There were no significant differences (p>0.05) between the M200, M400 and PLAC groups on a battery of measures drawn from the baseline ASI, TLFB, Ham A, Ham D, BDI, SCL-90, and CSSA. There were also no baseline differences when the male patients were analyzed separately. A selection of important baseline demographic, clinical, and psychosocial characteristics of randomized patients are listed in .
Baseline Characteristics as Percentages or Means (Standard Deviations) with p-Values
We obtained pill compliance data on 193 out of 210 patients. The other 17 dropped out of the study after fewer than four visits. There were no significant differences among the M200, M400 and PLAC groups on pill compliance or on the rate of return for extra pills supplied in the blister packs, which is a measure of possible overuse. A total of 147 patients (n=55 PLAC, n=41 M200, n=51 M400) had at least 1 week where they did not return all of the 8 extra pills in their weekly blister packs, which contained a 9-day supply to be used in the event of a missed session. A Chi-square test showed no difference between the groups in the proportion of patients with at least one such week (Chi-square(2)=1.79, p=0.41). An ANOVA test of the number of weeks of extra pill retention showed no differences among the three groups (F(2, 190)=0.38, p=0.69). Across weeks in which expected pills were not returned, the mean 8-week number retained in the PLAC (10.9, SD=11.1), M200 (9.8, SD=9.1) and M400 (10.1, SD=9.4) groups were very similar, and there was no difference found across the treatment groups by ANOVA test (F(2, 190)=0.23, p=0.80). When queried, patients typically stated they could not recall what happened to the missing pills. Three modafinil-treated and 4 placebo-treated patients retained more than 16 of the extra 64 pills supplied over 8 weeks.
Each participant could have had eight CBT sessions during the treatment phase. The mean numbers of sessions attended were 4.4 (SD=2.8) for the placebo group, 5.2 (2.6) for the modaf 200 group, and 5.2 (2.6) for the modaf 400 group. There was a significant difference between the groups (Chi-square(2)=6.85, p=0.03) with the plavcebo group having significantly fewer sessions than either of the two modafinil groups (p=0.02 in each case).
At the end of eight weeks, 40/65 (61.5%) of M200 patients and 43/70 (61.4%) of M400 patients were still retained in the study, compared to only 37/75 (49.3%) in the PLAC group. Alternatively, with dropout defined as failing to complete the first 2 weeks of treatment, there were 8 dropouts in the M200 group, 5 in the M400 group and 11 in the PLAC group (Chi-square(2)=2.10, p=0.35).
Our primary outcome was cocaine use measured by thrice-weekly quantitative urine BE levels and TLFB self-reported use. We followed the guidelines of Preston (Preston et al., 1997
) to create a set of “new use” indicators based on the combined TLFB and BE results, and to determine an initial set of cocaine use indicators for each day of the treatment period. However, we found very poor concordance between the TLFB and BE levels. The minimal recommended concordance of 70% was primarily seen in patients with very little cocaine use, and concordance decreased as BE levels increased. We ran all analyses on the simple BE-based indicators of cocaine use and on the Preston-rule indicators. While there were small numerical differences between the two sets of analyses, the inferences regarding the efficacy of modafinil were the same across the simple and Preston indicators. Accordingly, we opted to ignore the TLFB self reports in our analyses, and instead base all comparisons on the objective BE data.
Missing Data Issues
The design called for 24 urine samples from each of 210 patients, giving a total of 5040 planned BE tests. The actual total number of visits attended was 3318, or about 66% of those planned. The missed visits were split about equally between visits missed after drop out (872) and those missed intermittently (850). As described above, there were no significant differences in rates of treatment retention to the end of 8 weeks and the groups did not differ on other dropout measures. More than half of the modafinil-treated patients were retained through 8 weeks with a median time to the last visit of 24 for each group. The median time to the last visit for the PLAC group was 23 visits. The within group means for the last visit attended were PLAC (M=19.03, SD=7.69), M200 (M=20.23, SD=7.23), and M400 (M=20.37, SD=6.70). A log-rank test showed no significant difference in the distribution of time to drop out (Chi-square=2.96, df=2, p=0.23). The number of missing urines was greater in the PLAC group (M=9.68, SD=7.14, Med=9) than in the M200 (M=7.25, SD=6.95, Med=5) or the M400 (M=7.50, SD=6.71, Med=6.5) groups, but a negative binomial regression model found that the difference was not significant (Chi-square(2) = 3.84, p=0.15).
