Approach With Caution: A Pilot Randomised Controlled Trial of Approach Bias Modification for People Undergoing Residential Treatment for Methamphetamine Use Disorder

ABSTRACT

Introduction

Less than one‐quarter of people who attend residential rehabilitation for methamphetamine use disorder (MUD) remain abstinent for 1 year. Approach bias modification (ApBM), a cognitive training intervention, reduces relapse rates when delivered during residential rehabilitation for alcohol use disorder. However, there has been little research on ApBM for MUD.

Methods

Twenty‐four patients with MUD attending four residential treatment services in Melbourne were randomised, between March 2020 and August 2024, to 6 sessions of either ApBM (n = 10) or sham‐control training (n = 14). Safety was monitored during the intervention and acceptability was assessed at the end of training. A blinded researcher assessed methamphetamine use, craving, and dependence severity at follow‐ups 1 and 3 months after discharge from residential treatment.

Results

Three ApBM and five control participants experienced adverse psychological reactions to the training task, resulting in two ApBM and 1 control participants withdrawing. However, participants who completed training generally provided positive acceptability ratings (mean Endorsement and Discomfort Scale scores of 69.13 [SD = 10.93] in the ApBM group and 58.83 [SD = 13.41] in controls). Directions of differences in proportions reporting past‐month abstinence at 1‐month (ApBM: 5/7 [71%]; controls: 5/11 [46%]) and 3‐month follow‐ups (ApBM: 4/7 [57%]; controls: 3/9 [33%]) and continuous abstinence at 3‐month follow‐up (ApBM: 3/7 [43%]; controls: 2/11 [18%]) favoured ApBM; however, these differences did not approach significance in this very small sample. Craving and MUD severity outcomes also did not differ significantly between groups.

Discussion and Conclusions

Further research on ApBM's efficacy for MUD may be warranted, but safety concerns suggest caution.

Trial Registration: Australian New Zealand Clinical Trials Registry ACTRN12620000072910. Registered on 30 January 2020 (prospectively registered): https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=378804&isReview=true.

Summary

• We tested approach bias modification for methamphetamine use disorder.

• In a small pilot, abstinence rates from amphetamines at follow‐up were encouraging.

• Safety data was concerning: one‐third of participants had adverse psychological reactions.

1. Introduction

After alcohol, amphetamines are the second‐most common primary drug of concern among individuals seeking alcohol and other drug (AOD) treatment in Australia, where methamphetamine use disorder (MUD) accounts for 40% of residential rehabilitation treatment episodes [1]. Less than one‐quarter of MUD clients remain abstinent for at least 1 year after commencing residential rehabilitation treatment [2]. Approach bias modification (ApBM) is a cognitive training intervention designed to reduce “approach bias” – the automatic motor tendency to approach substance‐related stimuli [3]. ApBM (also known as approach‐avoidance training) is a computerised task where participants are presented with substance‐related images, to which they repeatedly make an “avoidance” movement (e.g., pushing a joystick away) to “shrink” a substance‐related picture; and repeatedly make an “approach” movement (e.g., pulling a joystick toward themselves) to “enlarge” neutral/positive non‐substance stimuli.

Several randomised controlled trials (RCT) have shown that when administered to people with alcohol dependence during residential withdrawal [4, 5, 6] or rehabilitation treatment [7, 8, 9, 10, 11], ApBM significantly reduces relapse rates for up to one year post‐discharge [12]. However, in contrast to alcohol, RCTs of ApBM for people seeking to reduce or cease cigarette use have mostly reported null effects on tobacco use [13, 14, 15, 16, 17, 18], with some exceptions [19, 20, 21, 22]. Small pilot studies on cannabis have reported positive signals but remain inconclusive [23, 24, 25]. With only two studies of ApBM for MUD to date, there remains a need to establish its effectiveness as an adjunctive treatment for people seeking methamphetamine treatment.

