The calculation of arithmetic mean according to the following Formula 1: M j = 1 m j ∑ i = 1 m j C ij (1)

M_j represents the arithmetic mean of all collected evaluation scores for each indicator j (j = 1, 2, …, n), calculated across all expert responses. C_ij denotes the score assigned by expert i to indicator j, and m_j refers to the number of experts who provided identical ratings for a given indicator. The value of M_j is computed as the sum of all C_ij values divided by the total number of evaluations. A higher M_j value indicates greater relative importance of the corresponding indicator in the assessment framework.

The calculation of frequency of maximum score according to the following Formula 2: K j = m j ′ / m j (2)

The ratio of the number of experts who assigned full scores to the importance and operability of a given indicator to the total number of experts evaluating that indicator. A higher full-score frequency indicates a greater proportion of experts assigning full scores, reflecting stronger consensus on the indicator’s significance.

The calculation of coefficient of variation according to the following Formula 3: V j = δ j / M j (3)

Let δ denote the standard deviation of the scores for a given indicator, and let M_j represent the arithmetic mean of those scores. The coefficient of variation, calculated as δ_j/M_j, serves as a measure of dispersion across expert evaluations. A smaller coefficient of variation indicates a higher level of consensus among experts, reflecting greater consistency and coordination in their assessments.

The weight coefficient for each criterion was calculated using the percentage weighting method in order of criterion ranking, according to the following Formula 4: K j = ∑ w i N i N ∑ w i (4)

K_j denotes the percentage weight of a given indicator. W_i represents the weighting coefficient corresponding to the importance ranking of an indicator; specifically, the 10 indicators selected in this study are ranked in descending order of importance, with the top-ranked indicator assigned a value of 10 and the lowest-ranked a value of 1. N_i refers to the frequency with which an indicator is assigned to a particular rank position. N denotes the total number of validly recovered questionnaires.

Finally, to facilitate the summation of index values with different dimensions, the optimal and worst expected values for each criterion were set based on actual clinical scenarios.

A total of 22 questionnaires were collected in the Delphi survey, yielding an effective response rate of 73.3% (22/30), which reflected a satisfactory level of expert engagement and confirmed that the number of valid responses met the methodological requirements for the study. All participating experts possessed professional backgrounds in medicine or pharmacy. Medical experts included specialists in integrated traditional Chinese and Western medicine, Western medicine, and TCM practitioners, primarily engaged in clinical and technical work within TCM internal medicine, rheumatology and immunology. Pharmaceutical experts held expertise in Chinese Pharmacy, and were mainly engaged in Chinese pharmaceutical compounding, clinical Chinese pharmacy, and rational drug use evaluation.

The average age of 22 experts was 37 ± 5.33, with an average working duration of 8.1 ± 3.09. All participants held intermediate or senior professional titles and were affiliated with hospitals across various provinces and municipalities, including Beijing, Tianjin, Hebei, Henan, Jilin, and others. Based on self-assessment, the experts’ authority coefficient was 0.71, meeting the requirements of the Delphi study and ensuring stable and reliable results. Further details on expert demographics are presented in Table 2.

The correlation coefficients of the 22 experts, who were involved in the study on the importance and feasibility of primary criteria of benefit and risk assessment, were relatively concentrated, with values of 0.4091 and 0.2475 respectively, and the differences were significant (P < 0.01).

Based on experts’ ratings of the importance and feasibility of primary and secondary BRA criteria, 10 criteria were identified as having high levels of importance and feasibility (mean score ≥3.5, frequency of maximum score ≥0.2) and good coordination (coefficient of variation <0.25). These included the incidence of ADR/ADE, cardiovascular damage, hematological damage, liver and kidney damage, and allergic reactions among the risk criteria, and the number of swollen joints, the number of tender joints, morning stiffness duration, erythrocyte sedimentation rate (ESR), rheumatoid factor (RF) among the benefit criteria. The mean, frequency of full score, and coefficient of variation for each risk and benefit criterion were detailed in Tables 3, 4.

According to the scoring and ranking of selected criteria by experts, calculated the corresponding percentage weights of each criterion, established an MCDA-BRA model for Fuzi in treating RA, and set the optimal and worst values of each criterion to express the treatment expectations, as shown in Figure 1.

A total of 40 RCTs on Fuzi-containing decoctions for RA were included. These RCTs were published in Chinese journals between 2005 and 2024. A total of 3188 participants were involved, with an age range of 16–75 years. Female participants significantly outnumbered male participants. The intervention measures in 11 studies were the sole application of the decoction containing Fuzi, while the intervention measures in the remaining trials were the combination of the Fuzi decoctions with standard biomedical treatments such as methotrexate, diclofenac sodium and leflunomide. All the measures in the control groups were merely these standard biomedical treatments. The quality of the included studies was moderate. 21 RCTs described the methods for sequence generation, however, none of the studies described blinding of participants, personnel, or outcome assessors. There was no incomplete data or selective reporting. More details of the included RCTs and references were shown in the attached file.

