Evidence Map Analysis on Randomized Controlled Trials: Chinese Patent Medicines and Classic Traditional Chinese Medicine Prescriptions

The study was supported by National Natural Science Foundation of China (82505806), the First Batch of Scientific and Technological Projects of Haihe Laboratory of Modern Traditional Chinese Medicine (25HHZYSS00023)

The Artificial Intelligence‐Clinical Evidence Database of Chinese Patent Medicine (AICED‐CPM, https://admin.tcmevd.com/index#/login ) [ 12 ] was searched. Supplementary searches were conducted in electronic databases including China National Knowledge Infrastructure (CNKI), Wan Fang Data, VIP, SinoMed, PubMed, and Web of Science. RCTs related to CPMs and CTCMPs published between January 1, 2024 and December 31, 2024 were included. Study type: RCT; participants: no restrictions on research subjects; intervention and comparison: the intervention measures in the experimental group include CPM/CTCMP, or CPM/CTCMP + other interventions, with no requirements on the dosage form of CPM/CTCMP. The intervention measures in the control group include CPM/CTCMP, other interventions, combinations of CPM/CTCMP + other interventions, blank control, or placebo; outcomes: no restrictions on evaluation indicators. The exclusion criteria are as follows: duplicated published studies; studies where the full text cannot be obtained or data are incomplete; studies where intervention measures are combined with acupuncture, acupoint application, physical exercise (such as Baduanjin, Tai Chi, and other sports), and surgery; conference papers and dissertations. Structured data related to the literature included in AICED‐CPM was exported by the system [ 13 ]. For literature from supplementary databases, two researchers independently reviewed records using NoteExpress and screened the literature according to pre‐specified inclusion and exclusion criteria. Any disagreements were resolved by discussion with a third researcher until consensus was reached. Before formal data processing, the two researchers involved in extraction received unified training and held detailed discussions on the definitions and standards for data extraction. Formal extraction was conducted independently by the two researchers strictly in accordance with a pre‐established extraction form. Extracted information included key items such as publication details, institution location, journal, sample size, disease type, interventions, outcome measures, and methodological quality assessment. After data extraction, inter‐rater reliability between the two reviewers was assessed using the irr package in R software (version 4.5.2). For categorical variables, Cohen's Kappa coefficient was calculated based on the initial independent extraction results to evaluate agreement between reviewers. Kappa values ≥0.75 were considered indicative of good agreement, values between 0.40 and 0.75 indicated moderate agreement, and values <0.40 indicated poor agreement. For continuous variables (e.g., sample size and impact factor), inter‐rater reliability was evaluated using the intraclass correlation coefficient (ICC) based on a two‐way mixed‐effects model with single measurements and absolute agreement. ICC values >0.75 were considered indicative of good agreement, values between 0.40 and 0.75 indicated moderate agreement, and values <0.40 indicated poor agreement. Any discrepancies between reviewers were resolved through discussion, and unresolved disagreements were adjudicated by a third reviewer until consensus was reached. The methodological quality of the included studies was assessed using the Cochrane Risk of Bias (RoB) assessment tool [ 14 , 15 ], which covers seven aspects: (1) method for generating the allocation sequence; (2) concealment of the allocation sequence; (3) blinding of participants; (4) blinding of outcome assessment; (5) completeness of outcome data; (6) selective reporting of study results; (7) other biases. The assessment results were classified into “low risk” “high risk,” or “unclear risk” [ 16 ]. For quantitative analysis, scores were assigned as follows: low risk = 3, unclear risk = 2, high risk = 1. The sum of scores across all domains was calculated as the overall RoB score, with a higher score indicating a lower risk of bias. All statistical analyses in this study were conducted separately for CPMs and CTCMPs, aiming to explore potential influencing factors of the methodological quality (RoB score) of the included studies. Data processing and statistical inference were performed using R software (version 4.5.2). Sample size was defined as a continuous variable. The impact factor (IF) of Chinese journals was obtained from the Chinese Academic Journal Impact Factor Annual Report (2025), and that of English journals was derived from the 2025 edition of Journal Citation Reports (JCR). Journals without a current IF were recorded as missing values, and studies were accordingly divided into two groups: “with IF” and “without IF.” Multicenter studies were classified as “yes” or “no” based on explicit reporting in the original articles. Funding support was categorized as “funded” or “no funding/unreported” according to whether funding was stated in the original text. Methodological quality was indicated by the overall RoB score (a continuous variable; a higher score represents better quality). Correlation analysis of continuous variables: Spearman's rank correlation was used to examine the associations between continuous variables (sample size, journal IF) and RoB scores. Given the presence of missing values in IF, this analysis was performed only among studies with available IF. The correlation coefficient ρ and p ‐value were reported. Between‐group comparison of categorical variables: The Wilcoxon rank‐sum test (Mann–Whitney U test) was applied to compare differences in RoB scores across groups of dichotomous variables (funding support, multicenter status, presence or absence of IF). Results were presented as median with the interquartile range (IQR), specifically as median (Q1, Q3), where Q1 represents the 25th percentile and Q3 represents the 75th percentile, along with the W statistic. Studies with missing journal IF were assigned to the “without IF” group and compared with the “with IF” group; no imputation was conducted for missing values. All tests were two‐sided, and p < 0.05 was considered statistically significant.

