Therapeutic Communities for Addictions: A Review of Their Effectiveness from a Recovery-Oriented Perspective

Therapeutic communities (TCs) for addictions are drug-free environments in which people with addictive problems live together in an organized and structured way to promote change toward recovery and reinsertion in society. Despite a long research tradition in TCs, the evidence base for the effectiveness of TCs is limited according to available reviews. Since most of these studies applied a selective focus, we made a comprehensive systematic review of all controlled studies that compared the effectiveness of TCs for addictions with that of a control condition. The focus of this paper is on recovery, including attention for various life domains and a longitudinal scope. We searched the following databases: ISI Web of Knowledge (WoS), PubMed, and DrugScope. Our search strategy revealed 997 hits. Eventually, 30 publications were selected for this paper, which were based on 16 original studies. Two out of three studies showed significantly better substance use and legal outcomes among TC participants, and five studies found superior employment and psychological functioning. Length of stay in treatment and participation in subsequent aftercare were consistent predictors of recovery status. We conclude that TCs can promote change regarding various outcome categories. Since recovering addicts often cycle between abstinence and relapse, a continuing care approach is advisable, including assessment of multiple and subjective outcome indicators.


Introduction
Drug addiction is a complex mental health problem that is oen associated with difficulties in various life domains such as unemployment, homelessness, relational con�icts, problems with the courts, and psychiatric comorbidity [1,2]. While some of these problems certainly evolve from the abuse of substances, many eventual addicts suffer from these problems prior to the onset of their drug use [3]. In both cases, drug addiction has generally been treated as an acute condition during brief episodes of residential care or several months of outpatient treatment, where the primary if not exclusive focus has been on abstinence to the exclusion of other concerns [4]. In contrast, addiction is increasingly regarded as a chronic relapsing disorder where recovery is possible [5], but oen the one that requires intensive or even multiple treatment episodes and/or strong personal or community resources. A continuing care approach is needed to initiate and maintain recovery [6,7]. e recovery movement focuses on individuals' perceived needs and objectives and sees abstinence as a potential resource, but not as a prerequisite, for recovery [8][9][10]. A recovery-oriented approach in addiction research implies attention for the evolutions on various life domains and individuals' subjective well being as well as the adoption of a longitudinal perspective to understand the complexity of individuals' substance use careers and recovery processes [11].
A wide range of treatment and support services are available for persons with alcohol or drug addiction problems: detox programs, drug-free outpatient treatment, methadone maintenance therapy, long-term residential treatment programs, and harm reduction services. erapeutic communities (TCs) for addictions, also called drug-free or concept TCs, aim at the reinsertion into society of former drug addicts and were one of the �rst specialized treatment initiatives for individuals with addiction problems, that evolved outside-and oen in reaction to-the traditional mental health care. e TC history dates back to Synanon, a selfsupporting community of ex-addicts founded in 1958 in Santa Monica (California) [12]. A TC can be de�ned as "a drug-free environment in which people with addictive problems live together in an organized and structured way to promote change toward a drug-free life in the outside society" [13]. Until the mid-1980s, TCs had a predominant position in most Western addiction treatment systems, but due to the drug and HIV epidemic larger scale harm reduction initiatives (e.g., methadone maintenance, needle exchange programs) became the central focus of most West European drug policies. Despite the long-standing and worldwide availability of TC treatment, TCs were criticized for their limited coverage of drug addicts, the high costs of long-term residential treatment, and the lack of evidence of effectiveness resulting from randomized controlled trials. Moreover, high drop-out and relapse rates, altered client expectations and social norms and criticism on the impact of lengthy stays in closed communities further questioned the appropriateness of TC treatment around the turn of the century [14].
Although outpatient, medically-assisted (substitution) therapy is currently the most common addiction treatment modality [15,16], one out of three clients in the European Union is engaging in other types of treatment, including therapeutic communities [17]. Recovery-oriented treatment in TCs starts from the widely accepted concept "community as a method" [18] and has been implemented on all continents. e standard TC model has been modi�ed to address the needs of speci�c populations (e.g., women with children, persons with comorbid psychiatric disorders) or new phenomena (e.g., TCs in prisons, methadone substitution during TC treatment) in the so-called modi�ed TCs (MTC) [19]. e TC method and objectives match well with the emerging recovery movement, since TC treatment can be regarded as an educational process where individuals are supported on their personal journey towards recovery and a drug-free lifestyle and to gain back control over their own lives [20].
Despite a long research tradition in TCs [21,22], the evidence base for the effectiveness of TCs is limited according to the prevailing Cochrane hierarchy of scienti�c evidence [23]. Available reviews have been biased by a selective focus on some types of TCs or study designs and a predominant focus on drug abstinence. e frequently cited Cochrane review by Smith and colleagues [23] only included randomized trials, while random group allocation appeared to be either not feasible (i.e., signi�cantly higher drop-out among controls) or advisable (i.e., motivation and self-selection are considered to be crucial ingredients of the treatment process) in several studies [24,25]. Consequently, this meta-analysis included some studies without true randomization and excluded a large number of good quality quasi-experimental studies. A recent review by Malivert and colleagues [26] le out studies on prison TCs, while this type of modi�ed TC has been the most frequently studied TC model during the last decade. Moreover, abstinence and treatment completion were the sole outcome measures in this study. Finally, the meta-analysis by Lees and colleagues [27] can be regarded as outdated, as it does not include any published study since 1999.
Since sound scienti�c evidence is needed to inform service users, treatment providers, and policy makers about TCs' potential to promote recovery, the aim of this paper is to review the effectiveness of TCs for addictions, based on a comprehensive systematic review of available randomized and nonrandomized controlled studies. e paper is limited to studies with a controlled design, as these are robust study designs that generate a high level of evidence. Also, nonrandomized studies were included, since the number of randomized studies was very small ( ) and true randomization was compromised in several studies. Given the focus on recovery, a range of outcome measures-apart from abstinence-will be evaluated and a long-term outcome perspective will be applied, including an assessment of the in�uence of aercare or continuing support.

