Therapeutic Potential of Repetitive Transcranial Magnetic Stimulation (TMS) in Long COVID: A Systematic Review and Meta-Analysis Protocol

Abstract

The cumulative global incidence of long COVID is around 400 million individuals, yet effective treatment options remain limited. A recent systematic review published in 2025 highlighted promising results for non-invasive brain stimulation in alleviating long COVID symptoms. Given the growing use of repetitive transcranial magnetic stimulation (rTMS) for people with long COVID, a focused meta-analysis on the therapeutic effectiveness of rTMS is warranted. To address this gap, this protocol outlines the planned procedures for a systematic review and meta-analysis. A comprehensive search will be conducted across CINAHL Ultimate, MEDLINE, ScienceDirect, and Scopus. Retrieved studies will be managed using Rayyan, with two independent reviewers screening titles and abstracts, followed by full-text review. Data extraction will follow PRISMA and Cochrane guidelines using a standardised form, with dual independent extraction and reconciliation of discrepancies. Risk of bias will be assessed using Cochrane RoB 2.0. Meta-analytical procedures will include calculation of standardised effect sizes (e.g., Hedges’ g), use of random-effects models, and assessment of heterogeneity via I², Cochran’s Q, and tau². Subgroup and moderator analyses will explore variations in rTMS protocols, participant characteristics, and symptom domains. Sensitivity analyses and meta-regression will be conducted where data permit. Results will be visualised using forest and funnel plots, and the GRADE framework will be used to assess the quality of evidence.

1. Introduction

Al-Aly et al. [1] estimated a cumulative global incidence of long COVID of approximately 400 million. Given this estimation was up until the end of 2023, it is possible this number will be even greater by the end of 2025. Due to considerable variability in how long COVID is defined across the literature, this protocol adopts the NICE guidelines’ definition, which characterises long COVID as the presence of a range of debilitating symptoms lasting from 4 weeks to over 12 weeks following acute COVID-19 infection [2]. A range of terms have been used to describe the prolonged symptoms following acute COVID-19 infection, including “long COVID”, “post-COVID condition”, “post-acute sequelae of SARS-CoV-2 infection” (PASC), “long-haul COVID”, and “COVID-19 recovery syndrome” [3,4,5]. These terms reflect the evolving understanding of the condition and its diverse symptomatology. While some clinical and research communities have adopted more technical nomenclature such as PASC, the term “long COVID” remains the most widely recognised and commonly used in public discourse, media, and patient advocacy. Importantly, our Public and Patient Involvement (PPI) group consistently uses and identifies with the term “long COVID”, citing its clarity, accessibility, and resonance with lived experience. Therefore, in this work, we adopt “long COVID” as the preferred terminology to align with patient perspectives and ensure relevance to the communities most affected. Long COVID is characterised by a wide range of symptoms, and over 100 have been identified, including muscle pain, severe fatigue, post-exertional malaise, sleep disturbances, breathlessness, and neurological and cognitive impairments, with significant symptom overlap observed with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS), although not all manifestations of long COVID clearly overlap with ME/CFS [6,7,8,9,10,11,12]. In one study, musculoskeletal pain has been reported as the most prevalent and persistent symptom for those living with long COVID [7]. In a review synthesising 21 studies, Sandler et al. (2021) reported approximately 46% of individuals with long COVID reported fatigue as a pertinent symptom, often lasting weeks or months [8]. In addition, in one study 59% of those with long COVID also met the criteria for post-exertional malaise, defined as a prolonged worsening of symptoms after minimal physical or mental activity; in another study approximately 65% reported poor sleep quality, in another study breathlessness was another frequently reported symptom, and in another study 40% reported neurological abnormalities [9,10,11,12]. Several studies have also reported well-being outcomes linked to long COVID, including diminished quality of life and higher levels of anxiety and depression [13,14,15]. Health-related quality of life has been shown to be significantly lower in individuals with a history of COVID-19 infection [13,14]. Symptoms of anxiety and depression are also prevalent and have been observed as early as three months post-infection [15]. However, findings remain inconsistent, with some studies reporting elevated psychological distress, while others indicate variability depending on symptom severity and duration [15]. Despite the global burden and complex symptomatology of long COVID, effective treatment options remain scarce.

