Table 1: Overview of included studies ( ).

Authors Study design + measurement(s)Participants Intervention + comparison groupOutcome measures
Correlates of relapse/abstinence
RetentionSubstance useCrim activityEmploymentOther

Author names,
year of publication,
and place (country + state)
Type of study (RCT, quasi-experimental, etc.)
Length of followup period, for example, outcomes 12 months after discharge
Number and type of participants
Attrition rate: how many of the original participants retrieved at followup?
Specific inclusion criteria
Type of TC (traditional, modified, prison TC) + type of control condition
Treatment length
For example, completion rates,
months in Tx
For example, abstinence/
relapse rates
% use in last month
time to first use after Tx
For example,
reconviction/rearrest rates
ASI scores
% reincarceration
For example,
% employment
For example,
% homeless
% psychiatric disorders
quality of life
family relations
Which variables were correlated with improved outcomes?

(1) Sacks et al., 2012
(Colorado, USA) [19]
Prospective controlled study design (partial randomisation, since assignment ratio changed during the study)
Outcomes 12 months after TC-entry
127 male offenders with substance use and mental disorders who participated in various types of prison Tx
FU: 86.6%
Reentry MTC ( )
6 month program
Controls: parole supervision case management ( )
SR drug offences: 37 versus 58%*; reincarc. rate: 19 versus 38%*  
SR crim activity: 39 versus 62%*; days till reincarc.: 161 versus 168*

(2) Zhang et al.,
2011 (California, USA)
[44]
Prospective controlled study design (QES)
Outcomes 1 and 5 years after prison release
798 male offenders with documented history of substance abuse
FU: 100% (data from official records)
Prison-based TC ( )
18 month program
Controls: matched group of untreated inmates in nearby prison ( )
1 year FU
Rearrests: 54.0 versus 47.6% (ns); reincarc.: 54.7 versus 51.9% (ns); days in prison: 79.1 versus 77.4 Participation in aftercare mediated reincarc. rates (after 1 year (ns), not after 5 years) and time in prison (after 1* and 5 years (ns))
5 year FURearrests: 80.4 versus 78.2% (ns); reincarc: 72.4 versus 72.5% (ns); days in prison: 450.4 versus 412.7

(3) Messina et al., 2010 (California, USA) [45]Prospective randomised controlled study design
Outcomes 6 and 12 months after release
115 female offenders with documented history of substance abuse
FU: 83% after 6 months
FU: 76% after 12 months
Gender-responsive MTC in prison ( )
6 month program
Controls: standard prison TC ( )
6 month program
Months in aftercare: 2.6 versus 1.8*  
MTC group had higher OR (4.60*) of successful aftercare completion
No regard. alcohol and drug ASI composite scores + self-efficacyNo in (time to) return to custody: 31 versus 45% No regard. family and psychological ASI composite scoresGreater reduction in drug use among MTC group*, when controlling for race, employ. + marital status
Return to custody less likely among MTC group, when controlling for race, employ. + living status

(4) Welsh, 2007
(Pennsylvania, USA)
[37]
Prospective controlled study design (QES in 5 state prisons)
Outcomes up to 2 years after release (on average after 17 months)
708 male inmates admitted to drug Tx in prison
FU: 100% (based on official records)
5 prison TCs ( )
Length varied from 9 ( ) to 12 ( ) and 16 months ( );
controls ( ): 3 other types of drug Tx (drug education, outpatient Tx, self-help groups)
No in number of pos. drug tests (35 versus 38%)Lower reincarc and rearrest rates, respectively 30 and 24% versus 41 and 34%*  
Reincarc. and rearrest respectively 1.6* and 1.5* times higher among controls
Higher employ.: 39.2 versus 25.9%***Reincarc. predicted by post-release employment status
Drug relapse predicted by age and employment (older and employed persons less likely to relapse)

