Efficacy of transcranial magnetic stimulation in anorexia nervosa: a systematic review and meta-analysis

Abstract

Purpose

Transcranial magnetic stimulation (TMS) has emerged as a promising treatment for various neuropsychiatric conditions, including depression, obsessive–compulsive disorder, and Parkinson's disease. Recent research has focused on evaluating its effectiveness in treating patients with anorexia nervosa (AN). This systematic review and meta-analysis examined the impact of TMS on patients with AN and evaluated any potential adverse effects.

Methods

We conducted search according to PRISMA guidelines and comprehensively analyzed data from multiple databases, including Pubmed, Scopus, Embase, Web of Science, and the Cochrane Library, up to September 13th. Statistical analysis utilized the Comprehensive Meta-analysis software version 3.0.

Results

The systematic review encompassed 17 studies, with nine undergoing meta-analyses. The primary target for TMS was the dorsolateral prefrontal cortex, with two studies targeting the dorsomedial prefrontal cortex, one targeting the insula and one targeting the inferior parietal lobe. The findings revealed a significant increase in body mass index (BMI) following TMS (SMD: −0.025, 95% CI: −0.0505 to −0.005, P-value = 0.045). Additionally, the Eating Disorder Examination Questionnaire (EDE-Q) score was quantitatively reported in six studies, which permitted its inclusion in the meta-analysis. The analysis exhibited a significant decrease in EDE-Q score after TMS (SMD: 0.634, 95% CI: 0.349–0.919, P-value < 0.001). Subgroup analysis based on TMS session duration indicated that the effect size of TMS on EDE-Q score is more pronounced when the session duration exceeds 20 min.

Conclusion

TMS represents an effective therapy for patients with AN, leading to improvements in both BMI and core symptoms of AN, with minor and transient side effects.

Level of evidence

Level I, systematic reviews and meta-analyses.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40519-025-01716-5.

Introduction

Anorexia nervosa (AN) is a psychiatric eating disorder (ED) that is characterized by severe malnutrition and physical comorbidities [1, 2]. Patients with AN suffer from an abnormal concern about their weight and body image regardless of their substantially reduced weight [3]. Body mass index (BMI) is a crucial variable in diagnosing patients with AN, and patients with BMI <17 kg/m² are considered AN based on the DSM-5 [4, 5].

Eating disorders (ED), particularly AN, have the highest mortality rate among all mental disorders, with a weighted mortality rate of 5.1 deaths per 1000 person-years [6]. Prolonged malnutrition in AN can result in severe complications such as osteoporosis, gastrointestinal and cardiac issues, hepatic dysfunction, hypoglycemia, and bone marrow dysfunction [7]. Common psychiatric comorbidities of AN include MDD with a prevalence of 50–70%, anxiety disorder with a prevalence of over 60%, and OCD with a prevalence of over 40% [8, 9]. Additionally, there is a 22% prevalence of suicide attempts among individuals with AN [10].

AN patients exhibit neurocognitive impairments, including compromised executive functioning, diminished central coherence, and a lack of theory of mind, which may persist following weight restoration [11]. Distinct cognitive features and personality traits in individuals with AN, such as perfectionism and obsessive personality correlated with dorsolateral prefrontal cortex (dlPFC) activation that regulates information processing through reward system [12].

The insula cortex has role in the rewarding system, in addition, it integrates the processing of the limbic system, nucleus accumbens, and striatum so it also contributes to AN through dysfunctional neural activity integration [13]. Patients with AN manifest heightened activation in the bilateral amygdala, medial prefrontal cortex (MPFC), anterior cingulate cortex (ACC), and striatum, all implicated in fear arousal and avoidance response to food cues and weight gain [14]. It has been postulated that both structural and functional changes in the insula and frontal cortex, associated with reward and anxiety processing, may play a contributory role in the genesis of eating disorders [15].

The severity of ED symptoms is assessed using the Eating Disorder Examination Questionnaire (EDE-Q) [16]. EDE-Q is a self-reporting questionnaire that evaluates four parameters: restraint, eating concern, shape concern, and weight concern in participants [17]. The most essential symptoms evaluated in EDE-Q are shape and weight overvaluation, which can be affected by AN disorder [18].

