REVIEW

Pharmacological treatment of cocaine dependence: a systematic review

Maurcio Silva de Lima1,2, Bernardo Garcia de Oliveira Soares1,3, Anelise Alves Pereira Reisser1 & Michael Farrell4
Centro de Medicina Baseada em Evidncias, Universidade Federal de Pelotas, Brazil1, Mestrado em Sade e Comportamento, Universidade Catlica de Pelotas, Brazil2, Departamento de Psiquiatria, Universidade Federal de So Paulo, Brazil3 and National Addiction Centre, Institute of Psychiatry, London, UK4

Correspondence to: Dr Maurcio Silva de Lima Centro de Medicina Baseada em Evidncias Departamento de Sade Mental Faculdade de Medicina UFPEL Avenida Duque de Caxias 250 Pelotas RS, Brazil Submitted 17 July 2001; initial review completed 30 October 2001; nal version accepted 6 March 2002

ABSTRACT Aims Cocaine dependence is a common and serious condition, associated with severe medical, psychological and social problems, including the spread of infectious diseases. This systematic review assesses critically the efcacy of pharmacotherapy for treating cocaine dependence. Methods The literature search strategy included: electronic searches of Cochrane Library holdings, EMBASE, MEDLINE, PsycLIT, Biological Abstracts and LILACS; scans of reference lists of relevant articles, personal communications, conference abstracts, unpublished trials from the pharmaceutical industry and book chapters on the treatment of cocaine dependence. Randomized controlled trials (RCTs) focusing on the use of antidepressants (ADs), carbamazepine (CBZ), dopamine agonists (DAs) and other drugs used in the treatment of cocaine dependence were included. The reviewers extracted data independently, and relative risks (RR) with 95% condence interval (CI) were estimated. Qualitative assessments were carried out using a Cochrane validated checklist. Where possible, analysis was carried out according to intention-totreat principles. Findings The search strategy generated 45 different trials. Most studied drugs were ADs (20 studies), DAs and CBZ. Data were very heterogeneous, with dropout rates within the studies between 0 and 84%. A non-signicant trend favoring CBZ was found in terms of dropouts (RR 0.88; 95% CI 0.751.03) and results from one trial suggest that uoxetine patients are less likely to drop out. The main efcacy outcome reported in the studies was the presence of cocaine metabolites in the urine. No signicant results were found, regardless the type of drug or dose used for all relevant outcomes assessed. Conclusions There is no current evidence supporting the clinical use of CBZ, antidepressants, dopamine agonists, disulram, mazindol, phenytoin, nimodipine, lithium and NeuRecover-SA in the treatment of cocaine dependence. Larger randomized investigation must be considered, while taking into account that these time-consuming efforts should be reserved for medications showing more relevant and promising evidence. Given the high dropout rate among the test population, clinicians may wish to consider adding psychotherapeutic supportive measures aimed at keeping patients in treatment programs. KEYWORDS Cocaine systematic review. dependence, meta-analysis, pharmacotherapy,

This review has been published partially in The Cochrane Library (Carbamazepine for Cocaine Dependence). It received the Kenneth Warren Prize given for relevant systematic reviews from developing countries in the last annual Cochrane Meeting (Cape Town, 2000).
2002 Society for the Study of Addiction to Alcohol and Other Drugs Addiction, 97, 931949

932

Maurcio Silva de Lima et al.

INTRODUCTION Cocaine consumption and related problems were epidemic in the 1920s in the United States, disappearing by the end of that decade (Musto 1992). However, the use of cocaine increased again between 1976 and 1979, mainly in North America and some countries of South America. Since the early 1980s, cocaine abuse in the United States has again been at epidemic levels. Estimates from a recent National Household Survey on Drug Abuse based on a sample of 28 832 subjects indicate that there are 1.3 million cocaine users in the United States, more than ve times the number of those addicted to heroin (Gold 1997). Cocaine problems in Europe have become more common since the early 1990s and now are part of the European drug scene (EMCDDA 2000). It has become a substantial public health problem, resulting in a signicant number of medical, psychological and social problems, including the spread of infectious diseases (e.g. AIDS, hepatitis and tuberculosis), crime, violence and neonatal drug exposure. In consequence, there is an urgent need to expand the treatment repertoire for this condition. Although there is no consensus regarding how to treat cocaine dependence (Carroll et al. 1994), effective pharmacotherapy can potentially play a major role within a broader treatment setting. The past decade has witnessed a sustained search for an effective pharmacotherapeutic agent for the treatment of cocaine dependence. While the administration of cocaine increases intercellular dopamine, serotonin and norepinephrine levels acutely by blocking their presynaptic reuptake (Gold 1997), chronic cocaine abuse leads to downregulation of monoamine systems. Post-cocaine use depression and cocaine craving may be linked to this downregulation. These preclinical ndings are the theoretical foundations on which the use of antidepressants for the treatment of cocaine dependence is based. Under this assumption, antidepressant pharmacotherapy, by increasing monoamine levels, may alleviate cocaine abstinence symptomatology, as well as relieving dysphoria and associated craving by general antidepressant action (Margolin, Avants & Kosten 1995b). There is a clear understanding that any discussion of pharmacotherapy for cocaine dependence should be contextualized within a framework of broader psychosocial interventions which could include therapeutic community, day programme or out-patient drug free treatment. A recent work (Simpson et al. 1999) looking at outcomes from a range of psychosocial treatments indicates that there is substantial positive impact from such treatments. The question of combining and evaluating psychosocial and pharmacotherapeutic interventions to improve
2002 Society for the Study of Addiction to Alcohol and Other Drugs

overall treatment outcomes will be a major challenge to researchers and clinicians in the drug treatment eld over the coming decade. Systematic reviews of scientic research allow for the efcient integration of valid information and provide a basis for rational decision-making. The use of explicit and consistent methods in reviews limits bias (systematic errors) and reduces random errors (simple mistakes), thus providing more reliable results upon which to draw conclusions and make decisions. In addition, metaanalysis, or the use of statistical methods to summarize the results of several independent studies, can provide a more precise estimate of the effects of health care than that which can be derived from the individual studies included in a review. In this paper, we report the main results of a series of Cochrane systematic reviews about pharmacotherapy of cocaine dependence. The various types of pharmacological interventions were grouped as follows: antidepressant (ADs); carbamazepine (CBZ), dopamine agonists (DAs); and miscellaneous, including other drugs such as naltrexone, mazindol and lithium.

METHODS In order to select all relevant randomized controlled trials (RCTs) in the pharmacological treatment of cocaine use disorders, a wide search strategy was performed. Electronic searches of Cochrane Library (2001 issue 4), EMBASE (from 1980 to October 2000), MEDLINE (from 1966 to October 2000), PsycLIT (from 1974 to July 2000), Biological Abstracts and LILACS (from 1982 to 2000) were completed with scanning of the reference lists of relevant articles and personal contacts with authors and to the pharmaceutical industry. RCTs focusing on the use of drugs for the treatment of cocaine dependence were included, irrespective of the diagnostic criteria. Trials where patients had an additional diagnosis such as opiate dependence were also eligible. Outcomes of interest were retention in treatment and the number of people reporting adverse events; efcacy as measured by urine samples positive for cocaine metabolites; self-reported craving; severity of dependence; amount of cocaine consumed; and quality of life measures. If assessment was not possible and no other outcomes were provided (such as continuous data, without mean and standard deviation), authors were contacted. Quality of trials was assessed using a criterion based on the evidence of a strong relationship between the potential for bias in the results and the allocation concealment (Schulz et al. 1995). Trials were considered with a low risk of bias if they had adequate allocation
Addiction, 97, 931949

Pharmacological treatment of cocaine dependence

933

concealment, and with a high risk of bias if an inadequate allocation of concealment was performed. Data were extracted independently by reviewers. For dichotomous data a standard weighted estimate of the typical treatment effect across studies, the relative risk (RR)that is, the risk of an event occurring among treatment-allocated individuals versus the corresponding risk in the control groupwas calculated. If the relative risk equals 1, this indicates no difference between the groups compared. The relative risks and its 95% condence interval (CI) were calculated with Review Manager 4.1 software, with the use of the DerSimonian Laird random effects model (DerSimonian & Laird 1986). The random effects model includes both within-study sampling error and between-studies variation in the assessment of the condence interval of the results, but the xed effects model takes only within-study variation to inuence the condence interval. Therefore, the random effects model is a rather more conservative approach. Whenever possible, analyses were performed according to intention-to-treat principles. To avoid the pitfall of applying parametric tests to non-parametric continuous outcome data, standard deviations and means were required to be reported in the paper or obtainable from the authors, and the standard deviation, when multiplied by 2, had to be less than the mean (otherwise, the mean is unlikely to be an appropriate measure of the centre of distribution) (Altman & Bland 1996). Data not satisfying these standards were not included in the data analysis. Heterogeneity in the results of the trials was assessed both by graphical inspection and by calculating a test of heterogeneity. Heterogeneity refers to the variation observed between the results of the individual studies. Statistically signicant heterogeneity is said to be present when more variation between the studies occurs than can be explained by the play of chance alone. Possible reasons for heterogeneity were prespecied. Responses differ (Alim et al. 1995): according to the different drugs (Alterman et al. 1992); when psychosocial therapies are provided in conjunction with prescribed drugs (Altman & Bland 1996); according to the characteristics of patients participating in trials; and (Arndt et al. 1992) depending on length of treatment. Heterogeneity was then assessed visually from graphs and by the c2 test of heterogeneity. A signicance level of less than 0.10 was interpreted as evidence of heterogeneity.

