Pharmacotherapy Plus Psychotherapy for Treatment of Depression in ActiveInjection Drug Users FREE

Depressive disorders are common among opiate abusers, with rates rangingfrom 14% to 55%.^1- 3 Asignificant challenge in the study of depression in opiate users is that opiatesmay induce symptoms that are difficult to distinguish from symptoms of primarymood disorders. Depressive symptoms may be attributable to an opiate's toxiceffects, drug withdrawal, or short-term life crises. Although there have beenrecent attempts to develop diagnostic criteria and interview instruments thatdifferentiate opiate-related symptoms from independent mood disorders includingmajor depression, there is general agreement that the task of establishinga primary-secondary distinction is especially difficult among patients withunrelenting, chronic, and severe substance abuse disorders.^4- 7

The impact of depression on opiate users is extensive. Comorbid depressionmay be associated with premature dropout from treatment, and it has been associatedwith poorer prognosis after drug treatment.^2,8- 11 Comorbiddepression also may serve as an important trigger for high-risk injectionpractices, continued drug use, or drug relapse.^12- 15 Situationsinvolving negative mood states are among the most frequently cited precipitantsof relapse across several types of addictive substances.¹³ Substanceabuse by psychiatric patients may be motivated by efforts to alleviate symptomsof the primary psychiatric disorder; conversely, mood disorders may be inducedby substance use.

Despite these psychiatric treatment needs, treatment of depression amongactive drug users who are not enrolled in drug treatment has never been clinicallytested. This may be in part because active drug users do not seek treatmentfor depression, and because of the belief that a diagnosis of major depressioncannot be reliably made in the context of daily drug use. It may also be dueto practical considerations. Drug injectors often lead significantly disruptedlives, making adherence to treatment difficult, which in turn may limit theeffectiveness of depression treatment.

Among opiate abusers receiving methadone maintenance treatment, however,4 of 5 pharmacologic trials have demonstrated favorable effects on mood.^16- 20 Nonetheless,there is little precedent for offering active drug users treatment for depressionwith pharmacotherapy. Drug injectors enrolled in methadone maintenance treatmenthave received psychotherapy for depression, such as cognitive behavior therapy(CBT), to prevent relapse to opiate use.²¹ Someresearch suggests that CBT may be a viable option for out-of-treatment drugusers, despite the difficulty with adherence to treatment.²² Callahanet al²³ found improved retention and treatmentoutcome in those who participated in a combination psychological and pharmacologicintervention, as compared with heroin users receiving pharmacotherapy alone.

In the Multidisciplinary Intervention for Needle Users to Reduce ViralAcquisition (MINERVA) study presented herein, we recruited active injectiondrug users for a randomized study testing the efficacy of combined CBT andpharmacotherapy for the treatment of depression. In MINERVA, we took the positionthat requiring abstinence before diagnosing depression limits therapeuticoptions, that primary-secondary depression distinctions are often difficultto discern, and that the cause of depressive symptoms (drug-induced or endogenous)may not be critical given that comorbid depression confers a poor long-termprognosis in substance abusers, regardless of the primary-secondary distinction.When antidepressant treatment is considered, combining pharmacotherapy andpsychotherapy may be the most powerful intervention, particularly for thosewith more severe and/or chronic depression.²⁴ Ourgoal was to determine whether combined treatment would reduce reported depressivesymptom scores compared with an assessment-only condition among out-of-treatmentdrug injectors.

Between March 1, 2000, and June 30, 2003, we recruited participantsin Rhode Island for a randomized study of combined psychotherapy and pharmacotherapyfor the treatment of depression in out-of-treatment drug injectors. We advertisedthe study as a "quality-of-life" research project with a financial incentiveat various community agencies, through newspaper advertisements and on thestreet with flyers. Those interested were directed to call the study telephonenumber to be screened by study research assistants, who inquired about mostof the inclusion criteria. During the recruitment period, the study telephonereceived 1518 calls. If eligible after a telephone screening that includeddemographics, recent drug use and needle-sharing behavior, and use of prescribedpsychoropic medications questions, individuals were asked to come to the researchsite at Rhode Island Hospital, Providence, for a more detailed assessment.

