## INTRODUCTION

Cocaine and heroin dependence each pose serious and substantial public health, social, and economic problems (ONDCP, 2003). Unfortunately, many opioid-dependent individuals who are treated with agonist medications such as methadone (MTD) and buprenorphine (BUP) still abuse cocaine, either independently or in heroin/cocaine ‘speedball’ combination (Magura et al, 1991; Torrens et al, 1991; Schütz et al, 1994; Craddock et al, 1997; Beswick et al, 2001; Leri et al, 2004). Cocaine and opioid use disorders often co-occur and this conjunction, relative to opioid dependence only, is associated with worse treatment outcomes (Perez et al, 1997; Preston et al, 1998; DeMaria et al, 2000; Downey et al, 2000; Sofuoglu et al, 2003; Tzilos et al, 2009), antisocial personality disorder (King et al, 2001), and risky drug-injection and sexual behaviors (Grella et al, 1995; Hudgins et al, 1995). Cocaine/heroin abusers engage in a disproportionate share of income-generating crimes (Strug et al, 1985; Cross et al, 2001). These adverse sequelae provide a compelling reason to find effective treatments for heroin-dependent individuals who abuse cocaine.

Substitution pharmacotherapies are an attractive option because they can produce cross-tolerance to drug reinforcement, replace harmful drug use behaviors with safer doses and routes of administration, and suppress withdrawal symptoms, thereby facilitating extinction and prolonged abstinence (Gorelick et al, 2004; Shearer and Gowing, 2004). Although agonist replacement is safe and effective for treating dependence on opioids (Kreek, 2000), MTD and BUP—despite some supportive data (Schottenfeld et al, 1993; Montoya et al, 2004)—are usually considered to be ineffective for reducing cocaine abuse. Most human laboratory tests of agonist-like agents for reducing cocaine self-administration have yielded negative results (Haney and Spealman, 2008). However, promising candidate medications under investigation include sustained-release formulations of monoamine agonists, which act on dopamine (DA) and norepinephrine (NE) neurotransmission (Rothman et al, 2002; Grabowski et al, 2004b; Castells et al, 2007), mechanisms that underlie the reinforcing effects of cocaine.

In previous studies, the effects of acute d-amphetamine (AMP) administration on cocaine self-administration have been mixed (Mansbach and Balster, 1993; Glowa et al, 1995; Lynch et al, 1998; Foltin and Evans, 1999; Barrett et al, 2004). In contrast, repeated or continuous AMP administration consistently produces significant reductions in cocaine-reinforced responding in rats (Peltier et al, 1996; Chiodo et al, 2008; Chiodo and Roberts, 2009), and cocaine vs food choice in rhesus monkeys (Negus, 2003; Negus and Mello, 2003a, 2003b). Importantly, chronic relative to acute AMP produces greater effects on D1 and D2 receptor responses (Ginovart et al, 1999; Kim et al, 2001) and glutamatergically mediated synaptic plasticity (Li and Kauer, 2004), although the impact of these neuroadaptations on responses to cocaine is not well understood. In human subjects, maintenance on sustained release AMP (SR-AMP) significantly reduced subjective effects, while increasing some physiological effects of intranasal cocaine (cumulative doses of 4, 34, and 94 mg cocaine after 3–5 days of stabilization on 0, 15, and 30 mg/day SR-AMP; Rush et al, 2009). At doses up to 60 mg/day, SR-AMP reduced cocaine use in two outpatient, placebo-controlled, randomized clinical trials, including one study with MTD-maintained, cocaine/heroin-dependent patients (Grabowski et al, 2001, 2004a).

Evidence from animal laboratory studies indicates that μ-opioid/cocaine combinations are self-administered in a dose-related manner (Mello et al, 1995; Rowlett and Woolverton, 1997; Duvauchelle et al, 1998; Rowlett et al, 1998, 2005; Cornish et al, 2005; Negus, 2005; Winger et al, 2006; Woolverton et al, 2008) and that speedball-maintained responding can be reduced by pretreatment with higher-dose BUP (Mello and Negus, 1998) or combined BUP/SR-AMP (Mello and Negus, 2007). Human laboratory studies with heroin-dependent participants have not examined actual seeking of cocaine or speedball combinations, nor interventions to reduce this drug use; this is a key feature of medication development efforts (Comer et al, 2008).