these citations and were included in the review: 20 on ADs, ve on CBZ, 13 on DAs and 11 on miscellaneous interventions. In four studies ADs were compared to both placebo and DAs or another drug. Disulram (antabuse) was compared to placebo in three RCTs, with participants presenting co-morbidity, opioid dependence in two trials (George et al. 2000; Petrakis et al. 2000) and alcohol abuse or dependence in Carrol et al. (1998). Mazindol, an amidazoline derivate that blocks reuptake of dopamine, was evaluated in two trials (Margolin et al. 1995a; Stine et al. 1995). Crosby et al. (1996) assessed the value of phenytoin, an anticonvulsant agent, as an anticraving medication. Rosse et al. (1994) compared nimodipine, a calcium channel blocker, to placebo. Lithium carbonate, a mood stabilizer used widely in the treatment of bipolar disorders, was studied by Gawin et al. (1989). Cold (1996) assessed the compound known as NeuRecoverSA, which includes l-tyrosine and various vitamins and minerals for which cocaine users present a deciency. A double-blind placebo controlled trial compared naltrexone with placebo (Somoza et al. 1998). In Grabowsky et al. (2000), two doses of risperidone (2 and 4 mg/day), an atypical antipsychotic, were compared to placebo.

Main features of included studies The main characteristics of the trials included in this review are displayed in Table 1.

Methodological quality Different methods of randomization were employed in clinical trials, but only nine author groups (Weddington et al. 1991; Kranzler & Bauer 1992; Covi et al. 1993; Moscovitz, Brookof & Nelson 1993; Kranzler et al. 1995; Margolin et al. 1995b; Nunes et al. 1995; Cold 1996) described adequate concealment of allocation (allocation was performed by a person who was not involved in recruitment of patients). These trials were rated as A: adequate allocation concealment; all remaining studies did not describe the concealment of allocation and were classied as B. Most trials included (90%) used a double-blind design. One study used diphenhydramine as an active placebo (Covi et al. 1993). An external evaluator blind to the interventions was used in Carrol et al. (1998), an open study which evaluated three different psychosocial interventions in addition to the disulram or no medication.

RESULTS The search More than 1000 citations were found throughout the search. Forty-nine different studies were identied from
2002 Society for the Study of Addiction to Alcohol and Other Drugs

Outcome reporting Many outcomes could not be summarized because they were presented in graphical form or only on statistical
Addiction, 97, 931949

934

Table 1 Pharmacotherapy of cocaine dependence: characteristics of included randomized controlled trials. Participants Interventions Outcomes used

Author/year

Methods

Maurcio Silva de Lima et al.

Carbamazepine studies Campbell et al. 1994 61 cocaine-dependent out-patients (DSM-III-R); mean age 32 years; 63% males; 90% black Co-morbidities: alcohol dependence (n = 16), major depression (11), antisocial personality disorder (16) 95 cocaine-dependent out-patients (DSM-III-R); age range 2151; 98% males; 98% black 183 cocaine-dependent out-patients (DSM-III-R); mean age 32.5; 71% males; 66.1% white, with at least an eighth grade education 40 cocaine-dependent out-patients (DSM-III-R); age range 1845; 100% males; 32% black. Subjects used at least 4 g of cocaine during the preceding month 72 cocaine-dependent out-patients (DSM-lll-R); mean age 33.2; 79% males; 68% black. At least 14 g of self-reported cocaine use in the previous 3 months 79 cocaine abuser out-patients (DSM-III) in methadone-maintained treatment for heroin dependence Mean age 40.5; 100% males, 90% black; 51% of subjects with APD (DSM-III) 1. Carbamazepine 400 mg (n = 62) 2. Carbamazepine 800 mg (n = 58) 3. Placebo(n = 63) Psychosocial interventions were also offered to participants 1. Carbamazepine 600 mg (n = 20) 2. Placebo (n = 20) 1. Carbamazepine (n = 37) Dose from 200 mg/day to reach serum levels 412 g/ml 1. Desipramine (n = 21; dose unknown) 2. Carbamazepine (n = 19; dose unknown) 3. Placebo (n = 25) Duration of treatment

Random allocation (computergenerated list), double-blind, 6 months duration, three parallel groups, non-ITT

Cornish et al. 1995

Random allocation; double-blind; 10 weeks duration; two parallel groups; non-ITT; two study sites

No retention in treatment; urine samples

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Halikas et al. 1997

Random allocation (block design); double-blind; 12 weeks duration; three parallel groups; non-ITT (33 replacements)

No retention in treatment; urine samples

Kranzler et al. 1995

Random allocation (undertaken by a research pharmacist not involved in the clinical care); double-blind; 12 weeks duration; two parallel groups; ITT

Urine samples; no retention in treatment; people with at least one side effect; ASI, BDI, SAS

Montoya et al. 1994

Random allocation; double-blind; 8 weeks duration; two parallel groups; non-ITT

1. Carbamazepine 800 mg/day (n = 28) 2. Placebo (n = 34)

No retention in treatment

Antidepressant studies Arndt et al. 1992*

Random allocation (2:1 ratio); double-blind; 12 weeks duration; two parallel groups; non-ITT

1. Desipramine 250300 mg/day (n = 36) 2. Placebo (n = 23)

No retention in treatment; no retention in treatment for side effect

Addiction, 97, 931949

Batki et al. 1996

Random allocation; double-blind; 12 weeks duration; two parallel groups; non-ITT

32 cocaine-dependent out-patients (DSM-III). Age ~34 years; 66% males, 53% African-American 12 participants with major depressive disorder, six with APD, six with alcohol abuse dependence (DSM-III-R) 65 cocaine-dependent out-patients (DSM-III-R); mean age 32; 63% male; 90% black 16 subjects with alcohol dependence, 11 major depression, two generalized anxiety disorder and 16 APD 1. Desipramine (n = 21; dose unknown) 2. Carbamazepine (n = 19; dose unknown) 3. Placebo (n = 25) Duration of treatment (no SD)

1. Fluoxetine 40 mg/day (20 mg in the rst week) (n = 16) 2. Placebo (n = 16)

No retention in treatment; QCI modied version (days used, craving, quality of high)

Campbell et al. 1994

Random allocation (computergenerated list); double-blind; 6 months duration, three parallel groups; non-ITT

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Carroll et al. 1994

Random allocation; double-blind; 12 weeks duration; four parallel groups; non-ITT. Subjects received RP or CM

No retention in treatment Cocaine use during treatment

139 cocaine-dependent out-patients (DSM-III-R) Age ~29 years, 73% males, 46% Caucasians; 49% with APD; 65% any other personality disorder, 48% alcohol dependence, 20% affective disorder, 13% anxiety disorder 45 cocaine-dependent out-patients (DSM-III-R). Mean age 30; 80% males, 56% white. 67% with tobacco dependence, 22% alcohol dependence, 18% anxiety disorders, 7% APD 80 cocaine-dependent out-patients (DSM-III-R). Mean age 37; 100% males, 95.5% African-aAerican; subjects had regular cocaine use ~6 years 100 cocaine-dependent out-patients (DSM-III-R). Mean age 29; 76% males, 71% Caucasians. Most participants had other diagnosis such as ADD 1. Fluoxetine 20 mg (n = 10) 2. Fluoxetine 40 mg (n = 11) 3. Fluoxetine 60 mg (n = 10) 4. Active placebo (diphenhydramine) (n = 14) 1. Ritanserin 10 mg/day (n = 40) 2. Placebo (n = 40)

1. Desipramine (mean 200 mg/day) + RP (n = 29) 2. Desipramine (mean 200 mg/day) + CM (n = 25) 3. Placebo + RP (n = 29) 4. Placebo + CM (n = 27)

ASI

Covi et al. 1993

Random allocation (performed by a person who was not involved in recruitment of patients); double-blind; 12 weeks duration; four parallel groups; non-ITT

Urine positive for cocaine metabolites

Ehrman et al. 1996

Random allocation; double-blind; 4 weeks duration (medical phase); two parallel groups; non-ITT

No retention in treatment

Pharmacological treatment of cocaine dependence

Gawin et al. 1989

Random allocation; double-blind; 6 weeks duration; three parallel groups; non-ITT

1. Desipramine 2,5 mg/kg (n = 31) 2. Lithium carbonate 600 mg (n = 37) 3. Placebo-atropine 0.1 mg (n = 32)

Urine samples; duration of treatment

935

Addiction, 97, 931949

936

Table 1 (Continued) Participants 40 out-patients, 20 chronic cocaine abusers and 20 chronic phencyclidine abusers; age range 2034 years, 100% male, 100% Caucasians Subjects used only that particular drug at least three times weekly and had a history of abuse of at least 1 year 20 patients on private-practice psychiatry group, with space-base (combination of free-base cocaine and phencyclidine) abusers. Age range 2128 years, 100% male, 100% Caucasians 94 in-patients at beginning (~2 weeks), with cocaine dependence (DSM-III-R). Mean age 38 years, 100% males, 85% black. Primary cocaine abusers; patients who abused other drugs were also included 60 cocaine-dependent in-patients (DSM-III-R) in the rst week of trial, mean age 30.8; 95% males; 66% black. Most subjects were employed living in a moderately stable social situation, with no APD 24 cocaine-dependent out-patients with methadone-maintained (DSM-III-R). Mean age 34.8; 85% males, 68% black. Average use of cocaine: 10 years. Comorbidities: attention decit disorder, affective and anxiety disorders 1. Desipramine 200 mg: continuity (n = 23) 2. Desipramine 200 mg; standard (n = 22) 3. Placebo: continuity (n = 28) 4. Placebo: standard (n = 21) 1. Gepirona: mean dose 16.25 mg/day (n = 20) 2. Placebo (n = 21) 1. Desipramine 200 mg/day (n = 10) 2. Placebo (n = 10) Non-abstinent No retention in treatment 1. Desipramine 150 mg/day (n = 10) 2. Placebo (diphenhydramine) (n = 10) No retention in treatment Interventions Outcomes used

Author/year

Methods

Maurcio Silva de Lima et al.

