The participants in this study were 111 opioid-dependent individuals who were under legal supervision in the community. All subjects were enrolled in drug treatment for at least six months at either the intensive outpatient program (IOP) offered by the research study or were treated at one of several community-based IOPs.
Subjects were eligible for participation if they: (1) signed an informed consent form agreeing to randomization to one of the two treatment groups; (2) were between the ages of 18 and 55; (3) had a diagnosis of opioid dependence based on DSM-IV criteria and a structured psychiatric interview; (4) were in good general health as determined by a complete physical examination and laboratory tests; (5) had been assigned to probation/parole for a minimum of six months; and (6) had a negative result for urinary opioids and reported being at least 3 days opioid-free prior to randomization.
Subjects were excluded if they had: (1) current severe alcohol dependence that required medical supervision for alcohol withdrawal symptoms; (2) current psychosis, dementia, mental retardation, or history of schizophrenia; (3) clinically significant abnormalities in hematology, chemistry, or urinalysis; (4) clinically significant cardiovascular, neurological, hepatic, renal, pulmonary, metabolic, endocrine, or gastrointestinal disorders; (5) a diagnosis of chronic pain disorder; or (6) taken an opioid antagonist within the prior 6 months. Female subjects who were pregnant or lactating, or women of childbearing potential who were not using birth control were also excluded.
Participants were recruited from various sources. Referrals were made by county and federal probation/parole officers, a drug treatment court, the public defenders office, community-based IOPs and two inpatient programs. Referrals were also obtained from an alternative disposition program that offered early parole, if participants agreed to attend a mandated substance abuse treatment program, and an intermediate punishment program (IPP) that provided offenders the option of attending mandated treatment in lieu of incarceration.
The project was approved by the Institution Review Boards of the University of Pennsylvania and the City of Philadelphia. The research was also approved by an administrative board consisting of the Chief Federal District Court Judge in Philadelphia and by a research review committee of one of the inpatient hospital programs. Interested participants were given a full description of the study by research staff. Special efforts were made to assure offenders that participation in the study was voluntary. Individuals were instructed that the research study was an additional service they could receive and that choosing to participate or not participate would have no effect on their probation/parole status and would not affect the duration of their probation/parole. Moreover, they could stop participation in this study at any time without any effect on their probation/parole, treatment or services. Subjects who agreed to participate in the study signed an informed consent document at the point of entry into the trial.
Enrolled subjects completed a two to three day screening process to determine eligibility into the study. Physical exams and laboratory tests were assessed to ensure that each subject was in good general health and had normal hepatic function. The Structured Clinical Interview for DSM-IV Axis-I (SCID)27
was administered to ensure each subject had a diagnosis of opioid dependence and to rule out any severe psychiatric disorders. During the screening process participants also completed a series of baseline research assessments. In order to enter the study, all subjects had to self-report that they had not used opioids for three days and had to provide a negative urine result for opioids immediately prior to randomization. Some participants who were unable to provide an opioid-free urine were referred to inpatient detoxification and returned to the study after having successfully completed detoxification. Candidates who passed these evaluations and provided a negative urine result, successfully passed this screening process and were randomly assigned to one of two treatment conditions (naltrexone + TAU or TAU only).
Randomization was balanced using six prognostic variables: gender, current marital status (yes/no), comorbid current alcohol abuse or dependence, comorbid current cocaine abuse or dependence, previous arrests and criminal charges (≤ 5 vs. >5), and previous drug treatments other than self-help groups and detoxification only (≤ 3 vs. > 3). The point of randomization of each subject was the actual starting point for the study. All subsequent scheduled events were calculated from the point of randomization. Subjects were assessed at baseline, twice weekly during the six-month treatment phase and then at six months post-treatment entry.
Participants in both groups were administered a standard side effects questionnaire that assessed any adverse events. Retention in the study was determined by the number of weeks participants remained in compliance with the protocol. Subjects who missed three consecutive weeks were dropped from the study.
Psychosocial Treatment Only (TAU Group)
Participants in the TAU group received six months of psychosocial treatment at one of several community-based treatment programs or at the university-based IOP provided by the research study. The university-based psychosocial treatment consisted of three hours of group therapy, one hour of individual therapy, and one hour of case management for six weeks of intensive outpatient treatment followed by 20 weeks of outpatient treatment (OP) consisting of one hour of individual and one hour of case management per week. The psychosocial therapy provided by the community-based programs was similar in content, but typically included additional hours of group therapy during the IOP phase.
Subjects in the TAU group did not receive a placebo. The rationale for this decision was based on our clinical experience with naltrexone and with placebo-controlled studies involving antagonist medications. It has been our experience that subjects on naltrexone will test the efficacy of the naltrexone one or more times during the early phase of treatment by using opioids and thus break the blind.17,18,28
Moreover, unlike placebo-controlled trials of other medications such as anti-depressants, the subject is easily able to determine if the study medication is active or inactive simply by using an opioid. We did not want subjects to test the study medication and increase the possibility of re-addiction for those receiving placebo. In addition, permission to study a criminal justice population is a very sensitive issue and the court system made it clear that the use of a placebo would be unacceptable. Therefore, for both ethical and practical reasons, our TAU group was not blinded and consisted of psychosocial treatment without additional medication. A total of 55 subjects were assigned to the TAU condition.