(5) Sullivan et al., 2007 (Colorado, USA) [46]Prospective randomized controlled study design
Outcomes 12 months post-release
139 male inmates with substance use and other psychiatric disorders
FU: 75% (82 versus 69%)
Prison MTC ( )
12 month program
Controls: standard mental health Tx in prison ( )
Rates of any substance use: 31 versus 56%**; any illicit drug use: 25 versus 44%*; alcohol intox.: 21 versus 39%*  
Time to relapse: 3.7 versus 2.6 months*
Sign. association between relapse and committing new (nondrug) offences
Increased OR for reoffending (4.2*) and reincarc. (5.8*) among persons who relapsed in substance use

(5) Sacks et al., 2004 (Colorado, USA) [36]Prospective controlled study design (no true randomisation, since 51 subjects moved from one condition to another)
Outcomes 12 months after prison release
185 male inmates with substance use and other psychiatric disorders
FU: 75% (82 versus 69%)
Prison MTC ( )
12 months
Controls: standard mental health Tx ( )
Lower reincarc. rates: 9 versus 33%**; no regarding other criminal outcomesMTC aftercare participants had superior outcomes regarding rates of reincarc.*, crim. activity* and drug-related crim activity* compared with controls
Time in Tx predicted absence of reincarc** and crim activity**

(6) Morral et al., 2004 (Los Angeles, USA) [47]Prospective controlled study design (cases assigned by probation)
Outcomes 12 months after start TC program
449 adolescent probationers with substance abuse problems
FU: 90.4% after 3 months
FU: 91.3% after 6 months
FU: 90.8% after 12 months
MTC in prison (Phoenix Academy) ( )
9 month program
Controls ( ): alternative probation disposition (res. group homes)
No in program retentionImproved substance use outcomes on substance problem index*, density index*, and involvement scale*Greater, nonsign. declines on various measures of crim involvementGreater reduction of somatic** and anxiety* symptoms + sign. larger reductions in psychological symptoms between 3 and 12 month FU

(7) Inciardi et al., 2004 (Delaware, USA) [28]Prospective controlled study design (group assignment by correctional staff) Outcomes 42 and 60 months after baseline690 male inmates with substance abuse problems, eligible for work release
FU: 69.8% after 48 months
FU: 63.8% after 60 months
Work-release (transitional) TC ( )
6 month program
Controls: standard work release, without Tx ( )
TC participation strongest predictor of drug-free status after 42 (OR 4.49***) and 60 months (OR 3.54***)TC participation strongest predictor of absence of rearrest after 42 (OR 1.71**) and 60 months (OR 1.61*)Older age predicted drug-free** and no rearrest status***, while frequency of prior drug use predicted relapse** after 48 months
No previous Tx experience predicted relapse after 60 months, while older age* and being female* predicted no rearrest
TC completion associated with no rearrest and being drug free after 42 and 60 months, with superior outcomes for persons who attended additional aftercare

   
   
   
   
   
   
      
   
   
(7) Martin et al., 1999 (Delaware, USA) [29]
    
   
Prospective controlled study design (partial randomization, since KEY program autom. followed by CREST)
Outcomes 6, 18, and 42 months after baseline (i.e., 1 and 3 years after TC period)
   
      
   
   
428 inmates with drug abuse problems
FU: approx. 80%
    
Transitional TC (CREST) ( )
6 month program
Controls:
prison TC (KEY) ( )
12 month program
Prison TC + transitional TC ( )
Regular work release ( )
     
    
    
    
18 month outcomes
    
    
    
    
31% drug-free versus 16%* in work-release group (versus 47% in KEY + CREST group)
    
    
    
57% not rearrested versus 46%* in work-release group (versus 77% in KEY + CREST group)
     While greater exposure to TC Tx led to better outcomes after 1 year, at 3 years after discharge no were found between various TC modalities
CREST drop-outs as likely to be rearrested as work-release group, but CREST completers** + CREST-completers who followed subsequent aftercare*** were least likely to be arrested
CREST drop-outs more likely to be drug free than work-release group*, but CREST completers** + CREST-completers who followed subsequent aftercare*** even more likely to be drug free
    
    
    
    
42 month outcomes
    
    
    
    
OR for being drug free 8.2 times higher in CREST-group**, 7.4 times in KEY-group**, and 6.7 times in KEY-CREST group* compared with work release group; 23% drug-free versus 6%* in work-release group  
    
