Efficacy and Key Materials of East Asian Herbal Medicine Combined with Conventional Medicine on Inflammatory Skin Lesion in Patients with Psoriasis Vulgaris: A Meta-Analysis, Integrated Data Mining, and Network Pharmacology

Psoriasis is a chronic inflammatory disease that places a great burden on both individuals and society. The use of East Asian herbal medicine (EAHM) in combination with conventional medications is emerging as an effective strategy to control the complex immune-mediated inflammation of this disease from an integrative medicine (IM) perspective. The safety and efficacy of IM compared to conventional medicine (CM) were evaluated by collecting randomized controlled trial literature from ten multinational research databases. We then searched for important key materials based on integrated drug data mining. Network pharmacology analysis was performed to predict the mechanism of the anti-inflammatory effect. Data from 126 randomized clinical trials involving 11,139 patients were used. Compared with CM, IM using EAHM showed significant improvement in the Psoriasis Area Severity Index (PASI) 60 (RR: 1.4280; 95% CI: 1.3783–1.4794; p < 0.0001), PASI score (MD: −3.3544; 95% CI: −3.7608 to −2.9481; p < 0.0001), inflammatory skin lesion outcome, quality of life, serum inflammatory indicators, and safety index of psoriasis. Through integrated data mining of intervention data, we identified four herbs that were considered to be representative of the overall clinical effects of IM: Rehmannia glutinosa (Gaertn.) DC., Isatis tinctoria subsp. athoa (Boiss.) Papan., Paeonia × suffruticosa Andrews, and Scrophularia ningpoensis Hemsl. They were found to have mechanisms to inhibit pathological keratinocyte proliferation and immune-mediated inflammation, which are major pathologies of psoriasis, through multiple pharmacological actions on 19 gene targets and 8 pathways in network pharmacology analysis. However, the quality of the clinical trial design and pharmaceutical quality control data included in this study is still not optimal; therefore, more high-quality clinical and non-clinical studies are needed to firmly validate the information explored in this study. This study is informative in that it presents a focused hypothesis and methodology for the value and direction of such follow-up studies.


Introduction
Psoriasis is a chronic inflammatory disease of the skin that affects millions of people worldwide with a wide range of clinical symptoms [1]. A 2013 global epidemiological study found that adult prevalence ranged from 0.91 to 8.5%. The lifetime prevalence estimated by physicians was 6.3%, and a subsequent Danish cohort study supported this estimate [2,3]. Psoriasis is characterized by chronic inflammation resulting from the uncontrolled proliferation and differentiation of keratinocytes [4,5]. Inflammation of the skin on exposed areas of the body, such as the face and limbs, has a significant negative impact on the daily lives of the majority of people with psoriasis [6]. Many people with persistent psoriasis have a number of complications that might shorten their lifespans [7,8]. Recent studies have found a connection between psoriasis and other chronic diseases that may reduce life expectancy, such as psoriatic arthritis, hypertension, type 2 diabetes, dyslipidemia, myocardial infarction, and stroke [1,[9][10][11]. In addition, a metaanalysis in 2022 revealed that the prevalence of psoriasis appears to increase the incidence of autoimmune thyroid disease [12]. According to these findings, psoriasis must be considered as a systemic disease that can increase the local skin condition of each patient's burden on society [13]. Therefore, to reduce the severe impact of psoriasis on physical, social, and psychological well-being, it is necessary to find a way to effectively treat psoriasis.
Psoriasis can be clinically classified into four main subtypes: erythrodermic, guttate, pustular, and plaque psoriasis [1]. Skin-related lesions make up the majority of psoriasis lesions, with psoriasis vulgaris accounting for approximately 90% of all cases [1,4,7]. Erythematous, itchy plaques coated in silvery plaques are features of psoriasis vulgaris [4,14]. The scalp, elbows, face, and lumbosacral area are the most common places for plaques to appear as scaly skin lesions. They can vary widely in width and thickness [1,4]. When these plaques cover only 3-5% of the body surface in mild cases, topical treatment or phototherapy is often helpful [15]. However, systemic oral medications are required for moderate-to-severe plaque psoriasis [1,15]. Since psoriasis is a chronic disease that requires long-term medication, people with psoriasis often require lifelong therapy [16]. Consequently, all treatment plans must adhere to strict standards for patient safety. Although many conventional medicines (CMs) are available, there are still certain limitations to systemic therapy for psoriasis that need to be addressed. For example, acitretin is contraindicated in women of reproductive age due to its teratogenicity and because adverse events (AEs) such as dose-dependent alopecia and xerosis have been observed. Meanwhile, the long-used drug methotrexate has side effects including hepatotoxicity and bone marrow suppression that can lead to cirrhosis [4,17]. On the other hand, although methotrexate has been used for a long time, the possibility of inducing liver cirrhosis along with side effects such as hepatotoxicity and bone marrow suppression is still a concern [1, 18,19].
Natural products have been considered as promising candidates for the treatment of various chronic diseases worldwide because they are safer than novel synthetic drugs even after prolonged administration and with high patient compliance [14,[20][21][22]. Of these trends, the most active area in the discovery of promising materials related to psoriasis is East Asian herbal medicine (EAHM) [23][24][25][26][27][28][29]. The term "EAHM" refers to herbal therapies approved for use as medicines in a number of East Asian countries, including China, Taiwan, Korea, and Japan [23][24][25][26][27][30][31][32][33][34]. EAHM is significantly different from natural resources in other parts of the world as many similar medicinal plants are commonly listed in the pharmacopeia of East Asian countries [24,32,35]. At the same time, integrative medicine (IM) studies of chronic disease management are often conducted in countries with a history of medicinal herb use [36][37][38][39][40][41][42][43][44][45]. IM is a comprehensive strategy that uses both complementary and conventional therapies simultaneously. As a subset of IM, the combination of CMs and natural products has shown superior therapeutic efficacy and safety compared to standard therapies for a number of ailments, including COVID-19, cancer, stroke, chronic pruritus, and rheumatoid arthritis [36,40,[46][47][48][49][50][51][52][53][54]. Meanwhile, a recent study using a bioinformatics approach reported that the effectiveness of widely used EAHM prescriptions in psoriasis is related to the suppression of oxidative stress and the alleviation of the resulting inflammatory pathology [55]. Considering several previous studies on this topic, it is reasonable to assume that IM may improve the psoriasis area and severity index (PASI) and clinical symptoms while reducing AEs associated with CMs in psoriasis patients [56][57][58][59][60]. However, several issues must be resolved before identifying candidate materials for optimal IM utilization and making robust clinical decisions. First, EAHM was administered in the form of a polyherbal prescription tailored to the clinical findings of each patient, and the maximization of the synergistic effect expressed in the appropriate combination of these herbs was estimated as the key mechanism of action [30,41,[61][62][63]. Therefore, it is challenging to identify candidate materials for optimal combination pharmacotherapy and sufficient discussion cannot be performed based only on the existing studies. Moreover, despite the considerable amount of evidence gathered, a detailed analysis of the dosage route and composition of diverse materials is lacking. These are important factors that prevent coherent conclusions from being drawn. Therefore, more research is needed to inform whether IM with EAHMs is a useful intervention with a benefit in psoriasis, and specifically which EAHMs should be used.
In recent years, "reliable data" and the "integration of a wide range of analytical methodologies" have been suggested as prerequisites for natural product-based drug research [64]. The authors established the following research objectives based on their understanding and expertise from previous studies: (1) a systematic review of randomized controlled clinical trials (RCTs) will determine whether IM using EAHM is worthy of investigation for the treatment of inflammatory skin lesions in plaque psoriasis, and (2) further data analysis of herbal prescription data collected through this review will lead to hypotheses regarding promising candidates for the best IM for plaque psoriasis. Through the above research, the authors have attempted to open a discussion on a multifaceted analysis method that can overcome the complex variable problem of EAHM and produce useful information that can be used in useful follow-up studies.

