Efficacy and safety of transcranial magnetic stimulation on cognition in mild cognitive impairment, Alzheimer’s disease, Alzheimer’s disease-related dementias, and other cognitive disorders: a systematic review and meta-analysis

Abstract

Objective:

We aim to analyze the efficacy and safety of TMS on cognition in mild cognitive impairment (MCI), Alzheimer’s disease (AD), AD-related dementias, and nondementia conditions with comorbid cognitive impairment.

Design:

Systematic review, Meta-Analysis

Setting:

We searched MEDLINE, Embase, Cochrane database, APA PsycINFO, Web of Science, and Scopus from January 1, 2000, to February 9, 2023.

Participants and interventions:

RCTs, open-label, and case series studies reporting cognitive outcomes following TMS intervention were included.

Measurement:

Cognitive and safety outcomes were measured. Cochrane Risk of Bias for RCTs and MINORS (Methodological Index for Non-Randomized Studies) criteria were used to evaluate study quality. This study was registered with PROSPERO (CRD42022326423).

Results:

The systematic review included 143 studies (n = 5,800 participants) worldwide, encompassing 94 RCTs, 43 open-label prospective, 3 open-label retrospective, and 3 case series. The meta-analysis included 25 RCTs in MCI and AD. Collectively, these studies provide evidence of improved global and specific cognitive measures with TMS across diagnostic groups. Only 2 studies (among 143) reported 4 adverse events of seizures: 3 were deemed TMS unrelated and another resolved with coil repositioning. Meta-analysis showed large effect sizes on global cognition (Mini-Mental State Examination (SMD = 0.80 [0.26, 1.33], p = 0.003), Montreal Cognitive Assessment (SMD = 0.85 [0.26, 1.44], p = 0.005), Alzheimer’s Disease Assessment Scale–Cognitive Subscale (SMD = −0.96 [−1.32, −0.60], p < 0.001)) in MCI and AD, although with significant heterogeneity.

Conclusion:

The reviewed studies provide favorable evidence of improved cognition with TMS across all groups with cognitive impairment. TMS was safe and well tolerated with infrequent serious adverse events.

Introduction

Dementia is a global challenge due to its profound negative psychosocial impact on individuals with dementia, their caregivers, and society at large. More than 55 million people live with dementia worldwide, and prevalence is expected to increase to 78 million by end of 2030 (Gauthier et al., 2021). Mild cognitive impairment (MCI) has a prevalence of 12% to 18% in people who are 60 years and older (Gaugler et al., 2021). Individuals with MCI have a higher risk of developing dementia, with dementia progression rates at 10% to 15% in the clinical setting and 8% to 18% per year in the community (Petersen et al., 2018). Currently, medications approved by the US Food and Drug Administration (FDA) for Alzheimer’s disease (AD) only temporarily treat cognitive and behavioral symptoms, although the latest approved drugs aducanumab and lecanemab may delay disease progression (Esang and Gupta 2021; van Dyck et al., 2022). Nonpharmacologic interventions such as risk reduction, cognitive training, psychosocial therapies, and nutraceuticals require further studies (Arvanitakis et al., 2019). More research is needed on novel therapies to improve cognitive impairments or delay progression in MCI or dementia.

Previously published clinical trials and systematic reviews with meta-analyses on the efficacy and safety of transcranial magnetic stimulation (TMS) are limited to focused groups as MCI, dementia due to AD, and AD-related dementias (Birba et al., 2017; Cheng et al., 2018; Dong et al., 2018; Nardone et al., 2014). These investigations suggest that TMS holds promise for enhancing cognitive functions. Much of the extant literature is confounded by methodological inconsistency despite such encouraging findings. For instance, treatment protocols vary considerably between investigations, with location, intensity, and frequency of magnetic stimulation differing across clinical trials. Additionally, outcome variables vary between studies, with some focusing on global cognition, while others measuring specific functions. Consequently, it is difficult to delineate clear and coherent conclusions from these disparate investigations, and a thorough systematic review may clarify matters.

To address these challenges, we conducted a systematic review to examine the efficacy and safety of TMS on cognitive functions in dementia and MCI and in populations with cognitive impairment not due to neurodegenerative disorders. In addition, we conducted a meta-analysis to assess the efficacy of randomized clinical trials (RCTs) of TMS compared to sham stimulation in MCI and AD populations.

