Treatment outcomes for methamphetamine use: Finding from the methamphetamine treatment evaluation study (MATES)

Evaluating the impact of community-based treatmentoptions on methamphetamine use: findings from theMethamphetamine Treatment Evaluation Study(MATES)

Rebecca McKetin1,2, Jake M. Najman3, Amanda L. Baker4, Dan I. Lubman5, Sharon Dawe6,Robert Ali7, Nicole K. Lee8,9, Richard P. Mattick2 & Abdullah Mamun3

Centre for Research on Ageing, Health and Wellbeing, The Australian National University, Canberra, Australia,1 National Drug and Alcohol Research Centre,University of New South Wales, Sydney, Australia,2 Queensland Alcohol and Drug Research and Education Centre, University of Queensland, Brisbane, Australia,3

Centre for Brain and Mental Health Research, University of Newcastle, Newcastle, Australia,4 Turning Point Alcohol and Drug Centre, Eastern Health and MonashUniversity, Melbourne, Australia,5 School of Psychology, Griffith University, Brisbane, Australia,6 University of Adelaide, Adelaide, Australia,7 The National Centrefor Education and Training on Addiction, Flinders University, Adelaide, Australia8 and National Drug Research Institute, Curtin University, Perth, Australia9

ABSTRACT

Aims To evaluate the impact of community-based drug treatment on methamphetamine use using inverse probabil-ity of treatment-weighted (IPTW) estimators to derive treatment effects. Design A longitudinal prospective cohortstudy with follow-ups at 3 months, 1 year and 3 years. Treatment effects were derived by comparing groups atfollow-up. IPTW estimators were used to adjust for pre-treatment differences between groups. Setting Sydney andBrisbane, Australia. Participants Participants were methamphetamine users entering community-based detoxifica-tion (n = 112) or residential rehabilitation (n = 248) services and a quasi-control group of methamphetamine users(n = 101) recruited from the community. Measurements Frequency of methamphetamine use between interviews(no use, less than weekly, 1–2 days per week, 3+ days per week), continuous abstinence from methamphetamine use,past month methamphetamine use and methamphetamine dependence. Findings Detoxification did not reducemethamphetamine use at any follow-up relative to the quasi-control group. Relative to quasi-control and detoxificationgroups combined, residential rehabilitation produced large reductions in the frequency of methamphetamine use at 3months [odds ratio (OR) = 0.23, 95% confidence interval (CI) 0.15–0.36, P < 0.001), with a marked attenuationof this effect at 1 year (OR 0.62, 95% CI 0.40–0.97, P = 0.038) and 3 years (OR = 0.71, 95% CI 0.42–1.19,P = 0.189). The greatest impact was for abstinence: for every 100 residential rehabilitation clients there was again of 33 being continuously abstinent at 3 months, with this falling to 14 at 1 year and 6 at 3 years.Conclusions Community-based residential rehabilitation may produce a time-limited decrease in methamphetamineuse, while detoxification alone does not appear to do so.

Keywords Amphetamine, longitudinal, methamphetamine, outcomes, psychiatric comorbidity, substance abuse,treatment, HIV risk, crime.

Correspondence to: Rebecca McKetin, Centre for Mental Health Research, ANU College of Medicine, Biology and Environment, The Australian NationalUniversity, Canberra, ACT 0200, Australia. E-mail: [email protected] 23 November 2011; initial review completed 5 January 2012; final version accepted 2 May 2012

INTRODUCTION

Methamphetamine dependence is estimated to cost atleast three-quarters as much as heroin dependenceper person [1], yet there is comparatively little evidencedocumenting treatment outcomes [2] and few specia-lized treatment options [3]. Notable is the lack of any

evaluation of whether generic community-based drugtreatment can reduce methamphetamine dependence.Existing studies in this area focus largely on opioidsand/or alcohol [4–6]—this research finding support forcommunity-based residential rehabilitation [4–6] but lessfor detoxification [4]. However, these findings may notapply to people who are dependent on methampheta-

RESEARCH REPORT

bs_bs_banner

doi:10.1111/j.1360-0443.2012.03933.x

© 2012 The Authors. Addiction © 2012 Society for the Study of Addiction Addiction, 107, 1998–2008

mine, who have high rates of psychiatric comorbidity[7–10] and who have different treatment needs thanpeople dependent on other drugs [11]. Given the scale ofmethamphetamine use internationally [12] and theseverity of problems it can incur [2], it is critical to knowwhether existing drug treatment infrastructure can beused to treat methamphetamine dependence.

A broader issue is the lack of control groups in previouscommunity-based treatment outcome studies [4–6],making it difficult to attribute changes in drug use totreatment (cf. natural remission from drug use or othersecular changes). Even where comparison groups exist,their non-equivalence makes it difficult to infer treat-ment effects [4]. In other areas of public health, treatmenteffects have been derived from observation treatmenttrials using ‘propensity matching’ procedures. Propensitymatching provides a method of adjusting for bias in quasi-experimental studies that is superior to conventionalregression adjustment. This method involves deriving acomposite ‘propensity score’ for each individual in thestudy, which reflects the probability that they wouldreceive treatment based on their clinical characteristics[13]. Treatment effects reflect the difference in outcomesfor participants in each group who have a similar propen-sity score. This difference is then averaged across thesample to derive an ‘average treatment effect’.

