Digital intelligent evidence-based medicine towards evidence-informed decision-making in Chinese medicine
Year: 2026
Type: article
Abstract: 1. Introduction For millennia, Chinese medicine (CM) has operated within a sophisticated system of personalized care. The core tenet of syndrome differentiation and treatment (Bian Zheng Lun Zhi) holds that therapeutic interventions must be tailored not merely to a disease entity, but to the specific pattern of disharmony (syndrome) manifesting in a unique individual at a particular moment in time.[1] This holistic perspective, which considers the patient’s constitution, environmental context, and dynamic physiological state, can be regarded as an early archetype of precision medicine. The ascendancy of evidence-based medicine (EBM) in the late 20th century established the randomized controlled trial (RCT) as the gold standard for clinical validity. While EBM successfully shifted clinical authority from subjective experience to objective empiricism, its methodological emphasis on internal validity—frequently achieved through strict inclusion criteria and standardized protocols—has inadvertently marginalized the complexity inherent in real-world practice.[2,3] For CM, this creates a tension: the standardization required by RCTs can lead to more rigid treatment protocols, potentially underrepresenting the adaptive and individualized nature of CM in routine clinical practice.[4] Consequently, an efficacy-effectiveness gap has emerged. High-quality evidence in the CM field remains scarce—not necessarily due to a lack of therapeutic efficacy, but because prevailing evaluation tools are often ill-suited to the complex nature of CM interventions.[5] Traditional EBM generates “average” evidence for the “average” patient, which contrasts sharply with CM’s emphasis on individual variability and multifaceted clinical presentations. To bridge this gap, a paradigm shift may be required—one that accommodates complexity rather than reducing it excessively. Digital intelligent evidence-based medicine (i-EBM), an approach we have recently proposed, is characterized by the integration of multisource data, intelligent evidence processing, and individualized decision support (Fig. 1). This framework may serve as a critical enabler for the modernization of CM.[6]Figure 1.: Conceptual framework of digital intelligent evidence-based medicine (i-EBM). NLP, Natural Language Processing; CDSS, Clinical Decision Support System2. Limitations of traditional EBM for CM The constraints of traditional EBM in the context of CM are often structural. The “evidence pyramid,” which places systematic reviews of RCTs at its apex, implicitly assumes that the most valid knowledge is derived from controlled, homogeneous environments. In clinical CM practice, interventions are inherently multifaceted. A herbal prescription typically comprises multiple ingredients and is modified (Jia Jian) in response to the patient’s evolving symptomatology. Standardized RCT protocols, which often preclude such flexibility, may evaluate only a relatively static approximation of CM rather than its full, adaptive practice. Furthermore, the evidence lifecycle suffers from prohibitive latency, with significant time lags between research completion and guideline implementation.[7] Additionally, the data underpinning CM are highly heterogeneous, encompassing ancient classical texts, historical case records, modern electronic medical records (EMRs), and omics datasets.[2] Traditional EBM has limited capacity to synthesize these disparate data modalities, often prioritizing published literature while underutilizing real-world data that capture the nuances of individualized care.[8] This fragmentation can contribute to evidence silos, in which clinical insights on specific CM syndromes remain insufficiently linked to contemporary biomedical findings concerning related physiological states. 3. Integration: reconstructing the holistic view with computable knowledge The first pillar of i-EBM, integration, seeks to address the fragmentation of CM knowledge by broadening the definition of evidence. Within the i-EBM framework, evidence encompasses a pan-data ecosystem that includes EMRs, genomic data, and outputs from wearable devices, rather than being limited to published literature.[9,10] Within this pillar, knowledge graph technology can serve as a foundational architecture.[11] Unlike traditional relational databases, knowledge graphs map concepts within a semantic entity-relationship-entity network, enabling a computable representation of CM’s holistic worldview.[12] This approach aligns conceptually with the principles of network medicine,[13] which views disease not as a localized defect but as a perturbation of complex biological networks. By ingesting and synthesizing large collections of ancient medical texts alongside modern clinical data and pharmacological databases, i-EBM may facilitate the semantic alignment of CM concepts (eg, kidney Yang deficiency) with contemporary biomedical phenotypes (eg, neuroendocrine or metabolic features).[14] This integrative capability can help bridge the epistemological divide between classical knowledge and modern science. The construction of a syndrome-disease-formula-target-pathway network allows an i-EBM system to present a more comprehensive evidentiary landscape. Upon diagnosis, the system can retrieve analogous historical cases, relevant biological pathways, and potential drug-herb interactions. Such functionality supports CM’s holistic perspective by providing a global view of data, ensuring that individualized clinical decisions are informed by a synthesis of all available knowledge. 4. Intelligence: validating complex interventions via causal artificial intelligence (AI) The second pillar, intelligence, seeks to address the methodological challenges associated with evaluating complex CM interventions. To complement the rigidity of traditional RCTs, i-EBM can leverage causal AI to advance evidence generation from passive retrieval toward more computable, data-driven approaches.[15] By utilizing advanced machine learning and causal inference techniques, i-EBM can mine large-scale real-world observational data to support target trial emulation.