Methods
is narrative review focuses on controlled studies (randomized trials as well as quasi-experimental designs) of therapeutic communities for addictions. We opted for a narrative review instead of a meta-analysis, given the heterogeneity of the study methodologies and the variety in data reporting. Studies were eligible if they met the following inclusion criteria.
(i) Intervention: therapeutic communities for the treatment of drug addiction that are long-term hierarchically structured (residential) educational environments, where former drug users live together and work towards recovery, and which are based on selfhelp and mutual help principles [12,21]. (ii) Target population: adults addicted to illegal drugs (mostly heroin, cocaine, or amphetamines), oen in combination with an addiction to other (legal) substances (e.g., alcohol, prescription drugs (iv) Study design: randomized controlled trials and quasiexperimental studies that have compared prospectively residents that followed TC treatment with a control group that was treated in a usual care setting ("treatment as usual"/standard of care) or another type of TC (e.g., shorter program/day TC) or with a control group out of treatment (e.g., in prison/waitlist controls). Studies needed to report �ndings on TC outcomes separately from these of other types of interventions (e.g., aercare).
Available reviews and meta-analyses were not included, but all studies selected for the reviews were screened based on the aforementioned inclusion criteria. Studies that did not focus on TC treatment, but on another type of residential care, were excluded from the paper. If several publications concerned the same baseline sample and study design, these publications were regarded as one single study.

Search Strategy.
We searched the following databases: ISI Web of Knowledge (WoS), PubMed, and DrugScope, up to December 31, 2011. ere were no language, country, or publication year restrictions. Search strategies were developed for each database, based on the search strategy developed for ISI Web of Knowledge, but were revised accordingly to take into account differences in controlled vocabulary and syntax rules. e key words we searched for were "therapeutic communit * " AND "drug * or addict * or dependen * or substance use" AND "outcome * or evaluation or follow-up or effectiveness. " e reference lists of retrieved studies and of available reviews were checked for relevant studies. In addition, the index of the International Journal of erapeutic Communities, a specialized peer-reviewed journal on therapeutic communities and other supportive organisations, was screened for relevant publications.
Our search strategy revealed 997 hits, which resulted in a �rst selection of 185 records, based on title and abstract (see Figure 1). orough analysis of these abstracts by two independent reviewers (Mieke Autrique and Wouter Vanderplasschen) led to the selection of 46 studies.
In addition to the database search, conference abstracts of European Federation of erapeutic Communities (EFTC), World Federation of erapeutic Communities (WFTC), and European Working Group on Drugs Oriented Research (EWODOR) conferences and the grey literature were scanned for relevant (un)published studies. We made a search of the registry of ongoing clinical trials to identify any ongoing RCTs. In case a publication could not be tracked through the Ghent University online library system, the study authors were contacted for a copy of the original manuscript. Finally, TC experts in various countries as well as the European Monitoring Centre for Drugs and Drug Addiction (EMCDDA) national focal points were contacted to retrieve additional (un)published or ongoing studies that have assessed the effectiveness of TCs for addictions.