GEE Comparisons of Abstinence Rates
Since we found poor concordance between patient self-reports and our objective UDS measure of abstinence, which is consistent with our prior experience (Dackis et al., 2005
), we focused exclusively on the UDS results. Our primary analysis of the 24 UDS indicators was based on missing samples being imputed as cocaine positive. This assumption is commonly made in cocaine treatment research because missing samples are not ignorable, given the tendency for active users to miss clinic appointments. Our initial model included binary factors contrasting M200 and M400 with PLAC as well as linear, quadratic, and cubic time trends. We found no significant effects for either modafinil group relative to placebo (M200 vs PLAC: Chi-square=0.12, p=0.73; M400 vs PLAC: Chi-square=0.75, p=0.39). We extended the model to include group by time interactions, as the graphs suggested that the relative ordering of abstinence rates between the three groups varied across time. When these group by time interactions were included in the model, there was some evidence of significant differences among the time courses of the three groups, with significant or nearly significant linear (M200 vs PLAC: Chi-square(1)=7.95, p=0.005; M400 vs PLAC: Chi-square(1)=4.58, p=0.03), quadratic (M200 vs PLAC: Chi-square(1)=5.12, p=0.02; M400 vs PLAC: Chi-square=3.59, p=0.06), and cubic (M200 vs PLAC: Chi-square(1)=3.61, p=0.06; M400 vs PLAC: Chi-square(1)=3.47, p=0.06) time trends.
The differences in time course resulted from initial lower cocaine abstinence in the M200 versus PLAC group and very little subsequent separation among the three groups. In particular, within-time point contrasts showed that the difference between M200 and PLAC was significant at the first and second visits, with the odds for abstinence in M200 being 0.40 times that of the PLAC group (chi-square(1)=8.80, p=0.003) at the first visit, with a corresponding odds ratio of 0.55 at the second visit (chi-square(1)=5.11, p=0.02). Similar contrasts at other time points showed that each of the modafinil groups had slightly higher odds of abstinence than the PLAC group, but the differences were generally not significant with (p-values ranging from 0.07 to 0.80). Only the comparison of M400 with PLAC at visit 24 was significant (OR=2.13, 95% confidence interval (1.04, 4.35), p=0.04). The corresponding GEE and mixed effects analyses with missing visits ignored showed similar patterns, but the differences were smaller and there were fewer significant comparisons. Very similar results were found when both modafinil groups were combined and compared to placebo, and when an 8-week time scale was used rather than a 24-visit time scale.
The Role of Gender
There was a significant gender by M400 interaction (Chi-square=4.28, p=0.04) with respect to UDS outcomes. To investigate this suggestion that modafinil efficacy might vary across gender, we fit models separately to the males and females in the sample. The study randomized a total of 155 males (PLAC=56, M200=47, M400=52) and 55 females (PLAC=19, M200=18 M400=18). We observed a similar pattern to that of the full sample analyses, with the placebo group doing better than the two modafinil groups in the very early visits, followed by the two modafinil groups doing better than placebo for the rest of the treatment phase. There was some evidence that high dose modafinil was associated with higher rates of abstinence among males. For a main effects model, the estimated odds ratio for abstinence in the M400 group versus PLAC was 1.77, with 95% confidence interval (0.99, 3.19) and a p-value of 0.06. The corresponding odds ratio for the M200 group was 1.32, with 95% confidence interval (0.76, 2.29) and a p-value of 0.32, so there is little evidence in favor of the M200 group. shows the plot of abstinence rates for the male patients. An entirely different pattern was observed for the females (see ). Group by time effects did not approach significance and a main effects model suggested that the PLAC group had the highest rates of abstinence and the M400 group had the lowest rates of abstinence. However, the overall effect was not significant (Chi-square(2)=2.28, p=0.32). The corresponding GEE and mixed effects models when missing UDS responses were ignored showed no significant effects.