Giannini et al. 1986

Random allocation (using Texas Instrument Programmable 58 random selection program); double-blind; 40 days duration four parallel groups; ITT

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Giannini et al. 1987

Random allocation (using Texas Instrument Programmable 68 random program); double-blind; 45 days duration; two parallel groups; ITT

Hall et al. 1994

Random allocation; double-blind; 12 weeks duration; four parallel groups; ITT

Urine samples;-non-entrance in out-patient program (not leaving the hospital)

Jenkins et al. 1992

Random allocation; double-blind; 12 weeks duration; two parallel groups; non-ITT; multi-center (four centers)

No retention in treatment; duration of treatment; urine samples; no global improvement (CGI)

Kolar et al. 1992*

Random allocation; double-blind; 12 weeks for desipramine and placebo; 8 weeks for amantadine followed by 4 weeks of placebo; three parallel groups; non-ITT

1. Desipramine 200 mg (n = 8) 2. Amantadine 200 mg followed by placebo (n = 5) 3. Placebo (n = 9)

No retention in treatment; urine samples; participants presenting at least one side-effect

Addiction, 97, 931949

Kosten et al. 1992*

Random allocation; double-blind; 12 weeks duration; three parallel groups; non-ITT

94 out-patients with opioid and cocaine dependence (DSM-III-R) Mean age 32 years, 52% males, 82% Caucasians. Other diagnosis: APD(20%); major depression (5%), dysthymia (22%) 149 out-patients with cocaine dependence (DSM-III-R), methadonemaintained for heroin dependence. Mean age 37.2 years, 62% males, 45% Caucasians. About 50% met criteria for antisocial personality disorder 15 in-patients, with cocaine dependence (DSM-III-R) and also meeting criteria for current or past abuse of other drugs. Mean age 29.5; 73% males. Average duration of cocaine use 5 years 113 out-patients with cocaine abuse (DSM-III-R). Mean age 32 years, 73% males, 52% Caucasians Depressive disorder present in 61% of participants 47 male subjects with cocaine dependence (DSM III R) in methadone maintenance, using cocaine for 3 or more months Age range 2950; those with sedative or alcohol dependence were excluded 22 subjects with cocaine dependence. Age, sex, race and setting unknown. Patients with other drug dependence were excluded 1. Imipramine 150300 mg (n = 59) 2. Placebo (n = 54) No response 1. Desipramine 200 mg (n = 9) 2. Placebo (n = 6) 1. Bupropion 300 mg/day (n = 74) 2. Placebo (n = 75)

1. Desipramine 150 mg (n = 30) 2. Amantadine 300 mg (n = 33) 3. Placebo (n = 31) Urine samples; no retention in treatment; no retention in treatment for side effect

Margolin 1995a*

Quasi-randomized (stratication by presence of APD and sequentially randomized); double-blind; 12 weeks duration; two parallel groups; ITT; multicenter study

No retention in treatment; no retention in treatment for side-effect; Ham-D; ASI; craving

2002 Society for the Study of Addiction to Alcohol and Other Drugs

McElroy et al. 1989

Random allocation; double-blind; 24 weeks duration; cross-over (at 12 weeks); non-ITT

Urine samples; no retention in treatment; no retention in treatment for side effect

Nunes et al. 1995

Random stratied allocation, administered by a nurse not involved in recruitment; double-blind; 12 weeks duration; two parallel groups; ITT

No retention in treatment No retention in treatment for side effect

OBrien et al. 1988*

Random allocation; double-blind, assessed at follow-up; 12 weeks duration; two parallel groups; non-ITT

1. Desipramine (n = 24) 2. Placebo (n = 14)

No retention in treatment

Pharmacological treatment of cocaine dependence

Tennant & Tarver 1985

Random allocation; double-blind; 12 days for patients on desipramine, 15 for placebo; two parallel groups; analysis unclear

1. Desipramine (n = 11) 2. Placebo (n = 11)

Positive urine samples for cocaine metabolites; no retention in treatment

937

Addiction, 97, 931949

938

Table 1 (Continued) Participants 82 male in-patients (for the rst 2 weeks) with crack-cocaine dependence. Age unclear, 85% black, 76% unemployed and 32% homeless 83 out-patients with cocaine dependence (DSM-III-R). Mean age 30 years, 76% males, 69% white. Additional diagnosis included attention decit disorder, affective and anxiety disorders 42 day hospital patients with cocaine dependence (DSM-III-R); mean age 35; 100% males, 90% black. Use of cocaine 15 days in the past 30 days and regularly for about 3 years 63 male in-patients with cocaine dependence (DSM-III-R); mean age 35.6; 86% black. Last cocaine use within 6 days before study entrance 30 male out-patients, cocaine abusers. Age range 2432; 100% Caucasians. Intranasal use on a daily basis for at least 6 months before study. Four subjects with APD 67 male out-patients with cocaine and heroin dependence (DSM-III-R), methadone-maintained; mean age 36 1. Amantadine 200400 mg (n = 21) 2. Placebo (n = 21) 1. Desipramine 200 mg (n = 32) 2. Amantadine 400 mg followed placebo (n = 23) 3. Placebo (n = 28) 1. Desipramine 200 mg qD (n = 41) 2. Placebo (n = 41) Duration of treatment Interventions Outcomes used

Author/year

Methods

Trifeman et al. 1992

Random allocation (2 2); doubleblind; 8 weeks duration; two parallel groups; non-ITT

Maurcio Silva de Lima et al.

Weddington 1991

Random allocation (administered by a person not involved in recruitment, using serially numbered, opaque, envelopes); single-blind; 12 weeks duration; three parallel groups; non-ITT

Urine samples; no retention in treatment; duration of treatment; participants presenting at least one sideeffect

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Dopamine agonist studies Alterman et al. 1992

Random allocation (completed by a research technician using a constrained block); double-blind; 2 weeks duration; TWO parallel groups; ITT

No retention in treatment; urine samples; BDI

Eiler et al. 1995

Random allocation; double-blind; 18 days duration; two parallel groups; non-ITT

1. Bromocriptine 2.510 mg/day (n = 32) 2. Placebo (n = 31)

No retention in treatment; no retention in treatment for sideeffect

Giannini et al 1989

Random allocation (Texas Instrument Programmable 68 random computer program); double-blind; 30 days duration; three parallel groups; ITT

1. Amantadine 400 mg/day (n = 10) 2. Bromocriptine 10 mg/day (n = 10) 3. Placebo (n = 10)

Craving Side effects

Addiction, 97, 931949

Handelsman et al. 1995*

Random allocation; placebo washout (1 week); double-blind; 8 weeks duration; three parallel groups; non-ITT

1. Amantadine 200 mg/day (n = 19) 2. Amantadine 400 mg/day (n = 23) 3. Placebo (n = 25)

No retention in treatment BDI SCL-90

Handelsman et al. 1997*

Random allocation; double-blind; 5 weeks duration; two parallel groups; non-ITT

60 male out-patients with cocaine abuse/dependence (DSM-III-R), methadone-maintained and using cocaine in the last 30 days. Mean age 39; 24% black 61 out-patients with cocaine dependence (DSM-III-R) Mean age 35 years, 80% males; 70% black. Cocaine use within 10 days prior to study. Alcohol and marijuana dependence were included 24 out-patients with cocaine dependence (DSM-III-R), methadone-maintained. Mean age 34.8; 85% males, 68% black Average use of cocaine: 10 years Co-morbidities: ADD, affective and anxiety disorders 94 out-patients with opioid and cocaine dependence (DSM-III-R) Mean age 32; 52% males, 82% white. ADD (20%); depression (5%), dysthymia (22%) 20 male in-patients with cocaine dependence (DSM-III-R). Mean age 27.7 years, 85% white Subjects had used an average of 3 g of cocaine during the month prior to the admission 29 male out-patients, cocaine users Mean age 37 years. Participants used cocaine at least four times per week for the previous month 1. Desipramine 150 mg (n = 30) 2. Amantadine 300 mg (n = 33) 3. Placebo (n = 31) 1. Desipramine 200 mg (n = 8) 2. Amantadine 200 mg followed by placebo (n = 5) 3. Placebo (n = 9) 1. Amantadine 300 mg (n = 30) 2. Placebo (n = 31) No retention in treatment;-non-abstinent; ASI; BDI; BAI

1. Bromocriptine 5 mg (n = 24) 2. Placebo (n = 26) No retention in treatment

Kampman et al. 1996

Random allocation (stratied block procedure); double-blind; 4 weeks duration; two parallel groups; ITT for urine samples