The present human laboratory study evaluated the potential efficacy and safety of a dual-agonist pharmacotherapy approach for non-treatment-seeking individuals with concurrent opioid and cocaine use disorders, using BUP and SR-AMP. The partial μ-opioid agonist BUP is an alternative to MTD that could be paired with SR-AMP (providing clinical flexibility), and BUP may be safer for and preferred by some patients. Our primary hypothesis was that, during BUP maintenance, SR-AMP would dose-dependently attenuate cocaine-seeking behavior and might reduce ‘speedball’-like (cocaine+hydromorphone), but not hydromorphone-seeking behavior. Our secondary hypotheses were that SR-AMP would suppress cocaine withdrawal symptoms and reduce cocaine's subjective and physiological effects. Finally, the safety of this medication (and experimental drug) combination also was evaluated.

## MATERIALS AND METHODS

### Participants

The local Institutional Review Board approved this study, which was conducted according to the Declaration of Helsinki. Volunteers, aged 18–55 years, were recruited by advertisements and word-of-mouth and not seeking drug abuse treatment. All provided informed consent. Screening included medical history, blood and urine chemistry, electrocardiogram and tuberculin testing, physical exam, and psychiatric interview (SCID-IV; First et al, 1996).

Volunteers met the DSM-IV criteria for current opioid dependence and cocaine abuse or dependence. Volunteers had to provide a supervised urine sample positive for opioids and cocaine and negative for MTD, amphetamines, and barbiturates. Benzodiazepine- or THC-positive urine samples were allowed, but sedative and cannabis use disorder diagnoses were exclusionary. Volunteers had to provide alcohol-free breath samples (<0.002%).

Volunteers were excluded if they: met the DSM-IV criteria for current axis I disorders, except opioid and nicotine dependence and cocaine abuse/dependence; were taking prescribed medications; had chronic health problems; were cognitively impaired (IQ <80) based on Shipley Institute of Living Scale (Zachary, 1991); or scored >15 on Medical Fear Survey Injection and Blood Withdrawal Phobia subscale (Kleinknecht et al, 1999).

### Study Design

#### Vital signs

HYD significantly reduced trough respiration rate and oxygen saturation; neither measure was influenced by COC or SR-AMP.

Cocaine significantly increased HR. SR-AMP increased HR by several beats per min, significantly more so in the presence of hydromorphone. SR-AMP did not potentiate cocaine tachycardia, but increased HR in the absence of cocaine (see Figure 3).

Cocaine significantly increased systolic and diastolic BP. SR-AMP did not enhance COC-induced increases in systolic and diastolic BP, but significantly increased systolic and diastolic BP by several mm Hg in the absence of cocaine (see Figure 3).

### Drug Reinforcing Effects (Choice PR)

Tables 1 and 4 provides descriptive statistics (means±1 SEMs) and overall ANOVA summaries for all measures of drug reinforcing efficacy.

#### Drug choices

In the overall analysis, the number of drug choices was significantly greater for all drug conditions relative to placebo, and there was a significant main effect for HYD. SR-AMP significantly reduced drug choices. In the additional planned hypothesis test (which excluded dual-placebo and HYD-alone conditions), SR-AMP significantly reduced all choices involving cocaine, dose F(2,14)=6.46, p<0.03, and more selectively reduced COC, but not speedball choices, SR-AMP × HYD F(2,14)=4.35, p<0.04. Participants tended to choose the speedball combination more often than cocaine alone, but this trend was not significant, HYD F(1,7)=5.04, p<0.06.

#### Breakpoints

Relative to placebo, log10 breakpoints were significantly greater for all drug conditions, and there was a significant main effect of HYD. SR-AMP significantly reduced log10 breakpoints. In the additional planned hypothesis test, SR-AMP significantly reduced log10 breakpoints, F(2,14)=9.81, p<0.02, and more selectively reduced COC, but not speedball breakpoints, SR-AMP × HYD F(2,14)=4.05, p<0.05 (see Figure 1).

#### Cumulative responding

Relative to placebo, cumulative responding was significantly greater for all drug conditions, and there was a significant effect of HYD. SR-AMP significantly reduced cumulative responding. In the additional planned hypothesis test, SR-AMP significantly reduced cumulative responding involving all cocaine choices, F(2,14)=10.54, p<0.01, and more selectively reduced COC, but not speedball responding, AMP × HYD F(2,14)=4.22, p<0.04.

## DISCUSSION

This study showed that when heroin/cocaine-dependent research volunteers were maintained on a moderate BUP dose (8 mg/day), stabilization on ascending doses of SR-AMP 30 mg/day then 60 mg/day (relative to initial placebo) significantly reduced cocaine- but not speedball-like or opioid-seeking behavior. This is the first human laboratory study to show that SR-AMP can reduce cocaine-seeking behavior in this comorbid population.

