate up to 6 months after active treatment had ended, but this group also failed to find any effect for acamprosate during the latter part of their active trial.

Based on the large number of subjects and the consistent finding in the trials that acamprosate reduces drinking frequency, we gave acamprosate an efficacy rating of A.

Serotonergic Agents (Dosages Varied by Drug). We reviewed 3 classes of serotonergic agents: selective serotonin reuptake inhibitors (SSRIs, principally fluoxetine), buspirone hydrochloride, and ondansetron hydrochloride (Table 1). The US Food and Drug Administration approved SSRIs for use in depression and anxiety disorders, buspirone for anxiety, and ondansetron for nausea. In recent years, these drugs have been evaluated off label in alcohol dependence, especially when patients exhibit comorbid mood or anxiety disorders. Thus, an adequate examination of their effects on alcohol dependence must take into account whether an affective or anxiety disorder is present. We divided trials (Table 4) into those that studied alcohol-dependent individuals with varying rates of psychiatric comorbidity (including none)^31,40- 47 and those that prospectively selected alcohol-dependent subjects with a known depressive or anxiety disorder.^48- 50

Trials in Subjects With Varying Rates of Comorbid Depression or Anxiety. The SSRI fluoxetine has been studied in 5 double-blind placebo-controlled trials (n = 227 subjects); 1 used a crossover design,³¹ 2 used randomization,^41- 42 and 2 did not use randomization.^40,43 Outcomes are mixed regarding the efficacy of fluoxetine for alcohol-dependent patients. Citalopram had no effect on GGTP or episodes of heavy drinking.⁴⁶ The 1 trial of buspirone (n = 57 subjects, using a double-blind placebo-controlled design) showed no therapeutic effect on several measures, including GGTP levels.⁴⁷ One trial of ondansetron (n = 71 subjects, using a double-blind placebo-controlled design, with random assignment to high dosage [4 mg/d] and low dosage [0.5 mg/d]) found that low-dosage ondansetron reduced drinking only among light drinkers.⁴⁵

Trials of Alcohol-Dependent Subjects With Depression or Anxiety. Trials among patients who also had depression or anxiety totaled 179 subjects (n = 51 for a fluoxetine trial⁵⁰; n = 128 for 2 buspirone trials^48- 49). Depressed alcohol-dependent patients taking fluoxetine showed improved amount and frequency of drinking compared with those receiving placebo; the buspirone studies showed only 1 positive change in drinking outcomes (time to first drink).

Generally, the studies of fluoxetine involved a modest number of patients and results are at best mixed, as is true of the buspirone and ondansetron studies. We judged that the limited nature of these data warranted efficacy ratings of I.

Lithium (Dosages and Blood Levels Varied). Six trials and 3 related substudies (n = 823 subjects) have appeared on the effects of lithium in alcohol-dependent subjects with and without depression or bipolar disorder^51- 59; about 55% of all subjects were studied in the large cooperative trial sponsored by the Department of Veterans Affairs.⁵⁸ All trials (Table 4) reported comorbid depression in subjects; only 2 controlled for depression in their data analyses.^52,58 The strongest and largest trial found no positive effect for lithium in either depressed or nondepressed alcoholic-dependent subjects.⁵⁸ One study (not controlling for comorbidity) found improved drinking outcomes in subjects with high lithium levels (compliant by definition).⁵⁷

Based on the negative findings of the large well-conducted study and the fact that all positive findings with lithium were limited to studies that did not control for comorbidity, we gave lithium an efficacy rating of C.

Compliance. In a clinical trial, the extent to which patients are compliant may affect whether a drug effect will be found. In the reviewed studies, investigators monitored compliance in diverse ways, including self-report, proxy report, pill counts at return visits, urine markers, and blood levels of active medication. The effect of compliance on outcome measures was rarely analyzed. In 2 cases in which compliance was studied, Fuller et al¹³ found that compliant subjects had improved outcomes regardless of assigned drug and Volpicelli et al²⁴ found a positive outcome for naltrexone only in compliant subjects.