Our open‐label pilot study [26] of ApBM during residential methamphetamine withdrawal treatment (typically 7–10 days duration) found promising abstinence rates at follow‐ups, and good acceptability in participants who completed the 4‐session protocol. However, feasibility and safety concerns arose, with 38% of participants failing to complete all 4 sessions, frequent increases in craving following ApBM sessions, and 9% withdrawing due to finding ApBM “triggering” or distressing. We concluded that delivering ApBM during residential rehabilitation, a more stable treatment environment, after the first week of acute withdrawal symptoms has subsided, may mitigate these challenges. The only other published MUD ApBM study is an RCT (N = 136) in a community‐based rehabilitation centre in China, which found a significant group × time interaction for cue‐induced craving scores, whereby adaptive ApBM (where difficulty was dynamically adjusted according to performance) decreased craving more than both standard ApBM and a no‐intervention control group post‐intervention and at a 12‐week follow‐up [27]. However, since no participants reported relapse during the follow‐up period, its impact on methamphetamine use could not be established.

ApBM has typically used fixed image sets whereby all participants are trained to avoid the same drug stimuli (e.g., alcoholic beverage types/brands) and approach the same neutral stimuli (e.g., non‐alcoholic drinks or neutral stimuli). However, since approach bias develops through repeated associative learning, it is likely to be specific to drug cues that closely resemble the drug type and paraphernalia used by the individual. For example, being trained to avoid images of syringes may have minimal impact for someone who only smokes methamphetamine. Thus, use of individually‐tailored image sets could increase engagement in, and the effectiveness of, ApBM training. As such, for the present pilot trial, we developed a novel “personalised” form of ApBM training that allowed participants to select personally‐relevant drug‐related images (e.g., representing the paraphernalia and types of methamphetamine they typically used), which they were trained to avoid; and personally meaningful, goal‐relevant images (e.g., representing healthy activities or values) to approach [28]. We aimed to collect pilot data on whether personalised ApBM increased rates of abstinence from methamphetamine 1 and 3 months after discharge from rehabilitation, relative to a sham‐training control group. Secondary outcomes included craving, severity of MUD, time to first lapse to illicit amphetamine use, acceptability, and safety.

2. Methods

2.1. Design

We conducted a pilot randomised, double‐blind, sham‐controlled, parallel group trial.

2.2. Setting and Participants

We recruited participants from four substance use treatment services (one public, 3 private) in metropolitan Melbourne, Australia. These services typically delivered 1–2 month residential treatment programmes. All sites' programmes involved medical management of withdrawal symptoms; predominantly group therapy, supplemented by individual care; assessment and treatment of psychiatric comorbidities; referrals for post‐discharge outpatient care; and all offered optional post‐discharge day programmes and/or peer support groups. Group and individual interventions varied but, depending on the site, could include psychoeducation, cognitive therapies (e.g., CBT, ACT, DBT), mindfulness, emotion‐focused therapy, family/relationship therapy, process group therapy, psychodynamic therapy, art or music therapy, 12‐step facilitation, physical exercise, relapse prevention, financial counselling, family violence risk assessment, and/or pharmacotherapy. Depending on the site, individual care was provided by psychiatrists, addiction medicine specialists, nurses, peer support workers, general practitioners, and/or psychotherapists.

Eligibility criteria required participants: to be aged at least 18 years; to meet criteria for current moderate or severe methamphetamine use disorder during the past 3 months, according to Diagnostic and Statistical Manual (DSM‐5) criteria (i.e., 4 or more symptoms present in past 3 months); to have used methamphetamine on at least 4 of the 28 days prior to commencing residential treatment; to have sufficient English language proficiency to understand the participant information sheet, questionnaires, and intervention task instructions; to be able to provide a phone number for follow‐ups; and, at the time of recruitment, to be intending to stay in residential treatment long enough to complete the 6‐session training protocol. Clinical staff were instructed not to refer clients to the research team if, according to their judgement, the client was too physically or mentally impaired to provide informed consent or safely participate in the intervention.