The efficacy and safety outcomes of the 40 RCTs were combined and are shown in Table 5. The efficacy of decoctions containing Fuzi for the treatment of RA was superior to that of standard biomedical treatment in the control groups. Fuzi decoction significantly reduced physical symptoms such as joint tenderness, swelling, and morning stiffness. They also significantly reduced laboratory test criteria such as ESR and RF. In addition, Fuzi decoctions reduced the overall incidence of ADRs/ADEs, as well as the incidence of serious ADRs/ADEs involving the hematology, liver and kidney systems. No cardiovascular damage was observed in the included studies.

Based on the clinical application characters of Fuzi, the differences in BRA value were compared from the perspectives of syndrome, dosage, treatment duration, and combination of standard biomedical treatment. Given that all aconite products used in the included studies were processed forms, a separate benefit-risk assessment was not conducted.

A sensitivity analysis was carried out on the BRA model of the combined and single use of Fuzi decoctions in the treatment of RA. Under the current weighting setting, the BRA value of single use of Fuzi decoctions was higher than that of combined use. Regardless of how the weights were altered, it did not lead to a change in the BRA comparison, thereby inferring that the evaluation framework was relatively stable, as shown in Figure 6.

BRA is a critical component of post-marketing drug re-evaluation (Working CIOMS Group Ⅳ, 1998). Although various international methods exist for conducting BRA, including qualitative, semi-quantitative, and quantitative research methods, there remains no universally consensus method (Lu et al., 2013; Mt-Isa et al., 2014; Hughes et al., 2016). Among these, qualitative frameworks such as BRAT and PrOACT-URL, semi-quantitative models like TURBO, and a wide range of quantitative methods, including Quantitative Framework for Risk and Benefit Assessment, MCDA and so on, have been developed. More than 200 distinct algorithms have been specifically designed for MCDA (Trung et al., 2024). Among all BRA methodologies, MCDA, also known as multi-criteria decision making (MCDM), is a widely adopted quantitative decision-making tool that integrates multiple, conflicting evaluation criteria through a comprehensive evaluation process (Tsoukias, 1994; Xiong and Dong, 2017). Initially applied in fields such as engineering, economics, environmental science, and the military, MCDA has gradually been adopted in healthcare research and management for supporting evidence-based decision-making (European Medicines Agency, 2007; Madhavan et al., 2012). Compared with other BRA methods, MCDA-BRA offers a structured and systematic framework that incorporates diverse evaluation criteria and stakeholder perspectives into the decision-making process. It enables the identification of optimal solutions by simultaneously evaluating multiple criteria across complex scenarios—a key advantage in diverse fields (Truong et al., 2023; Trung et al., 2024b). Therefore, the ability of MCDA to integrate subjective and objective indicators aligns closely with the holistic and individualized principles of TCM, making it particularly well-suited for assessing complex therapeutic interventions such as toxic herbal medicines. The MCDA-BRA model has been applied in various pharmaceuticals BRA, including Rocasian for obesity, Fexofenadine for allergic rhinitis, Shuanghuanglian injection for upper respiratory tract infections, and oral anticoagulants for non-valvular atrial fibrillation (Hsu et al., 2015; Guo et al., 2016; Zhou et al., 2013; Lu et al., 2017).

The MCDA model primarily consists of evaluation criteria, criterion weighting, and value assignments (He and Sun, 2016). In the study, the MCDA-BRA criteria regarding Fuzi for treating RA included both risk criteria based on ADRs/ADEs associated with Fuzi and benefit criteria related to RA symptom improvement. During criteria selection, a variety of benefit criteria were considered, such as subjective pain scores, objective laboratory measurements, overall clinical efficacy rates, and physician evaluations, in order to comprehensively evaluate therapeutic effectiveness. To demonstrate the clinical superiority of toxic botanical drugs in treating RA, the selected benefit criteria were required to be both clinical significant and practically meaningful for patients. Risk criteria encompassed all potential ADRs/ADEs associated with RA treatments, with particular attention to well-documented cardiotoxicity of Fuzi.

Based on a comprehensive literature review, an initial list of potential BRA criteria was compiled. The Delphi Survey involving 22 multidisciplinary participant experts was conducted to finalize the criteria and assign their respective weights. Experts from diverse professional backgrounds, including TCM and western treatment, have convened to develop a comprehensive understanding of the benefits and risks associated with the use of Fuzi in the treatment of RA. The survey achieved a participation rate exceeding 70%, and the high degree of consensus among experts indicated that clinics had prioritized and focused on the BRA of toxic botanical drugs. The final model incorporated 5 benefit criteria and 5 risk criteria, with weights derived via the percentage weighting method (Ben et al., 2004; He and Sun, 2016). This process ensured that the model accurately reflected real-world clinical priorities while preserving operational feasibility, thereby establishing a transparent and reproducible framework for quantifying the benefit-risk profile of Fuzi across diverse clinical scenarios.