Study Selection.
In total, 46 controlled studies were identi�ed (28 based on the previously mentioned search strategy and 18 additional titles were selected based on the reference lists of selected studies and available reviews). Aer reading the full texts of these articles, 16 studies were excluded, because only in-treatment outcomes were reported ( ), because the treatment provided was not in line with the TC de�nition we put forward ( ), or because the study design was deemed not a controlled design ( ). Four studies were excluded as they concerned secondary analyses of previously published data, usually with a focus on a speci�c subsample. Two studies did not compare TC treatment with a control intervention but rather compared outcomes related to speci�c client characteristics.

Data Extraction and Analysis.
Two reviewers (Mieke Autrique and Wouter Vanderplasschen) extracted data on the characteristics and results from the selected studies into a large summary table (cf. Table 1). e following study characteristics were extracted: (1) author, country (state), and year of publication; (2) type of study design and timing of follow-up measurements; (3) inclusion criteria and characteristics of the study participants + attrition rates at followup; (4) type of TC (including length of treatment) and type of control condition; and (5) outcome categories: retention and completion rates, substance use outcomes (drug and alcohol use), criminal involvement, employment, and other outcomes like health status, housing situation, and a column including determinants/correlates of abstinence/retention. Findings from studies including multiple follow-up assessments were grouped and numbered accordingly (cf. Table 1). We compared reported outcomes in various categories at all reported follow-up moments post treatment (cf. Table 2.). In this summary table, "+" indicates a signi�cant difference regarding the outcome category in favor of the experimental condition, while "-" indicates a signi�cant difference in favor of the control group. "�" means that no signi�cant between group differences were reported; alternatively text can be rephrased as follows: that no signi�cant differences were reported between the experimental and the control group.

Results
Based on our review of controlled studies of TC effectiveness, we identi�ed 30 publications that included a longitudinal evaluation of TCs for addictions and applied a prospective controlled study design (cf. Table 1). ese 30 publications are based on-in total-16 original studies, since several articles referred to the same (large) study and/or to various measurements regarding one single study (e.g., the Delaware study (no. 7) by Inciardi and colleagues [28][29][30][31][32]; the Amity prison study (no. 8) by Prendergast and colleagues [33][34][35]). orough methodological screening revealed that only �ve studies could be regarded as truly randomized (cf. Table 1), since in most studies the random group allocation process was compromised at some point [25,36] or was not possible/advisable at all [24,37,38]. e methodological quality of the studies varied but was oen rather poor due to high attrition rates, lack of objective veri�cation of study  Table 2). Overall, great within-group reductions in problem severity were observed between baseline and followup assessments, in particular regarding drug use, criminal involvement, and employment. e two outcome measures that were assessed in most studies are "substance use" and "criminal involvement. " All included studies reported at least one outcome measure in one of both categories. Eight out of 13 (note that this number is lower than 16, as not all studies reported outcomes concerning all categories) studies reported at least one positive signi�cant difference between e Scienti�c �orld �ournal 5 the TC and control group regarding legal outcomes at the one-year follow-up, while 9/14 studies found signi�cantly better substance use outcomes among the TC group at that time (cf. Table 2). All studies included multiple outcome indicators (also within one category), but only one study succeeded to �nd several signi�cant, positive outcomes regarding most legal outcome measures (i.e., reincarceration rate, days to �rst illegal activity/incarceration, and length of prison sentence) [34]. Most studies found only one signi�cant between group difference per category (e.g., time to drug relapse), while other outcome indicators within this category did not differ between groups (cf. Table 1). Signi�cantly better outcomes in one category (e.g., substance use, criminal involvement) are not necessarily accompanied by improved outcomes on other domains (e.g., employment, psychological health). Only four studies found signi�cant differences regarding three or more outcome categories [24,32,42,43].