A total of 24 patients with MUD were randomised. Since the intention of this pilot trial was to establish effect size estimates for future larger trials, the intended sample size of 100 was not based on an a priori power calculation, but rather a pragmatic estimate of likely recruitment. A total of 100 was considered appropriate, given we had previously observed significant group differences in relapse rates in our pilot ApBM RCT with only 83 participants undergoing alcohol withdrawal treatment [6]. However, with prolonged COVID‐19 pandemic‐related restrictions preventing researchers from attending recruitment sites, the early withdrawal of one recruitment site, and low rates of patients meeting eligibility criteria at the remaining sites, a smaller sample size was achieved when funding terminated. The study was approved by the Eastern Health Human Research Ethics Committee (project number: E19/022/58264), Monash University Human Research Ethics Committee (project number: 23080), Ramsay Health Research Governance (approval number: RHC RG–2020.021), and The Melbourne Clinic Human Research Ethics Committee (project number: 342).

2.3. Measures

2.3.1. Demographic and Clinical Characteristics

Researchers administered a structured questionnaire assessing demographic characteristics, history of methamphetamine use and MUD treatment, other drugs of concern, and psychiatric diagnoses.

2.3.2. MUD Severity

The MUD module of the Structured Clinical Interview for DSM‐5 Disorders–Research Version (SCID–5–RV) [29] confirmed eligibility at baseline and current MUD at 3‐month follow‐up. Questions were modified to refer to the past 3 months, rather than the past year, to allow assessment of equivalent non‐overlapping periods at these 2 timepoints. The Severity of Dependence Scale (SDS) [30] assessed severity of past‐month psychological dependence on methamphetamine. At follow‐ups, if participants reported no past‐month use, the SDS was not administered and a score of zero was assigned.

2.3.3. Substance Use

A timeline follow‐back (TLFB) interview [31] measured frequency and quantity of methamphetamine and other substance use (including non‐prescribed use of other amphetamine‐type stimulants) during: the 28 days preceding admission (baseline); the 28 days beginning with the date of discharge (1‐month follow‐up); and the past 28 days (3‐month follow‐up).

2.3.4. Methamphetamine Craving

The Craving Experience Questionnaire (CEQ) [32] measured frequency of methamphetamine craving (frequency scale) and strength of the strongest craving experienced (strength scale) in the past week. Additionally, a single‐item visual analogue scale (VAS) of craving intensity was administered before and after each training session for safety monitoring.

2.3.5. Task Acceptability

Participants self‐administered the Endorsement and Discomfort Scale (EDS) [33] to measure acceptability of the training. Scoring of the EDS followed methods described by Milosevic et al. [34].

2.3.6. Blinding

Blinding was assessed by asking participants whether they believed they had received “a form of training that was intended to reduce impulses to use methamphetamine” or “a ‘placebo’ version of the training, which was not designed to change impulses”.

2.4. Intervention

2.4.1. ApBM

Prior to their first training session, each participant viewed and rated 40 methamphetamine‐related images and 40 positive images (e.g., depicting people socialising; financial security; employment; exercise, sports, and recreational activities; healthy foods; pets). The methamphetamine and positive image sets had been co‐designed prior to commencing the trial with a focus group comprised of people with lived experience of treatment for MUD. Participants in the present trial rated each methamphetamine image, using a VAS in response to the question: “How much does this image remind you of the times you've used methamphetamine?”; and rated each positive image using a VAS in response to the question, “How much does this image remind you of your goals, things you enjoy, or things you would like to spend more time doing?”. The 10 most highly‐rated images from each category were then used in the ApBM task—hence each participant's training image set was personalised. Each session comprised 12 presentations of each of the 20 images (240 trials total) on a computer screen. Participants used a joystick to respond, with instructions to “push” away images with landscape‐oriented frames (causing them to shrink) and “pull” portrait‐oriented images toward themselves (causing them to enlarge). A total of 95% of landscape‐framed images were methamphetamine‐related (and 5% were positive), while 95% of portrait‐framed images were positive (and 5% were methamphetamine‐related). Hence, instructions required participants to ‘avoid’ 95% of methamphetamine images and ‘approach’ 95% of positive images. Further task parameter details are described by Garfield et al. [28].

