## INTRODUCTION

Cocaine dependence is a major public health problem that is characterized by recidivism and a host of medical and psychosocial complications. An effective pharmacotherapy has long been sought to improve treatment outcomes, particularly since this disorder has a significant neurobiological basis (Volkow and Fowler, 2000; Dackis and O'Brien, 2001; Ahmed et al, 2002). Although proven pharmacotherapies are available for alcohol and heroin dependence (O'Brien, 2001), none exist for cocaine dependence despite two decades of clinical trials primarily involving antidepressants, anticonvulsants, and dopaminergic medications. Testing glutamate-enhancing agents is a new and promising strategy based on recent findings that cocaine dysregulates reward-related glutamate pathways (Thomas et al, 2001; Dackis and O'Brien, 2003; Kalivas et al, 2003).

Modafinil is a wake-promoting agent that is approved for narcolepsy. Its glutamate-enhancing action (Ferraro et al, 1998; 1999) might be clinically advantageous in cocaine dependence because the repeated administration of cocaine depletes extracellular glutamate levels (Keys et al, 1998; Bell et al, 2000; Hotsenpiller et al, 2001; Kalivas et al, 2003), and reduces glutamatergic synaptic strength in the nucleus accumbens (Thomas et al, 2001). Furthermore, it has recently been demonstrated that normalizing extracellular glutamate levels with N-acetylcysteine abolishes cocaine-induced reinstatement, an animal model of relapse (Baker et al, 2003). Modafinil has also been proposed as a cocaine detoxification agent (Dackis et al, 2003). Cocaine withdrawal symptoms (see Table 1) predict poor outcome (Kampman et al, 2001), and modafinil would be expected to reverse these symptoms as it is activating (Dackis et al, 2003). In addition, it has recently been reported that patients with baseline cocaine withdrawal (Sofuoglu et al, 2003), and abstinence-related depression and irritability (Newton et al, 2003), report enhanced euphoria after cocaine administration under controlled conditions, suggesting that the reversal of cocaine withdrawal might be clinically advantageous. Modafinil is a reasonable medication for cocaine-dependent individuals because it has low abuse potential (Jasinski, 2000; Jasinski and Kovacevic-Ristanovic, 2000; Rush et al, 2002a; 2002b), neurochemical effects that differ markedly from those of psychostimulants (Lin et al, 1996; Ferraro et al, 1997), and good tolerability (Becker et al, 2004).

Given these considerations, we began to investigate modafinil in cocaine-dependent patients some time ago by conducting a drug interaction study. This study not only found that modafinil was safely co-administered with intravenous cocaine, but also reported that modafinil significantly blunted cocaine-induced euphoria under controlled conditions (Dackis et al, 2003). The consistency of this interesting property was strengthened by another controlled study reporting attenuated cocaine euphoria in modafinil-treated patients (Malcolm et al, 2002). We subsequently conducted an open-label modafinil trial and reported high levels of cocaine abstinence in patients experiencing severe cocaine withdrawal at baseline (Dackis and O'Brien, 2003). The current study is a randomized, double-blind, placebo-controlled, 8-week trial of modafinil in 62 cocaine-dependent outpatients.

## PATIENTS AND METHODS

### Outcome Measures, Schedule of Assessments, and Sample Size

The primary efficacy measure for this trial was cocaine abstinence based on thrice-weekly qualitative urinary benzoylecgonine (BE) levels, and we hypothesized that modafinil-treated patients would supply more BE-negative urine samples than placebo-controlled patients over the 8-week trial. Urine temperatures were measured at the time of collection and acceptable samples (temperatures between 94° and 99° Fahrenheit) were analyzed with a fluorescent polarization assay. Urine samples containing BE at concentrations 300 ng/ml were considered to be positive for cocaine. Secondary efficacy measures included the weekly physician-rated Clinical Global Impression—Objective Scale (CGI-O), (Guy, 1976) self-reported cocaine use on the Timeline Follow-Back Interview (TLFB) (Sobell and Sobell, 1992) and Clinical Global Impression Scale—Subjective (CGI-S) (Guy, 1976), and cocaine withdrawal/craving measured by the Cocaine Selective Severity Assessment (CSSA) (Kampman et al, 1998), Brief Substance Craving Scale (BSCS) (Somoza et al, 1995), and Cocaine Craving Questionnaire (CCQ) (Tiffany et al, 1993). In addition, baseline demographics and clinical characteristics were assessed with the Addiction Severity Index (ASI) (McLellan et al, 1992), Hamilton Anxiety Scale (Ham A) (Hamilton, 1969), and Hamilton Depression Scale (Ham D) (Hamilton, 1967), Beck Depression Inventory (BDI) (Beck et al, 1974), and the Symptom Checklist 90 Revised (SCL90) (Derogatis, 1977). Safety data were collected weekly, and included vital signs (blood pressure, pulse, temperature), body weight, and adverse events documented by the study physician on a standardized form.