Inclusion criteria included the following: (1) age between 18 and 70years; (2) injection of opiate or cocaine during the preceding 30 days; (3)injection-related risk behavior (ie, needle sharing) during the preceding30 days; (4) meeting criteria for a DSM-IV diagnosis(Structured Clinical Interview for DSM-IV–PatientVersion [SCID-P])²⁵ of either (a) major depression, (b) dysthymia, (c) substance-induced mood disorder with symptoms persistingat least 3 months, or (d) major depression plus dysthymia;(5) having a score on the Modified Hamilton Rating Scale for Depression (HAM-D)²⁶ greater than or equal to 14; (6) not currently enrolledin drug treatment (methadone, residential, outpatient); (7) absence of currentor past diagnosis of bipolar disorder, schizophrenia, schizoaffective disorder,schizophreniform disorder, or delusional disorder; (8) able to speak and readEnglish sufficiently to complete the procedures of the study; (9) having averifiable mailing address; (10) providing contact names for follow-up assessments;and (11) providing written informed consent for this study, which was approvedby the Rhode Island Hospital/Lifespan Institutional Review Board.

At the baseline visit, participants underwent urine toxicologic testingfor cocaine metabolites and heroin to confirm active drug use status, andwomen received urine pregnancy testing. Participants were then administereda 90-minute structured interview, which included sections on demographics;SCID-P mood, drug, and alcohol modules; and human immunodeficiency virus (HIV)risk behaviors. At the conclusion of the interview, participants were escortedto the blood laboratory for HIV antibody testing. Refusal to undergo HIV testingdid not exclude eligible persons from participation in the study.

On the basis of the telephone screen, 161 individuals were invited andcame to the study office for a full screening interview; 52 did not meet fulleligibility criteria. One hundred nine persons were eligible for the studyand enrolled, with 53 assigned to the combined therapy group and 56 to theassessment-only control group. Participants assigned to the control groupwere given a 2-week follow-up appointment to receive the results of theirHIV tests (including posttest counseling) and 3-, 6-, and 9-month follow-upappointments for face-to-face research assessments. At each assessment visit,subjects also received any medical or mental health referrals requested.

Participants assigned to the intervention group were scheduled to receivea total of 8 individual psychotherapy (CBT) visits (during the course of 3months) and 3 pharmacotherapy visits (monthly for 3 months). After this acute,combined 3-month phase, subjects were offered continuation pharmacotherapyalone for study months 3 to 6. Treatment group participants were scheduledfor their first psychotherapy appointment and medication appointment withina week of the baseline assessment. Participants in the treatment group alsoreturned to Rhode Island Hospital for follow-up assessment interviews at 3,6, and 9 months after baseline.

To maximize study retention, the protocol strongly emphasized the importanceof follow-up, and, at each interview, research assistants obtained contactinformation for at least 2 other people who could be contacted to help locatethe participant. At each psychotherapy visit, participants in the treatmentgroup had brief contact with research assistants to receive their financialcompensation of $15. Pharmacotherapy visits were not financially compensated,as they typically co-occurred with a psychotherapy or assessment visit, whichwas financially compensated. Participants in the intervention group couldreceive up to $240 ($120 for assessment and $120 for attending all psychotherapyappointments). Remuneration at assessments was not dependent on attendanceat therapy sessions.

Although we scheduled all therapy appointments at the time of a visit,we maintained a flexible schedule for persons missing visits. Study staffwere available by mobile telephone and toll-free numbers, which allowed participantsto call and to change appointments. Research assistants called study participantson the day before and the morning of all scheduled therapy appointments. Whena participant failed to attend a session, a call to reschedule the appointmentwas made. Treating clinicians were accommodating in attempts to arrange thefull complement of CBT and medication visits within the first 3 study months.In addition, we provided transportation via cabs or bus tokens for all studyvisits.

