A Systematic Review of Hypnosis for Clinical Pain Relief.

This review was designed by T.H.Y. and S.W.G.D. The searches, assessments and date extraction were completed by T.H.Y. and S.W.G.D. The data were analysed by T.H.Y. and the results interpreted by both authors. S.W.G.D. had a primary role in preparing the manuscript, which was edited by T.H.Y. Both authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work.

The authors have nothing to report.

The review protocol was prospectively registered on PROSPERO (ID: CRD42024608722) and conducted in accordance with PRISMA guidelines (Moher et al.  2009 ) (see Data  S1 for the PRISMA checklist). Two consecutive searches (initial and updated) were conducted across PsycINFO, PubMed, Scopus, Web of Science and the Cochrane Library Register of Controlled Trials from database inception to 20 October 2025. Search terms combined variants of hypnosis and non‐pharmacological intervention with terms related to pain reduction (full search strategy in Data  S5 ). Searches were applied to titles, abstracts and keywords (Scopus: title only). The full inclusion and exclusion criteria are detailed in Table  1 . Inclusionary and exclusionary criteria. Studies directly assessing hypnosis to deliver pain relief with or without specific suggestions for pain relief and delivered in any format (e.g., in‐person or via a pre‐recorded audio file, individually or in a group setting, alone or as a supplement to other forms of therapy) The study will be considered a hypnosis study if the authors specifically refer to the procedure as involving hypnosis, or the patients specifically consent to being hypnotised, or the study clearly includes an induction procedure intended to induce a hypnotic ‘trance’ state. Outcome measures for the intervention are reported. Studies that utilise a Visual Analogue Scale (VAS) to measure pain. Studies can either present pre‐intervention‐post‐intervention measures or post‐intervention measures only Eligible studies included adults (≥ 18 years) experiencing acute or chronic clinical pain. Studies involving experimentally induced pain, participants under 18 years, or samples selected for high or low hypnotic responsiveness were excluded. We included peer‐reviewed randomised controlled trials (RCTs) comparing hypnosis with non‐active/minimal controls (e.g., waitlist, attention and treatment as usual) or active comparators. Non‐randomised studies, studies without control groups, qualitative or descriptive designs, single‐session laboratory studies, case reports and review articles were excluded. Studies in all languages were eligible and translated where required. Included studies directly evaluated hypnosis for reducing pain intensity, with or without specific analgesic suggestions, delivered in any format (e.g., therapist‐delivered, audio‐recorded, self‐hypnosis, individual or group). Studies in which hypnosis was combined with another intervention, combined therapies (e.g., hypnosis with CBT), were included only when an appropriate control was included (e.g., hypnosis alone or CBT alone) to allow for the additional contribution of hypnosis or the other intervention to be evaluated. Studies were required to report pain intensity using a Visual Analogue Scale (VAS) or Numerical Rating Scale (NRS). Multi‐item pain questionnaires not reporting pain intensity separately were excluded. Studies had to provide means and standard deviations for post‐intervention scores and/or pre–post change scores; studies lacking sufficient variance data were excluded. Screening was conducted by the first author (TY) and verified by the senior author (SD). Title/abstract and full‐text screening were performed independently by both authors, with disagreements resolved by consensus. Inter‐rater reliability was assessed using Cohen's kappa (McHugh  2012 ). Data were extracted using a standardised spreadsheet and verified by both authors. Extracted variables included study characteristics (author, year and country), sample size and demographics, pain type, intervention characteristics, comparator type, number of sessions, pain measures and pain intensity outcomes (means and standard deviations or effect sizes). Hypnosis modality was coded a priori (see Data  S5 ). Information required for risk of bias assessment was also extracted. Unreported data were requested from study authors. Risk of bias (RoB) was independently assessed by both authors, with discrepancies resolved by discussion. Inter‐rater agreement was evaluated using Cohen's kappa (McHugh  2012 ). A 21‐item RoB tool was applied, based on the Revised Cochrane RoB tool for randomised trials (Sterne et al.  2019 ) and the validity scale developed by Thompson et al. ( 2019 ). The tool assessed methodological rigour, selection processes and reporting bias, including factors particularly relevant to hypnosis research (e.g., hypnotisability measurement, medication use, treatment expectations and intervention fidelity). Items were rated as ‘yes’, ‘no’ or ‘uncertain/unknown’. Overall RoB was classified as low (≥ 15), moderate (11–14) or high (≤ 10). Full item ratings are provided in Data  S2 . Overall certainty of evidence was evaluated using the GRADE framework, which provides a structured approach for judging how much confidence should be placed in an estimated treatment effect. With randomised trials initially rated as high‐certainty evidence, this certainty may then be downgraded when there are important concerns about risk of bias, inconsistency, indirectness, imprecision and publication bias domains (Prasad  2024 ). Non‐standard VAS scales (e.g., 0–240; VanDyck et al.  1991 ) were rescaled to a 10‐point metric prior to analysis. No distinction was made between 10‐point and 11‐point scales. The zero in 11‐point scales indicates ‘no pain,’ which is rarely reported, and the one in both scales refers to a comparable minimal level of pain. Given anticipated heterogeneity, random‐effects meta‐analyses were conducted. Between‐study variance (τ 2 ) was estimated using restricted maximum likelihood (Viechtbauer  2005 ). Heterogeneity was assessed using Cochran's Q and I 2 statistics. Effect sizes were calculated as Hedges' g with 95% confidence intervals (Hedges and Olkin  1984 ), applying the Knapp–Hartung adjustment to account for small sample bias (Knapp and Hartung  2003 ; IntHout et al.  2014 ; Langan et al.  2019 ). Analyses were conducted in R using the metafor package (Viechtbauer  2010 ; R Core Team  2017 ). Separate meta‐analyses were conducted for studies reporting post‐intervention scores and those reporting pre–post change scores. Subgroup analyses examined risk of bias, pain type (acute vs. chronic) and hypnosis modality where sufficient studies were available. Studies not suitable for pooling were narratively synthesised (see Data  S5 ). Publication bias and small study effects were assessed using funnel plots for analyses including ≥ 10 studies (Sterne et al.  2011 ), and Egger's regression test for analyses including ≥ 30 studies (Lau et al.  2006 ), respectively. Leave‐one‐out analyses were conducted for studies with large or anomalous effect sizes and moderate‐to‐high RoB to assess robustness.

