Efficacy of Disulfiram and Cognitive Behavior Therapy in Cocaine-DependentOutpatients: A Randomized Placebo-Controlled Trial

Context  Disulfiram has emerged as a promising treatment for cocaine dependence, but it has not yet been evaluated in general populations of cocaine users.

Objectives  To compare the effectiveness of disulfiram therapy with that of a placebo condition in reducing cocaine use and to compare the effectiveness of 2 active behavioral therapies—cognitive behavior therapy (CBT) and interpersonal psychotherapy (IPT)—in reducing cocaine use.

Design  Randomized, placebo-controlled, double-masked (for medication condition), factorial (2 × 2) trial with 4 treatment conditions: disulfiram plus CBT, disulfiram plus IPT, placebo plus CBT, and placebo plus IPT.

Setting  A community-based outpatient substance abuse treatment program.

Patients  A total of 121 individuals meeting the criteria for current cocaine dependence.

Interventions  Patients received either disulfiram (250 mg/d) or placebo in identical capsules. Medication compliance was monitored using a riboflavin marker procedure. Both behavioral therapies (CBT and IPT) were manual guided and were delivered in individual sessions for 12 weeks.

Main Outcome Measures  Random regression analyses of self-reported frequency of cocaine use and results of urine toxicology screens.

Results  Participants assigned to disulfiram reduced their cocaine use significantly more than those assigned to placebo, and those assigned to CBT reduced their cocaine use significantly more than those assigned to IPT (P<.01 for both). Findings were consistent across all study samples (eg, intention to treat, treatment initiators, and treatment completers). Benefits of disulfiram use and CBT were most pronounced for participants who were not alcohol dependent at baseline or who fully abstained from drinking alcohol during treatment. Adverse effects experienced by participants who received disulfiram were mild and were not considerably different from those experienced by participants who received placebo.

Conclusions  Disulfiram and CBT are effective therapies for general populations of cocaine-dependent individuals. Disulfiram seems to exert a direct effect on cocaine use rather than through reducing concurrent alcohol use.

The past 10 years have been marked by important advances in the developmentof effective treatments for cocaine dependence.¹ Disulfiramhas recently emerged as one of a few promising pharmacotherapeutic approaches.Given clinical observations of high rates of alcohol dependence among cocaine-dependentoutpatients,^2-8 theinitial rationale for disulfiram therapy was as a strategy to reduce alcoholuse among cocaine users.^9,10 Reducedalcohol use was hypothesized to concomitantly reduce patients' exposure toalcohol, which can be a potent cue for cocaine use; lessen alcohol-relatedimpairments in judgment and in the ability to resist cravings and offers ofdrugs¹¹; and reduce exposure to cocaethylene,a pharmacologically active metabolite of cocaine and alcohol when they areused concurrently.^5,12

An initial randomized trial¹³ of disulfiramtherapy in individuals who were cocaine and alcohol dependent supported theeffectiveness of disulfiram therapy for cocaine dependence. However, thatstudy was not masked because it used a no-medication comparison to controlfor patients' expectations of the ethanol-disulfiram reaction.¹⁴ Twosubsequent randomized trials^15,16 ofcocaine-dependent individuals undergoing agonist therapies (either methadonehydrochloride or buprenorphine hydrochloride) suggested that disulfiram maybe an effective treatment for cocaine dependence regardless of whether patientsare concurrently alcohol dependent. These findings suggest that disulfirammay have some direct effects on cocaine dependence, which seems to be consistentwith new evidence from laboratory studies^17-19 suggestingthat disulfiram may affect subjective and physiologic responses to cocaine.

The present study was conducted to address several gaps in knowledgeregarding the effects of disulfiram therapy on cocaine use by being the firstto (1) evaluate a large general outpatient sample, (2) include a placebo conditionto control for medication expectancies, (3) include alcohol-dependent andnon–alcohol-dependent cocaine abusers to evaluate possible effects ofalcohol on the effects of disulfiram, and (4) to combine disulfiram therapywith either a highly structured coping skills treatment or a less structuredbehavioral approach to evaluate the impact of concomitant behavioral treatmentswith very different rationales.