Research Workflow through Integrated Methodology
This study was carried out according to the following steps: (1) Clinical trial information was collected according to the systematic review methodology. At this stage, statistically valid evidence related to IM was obtained for efficacy and safety. Through a meta-analysis of the collected clinical trials, we first determined whether the IM approach using EAHM was a data pool worth searching for useful candidates. Once this condition was met, multiple data mining were performed as the second step, (2) deriving core materials through multiple data mining of drug information, was carried out. Through this, useful core materials predicted to exert the greatest weight in the pharmacology of EAHM theory were selected. Then, the appropriate dosage and duration of administration were investigated. (3) Prediction of the mechanism supporting the efficacy of the derived core material. The mechanism of action supporting the efficacy of the identified core herbs on inflammatory skin damage in psoriasis was predicted using systems biology methodology. Finally, based on the above steps, new IM information derived from clinical trial data was searched for crude drug usage patterns, doses, administration periods, compounds, targets, and pathways. This study was conducted as a process of building multidisciplinaryintegrative-decision making-actual achievement-scientific creativity (M.I.D.A.S) research platform. The workflow of the methodology described above is summarized in Figure 1.

Data Sources and Search Strategy
The systematic exploration of clinical trial data for this study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) 2020 statement [65], and the protocol was pre-registered in PROSPERO (registration number: CRD 42022296852, available from https://www.crd.york.ac.uk/PROS-PERO/display_record.php?RecordID=296852 ). In addition, the protocol of this study has been formally published [66].
RCTs that evaluated the efficacy and safety of IM for psoriasis vulgaris were searched in the following 10 electronic databases from their inception until 29 July 2021: three English databases (PubMed, Cochrane Library, and Embase), four Korean databases (Korean Studies Information Service System, Research Information Service System, Oriental Medicine Advanced Searching Integrated System, and Korea Citation Index), two Chinese databases (Chinese National Knowledge Infrastructure Database, Wanfang data), and one Japanese database (Citation Information by National Institute of Informatics). The overall literature search procedure was carried out independently by two researchers (HGJ and HK). Detailed search strategies are presented in Supplementary Table S1.

Type of Studies
Only RCTs evaluating the efficacy and safety of IM for inflammatory skin lesion in plaque psoriasis were included. There were no restrictions on the language or publication time. Studies that meet the following criteria were excluded: (a) studies that are not RCTs or quasi RCTs; (b) studies not related to plaque psoriasis or related diseases; (c) primary intervention not related to IM; (d) no oral administration of medications; (e) not a clinical trial; (f) case reports or reviews; (g) studies not published in scientific peer-reviewed journals, including postgraduate theses or dissertations; and (h) studies in which the experimental intervention was not based on an IM approach, such as EAHM monotherapy.

Data Sources and Search Strategy
The systematic exploration of clinical trial data for this study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) 2020 statement [65], and the protocol was pre-registered in PROSPERO (registration number: CRD 42022296852, available from https://www.crd.york.ac.uk/PROSPERO/ display_record.php?RecordID=296852). In addition, the protocol of this study has been formally published [66].
RCTs that evaluated the efficacy and safety of IM for psoriasis vulgaris were searched in the following 10 electronic databases from their inception until 29 July 2021: three English databases (PubMed, Cochrane Library, and Embase), four Korean databases (Korean Studies Information Service System, Research Information Service System, Oriental Medicine Advanced Searching Integrated System, and Korea Citation Index), two Chinese databases (Chinese National Knowledge Infrastructure Database, Wanfang data), and one Japanese database (Citation Information by National Institute of Informatics). The overall literature search procedure was carried out independently by two researchers (HGJ and HK). Detailed search strategies are presented in Supplementary Table S1.

Study Selection 2.3.1. Type of Studies
Only RCTs evaluating the efficacy and safety of IM for inflammatory skin lesion in plaque psoriasis were included. There were no restrictions on the language or publication time. Studies that meet the following criteria were excluded: (a) studies that are not RCTs or quasi RCTs; (b) studies not related to plaque psoriasis or related diseases; (c) primary intervention not related to IM; (d) no oral administration of medications; (e) not a clinical trial; (f) case reports or reviews; (g) studies not published in scientific peer-reviewed journals, including postgraduate theses or dissertations; and (h) studies in which the experimental intervention was not based on an IM approach, such as EAHM monotherapy.