Methods

This systematic review and meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Moher et al., 2009) and registered with PROSPERO (CRD42022326423).

Search strategy and selection criteria

We conducted a comprehensive search of several databases from January 1, 2000, to May 26, 2021, limited to the English language and excluding animal studies. The search was updated on February 9, 2023. Databases searched were Ovid MEDLINE, Ovid Embase, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews (2005+), Ovid APA PsycINFO, and Scopus via Elsevier. The search strategy was designed and conducted by a medical librarian (L.C.H.) with investigators input. Controlled vocabulary supplemented with keywords was used to search for studies describing TMS in AD and related disorders. The actual strategy listing all search terms used and how they are combined is available in Supplemental Table 1.

Included studies met the following criteria: (1) study population with cognitive impairment or dementia regardless of underlying cause, or healthy older adults (HOAs); (2) TMS as an intervention; (3) cognitive functions as outcomes; (4) study design: controlled or uncontrolled studies, including RCTs, open-label trials, case-control studies, or case series; and (5) English language. Studies on HOAs were included if TMS was used as an intervention to improve cognition. Single case studies, preclinical studies, abstracts only, and clinical trial registries without results were excluded.

Four reviewers (M.I.L., S.R.P., R.K., and L.C.H.) worked independently in pairs to identify and screen titles and abstracts using a standardized protocol. Subsequently, the full texts were reviewed separately by two reviewers (S.R.P., R.K.) Excluded articles and reasons for exclusion were logged (Supplemental Table 2). Disagreements were resolved through consensus. If there were multiple studies from the same cohort, only the study with a larger sample size was included.

Data collection and quality assessment

Data were extracted by two reviewers for each article (S.R.P. and R.K.) and discrepancies adjudicated by a third reviewer (M.I.L.). To check for reliability, 10% of the data extracted was randomly selected and verified for accuracy by three other reviewers (P.E.C., S.K., B.N.L.). Information extracted includes authors, year, country, study design (RCT, open-label, case series), study population (diagnosis), sample size, demographic characteristics of study participants, inclusion and exclusion criteria, TMS protocols and treatment parameters, cognitive outcome measures, adverse events, and study funding.

Studies were divided into six diagnostic groups − (1) dementia due to AD, (2) MCI and dementia due to AD (studies that included patients with AD and MCI), (3) MCI, (4) dementia due to non-AD, (5) other nondementia conditions with comorbid cognitive impairment, and (6) HOAs (including subjective cognitive decline). Studies that included more than one type of study population are each represented only once in our data set. Studies with combined patient population of MCI and dementia due to AD were grouped as “MCI and dementia due to AD.” The group of “other nondementia conditions with comorbid cognitive impairment” included psychiatric disorders such as schizophrenia, depression, bipolar disorder, and other brain disorders.

Two reviewers (S.R.P. and R.K.) independently assessed the quality of RCTs using the Cochrane risk of bias tool (Schünemann et al., 2019) and the Methodological Index for Non-Randomized Studies (MINORS) criteria (Slim et al., 2003) for nonrandomized studies.

Meta-analysis

Given heterogeneity in study designs, repetitive TMS (rTMS) protocols, and cognitive outcome measures, including all of the studies in meta-analysis was not feasible. We therefore only analyzed RCTs with common global cognitive outcomes (Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS-Cog)) in MCI and AD compared to sham stimulation. In instances of studies with multiple treatment groups, each treatment group was treated as an individual study. Change from baseline means and SDs was calculated for studies which only provided pre- and post-treatment means and SDs following standard formulas (Supplemental Table 3) (Higgins, 2011).

Overall heterogeneity was assessed using the Cochrane Q test and I ² statistic, and two-tailed P values reported (Cooper et al., 2009). Cochrane Q test P values of <0.1 and I² > 50% were deemed thresholds of study heterogeneity. Fixed-effect models were fit when study heterogeneity was absent, and random-effect models were fit when study heterogeneity was observed (Riley et al., 2011). Data analyses were performed in R version 4.2.2 (RStudio Team 2021, Boston, Massachusetts).

