This publication is provided for historical reference only and the information may be out of date.
Based on the data in the evidence tables, this chapter presents the major findings from our review of the literature on pharmacotherapies for alcohol dependence. This chapter begins with the definitions for grading the efficacy and harms data.
Subsequent sections for each key therapy provide a detailed discussion of four issues: the efficacy results (benefits), subgroup and followup studies, limitations, and side effect data (harms). At the end of each key therapy section, we grade the efficacy and harms literature.
The development and definition of our grading system have been discussed previously (see Chapter 2, "Grading the Evidence").
The disulfiram literature that addresses efficacy consists of six implant studies and five studies that use orally ingested disulfiram versus a placebo or no-drug comparison (Evidence Tables 1 and 2). One of the five studies of orally ingested disulfiram was a retrospective cohort design that evaluated social stability and control of drinking, comparing disulfiram to a no-drug control group (Bischof, Bucher, Battig, et al., 1995). This study will not be discussed in detail because its main outcome did not address drug efficacy.
Main Efficacy Results
Six controlled studies of disulfiram implants have been conducted over the past 24 years. Disulfiram implant studies have used disulfiram tablets inserted under the skin, commonly in the abdominal area. Only two studies assessed the blood level of disulfiram after implantation, and neither found detectable levels (Wilson, Davidson, Blanchard, et al., 1978; Wilson, Davidson, and Blanchard, 1980). Two studies (Wilson, Davidson, and White, 1976; Wilson, Davidson, Blanchard, et al., 1978.) performed ethanol challenges 5 days after implantation and failed to find evidence for a disulfiram-ethanol reaction; one study (Johnsen, Stowell, Bache-Wiig, et al., 1987) performed ethanol challenges at 1, 2, 4, 8, and 13 weeks after implantation and also failed to find evidence for a disulfiram-ethanol reaction in the disulfiram implant subjects. Interestingly, in the two reports from the Wilson team (Wilson, Davidson, and White, 1976;Wilson, Davidson, Blanchard, et al., 1978), subjects did not experience a reaction to ethanol under controlled conditions, but many more of the disulfiram-implanted patients than sham-implanted patients were reported to have had reactions to drinking in uncontrolled settings. Often, investigators inserted saline under the skin of placebo subjects to simulate a tablet implantation.
Several different alcohol outcomes have been evaluated to assess the efficacy of disulfiram implants. These included: mean number of standard drinks, mean number of drinking/nondrinking days over the trial period, mean number of heavy drinking episodes, abstinence rate, time to first drink, and biomarkers such as serum gamma-glutamyl transpeptidase (GGT) and mean corpuscular volume (MCV). Not all alcohol outcomes were assessed in all studies.
The data for mean number of standard drinks per day was measured as the grams per day in two studies (Johnsen, Stowell, Bache-Wiig, et al., 1987; Johnsen and Morland, 1991). These two studies reported a decline in the mean grams ingested per day for both the disulfiram and placebo groups, but the difference in the change in consumption during the trial between the two groups (disulfiram vs. placebo) was not statistically significant.
The alcohol outcomes used to address drinking/nondrinking days were inconsistent across the disulfiram implant literature. For example, Johnsen, Stowell, Bache-Wiig, et al. (1987) measured the mean number of moderate-drinking weeks both at baseline (placebo: 24.3; disulfiram: 22.5) and at the end of the 20-week trial (placebo: 8.5; disulfiram: 4.0) and reported no statistically significant differences between the intervention and placebo groups. In a later study (Johnsen and Morland, 1991), the authors reported the mean number of nondrinking weeks during the 40-week study (25.6 and 29.7 for disulfiram and placebo, respectively), a nonsignificant difference. In a third implant study, Wilson, Davidson, and Blanchard (1980) reported the mean number of nondrinking days during the 48-week study. The authors found a statistically significant difference (p<0.01) when comparing the mean number of nondrinking days for the disulfiram group (n=361) with that for the placebo implant group (n=307) and the two additional control groups (no operation and pseudocontrols), which had 24 and 31 nondrinking days, respectively.
The two Johnsen (1987 and 1991) studies also reported the number of episodes of heavy drinking. The earlier article found a study effect, i.e., the number of episodes of heavy drinking (as defined by the patient, not by standard criteria) was reduced in both groups, but the reduction for the disulfiram group was not greater than that for the placebo group. The later study did not show a difference in heavy drinking episodes over the trial period.
Three studies examined the effects of disulfiram implants on the rates of resumption of drinking (Wilson, Davidson, and White, 1976;Wilson, Davidson, and Blanchard, 1980;Johnsen and Morland, 1991). No statistical tests were reported, and the difference between the placebo and disulfiram groups on rates of abstinence did not appear to be clinically significant.
The time between study initiation and first drink was evaluated in four studies (Whyte and O'Brien, 1974 ;Wilson, Davidson, and White, 1976; Johnsen, Stowell, Bache-Wiig, et al., 1987; Johnsen and Morland, 1991). Only Whyte and O'Brien (1974) found a statistically significant increase (p<0.001) in time to first drink for the disulfiram implant group (5.4 months) vs. the no-drug control group (1.9 months).
Two studies examined the effects of disulfiram implants on biological measures of alcohol consumption (Johnsen, Stowell, Bache-Wiig, et al., 1987; Johnsen and Morland, 1991). In the later study, change in MCV during the trial was not different between the disulfiram and placebo groups, but GGT was improved in the disulfiram group as compared with placebo (p<0.02), a result that was not shown in the earlier Johnsen, Stowell, Bache-Wiig, et al. (1987) study.