A recent systematic review by Markser et al. [16] synthesised current evidence on the efficacy of non-invasive and minimally invasive brain stimulation techniques including rTMS for alleviating symptoms associated with long COVID and reported that existing studies show promising initial results with improvement in clinical outcome measures. Despite the multi-level symptomology associated with long COVID, several symptoms appear to be of a neurological nature (e.g., depression, anxiety, psychological distress, sleep disturbance, and cognitive impairments). Such manifestations, whilst not representing the complete symptomology of long COVID, may present as good candidates for treatment and/or management though non-invasive neuromodulation techniques, which have been shown to be effective in conditions with similar neurological symptoms [17]. With the increasing use of non-invasive brain stimulation techniques such as rTMS in this population [18], there is growing interest in its potential as a therapeutic tool for managing long COVID symptoms. Transcranial magnetic stimulation (TMS) is a non-invasive brain stimulation technique that uses a coil placed on the scalp to deliver magnetic pulses. Through electromagnetic induction, these pulses generate a magnetic field that induces an electrical current in the cortical tissue beneath the coil [19]. The effects of TMS can be either acute or prolonged, depending on stimulation parameters such as intensity, coil shape and orientation, and the frequency and pattern of pulses. Single-pulse TMS is primarily used to investigate brain function. For instance, delivering a single pulse over the primary motor cortex (M1) can evoke motor responses (motor evoked potentials; MEPs) in target muscles, which are measured using electromyography [20]. The amplitude and latency of these MEPs provide insights into motor cortex excitability [21]. In contrast, repetitive TMS (rTMS), which is the main focus of this review protocol, can modulate neuronal activity in a way that produces effects lasting beyond the stimulation period [22]. The effects of rTMS on neural activity depend on the frequency and pattern of stimulation, with certain protocols producing inhibitory effects and others excitatory. Repeated sessions have been explored as treatments for various psychiatric and neurological disorders, owing to their potential to induce long-lasting changes in neural plasticity [23,24]. Moreover, rTMS is currently being investigated for use in other post viral conditions such as ME/CFS to treat symptoms [25,26,27,28], with some encouraging preliminary findings [28].

Whilst rTMS is the most common neuromodulation technique with established clinical applications and approval from health and social care government bodies (e.g., the National Institute for Health and Care Excellence [NICE] approves its use in treatment-resistant depression [29] and has published guidelines for use in obsessive–compulsive disorder [OCD] and migraine treatment), there are alternative brain stimulation techniques. For example, transcranial direct current stimulation (tDCS) has been explored as a treatment option for depression, yet, whilst NICE has acknowledged that evidence exists to suggest tDCS is effective, they recommend that the procedure should only be used with special arrangements for clinical governance, consent, audit, and research. Unlike rTMS, which uses magnetic fields to modulate neuronal activity and causes neurons to fire, tDCS uses a weak electric current which modulates the membrane potential of neurons, making them more or less likely to fire [30]. Research has applied tDCS to long COVID, with specific focus on managing fatigue symptoms, but results have been inconsistent, and researchers suggest that the mechanism of tDCS may be insufficient to treat these conditions [31]. The effect of tDCS is both more subtle and less focal than rTMS, raising additional challenges [32]. Due to a relativity small number of clinically relevant research studies, the specific parameters required for effective application of tDCS in clinical settings are currently unclear [32]. As such, the focus of this review will be rTMS.

Given current evidence on the efficacy of brain stimulation techniques, including rTMS in individuals with long COVID on clinical outcomes [16], and the growing application of rTMS in individuals with long COVID [18] and related conditions [25,26,27,28], a focused meta-analysis on the therapeutic effectiveness of rTMS is warranted. There is also need for enhanced review of safety implications and guidelines for use of non-invasive brain stimulation techniques for long COVID. Paying specific attention to factors unique to long COVID which may interact with rTMS, for example, patients use of medications, is necessary to alleviate certain symptoms which may present as contraindications for TMS [32].

This protocol paper will be the first step in conducting a systematic review and meta-analysis with the aim to assess the effectiveness and suitability of rTMS for treatment of long COVID symptoms.