Recent statistical developments provide a more parsi-monious solution to the non-equivalence of treatmentgroups by using inverse probability of treatment-weighted estimators (IPTW) [14]. Participants areweighted according to the inverse probability that theywould receive treatment based on their observable char-acteristics (e.g. demographics, prior treatment history,severity of illness). This creates a pseudo-population ofgroups that are equivalent on their pre-existing charac-teristics. Their benefit over conventional matching proce-dures is that they provide treatment effects that are moreamenable to interpretation, they eliminate residual con-founding that arises from using a fixed number of stratato match clients and they provide better adjustment fortime-dependent confounders (e.g. heavier drug usersseek more intensive treatment and are more likely torelapse) [14,15].

Although IPTW estimators have been appliedincreasingly in public health research to establish theefficacy of clinical protocols in situ [16–18], this meth-odology has not yet been applied to drug treatmentoutcome studies. The aim of this study was to use IPTWestimators to determine the effect of community-basedtreatment (residential rehabilitation and detoxification)on methamphetamine use. Outcomes for these treat-ment modalities were compared to a quasi-controlgroup of methamphetamine users recruited from thecommunity. Data were taken from the Methampheta-

mine Treatment Evaluation Study (MATES): a longitu-dinal prospective cohort study of methamphetamineusers.

METHOD

Participants and procedure

Treatment participants (n = 360) were recruited on entryto 15 residential rehabilitation facilities and 11 detoxifi-cation units located in Sydney (n = 19) and Brisbane(n = 7), Australia. These agencies were selected from theAustralian Minimum Dataset for Alcohol and Other DrugTreatment Services (MDS–AODTS) [19] which includesall publicly funded government and non-governmentdrug and alcohol treatment agencies in Australia.Whether services provided detoxification or residentialrehabilitation was based on how they were classified inthe MDS–AODTS [19]. Being community-based drugtreatment services, there were no constraints on the typeof treatment provided within each treatment modality.Detoxification typically involved brief (e.g. 1 week)in-patient stays with medical support to manage with-drawal symptoms. Residential rehabilitation typicallyinvolved longer stays (e.g. several weeks to months) in adrug-free residential setting that provided an intensiveprogramme of integrated services and therapeutic activi-ties (e.g. behavioural treatment approaches, recreationalactivities, social and community living skills, group workand relapse prevention) [19].

Inclusion criteria for treatment participants were: (i)methamphetamine or amphetamine recorded as theprimary or secondary drug problem in the MDS–AODTS,(ii) being aged 16+ years, (iii) willingness to participate infollow-up interviews and (iv) comprehension of English.Exclusion criteria were having received methampheta-mine treatment or any in-patient drug treatment, orimprisonment, in the month prior to entering the study.These exclusion criteria were necessary to obtain a natu-ralistic level of drug use at the baseline interview. Ineligi-bility (n = 195) was due mainly to drug treatment (58%)or incarceration (28%) in the month prior to recruitment,while 10% declined participation. Differences betweenthis treatment sample and all recorded methampheta-mine treatment episodes in Australia were small (seeAppendix S1; details are given at the end of the paper).

The quasi-control group (n = 101) was recruitedthrough community health services and needle andsyringe programmes from the greater Sydney region fromJanuary 2006 until May 2008. Inclusion and exclusioncriteria were the same as for the treatment group. In addi-tion, quasi-control participants needed to have aminimum level of methamphetamine use to ensure thatthey were comparable to the treatment group: either

Methamphetamine use treatment outcomes 1999


screening positive for methamphetamine dependence (ascore of 4+ on the Severity of Dependence Scale [20,21])or using 3+ days/week in the past month. Most ineligiblequasi-control participants (n = 109) failed to meet themethamphetamine use inclusion criteria (43%) or hadbeen in drug treatment in the past month (35%), while16% declined participation.

Baseline interviews took 1.5 hours and were con-ducted face to face. Follow-up interviews (3 months, 1year, 3 years) took 45 minutes and were conducted faceto face or by telephone. Interviews were conducted attreatment centres or mutually convenient locations (e.g.cafes, parks, local health centres). All participants pro-vided informed consent and were reimbursed (AU$30 at3 months and 1 year, and AU$40 at 3 years). Additionalinformed consent and reimbursement (AU$10) was pro-vided for hair samples at 1 year.

