The Unsolved Problem of Attrition Rates on Randomized Clinical Trials for Cocaine Use Disorders: A Scoping Review

Abstract

Background:

Cocaine use disorder (CUD) is a significant and insufficiently studied public health issue, especially considering that the global prevalence of CUD is estimated to be higher than ever. There is still no consensus on effective treatments for CUD. Important barriers for research in the field include the high attrition levels observed in randomized controlled trials (RCTs) for CUD treatment and the lack of emphasis on methods to reduce attrition in CUD RCTs.

Methods:

The goal of this study was to systematically review over 2 decades of CUD RCTs, with the objective of evaluating the reporting of attrition bias and methods used to mitigate attrition.

Results:

Our scoping review extracted information from 106 RCTs, of which only 82 explicitly evaluated attrition as an outcome. Thirty-eight studies had an attrition rate above 50%, and five 16 studies had medium attrition bias, 6% to 19%. The remaining 68 had large attrition bias.

Conclusion:

Across all included studies, discussion of attrition as a limitation was uncommon. Overall, these analyses suggest that most RCTs evaluating CUD treatments have not adequately accounted for attrition in their analyses or employed approaches to mitigate attrition.

Introduction

While much attention has been drawn to opioid and amphetamine use disorders in recent years, cocaine use disorder (CUD) remains one of the most common and fastest-growing substance use disorders (SUDs) worldwide. In the United States, cocaine-related overdoses have been found to have mortality rates as high as those opioid-related overdoses, particularly in minoritized patient groups.¹ In South America, Brazil has the highest CUD-related disease burden globally, with its 2019 per 100 000 rates nearly doubling its 1990 rate.² Worldwide, cocaine manufacturing is estimated to be at its highest level ever.³

Relative to opioids and amphetamines, CUD remains an understudied public health problem worldwide, even though cocaine-related problems are as likely to result in hospital admission as opioid-related problems.⁴ Despite many years of effort, no Food and Drug Administration pharmacologic treatments were approved for CUD.⁵ To date, there remains a limited consensus on medications or psychotherapies deemed to be effective in reducing cocaine use, promoting abstinence, and avoiding relapses among people with CUD. Throughout the past 40 years, many medications have been studied, with different levels of success, but until now none of them has proven its efficacy. Dopamine agonists and antagonists, antidepressants, anticonvulsants, prescription psychostimulants, naltrexone, among others have already been tested.⁶

CUD trials have historically struggled with low attrition rates, which adds heavy bias to the trials’ results.⁶ A major limitation in research concerning treatment for CUD is the inadequate understanding of the factors related to the attrition rate, whose extent can be as high as 80% on the treatment of CUD and its impact on patients’ outcomes.⁷ Broadly speaking, the estimated effectiveness of SUD treatments is often biased by treatment discontinuation, a significant obstacle to assessing whether the results of randomized controlled trials (RCTs) can be generalized to the real-world.⁸ In clinical settings, research from several other SUDs (not CUD) has indicated that treatment discontinuation is often the norm rather than the exception. For instance, an analysis of real-world administrative data in an outpatient addiction treatment facility in Brazil found an attrition rate of 82% from November 1997 to April 2014, with a median treatment duration of 15 days.⁹ In an American sample of patients receiving pharmacological treatment for opioid use disorder, 64% were discontinued treatment before 6 months of initiation.¹⁰ The co-occurrence of other SUDs together with CUD was also found to exacerbate attrition rates.¹¹

These high attrition rates in real-world settings raise important questions about the significance of treatment discontinuations and whether CUD RCTs are factoring attrition into their study designs and manuscript discussion sections. It is known that attrition can introduce bias if the characteristics of participants with missing data differ between the randomized groups. For instance, ineffective treatment may appear ostensibly effective if subjects with a more severe burden of illness discontinue their assigned treatment.¹¹ High attrition rates may hinder the external validity of RCT data, especially if attrition is not explicitly discussed in the limitations section of RCTs.

A vivid example of the severe impact of study attrition on RCTs in addiction science can be seen in¹² the X: BOT trial, which compared extended-release naltrexone (XR-NTX) to buprenorphine-naloxone (BUP-NX) for opioid relapse prevention. Although the study considered both drugs to be equally effective toward relapse-free survival and overdose reduction, what ostensibly appears to be equipoise XR-NTX and BUP-NX, this conclusion is complicated by the high treatment discontinuation rates associated with XR-NTX. In other words, even though XR-NTX and BUP-NX could be considered equally effective once initiated, the high attrition rates associated with XR-NTX (28% vs 5.9% in BUP-NX) rendered the medication to be unpractical in real-world scenarios due to its demanding detoxification requirements compared with BUP-NX. Reanalyses of the X: BOT trial using statistical measures that account for induction failures (attrition) have shown an increased risk of overdose in the XR-NTX arm, contrasting with the results of the original X: BOT report.¹³ Ultimately, the X: BOT trial suggests that the inability to account for study attrition in RCTs is an important obstacle to high-quality comparative efficacy research in drug addiction.

We thus sought to evaluate the reporting of attrition in CUD RCTs. Because it is important for RCT protocols to establish strategies to minimize bias in estimating treatment effects, we conducted a scoping review of RCTs for CUD treatment. We searched whether the RCTs accounted for attrition bias in their statistical analyses (ie, imputation methods), employed attrition-mitigation measures in their study designs (ie, psychosocial supports, contingency management [CM]), and addressed attrition in their discussion of limitations.

The Cochrane Handbook for Systematic Reviews of Interventions, a standard methodology for evidence synthesis, ascertains that assessing the risk of bias (RoB) is of utmost importance for the included studies in a Systematic Review since a large RoB in this context can negatively impact the accuracy of what is being estimated. One of the evaluated domains is incomplete outcome data (attrition bias), which arises from incomplete data due to the nature or the way of handling missing outcomes. The ability to assess the RoB will be limited if the authors do not collect and report reasons that outcome data were missing. One technique commonly used for handling missing data is imputation. Although it is commonly used, the Cochrane Handbook^14,15 emphasizes skepticism toward imputation methods for addressing attrition, highlighting their limitations in mitigating bias when missingness in outcomes is related to their true values, unless such missingness can be accounted for by measured variables.

Methods

Study Selection

We elected to use a scoping review methodology due to our broad aim of assessing how a heterogeneous CUD literature base incorporated attrition into their analyses.¹⁶ Specifically, scoping reviews are useful for assessing “emerging evidence when it is still unclear what other, more specific questions can be posed and valuably addressed by a more precise systematic review.”¹⁶ We defined study eligibility criteria based on the PCC mnemonic (population, concept, and context) for scoping reviews. The population of interest was people with CUD; the concept of interest was attrition rate. With regard to context, we were interested in assessing the reporting of attrition among RCTS.

RCT that evaluated pharmacological, psychosocial, or both interventions for the treatment of CUD and that reported data on adherence, compliance, or attrition rate published until May 1, 2024, were thus searched from the PubMed database, LILACS, and PsycINFO. Including clinical trial studies were thoroughly examined.

The search strategy included the following key terms adapted to each database: “cocaine-related disorders,” “cocaine abuse,” “cocaine-related disorder,” “cocaine addiction,” “cocaine dependence,” “patient compliance,” “patient adherence,” “patient cooperation,” “patient noncompliance,” “treatment adherence,” “therapeutic compliance,” “therapeutic adherence,” “treatment adherence,” and “compliance.” Citations were de-duplicated, and this strategy was adapted to each database. The specific search strategy for each database is available in the Supplemental Material.

Search Criteria

The scoping review was completed in 2 stages: (1) screening of titles and abstracts and (2) full-text review. We imported references and de-duplicated results, and 2 independent, parallel reviewers subsequently screened titles and abstracts. Next, the same 2 reviewers conducted a full-text review and met again to reach a consensus on the final list of included articles, erring on inclusion.

Inclusion criteria consisted of RCTs that evaluated pharmacological, psychosocial, or both interventions for the treatment of CUDs. Studies that did not have information on attrition and retention were excluded from this review.

Efforts were made to contact selected authors to clarify methodological questions. Twenty-four papers were not included in Table 2, as they did not provide additional information for an in-depth analysis of their attrition rate. To cover the gray literature, all the references of the papers included were reviewed. In addition, experts in the field (professors and researchers from Australia, Canada, Europe, and the United States, who actively collaborate with our research group) were requested to send any additional paper suggestions.

Table 2.

Classification According to Cochrane Handbook for Systematic Reviews of Interventions (Number of Studies = 84^a).

Author	Medication	Psychotherapy	CM	IOP/TAU/ST	Intensive treatment	Counseling/community reinforcement	Attrition rate (%)b
Carroll, 2016a	X		X				Medium
Nuijten, 2016	X						Medium
Petry, 2015			X	X			Medium
Petitjean, 2014		X	X				Medium
McKay, 2014			X	X		X	Medium
Petry, 2012			X	X			Medium
Petry, 2007			X	X			Medium
Ciraulo, 2005 CREST 1	X	X					Medium
Petry, 2005b			X	X			Medium
Winhusen, 2005	X	X					Medium
Higgins, 2003			X			X	Medium
Marlowe, 2003			X	X	X		Medium
Rawson, 2002	X	X	X				Medium
Johnson, 1997	X			X			Medium
Silverman, 1996			X				Medium
Margolin, 1995	X						Medium
Levin, 2020	X						High
Carroll, 2018	X	X		X			High
Levin, 2015	X	X					High
Nuijten, 2015	X	X					High
Tapp, 2015	X						High
Mancino, 2014	X						High
Nuijten, 2014	X	X					High
Umbricht, 2014	X		X				High
Hagedorn, 2013			X	X			High
Johnson, 2013	X						High
Kampman, 2013	X	X					High
Somoza, 2013	X						High
McKay, 2012				X	X		High
Dackis, 2012	X	X					High
García-Fernández, 2011			X			X	High
Schottenfeld, 2011		X	X			X	High
Winstanley, 2011	X		X				High
Brodie, 2009	X	X					High
Garcia-Rodriguez, 2009			X	X			High
Mooney, 2009	X	X	X				H/igh
Loebl, 2008	X						High
Pettinati, 2008	X	X					High
Shoptaw, 2008	X					X	High
Ghitza, 2007			X				High
Gonzalez, 2007	X						High
Levin, 2007	X	X					High
McKee, 2007		X					High
Winhusen, 2007a	X						High
Winhusen, 2007b	X						High
Elkashef, 2006	X	X					High
Petry, 2006		X		X			High
Pierce, 2006			X	X			High
Poling, 2006	X		X				High
Ciraulo, 2005	X						High
Ciraulo, 2005 CREST 2	X	X					High
Dackis, 2005	X	X					High
Lambert Passos, 2005	X	X					High
Malcom, 2005	X	X					High
McDowell, 2005	X						High
Petry, 2005a			X	X			High
Grabowski, 2004 I	X						High
Grabowski, 2004 II	X						High
Jones, 2004	X		X				High
Petry, 2004			X	X			High
Campbell, 2003	X						High
Schubiner, 2002	X	X					High
Cornish, 2001	X						High
Grabowski, 2001	X						High
Schmitz, 2001a	X	X					High
Schmitz, 2001b	X	X					High
Higgins, 2000			X				High
Malcom, 2000	X	X					High
Crits-Christoph, 1999		X				X	High
Kirby, 1998a			X				High
Kirby, 1998b			X				High
Grabowski, 1997	X						High
Halikas, 1997	X		X				High
Batki, 1996	X						High
Crosby, 1996	X					X	High
Hoffman, 1996		X		X	X		High
Kampman, 1996	X					X	High
Cornish, 1995	X					X	High
Carroll, 1994a	X	X					High
Higgins, 1994			X				High
Montoya, 1994	X					X	High
Arndt, 1992	X						High
Kosten, 1992	X						High
Weddington, 1991	X						High

Abbreviations: CM, Contingency management; IOP, intensive outpatient program; TAU, Treatment as usual; ST, Standard treatment.