Four controlled trials of oral disulfiram versus placebo or no-drug were reviewed (Fuller and Roth, 1979; Schuckit, 1985; Fuller, Branchey, Brightwell, et al., 1986; Chick, Gough, Falkowski, et al., 1992). With a total sample size over all trials of 1,207 randomized patients, these trials are among the larger studies of alcohol-dependent patients.
Only two studies investigated a standard unit of drinking, either as units per week (Chick, Gough, Falkowski, et al., 1992) or as drinks per drinking day (Schuckit, 1985). Chick, Gough, Falkowski, et al. (1992) reported a statistically significant decrease (p=0.05) in the number of units ingested at trial end (24 weeks), but Schuckit (1985) did not report a change from baseline in either the treatment or placebo groups.
All four studies evaluated the effect of disulfiram on some variant of drinking days, either nondrinking days (Chick, Gough, Falkowski, et al., 1992), percentage of drinking days during the year (Fuller and Roth, 1979), drinking days per month (Schuckit, 1985), or drinking days over the 52-week trial (Fuller, Branchey, Brightwell, et al., 1986). Only two trials found a statistically significant improvement for the disulfiram group compared with the placebo group (Fuller, Branchey, Brightwell, et al., 1986; Chick, Gough, Falkowski, et al., 1992).
Of the two trials that investigated the percentage of patients who resumed drinking (Fuller and Roth, 1979; Fuller, Branchey, Brightwell, et al., 1986), neither found that disulfiram improved the abstinence rate compared with placebo.
Only the Chick, Gough, Falkowski, et al. (1992) study evaluated biomarkers as evidence of disulfiram efficacy. This group reported a statistically significant improvement in GGT for the disulfiram group (-21 IU/l change from baseline vs.+13 IU/l change for placebo [p=0.02]).
The two Fuller studies confirmed compliance with oral disulfiram by riboflavin (Fuller and Roth, 1979; Fuller, Branchey, Brightwell, et al., 1986). In the earlier research, medication compliance rates were generally low across the placebo, 1 mg/day disulfiram, and 250 mg/day disulfiram groups: 50 percent, 67 percent, and 57 percent, respectively. In contrast, for the 1986 work, compliance was very low for all three treatment groups; compliance for the higher dose disulfiram group was lower than that for either of the control groups (placebo or 1 mg/day disulfiram).
Subgroup and Followup Studies
The two Fuller studies evaluated rates of abstinence in compliant patients on oral disulfiram as a subgroup analysis (Fuller and Roth, 1979; Fuller, Branchey, Brightwell, et al., 1986). The 1979 study found that those who abstained were significantly more likely to have been compliant (70 percent) than those who returned to drinking (32 percent), regardless of treatment group. Similarly, in the 1986 study, men who were compliant experienced 1-year abstinence rates between 38 and 50 percent, whereas men who were noncompliant had abstinence rates between 6 and 10 percent, regardless of treatment group.
The main limitation of the disulfiram implant literature is the very small sample sizes for each of the studies, ranging from a low of 20 patients randomized to two groups to a high of 100 patients randomized to four groups. In addition, only two of the six implant studies assessed the bioavailability of disulfiram by measuring drug blood levels. As with many alcohol studies, alcohol consumption was unverified.
Of the four efficacy studies of oral disulfiram, the total number of patients who may have benefited from therapy was unclear. Two studies did not provide the number of patients who completed therapy (Fuller and Roth, 1979; Fuller, Branchey, Brightwell, et al., 1986) and a third study (Chick, Gough, Falkowski, et al., 1992) had a 46-percent dropout rate. Other limitations included nonrandom allocation to therapy (Schuckit, 1985) and unblinding of patients post-randomization (Chick, Gough, Falkowski, et al., 1992).
Disulfiram has been in use in treatment settings for more than 40 years. Over that time a number of side effects have been reported, including some that are considered to be serious (such as liver toxicity, peripheral neuropathy, and psychosis). In addition, the disulfiram-ethanol reaction can be serious and require medical intervention, especially in patients with other conditions such as esophageal varices. In the reviewed efficacy trials (Evidence Tables 3a and 3b), disulfiram appeared to be reasonably well tolerated. Side effects reported to occur significantly more often with the drug than with placebo included stiffness of neck, drowsiness, and sexual dysfunction. Chick, Gough, Falkowski, et al. (1992) reported a 3.1 percent dropout rate because of dizziness and nausea, 1.6 percent secondary to possible neuropathy, and 3.1 percent because of skin rash. Overall, the harms profile was low, although clinicians should be aware of the potential for rare, serious side effects.
Efficacy grade: B; Harms grade: Low.
Papers generated from three main studies and several substudies were examined for evidence of naltrexone efficacy in maintaining abstinence (Evidence Tables 4 and 5).
Main Efficacy Results
Only the study by O'Malley, Jaffe, Chang, et al. (1992) evaluated the standard number of drinks per drinking day among four treatment groups: naltrexone and coping therapy, naltrexone and supportive therapy, placebo and coping therapy, and placebo and supportive therapy. The authors reported that the naltrexone and coping therapy group consumed fewer drinks per drinking day (3.7 drinks) compared with the placebo groups (coping: 7.1 drinks; supportive: 6.5 drinks) or the naltrexone and supportive therapy group (6.4 drinks), a result that was not statistically significant for the entire study population but was for the treatment completers (p<0.05).