Cocaine Abstinence Rates in the Male Patients
Cocaine Abstinence Rates in the Female Patients
Since we found these gender specific results for the primary UDS outcome, we conducted separate analyses for males and females on the secondary outcomes during treatment and at the follow-up time points. There were no significant effects for males or females during treatment on the CSSA, CGI-O, CGI-S or BSCS. With regard to follow-up time points, there were stronger effects on the UDS outcomes in the males than in the overall sample. At visit 21, 48% (23/48) of the M400 patients but only 16% (5/32) of the PLAC patients had cocaine-free urines. In addition, the M400 group had higher rates of abstinence than the PLAC group at each follow up point; week 10 (OR=2.23, 95% confidence interval (0.91, 5.48), p=0.08); week 13 (OR=2.91, 95% confidence interval (1.13, 7.51), p=0.03); week 21 (OR=3.75, 95% confidence interval (1.47, 9.54), p=0.01). The differences between M200 and PLAC did not approach significance at any of the follow up points.
The rates of achieving three-weeks of continuous abstinence, with missed visits regarded as use, were not significantly different across the PLAC (30.7%), M200 (35.4%) and M400 (32.9%) groups (Chi-square=0.35, df=2, p=0.84).
Patient-Reported Cocaine Severity (CGI-S)
Clinical Global Improvement Scale-Self (CGI-S) ratings showed no differences between the M200, M400 and PLAC groups in reported severity of cocaine dependence or associated functional impairment (GEE model M200 Z-score=1.42, p=0.16, M400 Z-score=0.77, p=0.44). There were no differences when male patients were analyzed separately.
Physician-Rated Assessments (CGI-O)
The study physician rated the Clinical Global Improvement Scale-Observer (CGI-O) summary scales each week. For the Global Severity of Cocaine Dependence Scale there were no treatment group differences found (M200 GEE model Z-score=1.41, p=0.16, M400 Z-score=−0.25, p=0.80). For the Global Improvement of Cocaine Dependence of the CGI-O, there were significant group by linear time effects (GEE model M200 Z-score=−2.58, p=0.01, M400 Z-score=−2.10, p=0.04) but with the PLAC group showing greater clinical global improvement than the M200 group. When males were analyzed separately, the PLAC group still had higher scores on clinical improvement (beta=0.3812, p=0.01, 95% confidence interval (0.1043, 0.6581)) but the M400 group had a trend for lower severity than placebo at week 21 (beta=0.3030, p=0.07, 95% confidence interval (−0.0285, 0.6346)).
Cocaine Craving and Withdrawal (CSSA, BCSC)
There were no treatment group differences in the total Cocaine Selective Severity Assessment (CSSA) scores (M200 chi-square(1)=0.18, p=0.68; M400 chi-square(1)=0.01, p=0.94). These statistics were based on GEE models of log transformed CSSA total scores over the 8 weeks of medication in the study. Similar analyses for the “Intensity of Craving” item (M200 chi-square(1)=0.46, p=0.50; M400 chi-square(1)=1.93, p=0.16) and the “Frequency of Craving” item (M200 chi-square(1)=0.01, p=0.91; M400 chi-square(1)=0.96, p=0.33) also showed no significant group effects.
There were no significant treatment group differences in the BSCS Intensity (M200 chi-square(1)=0.39, p=0.53; M400 chi-square(1)=1.32, p=0.25), Frequency (M200 chi-square(1)=0.03, p=0.85; M400 chi-square(1)=0.66, p=0.42), Craving Duration (M200 chi-square(1)=0.69, p=0.41; M400 chi-square(1)=0.56, p=0.45) or Craving Frequency scales (M200 chi-square(1)=0.15, p=0.70; M400 chi-square(1)=0.79, p=0.37) based on log transformed number of times.
We found the same pattern of results when these CSSA and BSCS analyses were repeated within the male patients.