Type of Participants
There were no restrictions on age, gender, or race, and studies were only eligible for inclusion if they were conducted in patients with a diagnosis of psoriasis vulgaris. Only studies that provided official or validated diagnostic criteria were included. Studies that included patients with additional psoriasis subtypes, such as psoriatic arthritis, guttate psoriasis, palmoplantar pulposus, and erythrodermic psoriasis, were excluded from the review because the focus of the review was on plaque psoriasis.

Type of Interventions
RCTs evaluating the active intervention of IM (EAHM combined with CMs) in the treatment group versus CMs alone in the control group were included. For inflammatory skin lesions in psoriasis, all dosage forms of IM intervention were considered, including decoction, granules, capsules, and tablets. There were no restrictions on dosage or duration of therapy, but oral ingestion was the only acceptable method of administration. Trials that included non-drug treatment, acupuncture, massage, or other complementary therapies only in the experimental group and not in the control group were excluded. Studies that could not confirm the composition of specific herbal ingredients included in the EAHM formula were excluded.

Type of Outcome Measures
The primary endpoint was the response rate of patients with a PASI improvement of 60% (PASI 60). The primary outcome was also the absolute difference in PASI scores between the groups. In the first set of secondary outcomes, an improved PASI of 70% (PASI 70), recurrence rate, dermatological quality of life index (DLQI), and visual analog scale (VAS) were adopted to measure the clinical response of inflammatory skin lesions in patients with psoriasis. As the second group of secondary outcomes, TNF-α, IL-8, IL-17, IL-22, IL-23, and IFN-γ were selected to evaluate changes in inflammation-related biomarkers that support improvement of inflammatory skin lesions. The safety of IM was assessed using the incidence rates of adverse events (AEs) in each group as the third set of secondary outcomes.

Data Extraction and Management
Two researchers (HGJ and HK) independently retrieved the titles and abstracts of potentially eligible articles using the search approach described above. The inclusion and exclusion criteria were then used to guide a full-text review. Two reviewers (HGJ and HK) independently retrieved data from the included studies. Two reviewers independently extracted the following data for the selected trials: 1.
Publication information (title, first author, year of publication, and funding source).
Participants (age, sex, diagnostic criteria, and number of participants in each group). 4.
Intervention (experimental intervention, comparator, ingredients, and detailed information on intervention frequency of medication, dosage, mode of delivery, and course of treatment).

5.
Outcomes (primary and secondary outcomes, measurement point, blinding of outcome assessment, and AEs).
All disagreements were resolved through discussions with the researchers and the other author (DL).

Methodological Quality Assessment
The methodological quality of each included study was independently evaluated by two investigators (HGJ and HK) according to the revised version of risk of bias in randomized trials (RoB 2.0) [67]. Five areas of bias are addressed by RoB 2.0: bias arising from the randomization process, bias deviating from the intended intervention, bias due to where n is the number of nodes and E is the number of links. By evaluating the power of specific materials on the association between frequently prescribed EAHMs, centrality was used to discover EAHM with relatively greater influence. The eigenvector centrality scale was used to analyze the association between each of the concurrently administered herbs.
where N(i) represents the collection of herbs that are close to material i and λ is the eigenvalue of material i, a constant determined by the algorithm. If materials i and j are connected in the n × n-direction adjacency matrix A, Aij becomes "1"; otherwise, it becomes "0". Herb i and its neighbors constitute herb j, which is the eigenvector centrality value of Cj. Centrality measurements were performed on materials showing a frequency of use in more than 5% of the included trials. Third, association rule mining was performed to discover a meaningful combination pattern among all materials included in the IM prescription [76]. The frequent combination pattern itself may be a relevant herb unit representing clinical tacit knowledge, as EAHM is administered in a combination to optimize synergy. Support, confidence, and lift were the primary measures used in association rule mining. The support of the itemset is the proportion of transactions in the dataset that contain it. Itemsets whose support exceeds a user-defined minimum support level are considered frequent. A rule X ⇒ Y's confidence is expressed as conf (X ⇒ Y) = support(X∪Y)/support(X). Assuming that the transactions in question also contain the antecedent, this may be explained as an estimate of the probability P (Y|X) or the likelihood of finding the rule's consequent in those transactions. Another index called the lift can be represented by the following equation: The lift can be explained as the deviation of all rules supports from the support anticipated under independence, given the support of the rule on both sides. Stronger associations are indicated by higher lift values. targets were uploaded to the Metascape platform [87]. The p < 0.05 data selection threshold was selected. The KEGG mapper (https://www.genome.jp/kegg/mapper/) was used to explore their underlying molecular mechanisms [88].

Study Identification
By implementing the search strategy, the electronic search of the 10 databases identified 2434 potentially relevant articles. After removing 638 duplicate records, a total of 1796 records were collected. After screening for titles and abstracts, 1115 articles that met at least one of the exclusion criteria were excluded. The full-text assessment was performed on the remaining 460 studies, and 334 articles were excluded for the reasons listed in Figure 2. Finally, 126 eligible studies were included in this meta-analysis . The screening process is summarized in the PRISMA 2020 flow diagram ( Figure 2).

Study Characteristics
The basic characteristics of the 126 included studies are summarized in Table 1. In general, the sample size of the included studies ranged from 34 to 260, and 11,139 participants were divided into an experimental (n = 5624) and a control group (n = 5515). Acitretin (78 trials), topical corticosteroids (28 trials), immunosuppressants (8 trials), methotrexate (4 trials), topical urea (3 trials), eritretin (1 trial), topical retinoid (1 trial), topical pyrithione zinc (1 trial), topical boric acid (1 trial), and an ascorbate and pyridoxine combination (1 trial) were administered to the control group in all the included studies. PASI 60 was measured as the primary outcome in 96 studies. The PASI score, a primary outcome measure of the extent of inflammatory skin lesions, was used in 69 studies. In terms of inflammatory skin lesion-related secondary outcomes, PASI 70 was the score measured in 13 studies. Another secondary endpoint, the recurrence rate, was reported in 12 studies. A change in the DLQI was observed in ten studies. The VAS scores were observed in three studies. As a secondary outcome to measure laboratory findings related to inflammatory skin lesions, TNF-α, IL-8, IL-17, IL-22, IL-23, and IFN-γ were reported in twelve, eight, ten, four, five, and six studies, respectively. As a secondary outcome to evaluate the safety of IM versus CM, 75 trials reported information that could be used to compare the incidence rates of AEs in the experimental and control groups. All the included studies reported the treatment duration. It ranged from 2 to 16 weeks, with 31 studies adopting a treatment period of ≥12 weeks.