The initial database search (18 November 2024) identified 17,525 records, with 5790 duplicates removed. We excluded 141 non‐journal articles and 237 records unavailable for retrieval, leaving 11,357 records for title/abstract screening. After applying eligibility criteria, 10,940 records were excluded (Figure  1 ). Two additional studies identified from recently published reviews were added, yielding 419 reports for full‐text review. Inter‐rater agreement at full text was moderate (Cohen's kappa = 0.58; SE = 0.04; 95% CI 0.51–0.66). One exception was made: Ter Kuile et al. ( 1994 ) was included as a hypnotic intervention despite being labelled ‘cognitive self‐hypnosis’, because the protocol primarily consisted of hypnosis‐related induction and strategies (e.g., relaxation, imaginative inattention and pain displacement/transformation) with limited cognitive restructuring. After consensus, 100 studies met criteria from the initial search. PRISMA flowchart. The updated search (October 2025) identified 1046 additional records; 12 underwent full‐text screening and six were included. In total, 106 studies were included, from which 186 effect sizes were extracted. Overall, 104 effect sizes from 84 studies were pooled across meta‐analyses; the remaining 22 studies were primarily narratively synthesised (see Data  S5 ). Study characteristics and findings are summarised in Tables  2 and 3 . Characteristics of all 106 studies included in this review. Abbreviations: A, acute pain; ACU, acupuncture; ADD, addition of stated intervention on top of hypnosis; AT, autogenic training; BIO, bio‐feedback relaxation; C, chronic; CBT, cognitive‐behavioural therapy; COM, comparison between hypnosis and stated control intervention; DIS, distraction; EDU, pain education; F, female; H, hypnosis; IPT, in‐person taped/audio hypnosis; M, male; MF, mindfulness‐based intervention; MSG, massage; MUS, music; NB, non‐binary; PHY, physiotherapy; PL, placebo; RLX, relaxation; SELF, self‐hypnosis only; SRS, stress‐reducing strategies; TAU, treatment as usual; TENS, transcutaneous electrical nerve stimulation; TRP, in‐person therapist‐led hypnosis; TRP + SELF, in‐person therapist with regular self‐hypnosis; V, virtual‐reality hypnosis; WL, waiting list. Indicates studies that have been included in at least one meta‐analysis. Unless otherwise stated (i.e., when ‘ADD’ is stated), all controls did not concurrently introduce hypnotic suggestions and were compared directly to hypnosis. Due to experimental design considerations, Patterson et al. ( 2010 ) combined the distraction group with the treatment‐as‐usual group to form a single control group. Study findings of all included 106 studies in this review. COM/P (↓ HYP vs. EDU): 4.5 (2.72) vs. 6.6 (2.32); p  = 0.03 COM/P (↓ HYP vs. TAU): 4.5 (2.72) vs. 6.9 (2.56); p  = 0.01 COM/P (↓ HYP vs. PL): 3.06 (2.07) vs. 4.61 (2.48); p  = 0.031 COM/P (↓ HYP vs. CBT): 3.06 (2.07) vs. 4.92 (2.26); p  = 0.031 COM/P (HYP vs. RLX/AT ↓): 4.5 (2.96) vs. 3.24 (2.42) COM/P (↓ HYP vs. WL): 4.5 (2.96) vs. 5.08 (3.2) COM/P (↓ HYP vs. WL): 1.97 (1.31) vs. 5.05 (1.79) COM/P (↓ HYP vs. RLX): 1.97 (1.31) vs. 3.76 (1.77) COM/FUL3 (↓ HYP vs. WL): 1.66 (1.49) vs. 4.29 (1.31) COM/FUL3 (↓ HYP vs. RLX): 1.66 (1.49) vs. 2.75 (1.91) COM/FUL1W (HYP vs. TAU ↓): 3.17 (0.52) vs. 2.25 (2.63) COM/FUL3 (↓ HYP vs. TAU): 0.33 (0.82) vs. 1.83 (3.06) COM/FUL1W (↓ HYP vs. PL): 3.17 (0.52) vs. 3.83 (2.23) COM/FUL3 (↓ HYP vs. PL): 0.33 (0.82) vs. 1.67 (3.14) COM/P (↓ HYP vs. PL): 2.91 (2.53) vs. 4.73 (2.65); p  = 0.046 COM/P (↓ HYP vs. PL): −3.05 (2.42) vs. −1.33 (1.84); p  = 0.046 b COM/FUL3 (↓HYP vs. TAU): −2.12 (2.15) vs. −0.68 (2.56); p  = 0.02 b COM/FUL12 (↓ HYP vs. TAU): −1.63 (2.59) vs. −1.57 (3.2); p  = 0.76 (ns) b COM/P (HYP vs. BIO/RLX ↓): 5.09 (1.92) vs. 3.36 (1.28) COM/FUL3 (HYP vs. BIO/RLX ↓): 4.9 (2.13) vs. 3.13 (1.42) ADJ/P (↓ HYP vs. CBT): 5.3 (0.3) vs. 5.6 (0.3) ADJ/FUL3 (↓ HYP vs. CBT): 5.7 (0.4) vs. 5.9 (0.3) ADJ/FUL6 (↓ HYP vs. CBT): 5.6 (0.4) vs. 5.7 (0.4) COM/FUL3 (HYP vs. WL ↓): 6.14 (2) vs. 6 (2.43) COM/FUL3 (↓ HYP vs. WL): −1.14 (1.77) vs. −0.84 (2.11); p  = ns b COM/P (↓ HYP vs. WL): 4.6 (2) vs. 5.7 (2.3) COM/P (↓ HYP vs. EDU): 4.6 (2) vs. 5.5 (2.3) COM/P (↓ HYP vs. MSG): 4.6 (2) vs. 5.3 (2.2) COM/P (↓ HYP vs. MSG): 2.6 (2.8) vs. 3.6 (2.6) COM/FUL3 (↓ HYP vs. MSG): 3 (2.8) vs. 3.6 (2.1) COM/FUL3 (↓ HYP vs. DIET): −2.7 (2.68) vs. −2.6 (3.12); p  = ns b COM/FUL6 (↓ HYP vs. DIET): −3.3 (3.32) vs. −3 (2.62); p  = ns b ADJ/FUL3 (↓ HYP vs. DIET): −3.1 (2.81) vs. −2.7 (2.68); p  = ns b ADJ/FUL6 (HYP vs. DIET ↓): −2.9 (3.19) vs. −3.3 (3.32); p  = ns b COM/P (↓ HYP vs. EDU): 4.38 (2.9) vs. 4.67 (2.84); p  < 0.001 COM/P (↓ HYP vs. MF): 4.38 (2.9) vs. 4.56 (2.79); p  = ns COM/P (↓ HYP vs. TAU): 3.7 (1.7) vs. 5.5 (5.4); p  < 0.005 COM/FUL1 (↓ HYP vs. TAU): 4.9 (0.9) vs. 5.4 (1.5) ADD/P (↓ HYP vs. BIO): 2.56 (1.2) vs. 4.5 (2.61) ADD/FUL1 (↓ HYP vs. BIO): 2.42 (1.23) vs. 4.48 (2.17) ADD/P (↓ HYP vs. MF): 2.3 (2.42) vs. 4.5 (2.61) ADD/FUL1 (↓ HYP