Skip to main content
NCBI home page
Alcohol and Alcoholism (Oxford, Oxfordshire) logoLink to Alcohol and Alcoholism (Oxford, Oxfordshire)
. 2026 Jun 10;61(4):agag035. doi: 10.1093/alcalc/agag035

Functional magnetic response imaging predictors of alcohol use disorder treatment outcome: a systematic review

Mahmoud Elsayed 1,2, Ryan Siroosi 3,4, Peter Najdzionek 5,6, Carly McIntyre-Wood 7,8, Tegan Hargreaves 9,10, Ashley Blakely 11,12, Emily Mote 13,14, James MacKillop 15,16,✉
PMCID: PMC13253039  PMID: 42268834

Abstract

Background

Despite evidence-based pharmacological and behavioural interventions, alcohol use disorder (AUD) is associated with highly variable treatment outcomes. Functional magnetic resonance imaging (fMRI) may identify neural markers that predict treatment response, ultimately supporting a precision medicine approach to AUD.

Objectives

This systematic review synthesized evidence on fMRI predictors of treatment outcomes in individuals with AUD, evaluated methodological consistency, and identified gaps to guide biomarker development.

Methods

A comprehensive search of PubMed/MEDLINE, Embase, and PsycINFO combined terms related to fMRI, AUD, and treatment outcomes. Eligible studies included participants with AUD receiving pharmacological, behavioural, or neuromodulatory interventions with fMRI measures collected before or early in treatment to predict clinical outcomes. Screening and extraction were conducted in duplicate using Covidence, and study quality was assessed with the Grading of Recommendations Assessment, Development, and Evaluation framework.

Results

Of 342 records, 15 studies met the inclusion criteria. Most used alcohol cue reactivity tasks (k = 11), with others using resting-state fMRI (k = 2), a monetary reward task (k = 1), or an alcohol-specific Go/No-Go task (k = 1). Pharmacological treatments were most common (k = 8), followed by behavioural therapies (k = 6) and one neuromodulation trial. Across paradigms, neural activity in the ventral striatum, orbitofrontal cortex, and anterior cingulate cortex commonly predicted outcomes. Greater prefrontal engagement predicted improvement, while heightened striatal cue reactivity predicted relapse. Resting-state findings suggested reduced reward- and stress–network connectivity corresponded with better outcomes. Across studies, however, considerable heterogeneity and inconsistency were present and sample sizes tended to be small.

Conclusions

Evidence implicates frontostriatal and salience circuitry in predicting AUD treatment outcomes, but inconsistency and underpowered studies limit firm conclusions. Larger longitudinal studies are needed to robustly validate clinically useful biomarkers.

Keywords: alcohol use disorder, fMRI, treatment outcome, biomarkers, precision medicine

Short summary Functional MRI shows promise for predicting treatment outcomes in alcohol use disorder, with frontostriatal and salience network activity linked to prognosis. However, methodological heterogeneity and small samples limit conclusions, highlighting the need for larger-scale studies to develop reliable biomarkers.

Introduction

Alcohol use disorder (AUD) is characterized by compulsive alcohol seeking, loss of control over drinking, and continued use despite harmful consequences. It remains one of the most prevalent and burdensome psychiatric disorders worldwide (MacKillop et al. 2022, Kranzler 2023). According to the World Health Organization, harmful alcohol use contributes to over 3 million deaths annually and accounts for nearly 5% of the global burden of disease (World Health Organization 2024). Despite the availability of evidence-based treatments, such as cognitive behavioural therapy (CBT); pharmacotherapies, including naltrexone, acamprosate, and disulfiram; and emerging neuromodulation techniques, treatment outcomes for AUD remain highly variable (Magill et al. 2023, McPheeters et al. 2023, Mehta et al. 2024). Relapse rates often exceed 60% within the first 6 months following treatment completion (Durazzo and Meyerhoff 2017, Nguyen et al. 2020). This variability highlights an urgent need to identify reliable biomarkers that can predict treatment response, optimize intervention selection, and ultimately personalize care (Boness and Witkiewitz 2022).

Neuroimaging, and functional magnetic resonance imaging (fMRI) in particular, has provided valuable insights into the neurobiological mechanisms underlying addiction (Volkow et al. 2012, Fang et al. 2022). More specifically, fMRI enables noninvasive assessment of neural activity and connectivity associated with cognitive, emotional, and motivational processes central to addiction, including craving, inhibitory control, and reward valuation (Strosche et al. 2021, Weinstein 2023, Elsayed et al. 2024, 2025, Nawawi et al. 2024). Converging evidence suggests that AUD is associated with dysregulation within fronto-striatal-limbic circuits encompassing the ventral striatum, medial prefrontal cortex (mPFC), orbitofrontal cortex (OFC), anterior cingulate cortex (ACC), and insula, regions implicated in reward processing, salience attribution, and self-regulation (Koob and Volkow 2009, Suckling and Nestor 2016, Roberge et al. 2025). Functional abnormalities in these networks have been observed both during active drinking and early abstinence, suggesting that they may serve as state-dependent markers of relapse risk (Heinz et al. 2008, Zilverstand et al. 2018, Zheng et al. 2024).

Recent fMRI studies have begun to explore whether neural activation patterns measured before or early during treatment can prospectively predict therapeutic outcomes in AUD. Several cue-reactivity studies have shown that elevated activation in reward-related areas such as the ventral striatum or caudate prior to or early in treatment is associated with faster relapse or heavier post-treatment drinking (Mann et al. 2014, Schacht et al. 2017, Blaine et al. 2020). In contrast, greater engagement of prefrontal control regions or attenuation of cue-elicited responses over time has been linked to improved treatment success and prolonged abstinence (Becker et al. 2018, Bartholdy et al. 2019, Bach et al. 2021, Wetherill et al. 2021). Beyond task-based paradigms, resting-state studies have identified functional connectivity within large-scale networks, such as the default mode, salience, and frontoparietal control systems, that differentiate treatment responders from nonresponders (Srivastava et al. 2021, Syan et al. 2023). Together, these findings suggest that both task-evoked and intrinsic neural dynamics may provide biomarkers for predicting and understanding individual differences in therapeutic response among individuals with AUD. However, findings across studies have not been entirely consistent, with some investigations reporting weak or nonsignificant associations between cue-reactivity measures and subsequent treatment outcomes and others observing predictive effects that vary depending on analytic approach, sample characteristics, or stage of treatment (Schacht et al. 2017, Becker et al. 2018, Wetherill et al. 2021). Moreover, discrepancies have emerged regarding which specific neural circuits or connectivity patterns most reliably predict response, highlighting ongoing uncertainty about the robustness and generalizability of proposed neurobiological markers (Srivastava et al. 2021, Syan et al. 2023).

Thus, despite a growing evidence base, the literature on fMRI predictors of AUD treatment outcomes is inconsistent and, in some cases, conflicting, in part due to heterogeneity in designs. Therefore, a systematic review is warranted to synthesize current evidence on functional MRI predictors of AUD treatment outcomes, parse findings across studies, identify methodological strengths and limitations, and outline future research directions to guide biomarker development and ultimately advance precision medicine for AUD.

Methods

Search strategy

Relevant primary research was systematically sourced from two major databases (i.e. PubMed/MEDLINE, Embase, and PsycINFO) through the Ovid search engine (https://ovidsp.ovid.com/). A comprehensive search strategy was employed, encompassing terms relevant to functional MRI and AUD treatment outcome using Boolean operators (see Table 1 for details).

Table 1.

Search strategy.