There is ongoing debate regarding the most effective treatment for patients with AN due to the higher likelihood of relapse with current treatment methods [19, 20]. Some potential treatments for AN patients include transcranial magnetic stimulation (TMS), a noninvasive form of brain stimulation, and deep brain stimulation (DBS), an invasive technique. TMS can deliver electrical impulses to specific brain areas through the scalp [21]. Single-pulse TMS is typically used for assessing brain function, while repetitive TMS (rTMS) is employed to modulate the function of different brain regions [21]. rTMS can enhance neuroplasticity and the brain's capacity to form new connections [22]. Multiple studies have indicated a reduction in depression and anorexia nervosa symptoms, as well as an improvement in BMI, in patients undergoing rTMS targeting the left DLPFC [22–24]. Studies have recently explored alternative brain regions beyond the dlPFC as potential targets for TMS, and their results have been significant [25–28]. These findings indicate the promising potential of rTMS as a noninvasive treatment for individuals with anorexia nervosa.

Aim

The main objective of this study is to conduct a systematic review of the existing literature concerning the efficacy and safety of TMS in individuals diagnosed with AN and to examine its impact on EDE-Q global score and BMI.

Research questions

1. What is the effect of TM on BMI and EDE-Q global score of patients with AN?

2. What is the impact of TMS session’s durations on BMI and EDE-Q global score among patients who had AN?

3. What is the adverse effect of TMS on patients with AN?

Methods and materials

The current systematic review and meta-analysis was conducted to investigate the effects of TMS on patients with anorexia nervosa. This study was designed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines for 2020 [29]. The study protocol was recorded in the International Prospective Register of Systematic Reviews (PROSPERO) with the registration number CRD42024586170.

Search strategy

PubMed, Scopus, Embase, Web of Science, and Cochrane Library databases were systematically searched for records by using a combination of Entrée terms, MeSH terms, and keywords such as “Anorexia Nervosa,” “Body Mass Index,” “Weight Loss,” “Transcranial Magnetic Stimulation,” “TMS,” and “Theta-Burst Stimulation,” in conjunction with appropriate Boolean operators (AND, OR). Supplementary “A” details the search strategy for each database. In addition, the references of the included studies were thoroughly examined to encompass all pertinent research.

Eligibility criteria

The study's inclusion criteria included original English articles focused on human subjects who underwent TMS for anorexia nervosa. This study encompassed cohorts, cross-sectional studies, case–control studies, case series, and randomized and non-randomized clinical trials.

On the other hand, the exclusion criteria consisted of non-English and non-human studies, studies with less than four participants, case reports, reviews, letters, editorials, comments, and conference abstracts. Additionally, studies involving participants with disorders other than anorexia nervosa or investigating forms of neurostimulation other than TMS were also excluded.

Study selection

The obtained articles were collected using Endnote software version 20. After removing duplicate articles, two researchers (S.B., R.Z.) independently screened the articles based on the title and abstract. Subsequently, the retrieved articles were screened based on the full text by the same researchers, using the previously mentioned criteria. A third researcher (A.R.B.) resolved any discrepancies.

Data extraction

Two independent researchers (S.B., T.T.) carefully extracted the data from the gathered records and systematically organized it into structured Excel documents. This information encompasses general demographic characteristics, details of the TMS protocol (including frequency, duration, and target), patients' BMI, EDE-Q global score before and after TMS, and any reported adverse events. In addition, in the case of disagreements and validation, the third author (A.R.B.) was checked the data.

Risk of bias assessments

We used the Cochrane Collaboration tool to evaluate the risk of bias. Two researchers (R.Z., T.T.) independently weighed the risk of bias in the included studies. Studies were categorized into subtypes: low risk if they mentioned risks well and clarified the study design, unknown risks if they contained certain risks by not properly clarifying study designs, and high risk if they were serious due to the biased study design. A third researcher resolved inconsistencies (A.R.B.).

Statistical analysis

In conducting our meta-analysis for the systematic review, we followed the guidelines outlined which published by the Cochrane Handbook [30]. Cochran’s Q test and I² statistic was used to determine the statistical heterogeneity. It should be mentioned that an I² value of more than 30% was considered to indicate high degree of heterogeneity. In the case of low heterogeneity, data were assessed using both fixed and random effects models to ensure robustness. The results were consistent between the two models. Also, we utilized the Egger test to investigate publication bias among the included studies. In addition, subgroup analysis was performed the evaluate the effect of TMS sessions’ duration on the BMI and the Eating Disorder Examination Questioner (EDE-Q global) score of participants. All statistical analyses were performed using Comprehensive Meta-analysis software (CMA) version 3.0.