vs. MF): 3.31 (1.28) vs. 4.48 (2.17) ADJ/P (↓ HYP vs. EDU): 4.4 (2.14) vs. 5.6 (2.21) ADJ/FUL3 (↓ HYP vs. EDU): 4.4 (2.93) vs. 5.1 (2.48) COM/P (↓ HYP vs. TAU): 1.77 (0.83) vs. 2.59 (1.47); p  = 0.316 (ns) COM/FUL1 (HYP vs. TAU ↓): 3.7 (1.54) vs. 3.33 (1.07) COM/FUL3 (↓ HYP vs. TAU): 2.43 (1.79) vs. 2.45 (0.56) COM/FUL6 (↓ HYP vs. TAU): 1.4 (0.89) vs. 2.23 (1.41) COM/P (HYP vs. TAU ↓): 4.13 (1.09) vs. 3.81 (1.51) COM/P (HYP vs. TAU ↓): −0.09 (1.48) vs. −0.46 (2.11); p  = 0.686 (ns) b COM/P (HYP vs. ACU ↓): 3.1 (1.25) vs. 2.85 (1.31) b COM/FUL6 (HYP vs. ACU ↓): 2.2 (1.11) vs. 1.35 (0.81) b COM/P (HYP vs. ACU ↓): 4.95 (1.76) vs. 4.4 (2.19) COM/FUL6 (HYP vs. ACU ↓): 5.9 (1.37) vs. 5.85 (1.39) COM/P (↓ HYP vs. TAU): 2.5 (3.2) vs. 3.2 (3.4); p  = 0.21 (ns) COM/P (HYP vs. MUS ↓): 2.5 (3.2) vs. 2 (2.8); p  = 0.18 (ns) COM/P (↓ HYP vs. EDU): −0.78 (1.12) vs. −0.76 (1.26) b COM/P (HYP vs. CBT ↓): −0.78 (1.12) vs. −1.32 (1.73) b ADJ/P (↓ HYP vs. CBT): −1.54 (1.93) vs. −1.32 (1.73) b COM/P (↓ HYP vs. EDU): 3.67 (1.8) vs. 3.72 (1.91) COM/P (HYP vs. CBT ↓): 3.67 (1.8) vs. 3.58 (1.86) ADJ/P (↓ HYP vs. CBT): 3.37 (2.07) vs. 3.58 (1.86) COM/FUL6 (HYP vs. EDU ↓): 3.88 (2.09) vs. 3.66 (1.79) COM/FUL6 (HYP vs. CBT ↓): 3.88 (2.09) vs. 3.65 (1.79) ADJ/FUL6 (↓ HYP vs. CBT): 3.35 (2.08) vs. 3.65 (1.79) COM/P (↓ HYP vs. TAU): 6.7 (2.9) vs. 7.2 (2.5) COM/P (↓ HYP vs. RLX): 6.7 (2.9) vs. 7.1 (2.5) COM/P (HYP vs. CBT): 5.88 (1.62) vs. 5.88 (1.65) COM/FUL6 (↓ HYP vs. CBT): 5.57 (1.91) vs. 6.11 (1.52) COM/P (↓ HYP vs. MUS): 5.88 (1.62) vs. 5.91 (1.45); p  = ns COM/FUL6 (↓ HYP vs. MUS): 5.75 (1.91) vs. 5.83 (1.44) ADJ/P (HYP vs. SC ↓): 5.88 (1.62) vs. 5.42 (1.66) ADJ/FUL6 (HYP vs. SC ↓): 5.75 (1.91) vs. 5.51 (1.82) COM/P (↓ HYP vs. CBT): 3.45 (1.52) vs. 4.11 (1.72); p  = significant COM/P (HYP vs. MF ↓): 3.45 (1.52) vs. 3.38 (2); p  = ns COM/P (HYP vs. PL ↓): 2.59 (2.03) vs. 2.51 (2.17) COM/P (↓ HYP vs. PL): −0.76 (2.34) vs. −0.25 (0.92) b COM/P (HYP vs. TAU ↓): 4.13 (1.09) vs. 3.81 (1.51); p  = ns COM/FUL3 (↓ HYP vs. TAU): 3.63 (2.15) vs. 4.07 (1.41); p  = ns ADJ/P (↓ HYP vs. TENS): −2.36 (2.25) vs. −2.26 (2.04); p  = ns b ADJ/P (↓ HYP vs. TENS): 3.43 (2.13) vs. 3.49 (2.03); p  = ns COM/P (↓ HYP vs. PL): 4.52 (2.21) vs. 5.1 (2.08) COM/P (HYP vs. TAU ↓): 4.52 (2.21) vs. 4.44 (1.95) COM/P (HYP vs. TAU ↓): 1.9 (2.3) vs. 1.6 (2.1); p  = 0.825 (ns) COM/P (HYP vs. TAU ↓): 0.9 (2.7) vs. 0.8 (2.1) b COM/P (↓ HYP vs. EDU): 2.54 (1.21) vs. 3.92 (2.13) COM/P (↓ HYP vs. EDU): −1.58 (1.74) vs. −0.18 (1) b COM/P (↓ HYP vs. CBT): 2.54 (1.21) vs. 3.65 (2.43) COM/P (↓ HYP vs. CBT): −1.58 (1.74) vs. −0.7 (1.41) b ADJ/P (↓ HYP vs. CBT): 2.69 (2.09) vs. 3.65 (2.43) ADJ/P (↓ HYP vs. CBT): −1.55 (1.77) vs. −0.7 (1.41) b COM/P (↓ HYP vs. TAU): −1.5 (1.54) vs. −1.25 (1.92); p  = 0.360 (ns) b COM/FUL1W (↓ HYP vs. TAU): −1.75 (2.42) vs. −0.4 (2.26); p  = 0.07 b COM/P (↓ HYP vs. EDU): −0.61 (0.29) vs. −0.57 (0.29); p  = 0.39 (ns) b COM/FUL3 (↓HYP vs. EDU): −0.95 (0.37) vs. −0.48 (0.35); p  = 0.05 b COM/FUL6 (↓HYP vs. EDU): −1.04 (0.35) vs. −0.28 (0.32); p  < 0.001 b COM/P (HYP vs. MF ↓): −0.61 (0.29) vs. −0.85 (0.29); p  = 0.39 (ns) b COM/FUL3 (↓HYP vs. MF): −0.95 (0.37) vs. −0.67 (0.34); p  = 0.05 b COM/FUL6 (↓HYP vs. MF): −1.04 (0.35) vs. −0.86 (0.32); p  < 0.001 b COM/P (↓ HYP vs. RLX): 4.49 (2.11) vs. 4.5 (1.95) COM/P (↓ HYP vs. RLX): −1.61 (2) vs. −1.35 (1.69) b COM/P (↓ HYP vs. EDU): −0.78 (1.12) vs. −0.76 (1.26) b COM/P (HYP vs. CBT ↓): −0.78 (1.12) vs. −1.32 (1.73) b ADJ/P (↓ HYP vs. CBT): −1.54 (1.93) vs. −1.32 (1.73) b COM/P (HYP vs. TAU ↓): 1.8 (2.25) vs. 1.3 (1.42); p  = ns COM/FUL1 (HYP vs. TAU ↓): 1.6 (2.01) vs. 1.1 (1.1); p  = ns COM/P (HYP vs. EDU ↓): 5.46 (2.01) vs. 4.68 (2.09); p  = ns COM/P (HYP vs. EDU ↓): 0.17 (1.28) vs. −0.48 (2.39); p  = ns b COM/P (HYP vs. MF ↓): 5.46 (2.01) vs. 4.33 (2.19); p  = ns COM/P (HYP vs. MF ↓): 0.17 (1.28) vs. −0.17 (1.55); p  = ns b COM/FUL1W (↓ HYP vs. TAU): 3.12 (2.28) vs. 3.42 (1.69); p  = ns COM/FUL3 (↓ HYP vs. TAU): 2.5 (2.50) vs. 3.0 (2.06); p  = ns COM/P (↓ HYP vs. TAU): 5.6 (2.1) vs. 6.4 (1.2); p  = 0.31 (ns) COM/FUL6 (↓ HYP vs. TAU): 5.3 (3.4) vs. 6.2 (2.5); p  = 0.7 (ns) COM/P (↓ HYP vs. PL): 4.4 (0.9) vs. 9 (1.3); p  < 0.05 COM/FUL3 (↓ HYP vs. PL): 4.5 (1.2) vs. 9.1 (1); p  < 0.05 COM/P (↓ HYP vs. TAU): 5.06 (2.32) vs. 5.67 (1.85) COM/P (↓ HYP vs. TAU): −1.06 (1.73) vs. −0.5 (1.46) b COM/P (V vs. TAU ↓): 2.15 (18.81) vs. 0.32 (13.97); p  = 0.71 (ns) b COM/P (↓ IPT vs. TAU): −1.97 (21.9) vs. 0.32 (13.97); p  = 0.71 (ns) b Abbreviations: A, acute pain; ACU, acupuncture; ADD, addition of stated intervention on top of hypnosis; ADJ, adjunctive use of hypnosis on top of stated intervention; AT, autogenic training; BIO, bio‐feedback relaxation; C, chronic; CBT, cognitive‐behavioural therapy; COM, comparison between hypnosis and stated control intervention; DIS, distraction; EDU, pain education; FUL1, follow‐up less than 1 month; FUL12, follow‐up 6–12 months; FUL1W, follow‐up less than 1 week; FUL3, follow‐up 1–3 