The Canadian Journal of Psychiatry /
La Revue Canadienne de Psychiatrie
2021, Vol. 66(9) 763–773
© The Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0706743720982432
TheCJP.ca | LaRCP.ca
Background: Repetitive transcranial magnetic stimulation (rTMS) is a promising treatment modality for Post-traumatic stress disorder (PTSD). Several targets and stimulation parameters have been investigated, and while previous meta-analyses have suggested that rTMS is efficacious, these have pooled different stimulation parameters and targets, and the relative efficacy of each is unknown.
Methods: We therefore performed a systematic review and network meta-analysis of randomized controlled trials (RCTs) by searching MEDLINE, EMBASE, CENTRAL, and PsycINFO and retaining RCTs with at least 5 individuals per arm and clinician-rated PTSD symptoms (PROSPERO CRD42019134984). We adhered to PRISMA guidelines, and 2 independent reviewers screened studies for eligibility and extracted the primary outcome of clinician-rated PTSD symptoms. Dropouts were extracted as a proxy for acceptability. Random effects pairwise meta-analyses and a network meta-analysis were performed.
Results: We synthesize data from 10 RCTs with a total of 421 participants. Two rTMS interventions targeting the right dorsolateral prefrontal cortex (DLPFC) improved PTSD symptoms relative to sham: low-frequency stimulation (SMD = 0.70; 95% CI, 0.22 to 1.18) and high-frequency stimulation (SMD = 0.71; 95% CI, 0.11 to 1.31). Medial PFC dTMS, right DLPFC intermittent theta-burst stimulation, and left DLPFC high-frequency stimulation did not separate from sham. Dropouts as a proxy for acceptability revealed no differences between any of the active conditions or sham nor did any of the active conditions differ from each other.
Conclusion: The current literature does not support efficacy differences between interventions; however, protocols stimulating the right DLPFC appear superior to sham. It is unclear whether this reflects heterogeneity in pathology requiring a personalized medicine approach or nonspecific mechanisms of rTMS.
Abrégé
Contexte : La stimulation magnétique transcrânienne répétitive (SMTr) est un mode de traitement prometteur pour le TSPT. Plusieurs cibles et paramètres de stimulation ont été investigués, et bien que des méta-analyses précédentes aient suggéré que la SMTr est efficace, celles-ci ont regroupé différentes cibles et paramètres de stimulation, et l’efficacité relative de chacun est inconnue.
Méthodes : Nous avons donc procédé à une revue systématique et une méta-analyse de réseau d’essais randomisés contrôlés (ERC) en cherchant dans MEDLINE, EMBASE, CENTRAL, et PsycINFO et en retenant les ERC comportant au moins 5 personnes par volet et des symptômes de TSPT jugés par un clinicien (PROSPERO CRD42019134984). Nous avons suivi les directives de PRISMA, et deux réviseurs indépendants relevaient l’admissibilité des études et en tiraient les premiers résultats des symptômes de TSPT jugés par un clinicien. Les abandons étaient réservés comme indicateurs d’acceptabilité. Des méta-analyses par paires d’effets aléatoires et une méta-analyse en réseau ont été effectuées.
Résultats : Nous résumons les données de 10 ERC totalisant 421 participants. Deux interventions de SMTr ciblant le cortex droit dorsolatéral préfrontal (DLPFC) ont amélioré les symptômes du TSPT relatifs à un subterfuge: stimulation de basse fréquence (SMD = 0,70, IC à 95% 0,22 à 1,18) et stimulation de haute fréquence (SMD = 0,71; IC à 95% 0,11 à 1.31). Stimulation haute fréquence PFC médiale dTMS, du CDL DLPFC iTBS droit et DLPFC gauche et la stimulation de haute fréquence n’a pas séparé le subterfuge. Les abandons comme indicateurs d’acceptabilité n’ont pas révélé de différences entre toutes les conditions actives ou le simulacre, et les conditions actives ne différaient pas les unes des autres.