This treatment was aimed at reducing the participant's level of depressionand increasing his or her coping skills regarding depression. This 8-sessiontherapy was a modified version of a "Coping With Depression" protocol treatmentintervention used by Brown et al²⁷ for treatmentof depression in alcoholic patients. The first sessions were devoted to presentinga social learning view of depression and guiding the participant in learninghow to identify and differentiate mood states. The next sessions addressedacquiring skills in 3 specific areas: increasing pleasant activities, changingnegative cognitions, and improving social skills and increasing positive socialinteractions.²⁸ Each participant was encouragedto develop a personalized plan to work on problem areas by means of a participantworkbook. Each session consisted of a review of material from the previoussession, presentation of new material, discussion and exercises related tothe new material, and a homework assignment. The treatment was administeredin 60-minute individual sessions by PhD-level clinical psychologists trainedand supervised in the protocol by 2 of us (D.S.H. and S.E.R.) at weekly trainings.

Participants were scheduled to meet with a psychiatrist monthly for3 months. Whenever possible, psychiatric visits took place immediately afterCBT sessions. Psychiatrists followed the clinical management guidelines developedby Fawcett and colleagues.²⁹ Participants werefirst prescribed citalopram, 20 mg. The dose and dosage schedule were determinedby the clinical judgment of the treating psychiatrist (up to 60 mg). If citalopramwas not effective or produced adverse effects, the psychiatrist could prescribevenlafaxine hydrochloride, 37.5 mg (up to 375 mg), or bupropion hydrochloride,150 mg (up to 300 mg). Once a satisfactory response was achieved with anyof the aforementioned medications, participants continued to receive thatdose for the remainder of the acute combined treatment period, that is, upto 3 months. At each medication visit, subjects received 1 month's worth ofstudy medication. Medication study visits were limited to 15 minutes to ensurethat the study psychiatrist was not providing other active psychotherapy ingredientsto the treatment group. Eleven participants were switched from citalopramto venlafaxine and 2 were switched from citalopram to bupropion during thestudy.

Study medication use was confirmed by pill counts, self-report, and,at the 3-month visit, serum levels (citalopram only) for those who reportedcontinued use of study medication. All participants who reported citalopramuse at the 3-month assessment had serum levels verifying self-report.

The training for the 2 PhD-level clinical psychologists began with astudy of the project CBT manual. They also received 2 hours of didactic trainingin HIV risk reduction, injection behaviors, and local substance abuse treatmentprograms. Therapist manuals developed for the project were used in both initialtraining and continuing supervision. Each therapist then performed the interventionwith 3 pilot training cases before study initiation. Adherence to the interventionprotocol was monitored by having the therapist complete a checklist aftereach intervention session.³⁰ All interventionsessions were audiotaped, and 10% were randomly reviewed by the therapy supervisor(S.E.R.) for internal quality control to prevent deviation from the protocol.

Adherence to the treatment protocol was assessed as the number of CBTappointments the participant attended and the percentage of days on whichprescribed medications were taken during the first 3 study months (acute phase).We also constructed a dichotomous indicator, "fully adherent dose," defineda priori by consensus of the study investigators as attendance at 6 or moreof the 8 scheduled CBT sessions or adherence to the pharmacotherapy regimenon 75% or more of the days on which medications were prescribed.

We used the SCID-P to diagnose depression type (major depression, dysthymia,"substance-induced mood disorder" with symptoms persisting for at least thepast 3 months, or major depression plus dysthymia).²⁶ Frequencyof heroin use was assessed as the reported number of days on which heroinwas used in the past 30 days according to the Addiction Severity Index.³¹

The modified HAM-D scale was used to assess depression severity at bothbaseline and follow-up.²⁶ Treatment remissionwas defined as a 3-month HAM-D score of 8 or less.³² Asa secondary outcome, we defined treatment responders as those with a 50% orgreater improvement relative to baseline HAM-D scores.³³ Wetested as outcome measures both the continuous 3-month HAM-D score and thedichotomous indicators assessing remission and relative reductions of 50%or greater.

We present an "intent-to-treat" analysis based on 3-month assessmentdata from the end of combined treatment. There were 94 subjects for whom 3-monthfollow-up data were available, and our primary analyses used multiple imputationof missing data.^34- 35 We usedProc MI as implemented in SAS release 8.2 to generate 20 imputed data setsunder the multivariate normal model.³⁶ Variablesincluded in the imputation process were sex, age, race, educational attainment,baseline HAM-D score, "full" treatment adherence, and 3-month HAM-D score.We used the companion procedure Proc MIAnalyze to generate parameter estimatesand corrected SEs.³⁶ The imputation proceduresassumed that data were missing at random, implying that the probability ofbeing missing at follow-up was unrelated to depression severity at follow-up.Since it was possible that participants whose depression severity increasedmight be more likely to be missing at follow-up, we conducted a sensitivityanalysis in which imputed values were increased (or decreased, depending onthe substantive problem) by some constant.³⁵ Thus,we also report on analyses using "worst-case scenario" substitution in whichthe missing observations were assigned the highest observed 3-month HAM-Dscore (score, 34).