Regarding behavioral therapies, cognitive behavior therapy (CBT) hasbeen demonstrated to be effective in cocaine-dependent samples.^13,20 Thereis also some evidence^21-27 suggestingthat CBT may be more effective when combined with medication and that it oftenhas a delayed main effect that grows stronger at posttreatment follow-up ofup to a year. However, studies evaluating the effectiveness of CBT have oftencompared it with nonspecific control conditions. The present study was alsodesigned to build on previous literature by comparing CBT with another activebehavioral approach, interpersonal psychotherapy (IPT). Moreover, the initialstudy¹³ evaluating disulfiram therapy in cocaine-and alcohol-dependent individuals provided some evidence to suggest that disulfiramwas more effective when delivered in conjunction with behavioral therapiesthat focused on facilitating abstinence (eg, CBT or 12-step facilitation)compared with a supportive clinical management control condition. Thus, thepresent study was also designed to extend those findings by comparing CBTwith a less structured behavioral therapy (IPT).

In this article, we present findings from a 2 × 2 factorial trialevaluating 2 medication conditions (disulfiram and placebo) and 2 types ofbehavioral therapy (CBT and IPT) in a general outpatient sample of cocaine-dependentindividuals. We hypothesized that (1) disulfiram therapy is more effectivethan placebo use in reducing cocaine use, regardless of whether patients alsouse alcohol, and (2) CBT is more effective than IPT in reducing cocaine use.Secondary, exploratory hypotheses included the following: (1) disulfiram therapyis effective without reference to alcohol intake among cocaine abusers whodo not drink alcohol, (2) disulfiram therapy is effective through reductionof alcohol use among cocaine abusers who are concurrently alcohol dependent,(3) medication compliance improves the efficacy of disulfiram therapy, and(4) disulfiram therapy is more effective when combined with CBT than withIPT.

Participants were recruited from individuals seeking treatment at theCentral Treatment Unit of the APT Foundation, a nonprofit outpatient substanceabuse treatment center in New Haven, or from respondents to newspaper advertisements.Individuals who met DSM-IV criteria for current cocainedependence were included. Individuals were excluded if they (1) were currentlyphysically dependent on opiates or barbiturates or if their principal drugof dependence was not cocaine, (2) met lifetime DSM-IV criteriafor a psychotic or bipolar disorder or expressed strong current suicidal orhomicidal ideation, (3) had a current medical condition that would contraindicatedisulfiram treatment (eg, hepatic or cardiac problems, hypertension, or pregnancy),or (4) had been treated for substance use during the previous 2 months. Individualswho were physically dependent on alcohol were eligible for the protocol afterthey completed alcohol detoxification.

Participants were administered the Structured ClinicalInterview for DSM-IV²⁸ and underwenta physical examination, including electrocardiography, urinalysis, and laboratoryblood work. A total of 121 of the 154 individuals screened were determinedto be eligible for the study, provided informed consent, and were randomized.Primary reasons for ineligibility were (1) failure to complete the pretreatmentevaluation (n = 14), (2) liver or heart conditions that contraindicated disulfiramtreatment (n = 8), (3) unwillingness to take disulfiram (n = 4), (4) severepolysubstance dependence that required inpatient treatment (n = 2), (5) workconflicts (n = 2), (6) failure to meet current cocaine dependence criteria(n = 1), (7) concurrent opioid dependence (n = 1), and (8) pregnancy (n =1). Urn randomization²⁹ was used to balancetreatment groups with respect to baseline severity of cocaine dependence,sex, and race.

Each of the study psychotherapies was manual guided and was deliveredto participants in weekly individual sessions offered over 12 weeks. Participantsalso met weekly with an independent clinical evaluator who collected urineand breath specimens, assessed recent cocaine use, and monitored other clinicalsymptoms. Methods previously demonstrated to be comparable in effectivenessto standard medication masking procedures³⁰ wereused to maintain single masking for the psychotherapy condition.

Participants assigned to disulfiram treatment were prescribed 250 mgof the drug daily; participants assigned to the placebo group received identicalcapsules. Participants were cautioned not to drink alcohol, and breathalyzersamples were collected at each contact and before dispensing of medicationeach week (results of all breathalyzer samples collected were negative). Adverseevents were monitored weekly. Medication compliance was monitored via a riboflavinmarker procedure³¹ and was assessed by a commerciallaboratory that evaluated whether each urine specimen collected was positivefor the riboflavin marker.