Adverse Events. In assessing the usefulness of a medication, clinicians must balance efficacy against risks (adverse effects, morbidity, and mortality). We noted risks (when described in the trials) and gathered information from 1 large safety and tolerability study of naltrexone.⁶⁰ Given the incomplete picture of risk provided by clinical trials, the risk-benefit ratio for any of the reviewed medications cannot be fully specified. Disulfiram, lithium, SSRIs, and buspirone, although generally well tolerated, are associated with adverse effects and occasional morbidity. These trials reported no deaths attributed to any of these agents; dropout rates due to adverse effects, when reported, were low. In the naltrexone trials, no mortality was reported and the most common adverse effects were nausea and dizziness. The naltrexone safety and tolerability study⁶⁰ reported higher rates for 2 adverse effects, new-onset nausea (9.8% for naltrexone vs 0% for nonmedication group) and headache (6.6% for naltrexone vs 1.7% for nonmedication group) and a 15% discontinuance rate for naltrexone subjects because of adverse effects. In the acamprosate trials, no mortality attributed to medication was reported. The most common adverse effect was diarrhea; dizziness, itching, and increased sexual desire were also significantly greater in the active vs placebo groups. Dropout rates were less than 1% for any 1 adverse effect.

The pharmacotherapy of alcohol dependence has significantly changed during the past few years. The discovery of new medications may positively affect core components of the alcohol dependence syndrome, such as loss of control of drinking, preoccupation with alcohol use, or protracted abstinence symptoms. The development of new classes of medicines specifically effective for the primary symptoms of alcohol dependence represents an advance of historic importance.

Results from our evidence report⁹ that are reviewed here are noteworthy in 2 broad areas: (1) understanding the clinical implications of the efficacy of the reviewed medications and (2) addressing the remaining gaps in the knowledge base. No study examined the efficacy of these agents compared with each other; variations in study design, duration of follow-up, and outcome measures prevented comparison of these medications for relative efficacy.

Although we attempted to identify the nature and extent of psychosocial interventions based on information provided by the authors, typically these interventions were poorly described. Thus, the literature did not allow us to evaluate the important question of efficacy of pharmacotherapy when used in conjunction with varying types and intensities of psychosocial therapies.

Naltrexone and Acamprosate. The interest generated by recent reports documenting that naltrexone and acamprosate are more effective than placebo in the treatment of alcoholism appears warranted by the strength of the evidence (evidence grade, A), showing their efficacy among alcohol-dependent patients. The naltrexone trials involved a modest number of subjects (n = 271), but the quality of the trials was high and the findings consistently showed that naltrexone reduced drinking frequency and relapse rates, controlling for compliance. Although requiring confirmation, naltrexone's primary effect appears to be its ability to reduce the patient's emotional response to alcohol consumption; this may deter patients who sample alcohol from progressing to heavy drinking (relapse) and decrease alcohol intake over time. Other actions of naltrexone are likely operative, given the evidence that abstinence was enhanced in 2 of the 3 trials. The clinical impact of naltrexone is not yet clear—for some patients it may be profound, for others it may bring no benefit. Current work will expand this knowledge base considerably; 18 National Institute on Alcohol Abuse and Alcoholism (NIAAA)–sponsored trials targeting approximately 3000 subjects and 1 multisite Department of Veterans Affairs study involving 600 subjects are under way or just completed.

The acamprosate trials involved a large number of subjects (n = 2170), were of moderate to good quality, and consistently showed positive effects on reducing drinking frequency, with some evidence of enhancing abstinence. The acamprosate trials have been limited to European populations, although results from a just-completed US-based, placebo-controlled, double-blind, multisite trial sponsored by Lipha Pharmaceuticals Inc, New York, NY (n = ≈450 subjects), should be published within the next year. To date, acamprosate has not been used clinically in the United States; its impact on US practice is not yet clear.

Disulfiram. The disulfiram trials (evidence grade, B) were less consistent in both quality and results; they provided moderate evidence of reduced drinking frequency but no evidence of enhanced abstinence. Supervised disulfiram administration has been posited as a particularly effective treatment method,⁶¹ but the only placebo-controlled trial with this approach did not maintain double-blind conditions.¹⁴ The bioavailability of implanted disulfiram was not shown; alcohol challenges after implantation did not produce a disulfiram-ethanol reaction.