[16] This methodology enables researchers to replicate key elements of randomized trials in silico using existing patient datasets. For instance, to evaluate the efficacy of a specific herbal formula on a subgroup of stroke patients exhibiting “wind-phlegm stasis,” i-EBM can computationally identify a treated cohort and a matched control group from EMRs, adjusting for high-dimensional confounding variables that conventional statistical methods might overlook.[17] Beyond causal inference, large language models can further strengthen this pillar by synthesizing unstructured clinical texts and literature, as well as identifying patterns that may be difficult to capture through manual review alone.[18,19] For CM, this capability is potentially transformative, enabling more timely and scalable evaluation of individualized treatments without necessitating RCTs for every formula variation. By supporting the causal evaluation of complex CM interventions and their clinical outcomes, i-EBM can contribute to a more robust and comprehensive evidence base for the field. 5. Individualization: digital renaissance of Bian Zheng Lun Zhi The ultimate objective of i-EBM is individualization, a goal that aligns closely with the core philosophy of CM. Whereas traditional EBM often produces guidelines applicable to the “average” patient, i-EBM is designed to support precise, individualized clinical decision-making for each specific patient.[20] In the context of CM, i-EBM can function as a cognitive extension for the physician. By analyzing multimodal data—ranging from traditional “four diagnostic” information (eg, tongue and pulse analysis) to modern omics profiles—i-EBM can generate high-dimensional predictive models. These models can forecast the trajectory of a specific syndrome or predict the likely response to a particular herbal combination.[21] This application can be viewed as a digital evolution of Bian Zheng. Whereas a human practitioner relies on memory and experience to differentiate syndromes, an i-EBM system leverages probabilistic analyses of large numbers of similar clinical trajectories. It does not replace the physician’s judgment; rather, it augments it with computational intuition. For example, in managing complex chronic conditions such as diabetic kidney disease, an i-EBM system can analyze a patient’s unique constitution and biomarkers to recommend precise modifications of a classical formula. This approach may help anticipate both efficacy and potential adverse events more systematically than empirical observation alone, thereby supporting more precise CM care informed by computable, data-driven evidence.[22] 6. Human–AI symbiosis in clinical practice The integration of i-EBM into CM may facilitate a more sophisticated form of human–AI collaboration.[23] The future CM clinic is envisioned not as algorithm-driven, but as physician-led and evidence-empowered. A cognitive division of labor may emerge, wherein AI handles high-throughput processing of heterogeneous data and probabilistic computations, while human practitioners provide ethical reasoning, contextual understanding, and compassionate care.[24] This transition necessitates a rigorous emphasis on data governance. To fully realize the potential of i-EBM, the CM community must address challenges such as algorithmic bias, which can arise if training datasets lack diversity.[25] Furthermore, robust data privacy frameworks, such as federated learning, can be implemented to protect patient information while facilitating cross-institutional collaboration.[26] 7. Conclusions i-EBM represents more than a technological advancement; it may also constitute an epistemological shift that helps bridge the divide between the art of CM and the science of modern evidence. By integrating a broader spectrum of medical data, employing causal inference to evaluate complex interventions, and prioritizing the individual alongside population-level evidence, i-EBM provides a methodological framework that aligns with the holistic, personalized nature of CM. These advantages are not limited to CM: i-EBM also holds promise for medicine more broadly, supporting evidence-based, individualized care for aging populations with increasing multimorbidity and clinical complexity. Overall, i-EBM represents a promising pathway for the modernization of CM. It can enhance the evaluation of Bian Zheng Lun Zhi, not by forcing it into the mold of reductionist science, but by advancing methodological approaches that accommodate individual nuance. In this way, i-EBM supports CM practitioners in making decisions that draw on classical knowledge while being informed by modern computational methods. More broadly, i-EBM may strengthen the role of CM and other traditional medicine systems globally by providing a transparent, computable scientific foundation, thereby expanding safe and tolerable complementary therapeutic options. This evolution points toward a more integrated, intelligent, and individualized evidence ecosystem. Statement of ethics None declared. Conflict of interest statement The authors declare that they have no conflicts of interest with regard to the content of this report. Funding source This study was funded by the Traditional Chinese Medicine Innovation Team and Talent Support Program—National Traditional Chinese Medicine Multidisciplinary Cross-Innovation Team Project (ZYYCXTD-D-202401). Data availability statement All data generated or analyzed during this study are included in this published article. Author contributions Honghao Lai and Long Ge conceived the study and developed the conceptual framework of Digital Intelligent Evidence-based Medicine. Honghao Lai performed the literature search and drafted the original manuscript. Janne Estill and Long Ge provided critical review and significantly revised the manuscript for intellectual content. Long Ge supervised the study and acquired the funding. All authors have read and approved the final manuscript.
Authors: Honghao Lai, Long Ge, Janne Estill
Institutions: Lanzhou University, University of Geneva
Language: English
Subfield: Complementary and alternative medicine
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7130516105