Treatment Retention, Health, and Social Functioning.
As opposed to all other outcome categories, TC participants scored worse in comparison with controls on treatment retention/completion. Only two studies showed higher retention rates for the TC group, while three studies found signi�cantly worse completion rates among TC-participants, and six studies found non-signi�cant between group differences, mostly in favor of the control condition (cf. Table 1). Substantial drop-out has been observed in most long-term TC programs, especially in the early phases of treatment [48]. Studies that have compared longer and shorter TC programs usually found lower completion rates in longer and more intensive programs [38,51].
Five out of six studies that have reported employment outcomes found signi�cantly better employment rates among TC participants. Also, �ve studies (out of 7) showed superior outcomes on psychological symptoms, as compared with controls. Other outcomes that were studied are risk behavior ( ) and family and social relations ( ), which were found to be better in two studies [32,48].

Substance Use Outcomes.
Although TC participants had at some point posttreatment better substance use outcomes than controls in 10 studies, substance use levels varied greatly and overall, between 25% and 55% of the respondents relapsed to drug use aer 12 to 18 months. Some studies found very low initial relapse rates (e.g., 4% [38], 9% [42] and 15% [43]), while others found much higher relapse rates (e.g., 53% [34] and 69% [29]). Usually, time to relapse was longer among TC participants [52]. One of the few controlled studies that followed prison TC-participants up to three years aer their release found a relapse rate of 77% in the TC and 94% in the control condition [29]. Lower relapse rates were usually associated with longer treatment exposure (length of stay in treatment/retention) [24,31,39,41,52] and participation in subsequent treatment or aercare [32,35]. Treatment dropout and relapse aer treatment were predicted in at least two studies by the severity of substance use at baseline [28,38].

Legal
Outcomes. e majority of studies found a positive impact of TC treatment on diverse legal outcomes, such as recidivism, rearrest, and reincarceration. Recidivism rates (self-reported criminal involvement) of TC participants aer one year are usually around 40%-50% [19,31], as well as rearrest rates [29,44], although one study reported a rearrest rate of only 17% 18 months aer the start of TC treatment [42]. Reincarceration rates 12 to 18 months aer release are between 30% and 55% in most studies, although Sacks and colleagues have reported clearly lower rates (19% and 9%, resp.) in two studies [19,36]. Long-term follow-up measurements of prison TC participants indicate rearrest rates of 63% aer three years [29] and 80% aer �ve years [44] and reincarceration rates of over 70% aer 5 years [33,44]. Again, time to reincarceration was lower in the TC group and treatment completion and/or time in treatment predicted absence of recidivism [28,31,33,36,42,49]. Treatment completion was found to be associated with (older) age, single (instead of poly) drug dependence and being on parole [42].

Long-Term Outcomes and Outcome Predictors.
Six controlled studies have investigated the outcomes of TC participants in comparison with controls beyond a period of 12 to 18 months (cf. Table 2). Five of these studies show signi�cantly better legal outcomes in favor of the TC group, while only three studies could demonstrate signi�cantly lower levels of illegal drug use two years aer TC treatment. One of these studies [40] found a higher prevalence of alcohol problems among TC participants at the two-year follow-up, when compared with controls who only followed a short detoxi�cation period.
Several studies have identi�ed correlates of relapse and recidivism aer TC treatment. Participation in aercare [28,35,44], posttreatment employment [37], and older age [28,33] were found to be the most common predictors of abstinence and absence of rearrest (cf. Table 1). e effectiveness of completing treatment was shown in several studies, as TC + aercare completers had better outcomes than aercare drop-outs, who had in turn better outcomes than TC completers and TC drop-outs [33,35]. Martin and colleagues [32] even found no differences between inmates who followed in-prison TC treatment without subsequent aercare and controls who received usual work release. Relapse to drug use is oen associated with reoffending and reincarceration [46].