^aEighty-four of the 108 included studies are included in this table. The other 24 did not present the percentage of attrition rate (they presented, for instance, number of days under treatment)

^bPercentage of de intervention group attrition, according to the Cochrane Handbook for Systematic Reviews of Interventions (high >20%, medium 6%-19%, low <5%).

Data Extraction

Two double-blinded reviewers independently extracted data using a structured form containing the following variables: (a) author and publication year, (b) study setting, (c) primary outcomes, (d) available information on retention to treatment and attrition rates, (e) details of the evaluated data, (f) main findings, and (g) study limitations. Discrepancies were resolved by a consensus between the 2 double-blind reviewers and consultation with the senior author.

Data Analysis

A narrative summary of the findings from the studies included in the review was provided. PRISMA guidelines are included in the Supplemental Material.

The Cochrane Handbook for Systematic Reviews of Interventions considers that an attrition rate of less than 5% characterizes a “small” RoB.¹⁴ In contrast, greater than 20% attrition has been regarded as having a “high” RoB. This recommendation was used to classify the attrition rates of the intervention groups of the studies in “small,” “medium,” and “high.” Among studies with data on attrition, we subsequently counted the number of studies that reported their attrition rates as potential limitations of their analyses. We also assessed the number of studies that used imputation procedures to adjust for attrition and employed attrition-mitigation strategies in their RCT design (ie, CM to improve attrition in a pharmacologic trial, high-contact structured follow-up between study staff and participants).

Results

Characteristics of Included Studies

We identified 341 studies in our preliminary search. Of these, 24 were excluded for duplicates, and after reviewing the titles, 143 were excluded because they did not meet the inclusion criteria. The 174 articles were subject to thorough examination, of which 68 were subsequently excluded based on comprehensive full-text scrutiny, data not available, and pilot studies. Key reasons for exclusion were non-RCT status and insufficient attrition rate reporting. Detailed information is presented in Figure 1. This process yielded 106 articles that aligned with our predetermined inclusion criteria. A comprehensive breakdown of each article is presented in Table 1. Furthermore, it is worth noting that 2 articles encompassed 2 distinct constituent studies, the data of which were analyzed separately, resulting in a total of 108 trials.

Figure 1.

Flowchart for selection clinical trial articles.

Table 1.

Data Extraction Table of Randomized Controlled Clinical Trial Articles Design to Evaluate Treatments for Cocaine Use Disorders.

