Efficacy of N,N-dimethyltryptamine (DMT) psychedelic therapy for substance misuse: A systematic review and meta-analysis

Psychedelic therapies

Since the 2000s, psychedelic-assisted therapy has reemerged in research as a promising intervention for treatment-resistant psychiatric conditions, including SUDs (Hall, 2022). Therapeutic mechanisms of psychedelics are likely at the neural, biochemical, and psychological level. Classic psychedelics act as serotonergic 5-hydroxytryptamine 2A (5-HT2A) receptor agonists, producing altered states and increased neuroplasticity that facilitate therapeutic change through enhanced meaning-making and psychological flexibility (Madsen et al., 2019; Preller et al., 2017; Vargas et al., 2023).

Under these conditions, therapeutic outcomes are influenced by the client’s mindset (“set”) and contextual elements (“setting”), with suggestibility contributing to cognitive restructuring through neuroplasticity-enhanced bottom-up information processing, which helps adjust maladaptive thought and behavioral patterns (Cavarra et al., 2022).

Psychedelic-assisted psychotherapy (PAP) represents a specialized form of medication-assisted psychotherapy that incorporates usually discrete psychedelic dosing sessions rather than sustained pharmaceutical administration. It incorporates preparatory sessions to establish set and setting, supported psychedelic experiences, and subsequent integration sessions (Greer and Tolbert, 1998; Pahnke et al., 1970). CBT with psychedelic therapy demonstrates potential for SUDs, though benefits correlated most strongly with mystical experience intensity rather than specific cognitive interventions (Bogenschutz et al., 2015; Dakwar et al., 2019; Garcia-Romeu et al., 2014; Rothberg et al., 2021). The relative contributions of pharmacological versus psychotherapeutic elements remain debated, with some researchers emphasizing contextual influence (Hartogsohn, 2017; Leary, 1961) and others highlighting enhanced emotional breakthroughs (Roseman et al., 2019) or deepened psychological insights over neurobiological mechanisms alone (Peill et al., 2022). No studies have isolated the effects of these components through controlled trials, nor established the method’s advantages over psychedelics alone.

N, N-dimethyltryptamine (DMT) and its structural analog 5-methoxy-N, N-dimethyltryptamine (5-MeO-DMT) are emerging psychedelic candidates for SUD treatment. Plant-derived DMT is the psychoactive component in traditional Amazonian ayahuasca preparations, while 5-MeO-DMT occurs naturally in Incilius alvarius toad secretions and certain plants (Sherwood et al., 2020). Despite long indigenous healing histories (Delgrasso, 2024; Miller et al., 2019), these tryptamine alkaloids show only preliminary evidence linking their induced mystical experiences to improved psychological health and reduced tobacco use (Ferreira and Reis-Pina, 2025; Timmermann et al., 2024). Even so, traditional ayahuasca ceremonies have gained popularity in contemporary addiction treatment, attracting international participants to specialized healing centers (Kavenská and Simonová, 2015; Politi et al., 2018).

Recent reviews indicate that while ketamine (a non-classic psychomimetic anesthetic), lysergic acid diethylamide (LSD), and psilocybin have accumulated substantial evidence for SUD treatment, with LSD demonstrating clinical efficacy, research on DMT/5-MeO-DMT remains limited and inconclusive (Richard and Garcia-Romeu, 2025; Piper et al., 2025). Sparse and inconclusive evidence consists primarily of qualitative reports, ethnographic studies, case reports, cross-sectional surveys, and observational studies, with no randomized controlled trials (RCTs) examining SUD outcomes (Apud et al., 2022; Jiménez-Garrido et al., 2020; Talin and Sanabria, 2017). Both DMT analog compounds also show potential beyond SUDs for treating eating disorders, mood disorders, anxiety, Post-Traumatic Stress Disorder (PTSD), Obsessive-Compulsive Disorder, personality disorders, and suicidality (Gómez-Sousa et al., 2021; Jiménez-Garrido et al., 2020; Shinozuka et al., 2024), with 5-MeO-DMT also studied for pain and postpartum depression (Kargbo, 2024). Initial findings suggest benefits for treating substance abuse, but methodological limitations preclude definitive conclusions regarding the magnitude of efficacy or its mechanisms. Evidence remains insufficient to support clinical implementation of DMT compounds for SUDs.

Current study aims

This systematic review with meta-analysis is the first to address critical research gaps regarding the therapeutic effects of DMT/5-MeO-DMT on drug and alcohol substance use severity.

By pooling available quantitative data, the statistical analysis refines previously calculated effect sizes while attempting to distinguish pharmacological from psychological modality contributions. Objectives include (1) determining efficacy in reducing substance misuse, (2) evaluating whether psychotherapy enhances outcomes beyond DMT’s pharmacological effects, (3) exploring how treatment design variations influence clinical outcomes, and (4) synthesizing evidence to develop a conceptual understanding of likely therapeutic mechanisms. By addressing existing evidential limitations, this study aims to advance understanding of the therapeutic potential of DMT/5-MeO-DMTand its assisted psychotherapy for substance abuse.

Study design

This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) standards (Page et al., 2021) and the Cochrane Handbook guidelines (Higgins and Thomas, 2019). The protocol was preregistered with the Open Science Framework (OSF.io/75KD9) prior to data collection. A meta-analytic approach was justified to increase statistical power, resolve study inconsistencies, and quantify more precise magnitude of effects for DMT/5-MeO-DMT interventions, given the relatively small samples in available studies. This systematic pooling of data allows for more precise effect estimates than individual studies can provide alone.

The research questions were formulated using the Population, Intervention, Comparison, Outcome, and Study design (PICOS) framework (Hosseini et al., 2024). Adults over 18 years old with SUDs or problematic substance use served as the population of interest, with DMT or 5-MeO-DMT administered as the investigated intervention, with or without adjunct psychotherapy. Placebo, active control, treatment as usual, or pre-post intervention observations served as acceptable comparators. Study designs considered included RCTs, quasi-experimental designs, and pre-post observational studies.

