Effect of conditioned stimuli-triggered memory retrieval-extinction in patients with methamphetamine use disorder

Abstract

Methamphetamine (METH) is a widely abused stimulant that affects the central nervous system. The persistent maladaptive conditioned stimuli (CS, drug cues)-drug associative memories represent a primary factor precipitating relapse. Interfering with the reconsolidation of these memories may help disrupt and modify these maladaptive CS-drug associations, potentially reducing their influence on drug-seeking behavior. The present study explored the effect of CS-triggered memory retrieval-extinction on METH craving, potentially offering a new treatment strategy for addiction. This was a single-center, randomized, controlled trial involving individuals with METH use disorder (MUD). Participants completed one of three interventions on consecutive days (days 2 and 3): CS-triggered memory retrieval followed by extinction after a 10-min interval, CS-triggered memory retrieval followed by extinction after a 6-h interval, or extinction without prior retrieval. Self-report cue-induced craving for METH, salivary cortisol and sympathetic responses were measured at baseline (day 1), post intervention (day 4) and two follow-up timepoints (days 34 and 184), with cue-induced craving and salivary cortisol as primary outcomes. Ninety-eight MUD individuals (mean age 28.15 ± 6.31) were analyzed. After two-day’s interventions, cue-induced METH craving (time × cue interaction: F_(1,94) = 60.02, p < 0.001) reduced in all groups. Results from follow-up data indicated, when the extinction was performed 10 min, but not 6 h after memory retrieval or no retrieval, the intervention decreased experimental cue-induced METH craving (intervention × time × cue: F_(2,94) = 14.32, p < 0.001; intervention: F_(2,94) = 24.28, p < 0.001) and saliva cortisol increases (F_(2,90) = 9.51, p < 0.001), with effects lasting up to 6-month follow-up. The results revealed a substantial reduction in cue-elicited craving and saliva cortisol in the retrieval-10 min-extinction group over the 6-month follow-up. These findings provide compelling evidence that a brief reconsolidation-based intervention can effectively diminish METH-related craving and cortisol levels, underscoring its potential as a supportive measure in METH treatment. Salivary cortisol is a readily accessible and sensitive biomarker for evaluating intervention effects.

Introduction

Methamphetamine (METH) is a widely abused stimulant that affects the central nervous system. According to the United Nations Office on Drugs and Crime (UNODC), approximately 36 million people consumed amphetamines in 2021, with China among the leasing for METH seizures [1]. For the past twenty years, METH has consistently ranked as the second most commonly abused drug in Southeast and East Asia, particularly in China, and the trend of increasing METH usage persists [1, 2]. METH addiction presents significant public health challenges, as high doses can cause severe neuropsychiatric effects, including agitation, anxiety, hallucinations, paranoia and psychosis. Individuals with METH use disorder (MUD) frequently exhibit cognitive impairments affecting attention, memory, and executive function alongside persistent sleep disturbances [3]. The prevalence of psychiatric symptoms among patients with MUD is 11 times greater than in the general population, often leading to violence and suicide [4]. Additionally, METH abuse is associated with various cardiovascular complications, such as hypertension, tachycardia, acute coronary syndrome, pulmonary arterial hypertension, cardiomyopathy, and heart failure [5, 6]. Despite these severe health consequences, the therapeutic landscape for MUD remains remarkably limited. Currently, no FDA-approved pharmacotherapies exist for either MUD or its associated psychiatric complications [7, 8]. Available treatment modalities rely exclusively on psychosocial interventions, which demonstrate variable and often insufficient efficacy [7].

The primary challenge in MUD is relapse, which is closely linked to mechanisms of learning and memory [9]. Within the framework of maladaptive memory formation, repeated associations between the rewarding effects of drug consumption (the unconditioned stimulus [US] from METH administration) and environmental cues associated with METH use (conditioned stimulus [CS]) progressively hijack normal neural circuitry. This Pavlovian conditioning represents a critical aberrant engagement mechanism underlying substance use disorders (SUDs) [9]. Through this associative learning process, drug-paired CS memories evolve into potent triggers for relapse in individuals striving to maintain abstinence [10, 11]. The rediscovery of memory reconsolidation in 2000 has sparked interest in memory-focused addiction treatments [12]. Memory reconsolidation represents a temporally constrained process (occurring within a 6-h window) wherein, under specific retrieval conditions, previously consolidated memories transition into a labile state before restabilizing [13, 14]. Memory retrieval can occur through various methods, typically involving brief re-exposure to either the CS or the US [15–20]. Reconsolidation is a protein synthesis-dependent process during which memories become susceptible to modification through pharmacological or behavioral interventions. The resultant amnesia may reflect either complete erasure of the memory trace or targeted degradation of specific memory components [21]. The fundamental principle underlying reconsolidation-based therapeutic approaches involves two critical steps: first, reactivating maladaptive drug-seeking memories to induce lability in established neural representations; and second, administering amnestic agents or behavioral interference protocols to disrupt the reconsolidation process. This strategic intervention aims to attenuate maladaptive memory strength, thereby producing long-term reductions in drug-seeking behavior [22].

Several studies have investigated the potential of behavioral interventions to disrupt the reconsolidation of drug-related memories in humans. Research has shown that extinction, a specific type of intervention, is effective. Extinction training involves protracted, unreinforced exposure to cues that trigger conditioned responses, leading to a reduction in cue-elicited responses. In addiction treatment contexts, extinction-based therapy comprises repeated exposure to drug-associated stimuli (CS) without the accompanying drug reward (US) [23]. Extinction facilitates the learning of a new stimulus-no-reward association that requires its consolidation phase. Subsequently it competes with or suppresses initial drug-reward learning, but does not erase original associative memories [24]. The pioneering demonstration of retrieval-extinction efficacy in SUDs was reported by Xue and colleagues, who established that brief re-exposure to heroin-associated CS, followed by systematic exposure therapy in outpatients with heroin use disorder, significantly reduced cue-induced craving with effects persisting for at least six months [18]. A subsequent study showed that retrieval-extinction paradigms effectively reduce self-reported cigarette consumption among treatment-seeking tobacco-dependent individuals [23]. Our recent investigations further established that US (methadone) exposure in individuals with opioid use disorder, when, followed by extinction procedures, promoted sustained heroin abstinence while concurrently attenuating both cue-induced craving and blood pressure (BP) elevations throughout a six-month follow-up period [20]. Chen and colleagues provided compelling evidence that CS-triggered memory retrieval-extinction protocols significantly diminished subsequent drug priming-induced reinstatement, spontaneous recovery, and contextual renewal of methamphetamine-seeking behaviors in rats [25]. However, the therapeutic efficacy of memory reconsolidation interference specifically in MUD populations remains elusive. Consequently, the present study sought to explore the effect of CS-based memory reconsolidation and extinction in MUD individuals, offering a potential strategy for intervention.