Risk of Bias
The methodological quality of the 126 included studies is summarized in Figure 3 and Table 2. The risk of bias in the studies was assessed using the Rob 2.0 tool [67]. In 101 studies, the overall risk of bias, in one or more domains, was evaluated to be of "some concern". In all other studies, the overall risk of bias was evaluated to be "high". Twenty-six studies reported detailed information on the randomization process, and only five studies explicitly revealed that a single-blind design was adopted [129,152,161,163,202]. In the remaining studies, it was difficult to obtain specific information other than the fact that randomization was performed, and the adoption of a blinded design was uncertain. In addition, a potential common factor that could affect the risk of bias was the absence of a preregistered protocol that could prevent selective outcome reporting.

Risk of Bias
The methodological quality of the 126 included studies is summarized in Figure 3 and Table 2. The risk of bias in the studies was assessed using the Rob 2.0 tool [67]. In 101 studies, the overall risk of bias, in one or more domains, was evaluated to be of "some concern". In all other studies, the overall risk of bias was evaluated to be "high". Twentysix studies reported detailed information on the randomization process, and only five studies explicitly revealed that a single-blind design was adopted [129,152,161,163,202]. In the remaining studies, it was difficult to obtain specific information other than the fact that randomization was performed, and the adoption of a blinded design was uncertain. In addition, a potential common factor that could affect the risk of bias was the absence of a preregistered protocol that could prevent selective outcome reporting.            Figure 4A,B).

PASI Score
In the 69 studies comparing the effect of IM with that of the CM control, IM exhibited a significantly improved PASI score compared to the CM control (69 trials, n = 5801; MD:

Secondary Outcomes Group 3: Safety Assessment
A total of 79 trials (79/126, 62.7%) reported detailed information on AEs in the IM and CM control groups. In four trials, detailed information for each group of AEs was not reported [95,106,126,162]. Therefore, a safety assessment was performed by extracting AEs data from 75 trials. The individual symptoms of the reported AEs for each trial are presented in Table 1. AEs that occurred during the treatment of psoriasis could be classified into the following four categories: drug-induced liver injury, including findings of impaired hepatic function and elevated blood aspartate transaminase levels; cutaneous symptoms suggesting worsening of skin findings, such as xerostomia, xeroma, and xeroderma; alimentary symptoms, including abdominal discomfort and diarrhea after taking medication; and metabolic disorders reported primarily in the form of hyperlipidemia. Hence, the incidence rates were compared between the groups by dividing the reported AEs into the four categories mentioned above and one category that included other symptoms that could not be classified into these categories in this review.
A meta-analysis of 19 studies showed that IM significantly reduced the incidence of drug-induced liver injury compared to the CM control (19 trials, n = 1728; OR: 0.4219; 95% CI: 0.2482-0.7170; p = 0.0014, heterogeneity chi-square =11.90, df = 18, I 2 = 0%; p = 0.8523; Figure 7A). The pooled results of 31 studies comparing IM and CM controls suggested that the incidence of AEs was significantly reduced by IM administration. A total of 57 trials evaluated the effect of IM on reducing the incidence rate of cutaneous AEs compared to the CM control. The meta-analysis revealed a significant reducing effect of IM on AEs (57 trials, n = 2723; OR: 0.3578; 95% CI: 0.2687 to 0.4763; p < 0.0001, heterogeneity chi-square = 205.79, df = 56, I 2 = 72.8%; p < 0.0001; Figure 7B). The incidence rate of alimentary AEs in IM was similar to that of the CM controls, and there was no statistically significant difference between the two groups, according to a meta-analysis (31 trials, n =2723; OR: 0.9552; 95% CI: 0.6516 to 1.4002; p = 0.8143, heterogeneity chi-square = 30.70, df = 30, I 2 = 2.3%; p = 0.4305; Figure 7C). In contrast, the pooled results of 21 studies comparing IM and the CM control suggested that the incidence of metabolic AEs was significantly reduced by IM (21 trials, n = 1793; OR: 0.3288; 95% CI: 0.2369-0.4562; p < 0.0001, heterogeneity chi-square = 13.61, df = 20, I 2 = 0%; p = 0.8497; Figure 7D). In the 21 studies comparing the effect of IM with the CM control, IM significantly decreased the incidence rate compared to the CM

Secondary Outcomes Group 3: Safety Assessment
A total of 79 trials (79/126, 62.7%) reported detailed information on AEs in the IM and CM control groups. In four trials, detailed information for each group of AEs was not reported [95,106,126,162]. Therefore, a safety assessment was performed by extracting AEs data from 75 trials. The individual symptoms of the reported AEs for each trial are presented in Table 1. AEs that occurred during the treatment of psoriasis could be classified into the following four categories: drug-induced liver injury, including findings of impaired hepatic function and elevated blood aspartate transaminase levels; cutaneous symptoms suggesting worsening of skin findings, such as xerostomia, xeroma, and xeroderma; alimentary symptoms, including abdominal discomfort and diarrhea after taking medication; and metabolic disorders reported primarily in the form of hyperlipidemia. Hence, the incidence rates were compared between the groups by dividing the reported AEs into the four categories mentioned above and one category that included other symptoms that could not be classified into these categories in this review.

Sensitivity Analysis
Considerable heterogeneity was found in the synthesis of the trial data for PASI scores between the two primary outcomes of this study. In this respect, trials estimated to

Sensitivity Analysis
Considerable heterogeneity was found in the synthesis of the trial data for PASI scores between the two primary outcomes of this study. In this respect, trials estimated to be potential outliers were also observed in the drapery plot. To ascertain whether particular research matching these outliers was the cause of heterogeneity, a sensitivity analysis using the leave-one-out method was performed. Sensitivity analysis revealed that no single study's omission considerably affected the change in heterogeneity ( Figure 8A,B). be potential outliers were also observed in the drapery plot. To ascertain whether particular research matching these outliers was the cause of heterogeneity, a sensitivity analysis using the leave-one-out method was performed. Sensitivity analysis revealed that no single study's omission considerably affected the change in heterogeneity ( Figure 8A,B).