Substance abuse severity served as the primary outcome and dependent variable responding to DMT interventions (independent variable) and was selected due to its prevalence across studies using psychometric instruments that quantify problematic consumption patterns (Higgins et al., 2024a). Tools like the Addiction Severity Index (ASI), Alcohol Use Disorder Identification Test (AUDIT), and Alcohol, Smoking, and Substance Involvement Screening Test (ASSIST) show high but variable reliability and validity due to self-reporting and less relevant subscale criteria (Mäkelä, 2004; Reinert & Allen, 2002; Selin, 2003; WHO ASSIST Working Group, 2002). These limitations were mitigated by using only substance use pattern subscales and considering self-report reliability in interpretations.

Studies employing adjunctive psychotherapy were isolated as a subgroup to determine differential effect sizes. Additional subgroup analyses explored the potential influence of population type (community vs clinical samples), drug formulation, and intervention duration (single-session vs multi-session protocols). This design enables a comprehensive evaluation of both overall efficacy and specific moderating or confounding factors that influence treatment outcomes.

Procedures

Search, screening, and selection strategy

An extensive search in January 2025 across PubMed, PsycINFO, EBSCO, and Web of Science databases was supplemented by manual reference examination (Lefebvre et al., 2016; Page et al., 2021). Initial screening ended in January 2025, and newly published reviews were cross-referenced between February and March 2025 to ensure comprehensive study inclusion. Results were restricted to publications between January 1960 and December 2024, which was verified to span all DMT research. Additional sources were identified through forward and backward citation searching, expert consultation, and screening of relevant review articles. We also searched preprint servers (PsyArXiv, medRxiv) and clinical trial registries (ClinicalTrials.gov, EU Clinical Trials Register) to identify unpublished or ongoing studies. PICOS-directed search terms (Table 1) combined four conceptual blocks: (1) substance misuse (“addiction,” “SUD,” etc.), (2) DMT treatments (“dimethyltryptamine,” “ayahuasca,” etc.), (3) study designs (“clinical trial,” “observational study,” etc.), and (4) substance use outcomes (“addiction severity,” “abstinence,” etc.).

OPEN IN VIEWER

Table 1.

Search terms used in the systematic database queries.

Database	Search terms
PubMed	((addict*) OR (substance abuse) OR (substance use disorder) OR (substance misuse) OR (substance dependence) OR (dependence) OR (SUD) OR (alcohol*) OR (drug* AND abuse) OR (opioid*) OR (OUD) or (tobacco) OR (cocaine)) NOT ((MS) OR (multiple sclerosis) OR (stroke)) AND ((hallucinogen) OR (psychedelic*) OR (psychotomimetic agent*) OR (dimethyltryptamine) OR (DMT) OR (5-Meo-DMT) OR (5- methoxy- N,N-dimethyltryptamine) OR (O-methylbufotenin) OR (ayahuasca) OR (hoasca) OR (tryptamine*) OR (caapi) OR (banisteriopsis) OR (psychotria viridis)) AND ((random*) OR (placebo) OR (trial) OR (clinical study) OR (prospective) OR (cohort) OR (group*) OR (open-label) OR (open label) OR (pilot*) OR (observational)) NOT ((animal) OR (review) OR (vitro)) AND ((addiction severity) OR (dependence) OR (abstain) OR (abstinence) OR (crag*) OR (self-effic*) OR (withdraw*) OR (detox*))
PsycINFO	(((((“addict” or “substance abuse” or “substance use disorder” or “substance misuse” or “substance dependence” or “dependence” or “SUD” or “alcohol” or (“drug*” and “abuse”) or (“opioid*” or “OUD” or “tobacco” or “cocaine”)) not (“MS” or “multiple sclerosis” or “stroke”)) and (“hallucinogen” or “psychedelic*” or “psychotomimetic agent*” or “dimethyltryptamine” or “DMT” or “5-Meo-DMT” or “5-methoxy-N, N-dimethyltryptamine” or “O-methylbufotenin” or “ayahuasca” or “hoasca” or “tryptamine*” or “caapi” or “banisteriopsis” or “psychotria viridis”) and (“random*” or “placebo” or “trial” or “clinical study” or “prospective” or “cohort” or “group*” or “open-label” or “open label” or “pilot*” or “observational”)) not (“animal” or “review” or “vitro”)) and (“addiction severity” or “dependence” or “abstain” or “abstinence” or “crav*” or “self-effic*” or “withdraw*” or “detox*”)).af.
EBSCO	TX ((addict*) OR (substance abuse) OR (substance use disorder) OR (substance misuse) OR (substance dependence) OR (dependence) OR (SUD) OR (alcohol*) OR (drug* AND abuse) OR (opioid*) OR (OUD) or (tobacco) OR (cocaine)) NOT TX ((MS) OR (multiple sclerosis) OR (stroke)) AND TX ((hallucinogen) OR (psychedelic*) OR (psychotomimetic agent*) OR (dimethyltryptamine) OR (DMT) OR (5-Meo-DMT) OR (5-methoxy-N,N-dimethyltryptamine) OR (O-methylbufotenin) OR (ayahuasca) OR (hoasca) OR (tryptamine*) OR (caapi) OR (banisteriopsis) OR (psychotria viridis)) AND TX ( (random*) OR (placebo) OR (trial) OR (clinical study) OR (prospective) OR (cohort) OR (group*) OR (open-label) OR (open label) OR (pilot*) OR (observational)) NOT TX ((animal) OR (review) OR (vitro)) AND TX ((addiction severity) OR (dependence) OR (abstain) OR (abstinence) OR (crav*) OR (self-effic*) OR (withdraw*) OR (detox*))
Web of Science	((((((((ALL = ((addict*) OR (substance abuse) OR (substance use disorder) OR (substance misuse) OR (substance dependence) OR (dependence) OR (SUD) OR (alcohol*)))) OR ALL = ((drug*) AND (abuse))) OR ALL = ((opioid*) OR (OUD) or (tobacco) OR (cocaine))) NOT ALL = ((MS) OR (multiple sclerosis) OR (stroke))) AND ALL = ((hallucinogen) OR (psychedelic*) OR (psychotomimetic agent*) OR (dimethyltryptamine) OR (DMT) OR (5-Meo-DMT) OR (5-methoxy-N,N-dimethyltryptamine) OR (O-methylbufotenin) OR (ayahuasca) OR (hopsca) OR (tryptamine*) OR (carpi) OR (banistereopsis) OR (psychotria viridis))) AND ALL = ((random*) OR (placebo) OR (trial) OR (clinical study) OR (prospective) OR (cohort) OR (group*) OR (open-label) OR (open label) OR (pilot*) OR (observational))) NOT ALL = ((animal) OR (review) OR (vitro))) AND ALL = ((addiction severity) OR (dependence) OR (abstain) OR (abstinence) OR (crav*) OR (self-effic*) OR (withdraw*) OR (detox*))