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Kolar et al. 1992*

Random allocation; double-blind; 12 weeks for desipramine and placebo; 8 weeks for amantadine followed by 4 weeks of placebo; three parallel groups; non-ITT

Positive urine sample for cocaine metabolites; no retention in treatment

Kosten et al. 1992

Random allocation; double-blind; 12 weeks duration; three parallel groups; ITT

Urine samples; no retention in treatment; no retention in treatment for side effect

Kranzler & Bauer 1992

Random allocation (capsules identical in appearance administered by nursing staff, blind to group assignment); double-blind; 21 days duration; two parallel groups; non-ITT

1. Bromocriptine 2.5 mg (n = 10) 2. Placebo (n = 10)

Participants presenting at least one side effect Side effects

Pharmacological treatment of cocaine dependence

Moscovitz et al. 1993

Random allocation (performed by a pharmacist who had no contact with subjects); double-blind; 2 weeks duration; two parallel groups; non-ITT

1. Bromocriptine 3.75 mg/day (n = 14) 2. Placebo (n = 15)

Urine samples; no retention in treatment; participants presenting at least one side effect

939

Addiction, 97, 931949

940

Table 1 (Continued) Participants 14 out-patients with cocaine dependence. Age range 1638; Sex/race unclear; use of cocaine at least four times per day for the 30 days prior to the admission 83 out-patients with cocaine dependence (DSM-III-R). Mean age 30, 76% males, 69% white Cocaine use of 1 g or more per week for 12 weeks. Additional diagnosis included: ADD; affective and anxiety disorders 13 in-patients with cocaine abuse or dependence (DSM-III-R). Age range 2546 years, 75% male, race unclear; 10 patients also treated for alcohol dependence 60 out-patients with cocaine abuse or dependence (DSM-III-R). Mean age 34 years, 80% male, 55% AfricanAmerican 100 out-patients with cocaine dependence (DSM-III-R) seeking treatment. Mean age 29 years; 76% male; 71% Caucasians. Most participants had other diagnosis such as ADDr 1. NeuRecover-SA six capsules daily (n = 8) 2. Placebo (n = 4) 1. Desipramine 200 mg (n = 32) 2. Amantadine 400 mg followed placebo (n = 23) 3. Placebo (n = 28) 1. Amantadine 100 mg/day (n = 7) 2. Bromocriptine 2.5 mg/day (n = 7) Interventions Outcomes used Urine samples; no retention in treatment; side effects; side effects

Author/year

Methods

Maurcio Silva de Lima et al.

Tennant & Tarver 1987

Random allocation (using a coin ip); double-blind; 10 days duration; two parallel groups; ITT

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Weddington et al. 1991

Random allocation (by a person not involved with recruitment of patients, using serially numbered, opaque, envelopes); single-blind; 12 weeks duration; three parallel groups; non-ITT

Urine samples; no retention in treatment; duration of treatment; participants presenting at least one side effect

Miscellaneous drugs studies Cold 1996

Random allocation (by a person not responsible for recruiting patients); double-blind; 521 days duration; two parallel groups; non-ITT

Abstinence symptoms (ASE modied); cocaine craving (CCS)

Crosby et al. 1996

Random allocation; double-blind; 12 weeks duration; two parallel groups; non-ITT

1. Phenytoin 300 mg/day (n = 29) 2. Placebo (n = 31)

Cocaine use (self-report); no retention in treatment; craving; side-effects

Gawin et al. 1989

Random allocation; double-blind; 6 weeks duration, three parallel groups; non-ITT

1. Desipramine 2,5 mg/kg (n = 31) 2. Lithium carbonate 600 mg (n = 37) 3. Placebo-atropine 0.1 mg (n = 32)

Urine samples; duration of treatment

Addiction, 97, 931949

Margolin et al. 1995b*

Random allocation (a staff pharmacist, with no contact to participants, developed the randomization code and assigned the interventions); double-blind; 12 weeks duration; two parallel groups; ITT 70 male in-patients with cocaine dependence (DSM-III-R). Mean age 33 years, race unclear Average of cocaine use per week was 6.6 g 43 out-patients with cocaine dependence (DSM-III-R). Mean age 35 years; 86% male; 51% African-American. 67% and 58% of participants had alcohol and marijuana abuse, respectively 46 patients in a 28-day, substance abuse treatment program enrolled in the study during their rst week of in-patient treatment. On day 10 of hospitalization they were randomized to treatment groups After 3 months follow-up only six subjects in the naltrexone group and seven subjects in the placebo group remained active in the study 1. Naltrexone (n = 24) 2. Placebo (n = 22) 1. Mazindol 2 mg/day (n = 22) 2. Placebo (n = 21) 1. Nimodipine 90 mg/day (n = 33) 2. Placebo (n = 33)

37 out-patients with cocaine dependence and in methadone maintenance for heroine treatment Mean age 34 years, 43% male, 68% Caucasian. On average, subjects had been using cocaine for 11 years

1. Mazindol 1 mg/day (n = 18) 2. Placebo (n = 19)

Urine samples; no retention in treatment; participants presenting at least one side effect

Rosse et al. 1994

Random allocation; double-blind; 3 weeks duration; two parallel groups; non-ITT

Craving and urges (QCU); mood state (POMS-BI)

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Stine et al. 1995

Random allocation; double-blind; 6 weeks duration; two parallel groups; ITT

Urine samples; no retention in treatment

Pharmacological treatment of cocaine dependence

Somoza et al. 1998

Random allocation; double-blind; 3 months follow-up after hospitalization; two parallel groups; noninformation on analysis

Self-reported cocaine use; urine samples; subjective cocaine craving; BSI; no retention in treatment

*Studies including patients in methadone maintenance treatment: ADD, attention decit disorder; APD, antisocial personality disorder; ASI, Addiction Severity Index; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory; BPRS, Brief Psychiatric Rating Scale; CCS, Cocaine Craving Scale; CCI, cocaine craving intensity; CCF, cocaine craving frequency; CGI, clinical global improvement; CSR, cocaine status report; DSM-III, Diagnostic and Statistic Manual (American Psychiatric Association), 3rd edn; DSM-III-R, Diagnostic and Statistic Manual (American Psychiatric Association), 3rd edn revised; HAM-D, Hamilton Depression Scale; ITT, intention-to-treat analysis; MCCS, Minnesota Cocaine Craving Scale; NIMH DIS, National Institute of Mental Health Diagnostic Interview Schedule; PANAS, Positive Affect Negative Affect Scale; POMS, Prole of Mood States; QCI, Quantitative Cocaine Inventory; QCI modied, Yale Quantitative Cocaine Inventory, modied version (Batki et al. 1991, 1993, 1994); SSA, Spielberg State Anxiety score.

941

Addiction, 97, 931949

942

Maurcio Silva de Lima et al.

Study 01 Desipramine Gawin et al. 1989 Hall et al. 1994 Kolar et al. 1992 Kosten et al. 1992 McElroy et al. 1989 Tennant & Tarver 1985 Weddington et al. 1991 Subtotal (95% Cl) 02 Fluoxetine Covi et al. 1993 03 Gepirone Jekins et al. 1992 04 Amantadine Alterman et al. 1992 Kolar et al. 1992 Kosten et al. 1992 Weddington et al. 1991 Subtotal (95%Cl) 05 Bromocriptine Moscovitz et al. 1993 06 Carbamazepine Corinish et al. 1995 Halikas et al. 1997 Kranzler et al. 1995 Subtotal (95%Cl) 07 Disulfiram George et al. 2000 Petrakis et al. 2000 Subtotal (95%Cl) 08 Mazindol Margolin et al. 1995a Stine et al. 1995 Subtotal (95%Cl) 09 Lithium carbonate Gawin et al. 1989

Active drug n /N

Placebo n /N

RR (95%Cl Random)

Weight %

RR (95%Cl random)

10/24 13/22 1/8 22/30 2/9 4/11 24/32 76/136 13/31 6/12 1/15 4/5 28/33 18/23 51/76 10/14 13/37 44/58 14/20 71/115 6/11 31/36 37/47 10/18 17/22 27/40 18/24

20/24 11/21 7/9 27/31 3/6 3/11 19/28 90/130 5/14 5/14 6/15 7/9 27/31 19/28 59/83 13/15 11/45 53/62 16/20 80/127 5/9 29/31 34/40 13/19 15/21 28/40 20/24
0.1 0.2 Favors active drug 1 5 10 Favors placebo

18.1 17.1 2.6 27.8 4.1 5.4 25.0 100.0 100.0 100.0 1.5 15.3 51.5 31.7 100.0 100.0 7.4 69.2 23.4 100.0 3.9 96.1 100.0 32.0 68.0 100.0 100.0

0.50 [0.30,0.83] 1.13 [0.66,1,93] 0.16 [0.02,1.04] 0.84 [0.65,1.09] 0.44 [0.10,1.92] 1.33 [0.39,4.62] 1.11 [0.80,1.53] 0.82 [0.60,1.13] 1.17 [0.52,2.65] 1.40 [0.57,3.45] 0.17 [0.02,1.22] 1.03 [0.59,1.80] 0.97 [0.80,1.19] 1.15 [0.83,1.61] 1.01 [0.79,1.29] 0.82 [0.56,1.21] 1.44 [0.73,2.82] 0.89 [0.74,1.06] 0.88 [0.61,1.26] 0.92 [0.76,1.10] 0.98 [0.44,2.17] 0.92 [0.78,1.08] 0.92 [0.79,1.08] 0.81 [0.49,1.36] 1.08 [0.76,1.54] 0.99 [0.74,1.32] 0.90 [0.67,1.21]

Figure 1 Positive urine samples for cocaine metabolites

tests and P-values. For most of the continuous variables, standard deviation was not provided or data were skewed. Variation was seen regarding trialists denition of efcacy (positive urine samples for cocaine metabolites, self-report of cocaine use, craving, abstinence symptoms).