1.1. Aims of Meta-Analysis

• To evaluate the therapeutic effectiveness of rTMS in the treatment of long COVID symptoms.
• To synthesise and assess clinical outcomes reported in existing studies using rTMS interventions for long COVID.
• To identify patterns in treatment response, including symptom domains most affected (e.g., fatigue, cognitive impairment, mood disturbances).
• To explore potential moderating factors (e.g., stimulation parameters, number of sessions, target brain regions) that may influence treatment outcomes.

Research questions for the future systematic review and meta-analysis:

The primary research question of the future systematic review and meta-analysis will be as follows: What is the effect of rTMS on symptoms of long COVID?

Secondary research questions are designed to align with planned subgroup analyses and include the following:

1. Are there differences in the effectiveness of rTMS based on stimulation parameters (e.g., frequency, intensity, targeted brain region)?

2. Do participant characteristics (e.g., age, sex, duration since acute COVID-19 infection) moderate effectiveness of rTMS?

3. What is the risk of bias and methodological quality of studies investigating rTMS for long COVID?

We will approach the proposed systematic review and meta-analysis with the hypothesis that rTMS may offer therapeutic benefits for individuals experiencing persistent symptoms of long COVID, particularly in domains such as fatigue, depression, and cognitive dysfunction, which are commonly targeted in other neuropsychiatric conditions treated with rTMS. We also expect that the evidence base will be limited but emerging, given the recency of long COVID as a clinical entity and the time required to conduct and publish controlled trials, and most studies will be small-scale or pilot RCTs, possibly with methodological variability in study design, outcome measures, and stimulation parameters. Moreover, we expect that symptom improvements may vary by domain, with stronger effects hypothesised for mood-related symptoms (e.g., depression) than for more complex or diffuse symptoms like fatigue or brain fog. Given the novelty and evolving definition of long COVID, as already mentioned above we anticipate some heterogeneity in how populations are defined across studies, which may influence both inclusion decisions and the strength of conclusions.

1.2. Eligibility Criteria

• Participants

The proposed systematic review and meta-analysis will investigate interventions involving rTMS in individuals diagnosed with long COVID or other terminologies used for long COVID, without imposing any demographic restrictions. We acknowledge the variability in definitions across the literature and will accommodate this by clearly documenting and analysing the definitions used in each included study. This will allow us to account for potential heterogeneity in symptom classification and diagnostic criteria.

We will exclude articles that (a) use TMS for purposes other than therapeutic intervention, (b) do not involve rTMS, (c) focus on populations other than those with long COVID, and (d) are review articles or case studies. We will include grey literature alongside peer-reviewed sources. This includes preprints (e.g., PsyArXiv, arXiv), dissertations (via ProQuest), conference proceedings (e.g., Web of Science), and institutional repositories such as DSpace. Furthermore, we will supplement database searches by reviewing reference lists (forward citation searching) and identifying articles that cite included studies (backward citation searching) using tools such as CrossRef and Google Scholar. We will also use CoCites, if available, to find articles with similar citation patterns.

• Types of Sources

The proposed systematic review and meta-analysis will consider articles that present quantitative data. Articles must present at least pilot data (including case series of rTMS with n ≥ 5). Review articles and case studies should not be included.

2. Methods

The proposed systematic review and meta-analysis will be conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and checklist (Supplementary Material File S1). This protocol has been conducted in accordance with the PRISMA-protocol (PRISMA-P) guidelines and checklist (Supplementary Material File S2). The protocol has been registered with Open Science Framework: https://doi.org/10.17605/OSF.IO/RE235. Any deviation from the protocol will be clearly logged with dates, reasons, and impacts noted.