    
    
    
37% not rearrested versus 30% in work-release group after 42 months

(7) Lockwood et al., 1997 (Delaware, USA) [30]Prospective controlled study design
Outcomes 6 months post-release
483 inmates with history of substance abuse
FU: approx. 80%
Transitional TC (CREST) ( )
6 month program
Controls:
prison TC (KEY) ( )
12 month program
Prison TC + transitional TC ( )
Regular work release ( )
87% drug-free, versus 71% in KEY, 73.7% in work release and 93.3% in KEY-CREST group86.5% no arrest, versus 75% in KEY, 59.9% in work release and 97.1% in KEY-CREST group

(7) Nielsen et al., 1996 (Delaware, USA) [31]Prospective controlled study design (QES)
Outcomes after 6 and 18 months
689 inmates with history of substance abuse
FU: 77 versus 72.6% after 6 months
FU: 58.5 versus 36.7% after 18 months
Transitional TC (CREST) ( )
6 month program
Controls: conventional work release ( )
Sign. lower relapse after 6 (16.2 versus 62.2)*** and 18 months (51.7 versus 79%)***Sign. lower recidivism after 6 (14.7 versus 35.4)*** and 18 months (38.2 versus 63%)***Age, race, and gender do not affect outcomes, but length of time in program reduced relapse and recidivism rates (ns)
Program completion associated with fewer relapse*** after 6 and fewer recidivism after 6*** and 18 months*

(7) Martin et al., 1995 (Delaware, USA) [32]Prospective controlled study design (QES)
Outcomes 6 months after release
483 inmates with history of substance abuse
FU: approx. 80%
Transitional TC (CREST) ( )
6 month program
Controls:
prison TC graduates (KEY) ( )
Prison TC + transitional TC ( )
Regular work release ( )
Probability of being drug free the highest among CREST (0.84)*** and KEY + CREST group (0.94)***Prob. of being arrest free the highest among CREST (0.86)*** and KEY + CREST group (0.97)***Prob. of no longer injecting the highest in CREST (0.97)*** and KEY + CREST group (0.97)No between TC only and work release group on any of the outcome measures
Longer time in (subsequent) Tx the best predictor of drug-free*** and arrest-free* status after Tx, as well as participation in a longer TC program

(8) Prendergast et al., 2004 (California, USA) [33]Prospective randomized controlled study design
Outcomes 5 years after release
715 male inmates with substance abuse problems
FU: 81.2%
Amity prison TC ( )
9–12 month program
Controls: no Tx condition (waitlist)
( )
Months receiving Tx post-release: 4.6 versus 1.7***Heavy drug use past year: 24.9 versus 22.6%Reincarcerated within 5 years: 75.7 versus 83.4%*  
Days to reincarc: 809 versus 634***
Stable job in past year: 54.8 versus 52.3%Psychologic. distress: 31.8 versus 44.6Reincarc. predicted by younger age* and fewer months in Tx after release***  
Completion of TC and aftercare predicted absence of reincarc.***

(8) Prendergast et al., 2003 (California, USA) [34]Prospective randomized controlled study
Outcomes 12 months after release
715 male inmates with substance abuse problems
FU: 74%
Amity prison TC ( )
9–12 month program
Controls: no Tx condition (waitlist) ( )
Longer time to first drug use: 77 versus 31 days***  
No in pos. drug tests (52.9 versus 61%)
1 year reincarc. rate: 33.9 versus 49.7%*; more days to first illegal act. (138 versus 71 days)***; no in type of arrest; more days to first incarcer. (285 versus 243 days)***; fewer months in prison (3 versus 4.7)***Participation in Tx associated with more days to reincarc.
Aftercare completers had the lowest reincarc rate + the longest time to first illegal activity and to reincarceration and fewer days in prison
Prison TC drop-outs had the shortest time to SR drug use (32 days), followed by TC completers (62 days), aftercare dropouts (91 days), and aftercare completers (184 days) + more pos. drug tests