Efcacy ndings The main efcacy outcome reported in the studies was the presence of cocaine metabolites in the urine. Usually, the cut-off value considered to determine a positive test sample was 300 mg/ml. Figure 1 shows these ndings for all drugs. No signicant results were found, regardless the type of drug or dose used. Desipramine was evaluated in 14 trials. Some degree of heterogeneity was found when all studies were con 2002 Society for the Study of Addiction to Alcohol and Other Drugs

sidered, regardless of whether or not cocaine dependence was a primary diagnosis. In a preliminary analysis, when all desipramine trials with urinalysis data available were included, signicant heterogeneity was present (c2 = 12.7; d.f. = 6; P = 0.048). This result favored desipramine, but it was statistically non-signicant (RR = 0.82; 95% CI 0.61.13). When trials were analysed separately according to the presence of an additional diagnosis of opioid dependence, the result on heterogeneity held for the studies, including patients on methadone maintenance treatment (Kolar et al. 1992; Kosten et al. 1992; c2 = 4.24; d.f. = 1; P = 0.04). A small RCT (n = 17) (Kolar et al. 1992) showed an extremely non-signicant positive result favoring desipramine (RR = 0.16; 95% CI 0.021.04). The results for trials with primary cocaine dependence patients were more homogeneous, with no signicant differences between desipramine and placebo.
Addiction, 97, 931949

Pharmacological treatment of cocaine dependence

943

Regarding the pattern of cocaine use and craving, the following outcomes were used: duration of treatment in weeks; amount of grams used per day, days of cocaine use per week; percentage of abstinent days; Addiction Severity Index (ASI) drug scores at the end of the trial; craving intensity and frequency. Most of these data, however, were skewed and were not summarized. In other words, there was such a high level of variation in measures used across these studies that it was not possible to incorporate these data into the meta-analysis.

In a large RCT on pergolide (n = 464), a dopamine agonist, with 40% of participants presenting co-morbid alcohol and cocaine dependence (Malcolm et al. 2000), dropout rates favored placebo, without reaching statistical signicance. Figure 2 shows the dropout rates across different types of drugs. Side effects In general, trials did not show signicant results in terms of specic side effects between medications and placebo, which can be due to the small number of trials for some drugs and different methods of collecting information on this outcome. In CBZ studies, several side effects were described, including dermatological hypersensitivity reaction, dizziness, drowsiness, dry mouth, headache, nausea and vomiting, but there were no statistically signicant differences between CBZ and placebo. Only two ADs trials (McElroy 1989; Weddington et al. 1992) reported relevant data on side effects. A trend was found suggesting that patients on desipramine were more likely to present at least one side effect during the trial than those on placebo (RR = 1.65; 95% CI 0.982.79). Giannini et al. (1989) did not nd statistically signicant differences between amantadine and placebo in terms of side effects (diarrhoea, headache, nausea and rash). No signicant results were found regarding bromocriptine in terms of side effects and placebo. Direct comparisons Amantadine versus bromocriptine Results from two trials (Tennant & Tarver 1987; Giannini et al. 1989) did not nd any signicant difference between these two dopamine agonists for all efcacy measures, including patients subjective reports. However, a trend was observed favoring amantadine in terms of retention in treatment and the occurrence of side effects. When craving was assessed, a trend favoring bromocriptine was found in Gianninis trial. Amantadine versus desipramine Three trials compared these products (Weddington et al. 1991; Kolar et al. 1992; Kosten et al. 1992). The rst two included methadone-maintained subjects while Weddington evaluated cocaine dependence only. In general, there were no differences when retention in treatment and side effects were evaluated. A nonstatistically signicant difference favoring desipramine was found in methadone-maintained subjects, as revealed by positive urine samples for cocaine metabolites at the end of the trial.
Addiction, 97, 931949

Other clinical ratings A number of clinical ratings and scales were used in the studies, including the Beck Depression Inventory (BDI), the Spielberg State Anxiety Inventory (SSA), the Symptom Check List 90Revised (SCL-90-R) and the Patient Global Improvement (PGI), among others. However, relevant data were provided for a limited number of scales. Many continuous outcomes were described in terms of means without corresponding standard deviations, or actual measure units for baseline point and linear change per visit. Such data could not be summarized or presented in graphical form.

Dropoutsno retention in treatment In 10 studies, authors did not include descriptions of those who dropped out before the end of the trial protocol (Gawin et al. 1989; Giannini et al. 1989; Kranzler & Bauer 1992; Trifeman et al. 1992; Covi et al. 1993; Campbell et al. 1994; Hall et al. 1994; Rosse et al. 1994; Cold 1996; Carrol et al. 1998). A huge variation was seen in terms of dropout rates in the following studies, ranging from 0% (Giannini & Billet 1987patients from a private practice psychiatric group) to 84% (Weddington et al. 1991), suggesting that a very heterogeneous set of populations were compared. However, for most of studies a high number of patients who were not able to complete the protocol study was found. In general, the rate of patients remaining in treatment was similar for those taking medications or placebo. A non-signicant trend favoring CBZ was found in terms of dropouts (RR 0.88 95%CI 0.751.03), with four studies and 313 individuals included in the metaanalysis. No signicant differences on retention in treatment were obtained for desipramine, gepirone, ritanserin and imipramine trials compared with placebo. Similar results were found for dropout rates due to side effects. One signicant result was found by Batki et al. (1996): a higher percentage of patients on uoxetine completed the study, in comparison with placebo (RR = 0.53; 95% CI 0.320.88, n = 32).
2002 Society for the Study of Addiction to Alcohol and Other Drugs

944

Maurcio Silva de Lima et al.

Study 01 Bupropion Margolin et al. 1995a 02 Desipramine Arndt et al. 1992 Giannini et al. 1986 Kolar et al. 1992 Kosten et al. 1992 McElroy et al. 1989 O'Brien et al. 1988 Tennant & Tarver 1985 Weddington et al. 1991 Subtotal (95%Cl) 03 Fluoxetine Batki et al. 1996 04 Gepirone Jenkins et al. 1992 05 Imipramine Nunes et al. 1995 06 Ritanserin Ehrman et al. 1996 07 Amantadine Alterman et al. 1992 Handelsman et al. 1995 Kampman et al. 1996 Kolar et al. 1992 Kosten et al. 1992 Weddington et al. 1991 Subtotal (95%Cl) 08 Bromocriptine Eiler et al. 1995 Handelsman et al. 1995 Moscovitz et al. 1993 Subtotal (95%Cl) 09 Pergolide Malcolm et al. 2000 10 Carbamzaepine Cornish et al. 1995 Halikas et al. 1997 Kranzler et al. 1995 Montoya et al. 1995 Subtotal (95%Cl) 11 Disulfiram George et al. 2000 Petrakis et al. 2000 Subtotal (95%Cl) 12 Mazindol Margolin et al. 1995b Stine et al. 1995 Subtotal (95%Cl) 13 Naltrexone Somoza et al. 1998 14 Phenytoin Crosby et al. 1996 15 Risperidone 2mg/day Grabowski et al. 2000

Active drug n /N

Placebo n /N

RR (95%Cl Random)

Weight %

RR (95%Cl random)

11/74 17/53 2/10 0/8 8/30 5/9 8/32 5/11 27/32 72/185 8/16 9/20 24/59 11/40 5/21 2/23 24/30 2/5 8/33 27/32 68/144 17/32 6/24 9/14 32/70 111/156 29/46 35/58 9/20 19/28 92/152 3/11 11/36 14/47 3/18 7/22 10/40 18/24 23/29 23/30

13/75 3/26 0/10 5/9 4/31 5/6 1/15 5/11 16/28 39/136 15/16 11/21 27/54 13/40 5/21 3/25 19/31 5/9 4/31 19/23 55/140 17/31 4/26 10/15 31/72 89/153 34/45 42/62 7/20 27/34 110/161 2/9 4/31 6/40 4/19 8/21 12/40 15/22 25/31 42/45
0.1 0.2 Favors active drug 1 5 10 Favors placebo

100.0 11.8 2.5 2.8 12.5 20.5 5.0 15.4 29.5 100.0 100.0 100.0 100.0 100.0 2.9 1.2 31.1 2.3 2.9 59.7 100.0 52.6 8.5 39.0 100.0 100.0 33.4 35.1 4.3 27.2 100.0 30.8 69.2 100.0 26.9 73.1 100.0 100.0 100.0 100.0