Results

Search results

A total of 1,199 abstracts were screened, of which 327 articles were selected for full-text review eligibility, and 143 studies met inclusion criteria for systematic review. Twenty-five studies met inclusion criteria for meta-analysis as shown in the PRISMA flow diagram (Figure 1). Inter-reviewer agreement during both phases of study selection was excellent (>95%).

Figure 1. PRISMA flow diagram.

Characteristics of included studies: diagnostic groups and study design

A composite sample size of 5,800 participants emerged from the 143 included studies (Table 1) worldwide, which comprised of 94 RCTs, 43 open-label prospective, 3 open-label retrospective, and 3 case series. Diagnostic groups included nondementia conditions with comorbid cognitive impairment (2,337 [40.3%]), dementia due to AD (1,827 [31.5%]), MCI and dementia due to AD (271 [4.7%]), dementia due to non-AD (720 [12.4%]), MCI (522 [9%]), and HOA (123 [2.1%]). Sex was reported in only 133 studies, of which 2 studies included only men, and there were 2,439 (45.6%) women. Mean ages ranged from 60 to 74 years for MCI, dementia due to AD, and non-AD; 38 to 47 years for nondementia conditions with comorbid cognitive impairment; and a mean age of 63.4 years for HOA.

Table 1. Characteristics of 143 studies in the systematic review by diagnostic groups (N = 5,800) ^a

Abbreviations: AD, Alzheimer’s disease; MCI, mild cognitive impairment; NR, not reported; OLP, open-label prospective; OLR, open-label retrospective; RCT, randomized clinical trial.

^a Studies not reporting mean age or sex were excluded from the analysis.

^b One study did not report sex.

^c One study did not report mean age.

^d Three studies did not report sex.

Characteristics of included studies: Efficacy, safety, and TMS protocols

Table 2 outlines author, publication year, country, study design, study population, sample size, TMS protocols, cognitive outcomes, and adverse events. Studies are listed by diagnosis and study type: dementia due to AD (n = 56), combined MCI and dementia due to AD (n = 6), MCI (n = 16), dementia due to non-AD (n = 26), nondementia conditions with comorbid cognitive impairment (n = 34), and HOA (n = 5). Detailed inclusion and exclusion criteria, mean ages, and financial support for the studies are listed in Supplemental Table 4. More than half the included studies reported were from China (n = 48), Italy (n = 16), and USA (n = 13) with 25 other countries reporting 1 to 6 studies each (Supplemental Table 5) representing different population types and global work.

Table 2. Summary of rTMS studies across diagnostic groups (N=143)

Abbreviations: AD, Alzheimer’s disease; ADAS-Cog, Alzheimer’s Disease Assessment Scale–Cognitive Subscale; aMCI, amnestic mild cognitive impairment; ACE, Addenbrooke Cognitive Examination; ADL, activities of daily living; AVLT, Auditory Verbal Learning Test; BNT, Boston Naming Test; CAVLT, Chinese version of Auditory verbal learning test; COG, Cognition ; COWAT, controlled oral word association test; cTBS, continuous theta burst stimulation; CVA, cerebrovascular accident; DLPFC, dorsolateral prefrontal cortex; ECT, electroconvulsive therapy; FTD, frontotemporal dementia; HC, healthy controls; HF, high frequency; HOA, healthy older adult; IFG, inferior frontal gyrus; iTBS, intermittent theta burst stimulation; LF, low frequency; MCI, mild cognitive impairment; MDD, major depressive disorder; MMSE, Mini-Mental Status Examination; MoCA, Montreal Cognitive Assessment; NPI, neuropsychiatric inventory; NR, not reported; OLP, open-label prospective study; OLR, open-label retrospective study; PANSS, positive and negative syndrome scale; PFC, prefrontal cortex; PD, Parkinson disease; PNFA, progressive nonfluent aphasia; PPA, primary progressive aphasia; PROAD, probable Alzheimer dementia; PSAC, primary somatosensory association cortex; PSCI, post-stroke cognitive impairment; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; RCT, randomized clinical trial; RT, reaction time; rTMS, repetitive transcranial magnetic stimulation; SCD, subjective cognitive decline; tDCS, transcranial direct current stimulation; TMS, transcranial magnetic stimulation; WCST, Wisconsin Card Sorting Test.

