Effects of high-frequency repetitive transcranial magnetic stimulation on memory of post-stroke cognitive impairment patients: A systematic review and meta-analysis

Abstract

Background

There is an evident interrelationship between stroke and Alzheimer's disease (AD). Post-stroke cognitive impairment (PSCI) is a frequently encountered and potentially disabling outcome of stroke. Memory impairment is an important component of the post-stroke cognitive syndrome, and high-frequency repetitive transcranial magnetic stimulation (HF-rTMS) has been widely used for memory in patients with PSCI.

Objective

In this study, we systematically evaluated the therapeutic effects of HF-rTMS on memory function in patients with PSCI, offering insights that may also inform the treatment of AD.

Methods

All relevant publications in Chinese and English were systematically searched from ten databases up to March 20, 2025. Retrieved articles were carefully screened. The quality of the included studies was assessed using the Cochrane Collaboration's risk of bias tool. The Review Manager 5.4 software was adopted for meta-analysis.

Results

Twenty-one studies of 1746 participants with PSCI were included. Meta-analysis revealed that HF-rTMS ameliorated memory of PSCI patients according to several outcome indicators: Rivermead Behavioural Memory Test [mean difference (MD) = 2.59, 95% confidence interval (CI) (2.08, 3.11), p < 0.00001], forward digit span [MD = 1.79, 95% CI (1.36, 2.22), p < 0.00001] and backward digit span [MD = 1.18, 95% CI (0.77, 1.59), p < 0.00001] of digit span test, Delayed Recall of the Montreal Cognitive Assessment [MD = 0.53, 95% CI (0.47, 0.59), p < 0.00001]; all p < 0.05.

Conclusions

The HF-rTMS might enhance memory in patients with PSCI, with the left dorsolateral prefrontal cortex being the most common stimulation site.

Keywords

Alzheimer's disease cognitive impairment memory meta-analysis randomized controlled trials repetitive transcranial magnetic stimulation stroke systematic review

Introduction

The findings of the Global Burden of Disease (GBD) study in 2021 reported stroke as the third leading cause of death and fourth leading cause of disability-adjusted life years (DALYs) globally, with significant public health implications.¹ Stroke and Alzheimer's disease (AD) represent two pathological conditions that share common underlying risk factors and can result in similar cognitive deterioration.² Post-stroke cognitive impairment (PSCI) is a common and potentially disabling consequence of stroke.^3,4 A multi-center cross-sectional study conducted in China enrolled 24,055 first-ever ischemic stroke patients and estimated that the prevalence of PSCI in this population was 78.7%.⁵ Memory impairment is an important component of the post-stroke cognitive syndrome.⁶ More than one-third of patients experience memory impairment after symptomatic stroke.⁷ Post-stroke memory impairment (PSMI) is a novel sub-type of PSCI characterized by memory dysfunction.⁸ Memory is one of the most complex and multifaceted cognitive domains. It comprises multiple subdomains, for most of which formal assessments have been developed.⁸ For instance, the digit span task (DST) mainly evaluates working memory, while the Rivermead Behavioural Memory Test (RBMT) assesses everyday and episodic memory. Memory impairment can seriously hinder patients’ ability to acquire new knowledge, recognize individuals, retrieve key information, and preserve their sense of self-awareness. Previous studies have shown that memory decline could significantly affect the long-term cognitive function of post-stroke patients.⁹ However, the treatment options for the memory of patients with PSCI are still limited.

In recent years, the technology of non-invasive brain stimulation has garnered significant attention in stroke rehabilitation, particularly transcranial direct current stimulation and repetitive transcranial magnetic stimulation (rTMS).¹⁰ Due to its non-invasive nature and ability to focally modulate brain activity, rTMS has been employed in the management of neurological and psychiatric disorders.¹¹ The initial and most well-established clinical application of rTMS involves using high-frequency stimulation on the left dorsolateral prefrontal cortex (L-DLPFC) to treat medication-resistant depression. To date, the Food and Drug Administration (FDA) has cleared six different TMS devices for medication-resistant depression.¹² Meanwhile, research exploring other potentially valuable clinical applications, such as in stroke and AD, is advancing at a swift pace. Nevertheless, for most of these potential indications, the relevant devices have not yet obtained FDA clearance for on-label use. High-frequency rTMS (HF-rTMS), defined as stimulation at a frequency >1 Hz, has been widely used in patients with PSCI. Previously, we had conducted systematic reviews and meta-analyses of HF-rTMS on overall cognition and activities of daily living in patients with PSCI, confirming its effectiveness.^13,14 However, regarding specific cognitive subdomains such as memory, attention, and executive function, related system reviews and meta-analyses are rare. Recently, one study systematically evaluated the efficacy of transcranial magnetic stimulation (TMS) on executive function, memory, and attention in stroke,¹⁵ and the other study evaluated the efficacy of rTMS on overall cognitive function, memory and activities of daily living in patients with PSMI.¹⁶ However, both studies searched only seven databases and included 5 to 13 articles, with intervention in the experimental groups including various TMS protocols (HF-rTMS, low-frequency rTMS, and theta-burst stimulation). By contrast, the present study conducted a more thorough and comprehensive search across ten databases to identify relevant randomized controlled trials (RCTs). We specifically focus on assessing the effectiveness of HF-rTMS in improving memory in patients with PSCI, aiming to provide a more targeted and in-depth analysis. Our goal is to offer evidence-based medical evidence for HF-rTMS in improving memory in PSCI, providing insights that may also inform the treatment of AD.

Methods

Literature search

The relevant studies published in Chinese or English were retrieved from ten databases involving Chinese National Knowledge Infrastructure (CNKI), Wanfang, VIP Periodical, SinoMed, PubMed, Cochrane Library, Web of Science, EBSCOhost, OVID, and Embase, with a search period from the database establishment date to March 20, 2025. We used MeSH terminology and free terminology to construct the retrieval strategy and determined through repeated retrieval. The MeSH terms were rTMS (重复经颅磁刺激), cognitive impairment (认知障碍), cognitive dysfunction, cognitive disorders, cognitive decline, stroke (中风/脑卒中), cerebrovascular accident (脑血管意外), and PSCI.