0.86 [0.41,1.79] 2.78 [0.89,8.64] 5.00 [0.27,92.63] 0.10 [0.01,1.58] 2.07 [0.69, 6.15] 0.67 [0.34,1.32] 3.75 [0.51,27.33] 1.00 [0.40,2.50] 1.48 [1.04,2.10] 1.33 [0.82,2.15] 0.53 [0.32,0.88] 0.86 [0.46,1.62] 0.81 [0.54,1.22] 0.85 [0.43,1.66] 1.00 [0.34,2.95] 0.72 [0.13,3.96] 1.31 [0.94,1.82] 0.72 [0.21,2.44] 1.88 [0.63,5.62] 1.02 [0.80,1.30] 1.11 [0.92,1.33] 0.97 [0.61,1.53] 1.63 [0.52,5.07] 0.96 [0.57,1.64] 1.01 [0.73,1.41] 1.22 [1.03,1.45] 0.83 [0.63,1.10] 0.89 [0.68,1.17] 1.29 [0.60,2.77] 0.85 [0.63,1.16] 0.88 [0.75,1.03] 1.23 [0.26,5.82] 2.37 [0.84,6.69] 1.93 [0.82,4.59] 0.79 [0.21,3.06] 0.84 [0.37,1.90] 0.82 [0.41,1.66] 1.10 [0.76,1.59] 0.98 [0.76,1.27] 0.82 [0.66,1.02]

Figure 2 Dropout rates

Subgroup analysis Given the heterogeneity of patients, diagnoses and settings, subgroup analysis was performed separately for
2002 Society for the Study of Addiction to Alcohol and Other Drugs

two groups. Trials where patients had a primary diagnosis of cocaine dependence were compared to those where patients had diagnoses for both cocaine and opioid dependence or were in methadone maintenance treatAddiction, 97, 931949

Pharmacological treatment of cocaine dependence

945

ment. Results were very similar to those obtained for the main analysis, regardless type of drug, suggesting that there is no current evidence to support distinctions among patients with primary cocaine dependence or in methadone maintenance as far the pharmacological treatment of cocaine dependence is concerned. A number of miscellaneous trials are summarized in the gures. None of these trials were of substantial size and none of the results suggested any signicant differences. Available data in the original reports did not allow further subgroup analysis.

Antidepressants Methodological considerations A meta-analysis of desipramine for the treatment of cocaine addiction (Levin & Lehman 1991), including seven randomized studies with a total of 200 patients, found that desipramine is no better than placebo in retaining patients in treatment. However, it was also suggested that while patients were in treatment, desipramine is helpful in promoting abstinence (data from six trials). The authors state that the Fail-safe Nwhich provides an estimate of the number of additional negative studies that would be required to reverse a positive metaanalytical ndingin connection with desipramine efcacy is 16. In other words, 16 studies nding no advantage to desipramine compared with placebo would be necessary to negate this nding. These ndings were discussed by Delucchi (1992), who pointed out a number of limitations in the metaanalysis. The main problem was the use of an incorrect method for combining the studies chosen by the authors. Moreover, the main outcome measureabstinence was dened in many different ways across the studies. There are clear discrepancies between the results and conclusions of Levin & Lehman (1991) review and the present one. Different methods of combining studies were used, but other differences also need to be mentioned: 1 The rst review was published 8 years ago, and further randomized evidence has since become available: this review includes 14 RCTs (eight studies with available data for efcacy); 2 Efcacy data used in this review were combined only when authors dened efcacy in a similar and comparable way. All eight studies used the number of patients with positive samples for cocaine metabolites at the end of the trial. Therefore, the estimates from the rst metaanalysis could be inated by less restrictive denitions of efcacy. Alternatively, our results can be considered conservative ones, given the selection of outcomes, criteria for pooling analyses and the use of the random effects model, which takes into account heterogeneity between trials.

DISCUSSION Although in this review a large number of RCTs on the pharmacological treatment of cocaine dependence has been identied, no evidence of efcacy was found, regardless of the type of drug or dose used for all relevant outcomes assessed. A range of factors makes it difcult to draw conclusions from any synthesis of treatments for cocaine dependence. These include differences in psychiatric and substance use diagnoses, study quality and design, denitions of outcome variables, varying amounts of psychotherapy provided in conjunction with medications and the limitations of meta-analysis per se. Metaanalysis may be affected by biases of original studies and it may often not be the best method for studying the diversity of elds for which it has been used; this could be the case as far the addiction eld is concerned. Typical problems related to the use of meta-analytical approaches include: regressions are often non-linear; effects are often multivariate rather than univariate; coverage can be restricted; faulty studies may be included; the data summarized may not be homogeneous; grouping different causal factors may lead to meaningless effect sizes estimates; and the failure to relate data to theories may obscure discrepancies (Eysenck 1995). In this review, every effort was made to reduce biases, although the generally poor quality of reporting regarding both trial performance and outcome may have limited this aim. For example, the method of random assignment to treatment was seldom described and so 37 of the 46 studies were classied as having a moderate risk of selection bias, thus being more likely to favor the experimental treatment (Schulz et al. 1995). The key part of this discussion will focus on ADs including desipramine, and CBZ because the largest number of studies have been conducted on these medications. The other drugs reviewed included DAs and miscellaneous drugs but, overall, size of studies was limited, and there was little evidence to suggest benet from these medications.
2002 Society for the Study of Addiction to Alcohol and Other Drugs

Efcacy ndings Data from 18 RCTs involving 1379 subjects suggest lack of efcacy of ADs compared to placebo. Only three trials (Giannini et al. 1986; Giannini, Loiselle & Giannini 1987; Gawin et al. 1989) showed statistically signicant differences favoring the intervention (desipramine) in efcacy measures when compared to placebo. Most of the trials showed a similar decrease in cocaine use and craving in both the placebo and the active drug groups. When
Addiction, 97, 931949

946

Maurcio Silva de Lima et al.

relapse in cocaine-free patients was evaluated, the same pattern of response was found (McElroy et al. 1989). It was suggested that patients who had a history of opioid use were more likely to relapse when treated with ADs (Arndt et al. 1992), but current results were unable to conrm any relevant differences for any of the efcacy measures. There is still a limited number of RCTs assessing other agents such as uoxetine, ritanserin, gepirone, bupropion and imipramine. Previous experience with depression suggests they may have similar effects regardless of the proposed mechanism of action. When trials reported results on urinalysis (Jenkins et al. 1992; Covi et al. 1993) gepirone and uoxetine performances were very similar to those found for desipramine trials.

methods, the effects were modest, and the studies performed by a single group. Despite this rationale and the absence of a solid research base, the unfortunate consequence of the early publicity was that carbamazepine made its way into physicians practices (Johanson 1995). Halikas et al. (1997) found signicant results favoring the use 400 mg of carbamazepine, although therapeutic effects as assessed by clinicians showed that placebo performed signicantly better. In this same trial, a number of continuous outcome variables were subjected to sophisticated regression analysis. However, the most consistent efcacy outcomepositive urine analysis did not show any signicant difference between groups, even adopting the most positive results from Halikas et al. (400 mg/day).

Dropoutsno retention in treatment Although results from single trials suggest that patients on medications such as uoxetine are less likely to drop out, these results are from single trials and involve a limited number of patients. SSRIs are thought to have better acceptability, but this nding needs replication. Giannini et al. (1986) found the lowest dropout rate (10%). In this small study (n = 20), participants had a history of cocaine abuse of at least 1 year, and received supportive counselling at 5-day intervals. Batki et al. (1996) found the highest dropout rate: 72%. Most of the subjects in this trial (n = 32) had other psychiatric disorders such as major depressive disorder (40%), antisocial personality disorder (20%) and alcohol abuse and dependence (20%), and were paid to participate in the study: $10 for the intake and each of the weekly assessments. This nding is biased, possibly because of the very heterogeneous populations of clinical studies, and reinforces the view that a specic set of conditions may be associated with higher percentages of dropouts, including factors of co-morbidity, the absence of psychosocial support and the method of recruitment. It is plausible that the best results can be expected in highly motivated populations or associated with highly structured psychosocial interventions aimed at promoting and modulating motivation.

Dropoutsno retention in treatment Pooled results from ve RCTs involving 455 subjects suggest a lack of evidence regarding the efcacy of carbamazepine. Type II error could be an explanation for such ndings but other factors such as illness behavior must also be considered. However, the most prominent nding for these studies concern the high dropout rates. Although slightly favoring carbamazepine, these rates were high for both those taking carbamazepine (61%) and placebo (69%). These high dropout rates may be inherent to the type of drug problem and to its severity. The urgent demand by clinicians, patients, families and the community as a whole for an adequate treatment for cocaine dependence may lead to the adoption of therapeutic regimes even if the evidence of their efciency is weak. Alternatively, it is plausible that carbamazepine illustrates a common problem of extrapolating results from preclinical studies to clinical effects in adults, which are not necessarily related to demonstrated and signicant clinical effects. In the addiction eld, there is a high correlation between compliance with prescribed treatment and clinical outcome (Meyer 1992). In open trials such as the one conducted by Halikas et al. patients may be motivated and are not representative of cocaine-dependent patients in the general population. Cocaine-dependent subjects who participated in clinical trials may manifest vastly differing degrees of motivation for change. Such motivation may improve overall outcome signicantly but to date this has not been well demonstrated empirically. Ideally, successful pharmacological intervention should act independently of level of patient motivation for change. Theoretically, motivation need not be a prerequisite for the use of anticraving medication (Halikas et al. 1991). Motivation may also be inuenced signicantly by the overall quality of the program but this is equal for both active drugs and placebo subjects.
Addiction, 97, 931949