2.1. Search Strategy

Our search strategy was designed to achieve a balance between comprehensive coverage of the relevant literature and practical constraints while maintaining a high level of scientific rigour. We selected four databases—CINAHL Ultimate, MEDLINE, ScienceDirect, and Scopus—based on their relevance to the interdisciplinary nature of our topic, which spans psychology, health sciences, and behavioural research. CINAHL Ultimate was chosen for its coverage of nursing and allied health literature. MEDLINE (accessed via EBSCOhost) provides biomedical literature, including psychiatry and clinical psychology studies. ScienceDirect was included due to its full-text access to journals in psychology, neuroscience, and related disciplines. Scopus was selected for its comprehensive indexing of interdisciplinary research and citation tracking capabilities, allowing us to identify additional relevant studies. We will search grey literature alongside peer-reviewed sources. This includes preprints (e.g., PsyArXiv, arXiv), dissertations (via ProQuest, Michigan, USA), conference proceedings (e.g., Web of Science), and institutional repositories like DSpace. Finally, we will search trial registries (e.g., ClinicalTrials.gov, WHO ICTRP) to capture unpublished or ongoing studies, as per guidance [33].

Moreover, we will supplement database searches by reviewing reference lists, identifying citing articles via tools such as CrossRef and Google Scholar, and using CoCites to find articles with similar citation patterns. We will validate our search by testing it against known relevant studies to ensure key articles are captured. If too many irrelevant results appear, we will refine the search terms or criteria. This iterative process will continue until the strategy is optimised for relevance and accuracy. We will contact study authors via a standardised email when additional details are needed for eligibility or analysis. Follow-up emails will be sent after two and an additional two weeks if there is no response. All contact attempts and responses will be documented. To ensure transparency, we will report how many authors were contacted, how many responded, and how many provided the requested data. These details will appear in the main manuscript and in a Supplementary Table summarising the information requested and obtained. We do not plan a living review but will repeat database searches before submission if over six months have passed since the original search.

2.2. Query Strings

TI ((long COVID OR post-COVID OR “post-acute sequelae of SARS-CoV-2 infection” OR “post-viral fatigue” OR “COVID-19 recovery syndrome” OR “long-haul COVID” OR “long COVID syndrome”)) OR AB ((long COVID OR post-COVID OR “post-acute sequelae of SARS-CoV-2 infection” OR “post-viral fatigue” OR “COVID-19 recovery syndrome” OR “long-haul COVID” OR “long COVID syndrome”))

The proposed systematic review and meta-analysis will focus on clinical and functional outcomes related to long COVID symptoms following Repetitive Transcranial Magnetic Stimulation (rTMS). Key outcomes include fatigue, cognitive function, mood/psychological symptoms, sleep disturbances, quality of life, and overall symptom improvement. We will also gather data on adverse effects, rTMS tolerability, and treatment durability. The independent variable is rTMS as a treatment for long COVID, particularly fatigue, cognitive dysfunction, depression, and anxiety. Studies with various rTMS protocols will be included, such as high-frequency rTMS (≥10 Hz), low-frequency rTMS (1 Hz), theta burst stimulation (TBS), and other repetitive TMS (rTMS) paradigms, with different stimulation targets (e.g., dorsolateral prefrontal cortex) and treatment durations. Comparator groups may include sham rTMS, standard care, or no-treatment controls. We will also analyse variables that may influence treatment outcomes, including participant characteristics (age, sex, time since infection, COVID-19 severity, comorbid conditions), study-level factors (sample size, design, risk of bias, study setting), rTMS protocol variables (site, frequency, intensity, session number), and follow-up duration.

2.3. Study/Source of Evidence Selection

After the search, all citations will be uploaded and have their duplicates removed in Rayyan [34]. The screening process will consist of two rounds: (1) title and abstract screening and (2) full-text screening (Supplementary Material File S3). In both rounds, two independent reviewers will screen each record using pre-specified inclusion and exclusion criteria developed a priori. Blinding of reviewers during screening will be implemented to the extent possible using Rayyan, which allows for independent decisions without visibility into the other reviewer’s judgments. This helps reduce bias and increases objectivity in the initial phases of study selection. Conflicts between reviewers will be flagged automatically by Rayyan and subsequently resolved through discussion with two other reviewers. If consensus cannot be reached, a third reviewer (independent chair) will adjudicate. To assure consistency and transparency, all reviewers will undergo a calibration exercise using a small sample of studies before formal screening begins. At both the titles and abstracts stage and the full text stage, inter-rater reliability will be calculated and expressed via Cohen’s Kappa statistic; scores range from –1 to 1 with scores closer to 1 indicating stronger agreement. A value of >0.6 will be considered as acceptable inter-rater reliability, as Landis and Koch identified this value as ‘substantial agreement’ [35]. This exercise will help refine the application of the inclusion/exclusion criteria and ensure a shared understanding of borderline cases. For the proposed systematic review and meta-analysis, we will share the full list of sources from the database searches and screening decisions by individual screeners. Bibliographic data (titles, abstracts, metadata) will be exported in RIS and CSV formats, while screening decisions will be provided in a separate CSV/XLSX file. All files will be uploaded to an open-access repository, like OSF, upon manuscript submission or acceptance.