  1. ‘Alcohol Dependence’

 17. ‘Treatment’

  1. ‘Alcohol Addiction’

 18. ‘Recover*’

  1. ‘Problem Drinking’

 19. ‘Remission’

  1. ‘Excessive Drinking’

 20. ‘Intervention’

  1. ‘Heavy Episodic Drinking’

 21. ‘Treatment Response’

  1. ‘Compulsive Drinking’

 22. ‘Treatment Outcome’

  1. ‘Alcohol abuse’

 23. ‘longit*’

  1. ‘Alcohol misuse’

 24. ‘prospect*’

  1. ‘Alcohol use disorder’

 25. ‘predict*’

  1. ‘Binge drinking’

 26. ‘prognos*’

  1. ‘Harmful drinking’

 27. ‘Trial’

  1. ‘functional magnetic resonance imaging’

 28. Alcohol Use Bucket: 1 OR 2 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10 OR 11

  1. ‘fMRI’

 29. Functional MRI Bucket: 12 OR 13

  1. ‘Relapse’

 30. Treatment Outcome Bucket: 14 OR 15 OR 16 OR 17 OR 18 OR 19 OR 20 OR 21 OR 22

  1. ‘quit*’

 31. Study Type Bucket: 23 OR 24 OR 25 OR 26 OR 27

  1. ‘Abstinence’

 Final Search Terms: 28 AND 29 AND 30 AND 31
Open in a new tab

Inclusion/exclusion criteria

Studies were included if they met all of the following criteria: human participants with a diagnosed AUD; some form of treatment or intervention (pharmacological, behavioural, or neuromodulatory); and utilization of an fMRI measure (resting-state or task-based) to statistically predict treatment outcomes. The fMRI measure had to be obtained either before the start of treatment or within the first half of the treatment period. Additional inclusion criteria required that the study be published in a peer-reviewed journal and written in English. Studies that failed to meet one or more of these criteria were excluded.

Studies that compared pre- and post-treatment brain activity, or examined time-by-brain interactions without directly linking imaging findings to clinical outcomes, were excluded for this reason. Similarly, randomized control trials (RCTs) that reported group-level fMRI differences without a statistical association with treatment outcomes were not included. Finally, studies that used craving as the sole outcome were excluded, as craving is not typically considered a primary clinical endpoint in treatment research, in contrast to measures such as drinking frequency, amount consumed, symptom severity, or relapse/abstinence.

Review and data extraction procedures

All identified studies underwent title and abstract screening by M.E. and were simultaneously reviewed by one of four additional reviewers (C.M.W., T.H., P.N., or R.S.), ensuring that each study was independently screened by two reviewers, M.E. and one of the aforementioned reviewers, according to the predetermined inclusion and exclusion criteria. Any discrepancies between reviewers were resolved by consensus among M.E., C.M.W., T.H., P.N., and R.S. The full-text screening followed the same procedure. The number of included and excluded papers was documented using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram.

All screening procedures were conducted within Covidence, an online systematic review management platform (https://www.covidence.org/). Data extraction was performed using a customized Excel spreadsheet. Each study was independently extracted by two data extractors: M.E. and a second extractor (R.S., A.B., or E.M.). The extraction tables were subsequently reconciled by M.E. and the other data extractor. The extraction table was then reviewed by either C.M.W. or P.N. to ensure the accuracy and completeness of the extracted information.

Quality and risk of bias assessment

Two independent reviewers evaluated the overall quality of evidence using the Grading of Recommendations Assessment, Development, and Evaluation framework, which encompasses five domains: risk of bias, publication bias, consistency, directness, and precision (Guyatt et al. 2011). The mean of the two quality scores was calculated. Detailed evaluation criteria are provided in the supplementary materials.

Results

Overview

Of 342 studies identified in the search, 300 were excluded as ineligible during the title and abstract screening stage, and an additional 27 were excluded during the full-text review, resulting in 15 eligible studies (see Fig. 1 for the PRISMA diagram). These studies varied in design, treatment modality, analytic methods, and timing of neuroimaging assessments but collectively examined both task-based and resting-state functional correlates of treatment response. Of the 15 studies, 11 based their neurocorrelates on an alcohol cue-reactivity task, 2 utilized resting-state fMRI, 1 employed a monetary reward task, and 1 used an alcohol-specific Go/No-Go task. In terms of treatment type, the majority of studies investigated pharmacological therapy (8 studies), followed by behavioural interventions such as CBT, motivational interviewing, and cue exposure therapy (6 studies). Only one study examined a neuromodulation-based treatment (see Table 2 and Fig. 2).

Figure 1.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

PRISMA diagram. Note. Studies excluded for ineligible outcomes or indicators were those that assessed outcomes such as craving without formally reporting the number of alcohol use disorder (AUD) symptoms, drinking intensity, or drinking frequency, as well as studies that focused on mechanistic treatment biomarkers rather than predictive biomarkers. Studies excluded for ineligible design were those in which scanning was conducted after the midpoint of the treatment period, or those addressing a different research question, such as comparisons between responders and nonresponders, without directly linking fMRI findings to treatment outcomes.

Table 2.

The functional MRI predictors of the AUD treatment outcome.