Baseline interviews with treatment participantswere conducted on a median of 6 days after treatmententry (interquartile range 2–10 days) and measures per-tained to the month prior to treatment entry. Follow-upinterviews occurred at 3 months (median 98 days, inter-quartile range 88–117), 1 year (median 378 days, inter-quartile range 361–400) and 3 years (median 1136days, interquartile range 921–1299) after the baselineinterview. The time to follow-up was shorter for thequasi-control condition than for the treatment condi-tions (Fig. 1). However, there was no significant relation-

ship between days to follow-up and methamphetamineuse at any of the follow-up points.

Measures

Treatment exposure

The index treatment was defined as contiguous treatmentfrom recruitment, allowing up to 7 days’ gap in treatmentto accommodate for transfers between services. The dura-tion of the index treatment was defined as the first to thelast day of treatment, and was measured at the 3-monthfollow-up interview. For residential rehabilitation clients,the duration of treatment included any time spent indetoxification in preparation for residential rehabilitation(this occurred in 71% of cases). The number of treatmentepisodes initiated after the index treatment, and whetherthe participant received treatment for their metham-phetamine use during each of these treatment episodes,was recorded at each follow-up interview.

Methamphetamine use

Self-reported frequency of use was measured in the yearprior to entering the study and between each interviewusing the categories: no use, less than weekly, weekly,twice weekly, 3–4 days a week, 5+ days a week. Continu-ous abstinence was defined as no use of methampheta-mine since the baseline interview. Past month

Figure 1 Follow-up of participants by group

2000 Rebecca McKetin et al.


methamphetamine use was assessed at each interviewusing the Opiate Treatment Index (OTI) [22]. Self-reported methamphetamine use in the past month wasvalidated at follow-up in a subsample of 83 participants,and abstinence was confirmed in 94% of these cases (seeAppendix S2; details are given at the end of the paper).

Other measures

Motivation to reduce methamphetamine use wasmeasured at baseline using the Readiness to ChangeQuestionnaire and participants were scored as beingat the ‘pre-contemplation’, ‘contemplation’ or ‘action’stage [23].

Polydrug use was defined as the total number ofdrug classes used in the past month (including heroin,other opioids, cocaine, ecstasy, hallucinogens, cannabis,alcohol, inhalants and tobacco).

The Composite International Diagnostic Interview(CIDI) [24] was used to make DSM-IV diagnoses of meth-amphetamine dependence and other Axis I psychiatricdisorders. A DSM-IV diagnosis of conduct disorder wasmade using a modified version of the Diagnostic Inter-view Schedule [25].

Statistical analysis

Data were analysed using STATA version 11.1. All testswere two-sided, with significance set at P < 0.05. Mediansand interquartile ranges are reported for skewed data.Unadjusted treatment effects were derived for detoxifi-cation and residential rehabilitation compared to thequasi-control group. Comparisons between groups wereundertaken at each follow-up using logistic regressionfor dichotomous outcome measures and ordinal logisticregression for categorical outcome measures. Generalizedordinal logistic regression models were used where theassumption of proportional odds was breached. Subse-quent analyses used IPTW estimators to compare the out-comes for the residential rehabilitation group to the quasi-control and detoxification groups combined (n = 213).These groups were combined to increase statistical power,because their outcomes were not significantly different.IPTW treatment effects were derived by applying probabil-ity weights that represented the inverse probability thatthe participant would have received treatment based ontheir baseline characteristics. Weights were calculatedaccording to the procedure described by Robins and col-leagues [14] (see Appendix S3; details are given at the endof the paper). Confidence intervals (CI) were derived usingrobust standard errors to account for data clusteringwithin treatment centres and the IPTW weighting proce-dures. Missing data were imputed using multiple imputa-tion by chained equations, implemented using the STATA‘ice’ command (see Appendix S4; details are given at theend of the paper) [26].

RESULTS

Participant characteristics

Almost all participants (97%) met DSM-IV criteria formethamphetamine dependence at baseline; this did notdiffer significantly between treatment groups (Table 1).Most injected the drug, and the majority comprisedunemployed single males (Table 1). Participants hadused methamphetamine on a median of 16 days in thepast 4 weeks (interquartile range 8–23 days) with 3% ofparticipants not having used during this time. Otherdrug use in the past month consisted largely of tobacco(95%, 90% daily), cannabis (78%, 39% daily) andalcohol (71%, 15% daily), with a minority using heroin(26%, 3% daily) or cocaine (27%, <1% daily). In terms ofreadiness-to-change methamphetamine use, the major-ity of participants were in the action stage (55%). Mostparticipants had an extensive history of drug treatment(Table 1). Treatment participants were typically volun-tary admissions (85%) seeking complete abstinence(87%; 79% for detoxification and 91% for residentialrehabilitation).

There were a number of differences between partici-pants in the quasi-control group and those in the treat-ment groups, most notably that the quasi-control groupused methamphetamine less often, were less severelydependent on the drug and were less motivated to reducetheir methamphetamine use (Table 1).