The standardized mean difference (SMD) was calculated based on pre- and post-treatment measurements within the TMS subjects’ groups alone. While some studies included sham groups, not all did, leading us to focus on the pre- and post-TMS results of participants who received TMS treatment. This approach allowed us to include all relevant data. As well, it should be mention that a three-level meta-analysis was not adopted due to sample size limitations and the complexity it would add without yielding significantly different results.

Results

Following a comprehensive review of the databases, 2772 articles were initially gathered. Subsequent removal of duplicates resulted in 1784 studies, which underwent screening based on their titles and abstracts. Of these, 1734 articles were deemed irrelevant and were excluded, leaving 50 articles for full-text assessment. Finally, a total of 17 studies were subjected to a systematic review, and nine of them underwent meta-analysis [23, 26, 28, 31–36] (Fig. 1).

Fig. 1.

PRISMA flowchart

Baseline characteristics

The present meta-analysis consists of 129 participants with a mean age of 32.41 ± 7.27 years and a mean disease duration of 14.57 ± 6.66 years. The most common location of the TMS coil was the left dlPFC, in two studies was dorsomedial prefrontal cortex (dmPFC), in one study was inferior parietal lobe, and in one study was the insula. The mean number of TMS sessions was 18.92 ± 8.12 with a mean session duration of 24 ± 10.95 min. In addition, the patients' mean BMI increased from 15.96 ± 0.94 to 16.54 ± 1.15 during mean follow-up of 6 ± 5.32 weeks. Also, the mean score of EDE-Q global decreased from 4.60 ± 4.12 to 1.40 ± 1.25 among all subjects (Table 1). While some studies included sham groups, not all did, leading us to focus on pre- and post-treatment outcomes for TMS patients. The standardized mean difference (SMD) was calculated using only the TMS group data. Excluding non-sham-controlled studies would eliminate key research, potentially weakening the analysis.

Table 1.

Characteristic of studies

Author (year)	Country	Type of study	Sample size	Characteristic of disease	Mean age	Intervention	Number of sessions	Protocol of TMS	Outcome	Duration of follow up (week)
Hemmings et al. (2024) [40]	UK	RCT	66	Anorexia nervosa	NA	TMS (iTBS)	20	Left-DLPFC, Air-cooled coil, 4 min, 50 Hz	Baseline BMIa, EDDSb, EDE-Qc, PHQ-8d, PANASe, WSAS-Yf	4
Jaššová et al. (2019) [41]	Czech Republic	Observational study	56	Anorexia nervosa	53	TMS (rTMS)	9–24 (average 12.75)	Left-DLPFC/4 min/10 Hz	ZDSg	NA
Knyahnytska et al. (2019) [26]	Canada	Pilot study	8	Anorexia nervosa	33	TMS (dTMS)	42	Insula, H-coil, 20 min,	BMIa, YBC-EDSh, HDRSi, MADRSj, MOCAk, BAIl	6, 12, 24
McClelland et al. (2016) [23, 24]	Australia	RCT	21	Anorexia nervosa	25.29	TMS (rTMS)	NA	Left DLPFC,20 min, 10 Hz	BMIa, EDDSb, EDE-Qc, DASS-21m, VASn	NA
Muratore et al. (2021) [36]	USA	RCT	10	Anorexia nervosa	30.7	TMS (HF-rTMS)	1	Right DLPFC, blinded coils	BMIa, EDE-Qc	NA
Bartholdy et al. (2015) [38]	UK	RCT		Anorexia nervosa	NA	TMS (rTMS)	20	Left DLPFC, air-cooled coil, 40–60 min, 10 Hz	EDDSb, DASS-21m, VASn, SCIDo, FoFMp	12
Chastan et al. (2024) [28]	France	RCT	7	Anorexia nervosa	35	TMS (rTMS)	10	The right inferior parietal lobe (IPL), Air-cooled coil, 10 Hz	BMIa, BSQq	2
Ursumando et al. (2023) [42]	Italy	RCT	80	Anorexia nervosa	NA	TMS ( pp-TMS, rTMS, iTBS)	NA	NA	EAT-26r, BUTs, CBCL 6–18)t, (MASC-2)u, CDI-2v	NA
Eynde et al. (2013) [37]	UK	Pilot study	9	Anorexia nervosa	25	TMS (rTMS)	20	Left DLPFC, figure-eight coil, 10 Hz	BMIa, EDE-Qc, DASS-21m	NA
Woodside et al. (2017) [25]	Canada	Case series	14	Anorexia nervosa	39.80	TMS (rTMS)	20–30	DMPFC, 10 Hz	BMIa, EDE-Qc, DERSw, PCL-Cx	NA
Dalton et al. (2022) [39]	UK	Qualitative study	29	Anorexia nervosa	30.1	TMS (rTMS)	<17	DLPFC, 20 min, 10 Hz	NA	12
Dalton et al. (2018) [31]	UK	RCT	30	Anorexia nervosa	29.74	TMS (rTMS)	20	DLPFC, 30–60 min, 10 Hz	BMIa, EDE-Qc, OCI-Ry, POMS total mood disturbancez	4, 16
Dalton et al. (2020) [32, 33]	UK	RCT	26	Anorexia nervosa	30.2	TMS (rTMS)	20	DLPFC, 30–60 min, 10 Hz	BMIa, EDE-Qc, food choice questionaire	16
Dalton et al. (2020) [32]	UK	RCT	24	Anorexia nervosa	NA	TMS (rTMS)	20	DLPFC, 30–60 min, 10 Hz	BMIa, EDE-Qc	72
Dalton et al. (2021) [35]	UK	Pilot study	24	Anorexia nervosa	28.79	TMS (rTMS)	20	DLPFC, 30–60 min, 10 Hz	Baseline BMIa, EDE-Qc	4
Dalton et al. (2021) [35]	UK	RCT	26	Anorexia nervosa	29.56	TMS (rTMS)	20	DLPFC, 30–60 min, 10 Hz	BMIa, EDE-Qc	16