months; FUL6, follow‐up 3–6 months; H, hypnosis; IPT, in‐person taped/audio hypnosis; MF, mindfulness‐based intervention; MSG, massage; MUS, music; NB, non‐binary; ns, not statistically significant; P, post‐intervention; PHY, physiotherapy; PL, placebo; RLX, relaxation; SELF, self‐hypnosis only; SRS, stress‐reducing strategies; TAU, treatment as usual; TENS, transcutaneous electrical nerve stimulation; TRP, in‐person therapist‐led hypnosis; TRP+SELF, in‐person therapist with regular self‐hypnosis; V, virtual‐reality hypnosis; WL, waiting list. Indicates studies that have been included in at least one meta‐analyses. Indicates pre‐post mean difference scores. Studies were conducted in 22 countries, most commonly the USA (34) and France (12), followed by Iran and Australia (five each); other countries contributed four or fewer studies. Across 106 studies, 9872 participants were included. Excluding four studies without gender reporting (Patterson et al.  1992 ; Mackey  2018 ; Askay et al.  2022 ; Rhodes et al.  2025 ), samples comprised 3000 (31.6%) men, 6478 (68.3%) women, and four participants identifying as non‐binary or transgender; female over‐representation was expected given higher chronic pain susceptibility in women (Bartley and Fillingim  2013 ). Among 65 studies reporting mean age, mean ages ranged from 23.2 years (Ghoneim et al.  2000 ) to 80.6 years (Ardigo et al.  2016 ). Fifty studies involved chronic pain conditions (e.g., fibromyalgia, diabetic neuropathic pain, cancer pain and head/neck pain) and 56 involved acute/procedural pain (e.g., elective surgery, burns, tooth extraction and prostate needle biopsy). Delivery modality most commonly involved in‐person practitioner‐delivered hypnosis only (33 studies) or therapist‐delivered hypnosis supplemented by self‐hypnosis/self‐practice (31 studies). Thirteen studies primarily trained self‐hypnosis, 22 used in‐person tape/audio‐delivered hypnosis and seven used virtual reality. The mean number of sessions before outcome assessment was 4.81 (range: single perioperative session to 42 daily self‐hypnosis sessions; Anderson et al.  2025 ); session duration ranged from 10 min to 2 h. Many studies adapted established protocols, including Barber's Rapid Induction of Analgesia script (Barber  1977 ) (e.g., Everett et al.  1993 ; Abrahamsen et al.  2008 ; Slack et al.  2009 ; Jafarizadeh et al.  2018 ), Erickson and Rossi's indirect permissive approach (Erickson and Rossi  1979 ) (e.g., Gay et al.  2002 ; Akgul et al.  2016 ), or Torem's age progression/regression technique (Torem  1992 ) (e.g., Rizzo et al.  2018 ; Mendoza et al.  2022 ; Starosta et al.  2024 ). Across protocols, inductions typically involved attentional focus (e.g., eye fixation) and imaginative involvement; deepening commonly used progressive muscle relaxation (Frenay et al.  2001 ; Jafarizadeh et al.  2018 ), breathing exercises or hand levitation (Castel et al.  2009 ; Jafarizadeh et al.  2018 ). Analgesic suggestions included relaxation/well‐being, reduced pain intensity/unpleasantness, increased control or transformation/substitution of pain sensations (e.g., warmth, cooling and tingling), sometimes with post‐hypnotic suggestions to encourage durability (Eaton et al.  2023 ; Kaczmarska et al.  2024 ). Visualisation of a safe/quiet/preferred place was common (Stam et al.  1986 ; Abrahamsen et al.  2008 ; Abrahamsen et al.  2009 ; Jensen et al.  2009b ; Jensen et al.  2009a ; Ardigo et al.  2016 ; Amraoui et al.  2018 ; Garcia et al.  2020 ; Jensen et al.  2020 ; Bicego et al.  2021 ; Lang et al.  2021 ; Tonye‐Geoffroy et al.  2021 ), alongside other imagery such as a leaf floating (Castel et al.  2007 ) or an ‘analgesic glove’ (Bicego et al.  2021 ). Some studies incorporated self‐care (Vanhaudenhuyse et al.  2015 ), cognitive‐behavioural elements (Castel et al.  2009 ) or group formats (Castel et al.  2012 ). Relaxation was the most common active control (17 studies) (Stam et al.  1986 ; VanDyck et al.  1991 ; Spinhoven et al.  1992 ; Zitman et al.  1992 ; Ter Kuile et al.  1994 ; Gay et al.  2002 ; Askay et al.  2007 ; Castel et al.  2007 ; Abrahamsen et al.  2008 ; Abrahamsen et al.  2009 ; Jensen et al.  2009b ; Jensen et al.  2009a ; Phillips‐Moore et al.  2015 ; Garcia et al.  2020 ; Werner et al.  2020 ; Eaton et al.  2023 ; Anderson et al.  2025 ), typically incorporating body scanning and somatic suggestions to deepen relaxation; some used Jacobson's progressive muscle relaxation (Gay et al.  2002 ; Jensen et al.  2009b ) or autogenic training (VanDyck et al.  1991 ; Spinhoven et al.  1992 ; Zitman et al.  1992 ; Ter Kuile et al.  1994 ). One study used biofeedback to enhance relaxation (Jensen et al.  2009b ), and one blended relaxation with pain education and cognitive behavioural strategies (Anderson et al.  2025 ). Other common active controls were pain education (10 studies) (Patterson et al.  1992 ; Melzack et al.  1996 ; Vanhaudenhuyse et al.  2015 ; Garland et al.  2017 ; Rizzo et al.  