Compared with general medical software, traditional Chinese medicine (TCM) diagnosis and treatment software is a distinctive product category built upon TCM theories and possibly involving multi-dimensional indicators for TCM syndrome differentiation analysis. It is critically essential to establish a complete review-compliant evidence chain aligned with the product’s intended TCM clinical applications and algorithmic risk characteristics.
Compared with ordinary medical software, traditional Chinese medicine (TCM) diagnosis and treatment software is a special type of product. Constructed on models rooted in TCM theories, it may cover multi-dimensional indicators for TCM disease and syndrome analysis. It is crucial to build a full evidence chain complying with technical review requirements based on the product’s clinical TCM application scope and algorithm risk features.
Per existing published review frameworks, the NMPA and CMDE have issued guiding principles or key review points targeting AI-aided diagnosis, TCM four-diagnosis device software and auxiliary decision-making systems. Accordingly, the registration of TCM diagnosis and treatment software now focuses more on product definition, algorithm attributes and TCM clinical evidence. For teams conducting Class II/Class III medical device registration consultation and clinical evaluation, early integration of TCM theoretical logic and software engineering logic helps lower subsequent rework risks and relevant costs.
Many developers of TCM diagnosis and treatment software raise a core question upon project initiation: whether such products fall under Class II or Class III medical devices. A reliable classification cannot be given directly; instead, basic assessments shall be completed first to confirm core attributes: whether the product serves for auxiliary diagnosis or auxiliary treatment, whether it adopts rule-based reasoning or deep learning AI algorithms, whether it is filed as standalone software or a component of hardware systems, whether it involves TCM syndrome differentiation, whether it provides therapeutic regimen recommendations, and whether it interacts with patients’ private data. These determinations directly govern product classification and subsequent dossier compilation.
Published guiding principles embody differentiated review criteria. Distinctions exist in applicable regulatory documents, review priorities and dossier structures among standalone TCM knowledge base software, auxiliary decision-support systems, AI-powered TCM image analysis software, and embedded software bundled with integrated TCM medical devices. Misjudgment at this stage often leads to full restructuring of document logic rather than partial revisions. Therefore, priority at the preliminary phase lies in sorting out applicable clinical scenarios and confirming the filing mode (standalone software or system accessory), rather than drafting registration documents in advance.
Special attention shall be paid to AI-based TCM medical software: algorithm traceability and training data compliance constitute core review focuses, and relevant documents are recommended to be prepared in line with highest-risk specifications.
The core of technical review for TCM diagnosis and treatment software never rests solely on algorithm accuracy, but on whether the product can deliver sustained, stable and safe performance within its intended usage scenarios. Reviewers attach less importance to favorable accuracy figures than to how such figures are derived and whether the product bears genuine clinical value.
Traditional Chinese Medicine (TCM) diagnostic software usually adopts complex reasoning models or deep learning algorithms. The "black-box issue" is the most common concern raised during regulatory review. For decision-aiding software, if the algorithm outputs a diagnosis of "Heart-Spleen Qi Deficiency", it shall trace back the weighted contributing symptoms such as pale tongue, thin pulse and insomnia.
Pathological biopsy serves as the golden standard for Western medicine, whereas what counts as the golden standard for TCM? Expert consensus or reviews from senior veteran TCM physicians?
It is recommended to specify the reference standard as TCM physicians with associate senior professional title or above or consensus from an expert panel, together with elaboration on the formulation of such standard.
Numerous software products achieve outstanding accuracy under laboratory environments, yet their accuracy drops sharply in real clinical settings due to varying lighting conditions and tongue image acquisition devices.
The verification dataset of the product shall include real-world data collected from multiple clinical centers, various acquisition devices and diverse lighting environments. Sustained and stable performance, instead of peak performance under ideal lab conditions, is the valid evidence to verify product efficacy.
Clinical evaluation selection shall follow current regulatory frameworks rather than empirical judgment. The necessity of clinical trials is determined first by whether the product is covered under recommended equivalence pathways; for excluded products, developers further verify whether discrepancies against predicate devices can be fully demonstrated via non-clinical supporting data. The primary challenge for TCM software developers is not recognizing the availability of the equivalence pathway, but compiling sufficient discrepancy verification documents, algorithm validation records and TCM clinical evidence.
Per prevailing review logic, products with mature design, similar intended use to marketed predicates, and limited modifications confined to algorithm optimization, interface revision or incremental auxiliary functions qualify for the equivalence pathway provided all variances are substantiated by performance testing and software research data. In contrast, products adopting innovative TCM theoretical frameworks, brand-new core algorithms including original deep learning models, or expanded intended indications such as shifting from auxiliary diagnosis to adjunctive treatment require early feasibility assessment for formal medical device clinical trials when discrepancies cannot be fully justified by existing non-clinical data.
The following problems frequently occur in TCM diagnosis and treatment software projects. Rather than isolated deficiencies, they will jointly affect classification judgment, verification design and clinical evaluation route.
| Common Risk Items Prone to Supplementary Data Requests |
|---|
Directly equating the naming of "TCM Software" to a fixed administration category without judging based on intended use and clinical links (diagnosis/treatment/screening).The algorithm model is fully trained while data source, labeling criteria and data compliance verification are not scheduled synchronously.The IFU claims to "assist TCM practitioners in syndrome differentiation diagnosis" without supportive clinical trial data or equivalent product comparison evidence.Equivalent product comparison is only limited to functional level without item-by-item comparison on TCM theoretical framework, algorithm principle and differentiating items.Unsynchronized progress of R&D, verification, registration and quality system preparation resulting in messy software versions and conflicting conclusions in application documents at later stage. |
For projects planning registration of TCM diagnosis and treatment software, it is recommended to complete the following preparations in advance at minimum:
For projects lacking sufficient evidence on classification scope, system attributes or product differentiation, arrange pre-submission communication with review authorities at critical milestones. Consistent project logic formulated in early development effectively reduces risks of supplementary data requests and repeated revisions afterward.
The difficulties in the registration of traditional Chinese medicine diagnosis and treatment software lie in complex product boundaries, stringent requirements for evidence chains, and prominent systematic review characteristics. The accuracy of classification judgment, the closed-loop completion of software research and cybersecurity verification, and the rational selection of clinical evaluation routes determine the efficiency of project progress and the stability of the technical review stage.
The registration and application of TCM diagnosis and treatment software require enterprises to not only possess solid software engineering capabilities, but also deeply grasp the essence of TCM theories and medical device regulatory requirements. Enterprises should involve professional registration teams at the early stage of projects to fully clarify review requirements. Regulatory standards shall be taken as the benchmark throughout the whole project process to avoid registration failures caused by negligence of details in the later stage. From the perspective of review experts, applicants are encouraged to strengthen communication with professional registration teams and regulatory authorities, thoroughly understand relevant laws, regulations and guiding principles, and prepare application documents in a scientific and rigorous manner.
Deda Medical provides full-process supporting services covering preliminary direction judgment, research protocol design, clinical trial assistance and application document integration in the fields of medical device clinical trials, stability and effectiveness evaluation, registration document compilation and quality system docking.
Only by organically integrating original TCM thinking with modern technologies and regulatory standards can innovative products with outstanding safety, effectiveness and unique TCM characteristics be developed to pass technical review smoothly and contribute to the inheritance and innovative development of traditional Chinese medicine.