# Study Sample characteristics AUD characteristics at baseline Intervention AUD characteristics at follow-up fMRI Analysis type Brief summary of findings
Cue-reactivity Task fMRI
1 Schacht et al. 2013 N = 48 alcohol-dependent participants (Lower AW + Placebo = 14; Lower AW + BGP/FMZ = 22; Higher AW + Placebo = 6; Higher AW + GBP/FMZ = 6); age 18-70; mostly male (Lower AW + Placebo 78.6%; Lower AW + BGP/FMZ 77.3%; Higher AW + Placebo 66.7%; Higher AW + GBP/FMZ 83.3%). Alcohol Dependence Scale: (Lower AW + Placebo 12.4 ± 6.0; Lower AW + BGP/FMZ 13.4 ± 5.2; Higher AW + Placebo 13.0 ± 3.5; Higher AW + GBP/FMZ 18.5 ± 7.1).
Obsessive Compulsive Drinking Scale: (Lower AW + Placebo 13.9 ± 4.8; Lower AW + BGP/FMZ 15.0 ± 4.1; Higher AW + Placebo 21.2 ± 4.9; Higher AW + GBP/FMZ 19.3 ± 6.0). HDD on at least 70% of days in month prior to screening (5+ drinks per day): (Lower AW + Placebo 91.7 ± 9.6%; Lower AW + GBP/FMZ 86.8 ± 14.8%; Higher AW + Placebo 90.4 ± 18.9%; Higher AW + GBP/FMZ 92.4 ± 9.7%).		 6-week pharmacotherapy treatment.
Treatment group: daily oral GBP (300–1200 mg) for 6 weeks, FMZ (2 mg) infusion on Days 1 and 2.
Placebo group: daily oral placebo for 6 weeks, placebo infusions on Days 1 and 2.		 %HDD post-scan: Lower AW + Placebo 10.6(19.7); Lower AW + GBP/FMZ 23.1(25.5); Higher AW + Placebo 8.4(13.0); Higher AW + GBP/FMZ 4.6(9.7). 3T; A 720-s alcohol cue reactivity task utilizing 24 pseudorandom blocks of alcoholic beverage images, nonalcoholic beverage images, blurred control images, and a fixation cross. Scanned during the second or third week of the 39-day medication treatment (mean scan day = 15). Whole-brain general linear modelling. Greater activations in left middle frontal and precentral gyri (DLPFC) during alcohol cues were associated with higher post-scan %HDD.
2 Mann et al. 2014 64 Alcohol dependents who had been detoxified 14–21 days; 36 treated with NTX (Mage = 45 ± 10; 80.55% males); and 28 treated with acamprosate (Mage = 41 ± 6; 78.57% males) Drinks/day: NTX 12.3 ± 6.4, Acamprosate 13.9 ± 10.6, drink/day was defined as 12 g alcohol; AUDIT: NTX 27.0 ± 6.0, Acamprosate 24.9 ± 7.4; ADS: NTX 15.2 ± 6.7, Acamprosate 14.0 ± 7.1; AUQ: NTX 11.8 ± 6.3, Acamprosate11.2 ± 5.1; OCDS: NTX 14.1 ± 7.4, Acamprosate 12.9 ± 6.9; Patients were
randomized into
3 groups:
NTX (50 mg/day), acamprosate (2 g/day),
placebo group in proportions 2:2:1 for 12 weeks; patients received counselling over 6 months and additional follow-up assessment for another year, resulting in an observation period of 18 months after treatment initiation.		 Abstinence rates during the 12 weeks’ period of medical treatment did not differ between NTX (50.0%) and acamprosate (53.6%) patients. 1.5T; Visual cue perception task in a block design featuring alcohol-related and neutral stimuli (3 pictures per block, presented for 19.8 s). Scanned after withdrawal symptoms subsided but before the first dose of randomized medication was administered. ROI-based analysis restricted to the ventral striatum. NTX-treated patients with high VS cue reactivity prior to treatment showed significantly better outcomes than patients on NTX showing less VS cue reactivity. No association between VS cue reactivity and time to relapse for patients assigned to acamprosate
3 Reinhard et al. 2015 Subsample of (Mann et al. 2014) N = 49; 20.4% Female; Mage = 44 ± 9; Ages ranged between 27 andd63 years. Alcohol consumption immediately prior to detoxification: 151 ± 83 g /day Same as Mann et al. 2014 Days until first severe relapse: 54 ± 27 days. 20/49 (40.8%) suffered severe relapse. 1.5 T; Block design task with 15 alcohol-associated, 15 affectively neutral, and 15 abstract stimuli (19.8 s per block). Scanned 2 weeks after finishing a 3-week inpatient alcohol-detoxification programme, before randomization to outpatient medication. ROI-based (comparing aggregation measures in the ventral striatum, orbitofrontal cortex, and vACC). Cue-induced activation in the VS, OFC, and vACC was negatively associated with time to relapse, so greater cue reactivity predicted higher relapse risk.
VS cue-reactivity was the strongest and most consistent prognostic factor, explaining the most variance in relapse outcomes, while OFC and vACC effects were weaker and less reliable.
4 Herremans et al. 2016 N = 19 detoxified Alcohol dependent patients; 57.8% Male; age 18–65 Number of HDDs (during the last month before hospitalization): Relapsers 17.7 ± 10.6
Abstainers: 20 ± 10.0; Duration of alcohol addiction (years):
Relapsers: 14.5 ± 9.9
Abstainers: 9.8 ± 9.6		 All patients underwent a placebo-controlled HF-rTMS protocol targeting the DLPFC. Half of participants received active stimulation on first day and sham on second; half reverse order. Participants underwent 14 more accelerated sessions (15 total between Mon to Fri; 23,400 pulses). 68% of participants had relapsed within 4-weeks. 3T; Block paradigm (7 blocks of 5 alcohol/neutral stimuli) and an event-related cue exposure paradigm (48 alcohol/48 neutral stimuli presented randomly). Scanned at baseline (before stimulation), after the first single session, and a final scan after 15 accelerated HF-rTMS sessions spread over 3 days. ROI-based analysis (full factorial group analysis using an ROI mask). Abstainers had higher left dACC activation at baseline
compared to the relapsers. dACC activation was significantly decreased in abstainers and dACC activation was increased in relapsers following HF-rTMS. Rate dependent negative effect (i.e. a pattern in which the treatment effect is inversely related to the baseline level): Low rates of activation considerably increased, while high rates of activation could be unaffected or even decreased by the accelerated
HF-rTMS protocol.
5 Schacht et al. 2017 N = 146 treatment-seeking alcohol-dependent participants; Mage = 49.3 ± 10.1 (18-70); 69.2% male. Drinks per drinking day = 11.2 ± 4.8; Drinks per day = 9.6 ± 5.0; %HDD = 79.7 ± 22.3; Alcohol Dependence Scale = 15.4 ± 6.4; Obsessive Compulsive Drinking Scale = 25.6 ± 8.1; Drinker Inventory of Consequences = 41.4 ± 18.6. Pharmacotherapy; 16-week RCT; NTX (25 mg for first 2 days, 50 mg remainder) or placebo; follow-up medical management (MM) sessions at 9 timepoints. %HDD; assessed by TLFB at MM sessions; participants who received NTX engaged in fewer #HDD during RCT than those who received placebo (P = .023). 3T; A 12-min task passively viewing pseudorandom blocks of alcoholic beverages, nonalcoholic beverages, blurred control images, and a fixation cross. Scanned at baseline (immediately before the first medication dose) and following 2 to 6 weeks of medication treatment. ROI-based analysis (right and left ventral striatum). A greater reduction in the VS activation to the alcohol cues, in participants treated with NTX but not placebo, showed fewer #HDD.
6 Bach et al. 2020 N = 55 AUD+ group, N = 35 HC group, all males.
Mage:
AUD+ group = 45.6 ± 8.9;
HC group = 42.0 ± 9.8.		 Assessed over 90 days prior to detox: AUD + Ethanol consumption: 202.7 ± 134.6 g/day.
ADS: 14.2 ± 6.6.
OCDS: 16.6 ± 6.8.
Drinks/day: IWT 21.3 ± 15.5; IWT + NTX 15.0 ± 11.1.
%HDD: IWT 79.9 ± 29.9; IWT + NTX 76.4 ± 27.0.		 All patients were administered IWT (21-day detox programme)
25 patients also received adjunct NTX in a naturalistic open-label design (duration and dose not provided in the study). For longitudinal fMRI: IWT n = 13, IWT + NTX n = 22.		 Relapse to heavy drinking: IWT 25/29 (86%) vs IWT + NTX 13/23 (57%). Drinks/day: .9 ± 2.1 vs .1 ± .1. Abstinent days: 88.3 ± 21.7% vs 98.2 ± 5.0%. %HDD (trend): 8.1 ± 20.2% vs .6 ± 2.1%. 3T; Visual block-design alcohol cue-reactivity task presenting 12 blocks of alcohol pictures and 12 blocks of neutral pictures. Baseline scan scheduled after 2–4 weeks of controlled abstinence; follow-up scan scheduled 2 weeks into treatment. Whole-brain analyses followed by ROI analysis (left putamen). Higher baseline left putamen cue reactivity predicted an earlier return to heavy drinking. Changes over two weeks were also informative: increasing reactivity predicted earlier relapse, whereas decreasing reactivity predicted a longer time to relapse.
7 Blaine et al. 2020 N = 69 treatment-entering AUD patients; aged 21–60 years; 61% male Years of alcohol use = 14.2 ± 1.3
Days of alcohol use per month = 23.6 ± 4.7; Drinks per month = 139 ± 26.2		 Weekly behavioural counselling sessions with a training master’s-level alcohol and drug abuse counsellor utilizing standardized 12-step facilitation therapy; duration of treatment not specified but outcomes were measured during the first 2 weeks of treatment engagement. Participants drank on an average of 5.20 ± 4.36 days; range = 0–14) during the 2-week period, with an average of 4.58 ± 4.52 drinks per drinking day 3T; Sustained emotion/reward provocation task presenting successive blocks of highly stressful, alcohol cue, and neutral-relaxing images. Scanned during the intake phase, before the initiation of outpatient treatment. Voxelwise whole-brain analysis. lower vmPFC/rACC hypoactivation to
Stress images (stress-neutral contrast) and greater
vmPFC/rACC prefrontal hyperactivity in the neutral state (neutral-baseline contrast) predicted increased heavy-drinking days during early treatment. Furthermore, lower ventral
striatal hypoactivation in response to alcohol cues (alcohol cues–neutral contrast) was predictive of more heavy-drinking days during early treatment.