Of the three naltrexone studies that evaluated the alcohol outcome "abstinence" or "return to drinking" (O'Malley, Jaffe, Chang, et al., 1992; Volpicelli, Alterman, Hayashida, et al., 1992; Volpicelli, Rhines, Rhines, et al., 1997), only one showed a statistically significant benefit of naltrexone versus placebo (O'Malley, Jaffe, Chang, et al., 1992). In the O'Malley, Jaffe, Chang, et al. (1992) study, the patients who received naltrexone plus supportive therapy showed lower rates of return to drinking (39 percent) than those who received either naltrexone plus coping therapy (72 percent) or placebo plus either supportive (81 percent) or coping therapy (79 percent).
Reduction in drinking days during treatment also was evaluated in the three naltrexone studies (O'Malley, Jaffe, Chang, et al., 1992; Volpicelli, Alterman, Hayashida, et al., 1992; Volpicelli, Rhines, Rhines, et al., 1997). In Volpicelli and colleagues' 1992 study, compared with baseline, drinking days per week increased by a mean of 0.51 day in the placebo group but only by 0.09 day in the naltrexone group (p<0.03); at the end of the study, drinking days were 0.11 per week in the naltrexone group and 0.57 per week in the placebo group. A similar result was reported for the later Volpicelli, et al. (1997) study; percentage drinking days were 10.76 percent for the placebo group and 6.2 percent for the naltrexone group. This difference was statistically significant for the treatment completers (p=0.01) but not for the entire study sample. For the O'Malley, Jaffe, Chang, et al. (1992) study, regardless of the type of counseling, both naltrexone groups drank on fewer days during the trial than did either of the placebo groups (p<0.01).
All three naltrexone studies also examined craving, although they used different visual analog scales (VAS) to do so. In the O'Malley, Jaffe, Chang, et al. (1992) study, which used a 20-point VAS, the naltrexone and coping therapy groups exhibited the lowest rates of craving among the four intervention groups: 3.1 (naltrexone and coping), 5.3 (placebo and coping therapy), 4.7 (placebo and supportive therapy), and 4.4 (naltrexone and supportive therapy). Only the earlier of the two Volpicelli et al. studies (1992) demonstrated a statistically significant decrease in craving among subjects receiving naltrexone compared with those receiving placebo.
One of the stronger and more consistent findings from the naltrexone studies has been the drug's beneficial effect on relapse rates. In these studies, relapse was defined using very specific criteria (see Evidence Table 3). All three studies used five or more drinks on one occasion as a criterion, with the O'Malley, Jaffe, Chang, et al., (1992) trial using a slightly different definition for women (i.e., four or more drinks on one occasion). In the two Volpicelli, et al. (1992, 1997) articles, relapse also included occurrence of elevated breath alcohol levels (>100 mg/dL) or drinking on more than 5 days per week (1992 study only). In the earlier Volpicelli (1992) article and in the O'Malley, Jaffe, Chang, et al. (1992) article, naltrexone was significantly more efficacious than placebo based on relapse rates. In the 1992 Volpicelli study, relapse was 23 percent in the naltrexone group and 54 percent in the placebo group (p<0.01). In Volpolicelli et al.'s 1997 study, the relapse rate was 35 percent in the naltrexone group and 53 percent in the placebo group, a difference that was not statistically significant. However, analysis of the treatment completers from the later Volpicelli et al. (1997) study indicated that the relapse rate was statistically significantly lower in the naltrexone group (26 percent) compared with the placebo group (53 percent) (p=0.02).
Both Volpicelli et al. studies (1992 and 1997) also examined GGT levels over the course of treatment. In both studies, the naltrexone group had lower GGT values at the conclusion of the trial than did the placebo group, but only in the 1997 study did the difference reach statistical significance (p=0.03).
Various methods were used to measure compliance in these studies; one study used pill counts (Volpicelli, Rhines, Rhines, et al., 1997), and the other used riboflavin levels as a marker of use (O'Malley, Jaffe, Chang, et al., 1992). The O'Malley (1992) work found that 78 percent of the urine samples collected from subjects on placebo were positive for riboflavin, compared with 92 percent of those collected from subjects taking naltrexone. Although compliance was not specifically measured in the first Volpicelli et al. (1992) study, it was a major aspect of the 1997 study. In that study, the investigators used pill counts and patient self-report to determine the proportion of patients who were compliant for 90 percent or more of their visits - 47 percent of placebo patients and 58 percent of naltrexone patients were compliant (Volpicelli, Rhines, Rhines, et al., 1997).
Subgroup and Followup Studies
Three studies looked at subgroups and followup (O'Malley, Jaffe, Chang, et al., 1996; O'Malley, Jaffe, Rode, et al., 1996; Volpicelli, Watson, King, et al., 1995). O'Malley, Jaffe, Chang, et al., (1996) followed a subgroup of patients for 6 months after completion of a double-blind trial, while they were not taking medication. In examining the drinking outcomes of these patients over the entire 6-month followup period, they found that the naltrexone and coping therapy group had the lowest rates of heavy drinking in the followup period (43 percent) compared with the other three treatment groups (naltrexone and supportive [68 percent], placebo and coping [80 percent], and placebo and supportive [75 percent]) (p<0.05). Both the O'Malley and Volpicelli teams also studied a subgroup of their populations who had experienced relapse to examine their emotional response to having done so (O'Malley, Jaffe, Rode, et al., 1996; Volpicelli, Watson, King, et al., 1995). In both studies, patients receiving naltrexone reported statistically significant reduced craving or "highs" following alcohol consumption compared with patients in the placebo groups.