Inclusion and exclusion criteria

Inclusion criteria

(1) RCTs; (2) Patients with PSCI, including PSMI; (3) The control group interventions included routine treatment, rehabilitation therapy, cognitive stimulation, working memory training, pharmacotherapy and so on, without rTMS, whereas the experimental group added HF-rTMS to the intervention of the control group; or in sham-controlled designs, the experimental group received active HF-rTMS while the control group received sham HF-rTMS (sham-rTMS), with all other interventions being identical between groups; (4) Outcome indicators: at least one memory indicator was involved, such as RBMT, Wechsler Memory Scale (WMS), Addenbrooke's Cognitive Examination (ACE), auditory verbal learning test (AVLT), digit span test (DST) including forward digit span (FDST) and backward digit span (BDST), Cambridge prospective memory test (CAMPROMPT); and the global cognitive assessment scale with memory dimension like the Montreal Cognitive Assessment (MoCA).

Exclusion criteria

(1) Repeatedly published studies or data; (2) Narrative review, animal experimental studies, or case reports; (3) Even if the corresponding author is contacted, the full text cannot be obtained.

Literature screening and data collection

The retrieved literatures were independently sifted by two researchers (YK and GS). After reading titles and abstracts, irrelevant studies were deleted in accordance with inclusion and exclusion criteria. Subsequently, full-text readings were proceeded to further exclude irrelevant literature. Any disagreements were discussed by three researchers (YK, GS and FL) to reach a consensus. We extracted information such as title, the first author, publication year of literature, disease diagnostic criteria, number of participants, intervention of control and experimental groups, HF-rTMS parameters (frequency, intensity, stimulation sites), duration, outcome indicators, etc.

Assessment of risk-of bias

The Cochrane Collaboration's risk of bias tool contains 7 aspects and was employed to evaluate the risk of bias of the final included studies by two researchers (YK and JW). The risk of bias was classified into three distinct tiers: a “+” denoting low risk of bias, a “?” denoting an unclear risk of bias, and a “−” denoting high risk of bias. Any disagreements were discussed by three researchers (YK, JW, and FL) to reach a consensus.

Data analysis

Data analysis was performed with RevMan 5.4 software. Quantitative data were presented as mean differences (MDs). Study heterogeneity was assessed with the I² statistic. A threshold of I² ≤ 50% was considered to indicate low heterogeneity, and the fixed-effects model was employed for data synthesis. Conversely, when significant heterogeneity was detected (I²>50%), the random-effects model was applied, and subgroup analysis or sensitivity analysis was performed to identify potential sources of high heterogeneity. If heterogeneity remained substantially high, we conducted only a descriptive analysis without pooling. The statistically significant level is p < 0.05.

Results

Literature sifting

A preliminary search identified 4856 studies. After removing duplicates, 2284 studies remained. Further literature sifting was carried out in two steps: reading the titles and abstracts, and reading the full text carefully. Ultimately, 21 studies were included in this study. Figure 1 showed the literature sifting procedure.

Study characteristics

The 21 studies^17–37 included 1746 participants with PSCI. Four studies^18,20,26,30 did not report diagnostic criteria, 14 studies clearly reported diagnostic criteria for PSCI, and 3 studies^23,27,37 reported diagnostic criteria for PSMI. Regarding stroke type, 12 studies^{17,19,21–23}^{,25,28–29,31–32,34,37} included all types of stroke, 3 studies^33,35–36 only included ischemic stroke, one of the main subtypes of stroke, and 6 studies did not reported the stroke type. As for the course of stroke, the units used varied across studies (days, weeks, or months). None of the included studies reported the lesion side or lesion volume/location. One study¹⁹ encompassed two RCTs that met the inclusion criteria, therefore, the 21 studies actually comprised 22 RCTs. In 7 RCTs from 7 studies,^{17,19,21,29,34,36,37} the experimental group received rTMS, while the control group received sham-rTMS. The sham stimulation was delivered using a tilted coil used in 4 studies,^17,21,34,37 the sham coil in 2 studies,^19,34 and a flipped coil in 1 study.²⁹ In 15 RCTs, the control group received standard therapy, rehabilitation therapy, cognitive stimulation, working memory training, pharmacotherapy, or other interventions without rTMS, whereas the experimental group received rTMS in addition to the same interventions as the control group. All the frequencies of rTMS were higher than or equal to 5 Hz, i.e., HF-rTMS, specifically: 5 Hz in 5 studies,^{17,20,22,27,28} 6 Hz in 1 study,³³ 10 Hz in 13 studies,^{18,21,23,25,26,29–31}^,34–37 20 Hz in 2 studies.^19,24 Except for 3 studies that did not report the stimulation site,^20,25,26 rTMS was applied to the L-DLPFC in 15 studies,^{17–19,22,23,28–35,37} to bilateral DLPFC in 2 studies,^25,36 and to the DLPFC of lesion side in 1 study.¹ The treatment duration of HF-rTMS ranged from 2 to 8 weeks, with sessions once daily, 5∼7 times per week. Outcome indicators used to assess memory function included RBMT in 14 studies,^{18–24,27,29,30,32,34,36,37} the DST in 2 studies,^17,28 and global cognitive assessment scales that include memory subdomains, such as the Montreal Cognitive Test (MoCA), in 10 studies.^{21,25–29,31,32,35,36} Table 1 summarizes the features of the included studies.

Table 1.

Features of the included studies.