8 Bach et al. 2021 N = 44 males with DSM-IV alcohol dependence; IWT+ NTX = 22, IWT = 22; aged 18–65 Assessed over 90 days prior to detox: Ethanol consumption (g/day): IWT 276.3 ± 128.1; IWT + NTX 217 ± 127.5. Drinks/day: IWT 18.8 ± 11.3; IWT + NTX 15.6 ± 20.6.
%HDD: IWT 81.9 ± 23.5; IWT + NTX 76.9 ± 26.6.		 Naturalistic open-label IWT (21-day multiprofessional programme) with or without add-on oral naltrexone 50 mg daily, chosen by patients after baseline 32/44 (72%) of participants relapsed to heavy drinking within 90 days after baseline assessment. 20/22 in the IWT (90%) and 12/22 (55%) in the IWT + NTX group relapsed. 3T; Visual block-design cue-reactivity task featuring 12 blocks of neutral pictures and 12 blocks of alcohol pictures. Baseline scan before the start of treatment. ROI-based analysis (ventral striatum). Greater baseline cue–induced ventral striatal activation predicted a longer time to heavy relapse among patients receiving naltrexone, but across the full sample, higher ventral striatum activation was associated with a higher relapse risk, which implies poorer outcomes under IWT-only, and possible medication interaction.
9 Wetherill et al. 2021 N = 20 treatment-seeking patients with AUD. Placebo = 8 (Mage = 45.1 ± 8.1;75% males); Topiramate = 12 (Mage = 50.5 ± 8.1; 66.67% male) Topiramate group: drinking days in past month = 24.3 ± 5.1; drinks per drinking day in past month = 7.0 ± 3.3; #HDD in past month = 19.9 ± 7.7.
Placebo group: drinking days in past month = 27.0 ± 3.5; drinks per drinking day in past month = 5.8 ± 1.6; #HDD in past month = 17.9 ± 8.0.		 Pharmacotherapy; 12-week RCT of topiramate; max dose of 100 mg 2× per day; medical management sessions with physician or nurse practitioner weekly for first 6 weeks, then biweekly to increase adherence. Participants treated with topiramate, compared to placebo, reported significant reduction in alcohol-cue induced craving (z = −2.00, P = .05) and #HDD (z = −2.10, P = .04) between baseline and second scan. 3T; Pseudo-continuous arterial spin labelling (pCASL) perfusion fMRI while viewing a 10-min alcohol cue video and a 10-min nonalcohol cue video. Scanned at baseline (before starting medication) and after 6 weeks of double-blind treatment. ROI-based analysis (bilateral and medial orbitofrontal cortex, and bilateral ventral striatum). Reduction in right VS, right OFC, and medial OFC associated with fewer HDD.
10 Logge et al. 2021 N = 30; Placebo 45.45% Female; Mage = 51.73 ± 12.06; BAC 26.32% Female; Mage = 49.23 ± 10. Drinks per drinking day: Placebo 12.12 ± 4.34, BAC 10.58 ± 5.48; Years since alcohol-related problems began: Placebo 20 ± 13.42, BAC 15.83 ± 11.2; ADS score: Placebo13.73 ± 6.18, BAC 16.07 ± 7.65; PACS craving: Placebo 17.91 ± 5.11 BAC 16.7 ± 5.34 Participants were
randomized into
3 groups:
placebo, low-dose baclofen
(30 mg/day), high
(75 mg/day) for
12 weeks.		 Post-scan % HDD: Placebo 56.9 ± 36.33 BAC 32.2 ± 33.34 3T; Visual alcohol cue reactivity task utilizing blocks of alcohol-related pictures, neutral pictures, and scrambled control images. Scanned at approximately Week 3 of the trial, 120 min following treatment administration. Whole-brain general linear models (GLM). Placebo vs. 75 mg/day Baclofen: Post-scan %HDD positively correlated with increased cue-reactivity in a cluster encompassing the bilateral caudate nucleus and dorsal anterior cingulate cortex (dACC). Additionally, placebo vs. 30 mg/day Baclofen: Correlation between higher post-scan %HDD and cue-reactivity involved clusters in the left postcentral gyrus, insula, superior frontal gyrus, mid cingulate cortex, and precentral gyrus
11 Naqvi et al. 2024 N = 22; 36.4% Female; Treatment-seeking participants with AUD; Mage = 46.68 ± 12.4. %HDD as determined by TLFB: 81.5 ± 15.3; Baseline ADS: 13.6 ± 6.74 Participants received up to 12 CBT sessions lasting 1 h each. 10/22 (45.45%) Participants ceased heavy drinking throughout the CBT sessions, self-reported craving was significantly greater at the pre-CBT timepoint than post-CBT timepoint (t(87) = 5.35; P < .0001), 3 T; Regulation of craving cue-reactivity task directing participants to focus on immediate vs. long-term negative consequences, followed by picture cues of alcoholic beverages or high-calorie food. Scanned before and after a 12-week trial of cognitive behavioural therapy (CBT). Parcel-wise whole-brain analysis and ROI-based analysis (DLPFC parcels). Among participants who ceased heavy drinking, cue-induced craving was consistently associated with cue-induced activity in the left DLPFC, at both pre- and post-CBT assessments
Noncue-reactivity Task fMRI
12 Becker et al. 2018 N = 58 detoxified AUD patients (CET + TAU = 40; TAU = 18); aged 18–65; 77.59% males (CET = 70.0%, TAU = 94.4%). DSM criteria count: CET + TAU 6.1 ± 1.0; TAU 5.8 ± 1.4.
ADS: CET + TAU 15.9 ± 6.7; TAU 15.8 ± 7.4. Alcohol Abstinence Self-Efficacy Scale (AASE) – Temptation: CET+ TAU 31.8 ± 17.4; TAU 29.3 ± 18.2. AASE – Self-efficacy to abstain: CET + TAU 58.5 ± 17.5; TAU 53.8 ± 24.9. Baseline %HDD: CET + TAU 20.4 ± 27.4; TAU 21.2 ± 17.5. Baseline drinks/ drinking day: CET + TAU 18.6 ± 13.7; TAU 24.6 ± 25.1		 All patients: 3-week inpatient. TAU group: consisted of detoxification treatment, health education, psychotherapy, and occupational and sociotherapy. CET group: 5–9 individual CET sessions (~30–90 min) AASE (post-treatment) collected; group means not reported. 3T; monetary reward anticipation task reacting to a flashed light preceded by an arrow indicating potential monetary reward, punishment avoidance, or verbal feedback. Scanned at baseline and after 3 weeks of the treatment intervention. ROI-based analyses (ventral striatum). No significant results were found in the association between neural activities and drinking behaviour after the clinic release. However, greater baseline reward sensitivity, which showed larger increases in frontal activation (including superior frontal gyrus and anterior cingulate cortex), was associated with higher post-treatment self-efficacy to abstain and lower temptation to drink
13 Grieder et al. 2022 N = 45;
AUD Alc-IT = 25 (9F/16M); Mage = 43.6 ± 10.4.
AUD Control training = 20 (7F/13M); Mage = 43.1 ± 8.2.		 AUDIT: Alc-IT = 24.6 ± 6.9, Control = 24.9 ± 7.5; years of problematic drinking: Alc-IT = 13.3 ± 14.1, control = 11.9 ± 9.7; OCDSsum: Alc-IT = 23.8 ± 7.5, control = 23.9 ± 8.1; AUD-S: Alc-IT = 28.1 ± 7.5, control = 30.0 ± 7.9. All patients: 12-week residential, abstinence-oriented TAU (components NR). Alc-IT group: computerized alcohol-specific inhibition training, 6 sessions of ~10–15 min for weeks 4–5 of treatment, Control group: active nonspecific inhibition training, 6 sessions of ~10–15 min during weeks 4–5. Change in percentage of days abstinent at 3-month follow-up:
Alc-IT = 72.5 ± 20.8, n = 22, Control = 56.1 ± 45.9, n = 15		 3 T; Go-NoGo event-related task utilizing an alcoholic bottle with a glass or a neutral (water) bottle with a glass. Scanned at a pretraining assessment (1–2 weeks post-baseline) and a post-training assessment (1–4 days after the last training session). Whole-brain linear regression with resulting regions extracted as ROIs. Higher alcohol specific inhibition training activation in the rIFG (pars opercularis) was associated with better drinking outcomes
Resting-state fMRI
14 Syan et al. 2023 N = 46 participants with AUD; 65.2% female; Mage = 34.07 ± 10.70. Number of AUD symptoms = 6.57 ± 2.58. Baseline Drinks/week: Male Responders = 34.25, Male Nonresponders = 39.83, Female Responders = 20.29, Female Nonresponders = 18.69. Brief motivational interview; each session lasts 30–45 min. 56.52% of participants were responders (WHO risk drinking level decreased by ≥1 level); Follow Up Drinks/Week: Male Responders = 15.48; Male Nonresponders = 39.22, Female Responders = 6.37, Female Nonresponders = 23.99. 3T; Resting-state fMRI involving passively observing a fixation cross for 9 min. Baseline scan immediately followed by the brief intervention session. Seed-to-voxel whole-brain analysis using predefined network regions of interest. Treatment responders showed greater resting state functional connectivity than nonresponders between: ACC (seed) and left postcentral gyrus/right supramarginal gyrus; right posterior parietal cortex (seed) and right vACC; right IFG pars opercularis (seed) and right cerebellum/right occipital fusiform gyrus;
Treatment responders showed lower resting state functional connectivity than nonresponders between right rostral PFC (seed) and left IFG pars opercularis; left IFG pars triangularis (seed) and right cerebellum; right IFG pars triangularis (seed) and right frontal eye fields/right angular gyrus; right NA (seed) and right OFC/right insula.
15 Srivastava et al. 2021 N = 18; 44.4% Female; treatment-seeking participants with AUD; 21–65; Mage = 47.11 ± 11.09 #HDD as determined by TLFB 28 days: 22.23 ± 4.20 Participants received up to 12 weekly CBT sessions lasting 1 h each. #HDD was significantly reduced from pre- to post-CBT; 4.22 ± 5.89 3T; Resting state fMRI in which participants were instructed to rest with eyes open while looking at a fixation cross for 10 min. Scanned at baseline (pre-CBT) and after the completion of 12 weeks of outpatient CBT (post-CBT). ROI-to-ROI functional connectivity analysis (BNST and insula regions). Reduction in RSFC between the dorsal AI and BNST from pre- to post-CBT was associated with reductions in #HDD from pre- to post-CBT.
Open in a new tab