Methods

Ethics

The study protocol was approved by the Biomedical Ethics Committee of Peking University. All participants provided written informed consent. The study strictly adhered to the Declaration of Helsinki. The authors designed the study, collected and analyzed the data and wrote the manuscript.

Study design and participants

This study was a single-center, randomized, controlled trial involving individuals with MUD. All the participants were recruited from a drug rehabilitation center in Guangdong Province, China, from December 2016 to October 2017. The inclusion criteria were as follows: (1) participants aged 18–55 years; (2) those with at least an elementary school education; (3) a diagnosis of MUD according to the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM- V); (4) those with no medication use during the study period. The exclusion criteria were as follows: (1) individuals who had experienced a coma lasting more than 30 min due to head injuries; (2) those with severe mental disorders such as major depressive disorder, generalized anxiety disorder, etc.; (3) those with neuropathic disorders; (4) those who exhibited systolic blood pressure (SBP) below 90 mmHg or above 140 mmHg, diastolic blood pressure (DBP) below 50 mmHg or above 90 mmHg or heart rate (HR) below 50 bpm or above 100 bpm; and (5) participants who reported the use of addictive drugs other than METH (excluding nicotine) in the past 12 months.

Procedures and interventions

All participants underwent pre-screening and provided sociodemographic and clinical data. They completed a baseline assessment on Day 1, before intervention, followed by two consecutive days of intervention on Days 2 and 3. A post-intervention assessment was conducted on Day 4, with two subsequent follow-up evaluations on Days 34 and 184.

Pre-screening and measures

Initially, all participants underwent a comprehensive screening to exclude those with severe neuropsychiatric diseases or physical conditions. After pre-screening for inclusion and exclusion criteria, each participant participated in a clinical interview with M.I.N.I to assess for psychiatric disorders (SJ) [26]. After completion of the interview, participants completed self-assessments related to sociodemographic and clinical factors, while the assessor (CSJ) evaluated two additional clinical measures (working memory and anxiety). Subsequently, participants were assigned to one of three groups in a ratio of 1:1:1: (1) extinction without prior retrieval (No-retrieval extinction group, control); (2) extinction conducted within the reconsolidation window (10 min after retrieval; retrieval-10 min-extinction group, experimental); or (3) extinction conducted outside the reconsolidation window (6 h after retrieval; retrieval-6h-extinction group, sham comparator). Randomization was implemented by an independent blinded team member (BYP) who was exclusively involved in data analysis, using a randomization code table generated with Microsoft Excel software. Previous research has shown gender differences in cardiovascular and corticoadrenal responses to drug-associated cues among individuals with stimulant use disorders [27, 28]. To ensure balanced distribution of critical demographic and clinical variables across the three intervention groups, we implemented an urn randomization procedure with stratification for educational attainment, addiction severity, and gender [29]. This adaptive randomization approach minimizes allocation bias while maintaining group equivalence on key prognostic factors. Other team members who remained independent from assessment and data analysis (WRJ, GXJ and SYK) were responsible for the random allocation of participants to preserve allocation concealment.

Tests and interventions

At baseline assessment (Day 1), participants were exposed to a 5-min neutral video clip followed by a 5-min METH-related video clip. Self-reported cue-induced craving was measured before and after each video presentation. Concurrently, trained assessors (CSJ and LX) collected salivary samples and recorded cardiovascular parameters (BP and HR) at corresponding timepoints. Twenty-four hours later, participants underwent their assigned memory retrieval-extinction intervention protocol over two consecutive days (Days 2 and 3). Post-intervention assessment (Day 4) replicated the baseline procedure, with participants viewing identical video clips while assessors measured cue-induced craving, salivary biomarkers, and cardiovascular responses before and after each video presentation. This assessment protocol was designed to evaluate immediate intervention effects on primary (craving, cortisol) and secondary (BP/HR) outcome measures. To assess long-term intervention efficacy, identical assessment procedures were conducted at one-month and six-month follow-up timepoints, enabling comprehensive evaluation of the durability of intervention effects in MUD individuals (Fig. 1).

Fig. 1. Experimental design flowchart.

The study protocol began with a baseline assessment on day 1, which included measurements of methamphetamine (METH) cue-induced craving, salivary cortisol levels, and sympathetic nervous system activation. Participants were randomly allocated to one of three intervention groups: (1) experimental group (METH retrieval+10 min interval+60 min extinction), (2) sham comparator group (METH retrieval+6 h interval+60 min extinction), (3) control group (neutral retrieval+10 min interval+60 min extinction). Following the interventions, a post-intervention assessment was conducted on day 4, with subsequent follow-up evaluations at day 34 (1-month) and day 184 (6-month). All assessment time points included identical measurements of METH cue-induced craving, salivary cortisol levels, and sympathetic activation parameters.