Meta-Regression and Subgroup Analysis
Sensitivity analysis demonstrated that specific individual studies corresponding to outliers did not affect heterogeneity. Consequently, a meta-regression analysis was carried out on moderators predicted to have an impact on the results to identify other possible sources of heterogeneity. The following moderators were examined in this study: (i) comparator drug, (ii) treatment duration, (iii) source of investigational medicine, (iv) formulation type, (v) sample size, (vi) overall risk of bias, and (vii) randomization method. Meta-regression analysis revealed that the variables "source of investigational medicine" (p = 0.0240) and "formulation type" (p = 0.0145) had a statistically noticeable influence on the pooled results ( Figure 9A,B). However, the results of the meta-analysis were not significantly influenced by the comparator drug (p = 0.6215, Supplementary Figure S1A Figure S1E). Based on the meta-regression results, a subgroup analysis was performed on the source of investigational medicine and formulation type. However, subgroup analysis showed that neither moderator had a significant influence on the magnitude of the effect or degree of heterogeneity (Table 3).

Assessing Publication Bias
A contour-enhanced funnel plot and Egger's and Begg's tests were used to assess the potential publication bias of the primary outcomes in this meta-analysis. Asymmetric shapes were observed in the contour-enhanced funnel plot for PASI 60, suggesting potential bias ( Figure 10A). The publication bias was statistically significant in the Egger's and Begg's tests for PASI 60 (Egger's test: t = 10.67, df = 94, p < 0.0001; Begg's test: z = 7.48, p < 0.0001). Unlike PASI 60, a distinct asymmetric shape suggestive of publication bias was not observed in the contour-enhanced funnel plot for the PASI score ( Figure 10B). The publication bias of the continuous PASI score was significant in Egger's test, but no significant bias was confirmed in Begg's test (Egger's test: t = −6.58, df = 67, p < 0.0001; Begg's test: z = 1.19, p = 0.2335).

Summary of Evidence according to Outcome Measures
For all outcome measures in the IM compared to CM trials, the overall quality of the evidence ranged from very low to moderate. Table 4 shows the results of the GRADE assessments.

Summary of Evidence according to Outcome Measures
For all outcome measures in the IM compared to CM trials, the overall quality of the evidence ranged from very low to moderate. Table 4 shows the results of the GRADE assessments.    Table 5 provides detailed information on the frequency distributions of EAHM materials.

Social Network Analysis
Social network analysis was conducted on materials that were utilized regularly in more than 5% of trials to examine the correlation and centrality among herbal medicines prescribed to patients, as well as to identify the candidate sovereign herbs (Figure 11). In more than 5% of the trials, a total of 45 herbal medicines were included in the analysis. Centrality analysis showed that the range of the eigenvector centrality index of all materials was in the range of 0.338 to 1, and the range of the centrality index according to the PageRank method was found to be 0.0106 to 0.0284 (Table 5 and Table S3). The core materials that exceeded 0.9 in the eigenvector centrality were as follows: Rehmannia glutinosa (Gaertn.) DC., Salvia miltiorrhiza Bunge, Smilax glabra Roxb., Glycyrrhiza uralensis Fisch., Paeonia anomala subsp. veitchii (Lynch) D.Y.Hong and K.Y.Pan, Angelica sinensis (Oliv.) Diels, Paeonia × suffruticosa Andrews, Dictamnus dasycarpus Turcz., Arnebia euchroma (Royle) I.M.Johnst., Lonicera japonica Thunb., Saposhnikovia divaricata (Turcz.) Schischk., Cryptotympana dubia (Haupt), Isatis tinctoria subsp. athoa (Boiss.) Papan., Spatholobus suberectus Dunn, Sophora flavescens Aiton, and Scrophularia ningpoensis Hemsl. Only these materials exceeded 0.026 in the PageRank centrality, and there was no change in the ranking trend. These materials have the strongest reciprocity based on the properties of the prestige centrality. Therefore, they were considered as candidate sovereign herbs in this review. more than 5% of the trials, a total of 45 herbal medicines were included in the analysis. Centrality analysis showed that the range of the eigenvector centrality index of all materials was in the range of 0.338 to 1, and the range of the centrality index according to the PageRank method was found to be 0.0106 to 0.0284 (Tables 5 and S3). The core materials that exceeded 0.9 in the eigenvector centrality were as follows: Rehmannia glutinosa (Gaertn.) DC., Salvia miltiorrhiza Bunge, Smilax glabra Roxb., Glycyrrhiza uralensis Fisch., Paeonia anomala subsp. veitchii (Lynch) D.Y.Hong and K.Y.Pan, Angelica sinensis (Oliv.) Diels, Paeonia × suffruticosa Andrews, Dictamnus dasycarpus Turcz., Arnebia euchroma (Royle) I.M.Johnst., Lonicera japonica Thunb., Saposhnikovia divaricata (Turcz.) Schischk., Cryptotympana dubia (Haupt), Isatis tinctoria subsp. athoa (Boiss.) Papan., Spatholobus suberectus Dunn, Sophora flavescens Aiton, and Scrophularia ningpoensis Hemsl. Only these materials exceeded 0.026 in the PageRank centrality, and there was no change in the ranking trend. These materials have the strongest reciprocity based on the properties of the prestige centrality. Therefore, they were considered as candidate sovereign herbs in this review. Figure 11. IM herbal materials network used in more than 5% of trials for inflammatory skin lesion of psoriasis.

A Priori Algorithm-Based Association Rule Analysis
Nine association rules were discovered via a priori algorithm analysis of the ingredient data for the 137 herbs included in this review (Table 6). Among the investigated association rules, the following three patterns showed the highest confidence  :2.342). Therefore, four association rules with excellent characteristics, as described above, were selected as the most important patterns that can be confirmed among IM prescriptions. The relationship diagram of all the association rules is presented as a separate network graph ( Figure 12).