During the data extraction process, the two researchers demonstrated good to excellent agreement in extracting the characteristics of the included studies. Taking methodological quality assessment (RoB tool) as an example, the Cohen's Kappa value for the “randomization process” domain between the two researchers was 0.92 (95% CI: 0.89–0.95), indicating almost perfect agreement. For continuous variables, using sample size as an example, the ICC was 0.97 (95% CI: 0.95–0.98), suggesting extremely high reliability of data extraction. All discrepant items were resolved through discussion or third‐party adjudication to reach final consensus. After a comprehensive search of Chinese and English databases and subsequent screening, a total of 1874 articles were included. All of them were RCTs, comprising 1550 studies on Chinese proprietary medicines and 324 on classical Chinese medicine formulas. The literature selection process is illustrated in Figure  1 . Literature screening process. In 2024, a total of 1550 RCTs on clinical research of CPMs were published, among which 1497 were in Chinese and 53 were in English. The RCT studies on CPMs covered a total of 32 regions at home and abroad (23 provinces, 4 autonomous regions, 4 municipalities directly under the Central Government, 1 special administrative region, and 1 foreign country). Among these regions, Henan Province ranked first with 271 published papers, followed by Jiangxi Province (169 papers) and Shandong Province (142 papers). In addition, 2 RCT studies on CPMs were also conducted in Pakistan [ 17 , 18 ] (Figure  2 ). Geographical distribution of CPMs and CTCMPs RCT studies in 2024. The participating institutions in this research included various types, such as tertiary hospitals, secondary hospitals, primary hospitals (including township health centers), medical colleges and universities, scientific research institutions, and private hospitals. Among the Chinese literature, papers with tertiary hospitals as the main research institutions accounted for 73.28% (1097/1497); while in the English literature, all the first research institutions of the studies were tertiary hospitals, accounting for 100%. In 2024, a total of 324 clinical RCTs on CTCMP were published, including 320 in Chinese and 4 in English. The regions conducting RCT studies on CTCMP covered 27 domestic provinces. The top 3 regions in terms of publication quantity were Henan Province (35 papers), Shandong Province (33 papers), and Jiangsu Province (32 papers) (Figure  2 ). The research institutions included tertiary hospitals, secondary hospitals, primary hospitals (township health centers), medical colleges and universities, scientific research institutions, and private hospitals. In 2024, the published literature on CPMs RCTs involved 344 journals. There were 302 Chinese journals, among which 30 were included in the Peking University Chinese Core Journals, with 99 papers published. CPM had the highest number of published papers. The top 10 Chinese journals by publication output on CPMs are detailed in Table  1 . There were 42 English journals, 34 of which were included in SCI, with IF ranging from 0.2 to 168.9. The journal with the highest IF was Lancet (IF: 168.9, 1 paper) [ 19 ]. The publication status of English journals with an IF of 5 or higher is shown in Table  2 . Number of publication for CPM and CTCMP in Chinese journals (top 10). Publication status in English journals (referring to the 2024 IF, IF ≥ 5.0). In 2024, the published literature on RCTs of CTCMP involved 135 journals, including 132 Chinese journals and 3 English journals. Among the Chinese journals, 10 were included in the Peking University Core Journals, with 16 publications in total—Chinese Archives of Traditional Chinese Medicine had the highest number of publications (5 papers). Seven journals were included in the Chinese Science Citation Database (CSCD), with 8 publications in total, among which Journal of Pathogen Biology had the most publications (2 papers). The top 10 Chinese journals by publication output on CTCMPs are detailed in Table  1 . There were 3 English journals in total, all indexed in SCI, with IF ranging from 1.1 to 5.4. The journal with the highest IF was the Journal of Ethnopharmacology (IF = 5.4, 2 papers). According to statistics from the latest released ICD‐11 disease classification, the CPMs‐related RCTs published in 2024 involved 16 categories of diseases, among which diseases of the circulatory system (including cardiovascular‐related diseases) ranked first in the number of published papers (259 papers, 16.71%), followed by diseases of the respiratory system (249 papers, 16.06%) and diseases of the nervous system (including cerebrovascular‐related diseases) (185 papers, 11.94%); while the CTCMPs‐related RCTs published in 2024 involved 14 categories of diseases, with diseases of the digestive system taking the first place in the number of published papers (62 papers, 19.13%), followed by diseases of the respiratory system (53 papers, 16.35%) and diseases of the nervous system (including cerebrovascular‐related diseases) (53 papers, 16.35%). Specific details of top 10 are shown in Table  3 . Disease classifications of RCTs in 2024 based on ICD‐11 (Top 10). Statistics on the names of diseases involved in CPM RCTs in 2024 showed that among the top 5 disease systems by the number of published papers: in circulatory system (cardiovascular system) diseases, coronary heart disease had the largest number of publications (50 papers, 3.23%); in respiratory system diseases, chronic obstructive pulmonary disease ranked first (35 papers, 2.26%); in nervous system (cerebrovascular system) diseases, acute cerebral infarction had the highest number of papers (48 papers, 3.10%); in reproductive system diseases, uterine fibroids accounted for the most (13 