We performed our protocol in accordance with PRISMA-P (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols) guidelines [31] (Multimedia Appendix 1). This study was registered with PROSPERO (CRD42023475178).

Studies using fMRI to investigate the brain mechanism of depression based on TCM syndromes are included. All TCM syndrome subtypes must follow the classification and codes of diseases and patterns of TCM (GB/t 15657-2021) and be classified into DP and EP. Depression is diagnosed based on the standard diagnostic criteria, including the Diagnostic and Statistical Manual of Mental Disorders, the International Classification of Diseases, and the Chinese Classification of Mental Disorders. TCM syndrome differentiation follows the criteria for the diagnosis and the therapeutic effects of diseases and syndromes in TCM. Studies that meet the following criteria are included in the meta-analysis: (1) used an fMRI voxel-wise whole-brain approach of resting-state activation data; (2) reported altered spontaneous functional activities in local brain regions; and (3) reported brain regions with abnormal activations as x/y/z coordinates in standard space: Montreal Neurological Institute template or Talairach atlas.

The following types of studies are excluded: (1) reviews/meta-analyses/single-case studies/dissertations/conference papers/abstracts; (2) letter/comments/viewpoints/summaries for patients/editorials and opinion papers; (3) no assessment of TCM syndromes as exposure; (4) exposure to any active TCM interventions and non-TCM ones such as acupuncture, herbal medicine, moxibustion, massage, psychotherapy, pharmacological treatments, or brain stimulation techniques; (5) animal studies; and (6) duplicate publications of the same data.

Every study obtained from the search will be loaded into EndNote, and any duplicates will be eliminated. Two authors (LC and XJ) will independently review the titles, abstracts, and keywords of these studies based on the predefined criteria for inclusion and exclusion. Following that, the studies selected will undergo a second screening process by thorough reading of the complete text. Studies that do not match the specified review criteria will be excluded. If there is a disagreement, a third author (HL) will be responsible for making the ultimate decision.

Data extraction will also be performed by 2 independent reviewers (LC and XJ). We will design a structured data extraction form based on the fMRI study reporting criteria in order to ensure standardization in the process of extracting data. Any ambiguity or disagreement will be resolved by consulting a third investigator (HL). The following data will be extracted: first author, publication year, study type, sample size per group, age, gender, drug history, diagnosis criteria, ratings of depression and anxiety, modality and metrics of fMRI, type of task performed during functional imaging, fMRI device type, sequence parameters, software for fMRI data analysis, quality score, and reported results. Moreover, the significance level and the peak coordinates with its 2-sided t value or z score or P value will be extracted for the meta-analysis.

We will use the Agency for Healthcare Research and Quality scale to evaluate the quality of the clinical research based on the specific types of studies included [32]. Additionally, an fMRI checklist will be used to access the imaging methodology quality. This checklist involves the inclusion and exclusion criteria, population characteristics, diagnosis criteria, sample size, methods for image acquisition and analysis, and the quality of reporting of statistical results and research conclusions [33]. Similarly, the evaluation of quality will be conducted independently by 2 reviewers (LC and XJ). A third researcher (HL) will resolve any disagreement that arises during the evaluation procedure. We will categorize the literature into high, medium, and low quality according to their scores.

The synthesis of data from the selected studies will involve a combination of quantitative and qualitative methods. We will systematically summarize the clinical and methodological characteristics of the studies through tables and narratives, detailing imaging techniques, statistical approaches, TCM syndromes, and fMRI biological substrates associated with each syndrome. To improve data clarity, we concentrate on the primary DP and EP within the framework of syndrome subtypes methodology as applied in TCM clinical practice. The qualitative synthesis will follow a thematic content analysis approach, using a structured data extraction matrix to categorize imaging findings by brain region, activation direction (increased/decreased), and syndrome type (DP vs EP). Findings will be coded and grouped into emergent neurofunctional patterns that characterize each TCM subtype. The quantitative investigation will assess and contrast the neuroimaging characteristics among the 3 subgroups. We will also perform a direct comparison between the DP and EP groups to identify specific neural activation differences associated with TCM-defined subtype distinctions. Three sets of comparisons will be conducted in the meta-analysis: (1) DP versus healthy control (HC), (2) EP versus HC, and (3) DP versus EP.