Author, year	Country	Sample Size	Age range	Sample details	Objective	Follow-up	Groups studied	Outcome	Primary findings	Attrition rate	Study limitation
Weddington, 1991	USA	54	28.4–31.2	Subjects recruited subjects by newspaper and radio advertisements and emergency rooms	To examine the results of initial treatment with Amantadine and desipramine as an adjunct to counseling	12 weeks	• Desipramine • Amantadine • Placebo	• Cocaine use • Abstinence • Craving • Retention	• Groups demonstrated decreases in cocaine use, craving for cocaine, and psychiatric symptoms • Trend for subjects treated with desipramine to maintain longer periods of cocaine abstinence • Trend for more dropouts by subjects taking desipramine	• Desipramine: 70.6% • Amantadine: 75% • PLA: 57.2%	• Retention
Arndt, 1992	USA	59	20–50	Methadone-maintained subjects	To evaluate the effectiveness of Desipramine hydrochloride in methadone-maintained subjects	12 weeks	• Desipramine hydrochloride • Placebo	• Cocaine use • Attrition rate	• There were significantly more dropouts in the desipramine than in the placebo group • Desipramine were significantly better psychiatric status than the placebo group	• 25.3% • Desipramine: 32.1% • PLA: 11.5%	• Only included men
Kosten, 1992	USA	94	32	Methadone-maintained subjects	To compare amantadine, desipramine, and placebo in the treatment of cocaine abusing methadone-maintained patients	12 weeks	• Amantadine • Desipramine • Placebo	• Cocaine use • Craving • Dropout	• Cocaine abuse was significantly lower in the medicated groups at week 4	• Amantadine: 24% • Desipramine: 27% • PLA: 13%	• Do not consider buprenorphine • Heterogeneous subgroup
Carroll, 1994a	USA	139	28.8	Subjects were outpatient clinic and respondents to newspaper	To evaluate psychotherapy and pharmacotherapy, alone and in combination	12 weeks	• RP + Des • Cl.M + Des • RP + PLA • Cl.M + PLA	• Attrition rate • Cocaine use	• Significant reductions in cocaine use in all the groups • Baseline severity was found to interact differently with psychotherapy and pharmacotherapy	‐ 60% • RP + Des: 51% • Cl.M + Des: 63% • RP + PLA: 64% • Cl.M + PLA: 61%	• Associated outcome with severity baseline cocaine use • Attrition
Grabowski, 1994	USA	7		Subjects previously treated unsuccessfully in a double-blind fluoxetine study	To examine methylphenidate for initial agonist treatment	10 weeks	• PLA (n = 3) • MP (n = 4)	• Retention • Craving • Cocaine use	• No statistically difference in abstinence (cocaine-free urines) • MP:  ‐ Decreased desire to use and preoccupation to use	• 14.2%	• Low doses • Heterogenous patients
Hall, 1994	USA	94	38.1	Male vetera community reinforcement approach ns	To determine the efficacy of enhanced continuity of care and desaprimine in increasing treatment attendance and abstinence	12 weeks	• Desipramine continuity • Desipramine standard • Placebo continuity • Placebo standard	• Cocaine use • Treatment attendance	• No main effects for desipramine	• Days ‐ DC: 70.64 days ‐ DS: 80.14 days ‐ PC: 54.54 days ‐ PS: 65.36 days	• Subjects that failed to comply with instructions to return unused drug
Higgins, 1994	USA	40	31.3	Recruited by advertisements in local newspaper, public service announcements	To assess whether incentivism proved treatment outcome in ambulatory	24 weeks	• Voucher group • No-voucher group	• Abstinence • • Retention • Treatment acceptability	• Incentives delivered contingent on submitting cocaine-free urine specimens significantly improve treatment outcome	• Voucher group: 25% • No-voucher group: 60%	• Methodological • Practical issue cost
Montoya, 1994	USA	62	33.2	Recruited by advertisements and newspaper	To determine the effectiveness of carbamazepine	8 weeks	• Carbamazepine + counseling • Placebo + counseling	• Cocaine use • Craving • Retention	• Significantly decreased self-reported cocaine use • No significant differences between CBZ and placebo	• Carbamazepine: 75% • PLA: 79.4%	• Dropout • CBZ could be effective at higher doses or plasma levels
Cornish, 1995	USA	82	36.1	subjects were recruited from individuals seeking outpatient	To determine the effectiveness of carbamazepine (CBZ)	10 weeks	• Carbamazepine + counseling • Placebo + counseling	• Retention • Cocaine use • Craving	• Groups did not differ for primary outcome measure • CBZ was not an effective treatment	• Carbamazepine: 64% • PLA: 67%	• Dropout
Covi, 1995	USA	72	30	Outpatient treatment and cocaine-only dependence	To determine the effects of the combination of counseling with fluoxetine (20, 40, or 60 mg)	12 weeks	• Fluoxetine 20 mg + counseling • Fluoxetine 40 mg + counseling • Fluoxetine 60 mg + counseling • Placebo + counseling	• Craving • Cocaine use • Relapse prevention	• Fluoxetine group showed the least effectiveness, the group with detectable blood levels had less cravings • Fluoxetine subjects who were abstinent at the start of treatment were somewhat less likely to avoid relapses than those on placebo	• 19.4%	• Small size of the groups • Early dropout • Qualitative urine toxicology
Kranzler, 1995	USA	40	34.3	Subjects were male	To determine the efficacy of carbamazepine 600 mg	12 weeks	• Carbamazepine • Placebo	• Cocaine use • Retention	• Dosage of carbamazepine that was employed in the present study was insufficient to demonstrate a therapeutic effect	• 60%	• Sample size • Larger proportion of subjects smoked cocaine • Dropout
Margolin, 1995	USA	149	37.2	Methadone-maintained subjects	To evaluate the efficacy of bupropion 300 mg	12 weeks	• Bupropion • Placebo	• Cocaine use • Retention	• No significant differences between bupropion and placebo • Exploratory analyses suggested a medication effect for the subset of subjects depressed at study entry	• Bupropion: 15% • PLA: 17%	• Do not consider assessing motivation for behavior change
Nunes, 1995	USA	113	31.7	Subjects randomized, stratified by route of use and depression	To evaluate imipramine as a treatment, and to examine whether its effect may be limited to subgroups defined by route of use or by diagnosis of depression	12 weeks	• Imipramine + counseling • Placebo + counseling	• Treatment compliance • Abstinence • Cocaine use • Craving • Depression	• Imipramine group showed greater reductions in cocaine craving, cocaine euphoria, and depression • Effect of imipramine on cocaine use was less clear • Imipramine effect was greater among nasal users	• 85%	• Dropout • Small size
Batki, 1996	USA	32	34.3	Thirty-one subjects smoked the crack form of cocaine	To teste the efficacy of the selective serotonergic reuptake inhibitor fluoxetine	12 weeks	• Fluoxetine • Placebo	• Cocaine use • Craving • Retention	• Fluoxetine was significantly longer retention	• Fluoxetine: 50% • PLA: 94%	• Dropout
Crosby, 1996	USA	60	33.3–35.8	Subjects without psychiatric comorbidity	To evaluate the effectiveness of phenytoin	12 weeks	• Placebo (n = 24) • 300 mg Phenytoin (n = 20)	• Cocaine use • Global improvement • Craving • Retention	• Phenytoin:  ‐ Decrease 22% in the rate of positive urinalyses  ‐ Reduce 19.4% cocaine use • Greater abstinence	• Phenytoin: 79.3% • PLA: 80.6%	• Dropout 80% • Frequency and pattern of urine collection • Study blind • Sample size
Hoffman, 1996	USA	184	>18	93% were crack smokers	To examine comparative outcomes with more intensive group and standard group	16 months	• SGT • SGT + IP • SGT + IP + FT • IGT • IGT + IP • IGT + IP + FT	• Treatment exposure • Treatment retention • 12 months treatment outcomes	• IGT and SGT: no impact in long term • 23% with cocaine use after 12 months • Regular use: • Female/lees educated/attended fewer treatment sessions	• SGT: 80.9% • SGT + IP: 61.5% • SGT + IP + FT: 53.2% • IGT: 65.7% • IGT + IP: 65.7% • IGT + IP + FT: 61.5%	• Examine other populations using culturally relevant incentives
Kampman, 1996	USA	61	35	Subjects dependent on any drug except cocaine, nicotine, marijuana, and alcohol were excluded	To evaluate the efficacy of Amantadine	4 weeks	• Amantadine + counseling • Placebo + counseling	• Craving • Retention • Depression and anxiety scores	• No significant differences between groups in treatment retention, or in the number of positive urine sample	• Amantadine: 37% • PLA: 39%	• Setting of minimal psychosocial treatment
Schneider, 1996	USA	74	32.8–35.4	Recruited during a 5-day detoxification-stabilization admission.	To explore whether day treatment is a variable alternative to inpatient treatment	3 and 6 months posttreatment	• Day treatment group • Inpatient group	• Attrition • Retention • Abstinence	• Day treatment rates of initiation and completion of treatment	• 3 months: • DTG: 4% • IG:5% • 6 months • DTG:8% • IG:7%	• Small sample • Self-report • Generalizability
Silverman, 1996	USA	52	36	Injecting heroin abusers in a methadone maintenance treatment program	To assess the effectiveness of a voucher-based reinforcement contingency in producing sustained cocaine abstinence	12 weeks	• Reinforcement group • Control group	• Retention • Cocaine use	• Voucher-based reinforcement can produce sustained cocaine abstinence in inner-city injecting poly drug	• Reinforcement group: 11% • Control group: 17%	• Study duration • Sample size
Grabowski, 1997	USA	49	34.3	Frequency of cocaine use ‐ once per week (9%) ‐ once per week (23%) ‐ several times per week (49%) ‐ once per day (9%) ‐ 3 times per day (7%)	To investigate the efficacy of methylphenidate	11 weeks	• MP (25) • Placebo (24)	• Retention • Cocaine use	• The dosing MP was not associated with increased cocaine use • MP were no untoward effects	• MP: 42% • PLA: 48%	• Need to examine medication with profiles that are more favorable to the goal of replacement therapy
Halikas, 1997	USA	150	32.5–34.2	Subjects with any psychotropic medication	To assesses the effectiveness of carbamazepine	12 weeks	• Placebo • Carbamazepine 400 mg • Carbamazepine 800 mg	• Retention • Cocaine use • Craving	• Carbamazepine: ‐ Decrease in the rate of positive urinalyses • Reduce intensity and duration of craving	• PLA: 67.7% • Carbamazepine 400 mg: 60.3% • Carbamazepine 800 mg: 77.8%	• Motivated subjects
Johnson, 1997	USA	65	35.1	Excluded if had other psychiatric disorder other than nicotine dependence	To evaluate Ritanserin efficacy as adjunctive medication to a structured lOP program	10 weeks	• Ritanserin • Placebo	• Craving • Cocaine use • Retention	• Ritanserin produced no significant benefits in cocaine use, craving, or improvement in clinical outcome	• 27% • Ritanserin: 15.6% • PLA: 18.2%	• Study duration
Kirby, 1998a	USA	90	31.7	The value of each voucher was low at the beginning but increased as the patient progressed	To examine whether voucher delivery arrangements affect treatment outcome	12 weeks	• Voucher group • Non-voucher group	• Retention • Abstinence	• No significant differences between the voucher group and the no-voucher group on any of the treatment-related or drag use variables	• Voucher group: 77% • Non-voucher group: 83%	• Focused only on the ability of these voucher incentive programs to initiate cocaine abstinence.
Kirby, 1998b	USA	23	34.7	Voucher Schedule 2 began with high voucher values, but requirements for earning vouchers increased as the patient progressed.	To examine whether voucher delivery arrangements affect treatment outcome	12 weeks	• Voucher schedule 1 • Voucher schedule 2	• Retention • Abstinence	• The number of counseling sessions attended and the number of urine specimens delivered did not differ significantly • Voucher Schedule 2 group provided significantly more weeks of verified continuous cocaine abstinence	• Voucher schedule 1: 75% • Voucher schedule 2: 45.5%	• Focused only on the ability of these voucher-incentive programs to initiate cocaine abstinence
Silverman, 1998	USA	59	37.8	Methadone-maintained subjects	To evaluate the efficacy of escalating contingency interventions	12 weeks	• ES • ES with start-up bonuses (ES + SB) • Noncontingent schedule	• Methadone dose • Retention • Drug use • Craving	• Voucher-based reinforcement: long period: • Sustained cocaine abstinence	• Period Average (weeks) • ES: 11.2 • ES with start-up bonuses (ES + SB): 11.3 • NS: 10.3	• Report data on the administration of the voucher • Craving and lifestyle questionnaires
Crits-Christoph, 1999	USA	487	33.9	Multicenter investigation	To examine the efficacy of 4 psychosocial treatments	6 months	• Individual drug counseling + GDC • Cognitive therapy + GDC • Supportive expressive therapy + GDC • GDC	• Attrition • Drug use	• Supportive expressive + GDC and Cognitive therapy + GDC retained patients better • Individual counseling group + GDC produce superior reductions of overall drug and cocaine use	• Individual drug counseling + GDC: 23.1% • Cognitive therapy + GDC: 21% • Supportive expressive therapy + GDC: 17% • GDC: 5%	• Low psychiatric severity • Lack of a biological outcome measure
Higgins, 2000	USA	70	30.4	Seeking treatment at a university-based outpatient research clinic for treatment	To assess whether contingent incentives can be used to reinforce cocaine abstinence	24 weeks	• Contingency condition + CRA • Non-contingency condition + CRA	• Abstinence • Retention • Follow-up compliance	• Abstinence-contingent • Increased cocaine abstinence during treatment and 1 year of follow-up	• Contingency condition + CRA: 28% • Non-contingency condition + CRA: 21%	• Cost of the vouchers
Jones, 2000	USA	93	28.9	Subjects were pregnant women and methadone-treated	To examine the effectiveness of low-magnitude behavioral incentives	4 weeks	• MM • Abstinence-treated	• Retention (days/weeks) • Compare treatments of MM and AT program women	• Low-magnitude voucher incentives:  ‐ Did not impact treatment attendance for abstinence-based • Enhanced treatment attendance by methadone-treated	• AT: no difference • MM: non-incentive days and weeks of treatment • 6 months: 9% • 9 months: 11% • 12 months: 13%	• Lower valued • Shorter duration
Malcolm, 2000	USA	466	18–52	Current alcohol dependence in addition to cocaine dependence were accepted, as long as they did not require medications for alcohol detoxification.	To investigate the use of Pergolide	12 weeks	• Pergolide low dose + PSYC • Pergolide high dose + PSYC • PLA + PSYC	• Craving • Cocaine use • Retention	• Pergolide in the currently studied doses does not appear to have clinical value • Pergolide had no significant effect on self-reported alcohol use	• Pergolide low dose + PSYC: 71.6% • Pergolide highdose + PSYC: 71.2% • PLA + PSYC: 58.2%	• Subjects were seen only once weekly • Psychological measures were limited to anxiety and depression
Cornish, 2001	USA	80	37.7	Seeking outpatient treatment	To evaluate the efficacy of Ritanserin, a 5-HT2 antagonist, as an adjunct therapy	4 weeks	• Ritanserin • Placebo	• Retention • Cocaine use • Craving	• Ritanserin had no effect on the clinical outcome • No differences in treatment retention in both groups	• Ritanserin: 25% • PLA: 22%	• Dose of Ritanserin • Medication effect
Grabowski, 2001	USA	128	36	Subjects without other psychiatric diagnoses	To examine an amphetamine analog as a prototype agonist	12 weeks	• Dextroamphetamine (15–30) • Dextroamphetamine (30–60) • Placebo	• Retention and compliance • Cocaine use • Compliance • Dropouts • Cocaine use	• The optimal dose is likely between 30 and 60 mg for this preparation	• Dextroamphetamine (15–30 mg): 59.6% • Dextroamphetamine (30–60 mg): 93.3% • PLA: 77.1%	• Subject selection • Number of BZ-positive urine screens during stabilization
Schmitz, 2001a	USA	85	32.9–35.6	Recruited through advertisements in the local media	To examine the joint action of Naltrexone (NTX) in combination with RP therapy	12 weeks	• RP + PLA • RP + NTX • DC + PLA • DC + NTX	• Retention • Cocaine use • Craving • Medication compliance	• RP:  ‐ Did not increase the time until first use or “lapse” • RP + NTX produced less cocaine use	• RP + PLA: 50% • RP + NTX: ‐ 54.2% • DC + PLA: 50% • DC + NTX: 50%	• Select sample
Schmitz, 2001b	USA	68	37.3	Major depressive disorder	To test Fluoxetine efficacy for cocaine dependence and depression	12 weeks	• Fluoxetine + CBT • PLA + CBT	• Retention • Cocaine use • Depression	• Cocaine use and depressive symptoms significantly correlated • Dually diagnosed with severe depression and cocaine use tend to have worse treatment outcomes than those with less severe issues	• Fluoxetine: 47.1% • PLA: 59%	• Do not include appropriate control groups for each treatment
Rawson, 2002	USA	120	42–45.2	Subjects also receiving methadone	To compare the effectiveness of CM and CBT, alone and combination	16 weeks and 17, 26, 52 weeks follow-up	• CM • CBT • CM + CBT • MMTP only	• Retention • Compliance • Cocaine use	• CM: greater during treatment • CBT: comparable long-term outcomes • The order of the treatment might have an impact	• Weeks: ‐ CBT: 23.1% ‐ CBT + CM: 18.9% ‐ CM: 9.4% • MMTP: 24.4%	• Self-report at some phases of the study • Two interventions in parallel
Schubiner, 2002	USA	48	18–55	Subjects cocaine-dependent ADHD	To determine effectiveness of methylphenidate (MTP) control ADHD symptoms and cocaine use	13 weeks	• PLA + CBT • MTP + CBT	• Retention • Safety • ADHD symptoms outcomes • Substance use	• MTP:  ‐ Did not worsen cocaine use • Improved subjective reports of ADHD symptoms	• PLA + CBT: 42% • MTP + CBT: 55%	• Not designed for depression comorbidity • Sample size
Campbell, 2003	USA	146	31.2	Outpatients admitted to the mental health treatment center	To compare the ejects of desipramine or carbamazepine to placebo in an intensive outpatient program	8 weeks	• Desipramine • Carbamazepine • Placebo	• Cocaine use • Craving • Retention	• Carbamazepine did not retain subjects significantly longer than PLA • Males and females may have different response to desipramine	• Desipramine: 61% • Carbamazepine: 71% • PLA: 69%	• Small size to avail gender of cocaine-dependent patient
Higgins, 2003	USA	100	33.8–33.9	Predominantly white subjects	To examine the contributions CRA + voucher	24 weeks	• Voucher group • Voucher group + CRA	• Retention • Compliance • Cocaine use	• CRA + voucher: therapeutic effects  ‐ Substance abuse • Psychosocial function during treatment and follow-up	• Voucher group: 49% • Voucher group + CRA: 16%	• Does not whether CRA is superior to other therapies • Almost exclusively white population
Marlowe, 2003	USA	79	>18	Subjects urban and poor	To evaluate the efficacy of administering voucher incentive for full-day treatment	3 months	• Outpatients + voucher incentive • Day treatment + voucher incentive	• Abstinence • Treatment retention	• Psychoeducational day treatment: not the most efficient way to initiate cocaine abstinence	• Intensive outpatient group: 10% • Control group: 28%	• Psychoeducational and recreational groups • Initiation cocaine abstinence over the first 3 months
Carroll, 2004	USA	121	34.6	Seeking treatment, nonprofit outpatients, newspaper and advertisements	To compare the effectiveness of disulfiram therapy, placebo, CBT, and interpersonal psychotherapy (IPT)	12 weeks	• Disulfiram + CBT • Disulfiran + IPT • PLA + CBT • PLA + IPT	• Retention • Cocaine use • Medication compliance	• Disulfiram was more effective in reducing the frequency of cocaine use and the frequency of cocaine positive urine samples over time • CBT was more effective than IPT in reducing cocaine use during treatment	• 57%	• Dropout