Following result count verification, two investigators confirmed that the number of results and the first 20% of the publications matched to ensure replicability. Two people employed Rayyan software made by Rayyan Systems for manual deduplication and screening, first examining titles/abstracts before reviewing full texts of potential selections (Lefebvre et al., 2016). Inclusion and exclusion criteria were developed based on PICOS parameters underlying the research questions, informed also by previous reviews on psychedelic therapies and SUD treatments (McKenzie et al., 2024a). Selection criteria included: (1) adult participants with problematic substance use, (2) DMT or 5-MeO-DMT interventions, (3) substance use severity outcomes as continuous measures, and (4) clinical trials or prospective observational cohorts making within-group or between-group comparisons. Both monotherapy and the use of these psychedelic compounds as an add-on polytherapy treatment were considered. Due to blinding challenges with psychedelics, non-RCT studies were included. Exclusions included non-original research, retrospective studies, animal models, populations under the age of 18, cross-sectional surveys, case reports, and studies lacking baseline comparison measures or extractable data for calculating effect sizes. Finalist selections were subsequently evaluated by all investigators, with discrepancies resolved through consensus meetings.

The screening process was completed before the 2025 publication of reviews by Richard and Garcia-Romeu (2025). Later cross-referencing of these reviews identified 13 additional studies that were subsequently excluded due to methodological limitations. As illustrated in the PRISMA flow diagram (Figure 1), eight studies across seven reports met eligibility criteria for meta-analysis. These examined ayahuasca DMT or 5-MeO-DMT interventions on substance use outcomes through pre-post and longitudinal observational designs, with three including comparison groups. Settings varied from clinical programs to naturalistic ceremonial contexts.

OPEN IN VIEWER

Figure 1.

PRISMA 2020 flow diagram for systematic screening and review (adapted from Page et al., 2021). Licensed under CC BY 4.0.

Analysis approach

A meta-analysis of pooled data was conducted to determine, with higher statistical power, the true effect size of DMT/5-MeO-DMT in reducing substance misuse. Sensitivity analyses examined potential therapeutic differences with adjunctive psychotherapy. A critical review and quality assessment, accompanied by exploratory subgroup analyses, were conducted to evaluate whether methodological variations contributed to outcome differences. The subgroups included: (1) interventions involving PAP, (2) community and clinical populations, (3) long-term therapy with more than one treatment session, and (4) polytherapy with concomitant use of a second psychedelic agent.

Study selection

The electronic and hand searches of the literature yielded 1132 results, 116 of which were eliminated as duplicates. A total of 1009 reports were removed from screening because they failed to meet the required inclusion criteria. The PRISMA flow diagram in Figure 1 illustrates the study selection and screening outcomes, in which eight studies across seven reports were ultimately eligible for inclusion in the meta-analysis. These studies, published between 2010 and 2024, examined the effects of ayahuasca DMT or 5-MeO-DMT interventions on substance use outcomes using a mix of pre-post and longitudinal observational designs. Three studies included comparison or control groups, while the others employed single-group designs. One between-group comparison study employed a single-blinded design using a psychoactive placebo (Rodrigues et al., 2024), while two others in a single report longitudinally examined the effects of ritual use in urban and jungle communities, comparing outcomes with those of urban and rural general population samples, respectively (Fábregas et al., 2010). Intervention settings ranged from structured clinical programs to naturalistic or ceremonial community contexts.

After thorough screening, the more recently published reviews by Richard and Garcia-Romeu (2025) and Piper et al. (2025) were cross-referenced, wherein 13 additional DMT and 5-MeO-DMT studies were considered but ultimately excluded due to methodological limitations. Five studies (in four reports) originally selected initially for the current meta-analysis were also included in one or both of the 2025 reviews: Armstrong et al. (2023), Berlowitz et al. (2019), Fábregas et al. (2010), and Thomas et al. (2013). Three other studies included in our meta-analysis—Perkins et al. (2022), Rodrigues et al. (2024), and Rush et al. (2024)—were not captured in either of the other recent reviews.

Characteristics of the included studies

Participant characteristics

The pooled sample (Table 2) was comprised of 456 baseline participants (mean age 36.6 years, 79% male) with diagnosed or reported drug or alcohol misuse, with four of the eight studies being all-male. Substances of abuse included alcohol (exclusively examined in two studies) and various drugs across stimulant, sedative, cannabis, opioid, and psychotropic categories. Only one study by Armstrong et al. (2023) included individuals diagnosed with both PTSD and comorbid alcohol abuse, while other studies varied in their reporting of psychiatric comorbidities. Post-treatment assessments retained only 287 participants in total, representing a 37% attrition rate. While Hedge’s g effect size statistic was used here to compensate for small, changing samples, none of the studies employed intention-to-treat (ITT) methods to account for participant data losses in a conservative analysis that interprets dropouts as treatment failure (Gupta, 2011).

OPEN IN VIEWER

Table 2.

Participant characteristics for each included study.

Publication		Dependency status	Substance misuse		Demographics
First author, year	Location	Diagnoses	Substance category	Substances used	Population	Baseline sample	Final Sample	Mean age	Women (%)	Men (%)
Thomas et al. (2013)	Canada	Self-reported substance misuse	Drugs and alcohol	Alcohol, cannabis, cocaine, sedatives, tobacco, opioids	Clinical	18	12	Not reported	Not reported	Not reported
Berlowitz et al. (2019)	Peru	Substance dependence diagnosis (DSM IV-TR) assessed via MINI	Drugs and alcohol	Alcohol, amphetamines and other stimulants, cannabis, cocaine, opioids, tranquilizers	Clinical	53	36	30.86	0%	100%
Armstrong et al. (2023)	Mexico	86 diagnosed with PTSD; 45 of these met risky alcohol use criteria (AUDIT-C scores of ⩾4)	Alcohol	Alcohol	Clinical	99	45	44	0%	100%
Fábregas et al. (2010; Jungle Study 1)	Brazil	N/A—none required	Drugs and alcohol	Alcohol, cannabis and, 9 different psychotropic drugs, including ayahuasca	Community	56	39	36	51.79%	48.21%
Fábregas et al. (2010; Urban Study 2)	Brazil	N/A—none required	Drugs and alcohol	Alcohol, cannabis and, 9 different psychotropic drugs, including ayahuasca	Community	71	39	37.32	46.48%	53.52%
Perkins et al. (2022)	North America	N/A—none required	Drugs and alcohol	Alcohol and cannabis	Community	53	53	37	60.40%	39.60%
Rodrigues et al., 2024	Brazil	AUD (mild, moderate, full)	Alcohol	Alcohol	Clinical	11	11	25.9	0%	100%
Rush et al., 2024	Peru	Substance use and addiction related mental health challenges (depression, anxiety)	Drugs and alcohol	Alcohol, opioids, cocaine/crack, cannabis, “other”	Clinical	113	52	33.5	0%	100%