2.4.2. Sham Training

Sham training resembled ApBM, except that: (i) training was not personalised, with participants responding to all 40 methamphetamine and 40 positive images (presented 3 times each per session); and (ii) both methamphetamine‐related and positive images each comprised 50% of portrait‐oriented and 50% of landscape‐oriented images. As a result, participants were trained to approach and avoid both methamphetamine and positive images an equal number of times, such that there was no systematic avoidance or approach training to either image category. Further sham training details are described by Garfield et al. [28].

2.5. Randomisation and Blinding

Randomisation occurred automatically when participants commenced their first training session. Separate 1:1 allocation sequences, using blocks of six, were computer‐generated for each site and pre‐programmed into each site's training computer by a programmer who was not involved in recruitment or data collection. Hence, randomisation was stratified by site. Participants were assigned sequential numbers as they were recruited. Entering a participant's number into the ApBM programme at the start of each session caused the programme to load either the personalised ApBM or the sham training, depending on the number's allocation in the randomisation sequence. Research staff administering the training did not have access to the randomisation sequences and hence were blinded to allocation prior to randomisation, although they were no longer blinded once training commenced since they could see the training task. However, they were instructed not to tell the participant which condition they were in, and participant information did not include training details that would allow participants to discern their condition.

2.6. Procedure

Clinicians screened patients for likely eligibility during the first few days following admission to the treatment service and referred them to research staff if they were interested. Approximately 1 week after admission, a researcher met with interested patients in person to provide trial information. If the patient provided written informed consent, the researcher administered demographics questions, SCID‐5‐RV, TLFB, and collected contact details for follow‐ups, and the participant self‐administered the SDS and CEQ.

Researchers aimed to administer three training sessions per week for 2 weeks (6 sessions total), with flexibility in session frequency where necessary. The craving VAS was assessed before and after each session and, if participants reported high craving or distress, the researcher reminded them they could withdraw, offered referral to clinical staff for support, and recorded adverse event details. Following the 6th session, participants completed the EDS and received a A$60 supermarket voucher.

A researcher who was not involved in the participant's training sessions, and therefore blind to treatment condition, conducted follow‐ups by phone 1 month and 3 months post‐discharge from the residential treatment service, administering the TLFB, CEQ, and (if the participant had used methamphetamine) the SDS. The 3‐month follow‐up also included the SCID‐5‐RV, condition blinding check, and date of first illicit amphetamine use (if applicable). Participants received $20 supermarket vouchers for each follow‐up completed.

2.7. Statistical Analyses

Analyses were conducted in SPSS version 29.0.2.0. Due to the small sample size, we primarily focus on descriptive data for outcomes, however, inferential tests for between‐group comparisons included: 2‐tailed t‐tests for EDS and CEQ scores; Fisher's exact test for dichotomous outcomes (abstinence, MUD, proportion guessing they received the active intervention); a Kaplan–Meier test for time to first illicit amphetamine use after discharge; and Mann–Whitney U tests for SDS scores, number of days of past‐month methamphetamine use, and grams of methamphetamine used in the past month at each follow‐up. For analyses of the ‘abstinence’ and ‘time to first use’ outcomes, we considered any non‐prescribed use of amphetamine‐type stimulants (e.g., methamphetamine, non‐prescribed pharmaceutical amphetamines, etc.) to constitute a “lapse”.

3. Results

Participants were randomised between 3 March 2020 and 20 August 2024. The final outcome follow‐up assessment was completed on 29 November 2024. The number of patients screened, recruited, randomised, who completed training, and who provided data are shown in Figure 1. Demographic and clinical characteristics of the sample are shown in Table 1. All but one participant was born in Australia, and none were Aboriginal or Torres Strait Islanders. Among participants who completed the 1‐month follow‐up, the mean time from discharge to follow‐up was 33.50 days (range: 28–48) in the ApBM group and 32.18 days (range: 28–41) in controls. Three‐month follow‐ups were completed a mean of 89.57 days after discharge (range: 84–106) in the ApBM group and 96.40 days (range: 84–141) in controls.