Type of Controls and TC Modalities.
Eleven studies have compared TC treatment with some form of usual care (e.g., case management, standard treatment, and probation), and �ve studies compared one type of TC with another form of TC treatment (modi�ed versus standard TCs, or short versus long TC programs). In the latter case, the longest/most comprehensive TC program was regarded as the experimental condition, while the shorter/least intensive program was seen as the control condition. Only three comparisons of longer and shorter TC programs yielded signi�cantly better substance use outcomes at the �rst follow-up moment [25,41,42], while overall few signi�cant differences were observed in comparison with other TC modalities. Two studies found better employment outcomes compared with lower intensity TC models, and one study found fewer

18
e Scienti�c World �ournal psychological symptoms and relational problems among the higher intensity treatment group. Some studies have included multiple control conditions [29], but usually signi�cant differences were only observed when the most intensive intervention was compared with the least intensive treatment condition.
Most controlled studies of TC effectiveness have focused on TCs in prison settings ( ) that prepare inmates for reintegration in society, while seven studies concerned TCs in the community. Whereas a substantial number of residents enter community TCs under legal pressure, TC treatment in prison can be regarded as a different context given the compulsory custody and conditional release term and privileges. Substance use outcomes in community TCs were signi�cantly better than those among controls in �ve (out of 6) studies, while legal outcomes were found to be superior in three (out of 4) studies of community TCs. On the other hand, only in four (out of 7) studies of prison TCs, the experimental group scored signi�cantly better than the control group, and only one study could demonstrate this difference beyond the one-year follow-up assessment [28]. Six (out of 9) prison TC studies found signi�cantly better legal outcomes among TC participants. ree studies could demonstrate these gains aer two years, and two studies found these bene�ts maintained up to �ve years aer prison TC treatment [28,33].

Main
Findings. is narrative review was based on 16 studies that have evaluated the effectiveness of TCs as compared with other viable interventions regarding various indicators related with recovery: substance use, criminal involvement, employment, psychological well being, and family and social relations. �ased on the study �ndings, we can conclude that there is some evidence for the effectiveness of therapeutic community treatment. Almost two out of three studies have shown signi�cantly better substance use and legal outcomes at the �rst follow-up moment aer treatment among persons who stayed in a TC as compared with controls. Five studies found superior employment outcomes among TC participants, while another �ve studies showed signi�cantly fewer psychological problems in the experimental group. Only four studies have reported significantly better differential outcomes in at least three outcome categories. is does not mean that TC participants do not improve equally on all life domains, but these outcomes oen remained unreported or the observed progress did not differ signi�cantly from that among the control group. Several reviews [22,26,27,54] have addressed the question whether TCs generate better outcomes than other interventions, oen leading to con�icting and not really convincing conclusions.
Although several studies included in this paper showed improved differential outcomes [28,33,38,39,42,50], these �ndings were observed among varying populations in diverse settings, and few studies have succeeded to replicate the �ndings from other studies in exactly the same conditions. Moreover, some studies [25,42,45] have compared modi�ed TCs with standard TCs that were not speci�cally adapted to address the needs of special target groups. In general, such comparisons of one type of (modi�ed) TC with a less intensive (standard) TC model did not demonstrate much between group differences, given the strong similarities between both treatment conditions. Consequently, the main question is not whether one type of TC is better than another intervention/type of TC, but rather which persons bene�t most from (what type of) TC treatment at what point in the recovery process [22]. Also, uncontrolled treatment outcome studies have repeatedly shown fairly similar effects of various types of residential treatment [55,56] and when compared with outpatient methadone treatment [40,57,58], demonstrating that-from a longitudinal perspective-no single intervention is superior to another. Not the differential effectiveness of TCs, but rather individuals' assets and community resources and their personal needs and goals will determine whether TC treatment is indicated on the road to recovery.