Derivation of Core Herbs
Based on the above three-step data mining, four core herbs, Rehmannia glutinosa (Gaertn.) DC., Isatis tinctoria subsp. athoa (Boiss.) Papan., Paeonia × suffruticosa Andrews, and Scrophularia ningpoensis Hemsl., were selected to simultaneously satisfy the three conditions defined in the research method section. Table 7 shows the results calculated using the inter-quartile range, with the extreme values removed for the range of dose and administration period of each material based on the data of the included trial. The chemical ingredients of four core herbs were retrieved from a traditional Chinese medicine system pharmacology database (TCMSP), an encyclopedia of traditional Chinese medicine (ETCM), a high-throughput experiment-and-reference-guided database of traditional Chinese medicine (HERB), and a traditional Chinese medicine integrative database (TCMID). For the above compounds collected from multiple databases, after completing the first screening by utilizing OB and DL information from TCMSP DB, we conducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A).

PPI Network Construction
Using the STRING 11.5 platform, we imported common targets and constructed the PPI network model (minimum required interaction score: 0.4) by restricting the organism to "homo sapiens". A total of 41 nodes with 462 edges were obtained with an average degree of 11.268. Two targets (CA3, CHRNA7) were excluded from the PPI network because they did not interact with other targets. The MCC algorithm was used by cytoHubba to screen the main hub genes. The outcome was the identification of 19 nodes as hub gene targets, each of which had an MCC score at least three times higher than the median value 3.93 × 10 5 ). Table 9 lists the 19 hub gene targets according to their degree centrality. Figure 13B shows the PPI network of hub targets. finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). ducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). ducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). ducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). ducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). ducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A). ducted a second screening by performing an ADME prediction for each compound and finally derived a list of 22 potential active compounds. (Table 8) We then performed target prediction by compound on the SwissTargetPrediction platform and obtained pieces of 473 compound-target relationship data after removing duplicates (Supplementary Table  S4). Additionally, the GeneCards database contained information on 584 human target genes associated with psoriasis vulgaris (Supplementary Table S5). After intersection mapping, 43 consensus genes were identified as potential therapeutic targets for four core herbs against psoriasis ( Figure 13A).

PPI Network Construction
Using the STRING 11.5 platform, we imported common targets and constructed the PPI network model (minimum required interaction score: 0.4) by restricting the organism to "homo sapiens". A total of 41 nodes with 462 edges were obtained with an average degree of 11.268. Two targets (CA3, CHRNA7) were excluded from the PPI network because they did not interact with other targets. The MCC algorithm was used by cytoHubba to screen the main hub genes. The outcome was the identification of 19 nodes as hub gene targets, each of which had an MCC score at least three times higher than the median value 3.93 × 10 5 ). Table 9 lists the 19 hub gene targets according to their degree centrality. Figure 13B shows the PPI network of hub targets.

PPI Network Construction
Using the STRING 11.5 platform, we imported common targets and constructed the PPI network model (minimum required interaction score: 0.4) by restricting the organism to "homo sapiens". A total of 41 nodes with 462 edges were obtained with an average degree of 11.268. Two targets (CA3, CHRNA7) were excluded from the PPI network because they did not interact with other targets. The MCC algorithm was used by cytoHubba to screen the main hub genes. The outcome was the identification of 19 nodes as hub gene targets, each of which had an MCC score at least three times higher than the median value 3.93 × 10 5 ). Table 9 lists the 19 hub gene targets according to their degree centrality. Figure 13B shows the PPI network of hub targets.

PPI Network Construction
Using the STRING 11.5 platform, we imported common targets and constructed the PPI network model (minimum required interaction score: 0.4) by restricting the organism to "homo sapiens". A total of 41 nodes with 462 edges were obtained with an average degree of 11.268. Two targets (CA3, CHRNA7) were excluded from the PPI network because they did not interact with other targets. The MCC algorithm was used by cytoHubba to screen the main hub genes. The outcome was the identification of 19 nodes as hub gene targets, each of which had an MCC score at least three times higher than the median value 3.93 × 10 5 ). Table 9 lists the 19 hub gene targets according to their degree centrality. Figure 13B shows the PPI network of hub targets.

Yes
Yes No Yes No ADME: absorption, distribution, metabolism, and excretion.

GO and KEGG Pathway Enrichment Analyses
The results of GO and KEGG analyses of the top 19 hub targets are shown in Figure 13A-E. A total of 312 biological processes were identified, including the response to hormones, response to UV, response to estradiol, response to lipopolysaccharide, positive regulation of cell migration, the regulation of inflammatory response, positive regulation of cytokine production, response to glucocorticoid, cellular response to an organic cyclic compound, and the positive regulation of hydrolase activity ( Figure 14A,B). A total of 21 molecular functions were identified, including protein phosphatase binding, kinase binding, protein kinase activity, protease binding, protein homodimerization activity, cytokine receptor binding, chromatin binding, enzyme activator activity, heme binding, and endopeptidase activity ( Figure 14A,B). Five cellular components were identified, including nuclear membrane, vesicle lumen, membrane raft, transcription regulator complex, perinuclear region of cytoplasm ( Figure 14A,B). A total of 104 pathways were identified using KEGG pathway analysis. The results suggested that the mechanisms of four core herbs were mainly linked to pathways in cancer, the AGE-RAGE signaling pathway in diabetic complications, pancreatic cancer, small cell lung cancer, MicroRNAs in cancer, Endocrine resistance, Leishmaniasis, the IL-17 signaling pathway, HIF-1 signaling pathway, MAPK signaling pathway, Chagas disease, Acute myeloid leukemia, Thyroid hormone signaling pathway, and the NF-kappa B signaling pathway ( Figure 14C-E). To intuitively illustrate the compound-target-pathway relationships between the four core herbs and psoriasis, we used a network visualization method, Alubian plots. As shown in Figure 15, the network contained 45 nodes (including 12 compound nodes, 19 target nodes, and 14 pathway nodes) and 652 edges. The compounds in the network, qingdainone, Dihydrochelerythrine, Imperatorin, Hesperetin, kaempferol, Salutaridine, Sugiol, Glycyrol, quercetin, 5-[5-(4-Methoxy-phenyl)-furan-2-ylmethylene]-pyrimidine-2,4,6-trione, indirubin, and Sitogluside, showed degree centrality values of 69,51,42,32,29,25,22,20,15,8,7, and 6, respectively. Among therapeutic targets, PTGS, MAPK8, and NOS2 were identified as important nodes with high-degree centrality values of 80, 80, and 64, respectively. In the case of pathways, pathways in cancer (hsa05200) and the AGE-RAGE pathway in diabetic complications (hsa04933) showed high values of 39 and 42, respectively.   2,4,6-trione, indirubin, and Sitogluside, showed degree centrality values of 69,51,42,32,29,25,22,20,15,8,7, and 6, respectively. Among therapeutic targets, PTGS, MAPK8, and NOS2 were identified as important nodes with high-degree centrality values of 80, 80, and 64, respectively. In the case of pathways, pathways in cancer (hsa05200) and the AGE-RAGE pathway in diabetic complications (hsa04933) showed high values of 39 and 42, respectively.