These studies were conducted with modest sample sizes - a total of 271 subjects across the three trials. In addition, patients were further allocated to placebo and drug groups, and the O'Malley, Jaffe, Chang, et al., 1992 study further subdivided patients into two types of psychosocial therapy. In addition, dropout rates in all three studies were high, resulting in small numbers of subjects completing the trials. Finally, compliance with drug therapy was assessed in only two of the three naltrexone studies (O'Malley, Jaffe, Chang, et al., 1992; Volpicelli, Rhines, Rhines, et al., 1997), a factor that appears to be important for understanding the benefits of therapy.
All three efficacy studies just discussed evaluated potential side effects of naltrexone therapy (Evidence Tables 6a and 6b). In the controlled trials, side effects that were more prevalent with naltrexone than placebo included nausea, dizziness, and weight loss. O'Malley, Jaffe, Chang, et al. (1992) reported that 9.6 percent of subjects receiving naltrexone dropped out of the study because of nausea or dizziness. The large safety and tolerability study (Croop, Faulkner, Labriola, et al., 1997) compared 570 subjects who were receiving naltrexone with 295 subjects in a nonmedicated reference group. New-onset nausea was reported in 9.8 percent of the naltrexone subjects compared with 0 percent of the reference subjects and headache in 6.6 percent of naltrexone subjects compared with 1.7 percent of reference subjects. In addition, 15.0 percent of subjects on naltrexone discontinued the study because of side effects. In short, no serious side effects were reported in any of the trials. Despite this, clinicians might evaluate patient comorbidities such as acute hepatitis, liver failure, or opiate abuse when prescribing naltrexone.
Efficacy grade: A; Harms grade: Low.
Only one published article addressed the efficacy of nalmefene for alcohol dependence (Mason, Ritvo, Morgan, et al., 1994). This pilot study began with 21 patients; 9 remained at the completion of the study.
Main Efficacy Results
This pilot study evaluated several different measures of return to drinking: mean number of standard drinks per drinking day, number of nondrinking days, and relapse where relapse was defined similarly to that for the naltrexone studies - five or more drinks on one occasion or drinking more than 5 days per week (Evidence Tables 4 and 5). Fewer of the patients in the higher-dose nalmefene treatment group met the criteria for relapse at the end of the 12-week study (two of seven) compared with the placebo (four of six) and lower-dose nalmefene groups (six of six) (p< 0.05). These investigators also evaluated craving but found no differences between the nalmefene and placebo groups.
Of the four opiate antagonist efficacy studies, this study had the most stringent criteria for compliance. Subjects who did not take at least 75 percent of their prescribed medications (based on pill count) for two consecutive visits were considered noncompliant and dropped from the study.
The most obvious limitation is the very small sample size, with only nine treatment completers. Compared with the other opiate antagonist studies that provided a structured counseling program, this pilot study only encouraged subjects to attend Alcoholics Anonymous.
The side effects of nalmefene therapy are derived from this study as well (Evidence Tables 6a and 6b). Headache, nausea, rash, and sleep disturbance were reported as side effects, with dizziness and rash as problems that caused study withdrawal. There were no statistically significant differences between treatment groups for any reported side effect.
Efficacy grade: Insufficient evidence; Harms grade: Insufficient evidence.
The efficacy data for acamprosate in this evidence report are derived from nine published clinical trials conducted in Europe (Evidence Tables 7 and 8). The primary alcohol measures in these studies include abstinence, cumulative drinking days, and time to initiation of drinking. Unlike many American trials (particularly of naltrexone), these studies did not use an a priori criterion for relapse.
Main Efficacy Results
Of the nine studies, seven evaluated the number of nondrinking days over the course of the trial. In all seven, acamprosate showed a statistically significant improvement over placebo (Ladewig, Knecht, Leher, et al., 1993; Paille, Guelfi, Perkins, et al., 1995; Sass, Soyka, Mann, et al., 1996; Whitworth, Fischer, Lesch, et al., 1996; Geerlings, Ansoms, and Van den Brink, 1997; Pelc, Verbanck, Le Bon, et al., 1997; Poldrugo, 1997).
Abstinence rates during the trials were evaluated in eight of the studies; the results are less consistent than those for nondrinking days. For example, four of seven trials (Sass, Soyka, Mann, et al., 1996; Whitworth, Fischer, Lesch, et al., 1996; Pelc, Verbanck, Le Bon, et al., 1997; Poldrugo, 1997) showed that significantly fewer patients taking acamprosate resumed drinking during the trial compared with those receiving placebo. Two additional trials (Ladewig, Knecht, Leher, et al., 1993; Paille, Guelfi, Perkins, et al., 1995) show advantage for acamprosate at selected time points within the trial but not for the overall study. The remaining two studies (Roussaux, Hers, and Ferauge, 1996; Geerlings, Ansoms, and Van den Brink, 1997) did not show statistically significant differences between acamprosate and placebo although in the Geerlings, et al. (1997) trial, the patients receiving acamprosate had higher abstinence rates compared with those on placebo (25 vs. 13 percent, p=0.06).
Five acamprosate studies examined average time to first drink. Four of the studies showed a statistically significant improvement for the acamprosate group; compared with patients on placebo, those taking acamprosate took longer to sample alcohol (Paille, Guelfi, Perkins, et al., 1995; Sass, Soyka, Mann, et al., 1996; Pelc, Verbanck, Le Bon, et al., 1997; Poldrugo, 1997). The Geerlings team noted no significant difference between placebo and acamprosate for time to first drink, although the trend was in the same direction as the other four studies.