Abbreviations: OCDSsum, sum of Alcohol Compulsive Drinking Scale; AUDIT, Alcohol Use Disorder Identification Test; TLFB, Timeline Follow-Back; HDD, heavy-drinking days; AW, alcohol withdrawal; IWT, intensive withdrawal training; NTX, naltrexone; BAC, baclofen; GBP, gabapentin; FMZ, flumazenil; CBT, cognitive behavioural therapy; CET, cue exposure–based extinction training; TAU, treatment as usual; Alc-IT, alcohol-specific inhibition training; RCT, randomized clinical trial; AW, alcohol withdrawal; PFC, prefrontal cortex; vmPFC, ventromedial prefrontal cortex; DLPFC, dorsolateral prefrontal cortex; rACC, right anterior cingulate cortex; vACC, ventral anterior cingulate cortex; IFG, inferior frontal gyrus; NA, nucleus accumbens; HC, healthy control; ADS, alcohol dependency scale; VS, ventral striatum; OFC, orbitofrontal cortex; T, Tesla, demonstrating the scanner’s field strength.

Figure 2.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

Distribution of studies across different treatments. Abbreviations: MI, motivational interviewing; CET, cue exposure training; Alc-IT, alcohol-specific inhibition training; BC, behavioural counselling, CBT, cognitive behavioural therapy.

Task-based fMRI predictors

Eleven studies employed alcohol cue-reactivity tasks to assess neural activation related to relapse and treatment response. Across pharmacological interventions (including naltrexone, acamprosate, baclofen, gabapentin/flumazenil, and topiramate), task-based studies commonly identified activity within striatal and prefrontal regions as related to clinical outcomes. Specifically, cue-elicited activation in the ventral striatum, putamen, OFC, and ACC at baseline or following treatment was associated with subsequent relapse risk or reductions in heavy drinking. The association with ventral striatum was the most reported, appearing in 54.54% (6 out of 11) of cue-reactivity studies, though the direction of this association was not consistent; see Table 2 and Fig. 3. Beyond cue reactivity, Becker et al. (2018) applied a reward anticipation task and found that baseline striatal reward sensitivity predicted treatment-related increases in frontal cortical activation and improvements in abstinence self-efficacy. Grieder et al. (2022) used an alcohol-specific Go/No-Go inhibition task and identified right inferior frontal gyrus activation as a functional correlate of better drinking outcomes following inhibition training. Similarly, Herremans et al. (2016) observed differential dorsal ACC activation changes following high-frequency repetitive transcranial magnetic stimulation (rTMS) that corresponded with relapse status, see Table 2, and Fig. 3.

Figure 3.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

Number of implications for each brain region in the included studies. Abbreviations: BNST, bed nucleus of the stria terminalis; PPC, posterior parietal cortex; rPFC, rostral prefrontal cortex; IFG, inferior frontal gyrus; left DLPFC, left dorsolateral prefrontal cortex; vCC, ventral cingulate cortex; mCC, middle cingulate cortex; OFC, orbitofrontal cortex; ACC, anterior cingulate cortex; sFG, superior frontal gyrus; VS, ventral striatum. Other regions include the supramarginal gyrus, fusiform gyrus, cerebellum, primary motor cortex, thalamus, precentral gyrus, postcentral gyrus, insula, caudate nucleus, and putamen.

Resting-state fMRI predictors

Two studies used resting-state functional connectivity (rsFC) to evaluate treatment effects or predictive markers. Srivastava et al. (2021) observed that decreased connectivity between the anterior insula and bed nucleus of the stria terminalis (BNST) accompanied reductions in heavy drinking after CBT. Syan et al. (2023) found that baseline resting-state connectivity within default mode, frontoparietal, and salience networks differentiated responders from nonresponders to a brief intervention. See Supplementary Material 2 for an interactive summary of the results.

Quality and risk of bias assessment

Quality ratings (Fig. 4) indicated that most studies achieved moderate to high quality across domains of bias, directness, precision, and outcome measurement.

Figure 4.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

Quality and Risk of Bias Assessment. Note: See the Supplementary Material 1 for the full scale.

Discussion

The reviewed literature collectively demonstrated that fMRI markers provide meaningful information about treatment outcomes in AUD. Across 15 studies, both task-based and resting-state fMRI paradigms identified patterns of brain activation and connectivity associated with relapse, heavy-drinking days, or abstinence maintenance. Task-based investigations, particularly those using alcohol cue-reactivity paradigms, consistently implicated the ventral striatum, OFC, and ACC as central regions whose activation at baseline or changes during treatment predicted subsequent drinking behaviour. In several studies involving pharmacotherapies such as naltrexone, acamprosate, baclofen, topiramate, and gabapentin–flumazenil combinations, modulation of these regions during treatment corresponded with reductions in alcohol consumption. However, the consistency of predictive regions appeared to vary by medication mechanism: studies of opioid antagonists, such as naltrexone, most consistently implicated ventral striatal cue-reactivity, whereas findings for other agents (e.g. acamprosate, baclofen, topiramate) were more heterogeneous and implicated prefrontal and cingulate regions, suggesting that medications targeting craving/reward processes may preferentially engage striatal predictors, while others may rely more on regulatory or stress-related circuitry (Mann et al. 2014, Schacht et al. 2017, Bach et al. 2021).

Importantly, analytic approaches varied across studies, with some employing region-of-interest (ROI) analyses targeting a priori reward and control regions, while others used whole-brain exploratory approaches. This variation likely contributes to differences in the spatial extent and consistency of reported findings. For example, studies using ROI-based approaches frequently focused on ventral striatal and prefrontal regions and tended to report more consistent associations with treatment outcomes. In contrast, whole-brain analyses often identified more distributed activation patterns across cortical and subcortical regions, though these findings were less consistently replicated across studies. In the resting state domain, although only two studies were included, they differed in analytic strategy, including seed-based whole-brain and ROI-restricted connectivity approaches, further highlighting the influence of methodological variability on observed findings.

The literature was heavily lopsided toward cue-reactivity tasks, but additional studies have examined reward anticipation, inhibitory control, and stress provocation tasks and revealed the importance of prefrontal and cingulate regions in predicting treatment response. For instance, stronger prefrontal engagement during inhibitory control and regulation tasks predicted lower relapse rates and greater behavioural improvement following cognitive or pharmacological interventions. Notably, relatively few studies directly probed negative affect, stress reactivity, or executive function despite their central role in addiction models, representing an important gap. Frameworks such as the Addictions Neuroclinical Assessment (ANA) emphasize domains of incentive salience, negative emotionality, and executive function, and future fMRI work should more systematically sample across these domains to improve mechanistic specificity and clinical translation (Kwako et al. 2016, 2017, Gunawan et al. 2023, Elsayed et al. 2026, Garber et al. 2026).