Study characteristics

The included studies likewise exhibited heterogeneity in substances addressed and intervention design (Table 3).

OPEN IN VIEWER

Table 3.

Study characteristics, quality index (QI), and risk of bias.

Authors	Place	Study design	Psycho-therapy	Formula	Doses	Dose freq.	Duration	Other substances	Outcome measure	Alcohol effect size (g)	Drug effect size (g)	Total effect size (g)	QI	Risk of bias
Thomas et al. (2013)	Canada	Pre/post pro-spective cohort	Yes	Ayahuasca beverage	2	Daily	4 days	None	4WSUS	0.35 (0.03–0.67), p < 0.001	1.46 (1.14–1.78), p < 0.001	0.90 (0.58–1.22), p < 0.001	11/27	High
Berlowitz et al. (2019)	Peru	Pre/post pro-spective cohort	Yes	Ayahuasca beverage	Not reported	2×/month	3–12 months (mean 217.2 days; SD: 85.0 days)	Medicinal herbs	ASI	1.42, (1.24–1.60), p < 0.001	1.99, (1.81–2.18), p < 0.001	1.71, (1.52–1.89), p < 0.001	14/27	High
Armstrong et al. (2023)	Mexico	Pre/post pro-spective cohort	Yes	Inhaled vaporous 5-MeO-DMT	3+	3×/1 day	3 days	Oral ibogaine hydro-chloride 10 mg/kg	AUDIT-C	1.32, (1.18–1.46), p < 0.001	N/A	1.32, (1.18–1.46), p < 0.001	15/27	High
Fábregas et al. (2010; Jungle study)	Brazil	Longitudinal survey	No	Ayahuasca beverage	8–12	6×/month	8–12 months	None	ASI	0.33, (0.15–0.51), p < 0.001	0.17, (−0.01, 0.34), p < 0.001	0.25, (0.07–0.43), p < 0.001	10/27	High
Fábregas et al. (2010; urban study )	Brazil	Longitudinal survey	No	Ayahuasca beverage	8–12	8×/month	8–12 months	None	ASI	0.18, (0.02–0.35), p < 0.001	2.54, (2.37–2.70), p < 0.001	1.36, (1.20–1.52), p < 0.001	10/27	High
Perkins et al. (2022)	North America	Pre/post pro-spective cohort	No	Ayahuasca beverage	1–3	Daily	1–3 days	None	AUDIT and ASSIST	0.36, (0.20–0.53), p < 0.001	0.49, (0.32–0.65), p < 0.001	0.42, (0.26–0.59), p < 0.001	12/27	High
Rodrigues et al. (2024)	Brazil	Single-blind pre/post pro-spective cohort	No	Ayahuasca beverage	1	1–3×/1 day	1 day	None	Alcohol intake and urge questionnaire	0.35, (−0.03, 0.73), p < 0.001	N/A	0.35, (−0.03, 0.73), p < 0.001	15/27	High
Rush et al. (2024)	Peru	Pre/post pro-spective cohort	Yes	Ayahuasca beverage	13.9 avg	2×/month	Mean 27.9 weeks (SD: 12.9 weeks), up to 9 months	Medicinal herbs	ASI-5	0.82, (0.68–0.95), p < 0.001	1.46, (1.32–1.59), p < 0.001	1.14, (1.00–1.27), p < 0.001	14/27	High

Alcohol and drug abuse treatment

Six studies examined changes in both alcohol and drug misuse outcomes (Berlowitz et al., 2019; Fábregas et al., 2010; Perkins et al., 2022; Rush et al., 2024; Thomas et al., 2013), while two studies focused exclusively on alcohol-related measures (Armstrong et al., 2023; Rodrigues et al., 2024). Because psychedelic treatment may interact differently with substance-specific psychological patterns and addiction narratives, these differences were accounted for by calculating separate pooled effect sizes for abuse of drugs and alcohol.

Intervention design

Psychedelic formulations varied across studies, with six studies employing DMT administered orally as ayahuasca and one using inhaled 5-MeO-DMT 2 days after oral ibogaine. While two ayahuasca studies from Peru’s Takiwasi Center employed identical protocols that integrated traditional Amazonian plant medicine and Western addiction rehabilitation approaches (Berlowitz et al., 2019; Mabit et al., 1996; Rush et al., 2021, 2024), the remaining studies applied diverse methodologies and therapeutic frameworks. Four studies provided post-session psychotherapy in individual and group settings (Armstrong et al., 2023; Berlowitz et al., 2019; Rush et al., 2024; Thomas et al., 2013), while four provided no psychotherapeutic support (Fábregas et al., 2010; Perkins et al., 2022; Rodrigues et al., 2024). No details were provided about the specific therapeutic frameworks used.

Outcome measures

Substance use severity was assessed with several validated instruments, including ASI, AUDIT-C, ASSIST, and alcohol intake questionnaires such as the 4-week substance use scale (4WSUS). Overall, these scales have been ruled adequately valid and reliable for clinical and research use, contributing to study homogeneity (Mäkelä, 2004; Reinert and Allen, 2002; Selin, 2003; WHO ASSIST Working Group, 2002).

Treatment length and follow-up

All eight studies included in this study indicated at least one follow-up assessment after the intervention. For the meta-analysis, only the first post-intervention outcome was used to guarantee consistency and avoid confusing effects from long-term or repeated measures. Treatment duration varied substantially (1 day to 12 months), reflecting diverse dosing schedules ranging from single psychedelic administration sessions (k = 3) versus multiple sessions over extended weeks or months (k = 4). Changing the intensity and duration of treatment is an indication of the variance of therapeutic philosophies and resource availability.