Craving was evaluated in six of the nine acamprosate trials (Gerra, Caccavari, Delsignore, et al., 1992; Paille, Guelfi, Perkins, et al., 1995; Roussaux, Hers, and Ferauge, 1996; Sass, Soyka, Mann, et al., 1996; Geerlings, Ansoms, and Van den Brink, 1997; Pelc, Verbanck, Le Bon, et al., 1997). In only two of those trials are sufficient data available to examine the effects of acamprosate compared with placebo. Using a categorical craving scale, significantly more patients receiving acamprosate reported no desire for alcohol compared with those on placebo (Pelc, Verbanck, Le Bon, et al., 1997). The Sass team (1996) reported no statistically significant difference between placebo and acamprosate subjects based on a visual analog scale for craving. Reports from the Gerra (1992), Paille (1995), Roussaux (1996), and Poldrugo (1997) teams did not provide sufficient information to evaluate craving for alcohol.
Compliance with treatment was assessed by pill count in four trials (Paille, Guelfi, Perkins, et al., 1995; Roussaux, Hers, and Ferauge, 1996; Geerlings, Ansoms, and Van den Brink, 1997; Pelc, Verbanck, Le Bon, et al., 1997) and by pill count and a urine marker in one other trial (Sass, Soyka, Mann, et al., 1996). The Geerlings team (1997) found similar rates of treatment compliance in the acamprosate and placebo groups (86 and 88 percent, respectively). Compliance appeared to be high in the Sass (1996) study, but data were not reported by treatment group; however, the authors did indicate that pill counts did not differ between the acamprosate and placebo groups. Also, compliance rates in the acamprosate-treated patients differed between subjects who abstained and those who relapsed, with positive urines in 82 percent of those who abstained and in 67 percent of those who relapsed.
Mean corpuscular volume (MCV), carbohydrate-deficient transferrin (CDT), and GGT were measured in several of the acamprosate clinical trials (Paille, Guelfi, Perkins, et al., 1995; Roussaux, Hers, and Ferauge, 1996; Sass, Soyka, Mann, et al., 1996; Poldrugo, 1997). The Sass (1996) and Roussaux (1996) teams found no significant differences between the acamprosate and placebo groups for GGT, MCV, or CDT. The Paille group (1995) reported that the percentage of patients with GGT levels within the normal range was significantly higher for the acamprosate than the placebo group, at both 6 and 12 months followup. They also reported a statistically significant improvement in MCV for the acamprosate group over placebo at 6 months but not at 12 months. Poldrugo (1997) also reported a greater percentage of acamprosate patients with GGT within 1.3 times the upper limit of normal at 6 months compared with controls (p=0.0017).
Subgroup and Followup Studies
Four acamprosate studies reported on followup analyses to assess return to drinking after the trial concluded. The most promising showed a greater number of cumulative nondrinking days at both 12 and 24 months, where the 24-month result is the sum of the nondrinking days over the entire 24-month period (Sass, Soyka, Mann, et al., 1996). Patients taking acamprosate averaged 387 nondrinking days and those on placebo averaged 250 nondrinking days, a statistically significant difference (p<0.001). This study also examined the number of patients who resumed drinking at 24 months. Again, significant differences were noted in favor of acamprosate over placebo for a resumption of drinking, with rates of 60.0 and 82.7 percent, respectively (p<0.003). Similarly, Poldrugo (1997) reported significant findings for both nondrinking days and for abstinence at the 52-week followup. Patients on acamprosate reported a mean of 167.7 nondrinking days compared with 120.5 for those on placebo (p=0.014). Similarly, 56 percent of the acamprosate group and 70 percent of those on placebo had resumed drinking at 52 weeks (p=0.05). In a 12-month followup, Whitworth, Fischer, Lesch, et al. (1996) did not find a significant difference between the acamprosate and placebo groups for the number of nondrinking days. Although these investigators found that more patients remained abstinent in the acamprosate group (11.9 percent) than in the placebo group (4.9 percent) (p<0.05), the analysis conducted by Ladewig, Knecht, Leher, et al. (1993) failed to find an advantage for acamprosate vs. placebo at 90 or 180 days post-trial followup with respect to the number of subjects who remained abstinent.
The acamprosate studies have several limitations, most of which are generic to alcohol studies. They include high or unreported dropout rates, with some studies losing approximately 60 percent of their subjects. Although most of the studies did assess compliance with therapy based on pill counts, the extent of individual patient compliance often was not reported. In addition, most of the acamprosate studies either did not specify the type of psychosocial co-intervention or did not provide group and/or individual counseling. Lastly, alcohol consumption was typically measured by self-report, a technique where underreporting is a major concern.
Information on the side effects of acamprosate therapy is derived from the efficacy studies that included more than 2,000 patients (Evidence Tables 9a and 9b). The side effect profile of acamprosate appears to be acceptable from a clinical standpoint, with the most frequently reported problem being diarrhea (approximately 10 to 40 percent of patients). Additional side effects reported more often in acamprosate patients than in placebo patients included dizziness, itching, and increased sexual desire. Dropouts due to any single side effect were less than 1 percent. No life-threatening side effects were reported in the trials.
Efficacy grade: A; Harms grade: Low.
Nine studies constitute the literature reviewed on the efficacy of serotonergic agents in maintaining abstinence in alcohol-dependent patients. Evidence Tables 10 and 11 address the studies of patients with and without comorbid psychiatric conditions, where the analysis does not distinguish between these two patient groups. The studies that evaluated patients with alcohol dependence and depression or anxiety as a distinct group are found in Evidence Tables 12 and 13.