Animal laboratory studies that have used a chronic AMP administration protocol—similar to maintenance treatment—reliably showed that AMP reduced cocaine-reinforced operant responding in rats (Peltier et al, 1996; Chiodo et al, 2008; Chiodo and Roberts, 2009), and cocaine vs food choice in rhesus monkeys (Negus, 2003; Negus and Mello, 2003a, 2003b). In a human laboratory study (Rush et al, 2009), short-term (3–5 days) maintenance on SR-AMP (15 and 30 mg/day vs placebo) significantly reduced subjective and physiological effects of intranasal cocaine (at cumulative doses up to 94 mg). A recent follow-up study with nine cocaine-dependent individuals showed that maintenance on SR-AMP (40 mg/day vs placebo) modestly, but significantly reduced the number of choices (up to 6 units available at 45 min intervals) of an intermediate cocaine unit dose (20 mg), but not a lower (10 mg) or higher (30 mg) unit dose (Rush et al, 2010). Finally, two placebo-controlled, randomized clinical trials showed that SR-AMP doses ranging from 15 to 30 mg twice a day (30–60 mg/day) significantly increased treatment retention, whereas only SR-AMP 60 mg/day significantly reduced cocaine use (Grabowski et al, 2001, 2004a). Importantly, the combined use of MTD (1.1 mg/kg/day) and SR-AMP (up to 60 mg/day) was effective in one clinical trial (Grabowski et al, 2004a); however, its ability to reduce speedball-maintained responding has not yet been studied in the human laboratory setting. This study's finding is consistent with the above literature in showing that SR-AMP (within a similar dose range) significantly reduces cocaine-reinforced responding. This scientific demonstration is important because it contributes evidence toward the predictive validity of laboratory-based medication screening for cocaine dependence (Comer et al, 2008; Herin et al, 2010). Other than this collection of SR-AMP findings, the only other instance of concordance between human laboratory and clinical trial efficacy findings for an anti-cocaine medication has occurred with modafinil (Dackis et al, 2005; Hart et al, 2008; Anderson et al, 2009).

Our expectation for SR-AMP, as an agonist replacement approach, is that relatively higher doses should reduce cocaine reinforcement just as relatively higher BUP doses should reduce μ-opioid reinforcement. Attenuating the self-administration of cocaine/opioid combinations is a desirable, but not guaranteed, consequence of adequate medication combination doses because the reinforcing effects of these abused drugs are mediated by partially distinct neurobiological mechanisms. The results of this study indicate that SR-AMP did not significantly alter drug seeking for HYD, either alone or combined with cocaine. This occurred during maintenance on a moderate BUP dose (8 mg/day), under which condition HYD serves as a reinforcer (Greenwald and Hursh, 2006; Greenwald and Steinmiller, 2009; Greenwald, 2010).

This study's design did not provide a complete test of the dual-agonist replacement hypothesis. Specifically, we used a single moderate BUP dose to suppress opioid withdrawal while enabling HYD μ-agonist effects to serve as a positive control condition. A full test of the hypothesis would have used additional higher BUP doses. Under our test conditions, SR-AMP doses did not significantly alter the reinforcing efficacy of HYD or the speedball analog. This pattern indicates that the reduction in cocaine-seeking behavior was pharmacologically specific and that the laboratory model was reliable (ie, subjects consistently responded for HYD and speedball across protocol weeks). Previous human laboratory studies have shown that higher BUP maintenance doses (16–32 mg/day) significantly decrease opioid seeking (Greenwald et al, 2002; Comer et al, 2005). Thus, our working hypothesis—which requires further research—is that increasing the BUP dose in combination with SR-AMP could decrease the reinforcing efficacy of μ-opioids and cocaine individually and, potentially, opioid/cocaine combinations. Such an outcome would also be consistent with the results of one clinical trial showing that very high BUP doses (24–32 mg/day) alone could reduce but not eliminate cocaine use (Montoya et al, 2004). Experimental confirmation of this hypothesis could point the way toward a dual agonist-replacement approach for this polydrug-using population. Although concerns about the safety of, and compliance with, a dual-pharmacotherapy regimen poses challenges, the benefit/cost ratio of this approach is favorable given the disproportionate harms in this population.