The present investigation implemented a two-day CS-triggered retrieval-extinction procedure based upon the foundational work of Xue et al. and Germeroth et al., along with our preliminary experimental findings [18, 23]. All participants were instructed to abstain from cigarette smoking for a minimum of two hours prior to experimental sessions to control for potential nicotine-related confounding effects. The study design incorporated three distinct intervention conditions. In the experimental group (retrieval-10 min-extinction), participants first viewed a 5-min METH-related video to activate drug-associated memory, followed by a 10-min interval, and subsequently underwent a 60-min extinction training session. This timing was specifically designed to target the reconsolidation window. In the sham comparator group (retrieval-6h-extinction), participants also viewed a 5-min METH-related video for memory retrieval, but extinction training was deliberately delayed by 6 h to ensure it occurred outside the putative reconsolidation window. Participants in the control group (no-retrieval extinction) initially watched a 5-min neutral video (containing no drug-related content), followed by a 10-min interval, and subsequently underwent an identical 60-min extinction training. The extinction training consisted of four sequential blocks of METH-related stimuli, including 5-min video segments, 5-min picture presentations, and 5-min periods of inspection and handing of drug paraphernalia previously associated with METH [18, 30]. To minimize order effects, we implemented a randomization procedure for stimulus presentation. Six distinct stimulus blocks were available (comprising video, picture, and in vivo paraphernalia handling), and participants were instructed to randomly select the initial block during the 10-min or 6-h interval period. Following determination of the first block, researchers systematically selected the remaining three blocks according to predefined criteria: (1) the terminal stimulus of each block differed from the initial stimulus of the subsequent block, and (2) no block could be presented more than once within the four-block sequence. This procedure ensured both standardization and appropriate stimulus variation across participants. Upon completion of experimental procedures and prior to laboratory discharge, participants engaged in a brief mindfulness exercise specifically designed to attenuate residual cue-induced craving and facilitate return of physiological parameters to baseline levels [31].

Measures

Participants completed a comprehensive assessment battery that included detailed demographic information (gender, age, ethnicity, educational attainment) and substance use parameters. METH use history was systematically documented, including duration of use, abstinence period (time elapsed from last METH use to study enrollment), and average daily consumption. The DSM-V criteria classified the severity of MUD as mild (2–3), medium (4–5) and severe (≥6). Given the high comorbidity between stimulant and nicotine use, we assessed nicotine dependence severity using the Fagerstrom Test for Nicotine Dependence (FTND), an instrument with established psychometric properties including strong internal consistency and construct validity. In accordance with validated clinical thresholds, FTND scores ≥6 indicated high nicotine dependence [32, 33]. Depressive symptomatology was assessed using the 13-item Beck Depression Inventory (BDI-13), with scores ≥8 considered clinically significant. Anxiety symptoms were evaluated using the Hamilton Anxiety Scale (HAMA), where scores ≥14 indicated clinically relevant anxiety [34, 35]. Cognitive functioning was evaluated using the Digit Span Test (DST), which provides a validated assessment of attention, working memory, and short-term auditory [36]. To characterize impulsivity dimensions-a critical construct in substance use disorders-we administered the Chinese version of the Barratt Impulsiveness Scale (BIS), which yields distinct subscale scores for cognitive impulsiveness (CI), motor impulsiveness (MI), and non-planning impulsiveness (NPI) [37, 38].

The primary outcome measures comprised cue-induced METH craving and salivary cortisol. Craving intensity was quantified using a self-report visual analog scale (VAS)-a 10-cm horizontal line anchored at the extremes with 0 (“not at all”) and 10 (“extremely high”). Participants indicated their subjective craving level by marking this continuous scale immediately before and after exposure to cues at four assessment timepoints: baseline (Day 1), post-intervention (Day 4), 1-month and 6-month follow-up tests (Day 34 and Day 184). Salivary cortisol, a physiological marker of hypothalamic-pituitary-adrenal (HPA) axis reactivity, was collected using standardized saliva collection tubes (Sarstedt, Nümbrecht, Germany) immediately before and after cue exposure at all four assessment timepoints. To minimize potential confounding effects of nicotine on cortisol secretion, participants were required to abstain from smoking for a minimum of two hours prior to each assessment session. Salivary samples were analyzed using a commercially available enzyme-linked immunosorbent assay (Salivary Cortisol ELISA SLV-2930, DRG Instruments GmbH, Germany) with established sensitivity and specificity for cortisol quantification.

Secondary outcome measures included cardiovascular parameters-SBP, DBP and HR-which were assessed as physiological indices of autonomic arousal in response to drug-associated cues. These physiological measurements were obtained using a calibrated EM-6 multi-parameter patient monitor with participants maintained in a standardized seated position pre- and post-cue exposure at four timepoints (before/post intervention and two follow-ups). To minimize potential measurement bias, strict blinding procedures were implemented throughout the study. Team members responsible for outcome assessment, test administration, and data entry remained blinded to participant randomization status and intervention condition. The integrity of this blinding was maintained until after database lock and completion of planned statistical analyses by the study statistician, ensuring unbiased evaluation of intervention effects.

Statistical analysis

The sample size was established based on previous research in this field. Power analysis indicated that a minimum of 82 participants would be required to detect an effect size of 0.60 for cue-elicited craving during post-intervention assessment, with 90% statistical power at α ≤ 0.05. This calculation incorporated an anticipated 20% attrition rate during the follow-up period [18, 20].

A total of 107 eligible participants were recruited. However, nine participants were excluded due to incomplete intervention adherence, resulting in an 8.41% exclusion rate. The final analytical sample consisted of 98 participants distributed across the three intervention conditions: control group (n = 36), experimental group (n = 33), and sham comparator group (n = 29). This sample was used for analysis of baseline demographic and clinical characteristics, as well as primary and secondary outcomes (cue-induced METH craving, BP and HR). Continuous variables-including age, educational attainment, duration of METH use, abstinence duration, average METH dosage, BDI score, HAMA score, DST score, BIS score, and FTND score-were reported as mean (standard error of the mean) and analyzed using one-way analyses of variance (ANOVAs). Categorical variables-including gender, METH dependence severity, presence of clinically significant depression or anxiety, and smoking status-were reported as proportions (%) and analyzed using Chi-square test (χ² test).