^aBMI body mass index

^bEDDS eating disorder diagnostic scale

^cEDE-Q eating disorder examination questionnaire

^dPHQ-8 patient health questionnaire‐8

^ePositive and negative affect schedule

^fWork and social adjustment scale

^gZung’s self-rating depression scale

^hYale-Brown-Cornell eating disorder scale

ⁱHamilton depression rating scale

^jMontgomery–Asberg depression rating scale (MADRS)

^kMontreal cognitive assessment scale

^lBeck anxiety inventory

^mThe depression anxiety and stress scales

ⁿVisual analogue scales

^oStructured clinical interview for DSM disorders

^pThe fear of food measure (FoFM)

^qThe body shape questionnaire

^rEating attitudes test

^sBody uneasiness test (BUT)

^tChild behavior checklist for ages 6–18

^uMultidimensional anxiety scale for children—second edition

^vChildren’s depression inventory—second edition

^wHe diffi-culties with emotional regulation scale

^xPTSD checklist-civilian version

^yrevised obsessive–compulsive inventory

^zProfile of mood states

Statistical analysis

BMI

We finally assessed 9 studies to determine the BMI changes after TMS. The forest plot illustrated that the BMI of patients significantly increased after TMS (SMD = −0.255, 95% CI: −0.505 to −0.005, and P-value = 0.045) (Fig. 2) without any heterogeneity (Q = 3.234, P-value = 0.919; I² = 0%). Moreover, the result of Egger’s test depicted no publication bias (P-value = 0.272) (Fig. 3). Additionally, we conducted a subgroup analysis to illustrate the impact of TMS duration (less and more than 20 min) on the BMI of participants. The between-group comparison indicated no significant differences between the two subgroups (Q = 0.031, P-value = 0.860).

Fig. 2.

Meta-analysis of the impact TMS on BMI among included studies

Fig. 3.

The funnel plot of TMS effects on BMI

EDE-Q

We ran another meta-analysis to assess the impact of TMS on the EDE-Q global score of subjects. Only six studies provided sufficient statistical data for inclusion in this meta-analysis. The other studies mentioned results qualitatively without providing detailed statistical data or numerical values, making them ineligible for inclusion. The forest plot illustrated that among subjects who performed TMS the post-stimulation EDE-Q global score was significantly diminished (SMD: 0.634, 95% CI; 0.349 to 0.919, and P-value < 0.001) without any heterogeneity (Q = 2.559, P-value = 0.768; I2 = 0%) (Fig. 4). Moreover, owing to Eggers’ test and asymmetric funnel plot, publication bias was not detected (P-value = 0.975) (Fig. 5).

Fig. 4.

The forest plot of TMS on EDE-Q global among included studies

Fig. 5.