papers, 0.84%); in digestive system diseases, chronic atrophic gastritis was the most studied (9 papers, 0.58%). Diseases with high research popularity (top 10 in terms of the number of published papers) were in the following order: diabetic nephropathy (53 papers, 3.42%), coronary heart disease (50 papers, 3.23%), acute cerebral infarction (48 papers, 3.10%), chronic obstructive pulmonary disease (35 papers, 2.26%), chronic heart failure (32 papers, 2.06%), type 2 diabetes mellitus (29 papers, 1.87%), Mycoplasma pneumoniae pneumonia (28 papers, 1.81%), coronary heart disease with angina pectoris (23 papers, 1.48%), cerebral infarction (20 papers, 1.29%), and acute myocardial infarction (18 papers, 1.16%). Compared with the research hotspots of diseases in 2023, diabetic nephropathy jumped from the fourth place to the first place (with its proportion increasing from 2.22% to 3.42%). Meanwhile, the number of papers on diseases such as acute cerebral infarction and Mycoplasma pneumoniae pneumonia increased significantly, while the number of papers on fractures and influenza decreased. For the annual disease distribution of CPM RCTs, see Figure  3 . Distribution of diseases in CPM RCTs in 2024. A statistical analysis was conducted on the disease names involved in the RCTs of CTCMP in 2024. Among digestive system diseases, chronic gastritis had the highest number of publications (8 papers, 2.47%); among respiratory system diseases, acute exacerbation of chronic obstructive pulmonary disease (AECOPD) ranked first in publication quantity (11 papers, 3.40%); among neurological (cerebrovascular system) diseases, hypertension had the most publications (17 papers, 5.25%); and among circulatory system (cardiovascular system) diseases, coronary heart disease led in publication quantity (15 papers, 4.63%). Diseases with high research attention (publication quantity > 5 papers) were in the following order: hypertension (17 papers, 5.25%), stroke (16 papers, 4.94%), coronary heart disease (15 papers, 4.63%), diabetic peripheral neuropathy (14 papers, 4.32%), acute exacerbation of chronic obstructive pulmonary disease (11 papers, 3.40%), heart failure (11 papers, 3.40%), vertigo (11 papers, 3.40%), chronic gastritis (8 papers, 2.47%), and Helicobacter pylori ‐associated gastritis (7 papers, 2.16%). Digestive system diseases and respiratory system diseases were the research hotspots. Only the top 2 diseases in each disease system are displayed (Figure  4 ). Distribution of diseases in CTCMP RCTs in 2024. The included RCTs involved 567 types of CPMs (501 types of oral CPMs, 51 types of TCM injections, and 15 types of external‐use CPMs). The top 10 oral CPMs in terms of the number of publications are as follows: Fufang Danshen Dripping Pill/Tablet (25 papers), Shexiang Baoxin Pill (25 papers), Guizhi Fuling Capsule/Pill (21 papers), Tongxinluo Capsule (20 papers), Shenqi Jiangtang Capsule/Granules/Tablet (19 papers), Shensong Yangxin Capsule (19 papers), Qishen Yiqi Dripping Pill (17 papers), Huangkui Capsule/Granules (16 papers), Xiao'er Chiqiao Qingre Granules (16 papers), and Xuefu Zhuyu Capsule/Oral Liquid/Pill (15 papers). The top 10 TCM injections in terms of the number of publications include Shenfu Injection (29 papers), Shenmai Injection (19 papers), Tanreqing Injection (18 papers), Danhong Injection (17 papers), Fufang Kushen Injection (16 papers), Shuxuening Injection (15 papers), Xiyanping Injection (14 papers), Xingnaojing Injection (14 papers), Danshen Injection (11 papers), and Xueshuantong Injection (10 papers). The top 5 external‐use CPMs in terms of the number of publications are as follows: Shirun Shaoshang Ointment (4 papers), Mayinglong Shexiang Zhichuang Ointment (3 papers), Jiedu Shengji Ointment (2 papers), Luofushan Fengshi Plaster (2 papers), and Compound Tongye Shaoshang Oil (2 papers). The included RCTs involved a total of 75 types of CTCMPs, mainly used for treating digestive system diseases and respiratory system diseases. Among the CTCMPs with relatively high publication quantities, the specific numbers were as follows: Banxia Xiexin Decoction (40 papers), Banxia Baizhu Tianma Decoction (30 papers), Taohong Siwu Decoction (21 papers), Huangqi Guizhi Wuwu Decoction (17 papers), Qingjin Huatan Decoction (14 papers), Sangbai Pi Decoction (12 papers), Zhenwu Decoction (12 papers), Gualou Xiebai Banxia Decoction (11 papers), Wendan Decoction (11 papers), Shenling Baizhu Powder (9 papers), Banxia Houpo Decoction (8 papers), Danggui Liuhuang Decoction (8 papers), Wenjing Decoction (8 papers), and Danggui Sini Decoction (7 papers). For the convenience of readers, the abbreviations used in this section and their meanings are listed in Table  4 . Abbreviations for intervention‐control settings. A total of 1550 RCTs were included, consisting of 1507 two‐arm trials and 43 multi‐arm trials (including 7 four‐arm trials and 36 three‐arm trials). The two‐arm trials involved 15 types of intervention‐control designs. The most common design was “CPM + Western medicine vs. Western medicine” (710 trials, 47.11%), followed by “CPM + Western medicine + routine treatment vs. Western medicine + routine treatment” (416 trials, 27.6%) and “CPM + routine treatment vs. routine treatment” (225 trials, 14.93%). Among the 50 English two‐arm RCTs, 33 studies used a placebo‐controlled design, and 1 study used a blank control [ 20 ]. The intervention‐control settings of two‐arm RCTs on CPMs in 2024 are illustrated in Figure  5 . CPM intervention‐control settings of two‐arm RCTs. C: conventional treatment; P: placebo; B: blank control; TP: CPM placebo; T: CPM; W: Western medicine. A total of 324 RCTs were included, comprising 319 two‐arm trials and 5 three‐arm trials. The two‐arm trials involved 12 types of intervention‐control designs, including: “CTCMP + Western medicine vs. Western