Interestingly, the Rush et al (2009) study did not find a significant effect of SR-AMP on intranasal cocaine money value based on crossover points in the MCP or ratings of ‘willing to pay for’ cocaine. However, mean cocaine value in the placebo SR-AMP condition of this study, which used a hypothetical, non-reinforced MCP (US$9.36; Table 3), was considerably higher than cocaine's value in the Rush et al (2009) study (≈US$2.00). In our planned hypothesis test that included the cocaine and speedball conditions, SR-AMP did not significantly reduce cocaine money value. This finding that SR-AMP significantly decreases cocaine-seeking behavior, but not its monetary value, suggests that the choice PR procedure may be a more sensitive assay under these conditions. The lack of a significant SR-AMP effect on cocaine value observed by Rush et al (2009) could have been owing to a floor effect in that subject sample. The inconsistent effect of SR-AMP on reducing cocaine choices across unit doses (Rush et al, 2010) could, as those authors explain, be related to the use of a potentially less-sensitive discrete choice procedure.

These data also show that SR-AMP (1) partially attenuated negative baseline symptoms (possibly related to short-term cocaine abstinence while living on the in-patient unit), and (2) significantly attenuated subjective responses to cocaine. First, independent of the test drug (HYD or cocaine) condition, SR-AMP generally decreased ratings of stimulation (ie, dose main effect for total scores on the SSARS, and VAS ‘stimulated’). SR-AMP also decreased the ‘want drug again’ VAS rating and, for reasons that are not entirely clear, total scores on the Heroin Craving Questionnaire. SR-AMP tended to reduce baseline cocaine abstinence symptoms on the CSSA, although this was not statistically significant. SR-AMP did not significantly increase baseline opioid withdrawal symptoms (which were low) during BUP maintenance, nor did it increase ratings of ‘bad drug effect’ or sedation. Second, several questionnaire-based indices of cocaine drug effect were reduced during active SR-AMP treatment. Cocaine craving, which increased after cocaine sampling doses, was attenuated during maintenance on SR-AMP 30 mg/day but not 60 mg/day. SR-AMP treatment significantly attenuated cocaine-induced ratings on the ARCI–BG scale, and significantly attenuated cocaine-induced ratings of ‘good drug effect’, ‘high’, and ‘liking’.

The safety profile of SR-AMP in combination with BUP was also evaluated. SR-AMP modestly increased baseline HR and BP in the absence of cocaine or HYD, but SR-AMP did not significantly potentiate the cardiovascular effects of cocaine. These findings are consistent with data from Rush et al (2010), and suggest that SR-AMP may be generally safe, especially when careful screening precautions are implemented. On the other hand, two participants were excluded for adverse events at the start of in-patient week 2, which coincided with induction onto the first active dose of SR-AMP (30 mg/day). One participant experienced temporary nausea/vomiting soon after admission (placebo SR-AMP) and continued, but was later discharged upon reporting feelings of depression after 3 days on SR-AMP 30 mg/day. The other participant exhibited tachycardia (HR >90 b.p.m.) and reported headache (unresponsive to acetaminophen and ibuprofen) and photophobia once SR-AMP 30 mg/day was started. These signs and symptoms for both individuals quickly resolved once participation was terminated. This also emphasizes the virtue of dose escalation in laboratory studies and for potential clinical use (see Herin et al, 2010, for a discussion of risk minimization).

There is presently no basis on which to expect a pharmacokinetic interaction between SR-AMP and BUP. First, these medications are primarily metabolized by different pathways. In humans, the primary mode of AMP metabolism occurs through deamination to phenylacetone (Green et al, 1986), with 4-hydroxylation by cytochrome P450 (CYP) 2D6 as a minor pathway (Bach et al, 1999), which differs from rats (Tomkins et al, 1997). In contrast, BUP is metabolized to nor-BUP by CYP3A4 (eg, Kobayashi et al, 1998; Moody et al, 2002), with additional involvement of CYP2C8 (Moody et al, 2002). Second, while there is in vitro evidence that BUP can weakly inhibit CYP2D6-catalyzed reactions (eg, Umehara et al, 2002; Zhang et al, 2003), BUP concentrations used in those studies substantially differed from therapeutic concentrations. BUP is therefore unlikely to have a major effect in altering clearance of AMP.