Disulfiram presents an interesting clinical dilemma. It has been used for close to 50 years and many practitioners believe it has great value. Nevertheless, the controlled clinical trials yield ambiguous conclusions about its efficacy. Is this divergence between clinical practice and clinical trials a real phenomenon? On one hand, given the appropriate subjects and research design (eg, supervised administration), disulfiram might demonstrate greater efficacy than in the data we reviewed. On the other hand, a large portion of the disulfiram effect is related to fear of a disulfiram-ethanol reaction; hence, in a placebo-controlled trial, an active drug effect would be more difficult to ascertain.

Serotonergic Agents and Lithium. The controlled studies of serotonergic agents were limited and yielded mixed results (evidence grade, I). Coexisting mood and anxiety disorders greatly complicate interpretation of the findings, but the limited evidence available does not support the use of serotonergic agents for treating primary alcohol-dependent patients. The 2 studies that evaluated buspirone in alcohol-dependent subjects with anxiety had primarily negative effects, although 1 was suggestive for some positive effects.⁴⁹ The 1 fluoxetine trial in alcohol-dependent subjects with depression was positive.⁵⁰ This minimal database on these medications will be greatly expanded by a Pfizer Inc (New York, NY) multisite trial of the SSRI sertraline hydrochloride for clinically depressed alcohol-dependent subjects and by 7 NIAAA trials (n = ≈1100 subjects).

The controlled trials of lithium did not demonstrate efficacy in either nondepressed or depressed alcohol-dependent subjects (evidence grade, C). We concluded that the literature does not support use of lithium for treatment of primary alcohol dependence.

Our evidence tables document significant variability in trial design.⁹ First, the alcohol outcomes measured in each study vary substantially. Of the work reviewed, 24 measured number of drinking/nondrinking days; 22, abstinence; 14, drinks per unit of time; 13, time to first drink; 12, at least 1 biological measure (eg, GGTP); and 6, a criteria-based definition of relapse. Second, the meaning and use of the term relapse vary considerably. In the naltrexone trials, a relapse is a bout of heavy use with an a priori definition, such as consuming at least 5 drinks per occasion (for men), whereas in the European acamprosate trials, relapse refers to return to any drinking. Third, the concept of craving is not often addressed; of 41 studies reviewed, 17 measured craving using many different assessment techniques. Fourth, compliance with medication was irregularly addressed. Fifth, the type and intensity of psychosocial treatment, as well as compliance with that treatment, were typically not described. Moreover, the presence of psychiatric disorders was sometimes not mentioned or, when acknowledged, was not controlled for in the analysis. Sixth, external validity and generalizability are problems because subjects were commonly recruited from specialized substance abuse settings; the requirements of many protocols would inevitably produce subject selection bias. In addition, dropout rates were, with some exceptions, in the 20% to 50% range and sometimes higher.

The lack of consistency in research design across these trials is a major issue that prohibits cross-study comparisons, and it deterred us from conducting meta-analyses of the pooled efficacy data in these studies. These methodological problems appear to be improving, but additional standardization (eg, uniform definitions, recommended core outcome measures for clinical trials) would clearly be an important advance, to which funding and professional organizations such as the NIAAA, the Research Society on Alcoholism, and the American Society of Addiction Medicine could contribute greatly.

Efficacy trials examine whether a health intervention works in an ideal treatment setting; effectiveness studies focus on the impact of an intervention in everyday settings. Now that efficacy has been established for at least some of the medications reviewed here, effectiveness issues in primary care settings should be examined, including:

The work we reviewed invariably combined medications with some type of psychosocial intervention. Thus, these medications have been tested only in conjunction with psychosocial interventions; their effectiveness used in isolation is unknown. Although 2 medications, naltrexone and acamprosate, did bring clinically and statistically significant improvement compared with placebo in drinking outcomes, most subjects did resume drinking during treatment. Clinicians and researchers remain challenged by the problem of alcohol dependence; these pharmaceutical agents are only the beginning of improved treatment strategies.

The pharmacotherapy of alcohol dependence is experiencing a major shift in direction. Traditional medications (such as disulfiram) that show limited efficacy may be used less frequently, although they may still have value. New medications, including naltrexone and acamprosate, are likely to be used at increasing rates, although additional clinical experience is needed to ascertain their range of effectiveness. We herald the advances in the pharmacotherapy of alcohol dependence as valuable developments, but we note that treatment of patients with alcoholism should continue to incorporate a biopsychosocial perspective in an effort to change a life from a pattern of addiction to a pattern of sobriety and improved physical, mental, and social health.