Grabowski, 2004—Study I	USA	94	36.7	59.6% were unemployed	To compare sustained-release d amphetamine and placebo (15/30 or 30/60 mg)	6 months	• PLA + MTD • d-Amphet (15/30 mg) + MTD • d-Amphet (30/60 mg) + MTD	• Retention • Compliance • Dropouts and side effects • Cocaine use	• Cocaine use for the 30/60 mg than 15/30 mg and Placebo	• PLA: 75% • d-Amphet (15/30 mg): 50% • d-Amphet (30/60 mg): 60.7%	• Study with low dose of d-Amphetamine
Grabowski, 2004—Study II	USA	96	36.7	40.6% female	To examine Risperidone and placebo	6 months	• PLA + MTD • Risperidone (2 mg) + MTD • Risperidone (4 mg) + MTD	• Retention • Compliance • Dropouts and side effects • Cocaine use	• Cocaine use did not change in the Risperidone or placebo group	• PLA: 65% • Risperidone (2 mg): 50% • Risperidone (4 mg): 36.4%	• Dose could not be adjusted directly based on drug screen results
Jones, 2004	USA	199	18–55	Subjects were enrolled between August 1998 and September 2000	To evaluate safety and efficacy of L-tryptophan and behavioral incentives	16 weeks	• Tryptophan • Tryptophan + contingent vouchers • Placebo • Placebo + contingent vouchers	• Treatment retention • Cocaine use • Side effects	• No evidence found for the clinical usefulness of tryptophan • Did find efficacy of voucher incentives	• Tryptophan: 75% • Tryptophan + contingent vouchers: 68.9% • PLA:72.5% • Placebo + contingent vouchers: 70.7%	• Different severity of cocaine addiction among patients
Petry, 2004	USA	120	33.6–35.7	Subjects from community-based treatment centers	To evaluate the efficacy of low-cost prize reinforcement CM	12 weeks	• ST • ST + CM ($80) • ST + CM ($240)	• Cocaine use • Abstinence • Retention	• CM is efficacious in reducing cocaine use • ST + CM ($240) more abstinence than ST • CM had most responsive positive urinalysis	• ST: 86.5% • ST + CM ($80): 80% • ST + CM ($240): 68.4%	• Low adherence at the end
Schmitz, 2004	USA	80	35.9	Subjects with cocaine dependence and alcohol dependence	To evaluate benefit from naltrexone in conjunction with psychotherapy RP and DC	12 weeks	• NTX + RP • NTX + DC • PLA + RP • PLA + DC	• Cocaine use • Retention • Medication compliance • Adverts effects	• Lack of efficacy of NTX for patients • NTX effectively reduced craving and alcohol consumption • RP reducing cocaine use initially	• 67%	• Medication doses
Berger, 2005	USA	60	36.3–41.1	Subjects from clinic referrals, flyers sand television, radio, and newspaper	To evaluate the safety and efficacy of reserpine, gabapentin, or Lamotrigine	10 weeks	• Reserpine • Gabapentin • Lamotrigine • Placebo	• Cocaine use • Retention • Compliance • Dose tolerability • Safety • Depression ratings • Efficacy	• Reserpine indicated significant improvement for all study groups	• 17%	• Sample size • Unmatched placebo
Ciraulo, 2005	USA	69	40	Subjects with depressive symptoms	To determine the efficacy of nefazodone	8 weeks	• Placebo • Nefazodone	• Cocaine use • Psychiatric symptoms • Craving • Retention • Compliance	• Nefazodone can reduce cocaine craving	• PLA: 31.5% • Nefazodone: 20.6%	• Sample size
Ciraulo, 2005 CREST 1	USA	65	40.3	Volunteers who received monetary compensation for participating	To assess the efficacy of paroxetine, pentoxifylline, riluzole, and placebo	8 weeks	• Paroxetine + CBC • Pentoxyphylline + CBC • Riluzole + CBC • PLA + CBC	• Cocaine use • Retention • Craving	• Paroxetine, Riluzole, Pentoxifylline were not associated with significant changes from baseline urine BE compared with placebo • Only the pentoxifylline group showed even a slight signal for efficacy	• Paroxetine + CBC: 15.4% • Pentoxyphylline + CBC: 6.8% • Riluzole + CBC:13.3% • PLA + CBC: 14.3%	• Small size • Lack of homogeneity of experimental groups • Study duration
Ciraulo, 2005 CREST 2	USA	60	43	Volunteers who received monetary compensation for participating	To assess the efficacy of pramipexole, Vnelafaxine, and placebo	8 weeks	• Pramipexole + CBC • Venlafaxine + CBC • PLA + CBC	• Cocaine use • Retention • Craving	• Venlafaxine and Pramipexole were not associated with significant changes from baseline urine BE compared with placebo	• Pramipexole + CBC: 25% • Venlafaxine + CBC: 50% • PLA + CBC: 35%	• Small size • Lack of homogeneity of experimental groups • Study duration
Dackis, 2005	USA	62	44.5	Subjects used ate least $200in the past 30 days	To evaluate the effectiveness of Modafinil (400 mg)	8 weeks	• PLA + CBT • Modafinil (400 mg) + CBT	• Treatment retention • Abstinence • Craving • Cocaine withdrawal • Tolerability	• Modafinil:  ‐ Significantly greater proportion of BE-negative urine samples  ‐ Important clinical goal of protracted cocaine abstinence • No evidence for reduced craving	• Modafinil (400 mg) + CBT: 36.7% • PLA + CBT: 34.4%	• Sample size • Subjective nature of craving and withdrawal • Method for medication adherence
Lambert Passos, 2005	BRA	210	18–65	Inhaled cocaine users	To assess the efficacy of oral nefazodone (N) and other treatment modalities (OTM) as psychotherapy, occupational therapy, and family therapy	10 weeks	• Nefazodone + OTM • PLA + OTM	• Abstinence • Craving • Retention	• Abstinence from cocaine for 3 weeks or more was achieved by 49.5% (N) and 45.7% (PLA) • The results do not support the indication of nefazodone	• Nefazodone + OTM: 77.1% • PLA + OTM: 81%	• Self-reporting • Do not use urine screens
Malcom, 2005	USA	195	21–50	Subjects of all ethnicities	To test the efficacy of amlodipine	12 weeks	• PLA + CBT • Amlodipine + CBT	• Retention • Cocaine use	• Amount of therapy: not significant • Amlodipine did not:  ‐ Confer any advantage • Alter cocaine use	• 2-week baseline: 41%  ‐ Amlodipine: 65% • PLA: 71.4%	• Randomization upon the completion of baseline Placebo phase • No comorbid • Disorders
McDowel, 2005	USA	111	18–65	Subjects with comorbid depression disorder	To test the efficacy of desipramine in treatment with current depression disorders	12 weeks	• Placebo • Desipramine	• Attrition • Adverse effects • Depression treatment • Cocaine use	• Desipramine was associated with higher rate of depression response than placebo • Groups did not differ in rate of cocaine response	• Desipramine: 47.6% • PLA: 41.9%	• Sample size • Self-report • Noncompliance • Side effects
Petry, 2005a	USA	415	35.8	Outpatient psychosocial treatment programs	To evaluate the efficacy of an abstinence-based CM intervention as an addition to usual care in community treatment settings	12 weeks	• Incentive condition • Usual care condition	• Drug use • Counseling use • Retention	• Incentive condition:  ‐ Improved drug use out comes related to treatment retention  ‐ Higher overall number of negative samples • Longer duration of documented continuous abstinence compared	• Incentive condition: 51% • Usual care condition: 65%	• Relative high rates of missing data • Single set of incentive parameters
Petry, 2005b	USA	77	39.5	Methadone-maintained subjects	To evaluate the efficacy of prize CM reinforcing urinalysis results and group therapy attendance	12 weeks	• ST • CM treatment	• Cocaine abstinence	• Prize CM procedure decreased cocaine use and enhanced therapy attendance	• 14.2% • ST: 16.2% • CM treatment: 12.5%	• Cost-effectiveness of the approach
Reid, 2005	USA	68	38.72	CREST inclusion and exclusion criteria	To screen the efficacy of olanzapine, valproate, or coenzyme Q10/L-carnitine combination versus placebo	8 weeks	• Olanzapine • Valproate • Coenzyme Q10 and L-cartinine combination • Placebo	• Abstinence • Global impression scores • Craving • Retention • Related psychosocial measures • Adverse events	• Study retention was similar across all treatment groups • None of the study medications were superior to placebo on any of the outcome measures	• Approximately 40% to 42%	• Small sample size • Retention rate • Medication dose
Winhusen, 2005	USA	67	40	CREST inclusion and exclusion criteria	To evaluate the safety and efficacy of tiagabine, sertraline, or donepezil versus an unmatched placebo control	10 weeks	• Sertraline + CBT • Donepezil + CBT • Tiagabine + CBT • PLA + CBT	• Urine BE level • Self-report of cocaine use • Global impression score • Adverse events	• Subjective improvement in cocaine dependence for all groups • Tiagabine group showed a trend in decreased urine BE levels • No treatment-related physical or laboratory abnormalities identified	• 17.9% • Sertraline + CBT: 6% • Donepezil + CBT: 24% • Tiagabine + CBT: 12% • PLA + CBT: 29%	• Small sample size
Bisaga, 2006	USA	99	39	Cocaine smokers and intranasal users	To evaluate the efficacy and safety of gabapentin with relapse prevention therapy	12 weeks	• Placebo • Gabapentin	• Cocaine use • Craving • Retention • Abstinence • Adverse events • Medication compliance	• Gabapentin: • When combined with IRPT was no more effective than Placebo	• 51% • Smokers:52% • Intranasal users: 35%	• Self-reported
Elkashef, 2006	USA	300	40.7	Subjects recruited at sites Cocaine dependence	To assess the efficacy and safety of selegiline in treating cocaine dependence	8 weeks	• STS patches + PSYC • PLA + PSYC	• Cocaine use • Abstinence • Craving	• Analysis did not show a significant effect for Selegiline over placebo	• STS + PSYC: 34.9% • PLA + PSYC: 26.7%	• Selegiline transdermal formulation • Group differences at baseline
Gonzalez, 2006	USA	123	35.9	The majority completed high school (82%)	To examine effects of homework in CBT compliance for relapse prevention therapy	12 weeks	• ST • Homework group	• Motivation • Cocaine use	• Homework: • Early in therapy better retention Predicted less cocaine use during treatment for participants higher in readiness to change	• 33.3% approximately	• Measure of homework • Homework just during treatment • Sample heterogeneity
Peirce, 2006	USA	388	42	Methadone-maintained subjects	To compare outcomes of lower cost prize-based CM treatment is added to usual care in community methadone hydrochloride maintenance treatment	12 weeks	• Usual care with abstinence incentives • Usual care without abstinence incentives	• Abstinence • Retention • Counseling attendance	• An abstinence incentive approach that paid $120 in prizes per participant effectively increased stimulant abstinence in community-based MM treatment clinics	• Incentive participants: 32.9% • UC : 35.2%	• Voluntary nature of study participation • Missing data
Petry, 2006	USA	131	36.5	Outpatients who initiating treatment between 2000 and 2003	To evaluate the efficacy of prize-based CM	12 weeks	• ST • ST with CM + abstinence condition • ST with CM + goal-activities condition	• Retention • Abstinence • Follow-up effects	• Abstinence at 6 to 9 months follow-up did not differ by group • Longest duration of abstinence achieves was associated with abstinence posttreatment	• ST: 22,5%• ST + CM + abstinence: 27.27% • ST + CM + activities: 29.7%	• Heterogenous drug-abusing population • Infrequently collected and tested urine samples
Poling, 2006	USA	106	34.6	Methadone-maintained subjects	To compare the efficacy of bupropion hydrochloride and CM for reducing cocaine use in methadone hydrochloride-maintained individuals	25 weeks	• CM + PLA (CMP) • CM + 300 mg/day of Bupropion hydrochloride (CMB) • Voucher control + PLA (VCP) • Rvoucher control + bupro pion (VCB)	• Cocaine and heroin use	• These findings suggest that combining CM with Bupropion for the treatment of cocaine addiction may significantly improve outcomes relative to Bupropion alone	CMP: 40% • CMB: 44% • VCP: 37% • VCB: 44%	• This study did not use methods to track participant compliance
Ghitza, 2007	USA	116	37	Methadone-maintained subjects	To examine the effect of reinforced density in prize-based abstinence reinforcement, heroin/cocaine users in MM	20 weeks	• Manual • Computerized standard • Computerized • High prize density • Noncontingent	• Opioid and cocaine abstinence	• A computer-automated, high-density prize-based reinforcement may effectively reduce cocaine use • Related problems in MM patients who continue using during treatment	• Manual: 30% • Computerized standard: 31% • Computerized • High prize density: 40% • Noncontingent: 27%	• Baseline drug use imbalance • Lack of assessment of participant awareness • Limited generalizability • Nonindividualized methadone dosing • High pre-randomization attrition
Gonzalez, 2007	USA	76	37	Cocaine dependent, methadone treated	To evaluate gabapentin and Tiagabine compared with placebo in reducing cocaine taking behavior	10 weeks	• Tiagabine • Gabapentin • Placebo	• Thrice-weekly drug free urine samples	• Gabapentin showed poor treatment retention and ineffectiveness in reducing cocaine use. Tiagabine significantly reduced cocaine • Taking behavior compared with placebo or gabapentin among methadone-stabilized cocaine abusers	• Tiagabine: 20% • Gabapentin: 35% • PLA: 8%	• Interim analysis and change in randomization • Qualitative urine BE testing • Lack of craving and withdrawal measures • Inconsistent urine toxicology and self-reported opiate use
Levin, 2007	USA	106	37	Subjects with cocaine dependence and persistent adult ADHD	To compare efficacy of sustained-release methylphenidate (MPH) to placebo (PBO) in treating adult with ADHD	14 weeks	• MPH + CBT • PBO + CBT	• ADHD symptom improvement • Retention • Efficacy of MPH for adult ADHD • Abstinence • Proportion of cocaine positive weeks	• Low percentage of prolonged cocaine abstinence • MPH maintenance may lead to reductions in cocaine use ADHD symptom improvement	• 14 weeks: 56% • MPH: 57%PBO: 55%	• PBO response rate based on a self-report • Subjects generally healthy • MPH dose
McKee, 2007	USA	74	34.9	Cocaine abuse or dependence Opiate abuse or dependence were excluded	To investigate the impact of enhancing brief CBT with motivational interviewing techniques for cocaine abuse or dependence	16 weeks	• CBT • MET + CBT	• Treatment retention• Process measures (eg, commitment to abstinence, satisfaction with treatment) • Cocaine use	• MET + CBT attended more:  ‐ Treatment sessions  ‐ Reported higher desire for abstinence  ‐ Expectation of successAnticipated greater difficulty in maintaining abstinence	• CBT: 40% • MET + CBT: 35%	• Small size • Same therapists provided both interventions • Sample homogeneity and high therapist education may limit the generalizability
Petry, 2007	USA	74	41.6	Cocaine-dependent methadone outpatients	To compare voucher and prize CM interventions	9 months	• ST • ST plus prize • ST plus voucher	• Cocaine nonuse	• Reinforcement with prizes was similar to voucher CM in promoting IVR	• ST plus prize: 14.3% • ST plus voucher: 18.6% • ST: 26.3%	• Did not examine psychotherapies, such as CBT
Winhusen, 2007a	USA	119	40	Exclusion for dependence on substances other than cocaine, alcohol, nicotine, or marijuana	To evaluate the safety and efficacy of reserpine treatment for cocaine dependence	12 weeks	• Reserpine • Placebo	• Cocaine nonuse days—self-report confirmed with urine BE level • Craving • Addiction severity index scores • Clinical global impression scores • Safety measures	• These results do not support the efficacy of reserpine as a cocaine dependence treatment	• 34% • Reserpine: 30% • PLA: 37%	• Dose of reserpine evaluated
Winhusen, 2007b	USA	141	40	Cocaine dependent	To evaluate the safety and efficacy of Tiagabine treatment for cocaine dependence	12 weeks	• Tiagabine • Placebo	• Cocaine nonuse • Safety measures	• Tiagabine, at a dose of 20 mg/day, did not have a robust effect in decreasing cocaine use	• 44% • Tiagabine: 46% • PLA: 42%	• Dose of Tiagabine • Sample: primary cocaine dependence
Loebl, 2008	USA	31	43.25	Cocaine dependent	To test the efficacy of a long-acting, injectable preparation of Risperidone on cocaine use in active cocaine users	12 weeks	• Risperidone • Placebo	• Cocaine use • Craving • Depressive symptoms • Adverse events	• Treatment with long-acting injectable Risperidone was not associated with reduction in cocaine use or craving • Associated with worsening of depressive symptoms and weight gain	• Risperidone: 50% • PLA: 60%	• Urine BE level sampling frequency • Baseline differences • High attrition rates
Pettinati, 2008	USA	208	41	Excluded if they met criteria for dependence for any substance other than cocaine and alcohol	To evaluate the efficacy of disulfiram, naltrexone, and their combination in patients with co-occurring cocaine and alcohol dependence	11 weeks	• Disulfiram + CBT • NTX + CBT • Combination + CBT • PLA + CBT	• Abstinence from cocaine and/or alcohol (self-report)	• There was an association between disulfiram treatment and abstinence from cocaine and alcohol • More patients taking the disulfiram-naltrexone combination achieved 3 consecutive weeks of abstinence in treatment than placebo-treated patients	• Disulfiram + CBT: 22.6% • NTX + CBT: 32.7% • Combination + CBT: 40.8% • PLA + CBT: 40.7%	• Patients in the trial may differ from typical clinical patients • High levels nonadherence
Shoptaw, 2008	USA	70	36.9	Exclusion for dependence on alcohol, opiates, benzodiazepine, or other sedative-hypnotics	To investigate bupropion as a treatment for cocaine dependence	16 weeks	• Bupropion + counseling • PLA + counseling	• Cocaine use • Treatment retention • Craving • Depressive symptoms • Adverse events	• No statistically significant effect for bupropion relative to placebo for the treatment of cocaine dependence	• Bupropion + counseling: 81% • PLA + counseling: 85%	• Small sample size • High attrition rate
Brodie, 2009	Mexico	103	29.1	Exclusion for dependence on substances other than cocaine, alcohol, nicotine, or marijuana	To assess the efficacy of Vigabatrin for short-term cocaine abstinence in cocaine-dependent individuals	13 weeks	• Vigabatrin + CBT • PLA + CBT	• Cocaine use (urine screens) Abstinence from marijuana and alcohol • Craving • Adverse events	• This trial supports the safety and efficacy of short-term Vigabatrin treatment of cocaine dependence	• Vigabatrin + CBT: 38% • PLA + CBT: 58.5%	• Urine analysis frequency • Nicotine use was not assessed • Did not ascertain reasons for dropout • Nonacademic setting in a resource-poor nation
Garcia-Rodriguez, 2009	Spain	96	29.5	Exclusion for dependence on opioids	To assess the impact of voucher magnitude on cocaine abstinence and retention, and the effectiveness of CM in a European outpatient setting	24 weeks	• Standard outpatient treatment • CRA + low monetary value vouchers • CRA + high monetary value vouchers	Cocaine abstinence (urine specimens) • Treatment retention	Combining CRA with vouchers was more effective than ST in community outpatient programs in Spain	•CRA + low monetary value vouchers: 46.7 • CRA + high monetary value vouchers: 31% • Standard outpatient treatment: 63.5%	• Effectiveness may vary more between clinics than between groups
Hamilton, 2009	USA	48	45.8	Male veterans Methadone-maintained subjects	To determine whether Olanzapine helped cocaine-dependent patients	(16+4) 20 weeks	• Olanzapine • Placebo	• Cocaine abstinence • Craving • ASIExtrapyramidal symptom scales	• Olanzapine appears ineffective for cocaine dependence	• Subjects who completed at least 8 visits (N = 28) • Subjects who completed fewer than 8 total visits (N = 20)	• Subjects were all male veterans • High Placebo response • Small sample size