(8) Wexler et al., 1999 (California, USA) [35]Prospective randomised controlled study design
Outcomes 12 and 24 months after release
715 male inmates who volunteered for TC treatment in prison
FU: 100% after 12 months
FU: 36.8% after 24 months
Amity prison TC ( )
8–12 month program
Controls: no Tx condition (waitlist) ( )
Lower reincarc rates after 12 (33.9 versus 49.7%)*** and 24 months (43.3 versus 67.1%)***   
More days to reincarc after 12 (192 versus 172*) months
OR for reincarc sign lower: 0.52*** after 12 and 0.63** after 24 months
Reincarc. rates sign lower after 12 and 24 months among TC + aftercare completers, as opposed to persons who dropped out previously
Aftercare completion positively related to time to reincarc. + the strongest predictor of positive outcomes

(9) Greenwood et al., 2001 (San Francisco, USA)
[25]
Prospective controlled study design (only partial randomisation, since sign. drop-out among control before Tx start)
Outcomes 6, 12, and 18 months after admission
261 substance abusers seeking treatment at Walden House
FU: 82.4% at 6 months
FU: 82.7% at 12 months
FU: 82.7% at 18 months
Residential TC ( )
12 month program
Controls: day TC
program (same TC, but returned home at the end of the day) ( )
Time in program: 109.8 versus 102.7 daysTotal abstinence after 6 (62.6 versus 47%), 12 (47.9 versus 49%) and 18 months (50.4 versus 55.2%)
OR for relapse at 6 months = 3.06*, not sign. at 12 and 18 months
Relapse after 18 months predicted by employment status prior to Tx start***, injecting drug use** and having >1 sexual partner**

(9) Guydish et al. 1999 (San Francisco, USA)
[48]
Prospective controlled study design (only partial randomisation, since sign. drop-out among controls before Tx start)
Outcomes at 6, 12, and 18 months
188 substance abusers seeking treatment at Walden House who participated in all 3 FU-interviews
Residential TC ( )
12 month program
Controls: day TC
program (same TC, but returned home at the end the day) ( )
No in time to drop-out (119.7 versus 108.1 days)
12 month retention in day TC: 17% versus 9%
Lower SCL scores at 6**, 12*, and 18* months, lower BDI scores after 12 months*, higher social support scores at 18 months*; lower social problem severity (ASI)*Most changes observed during first months of Tx, followed by maintenance of change

(9) Guydish et al. 1998 (San Francisco, USA)
[49]
Prospective controlled study design (only partial randomisation)
Outcomes at 6 months
261 substance abusers starting treatment at Walden House
FU: 82.4% at 6 months
Residential TC ( )
12 month program
Controls: day TC
12 month program ( )
Tx adherence after 6 months: 29 versus 34% in day TC;
Time in Tx: 109.8 versus 102.7 days
Lower ASI severity scores for social* and psychological problems**Persons who stayed >6 months in Tx had sign lower legal, alcohol, drug, and social severity scores
Alcohol severity reduced sign. if persons stayed >6 months in residential TC

(10) Nemes et al. 1999 (Washington, USA) [42]Prospective, randomised controlled study design
Outcomes 18 months after admission
412 substance users seeking Tx at a central intake unit
FU: 93%
Standard TC ( )
12 month program (10 months inpatient, 2 outpatient)
Controls: abbreviated TC ( ): 12 month program (6 month inpatient, 6 month outpatient + extra services)
Completion rates: 33 versus 38% (ns), and similar time in Tx (8.2 versus 8.6 months)Lower SR heroin use: 9 versus 15%*Lower rearrest rates: 17 versus 26%** + longer time to arrest (9.4 versus 6.9 months)*Employment rate higher in standard TC: 72 versus 56%**Lower heroin and cocaine use levels + lower rearrest rates among treatment completers versus noncompleters
Positive cocaine tests were associated with premature Tx drop-out
Treatment completion was predicted by age, single heroin dependence, and parole status