Cognitive behavior therapy was based on the relapse prevention modelof Marlatt and Gordon³² and adapted for usewith cocaine users. The goal of this treatment was abstinence from cocaineand other substances through functional analysis of high-risk situations forsubstance use and the development of effective coping strategies through skillstraining. As described in the manual,³³ skilltraining addressed (1) understanding patterns of drug use through functionalanalyses, (2) identifying and coping with cravings, (3) managing thoughtsabout drugs and alcohol, (4) developing effective drug-refusal skills, (5)developingproblem-solving skills, (6) developing emergency coping plans, (7) improvingdecision-making skills, and (8) using strategies to reduce the risk of humanimmunodeficiency virus.

Interpersonal psychotherapy³⁴ adaptedfor use with cocaine abusers³⁵ has 4 basiccharacteristics: (1) adherence to a medical model of psychiatric disorders,(2) a focus on patient difficulties in current interpersonal functioning,(3) brevity and consistency of focus, and (4) use of an exploratory stanceby the therapist that is similar to that of supportive and expressive therapies.The goals of IPT include the cessation of cocaine use and the developmentof more productive strategies for dealing with social and interpersonal problemsassociated with the onset and perpetuation of cocaine dependence.

The 16 doctoral-level therapists (7 in IPT and 9 in CBT) who deliveredthe study treatments were experienced in and committed to the type of treatmentthey delivered in the trial and in treating substance users. Therapists hada mean (SD) of 9.6 (8.4) years of postdoctoral experience and received trainingthat included (1) a didactic seminar and (2) completion of at least 1 closelysupervised training case. To promote adherence to manual guidelines, therapistsin each condition met regularly with supervisors to discuss case materialsand review session videotapes.

Participants were assessed before treatment, weekly during treatment,and at the 12-week treatment termination point by an independent clinicalevaluator who was masked to treatment condition. Primary outcome measureswere frequency of cocaine use (operationalized for the random regression analysesas the number of days per week the participant reported using cocaine) andresults of urine toxicology screens (operationalized as the likelihood ofsubmitting a cocaine-positive urine sample each week). The Substance AbuseCalendar, similar to the Timeline Followback method,^36,37 wasadministered weekly during treatment to collect detailed day-by-day self-reportsof cocaine use, alcohol use, other drug use, and medication compliance throughoutthe 84-day treatment and for the 28 days before randomization.

Participant self-reports of cocaine use were verified through urinetoxicology screens obtained at every visit. Of 851 urine specimens collectedduring the treatment phase, 713 (84%) were consistent with self-report, 27(3%) were negative for cocaine although the participant reported recent cocaineuse, and 111 (13%) were positive for cocaine in cases in which the participanthad denied use. This rate compares favorably with previous studies^37-39 of cocaine-dependentsamples, which have supported the accuracy of self-report data using the methodsdescribed herein.

The principal analytic strategy was the use of random regression models⁴⁰ for the 2 primary outcome variables, with time asa log-transformed variable to represent the greater slope anticipated in theearly weeks of treatment. Baseline frequency of cocaine use (operationalizedas the number of days of cocaine use in the 28 days before randomization)was used as a covariate for these analyses. In addition to the principal analysisconducted on the 121 participants randomized to treatment (intention-to-treatsample), supplemental analyses also evaluated treatment effects for the 112participants who initiated treatment, the 87 who completed 3 or more weeksof treatment, and the 53 who completed all treatment sessions. Results werehighly consistent across analysis samples; therefore, results from only theintention-to-treat sample are presented (a full report of all analyses isavailable from Dr Carroll).

Randomized participants who did not initiate treatment or who droppedout of treatment were followed and interviewed at the 12-week terminationpoint. Of these 68 participants, 49 (72%) were successfully tracked and interviewed.Thus, complete self-report outcome data were available for 84% of the randomizedsample (Figure 1).