medicine” (230 trials, 72.10%); “CTCMP vs. Western medicine” (41 trials, 12.85%); and “CTCMP + Conventional treatment vs. Conventional treatment” (20 trials, 6.27%). Among the 4 English‐language two‐arm RCTs, 2 studies employed a placebo control design. The intervention‐control settings of two‐arm RCTs on CTCMP in 2024 are illustrated in Figure  6 . CTCMP intervention‐control settings of two‐arm RCTs. C: conventional treatment; P: placebo; T: CPM; W: Western medicine; D: CTCMP; O: other treatments. In this study, the outcome indicators of 1550 CPM RCTs were systematically sorted and classified based on 7 core indicator domains (Physical and Chemical Examination, Symptoms and Signs, Traditional Chinese Medicine Symptoms/Syndromes, Quality of Life, Economic Evaluation, Safety Events, and Long‐term Prognosis). A total of 257 outcome indicators were included, with a cumulative application frequency of 13,139 times. Among them, there were 131 physical and chemical examination indicators, which was the core domain with the largest number of indicator types; 92 symptom and sign indicators, including only 3 traditional Chinese medicine symptom/syndrome indicators; 25 quality of life indicators; and the number of indicators in the domains of long‐term prognosis, safety events, and economic evaluation was 3, 2, and 1, respectively. Details are shown in Table  5 . The index domain situation of CPM. The outcome indicators of the 324 included CTCMP literatures were systematically sorted and classified in accordance with the aforementioned core indicator domains. The results showed that a total of 591 outcome indicators were covered, with a cumulative application frequency of 3272 times. Among them, symptom and sign indicators were the most prominent, totaling 329; there were 167 physical and chemical examination indicators; 85 quality of life indicators; 3 Traditional Chinese Medicine symptom/syndrome indicators, which were consistent in quantity with the indicators related to CPM; 3 long‐term prognosis indicators and 2 safety event indicators; and 2 economic evaluation indicators. Details are shown in Table  6 . The index domain situation of CTCMP. After deduplication and standardization of the outcome indicators of all included literatures ( n = 1874), a total of 848 outcome indicators were extracted, among which 169 were patient‐reported outcomes (PROs) [ 21 ], covering two domains: symptoms and signs (62 indicators) and quality of life (107 indicators). These PROs reflect patients' subjective evaluations of their own health status, functional status, and treatment experience. Details are shown in Table  7 . Patient‐reported outcome. A total of 103,519 patients were involved in 1550 RCTs, with the sample size of each study ranging from 15 to 3110 cases. Studies designed with a sample size of 40 cases were the most common (157 studies, 10.13%). There were 1446 studies with a sample size of ≤100 cases, accounting for 93.29%; 65 studies with a sample size ranging from 100 to 199 cases, accounting for 4.19%; and only 14 studies with a sample size of ≥500 cases, accounting for less than 1%. For studies with a sample size of ≥500 cases, see Table  8 . Varieties of CPMs and diseases involved in CPM RCTs with a sample size of ≥500 cases in 2024. n , number of RCTs. There were 81 multicenter RCTs (5.23%), involving 74 types of oral CPMs; the main types of diseases studied were respiratory system diseases, circulatory system diseases, and reproductive system diseases. The minimum number of centers was 2, and the maximum was 133. There were 13 studies indicating they were multicenter studies but did not specify the number of centers; the number of studies with collaboration between 2 centers was the largest (16 studies). For the CPMs and diseases involved in RCTs with more than 15 collaborating centers, see Table  9 . Distribution of CPMs and diseases involved in multicenter RCTs in 2024 (with ≥15 centers). n , number of RCTs. A total of 28,369 patients were enrolled in the 324 studies, with the sample size of each study ranging from 40 to 310 cases. Studies with a sample size of < 100 cases were the most common (257 studies, 79.32%), followed by studies with a sample size of 100–199 cases (65 studies, 20.06%) and studies with a sample size of 200–499 cases (1 study, 0.31%). There were no studies with a sample size of ≥ 500 cases. There were 3 multicenter studies in total (0.93%), each involving 2 centers. The details of the top 10 CTCMPs by sample size in 2024, along with the corresponding information of their RCTs, are provided in Table  10 . Top 10 CTCMPs by sample size in 2024 and the corresponding information of their RCTs. The duplicate entries in the CTCMP column are present because the table is sorted by sample size, and studies involving the same intervention (i.e., the same CTCMP) have different sample sizes. The methodological quality assessment results revealed significant variations in the risk of bias distribution among the 1550 included studies. Specifically, the random sequence generation methods demonstrated a high level of standardization, with a low‐risk rate of 77.0% (1194 studies). However, the allocation concealment aspect had serious flaws, as the medium/high‐risk rate reached 93.7%, and 82.5% of these studies were classified as medium risk. In terms of blinding implementation, the rates were generally low: the proportion of high risk in participant blinding was as high as 85.7% (1329 studies), and the medium‐risk proportion in outcome assessment blinding reached 82.3% (1276 studies), with only 11.3% of studies (175 studies) implementing this aspect in a standardized way. Additionally, data completeness and selective reporting performed well, with low‐risk rates of 86.0% and 97.8%, respectively, while other risks of bias were prominent, accounting