Recently it has proven useful to test potential cocaine medications in the context of specific behavior therapy platforms. We (Grabowski et al., 1995) and others (Covi et al., 1995) reported fluoxetine combined with cognitive psychotherapy to be ineffective but later demonstrated efficacy when evaluating fluoxetine in combination with abstinence contingent incentives (Schmitz et al., 1998). Kosten and colleagues (Kosten et al., 2003) reported an additive effect of the antidepressant desipramine and CM in the treatment of cocaine use in buprenorphine-maintained patients. In that study, the combination treatment had almost twofold greater efficacy in improving cocaine-negative urines than either treatment alone. Poling and colleagues (Poling et al., 2006) reported a synergistic effect of CM and bupropion for the treatment of cocaine dependence in methadone-maintained patients. Cocaine use was lower in the group receiving CM treatment with 300 mg/d of bupropion than in the group receiving CM with placebo, or in the two voucher control conditions with or without bupropion. Thus, the behavioral therapy context may be a particularly important determinant of effectiveness for some medications.

Levodopa, the dopamine precursor considered uniquely efficacious in treatment of Parkinson's disease, has particular relevance as a candidate medication in the treatment of stimulant abuse since dopamine plays a critical role in mediating the acute reinforcing effects of cocaine (Koob, 1992; Volkow et al., 2004), and the depletion of dopamine associated with chronic cocaine use is thought to underlie cocaine craving and relapse following periods of abstinence (Dackis & Gold, 1985; Shalev et al., 2002). Early evidence that chronic cocaine administration might result in dopamine depletion came from reports of hyperprolactinemia in abstinent patients following chronic cocaine abuse (Dackis et al., 1984; Dackis & Gold, 1985; Teoh et al., 1990). The clinical observation that dopamine-regulated prolactin levels vary inversely with available dopamine in the synapse provided the rationale for studying dopamine agonist drugs like bromocriptine which presumably worked by reversing dopamine supersensitivity caused by dopamine depletion (Dackis & Gold, 1985; Giannini et al., 1987; Giannini & Billett, 1987). Neurochemical studies in animals have shown changes in brain dopamine function consistent with the dopamine depletion hypothesis (e.g., Pradhan et al., 1978; Mateo et al., 2005). Imaging studies of the human brain have linked cocaine-induced decreases in dopamine concentrations with disruptions of prefrontal regions involved with reward, motivation, memory, and cognitive control (reviewed in Volkow et al., 2003). Thus, the collective findings support the hypothesis that chronic cocaine exposure produces dopaminergic dysfunction and that successful pharmacological interventions might be those having a restorative effect on this neurotransmitter system.

Levodopa, converting to dopamine in the brain, should increase CNS levels thus ameliorating deficits and enhancing function following cocaine discontinuation. Cocores and colleagues (Cocores et al., 1989) were the first to report reduced cocaine withdrawal symptomatology when levodopa or bromocriptine was administered in addition to standard outpatient treatment. Wolfsohn et al. (Wolfsohn and Angrist, 1990) reported similar encouraging results in an open trial, but these findings were not replicated in a subsequent placebo-controlled double-blind trial (Wolfsohn et al., 1993). More recently in a NIDA sponsored Clinical Rapid Efficacy Screening Trial (Shoptaw et al., 2005), levodopa/carbidopa (300/75 mg/d) was not superior to placebo on urine benzoylecgonine (BE) or self-reported measures of cocaine use. Our exploratory pilot and subsequent dose-ranging studies (placebo vs. levodopa/carbidopa 400/100 mg/d or 800/200 mg/d) indicated safety and tolerability but no reduction in cocaine use (Mooney et al., 2007). Thus, a study combining levodopa with a more potent behavioral therapy was predicated on work by us and others indicating instances of CM enhanced medication effectiveness (e.g., Kosten et al., 2003; Poling et al., 2006; Schmitz et al., 1998).

Levodopa-treated subjects had lower cocaine use rates, longer periods of continuous abstinence, and less craving compared to patients assigned to placebo. These findings stand in contrast to those reported in an earlier medication screening trial in which levodopa, (tested at a lower dose, 300/75 mg/d) showed no advantage over placebo (Shoptaw et al., 2005). Additionally, Mooney et al. (Mooney et al., 2007) documented the safety and feasibility of levodopa/carbidopa therapy (800/200 mg/d) in an initial safety trial and in a subsequent dose ranging study but reported no clear decrease in cocaine use as a function of levodopa dose. Given the apparent safety and tolerability of levodopa (Mooney et al., 2007), the present results are highly encouraging with regard to the potential utility of levodopa pharmacotherapy in reducing cocaine use when combined with appropriate behavioral interventions.