2018 ; Jensen et al.  2020 ; Mendoza et al.  2022 ; Williams et al.  2022 ; Jensen et al.  2023 ; Anastas et al.  2024 ) and CBT (8 studies) (Syrjala et al.  1992 ; Castel et al.  2009 ; Castel et al.  2012 ; Jensen et al.  2020 ; Bicego et al.  2021 ; Hanley et al.  2021 ; Mendoza et al.  2022 ; Jensen et al.  2023 ). Pain education involved psychoeducation about pain and modulation (e.g., spinal cord mechanisms, pain ≠ harm) (Knafo and Weinberger  2024 ) and coping strategies; CBT typically targeted maladaptive appraisals and skills development, including behavioural strategies (Williams et al.  2020 ). One study blended CBT with pain education (Anderson et al.  2025 ). Additional active controls not pooled included music (Nilsson et al.  2001 ; Nilsson et al.  2003 ; Mackey  2010 ; Téllez et al.  2016 ; Mackey  2018 ; Sánchez‐Jáuregui et al.  2019 ; Bicego et al.  2021 ), mindfulness (Garland et al.  2017 ; Jensen et al.  2018 ; Hanley et al.  2021 ; Williams et al.  2022 ; Anastas et al.  2024 ), self‐care (Bicego et al.  2021 ) and stress‐reduction (Frenay et al.  2001 ). Physiological controls included biofeedback (Jensen et al.  2018 ), low FODMAP diet (IBS) (Peters et al.  2016 ), massage (Vanhaudenhuyse et al.  2015 ; Ardigo et al.  2016 ), acupuncture (Tastan et al.  2018 ; Razak et al.  2019 ) and TENS (Tonye‐Geoffroy et al.  2021 ). Additionally, as none of the comparisons between combined therapy (i.e., hypnosis combined with an active intervention) and hypnosis or the active intervention only reached the critical mass required for meta‐analysis, the effect sizes were not pooled and were instead narratively reported. Narrative synthesis of studies not suitable for pooling provided no consistent evidence that hypnosis produced greater pain reduction than other active comparators. Findings were mixed for music and physiological treatments, while hypnosis generally did not differ significantly from mindfulness, stress‐reduction training, self‐care, TENS, acupuncture, massage or low FODMAP diet, with isolated study‐level advantages insufficient to alter the overall conclusion (see Data  S5 for full details). Inter‐rater reliability for RoB criteria ranged from κ = −0.02 to 0.85, with weak to almost perfect agreement across 11 domains and perfect agreement in one domain (Data  S2 ). No/minimal agreement (0 ≤ κ ≤ 0.39) occurred in 10 domains (inclusion/exclusion met, pre‐existing pain, abstinence/reporting of analgesic use, hypnotic intervention, matching, complete outcome data, random allocation, training for hypnotist/therapist, active engagement and deviation from protocol) and disagreement (κ < 0) occurred for reporting participant characteristics; ambiguity and inconsistent reporting likely contributed, and discrepancies were resolved by consensus. After resolution, 58 studies (54.72%) were rated low RoB, 44 (41.51%) moderate and 4 (3.77%) high (Data  S3 ). Most notably, many studies did not report treatment expectations (92; 86.8%), active engagement (85; 80.2%), hypnotisability (84; 79.2%), investigator blinding (71; 67.0%), data collection blinding (65; 61.3%) or abstinence from medication (56; 52.8%; Figure  2 ). Graph showing the percentage of studies meeting each of the 21 RoB assessment criteria. n.c, sample size of control group; n.e, sample size of experimental group; SE, standard error of standardised mean difference; SMD, standardised mean difference. Fifty‐eight studies ( n  = 4621; hypnosis n  = 2351; control n  = 2270) reported post‐intervention pain intensity comparing hypnosis with non‐active/minimal controls (e.g., distraction, treatment as usual and waitlist). Hypnosis reduced post‐intervention pain versus controls (SMD −0.38; 95% CI −0.56 to −0.21; t  = −4.37; p  < 0.0001), with high heterogeneity ( τ 2  = 0.29; 95% CI 0.22–0.63; I 2  = 76.4%; 95% CI 69.7%–81.6%; p  < 0.0001). One outlier (Dorta et al.  2024 ) (SMD −4.03; moderate RoB) contributed substantially. Excluding the outlier yielded 57 studies ( n  = 4572; hypnosis n  = 2327; control n  = 2245), with a slightly smaller but still significant effect (SMD −0.33; 95% CI −0.47 to −0.19; t  = −4.79; p  < 0.0001; Figure  3 ) and reduced heterogeneity ( τ 2  = 0.16; 95% CI 0.10–0.34; I 2  = 70.3%; 95% CI 61.1%–77.2%; p  < 0.01). Forest plot comparing effects of hypnosis or non‐active controls on post‐intervention pain intensity (Abbreviations as for Figure  2 ). Subgroup analyses showed no significant differences when comparing low (27 studies) with moderate (30 studies) RoB (low: SMD −0.25; 95% CI −0.44 to −0.07; moderate: SMD −0.40; 95% CI −0.61 to −0.20; p  = 0.26); control type (placebo, 17 studies: SMD −0.31; 95% CI −0.47 to −0.15; waitlist, 9 studies: SMD −0.52; 95% CI −0.83 to −0.21; treatment‐as‐usual, 31 studies: SMD −0.30; 95% CI −0.53 to −0.07; p  = 0.36); pain type (acute pain, 40 studies: SMD −0.30; 95% CI −0.45 to −0.15; chronic pain, 17 studies: SMD −0.42; 95% CI −0.75 to −0.07; p  = 0.50); or hypnosis modality (self‐hypnosis, eight