FIGURE 1.

Number of patients screened, referred, recruited, randomised, withdrawn, completing training and providing data at each follow‐up. (*Note that, although only 6 participants in the ApBM group completed 1‐month follow‐up, amphetamine abstinence status for a 7th participant during the first 4 weeks post‐discharge was ascertained by retrospective self‐report when they completed the 3‐month follow‐up).

TABLE 1.

Demographic, clinical and treatment characteristics of participants at baseline.

	Whole sample (N = 24)	ApBM (n = 10)	Sham (n = 14)
Age: mean (SD)	31.82 (7.89)	27.09 (3.41)	35.20 (8.51)
Male gender: n (%)	15 (62)	6 (60)	9 (64)
Employment: n (%)
Full‐time	5 (21)	2 (20)	3 (21)
Part‐time	2 (8)	1 (10)	1 (7)
Unemployed	15 (62)	7 (70)	8 (57)
Disability or sickness benefits	1 (4)	0	1 (7)
Carer or home duties	1 (4)	0	1 (7)
Age of first illicit amphetamine use: mean (SD)	18.71 (4.18)	17.10 (1.66)	19.86 (5.05)
Days on which methamphetamine was used in 4 weeks prior to admission: mean (SD)	21.25 (6.99)	20.50 (7.50)	21.79 (6.84)
Grams of methamphetamine used in 4 weeks prior to admission: mean (SD)	17.01 (15.17)	23.40 (20.08)	12.44 (8.59)
Main route of methamphetamine administration: n (%) a
Smoking	19 (86)	8 (89)	11 (85)
Injecting	3 (14)	1 (11)	2 (15)
Severity of Dependence Scale score: mean (SD)	9.83 (2.62)	10.50 (2.76)	9.36 (2.50)
CEQ – Frequency score: mean (SD)	4.38 (2.98)	5.42 (2.77)	3.64 (2.99)
CEQ – Strength score: mean (SD)	5.07 (2.95)	6.22 (2.37)	4.24 (3.13)
Has additional drugs of concern (other than amphetamines) b : n (%)	18 (75)	8 (80)	10 (71)
Using tobacco or nicotine vapour daily: n (%)	22 (92)	10 (100)	12 (86)
Any previous residential addiction treatment: n (%)	14 (58)	5 (50)	9 (64)
Any current non‐substance psychiatric diagnosis: n (%)	18 (75)	7 (70)	11 (79)
Days from admission to 1st training session: mean (range)	14.88 (6–48)	16.00 (6–48)	14.07 (6–28)
Duration of residential admission (days): mean (range) c	33.00 (7–76)	40.88 (28–63)	28.15 (7–76)

Abbreviations: ApBM, approach bias modification group; CEQ, Craving Experience Questionnaire; SD, standard deviation.

^a N = 22 due to missing data.

^b Additional drugs of concern included GHB (n = 12), cannabis (n = 7), alcohol (n = 5), opioids (n = 3), alprazolam (n = 2), ketamine (n = 1), and pregabalin (n = 1). Amphetamines were the primary drug of concern for all but 3 participants (GHB = 2; alcohol = 1).

^c N = 21 due to missing information on some participants who withdrew consent.