Risk of bias

Cochrane’s systematic assessment of bias revealed that all studies demonstrated high overall risk, wherein as little as one high-risk domain automatically results in a high overall rating (Cochrane Methods, 2017). Studies uniformly demonstrated high risk in domains one (randomization), two (intended interventions), and four (outcome measurement) due to non-randomized designs and unblinding (Figure 2). These pervasive limitations indicate consequential risk of selection, performance, and attrition bias across all included studies. Detection bias (domain three) varied based on the randomness of missing data, with high-risk designation for unaddressed data absences. Within-study reporting bias (domain five) was frequently problematic due to inadequate reported protocol adherence. These deficiencies warrant a conservative interpretation, as directional bias, unblinded expectancy effects, and selective reporting may inflate effect sizes and compromise confidence intervals—considerations that are integrated into our results interpretation.

OPEN IN VIEWER

Figure 2.

Risk of bias for each of the included studies (McGuinness and Higgins, 2020). Licensed under CC-BY-NC-ND 4.0.

Quality index assessment

The Downs and Black Checklist (Supplemental Appendix A) revealed variable methodological quality across studies, as indicated by quality index (QI) scores in Table 3. Reporting clarity and design descriptions were generally adequate. However, studies predominantly used uncontrolled, single-arm, unblinded designs and neglected ITT analytic methods that address participant attrition (Gupta, 2011). External validity was limited but similar across studies, which did not report on the representativeness of the sample compared to the populations from which they recruited. Internal validity weaknesses included predominantly unblinded designs (except Rodrigues et al., 2024) and single-arm interventions. Internal validity was similarly limited across studies due to the non-randomized design and the inclusion of only one intervention group without controls. Participant loss to follow-up was generally accounted for across studies unless unspecified or undeterminable (Fábregas et al., 2010; Perkins et al. 2022). All studies except Rodrigues et al. (2024) achieved >80% statistical power with an adequate sample size. The single-study deficit and generally small studies are compensated for by combining study data and using Hedge’s g effect size equation for small samples.

Meta-analysis

Treatment effects

The eight selected studies measured effect sizes by comparing post-treatment substance use severity to pre-treatment baseline measures (Supplemental Appendix B). Figure 3 presents forest plots of individual studies and pooled effect sizes with 95% confidence intervals for combined substance use, drug, and alcohol use. As seen in Figure 3(a), the total overall pre- to post-treatment effect size for reducing substance use severity was large at g = 0.94 (95% CI: 0.56–1.31, p < 0.0001; Durlak, 2009). When isolated to drug use alone, as seen in Figure 3(b), the effect size was even greater at g = 1.35 (95% CI: 0.63–2.07, p < 0.0001), compared to a moderately high alcohol use effect size of g = 0.65, as seen in Figure 3(c) (9%% CI: 0.31–0.99, p < 0.0001). Synthesized results were verified through a sensitivity analysis, systematically removing each study one at a time from calculations to identify changes in overall effect sizes. This revealed no significant alterations to the overall effects, confirming the accuracy of the findings. Egger’s test of the intercept revealed no significant publication bias (β₀, p = 0.51), indicating a symmetrical distribution of results, which was confirmed by inspection of the funnel plot (Figure 4).

OPEN IN VIEWER

Figure 3.

Forest plots of effect sizes. (a) Forest plot of total combined substance use effect sizes (r = 0.5), (b) Forest plot of drug use effect sizes (r = 0.5), and (c) Forest plot of alcohol use effect sizes (r = 0.5).

OPEN IN VIEWER

Figure 4.

Funnel plot of all study results illustrating the symmetrical distribution of effect sizes (in Hedge’s g at r = 0.5).

Overall heterogeneity before analysis by subsets indicated moderate between-study variance of τ² = 0.28 (95% CI: 0.12–1.19, p < 0.0001), and a high proportion of variance due to heterogeneity at I² = 96.9% (95% CI: 95.5%–97.9%, p < 0.0001. The use of psychotherapy as an adjunct treatment is a primary source of heterogeneity, as are other aspects of intervention and study design.

Psychotherapy’s moderating effects

When stratified by the presence of psychotherapy as seen in Figure 5(a), studies that incorporated adjunct psychotherapy (k = 4) demonstrated a very large and statistically significant pooled effect size (g = 1.28, 95% CI: 0.96–1.60, p < 0.0001) and substantial heterogeneity (I² = 90.3%, τ² = 0.096, p < 0.0001). In contrast, studies without psychotherapy (k = 3) yielded a medium but still statistically significant effect size (g = 0.60, 95% CI: 0.09–1.12, p < 0.0001) with substantial heterogeneity (I² = 97%, τ² = 0.26, p < 0.0001). The test for subgroup differences was statistically significant (p = 0.0287), indicating that psychotherapy may have significantly moderated the effect of DMT treatments on substance use severity.

OPEN IN VIEWER

Figure 5.

Forest plots of the effect sizes for substance misuse severity reduction in groups of studies with and without adjunct psychotherapy use. (a) Forest plot of total substance use (alcohol and drug) effect sizes (r = 0.5), (b) Forest plot of the drug misuse reduction effect sizes in groups of studies with and without adjunct psychotherapy use, and (c) Forest plot of the alcohol misuse reduction effect sizes in groups of studies with and without adjunct psychotherapy use.

As seen in Figure 5(b), drug use severity was reduced with the largest substance-specific effect size when psychotherapy was employed (k = 4, g = 1.64, 95% CI: 1.28–2.00, p < 0.0001), demonstrating substantial heterogeneity as usual (I² = 91.3%, τ² = 0.088, p < 0.0001). In contrast, studies without psychotherapy (k = 3) yielded a smaller but still large effect size (g = 1.06, 95% CI: −0.39, 2.52, p < 0.0001), also with substantial heterogeneity (I² = 99.6%, τ² = 1.65, p < 0.0001). However, the chi-squared (χ²) test for subgroup differences was not statistically significant (p = 0.45), implying that substance-specific inferences cannot be made about psychotherapy’s mediating role in addressing drug misuse with DMT specifically.