This study employed a three-factor mixed design, with repeated measures for two within-subject factors (time and cue) and one between-subject factor (intervention group). To analyze changes in cue-induced METH craving, BP, and HR, we conducted three-way mixed analyses of covariance (ANCOVAs) incorporating appropriate within- and between-subject factors, with daily cigarette consumption included as a covariate to control for potential nicotine effects. For salivary cortisol analyses, two-way mixed ANCOVAs were implemented. Five participants were excluded from cortisol analyses due to missing saliva samples, resulting in a final sample of 93 participants for these analyses (control group: n = 33; experimental group: n = 33; sham comparator group: n = 27). Significant ANCOVA effects were further examined using Bonferroni-corrected post hoc comparisons to control for multiple testing. A 2-tailed p < 0.05 was considered statistically significant. All statistical analyses were performed using IBM SPSS Statistics (version 22.0).

Results

Participant characteristics

The final analytical sample comprised 98 participants. Baseline comparisons revealed no significant between-group differences in demographic characteristics, including age, educational attainment, duration of METH use, abstinence period, average daily METH dosage, and years of tobacco use, as assessed by one-way analysis of variance (all p > 0.05). Similarly, clinical characteristics-including BDI, HAMA, DST and FTND-showed no significant differences across intervention groups (all p > 0.05). Chi-square analysis indicated comparable gender distribution across the three groups (p = 0.14). These results confirm successful randomization and baseline equivalence across intervention conditions (Table 1).

Table 1.

Demographic and clinical characteristics of patients with methamphetamine use disorder.

Variables	Participants (N = 98)	No-retrieval extinction (N = 36)	Retrieval-10 min-extinction(N = 33)	Retrieval-6h-extinction (N = 29)	p
Age (year, Mean ± SD)	28.15 ± 6.31	28.38 ± 6.19	27.80 ± 5.69	28.3 ± 7.37	0.92
Male-gender (N, %)	76 (79.17)	27 (79.41)	29 (82.86)	20 (74.07)	0.70
Education (year, Mean ± SD)	7.74 ± 1.57	8.06 ± 1.63	7.66 ± 1.59	7.44 ± 1.45	0.30
METH use
Duration of use (month, Mean ± SD)	52.20 ± 29.84	56.82 ± 33.21	45.43 ± 26.42	55.15 ± 29.0	0.24
Abstinent time (month, Mean ± SD)	4.14 ± 2.60	4.88 ± 3.08	3.97 ± 2.29	3.41 ± 2.12	0.08
Average dosage (g, Mean ± SD)	0.57 ± 0.48	0.59 ± 0.49	0.57 ± 0.53	0.54 ± 0.43	0.93
METH addiction severity (N, %)					0.14
Mild	15 (15.63)	4 (11.76)	8 (22.86)	3 (11.11)
Medium	25 (26.04)	7 (20.59)	12 (34.29)	6 (22.22)
Severe	56 (58.33)	23 (67.65)	15 (42.86)	18 (66.67)
BDI score (Mean ± SD)	7.30 ± 5.27	7.56 ± 5.72	7.57 ± 4.95	6.63 ± 5.21	0.74
≥8 (N, %)	41 (42.71)	14 (41.18)	16 (45.71)	11 (40.74)	0.90
HAMA score (Mean ± SD)	4.45 ± 4.53	5.41 ± 5.69	3.51 ± 3.07	4.44 ± 4.38	0.22
≥14 (N, %)	5 (5.21)	4 (11.76)	0 (0)	1 (3.70)	0.08
DST score (Mean ± SD)	11.92 ± 1.75	11.79 ± 1.86	12.09 ± 1.72	11.85 ± 1.70	0.77
BIS score (Mean ± SD)	86.54 ± 13.33	88.38 ± 12.89	84.63 ± 14.72	86.7 ± 12.05	0.51
CI score	32.28 ± 6.50	32.47 ± 5.99	32.23 ± 6.94	32.11 ± 6.77	0.98
MI score	23.94 ± 6.92	24.85 ± 6.56	22.66 ± 6.97	24.44 ± 7.31	0.38
PI score	30.32 ± 8.20	31.06 ± 8.88	29.74 ± 8.65	30.15 ± 6.82	0.80
Smoking (N, %)	90 (91.84)	33 (91.67)	31 (93.93)	26 (89.66)	0.28
Duration (year, Mean ± SD)	10.38 ± 5.82	11.27 ± 5.83	10.68 ± 6.10	8.88 ± 5.38	0.28
Number of cigarettes per day (Mean ± SD)	17.22 ± 9.88	20.58 ± 9.74	16.03 ± 9.48	14.38 ± 9.66	0.04*
FTND score (Mean ± SD)	5.08 ± 2.35	5.52 ± 2.36	4.81 ± 2.56	4.85 ± 2.07	0.41

Continuous variables as Mean ± SD, categorical variables as N (%).

METH methamphetamine, BDI beck depression inventory, HAMA hamilton anxiety scale, DST digit span test, BIS barrett impulsivity scale, CI cognitive impulsiveness, MI motor impulsiveness, PI non-planning impulsiveness, FTND fagerstrom test for nicotine dependence.

^*p < 0.05.

The study population had a mean age of participants was 28.15 ± 6.31 years, with males comprising 79.17% (n = 76) of the sample, reflecting the gender distribution typically observed in MUD populations in China. Participants reported a substantial history of METH use, with a mean duration exceeding four years, and were in early recovery as indicated by a mean abstinence period of approximately four months. Notably, 58.33% (n = 56) of participants met criteria for severe MUD according to DSM-V.

CS-triggered memory retrieval-extinction procedure causes a long-lasting attenuation of experimental cue-induced METH craving

Analysis of acute intervention effects on METH craving revealed significant time × cue interaction (F_(1,94) = 60.02, p < 0.001), indicating differential craving responses to drug versus neutral cues across assessment timepoints. Significant main effects were observed for both time F_(1,94) = 91.11, p < 0.001) and cue (F_(1,94) = 27.43, p < 0.001). However, no significant intervention-related effects were detected, as evidenced by non-significant intervention × time × cue interaction (F_(2,94) = 0.13, p = 0.878), intervention × time interaction (F_(2,94) = 0.94, p = 0.396), cue × intervention interaction (F_(2,94) = 0.58, p = 0.564) and intervention main effect (F_(2,94) = 1.92, p = 0.152) when comparing pre-intervention versus immediate post-intervention assessments (Fig. 2A, B).