Follow-up of the cohort

The follow-up rates were 80% at 3 months, 74% at 1 yearand 66% at 3 years, with 88% of participants followed-upat least once; loss to follow-up included participants whowere incarcerated or deceased (Fig. 1). The main predic-tors of attrition were low education and a prison history.Analyses presented below use imputed missing data. SeeAppendix S4 for details of attrition, the imputation pro-cedure and results without data imputation.

Treatment exposure

Sixty-two per cent received residential rehabilitation(n = 248) and the remainder (28%, n = 112) receiveddetoxification as their index treatment. The median dura-tion of residential rehabilitation was 62 days (interquar-tile range 29–98) and detoxification 5 days (interquartilerange 4–7). Most participants (83%) had left their indextreatment by the 3-month follow-up.

Additional treatment during the follow-up period wascommon in all groups (Fig. 2), with 47% of the samplereceiving additional treatment and 43% receiving addi-tional treatment for methamphetamine use.



Treatment outcomes

Unadjusted outcomes

There was a reduction in the frequency of methampheta-mine use over the 3-year follow-up period in both thequasi-control group and the treatment groups (Fig. 3).There was no significant difference in frequency of meth-amphetamine use between the detoxification group andthe quasi-control group at any follow-up [3 months: oddsratio (OR) 0.96, 95% CI 0.56–1.63, P = 0.882; 1 year:OR 1.01, 95% CI 0.59–1.75, P = 0.967; 3 years: OR0.79, 95% CI 0.44–1.40, P = 0.418]. The residentialrehabilitation group showed a large reduction in meth-amphetamine use relative to the quasi-control group atthe 3-month follow-up (OR = 0.20, 95% CI 0.13–0.33,P < 0.001). This effect was attenuated but still significantat 1 year (OR 0.58, 95% CI 0.36–0.95, P = 0.031) and 3years (OR = 0.57, 95% CI 0.33–0.99, P = 0.045). The

residential rehabilitation group showed a similar patternof results relative to the detoxification group, who weremore similar in their baseline methamphetamine use(Fig. 3), although the effect was not significant at 3 years(3 months: OR 0.24, 95% CI 0.15–0.40, P < 0.001; 1year OR 0.59, 95% 0.36–0.97, P = 0.038; 3 yearsOR = 0.72, 95% CI 0.40–1.32, P = 0.289).

IPTW treatment effects

When the residential rehabilitation group was comparedwith the quasi-control and detoxification groups com-bined (n = 213), unadjusted effects for residential reha-bilitation remained significant at all follow-ups (Table 2).IPTW estimators were used to adjust for between groupdifferences in the number of previous drug treatmentepisodes, previous treatment for methamphetamine use,severity of methamphetamine use prior to treatment

Table 1 Characteristics of participants by treatment modality.

Quasi-control(n = 101)

Detoxification(n = 112)

Residentialrehabilitation(n = 248)

Methamphetamine useDSM-IV diagnosis of dependence (%) 94 96 98Duration of use (mean years) 16.6 13.0** 12.1***Severity of dependence (mean SDS score)a 7.6 9.1** 8.9**Days used in the past month (median) 14 20*** 16†

Injecting methamphetamine (%) 86 73* 67***Polydrug use (mean) 3.8 3.4* 3.6Demographics

Age (mean years) 35.3 31.7** 30.5***Male (%) 67 72 77Unemployed (%) 84 74 89†

Immigrant (%) 13 12 15Incomplete schoolingb (%) 39 40 29†

Tertiary qualification (%) 42 46 49No fixed address (%) 5 13* 5†

Prison history (%) 61 46* 38***Ever had children (%) 53 52 50Single (%) 72 79 72

Previous drug treatment (%)Nil 25 30 301–4 episodes 36 40 425+ episodes 40 29 28

Previous treatment for methamphetamine use (%) 20 42*** 46***Motivation (%, action stage) 30 54*** 66***†

Psychiatric disorders (%)Major depressionc (%) 63 38*** 40***Social phobiac (%) 27 21 23Panic disorderc (%) 33 15** 29†

Schizophrenia or maniad (%) 15 8 13Conduct disorderd (%) 73 71 80

***P < 0.001; **P < 0.01; *P < 0.05, compared to the quasi-control group. †P < 0.05 compared to the detoxification group. aHigher scores indicate moresevere dependence. bCompleted less than 10 years of schooling. cPast year diagnoses taken at baseline. dLife-time diagnoses taken at follow-up. SDS:Severity of Dependence Scale.



entry, motivation to reduce methamphetamine use, poly-drug use, unemployment and prison history (See Appen-dix S3 for details).

IPTW estimators reduced the magnitude of the treat-ment effects which were no longer significant at the3-year follow-up (Table 2). The greatest change in meth-amphetamine use was seen for abstinence, with 34%more participants in the residential rehabilitation groupreporting no use at the 3-month follow-up, this droppingto 9 and 6% at the 1- and 3-year follow-ups, respectively(Table 2). Similar results were seen for dependence andpast month use (see Appendix S5; details are given at theend of the paper).