Summary of the Main Findings and Comparision with Previous Research
According to the clinical trial data analyzed in this study, IM can have a superior effect compared to CM in the improvement of skin damage as measured by PASI as well as inflammation-related biomarkers including TNF-α, IL-8, IL-17, IL-22, IL-23, and IFN-γ. Through integrated data mining to explore the core herbs that contribute most to this effect, Rehmannia glutinosa (Gaertn.) DC., Isatis tinctoria subsp. athoa (Boiss.) Papan., Paeonia × suffruticosa Andrews, and Scrophularia ningpoensis Hemsl. were identified as the four core herbs. To explore their therapeutic mechanisms for psoriasis from a holistic perspective, we further conducted network pharmacology analysis and found that the multiple bioactive compounds contributed to the clinical effects, mainly based on their actions on 19 gene targets and 14 signaling pathways.
The above results can be compared to previous studies we have conducted. Previously, we analyzed data from clinical trials comparing the efficacy of CM with the oral administration of EAHM alone and explored the core herbal materials through frequency analysis and network analysis [215]. In that study, EAHM showed superiority in PASI 70, PASI 60, and continuous PASI scores compared to the CM group. It also showed significant results in immune-mediated inflammatory markers such as IL-17, IL-23, and TNF-α, and in the incidence of adverse events. This supports the clinical benefit of IM identified in this study. On the other hand, the data mining used to extract the core drugs in the previous study is different from the methodology of this study, and no separate pharmacopredictive analysis was performed, making a direct comparison difficult. However, Rehmannia glutinosa (Gaertn.) DC. and Paeonia × suffruticosa Andrews were classified as core drugs in EAHM monotherapy. Isatidis radix, a plant of the same origin as Isatis tinctoria subsp. athoa (Boiss.) Papan. in this study, but administered at different sites and with almost similar traditional indications, was also a key drug in EAHM monotherapy [216]. Despite the completely different nature of EAHM monotherapy and its combination with CM, the findings of previous studies that similar drugs are important in psoriasis provide some support for the conclusion that the four core herbs in this study exert important pharmacological effects that represent the overall clinical efficacy of IM.
Compared to other relevant previous studies, the favorable effect of IM on psoriasis demonstrated in this study reveals a consistent conclusion [56][57][58][59][60]. However, the existing systematic reviews for simply evaluating the effectiveness of IM or EAHM monotherapy performed an analysis of the overall clinical endpoints related to psoriasis. Additionally, these previous studies did not limit the administration route and application form of the interventional drugs. Considering this, here we evaluated whether EAHM could be a useful candidate for a new IM drug, with a specific target, by setting strict criteria for the administration route and whether to use it as a combined treatment. As for the evaluation index, only the range that specifically supports the indication of "inflammatory skin lesion" was adopted, and not overall psoriasis. The results of this perspective analysis provided a favorable answer to the question of whether the whole EAHM used as an IM, including core herbs, may be a promising candidate material for inflammatory skin lesions caused by psoriasis.
On the other hand, owing to the nature of psoriasis, which has a complex underlying mechanism and a high incidence of comorbid diseases, safety concerns due to polypharmacy are important topics to be addressed [217,218]. Additionally, a detailed evaluation of the relationship between this dosage form and certain adverse effects, particularly druginduced liver damage, is required owing to the nature of IM, which requires combination therapy with CM. In this study, taking into consideration the aforementioned issues, the incidence rates of AEs were grouped and compared by the lineage in which the symptoms developed. The findings of the analysis demonstrate that IM can be a safer option than CM in the treatment of psoriasis by reducing the incidence of several AEs, such as drug-induced liver damage, cutaneous symptoms, and metabolic disorders. In addition, this justifies the development of a new EAHM drug for patients with poor adherence to CM.
Overall, based on the information from this study, EAHM can be considered worthy to continue its research, as a more effective and safer IM drug candidate for inflammatory skin lesions in psoriasis.