Conclusion : La littérature actuelle nesoutient pas lesdifférences d’efficacitéentreles interventions, quoiqueles protocoles stimulantleDLPFCdroitsemblentsupérieursausimulacre. Iln’est pasdéfini sicelareflète l’hétérogénéitédelapathologiequi nécessiterait une approche médicale personnalisée ou des mécanismes non spécifiques de SMTr.
repetitive transcranial magnetic stimulation, rTMS, post-traumatic stress disorder, systematic review, meta-analysis, network meta-analysis
Post-traumatic stress disorder (PTSD) develops in up to a third of individuals after trauma.1 The lifetime prevalence is 7.5% in the United States,2 and certain groups, such as military personnel, have elevated rates up to 15%.3 Psychotherapy4 and psychopharmacological interventions5 are efficacious; however, not all patients experience clinically meaningful symptomatic improvements. Additional interventions are required to augment conventional treatments, and noninvasive brain stimulation interventions hold great promise.
Repetitive transcranial magnetic stimulation (rTMS) is a noninvasive brain stimulation modality that is well established in the treatment of major mental health conditions such as depression6 and OCD,7 with a growing evidence base for the treatment of PTSD.8-11 rTMS utilizes magnetic fields to induce current in underlying cortex and therefore is well suited to disorders of neural circuitry such as PTSD. The randomized controlled trial (RCT) literature examining rTMS for PTSD has utilized several variants of the intervention, with different stimulation targets and stimulation parameters. These include low-frequency stimulation (LFS) protocols12-18 that decrease the excitability of cortex, and conversely high-frequency stimulation (HFS) protocols12,15,16,19 and novel protocols such as intermittent theta-burst stimulation (iTBS) protocols20 that also enhance the excitability of cortex. To date, the most common target in PTSD has been the right dorsolateral prefrontal cortex (DLPFC);12-18,20 however, other targets include the left DLPFC12 and the dorsomedial PFC (DMPFC).19
Meta-analyses have supported the efficacy of rTMS in the treatment of PTSD by pooling all interventions9-11 and in pooling protocols limited to the right DLPFC.8 Only 2 have been limited to RCTs examining clinician-rated PTSD symptoms, synthesizing an embryonic literature with 3 RCTs (n = 64)8 and 5 RCTs (n = 74), respectively. More recent meta-analyses have been less stringent, pooling effect sizes from open-label studies and RCT data,9 as well as retrospective case-series data.11 More importantly, none of the quantitative syntheses to date resolve the comparative efficacy and acceptability of rTMS interventions in the treatment of PTSD as they have all pooled targets and interventions.
Here, we conduct a systematic review of the RCT literature and update the quantitative synthesis of this field focusing on RCTs with clinician rated PTSD symptoms with a total sample more than 5-fold that of previous meta-analyses. We present novel data by comparing multiple treatments using a network meta-analysis (NMA). The benefit of an NMA is that it allows for indirect comparisons between treatment arms in an RCT that have not been compared before in a real clinical trial by using a common comparator. These new results can offer a framework for clinical decision making.
We registered our protocol a priori with PROSPERO (CRD42019134984) and adhered to PRISMA guidelines.
We searched CENTRAL, EMBASE, MEDLINE, and PsycINFO from inception to July 12, 2019, and then updated March 28, 2020. The full search strategy is detailed in the Supplemental Materials. Two authors (AM & AR) independently screened the identified studies for RCTs of individuals with a primary diagnosis of DSM-IV or DSM-V PTSD. Disagreements resolved with a third author (MTB).
Eligibility criteria were as follows: blinded randomized controlled design, minimum of 2 weeks of daily treatments, and a minimum of 5 cases in each arm. RCTs in PTSD populations with a primary outcome other than PTSD symptoms (such as depressive symptoms or mild-traumatic brain injury) that did not report clinician-rated PTSD symptoms or did not report aggregate pre–post data or effect sizes/mean differences were excluded. Corresponding authors were contacted 3 times at 2-week intervals in an effort to obtain this data if it was not reported in the manuscript.