studies: SMD −0.24; 95% CI −0.46 to −0.02; therapist‐delivered, 24 studies: SMD −0.35; 95% CI −0.57 to −0.13; in‐person tape, nine studies: SMD −0.33; 95% CI −0.74 to 0.08; therapist‐delivered with self‐hypnosis practice, 10 studies: SMD −0.39; 95% CI −0.99 to 0.22; virtual‐reality, six studies: SMD −0.31; 95% CI −0.75 to 0.13; p  = 0.94). Eight studies ( n  = 520; hypnosis n  = 261; control n  = 259) reported pre–post change scores (Figure  4 ). Hypnosis showed a larger decrease than controls, but was not significant (SMD −0.31; 95% CI −0.73 to 0.11; t  = −1.76; p  = 0.12), with substantial heterogeneity ( τ 2  = 0.11; 95% CI 0.01–1.24; I 2  = 58.7%; 95% CI 9.7%–81.1%; p  = 0.002). Forest plot comparing the effects of hypnosis or non‐active controls on pain intensity mean difference (Abbreviations as for Figure  2 ). Seven studies ( n  = 223; hypnosis n  = 113; control n  = 110) reported post‐intervention pain at 1–3 months follow‐up. The pooled effect was large but non‐significant (SMD −0.96; 95% CI −2.35 to 0.43; t  = −1.70; p  = 0.14) with high heterogeneity ( τ 2  = 2.02; 95% CI 0.71–10.84; I 2  = 90.1%; 95% CI 82.2%–94.5%; p  < 0.0001), again influenced by outlier (Dorta et al.  2024 ) (SMD −4.11; moderate RoB). Excluding the outlier (Dorta et al.  2024 ) left six studies ( n  = 174; hypnosis n  = 89; control n  = 85; Figure  5 ), with a smaller non‐significant effect (SMD −0.39; 95% CI −1.08 to 0.29; t  = −1.48; p  = 0.20) and lower heterogeneity ( τ 2  = 0.21; 95% CI 0.00–2.63; I 2  = 53.9%; 95% CI 0.0%–81.6%; p  = 0.05). Forest plot comparing effects of hypnosis or non‐active controls on post‐intervention pain intensity at 1–3 months follow‐up (Abbreviations as for Figure  2 ). Fourteen studies ( n  = 940; hypnosis n  = 482; relaxation n  = 458) compared hypnosis with relaxation on post‐intervention pain (Figure  6 ). The pooled effect was small and non‐significant (SMD −0.13; 95% CI −0.42 to 0.16; t  = −0.98; p  = 0.34), with substantial heterogeneity ( τ 2  = 0.14; 95% CI 0.04–0.63; I 2  = 65.4%; 95% CI 39.0%–80.4%; p  = 0.0003). Subgroup analyses by RoB showed no significant difference ( p  = 0.64), with low RoB (SMD −0.08) and moderate RoB (SMD −0.31). Forest plot comparing the effects of hypnosis or relaxation on post‐intervention pain intensity (Abbreviations as for Figure  2 ). Seven studies ( n  = 606; hypnosis n  = 353; pain education n  = 253) compared hypnosis with pain education on post‐intervention pain (Figure  7 ). The effect was small and non‐significant (SMD −0.19; 95% CI −0.64 to 0.26; t  = −1.05; p  = 0.33), with substantial heterogeneity ( τ 2  = 0.13; 95% CI 0.01–1.30; I 2  = 65.2%; 95% CI 21.8%–84.5%; p  = 0.008). Forest plot comparing the effects of hypnosis or pain education on post‐intervention pain intensity (Abbreviations as for Figure  2 ). Five studies ( n  = 505; hypnosis n  = 255; pain education n  = 250) reported pre–post change scores (Figure  8 ). The pooled effect was small and non‐significant (SMD −0.17; 95% CI −0.76 to 0.42; t  = −0.81; p  = 0.46), with substantial heterogeneity ( τ 2  = 0.16; 95% CI 0.02–1.92; I 2  = 73.3%; 95% CI 33.3%–89.3%; p  = 0.005). Forest plot comparing the effects of hypnosis or pain education on pain intensity mean difference (Abbreviations as for Figure  2 ). Five studies ( n  = 399; hypnosis n  = 204; CBT n  = 195) compared hypnosis with CBT on post‐intervention pain (Figure  9 ). The pooled effect was not statistically significant (SMD −0.32; 95% CI −0.72 to −0.08; t  = −2.21; p  = 0.09), with low‐to‐moderate heterogeneity ( τ 2  = 0.03; 95% CI 0.00 –1.04; I 2  = 38.6%; 95% CI 0.00%–77.2%; p  = 0.16). Forest plot comparing the effects of hypnosis or cognitive‐behavioural therapy on post‐intervention pain intensity (Abbreviations as for Figure  2 ). Excluding high‐RoB studies did not change directions of effect or statistical significance. Visual inspection of funnel plots for primary analyses with > 10 studies suggested no substantial asymmetry (Figure  10 ). For hypnosis versus non‐active controls (post‐intervention pain), Egger's test was non‐significant ( t  = −1.16, df = 55, p  = 0.25), indicating no evidence of small‐study effects. Funnel plots of studies comparing the effects of hypnosis and non‐active control (left) or relaxation (right) on post‐intervention pain intensity. Because included studies were RCTs, the evidence base received an initial ‘high certainty’ rating. Evidence was not downgraded for limitations in study design/execution, indirectness or dissemination bias. Evidence was downgraded for inconsistency (substantial heterogeneity across most meta‐analyses) and imprecision (most comparisons had small‐to‐moderate sample sizes and/or non‐significant effects, except hypnosis versus non‐active controls on post‐intervention pain, n  = 4572). Final certainty was ‘moderate’ for hypnosis versus non‐active controls and hypnosis versus CBT (post‐intervention pain), and ‘low’ for other meta‐analytic conclusions (Data  S4 ).