2.4. Data Extraction

The data extraction process will follow PRISMA guidelines and Cochrane Handbook procedures, using a standardised form available in the OSF project (Data_Extraction_Form_TMS_LongCOVID.xlsx). It will be conducted in four stages:

Training and Calibration: A small sample of studies (5–10%) will be independently extracted by all reviewers using the draft form. Discrepancies will be discussed to refine the protocol.

Primary Extraction: Two reviewers will independently extract key data (e.g., means, standard deviations [SDs], rTMS parameters, demographics, outcomes) using the finalised form.

Risk of Bias Assessment: Two reviewers will independently assess study-level risk of bias using tools such as the Cochrane RoB 2.0. Discrepancies will be resolved through discussion.

Reconciliation and Verification: All discrepancies will be resolved through discussion or adjudication by a third reviewer. Final data will be entered into the meta-analysis dataset. Optional AI/computer-assisted tools may support text identification and bibliographic metadata extraction, but all outputs will be verified by human reviewers. All stages will be human led, with AI support supervised accordingly. Only data related to long COVID, as defined by study authors, will be extracted. Missing data will be marked as “NR” and flagged for follow-up.

3. Results

3.1. Extracted Data Includes

Study Information: Author(s), year, title, journal, country, institution, study design (e.g., RCT, quasi-experimental), and setting.

Participant Characteristics: Mean age, age range, sex/gender distribution, time since acute COVID-19, severity of initial infection, and comorbidities.

Intervention Details: rTMS protocol type, stimulation site, frequency, intensity, session number, neuronavigation use, treatment schedule, and consideration of safety parameters.

Comparators: Control conditions (e.g., sham rTMS, standard care, waitlist).

Outcomes: Clinical measures (e.g., fatigue, cognition, mood, sleep, quality of life), statistical data (e.g., means, SDs, sample sizes, p-values, effect sizes), follow-up data, and adverse effects.

Risk of Bias Domains: Randomisation, blinding, attrition, and selective reporting.

Ambiguous data will be flagged for discussion, with clarifications noted in comments. No imputation of missing data will occur unless explicitly instructed. All numerical entries will be double-checked, and a second reviewer will verify the data. Completed forms will be saved and uploaded to the OSF folder using the following naming convention: StudyID_ExtractorInitials_Extraction.xlsx.

Each round of data extraction will be conducted by two independent reviewers working in parallel. They will extract study characteristics, rTMS intervention details, participant data, clinical outcomes, and risk of bias assessments. Reviewers will receive standardised training and work independently, with discrepancies resolved through discussion. If consensus cannot be reached, a third reviewer will adjudicate.

During training, inter-rater reliability will be assessed using Cohen’s kappa for categorical variables and intraclass correlation coefficients (ICCs) for continuous variables, based on a subset of studies. These results will guide refinements to the extraction form and training procedures.

In each round, reviewers will compare entries side by side, with differences automatically highlighted. Discrepancies will be discussed using source material and extraction guidelines. Unresolved disagreements will be adjudicated by a senior reviewer, with decisions recorded in a reconciliation log. All discrepancies and resolutions will be documented.

Systematic issues will be analysed to inform updates to the extraction protocol. Once reconciled, data will be finalised and added to the meta-analysis dataset.

3.2. Data Analysis and Presentation

The proposed meta-analysis will focus on estimating the overall effect of rTMS on long COVID symptoms, such as fatigue, cognitive dysfunction, and mood disturbances. Effect sizes from continuous outcomes will be standardised (e.g., Hedges’ g or standardised mean difference), and odds ratios will be used for dichotomous outcomes when applicable. A random-effects model will be used throughout, with analyses conducted in R Studio version 3.6.0+ using the metafor package. Heterogeneity will be assessed via I², Cochran’s Q, and tau2 statistics. Studies with very small sample sizes (e.g., n < 10) will be included if they meet eligibility criteria. Their influence will be moderated through inverse-variance weighting, which naturally assigns lower weights to studies with greater uncertainty. Sensitivity analyses will assess the impact of these studies on pooled estimates, and their contribution to heterogeneity will be explored through subgroup and moderator analyses.