(11) De Leon et al., 2000 (New York, USA) [24] Prospective controlled study design (QES: sequential group assignment)
Outcomes 12 and on average 24 months after baseline
342 homeless mentally ill substance abusers
FU: 68% at 12 months
FU: 82% at latest FU
MTC1 for homeless persons ( )
12 month program
MTC2: lower intensity, flexible program ( )
12 month program
Controls: treatment as usual ( )
12 monthsMTC2 had less alcohol intox* + fewer illegal drug use** + used less substances** than TAU
No between MTC1 and TAU
No between MTC1 or MTC2 and TAUMTC1*** and MTC2*** more likely to be employed than TAUNo between MTC1 or MTC2 and TAU regard. HIV risk behavior and psychological dysfunctions
24 monthsMTC2 had less alcohol intox* + used less substances* compared with TAU
No between TC1 and TAU
MTC1* and MTC2*** committed fewer crimes than TAUMTC1** and MTC2*** more likely to be employed than TAUMTC2 had less symptoms of depression*** and anxiety* than TAU MTC2 improved more on several outcomes measures than MTC1
MTC completers scored sign better than MTC drop-outs and TAU

(11) French et al., 1999 (New York, USA) [50]Prospective controlled study design (QES: sequential group assignment)
Outcomes at last FU-point (on average 24 months after baseline)
342 homeless mentally ill substance abusers
FU: 82%
MTC for homeless persons ( )
12 month program
Controls: treatment as usual ( )
No regard. substance use outcomesFewer criminal activity**Better employm. outcomes (ns)Lower scores on BDI*, no regard. other psychological symptoms or risk behavior

(12) Nuttbrock et al., 1998 (New York, USA) [38]Prospective controlled study design (QES, as allocation based on availability + client preference)
Outcomes 12 months after start Tx
290 homeless men with major mental disorder and history of substance abuse
FU: not reported
Modified TC ( )
18 month program
Controls: 2 homeless community residences ( )
18 month program
43% stayed 6 months in TC (versus 55%); 25% stayed 12 months (versus 37%)4.1 versus 30.1% pos. urine tests*; SR alcohol use: 0 versus 14.3%*; SR marijuana use: 2.6 versus 2.9%; SR crack use: 7.7 versus 14.2*Greater (ns) reductions in psycho-pathology (depression, anxiety, psychiatric distress) MTC participation predicted lower levels of anxiety** and better GAF-scores**Drop-out after 6–12 months in community residences was predicted by substance use severity*

(13) McCusker et al. 1997 (New England, USA)
[51]
Prospective controlled study design (no real randomisation, since interventions at both study sites)
Outcomes 3 months after discharge and 18 months after admission
539 drug abusers entering residential Tx at 2 sites
FU: 86% after 18 months
Traditional TC program (6 ( ) and 12 month alternative ( ))
Controls: MTC program (relapse prevention) 3 ( ) and 6 month ( ) alternatives
Tx completion: 23% in long TC, 34% in shorter TC, 31% in long MTC and 56% in short MTCTime to drug use not between TCs and than MTC
Stronger effect of long TC versus short TC and MTCs regard. drug and alcohol severity (ns)
Stronger effect of long TC versus short TC and MTCs regarding legal problems Effect of TC on employm. stronger than in MTC*Small effects of long TC versus short TC and MTCs regard. other ASI domains

(13) McCusker et al., 1996 (Massachusetts, USA) [52]Prospective randomized controlled study design
Outcomes 6 months after Tx
444 drug abusers entering one residential Tx facility
FU: 74%
Long MTC ( )
6 month program
Controls: short MTC ( )
3 month program
Program completion: 30 versus 56% in short TC programRelapse to drug use in first week after leaving Tx: 33 versus 70%*  
No group in heroin or cocaine use
Greater improvement in levels of depression among persons staying >80 days in TC***  
Length of stay in TC** + program completion* pos. associated with levels of precontemplation
Persons staying >80 days in TC had lower drug use***