To the best of our knowledge, this is the largest meta‐analysis to date critically summarising the analgesic efficacy of hypnosis compared with minimal or non‐active controls (placebos, waitlists and treatment as usual) and active controls (relaxation, CBT and pain education). Across the primary and secondary aims of this systematic review, results converged on a consistent conclusion: hypnosis produces, at best, modest reductions in pain intensity that are comparable to those achieved by other psychological interventions. Compared to non‐active controls or minimal intervention, aggregate post‐intervention pain intensity following hypnosis was reduced by less than one point on a 1–10 scale. Analyses based on pre–post difference scores, arguably a more conservative indicator of treatment‐related change, yielded a similarly small and non‐significant effect. These modest effects persisted after excluding outlier studies and studies at high risk of bias. Taken together, the observed reductions are unlikely to be experienced by most patients as meaningful improvement, if detected at all (Ferreira et al.  2012 ; Strijkers et al.  2021 ). Beyond comparisons with non‐active controls, hypnosis did not demonstrate superior analgesic effects relative to active interventions, including relaxation, pain education and CBT; effect size differences were generally smaller than those observed against non‐active controls. Narrative synthesis of studies comparing hypnosis with other sensory (e.g., music), psychological (e.g., mindfulness) and physiological interventions (e.g., massage, acupuncture, TENS and biofeedback) likewise provided no basis to revise the central conclusion. Notably, mindfulness and music produced reductions in pain intensity comparable to hypnosis, reinforcing the interpretation that hypnosis does not occupy a privileged position among non‐pharmacological pain treatments. Subgroup analyses did not identify reliable factors influencing hypnosis efficacy. More extensive intervention protocols (i.e., therapist‐delivered hypnosis with self‐hypnosis practice) produced the largest reduction in pain intensity (SMD −0.39). Hypnosis was also more effective at reducing chronic pain than acute or procedural pain (acute pain: SMD −0.30 vs. chronic pain: SMD −0.42). Nevertheless, all observed effect sizes were small, and neither more extensive intervention protocols (e.g., therapist‐delivered hypnosis with self‐practice) nor pain type (acute vs. chronic) significantly improved the small effect sizes observed. Future studies or reviews could consider whether the brevity of most hypnosis interventions provides sufficient opportunity to learn or practise hypnotic analgesia, and whether chronic pain patients have greater motivation and more opportunity to practise self‐regulatory strategies. Patient characteristics, pain context and intervention features remain potential moderators of hypnosis response and are worthy of further exploration. The absence of identifiable moderators in this review may reflect the high heterogeneity of effect sizes across analyses, indicating that hypnosis effects are highly variable. An umbrella review reported wide‐ranging effects of hypnosis on clinical and procedural pain, with effect sizes spanning d  = 0.37–0.81 and d  = 0.10–2.53, respectively (Rosendahl et al.  2024 ). In some circumstances, hypnosis can provide notable pain relief: hypnotic suggestions can effectively modulate experimental pain (Thompson et al.  2019 ), laboratory studies report substantial control over clinical pain (Derbyshire et al.  2009 ), and there are reports of childbirth, oral surgery or coronary artery bypass grafting proceeding with minimal pain using hypnoanalgesic procedures alone (Akgul et al.  2016 ; Venkiteswaran and Tandon  2021 ; Fernández‐Gamero et al.  2024 ). Nevertheless, our meta‐analysis provides little evidence for consistent or reliable aggregate pain relief. The high heterogeneity and absence of clear moderators suggest that hypnosis efficacy cannot be predicted from clinical presentation or treatment characteristics, and thus undermine the utility of hypnosis as a scalable, targeted intervention. Overall, pain relief following hypnosis is comparable to that achieved with other psychological therapies, which typically yield small effect sizes (0.1–0.4) corresponding to minimal changes in pain reports (Morley et al.  1999 ; Veehof et al.  2016 ; Williams et al.  2020 ; Strijkers et al.  2021 ). Our results, alongside other systematic reviews, suggest that any psychological processes exacerbating pain intensity may be difficult to modify with psychological therapy (Veehof et al.  2016 ; Vlaeyen and Linton  2000 ; Williams et al.  2020 ; Strijkers et al.  2021 ; Turk  2003 ; Moseley and Butler  2015 ). These results align with broader concerns about the limited population‐level impact of many medical interventions despite plausible mechanisms (Stegenga  2018 ). Examples include antibiotics for acute sinusitis, which only cure one in 17 patients within the recommended treatment period (Lemiengre et al.  2018 ), statins, which do not prevent death in patients with no known heart disease (Ray et al.  2010 ), prostate‐specific antigen tests, which do not protect against death from prostate cancer (Djulbegovic et al.  2010 ), and routine health checks, which do not reduce mortality (Krogsbøll et al.  2019 ). In this context, modest and inconsistent effects of hypnosis are not anomalous; rather, they reflect a recurring pattern in evidence‐based medicine. This does not imply that hypnosis should be excluded from clinical care. Unlike medical interventions with meaningful adverse‐effect profiles (e.g., statin‐associated muscle symptoms or myopathy; opioid addiction risk; NSAID‐related gastrointestinal bleeding, renal impairment and cardiovascular risk), psychological interventions are often part of a ‘gentle’ medicine involving less aggressive approaches to ill health (Stegenga  2018 ). Recommendations to use hypnosis should therefore be grounded in an honest appraisal: it is generally low risk and may help some patients manage pain‐related distress or improve coping even when pain intensity is unchanged (Stegenga  2018 ). There also remains the possibility that further development of psychological methods arising from better mechanistic understanding and more personalised treatment will yield more successful psychological approaches to pain management. A key strength of this