Subgroup analyses will explore variations in rTMS protocols, participant characteristics (e.g., age, sex, COVID-19 severity), study features, and symptom domains. Moderator analyses will assess how rTMS parameters and participant traits influence outcomes, and the impact of risk of bias will be tested through sensitivity analyses. Additional sensitivity checks will examine the effects of excluding high-risk studies, handling missing data, and methodological quality. Where available, follow-up data will be used to assess treatment durability, and adverse effects will be summarised qualitatively or quantitatively. If data allow, meta-regression will be used to examine relationships between continuous variables (e.g., number of sessions) and effect sizes. In cases of missing or insufficient data, planned analyses will be adjusted and limitations clearly reported. Results will be presented using forest and funnel plots, and a GRADE profile will summarise the quality of evidence for each outcome. Conclusions will be based on α level, confidence intervals, and effect sizes. Effect sizes will be interpreted using thresholds: Cohen’s d ≥ 0.20 as small but meaningful, ≥0.50 as moderate, and ≥0.80 as large; ORs or RRs ≥ 1.5 will indicate clinical relevance for dichotomous outcomes. High heterogeneity (I² > 50%) will prompt cautious interpretation and exploration through subgroup or moderator analyses. The GRADE framework will guide the confidence in conclusions, depending on evidence quality. For rTMS to be considered clinically useful for long COVID, we expect consistent small-to-moderate effects across key outcomes like fatigue and cognitive dysfunction. If subgroup findings show stronger effects in specific populations or symptom domains, these will be highlighted. Adverse effects, if present in ≥10% of participants, will temper conclusions. In the purposed meta-analysis, a saturation point will be reached (5–10 studies) when additional data no longer significantly alter the results, indicating that the findings are stable. Long-term follow-up data (≥3 months) will be incorporated to strengthen the conclusions regarding the durability of treatment effects. Additionally, if evidence of publication bias is detected, the interpretation of results will be approached with greater caution, particularly in terms of their generalisability.

4. Discussion

The meta-analysis outlined in this protocol will assess the therapeutic effectiveness of repetitive transcranial magnetic stimulation (rTMS) in the treatment of long COVID symptoms. With growing application of rTMS, in individuals with long COVID [16] and related conditions [25,26,27,28], an up-to-date meta-analysis is warranted. The decision to conduct a meta-analysis arose from the need to identify key symptom outcomes of rTMS, including parameters such as frequency, intensity, safety, and target sites, and to address existing gaps in the literature. The decision to concentrate on rTMS stems from its growing clinical adoption, mechanistic specificity, and the need to clarify its role in managing symptoms such as fatigue, cognitive dysfunction, and mood disturbances. By isolating rTMS from other modalities, the proposed systematic review and meta-analysis will better identify effective stimulation parameters (e.g., frequency, intensity, target sites), assess durability of effects, and highlight gaps in the literature that may be obscured in broader analyses.

5. Project Timeline

The estimated project timeline includes screening of titles, abstracts, and full texts from November to December 2025, followed by data extraction from January to March 2026. Results are expected by May 2026. During this period, database searches will be updated to identify any newly published studies which meet the inclusion criteria. Authors will discuss the results and how they can be interpreted from the perspective of previous studies and of the working hypotheses. The findings and their implications will be discussed in the broadest context possible. It is likely that future research directions will also be highlighted.

6. Conclusions

The meta-analysis outlined in this protocol will not be a critical appraisal of studies due to the emerging nature of the field but will aim to provide a groundwork on rTMS in long COVID. The review will highlight the efficacy of rTMS in managing long COVID symptoms like fatigue, cognitive dysfunction, and mood disturbances, helping to inform future research directions and clinical practice. By providing context to current findings, this research will guide future studies to enhance treatment effectiveness for long COVID.