Continuous abstinence

A problem in interpreting the above result is that that theresidential rehabilitation group had higher rates of treat-ment exposure during the follow-up period (Fig. 2), andthis might account for their lower levels of metham-phetamine use. To isolate the impact of the index treat-ment episode, we used continuous abstinence (measuredat each follow-up) as the outcome measure. This was notaffected by additional treatment exposure during thefollow-up period because participants typically did notseek further treatment until they had relapsed to use. Asmall number of participants re-entered treatment

during the follow-up period despite being continuouslyabstinent (n = 19). We censored data from these partici-pants from the analysis at the point when they re-enteredtreatment because it would not be accurate to attributetheir subsequent outcomes to their index episode of resi-dential rehabilitation. Again, the detoxification andquasi-control groups were combined because thesegroups did not differ in their rates of continuous absti-nence (P > 0.05).

Unadjusted analyses showed that the residentialrehabilitation group had higher rates of continuousabstinence than the combined quasi-control and detoxi-fication comparison group at all three follow-up points(Table 2). Again, IPTW estimators reduced the magni-tude of these treatment effects (Table 2). Residentialrehabilitation produced a 33 percentage point increase inthe number of participants who remained abstinent at 3months compared to the quasi-control and detoxificationgroups combined, with this benefit dropping to 14 and 6percentage points at 1 and 3 years, respectively. Thesetreatment effects were significant at 3 months and 1 year,but not 3 years (Table 2).

DISCUSSION

Community-based residential rehabilitation produced alarge but time-limited reduction in methamphetamine

Figure 2 Exposure to drug treatmentover the follow-up period (excluding theindex treatment) by group: (a) the cumula-tive proportion of participants whoreceived any drug treatment during thefollow-up period; (b) the average cumula-tive number of treatment episodes initiatedper participant during the follow-up period



use. The largest gains were seen for abstinence at 3months after treatment, with a 33 percentage pointincrease in continuous abstinence compared to partici-pants who received no treatment or detoxification only.However, by 1–3 years after treatment, the vast majorityof residential rehabilitation clients reported similar meth-amphetamine use levels to the level that would beexpected if they had not received treatment or had onlyreceived detoxification. Detoxification alone did notchange methamphetamine use at any follow-up relativeto no treatment. These findings should be applied cau-tiously to locations where methamphetamine use pat-terns and drug treatment differ from those seen inAustralia.

These findings highlight the chronic relapsing natureof methamphetamine dependence and the need for atreatment approach with a more sustained impact.Although residential rehabilitation facilities provide astructured drug-free environment to initiate abstinence,this approach may not address factors that are likely totrigger relapse once clients re-enter the community (e.g.

cravings for the drug, socializing and living with drugusers, conflict and stress). Further research is also neededto determine whether the time-limited benefits of resi-dential rehabilitation outweigh their expensive runningcosts (which have been cited at US$11 016 per person inthe United States [27]). While existing research showsthat residential rehabilitation is cost-effective [28], thelack of control groups in previous studies means thatthey may overestimate the benefits of treatment [29].

Detoxification conveyed no benefit in reducing meth-amphetamine use at any follow-up relative to not receiv-ing treatment. This is consistent with previous research[4], and suggests that detoxification should not be pro-vided as a stand-alone service. We found that mostdetoxification clients were highly motivated to reducetheir methamphetamine use and sought abstinence,indicating a need to educate detoxification clients thataddressing physical withdrawal symptoms may not alle-viate methamphetamine dependence in the longer term,and that further treatment is needed to address thebroader psychosocial issues (e.g. coping and interper-

Figure 3 Frequency of methampheta-mine use reported at each follow-up bygroup



Tabl

e2

Effe

ctof

resi

den

tial

reh

abili

tati

onon

freq

uen

cyof

met

ham

phet

amin

eu

sean

dco

nti

nu

ous

abst

inen

cefr

omm

eth

amph

etam

ine.

Una

djus

ted

OR

(95

%C

I)P

-val

ue

Pre

dict

edva

lues

(%)

IPTW

OR

(95

%C

I)P

-val

ue

Pre

dict

edva

lues

(%)

Con

trol

RR

Abs

olut

ech

ange

Con

trol

RR

Abs

olut

ech

ange

Freq

uen

cyof

use

3m

onth

s0

.21

(0.1

4–0

.32

)a<0

.00

10

.23

(0.1

5–0

.36

)a<0

.00

1N

ou

se1

85

73

91

85

33

4Le

ssth

anw

eekl

y3

42

4-1

03

62

9-7

1–2

days

/wee

k2

71

2-1

42

41

1-1

33

+da

yspe

rw

eek

22

8-1

42

27

-15

1ye

arN

ou

se0

.56

(0.3

8–0

.81

)0

.00

22

03

11

10

.62

(0.4

0–0

.97

)0

.03

82

13

09

Less

than

wee

kly

41

43

24

44

51

1–2

days

/wee

k1

91

4-5

16

12

-43

+da

yspe

rw

eek

20

12

-81

91

3-6

3ye

ars

No

use

0.6

2(0

.42

–0.9

1)