Main Efficacy Results
Trials of medications that affect serotonergic neurotransmission also have been conducted for the treatment of primary alcohol dependence. These studies have evaluated the selective serotonin reuptake inhibitors (SSRIs) fluoxetine, citalopram, and fluvoxamine, along with the anxiolytic 5-hydroxytryptamine-1A (5-HT1A) receptor agonist buspirone, and the 5-hydroxytryptamine-3 (5-HT3) receptor antagonist ondansetron. Additional serotonergic agents have been studied but not with well-conducted controlled trials (Evidence Tables 10 and 11).
Five studies have examined the efficacy of fluoxetine for maintaining abstinence in alcohol-dependent patients. Two were randomized parallel group trials (Kranzler, Burleson, Korner, et al., 1995; Janiri, Gobbi, Mannelli, et al., 1996); two others were parallel group nonrandomized trials (Gorelick and Paredes, 1992; Kabel and Petty, 1996); and the fifth was a randomized cross-over study (Gerra, Caccavari, Delsignore, et al., 1992).
Unlike the acamprosate and naltrexone literature, the efficacy studies of the serotonergic agents do not evaluate a consistent set of alcohol outcomes. Of the three fluoxetine studies that assessed a standard quantity of alcohol consumed, all had different definitions for this measure, i.e., standard drinks, milliliters of alcohol consumed, or drinks per drinking day. The Gerra team (1992) reported a statistically significant difference between the placebo and fluoxetine groups in reduction of drinks (p<0.05) but only in those with familial alcoholism. In the study by the Kranzler group (1995), both the fluoxetine and placebo groups reported a statistically significant decline for two alcohol outcomes during the trial: the standard number of drinks per drinking day and the mean number of drinking days over the trial period. The reduction from baseline was equal in both treatment groups for both of these alcohol outcomes, indicating that fluoxetine was not superior to placebo for either alcohol outcome. In a subgroup analysis, these authors showed that type B alcoholics, i.e., those who have a greater severity of dependence, actually did worse on fluoxetine than those on placebo based on drinking days, drinks per day, and GGT levels (Kranzler, Burleson, Brown, et al., 1996). Comparing the results from baseline to end of trial, the Gorelick and Paredes (1992) study reported similar decreases in milliliters consumed between the fluoxetine and placebo groups. This indicates a study but not a treatment effect.
In the only fluoxetine study to examine the percentage of patients who resumed drinking during the trial (Janiri, Gobbi, Mannelli, et al., 1996), fewer fluoxetine patients than placebo subjects resumed drinking (38.1 and 65.5 percent, respectively) (p=0.05).
Craving was evaluated by three sets of investigators (Gorelick and Paredes, 1992; Janiri, Gobbi, Mannelli, et al., 1996; and Kabel and Petty, 1996). Only Kabel and Petty (1996) (using a craving questionnaire) reported a statistically significant reduction in craving in the fluoxetine group compared with the placebo group. Gorelick and Paredes (1992) also reported less alcohol craving in fluoxetine patients, but their results were not statistically significant. The Janiri team (1996) reported that craving for alcohol was actually higher in the fluoxetine group than in the placebo group at the end of the trial (p<0.05).
Only two fluoxetine studies specified relapse definitions with criteria similar to those in the naltrexone studies (Janiri, Gobbi, Mannelli, et al., 1996; Kabel and Petty, 1996). Neither study found significant differences between the fluoxetine and placebo groups, and Kabel and Petty (1996) reported a higher relapse rate for the fluoxetine treatment group.
One fluoxetine trial assessed compliance using riboflavin as the indicator (Kranzler, Burleson, Korner, et al., 1995). Compliance with treatment was greater in those on placebo (90 percent) compared with those on fluoxetine (77 percent), a statistically significant difference (p=0.03).
In the one study that investigated the SSRI citalopram, 62 patients were randomized to therapy but only 33 (53 percent) completed the study (Tiihonen, Ryynanen, Kauhanen, et al., 1996). The study investigated the number of episodes of heavy drinking and reduction in GGT but found no statistically significant differences between citalopram and placebo subjects.
A very small study of fluvoxamine that evaluated episodes of heavy drinking found no difference between the placebo and fluvoxamine treatment groups (Kranzler, Del Boca, Korner, et al., 1993).
In the only buspirone study that did not differentiate patients by comorbid anxiety, 57 patients were randomized to treatment, but only 36 (63 percent) remained in the trial (Malec, Malec, Gagne, et al., 1996). The alcohol outcomes were evaluated only for treatment completers and included the mean number of standard drinks, mean drinking days over the trial period, resumption of drinking, and GGT levels. Although all four outcomes declined from baseline in both the placebo and buspirone groups, the reduction was the same for both groups.
Ondansetron was evaluated for efficacy in only one study, which randomized a total of 71 patients to placebo, 0.5 mg ondansetron, or 4.0 mg ondansetron (Sellers, Toneatto, Romach, et al., 1994). The one alcohol outcome measured was mean number of standard drinks per day. Compared with placebo, a statistically significant difference was found for the low-dose group, but only in light drinkers. The study did assess treatment compliance by both urine marker and pill count, with compliance rates of >90 percent.
Subgroup and Followup Studies
Only one subgroup analysis was conducted (Janiri, Gobbi, Mannelli, et al., 1996). Of 10 depressed patients, all of the six patients on fluoxetine remained abstinent whereas only two of the four patients on placebo remained abstinent. Although encouraging, these results should be viewed with caution in light of the very small sample size.