Towards a Recovery
Perspective on TC Treatment. While looking beyond abstinence and desistance is warranted from a recovery perspective [8], six of the selected controlled studies did not report other than substance use and legal outcomes. Stable recovery in opiate addicts has been primarily associated with social participation and having meaningful activities and purposes in life, rather than with drug abstinence or controlled drug use [59]. Focus groups with drug users regarding their perceived quality of life revealed few speci�c but mostly generic aspects of �o� like well being, social inclusion, and human rights [60]. Still, a predominant focus on objective socially desirable outcome measures (e.g., work, alcohol and drug use, and recidivism) prevails in addiction research, while more subjective outcome indicators like emotional well being, quality of life, or job satisfaction have largely been disregarded [61]. Such a broad perspective is also needed in TC research, as it allows a more accurate evaluation of individuals' personal growth and well being aer TC treatment. Up to now, recovery has primarily been measured based on abstinence rates aer TC treatment, while abstinence is not a synonym of nor a prerequisite for recovery [8]. Total abstinence-as required during and expected aer TC treatment-appears not to be self-evident, not even aer a lengthy treatment episode in a TC and subsequent continuing care. TC participants typically improve on most life domains during the �rst months of treatment and are usually able to maintain this status until they leave treatment [26,48]. However, once individuals leave the TC, success rates tend to drop quickly, especially during the �rst month(s) aer treatment. A recent review of longitudinal (mostly uncontrolled) TC studies showed that 21% to 100% relapsed into drug use six months to six years aer leaving treatment [26]. We found substantial relapse rates (25%-70%) 12 to 18 months aer leaving treatment, which indicate that 30% to 75% of the studied TC sample did not relapse within one year aer TC treatment. Although the de�nition of �relapse� varied largely between studies (e.g., any substance use, illicit drug use, regular use, and last month use), relapse can be addressed in at least two different ways, depending whether one starts from an acute e Scienti�c �orld �ournal 19 or a continuing care perspective. e former approach sees relapse as a failure as treated individuals did not succeed to abstain from drug use aer intensive treatment. e latter perspective acknowledges the chronic relapsing nature of drug addiction and assumes that relapse is part and parcel of the recovery process and should rather be considered as a learning moment to keep the precarious balance between abstinence and relapse [62]. Factors that may contribute to recovery are longer length of stay in the TC (retention) and participation in subsequent aercare, since both variables have been consistently identi�ed as predictors of improved substance use outcomes [23,26]. Surprisingly, treatment completion was not found to be a predictor of abstinence, but it was associated with reduced recidivism rates in several studies of prison TCs [33,36].
Treatment in TCs for addictions takes time, usually around 6 to 12 months, which heightens the possibility that residents leave prematurely [27]. Retention in (longer term) TCs is typically lower than in shorter term programs [42,51,55], but in general TC residents who stayed longer in treatment had signi�cantly better outcomes than persons who dropped out early. is has led to concerns with enhancing retention through the involvement of the family and social network and the use of senior staff [63] and with promoting initial engagement through motivational interviewing, contingency management, and induction interventions [64][65][66]. An alternative promising way of looking at retention may be to see it as the sum of treatment episodes in different services and the accumulation of associated treatment experiences instead of de�ning retention as a single uninterrupted stay in one treatment program [67]. Reentry in the community appears to be a critical point aer TC treatment, if not prepared adequately (e.g., by providing aercare) or if drug users go back to their old neighborhoods [68]. Some type of continuing support is warranted aer TC treatment not only to prevent relapse, but also to link with employment/training and to engage in community-based activities. Moreover, treatment discharge should be dealt with in a �exible and individualized way, since some persons will need to be further supported or to reenter the community if they are doing poorly. e recovery movement starts from a longitudinal approach to addiction and other mental health problems [69], but few controlled studies have assessed TC outcomes beyond a two-year follow-up period. Available studies suggest that-despite a fading effect of TC treatment over time-recidivism rates continued to be signi�cantly better than these of controls in three studies of prison TCs [28,33,37], while �ndings regarding substance use outcomes indicated fewer between group differences. e three-year follow-up outcomes of the Delaware prison study showed a 94% relapse rate among the usual care group (traditional work release) compared with a 77% relapse rate in the prison aercare TC group [29]. ese �gures do not only illustrate the relapsing nature of addiction problems, but also point at the relatively poor effectiveness of treatment programs. Although robust study designs including substantial followup periods that are able to retain most respondents in the analyses are needed, one may not overestimate the lasting effects of one single (prolonged) treatment episode. Recovery is considered to be a lengthy process, and continuing care is needed to maintain recovery that has been initiated during, for example, TC treatment. Some studies have shown that the provision of aercare was as or even more effective than initial TC treatment [29,70], and the combination of TC treatment and subsequent aercare has generated the best results [33,71].
Finally, the study �ndings show that TC treatment has generated bene�cial outcomes in diverse treatment settings and may have particularly strong effects among severely addicted individuals like incarcerated, homeless, and mentally ill drug addicts [22,36,37,46]. erefore, treatment in TCs should be considered as a speci�c intervention, reserved for drug addicts with multiple and severe problems. Although outpatient methadone maintenance therapy is the mainstream addiction treatment worldwide, therapeutic communities for addictions can be regarded as a valuable alternative for persons who do not do well in outpatient treatment due to the lack of structure and supports in the community and the fact that they live in neighborhoods that are pervasively affected by drug abuse [68]. TCs can be supportive places where clients can learn some of the internal control and refusal skills conducive to stable recovery. Motivation, social support and coping with stress without using substances appear to be key factors in successful recovery [72].