TMS efficacy across diagnostic groups

The studies in each diagnostic group are further classified by the study design type and report the number of patients and mean age (Table 1). The TMS protocol parameters are reported in Supplemental Figure 1. High-frequency (HF) stimulation is defined as 5 Hz or greater, while all stimulation frequencies less than 5 Hz is labeled as low frequency (LF).

Dementia due to AD

Among all AD studies, the most used cognitive outcomes were measures of global cognition such as the MMSE (n = 30), ADAS-Cog (n = 26), and MoCA (n = 15). Thirty-four of the 37 RCT studies compared TMS to sham stimulation, among which 31 (91%) showed significant improvement in cognitive measures. Three other studies (8%) reported the following: no overall efficacy (Saitoh et al., 2022), no statistically significant improvement (Vecchio et al., 2021), and low improvement rates in ADAS-Cog scores noted in only 13 of 27 patients with AD (Lithgow et al., 2021). Among AD open-label studies, 18 of 19 studies (95%) showed improvement in global cognition (MMSE, MoCA, ADAS-Cog) and other specific cognitive functions measured (memory, learning, naming, executive function). Teti Mayer et al., noted no impact on MMSE, but improved semantic and visual memory (Teti Mayer et al., 2021). Overall, a majority of AD studies report improvement in different cognitive measures with TMS.

MCI and dementia due to AD

Two of the six studies were RCTs. Five of the six studies (83%) reported improved memory, executive function, and global cognition with TMS. One study analyzing the TMS impact on AD progression, using continuous theta burst stimulation (TBS) and intermittent TBS (iTBS), found that AD progression was faster in patients with cerebrospinal fluid–positive AD (positive CSF biomarkers and presence of dementia) or prodromal AD (positive CSF biomarker and absence of dementia) than MCI (negative CSF biomarker and absence of dementia) patients, as measured by MMSE over 36 months (Di Lorenzo et al., 2020).

Mild cognitive impairment

There were 16 MCI studies, comprised of 12 RCTs and 4 open-label prospective studies. Diagnoses included MCI (n = 12), vascular MCI (n = 3), and MCI-Parkinson disease (PD) (n = 1). Of the 12 RCTs, 11 studies (92%) reported improved cognitive outcomes, while 1 study (Sedlackova et al., 2008) in vascular MCI participants reported no change. All open-label studies reported improvement in MMSE and recognition memory with TMS.

Dementia due to non-AD

Diagnoses for dementia due to non-AD included stroke (n = 10), frontotemporal dementia (n = 7) (including primary progressive aphasia and progressive nonfluent aphasia), PD (n = 6), multiple sclerosis (n = 1), Huntington disease (n = 1), and corticobasal degeneration (n = 1). Five of nine RCTs in stroke patients used LF (1 Hz) stimulation. HF stimulation was used in four studies, which included two iTBS protocols (Chu et al., 2022; Tsai et al., 2020). All nine RCT studies in stroke patients showed improvement in cognitive function. Among PD studies, all studies demonstrated cognitive improvement with TMS except for 1 study that only applied a single iTBS session to the L-DLPFC (Hill et al., 2020). In progressive nonfluent aphasia, LF stimulation (1 Hz) on the right Broca’s area showed significant improvement in cognition compared to HF stimulation (10 Hz) (Hu et al., 2018). One study in Huntington disease did not show significant cognitive improvement with a single session, M1 motor area stimulation utilizing 200 pulses (Groiss et al., 2012). Overall, a majority of non-AD studies (24 out of 26 studies) demonstrate that TMS has a positive impact on cognitive functions.