Oral Naltrexone Plus Psychosocial Treatment
Subjects randomized to naltrexone received a challenge test consisting of 0.8 mg of naloxone administered intravenously or intramuscularly followed by a 20-mininute observation period during which the subject was evaluated for signs and symptoms of opioid withdrawal. Subjects who failed the naloxone challenge test were re-evaluated to determine if they required additional days of abstinence prior to re-challenge or were referred to a substance abuse detoxification program. Relatively few subjects failed the challenge since all participants had to be opioid-free prior to randomization. Subjects who successfully passed the naloxone challenge test were started immediately on naltrexone. The initial naltrexone dose was 25 mg. During the first week, subjects returned for two more visits, and on the second visit the dose was increased to 50 mg and on the third visit the dose was 100 mg. Beginning in the second week, subjects receive 150 mg of naltrexone twice a week (300 mg per week) for a total of 26 weeks or six months. The dispensing of all medication was directly observed. Participants in the naltrexone group received the same six months of psychosocial treatment as the TAU condition. Fifty-six participants received oral naltrexone plus TAU.
The instruments listed below were administered in order to obtain measurements of substance abuse, comorbid psychiatric disorders, psychosocial functioning, human immunovirus (HIV) risk behaviors, depression, antisociality, and criminal behavior. Both groups received the same schedule of assessments and were compensated for completing research paperwork at each evaluation time point.
The Addiction Severity Index (ASI) was administered by a trained technician at baseline to assess lifetime and recent (past 30 days) functioning in seven potential problem areas: medical, employment/economic, drug use, alcohol use, legal, family/social and psychiatric.29-31
Composite scores (CS) computed in each of the seven areas provided an indication of overall problem severity. The ASI also yields relevant sociodemographic information. The abbreviated, follow-up version of the ASI was administered six months after treatment entry.
The Structured Clinical Interview for DSM-IV (SCID) was administered during screening to exclude subjects with severe psychiatric disorders and to verify an opioid substance dependence diagnosis.27,32
The Beck Depression Inventory (BDI) was administered at baseline to assess level of depression and measure potential effects of depression on treatment outcome.33
The California Psychological Inventory-Socialization Scale (CPI-So) is a self-report inventory that was administered as an independent instrument (i.e., unembedded in the larger CPI) at the baseline assessment.34-38
The socialization scale is a measure of socialization, social judgment, and normative behaviors during childhood and adolescence that yields one summary measure of a disposition to antisociality. Each of the 46 items is rated true or false (range 0-46) with lower scores reflecting poorer social judgment, less empathy, and less conformance with social norms. A score of 22 and below was considered diagnostic of severe problems in rule-following and norm-accepting behavior.39
The Risk Assessment Battery (RAB) is a self-report measure that was used at baseline and six months to assess both sex and drug risk factors associated with HIV acquisition and transmission such as needle sharing and unsafe sexual practices.40-41
With subject’s informed consent, criminal records data were obtained from the City of Philadelphia and the Pennsylvania Commission on Crime and Delinquency. These data included the total number and type of offenses a participant was charged with prior to and after randomization. Self-report measures of criminal activities from the legal section of the ASI that included months incarcerated in lifetime, days of illegal activity in the last month, and parole/probation status were also obtained.
Monitored urine drug screens were collected by research staff and were analyzed using the Enzyme Multiplied Immunoassay Technique (EMIT) system in our urine toxicology lab. Five drugs were tested: opioids, cocaine, methadone, benzodiazepines, and cannabis. Urine drug screens were collected at baseline, twice weekly during the six-month treatment phase (to coincide with the naltrexone dosing schedule) and at the six-month follow-up.
Subjects who were assigned to the medication group received their naltrexone dose at the research offices at the University of Pennsylvania. Psychosocial treatment for both the TAU and naltrexone groups were provided at either the university research office for those who received treatment as part of the research program, or at the community-based clinic that they were attending. All urine specimens and research data were collected at the research offices at the University of Pennsylvania.
Baseline comparisons between the randomization groups used t-tests for continuous measures and chi-square tests for binary measures. Logistic regression models were used to predict treatment completion from baseline characteristics. Cox Proportional Hazards regression models were used to predict time to dropout from treatment.
Our primary comparisons were on the rates of opioid use across the two groups. Rates of opioid positive urines were compared across the two groups using Generalized Estimating Equations (GEE) methods for repeated binary outcomes. The GEE approach to comparing two sets of repeated measures assumes a ‘working’ correlation matrix for the within subject correlations across time, and yields regression coefficients that are valid even if that assumption is incorrect. Thus, the methods are robust to misspecification of the within subject correlations. The primary factor in these GEE models was treatment group assignment, and the model also included time trends, and some group by time interactions.
There was a great deal of missing data in this trial. While the GEE models are insensitive to model misspecification of the within subject correlation, they may be sensitive to drop out that is related to treatment outcomes. As a result, the conclusions of the GEE models must be very sensitive to what is assumed about the missing data. To determine sensitivity of the results to the high rates of missing data, we performed three versions of the GEE analyses:
First, we assumed that the missing data are ignorable, which in the context of our GEE models means that we assumed that they were missing completely at random. This is a stronger assumption than simply assuming that the missing data can be predicted from observed variables and prior outcomes, which are referred to as missing at random. In our initial analyses, we found no significant predictors of completion or of time to drop out (except for age) or first use, so the distinction between the two types of missing at random is of no practical importance for this dataset.
Our second analysis imputed missing urine drug screen responses as being positive for opioid use. This is a common strategy in analyzing data from substance abuse studies, and can be regarded as defining a new outcome variable: a treatment visit is defined to be successful if a patient makes the visit and provides an opioid negative urine, otherwise the visit is regarded as a failure. Thus, our second analysis can be regarded as a GEE analysis of the repeated binary outcomes indicating successful visit, where our definition of response yields complete data for all subjects.
Our third strategy used a pattern mixture approach42
and extended the first analysis approach by including variables describing the dropout process as main effects and as interactions with treatment group in the model. A significant interaction between treatment and these variables would indicate that treatment efficacy was different at different levels of dropout, which would suggest that the missing visits are non-ignorable, and would yield different inference for completers compared to non-completers.