Limitations of the Paper.
First, most selected studies were published in peer reviewed journals. Although the restriction of peer-review guarantees some form of quality control, it may have induced a selection bias as the likelihood of retrieving non-English language articles was limited in this way. Only results that were reported in the published papers could be included, while it was oen unclear whether the nonreporting of some speci�c outcomes (e.g., recidivism, alcohol use) meant that this information was not collected, not analyzed, or did not yield signi�cant �ndings. Second, substantial heterogeneity has been observed between the included studies, not only regarding program and setting characteristics, but also regarding sample characteristics and outcome measures. Despite the common "community as method" principle [18], TCs for addictions consist of various practices and programs with varying treatment length. Standard and modi�ed TC programs have been evaluated in this paper, as well as TCs in prison settings and aercare TCs. is heterogeneity should be taken into account when interpreting the study outcomes. Although the underlying elements may be fairly similar across TC programs, the dosage of the program and �delity to the concept may have varied considerably [22,27]. Also, types of controls varied across studies from waitlist controls to interventions that differed only slightly from the experimental group (e.g., residential/longer versus day/shorter TC programs). Another limitation is the use of varying outcome measures and instruments across studies, which further hampers the replication and generalization of the �ndings. ird, this systematic review was not restricted to randomized trials, although the Cochrane collaboration and other proponents of the evidence-based paradigm regard this type of study design as the gold standard for the evaluation of evidence of effectiveness [73]. Given the difficulties to apply this design to long-term and comprehensive multi-interventions like TCs and the low number of randomized controlled trials on the effectiveness of TCs, a comprehensive review of randomized and nonrandomized controlled studies was deemed to be of surplus value in comparison with available reviews, still generating an acceptable level of evidence [68]. Treatment drop-outs may further compromise the validity of the reported results. Several studies only included substance users who stayed for a substantial period in the TC or who completed treatment but made no intent-to-treat analysis of everyone who started TC treatment (cf. Table 1). Finally, this is a narrative review of controlled studies that does not allow to weigh the �ndings from different studies or to estimate effect sizes. A meta-analysis was not possible at this point, given the substantial heterogeneity between programs and the diverse outcome categories and measures that were reported in the selected studies.

Conclusion
erapeutic communities for addictions can be regarded as recovery-oriented programs that produce change regarding substance use, legal, employment, and psychological wellbeing outcomes among drug addicts with severe and multiple problems. Despite various methodological constraints, TCs appeared to generate signi�cantly better outcomes in comparison with other viable interventions in two out of three studies. TC programs have usually been evaluated from an acute care perspective with a primary focus on abstinence and recidivism, while a continuing care approach including multiple and more subjective outcome indicators is necessary from a recovery perspective. If residents stay long enough in treatment and participate in subsequent aercare, TCs can play an important role on the way to recovery. Abstinence may be just one resource to promote employment or enhance personal well being which can in turn contribute to recovering addicts' participation in community-based activities and their social inclusion.