Studies & year	Participants, stroke types	Course of stroke	Diagnostic criteria	Age (years)	Sample size (C/E)	Interventions		HF-rTMS parameters			Intervention FRQ (per week)	Duration (week)	Outcome indicators
Studies & year	Participants, stroke types	Course of stroke	Diagnostic criteria	Age (years)	Sample size (C/E)	C	E	FRQ (Hz)	Intensity (MT)	Stimulation sites, coil type	Intervention FRQ (per week)	Duration (week)	Outcome indicators
Chang JP 2024¹⁷	PSCI, All types	C: 8.43 ± 2.49w E: 8.59 ± 2.79w	①; CT/MRI; MMSE	C: 59.40 ± 11.26 E: 57.82 ± 11.96	69 (35/34)	ST + CS + sham-rTMS (Tilted coil)	ST + CS + rTMS	5	80%	L-DLPFC, no reported	10min × 5times	4	ACE; DST
Luo W 2024¹⁸	PSCI, No reported	No reported	No reported	C: 61.71 ± 3.80 E: 62.95 ± 4.56	86 (43/43)	CS	CS + rTMS	10	80%	L-DLPFC, no reported	20min × 5times	4	RBMT (total)
Liu J 2024¹⁹	PSCI, All types	C:18.44 ± 8.36d E:18.28 ± 8.32d	CT/MRI; MMSE	C1: 68.12 ± 5.56 C2: 68.04 ± 7.31 E1: 66.68 ± 6.02 E2: 68.04 ± 4.84	100 (25/25/ 25/25)	C1: RT + sham-rTMS (Sham coil) C2: RT + WMT	E1: RT + rTMS E2: RT + WMT + rTMS	20	80%∼120%	L-DLPFC, F8c	20min × 6times	4	RBMT (total)
Li YQ 2024²⁰	PSCI, No reported	C:16～57d E:17～56d	No reported	C: 54.38 ± 5.64 E: 54.16 ± 5.49	80 (40/40)	ST	ST + rTMS	5	80%	Unclear, Cc	20min × 5times	2	RBMT (total)
Lu F 2023²¹	PSCI, All types	C:56.84 ± 9.81d E:57.43 ± 10.76d	①	C: 58.34 ± 6.15 E: 59.11 ± 7.24	80 (40/40)	DT + RT + sham-rTMS(Tilted coil)	DT + RT + rTMS	10	80%	DLPFC of lesion side, F8c	20min × 5times	4	MoCA; RBMT (total)
Hong MQ 2023²²	PSCI, All types	C:33.2 ± 23.77d E:32.8 ± 18.84d	④; CT/MRI; MoCA	C: 55.0 ± 13.0 E: 56.5 ± 8.23	60 (30/30)	DT + RT	DT + RT + rTMS	5	80%∼90%	L-DLPFC, no reported	20min × 5times	3	RBMT（screening score）
Chen YP 2023²³	PSMI, All types	C:3.20 ± 1.5m E:3.36 ± 1.38m	④; RBMT	C: 69.98 ± 6.47 E: 69.62 ± 5.97	90 (45/45)	MT +SFT + HG	MT + SFT + HG + rTMS	10	Unclear	L-DLPFC, F8c	20min × 7times	4	RBMT (sub-tests)
Xu BY 2022²⁴	PSCI, No reported	C:15.64 ± 3.72d E:16.17 ± 3.54d	②; ⑤; CT/MRI; MoCA	C: 63.52 ± 4.27 E: 64.09 ± 3.86	130 (65/65)	ST + SAFT	ST + SAFT + rTMS	20	80%	bilateral DLPFC, F8c	15 min/per side×5times	8	RBMT (total)
Shen LM 2022²⁵	PSCI, All types	C:2.51 ± 0.48m E:2.47 ± 0.36m	②; CT/MRI; MoCA	C: 61.59 ± 5.82 E: 61.14 ± 4.97	300 (150/150)	ST + RT + CS	ST + RT + CS + rTMS	10	80%	Unclear, no reported	20min × 5times	4	MoCA
Liu L 2022²⁶	PSCI, No reported	No reported	No reported	C: 61.14 ± 1.18 E: 61.25 ± 1.07	60 (30/30)	CS	CS + rTMS	10	80%	Unclear, no reported	20min × 5times	4	MMSE; MoCA
Huang CY 2021²⁷	PSMI, No reported	C:105.83 ± 44.20d E:96.25 ± 29.11d	⑥; CT/MRI; MMSE; RBMT	C: 61.48 ± 9.08 E: 63.87 ± 6.31	24 (12/12)	DT + RT	DT + RT + rTMS	5	80%∼120%	L-DLPFC, Cc	20min × 5times	4	MoCA; RBMT (total and sub-tests)
Xia JY 2021²⁸	PSCI, All types	C:3.15 ± 1.60m E:3.2 ± 1.80m	②; MMSE	C: 67.15 ± 6.352 E: 65.30 ± 9.04	40 (20/20)	ST + AA	ST + AA + rTMS	5	80%	L-DLPFC, F8c	10min × 5times	4	MoCA; DST
Zheng CJ 2020²⁹	PSCI, All types	C:47.3 ± 11.8d E:48.7 ± 14.4d	⑦; CT/MRI; MoCA	C: 59.7 ± 6.3 E: 58.3 ± 7.9	92 (45/47)	DT + RT + sham-rTMS (flipped coil)	DT + RT + rTMS	10	80%	L-DLPFC, F8c	20min × 5times	4	MoCA; RBMT (total)
Liu DX 2020³⁰	PSCI, No reported	No reported	No reported	C: 57.31 ± 3.62 E: 57.26 ± 3.74	80 (40/40)	ST + RT + CS	ST + RT + CS + rTMS	10	80%	L-DLPFC, no reported	20min × 5times	4	RBMT (total)
Yu ZY 2019³¹	PSCI, All types	C:12.28 ± 6.45m E:13.25 ± 5.80m	⑦; MoCA	C: 62.15 ± 13.49 E: 65.13 ± 10.32	100 (49/51)	DT + RT	DT + RT + rTMS	10	80%	L-DLPFC, F8c	20min × 5times	8	MoCA; AVLT
Chen ZY 2019³²	PSCI, All types	C:2.8 ± 0.8m E:2.9 ± 0.8m	⑦; CT/MRI; MoCA	C: 56.8 ± 4.8 E: 57.1 ± 5.0	140 (70/70)	ST + CS	ST + CS + rTMS	10	80%	L-DLPFC, F8c	20min × 5times	4	MoCA; RBMT (total)
Bao YY 2019³³	PSCI, IS	C:24.79 ± 4.59m E:25.15 ± 4.70m	④; CT/MRI	C: 68.47 ± 6.87 E: 68.74 ± 7.12	80 (40/40)	ST + DT	ST + DT + rTMS	6	120%	L-DLPFC, no reported	5times	8	WMS
Yin MY 2018³⁴	PSCI, All types	C:56.15 ± 23.74d E:59.83 ± 30.59d	⑦; CT/MRI; MoCA	C: 60.15 ± 10.29 E: 58.58 ± 11.98	25 (13/12)	DT + RT + sham-rTMS(Tilted coil)	DT + RT + rTMS	10	80%	L-DLPFC, F8c	20min × 5times	2, 4	RBMT (total)
He YG 2017³⁵	PSCI, IS	C:1.80 ± 0.77m E:1.56 ± 0.73m	⑦;⑧; CT/MRI; MoCA; CDR; HIS	C: 47.43 ± 13.67 E: 48.00 ± 12.42	30 (14/16)	DT + RT + CS	DT + RT + CS + rTMS	10	80%∼120%	L-DLPFC, Cc	20min × 5times	4	MoCA
Zhou Q 2017³⁶	PSMI, IS	C:1.63 ± 0.76m E:1.43 ± 0.68m	⑤; ⑧; CT/MRI; MoCA; RBMT;	C: 55.32 ± 7.58 E: 52.57 ± 8.53	40 (19/21)	DT + RT + CS +sham-rTMS (Sham coil)	DT + RT + CS + rTMS	10	80%∼120%	bilateral DLPFC,Cc	15 min/per side×5times	4	MoCA; RBMT (total); CAMPROMPT
Wang K 2016³⁷	PSCI, All types	C:2.5 ± 0.8m E:2.7 ± 0.7m	⑦; CT/MRI; MoCA	C: 58.3 ± 5.5 E: 57.9 ± 6.3	40 (20/20)	DT + CS + sham-rTMS(Tilted coil)	DT + CS + rTMS	10	80%	L-DLPFC, Cc	20min × 5times	8	RBMT (total)