Resting-state studies further extended these findings by showing that functional connectivity within and between networks implicated in cognitive control, salience detection, and reward processing (such as the mPFC, dorsal ACC, and caudate) changed in ways that tracked with treatment-related improvements. Importantly, these findings also highlight the value of network-level metrics (e.g. default mode, salience, and frontoparietal control networks, or ‘triple network’ models) over isolated regional activation, particularly given that addiction is increasingly conceptualized as a disorder of large-scale network dysfunction (Elsayed et al. 2024, 2025). Collectively, these findings suggest that both task-evoked activation and intrinsic network organization may serve as promising neural markers of treatment prognosis in AUD.

The overlapping evidence across different paradigms supports a model in which the balance between reward-driven and control-related neural systems plays a central role in treatment outcomes for AUD. Elevated ventral striatal cue reactivity has repeatedly been observed in individuals with high craving or relapse risk, consistent with longstanding theories that exaggerated reward anticipation and sensitization contribute to the ongoing substance use, without remission (Koob and Volkow 2009). Additionally, increased prefrontal and ACC engagement may reflect enhanced regulatory capacity or cognitive control, which could facilitate abstinence and treatment adherence (D’souza 2019). The studies reviewed here also indicated that the direction and meaning of neural predictors may depend on treatment type. For example, individuals with higher ventral striatal reactivity at baseline tended to respond more favourably to naltrexone, suggesting a potential medication-by-brain interaction consistent with naltrexone’s modulation of mesolimbic dopamine signalling (Reinhard et al. 2015, Schacht et al. 2017, Bach et al. 2021). In contrast, for non-opioidergic pharmacotherapies and behavioural treatments, predictive signals appeared more distributed and may reflect broader regulation of stress, salience, or executive control systems, underscoring the importance of considering pharmacological mechanisms when interpreting neural predictors (Logge et al. 2021, Wetherill et al. 2021, Naqvi et al. 2024).

Resting-state studies identified that decreased connectivity within hyperactive reward circuits, or between stress and control networks following interventions such as motivational interviewing or CBT, was associated with improved outcomes, which shows the potential of this approach to reveal both mechanistic and prognostic neural markers of treatment response in AUD. At this stage, however, with only two studies to date, the literature is highly preliminary and firm conclusions cannot be drawn.

Despite encouraging convergence, the current body of research is limited by several methodological and conceptual challenges. Considerable heterogeneity existed in sample characteristics, imaging protocols, analytic pipelines, and outcome measures. Sample sizes were often small to moderate (N = 18–146), and many studies disproportionately included male participants or inpatient populations who were abstinent before imaging (Regner et al. 2015, Cornish and Prasad 2021). Such restrictions limit the generalizability of findings to more diverse, outpatient, or actively drinking populations and may also impact the reproducibility (Marek et al. 2022). The time interval between detoxification and baseline scanning also varied widely, which can influence neural responses to cues or stress due to transient withdrawal or craving states. Additionally, discrepancies across studies likely reflect differences in fMRI task design (e.g. block vs. event-related paradigms), stimulus selection, and acquisition parameters such as temporal resolution (TR), spatial resolution (voxel size), and preprocessing choices (e.g. smoothing), all of which can substantially affect sensitivity to subcortical versus cortical signals (Amaro and Barker 2006, Carp 2012, Turner et al. 2018). The diversity of tasks, analytic choices, and treatment heterogeneity poses another layer of complexity that prevents direct comparisons and meta-analytic synthesis. Furthermore, despite the strong biological rationale, integration of multimodal biomarkers, such as genetic, neurochemical, or physiological indices, with fMRI data remains rare, as does replication across independent samples. Future work would benefit from integrating complementary modalities, including structural MRI (e.g. cortical thickness, diffusion imaging), cerebral blood flow measures, and non-MRI approaches such as PET (Positron Emission Tomography; neurotransmitter systems), EEG/fNIRS (electroencephalogram/Functional near-infrared spectroscopy; temporal dynamics), and neuromodulation tools (e.g. Transcranial Magnetic Stimulation [TMS]/Transcranial Direct Current Stimulation [tDCS]), to provide a more comprehensive and mechanistic account of treatment response (Volkow et al. 1999, Zilverstand et al. 2018).

There remain several critical gaps that limit progress toward clinical translation. Few studies explicitly test brain-by-treatment interactions beyond naltrexone, leaving unclear whether similar predictive relationships exist for other pharmacotherapies such as acamprosate, topiramate, baclofen or even behavioural and neuromodulation therapy. Network-level interactions between reward, stress, and control systems have not been thoroughly mapped within the same individuals, despite growing evidence that substance use disorders involve dynamic network dysfunction rather than isolated regional abnormalities. Future work should prioritize methodological harmonization and multi-site collaboration to achieve adequate statistical power and reproducibility. Specifically, harmonization should extend beyond scanners to include standardized task paradigms (ideally spanning incentive salience, negative affectivity, and executive function domains), shared stimulus sets, consistent quality control, and preprocessing pipelines, and reporting standards for acquisition parameters and analytic strategies (Poldrack et al. 2017, Marek et al. 2022). The field would benefit from standardized fMRI paradigms, shared cue sets, and preregistered analytic pipelines that include whole-brain correction for multiple comparisons alongside clearly justified ROI analyses (Poldrack et al. 2017, Marek et al. 2022). Importantly, subcortical regions such as the ventral striatum may require tailored preprocessing approaches (e.g. reduced smoothing), which should be explicitly reported and standardized where possible (Sacchet and Knutson 2013, Alakörkkö et al. 2017). Large-scale prospective studies should be designed to test predictive validity in independent samples, ideally comparing the incremental value of neural markers over conventional clinical and behavioural predictors.

A key methodological constraint of the present review is the absence of a quantitative meta-analysis. Although meta-analytic techniques can generate pooled effect size estimates and formally assess heterogeneity, the included studies varied substantially in task paradigms, imaging acquisition and preprocessing procedures, analytic strategies (e.g. ROI vs. whole-brain approaches), treatment modalities, outcome definitions, and timing of neuroimaging assessments. This degree of methodological and clinical heterogeneity, together with the relatively small number of eligible studies, limited the feasibility and interpretability of statistical aggregation. As a result, the current synthesis remains primarily qualitative, restricting the ability to quantify the strength and consistency of neural predictors, evaluate publication bias, or identify potential moderators of treatment response. As the literature expands and becomes more methodologically harmonized, future systematic reviews should incorporate meta-analytic approaches to provide more precise and generalizable estimates of the prognostic utility of fMRI markers in AUD treatment outcomes.

There is also a pressing need for head-to-head comparisons across treatments to establish whether specific neural signatures predict preferential response to certain pharmacologic, behavioural, or neuromodulation interventions. At the same time, it remains an open question whether a single set of common neuroimaging biomarkers is feasible given the heterogeneity of AUD and treatment mechanisms, or whether embracing this heterogeneity, by identifying treatment-specific or domain-specific biomarkers, may be a more realistic and clinically useful approach. Subgroup interactions must also be prioritized to ensure that predictive models generalize across sex, age, ethnicity, and comorbidity. Finally, studies employing neuromodulation techniques such as TMS should leverage fMRI to test causal mechanisms, determining whether induced network changes can prospectively guide individualized treatment planning.

Collectively, the accumulating evidence suggests that fMRI markers, particularly within fronto-striatal circuits, could enhance clinical prediction of relapse and treatment response in AUD. Baseline ventral striatal cue reactivity and its early attenuation during therapy appear to be promising indicators of treatment prognosis. Likewise, resting-state connectivity changes in prefrontal and salience networks may reveal dynamic markers of recovery. At present, fMRI is best conceptualized as a high-value research tool for mechanistic insight and clinical trial stratification rather than a standalone clinical predictor, with key considerations including cost, reproducibility, sensitivity to design choices, and the need for multimodal integration (Woo et al. 2017, Ekhtiari et al. 2024). While such biomarkers are not yet ready for routine clinical use, these findings offer proof of concept that fMRI indices may ultimately be integrated into a precision-medicine framework for AUD.

Supplementary Material

Supplementary_materials_agag035
supplementary_materials_agag035.docx (218KB, docx)

Acknowledgements

The authors wish to express their sincere gratitude to the researchers and participants of all cited studies, whose contributions form the foundation of this work. We also extend our appreciation to the journals’ editors and peer reviewers, whose dedication, rigour, and thoughtful evaluation are essential to the advancement of scientific discovery and the dissemination of knowledge.