3.1. Abstinence From Illicit Amphetamines at Follow‐Up

At 4‐week follow‐up, abstinence from illicit amphetamines was reported by 5 of the 7 participants (71%) in the ApBM group, and 5 of 11 (45%) controls, whose abstinence status was measured (Fisher's exact test p = 0.367). Applying a conservative “intention‐to‐treat” (ITT) assumption that all participants lost to follow‐up were non‐abstinent, these proportions would be 5/10 (50%) in the ApBM group and 5/14 (36%) in controls (p = 0.678). At the 3‐month follow‐up, past‐month abstinence was reported by 4 of the 7 ApBM participants (57%), and 3 of 9 controls (33%) (p = 0.615). Following the ITT assumption, these proportions would be 40% for the ApBM group and 21% for controls (p = 0.393). Continuous abstinence since discharge at the 3‐month follow‐up was reported by 3 of the 7 (43% observed, 30% ITT) ApBM participants, and 2 of the 11 (18% observed, 14% ITT) controls in whom this outcome was assessed (p = 0.326 observed; p = 0.615 ITT). Survival functions for time to first lapse are shown in Figure 2. Median time to first lapse after discharge was 35 days in the ApBM group (95% confidence interval 19–50) and 14 days in controls (95% confidence interval 0–55), with no significant difference between group's survival functions (log‐rank test χ²(1) = 1.43, p = 0.232).

FIGURE 2.

Proportion of participants in each group maintaining continuous abstinence since discharge as a function of days since discharge.

3.2. MUD Severity, Craving, and Methamphetamine Use at Follow‐Up

Of the participants who completed the SCID‐5‐RV at 3‐month follow‐up, 3 of the 7 ApBM participants (43% observed, 60% ITT) met current MUD criteria within the past 3 months, as did 4/10 controls (40% observed, 57% ITT) (ps > 0.999). Tables 2 and 3 show statistics at 1‐month and 3‐month follow‐ups, respectively, for: SDS scores; number of days methamphetamine was used; and total grams used (in the first 28 days following discharge, for 1‐month follow‐up; in the past 28 days at 3‐month follow‐up); and CEQ scores.

TABLE 2.

Severity of methamphetamine dependence, quantity of use and craving at the 1‐month follow‐up.

Outcome	ApBM	Control	Test statistic	p
SDS score: median (IQR)	0 (0–0.5)	1 (0–5)	U = 1.497	0.220
Number of days methamphetamine used: median (IQR)	0 (0–5)	1 (0–6)	U = 1.182	0.301
Grams of methamphetamine used: median (IQR)	0 (0–0.76)	0.05 (0–1.23)	U = −0.935	0.428
CEQ Frequency: mean (SD)	2.25 (2.52)	2.06 (2.04)	t(14) = −0.165	0.871
CEQ Strength: mean (SD)	3.68 (3.08)	3.05 (2.64)	t(14) = −0.438	0.668

Note: Test statistic is the Mann–Whitney U standardised test statistic for variables where median (IQR) are reported, and t value for variables where mean (SD) are reported.

Abbreviations: CEQ, Craving Experience Questionnaire; IQR, interquartile range; SD, standard deviation; SDS, Severity of Dependence Scale.

TABLE 3.

Severity of methamphetamine dependence, quantity of use and craving at the 3‐month follow‐up.

Outcome	ApBM	Control	Test statistic	p
SDS score: median (IQR)	0 (0–12)	2.5 (0–9)	U = 0.052	> 0.999
Number of days methamphetamine used: median (IQR)	0 (0–20)	1.5 (0–14)	U = −0.413	0.740
Grams of methamphetamine used: median (IQR)	0 (0–8.75)	0.15 (0–5.05)	U = 0.170	0.918
CEQ Frequency: mean (SD)	2.94 (3.05)	3.06 (2.93)	t(14) = 0.075	0.941
CEQ Strength: mean (SD)	3.43 (3.15)	3.44 (3.31)	t(14) = 0.010	0.992

Note: Test statistic is the Mann–Whitney U standardised test statistic for variables where median (IQR) are reported, and t value for variables where mean (SD) are reported.

Abbreviations: CEQ, Craving Experience Questionnaire; IQR, interquartile range; SD, standard deviation; SDS: Severity of Dependence Scale.

3.3. Acceptability

Among the participants who completed 6 training sessions, mean EDS scores were 69.13 (SD = 10.93) in the ApBM group and 58.83 (SD = 13.41) in controls (t(12) = −1.58, p = 0.139, Cohen's D = 0.86). All participants' total scores, except for one ApBM and one control participant, were above the scale mid‐point (50), indicating generally good acceptability.