Nondementia conditions with comorbid cognitive impairment

Of the 20 RCTs, conditions with comorbid cognitive impairment included psychiatric (schizophrenia [n = 9], major depressive disorder [MDD] [n = 9], generalized anxiety disorder [n = 1]), and nonpsychiatric (traumatic brain injury, n = 1) diagnoses. In schizophrenia, there was no benefit in cognitive function when 10 Hz was applied to the L-DLPFC (Guse et al., 2013; Hasan et al., 2016; Wolwer et al., 2014; Xiu et al., 2020), although 1 study (Du et al., 2022) found a higher pattern in recognition memory at week 8 despite no improvement at week 4. Results from using 20 Hz were mixed, with some cognitive benefit in two studies (Xiu et al., 2020; Guan et al., 2020) but none in another study (Zhuo et al., 2019). Stimulation with 10 Hz as a supplement to antipsychotics resulted in improved recall in 1 study (Wen et al., 2021). In MDD, seven studies reported improved cognition (Buchholtz et al., 2020; Cheng et al., 2016; Hou et al., 2022; Jagawat et al., 2022; Myczkowski et al., 2018; Nadeau et al., 2014; Yu et al., 2022), while two studies (Holczer et al., 2021; Hausmann et al., 2004) did not.

Most of the open-label studies involved patients with MDD, except for one study each on schizophrenia (Zhuo et al., 2022), traumatic brain injury (Zhou et al., 2021), and long-COVID (Noda et al., 2022). Patients whose depressive symptoms decreased in response to TMS sustained improvement in cognition (Abo Aoun et al., 2019; Furtado et al., 2013). Only two open-label MDD studies, with stimulation over bilateral DLPFC, noted no improvement in cognition (Galletly et al., 2016; Hoy et al., 2012). Other MDD studies noted cognitive benefit, independent of the improvement in depression.

HOAs and subjective cognitive decline

Three of the five studies are RCTs. All five studies reported improvement in cognition following TMS (Cotelli et al., 2010; Chen et al., 2020; Hermiller et al., 2022; M. Liu et al., 2021; Sole-Padulles et al., 2006).

Meta-analysis: TMS effect on global cognition, compared to sham stimulation in MCI and AD subgroups

Twenty-five RCTs on MCI and AD were included in the meta-analysis. TMS significantly improved cognition in MCI and AD, when compared to sham stimulation, across all three of the most used global cognitive outcome measures. MMSE (n = 24, SMD = 0.80 [0.26, 1.33], p = 0.003), MoCA (n = 10, SMD = 0.85 [0.26, 1.44], p = 0.005), and ADAS-Cog (n = 14, SMD =−0.96 [−1.32, −0.60], p < 0.001) all showed large effects of improvement on global cognition (Figure 2). There was significant heterogeneity in the subgroup analyses (MMSE, I ² = 96.68%; MoCA, I ² = 82.09%; ADAS-Cog, I ² = 82.09%) (Supplemental Table 6a, b, c). Of the 25 studies included in meta-analysis, 10 studies were from China, 4 from Italy, 3 from USA, while other countries namely Iran, Mexico, Taiwan, Japan, Korea, Israel, Egypt, and Turkey had one study each. This represents the diverse regional representation of studies in the meta-analysis. We have not noticed specific differences in results across studies by region.

Figure 2. Forest plot analysis of different cognitive outcomes. A, Mini-Mental Status Examination (MMSE). B, Montreal Cognitive Assessment (MoCA). C, Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS-Cog).

Safety

Most of the studies demonstrated no major safety concerns (Table 2). Of 143 studies, there were 2 studies that reported 4 serious adverse events as seizures. In 1 RCT, there were 3 instances of seizures that occurred 6 to 12 months after TMS (J. Cheng et al., 2021), with 2 of cases occurring in the sham group, and none were deemed related to rTMS. In another study (Tumasian and Devi, 2021), 1 patient experienced motor movements during parietal rTMS deemed to be focal motor seizures which resolved with coil positioning (Tumasian and Devi, 2021). Two other AD studies reported serious adverse events of acute myocardial infarction (Leocani et al., 2020) and urinary sepsis (Bentwich et al., 2011) all unrelated to TMS. Overall, 47 studies (33%) reported adverse events, most commonly headache, local skin or scalp discomfort, and fatigue. Only 2 patients discontinued the study due to side effect intolerance. Forty (28%) studies reported no adverse events, and 52 (36%) studies did not have information on adverse events.