meta‐analysis is the robustness of results across sensitivity analyses. Removing all studies rated as high risk of bias did not materially alter pooled estimates or change conclusions regarding statistical significance, suggesting that findings were not driven by methodologically weaker studies. In addition, we found no strong evidence of publication bias across primary analyses, reducing concern that the results reflect selective reporting. Together, these features strengthen confidence in the overall pattern of findings. Several limitations warrant consideration. First, most analyses relied on post‐intervention pain intensity rather than pre–post change scores. This approach implicitly assumes baseline equivalence between groups, an assumption that may not always hold (26.4% of included studies did not report evidence of group matching). Second, pooling self‐reported VAS or NRS measures across studies can be questioned: pain is multidimensional and cannot be captured by intensity alone. Future reviews should evaluate additional outcomes, including pain interference, pain unpleasantness and related functional or affective measures. In addition, while our systematic review does not provide evidence of hypnosis substantially reducing pain intensity, the modest average effect sizes likely obscure a mix of effect sizes for different people and different subgroups. Some patients do experience meaningful pain relief, while others experience other beneficial effects, such as a return to work, elevated mood and a general ability to manage their lives despite their pain. Future investigations might focus less on aggregate pain intensity measures and more on what makes the experience of patients different. Finally, risk‐of‐bias assessment inevitably involves judgement. The selected criteria and scoring are open to dispute, and disagreements arose both between—and at times within—the current authors. For instance, it is unclear what behavioural observations or procedural requirements are suitable to indicate patient engagement with hypnosis therapy (e.g., pain diaries, engagement questionnaires, qualitative indicators or other indicators of compliance). Similar issues arise in determining what counts as ‘adequate detail’ or ‘clearly reported’, underscoring the subjective elements and interpretation that accompany RoB assessments in complex behavioural interventions. Nevertheless, subgroup analysis based on RoB ratings did not affect the outcomes, and previous systematic reviews while being diligent about assessing RoB still focus on the overall findings without excessive concern about the variable RoB. Reviewers frequently disagree on study quality, leaving comprehensiveness as the most defensible strategy, which this review adopted (Stegenga  2018 ).

To the best of our knowledge, this is the largest systematic review and meta‐analysis examining the analgesic effects of hypnosis on clinical pain compared with non‐active/minimal interventions and active controls. Across a wide range of pain conditions, delivery modalities and comparators, hypnosis was associated with small and inconsistent reductions in pain intensity that rarely reached thresholds of clinical importance and did not significantly exceed the effects of other psychological interventions. Future primary studies and systematic reviews should prioritise broader pain outcomes (e.g., pain unpleasantness, pain interference and perceived control) and longer‐term follow‐up to better characterise the role of hypnosis within longer‐term and comprehensive pain management. Overall, these findings support a recalibration of expectations regarding hypnosis for pain intensity reduction: clearer framing of its likely benefits will better serve patients, clinicians and the scientific field.

Chronic pain imposes a substantial global burden, with prevalence estimates ranging from 20% to 40% that generate major contributions to disability and economic costs approaching $600 billion annually (GBD 2016 Disease and Injury Incidence and Prevalence Collaborators  2017 ; Cohen et al.  2021 ; Jackson et al.  2016 ; U.S. Burden of Disease Collaborators  2013 ). Despite extensive pharmacological treatment, many patients achieve inadequate relief and concerns about opioid‐related harms have increased interest in safer psychological approaches (Sullivan and Ballantyne  2023 ). Non‐pharmacological interventions such as cognitive behavioural therapy (CBT) and acceptance‐based therapies typically yield modest reductions in pain intensity, averaging approximately one point on a 0–10 scale (Williams et al.  2020 ; Venkiteswaran and Tandon  2021 ; Jenkins et al.  2025 ). These effects are often borderline clinically meaningful and do not consistently exceed placebo controls (Ferreira et al.  2012 ; Strijkers et al.  2021 ), although some patients experience broader benefits in emotional or functional outcomes. Hypnosis has similarly been reported to reduce acute, procedural, and chronic pain by roughly one point on a 0–10 scale (Adachi et al.  2014 ; Langlois et al.  2022 ; Rosendahl et al.  2024 ; Yerzhan et al.  2025 ), with considerable variability across studies. As an additional intervention, hypnosis has shown modest enhancement of pain education or pharmacological treatments, though effects diminish when controlling for time and attention (Jones et al.  2024 ). Prior meta‐analyses, however, have often included small samples, pooled heterogeneous control conditions or focused exclusively on placebo comparisons or additional effects, limiting clinical interpretability (Ramondo et al.  2021 ; Langlois et al.  2022 ; Jones et al.  2024 ; Rosendahl et al.  2024 ; Yerzhan et al.  2025 ). The primary aim of the present meta‐analysis was to evaluate the efficacy of hypnosis in reducing clinical pain intensity compared with non‐active controls (e.g., waitlist, attention or treatment as usual) and relaxation‐based interventions. Second, we compared hypnosis with other active non‐pharmacological treatments (CBT and pain education). Finally, we examined whether delivery modality (recorded, therapist‐delivered, self‐hypnosis or mixed) and pain type (acute vs. chronic) moderated treatment effects.

The authors declare no conflicts of interest.

Data S1: PRISMA checklist. Data S2: Full RoB Assessment Ratings. Data S3: Full Inter‐rater Reliability for RoB Assessment. Data S4: GRADE Level of Evidence Assessment. Data S5: Narrative synthesis of studies excluded from the meta‐analysis.