As seen in Figure 5(c) for alcohol use, studies that incorporated adjunct psychotherapy (k = 4) demonstrated a large and statistically significant pooled effect size (g = 0.99, 95% CI: 0.52–1.46, p < 0.0001) with substantial heterogeneity (I² = 94.9%, τ² = 0.22, p < 0.0001). Studies without psychotherapy (k = 3) yielded a far smaller effect size (g = 0.29, 95% CI: 0.19–0.40, p < 0.0001) with no statistically significant determinable heterogeneity (I² = 0%, τ² = 0.0013, p = 0.45). The χ² test for subgroup differences, however, was statistically significant (p = 0.0047), indicating the significance of psychotherapy’s role in moderating the effects of DMT-based therapy for reducing alcohol use severity.

Clinical and community populations

A subgroup analysis examining whether having a community (k = 3) or clinical (k = 5) sample may have influenced substance use outcomes found that the total effect size for clinical samples was very large (g = 1.10, 95% CI: 0.68–1.53, p < 0.0001) and moderately high for community samples (g = 0.68, 95% CI: 0.00–1.35, p < 0.0001), as seen in Figure 6(a). The analysis revealed substantial heterogeneity of results across both community (I² = 92.6%, τ² = 0.22, p < 0.0001) and clinical groups (I² = 98%, τ² = 0.35, p < 0.0001), while the test for subgroup differences was not significant (p = 0.30).

OPEN IN VIEWER

Figure 6.

Forest plots of effect sizes segmented by groups of studies with community and clinical populations. (a) Forest plot of total substance misuse severity reduction effect sizes by population type, (b) Forest plot of drug use reduction effect sizes by population type, and (c) Forest plot of alcohol use reduction effect sizes by population type.

As seen in Figure 6(b), when isolated to drug use, the effect size was greater for both clinical samples (g = 1.64, 95% CI: 1.28–2.00, p < 0.0001) and community samples (g = 1.06, 95% CI: −0.39, 2.52, p < 0.0001), which reflects overall substance-specific effect size patterns. Substantial heterogeneity persisted across studies in both clinical (I² = 91.3%, τ² = 0.088, p < 0.0001) and community groups (I² = 99.6%, τ² = 1.65, p < 0.0001), and the test for subgroup differences was again not significant (p = 0.45).

The only statistically significant differences resulting from population type were for alcohol use reduction effect sizes (p = 0.013 for the Chi-squared test), where the clinical group had a moderately large effect size of g = 0.87 (95% CI: 0.43–1.31, p < 0.0001) and substantial heterogeneity across studies (I² = 94.5%, τ² = 0.24, p < 0.0001), as seen in Figure 6(c). Community samples had a medium-low effect size of g = 0.29 (95% CI: 0.18–0.40, p = 0.28) and far lower outcome heterogeneity (I² = 22%, τ² = 0.0023, p = 0.28).

Treatment length

As seen in Figure 7(a), the large total effect size for reducing all substance use when receiving short-term DMT treatment with up to three doses in one single session was g = 0.71 (95% CI: 0.09–1.33, p < 0.0001) with sustained high heterogeneity across these studies (I² = 97.2%, τ² = 0.28, p < 0.0001). The even larger total effect size for long-term multi-session therapy was g = 1.07 (95% CI: 0.59–1.56, p < 0.0001), also with substantial heterogeneity across the group’s studies (I² = 97.2%, τ² = 0.29, p < 0.0001). Despite the gap between effect sizes, the test for subgroup differences was not significant (p = 0.36), as seen in Figure 7(a).

OPEN IN VIEWER

Figure 7.

Forest plots of effect sizes segmented by groups of studies with short- and long-term treatment duration. (a) Forest plot of total substance misuse severity reduction effect sizes by treatment duration, (b) Forest plot of alcohol use reduction effect sizes by treatment duration.

As seen in Figure 7(b) for alcohol use, the effect size was larger for single-session therapy with an effect size of g = 0.69 (95% CI: 0.05–1.33, p < 0.0001), while that of multisession therapy was g = 0.62 (95% CI: 0.18–1.07, p < 0.0001). The analysis continued to reveal substantial heterogeneity across studies in both groups: short-term therapy (I² = 97.5%, τ² = 0.30, p < 0.0001) and long-term therapy (I² = 96.7%, τ² = 0.25, p < 0.0001). Again, the test for subgroup differences here showed that the difference in alcohol use outcomes due to treatment duration was not significant (p = 0.87).

Subgroup analysis for treatment length could not be conducted for drug use reduction, as there was only one single-session study addressing drug use (Perkins et al., 2022). However, a leave-one-out analysis omitting only Perkins et al. (2022) suggests that the study does slightly reduce the overall drug use effect size (Δg = −0.17, Δp = −0.0001), as represented by line 6 of Figure 8.

OPEN IN VIEWER

Figure 8.

Forest plot of drug use reduction pooled effect sizes resulting from the omission of each study. Omitting Perkins et al. (2022) in line 6 represents changes in the pooled drug use reduction effect size due to the removal of a short-term single-session treatment study.

Drug formulation

An exploratory leave-one-out sensitivity analysis omitted the single inhalant 5-MeO-DMT polydrug therapy study (Armstrong et al., 2023) to assess its impact on the pooled effect sizes for alcohol use outcomes. The Armstrong et al. (2023) study only assessed differences in alcohol abuse patterns and did not address drug use, so substance-agnostic generalizations cannot result from this approach. Removing the Armstrong et al. (2023) study led to a modest shift in both SMD (Δg = −0.10) and p-value (Δp = +0.0008). Overall, the pooled alcohol use effect size remained largely consistent after removal of this study with an alternative drug formulation, as charted by the forest plot in line 3 of Figure 9.

OPEN IN VIEWER

Figure 9.

Forest plot of alcohol use reduction pooled effect sizes resulting from the omission of each study. Omitting Armstrong et al. (2023) in line 3 represents changes in the pooled alcohol use reduction effect size due to the removal of a study treating with inhalant 5-MeO-DMT after oral ibogaine.