RA, in TCM, is classified as a form of Bi pattern. The therapeutic use of Fuzi can be traced back to the Fuzitang documented in Shanghanlun (Zhang, 1963), reflecting its long-standing history of medicinal use and demonstrated efficacy. The included RCTs in the study also demonstrated that Fuzi-containing decoctions significantly enhanced therapeutic outcomes for RA compared to conventional standard biomedical therapy. Pharmacodynamic studies also supported these findings, indicating that Fuzi may exert anti-inflammatory effects and reduce joint swelling by lowering serum levels of inflammatory mediators and modulating the expression of relevant cytokines (Wang and Zhu, 2010). These supported Fuzi as a viable complementary or alternative therapeutic option for the management of RA.

Despite its therapeutic benefits, ADRs/ADEs reports associated with Fuzi were relatively prevalent. Wang’s study analyzed 153 cases of ADRs/ADEs with clear causality from literature (Wang et al., 2016), while Lu’s epidemiological study indicated that over 5,000 ADRs/ADEs related to Fuzi were reported in China, Japan, and Germany between 2000 and 2010 (Lu et al., 2010). These findings underscore the necessity of recognizing the potential risks linked to Fuzi. Fuzi contains bisester-type aconitine, which is known to cause multi-organ toxicity, particularly cardiotoxicity. The bisester-type aconitine causes myocardial cell injury and stimulates the vagus nerve, thereby leading to arrhythmias and other cardiac toxicity that require close monitoring.

Remarkably, none of the included RCTs reported cardiotoxic ADRs/ADEs. This may reflect the effectiveness of established risk control strategies in TCM clinical practice. The favorable outcome may likely be attributable to three key factors. Firstly, strict adherence to the principle of “yougu-wuyun” (有故无殒), which guided the use of Fuzi in specific syndrome patterns such as mixed cold-heat and cold-dampness Bi syndrome, ensuring therapeutic appropriateness and minimizing unintended toxicity. Pharmacokinetic studies provided direct evidence, in patients with cold-dampness Bi syndrome who received standardized Fuzi-containing decoctions, serum analyses detected only four monoester-type alkaloids, all at concentrations well below established toxic thresholds (Hou et al., 2014). Second, the standardized use of processed Fuzi facilitates the complete hydrolysis of highly toxic diester-type diterpenoid alkaloids into less toxic monoester-type derivatives. Third, proper decoction preparation, particularly prolonged boiling, further promotes the degradation of residual toxic constituents. All these practices measures integrated “detoxification” protocol essential for the safe clinical application of Fuzi. Although the possibility of undetected rare adverse events due to limited sample sizes cannot be entirely excluded, the consistent absence of reported toxicity across multiple studies strengthens the robustness and credibility of the findings.

These findings indicated that when administered within the structured framework of TCM diagnosis and preparation, Fuzi exhibits a favorable benefit-risk profile in the treatment of rheumatoid arthritis (RA).

The overall BRA value of Fuzi in the treatment of RA was ranging from 5 to 54 across included RCTs, which mean its benefits generally outweigh the risks when compared with standard biomedical therapy. The variability in BRA values reflected the influence of multiple clinical factors, which were systematically evaluated using a MCDA model. Since all included RCTs used prepared Fuzi, the study was able to assess how key variables, including TCM pattern, dosage, treatment course, and drug combinations, modulated the BRA profile.

In terms of TCM pattern, Fuzi possesses strong heating properties, which can correct a cold pattern. In the included trials, however, it was applied not only in cold-damp obstruction, but also in damp-heat obstruction, liver and kidney deficiency, phlegm-damp obstruction, and cold-heat complex patterns. The highest BRA value was observed in the cold-heat complex pattern (54), followed by liver and renal deficiency pattern at 41, cold-damp obstruction at 36, damp-heat obstruction at 30, and phlegm-damp obstruction at 5. This may be related to the effectiveness of Fuzi, which supports yang, warms the meridians and channels.

Regarding dosage, 34 RCTs (85%) utilized Fuzi within a range of 3–15 g specified by the Chinese Pharmacopoeia (2020 edition) (National Pharmacopoeia Committee, 2020). Compared with the high-dose group (15 g), the 3–15 g group achieved a comparable benefit score (43) but exhibited a lower risk score (48 vs. 39), resulting in a more favorable BRA value (46 vs. 41). Higher doses of toxic herbs invariably carry increased intrinsic risk, and clinical practice should strictly adhere to pharmacopoeial dosage limits to preserve therapeutic efficacy while minimizing potential toxicity.

With regard to treatment duration, the “2–3 months” treatment group demonstrated the highest BRA value (45), surpassing both the <2-month (41) and >3-month groups (37). Prolonged treatment did not enhance clinical benefit but was associated with increased risk, reinforcing the clinical recommendation of a 2–3 months treatment course as an optimal balance between efficacy and safety.

Finally, monotherapy with Fuzi-containing decoction yielded a significantly higher BRA value (54) than combination therapy with standard biomedical drugs (43), with a 94% probability of a true difference. This intriguing finding may be partly due to the avoidance of ADRs associated with conventional drugs in the monotherapy group. Nonetheless, it highlights the potential of Fuzi as a standalone therapy and warrants further investigation in larger trials designed specifically to compare monotherapy and add-on therapy.