0.0

16

20

29

90

.71

(0.4

2–1

.19

)0

.18

92

02

66

Less

than

wee

kly

47

48

14

95

01

1–2

days

/wee

k1

91

4-5

17

14

-33

+da

yspe

rw

eek

14

9-5

14

11

-4C

onti

nu

ous

abst

inen

ce3

mon

ths

6.3

(3.8

–10

.4)

<0.0

01

15

53

38

5.1

(2.8

–9.3

)<0

.00

11

54

83

31

year

4.2

(1.9

–9.7

)0

.00

16

22

16

3.5

(1.3

–9.8

)0

.01

67

20

14

3ye

ars

3.4

(1.2

–9.4

)0

.00

74

13

92

.4(0

.7–7

.6)

0.1

47

51

26

Con

trol

:qu

asi-

con

trol

and

deto

xific

atio

ngr

oups

com

bin

ed;R

R:r

esid

enti

alre

hab

ilita

tion

grou

p.a O

ddso

fch

ange

diffe

red

acro

ssfr

equ

ency

ofm

eth

amph

etam

ine

use

.Un

adju

sted

resu

lts:

any

use

vers

usa

bsti

nen

ceod

dsra

tio

(OR

)0.1

7,

95

%co

nfid

ence

inte

rval

(CI)

0.1

0–0

.28

;mor

eth

anw

eekl

yve

rsu

sles

sth

anw

eekl

yO

R0

.26

,95

%C

I0.1

7–0

.42

;3+

days

vers

usu

pto

1–2

days

perw

eek

OR

0.2

9,9

5%

CI0

.14

–0.5

8.I

nver

sepr

obab

ility

oftr

eatm

ent-

wei

ghte

d(I

PT

W)

resu

lts:

any

use

vers

us

abst

inen

ceO

R0

.20

,95

%C

I0

.11

–0.3

7;m

ore

than

wee

kly

vers

us

less

than

wee

kly

OR

0.2

6,9

5%

CI

0.1

6–0

.44

;3+

days

vers

us

up

to1

–2da

yspe

rw

eek

OR

0.2

8,9

5%

CI

0.1

3–0

.60

.



sonal skills), as well as ongoing cravings for the drug, andtheir role in precipitating relapse [30,31].

The time-limited benefit of residential rehabilitation isinconsistent with previous treatment outcome studies,which suggest more sustained reductions in drug use[4–6]. This discrepancy largely reflects the use of a quasi-control group in the current study, with earlier studiesexamining pre-treatment versus post-treatment changesin drug use. While we also saw sustained drops inmethamphetamine use after treatment relative to pre-treatment levels, similar reductions occurred in thequasi-control condition, suggesting that they were attrib-utable to factors other than the index treatment (e.g.maturation out of methamphetamine dependence, theimpact of the study procedure itself, and reductions inmethamphetamine availability that occurred over thestudy period [32]). IPTW estimators further attenuatedtreatment effects, suggesting that it was partly the char-acteristics of clients who entered residential rehabilita-tion (e.g. high levels of motivation) that drove post-treatment improvements. Our findings highlight the needfor quasi-control groups in observational treatmentoutcome studies, and the importance of using IPTW orother matching procedures to derive treatment effects.

Methodological considerations

The capacity of IPTW estimators to produce unbiasedtreatment effects assumes that there are no unmeasuredconfounders [14,33]. We were able to adjust for treat-ment history, severity of methamphetamine use, motiva-tion, psychiatric disorders and basic demographics, butthere were undoubtedly factors that we did not measurethat may have impacted upon treatment outcomes. Theapplication of IPTW estimators also provides treatmenteffects that would arise should both the treatment and thenon-treatment group be provided with treatment. Thisapproach may underestimate the benefit of treatment asit applies to those people who are inclined to seek treat-ment (e.g. highly motivated clients), rather than the non-treatment-seeking control group used herein.

Self-reported methamphetamine use in the pastmonth was validated using hair toxicology in a subgroupof the sample and found to be accurate in the vast major-ity of cases. This is consistent with previous evidence thatself-report is generally found to be a reliable and validindicator of drug use [34]. Having said this, there were anumber of limitations in our validation procedures: toxi-cology could not be obtained on a random sample of par-ticipants; variability in the rate of hair growth can affectthe validity of results; and the concentrations of meth-amphetamine in hair are affected by exposure to deter-gents (hair washing) and other environmental factors[35,36].

Treatment outcomes reported herein are contingenton the follow-up times, which were often delayed due tothe difficulty in locating participants, particularly for the3-year follow-up. This may have attenuated treatmenteffects in so far as these treatment effects decayed overtime. Low rates of continuous abstinence at the 3-yearfollow-up also meant that we may not have had enoughstatistical power to detect a significant effect. Imputationof missing data would have reduced bias due to the highattrition rate at 3 years, but this procedure cannot elimi-nate bias when attrition is due to client outcomes(e.g. clients being lost to follow-up because they haverelapsed) [26].