Mooney, 2009	USA	82	36.4	Participants were recruited through advertisements in local social sources, and underwent a telephone interview	To evaluate the safety, tolerability and effectiveness of methamphetamine (MTPM)		• PLA + CBT + CM • 30 mg IR MTPM + CBT + CM • 30 SR MTPM + CBT + CM	• Retention • Cocaine use • Medication adherence • Craving	• Adherence for the initial daily dose was high, but decrease for subsequence doses • SR MTPM significantly reduce cocaine use and craving • CM had little discernible effect	• PLA: 70.3%• 30 mg IR MTPM: 70% • 30 SR MTPM: 68%	• High attrition • Disparity gender differences
Schmitz, 2009	USA	87	34.4	Subjects recruited by media advertisements	To compare medication condition (naltrexone, placebo) with behavioral therapy conditions (CBT, CBT + CM)	12 weeks	• NTX + CBT • NTX + CBT + CM • PLA + CBT • PLA + CBT + CM	• Cocaine and alcohol use • Retention and adherence	• Retention and medication compliance suboptimal adherence with the protocol • High-dose naltrexone did not improve cocaine and drinking	• Subjects:76% • Days of treatment: ‐ NTX + CBT:62.5% ‐ NTX + CBT + CM: 64.3% ‐ PLA + CBT: 59.5% • PLA + CBT + CM: 56.6%	• Small size of the treatment groups • Overall low retention in treatment
Schmitz, 2010	USA	136	41	53% were employed	To examine levodopa treatment effects across different CM platforms	12 weeks	• CM-URINE • CM-ATTEND • CM-MEDICATION	• Retention and attendance • Medication compliance • Cocaine use	• CM-ATTEND and CM-MEDICATION producing corresponding changes in targeted behaviors • The CM-URINE was moderated by medication	• 6 weeks: 49% • 12 weeks: 65%	• Sample size • High attrition • Rates of responding were less than robust
García-Fernández, 2011	SPA	58	>20	Adults seeking outpatient treatment as a community clinic in Spain	To compare the efficacy of the CSSA with and without an incentive program	24 weeks	CRA – 29 subjects • CRA + vouchers	Cocaine abstinence • Treatment retention • Phychosocial functioning	Voucher group have greater improvement in psychosocial functioning • Voucher incentives improve enrolled on treatment programs	CRA: 41.4% • CRA + vouchers: 34.5%	Difference between groups in urine specimens • Not include who failed to complete the trial
Schottenfeld, 2011	USA	145	31.1	Pregnant women or women with young children Excluded if they were using opioids	To evaluate the efficacy of CM, the CRA, and 12-step facilitation (TSF) for cocaine-dependent pregnant women or women with young children	24 weeks	• CRA + CM TSF + CM • CRA + VC • TSF + VC	• Abstinence • Retention	• CM for cocaine-dependent pregnant women and women with young children but do not support greater efficacy of CRA compared with TSF or differential efficacy of CM when paired with either CRA or TSF	• 12-week CRA + CM: 47.2%• TSF + CM: 45.9% • CRA + VC: 54.2% TSF + VC: 56.7%	• Missing data • Lower retention • Underreporting of cocaine use • Did not compare with no treatment or minimal counseling
Winstanley, 2011	USA	145	39	Cocaine and opioid dependent and treated with methadone	To assess whether fluoxetine could be an effective pharmacotherapy for the treatment of cocaine dependence	33 weeks	• Fluoxetine + voucher incentives • PLA + voucher incentives • Fluoxetine without vouchers • Placebo without vouchers	• Treatment retention • Cocaine use (thrice-weekly urine testing)	• Fluoxetine was not efficacious in reducing cocaine use in patients dually dependent on cocaine and opioids	• FV: 78% • PV: 76.4% • F: 82% • P: 77.6%	• Small size • Baseline differences • Early termination due to job-related reasons • High dropout rate • Inclusion of MM therapy
Dackis, 2012	USA	210	44.5	Used at least $200 worth of cocaine in the past 30 days	To test the efficacy of Modafinil	8 weeks	• Modafinil 200 mg + CBT • Modafinil 400 mg + CBT • P + CBT	• Cocaine abstinence • Craving • Retention • Tolerability	No significant differences between Modafinil and placebo • Significant gender difference in the male patients with 440 mg/day abstinence than placebo counterparts	• Modafinil 200 mg + CBT: 38.5% • Modafinil 400 mg + CBT: 38.6% • PLA + CBT: 50.7%	• Gender differences • Inadequate psychosocial treatment intensity
McKay, 2012	USA	152	42.3–43.4	No regular intravenous heroin use within the past 12 months	To teste whether ECC improved outcomes entering an IOP	12 months	• IOP (TAU) • IOP + ECC	• Abstinence • Frequency of abstinent days • Cocaine use days • Heavy drinking days	• Continuing care that is not well integrated with the primary treatment may interfere in some way with the therapeutic process	• IOP + (TAU): 21.8% • IOP + ECC: 29.7%	• Counselors who were not located in the IOP • Did not have data on the characteristic of dropout
Petry, 2012	USA	515	37.2	Cocaine negative (333) Cocaine positive (109)	To compare reinforcement magnitudes in cocaine-positive patients and evaluate attendance-based CM in cocaine-abstinent patients at treatment initiation	9 months	• Cocaine positive • Standasrt care (SC) • SC + $240Abs • SC + $560Abs • Cocaine negative • SC • SC + $240Abs • SC + $560Abs	• Retention • Number of sessions attended • Samples submitted • Abstinence LDA	• High magnitude abstinence-based reinforcement improved all abstinence outcomes in patients who began treatment while using cocaine • For patients initiating treatment abstinent, both attendance- and abstinence-based CM resulted in improvements on some measures	• Cocaine positive • SC: 8.8% • SC + $250Abs: 16.6% • SC + $560Abs: 12.5% • Cocaine negative • SC: 15.7% • SC + $250Abs: 12.7% • SC + $250Att: 16.8%	Missing data
Schmitz, 2012	USA	73	42	85% male	To test low doses combination of Modafinil and d-Amphetamine	16 weeks	• d-Amphetamine • Modafinil • Modafinil + d-Amphetamine • Placebo	• Retention • Cocaine use • Compliance • Adverse events	• No evidence that the dual-agonist combination had benefit • Relatively better cocaine outcomes in Placebo and d-Amphetamine groups	• Did not differ between groups: • week 12: 60% • week 16: 80%	• High attrition especially in week 12 • Preliminary study
Hagedorn, 2013	USA	332	50	Alcohol dependence or stimulant dependence Veterans	To compare the effectiveness trial of a voucher drawing CM intervention with usual care	12 months	• CM intervention • Alcohol • Stimulant • Usual care • Alcohol • Stimulant	• Total number of urine and breath samples that were negative • Retention • LDA • Self-report drug and • alcohol use	• Intervention effects were nonsignificant for the stimulant-dependent subgroup • The study provides support for the effectiveness of CM interventions for alcohol-dependent patients	• CM intervention: 25.5% • Alcohol: 21.3% • Stimulant: 31% • UC: 26.7% • Alcohol: 27.8% • Stimulant: 27.9%	• Alcohol use detection limited by breathalyzer sensitivity • Stimulant-dependent participants more likely to be African American and homeless • Lack of blinding
Johnson, 2013	USA	142	43.7	Cocaine dependent	To determine the efficacy of Topiramate versus placebo	12 weeks	• Topiramate • Placebo	• Abstinence • Craving • Clinical global impression scales	• Topiramate is more efficacious than placebo at increasing the mean weekly proportion of cocaine nonuse days and associated measures of clinical improvement among cocaine-dependent individuals	• 49.2% • Topiramate: 46.4% • PLA: 52.1%	• High attrition rate • Prolonged assessment needed
Kampman, 2013	USA	170	43–45	Subjects addicted in cocaine and alcohol	To test efficacy of Topiramate	13 weeks	• PLA + CBT • Topiramate + CBT	• Cocaine use • Cocaine and alcohol craving • Retention • Medication adherence	• Topiramate: ‐ Was not better than placebo primary ‐ Is better for who entered with more severe cocaine withdrawal symptoms • Retained more	• Topiramate: 35% • PLA: 47%	• Self-reported • Sample size • Abstinence before beginning the trial
Secades-Villa, 2013	SPA	118	31.07–31.44	Different socioeconomic status	To examine the impact of socioeconomic variables on the effectiveness of CM in a community setting	24 weeks	• CRA • CRA + vouchers	• Retention • Abstinence	• MAST scores and ASI scores were predictors of retention and abstinence variables • Income had no effect on retention and abstinence outcomes after 6 months of treatment	• CRA: 40.6% (weeks) • CRA + vouchers: 24.6% (weeks)	• Self-reported income • Combined sample from 2 treatment settings • Income was primary variable of interest
Somoza, 2013	USA	186	44.8	Cocaine dependent	To evaluate the safety and efficacy of Vigabatrin for the treatment of cocaine dependence in a US sample	24 weeks	• Vigabatrin • Placebo	• Abstinence • Cocaine use	• No protocol-defined differences in efficacy between Vigabatrin treatment and placebo were detected for any outcome variable	• 51.6% • Vigabatrin: 53.2% • PLA: 50%	• Medication adherence method: self-reported
Festinger, 2014	USA	222	37.1	Cocaine-dependent methadone-maintained subjects	To compare the efficacy of cash-based reinforcement therapy (CBRT) to voucher-based reinforcement therapy (VBRT) and a non-CM condition on cocaine abstinence and treatment attendance; examined whether CBRT resulted in greater levels of harm than VBRT	12 weeks	• VBRT • CBRT • Non-CM control condition	• Abstinence (urine analysis) • Attendance • Craving • High-risk behavior	• CBRT was as effective as VBRT when compared to the non-CM condition and that it did not increase rates of drug use, cravings, or high-risk behaviors	• Number of sessions • VBRT: 28.28 sessions • CBRT: 28.49 sessions • Non-CM control condition: 24.82 sessions	• Lack of treatment context • No attendance records
Mancino, 2014	USA	99	39.5	Depression symptom	To determine whether treatment with SERT or SERT + GC-MS Ham D improve treatment	12 weeks	• SERT • SERT + GBP • PLA	• Abstinence • Cocaine use • Retention • Adverse events • Compliance • Cocaine use • Depressive symptoms	• Relapse: 88.9% PLA compared with SERT (65.2%) • Retention: did not differ significantly between the groups • SERT + GBP may not be superior to SERT	• SERT: 42.9% • SERT + GBP: 44% • PLA: 55.3%	• Drop out
Petitjean, 2014	Switzerland	60	34.5	Methadone-maintained subjects	To investigate the acceptability and efficacy of CBT plus prize CM in the European context	24 weeks 6 months	• CBT + prize CM • Control group (CBT)	• Retention • Abstinence	• Adding prize CM to CBT enhances treatment effects, particularly early on, with both interventions significantly reducing cocaine use during treatment and sustaining effects at the 6-month follow-up	• 36.7% (abstract) • CBT + prize CM: 13.7% • Control group (CBT): 22.5%)	• Small sample size
Schmitz, 2014	USA	81	42.1	Excluded if they met criteria for dependence on substances other than cocaine, nicotine, or cannabis	To evaluate a screening paradigm for multiple medications versus placebo on cocaine use reduction and abstinence maintenance, and to assess baseline abstinence as a predictor and moderator of treatment outcomes	Phase 1 4 weeks Phase 2 12 weeks	• Modafinil + CBT • Levodopa + CBT • NTX + CBT • PLA + CBT	• Cocaine use	• Good retention (80%) and reasonable abstinence initiation (40%) during Phase I • Stratifying by Phase I abstinence status revealed that levodopa-carbidopa and Naltrexone benefited non-abstinent patients but showed no benefit for those who were abstinent at baseline	• 20% (Phase I) • Modafinil: 13.6% • Levodopa: 52% • NTX: 31.25% • PLA: 38.8%	• Small sample size
McKay, 2014	USA	321	43	Subjects who finished a period of IOP	To determine which patients engaged in an IOP were most likely to benefit from extended continuing care	24 months	• TAU • TAU + TMC • TAU + TMC + voucher	• Cocaine urine toxicology • Substance use	• TMC + better for participants:  ‐ In a controlled environment prior to IOP  ‐ With no days of depression early in IOP • With high family/social problem severity or high self-efficacy	• TMC = 27.4% • TMC + voucher = 16.8%	• Is not clear whether patients who were receiving standard rather than intensive outpatient treatment would have de the same results
Nuijten, 2014	Holand	74	41.3–43.3	Crack-cocaine-dependent outpatients	To evaluate the acceptance and effectiveness of Topiramate as an add-on to CBT	12 weeks	• Topiramate 200 mg/day + CBT • CBT	• Retention • Medication adherence • Safety • Cocaine and other substance use • Health • Social functioning • Patient satisfaction	• Topiramate’s accepting was generally low • Positive effects only in the subgroup with comorbid opioid dependence	• Topiramate + CBT: 70.2% • CBT: 74%	• Open-label character of the trial • Frequency of Urine samples
Schmitz, 2014	USA	118	41.97–42.90	Treatment-seeking	To the evaluation of a paradigm for screening multiple medications	4 weeks + 12 weeks	• Modafinil (400 mg/day) + CBT + CM • Levodopa/Carbidopa (800/200 mg/day) + CBT + CM • NTX (50 mg/day) + CBT + CM • PLA + CBT + CM	• Cocaine use • Medication compliance and adverse events • Retention	• Retention (80%) and reasonable rates of abstinence initiation (40%) during the baseline • Levodopa-carbidopa and naltrexone:  ‐ benefit who were non-abstinent at baseline • not benefit who achieved abstinence at baseline	• Before Phase I: 20% • Before Phase II: 5.9% • Phase II: 50%	• Low refinements to ensure more stable baseline responding • Sample size
Umbricht, 2014	USA	171	18–55	Cocaine-dependent MM patients	To evaluate whether Topiramate (TOP) would increase cocaine abstinence in methadone patients	18 weeks	• TOP + CM • TOP • PLA + MC • Placebo	• Retention • Cocaine abstinence • Cocaine craving • Other drugs and alcohol use • Medication Adherence • Adverse events	• TOP: • is not efficacious for increasing cocaine abstinence for methadone patients.	• Evaluation phase  ‐ TOP + MC: 32%  ‐ TOP: 42%  ‐ PLA + MC: 31% • PLA: 30%	• The CM groups in this study did not perform better than the non-CM groups
Kampman, 2015	USA	94	47	No alcohol dependence comorbid	To test the efficacy of Modafinil	8 weeks	• Modafinil 300 mg + IT • PLA + IT	• Cocaine use • Craving • Retention	• Modafinil group reported: ‐ be abstinent from cocaine during the last 3 weeks • very low levels of cocaine-craving intensity and duration	• Not significantly difference between the groups: 24% (general) • Modafinil group: 12 visits • Placebo group: 13 visits	• Small single site trial • Did not have comorbid alcohol dependence • Measure medication adherence
Levin, 2015	USA	126	38.7–43.90	Recruited by local advertising for treatment research or clinical referrals	To examine whether treatment of co-occurring ADHD and cocaine use disorder with extended-release mixed amphetamine salts	14 weeks	• Medication 1 (60 mg) + CBT • Medication 2 (80 mg) + CBT • PLA + CBT	• Treatment adherence • ADHD • Cocaine use • Averse effects	• Extended-release mixes amphetamine salts: ‐ reduction in cocaine use • benefit patient with ADHD and cocaine use combined CBT	• Medication 1: 25% • Medication 2: 20.9% • PLA: 32.6%	• Should test long-acting
McKay, 2015	USA	409	48.3	85% not currently married	To evaluate the effect of providing choice of treatment alternatives to patients who fail to engage in or drop out of IOP	8 weeks	• MI-IOP • MI-PC	• Motivation interviewing adherence • Treatment participation • Measure of alcohol and cocaine use at 1, 2, 3 and5 6 m	• MI-PC produced better attendance than conditions in patients than dropped out after initial engagement and in those re-randomized at the 8th week	• After 8 weeks: 56%	• Sample • Treatment was not offered at intake, but after first disengagement • Did not offer combinations of treatments
Nuijten, 2015	Holand	65	45.3–47.1	Chronic cocaine-dependent with a history of multiple treatments	To evaluate the acceptance, safety, and potential benefits of 400 mg/day Modafinil as an add on to CBT compared with CBT only	12 weeks	• CBT + 400 mg Modafinil • CBT	• Retention in CBT treatment • Crack-cocaine use	• 400 mg/day Modafinil for safety and benefits were not demonstrated • Medication adherence low, partly attributable to adverse events. • Mofafinil was not related to improvement in:  ‐ CBT retention  ‐ cocaine use • second outcomes	• CBT + Modafinil: 70% • CBT: 77.1%	• Chose an open-label • Non-placebo-controlled design
Petry, 2015	USA	240	40.3	Cocaine-dependent methadone patients	To evaluate whether increasing magnitudes of reinforcement will improve outcomes	12 weeks	• UC • UC + $300 standart prize CM • UC + $900 high magnitude prize CM • UC + $900 voucher	• Abstinence	• Providing the standard magnitude of $300 in prizes was as effective as larger magnitude CM in cocaine-dependent methadone patients	• UC: 1.7% • $300 Prize CM: 13.7% • $900 Prize CM: 1.6% • $900 Voucher: 7.9%	• Limited voucher options • Restricted prize choices • Limited generalizability
Tapp, 2015	USA	60	47.9	Cocaine dependence	To compare cocaine use and cravings between the Quetiapine and placebo groups	12 weeks	• Quetiapine • Placebo	• Cocaine use • Money spent on cocaine	• Did not find group differences between the Quetiapine and placebo arms, suggesting that Quetiapine is not an efficacious treatment for DSM-IV cocaine dependence	• 68% • Quetiapine: 65.5% • PLA: 74.1%	• High number of drop out • Gender discrepancy • Chronic cocaine users
Carroll, 2016a	USA	99	39.3	Excluded if they met criteria for dependence on another drug (except tobacco) or whose principal drug use was not cocaine	To evaluate the extent to which outcomes from CBT treatment of cocaine dependence could be enhanced by adding CM, disulfiram (Dis), and the combination of CM and Dis via a 2 × 2 factorial design	12 weeks 1 year	• Dis/CM • Dis/no CM • Placebo/no CM • Placebo/CM	• Cocaine use • Abstinence	• CM enhances outcomes for CBT treatment of cocaine dependence • Disulfiram provided no added benefit to the combination of CM and CBT	• 12 weeks • Dis/CM: 21.7% • Dis/no CM: 10.7% • Placebo/no CM: 11.5% • Placebo/CM: 13.6% • 12-mont follow-up • Dis/CM: 26% • Dis/no CM: 28.5% • Placebo/no CM: 30% • Placebo/CM: 9.0%	• High attrition rate • Small sample size
Nuijten, 2016	Netherlands	73	49	Patients who received oral methadone plus inhalable or injectable diacetylmorphine	To assess the acceptance, efficacy, and safety of a sustained-release dexamphetamine in chronic crack-cocaine-dependent with comorbid heroin dependence	12 weeks	• SRDG • Placebo	• Cocaine use • Cocaine negative urine samples • Craving	SRDG:  ‐ is well accept, safe, effective • superior to placebo in cocaine-related outcome and urine-based	• SRDG: 10.5% • PLA: 11.4%	• Sample size • Duration of the study • Overestimate the results
Carroll, 2018	USA	120	38.2	Methadone-maintained subjects	To examine the effectiveness of Galantamine, alone or with computerized CBT, in improving cocaine use outcomes and cognitive functioning	20 weeks	• Galantamine + CBT • Galantamine + TAU • PLA + CBT • PLA + TAU	• Cocaine use • Abstinence • Cognitive function	• Findings suggest benefits of Galantamine and computerized cognitive behavioral therapy for cocaine use in this sample	• Galantamine + CBT: 28.5% • Galantamine + TAU: 25.9% • PLA + CBT: 28.9% • PLA + TAU: 7.4%	• Long-term methadone use limits generalizability and may affect Galantamine’s cognitive impact
Levin, 2020	USA	127	46.2	48.8% current alcohol use disorder	To test the combination of mixed amphetamine salt extended-release (MAS-ER) and Topiramate or placebo	12 weeks	• MAS-ER and Topiramate • Placebo	• Abstinence • 3 consecutive weeks of abstinence • Continuous cocaine abstinence • Use • Craving • Retention	• MAS-ER:  ‐ achieving 3 abstinent weeks significantly larger (14.1%)  ‐ is more effective than either treatment alone	• MAS-ER and Topiramate: 34% • PLA: 41%	• Sample size • Study was concluded before full enrollment could be achieved • Dropout