C: control group; E: experiment group; FRQ: frequency; MT: resting motor threshold.; IS: ischemic stroke; HS: hemorrhagic stroke.

Diagnostic criteria: MMSE: Mini-Mental State Examination; MoCA: Montreal Cognitive Assessment; CDR: Clinical Dementia Rating; HIS: Hachinski Ischemic Score; RBMT: Rivermead Behavioural Memory Test. ① Diagnostic points of various major cerebrovascular diseases in China (2019); ② Expert consensus on the management of cognitive impairment after stroke (various year versions); ③ Guidelines for the diagnosis and treatment of dementia and cognitive impairment in China (2018); ④ Guidelines for the diagnosis and treatment of acute ischemic stroke in China (various year versions); ⑤ Guidelines for cerebrovascular disease prevention and treatment in China (various year versions); ⑥ Guidelines and consensus on the diagnosis and treatment of cerebrovascular diseases in China (2016); ⑦ Diagnostic points of various cerebrovascular diseases (1996); ⑧ Guidelines for the diagnosis and management of vascular cognitive impairment (2011).

Intervention: ST: standard therapy; RT: rehabilitation therapy; CS: cognitive stimulation; WMT: working memory training; DT: drug therapy; MT: memory training; SFT: symptom-focused therapy; HG: health guidance; SAFT: speech-auditory feedback training; AA: auricular acupuncture.

Stimulation site: L-DLPFC: left dorsolateral prefrontal cortex; DLPFC: dorsolateral prefrontal cortex; M1: primary motor cortex; F8c: figure-8 coil; Cc: circular coi.

Outcome indicators: RBMT: Rivermead Behavioural Memory Test (includes screening score and profile score, with a total score of 12 and 24 respectively); ACE: Addenbrooke's Cognitive Examination; WMS: Wechsler Memory Scale; AVLT: Auditory Verbal Learning Test; DST: Digit Span Test; CAMPROMPT: Cambridge Prospective Memory Test.

Study quality

The risk of bias assessments were shown in Figures 2 and 3. The method of random sequence generation was reported in 20 studies and was rated as “+”. Allocation concealment was reported in 2 studies,^27,28 blinding of participants and personnel was described in 5 studies, so these studies were rated as “+”.^{21,29,34,36,37} The blinding of outcome assessments was described in 8 studies and was rated as“+”.^{19,22,27–29}^,34,36,37 Except for one study²⁶ was rated as “-” for other bias due to an error in subdomains of outcome indicators, the remaining aspects of risk-of-bias were rated as “+”.

Meta-analysis of RCTs

For one study²² only the screening score of RBMT (range 0–12) was reported, while another only reported WMS.³³ As these outcome measures were each used in only one study, and one study²⁶ was judged to have a high risk of bias, these three studies^22,26,33 were not included in the meta-analysis. One study¹⁹ included two RCTs, therefore, a total of 18 studies (comprising 19 RCTs) were analyzed in this meta-analysis.

RBMT. RBMT is an ecologically valid memory test, which has been widely used for memory assessment. The test comprises 12 sub-tests covering a range of everyday memory capabilities; each sub-test includes screening scores (points 0–12) and profile scores (points 0–24).³⁸ The profile score (points 0–24) was used in 13 studies,^18–21^{,23,24,27,29,30,33,34,36,37} of which 11 studies only reported the total score, 1 study only reported the score for 12 sub-tests,²³ 1 study reported both the total score and the scores for 12 sub-tests.²⁷

Twelve studies,^18–21^{,24,27,29,30,33,34,36,37} of which one study¹⁹ included two RCTs, involved 13 RCTs including 917 participants using the total score of RBMT. A random-effect model was employed for data synthesis, revealing significant heterogeneity (p < 0.0001, I² = 87%). The meta-analysis showed that the HF-rTMS group achieved higher total scores of RBMT than the control group [MD = 3.19, 95% CI (2.17, 4.20), p < 0.00001] (Figure 4). Sensitivity analysis for heterogeneity dropped to 4% after excluding two studies,^18,30 and the meta-analysis showed [MD = 2.67, 95% CI (2.16, 3.17), p < 0.00001] (Figure 5). The heterogeneity might stem from differences in control group interventions and the absence of reported diagnostic criteria in two studies.^18,30 These above results suggested that HF-rTMS could improve memory in patients with PSCI. Sensitivity analysis by excluding two studies confirmed the robustness of these findings.