3.4. Blinding

Of the participants who guessed what condition they were randomised to, 6/7 (86%) of the ApBM group and 6/11 (55%) of controls thought they received the active intervention (p = 0.316).

3.5. Safety

Three participants (2 ApBM, 1 control), who were all from the same site, withdrew from the study because the training task triggered psychological distress. Both ApBM participants withdrew after one training session (with one participant requiring immediate treatment with quetiapine and psychological support) and the control withdrew after two sessions. These adverse events prompted the site to terminate their involvement in the trial. In addition, 5 other participants (1 ApBM, 4 controls) experienced triggering of craving or distress following at least one training session, but chose to complete subsequent sessions. Two of these participants (both controls) required support from clinical staff after two of their training sessions (including administration of diazepam in one case), while the other participants only reported being triggered after one of their sessions.

4. Discussion

Our results, while inconclusive due to the small sample size, are encouraging in terms of the directions of (non‐significant) differences between groups in abstinence rates, although safety outcomes suggest caution regarding further trialling or implementation of this intervention. A smaller proportion of the ApBM group reported illicit amphetamine use at follow‐ups and survival analysis indicated that they were slower to lapse, although these differences did not approach significance. Nonetheless, the finding that over half of ApBM participants reported past‐month abstinence at follow‐ups mirrors the 54% abstinence rate found in our previous open‐label study's 3‐month follow‐up [26].

If the direction of the differences between groups in abstinence rates and survival analysis were replicated in a well‐powered sample, the clinical implications would be substantial. Although our follow‐up period was relatively brief (3 months following discharge), it covers a crucial period of vulnerability to relapse and of neural and psychological recovery. After prolonged, heavy methamphetamine use, the acute (typically ~1‐week) phase of withdrawal is generally followed by a protracted, sub‐acute withdrawal phase characterised by dopamine receptor hypersensitivity [35], impaired decision‐making, elevated anxiety and depression symptoms, elevated impulsivity, and heightened vulnerability to cue‐induced craving [36]. These symptoms can remain elevated for 3–6 months [35, 36]. During this period, learning cognitive‐behavioural techniques, participating in mutual aid/peer‐support groups, and building social support networks are protective against relapse [35, 37]. However, it takes time to engage in and benefit from these therapeutic and social processes. Hence, preventing, or at least delaying any lapse as long as possible during this period, may support longer‐term recovery.

Despite encouraging results, one third of the participants had an adverse reaction to their allocated training task, with some requiring psychological or medication support following sessions, and 3 participants (12%) withdrew due to this distress. This led one site to discontinue participation in the trial, negatively impacting recruitment. While our previous pilot study [26] found low rates of ApBM protocol completion in patients receiving treatment for methamphetamine, relative to alcohol patients in the same setting [5, 6], we hoped that commencing training after at least 1 week of abstinence would minimise such problems in the present trial. However, this approach did not appear to be successful. It is possible that the more potent effect of methamphetamine on the dopaminergic system (relative to alcohol) renders people who use methamphetamine more easily “triggered” by drug‐associated cues. As observed by others [38], MUD treatment studies commonly have high rates of attrition and low power [39, 40]. While participants who completed 6 sessions of training considered the ApBM broadly acceptable, acceptability data are likely biased by the drop‐out of those who found it least acceptable.

Study limitations include apparent chance differences between groups at baseline in several variables, including mean age, quantity (in grams) of methamphetamine used in the past month at admission, and craving scores (see Table 1). Due to the small sample size, it was not practical to statistically control for potential effects of these variables in analyses. There was also a difference between groups in duration of residential treatment, with ApBM group participants staying a mean of 13 days longer than controls, which may potentially influence outcomes. Additionally, the difference between groups in proportions who guessed they received ApBM, while non‐significant, suggests blinding may not have been successful, which could produce confounding expectancy effects. Generalisability of the findings is limited by the fact that all recruitment was conducted in Melbourne, Australia, and all but 1 of the participants were Australian‐born. Moreover, with most recruitment occurring during the COVID‐19 pandemic, additional contextual factors (e.g., COVID‐related changes to sites' programmes; extended lockdowns and public gathering restrictions) may also influence outcomes and reduce generalisability of findings.