for 85.7% (1329 studies, among which 923 did not report funding sources and trial registration). Methodological quality assessment revealed significant heterogeneity in the risk of bias among the 324 included studies. In terms of the randomization process, the standardization of sequence generation was relatively high (low risk rate: 94.1%, 305 studies), while the allocation concealment link had prominent flaws, with 56.2% of studies (182 studies) at moderate risk, and the total proportion of studies at moderate/high risk reaching 63.3%. The implementation of blinding was severely inadequate: the proportion of studies with high risk of participant blinding reached 67.9% (220 studies), and the issue of outcome assessment blinding was particularly critical, as the total proportion of studies at moderate/high risk was 91.7%, with a standard implementation rate of only 8.3% (27 studies). In contrast, the data reporting dimension performed well, with effective risk control in outcome data completeness (low risk: 97.5%, 316 studies) and selective reporting (low risk: 98.5%, 319 studies). Additionally, other risks of bias were significant, with 66.4% of studies (215 studies that did not report funding sources and trial registration) at high risk. Among the 1550 RCTs, 1062 studies (1009 in Chinese/53 in English) reported ethical approval, and 59 studies (9 in Chinese/50 in English) reported registration information. Among the 324 RCTs, 198 studies (194 in Chinese and 4 in English) reported ethical approval, and 3 studies (1 in Chinese and 2 in English) reported registration information, indicating that there is still a large gap in the compliance with international norms in the current clinical research on CTCMPs. To explore the associations between factors including sample size, presence and value of journal IF, funding support, and multicenter design with the methodological quality (expressed as the overall RoB score) of the included studies, Spearman's rank correlation analysis and the Wilcoxon rank‐sum test were performed separately. Spearman's correlation analysis revealed a significant but very weak positive correlation between sample size and RoB score among CPM RCTs ( ρ = 0.10, p < 0.001). Journal IF (based on 1234 studies with available IF) was significantly and positively correlated with the RoB score ( ρ = 0.28, p < 0.001), indicating that studies published in journals with higher IF had relatively better methodological quality (Table  11 ). Spearman's correlation analysis of sample size, journal IF, and ROB score in CPM/CTCMP RCTs. The Wilcoxon rank‐sum test demonstrated that funded studies ( n = 627) had a median RoB score of 16 (Q1, Q3: 15, 17), which was significantly higher than that of unfunded studies ( n = 923, median score 15 (14, 15)), with a statistically significant difference ( W = 139,698, p < 0.001). Multicenter studies ( n = 86) had a median RoB score of 18 (16, 20), which was significantly higher than that of single‐center studies ( n = 1464, 15 (15, 16); W = 20,761, p < 0.001). Although the median RoB score was 15 for both studies published in journals with IF ( n = 1234) and those without IF ( n = 316), the interquartile range was 15–16 in the IF group and 14–16 in the non‐IF group. The overall distribution differed significantly between the two groups ( W = 235,571, p < 0.001), suggesting that studies in journals with IF had relatively better methodological quality (Table  12 ). Wilcoxon rank‐sum test for the association of funding support, the number of centers with ROB score in CPM/CTCMP RCTs. Spearman's correlation analysis showed no significant correlation between sample size and RoB score in CTCMP RCTs ( ρ = 0.06, p = 0.266). Journal IF (based on 209 studies with available IF) was significantly and moderately positively correlated with the RoB score ( ρ = 0.37, p < 0.001) (Table  11 ). The Wilcoxon rank‐sum test indicated that funded studies ( n = 109) had a median RoB score of 17 (16, 18), which was significantly higher than that of unfunded studies ( n = 215, median score 15 (14, 16); W = 4187.5, p < 0.001). Only 3 multicenter studies were included, with a median score of 17 (16, 17.5), while single‐center studies ( n = 321) had a median score of 16 (15, 17). No significant difference was observed between the two groups ( W = 317, p = 0.300), which might be attributed to the extremely small sample size of multicenter studies. Studies published in journals with IF ( n = 209) had a median RoB score of 16 (15, 17), which was significantly higher than that of studies in journals without IF ( n = 115, median score 15 (14, 16); W = 15,938, p < 0.001), suggesting that studies published in journals with IF had relatively better methodological quality (Table  12 ). A horizontal comparison between CPM and CTCMP RCTs was performed regarding study scale, methodological quality, and influencing factors (Table  13 ). Comparative analysis of study scale, methodological quality, and influencing factors between CPM and CTCMP RCTs. p < 0.001. Study scale and design: CPM RCTs ( n = 1550) substantially outnumbered CTCMP RCTs ( n = 324). The median sample size of CPM RCTs was 92 (Q1, Q3: 80–110), significantly larger than that of CTCMP RCTs (81.50, Q1, Q3: 70–100; W = 297,006, p < 0.001). The proportion of multicenter studies was also significantly higher in CPM RCTs (5.55%, 86/1550) than in CTCMP RCTs (0.93%, 3/324; χ 2 = 11.66, p < 0.001). Methodological quality: CTCMP RCTs exhibited slightly better methodological quality, with a median RoB score of 16 (Q1, Q3: 15–17) compared to 15 (Q1, Q3: 15–16) in CPM RCTs ( W = 206,006, p < 0.001). The proportion of studies achieving RoB scores ≥18 was significantly higher in CTCMP RCTs (12.65%, 41/324) than in CPM RCTs (4.97%, 77/1550; χ 2 = 16.75, p < 0.001). Influencing factors: Journal IF showed a significant positive correlation with the RoB score in both groups, with a stronger correlation in the CTCMP RCTs ( ρ = 0.37, p < 0.001) than in the CPM RCTs ( ρ = 0.28, p < 0.001). Funding support significantly improved methodological quality in both types. Among CPM RCTs, funded studies had a median RoB score of 16 (Q1, Q3: 15–17) versus 15 (Q1, Q3: 14–15) for unfunded studies ( W = 139,698, p < 0.001). Among CTCMP RCTs, funded studies had a median score of 17 (Q1, Q3: 16–18) versus 15 (Q1, Q3: 14–16) for unfunded studies ( W = 4187.5, p < 0.001).