And we performed sub-group analysis based on the different frequency of HF-rTMS: 2 studies^20,27 used 5 Hz [MD = 2.99, 95% CI (1.72, 4.25), p < 0.00001]; 6 studies^{21,29,32,34,36,37} used 10 Hz [MD = 2.24, 95% CI (1.54, 2.94), p < 0.00001]; 2 studies^19,24 including 3 RCTs used 20 Hz [MD = 3.20, 95% CI (2.31, 4.08), p < 0.00001] (Figure 6). Subgroup analyses were also conducted according to treatment duration: 2 studies^20,34 with a duration of 2 weeks [MD = 2.90, 95% CI (1.61, 4.19), p < 0.001]; 7 studies^{19,21,27,29,32,34,37} including 8 RCTs with a duration of 4 weeks [MD = 2.35, 95% CI (1.68, 3.03), p < 0.00001]; 2 studies^24,37 with a duration of 8 weeks [MD = 3.14, 95% CI (2.24, 4.05), p < 0.00001] (Figure 7). The above results indicated that different stimulation frequencies and durations of HF-rTMS could improve memory function in patients with PSCI.

The score for the 12 sub-tests were reported in 2 studies,^23,27 involving 114 participants. Pictures, as one of the sub-tests, showed high heterogeneity so that it was excluded from the meta-analysis. A fixed-effect model was used in the remaining 11 sub-tests in the meta-analysis (all I² < 50%) and revealed that the HF-rTMS group achieved significantly higher scores than the control group on eight of the 11 sub-tests: Names [MD = 0.37, 95% CI (0.09, 0.64), p = 0.008], Belongs [MD = 0.36, 95% CI (0.12, 0.59), p = 0.003], Faces [MD = 0.35, 95% CI (0.10, 0.61), p = 0.007], Orientation [MD = 0.48, 95% CI (0.24, 0.72), p = 0.0001], Date [MD = 0.42, 95% CI (0.15, 0.70), p = 0.003], Immediate Messages [MD = 0.29, 95% CI (0.04, 0.54), p = 0.02], Delayed Messages [MD = 0.36, 95% CI (0.08, 0.63), p = 0.01], and Immediate Story [MD = 0.30, 95% CI (0.04, 0.55), p = 0.02]. No significant differences were observed for Delayed Story [MD = 0.19, 95% CI (−0.11, 0.48), p = 0.22], Immediate Route [MD = 0.25, 95% CI (−0.00, 0.50), p = 0.05], Delayed Route [MD = 0.22, 95% CI (−0.03, 0.48), p = 0.08] (Figure 8). These above results suggested that HF-rTMS could improve most of everyday memory capabilities in patients with PSCI.

DST. Two studies^17,28 involving 109 participants assessed memory by DST. The meta-analysis conducted with a fixed-effect model (I² = 0%) showed that the HF-rTMS group achieved significantly higher on both the FDST [MD = 1.79, 95% CI (1.36, 2.22), p < 0.00001] and the BDST [MD = 1.18, 95% CI (0.77, 1.59), p < 0.00001] compared to the control group (Figure 9). The results manifested that HF-rTMS could facilitate memory rehabilitation in patients with PSCI.

Delayed recall of the MoCA

The MoCA is one of the primary tools for cognitive screening in both everyday clinical practice and research, evaluating multiple cognitive domains including visuospatial ability, memory, attention, executive function, orientation, and so on.³⁸ Nine^{21,25,27–29}^,31,32,35,36 studies involving 846 participants assessed memory function using the Delayed Recall subtest of the MoCA. A meta-analysis conducted with a fixed-effect model (I² = 25%) showed that the HF-rTMS group achieved significantly higher scores on Delayed Recall sub-test of the MoCA than the control group [MD = 0.53, 95% CI (0.47, 0.59), p < 0.00001] (Figure 10). The result indicated that HF-rTMS could effectively enhance memory function in PSCI patients.

Descriptive analysis

One included study,²⁶ which assessed the effectiveness of HF-rTMS using the Mini-Mental State Examination and MoCA, was rated as having high risk of bias. Other cognitive indicators, including the ACE,¹⁷ the screening score of the RBMT,²² the WMS,³³ the AVLT,³¹ and the CAMPROMPT,³⁶ were each used in a single study to evaluate memory function after HF-rTMS in patients with PSCI. These studies did not be merge, their individual findings consistently indicated that HF-rTMS might improve memory function in patients with PSCI.

Adverse events (AEs) were reported in 7 studies, three of which^17,27,32 reported no AEs during HF-rTMS. One study²⁹ reported that 3 patients experienced mild dizziness and headache during HF-rTMS treatment. One study³⁴ reported that 3 participants experienced scalp pain. In two studies,^29,33 intervention with oral medicines (such as Donepezil and Traditional Chinese Medicine Wuling capsules) were used in both the control group and the HF-rTMS group. In one study,³³ there was one case of nausea in the control group and three cases in the HF-rTMS group, and one case of abdominal pain in both groups. In the other study,²⁹ there was one case of mild dizziness and headache in the control group, and one case of inattention and one case of sleep disorders in the HF-rTMS group. Importantly, no statistically significant differences were observed between groups, and the pharmacotherapy received by both groups might have contributed to these AEs. All AEs were tolerable for patients and resolved after rest. Table 2 summarizes the AEs reported in four included studies. Therefore, HF-rTMS can be considered a relatively safe technique for the improvement of memory in patients with PSCI.

Table 2.

Adverse events of four included studies.

Studies & year	Interventions and AEs		HF-rTMS parameters			Duration (week)
Studies & year	C	E	FRQ (Hz)	Intensity (MT)	Intervention FRQ (per week)	Duration (week)
Shen LM 2022²⁵	ST + RT + CS no AEs	ST + RT + CS + rTMS 3 patients experienced mild dizziness and headache	10	80%	20min × 5time	4
Zheng CJ 2020²⁹	DT (Donepezil Hydrochloride Tablets)+RT + sham-rTMS (flipped coil) 1 case of mild dizziness and headache	DT(Donepezil Hydrochloride Tablets) + RT + rTMS 1 case of inattention and 1 case of sleep disorders	10	80%	20min × 5times	4
Bao YY 2019³³	ST + DT (Wuling capsule) 1 case of nausea in the control group and 1 case of abdominal pain	ST + DT (Wuling capsule) + rTMS 1 case of nausea and 1 case of abdominal pain	6	120%	5times	8
Yin MY 2018³⁴	DT (conventional drugs)+RT + sham-rTMS (Tilted coil) no AEs	DT (conventional drugs) + RT + rTMS 3 participants with scalp pain	10	80%	20min × 5times	2, 4

Discussion

This systematic review included a total of 21 studies, and the results of the meta-analysis demonstrated that HF-rTMS significantly increased scores on the RBMT (the total score and scores of sub tests), as well as on the DST and the Delayed Recall of the MoCA. Sensitivity analysis confirmed the robustness of these findings, with low heterogeneity observed across studies. Although various memory assessment tools were used across the included studies, all fundamentally evaluated memory function. Furthermore, among the included studies, 15 studies applied HF-rTMS to the L-DLPFC. These findings indicated that HF-rTMS has the potential to enhance memory function in patients with PSCI, with the L-DLPFC emerging as the most frequently targeted stimulation site, irrespective of stroke lesion side.