We used analysis of covariance to estimate treatment differences on3-month HAM-D scores adjusting for baseline depression severity. For conveniencein obtaining parameter estimates and adjusted SEs by means of Proc MIAnalyze,we used regression with treatment coded as a dummy variable (1, treatment;0, control) to conduct the analyses of covariance.

We present contingency tables to describe treatment differences on depressionremission. In each imputed data set, the cell frequencies are integers, butthe average of imputed cell frequencies are not so constrained. We used logisticregression in conjunction with Proc MIAnalyze to estimate logit coefficientsand corrected SEs. It should also be noted that the test statistic for parametersestimated by means of multiple imputation is reported as a t statistic in which the parameter estimate is divided by the correctedestimate of SE. To facilitate description, we report effects on odds ratherthan the raw logit coefficients.

Participants averaged 36.7 years of age and 11.5 years of education(Table 1). A majority (64.2%)were male, and 82% were white. Fifty-one percent had been unemployed for the3 months before study enrollment, 8% were married, and 8% were HIV positive.Injectors were primarily heroin users, with only 10% using cocaine more frequentlythan heroin and 2 persons not using any heroin in the past 30 days. The mean(SD) number of heroin use days in the preceding month was 21.8 (9.8), and91% met criteria for current opioid dependence. Depression types includedmajor depression only (63%), major depression and dysthymia (17%), substance-induceddepression (17%), and dysthymia only (2%). The average baseline score on theHAM-D was 20.7 (3.9). Fifteen participants (13.8%) were lost to follow-upat 3 months. The treatment groups did not significantly differ with respectto intake characteristics, heroin or cocaine use or dependence, baseline levelsof depression, or depression type. In addition, the proportion of subjectslost to follow-up was highly consistent across treatment conditions (Table 1). Two participants in the treatmentgroup had emergency department visits for suicidal ideation within the firststudy months; both halted the intervention. There were no adverse medicationeffects that led to treatment cessation.

To determine whether participants with more severe depression at baselinewere more likely to be lost to follow-up, we compared baseline HAM-D scoresfor subjects observed at 3 months with those of subjects lost to follow-up.The mean ± SD baseline HAM-D score for subjects observed at 3 months(20.71 ± 3.77) was trivially higher (t = .41, P = .68) than the mean for the 15 subjects lost to follow-up(20.26 ± 3.95).

We assessed both the number of CBT appointments that participants attendedand their daily use of prescribed depression medications. Only 26 (49%) ofthe 53 participants assigned to treatment attended 50% or more of their CBTappointments and only 20 (38%) attended 75% or more. Participants were evenless likely to adhere to their daily pharmacotherapy regimens. Only 18 (34%)took antidepressant medications on 50% or more of the days on which they wereprescribed, and 16 (30%) had pharmacotherapy adherence rates of 75% or greater.Participants who adhered to CBT tended to adhere to their prescribed pharmacotherapyregimens and vice versa. Of the 20 participants who attended 75% or more oftheir CBT appointments, 13 (65%) used antidepressant medications on 75% ormore of the days on which it was prescribed. Twenty-three (43%) of the 53subjects assigned to the combined treatment group were fully adherent to treatment.

As measured by mean HAM-D scores, significant improvements in depressionat 3 months were observed in both arms of the study. Among controls, mean3-month HAM-D scores were 3.48 points lower than at baseline (t = 3.48, P = .001). For participants assignedto combined treatment, HAM-D scores at 3 months averaged 6.04 points (t = 5.24, P<.001) lower thanat baseline. Eighteen (19%) of the 94 subjects with valid 3-month data werein remission (HAM-D score ≤8) at follow-up and 21 (22%) reported relativereductions of 50% or greater in their HAM-D scores.