These data suggest that, among cocaine/heroin-dependent individuals, precautions are necessary to avert potential side effects of SR-AMP. All participants underwent extensive medical and psychiatric screening, and SR-AMP doses were administered in ascending order (which may confound the results) to minimize safety problems, consistent with previous human studies (Grabowski et al, 2001, 2004a; Rush et al, 2009) and clinical practice (Herin et al, 2010). If SR-AMP were to be adopted in drug abuse treatment programs, similar safeguards would be needed to screen and monitor patients. Having noted this issue, it can also be concluded that for individuals who safely tolerated these SR-AMP doses, the medication was effective. Efficacy, as measured by reduced breakpoints, was observed in seven of eight completers (Figure 1). The maximum dose in this study (60 mg/day) exceeded the dose of 40 mg/day in the laboratory study by Rush et al (2010), but matched the upper dose in two clinical trials (Grabowski et al, 2001, 2004a). These data also show that the lower SR-AMP dose (30 mg/day) was equal in efficacy to the higher dose (60 mg/day), suggesting that it may not be necessary to escalate the dose further in all individuals.

The primary limitation of this study is the relatively small sample size, leading to concerns about generalizability of the study findings. Nonetheless, we observed significant and mostly selective effects of SR-AMP on reduction of cocaine-seeking (primary outcome) and cocaine-induced subjective effects (secondary outcome) without potentiation of cocaine-induced cardiovascular effects (safety outcome). Concerns about external validity are also mitigated by the fact that Rush et al (2009, 2010) observed SR-AMP attenuation of cocaine's reinforcing and subjective effects in cocaine-dependent individuals. Several animal studies also showed the ability of SR-AMP to reduce cocaine self-administration using only a few rats or rhesus monkeys. Finally, two clinical trials with moderately sized samples showed the ability of SR-AMP to reduce cocaine use. On the other hand, some findings of this study were statistically marginal (eg, trend for SR-AMP to reduce baseline cocaine abstinence signs on the CSSA), probably owing to a lack of power. However, the study was not designed to detect effects on these secondary outcomes, rather the intent was to include a broad range of measures for hypothesis testing in future studies. Another limitation is that the routes of experimental drug administration used here (intramuscular HYD and intranasal COC) do not match the typical routes used naturalistically by these study participants (see Table 2). Specifically, intravenous/intranasal heroin and smoked cocaine (predominant among these subjects) may produce faster-onset pharmacokinetics with more profound pharmacodynamic results compared with our experimental routes. On the other hand, we consider these concerns to be mitigated in at least three ways. First, these pharmacodynamic effects should be qualitatively similar. Second, the designed staggering of HYD and COC administration (by 15 min) succeeded in aligning peak responses to the two drugs (thus simulating speedball-like effects). Third, as noted above, these human laboratory results are concordant with both preclinical self-administration and clinical trial findings.

The scientific rationale for this study was based on two key ideas: (1) combining two agonist medications for treating cocaine/heroin abusers could be effective for reducing both types of illegal drug use, and (2) combining SR-AMP with a partial μ-agonist (BUP) could be safer than with a full μ-agonist (MTD). There may be several advantages to combining anti-cocaine medications such as SR-AMP with BUP, rather than MTD, aside from the important fact that BUP is another treatment option for the clinician and patient. First, BUP could be safer, better tolerated, or more preferred than MTD for some individuals. On the other hand, we recognize that some patients with more severe opioid physical dependence could benefit from MTD as a μ-opioid full agonist medication in combination with SR-AMP. Alternatively, patients with less severe opioid addiction could initially be offered BUP in combination with SR-AMP and, if BUP does not prove effective, the patient could be switched directly to MTD in combination with SR-AMP; Herin et al (2010) recently discussed this type of graded agonist approach. Second, BUP is longer acting than MTD, which could facilitate less-than-daily dosing; however, this is based on the non-trivial assumption that take-home doses of SR-AMP could be dispensed, which would have to be evaluated for safety and diversion concerns. Third, BUP can be prescribed by appropriately trained physicians in office-based settings, which could promote greater access to treatment and reduction in stigma to cocaine/heroin-dependent patients.

In conclusion, this study provides the first demonstration in the human laboratory setting that SR-AMP attenuates cocaine-seeking behavior, and that this effect is selective in cocaine/heroin-dependent individuals. These data are consistent with other preclinical and clinical data, suggesting that this is a robust effect. Given the large number of putative medications that have been evaluated and failed to reduce cocaine self-administration in the human laboratory setting and cocaine use in clinical trials, this is an important result. With proper attention paid to safety considerations, these novel findings suggest that a polypharmaceutical combination of higher-dose BUP and SR-AMP could be an effective approach for treating individuals who abuse cocaine and opioids.