There is consensus in the scientific community that combining behavioral treatments and pharmacological agents provides more effective treatment than either constituent therapy alone in most cases (Carroll et al., 2004b; Onken et al., 1995; Stitzer and Walsh, 1997; Volkow et al., 2004). Carroll (Carroll et al., 2004b) however, specifically stated that determinations of optimal behavioral therapy platform-pharmacotherapy combinations should be evidence based, not arbitrarily determined. The process should involve consideration of the specific strengths and weaknesses of the medication, as well as the potential targets of the behavioral therapy platform (Carroll et al., 2004b). Levodopa's strengths, i.e., safety, tolerability and compliance, are offset by modest direct pharmacotherapeutic effects (Mooney et al., 2007; Shoptaw et al., 2005). Still, our finding that a high intensity behavioral therapy platform enhances the pharmacotherapy efficacy provides the evidence for a balanced, tenable therapy combination using the criteria specified by Carroll (Carroll et al., 2004b). That we did not observe a linear relationship between therapy “dose” or intensity and medication effect suggests that levodopa was effective because of its unique interaction with VBRT. Findings from this trial add to other recent reports showing a significant and specific synergism between contingency management and cocaine pharmacotherapy (Kosten et al., 2003; Poling et al., 2006; Schmitz et al., 1998).

Although not designed to investigate treatment mechanisms, the trial permits speculation that might direct future research to elucidate the underpinnings of enhanced effectiveness of the levodopa contingency management combination. One explanation comes from recent models of dopamine and reward saliency. Koob and others (Koob and Le Moal, 2001) have proposed that chronic cocaine use produces persistent diminution of brain reward function, resulting in a condition in which the saliency of drug-related cues far exceeds the saliency value of natural non-drug reinforcers (Ahmed et al., 2002; Ahmed and Koob, 1998; Koob and Le Moal, 1997, 2005). Contingency management increases availability of non-drug or environmental reinforcers directly competing with drug use. To the extent that processes related to reward saliency are mediated via dopamine transmission (e.g. Berridge and Robinson, 1998; Volkow, 2004), levodopa administration might compensate for dopamine-related deficits and thus enhance saliency of contingently delivered reinforcers, thus improving responsiveness to this behavioral intervention. Conversely, this proposed mechanism suggests that in the absence of alternative reinforcers the therapeutic efficacy of levodopa would be diminished, precisely the observation in the non-voucher conditions. If confirmed as the treatment mechanism, levodopa and other dopamine-augmentation strategies could be used as a way to enhance the efficacy of contingency management.

Early onset of levodopa treatment benefit suggests that it may also attenuate cocaine withdrawal symptoms, necessarily by the same dopaminergic mechanism. This is supported by the decreased self-reported cocaine craving in subjects receiving active medication. Theoretical arguments concerning dopaminergic perturbations contributing to persistent craving have been posited by Gawin, Dackis and others (Dackis and Gold, 1985; Gawin and Ellinwood, 1988). Empirical evidence for the perspective comes from McDougle et al. (McDougle et al., 1992) who, using a levodopa challenge paradigm, demonstrated increased dopamine responsivity during early cocaine abstinence. Wolfsohn (Wolfsohn and Angrist, 1990) initially reported rapid resolution of abstinence symptoms in 8 patients receiving levodopa on a detoxification unit, but failed to replicate these preliminary results in a later placebo-controlled study (Wolfsohn et al., 1993). Further studies that include measures of cocaine withdrawal symptoms are needed to clarify the extent to which reduction of withdrawal is a mediator of levodopa's effects on cocaine use.

Interestingly, our study did not find strong and reliable advantages of contingency management procedures relative to the non-voucher therapies. In non-medication studies of behavior therapy for cocaine dependence, the relative advantage of VBRT in reducing cocaine use has been robust (e.g., Higgins et al., 1994; Higgins et al., 1991; Rawson et al., 2006). Possible explanations may have to do with differences in the implementation of VBRT. In the present study contingent rewards were distributed once per week, whereas standard procedures recommend delivery of the earned reinforcer immediately following the target behavior, i.e., urine test results. Given that immediacy of contingent reinforcement is a robust predictor of behavioral response in both human laboratory (Roll et al., 2000) and treatment research (Lussier et al., 2006), the delayed procedure used here may have weakened the contingency management effect. Our contingency management procedure also differed in that it was not integrated with the other behavioral interventions (i.e., ClinMan, CBT) but rather delivered in parallel as a standalone intervention. Therapists delivering CBT did not address whether or not their patients were earning voucher rewards, which may have removed an important source of support and encouragement toward achieving targeted behaviors.

As with other cocaine clinical trials, the majority of individuals who began treatment did not complete it. Although we applied current intent-to-treat analytic strategies for handling missing data (Bryk and Raudenbush, 1992), high rates of attrition can complicate interpretation of true treatment effects (Nich and Carroll, 2002). In this study, small group sizes, combined with slightly higher retention in the levodopa + VBRT, may have introduced bias in the analysis of certain outcomes, particularly those that rely on length of time in treatment (e.g., continuous abstinence rates). Interpretability of the present findings is further affected by the additive design used in constructing the levels of behavior therapy. Because VBRT was added to a behavioral platform consisting of ClinMan+CBT, the data do not allow conclusions regarding the separate and independent effects of this component. Moreover, we cannot rule out the possibility that better CBT attendance in the VBRT group accounted for observed treatment differences.