When outcomes were compared between the treatment and control groupson 3-month HAM-D means adjusted for baseline HAM-D scores, the observed meandifferences were directionally consistent with the hypothesized treatmenteffects, but were not statistically significant at conventionally acceptedlevels. When we used the subjects for whom complete data were observed, participantsin treatment averaged about 2.11 points lower than controls on 3-month HAM-Dscores (β = −2.11, t₉₁ = −1.42, P_α/2= .08). The standardized effect sizefor the completely observed data was 0.29, which could be described as a smallbut meaningful effect.³⁷ The estimated meandifference using multiple imputation (β = −2.17, t = −1.46, P_α/2= .07)was similar in both magnitude and statistical reliability. Not surprisingly,using worst-case substitution attenuated the estimated differences in treatmentmeans (β = −1.90, t₁₀₆ = −1.07, P_α/2= .14).

A more robust pattern of treatment effects was observed when rates ofremission were compared (Table 2).Subjects randomized to combined treatment were roughly 2.5 times more likelythan controls to be in remission at the 3-month follow-up. The odds ratio(OR) obtained by means of complete data (OR, 2.57; P_α/2= .047) was of similar magnitude to those obtained by multipleimputation (OR, 2.47; P_α/2= .049)and worst-case substitution (OR, 2.44; P_α/2= .047) (Table 2). We replicatedthis analysis by using the indicator of treatment response (not presentedin Table 2). The results werecompletely consistent statistically, and the magnitude of observed effectswas very similar to those reported for remission. The treatment response ratewas 30.4% in the combined treatment group and 14.6% in the control group.

We used logistic regression models to estimate treatment effects, controllingfirst for basic background characteristics (age, race, sex, and educationalattainment) and second for an indicator of treatment adherence. Herein, wereport only the effects estimated by means of multiple imputation of missingvalues; however, results using only complete data and worst-case substitutionwere substantively and statistically consistent. Adjusting for backgroundcharacteristics, participants assigned to the treatment condition were estimatedto be about 2.67 (t = 1.71, P_α/2= .04) times more likely than controls to be in remission;none of the background characteristics was a significant predictor. When treatmentadherence was added to the model, the magnitude of the estimated treatmenteffect was sharply attenuated and not statistically significant (OR, 1.39; t = 0.46, P_α/2=.32). However, participants who fully adhered to the treatment protocol weremore than 3 times (OR, 3.64; t = 1.80, P_α/2= .04) more likely to be in remission at follow-up.Thus, differences between subjects assigned to different treatment conditionswere mediated by adherence to the prescribed treatment protocols.

To examine this further, we constructed a 3-category indicator contrasting(1) participants randomized to the combined treatment arm who fully adheredto the treatment protocol with (2) subjects randomized to combined treatmentbut who had lower adherence and with (3) controls. Table 3 gives average cell frequencies and column percentages averagedacross the 20 imputed data sets. Almost 40% of participants who fully adheredto treatment were in remission at 3-month follow-up. By comparison, only 15.5%of participants randomized to treatment with low adherence, and only 12.4%of controls, reported such responses. We constructed dummy variables to compareparticipants with high treatment adherence with those with low treatment adherenceand with controls. We used SAS Proc Logistic in conjunction with Proc MIAnalyzeto obtain estimated ORs and tests of significance. Participants who fullyadhered to treatment were significantly more likely to be in remission thancontrols (OR, 6.41; t = 2.44, P_α/2= .008) or subjects randomized to treatment withlow adherence (OR, 5.00; t = 1.79, P_α/2= .04). Results observed when response to treatment(50% reduction in HAM-D scores) was analyzed were completely consistent withthese findings.

In addition, we used dummy variable regression to perform analysis ofcovariance comparing adjusted 3-month mean HAM-D scores across the 3-categorytreatment adherence indicator. After adjustment for baseline differences onthe HAM-D, participants with full treatment adherence had significantly lower3-month HAM-D scores than either low-adherence participants assigned to treatment(β = −4.77, t = −2.29, P_α/2= .01) or controls (β = −4.88, t = −2.63, P_α/2= .004).