Fig. 2. Temporal changes in methamphetamine cue-induced craving.

A Line graphs depicting craving responses across assessment timepoints, with METH cue responses represented by solid lines and neutral cue responses by dashed lines; B Acute intervention effects (pre-intervention versus post-intervention); C Long-term intervention effects (post-intervention versus 1-month follow-up; post-intervention versus 6-month follow-up). No significant between-group differences in craving were observed at baseline. Following intervention, all groups exhibited significant decreases in craving levels, although no significant between-group differences were detected. Notably, participants in the retrieval-10 min-extinction group maintained significantly lower craving levels at both 1-month and 6-month follow-ups compared to their post-intervention measurements, whereas participants in the other two groups demonstrated a return to baseline craving levels. ^ns, p > 0.05, between-subjects comparison at identical timepoints; ^***, p < 0.001, within-subjects comparison across different timepoints; ^###, p < 0.001, between-subjects comparison at identical timepoints. Craving change scores were calculated as the difference in visual analogue scale (VAS) values between post- and pre-cue exposures. Error bars represent standard errors of the mean (analyzed using three-way mixed ANCOVA).

Examination of long-term intervention effects on METH craving revealed robust treatment-specific outcomes. When comparing post-intervention with 1-month follow-up assessments, significant intervention × time × cue interaction (F_(2,94) = 14.32, p < 0.001), indicated that intervention conditions differentially affected cue-induced craving trajectories over time. Additional significant interactions included intervention × time (F_(2,94) = 14.52, p < 0.001), cue × intervention (F_(2,94) = 10.41, p < 0.001), and time × cue (F_(1,94) = 20.53, p < 0.001). Significant main effects were observed for intervention condition (F_(2,94) = 24.28, p < 0.001), time (F_(1,94) = 52.92, p < 0.001), and cue type (F_(1,94) = 16.75, p < 0.001). Importantly, these intervention effects persisted through the 6-month follow-up period (post-intervention versus 6-month follow-up), demonstrating the durability of intervention outcomes (Fig. 2A, C and Table S1).

Post-intervention assessments revealed significant reductions in METH craving compared to baseline across all intervention conditions (p < 0.001), demonstrating the immediate efficacy of the two-day intervention protocols. However, notable differences emerged in the durability of these intervention effects. Only participants in the retrieval-10 min-extinction group (experimental condition) maintained significantly lower craving responses compared to both control and sham comparator groups at 1-month and 6-month follow-up assessments (all p < 0.001). This pattern indicates spontaneous recovery of craving responses in both the no-retrieval extinction group (control) and the retrieval-6 h-extinction group (sham comparator) (Fig. 2A). These findings demonstrate that while all intervention protocols produced initial reductions in cue-induced craving among MUD individuals, only the retrieval-extinction procedure conducted within the reconsolidation window (10-min interval) yielded persistent therapeutic effects that maintained through the 6-month follow-up period. In contrast, extinction training without prior retrieval or with retrieval outside the reconsolidation window (6-h interval) produced only transient benefits that dissipated by the 1-month follow-up assessment.

CS-triggered memory retrieval-extinction procedure causes a long-lasting attenuation of cue-induced salivary cortisol

Analysis of immediate effect of the intervention effects on salivary cortisol revealed no significant intervention × time interaction (F_(2,90) = 1.09, p = 0.342) or intervention main effect (F_(2,90) = 1.05, p = 0.354), indicating that cortisol did not differ significantly across the three intervention conditions in the immediate post-intervention period. However, a significant main effect of time was observed (F_(1,90) = 4.87, p = 0.030) when comparing pre-intervention with post-intervention assessments, suggesting that salivary cortisol levels changed significantly over time regardless of intervention condition (Fig. 3A, B). This pattern indicates that while physiological stress responses as measured by salivary cortisol exhibited temporal changes during the study period, these changes were not differentially affected by the specific intervention protocols in the acute phase.

Fig. 3. Temporal changes in methamphetamine cue-induced salivary cortisol.

A Line graphs illustrating salivary cortisol responses across assessment timepoints, with METH cue responses represented by solid lines; B Acute intervention effects (pre-intervention versus post-intervention); C Long-term intervention effects (post-intervention versus 1-month follow-up; post-intervention versus 6-month follow-up). Throughout the entire follow-up period, significant decreases in METH cue-induced salivary cortisol levels were observed exclusively in the retrieval-10 min-extinction group. ^ns, p > 0.05, between-subjects comparison at identical timepoints; ^*, p < 0.05, within-subjects comparison across different timepoints; ^##, p < 0.01, between-subjects comparison at identical timepoints. Cortisol change scores were calculated as the difference in cortisol concentration between post- and pre-cue exposures. Error bars represent standard errors of the mean (analyzed using two-way mixed ANCOVA).

Analysis of long-term intervention effects on salivary cortisol revealed differential patterns across follow-up timepoints. When comparing post-intervention with 1-month follow-up assessments, a significant main effect of intervention was observed (F_(2,90) = 9.03, p < 0.001), indicating that cortisol levels differed significantly across the three intervention conditions. However, the intervention × time interaction (F_(2,90) = 1.20, p = 0.308) and time main effect (F_(1,90) = 3.79, p = 0.055) did not reach statistical significance during this period. Extended follow-up analyses comparing post-intervention with 6-month assessments demonstrated significant main effects for both intervention (F_(2,90) = 9.51, p < 0.001) and time (F_(1,90) = 6.30, p = 0.014), while the intervention × time interaction remained non-significant (F_(2,90) = 2.23, p = 0.114). These findings indicate that while intervention conditions produced significantly different cortisol responses that persisted through 6 months, the pattern of change over time was relatively similar across conditions (Fig. 3A, C). The significant time main effect at 6-month follow-up further suggests that cortisol levels continued to change over the extended follow-up period regardless of intervention condition.