TMS parameters

TMS parameters are summarized in Supplemental Figure 1, including site of stimulation, frequency, motor threshold, number of treatment sessions, and total pulses per session. Stimulation sites were classified into five different categories based on site of stimulation as L-DLPFC only, bilateral DLPFC, six sites (right DLPFC, left DLPFC, Broca’s area, Wernicke’s area, right parietal somatosensory association cortices (PSAC), and left PSAC), other sites, and L-DLPFC combined with other sites of stimulation. L-DLPFC is the most common stimulation site across all diagnostic groups. Most of the studies used HF stimulation. Percent motor threshold ranged from 70 to 120%, although 90 to 100% was the most used range. Number of TMS sessions ranged from 1 to 54, with 10 or 20 sessions being the common treatment duration. A total of 19 studies (4-HOA and SCD, 5-non-AD, 4-MCI, 2-AD & MCI, 4-AD) in the systematic review reported 4 or less TMS sessions that they administered in their study. Total number of pulses per session ranged from <600 to 4,000, with 1,000–2,000 per session being the most frequently used.

Quality assessment

The quality assessment is reported in Figures 3A and 3B, with overall quality being modest across the studies. Detailed quality assessments for each study are included in Supplemental Table 7 (Cochrane Risk of Bias) for RCTs and Supplemental Table 8 (MINORS criteria) for non-RCT studies.

Figure 3. Qualitative assessments. A, Cochrane Risk of Bias for RCT (n = 94). B, MINORS criteria for non-RCT (n = 49).

Discussion

We have three main findings from this study. First, there is evidence for improvement of global and specific cognitive functions with TMS across all diagnostic groups with cognitive impairment. Second, TMS was safe and well tolerated with minimal serious adverse events generally deemed unrelated to TMS. Third, there was a wide variability across studies, in TMS protocols and cognitive measures which limit the determination of optimal parameters in this population.

Efficacy of rTMS for cognitive impairment

Most of the reviewed studies in our systematic review provide evidence of improved cognitive functions with TMS. Meta-analysis of RCT studies in MCI and AD shows rTMS significantly improved global cognition (MMSE, MoCA, ADAS-Cog) compared to sham stimulation. Improvement in specific domains such as memory, working memory, or executive function was found in different studies, but this may reflect the dearth of studies that addressed such specific domains. Future research might transcend reliance upon general cognitive measures and focus on more sensitive measures of specific cognitive domains. In doing so, those neuropsychologic functions that are most likely to improve may be identified. Furthermore, research may reveal that TMS to specific regions may exert a more potent benefit upon certain cognitive domains. For example, stimulation of frontal regions may yield a more robust benefit of executive function and working memory than new learning. Ultimately, this would allow a more personalized approach, where the TMS intervention might be guided by each patient’s symptoms or cognitive disability.

Our study findings are consistent with previously published systematic reviews and meta-analyses reporting a range of effect sizes. A meta-analysis of 12 studies analyzing the effect of rTMS therapy on cognition in AD found a moderate effect size (SMD = 0.60; 95% CI, 0.35–0.85) (Lin et al., 2019). Additionally, multiple sites of stimulation improved cognition more than single-site stimulation, and more rTMS treatments (≥5) resulted in better cognitive improvement than less (≤3) rTMS treatments (Lin et al., 2019). Another review of 5 RCTs found significant improvement in cognition with high-frequency rTMS when measured by ADAS-Cog (SMD = −3.65; 95% CI, −5.82 to −1.48; P = 0.001) but not MMSE (SMD = 0.49; 95% CI, −1.45 to 2.42; P = 0.62) (Dong et al., 2018). A meta-analysis investigated the efficacy of two techniques of noninvasive brain stimulation (rTMS and transcranial direct current stimulation [tDCS]) on global cognition and neuropsychiatric symptoms in people with AD and MCI (Teselink et al., 2021). There was significant improvement of global cognition (MMSE, MoCA, ADAS-Cog) with active rTMS but not tDCS (Teselink et al., 2021). Improvement of global cognition was greater in patients with AD and MCI when the site of active stimulation was the L-DLPFC compared to sham stimulation (Teselink et al., 2021). Another review of efficacy of TMS and tDCS on cognitive functioning is similar to our current systematic review in that it included many brain disorders (Begemann et al., 2020). Meta-analysis from 82 studies showed small effect sizes (Hedges’ g) of both TMS (g = 0.17, P = 0.015) and tDCS (g = 0.17, P = 0.021) on working memory across all brain disorders (Begemann et al., 2020). Another recent meta-analysis by Yan et al., described similar results on the overall cognitive improvement with TMS compared to sham stimulation in patients with MCI and AD both short term (<3 days) and long term (>4 weeks) (Yan et al., 2023). In the study by Yan et al., all the cognitive outcomes namely MMSE, MoCA, ADAS-Cog, and Rivermead Behavioral Memory test have been combined into one Meta-analysis category (Yan et al., 2023). Our study analyzed the effect on each cognitive outcome (MMSE, MoCA, and ADAS-Cog) separately. All the RCT TMS studies in AD and MCI populations analyzed in the above different meta-analysis studies were all included in our study along with other new eligible studies.