We examined drug use reported for the 30 days before the 3-month assessment.Overall, participants reported using heroin 12.31 (SD, 12.33) days, representing8.72 fewer heroin use days than at baseline. Treatment groups did not differsignificantly with respect to the frequency of heroin use (P = .69) or cocaine use (P = .29) at follow-up.

To determine whether institutionalization was related to treatment adherence,we examined drug treatment and incarceration during follow-up. Within thetreatment group, all 5 participants who entered residential treatment hadlow adherence, while none of the highly adherent participants entered residentialtreatment (P = .06). Similarly, participants withlow treatment adherence reported an average of 13 days of incarceration duringfollow-up compared with 1.9 days among persons with high adherence (P = .04).

Regardless of treatment assignment, depression status was associatedwith frequency of heroin use. Participants in remission at 3 months, on average,reported 7.38 days of heroin use compared with 13.49 days of heroin use amongthose not in remission (t = 1.91, P = .06); they also reported fewer cocaine use days (2.17 vs 5.56; P = .14).

Twenty-four participants entered detoxification programs between baselineand the 3-month assessment (control, 18.8%, vs treatment, 32.6%; P = .16), and 10 participants entered residential treatment (5 eachin the control and treatment groups; P>.99).

Depression, a highly prevalent and debilitating disorder among substanceabusers with significant behavioral and quality-of-life implications, frequentlyremains untreated in this population. The targets of this combined pharmacotherapy-psychotherapyrandomized clinical trial were depressed, active injection drug users whocontinued high-risk injection practices. Our results demonstrate that combinedtreatment for depression is superior to a control condition of assessmentonly; however, only a minority of this population experienced depression remissionduring a 3-month follow-up. Not surprisingly, effective response to treatmentis driven by adherence to treatment.

The remission rate in the combined treatment group was 26% (the >50%response rate for this group was 30.4%), lower than the intent-to-treat responserates of 47% to 51% in meta-analyses of pharmacologic monotherapy trials thatexclude substance abusers.³⁸ The control groupremission and response rates were also lower than those typically reportedin the control arms of medication trials.³⁹ Thesefindings speak to the severe, chronic nature of depression in this population,and the limited generalizability of findings from many of the randomized controlledtrials of antidepressant medications.

Treatment resistance may have many causes, such as the presence of comorbidpsychiatric disorders. Of course, continuing drug abuse itself is associatedwith treatment resistance, and the outcome of treatment may depend on theefficacy of therapy for the comorbid disorder as well as for the depression.In addition to the severity of depression, this population is likely to haveother factors associated with poor response, including a family history ofaffective disorders, previous episodes of depression, negative life events,and poor social support.³³

While 15% of participants assigned to the combined treatment neitherused the prescribed medication nor attended any CBT session, nearly 40% metour definition of being fully adherent to treatment by attending at least75% of CBT sessions or using 75% of pharmacotherapy. Those who were fullyadherent had superior results, with 40% reporting a treatment remission, nearthe response rate reported in meta-analyses of non–substance abusers.

The low adherence to treatment may have been due in part to drug users'erratic lifestyles, one of the ongoing challenges of caring for this population.Our findings also suggest that entry into residential drug treatment and incarcerationmay lower adherence to treatment, although this remains speculative. Adherencemay also have been low, of course, because study participants did not perceivebenefits from treatment or did not engage well with therapists, or becauseappointments interfered with drug-seeking activities. Nonetheless, one ofthe important findings here is a demonstration that it is possible to engagea substantial proportion of depressed active injection drug users in a multisession,combination antidepressant therapy. Methods for engaging this population includedcollecting locator information at the start of the study and at visits, informingparticipants about follow-up assessment and therapy schedules, being flexibleand supportive about appointment scheduling, providing resources for travel,and hiring experienced staff. These methods are known to be useful in followingup substance users in research settings.⁴⁰ Inaddition, our protocol offered a small financial compensation for attendanceat CBT sessions, which may have increased treatment adherence.