CONCLUSION

Community-based drug residential rehabilitation pro-duced large short-term reductions in methamphetamineuse relative to no treatment or detoxification alone, butthere was no clear evidence that it conveyed a benefitat 3 years after starting treatment. Detoxification alonedid not reduce methamphetamine use relative to notreatment. Improved treatment approaches are neededto produce long-term reductions in methamphetamineuse.

Declarations of interest

DL has received speaker honoraria from Astra Zeneca andJanssen. The other authors have no interests to declare.

Acknowledgements

The data reported in this paper were collected throughthe Methamphetamine Treatment Evaluation Study(MATES), which was conducted by the National Drug andAlcohol Research Centre, University of New South Wales,and was funded by the National Health and MedicalResearch Council and the Australian GovernmentDepartment of Health and Ageing. The authors thankJoanne Ross, who was an investigator on the project,researchers working on the project (Erin Kelly, ShelleyCogger, Rachel Sutherland, Grace Ho, Cathie Sammut,Kate Hetherington, Sagari Sarkar, Julia Rosenfeld andMiriam Wyzenbeek), the participating treatment agen-cies and health services and the participants.

References

1. Nicosia N., Liccardo Pacula R., Kilmer B., Lundberg R.,Chiesa J. The Economic Cost of Methamphetamine Use in theUnited States, 2005. Los Angeles, CA: RAND Drug PolicyResearch Center, Rand Corporation; 2009.

2. Colfax G., Santos G.-M., Chu P., Vittinghoff E., PluddemannA., Kumar S. et al. Amphetamine-group substances andHIV. Lancet 2010; 376: 458–74.



3. Rawson R. A., Marinelli-Casey P., Anglin M. D., Dickow A.,Frazier Y., Gallagher C. et al. A multi-site comparisonof psychosocial approaches for the treatment ofmethamphetamine dependence. Addiction 2004; 99:708–17.

4. Teesson M., Mills K., Ross J., Darke S., Williamson A.,Havard A. The impact of treatment on 3 years’ outcome forheroin dependence: findings from the Australian TreatmentOutcome Study (ATOS). Addiction 2008; 103: 80–8.

5. Gossop M., Marsden J., Stewart D., Kidd T. The NationalTreatment Outcome Research Study (NTORS): 4–5 years’follow-up results. Addiction 2003; 98: 291–303.

6. Hubbard R. L., Craddock S. G., Anderson J. Overview of5-year followup outcomes in the drug abuse treatmentoutcome studies (DATOS). J Subst Abuse Treat 2003; 25:125–34.

7. Zweben J. E., Cohen J. B., Christian D., Galloway G. P., Sali-nardi M., Parent D. et al. Psychiatric symptoms in metham-phetamine users. Am J Addict 2004; 13: 181–90.

8. Glasner-Edwards S., Marinelli-Casey P., Hillhouse M., AngA., Mooney L. J., Rawson R. for the MethamphetamineTreatment Project. Depression among methamphetamineusers: association with outcomes from the Methampheta-mine Treatment Project at 3-year follow-up. J Nerv Ment Dis2009; 197: 225–31.

9. Glasner-Edwards S., Mooney L. J., Marinelli-Casey P., Hill-house M., Ang A., Rawson R. for the MethamphetamineTreatment Project. Clinical course and outcomes ofmethamphetamine-dependent adults with psychosis. JSubst Abuse Treat 2008; 35: 445–50.

10. Glasner-Edwards S., Mooney L. J., Marinelli-Casey P.,Hillhouse M., Ang A., Rawson R. for the Methamphe-tamine Treatment Project. Anxiety disorders amongmethamphetamine dependent adults: association withpost-treatment functioning. Am J Addict 2010; 19:385–90.

11. Wright S., Klee H., Reid P. Attitudes of amphetamine userstowards treatment services. Drugs Educ Prev Policy 1999; 6:71–86.

12. United Nations Office on Drugs and Crime. World DrugReport 2010. New York: United Nations; 2010.

13. Höfler M. Causal inference based on counterfactuals. BMCMed Res Methodol 2005; 5: 28.

14. Robins J., Hernan M., Brumback B. Marginal structuralmodels and causal inference in epidemiology. Epidemiology2000; 11: 550–60.

15. Curtis L. H., Hammill B. G., Eisenstein E. L., Kramer J. M.,Using A. K. J. Inverse probability-weighted estimators incomparative effectiveness analyses with observational data-bases. Med Care 2007; 45: S103–7.

16. Lipkovich I., Adams D. H., Mallinckrodt C., Faries D., BaronD., Houston J. P. Evaluating dose response from flexible doseclinical trials. BMC Psychiatry 2008; 8: 3.