### Statistical Analysis

The analysis was by intention-to-treat. The patients were compared on a variety of baseline characteristics, using χ2 tests for categorical characteristics, and t-tests for continuous characteristics, to assess how well the randomization had balanced patient characteristics across the two (moderately sized) treatment groups. The primary analyses did not include additional covariates, but characteristics that showed significant imbalance across the groups were considered for inclusion as covariates in supplementary analyses, together with characteristics known to be of importance such as cocaine positive urine at baseline.

The repeated binary outcomes obtained from the quantitative BE assays were analyzed using generalized estimating equation (GEE) models (Diggle et al, 2002). In these analyses, missing urine screens were imputed as positive, which is a standard practice in clinical trials for which cocaine abstinence is a primary outcome (Shoptaw et al, 2002), although it does create a bias against treatment dropouts. Other repeated outcomes (CGI-O, CGI-S, TLFB, CSSA, BSCS, CCQ) were also analyzed using GEE models for continuous or count responses. In all the repeated measures analyses, the models included terms for treatment group (placebo versus modafinil), together with linear and quadratic time effects, and some group by time interactions. Quadratic time effects were included in the models to allow for the possibility that rates of cocaine use might decrease (or increase) early in treatment, and remain at a lower (or higher) level through the rest of the trial. Group by quadratic time interactions allow these patterns to be different for the different groups. In fitting these models to the data, terms were included in the GEE models if they were significant at the 5% level, and lower order effects contained in a significant interaction effect were also included. Model-based standard errors (Wald statistics) were used to assess significance.

## RESULTS

### Baseline Measures

There were no significant (5% level) differences between the modafinil and placebo groups on a battery of measures drawn from the baseline ASI, TLFB, Ham A, Ham D, BDI, SCL-90, and CSSA. A selection of important baseline demographic, clinical, and psychosocial characteristics of enrolled patients are listed in Table 2. As race, mean weekly cocaine cost, mean days of weekly use, and years of lifetime cocaine use were significant at the 10% level, they were included as covariates in supplementary GEE analyses.

### Pill Compliance

There were no significant differences between the modafinil and placebo group on pill compliance or pill overuse. A total of 41 patients (21 placebo, 20 modafinil) had at least 1 week where they did not return all of the expected unused pills in their blister packs (9-day supply given weekly). A χ2 test showed no difference in the group proportions with at least one such week (χ2 (1)=0.01, p=0.93). In each group, the median number of weeks in which unprescribed modafinil pills were retained 2 (third quartile of 3), and a Mann–Whitney test of the number of weeks of unprescribed pill retention among these patients showed no differences between the two groups (Z=−0.14, p=0.99). Across weeks in which an excess of pills were used, the total numbers of excess pills were also very similar: the mean 8-week-total overused was 11.82 (SD=8.25) in the placebo patients, and 12.80 (8.52) in the modafinil patients (Mann–Whitney Z-score=−0.33, p=0.77). When queried, patients typically stated they could not recall, or were reluctant to disclose, why they did not return the extra blister packaged pills. Three modafinil-treated and four placebo-treated patients retained more than 16 of the extra 64 pills supplied over 8 weeks.

### Treatment Retention

In the modafinil group, 19/30 (63.3%) of patients were still present at the end of the eight weeks, compared to 21/32 (65.6%) in the placebo group. Also, for each group, the median time to drop out was 24 visits. When defined as failing to complete the first 2 weeks of treatment, there were two dropouts in the modafinil group, and three in the placebo group. A log-rank test showed no significant difference in the distribution of time to drop out (χ2 (1)=0.79, p=0.37). There was no significant difference with regard to CBT attendance for the modafinil (50.0%) compared to the placebo (54.6%) groups (t(60)=0.64, p=0.53).