Salivary cortisol levels demonstrated a consistent pattern of reduction from baseline through the post-intervention assessment across all conditions, indicating an immediate physiological stress-dampening effect of the interventions. During the 1-month and 6-month follow-up periods, cortisol levels continued to decline compared to post-intervention values. These findings suggest that all three intervention protocols produced sustained reductions in physiological stress responses among MUD individuals, with effects persisting through the 6-month follow-up period. Notably, between-group comparisons revealed that participants in the retrieval-10 min-extinction group exhibited significantly lower salivary cortisol concentrations compared to both the no-retrieval extinction group and the retrieval-6 h-extinction group. This differential response pattern suggests that extinction training conducted within the reconsolidation window (10-min interval) produced more robust attenuation of physiological stress reactivity than either extinction without retrieval or extinction conducted outside the reconsolidation window (6-h interval).

CS-triggered memory retrieval-extinction procedure causes an attenuation of cue-induced HR change

Analysis of acute intervention effects on DBP revealed a significant main effect of time (F_(1,94) = 8.13, p = 0.005), indicating that DBP values changed significantly from pre-intervention to post-intervention assessment regardless of intervention condition or cue type. However, no significant effects were observed for intervention × time × cue interaction (F_(2,94) = 0.88, p = 0.418), intervention × time interaction (F_(2,94) = 0.52, p = 0.599), cue × intervention interaction (F_(2,94) = 0.08, p = 0.921), or time × cue interaction (F_(1,94) = 2.31, p = 0.132). Similarly, no significant main effects were detected for intervention condition (F_(2,94) = 0.10, p = 0.901) or cue type (F_(1,94) = 1.55, p = 0.216) when comparing pre-intervention with immediate post-intervention assessments (Fig. 4A, B). The pattern of results for long-term intervention effects on DBP paralleled those observed in the acute phase, with time-related changes occurring independently of intervention condition or cue exposure (Fig. 4A, C and Table S1). These findings suggest that while DBP exhibited temporal changes during the study period, these changes were not differentially affected by the specific intervention protocols or cue exposure conditions in either the acute or long-term follow-up phases.

Fig. 4. Temporal changes in methamphetamine cue-induced diastolic blood pressure.

A Line graphs illustrating diastolic blood pressure responses across assessment timepoints, with (METH) cue responses represented by solid lines and neutral cue responses by dashed lines; B Acute intervention effects (pre-intervention versus post-intervention); C Long-term intervention effects (post-intervention versus 1-month follow-up; post-intervention versus 6-month follow-up). ^ns, p > 0.05, between-subjects comparison at identical timepoints; ^**, p < 0.01, within-subjects comparison across different timepoints. DBP change scores were calculated as the difference in DBP values between post- and pre-cue exposures. Error bars represent standard errors of the mean (analyzed using three-way mixed ANCOVA).

Analysis of acute intervention effects on SBP revealed a significant time × cue interaction (F_(1,94) = 7.03, p = 0.009), indicating differential SBP responses to drug versus neutral cues across assessment timepoints. Significant main effects were also observed for time (F_(1,94) = 8.83, p = 0.004) and cue type (F_(1,94) = 5.51, p = 0.021). However, no significant intervention-related effects were detected, as evidenced by non-significant intervention × time × cue interaction (F_(2,94) = 0.37, p = 0.691), intervention × time interaction (F_(2,94) = 0.46, p = 0.636), cue × intervention interaction (F_(2,94) = 0.08, p = 0.928), and intervention main effect (F_(2,94) = 0.70, p = 0.499) when comparing pre-intervention versus immediate post-intervention assessments (Fig. 5A, B). Examination of long-term intervention effects on SBP yielded no significant interaction effects or main effects for intervention condition, time, or cue type in either the post-intervention versus 1-month follow-up comparison or the post-intervention versus 6-month follow-up comparison (Fig. 5A, C and Table S1).These findings suggest that while SBP exhibited differential responses to drug cues versus neutral cues in the acute phase, these cue-elicited cardiovascular responses were not maintained during long-term follow-up, nor were they differentially affected by the specific intervention protocols at any timepoint.

Fig. 5. Temporal changes in methamphetamine cue-induced systolic blood pressure.

A Line graphs illustrating systolic blood pressure responses across assessment timepoints, with (METH) cue responses represented by solid lines and neutral cue responses by dashed lines; B Acute intervention effects (pre-intervention versus post-intervention); C Long-term intervention effects (post-intervention versus 1-month follow-up; post-intervention versus 6-month follow-up). ^ns, p > 0.05, between-subjects comparison at identical timepoints; ^**, p < 0.01, within-subjects comparison across different timepoints. SBP change scores were calculated as the difference in SBP values between post- and pre-cue exposures. Error bars represent standard errors of the mean (analyzed using three-way mixed ANCOVA).

Analysis of acute intervention effects on HR revealed significant intervention × time interaction (F_(2,94) = 3.33, p = 0.040), indicating differential temporal patterns of HR response across intervention conditions. Additionally, significant time × cue interaction (F_(1,94) = 11.31, p = 0.001) and cue main effect (F_(1,94) = 4.30, p = 0.041) were observed, suggesting that HR responses to drug versus neutral cues varied across assessment timepoints. However, no significant effects were detected for intervention × time × cue interaction (F_(2,94) = 0.19, p = 0.831), cue × intervention interaction (F_(2,94) = 0.20, p = 0.823), time main effect (F_(1,94) = 0.99, p = 0.322), or intervention main effect (F_(2,94) = 0.15, p = 0.861) when comparing pre-intervention versus immediate post-intervention assessments (Fig. 6A, B). Examination of long-term intervention effects on HR at the 6-month follow-up revealed a significant intervention × time × cue interaction (F_(2,94) = 7.20, p = 0.001), demonstrating that intervention conditions differentially affected cue-induced HR responses over the extended follow-up period. Significant intervention × time interaction (F_(2,94) = 5.11, p = 0.008) and time main effect (F_(1,94) = 5.14, p = 0.026) were also observed when comparing post-intervention with 6-month follow-up assessments (Fig. 6A, C and Table S1).These findings indicate that the three intervention protocols produced significantly different patterns of HR response during long-term follow-up, suggesting delayed emergence of intervention-specific effects on this autonomic measure.