Numerous, nonserious side effects were noted more frequently for those using serotonergic agents than for those on placebo (Evidence Tables 14a and 14b). Most of the reported side effects are typical for each class of drugs. For the SSRIs, the most frequently reported side effects were symptoms related to fatigue, nausea or vomiting, and sexual dysfunctions. The side effects for buspirone were those already known about this drug including dizziness, drowsiness, and nausea. None of the side effects or reasons for withdrawal appeared to be serious or clinically significant.
Efficacy grade: Insufficient data; Harms grade: Low.
Serotonergic Agents in Alcohol Dependence with Comorbid Depression and Anxiety
Only three studies evaluated serotonergic agents for their efficacy in reducing a return to drinking in alcohol-dependent patients who also suffered from clinical depression or anxiety (Evidence Tables 12 and 13).
Main Efficacy Results
One study of the efficacy of fluoxetine has been conducted in alcoholic patients who were also depressed (Cornelius, Salloum, Ehler, et al., 1997). That trial was a randomized, controlled, double-blind study over a 12-week treatment period in 51 patients, 46 of whom completed the study. During the 12-week period, the fluoxetine group had fewer standard drinks (p<0.03), drinking days (p<0.05), and days of heavy drinking (p=0.04) than did the placebo group, with a trend toward longer time to first drink for those on fluoxetine (p=0.08).
Two studies examined the effects of buspirone on symptoms in alcohol-dependent patients who also had anxiety disorder (Malcolm, Anton, Randall, et al., 1992; Kranzler, Burleson, Del Boca, et al., 1994). Compared with patients on placebo, the buspirone patients had fewer standard drinks per day and fewer drinking days during the trial, but neither of these differences in alcohol outcomes was statistically significant; however, time to first heavy drinking was longer in the buspirone group compared with placebo (p=0.03) (Kranzler, Burleson, Del Boca, et al., 1994). The Malcolm team (1992) conducted a 26-week study and reported no significant differences between placebo and buspirone subjects in drinks consumed over a given 28-day period, the percentage of patients who resumed drinking, time to first drink, or craving for alcohol.
Both these buspirone trials assessed compliance with therapy using urine riboflavin, and both reported very good compliance for the buspirone and placebo groups, with rates that did not differ between the two groups. The Cornelius group (1997) assessed compliance by pill count and serum drug levels, reporting that more than 99 percent of those on fluoxetine indicated compliance with drug therapy.
Subgroup and Followup Studies
One set of followup analyses at 6 months indicated that individuals who had received buspirone had significantly fewer drinking days than those who received placebo (p<0.01) (Kranzler, Burleson, Del Boca, et al., 1994).
In the fluoxetine study (Cornelius, Salloum, Ehler, et al., 1997), compliance by pill count was not reported for either the fluoxetine or placebo group, yet it appeared to be high for the fluoxetine patients based on serum drug levels. In addition, the sample was rather small (n=51) and composed of patients with severe depression based on their suicidal tendencies. This group may not be generalizable to the universe of alcohol-dependent patients who also suffer from depression, most of which is not as severe.
The buspirone studies also had a limited sample size (in total) of 128 subjects. Also, nearly half of the placebo patients in one study did not complete the trial (Kranzler, Burleson, Del Boca, et al., 1994). The dropout rate was even higher in another trial, in which nearly two-thirds of each treatment group left the study (Malcolm, Anton, Randall, et al., 1992).
The fluoxetine and buspirone side effect data were reviewed in the previous section on serotonergic agents (Evidence Tables 14a and 14b).
Efficacy grade: Insufficient data; Harms grade: Low.
Six clinical trials have investigated the efficacy of lithium for its actions in enhancing abstinence and decreasing alcohol intake and relapse rates in alcoholics with or without mood disorders (Kline, Wren, Cooper, et al., 1974; Merry, Reynolds, Bailey, et al., 1976; Pond, Becker, Vandervoort, et al., 1981; Clark and Fawcett, 1989; de la Fuente, Morse, Niven, et al., 1989; Dorus, Ostrow, Anton, et al., 1989) (Evidence Tables 15 and 16). In this context, study design is a critical factor because lithium is a standard therapy for certain mood and affect disorders. Some of the alcohol-dependence studies included patients who also had illnesses such as depression and/or anxiety, but the authors did not evaluate these patients separately from patients without such illnesses. Other studies did stratify by comorbidity, separately evaluating the effects of lithium in those patients with and without comorbid conditions.
Main Efficacy Results
The total sample size across these six studies was 750 subjects. The largest study assessed 286 nondepressed alcoholics and 171 depressed alcoholics (Dorus, Ostrow, Anton, et al., 1989). Trial lengths ranged from 12 to 96 weeks; the two largest studies lasted 52 and 72 weeks (Dorus, Ostrow, Anton, et al., 1989; Clark and Fawcett, 1989, respectively). Three trials administered lithium to achieve serum levels between 0.6 and 1.2 milliequivalents per liter (Kline, Wren, Cooper, et al., 1974; Merry, Reynolds, Bailey, et al., 1976; Pond, Becker, Vandervoort, et al., 1981); three other trials gave oral doses of lithium in the range of 600 to 1200 milligrams per day (Clark and Fawcett, 1989; de la Fuente, Morse, Niven, et al., 1989; Dorus, Ostrow, Anton, et al., 1989).
These lithium trials examined three alcohol outcome variables: drinking days, episodes of heavy drinking, and return to drinking. Of the three studies that evaluated drinking days, two showed no effect for lithium compared with placebo (de la Fuente, Morse, Niven, et al., 1989; Dorus, Ostrow, Anton, et al., 1989); the third study (Merry, Reynolds, Bailey, et al., 1976) reported fewer incapacitating days for depressed patients on lithium (0.7 days) than for depressed patients on placebo (5.1 days) (p<0.05), with no information available for those without depression.