Contributor Information

Mahmoud Elsayed, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Ryan Siroosi, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Medicine, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Peter Najdzionek, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Carly McIntyre-Wood, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Tegan Hargreaves, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Ashley Blakely, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Emily Mote, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

James MacKillop, Peter Boris Centre for Addictions Research, St. Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON L9C 1G, Canada; Department of Psychiatry and Behavioural Neurosciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4L8, Canada.

Author contributions

Mahmoud Elsayed (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing—original draft, Writing—review & editing [lead]), Ryan Siroosi (Data curation, Formal analysis, Validation, Writing—review & editing [equal]), Peter Najdzionek (Data curation, Formal analysis, Investigation, Writing—review & editing [equal]), Carly McIntyre-Wood (Data curation, Formal analysis, Validation, Writing—review & editing [equal]), Tegan Lee Hargreaves (Data curation, Investigation), Ashley Blakely (Formal analysis, Investigation [equal]), Emily Mote (Formal analysis, Investigation [equal]), and James MacKillop (Conceptualization, Funding acquisition, Methodology, Resources, Supervision, Writing—review & editing [lead], Formal analysis, Validation, Visualization [supporting])

Conflicts of interest

J.M. is a principal and senior scientist in Beam Diagnostics, Inc.; no other authors have financial disclosures.

Funding

The Peter Boris Chair in Addictions Research (J.M.), Canada Research Chair in Translational Addiction Research (CRC-2020-00170; J.M.), and the Juravinski Research Institute.

Data availability

Original bibliographic searches are available upon request.