The meta-analysis of resting-state fMRI studies will be conducted in Montreal Neurological Institute space using Seed-based d Mapping-Permutation of subject images software (version 6.21). For each comparison (DP vs HC, EP vs HC, and DP vs EP), we will generate effect size maps based on reported peak coordinates and statistical values. Quantitative outcomes of interest will include fractional amplitude of low-frequency fluctuations, amplitude of low-frequency fluctuations, regional homogeneity, and blood-oxygen-level–dependent signal changes in specific brain regions. Where available, other neuroimaging parameters such as functional connectivity measures may also be extracted and synthesized. We will conduct a comparison of brain activities between the DP group and the HC group, as well as between the EP group and the HC group. First, we will prepare the text files containing peak coordinates and effect dimensions. For data preprocessing, we will use those files to determine the resulting values of the minimum and maximum potential effect sizes for each voxel. Second, the mean images will be created by averaging the random effects datasets, considering the sample size, study variance, and variations between groups. The effect size maps from various studies will undergo multiple imputations. Third, a family-wise error correction will be used to account for multiple comparisons. If the number of eligible studies permits, we will also conduct subgroup analyses or meta-regression to examine potential moderators (eg, age, sex, imaging parameters).

This study protocol proposes a systematic review and meta-analysis to explore the neural mechanisms of TCM-based subtypes of MDD, specifically focusing on DPs and EPs. By applying both qualitative synthesis and quantitative meta-analytic techniques to resting-state fMRI studies, this protocol is designed to assess whether distinct brain activity patterns exist between TCM-defined subtypes. The significant heterogeneity of MDD has historically influenced its diagnosis and therapy [34]. Despite the various classification approaches for MDD, the neurobiological basis of MDD remains poorly understood, and there is still a dearth of compelling evidence for the clinical application value of MDD subtypes in real clinical settings [12]. Although the syndrome-based classification model of TCM (corresponding to the TCM syndrome differentiation and treatment system) is a clinically efficient categorization tool, the connection between this syndrome-based classification model of TCM and the neurological substrate requires further investigation. Our study is the first investigation to comprehensively clarify the neural mechanisms of depression’s neurobiological subtypes from the perspective of TCM, employing both quantitative and qualitative approaches.

Previous studies have focused extensively on Western medicine–based subtyping approaches to MDD, including symptom clusters, genetic polymorphisms, and machine learning–defined neurobiological patterns [35,36]. However, the clinical applicability of these subtypes has remained limited due to replication challenges and inconsistent outcomes [7]. The TCM syndrome classification offers a complementary, theory-driven alternative, with some preliminary studies suggesting that fMRI markers may differ between TCM-defined subtypes [37,38]. This protocol builds upon such observations and aims to systematically synthesize the evidence, which has not yet been undertaken comprehensively.

The primary strength of this protocol is its novelty in applying neuroimaging meta-analytic methods to examine brain-based correlations of TCM-defined subtypes of MDD—a field that remains underexplored. In addition, this study follows rigorous methodological frameworks (eg, PRISMA-P, PROSPERO registration) and integrates multiple databases and analytic strategies. Nonetheless, some limitations must be acknowledged. The heterogeneity among the included studies in terms of fMRI parameters, diagnostic criteria, and syndrome classification methods may affect the consistency of the results. The inclusion of non–peer-reviewed literature such as dissertations may also introduce variability in the study quality. Moreover, the current lack of standardized diagnostic criteria for TCM patterns in MDD may challenge cross-study comparability.

Upon completion, this review is expected to offer a neurobiological perspective on the TCM framework and may help identify reliable imaging biomarkers for MDD subtype differentiation. Future research should aim to develop standardized criteria for TCM syndrome diagnosis that are compatible with neuroimaging methodologies and increase sample sizes through multicenter collaboration. Additionally, further prospective clinical studies could validate whether neuroimaging-defined TCM subtypes predict differential treatment responses.