Two authors (AM & AR) independently extracted the data according to a predetermined extraction checklist (CRD42019134984). The primary outcome was change in clinician-rated PTSD symptoms at the end of the blinded phase, with the exception of 1 study that reported results after the ninth treatment though participants received between 12 and 15 treatments. Secondary outcomes were dropouts as a proxy for treatment acceptability.
Two authors (AM & AR) evaluated the risk of bias using the Cochrane Risk of Bias Tool for Randomized Controlled Trials (RoB2).21 The Grading of Recommendations Assessment, Development and Evaluation (GRADE) risk of bias tool was used to evaluate the quality of evidence associated with the results in the NMA.22
Transitivity was assessed by a global test for inconsistency23 and inconsistency plots assuming loop-specific heterogeneity to determine whether inconsistency existed in each NMA.24-27 Inconsistency was potentially mitigated a priori with the use of a strict inclusion criteria during the screening phases.
We first performed standard pairwise meta-analyses to estimate direct comparisons using random-effects models28 with Comprehensive Meta-Analysis 2.0 (Biostat, Englewood, NJ, USA). We then performed a random-effects multivariate NMA assuming consistency for each outcome with Stata statistical software (metan package, version 3.03; Stata- Corp). The RCT literature examining rTMS for PTSD is principally composed of small trials, therefore for network analyses we did not utilize final scores in our models and instead we calculated within arm standardized mean differences from pre to post for each treatment arm within a study. Based on heterogeneity parameters from pairwise comparisons, we assumed a common heterogeneity parameter across all direct comparisons. We present standardized mean differences (with 95% confidence interval [CI]) for every possible pairwise comparison for the primary outcome, and standardized mean differences or relative risk (with 95% CI) for the secondary outcome. Treatment hierarchy was estimated using the surface under the cumulative ranking curve (SUCRA), a means of comparing efficacy and acceptability. Statistical heterogeneity (t) was compared with empirical distributions, and the assumption of consistency was tested using the loop-specific approach.
Publication bias was tested in pairwise comparisons by visual inspection of the funnel plots and using Egger’s intercept.29-31
The results of our literature search are detailed in Figure 1 and in the Supplementary Material. Of 301 unique records, 31 full-text articles were reviewed and ten studies included in final quantitative analyses (Table 1). This represents data on 421 individuals with PTSD and includes the following rTMS interventions: 1 Hz stimulation of the right DLPFC (LFS-rTMS),13-18 10 Hz or 20 Hz stimulation of the right DLPFC (HFS-rTMS),12,13,16 10 or 20 Hz stimulation of the left DLFPC,12,32 iTBS of the right DLPFC,20 20 Hz deep TMS to bilateral DMPFC,19 and sham. Three of these studies were 3-arm designs,12,13,16 1 study was a 3-arm design, and therefore, we pooled the active-stimulation groups for analyses as the differentiating feature was psychotherapy,19 and the remaining 6 RCTs were 2-arm designs,14,15,17,18,20,32 which provides information on 21 direct comparisons. All but one of the RCTs included a sham condition.15 One study employed a crossover design.17 In one trial, the primary outcome was depressive symptoms; however, PTSD was highly prevalent in this sample of veterans, and we obtained clinician-rated PTSD symptoms from the subsample of participants with SCID-I diagnosed PTSD.32
LFS of the right DLPFC was the most common comparison and had the largest contribution to network estimations for the primary and secondary outcome (Figure 2).
The RCTs were 2,12,13,18,20 4,17 or 615,32 weeks in duration, with the exception of 2 trials that paired rTMS with psychotherapy for 12 treatments.14,19 All studies were conducted in males and females. Four studies were conducted in among veterans,14,15,20,32 3 in civilians,12,16,17 and 3 included both civilian and veteran populations.13,18,19 Nine of the studies utilized the Clinician-Administered PTSD Scale for DSM-IV or DSM-V, whereas 1 study utilized the Treatment Outcome PTSD Scale Score.12 There were no significant differences with respect to age or baseline severity in direct comparisons. The assumption of transitivity is therefore likely to hold in our data.