Proposed mechanisms

The magnitude of observed effects may stem from DMT’s neurobiological mechanisms, therapeutic alliance, contextual factors, or a combination of these factors. Like LSD, DMT analogs might outperform conventional medications through 5-HT2A receptor agonism, enhancing neuroplasticity, and creating opportunities for psychological change (Cavarra et al., 2022; Vargas et al., 2023). Across modalities, trusting therapeutic relationships predict outcomes (Del Re et al., 2021), while set and setting generate non-pharmacological effects that correlate with improved well-being (Borkel et al., 2023; Hartogsohn, 2016). Psychedelics’ enhanced environmental suggestibility potentially amplifies therapeutic effects within ceremonial contexts, explaining why traditional Amazonian healing centers demonstrated particularly robust outcomes (k = 2, g = 1.99, 1.46), with cultural and ritual elements perhaps optimizing responses (Berlowitz et al., 2019; Rush et al., 2024). These contextual factors, while therapeutic, contributed to heterogeneity, consistent with the influence of set and setting on psychedelic effects (Hartogsohn, 2016). Existing evidence cannot differentiate pharmacological effects from contextual influences or distinguish specific from non-specific mechanisms. As such, interpretations remain preliminary.

Substance-specific effects

The larger effect sizes observed for drug use compared to alcohol use may reflect differences in underlying mechanisms, different measuring instrument sensitivities, and drug diversity among stimulant, sedative, opioid, and psychotropic substances. Notably, two studies reported minimal or non-significant reductions in cannabis use following ayahuasca exposure (Fábregas et al., 2010; Thomas et al., 2013), while Perkins et al. (2022), which assessed only alcohol and cannabis outcomes, reported smaller drug-related effect sizes than most other studies. This pattern suggests that cannabinoids may possess distinct properties affecting their responsiveness to DMT intervention. Still, definitive conclusions cannot be drawn due to potential confounding from persisting cultural, religious, or medicinal cannabis practices—especially in Indigenous settings (Balant et al., 2021). DMT’s effect size for drug use reduction is likely understated when including this cannabis data. Overall, differences between drug and alcohol outcomes may reflect substance-specific neurobiological interactions (Johansson et al., 2006). These differences may result from varying neurotransmitter system involvement, as all abused substances affect dopaminergic reward systems to different degrees, while alcohol affects a wider range of neurotransmitters (Tomkins and Sellers, 2001).

Moderating and confounding factors

Study design implications

While gender-based differences in substance misuse prevalence (Substance Abuse and Mental Health Services Administration, 2018), instrument reliability (De Oliveira et al., 2014), and treatment response (Becker et al., 2016) may justify single-gender studies, the imbalanced demographic composition of the included studies and of the total pooled population limits generalizability. Primarily male participation is particularly problematic given that women are likely as vulnerable to misuse and may be more susceptible to the addiction cycle of abuse (Zakiniaeiz and Potenza, 2018). Inconsistency in including comorbid diagnoses introduces significant study heterogeneity, which complicates the interpretation of treatment effects and potentially obscures condition-specific treatment responses. A high participant attrition rate across all studies increases the risk of attrition bias and inflated treatment effect estimates.

Intervention design differences—particularly regarding psychedelic compounds, formulation, dosing, and psychological support—limit this review’s capacity to isolate specific therapeutic mechanisms and best practices. For example, the pharmacological heterogeneity presented using both DMT and 5-MeO-DMT analog variants, in both oral and inhaled formulations, as part of a monotherapy or poly-drug therapy regimen, represents several confounding variables. Although both substances are tryptamine-class psychedelics, they exhibit notable differences in their pharmacology, subjective effects, routes of administration, and contextual practices, limiting the extent to which we can interpret the outcome of their combined effect (Reckweg et al., 2022). Notably, our only study involving 5-MeO-DMT (Armstrong et al., 2023) administered it in combination with ibogaine. As a result, the independent contribution of 5-MeO-DMT cannot be isolated, and no generalizable claims can be supported based on the current evidence. Our exploratory leave-one-out analysis by drug formulation rendered it impossible to disentangle formulation effects from co-administration effects. Thus, these findings should be interpreted with caution until standardized designs for both DMT and 5-MeO-DMT are conducted to establish clearer comparative effects.

In addition to interventional heterogeneity posed by the introduction of psychotherapy, non-standardized therapist training and often unclear supervision protocols introduce potential therapist effects that may similarly confound treatment outcomes synthesized in this study. The lack of standardization extends to varied settings, ranging from traditional shamanic ceremonies to clinical environments. These widely varied treatment paradigm elements likely influence outcomes and remain insufficiently controlled or measured, making it impossible to isolate therapeutic mechanisms.

Treatment duration and dosing schedule variations likely capture different phases of therapeutic response, as psychedelic interventions typically exhibit both acute and sustained effect patterns (Nikolic et al., 2024; Preller et al., 2020). This temporal heterogeneity could theoretically illuminate dose–response relationships in well-designed trials; however, it instead introduces confounding variance within the meta-analytic framework.

While the substance use severity scales employed are valid and reliable overall, methodological limitations do exist (Reinert and Allen, 2002; Selin, 2003; WHO ASSIST Working Group, 2002). Variable gender-based reliability and ASI’s questionable capacity to differentiate between substance types (Humeniuk et al., 2008; Mäkelä, 2004; McLellan et al., 1985; Simon et al., 2023) potentially compromise measurement consistency across mixed-gender samples and substance categories in this review. Pride, shame, or other self-image concepts may introduce detection bias through self-reported measures. Temporal disparities between instruments (e.g., AUDIT’s 12-month timeframe versus more recent history surveys) may differentially affect the detection of treatment response, particularly for brief treatments. While timeframe-appropriate instrument selection mitigates this issue, Armstrong et al.’s (2023) application of AUDIT for a 3-day intervention represents a temporal mismatch, yielding underestimated treatment efficacy.

Study quality implications

Most studies’ lack of controlled conditions undermines causal inferences, and neglect of ITT analytic methods introduces systematic bias through the exclusion of dropped participants, which might represent treatment non-response. Inadequate sample representativeness limits external validity and likely reflects self-selection with positive treatment expectations rather than the general substance-using population. Consequently, effect estimates are most applicable to self-selected, treatment-seeking populations with positive expectations toward psychedelic interventions rather than the broader substance-using population. Observed changes lacking control conditions demonstrate low internal validity and are inadequate to support causal inferences attributed directly and entirely to interventions, leaving open questions of effects due to participant expectations, healing trends without treatment, or the influence of a supportive environment. Taken together, these limitations correspond to a low certainty of evidence as estimates are highly sensitive to bias, as informed by the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework (Goldkuhle et al., 2025). As such, the synthesized effects should be understood as preliminary indicators of possible rather than certain treatment efficacy.