Of the 121 participants randomized to treatment, 26% were female, 63%were white, 31% were African American, and 6% were Hispanic (Table 1). Seventy-seven percent of the participants were singleor divorced, and 55% were working full- or part-time. The mean age of theparticipants was 34.6 years. Mean monthly cocaine use at baseline was 16.6g, and participants reported using cocaine a mean of 13.0 days in the previous28 days. Participants reported drinking alcohol a mean of 9.4 days in theprevious 28 days. Fifty-four percent of the participants met DSM-IV criteria for a lifetime diagnosis of alcohol dependence, and26% met the criteria for a lifetime diagnosis of alcohol abuse. Fifty-twopercent of the patients met DSM-IV criteria for currentalcohol abuse or dependence. Analyses of variance and χ² testsrevealed a statistically significant difference between medication groupson rates of current alcohol abuse or dependence at baseline (disulfiram vsplacebo, 63% vs 41%; χ²₁ = 6.0; P = .01) but no other statistically significant differences by treatmentgroup for any other baseline variables.

Participants reported taking their medication on 76% of treatment days,and 72% of all urine specimens were positive for the urine riboflavin marker.Of 825 urine specimens collected during treatment that were matched to a participantself-report of medication compliance, 94% were consistent with the participant'sself-report. Medication compliance was not significantly different acrosspsychotherapy conditions (CBT mean [SD], 0.76 [0.26]; IPT mean [SD], 0.70[0.32]; F_1,99 = 1.1; P = .30) or acrossmedication conditions (disulfiram mean [SD], 0.72 [0.28]; placebo mean [SD],0.73 [0.30]; F_1,99 = 0.04; P = .85).

All psychotherapy sessions were videotaped for supervision and assessmentof fidelity to manual guidelines. To evaluate the discriminability of the2 types of behavioral therapy, 508 session videotapes (randomly selected early,middle, and late sessions for all participants who completed ≥1 session)were rated by evaluators who were masked to the patient's treatment condition.The Yale Adherence/Competence Scale, which has been demonstrated to have excellentreliability and concurrent and factorial validity in several previous studies,^41,42 was used for the process ratings.Estimates of interrater reliability were made based on a sample of 20 videotapesrated by all 11 raters. The model of Shrout and Fleiss⁴³ forrandom effects indicated a mean intraclass correlation coefficient estimateof 0.96, suggesting that the ratings were highly reliable. Random regressionanalyses suggested that study treatments were highly discriminable in that(1) CBT therapists used significantly more CBT interventions than IPT therapists(z = 12.4; P<.001) and(2) IPT therapists used significantly more IPT interventions than CBT therapists(z = 10.1; P<.01).

Moreover, to evaluate the potential effect of therapist effects on findingsreported in the following subsections, we conducted a 1-way analysis of varianceon continuous measures of outcome, including treatment retention, self-reportedcocaine use, urine test results, and medication compliance. No significanteffects for therapist were found for any of these variables, even using liberal P values (P<.25).⁴⁴

Of the 121 participants randomized, 112 (93%) initiated treatment. Themean (SD) number of sessions completed was 8.0 (5.1). One participant waswithdrawn from the trial because of continued high levels of substance usethat required inpatient treatment. Fifty-three participants remained in treatmentfor the full 12 weeks. Participants who remained in treatment for the full12 weeks did not differ from those who did not initiate treatment or who droppedout in terms of sex, race, route of administration, and the presence of lifetimeaffective disorder or antisocial personality disorder. However, participantswho were employed at baseline were more likely to complete treatment (54%vs 31%; χ²₁ = 6.01; P =.04), and those diagnosed as having a lifetime anxiety disorder were morelikely to drop out of treatment (27% vs 9%; χ²₁ =6.22; P = .01). There were no significant effectsof psychotherapy (χ²₁ = 0.07; P = .79) or medication condition (χ²₁ = 0.99; P = .32) on treatment completion.

Random-effects regression models evaluating the effects of the studytreatments on the frequency of self-reported cocaine use are presented in Table 2 and Figure 2. The overall effect for time was significant, indicatinga general reduction in the frequency of cocaine use over time for the participantsas a group (z = −7.48; P<.01). Participants assigned to disulfiram treatment reduced theircocaine use significantly more than those assigned to the placebo condition(medication × time, z = −2.82; P<.01). Participants assigned to CBT reduced their frequencyof cocaine use significantly more than those assigned to IPT (psychotherapy× time, z = −3.06; P<.01). The interaction of medication and psychotherapy was notstatistically significant (medication × psychotherapy × time, z = −1.12; P = .26). Whenpostattrition data from individuals who dropped out of treatment were includedin the analyses, all effects were consistent with the analyses limited todata collected before dropout (Table 2).