### Cocaine Abstinence

The trial was designed to obtain 24 urines from each patient during medication treatment. The percent obtained of expected urines for the modafinil group was 74.5%, and for the placebo group was 63.0%: a t-test showed that the average number of urines provided was not significantly different across the two groups (t(60)=1.58, p=0.12). Figure 1 shows the within-group percentages of BE-negative urines provided across the study. The mean proportion of BE-free urines supplied (calculated for each patient as a percentage of the 24 requested) was 42.3% for the modafinil group, and 24.0% for the placebo group. A full GEE model of the log-odds of a clean urine vs a missing or dirty urine screen showed significant group by quadratic time (Z=−2.25, p=0.02) and group by linear time (Z=2.49, p=0.01) effects. Examination of predicted probabilities of BE-negative urine (‘abstinent visit’) suggested that the groups had similar initial probabilities of a BE-negative urine (0.29 for modafinil, 0.25 for placebo), that these probabilities decreased steadily in the placebo group (0.23 at week 5 and 0.18 at week 8), and initially increased and later decreased in the modafinil group (0.46 at week 5 and 0.32 at week 8). (Note: If significance was assessed using the more conservative (Stokes et al, 2000) empirical standard errors, then the group by quadratic effect was not significant (p=0.09), and dropping further nonsignificant terms from the model yielded a final model with a significant main effect for cocaine abstinence in the modafinil group (odds ratio=2.41, 95% CI=(1.09, 5.31), p=0.03), compared to placebos, together with a significant quadratic time effect.) Analyses adjusting for race, and for baseline measures of mean weekly cocaine cost, mean days of weekly use, and years of lifetime cocaine use, yielded virtually identical estimates to those given by unadjusted analyses.

The baseline level of cocaine use has been reported to be an important predictor of treatment outcome (Kampman et al, 2001). In this study, there was a trend towards a higher proportion of BE-positive urines at baseline in the placebo group compared to the modafinil group. A complex model was formulated that included terms representing baseline urine BE status and its interactions with treatment group and time. There was a significant group by baseline status by quadratic time effect (p=0.01), indicating significantly different quadratic trends across the four groups (baseline BE status by treatment condition). Table 3 shows the estimated probabilities of an abstinent visit at the beginning, middle, and end of the 8 weeks, for each of the four groups.

As expected, based on previous work, the dominant effect is that people with a BE-negative baseline status are more likely to attain abstinent visits during the study. Within each level of baseline status, the modafinil effect is similar to that observed in the analyses described above: the groups are similar at the beginning, the modafinil group appears to improve more than the placebo group through the first 5 weeks, and the groups are similar at the end of the 8 weeks. In addition to this effect, a greater percent of the modafinil (33%) than placebo (13%) were able to attain prolonged abstinence, as defined by the achievement of 3 consecutive weeks of BE-negative urines (none missing) during any time period (χ2=3.84, df=1, p=0.05).

### Patient-Reported Use (TLFB, CGO-S)

There were no significant modafinil effects on self-reported (TLFB) rates of cocaine use (GEE model Z-score=0.35, p=0.73), or on dollars spent on cocaine (GEE model Z-score=−1.08, p=0.28). CGI-S showed no differences between modafinil and placebo groups in reported cocaine severity (GEE model Z-score=−0.37, p=0.72) or functional impairment.

### Physician-Rated Assessments

The CGI-O summary scales were rated weekly by the study physician. The modafinil group showed consistently lower overall severity scores, and consistently greater improvement scores, although these differences did not reach significance: for the Global Severity of Cocaine Dependence Scale (GEE model Z-score=−1.47, p=0.14); for the Global Improvement of Cocaine Dependence of the CGI-O (GEE model Z-score=−1.88, p=0.06).

### Cocaine Craving and Withdrawal (CSSA, BCSC, CCQ)

There were no treatment group differences in the total CSSA scores (mixed effects model F(1, 57)=0.75, p=0.34) over the 8 weeks. For the intensity of craving item, there were significant group by linear (GEE model Z-score=−3.16, p=0.002) and group by quadratic (Z-score=3.10, p=0.002) time effects. Each group showed a decrease in scores, and the modafinil group showing a greater initial decline in scores. However, while modafinil group means were consistently lower than those of the placebo group, within-time-point contrasts showed no significant group differences at any point. Similar results were found for the frequency of craving item: there were significant group by linear (Z-score=−2.91, p=0.004) and group by quadratic (Z-score=3.10, p=0.002) time effects, but no significant group differences at any point. There were no significant treatment group differences in the BCSC intensity (Z-score=0.47, p=0.64), frequency (Z-score=−0.35, p=0.73), length of time (Z-score=−0.08, p=0.94), and number of times (Z-score=−1.52, p=0.13) scales or in the CCQ total score (Z-score=−0.76, p=0.45).