Abbreviations: Des, desipramine; MM, methadone maintenance; ES, escalating schedule; ADHD, adult attention-deficit/hyperactivity disorder; ASI, Addiction Severity Index; GBP, Gabapentine; CBT, cognitive behavioral therapy; CM, contingency management; CM-ATTEND, clinic attendance; CM-MEDICATION, medication compliance; CM-URINE, cocaine negative urine toxicology; CRA, community reinforcement approach; CSSA, Cocaine Selective Severity Assessment; ECC, enhanced continuing care; FT, family therapy; GC-MS, chromatography-mass spectrometry; Ham D, Hamilton depression scale; IGT, intensive group treatment; IOPs, intensive outpatient programs; IP, individual psychotherapy; IT, individual therapy; IRPT, individual relapse prevention therapy; LDA, longest duration of continuous abstinence; IVR, interactive voice response; MAST, Michigan Alcoholism Screening Test; MI, motivational interview; MI-PC, motivational interviewing provided a choice of intensive outpatient care; MMTP, methadone maintenance treatment program; MM, medication management; MTD, methadone; NTX, naltrexone; PLA, placebo; RP, relapse prevention; SERT, sertraline; SGT, standard group treatment; ST, standard treatment; SRDG, sustained-release dexamphetamine group; TAU, treatment as usual; TMC, telephone monitoring counseling; TOP, topiramate; UC, usual care; VBRT, voucher-based reinforcement therapy; MI-IOP, motivational interviewing telephone calls for reengaging intensive outpatient care; DC, desipramine continuity; DS, desipramine standard; PC, placebo continuity; PS, placebo standard; DTG, day treatment group; IG, inpatient group; NS, noncontingent schedule; GDC, group drug counseling; AT, abstinence-treated; PSYC, psychotherapy; NTX, naltrexone; CBC, cognitive behavioral counseling; VC, voucher control; FV, fluoxetine + voucher incentives; PV, placebo + voucher incentives; F, fluoxetine without vouchers; P, placebo without vouchers; CDC, group drug counseling; Cl.M, clinical management; MP, methylphenidate.