References

  1. Alakörkkö  T, Saarimäki  H, Glerean  E. et al.  Effects of spatial smoothing on functional brain networks. Eur J Neurosci. 2017;46:2471–80. 10.1111/ejn.13717. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Amaro  E, Barker  GJ. Study design in fMRI: Basic principles. Brain Cogn. 2006;60:220–32. 10.1016/j.bandc.2005.11.009. [DOI] [PubMed] [Google Scholar]
  3. Bach  P, Weil  G, Pompili  E. et al.  Incubation of neural alcohol cue reactivity after withdrawal and its blockade by naltrexone. Addict Biol. 2020;25:e12717. 10.1111/adb.12717. [DOI] [PubMed] [Google Scholar]
  4. Bach  P, Weil  G, Pompili  E. et al.  FMRI-based prediction of naltrexone response in alcohol use disorder: A replication study. Eur Arch Psychiatry Clin Neurosci. 2021;271:915–27. 10.1007/s00406-021-01259-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bartholdy  S, O’Daly  OG, Campbell  IC. et al.  Neural correlates of failed inhibitory control as an early marker of disordered eating in adolescents. Biol Psychiatry. 2019;85:956–65. 10.1016/J.BIOPSYCH.2019.01.027. [DOI] [PubMed] [Google Scholar]
  6. Becker  A, Gerchen  MF, Kirsch  M. et al.  Striatal reward sensitivity predicts therapy-related neural changes in alcohol addiction. Eur Arch Psychiatry Clin Neurosci. 2018;268:231–42. 10.1007/s00406-017-0805-y. [DOI] [PubMed] [Google Scholar]
  7. Blaine  SK, Wemm  S, Fogelman  N. et al.  Association of Prefrontal-Striatal Functional Pathology with alcohol abstinence days at treatment initiation and heavy drinking after treatment initiation. Am J Psychiatry. 2020;177:1048–59. 10.1176/appi.ajp.2020.19070703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Boness  CL, Witkiewitz  K. Precision medicine in alcohol use disorder: Mapping etiologic and maintenance mechanisms to mechanisms of behavior change to improve patient outcomes. Exp Clin Psychopharmacol. 2022;31:769–79. 10.1037/PHA0000613. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Carp  J. The secret lives of experiments: Methods reporting in the fMRI literature. NeuroImage. 2012;63:289–300. 10.1016/j.neuroimage.2012.07.004. [DOI] [PubMed] [Google Scholar]
  10. Cornish  JL, Prasad  AA. Sex differences in substance use disorders: A neurobiological perspective. Front Glob Womens Health. 2021;2:778514. 10.3389/fgwh.2021.778514. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. D’souza  MS. Brain and cognition for addiction medicine: From prevention to recovery neural substrates for treatment of psychostimulant-induced cognitive deficits. Front Psych. 2019;10:509. 10.3389/FPSYT.2019.00509. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Durazzo  TC, Meyerhoff  DJ. Psychiatric, demographic, and brain morphological predictors of relapse after treatment for an alcohol use disorder. Alcohol Clin Exp Res. 2017;41:107–16. 10.1111/ACER.13267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Ekhtiari  H, Sangchooli  A, Carmichael  O. et al.  Neuroimaging biomarkers in addiction. Nat Ment Health. 2024;2:1498–1517. 10.1101/2024.09.02.24312084. [DOI] [Google Scholar]
  14. Elsayed  M, Marsden  E, Hargreaves  T. et al.  Triple network resting-state functional connectivity patterns of alcohol heavy drinking. Alcohol Alcohol. 2024;59:agae056. 10.1093/ALCALC/AGAE056. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Elsayed  M, Syan  SK, Belisario  KL. et al.  Resting-state functional connectivity and alcohol use disorder: A case-control study. J Neuropsychiatry Clin Neurosci. 2025;38:179–87. 10.1176/APPI.NEUROPSYCH.20240145. [DOI] [PubMed] [Google Scholar]
  16. Elsayed  M, Belisario  KL, Murphy  JG. et al.  Machine learning-based computational validation of the addictions Neuroclinical assessment framework in relation to hazardous drinking. J Psychiatry Neurosci. 2026;51:1–11. 10.1139/jpn-25-0116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fang  Y, Sun  Y, Liu  Y. et al.  Neurobiological mechanisms and related clinical treatment of addiction: A review. Psychoradiology. 2022;2:180–9. 10.1093/PSYRAD/KKAC021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Garber  ML, McIntyre-Wood  C, Elsayed  M. et al.  Diagnostic validity of addictions Neuroclinical assessment indices for alcohol use disorder: Findings from an observational case–control study. Addiction. 2026. 10.1111/add.70371. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Grieder  M, Soravia  LM, Tschuemperlin  RM. et al.  Right inferior frontal activation during alcohol-specific inhibition increases with craving and predicts drinking outcome in alcohol use disorder. Front Psych. 2022;13:1–13. 10.3389/fpsyt.2022.909992. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Gunawan  T, Kwako  LE, Diazgranados  N. et al.  Addictions neuroclinical assessment. Alcohol Alcohol-Relat Dis. 2023;215–34. 10.1007/978-3-031-32483-3_14. [DOI] [Google Scholar]
  21. Guyatt  GH, Oxman  AD, Schünemann  HJ. et al.  GRADE guidelines: A new series of articles in the journal of clinical epidemiology. J Clin Epidemiol. 2011;64:380–2. 10.1016/J.JCLINEPI.2010.09.011. [DOI] [PubMed] [Google Scholar]
  22. Heinz  A, Beck  A, Grüsser  SM. et al.  Identifying the neural circuitry of alcohol craving and relapse vulnerability. Addict Biol. 2008;14:108–18. 10.1111/J.1369-1600.2008.00136.X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Herremans  SC, De Raedt  R, Van Schuerbeek  P. et al.  Accelerated HF-rTMS protocol has a rate-dependent effect on dACC activation in alcohol-dependent patients: An open-label feasibility study. Alcohol Clin Exp Res. 2016;40:196–205. 10.1111/acer.12937. [DOI] [PubMed] [Google Scholar]
  24. Koob  GF, Volkow  ND. Neurocircuitry of addiction. Neuropsychopharmacology. 2009;35:217–38. 10.1038/npp.2009.110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kranzler  HR. Overview of alcohol use disorder. Am J Psychiatry. 2023;180:565–72. 10.1176/APPI.AJP.20230488. [DOI] [PubMed] [Google Scholar]
  26. Kwako  LE, Momenan  R, Litten  RZ. et al.  Addictions Neuroclinical Assessment: A neuroscience-based framework for addictive disorders. Biol Psychiatry. 2016;80:179–89. 10.1016/J.BIOPSYCH.2015.10.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kwako  LE, Momenan  R, Grodin  EN. et al.  Addictions Neuroclinical assessment: A reverse translational approach. Neuropharmacology. 2017;122:254–64. 10.1016/j.neuropharm.2017.03.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Logge  WB, Morris  RW, Baillie  AJ. et al.  Baclofen attenuates fMRI alcohol cue reactivity in treatment-seeking alcohol dependent individuals. Psychopharmacology. 2021;238:1291–302. 10.1007/s00213-019-05192-5. [DOI] [PubMed] [Google Scholar]
  29. MacKillop  J, Agabio  R, Feldstein Ewing  SW. et al.  Hazardous drinking and alcohol use disorders. Nat Rev Dis Primers. 2022;8:80. 10.1038/S41572-022-00406-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Magill  M, Kiluk  BD, Ray  LA. Efficacy of cognitive behavioral therapy for alcohol and other drug use disorders: Is a one-size-fits-all approach appropriate?. Subst Abus Rehabil. 2023;14:1–11. 10.2147/SAR.S362864. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Mann  K, Vollstädt-Klein  S, Reinhard  I. et al.  Predicting naltrexone response in alcohol-dependent patients: The contribution of functional magnetic resonance imaging. Alcohol Clin Exp Res. 2014;38:2754–62. 10.1111/acer.12546. [DOI] [PubMed] [Google Scholar]
  32. Marek  S, Tervo-Clemmens  B, Calabro  FJ. et al.  Reproducible brain-wide association studies require thousands of individuals. Nature. 2022;603:654–60. 10.1038/s41586-022-04492-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. McPheeters  M, O’Connor  EA, Riley  S. et al.  Pharmacotherapy for alcohol use disorder: A systematic review and meta-analysis. JAMA. 2023;330:1653–65. 10.1001/JAMA.2023.19761. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Mehta  DD, Praecht  A, Ward  HB. et al.  A systematic review and meta-analysis of neuromodulation therapies for substance use disorders. Neuropsychopharmacology. 2024;49:649–80. 10.1038/s41386-023-01776-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Naqvi  NH, Srivastava  AB, Sanchez-Peña  J. et al.  Neural correlates of drinking reduction during a clinical trial of cognitive behavioral therapy for alcohol use disorder. Alcohol Clin Exp Res. 2024;48:260–72. 10.1111/acer.15259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Nawawi  NAAM, Nasir  F, Abdullah  KA. et al.  Understanding drug craving: Evidence from fMRI studies. Neurosci Res Notes. 2024;7:294–1. 10.31117/NEUROSCIRN.V7I3.294. [DOI] [Google Scholar]
  37. Nguyen  LC, Durazzo  TC, Dwyer  CL. et al.  Predicting relapse after alcohol use disorder treatment in a high-risk cohort: The roles of anhedonia and smoking. J Psychiatr Res. 2020;126:1–7. 10.1016/J.JPSYCHIRES.2020.04.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Poldrack  RA, Baker  CI, Durnez  J. et al.  Scanning the horizon: Towards transparent and reproducible neuroimaging research. Nat Rev Neurosci. 2017;18:115–26. 10.1038/nrn.2016.167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Regner  MF, Dalwani  M, Yamamoto  D. et al.  Sex differences in gray matter changes and brain-behavior relationships in patients with stimulant dependence. Radiology. 2015;277:801–12. 10.1148/RADIOL.2015142541. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Reinhard  I, Leménager  T, Fauth-Bühler  M. et al.  A comparison of region-of-interest measures for extracting whole brain data using survival analysis in alcoholism as an example. J Neurosci Methods. 2015;242:58–64. 10.1016/j.jneumeth.2015.01.001. [DOI] [PubMed] [Google Scholar]
  41. Roberge  M, Boisvert  M, Potvin  S. A meta-analysis of task-based fMRI studies on alcohol use disorder. Brain Sci. 2025;15:665. 10.3390/BRAINSCI15070665. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Sacchet  MD, Knutson  B. Spatial smoothing systematically biases the localization of reward-related brain activity. NeuroImage. 2013;66:270–7. 10.1016/j.neuroimage.2012.10.056. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Schacht  JP, Anton  RF, Randall  PK. et al.  Effects of a GABA-ergic medication combination and initial alcohol withdrawal severity on cue-elicited brain activation among treatment-seeking alcoholics. Psychopharmacology. 2013;227:627–37. 10.1007/s00213-013-2996-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Schacht  JP, Randall  PK, Latham  PK. et al.  Predictors of naltrexone response in a randomized trial: Reward-related brain activation, OPRM1 genotype, and smoking status. Neuropsychopharmacology. 2017;42:2640–53. 10.1038/NPP.2017.74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Srivastava  AB, Sanchez-Peña  J, Levin  FR. et al.  Drinking reduction during cognitive behavioral therapy for alcohol use disorder is associated with a reduction in anterior insula-bed nucleus of the stria terminalis resting-state functional connectivity. Alcohol Clin Exp Res. 2021;45:1596–606. 10.1111/acer.14661. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Strosche  A, Zhang  X, Kirsch  M. et al.  Investigation of brain functional connectivity to assess cognitive control over cue-processing in alcohol use disorder. Addict Biol. 2021;26:e12863. 10.1111/ADB.12863. [DOI] [PubMed] [Google Scholar]
  47. Suckling  J, Nestor  LJ. The neurobiology of addiction: The perspective from magnetic resonance imaging present and future. Addiction. 2016;112:360–9. 10.1111/ADD.13474. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Syan  SK, McIntyre-Wood  C, Vandehei  E. et al.  Resting state functional connectivity as a predictor of brief intervention response in adults with alcohol use disorder: A preliminary study. Alcohol Clin Exp Res. 2023;47:1590–602. 10.1111/acer.15123. [DOI] [PubMed] [Google Scholar]
  49. Turner  BO, Paul  EJ, Miller  MB. et al.  Small sample sizes reduce the replicability of task-based fMRI studies. Commun Biol. 2018;1:62. 10.1038/s42003-018-0073-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Volkow  ND, Fowler  JS, Wang  GJ. Imaging studies on the role of dopamine in cocaine reinforcement and addiction in humans. J Psychopharmacol. 1999;13:337–45. 10.1177/026988119901300406. [DOI] [PubMed] [Google Scholar]
  51. Volkow  ND, Wang  GJ, Fowler  JS. et al.  Addiction circuitry in the human brain. Annu Rev Pharmacol Toxicol. 2012;52:321–36. 10.1146/annurev-pharmtox-010611-134625. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Weinstein  AM. Reward, motivation and brain imaging in human healthy participants – A narrative review. Front Behav Neurosci. 2023;17:1123733. 10.3389/fnbeh.2023.1123733. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Wetherill  RR, Spilka  N, Jagannathan  K. et al.  Effects of topiramate on neural responses to alcohol cues in treatment-seeking individuals with alcohol use disorder: Preliminary findings from a randomized, placebo-controlled trial. Neuropsychopharmacology. 2021;46:1414–20. 10.1038/s41386-021-00968-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Woo  CW, Chang  LJ, Lindquist  MA. et al.  Building better biomarkers: Brain models in translational neuroimaging. Nat Neurosci. 2017;20:365–77. 10.1038/nn.4478. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. World Health Organization . (2024). https://www.who.int/news-room/fact-sheets/detail/alcohol/
  56. Zheng  H, Zhai  T, Lin  X. et al.  The resting-state brain activity signatures for addictive disorders. Med. 2024;5:201–223.e6. 10.1016/J.MEDJ.2024.01.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Zilverstand  A, Huang  AS, Alia-Klein  N. et al.  Neuroimaging impaired response inhibition and salience attribution in human drug addiction. A systematic review. Neuron. 2018;98:886–903. 10.1016/j.neuron.2018.03.048. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary_materials_agag035
supplementary_materials_agag035.docx (218KB, docx)

Data Availability Statement

Original bibliographic searches are available upon request.

Articles from Alcohol and Alcoholism (Oxford, Oxfordshire) are provided here courtesy of Oxford University Press

RESOURCES