Our finding that persons who adhered to CBT tended to follow the prescribedpharmacologic regimen and vice versa may also offer insight into the possibleadvantage of combined treatment for depressed, active opiate users. Each treatmentmodality may offset perceived drawbacks of the other, and each works differently,addressing different symptoms and problem areas.^24,41- 42 Provisionof psychotherapy may enhance adherence to medication or foster retention whenmedication effects are delayed. Conversely, medications may encourage patientsto persist in therapy until a therapeutic alliance is formed and coping skillsare learned. Nonetheless, persons who adhere to treatment may differ fromnonadherers and may respond differently to treatments other than those testedin this study.

Baker et al²¹ demonstrated that a 6-sessionintervention for injection drug users in methadone maintenance was successfulin attracting and retaining clients. Previous or ongoing treatment for substanceabuse may be necessary to evaluate psychotherapeutic treatments, and thereis strong evidence in favor of using psychosocial interventions to augmentthe effectiveness of methadone maintenance.^11,43 Butwhat is the best approach for the majority of injection drug users who remainwithout substance abuse treatment? Are there criteria that can be used toscreen out drug users who are unlikely to adhere to or benefit from antidepressanttherapy? The MINERVA study was a first attempt at developing antidepressanttreatment for this population. Perhaps a lower-intensity approach such asmotivational interviewing, with fewer sessions, would benefit out-of-treatmentinjection drug users and might serve as a prelude to combined depression treatment.⁴⁴ Indeed, the cause of depressive symptoms (primaryor secondary) may have a bearing on optimal treatment duration. For patientswith concurrent depression and opiate abuse, more intensive treatment maybe needed.

The issue of adherence in trials of pharmacologic agents is importantbecause failure of adherence may lead investigators to invalid conclusionsabout the effectiveness of a medication, or could undermine a study's internalvalidity. We used self-report and pill counts to rate medication use, anddrug serum levels to confirm recent use, but other methods such as MedicationEvents Monitoring System caps may be more accurate in describing the patternof medication use. Eighty-five percent of participants used at least someof the offered treatment, either medication or psychotherapy. We defined 75%completion of either modality as a "full dose" on the basis of other clinicaltrials literature. Although this is not the subject of this article, we donot know the threshold for a treatment effect for either modality in our combinedtherapy regimen.

The MINERVA study had several additional limitations. First, all dataare self-reported. Second, we did not examine the therapeutic alliance betweenpatients and therapists, which may impact adherence to treatment and outcome.Third, the follow-up occurred at 3 months; the duration of recurrent moodepisodes for patients with major depression averages 20 weeks.⁴⁵ Theeffects of combined treatment on depressive symptoms may be attenuated orenhanced after longer periods of follow-up. Because depression represents,for most patients, a chronic or recurring condition, longer outcome studiesare required to assess the need for ongoing interventions. Fourth, this studytested only a limited number of antidepressant medications and a single formof psychotherapy; other studies might consider different treatment options,for instance, testing combinations of antidepressant agents or augmentationby using a nonantidepressant, or examining the comparative effectiveness ofgroup and individual interventions. Fifth, our study design was unable todetermine the active component of combined treatment. Disentangling the relativecontribution of pharmacotherapy and psychotherapy is not possible here, although,as previously noted, most persons receiving a full dose of treatment usedboth modalities. Finally, we did not design this study with a specific adherenceenhancement component, as we might in future studies.

Clearly, there is a need for strategies to improve drug abusers' adherencewith the range of outpatient treatment, including substance abuse treatment,mental health treatment, or a combination of both. Currently, there is a poorfit between dually diagnosed substance-abusing patients and the availabledrug and mental health treatment systems. Most mental health and drug treatmentprograms do not provide integrated treatment for their dually diagnosed clients.⁴⁶ Developing and improving programs for patients withdual diagnoses who do not seek drug treatment may be important both in termsof HIV risk reduction and as a potential prelude to drug treatment entry.

Corresponding author: Michael D. Stein, MD, Division of General InternalMedicine, Rhode Island Hospital, 593 Eddy St, Providence, RI 02903 (e-mail: mstein@lifespan.org).

Submitted for publication April 16, 2003; final revision received July3, 2003; accepted July 14, 2003.

This research was supported by grants R01-MH61141 and R01-MH62719 fromthe National Institute of Mental Health, Bethesda, Md. Dr Stein is a recipientof Mid-Career Investigator Award K24 DA 00512 from the National Instituteon Drug Abuse, Bethesda.