A strength of this trial that is lacking in most prior ApBM research was the use of personalised stimuli. Since only a small number of participants reported injecting as their primary route of administration, if we had used the same, generic, mixed set of stimuli for all participants, injection‐related images (e.g., syringes) would have been irrelevant to most. Allowing participants to select which images are most relevant to them is likely to maximise engagement in the training (by targeting the stimuli most likely to elicit cognitive biases). Qualitative data from our studies of personalised ApBM for alcohol use suggests that personalising the positive images participants are trained to approach reinforces motivations to engage in healthy activities that replace substance use [41, 42].

While the results encourage further research on ApBM's efficacy for MUD, we suggest that any investment in larger, fully‐powered trials should closely attend to improving its safety profile. Such strategies might include delaying ApBM until after the “early protracted withdrawal” phase [35] (i.e., > 1 month of abstinence), which would make longer‐term residential rehabilitation settings more appropriate environments for trials. Other potential strategies include closer clinical monitoring in the hours and days following ApBM sessions. Use of digital ecological momentary assessment/wearables could allow for real‐time safety alerts. Protocols which include craving reduction practises (e.g., mindfulness and breathing exercises) following training sessions may reduce adverse events. Moreover, practicing these skills in response to the arousing effects of training stimuli may further improve participants' ability to cope with encountering methamphetamine cues post‐discharge. Noting that most adverse events occurred in earlier sessions, the personalisation protocol could also be modified in future trials, by beginning ApBM (or sham) training with less‐personalised (i.e., lower‐rated) methamphetamine images to reduce the initial “shock” of intensive exposure to highly “triggering” images; and then introducing more “potent” personalised, highly‐rated drug images in subsequent sessions. In conclusion, ApBM delivered during residential rehabilitation may potentially help prevent relapse post‐discharge, however close attention is needed to improving its safety profile in any future, larger and statistically‐powered trial aimed at establishing its efficacy.

Author Contributions

Victoria Manning conceived the study in collaboration with Joshua B.B. Garfield. Victoria Manning led the funding application, which also involved Joshua B.B. Garfield, Shalini Arunogiri, Dan I. Lubman, Eli Kotler, Goke Okedara, and Malcolm Hopwood. Joshua B.B. Garfield and Hugh Piercy drafted the protocol. Joshua B.B. Garfield and Victoria Manning supervised the research team responsible for data collection (which included Hugh Piercy and Danielle Whelan), while Eli Kotler, Malcolm Hopwood, Michael Maloney, and Dan I. Lubman were principal investigators at recruitment sites. Hugh Piercy cleaned the data and Joshua B.B. Garfield conducted all statistical analyses. Joshua B.B. Garfield, Victoria Manning, and Danielle Whelan drafted the manuscript, and all authors read, provided input on, and approved the manuscript.

Funding

This project was funded by the National Centre for Clinical Research on Emerging Drugs (NCCRED) (project number: NCR2SF10). NCCRED played no role in the study design or in the collection, management, analysis, interpretation, or publication of data.

Conflicts of Interest

All authors have no conflicts of interest to declare. Shalini Arunogiri is supported by an Investigator Grant from the National Health and Medical Research Council (NHMRC: grant number 2008193).

Supporting information

Data S1: Supplementary Information.

Data Availability Statement

The trial protocol is publicly available via the Australian New Zealand Clinical Trials Registry (ACTRN12620000072910): https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=378804&isReview=true. We do not intend to make the dataset accessible to the general public. However, researchers interested in accessing deidentified individual participant data may contact Professor Victoria Manning with a proposal explaining their intended analyses. Granting access to other researchers to use deidentified data will require completion of a data sharing agreement between Monash University and the researchers seeking to access the data, and if there are any costs associated with this process, these costs will need to be borne by the researchers seeking access to the data. Data will be retained for 7 years after the completion of reporting of trial outcomes, but may be unavailable after that point.