The highlighted words in the table assisted in counting and creating data for the discussion.

Table 1 depicts the characteristics of the reviewed studies. Most of the studies were conducted in the United States (91.7%). The median age of included publications was 1991 to 2020. The median sample size of reviewed studies was approximately 125.6 individuals. Urine toxicology was the most used instrument to evaluate cocaine use, followed by self-report questionnaires. As shown in Table 2, for the CUD interventions tested, 29 studies included at least one form of psychotherapy (family based, individual, group, or relapse prevention), 29 studies specifically included CM, and 58 focused on pharmacologic interventions (desipramine, amantadine, carbamazepine, fluoxetine, bupropion, imipramine, Ritanserin, Pergolide, disulfiram, paroxetine, Pentoxyphylline, Pramipexole, nefazodone, Olanzapine, Valproate, Coenzyme Q10, sertraline, Donepezil, Tiagabine, gabapentin, reserpine, Risperidone, Vigabatrin, Topiramate, Modafinil, naltrexone, Quetiapine, Galantamine, methylphenidate, phenytoin, dextroamphetamine, d-Amphetamine, tryptophan, Lamotrigine, Amoldipine, levodopa); 72 studies combined at least 2 of these modalities in combination.

Reporting of Attrition Rates

Thirty-eight studies (Table 1) identified an overall attrition rate above 50% across the placebo or intervention groups. Eighty-two studies evaluated retention in any form of CUD treatment as one of the study outcomes (Table 1).

Data are presented in Table 1. The attrition rates across all included studies were analyzed, and differences between the placebo and intervention groups were calculated and presented in the summarized data in Table 1. The articles that did not report these detailed data, such as the attrition data of the groups, were not included in Table 1.

In Table 2, we illustrated that 16 studies had a medium attrition bias,^17–32 and the other 68^{24,26,33–95} were rated as having a high attrition bias. Thirty studies with high attrition bias,^{7,34–36,39,40,43,46,49,50,52,54,55,58,61,64–68,72,75,76,78,79,81,85,88,90,92,97} employed interventions that combined pharmacologic and psychosocial interventions. It is important to state that these interventions were not included in the trial design specifically to minimize the attrition rate.

Among all studies with data on attrition, we found that 31 specifically identified attrition rates as being potential limitations of the analyses.^{7,17,22,32,33,37,38,42,48,49,52,53,55,64,56,69, 70,78,86,89,90,92,95–103}

Reporting of Methods for Mitigating Attrition

We identified 31 studies that used imputation procedures to adjust for attrition in analyses.^{19,20,22,23,25,35,36,39–43,46, 50,51,53,54,56,57,63,66,68,70,84,100,101,104–108} The summary of the imputation techniques is presented in Table 3. It is interesting to note that all but one were published in the past 20 years.

Table 3.

Imputation Techniques.

Grabowski, 1997	In the realm of repeated-measures designs, the management of dropouts has typically involved either the removal of subjects with incomplete data or the application of data imputation techniques to account for and estimate the missing values
Dackis, 2005	Missing urine screens were imputed as positive, assuming missing urine samples are not ignorable (and more likely to be positive)
Petry, 2005a	We coded missing visits as negative if the most proximal samples before and after the missing value had negative results
Petry, 2005b	Patients (n = 11) who did not complete the urine collection portion of the study were considered positive for missed samples
Malcom, 2005	Missing data occurring before dropout were imputed using linear interpolation. After that, data were stratified on missing data patterns due to attrition
Reid, 2005	Incomplete or missing data were accounted for by the mixed-effects model. The modeled estimates were based on all available weekly data. The analyses performed assume that the missing values are missing at random
Pierce, 2006	Missing visits were coded as negative when both of the proximal samples collected in the week before and after the missing value were negative for stimulants and alcohol
Ghitza, 2007	Missed specimens counted as positive. They facilitate intent-to-treat analyses by interpolating missing values
Gonzalez, 2007	Incorporation of all data including intermittent and dropout missing data is a strength of these mixed models because they compare the outcomes (or trajectories) across the various patterns of missing data
Petry, 2007	Missing samples are not included in the denominators, although results look similar if missing samples are considered positive
Loebl, 2008	Missing values in this analysis were imputed as positive for cocaine
Pettinati, 2008	Missing weeks were imputed as positive/use
Brodie, 2009	Missing urine toxicology tests were imputed to be dirty
Garcia-Rodriguez, 2009	Considered missing urine sample positive only if a sample provided before or after the missing sample was positive
Schmitz, 2009	Intermittent missing data were imputed as drug-positive; missing data due to attrition were handled as missing
Dackis, 2012	Missing urines were imputed as positive. UDS indicators missing samples being imputed as cocaine positive
Petry, 2012	Proportion of negative samples was analyzed in 2 ways: • Including only submitted samples in the denominator • Assuming no missing samples
Hagedorn, 2013	Excused absences were counted as negative if the samples provided at the appointments immediately prior to and immediately following the excused absence were negative. No-shows for scheduled appointments were counted as positive
Johnson, 2013	Imputed data for all dropouts as relapse to each participant’s baseline measure to provide a conservative estimate for the difference in treatment effect between Topiramate and placebo. We therefore conducted analyses for data with and without imputing missing data for dropouts
Petry, 2013	If a patient refused or missed a sample, the string was broken. Proportions of negative samples were analyzed in 2 ways: first, including submitted samples in the denominator, and second, using anticipated samples in the denominator
Kapman, 2013	Urine Benzoylecgonine Tests measures were performed with missing weeks ignored and missing tests imputed as positive, using a pattern mixture model based on the number of available weeks as an indicator of missing data
Secades-Villa, 2013	Missing urine samples were treated as positive, which is a method traditionally recommended for addressing missing data in substance abuse treatment research
Festinger, 2014	Maximum duration of abstinence was calculated treating missing data as positive
Nuijten, 2014	Missing data of 4 participants in each group were estimated with multiple imputation, using 5 imputed datasets, with baseline variables as predictors
Petitjean, 2014	Missed urinalyses were coded as positive
Schmitz, 2014	Missing or refused readings were considered noncompliant and reset the voucher value to $2.50
Umbricht, 2014	A sensitivity analysis was also conducted in which we imputed data for all dropouts as relapse to each participant’s baseline cocaine use
Kampman, 2015	Missing weeks are imputed as non-abstinence weeks because the Modafinil group had slightly more missing weeks than the placebo group
Levin, 2015	Insufficient individual data in these weeks were imputed as non-abstinent
Nuijten, 2015	Missing data were estimated with multiple imputation, using 5 imputed datasets
Petry, 2015	We analyzed proportions of negative samples submitted using the actual number of samples submitted in the denominator, a conservative approach that does not presume missed samples are positive

Abbreviation: UDS, urine drug screen.