PTSD symptoms. Pairwise comparisons of direct evidence revealedthatLFSrTMSandHFSrTMSoftherightDLFPC were significantly more efficacious than sham stimulation at reducing PTSD symptoms (Supplemental Figure 1). HFS rTMS of the left DLPFC, dTMS, and iTBS did not separate from sham.
This was paralleled in the NMA results (Figure 3A), with LFS-rTMS and HFS-rTMS of the right DLPFC separating from sham, but no other significant differences relative to sham or between active interventions. According to the SUCRAs, LFS rTMS of the right DLPFC ranked first followed by HFS rTMS of the right DLPFC (Figure 3B).
Acceptability. Direct evidence suggested that no intervention was more acceptable than sham (Supplemental Figure 2). This was paralleled in the NMA model (Figure 3C). According to the SUCRAs, LFS rTMS of the right DLPFC ranked first (Figure 3D).
Study quality revealed overall some concerns for the majority of studies using the RoB2 instrument (Table 2). Examination of the funnel plots revealed broadly symmetrical distributions (Supplemental Figures 3 and 4), with 1 notable study with respect to PTSD symptoms with a large effect in favor of right DLPFC HFS.13
The overall quality of the NMA using GRADE for the PTSD symptoms outcome varied between very low quality (n = 11) to moderate quality (n = 1), largely due to small samples and high imprecision of the estimates. Accordingly, the majority of the evidence from the NMA was rated as having very low quality of evidence (Supplemental Tables 1 and 2).
Global tests of inconsistency found no statistically significant evidence of inconsistency in the PTSD symptom NMA (Supplemental Figure 5) and dropout NMA (Supplemental Figure 6). Inconsistency plots formed 2 triangular loops for each NMA, each of which produce no statistically significant evidence of inconsistency in the NMA (Supplemental Figures 5 and 6).
In this systematic review and NMA, we compare the efficacy and acceptability of rTMS from 10 RCTs including 421 civilians and veterans with DSM-IV or DSM-V diagnoses of PTSD. We compared 5 interventions: low-frequency rTMS to the right DLPFC, high-frequency rTMS to the right DLPFC, high-frequency rTMS to the left DLPFC, deep TMS, and iTBS to the right DLPFC. Our analyses indicate that only HFS and LFS of the right DLPFC separate from sham. In this population, none of the interventions separated from sham with respect to acceptability.
Previous standard meta-analyses of RCTs have supported the efficacy of rTMS in PTSD by pooling different targets and/or stimulation parameters.8-11 As this literature has increased, small studies with large effect sizes no longer dominate the literature, and our pooled effect size lies outside of the CI identified in previous meta-analyses limited to RCTs and clinician-rated PTSD symptoms.8,10 The effect size we observe is distinctly lower than meta-analyses that have pooled different methodologies.9,11 Thus, while rTMS is efficacious in PTSD, previous estimates may be inflated. Further, our analyses suggest that only HFS and LFS of the right DLPFC separates from sham; however, the limited number of studies examining other protocol variants should be highlighted. Many protocol variants that are efficacious in other conditions, such as priming rTMS,6 have not been studied in PTSD, and an RCT utilizing a bilateral HFS protocol was not retained in our analyses because it lacked a clinician-rated measure of PTSD symptoms.33 This trial compared unilateral HFS to the right DLPFC and bilateral HFS to sham and found significant improvements in self-reported PTSD symptoms in both active interventions relative to sham but no difference between active groups. An additional RCT was not retained in our analyses as we were unable to obtain data from the authors.34 Nevertheless, this RCT reported a significant effect of 1 Hz to the right DLPFC relative to sham, an effect that persisted 5 weeks after treatment.