The funnel plot depicted the impact of TMS on EDE-Q globa

TMS duration

Subgroup analysis was performed to determine the impact of TMS session duration (less and more than 20 min) on the EDE-Q global score of subjects. The mixed effect model showed, the duration of TMS in both subgroups significantly diminished the EDE-Q global score (less than 20 min; SMD: 0.726, 95% CI: 0.323–1.128, and P-value <0.001, and longer than 20 min: SMD = 0.542, 95% CI: 0.138 to 0.945, and P-value = 0.009). As well, owing to pooled effect size of 0.634 that indicated by mixed-effect model, the subjects who stimulated longer than 20 min have a significant better outcome.

Adverse events

In our systematic review just two studies reported adverse events [28, 37]. One study found that five patients experienced undernutrition and health deterioration in both sham and active rTMS conditions, with two in the rTMS group attempting suicide [28]. Another study reported significant discomfort for all participants during stimulation, which lessened over time; one patient withdrew due to severe pain from tapping sensations [37].

Quality assessment

We assessed the studies’ risk of bias and quality using the non-randomized studies of interventions (ROBINS-I) tool. Bias in the confounding aspects of the study (D1), bias in classification of intervention (D3), and bias owing to deviations from intended interventions (D4) were reported low among all studies [23, 25–28, 31–42]. The high and middle level of selection bias (D2) was detected among six [37–42] and two [25, 34] studies respectively. As well, three studies had middle [25, 34, 41] and five studies had high degree [37–40, 42] of measurement of outcome bias (D6), and reporting results bias (D7). On the other hand, 47% of studies had a high bias regarding missing data (D5) [25, 34, 37–42]. Finally, the overall quality assessments illustrated that the studies had a low risk of bias (Figs. 6, 7).

Fig. 6.

Risk of bias of included study. Based on authors’ judgment

Fig. 7.

Risk of bias assessment based on subscales for all included studies based on authors’ judgment

Discussion

TMS targets

TMS has been recently adopted as a treatment for AN, with the dlPFC being the primary target. Out of the 17 studies reviewed regarding the effects of TMS on patients with AN, 14 targeted the dlPFC, while two targeted the dmPFC, one targeted insula, and one targeted the inferior parietal lobule (IPL).

Dorsolateral prefrontal cortex

The primary target of TMS in AN is the dlPFC due to its established role in self-regulation and self-control behavior [43, 44] and AN is often characterized as a disorder of excessive self-control [32].

The dlPFC is interconnected with limbic regions such as the amygdala and insula. Dysregulation in this frontolimbic circuit, linked to habit and reward, are implicated in developing AN and restricted food intake [12, 45–47]. dlPFC-rTMS can modulate this circuit through top-down control, leading to improvements in symptoms of AN [43].

In patients with AN dietary food restriction may be more psychologically rewarding than the food itself, as prolonged food restriction can develop into a habitual behavior over time [48, 49]. Excitation of the dlPFC by high-frequency rTMS enhances cognitive control, allowing individuals to override maladaptive food restrictive choices by favoring low-value rewards that disrupt maladaptive habitual behaviors [32].

Furthermore, rTMS affects the modulatory impact of the dlPFC on amygdala function, which is linked to anxiety and negative emotional states [50]. Alleviation of negative mood contributes to a decreased inclination to restrict food consumption [32]. however, there are reports of no significant enhancement in mood among patients who underwent dlPFC-rTMS compared to the Sham group immediately post-treatment and during the subsequent observation period [38].

The TIARA study, a notable randomized clinical trial by Dalton and colleagues 34 patients with a history of over three years AN were randomly assigned to receive either sham or dlPFC-rTMS. The results were encouraging, demonstrating enhancements in participants' BMI, as well as improvements in eating disorder symptoms, including anxiety, feelings of fatness, mood, and quality of life, which were sustained in the long term [31]. It was also noted that in another study by Dalton et al. dlPFC-rTMS not have a significant impact on body image disturbance or BMI [32].

Dorsomedial prefrontal cortex

The dmPFC is the other target of TMS due to its involvement in regulating emotional responses and impulse control [51]. Studies have shown that targeting the dmPFC with rTMS (dmPFC-rTMS) has resulted in significant improvement in symptoms of conditions such as bulimia nervosa [52], obsessive–compulsive disorder [51], and PTSD [25]. These improvements are associated with changes in brain circuitry function as observed in resting-state fMRI studies [27].