17. Lee S.-W., Park S.-W., Yun S.-C., Kim Y.-H., Park D.-W., KimW.-J. et al. Triple antiplatelet therapy reduces ischemicevents after drug-eluting stent implantation: drug-elutingstenting followed by Cilostazol treatment REduces AdverseSerious cardiac Events (DECREASE registry). Am Heart J2010; 159: 284–91. e281.

18. Nishida Y., Takahashi Y., Nakayama T., Soma M., KitamuraN., Asai S. Effect of candesartan monotherapy on lipidmetabolism in patients with hypertension: a retrospectivelongitudinal survey using data from electronic medicalrecords. Cardiovasc Diabetol 2010; 9: 38.

19. Australian Institute of Health and Welfare. Alcoholand Other Drug Treatment Services National MinimumData Set 2011–12 Specifications and Collection Manual.Canberra: Australian Institute of Health and Welfare;2011, p. 1–183.

20. Gossop M., Darke S., Griffiths P., Hando J., Powis B., Hall W.et al. The Severity of Dependence Scale (SDS) in English andAustralian samples of heroin, cocaine and amphetamineusers. Addiction 1995; 90: 607–14.

21. Topp L., Mattick R. P. Choosing a cut-off on the Severity ofDependence Scale (SDS) for amphetamine users. Addiction1997; 92: 839–45.

22. Darke S., Hall W., Wodak A., Heather N., Ward J. Develop-ment and validation of a multi-dimensional instrumentfor assessing outcome of treatment among opiate users:the Opiate Treatment Index. Br J Addict 1992; 87:733–42.

23. Rollnick S., Heather N., Gold R., Hall W. Development of ashort ‘readiness to change’ questionnaire for use in brief,opportunistic interventions among excessive drinkers. B JAddict 1992; 87: 743–54.

24. Andrews G., Peters L. The psychometric properties of theComposite International Diagnostic Interview. Soc Psychia-try Psychiatr Epidemiol 1998; 33: 80–8.

25. Robins L. N., Helzer J. E., Ratcliff K. S., Seyfried W. Validity ofthe diagnostic interview schedule, version II: DSM-III diag-noses. Psychol Med 1982; 12: 855–70.

26. Royston P. Multiple imputation of missing values: update.Stata J 2005; 5: 1–14.

27. Flynn P. M., Kristiansen P. L., Porto J. V., Hubbard R. L. Costsand benefits of treatment for cocaine addiction in DATOS.Drug Alcohol Depend 1999; 57: 167–74.

28. Godfrey C., Stewart D., Gossop M. Economic analysis ofcosts and consequences of the treatment of drug misuse:2-year outcome data from the National Treatment OutcomeResearch Study (NTORS). Addiction 2004; 99: 697–707.

29. Cartwright W. S. Cost–benefit analysis of drug treatmentservices: review of the literature. J Ment Health Policy Econ2000; 3: 11–26.

30. Hartz D. T., Frederick-Osborne S. L., Galloway G. P. Cravingpredicts use during treatment for methamphetaminedependence: a prospective, repeated-measures, within-subject analysis. Drug Alcohol Depend 2001; 63: 269–76.

31. Brecht M. L., von Mayrhauser C., Anglin M. D. Predictors ofrelapse after treatment for methamphetamine use. J Psycho-act Drugs 2000; 32: 211–20.

32. Phillips B., Burns L. NSW Drug Trends 2009: Findings fromthe Illicit Drug Reporting System (IDRS). Australian DrugTrend Series no. 38. Sydney (Australia): National Drug andAlcohol Research Centre, University of New South Wales;2010.

33. Austin P. C. The performance of different propensity-scoremethods for estimating differences in proportions (risk dif-ferences or absolute risk reductions) in observationalstudies. Stat Med 2010; 29: 2137–48.

34. Darke S. Self-report among injecting drug users: a review.Drug Alcohol Depend 1998; 51: 253–63.

35. Miyaguchi H., Kakuta M., Iwata Y. T., Matsuda H., TazawaH., Kimura H. et al. Development of a micropulverizedextraction method for rapid toxicological analysis ofmethamphetamine in hair. J Chromatogr 2007; 1163:43–8.

36. Wennig R. Potential problems with the interpretation ofhair analysis results. Forensic Sci Int 2000; 107: 5–12.



Supporting information

Additional Supporting Information may be found in theonline version of this article:

Appendix S1. A comparison of treatment participantswith methamphetamine treatment clients in AustraliaAppendix S2. Details of hair toxicologyAppendix S3. Derivation of IPTW estimatorsAppendix S4. Predictors of attrition and data imputa-tion procedures

Appendix S5. Treatment effects for methampheta-mine dependence and past month methampheta-mine use

Please note: Wiley-Blackwell are not responsible for thecontent or functionality of any supporting materials sup-plied by the authors. Any queries (other than missingmaterial) should be directed to the corresponding authorfor the article.



Treatment outcomes for methamphetamine use: Finding from the methamphetamine treatment evaluation study (MATES)

Documents