The pathophysiological underpinnings of PSCI are not yet fully elucidated. Multiple mechanisms have been implicated, including increases in excitatory amino acid, oxidative stress, inflammatory responses, apoptosis, altered neurotrophic factor levels, and changes in gene expression that influence synaptic plasticity. Similar to AD,³⁹ the pathophysiology of PSCI remains complex and poorly understood. Emerging evidence suggests that PSCI and AD share overlapping pathological cascades, including the accumulation of amyloid-β and tau protein due to blood-brain barrier disruption, chronic neuroinflammation mediated by microglial activation, and impaired neurotrophic support (e.g., brain-derived neurotrophic factor, BDNF).⁴⁰ These shared mechanisms provide a biological rationale for the potential cross-disease efficacy of therapeutic interventions. Alternatively, PSCI may arise from the exacerbation of pre-existing vascular risk factors following a stroke, such as underlying white matter abnormalities or concurrent neurodegenerative changes, which become clinically manifest after the acute event.⁴¹ Infarction of the median thalamus is widely recognized as a significant cause of episodic memory deficits.⁴² Stroke can also affect other structures, including the internal capsule, deep white matter, and basal forebrain, predominantly in the left hemisphere, and may potentially leading to memory dysfunction.^43,44 The hippocampus and the medial thalamus are two key components of the extended hippocampal network, within which nerve fibers converge on specific brain structures identifiable on neuroimaging. Functional deficits following stroke arise from the disruption of these complex neural networks. In a strategic study, memory impairment was found in 4 out of 5 patients with first-ever left posterior cerebral artery territory infarcts, and the findings indicated that the common anatomical factor for memory impairment was interruption of the parahippocampal gyrus.⁴⁵ In addition, the mechanisms that lead to memory impairment include programmed cell death (apoptosis), inflammatory damage, and secondary damage caused by abnormal electrical activity and excitotoxicity.⁴⁶

Recent studies have uncovered potential therapeutic mechanisms underlying the application of non-invasive brain stimulation modalities in PSCI. A growing body of research had demonstrated that TMS can influence neuronal morphology, modulate glutamate receptors and neurotransmitter systems, and regulate the expression of synaptic-associated proteins. This effect is achieved partly by regulating the expression of BDNF expression, thereby influencing learning and memory functions.⁴⁷ An animal study revealed that HF-rTMS significantly improved cognition and mitigated white matter injury in a rat model of PSCI, and these effects were associated with the phenotypic reconfiguration of microglia, including a shift toward the neuroprotective M2 (anti-inflammatory) phenotype and increased anti-inflammatory cytokines levels.⁴⁸ One of the included studies in this review reported that HF-rTMS significantly enhanced cognitive function after stroke, an effect that may be improved cerebral blood perfusion in the frontal lobe, temporal lobe, and basal ganglia.²¹ Another study investigated the effects of HF-rTMS on memory function in elderly patients with stroke using functional MRI and found that its mechanism of action may be related to enhance regional blood flow and neuronal activity.²³ Collectively, these findings suggested that HF-rTMS can effectively ameliorate memory function in patients with PSCI through multiple potential mechanisms. Given that stroke and AD share several pathophysiological mechanisms, HF-rTMS may also represent an effective therapy for improving memory in patients with AD.

In addition, the DLPFC was the most commonly used stimulation site for improving memory of PSCI patients treated with HF-rTMS. Several brain regions including the occipital, fronto-parietal, and notably the DLPFC, anterior cingulate cortex, and parietal cortices, are critically implicated in the neural processes underpinning working memory (WM), reflecting their roles in attentional control, error monitoring, and sensory-spatial integration.^49,50 Neuropsychological studies suggest that the prefrontal cortex implements control processes that contribute to working memory and episodic long-term memory.⁵¹ This involvement is underscored by behavioral studies demonstrating that non-invasive stimulation of the DLPFC,⁵² anterior cingulate cortex,⁵³ and parietal cortex⁵⁴ directly modulates WM task performance. The DLPFC plays a vital role in the maintenance of information through a series of interconnected functions, such as retrieving stored information from memory, filtering out irrelevant stimuli, and integrating new data into existing memory frameworks to keep information relevant and up to date.^49,55,56 The DLPFC serves as a critical hub in brain networks mediating WM and executive control. The L-DLPFC, a key node of the central executive network, is closely related to advanced cognitive functions such as working memory, episodic memory, and selective attention.⁵⁷ Consequently, non-invasive stimulation of the L-DLPFC may modulate the functionality of these networks, offering a potential therapeutic avenue to enhance memory rehabilitation in patients with PSCI.

This systematic review has several limitations. First, the diagnostic criteria of PSCI were not uniform across the included studies. Additionally, variations in stroke type and course, lesion side, and control group interventions may have introduced clinical heterogeneity, potentially affecting the robustness of our findings. Second, treatment protocols such as sham control methodology, stimulation frequency, intensity, duration, and other clinical parameter, differed among studies, which precluded the aggregation and meta-analysis of these data. Third, the primary outcome measures were largely limited to neuropsychological scales, with a lack of objective biomarkers such as neuronal morphology, receptor expression, neurotransmitter levels, or inflammatory factors. Finally, the long-term efficacy of HF-rTMS could not be assessed, as none of the included studies reported follow-up outcomes beyond the immediate post-intervention period. Future studies should employ standardized treatment protocols and systematically compare the effects of different stimulation parameters (e.g., frequency, intensity, duration, and target sites) to identify the optimal regimen for memory rehabilitation in PSCI patients. And well-designed randomized controlled trials with extended follow-up periods are warranted to evaluate the durability of treatment effects and to determine whether HF-rTMS can confer sustained cognitive benefits or delay long-term cognitive decline in this population.