Fig. 6. Temporal changes in methamphetamine cue-induced heart rate.

A Line graphs illustrating heart rate responses across assessment timepoints, with (METH) cue responses represented by solid lines and neutral cue responses by dashed lines; B Acute intervention effects (pre-intervention versus post-intervention); C Long-term intervention effects (post-intervention versus 1-month follow-up; post-intervention versus 6-month follow-up). HR change scores were calculated as the difference in HR values between post- and pre-cue exposures. Error bars represent standard errors of the mean (analyzed using three-way mixed ANCOVA).

Discussion

Addiction is a chronic, relapsing condition, with individuals prone to relapse following periods of abstinence. A significant trigger for relapse is the re-exposure to cues associated with drug use, such as specific people, locations and paraphernalia consistently linked to drug consumption. These CSs acquire motivational salience through repeated pairings with drug effects, eventually eliciting powerful physiological and psychological responses that can undermine recovery efforts [39, 40]. The maladaptive CS-drug associative memories were traditionally considered permanent and resistant to modification through conventional interventions. However, the discovery of memory reconsolidation has provided a novel mechanistic framework for targeting and potentially disrupting these pathological memory traces. The primary findings from our investigation demonstrated that a CS-triggered memory retrieval-extinction procedure effectively attenuated both cue-induced craving and cortisol levels in individuals with MUD. Specifically, extinction training conducted within the reconsolidation window (10 min post-retrieval) produced robust reductions in cue-elicited METH craving and salivary cortisol levels compared to either extinction without prior retrieval or extinction conducted outside the putative reconsolidation window (6 h post-retrieval). Notably, the convergence of subjective (self-reported craving) and objective (salivary cortisol) outcome measures strengthens the validity of these findings. These results align with reconsolidation theory, which posits that brief reactivation of drug-associated memories (via 5-min exposure to METH-related cues) renders these memories labile and susceptible to modification through subsequent extinction training, thereby updating the cue-drug contingency representation and producing enduring behavioral and physiological changes. Overall, to the best of our knowledge, this is the first study to evaluate the effects of a brief reconsolidation extinction procedure in individuals with MUD. The sustained reductions in both subjective craving and physiological marker (cortisol) throughout the 6-month follow-up period provided compelling evidence for the durability of this intervention. These findings suggest that timed extinction training conducted within the reconsolidation window offers a promising, mechanistically-driven therapeutic strategy for METH addiction.

The retrieval-extinction paradigm was initially established as an effective approach for modulating fear memories in both preclinical animal models and human studies. It was subsequently demonstrated to exhibit comparable efficacy in disrupting drug-associated memories. The cross-domain effectiveness of this approach underscores the shared neurobiological mechanisms underlying both fear conditioning and addiction-related associative memories [18, 41–45]. The retrieval-extinction involves a precisely timed two-phase procedure: first, a brief re-exposure session to reactivate the target memory trace and initiate destabilization; followed by a temporally distinct, prolonged extinction training session. This second phase must be administered within a specific temporal window-typically 10–60 min after memory reactivation, but theoretically effective within the 3–4 h period during which the reconsolidation process remains susceptible to interference [39]. The present findings align with and extend a growing body of human laboratory studies demonstrating the efficacy of reconsolidation-based interventions [18, 20, 23, 46], with our results revealing significant clinical benefits. These convergent outcomes across multiple investigations provide compelling evidence for the translational potential of memory reconsolidation principles in addressing SUDs [47]. Despite promising experimental findings, the clinical implementation of retrieval-extinction procedures has been relatively limited in the treatment of SUDs. The present investigation significantly advances this field by extending the reconsolidation-based extinction paradigm to individuals with MUD-a population with particularly challenging treatment needs.

Our results documented robust attenuation of cue-elicited craving not only in immediate post-intervention assessments but also throughout the longitudinal follow-up period, with therapeutic effects maintained at both 1-month and 6-month timepoints. Notably, while all groups demonstrated initial reductions in cue-elicited craving immediately post-intervention, these therapeutic effects dissipated in both the control group (no-retrieval extinction) and sham comparator group (retrieval-6 h-extinction) at 1-month and 6-month follow-up assessments. This differential pattern of immediate versus sustained effects aligns with contemporary learning theory. The transient benefits observed in these two comparison conditions likely reflect standard extinction processes, wherein extensive cue exposure generates a new inhibitory association (“cue-no outcome”) that temporarily suppresses but does not fundamentally alter the original memory trace [48]. These findings suggest that conventional extinction effects were transient, with inhibitory control diminishing over time, resulting in the reemergence of conditioned responses-a phenomenon consistent with established extinction principles. In contrast, the persistent reduction in cue-reactivity observed exclusively in the retrieval-10 min-extinction group provides compelling evidence for reconsolidation-mediated memory updating rather than mere inhibitory learning. This sustained therapeutic effect demonstrated resistance to spontaneous recovery, a hallmark feature distinguishing reconsolidation-based interventions from standard extinction approaches. Importantly, these intervention-specific benefits cannot be attributed solely to abstinence duration, as previous research has established that while cue-induced craving typically attenuates after approximately six months of abstinence [30]. However, craving trajectories in the control conditions did not conform to the expected pattern of natural attenuation over time, particularly at the 1-month and 6-month follow-up assessments. One plausible explanation for this divergence involves the formation of novel associative memories between the experimental context and the stimuli (pictures and videos) employed in the study. When participants returned for follow-up evaluations, these context-stimulus associations may have been reactivated, potentially amplifying cue-induced craving responses. This phenomenon highlights the complex interplay between contextual factors and drug-associated memory processes in addiction. Future investigations should systematically examine these potential mechanisms, perhaps incorporating neuroimaging techniques to elucidate the neural substrates underlying differential responses to reconsolidation-based versus standard extinction interventions across varying temporal contexts.