This evidence map systematically analyzed 1550 CPM RCTs and 324 CTCMP RCTs published in 2024. Both study types were characterized by small sample sizes (predominantly <100 cases) and a predominance of single‐center designs. Circulatory system diseases were most frequently studied in CPM trials, while digestive system diseases dominated CTCMP research. Regarding outcome selection, CPM RCTs favored physical and chemical examination indicators, whereas CTCMP RCTs prioritized symptoms and signs. A total of 169 patient‐reported outcomes were identified across both types, indicating emerging attention to patient‐centered endpoints. Substantial methodological shortcomings were observed in both categories. Although random sequence generation was generally adequate, serious deficiencies existed in allocation concealment and blinding implementation. Ethics approval reporting rates were moderate, and trial registration rates were alarmingly low (3.8% in CPM, 0.9% in CTCMP). Funding support, publication in high‐IF journals, and multicenter design were positively associated with better methodological quality. These findings reveal a considerable gap between current TCM RCT practices and international methodological standards, highlighting an urgent need to enhance trial design rigor, strengthen ethics and registration awareness, and align outcome selection with patient‐centered and real‐world evidence priorities.

This study systematically reviewed the research status of CPM ( n = 1550) and CTCMP ( n = 324) RCTs in 2024 and explored the associations between sample size, journal IF, funding, multicenter design, and ROB scores. The results showed that the global influence and research topics of traditional Chinese medicine studies have been continuously expanding, yet long‐standing issues regarding study design, methodological quality, and trial registration remain unresolved. Both types of studies exhibited a disconnect between quantity and quality, and these challenges limit the reliability, generalizability, and ultimate clinical application value of the research evidence. This paper provides empirical evidence for the strengths and weaknesses of current clinical research on traditional Chinese medicine, and offers recommendations for the optimization of future studies. A noteworthy trend is the continuous growth in the number of randomized controlled trials of CPMs, with several studies published in high‐impact international journals such as The Lancet and Nature . This marks a crucial step forward in the internationalization of TCM research and demonstrates that rigorously designed clinical trials of TCM can fully meet the publication standards of top international medical journals. In contrast, the output of RCTs on CTCMP remains stable, which may reflect the necessary transition from quantitative expansion to qualitative and in‐depth improvement for mature formulations (e.g., Banxia Xiexin Decoction). However, methodological flaws have severely hindered relevant progress, among which inadequate allocation concealment and blinding implementation are particularly critical, largely leading to overestimation of intervention effects and weakened reliability of conclusions. Especially for complex interventions such as CPM and CTCMP, clinical heterogeneity in components, dosage, dosage forms, and syndrome differentiation poses inherent challenges to standardized implementation and blinding (e.g., the preparation of matching placebos). Therefore, when interpreting methodological quality scores, it should be recognized that the high risk observed for some bias items (e.g., blinding) may arise from the contradiction between intervention characteristics and the generality of assessment tools, rather than simply from defects in study design. The systematic classification of outcome indicators reveals distinct emphasis between CPM and CTCMP RCTs. CPM studies predominantly employ physical and chemical examination indicators, reflecting a preference for objective laboratory measures. In contrast, CTCMP studies prioritize symptom and sign indicators, aligning with the syndrome‐based diagnostic tradition of classic formulas. Notably, both types underutilize TCM symptom/syndrome‐specific indicators, suggesting a potential gap in capturing intervention‐specific effects. The identification of 169 PROs across two domains—symptoms and quality of life—indicates growing attention to patient‐focused endpoints. However, the limited use of long‐term prognosis, safety events, and economic evaluation indicators highlights an opportunity to broaden the evidence base toward outcomes relevant to real‐world decision‐making. These findings underscore the need to align RCT endpoint selection with real‐world evidence priorities. Incorporating PROs and patient‐important outcomes can enhance the clinical relevance and generalizability of trial results. Furthermore, integrating pragmatic design elements—such as broader eligibility criteria and longer follow‐up—may bridge the gap between explanatory RCTs and real‐world practice, facilitating the translation of traditional Chinese medicine interventions into routine care. Future research should balance mechanistic and patient‐centered indicators to comprehensively evaluate therapeutic value. In addition, this