### Tolerability

There were no medication-related serious adverse events, and none of the patients discontinued modafinil due to adverse events. There were no clinically significant differences between the two groups with regard to laboratory, vital sign, electrocardiogram, body weight, or physical examination findings. Adverse events occurring in at least 5% of modafinil patients, and with at least twice the incidence of occurrence in placebo patients, included: nausea (23%), upper respiratory symptoms (17%), anxiety (13%), tachycardia (13%), urinary tract infection (10%), dizziness (7%), reduced appetite (7%), racing thoughts (7%), and dry mouth (7%). None of the patients ascribed euphoria or cocaine-like effects to the study medications. Dose reductions due to adverse events were made in six of 30 modafinil-treated patients, from the initial dose of 400 to 300 mg/day (n=2) or 200 mg/day (n=4), and adverse events subsequently resolved in each case (see Table 4). Dose reductions were not made for any of the placebo-treated patients.

## DISCUSSION

The current study found that modafinil-treated patients provided a significantly greater proportion of BE-negative urine samples than placebo-treated patients over the 8-week clinical trial. Also, a significantly greater number of modafinil-treated patients attained the important clinical goal of protracted cocaine abstinence. It is important to note that missing urines were imputed as positive in our analysis. This approach is widely used in cocaine treatment research, based on the assumption that missing urine samples are not ignorable (and more likely to be positive), but it makes clinic attendance an essential factor in the definition of abstinence and may represent a limitation in our study. Both groups received an equivalent number of CBT sessions, further attributing improved outcome to modafinil administration. Although enrolled patients were predominantly African Americans living in an urban setting, other demographic and clinical characteristics are reasonably consistent with the general cocaine-dependent population. Modafinil was well tolerated with a dose titration option, there were no serious adverse events, and there were no medication-associated dropouts.

Even though modafinil is pharmacologically distinct from classic psychostimulants (Lin et al, 1996; Ferraro et al, 1997), it is listed on Schedule IV as having mild abuse potential. Consequently, we carefully monitored the possibility of modafinil overuse and our data showed no discernible evidence of this phenomenon. None of the patients enrolled in this study ascribed cocaine-like effects to their study medication, and blister pack compliance data showed no evidence of modafinil compared to placebo overuse. Both groups had a small and equivalent tendency to retain extra pills that were supplied in their blister packs, but it is unclear whether the pills were ingested, hoarded, lost, or otherwise disposed. Still, since pertinent animal (Gold and Balster, 1996; Deroche-Gamonet et al, 2002) and human (Jasinski, 2000; Rush et al, 2002a; 2002b) studies suggest that modafinil might be weakly reinforcing, overuse should be further assessed in future trials with this population.

None of the secondary efficacy measures attained significance, although physician-rated improvement showed a trend (p=0.06) for the modafinil group. There was no evidence that modafinil reduced cocaine craving, which is notoriously difficult to quantify. There was also no evidence that modafinil reversed cocaine withdrawal, but our study was not designed to assess this possibility and the majority of enrolled patients had insignificant cocaine withdrawal at baseline. Detoxification remains an important research question because patients with cocaine withdrawal symptoms have poor clinical outcome (Kampman et al, 2001), and enhanced euphoric responses to cocaine (Newton et al, 2003; Sofuoglu et al, 2003). Modafinil promotes feelings of well-being (Beusterien et al, 1999) has clinical effects that are largely opposite to cocaine withdrawal symptoms, and blunts cocaine-induced euphoria. It is possible that the steady improvement seen in modafinil-treated patients (see Figure 1) reflects the effect of diminished reward on their cocaine use.

Although we report significant objective improvement (clean urines) with modafinil, our results are limited by the sample size, the subjective nature of craving and withdrawal questionnaires, and the fact that a substantial proportion of our patients had a cocaine-negative sample at baseline. While our assessment of baseline cocaine use might have been different had we collected multiple urine samples during screening, it is also possible that our patients had only moderate addiction severity, and a greater likelihood of responding to placebo treatment. On the other hand, although not statistically significant, the modafinil group had fewer days of cocaine use, fewer cocaine-positive urines, and lower cocaine withdrawal scores (although more years of cocaine use), so baseline differences may have created a disadvantage for the placebo group. Another study weakness is that medication adherence was only confirmed by pill counts. Finally, our subjects received manual-guided CBT, which might not generalize to the clinical setting where this treatment is seldom available. Nevertheless, we did find that modafinil significantly improved cocaine abstinence in this randomized, controlled pilot study, suggesting that further research should be conducted to determine whether modafinil might become a first-line treatment for cocaine dependence.