Results of urine toxicology screens were consistent with the findingsbased on self-report data. For the full randomized sample, there was a significantdecrease in the frequency of cocaine-positive urine specimens submitted overtime (z = −3.74; P<.01).There was a significant medication × time effect (z = −2.06; P = .04), suggesting a greaterreduction in the likelihood of submitting a cocaine-positive urine specimenfor participants assigned to disulfiram treatment vs placebo use. The psychotherapycondition × time effect failed to reach statistical significance forthe urine results (z = −1.78; P = .07), but it was in the same direction as the self-report data.Again, there was no evidence of a statistically significant psychotherapy× medication interaction (z = −0.96; P = .34).

As noted previously herein, approximately half of the sample met thecriteria for current alcohol abuse or dependence at baseline. To evaluatethe significance of current alcohol dependence with respect to treatment response,we conducted supplemental analyses using the same model as described previouslyherein to evaluate outcomes for (1) participants who met the current criteriafor alcohol abuse or dependence (n = 63) and (2) participants who did nothave a current alcohol diagnosis at baseline (n = 58). For the subgroup thatmet the criteria for alcohol abuse or dependence at baseline, the overalleffect of time on cocaine use was statistically significant (z = −8.74; P<.001), as was the psychotherapy× time effect (z = −2.60; P = .01); however, the medication × time effect was not significant(z = −1.29; P<.20)(Table 3).

For the subgroup that did not meet the criteria for alcohol dependence(n = 58), there was a significant overall effect of time on frequency of cocaineuse (z = −7.15; P<.001),a significant medication × time effect (z =−2.10; P = .04), a significant psychotherapy× time effect (z = −2.33; P = .03), and a significant 3-way interaction (z = −1.97; P = .05). The latter suggestedthat the placebo plus IPT condition was significantly less effective thaneach of the other 3 treatment conditions.

Similarly, whether a participant used any alcohol at all during treatmentwas strongly associated with the effectiveness of disulfiram treatment inreducing cocaine use. That is, for the 66 individuals who did not drink alcoholon any treatment days, the medication × time effect was significant(z = −4.2; P = .01),as was the psychotherapy × time effect (z =−4.4; P = .01) for frequency of cocaine use.However, for the 35 participants who reported drinking alcohol on 1 or moredays during treatment, the medication × time and psychotherapy ×time effects were not statistically significant, and neither were the overalleffects of time. This suggests that the benefits of CBT and disulfiram usein reducing cocaine use were most pronounced in participants who abstainedfrom alcohol during treatment. Although individuals who abstained from alcoholduring treatment reported using alcohol on significantly fewer days in the28 days before treatment initiation than those who used alcohol during treatment(6.1 vs 15.4 days; F_1,103 = 30.6; P<.001),those who were abstinent during treatment were not significantly differentfrom participants who used alcohol during treatment in terms of whether theymet the criteria for current alcohol abuse or dependence at baseline (53%vs 47%; χ² = 1.1; P = .29).

Additional exploratory analyses of variance evaluating compliance bytreatment condition and whether the participant reported any alcohol use duringthe trial suggested that individuals who were assigned to disulfiram treatmentand who used alcohol had lower rates of medication compliance (61% of urinespecimens positive for riboflavin) than those assigned to the placebo conditionwho did not drink alcohol (71%), those assigned to placebo use who did drinkalcohol (76%), and those assigned to disulfiram treatment who did not drinkany alcohol during the trial (79%) (medication condition × alcohol useeffect, F_1,95 = 3.6; P = .06).

To evaluate the impact of medication compliance on cocaine use duringtreatment, compliance with medication was added as a time-varying covariateto the random regression model described previously herein. The covariatewas significant, suggesting that medication compliance was a significant predictorof greater reductions in cocaine use over time (z =−12.6; P = .001). Nevertheless, the medication× time (z = −2.4; P = .02) and psychotherapy × time (z =−3.6; P = .01) effects remained statisticallysignificant when controlling for medication compliance.