It should be noted that the Rodrigues et al. (2024) study’s inability to detect subtle effects may have been due to its underpowered design. While this may skew the pooled results toward diminished outcomes, the sample size adequacy for 80% powered results or greater among all other studies largely overpowers this influence, increasing the credibility of meta-analysis findings.

Finally, we imputed missing data from Thomas et al. (2013) with standard deviation values from Rodrigues et al. (2024), acknowledging that the study was underpowered and only reported on alcohol outcomes. Although these substitutions followed established approaches for handling missing data, the method introduces assumptions that the variability in outcomes is comparable to what could have been observed in the Thomas et al. (2013) study and thus may bias the resulting effect estimates. Notably, exploratory subgroup analyses were similar when the imputed Thomas et al. (2013) data were excluded, indicating minimal impact of these imputation methods on overall results. However, these results involving imputed standard deviations should be interpreted with caution.

Study evaluation

This rigorous systematic synthesis of DMT intervention studies yielded reliable effect size estimates through high-powered data pooling. Predetermined screening criteria ensured objectivity, and a comprehensive search methodology promoted unbiased reporting and reproducible analysis. The number of studies was adequate for meta-analysis, but including more homogeneous RCTs would yield more reliable conclusions. All studies used similar, validated measures, contributing to homogeneity, but self-reports reduced measure reliability. Some study characteristics varied, limiting the validity and certainty of the findings. Although we critically evaluated their influence—for example, through our exploratory leave-one-out analysis of different drug formulations—it remained impossible to disentangle the effects of formulation from the co-administration of 5-MeO-DMT and DMT. Marked heterogeneity in DMT-assisted psychotherapy outcomes compromised the generalizability of psychotherapy’s associated findings, representing average rather than predictable treatment outcomes. Differences in setting (traditional shamanic vs clinical) also complicated interpretations of therapeutic mechanisms due to the environmental variables. This investigation’s limited scope excluded valuable subgroup analyses by formulation, dosing, treatment duration, and setting, which could evaluate each characteristic’s influence on outcomes. While effect sizes appeared substantial, the exclusion of control groups limited causal attribution, so natural recovery, regression to the mean, or placebo effects might also contribute to outcomes. Pervasive high risk of bias due to methods likewise undermines confidence in effect size estimates and psychotherapy’s moderating effects, necessitating caution in interpreting efficacy claims. The absence of extended follow-up data suggests that observed effects could represent transient responses and does not inform treatment durability. These early estimates should inform future hypotheses rather than confirm treatment efficacy and require controlled research validation rather than representing actionable clinical evidence.

Future research

Future research should address these limitations through RCT designs and follow-up meta-analyses, excluding studies with high risk of bias. Other meaningful outcomes, like relapse rates, should be measured while comparing treatment to active control groups and eliminating confounding factors. Comparing a manualized or semi-standardized psychotherapy approach will ideally isolate pharmacological from psychotherapeutic effects. Various therapeutic frameworks (CBT, DBT, ACT, psychodynamic) should also be evaluated with DMT to identify the most advantageous approach to reducing substance misuse, with findings evaluated across other psychedelic therapies. Mixed-methods studies can identify productive elements of psychedelic and complementary psychotherapy, and longitudinal studies should measure long-term outcomes. To differentiate pharmacological effects from contextual influences, studies should measure specific contributions of DMT formulations, dosing regimens, environments, and set/setting elements. Comparative analysis of traditional ceremonial versus clinical settings could determine how ritual elements influence outcomes, with beneficial components integrated into treatment paradigms. Differential outcomes between alcohol, drugs, and cannabis warrant substance-specific research into DMT interactions with neurobiological phenomena, exploring varied dosing regimens to optimize treatment by substance type. Neuroimaging studies may specifically illustrate neural correlates of SUD recovery mechanisms and inform new lines of inquiry. Finally, research might address common neuropsychiatric comorbidities with enhanced vigilance and rescue protocols for adverse events, particularly with psychosis risk. Overall, there is a substantial opportunity to progress our knowledge of how DMT-assisted therapy rehabilitates substance abuse.

Clinical implications

Identified ethical deficiencies—inadequate informed consent, insufficient safety monitoring, and unclear boundary protocols—rendered findings ethically problematic for clinical translation. Psychedelic-specific ethical standards must be refined to reduce human subject risk and increase the credibility of future studies for direct translation to safe clinical guidelines. Safety monitoring protocols, enhanced informed consent, and contingency plans would reduce risks while enabling research with more diverse clinical populations that include medical comorbidities. With such safety measures in place, extended studies can determine optimal dosing schedules for sustained benefits. Evidence-based best practices emerging from safe, ethical research should inform policy recommendations and therapist training requirements, providing a strong foundation for responsible clinical implementation.

Establishing long-term efficacy and safety profiles remains paramount for clinical adoption. Addiction treatment efficacy is defined by sustained abstinence over months to years, not acute improvements. The absence of extended follow-up data, therefore, undermines the current findings’ clinical relevance. Furthermore, the lack of long-term safety monitoring means delayed adverse effects remain unknown, creating ethical constraints on clinical use.

The observed efficacy gap between alcohol and drug use challenged DMT’s perceived status as a universal SUD intervention. Without mechanistic clarification, clinicians cannot reliably predict treatment responses for specific substances of abuse. This reduces DMT’s immediate clinical utility and would require that treatment decisions be based on incomplete understanding, risking wasteful rounds of ineffectual treatment.

Standardization is crucial for the translation of psychedelic research into pragmatic clinical practice. Heterogeneity in DMT-assisted psychotherapy means that clinicians attempting to replicate findings without standardized protocols faced some degree of outcome uncertainty. Guidelines and procedures should be substance-specific, considering alternatives for cannabis dependence, which demonstrated lesser DMT response. Given superior efficacy for alcohol and other abused drugs, standard DMT or 5-Meo-DMT protocols may someday complement treatments, especially for treatment-resistant populations. If durable effects are confirmed, optimized DMT formulations with brief, targeted integrative psychotherapy may prove more cost-effective than ongoing pharmacotherapy and long-term psychotherapy.