Cognitive abnormalities are pervasive across a spectrum of neurological diseases and psychiatric disorders, substantially impairing learning capacity, social functioning, and overall quality of life. These deficits are often characterized by slow recovery trajectories and limited compensatory potential, posing significant challenges for clinical rehabilitation. While non-pharmacological interventions such as repetitive transcranial magnetic stimulation (rTMS) and cognitive training (CT) have been widely adopted, monotherapies exhibit inherent limitations. Specifically, rTMS effectively modulates cortical excitability and facilitates neuroplasticity but lacks precise functional specificity. Conversely, CT engages specific cognitive processes through repeated training, but it is constrained without sufficient cortical excitability or neural network involvement. Consequently, protocols combining rTMS with CT have emerged as a promising intervention tool. However, current clinical outcomes remain highly heterogeneous, and the synergistic neural mechanisms have not been systematically elucidated. Therefore, this study aims to explicate the underlying neural mechanisms and influential factors, as well as systematically synthesize recent advances in this combined intervention among populations with various neuro-psychiatric disorders. 
To achieve this objective, this study analyzed a broad range of recent empirical studies and mechanistic investigations regarding rTMS, CT, and their combined applications. We critically evaluated evidence across multiple dimensions, including neurophysiological mechanisms, clinical efficacy across different pathological phenotypes, and methodological variations in intervention protocols. We specifically dissected the synergistic pathways from three distinct perspectives: (1) The state-dependent modulation of brain, illustrating how one intervention modality alters the excitation/inhibition balance to prepare neural networks for the other modality; (2) Frequency-phase coupling, demonstrating the resonance between exogenous stimulation frequencies and endogenous neural oscillations; and (3) Neurometabolism and network reconfiguration, emphasizing long-term structural and functional connectivity changes. Furthermore, we assessed the influence of critical moderating variables, such as stimulation target localization, dose-response relationships, temporal sequencing of intervention modalities, and individual differences in baseline brain states.  
The synthesized findings from previous clinical trials indicate variations in therapeutic effects of this combined protocol in different clinical populations. In patients with Alzheimer’s disease and post-stroke cognitive impairment, the combination yields relatively consistent and robust improvements in global cognition and specific domains, such as memory and executive function. However, outcomes are significantly more variable in mild cognitive impairment (MCI) and severe psychiatric disorders. We highlight critical determinants of the synergistic efficacy from four aspects. First, outcomes depend on the precise alignment between the stimulation site and the functional network engaged by the cognitive task, with individualized functional localization showing potential superiority over anatomical coordinates. Second, the relationship between stimulation parameters (frequency, intensity) and cognitive gains is non-linear and subject to homeostatic regulation, where excessive cumulative dosage may reverse excitability states. Third, efficacy relies on the synchronization of rTMS with specific training phases, indicating that the training content should functionally recruit the specific neural circuit modulated by rTMS to avoid a mismatch effect regardless of adopting the simultaneous or sequential mode. Finally, treatment response is heavily moderated by individual differences, such as baseline brain states and demographic factors, especially in clinical populations.
To conclude, the combination of rTMS and CT represents a promising intervention model that integrates exogenous neuromodulation with endogenous cognitive drive. The synergy is not a simple additive effect but a dynamic interaction where rTMS provides the physiological capacity, and CT provides the functional direction. By optimizing this interaction, short-term neuromodulatory effects can be translated into sustained and transferable structural improvements, such as enhanced white matter integrity and optimized large-scale network efficiency. However, this field is still hindered by significant limitations, including small sample size, insufficient follow-up duration to detect long-term maintenance, heterogeneity in assessment tools, and a lack of standardized protocols. These methodological issues currently obscure the potential of combined treatment and contribute to the observed variability in clinical outcomes.  
Findings of this study provide crucial guidance for future research and clinical practice in cognitive rehabilitation. Methodologically, there is an urgent need to transition from small-scale exploratory trials to large-sample, multi-center, longitudinal follow-up designs to validate long-term efficacy and transferability. Ideally, future protocols should adopt reproducible principles for configuring synergistic parameters, moving away from the population-average approaches to personalized interventions based on individual neural connectivity and baseline states. We propose the development of a multilevel efficacy evaluation system that integrates behavioral performance with neurobiological biomarkers. By advancing mechanism-informed paradigms aligning stimulation timing and frequency with task-specific neural dynamics, researchers and clinicians can promote the precision and standardization of such combined intervention protocols. This shift is essential for ultimately delivering stable, functionally specific, and clinically meaningful cognitive rehabilitation outcomes for diverse patient populations.