As noted previously herein, all participants were cautioned not to drinkalcohol during the study because of the possibility of the ethanol-disulfiramreaction, and 66 participants reported no alcohol use during the trial. Forparticipants who reported drinking alcohol on 1 or more days, mean (SD) alcoholuse during the 84-day trial was 4.7 (6.3) days (range, 1-33 days). Random-effectsregression models evaluating effects of the study treatments on alcohol useindicated that as a group, participants reported reducing the frequency oftheir alcohol use over time (z = −0.53; P<.01). For the randomized sample (N = 121) and thesample that initiated treatment (n = 112), there were no statistically significanteffects of medication, psychotherapy, or their interaction on frequency ofalcohol use × time. However, for the subgroup that remained in treatmentfor 3 or more weeks (n = 87), there was a statistically significant effectfor disulfiram (medication × time, z = −2.68; P = .01). Moreover, when all postattrition data were includedin the analyses, there was a significant medication × time effect forthe full randomized sample (z = −2.33; P = 02).

Adverse effects were mild, and there were no significant differencesin rates of adverse effects by medication condition. The most frequently reportedadverse effect, headaches, was reported once or more during treatment by 33%of those assigned to placebo use and 34% of those assigned to disulfiram use.Rates of fatigue (28% for placebo vs 34% for disulfiram), nausea (26% vs 26%),and diarrhea (22% vs 21%) also were not significantly different by medicationcondition. There also were no significant differences in reported adverseeffects among the subgroups of participants who drank any alcohol or who usedany cocaine during treatment. There were no participant deaths or seriouscardiac events during the protocol.

The results of this randomized controlled trial of psychotherapies andpharmacotherapies for cocaine dependence suggest that (1) disulfiram treatmentwas more effective than the placebo condition in reducing the frequency ofcocaine use and the frequency of cocaine-positive urine samples over timeand (2) CBT was more effective than IPT in reducing cocaine use during treatment.

This is the first placebo-controlled trial, to our knowledge, to demonstratethat disulfiram therapy is effective in nonalcoholic cocaine-dependent outpatients.Moreover, these findings suggest that disulfiram therapy is especially effectivefor nonalcoholic cocaine users, as the effects of disulfiram treatment weremost pronounced in participants who did not meet the criteria for currentalcohol abuse or dependence and in those who abstained from alcohol duringthe trial. The findings disconfirm our hypothesis that disulfiram treatmentis equally effective in alcoholic and nonalcoholic patients but confirm ourhypothesis that medication compliance is associated with better outcomes inthe disulfiram group. However, the effect of medication compliance on outcomewas seen in the placebo group as well, and the effect of disulfiram treatmenton reducing cocaine use remained statistically significant even after controllingfor medication compliance.

The unexpected findings regarding the effects of disulfiram treatmenton the alcoholic subgroup might be explained by the differential compliancein the good vs poorer outcome alcoholic participants. Alcoholic participantswho drank while taking disulfiram were somewhat less likely to be compliantand seemed unable to refrain from continued alcohol use. When those patientsdiscovered a disulfiram reaction, they effectively broke the medication maskingand either discontinued taking disulfiram entirely or discontinued it whenthey wanted to drink alcohol or use cocaine. The patients who did not drinkalcohol did not break the medication masking and thus may have complied withthe medication treatment sufficiently to experience the anticocaine effect.

Paradoxically, this finding confirms (but in an opposite direction)our hypothesis that the effect of disulfiram treatment on alcoholic cocaineabusers is mediated by its impact on alcohol use. Instead of reducing cocaineuse through first stopping drinking, the powerful antidipsotropic effect ofdisulfiram led drinkers to stop taking disulfiram instead. These findingsfurther suggest that disulfiram treatment reduces the attractiveness of cocainein the same way for alcoholic and nonalcoholic cocaine abusers. This interpretation(ie, that the effect of disulfiram treatment on cocaine use is not mediatedby stopping alcohol use) is further supported by the finding that the disulfiram/placebodifferences were greater in participants who did not drink alcohol. If disulfiramexerted its effect on cocaine use by reducing the use of alcohol, a smallerdifference would be expected between disulfiram and placebo in the subgroupthat did not drink alcohol during the trial. Also, disulfiram treatment seemedto be safe in this population, and levels of reported adverse events weremild and comparable to those reported by patients in the placebo condition.