A clinically relevant change in MMSE scores is an important consideration in both clinical practice and research. While we found large effect sizes on global cognition in our meta-analysis, this does not always translate to clinical meaningfulness. Different studies have provided insights into what constitutes a significant change in MMSE scores or minimum clinically important difference (MCID). In one study of 451 cognitively unimpaired individuals and 292 people with MCI, a change of −1.5 to −1.7 points in MMSE was considered as MCID (Borland et al., 2022). Another study that used a distribution-based approach reported a similar range of mean changes in MMSE scores for MCIDs (Watt et al., 2021). One other study indicated that, in repeated assessments with 1.5-year intervals, a change in MMSE of at least 2–4 points indicated a reliable change at the 90% confidence level. However, it was emphasized that small changes in MMSE should be interpreted cautiously due to potential causes like measurement error, regression to the mean, or practice effect (Hensel et al., 2007). In a study of community-dwelling adults, a 3-point change in MMSE scores over a period of 3 years or more has been established as representative of a clinically meaningful decline in cognitive functioning (Pitrou et al., 2022). These studies collectively suggest that a change of 2–4 points in the MMSE score, especially over intervals of 1.5 to 3 years, can be considered clinically significant. However, the interpretation of these changes should be done cautiously, considering the potential for measurement error and individual variations. In our meta-analysis of 25 studies, we observed changes in MMSE scores that were lower than the conventional threshold for clinical significance. However, detecting small changes in MMSE scores even if not clinically significant can be valuable in understanding the subtle effects of TMS on cognitive function in people with MCI and dementia where any degree of cognitive improvement is meaningful.

Safety and tolerability of TMS in cognitively impaired populations

TMS was overall safe and well tolerated, with a low incidence of adverse events that were consistent with known adverse effects of TMS. Although rare, seizures are the most serious adverse event with TMS and the estimated risk is low at less than 1 in 30,000 (Rossi et al., 2021). The more common and expected adverse effects of TMS are transient headaches, scalp discomfort, and muscle twitches during stimulation (Rossi et al., 2021). In people with cognitive impairment, age is an important safety consideration for TMS given age-related physiologic changes, medical and neurologic comorbidities, presence of devices or implants, and polypharmacy, all factors that can affect response to TMS. However, the safety and tolerability of TMS is well-established when proper safety procedures are observed, even in older adults with depression (Iriarte and George, 2018). Following current TMS safety guidelines (Rossi et al., 2021) including proper screening of participants, ensuring stimulation parameters are within safety limits, and using qualified technicians and clinicians can help mitigate seizure risk (Fried et al., 2021; Pandis and Scarmeas, 2012; Targa Dias Anastacio et al., 2022). It is notable that there is significant underreporting as nearly one-third of studies did not report safety or adverse events. Inadequate documentation and disclosure of adverse events can distort the safety profile of TMS and hampers our understanding of the true benefits and risks in this population.