study found a low registration rate among the included studies, a phenomenon closely related to policy implementation, research practice, and institutional coordination. Although the International Committee of Medical Journal Editors (ICMJE) has required prospective registration of clinical trials before participant enrollment since 2005 [ 51 ], some domestic journals have only made registration a mandatory requirement for submission in recent years, weakening the effect of external constraints. Furthermore, the registration of Chinese medicine compound interventions involves more complex information than chemical drugs, including herbal origin, preparation technology, and other components, which increases the burden on researchers. Previous studies have indicated that TCM RCTs have certain limitations in methodological design, such as blinding implementation and sample size calculation. Some researchers have prioritized limited efforts on solving the standardization of syndrome differentiation and treatment, while paying insufficient attention to the registration process. To address these issues, we recommend that trial registration be set as a prerequisite for ethical review, and that data linkage be promoted between the Chinese Clinical Trial Registry (ChiCTR) and the National Medical Products Administration platform. In this way, once researchers complete registration, the information can be queried on both platforms, reducing the cost of duplicate registration [ 52 ]. Journals should actively adopt the ICMJE registration policy and use the publication process to reversely drive standardization at the research stage. Meanwhile, specialized training on registration procedures should be provided for TCM researchers, and information reporting templates should be developed for complex interventions to systematically improve the transparency and standardization of clinical trials. In the analysis of factors influencing ROB scores, comparison between journals with and without an IF showed that studies with IF generally had higher quality, although the effect size differed between the two types of research. A positive association between journal IF and ROB scores was observed, which was largely consistent with previous findings, indicating that the peer‐review process of high‐impact journals selects for and promotes methodological rigor. High‐IF journals typically impose stricter requirements on study design, sample size calculation, blinding implementation, and other key aspects. However, a study suggested that journal IF explains only approximately 60% of the variance in literature quality, with the remaining 40% dependent on the scientific merit and clinical novelty of the research itself [ 53 ]. Thus, in the specific field of Traditional Chinese Medicine, high‐quality studies may also be published in some domestic core journals that have not yet obtained an IF. The association between funding support and methodological quality was prominent in both types of studies. This echoed the results of Lima et al., who reported that RCTs receiving industry funding were more likely to implement blinding appropriately and register in trial platforms. Hughes et al. further noted that the methodological advantages of funded studies may stem from more adequate resource support and stricter regulatory requirements [ 54 ]. Nevertheless, this association is not inevitable. A study by Ghozy et al. found that funding status did not significantly influence the methodological quality of systematic reviews (measured by AMSTAR‐2 scores), suggesting that the effect of funding may be study‐type specific [ 55 ]. The impact of multicenter design on methodological quality differed between CPM and CTCMP studies. The quality of multicenter CPM RCTs was significantly superior to that of single‐center studies. However, multicenter CTCMP studies were extremely rare, with only 3 included in the present study, accounting for less than 1% of the total, resulting in insufficient statistical power to assess their true effect. Multicenter studies generally involve more complex coordination, unified quality control, and centralized data management, which should theoretically help reduce the risk of bias. However, Goungounga et al. also pointed out that multicenter design may introduce confounding if inter‐center variability is not properly addressed [ 56 ]. The present findings indicate that a certain multicenter research foundation has been formed in the field of CPMs, whereas studies on classic TCM prescriptions urgently need to break away from the current status of single‐center and small‐sample designs. The correlation between sample size and methodological quality was very weak. A review of cluster‐randomized trials by Rutterford et al. found that only 55% of studies reported sample size calculation, and merely 22% reported all components completely [ 57 ]. Sample size itself is not equivalent to methodological quality. A comparative analysis of CPM and CTCMP RCTs showed that the average sample size of CPMs was significantly larger than that of CTCMPs. However, CTCMPs, with relatively smaller sample sizes, showed a slight advantage in overall methodological quality. Large‐sample studies with flawed designs (e.g., inappropriate randomization, lack of blinding) can still yield unreliable conclusions. Conversely, well‐designed small‐sample studies can provide high‐quality evidence. The present results support this view and emphasize that when evaluating research credibility, priority should be given to core methodological details such as randomization, allocation concealment, and blinding, rather than relying solely on sample size. This study prompts us to reflect on the limitations of the current evaluation paradigm in TCM research. Existing risk of bias assessment tools (e.g., the Cochrane