This study also adds to the growing body of work supporting CBT as atreatment for cocaine dependence and extends it in 2 ways, first by comparingit with another active psychotherapy and second by findings that suggest thatthe effects of CBT, when delivered with placebo, were not statistically significantlydifferent from CBT or IPT delivered in conjunction with disulfiram. The formerfinding is compelling in that IPT was conceived and implemented as an activecontrol condition and was delivered and supervised by investigators who wereadherents of IPT.^45,46 That theeffectiveness of CBT relative to that of IPT was robust across analysis samplesand the alcoholic/nonalcoholic groups is also notable, given findings fromthe general psychotherapy literature^47-49 thatcomparisons of 2 active behavioral conditions often result in comparable effects.Regarding the former finding (the lack of evidence of a CBT-disulfiram interactioneffect), Figure 2 and the analysisof the non–alcohol-dependent subgroup suggested a pattern of findingsthat was consistent with additive^50,51 effectsof the study treatments. That is, results suggest poorest outcomes for thegroup assigned to IPT plus placebo, and offering either CBT or disulfiramwas associated with optimal outcomes in this sample; combining the two providedlittle additional incremental benefit.

Limitations of the present study include that, as in other trials incocaine-dependent populations, 50% of individuals who began treatment didnot complete it (although, on average, patients attended 67% of sessions offered).Although this underlines that available treatments could, and should, be mademore attractive to patients, it is also important that attrition was unlikelyto undercut the validity of the conclusions that may be drawn from the studybecause (1) differential attrition by treatment condition was not seen, (2)statistical methods were used that are less sensitive to missing data andallow use of all data collected from all participants, (3) postattrition datawere collected from 73% of the dropouts, and (4) analyses that included datacollected after dropout were highly consistent with analyses limited to datacollected from participants while they were in treatment.⁵²

In summary, this study provides the strongest evidence to date regardingthe effectiveness of disulfiram treatment in reducing cocaine use in thatit was the first to be placebo controlled and to include a comparatively largeoutpatient sample of cocaine-dependent individuals. Furthermore, by includingalcohol-dependent and non–alcohol-dependent cocaine abusers, the resultsof this trial strongly suggest that the effect of disulfiram treatment oncocaine use was not related to cessation of concurrent alcohol use becauseits superiority to the placebo condition was greater in those who did notdrink. Disulfiram treatment was efficacious in patients without a currentdiagnosis of alcohol abuse or dependence or who refrained from drinking alcoholbut was not statistically significantly different from placebo among patientswho drank alcohol during the trial. Hence, the use of disulfiram in alcoholiccocaine abusers should be combined with efforts to enhance patient commitmentto alcohol abstinence. Alcohol use during treatment was strongly related tococaine use,⁶ independent of the patient'smedication condition.

Moreover, if research continues to support the efficacy of disulfiramtreatment in varied samples of cocaine abusers, it is important to rememberthat even the most powerful pharmacotherapies for substance use disorderscan be rendered ineffective unless delivered with adequate psychosocial treatment(eg, McLellan et al⁵³) and that carefully targetedbehavioral therapies can dramatically enhance pharmacotherapy compliance andeffectiveness.^54,55 Although theeffects of disulfiram treatment were comparable when delivered with CBT orIPT, both approaches were active behavioral therapies delivered by experiencedclinicians. Disulfiram treatment compliance and hence outcome might be lowerwhen used in combination with less structured behavioral approaches.⁵⁶

Corresponding author: Kathleen M. Carroll, PhD, Yale University Schoolof Medicine, Division of Substance Abuse, VA CT Healthcare Center (151D),Yale University School of Medicine, 950 Campbell Ave, West Haven, CT 06516(e-mail: kathleen.carroll@yale.edu).

Submitted for publication March 24, 2003; final revision received August1, 2003; accepted August 26, 2003.

This study was supported by grants R01-DA10679, KO5-DA00089 (Dr Rounsaville),K05-DA 00457 (Dr Carroll), and P50-DA09241 from the National Institute onDrug Abuse, Bethesda, Md.

We thank Joanne Corvino, MPH, John Cecero, PhD, Sarah Church, PhD, CelesteMilligan, PhD, Robert Chang, MD, Meghan Biro, Sarah Fitzsimmons, Kia Samuels,Lorraine Myers, Lance Barnes, and the staff of the Central Medical Unit ofthe APT Foundation.