We also identified 29 studies that employed attrition-mitigation methods in their study design, such as CM.

There were 10 articles^109–118 in our review that did not have enough data to be included in any of the proposed analyses.

Discussion

In summary, this scoping review aimed to review factors contributing to attrition rates in CUD treatment RCTs. According to Table 2, our main findings are as follows: (i) 80.1% (68 studies out of the 84 that presented this data) of the articles presented high attrition rates, while 19.2% (16 studies out of the 84) had a medium attrition rate; (ii) the interventions that led to lower attrition rates included, in order, medication, psychotherapy, CM, treatment as usual, counseling, and intensive treatments. In general, the reporting of attrition rates and strategies to mitigate attrition was inconsistent across RCTs evaluating CUD treatments.

Attrition seems to impair the success of RCTs evaluating CUD treatments. This is illustrated by our finding that attrition rates in CUD trials ranged from 4%¹¹⁹ to 80%.⁷ Our scoping review also shows that only 31 studies reported their high attrition rates as a limitation of the study. If loss-to-follow-up data is missing from a study, the data from the sample may not permit readers to adequately assess the comparability of the trial groups.¹²⁰ For instance, Akt et al state that loss to follow-up altered the interpretation of 58% of the 235 analyzed RCTs published in major medical journals.¹²¹

Imputation techniques have been developed to mitigate the errors missing data can bring to analyses. Specifically, imputation techniques can fill out missing values by “generating plausible numbers derived from distributions of relationships among observed variables in the data set.”^122,123

For example, a single imputation can be applied to replace missing data by the mean or median of the variable through interpolation or linear regression, and the replacement of missing data is computed only once. Because single imputation has been critiqued for its uncertainty surrounding the completeness of its imputed data, multiple imputation techniques have been developed such that several values are imputed for each missing data point, with a point estimate of the parameter obtained through the average of multiple imputations.¹²² Such aggregated information has been posited to provide a more accurate representation of the underlying uncertainty due to missing data, furnishing more robust estimations of statistics and confidence intervals, and mitigating the potential biases arising from incomplete data.¹²⁴ However, some authors suggest that when data are missing only in the outcome, multiple imputations do not address the bias and may add unnecessary noise to the estimates.¹²⁵ In fact, it is always difficult to address the trade-off between imputation versus clinical representativeness. There is little research on how those participants who drop out of a trial and those who finish it differ, at least not beyond the classic analysis of severity, co-consumption and psychiatric comorbidity. Future research should focus, for instance, on cognitive or social-cognitive skills differences between these groups. Notably, in our review, 31 studies^{19,20,22,23,25,35, 36,39–43,46,50,51,53,54,56,57,63,66,68,70,84,100,101,104–108} described the imputation procedure used as presented in Table 3. It is important to emphasize, though, that, as better as an imputation technique is, it will never eliminate the bias induced by high attrition rate; rather, it will only mitigate it.¹²⁶

More recently, motivated by the ENGAGE study,¹²⁷ addressed the challenge of low compliance by defining the target parameter as the mean outcome under a dynamic treatment regimen. They proposed a Bayesian approach based on principal stratification, which proved promising in multistage trials through extensive simulation studies and demonstrated robustness in nonlinear and non-Gaussian scenarios. They also used a Dirichlet process mixture model for the treatment sequence-specific outcomes.¹²⁷ Another strategy was conducted 2 analyses³³: one with participants who dropped out treated as “relapsed or using” and the other as “missing.” There is an issue of missing data in a Generalized Estimation Equation model by assuming that missing data were missing completely at random. Although these strategies described here are not examples of multiple imputation techniques, they might be judged as appropriate ways to address the issue of attrition or missing data.¹⁸

Although such statistical methods for handling missing data can be valuable, arguably, the best way to manage missing data is not to have missing data in the first place.^3,126

The need to devote a significant amount of effort during the design phase of an RCT to plan strategies to diminish the amount of attrition, hence decreasing the impact of missing data.¹²⁸

In the case of CUD, our review seems to show that comprehensive multidisciplinary interventions involving a combination of medication, psychotherapy (particularly CM), harm reduction, psychosocial supports, and intensive or outpatient treatment are essential to decrease attrition and improve adherence in RCTs. In our review, the studies classified as having medium attrition bias had at least one multidisciplinary intervention. However, it should be noticed that the amount or treatment offered should not be exaggerated, as a high participant burden can be associated with an increase in attrition rate.¹²⁹

It was noted in this scoping review that combining CM with pharmacologic treatments can be a highly effective strategy for improving retention rates in RCTs of pharmacologic CUD treatments.¹³⁰ However, in Brazil⁹ and other low- and middle-income countries (LMIC), this may be a costly and thus challenging strategy to implement as CM is considered expensive and time-consuming to administer due to its reward escalation.¹³¹ In addition, CM requires a highly trained professional team, which is expensive to obtain in resource-limited settings. In some LMICs, there may also be a stigma around CM, as studies have suggested that part of the population may perceive monetary incentives CM as “unfair” and “unethical” given economic stressors facing large proportions of the general population.¹³¹ In another example¹³² was applied a series of strategies to improve follow-up and minimize the attrition rate in a cohort study of men reporting a history of injection drug use before release from prison, encompassing high-frequency contact with participants (ie, in-person meetings, written letters, social media, and communication through health plans), culminating in strongly reduced attrition rates.¹³² Finally, some new studies have been showing that decentralized and digital health methods, such as text messages automatically sent via cell phones or app notifications reminding patients to take their medication, can be used to decrease attrition rates.^133,134

High attrition rates are not something specific to CUD. As far as we are aware, there are no other reviews on attrition rates of clinical trials of pharmacotherapy of SUDs. However, the literature provides some examples¹³⁵: an attrition rate of 17%, both on the intervention and the placebo group, testing a fatty acid amide hydrolase inhibitor for the treatment of cannabis use disorders. An attrition rate of 23.1% among the placebo group and of 52.5% among the Topiramate group, when testing the medication for cannabis use disorders.¹³⁶ In a review from clinical trials for alcohol use disorders, attrition rates ranged between 10% and 35%.¹³⁷ As for other psychiatric disorders, such as depression, rates of attrition in the first 12 weeks can be up to 65% in naturalistic settings and up to 36% in clinical trials.¹²⁹ Finally, for schizophrenia, the rage varies from 4% to 71%.¹³⁸

This review encompasses several limitations. First, scoping reviews do not seek to produce critically appraised and synthesized results but to highlight a map of the evidence base. Thus, we do not assess methodological limitations or RoB as a systematic review would.¹⁶ Second, we find tremendous heterogeneity in the types of CUD interventions and analytic frameworks encapsulated by the present review, limiting our ability to conduct a meta-analysis. Third, as this was a streamlined scoping review, we did not register our review on PROSPERO or assess each paper’s study quality using a formal instrument. Ideally, questions about CUD outcomes would be answered with an RCT; however, RCTs are not always feasible in populations with high socioeconomic deprivation, where attrition is a substantial threat to the success of RCTs. It is also difficult to conduct RCTs with an adequate sample size to detect differences in events such as overdose among people with CUD. Thus, a well-powered observational study based on real-world data can emulate an RCT, enabling the marginal treatment effect estimation. We have seen important advancements in statistical methods, allowing the use of observational data to simulate intention-to-treat (ITT) and per-protocol (PP) estimates. Because target trial emulations can produce separate ITT and PP effects, these analyses can provide a valuable opportunity to understand the impact of study attrition on the real-world effectiveness of our treatments without recruiting human subjects for a clinical trial.

Target trial emulation methods^139,140 can be conducted in 2 steps. First, the framework requires the investigator to specify the protocol of a hypothetical pragmatic randomized trial (target trial) to answer the question of interest (ie, CUD outcomes secondary to an intervention). Second, the investigator will emulate the trial using observational data. One of the key strengths of the target trial emulation method is that it requires the investigator to compute both an ITT effect and a separate PP effect. In a standard RCT, ITT analyses estimate the effect of treatment assignment and are agnostic about treatment discontinuation (attrition). In contrast, PP analyses estimate the effect of treatment if all patients in a standard RCT were retained in treatment. In target trial emulation studies, an ITT analysis will estimate the study outcome among all persons with CUD without adjusting for variables that predict attrition, whereas the PP analysis will adjust for predictors of attrition. In other words, the ITT effect will apply inverse probability weighting to account for confounding that may influence people’s likelihood of receiving a treatment condition, whereas the PP effect will adjust for baseline and time-varying confounders that predict attrition.

To our knowledge, the use of target trial emulation to evaluate the comparative effectiveness of CUD outcomes has not been studied. Given the challenges facing RCT development in this area, it is a promising future avenue for CUD research.

Conclusion

The attrition rate in RCTs for CUD treatment is commonly a limiting factor in the evaluation of the effectiveness of the CUD treatments tested in RCTs. Therefore, this scoping review, evaluating over 2 decades of CUD research, aimed to evaluate factors contributing to a lower attrition rate in RCTs. Our results highlight the importance of optimizing RCT design to increase participation so that subjects do not discontinue treatment, as well as the need for more consistent reporting of attrition in RCTs. Our analyses suggest that most RCTs evaluating CUD treatments have not accounted for attrition in their analyses or employed approaches to mitigate attrition.

Considering the difficulties of the feasibility of RCTs assessing CUD treatment, we consider that the use of observational data to emulate RCTs and make robust causal inferences may hold promise in the field. Specifically, target trial emulation is increasingly becoming a popular method for addressing time-sensitive causal questions without conducting RCTs with human subjects.

Supplementary Material

Supplemental material for this article is available online at the SAJ website http://journals.sagepub.com/doi/suppl/10.1177/0010414018XXXXXX.

Highlights.

• Attrition rates in randomized controlled trials (RCTs) for cocaine use disorder (CUD) treatment are commonly a limiting factor when evaluating the effectiveness of CUD treatments.

• Our scoping review assessed RCTs for CUD and identified factors associated with lower attrition rates.

• Most included papers did not adequately account for attrition in their analyses or employ approaches to mitigate attrition.

• It is important to optimize CUD RCT design to decrease attrition, as well as to improve methods for managing missing data.