ROB tool) focuses primarily on the internal validity of studies, yet methodologically rigorous trials are not equivalent to high clinical value [ 58 ]. Current evaluation frameworks often fail to adequately assess whether studies address key clinical questions, adopt comparators aligned with clinical practice, incorporate patient‐centered outcomes, and whether their findings possess real‐world translatability. Therefore, while improving methodological quality, there is an urgent need to establish a dedicated evaluation framework consistent with the characteristics of TCM. We propose that this framework integrates three core dimensions: First, methodological quality: on the basis of existing ROB tools, assign differentiated domain weights by incorporating the unique features of TCM interventions (e.g., blinding challenges due to sensory characteristics). Second, clinical value: include indicators such as PROs, TCM syndrome efficacy, and health economic evaluation to respond to the patient‐centered research orientation. Third, TCM characteristics: evaluate the standardization of syndrome differentiation and treatment, the consistency of intervention protocols with TCM theories, and the formula‐syndrome correspondence. At the indicator system level, consider the applicability of Core Outcome Sets (COS), the reliability and validity of PRO quantification, and the integration of real‐world evidence. For the implementation pathway, the integrated evidence chain methodology can be referenced to collaboratively evaluate evidence from clinical experience, experimental research, and clinical trials [ 59 ]. On this basis, research resources can be allocated to studies that truly advance patient‐focused drug development (PFDD) and reflect the advantages of the holistic view of TCM. This study has several methodological limitations. First, the search was restricted to Chinese and English databases, which may miss literature in Japanese, Korean, and other languages, affecting the comprehensive portrayal of the research status in East Asia. Second, equal weighting was used in calculating ROB scores (high risk = 1, unclear = 2, low risk = 3), without differentiating the impact of various bias types; for example, the importance of inadequate randomization may have been underestimated. Third, the cross‐sectional design can only reveal associations rather than confirm causality. Future studies may expand multilingual searches, optimize weighting in bias assessment, and adopt longitudinal designs for further validation. In summary, although the scale and scope of clinical research in TCM continue to expand, there is an urgent need to further enhance research quality and develop patient‐centered clinical research methodological frameworks and quality assessment tools aligned with clinical value. To bridge this gap, we propose the following recommendations: (1) For researchers: Priority should be given to conducting adequately powered, multicenter, confirmatory trials. Methodological rigor in randomization, allocation concealment, and blinding must be non‐negotiable, and prospective trial registration should become a universal standard. (2) For regulators and funders: Oversight of clinical research should be strengthened to evaluate not only methodological rigor but also the clinical value of proposed studies, guiding the field toward patient‐centered research with practical impact. (3) For the TCM academic community: There is an urgent need to develop and validate a specialized evaluation framework for TCM that can coexist with conventional standards, ensuring that the unique therapeutic concepts and outcome measures of TCM are fully assessed and reported. By simultaneously enhancing methodological robustness and clinical relevance, the field can shift from generating large quantities of data to producing reliable, credible, and meaningful evidence that truly validates the clinical value of CPMs and CTCMPs.

The concept of “preventive treatment of disease” (encompassing “preventing the occurrence of disease before its onset and preventing the deterioration or complication of disease during its course”) constitutes the core theoretical system of Traditional Chinese Medicine (TCM) [ 1 , 2 ]. It stresses proactive prevention before the occurrence of a disease and effective interruption of its progression after onset. The unique advantages of this concept have demonstrated crucial value in the prevention and treatment of various diseases [ 3 ]. In recent years, with the vigorous development of the cause of TCM, the clinical value of Chinese patent medicines (CPMs) and Classic Traditional Chinese Medicine Prescriptions (CTCMPs) has been increasingly prominent, and the number of relevant randomized controlled trials (RCTs) has shown a significant growth trend [ 4 , 5 , 6 , 7 ]. The emergence of a large amount of clinical evidence has provided evidence‐based medical support for verifying their efficacy and safety, as well as guiding clinical practice [ 8 ]. Therefore, continuously conducting systematic and rigorous evaluation studies, and obtaining and integrating high‐quality evidence‐based evidence are of great significance for ensuring the safety and effectiveness of patients' medication, and promoting the modernization and internationalization of TCM [ 9 , 10 , 11 ]. This chapter reports the current status of clinical research evidence (mainly RCTs) related to CPMs and CTCMPs published at home and abroad in 2024, so as to provide the latest and reliable evidence‐based basis for the clinical application of CPMs and CTCMPs. It also analyzes the common problems existing in current research and the direction of improvement, aiming to promote the continuous improvement of the quality of design and implementation of TCM clinical research.

The authors declare no conflicts of interest.