Although our data indicate that targeting the right DLPFC is efficacious, there are several possible interpretations for the absence of efficacy or tolerability differences between stimulation targets and protocols. While it is possible that rTMS results in clinical improvements through nonspecific mechanisms, it is also increasingly evident that PTSD is a heterogeneous condition35 with heterogeneous functional circuit pathology.36 With such heterogeneity, there may be no a priori advantage to 1 stimulation site or protocol, a possibility that is becoming increasingly apparent for TMS indications with much larger literatures such as major depression.6 A personalized medicine approach, capable of targeting nodes that are specific to the individual, may resolve this clinical obstacle. Indeed, readily scalable markers, such as verbal learning, information processing, and working memory,37 have been linked to treatment resistance. Moreover, these prognostic factors may indicate a different target, the dorsomedial PFC, as a more suitable target for some.36 Additional research is required to not only optimize interventions but to identify and validate biomarkers to guide treatment selection.
Only 10 RCTs were included in the current meta-analyses, which reflects a lack of high-quality RCTs to date. This literature is predominantly characterized by small samples and therefore may be susceptible to small study effects. However, when considering smaller sample sizes, combining the NMA approach combines direct and indirect evidence to increase statistical power and exactness of treatment effect estimates.38,39
There are many sources of heterogeneity among the trials, including both clinical and methodological heterogeneity, as several studies focused on civilian or veteran populations and there are differences in the severity of PTSD symptoms between studies and differences between groups, likely resulting from small sample randomization issues. With respect to blinding integrity, the majority of studies utilized a sham coil; however, separate coils were utilized for active and sham stimulation, and in the absence of triple blinding across all studies, possible treatment delivery biases cannot be excluded. Two studies utilized an angled coil for their sham conditions, and this might have inadvertently stimulated the underlying cortex.
The majority of the data in this field utilizes a 2-week clinical trial design, but increasing evidence suggests that extending treatment courses leads to improvements in clinically relevant outcomes including hospitalization and suicide.40 Unfortunately, the number of trials available in PTSD is insufficient to resolve the optimal stimulation parameters; however, systematic study is required to establish gold standard stimulation protocols.
Some nodes with the network were not well connected, and therefore, it is possible that the lack of separation between active treatments reflects imprecise estimates. Indeed, all comparisons in the NMA were downgraded in overall quality using GRADE due to imprecision. Accordingly, the majority of evidence represented by GRADE in the NMA for PTSD symptoms was rated as having low-quality evidence with only 1 comparison representing a moderate quality. There are relatively few RCTs in this field, and additional studies are required to more robustly compare different rTMS interventions.
In our protocol registration, we aimed to also synthesize the impact of rTMS on clinician-rated depressive symptoms. While depressive symptoms were described in most of the included trials, these utilized a mix of clinician-rated and self-reported measures, and therefore, we did not proceed with analyses.
We do not find evidence for differences in clinical efficacy between rTMS interventions in PTSD, and only LFS and HFS of the right DLPFC separate from sham. Moreover, based on current data, different stimulation protocols do not differ in their acceptability. As this literature has grown, the extremely large effect sizes reported in early studies are counterbalanced by larger studies with smaller or null effect sizes, suggesting that earlier meta-analyses concluded inflated rTMS efficacy. Nevertheless, the RCT literature evaluating rTMS as a treatment for PTSD is limited and remains principally comprised of small trials. To definitely conclude clinical equipoise between rTMS interventions, additional fully powered high-quality RCTs are required to test both established and novel protocols.
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by Campus Alberta Innovates Neurostimulation Chair (AM)
Alexander McGirr, MD, PhD, FRCPC https://orcid.org/0000-0002-8425-3958
The supplemental material for this article is available online.
1 Department of Psychiatry, University of Calgary, Alberta, Canada
2 Hotchkiss Brain Institute, University of Calgary, Alberta, Canada
3 Mathison Centre for Mental Health Research and Education, Calgary, Alberta, Canada
4 Department of Clinical Neuroscience, University of Calgary, Alberta, Canada
5 Department of Psychiatry, McGill University, Montreal, Quebec, Canada
6 Douglas Mental Health University Institute, Montreal, Quebec, Canada
Corresponding Author:Alexander McGirr, MD PhD FRCPC, University of Calgary, 3280 Hospital Drive NW, TRW-4D68, Calgary, Alberta, Canada T2N 4Z6.Email: alexander.mcgirr@ucalgary.ca