Conclusion

The HF-rTMS might improve PSCI patients’ memory, and the L-DLPFC emerged as the most commonly used stimulation site. This study exhibits certain limitations, necessitating further exploration that incorporates more objective measurement criteria and optimized parameters to improve its practical applicability.

Footnotes

Acknowledgements

The authors have no acknowledgments to report.

ORCID iD

Fang Liu

Ethical considerations

Not applicable

Consent to participate

Not applicable

Consent for publication

Not applicable

Author contribution(s)

Fang Liu: Conceptualization; Data curation; Writing – original draft; Writing – review & editing.

Jianhong Wang: Software; Validation.

Yahong Kang: Formal analysis; Software; Validation.

Guiting Su: Data curation; Formal analysis.

Xiaofang Wang: Software; Validation.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by 2022 Fujian Provincial College Student Innovation and Entrepreneurship Training Program (grant number S202210393008), and 2024 The Research Projects on Lifelong Education in Fujian Province (grant number ZS24050).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The datasets generated during and/or analyzed during the current study are publicly available.

References

Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet 2024; 403: 2133–2161.

Sriram

Mehkri

Quintin

, et al. Shared pathophysiology: understanding stroke and Alzheimer's disease. Clin Neurol Neurosurg 2022; 218: 107306.

Quinn

Richard

Teuschl

, et al. European Stroke Organisation and European Academy of Neurology joint guidelines on post-stroke cognitive impairment. Eur J Neurol 2021; 28: 3883–3920.

Craig

Hoo

Yan

, et al. Prevalence of dementia in ischaemic or mixed stroke populations: systematic review and meta-analysis. J Neurol Neurosurg Psychiatry 2022; 93: 180–187.

Wang

, et al. Incidence of post-stroke cognitive impairment in patients with first-ever ischemic stroke: a multicenter cross-sectional study in China. Lancet Reg Health West Pac 2023; 33: 100687.

Lin

Wang

Chen

, et al. Involvement of P2X7R-mediated microglia polarization and neuroinflammation in the response to electroacupuncture on post-stroke memory impairment. Brain Res Bull 2024; 212: 110967.

O'Sullivan

Galligan

, et al. Cognitive recovery after stroke: memory. Stroke 2023; 54: 44–54.

Harvey

. Domains of cognition and their assessment. Dialogues Clin Neurosci 2019; 21: 227–237.

de Lima Ferreira

de Oliveira Lopes

Montoril

, et al. Clinical validation of the nursing diagnosis of impaired memory in patients with a stroke. Jpn J Nurs Sci 2019; 16: 136–144.

10.

Motolese

Capone

Di Lazzaro

. New tools for shaping plasticity to enhance recovery after stroke. Handb Clin Neurol 2022; 184: 299–315.

11.

Burke

Fried

Pascual-Leone

. Transcranial magnetic stimulation: neurophysiological and clinical applications. Handb Clin Neurol 2019; 163: 73–92.

12.

Lefaucheur

Aleman

Baeken

, et al. Evidence-based guidelines on the therapeutic use of repetitive transcranial magnetic stimulation (rTMS): an update (2014-2018). Clin Neurophysiol 2020; 131: 474–528.

13.

Chen

Xiu

Hou

, et al. High-frequency repetitive transcranial magnetic stimulation on overall cognition in patients with poststroke cognitive impairment: a systematic review and meta-analysis. Am J Phys Med Rehabil 2024; 103: 418–427.

14.

Chen

Liu

Lyu

, et al. High-frequency repetitive transcranial magnetic stimulation (HF-rTMS) impacts activities of daily living of patients with post-stroke cognitive impairment: a systematic review and meta-analysis. Neurol Sci 2023; 44: 2699–2713.

15.

Honma

Annaka

, et al. Effectiveness of transcranial magnetic stimulation on executive function, attention, and memory in stroke patients: a systematic review and meta-analysis. Cureus 2024; 16: e75194.

16.

Xie

Luo

Xiong

, et al. Efficacy and safety of repetitive transcranial magnetic stimulation for poststroke memory disorder: a meta-analysis and systematic review. J Integr Neurosci 2023; 22: 131.

17.

Chang

. Clinical Study of Intermittent Theta Burst Stimulation in The Treatment of The Post-Stroke Cognitive Impairment . Master Thesis, Zhengzhou University, China, 2024.

18.

Luo

Dai

. Analysis of the efficacy of cognitive training combined with repetitive transcranial magnetic stimulation in the treatment of cognitive impairment after stroke. CSTJ (Medicine) 2024; 6: 77–80.

19.

Liu

Zhao

Yang

, et al. TMS Combined with synchronous working memory training on cognitive function of stroke patients in the elderly. Chin J Geriatr Cardiol Cerebrovasc Dis 2024; 26: 539–543.

20.

. Exploring the effects of repetitive transcranial magnetic stimulation at different frequencies on cognitive impairment after stroke. CSTJ Med Hyg 2024; 4: 76–79.

21.

Wan

Zhao

, et al. Evaluation on curative effect of different frequencies of rTMS on cognitive dysfunction after stroke based on magnetic resonance 3D-ASL. Chin J CT MRI 2023; 21: 23–26.

22.

Hong

. Efficacy of music therapy combined with transcranial magnetic stimulation on cognitive function and activities of daily living in patients with post-stroke cognitive impairment. Chin J Rehabil Med 2023; 38: 1745–1748.

23.

Chen

Wang

, et al. Study on the mechanism of high-frequency repetitive transcranial magnetic stimulation in improving memory function of stroke patients by functional magnetic resonance imaging. Chin J Magn Reson Imaging 2023; 14: 128–133.

24.

Gong

Wang

, et al. Effects of high-frequency repetitive transcranial magnetic stimulation combined with speech-auditory feedback training on event-related potential P300 and serum NSE and S100β protein in patients with post-stroke cognitive impairment. Prog Mod Biomed 2022; 22: 4541–4545.

25.

Shen

, et al. Effects of cognitive rehabilitation training combined with high frequency repetitive transcranial magnetic stimulation on cognitive function and serum BDNF, VEGF in patients with post-stroke cognitive impairment. Prog Mod Biomed 2022; 22: 482–485.