Heterogeneity of TMS treatment parameters

There is a wide variation in the TMS parameters used in each study. The most common site of stimulation is the L-DLPFC, using high-frequency stimulation, i.e. more than 5 Hz frequency, with 1,000–1,500 pulses per session, at 90% to 100% resting motor threshold (RMT), and treatment duration of 10–20 sessions. The current US FDA approval of TMS for MDD uses the L-DLPFC site, with HF 10–20 Hz (1,800–3,000 pulses per session) or iTBS (600 pulses per session), at 120% RMT, and 30 sessions. There are interesting similarities and differences between studies reviewed here and the US FDA-approved parameters in MDD. The similarities are L-DLPFC as the stimulation site and HF stimulation. In contrast to MDD protocols, fewer pulses per session, lower intensity (%RMT), and shorter duration of treatment were noted. In a previous systematic review of 30 studies including patients with psychiatric and neurologic diseases or healthy volunteers, it was reported that TMS was most likely to significantly improve cognitive functions when applied over the L-DLPFC, administered at 10-, 15-, or 20-Hz intensity, dosed at 80% to 110% of motor threshold, and delivered in 10 to 15 successive sessions (Guse et al., 2010). While TMS has received the most attention for depression, its potential use for other conditions is being investigated. There is ongoing debate on the dual identity of TMS as a one-size-fits-all therapeutic intervention and a personalized intervention targeting individual substrate and symptom-specific targets. The question of standardized versus personalized approaches remains a crucial area of investigation.

Cognitive impairment and dementia are conditions that are distinct from depression such that different parameters will be needed when TMS treatment is considered. However, it is also possible that improvements in mood could lead to cognitive enhancements in people with dementia and comorbid depression, underscoring the intricate interplay between emotional well-being and cognitive function. Many studies investigating the effects of TMS on cognition target the L-DLPFC but fail to control for potential mood effects. Since L-DLPFC stimulation has known antidepressant effects, any cognitive improvement observed could be directly due to the stimulation of this region or indirectly due to alleviation of depressive symptoms, emphasizing the importance of controlling for depression in these studies to isolate the true cognitive effects of TMS. Cognition is attributed to specific areas of the brain and exploration of sites other than L-DLPFC should be considered. Stimulating at 1 site could affect brain functional connectivity and impact another site (Eshel et al., 2020). HF stimulation is excitatory, which is thought to be needed for depression and dementia, whereas LF (thought to be inhibitory) stimulation has been used for anxiety and depression disorders. Future rTMS studies for dementia could investigate rTMS at 120% of RMT and use higher pulses per session and total number of sessions. Having the knowledge that higher parameters are used for other clinical and research applications of rTMS can help shape future rTMS for dementia research.

The effects of TMS in cognitive impairment or dementia are multifaceted and reflect complex interactions between TMS parameters and targeted brain tissue, therefore resulting in variability of TMS parameters. Varying degrees of brain atrophy can affect the amount of current induced in the brain, necessitating individualized computational modeling of the brain to adjust for optimal therapeutic effects. The slowing of neural oscillatory activity in dementia can influence how the brain responds to TMS, adding another layer of complexity but also offers the opportunity for a more nuanced, individualized and potentially effective approach. Given these diverse anatomical and physiological changes in dementia, there is a critical need for individualized approaches to ensure optimal therapeutic outcomes for each individual.

Strengths and limitations

This study extends findings of previous systematic reviews and meta-analyses to include a broader population with non-AD dementia subtypes and nondementia conditions with cognitive impairment, incorporate newer recently published studies for a more comprehensive review, summarize adverse effects and safety profile in cognitively impaired populations, analyze the extent of heterogeneity in study characteristics that impact generalizability of findings, consolidate existing knowledge, and provide further insights on the impact and potential benefits on TMS in populations with cognitive impairment globally. Limitations include heterogeneity in study designs, variability in stimulation parameters and cognitive outcome measures that limited ability to perform quantitative analysis in other diagnostic groups, and limited long-term data. Despite these limitations, this systematic review and meta-analysis provide valuable insights into the existing literature.

Conclusion

Overall, the reviewed studies provide favorable evidence for improvement of global and specific domains of cognitive functions with rTMS across all diagnostic groups with cognitive impairment. Meta-analysis showed large effect sizes on global cognition in MCI and AD, although with significant heterogeneity. The most common TMS parameters use the left DLPFC as the site for HF stimulation, 1,000–1,500 pulses per session at 90–100% of RMT, and duration of 10–20 sessions. TMS was safe and well tolerated with minimal adverse events, although there may be underreporting of adverse events. Heterogeneity of study design, TMS protocols, and cognitive measures limit the determination of optimal parameters for cognitively impaired populations.