Kline, Wren, Cooper, et al. (1974) and Merry, Reynolds, Bailey, et al. (1976) examined the effect of lithium on heavy drinking. Heavy drinking was defined as "percent of time incapacitated by drink" by the Merry (1976) team. They reported that depressed alcoholics on lithium had a lower percentage of time incapacitated (0.4 percent) than did depressed patients on placebo (1.5 percent), a result that was statistically significant (p< 0.05). The Kline group (1974) defined heavy drinking as "disabling" episodes of drinking. Although the treatment and placebo groups in their trial did not differ on levels of disabling drinking at baseline, patients on lithium had a lower percentage of disabling episodes (25 percent) than did those on placebo (64 percent) at the end of lithium treatment (p<0.05).
Two studies (de la Fuente, Morse, Niven, et al., 1989; Dorus, Ostrow, Anton, et al., 1989) examined the effects of lithium on return to drinking. One 6-month trial found that 56 percent of alcohol-dependent patients on placebo and 21 percent of patients in the lithium treatment group returned to drinking (p<0.01) (de la Fuente, Morse, Niven, et al., 1989). The other trial did not report statistically significant differences for return to drinking between the placebo and lithium treatment groups in either depressed or nondepressed subgroups (Dorus, Ostrow, Anton, et al., 1989).
Compliance with active (lithium) treatment was assessed in all studies, either by blood level measures or pill counts. When lithium blood levels were used for compliance, there was no corresponding measure of compliance for the placebo subjects. To maintain the blinding of the patients and study personnel, study investigators changed a placebo patient's treatment regimen when they changed a lithium patient's dosage. The Dorus team (1989) provided rigorous definitions for medication compliance but reported compliance only by depression status, not by treatment group, which diminishes the usefulness of their compliance data. Pond, Becker, Vandervoort, et al. (1981) used pill count to monitor compliance, reporting high compliance with therapy.
Subgroup and Followup Studies
Three trials conducted subgroup analyses (Pond, Becker, Vandervoort, et al., 1981; Clark and Fawcett, 1989; de la Fuente, Morse, Niven, et al., 1989). In the Clark and Fawcett (1989) subgroup analysis, subjects were divided by treatment (lithium vs. placebo), blood level (high vs. low), and compliance. Lithium-compliant patients with high blood levels had the lowest rates of rehospitalization (7 percent); all other groups had admission rates in the range of 23 to 29 percent (p=.006). When abstinence was examined, those with high blood lithium levels had the highest abstinence rate (57 percent).
The limitations of the lithium trials are similar to those of other drug trials of alcoholism: small sample sizes, high dropout rates, and unconfirmed reports of alcohol consumption. The lithium data also are complicated by the fact that four of the six trials did not analyze results separately for depressed and nondepressed subjects (Kline, Wren, Cooper, et al., 1974; Pond, Becker, Vandervoort, et al., 1981; Clark and Fawcett, 1989; de la Fuente, Morse, Niven, et al., 1989). In addition, of the six lithium studies we evaluated, two (Kline, Wren, Cooper, et al., 1974; Pond, Becker, Vandervoort, et al., 1981) were cross-over designs, a study design that may not be appropriate for evaluating drug efficacy in alcohol-dependent populations (Peck, Pond, Becker, et al., 1981).
As noted, all the studies were of alcohol-dependent patients who also had depression or bipolar disorder, but only two (Merry, Reynolds, Bailey, et al., 1976; Dorus, Ostrow, Anton, et al., 1989) stratified their patient populations by the presence or absence of coexisting mental disorder (depression). Thus, in the studies that did not stratify by mood disorder, it is difficult to determine whether the results could be attributable to lithium's effects on the core symptoms of alcoholism or to its effects on mood disorder symptoms.
Lithium is known to have many side effects based on experience with its use in mood-disorder patients; these include gastrointestinal symptoms, weakness, tremor, thirst, increased urination, rash, weight gain, and hypothyroidism. In addition, lithium toxicity can be serious, even life-threatening; this is a rare occurrence, however, if the medication is used properly. In the alcohol trials, diarrhea, tremor, and difficulty in walking were the only side effects reported as more common in patients receiving lithium than in those receiving placebo; no dropout rates were reported (Evidence Table 17). We would anticipate that the side effects known to occur in mood-disorder patients also would occur in alcohol-dependent patients. Also, lithium-induced side effects might be exacerbated in patients with heavy alcohol use.
Efficacy grade: C; Harms grade: Low.
Comment on Harms Data
Because the data on harms were not a systematic focus of the evidence report, readers should be aware that many side effects that may occur with the medications reviewed have not been recorded in our harms summary. Nevertheless, the evidence for serious side effects appears uniformly low across all the reviewed trials; we believe this indicated that, on balance, harms are low with these classes of medication.
The results presented above represent the literature published through November 1997. Findings from several important ongoing trials on naltrexone, acamprosate, and SSRIs are expected to advance the field of pharmaco-therapy for treating alcohol dependence and to add to the knowledge base on potential side effects of these agents.
Agency for Health Care Policy and Research (US), Rockville (MD)
West SL, Garbutt JC, Carey TS, et al. Pharmacotherapy for Alcohol Dependence. Rockville (MD): Agency for Health Care Policy and Research (US); 1999 Jan. (Evidence Reports/Technology Assessments, No. 3.) 3, Results.