26.

Liu

. Clinical value of cognitive rehabilitation training combined with high-frequency repetitive transcranial magnetic stimulation in patients with post-stroke cognitive impairment. Chin Baby 2022; 27: 45–47.

27.

Huang

. rTMS combined with tDCS for memory dysfunction after stroke . Master Thesis, Tianjin University of Sport, China, 2021.

28.

Xia

. Effect of Auricular Acupuncture Combined with High Frequency Repetitive Transcranial Magnetic Stimulation on Cognitive Function of Stroke Patients . Master Thesis, Shanghai University of Traditional Chinese Medicine, China, 2021.

29.

Zheng

Xia

Duan

, et al. Repeated transcranial magnetic stimulation combined with donepezil can improve the cognition of cognitively impaired stroke survivors. Chin J Phys Med Rehabil 2020; 42: 32–36.

30.

Liu

Zhou

Song

. The effect evaluation of high-frequency repetitive transcranial magnetic stimulation combined with cognitive rehabilitation training in the treatment of cognitive impairment after stroke. Reflexology Rehabil Med 2020; 1: 129–132.

31.

Zhang

. Effects of repetitive transcranial magnetic stimulation on cognitive dysfunction in stroke patients. Rehabil Med 2019; 29: 20–26.

32.

Chen

Gong

, et al. The efficacy of repetitive transcranial magnetic stimulation combined with cognitive rehabilitation training in the treatment of post-stroke cognitive impairment. Chin J Phys Med Rehabil 2019; 41: 199–201.

33.

Bao

. Therapeutic effect of Wuling capsules combined with transcranial magnetic stimulator in the treatment of cognitive dysfunction during cerebral infarction recovery. Chin Mod Doct 2019; 57: 23–25.

34.

Yin

Luo

, et al. Effects of high frequency repetitive transcranial magnetic stimulation on post-stroke cognitive impairment. Chin J Rehabil Med 2018; 33: 763–769.

35.

Zhou

. Effect of repetitive transcranial magnetic stimulation on non-dementia vascular cognitive dysfunction. Chin J Phys Med Rehabil 2017; 39: 464–466.

36.

Zhou

. Effect of Repetitive Transcranial Magnetic Stimulation on prospective memory dysfunction after stroke . Master Thesis, Zhengzhou University, China, 2017.

37.

Wang

. Effect of Repetitive Transcranial Magnetic Stimulation on post-stroke cognitive impairment . Master Thesis, Zhengzhou University, China, 2016.

38.

Alexandrova

. Potential challenges to harmonize post-stroke cognitive assessment and its prognostic value: a narrative review. J Med Life 2024; 17: 963–977.

39.

Acharya

Grossman

Clifford

, et al. A chronic increase in blood-brain barrier permeability facilitates intraneuronal deposition of exogenous bloodborne amyloid-beta1-42 peptide in the brain and leads to Alzheimer's disease-relevant cognitive changes in a mouse model. J Alzheimers Dis 2024; 98: 163–186.

40.

Khan

Devlin

Urayama

, et al. Models and mechanisms of post-stroke dementia and cognitive impairment. Front Stroke 2025; 4: 1563924.

41.

Coco D

Lopez

Corrao

. Cognitive impairment and stroke in elderly patients. Vasc Health Risk Manag 2016; 12: 105–116.

42.

Danet

Barbeau

Eustache

, et al. Thalamic amnesia after infarct: the role of the mammillothalamic tract and mediodorsal nucleus. Neurology 2015; 85: 2107–2115.

43.

Godefroy

Roussel

Leclerc

, et al. Deficit of episodic memory: anatomy and related patterns in stroke patients. Eur Neurol 2009; 61: 223–229.

44.

Lim

Alexander

. Stroke and episodic memory disorders. Neuropsychologia 2009; 47: 3045–3058.

45.

Metzler-Baddeley

Jones

Belaroussi

, et al. Frontotemporal connections in episodic memory and aging: a diffusion MRI tractography study. J Neurosci 2011; 31: 13236–13245.

46.

O'Dwyer

Lamberton

Matura

, et al. Reduced hippocampal volume in healthy young ApoE4 carriers: an MRI study. PLoS One 2012; 7: e48895.

47.

Zhu

, et al. Research progress on the effect of transcranial magnetic stimulation on learning, memory and plasticity of brain synaptic. J Biomed Eng 2021; 38: 783–789.

48.

Chen

Zeng

Hong

, et al. Effects of HF-rTMS on microglial polarization and white matter integrity in rats with poststroke cognitive impairment. Behav Brain Res 2023; 439: 114242.

49.

Kim

Kroger

Calhoun

, et al. The role of the frontopolar cortex in manipulation of integrated information in working memory. Neurosci Lett 2015; 595: 25–29.

50.

Chai

Abd

Abdullah

. Working memory from the psychological and neurosciences perspectives: a review. Front Psychol 2018; 9: 401.

51.

Blumenfeld

Ranganath

. Dorsolateral prefrontal cortex promotes long-term memory formation through its role in working memory organization. J Neurosci 2006; 26: 916–925.

52.

Kehler

Francisco

Uehara

, et al. The effect of transcranial alternating current stimulation (tACS) on cognitive function in older adults with dementia. Annu Int Conf IEEE Eng Med Biol Soc 2020; 2020: 3649–3653.

53.

Turi

Mittner

Lehr

, et al. θ-γ cross-frequency transcranial alternating current stimulation over the trough impairs cognitive control. eNeuro 2020; 7: ENEURO.0126–20.2020.

54.

Wolinski

Cooper

Sauseng

, et al. The speed of parietal theta frequency drives visuospatial working memory capacity. PLoS Biol 2018; 16: e2005348.

55.

Bagherzadeh

Khorrami

Zarrindast

, et al. Repetitive transcranial magnetic stimulation of the dorsolateral prefrontal cortex enhances working memory. Exp Brain Res 2016; 234: 1807–1818.

56.

D'Esposito

Postle

. The cognitive neuroscience of working memory. Annu Rev Psychol 2015; 66: 115–142.

57.

Bigliassi

Filho

. Functional significance of the dorsolateral prefrontal cortex during exhaustive exercise. Biol Psychol 2022; 175: 108442.