Mechanistically, retrieval-extinction interventions targeting drug-associated memories engage specific molecular cascades involving α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) trafficking dynamics within the prefrontal cortex (PFC) [49]. Drug memory retrieval initiates a temporally precise cascade of synaptic modifications in prefrontal cortical neurons, characterized by transient reductions in AMPAR-mediated synaptic transmission. This process begins within 10 min of retrieval and persists for approximately 2–4 h, primarily through the rapid endocytosis of calcium-permeable AMPARs (CP-AMPARs) from synaptic membranes. Notably, this receptor trafficking mechanism is orchestrated by dopamine D1 receptor signaling, highlighting the critical interaction between dopaminergic and glutamatergic systems in addiction-related neuroplasticity. The subsequent reconsolidation phase is marked by the reinsertion of CP-AMPARs into synaptic sites approximately six hours following memory retrieval, effectively closing the reconsolidation window. These neurobiological dynamics provide a compelling mechanistic foundation for the 10-min and 6-h retrieval-extinction intervals employed in our experimental design, establishing both theoretical validity and translational relevance for the timing parameters of reconsolidation-based interventions in clinical applications.

The present investigation revealed a delayed intervention effect on salivary cortisol levels, with significant reductions observed at both 1-month and 6-month follow-up assessments compared to immediate post-intervention measurements. Notably, participants in the retrieval-10 min-extinction group exhibited significantly lower salivary cortisol concentrations relative to both control and sham comparator conditions. This attenuated neuroendocrine response to drug-associated cues in the reconsolidation-based intervention group parallels the observed reductions in subjective craving, establishing convergent validity across both psychological and physiological outcome measures. These findings align with previous research in SUDs demonstrating positive associations between salivary cortisol and craving intensity in cocaine-abstinent individuals [50]. Elevated cue-induced salivary cortisol has been independently associated with adverse treatment outcomes in SUDs.

Cortisol, a glucocorticoid hormone secreted by the adrenal cortex in response to HPA axis activation, plays a critical role in stress-related neurobiological processes relevant to addiction. Experimental studies in humans have demonstrated that acute intravenous cortisol administration elicits drug craving responses in cocaine-dependent individuals that parallel the craving intensity observed following cocaine priming doses [51]. Wanger and colleagues demonstrated that elevated salivary cortisol concentrations were positively correlated with increased neural activation in the right anterior insula and right dorsolateral prefrontal cortex (DLPFC) [52]. Elevations in cortisol levels correspond with increased reactivity in brain regions implicated in SUDs, particularly those involved in salience attribution, interoceptive processing, and executive function regulation. This heightened neural reactivity has been consistently associated with increased vulnerability to relapse. Consequently, salivary cortisol represents a readily accessible and sensitive biomarker for evaluating intervention efficacy and cue-induced neurobiological processes that contribute to the maintenance of substance dependence.

There was a notable dissociation between craving/cortisol outcomes and cardiovascular responses (such as BP) in the current study. Drug-induced elevations in cortisol are associated with mesolimbic dopaminergic transmission and drug-related euphoria ratings in humans, suggesting cortisol may play a significant role in appetitive drug-related states [53]. Cue-related alterations in cortisol might serve a similar function in MUD. In contrast, cardiovascular responses may be more specifically linked to the mobilization of energy resources through sympathetic nervous system activation [54].

Limitations

Several limitations warrant consideration in the current study. Although double-blinding would have been an optimal design feature, blinding participants to the intervention was fundamentally unachievable, and maintaining blinding among study personnel would have been exceptionally challenging and highly susceptible to compromise. Additionally, the single-center design conducted in China inherently constrains the generalizability of our findings to more diverse populations. However, a complementary investigation conducted in the United States with active cigarette smokers utilizing the reactivation-extinction procedure demonstrated that participants in the retrieval-extinction group experienced rapid and sustained reductions in smoking frequency, alongside more pronounced and enduring decreases in cue-induced craving [23]. These findings suggest cross-cultural and cross-regional potential for the reactivation-extinction procedure in disrupting maladaptive addiction-related memories. We did not employ urine drug screens (UDS) as a direct measure of METH use, despite their superior reliability as proxies for relapse. Instead, we selected cue-induced craving as our primary outcome measure because the study was conducted within a drug rehabilitation center where participants had no access to METH or other substances. UDS might not adequately capture intervention-induced changes in the present study. Hulse et al. demonstrated that craving can function as a predictor of relapse to heroin use, with their findings indicating that a mere one-point increase in an individual’s craving score from one month to the next was associated with a 5% elevation in heroin use risk [55]. An additional minor limitation concerns our instruction for participants to abstain from smoking for two hours prior to testing, which could potentially introduce the confounding variable of nicotine withdrawal.

Conclusions

The CS-triggered retrieval-extinction training for individuals with MUD resulted in a significant attenuation of cue-elicited craving and saliva cortisol levels. While the absence of measurable changes in BP was somewhat unexpected. These findings provide compelling evidence that even a brief reconsolidation-based intervention can effectively reduce METH-related craving and cortisol levels, highlighting its potential as an adjunctive approach in MUD treatment. Moreover, salivary cortisol serves as an accessible and sensitive biomarker for evaluating intervention effects.

Author contributions

LL, PW, and YXX conceived and designed the study. JLY, RJW, SJC, XL, QF, XJG, and YKS carried out the study and collected data. JLY, RJW and YPB analyzed and interpreted the data. JLY drafted the manuscript. KY, JS, YXX, PW, and LL were involved in the supervision of the study. All authors contributed to the writing and approved the final version of the manuscript.

Data availability

Data are available upon reasonable request to the corresponding authors.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-025-03474-5.