In a study conducted by Woodside et al., 19 subjects diagnosed with AN received 20–30 sessions of dmPFC-rTMS to address their comorbid MDD or PTSD. Post-intervention, a significant amelioration was noted in the EDE-Q, anxiety, with a moderate improvement in shape and weight concerns as well as depression [27]. the dmPFC is implicated in a "salience network" encompassing regions related to cognitive and impulse control [53]. Notably, a lower predilection for resting state functional connectivity from the dmPFC to the right frontal pole and left angular gyrus prior to treatment demonstrated a significant correlation with improvements in core symptoms of AN [27].

Insula

fMRI findings have indicated abnormal insula activity in emaciated and malnourished individuals when exposed to images of food [26, 54–56]. The insula integrates visual and somatic perception with emotions, feelings, and higher cognitive processes [55]. In a study conducted by Knyahnytska et al., deep TMS (dTMS) targeted at the insula in eight patients with severe enduring AN (SE-AN) resulted in a reduction of AN-related obsessive and compulsive behaviors as well as depression and anxiety, with no adverse effects and no impact on cognition. Although these findings must be interpreted cautiously due to the limited sample size of the study [26].

Inferior parietal lobe

IPL plays a role in self-other discrimination and self-perception processes [57]. In a study conducted by Chastan et al., 17 patients with AN, with an average age of 35 years, were randomly allocated to receive either real or sham high-frequency rTMS over the IPL for two weeks [28]. The findings indicated no significant variance in body shape perception or other related parameters. Considering the complex disruption in the frontoparietal neural network in patients with AN, targeting the PFC may have a more substantial impact on IPL activity modulation in addition to BMI, and behavior of AN [58].

TMS protocol

The most common TMS protocol used in studies was rTMS, with a frequency of 10 Hz and each session duration of 30–60 min for 20 sessions. One study used intermittent theta burst stimulation (iTBS) for patients with AN. iTBS takes less than four minutes, and when applied intermittently, it would be a facilitator for cortical activity; it requires a lower stimulation intensity and has a long-lasting effect on brain plasticity [59] iTBS was not inferior to rTMS in treating MDD in adults [60]. However, iTBS is safe and tolerable for children [61], but its discomfort during stimulation would be a barrier to its usage [40].

In a single study, researchers utilized deep TMS with an H-coil to target the insula [26]. The study involved 42 sessions of TMS, each lasting 20 min per day, 5 days a week, over 6 weeks. The only reported side effects were occasional mild headaches and discomfort associated with deep TMS, with no adverse events or changes in cognition [26].

Owing to the limited number of studies, small sample size, and the variation in TMS protocols, it was not feasible to compare the efficacy of different protocols with each other.

TMS session duration

In our study, we did a subgroup analysis based on the duration of each TMS session. The studies included in our meta-analysis employed varied TMS protocols with stimulation durations of 4, 20, 30–60, and 40–60 min. We categorized the duration as less than or more than 20 min. This cutoff is supported by several studies [62, 63], which suggest that 20 min is a common threshold used in TMS research to differentiate between short and long stimulation sessions. The results indicated that patients who underwent rTMS sessions lasting more than 20 min experienced better outcomes [31–35, 38, 64]. However, there is a lack of RCT or cohort studies encompassing patients from both groups to compare the efficacy of rTMS.

Comorbidities

Depression often manifests early on, potentially preceding the manifestation of AN symptoms, and is associated with unfavorable outcomes [65, 66]. Addressing depressive symptoms is considered a crucial approach in the treatment of severe and enduring AN [67]. Given the low responsiveness to antidepressant medication in individuals with AN [5], the antidepressant effects of rTMS present as a viable option for severely enduring patients with AN [33]. In this systematic review, two studies documented PTSD in AN patients who underwent rTMS [25, 27], and two studies indicated MDD as a comorbidity in AN patients [35, 40], with rTMS demonstrating beneficial outcomes in all cases.

In a study by Dalton et al., 26 AN patients underwent dlPFC-rTMS. Of these patients, 16 received concomitant antidepressant treatment, and this subgroup exhibited significantly improved ED symptoms compared to those who did not receive antidepressants [35]. Interestingly, the study noted that while antidepressant usage was associated with enhanced ED outcomes, it did not significantly affect mood following rTMS treatment [35]. The precise mechanism underlying the association between antidepressant usage and improved ED outcomes following rTMS treatment remains unclear. Some studies suggest that antidepressant usage may enhance rTMS efficacy during therapy for MDD [68], while others do not support this synergistic effect [69]. It is hypothesized that this synergistic effect could be attributed to the modulation of neural plasticity and cortical excitability [70, 71].