Carbamazepine Methodological considerations The efcacy of CBZ for the treatment of cocainedependent patients was rst suggested by an open trial with 35 participants (Halikas et al. 1989). These same authors conducted a cross-over randomized controlled study 2 years later conrming this nding (n = 32). Such promising results, however, were based on limited
2002 Society for the Study of Addiction to Alcohol and Other Drugs

Pharmacological treatment of cocaine dependence

947

Implications for practice With the evidence currently available, there are no data supporting the efcacy of antidepressants, carbamazepine, dopamine agonists, or other drugs such as disulram, naltrexone and mazindol for the treatment of cocaine dependence. In the absence of more reliable evidence, clinicians may consider prescribing alternative medications, where a higher number of studies and patients evaluated are available (for instance, ADs). Pharmacological treatments that affect the consumption of other substances such as alcohol may also have a role to play. Disulram as a treatment agent requires further exploration; however, given the difculty in determining its role in alcohol dependence, further studies are required before any conclusions can be drawn on its impact in cocaine dependence. It seems unlikely that, in the absence of motivation for behavior change, pharmacological agents, with the possible exception of cocaine-specic blocking or maintenance agents, are able to promote a signicant improvement in behavior. The value of medications such as ADs, prescribed in conjunction with a more potent psychosocial intervention, remains unknown. However, until further efcacy and effectiveness studies are available, clinicians may consider adding psychosocial and psychotherapeutic supportive measures aiming to keep patients in treatment. Such advice does not rely on any direct evidence from RCTs, but does consider the best available evidence, the high dropout rates and illness behavior associated with cocaine use. The use of ADs in the absence of depression is not supported by current evidence of impact. This does not preclude the use of ADs to treat major depressive illness that is associated with cocaine abuse and dependence but would clarify that the antidepressant does not have any specic anticraving effect.

Implications for research There is a need for further exploration of other medications but there is also a need to clarify the current available evidence for psychosocial interventions and to ensure that current psychosocial treatment programs are organized to deliver the greatest possible treatment impact (Carroll et al. 1994). Further studies need also to take account of the organization of the treatment setting when planning trials. The unusually high dropout rates across studies and drugs, rather than reecting a simple methodological aw, may suggest that specic compliance promoting approaches are needed to investigate clinical effects of drugs for the treatment of cocaine dependence. If compliance can be improved through psychosocial interven 2002 Society for the Study of Addiction to Alcohol and Other Drugs

tions, what then would be the optimal duration of pharmacological treatment? Is medication still useful after the natural resolution of the withdrawal phase of abstinence symptoms? In reviewing RCTs on the treatment of cocaine dependence, some outcome measures seem to be more reliable and useful than others. Trialists can choose to use continuous data from one craving or psychiatric scale, relying mainly on those that are widely used and validated. The ASI seems to be a useful instrument, given that it is designed to assess the problem of severity in seven areas affected commonly by addiction. Equivalent instruments such as the Maudsley Addiction Prole (Marsden 1998) could be considered, but overall it is critical that validated instruments are used in such studies. Dichotomous variables are interpreted more easily by clinicians, and provide estimations of effect size and number needed to treat/number needed to harm, particularly in respect to: 1 the number of subjects with positive urine samples for cocaine metabolites at end-point; 2 self-reported use of cocainealthough less reliable, these results can be compared to those from urinalysis; 3 retention in treatment (overall), and 4 retention in treatment for side-effects. The absence of clinical effects based on a small number of trials may lead to a general conclusion that larger RCTs are needed. However, in light of the data reviewed, it could be that this is not justied, due to the lack of any clinically relevant results regardless of the type of medication. More than trends and a small number of signicant results, on which the current evidence relies, are needed. Larger randomized investigations are expensive and time-consuming and should perhaps be reserved for medications showing more relevant and promising evidence. Given the relative lack of success of drugs assessed in this review, it may be necessary to conduct a new set of clinical studies, characterized by research designs that verify expressly hypotheses concerning possible interactions between treatment and relevant patient characteristics. There is a need for a systematic review of psychosocial interventions, as at present the prospects for health and social gain through psychosocial treatments appears substantially more promising than currently available pharmacotherapies.

REFERENCES
Alim, T. N., Roose, R. B., Vocci, F. J., Lindquist, T. & Deutsch, S. I. (1995) Diethylpropion pharmacotherapeutic adjuvant therapy for inpatient treatment of cocaine dependence: a test of the cocaine-agonist hypothesis. Clinical Neuropharmacology, 8, 183195.
Addiction, 97, 931949

948

Maurcio Silva de Lima et al.

Alterman, A. I., Droba, M., Antelo, R. E., Cornish, J. W., Sweeney, K. K., Parikh, G. A. & OBrien, C. P. (1992) Amantadine may facilitate detoxication of cocaine abusers. Drug and Alcohol Dependence, 31, 1929. Altman, D. G. & Bland, J. M. (1996) Detecting skewness from summary information. British Medical Journal, 313, 1200. Arndt, I. O., Dorozynsky, L., Woody, G. E., McLellan, A. T. & OBrien, C. P. (1992) Desipramine treatment of cocaine dependence in methadone-maintained patients. Archives of General Psychiatry, 49, 888893. Batki, S. L., Washburn, A. M., Deluchi, K. & Jones, R. T. (1996) A controlled trial of uoxetine in crack cocaine dependence. Drug and Alcohol Dependence, 41, 137142. Campbell, J. L., Thomas, H. M., Gabrelli, W., Liskow, B. I. & Powell, B. J. (1994) Impact of desipramine or carbamazepine on patient retention in outpatient cocaine treatment: preliminary ndings. Experimental Therapeutics in Addiction Medicine, 13, 191199. Carroll, K. M., Nich, C., Ball, S. A., McCance, E. & Rounsaville, B. J. (1998) Treatment of cocaine and alcohol dependence with psychotherapy and dissulram. Addiction, 93, 713728. Carroll, K. M., Rounsaville, B. J., Gordon, L. T., Nich, C., Jatlow, P., Bisighini, R. M. & Gawin, F. H. (1994) Psychotherapy and pharmacotherapy for cocaine abusers. Archives of General Psychiatry, 51, 177187. Cold, J. A. (1996) NeuRecover-SA in the treatment of cocaine withdrawal and craving. A pilot study. Clinical Drug Investment, 12, 17. Cornish, J. W., Manny, I., Fudala, P. J., Neal, S., Poole, S. A., Volpicelli, P. & OBrien, C. P. (1995) Carbamazepine treatment for cocaine dependence. Drug and Alcohol Dependence, 38, 221227. Covi, L., Hess, J. M., Kreiter, N. A. & Haertzen, C. A. (1995) Effects of combined uoxetine and counselling in the outpatient treatment of cocaine abusers. American Journal of Drug and Alcohol Abuse, 21, 327344. Crosby, R. D., Pearson, V. L., Eller, C., Winegarden, T. & Graves, N. L. (1996) Phenytoin in the treatment of cocaine abuse: a double-blind study. Clinical Pharmacology and Therapeutics, 59, 458468. Delucchi, K. L. (1992) Research on desipramine in the treatment of cocaine abuse: a critique of Levin and Lehmans metaanalysis. Journal of Clinical Psychopharmacology, 12, 367370. Dersimonian, R. & Laird, N. (1986) Meta-analysis in clinical trials. Controlled Clinical Trials, 7, 177178. Ehrman, R. N., Robbins, S. J., Cornish, J. W., Childress, A. R. & OBrien, C. P. (1996) Failure of ritanserin to block cocaine cue reactivity in humans. Drug and Alcohol Dependence, 42, 167174. Eiler, K., Schaefer, M. R., Salstrom, D. & Lowery, R. (1995) Double-blind comparison of bromocriptine and placebo in cocaine withdrawal. American Journal on Drug and Alcohol Abuse, 21, 6579. European Monitoring Centre for Drugs and Drug Addiction (EMCDDA) (2000) Annual Report on the State of the Drugs Problem in the European Union. Luxembourg: Ofce for Ofcial Publications of the European Communities. Eysenck, H. J. (1995) Problems with meta-analysis. In: Chalmers, I. & Altman, D. G., eds. Systematic Reviews, pp. 6774. London: BMJ Publishing Group. Gawin, F. H., Kleber, H. D., Byck, R., Rounsaville, B. J., Kosten, T. R., Jatlow, P. I. & Morgan, C. (1989) Desipramine facilitation of initial cocaine abstinence. Archives of General Psychiatry, 46, 117121.
2002 Society for the Study of Addiction to Alcohol and Other Drugs