(13) McCusker et al., 1995 (New England, USA)
[53]
Prospective controlled study design
Outcomes 3–6 months after discharge
628 drug abusers entering residential Tx at 2 sites
FU: 84% in TC versus 74% in MTC
Traditional TC program (6 ( ) and 12 month ( ) alternative)
Controls: MTC 3 ( ) and 6 month program ( )
40 day retention: respectively, 70, 85, 73, and 72%; Tx completion: respectively, 33, 21, 56, 30% (ns in 4 groups)Relapse: 50% in TC versus 44% in MTC
No in number of days of drug use

(14) Hartmann et al. 1997 (Missouri, USA) [43]Controlled study design (QES, self-selection for exp. intervention)
Outcomes at least 5 months after release
286 male offenders with a history of substance abuse
No information on FU-rate
Prison TC graduates ( )
No information on program length
Controls: comparison group of eligible persons who did not attend prison TC ( )
No substance abuse: 67.4 versus 62% (ns)No arrest: 85.4 versus 72%**  
Reincarc.: 16.4 versus 27.6*

(15) Bale et al., 1984 (California, USA) [39]Prospective controlled study design (only partial randomization due to substantial drop-out after group allocation)
Outcomes after 2 years
363 male veterans addicted to heroin entering withdrawal Tx
FU: 95.6%
3 TCs ( ): standard TC ( ) + two MTCs ( and )
6 month programs
Controls: 5-day withdrawal Tx ( )
Mean TIP longer in TC 1 (10.4 weeks) and TC3 (11.5 weeks)* than in TC2 (6.0 weeks)No heroin use: 40, 48.1, and 35.4% versus 33.3% of controls;
No other illegal drug use: 40, 41.6, and 53.3%* versus 39.3% of controls
More alcohol problems: 40, 47.3** and 30.8% versus 22.4% of controls
No conviction: 44*, 32.5 and 59.5%** versus 31.3% of controlsEmployed/attending school: 48*, 46.8 and 51.9%** versus 34% of controlsMortality: 1.7% in TCs versus 6.6% among controlsThe 3 TCs differed largely on program characteristics
Heroin use + other major outcomes sign. better among subjects staying longer in Tx

(15) Bale et al., 1980 (California, USA) [40]Prospective controlled study design (as treated analyses)
Outcomes after 1 year
585 male veterans addicted to heroin entering withdrawal Tx
FU: 93.2%
Veterans staying long (≥50 days) ( ) or short in TC program (<50 days) ( )
Controls: MMT ( ); detox only ( ); detox + other Tx ( )
1-year retention rate: <5% in TC versus 74.5% in MMT Recent heroin use + any illicit drug use lower in long TC subjects (37.3 and 29.3) than detox only-group (65.5 and 46.9)**, but not than MMT group (46.6 and 38.6%)Arrest (37.3%), conviction (21.3%) and reincarc (4%) rate sign lower than in detox only-group (54.5**, 38** and 21.1%***, resp.), but not than MMT group (49.2, 22, and 10.2%)Employm./school attendance: 65.3% of long TC group, 50.9% of MMT and 38.4%*** of detox only-groupSign. more subjects had good global outcome score in long TC (64%) and MMT (54.45%) versus detox only-group (33.8%)TC with confrontational style least successful
Short TC group did not score sign better than detox only-group
Twice the number of long TC subjects (29.3%) had the max. global outcome score than MMT clients (14%)

(29) Coombs, 1981 (California, USA) [41]Prospective controlled study design (group allocation by self-selection)
Outcomes 11–18 months after leaving TC
207 heroin addicts starting treatment in one of 2 TCs
FU: 78.5%
Long-term TC ( )
12 month program
Controls: short-term TC ( )
3 month program
Program completion: 63.6 versus 74.6%Total abstinence: 4.3 versus 0%;
Return to heroin use: 28.6 versus 53%
Program graduates used less often illicit drugs and were less likely to have relapsed or to be rearrested compared with splittees. Also higher employment rates among graduates

TC: therapeutic community, MTC: modified therapeutic community; SR: self-reported; QES: quasi-experimental study; Tx: treatment; TIP: time in program; BDI: Beck Depression Inventory; ASI: Addiction Severity Index; level of significance: *P < 0.05; **P < 0.01; ***P < 0.001.