Adverse events

In one study targeting the IPL for rTMS, headache and eyelid twitching were observed in the sham group, unrelated to rTMS. However, in this study, serious adverse events, including undernutrition and deteriorated general health, were seen in five patients across both sham and real rTMS groups, leading to the hospitalization of two patients. Additionally, two patients who underwent rTMS experienced suicidal attempts and did not show improvement in mood and anxiety. These findings suggest that IPL-rTMS may not effectively improve psychiatric comorbidities [28]. In another pilot study involving dlPFC-TMS at 10 Hz for 20 min, all patients reported significant discomfort during the stimulation. This discomfort gradually decreased during the rTMS sessions. One patient discontinued the study due to intense and painful tapping sensations from rTMS, which were not experienced in the sham group [37].

Eating disorder examination questionnaire/body mass index (EDE-Q/BMI)

In all studies that reported EDE-Q pre and post TMS, the questionnaire score diminished following TMS. All studies that evaluated participants' BMI before and after TMS reported that the BMI of AN patients increased following TMS except in one study, where despite a decrease in the EDE-Q score of patients with AN following TMS, the BMI of participants did not change. It is possible that the higher BMI results from changes in eating habits and behavior, which may necessitate longer follow-up periods to observe the effects [32].

Strength and limits

This is the first systematic review and meta-analysis evaluating the efficacy of different brain targets which stimulated by TMS on BMI and EDE-Q global score of patients with AN. This systematic review acknowledges several limitations; one is including observational and pilot studies alongside RCTs. Some studies included sham groups, but many did not, so we focused on pre-and post-TMS results. While this approach may not control for all confounding factors, it allowed us to analyze the relevant data comprehensively. Additionally, the number of studies with long-term follow-up was limited, particularly for brain targets other than dlPFC. Various TMS protocols used across different studies and no presence of comparative studies led to a lack of consensus on the most effective method. Moreover, the presence of comorbidities such as PTSD and depression in patients with AN was noted in some studies. Still, there is a lack of research comparing the efficacy of TMS in patients with or without comorbidities. Furthermore, there is no study to evaluate the effectiveness of TMS in AN patients based on race, sex, and age. Future large-scale prospective cohort studies or RCTs, designed uniformly with respect to patient characteristics (including comorbidities, age, and sex) as well as TMS protocols (encompassing target, duration, and pulse), complemented by appropriate control groups and extended follow-up periods, are necessary to further substantiate our findings.

Conclusion

The results of this systematic review and meta-analysis establish the efficacy of TMS as a treatment for patients with AN. TMS has been shown to raise patients' BMI by alleviating core symptoms of AN. While most studies have focused on the dlPFC as the target for TMS, further investigation into other potential TMS targets is necessary. Additionally, long-term cohort studies are needed to identify the optimal TMS target and stimulation protocol.

What is already known on this subject?

AN is a severe eating disorder involving extreme food restriction and distorted body image, with traditional treatments often showing limited success. TMS has been explored as a treatment option, targeting brain regions, especially dlPFC, which are involved in cognitive control and reward processing. Preliminary studies suggest that TMS may improve core symptoms and body mass index (BMI) in patients with AN.

What your study adds?

The current study provides the most comprehensive review and meta-analysis to date, highlighting the positive effects of TMS on increasing BMI and reducing core symptoms (which evaluated by EDE-Q global) of AN. It also identifies specific brain regions targeted by TMS, particularly the dorsolateral prefrontal cortex, and explores the duration of sessions, offering insights into optimal treatment protocols for improved patient outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

The authors confirm contribution to the paper as follows: study conception and design: S.R., A.R.B., and A.D.; search strategy: S.B., and A.R.B.; data collection: S.B., R.Z., and T.T.; analysis and interpretation of results: A.R.B., and A.D.; preparation of demographic table: T.T., and A.R.B., and preparation of quality assessment figures: A.D.; draft manuscript preparation: A.R.B., P.J., R.Z., and S.B. editing the manuscript and prepared the final version of manuscript: S.R., A.T., and S.S. All authors reviewed the results and approved the final version of the manuscript.

Funding

Not applicable.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.