George, T. P., Chawarski, M. C., Pakes, J., Carroll, K. M., Kosten, T. R. & Schottenfeld, R. S. (2000) Disulram versus placebo for cocaine dependence in buprenorphine-maintained subjects: a preliminary trial. Biological Psychiatry, 47, 1080 1086. Giannini, J. A., Baumgartel, P. & Dimarzio, L. R. (1987) Bromocriptine therapy in cocaine withdrawal. Journal of Clinical Pharmacology, 27, 267270. Giannini, J. A. & Billet, W. (1987) Bromocriptinedesipramine protocol in treatment of cocaine addiction. Journal of Clinical Pharmacology, 27, 549554. Giannini, J. A., Malone, D. A., Giannini, M. C., Price, W. & Loiselle, R. H. (1986) Treatment of depression in chronic cocaine and phencyclidine abuse with desipramine. Journal of Clinical Pharmacology, 26, 211214. Giannini, J. A., Loiselle, R. H. & Giannini, M. C. (1987) Spacebased abstinence: alleviation of withdrawal symptoms in combinative cocainephencyclidine abuse. Clinical Toxicology, 25, 493500. Giannini, J. A., Folts, J. D., Feather, N. J. & Sullivan, S. B. (1989) Bromocriptine and amantadine in cocaine detoxication. Psychiatry Research, 29, 1116. Gold, M. S. (1997) Cocaine (and crack): clinical aspects. In: Lowinson, J. H., Ruiz, P., Millman, R. B. & Langrod, J. G., eds. Substance Abuse: a Comprehensive Textbook, 3rd edn, vol. 16, pp. 181199. New York: Lippincott, Williams & Wilkins. Grabowski, J., Rhoades, H., Silverman, P., Schmitz, J. M., Stotts, A., Creson, D. & Bailey, R. (2000) Risperidone for the treatment of cocaine dependence: randomized, double-blind trial. Journal of Clinical Psychopharmacology, 20, 305310. Handelsman, L., Limpitlaw, L., Williams, D., Schmeidler, J., Paris, P. & Stimmel, B. (1995) Amantadine does not reduce cocainedependent methadone maintenance patients. Drug and Alcohol Dependence, 39, 173180. Handelsman, L., Rosenblum, A., Palij, M., Magura, S., Foote, J., Lovejoy, M. & Stimmel, B. (1997) Bromocriptine for cocaine dependencea controlled clinical trial. American Journal on Addictions, 6, 5464. Halikas, J. A., Crosby, R. D., Pearson, V. L. & Graves, N. M. (1997) A randomised double-blind study of carbamazepine in the treatment of cocaine abuse. Clinical Pharmacology and Therapeutics, 62, 89105. Hall, S. M., Tunis, S., Trifeman, E., Banys, P., Clark, H. W., Tusel, D., Stewart, P. & Presti, D. (1994) Continuity of care and desipramine in primary cocaine abusers. Journal of Nervous and Mental Disease, 182, 570575. Jenkins, S. W., Wareld, N. A., Blaine, J. D., Cornish, J., Ling, W., Rosen, M. I., Urschel, H., Wesson, D. & Ziedonis, D. (1992) A pilot trial of gepirone vs. placebo in the treatment of cocaine dependence. Psychopharmacology Bulletin, 28, 2126. Johanson, C. E. (1995) Why publish three negative articles on carbamazepine as a medication for the treatment of cocaine dependence? Drug and Alcohol Dependence, 38, 201202. Kampman, K., Volpicelli, P., Alterman, A., Cornish, J., Weinrieb, R., Epperson, L., Sparkman, T. & OBrien, C. P. (1996) Amantadine in the early treatment of cocaine dependence: a double-blind, placebo-controlled trial. Drug and Alcohol Dependence, 1996, 2533. Kolar, A. F., Brown, B. S., Weddington, W. W., Haertzen, C. C., Michaelson, B. S. & Jaffe, J. H. (1992) Treatment of cocaine dependence in methadone maintenance clients: a pilot study comparing the efcacy of desipramine and amantadine. International Journal of the Addictions, 27, 849868. Kosten, T. R., Morgan, C. M., Falcione, J. & Schottenfeld, R. S.
Addiction, 97, 931949

Pharmacological treatment of cocaine dependence

949

(1992) Pharmacotherapy for cocaine-abusing methadonemaintained patients using amantadine or desipramine. Archives of General Psychiatry, 49, 894898. Kranzler, H. R., Bauer, L. O., Hersh, D. & Klinghoffer, V. (1995) Carbamazepine treatment of cocaine dependence: a placebocontrolled trial. Drug and Alcohol Dependence, 38, 203211. Kranzler, H. R. & Bauer, L. O. (1992) Bromocriptine and cocaine reactivity in cocaine-dependent patients. British Journal of Addiction, 87, 15371548. Levin, R. L. & Lehman, A. F. (1991) Meta-analysis of desipramine as an adjunct in the treatment of cocaine addiction. Journal of Clinical Psychopharmacology, 11, 374378. Malcolm, R., Kajdasz, D. K., Herron, J., Anton, R. F. & Brady, K. T. (2000) A double-blind placebo controlled outpatient trial of pergolide for cocaine dependence. Drug and Alcohol Dependence, 60, 161168. Margolin, A., Kosten, T. R., Avants, S. K., Wilkins, J., Ling, W., Beckson, M., Arndt, I. O., Cornish, J., Ascher, J. A., Li, S. H. & Bridge, P. (1995a) A multicenter trial of bupropion for cocaine dependence in methadone-maintained patients. Drug and Alcohol Dependence, 40, 125131. Margolin, A., Avants, S. K. & Kosten, T. R. (1995b) Mazindol for relapse prevention to cocaine abuse in methadonemaintained patients. American Journal of Drug Abuse, 21, 469481. Marsden, J., Gossop, M., Stewart, D., Best, D., Farrell, M., Lehman, P., Edwards, C. & Strang, J. (1998) The Maudsley Addiction Prole (MAP), a brief instrument for assessing treatment outcome. Addiction, 93, 18571867. McElroy, S. L., Weiss, R. D., Mendelson, J. H., Teoh, S. K., McAfee, B. & Mello, N. K. (1989) Desipramine treatment for relapse prevention in cocaine dependence. National Institute on Drug Abuse Research Monograph Series, 95, 5763. Meyer, R. E. (1992) New pharmacotherapies for cocaine dependence . . . revisited. Archives of General Psychiatry, 49, 900904. Montoya, I. D., Levin, F. R., Fudala, P. J. & Gorelick, D. A. (1995) Double-blind comparison of carbamazepine and placebo for treatment of cocaine dependence. Drug and Alcohol Dependence, 38, 213219. Moscovitz, H., Brookof, D. & Nelson, L. (1993) A randomised trial of bromocriptine for cocaine users presenting to the emergency department. Journal of General Internal Medicine, 8, 14. Musto, D. F. (1992) Cocaines history, especially the American experience. Ciba Foundation Symposium, 166, 714. Nunes, E. V., McGrath, P. J., Quitkin, F. M., Ocepek-Welikson, K., Stewart, J. W., Koening, T., Wager, S. & Klein, D. F. (1995) Imipramine treatment of cocaine abuse: possible

boundaries of efcacy. Drug and Alcohol Dependence, 39, 185195. OBrien, C. P., Childress, A. R., Arndt, I. O., Mclellan, A. T., Woody, G. E. & Maany, I. (1988) Pharmacological and behavioural treatments of cocaine dependence: controlled studies. Journal of Clinical Psychiatry, Suppl. 49, 1722. Petrakis, I. L., Carroll, K. M., Nich, C., Gordon, L. T., McCanceKatz, E. F., Frankforter, T. & Rousanville, B. J. (2000) Disulram treatment for cocaine dependence in methadonemaintained opioid addicts. Addiction, 95, 219228. Rosse, R. B., Alim, T. N., Fay-McCarthy, M., Collins, J. P., Vocci, F. J., Lindquist, C. J., Hess, A. L. & Deutsch, S. I. (1994) Nimodipine pharmacotherapeutic adjuvant therapy of inpatient treatment of cocaine dependence. Clinical Neuropharmacology, 17, 348358. Schulz, K. F., Chalmers, I., Hayes, R. J. & Altman, D. G. (1995) Empirical evidence of bias: dimensions of methodological quality associated with estimates of treatment effects in controlled trials. Journal of the American Medical Association, 273, 408412). Simpson, D. D., Joe, G. W., Fletcher, B. W., Hubbard, R. L. & Anglin, M. D. (1999) A national evaluation of treatment outcomes for cocaine dependence. Archives of General Psychiatry, 56, 507514. Somoza, E., Carter, J., Upadhyaya, H., Mezinskis, J., Lutz, J., Geracioti, J., Dyrenforth, S. & Doan, G. (1998) A double-blind, placebo controlled clinical trial of naltrexone as a treatment for cocaine dependence. In: Harris, L. S., ed. Problems of Drug Dependence 1998: Proceedings of the 60th Annual Scientic Meeting, the College on Problems of Drug Dependence, Inc., p. 138. Research Monograph No. 179, p. 295. Bethesda, MD: National Institute on Drug Abuse. Stine, S. M., Krystal, J. H., Kosten, T. R. & Charney, D. S. (1995) Mazindol treatment for cocaine dependence. Drug and Alcohol Dependence, 39, 245225. Tennant, F. S. & Tarver, A. L. (1985) Double-blind comparison of desipramine and placebo in withdrawal from cocaine dependence. National Institute on Drug Abuse Research Monograph Series, 55, 159163. Trifeman, E., Deluchi, K., Tunis, S., Banys, P. & Hall, S. (1992) Desipramine in the treatment of crack cocaine dependence. Preliminary results. National Institute on Drug Abuse Research Monograph Series, 132, 317. Weddington, W. W., Brown, B. S., Haertzen, C. A., Hess, J. M., Mahaffey, J. R., Kolar, A. F. & Jaffe, J. H. (1991) Comparison of amantadine combined with psychotherapy for treatment of cocaine dependence. American Journal on Drug and Alcohol Abuse, 17, 137152.

2002 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction, 97, 931949