Abstinence in Final Three Weeks of Treatment
The following table shows the rates of complete abstinence in the final three weeks of the study. Patients were regarded as abstinent if they attended each of their last nine scheduled visits and provided cocaine-negative urines on each occasion. The overall test for the full table was not significant (Chi-square(2)=3.85, p=0.15) but the M400 group showed a trend towards higher rates of complete abstinence (OR=3.10, p=0.06).
End of Study Questionnaire
A questionnaire was added to our study at about midpoint to further delineate the effect of modafinil on craving and cocaine-induced euphoria. A total of 91 patients responded to this questionnaire and we analyzed the 88 who used cocaine during the trial. Patients were asked to rate (1) cocaine-induced euphoria, (2) cue-induced craving, (3) cocaine-induced craving and (4) general craving during their 8-week modafinil trial, with each item rated as “absent,” “less than usual,” “usual,” or “more than usual.” We combined “absent” with “less than usual” responses, as well as “usual” with “greater than usual” responses for our analysis. With regard to (1) cocaine-induced euphoria, the difference between modafinil (M200 and M400) compared to PLAC approached statistical significance (chi-square(1)=3.35, p=0.07). For each of the craving measures, the combined modafinil group had higher rates of “absent” or “less than usual” craving, but the differences were not significant for (2) cue-induced craving (chi-square(1)=1.60,, p=0.21), general craving (chi-square(1)=2.65, p=0.10) or cocaine-induced craving (chi-square(1)=2.74, p=0.10).
Follow-Up Time Points
Self-reported cocaine abstinence rates at follow-up (weeks 10, 13, and 21) did not differ across the three treatment groups. UDS abstinence was similar for weeks 10 and 13, but there were differences at week 21 in the percent of cocaine-free urines for the PLAC (16%), M200 (15%) and M400 (33%) groups. An overall test of association was significant (Chi-square(2)=8.15, , p=0.02), and the odds ratio for abstinence in the M400 group relative to the PLAC group at week 21 was 2.57 (p=0.02).
Within males, there were stronger effects. At visit 21, 48% (23/48) of the M400 patients but only 16 % (5/32) of the PLAC patients had cocaine-free urines. With a missing UDS imputed as positive, the M400 group had higher rates of abstinence than the PLAC group at each follow up point; week 10 (OR=2.23, 95% confidence interval (0.91, 5.48), p=0.08); week 13 (OR=2.91, 95% confidence interval (1.13, 7.51), p=0.03); week 21 (OR=3.75, 95% confidence interval (1.47, 9.54), p=0.01). The differences between M200 and PLAC did not approach significance at any of the follow up points.
Depression and Anxiety symptoms
The Beck depression inventory was obtained at weeks 0 through 8 and weeks 10, 13, and 21, while the Hamilton anxiety and depression scales were obtained at weeks 0, 10, 13, and 21. Scores on the total scores for all three scales decreases from baseline (week 0) through the treatment phase (to week 10), with little change between weeks 10 and 21. There were no significant differences in time course across the groups, and the main effects of group were not significant: BDI in treatment phase – F(2, 198)=0.07, p=0.93, BDI in follow-up phase – F(2, 154) = 0.04, p=0.96; HAM-A in follow-up phase – F(2,151) = 1.67, p=0.19; HAM-D in follow-up phase – F(2,137) = 0.71, p=0.50. We note that there were 161 subjects available for the BDI and 164 subjects available for the Hamilton scales for the follow-up analyses. The variation in degrees of freedom reflects the use of the Kenward-Rogers option for denominator degrees of freedom in the test statistics. There were no gender by group differences on any of these outcomes.
There were no clinically significant differences between the two groups with regard to laboratory, vital sign, electrocardiogram, body weight, or physical examination findings. There was only one serious adverse events thought to be related to study medications. This involved treatment emergent mania in a patient who was using large amounts of cocaine and required psychiatric hospitalization. He was diagnosed with bipolar affective disorder and discontinued from the study medications. Adverse events occurring in at least 5% of modafinil patients, and with at least twice the incidence of occurrence in placebo patients, included upper respiratory symptoms (25%), headaches (23%), insomnia (11%), weight loss (9%) and nausea (9%). None of the patients ascribed euphoria or cocaine-like effects related to the study medications.