A total of 11,369 records were identified through literature search. After duplicates removal, 7497 records were screened through titles and abstracts, and 206 reports were obtained for full-text screening. Two authors (EM, ST) independently evaluated the 206 selected reports for in-depth examination. Disagreement concerned three reports (inter-rater agreement: 97%) and was solved by discussion. Thirty-four reports fulfilled the inclusion criteria and were therefore included in the systematic review; fourteen (Makkos et al. 2016; Shin et al. 2016; Brys et al. 2016; Trung et al. 2019; Khedr et al. 2020; Zhuang et al. 2020; Lang et al. 2020; He et al. 2021; Wei et al. 2022; Jiang et al. 2023; Song et al. 2024; Wu et al. 2024; Zhang et al. 2025; Feng et al. 2025) of them were also included in the meta-analysis (Fig. 1).

PRISMA diagram of the study. Page et al. (2021); www.prisma-statement.org

Studies were grouped according to the NIBS technique used (i.e., rTMS: N = 27; tDCS: N = 7) and for the rTMS protocol (i.e., excitatory: N = 18; inhibitory: N = 6; different protocols: N = 3). No studies on tACS, tRNS, LIFU or tPS were found.

Twenty-seven reports explored rTMS protocols for cognitive and neuropsychiatric NMS (Okabe et al. 2003; Hamada et al. 2008; Arias et al. 2010; Pal et al. 2010; Benninger et al. 2012; NCT00955032 ReStore 2012; Shirota et al. 2013; Makkos et al. 2016; Shin et al. 2016; Brys et al. 2016; Buard et al. 2018; Cohen et al. 2018; Trung et al. 2019; Khedr et al. 2020, 2024; Zhuang et al. 2020; Li et al. 2020; Lang et al. 2020; He et al. 2021; Pan et al. 2022; Wei et al. 2022; Jiang et al. 2023; Song et al. 2024; Barboza et al. 2024; Wu et al. 2024; Zhang et al. 2025; Feng et al. 2025) (Table 1). All of them included patients with PD, except one on MSA (Pan et al. 2022). Twenty studies used excitatory rTMS protocols (HF, N = 15; intermittent TBS, iTBS, N = 3) to target the left (N = 8), left mid (N = 1) or bilateral (N = 1) DLPFC (Pal et al. 2010; Shin et al. 2016; Trung et al. 2019; Lang et al. 2020; He et al. 2021; Pan et al. 2022; Wei et al. 2022; Jiang et al. 2023), bilateral (N = 4) or unilateral M1 (N = 1) (Benninger et al. 2012; Makkos et al. 2016; Khedr et al. 2020; Li et al. 2020; Song et al. 2024), bilateral SMA (N = 1) (Hamada et al. 2008), bilateral parietal areas (N = 1) (Khedr et al. 2024), right posterior insula (N = 1) (Barboza et al. 2024), while two studies compared left DLPFC, bilateral M1 and their combined stimulation (Brys et al. 2016) and bilateral M1, SMA and their combined stimulation (Zhang et al. 2025). Six studies used inhibitory LF rTMS of the right DLPFC (N = 3) (Zhuang et al. 2020; Wu et al. 2024; Feng et al. 2025), bilateral M1 (N = 2) (Okabe et al. 2003; Arias et al. 2010), M1 and the prefrontal cortex (PFC, N = 1) (Buard et al. 2018). A single study compared HF and LF rTMS of the SMA (Shirota et al. 2013). There was a high heterogeneity of rTMS sessions across studies (i.e., 5–24). Tables S1 and S2 report clinical data of the PD and AP populations.

Cognitive outcomes were explored in 17 studies (primary, N = 3; secondary, N = 12; not specified, N = 2). Measures of global cognition (e.g., MoCA/MMSE) improved to real vs sham rTMS in 3 out of 12 studies (left DLPFC HF, N = 0/3 Pal et al. 2010; Wei et al. 2022; Jiang et al. 2023); bilateral M1 HF, N = 1/3 (Makkos et al. 2016; Khedr et al. 2020; Song et al. 2024); right posterior insula, N = 0/1 (Barboza et al. 2024); left DLPFC, bilateral M1, combined stimulation, N = 0/1 (Brys et al. 2016); left DLPFC iTBS, N = 1/2 (Trung et al. 2019; He et al. 2021); right DLPFC LF, N = 2/2 (Zhuang et al. 2020; Wu et al. 2024), with effects after right DLPFC LF lasting up to 3–6 mos. Several cognitive domains were explored as secondary outcomes in 10 studies (left/bilateral DLPFC HF rTMS, N = 4; Pal et al. 2010; Buard et al. 2018; He et al. 2021; Wei et al. 2022; bilateral M1 HF rTMS, N = 3; Benninger et al. 2012; Makkos et al. 2016; Khedr et al. 2020; left DLPFC iTBS, N = 2; Trung et al. 2019; Lang et al. 2020; combined M1 and PFC LF rTMS, N = 1; Cohen et al. 2018), with attention and executive function domains (Stroop test, Wisconsin card sorting test, attention network test) being improved after real vs sham treatment in 2 studies (Pal et al. 2010; Wei et al. 2022) and the effect on Stroop test lasting up to 1 month follow-up in a single study (Pal et al. 2010).

Depression was the most common NMS, being explored as a primary/secondary outcome in 24 studies. Three out of 6 RCTs found significant improvement in measures of depression up to 1 month follow-up after real vs sham HF rTMS of the left DLPFC (Pal et al. 2010; Shin et al. 2016; Jiang et al. 2023). Three out of 4 studies reported real vs sham HF fTMS over bilateral (N = 2) (Makkos et al. 2016; Song et al. 2024) or unilateral M1 (N = 1) (Li et al. 2020) to be effective on depression up to 1 month after the treatment. Real HF rTMS on bilateral parietal areas was found to improve depression vs sham with no follow-up data (Khedr et al. 2024). Two out of 4 studies found improvement in depression up to 3 months follow-up after real vs sham LF rTMS of the right DLPFC (Zhuang et al.