What Are the Limitations and Future Tasks of This Study?
The long history of EAHM administration on humans in numerous countries is perhaps the most substantial advantage of using them to identify potential new drug candidates. If the source data for drug discovery are confined to the classical literature or the experience of individual clinicians, the reproducibility of this information should be questioned. Encouragingly, the number of clinical trials using EAHM has increased sig-nificantly over the past decade. Our study reflects the quantitative growth of IM-related clinical and pharmacological data on psoriasis. However, due to data quality issues and methodological limitations not fully addressed in this study, considerable caution should be taken in interpreting the results.
The recent literature reflects the advances that have been made in research methodologies and suggests several requirements for optimized herbal medicine research [219]. Among them, the reproducibility/consistency of the results of the investigation is presented as an important factor. With regard to this requirement, the present study has limitations that cannot be overlooked. Despite the increase in the number of clinical trials identified in this review, there is still a significant risk of bias due to issues such as lack of pre-protocol enrollment, uncertainty in randomization methods, and selective outcome reporting. In addition, the lack of chemical and quality control information in all studies, which is essential for natural medicine trials, is an important reason to question the reproducibility of inferred IM efficacy. However, this is exactly the problem that this study aims to address. There are currently hundreds of EAHMs with pharmaceutical status, and it is unclear which individual herb or herbal preparations are worthy candidates for further study in psoriasis. Therefore, selecting more promising candidates and focusing resources on them according to the principle of selection and concentration to obtain higher quality and reproducible clinical evidence is a key challenge for future research and the aim of this manuscript. In the same sense, it is appropriate to accept the information in this study only at the level of a useful research hypothesis proposal that is relatively consistent with subsequent studies.
Second, the literature recommends the use of network pharmacology in herbal medicine research, but the results should be validated by in vivo experiments or clinical trials. The present study partially satisfies the above conditions in that it derived herbs that are believed to play a key role in a number of clinical trials and performed network pharmacology analysis on them. Although the four core herbs derived in this study showed a pattern of superiority in the data science characteristics based on clinical utilization information compared to other drugs, their pharmacological activity against psoriasis needs to be further verified by separate follow-up animal experiments to show their superiority over other drugs. Meanwhile, the methodology of network pharmacology itself needs to be improved. For example, during the course of this study, we recognized the risk that compound screening methodologies that rely solely on information provided by specific network pharmacology databases may not be updated or may erroneously lead to assays based on compounds that are not actually druggable. This would render the overall predictive results of Network Pharmacology unreliable. Therefore, future network pharmacology studies should be conducted in a manner that recognizes the potential for such errors in the information provided by existing databases and has measures in place to compensate for them.
Third, the specification of synergistic or antagonistic interactions of multi-component preparations was also presented as a very important condition for herbal medicine research. In particular, in the case of EAHM, which is the subject of this study, maximizing the synergistic effects between drugs in a multi-herb combination is a theoretical principle of clinical practice, and most clinical trials have adopted such a combination as an intervention. However, there are limitations due to the scope of the design of this study that prevented us from exploring these interaction effects in more depth. Furthermore, this study examined combination therapy with CM as an intervention, not EAHM monotherapy. The most appropriate CM that can be used in combination with core herbs could not be identified, and various expected interactions could not be considered. The most important issue to be addressed in relation to combined pharmacotherapy is the safety issue due to drug-drug interactions. Overall, this study's meticulous examination of all types of adverse events concluded that IM would benefit patients with psoriasis in terms of safety. Further studies of IM will require a more detailed analysis of the interactions between specific CMs and the candidate EAHM materials that will make up the actual IM.
Fourth, the four core herbs and their predicted pharmacodynamics in this study do not fully represent the full spectrum of EAHM efficacy for psoriasis. This approach was chosen to address some of the weaknesses of systematic reviews of EAHM efficacy in general, including a lack of information on the mechanisms of action and on key drug identification. This allowed us to filter out drugs that were likely to support the concept of a sovereign herb in EAHM, at least in terms of drug prescription data that reflected clinical implications, and that had prominent data science characteristics compared to many other individual herbs. We also used network pharmacology techniques to confirm that these potential sovereign herbs do indeed have pharmacological activities consistent with the efficacy of EAHM for psoriasis. However, this study cannot answer the question of whether herbs that exhibit prominent patterns and frequencies in data mining are likely to be actual theoretical EAHMs. Furthermore, due to the nature of psoriasis as an immune-mediated inflammatory disease, the targets and pathways involved in the pathology are far beyond the scope of modulation by the four core herbs in this study. Therefore, the methodology attempted in this study should be further developed through continuous follow-up and validation, and the results should not be interpreted as more than a high-quality research hypothesis.
In addition to the BLT4s discussed above, other factors important in the pathogenesis of psoriasis are being studied, including cytoplasmic phospholipase A2, glycerophosphodiester phosphodiesterase domain containing 3, arachidonate 12-lipoxygenase R type, phospholipase B-like 1, sphingomyelin phosphodiesterase 3, ganglioside GM2 activator, and serine palmitoyltransferase long chain subunit have been suggested as important gene targets to consider for future studies.

Implications of the Four Core Herbs That Emerged from This Study
The notable feature of this study is the discovery of four core herbs as a promising candidate material for further investigation using clinical trial data. It is anticipated that this idea can be employed as an effective research hypothesis in subsequent experiments for the following reasons.
EAHM is different from the experience of using other natural medicinal products in that the principle of use is to incorporate a polyherbal formulation that can maximize the synergistic effect. "Gun-Shin-Jwa-Sa" theory, which in WHO standard nomenclature is known as a sovereign-minister-assistant-courier, is the formulation theory for the ideal combination of EAHM [30]. Among them, the drug responsible for the main pharmacology is called a sovereign or monarch drug. Due to the impact of this sovereign drug, EAHM prescriptions may be taken as a single medication, and the remaining drugs both minimize the overall prescription's toxicity and enhance the activity of sovereign drugs [24,61,75].
The sovereign drug concept can be explained in two ways. First, identifying a sovereign drug in the EAHM formula, in which several materials are combined, is to screen candidates that represent the overall pharmacology of the prescription, so it is possible to effectively narrow down the dimension of the hypothesis for actual drug discovery. In addition, based on the premise that it is a sovereign drug that represents the mechanism of action of EAHM for a specific disease, an in-depth exploration of the mechanism of action of this material can provide more refined information for subsequent research.
However, the criteria used to identify this sovereign herb are ambiguous and the general system depends on the clinician's tacit knowledge. For example, the material with the highest capacity may not be a sovereign herb; conversely, the material used in a very small amount may be a sovereign herb. Consequently, the scientific research methodology to address this problem has not yet reached its full potential.
As a result, we focused on how to recognize such four core herbs as the most essential hypothesis for IM drug discovery in this work. In the authors' previous study, four core herbs was selected only by social network analysis of the frequently prescribed drug group and the centrality index calculated from it [220]. In contrast, this study offered four core herbs of a more specifically compressed range compared to previous research by combining several data-mining approaches and choosing materials that exhibit essential features across all procedures as intersections. The greatest strength of EAHM research is that it is currently used as a medicine in many East Asian countries. In other words, because crude EAHM is used as a drug, drug data such as human dose, administration period, and prescription trend by the disease can be obtained in a more direct form, giving it a comparative advantage in terms of safety as a candidate material for new drug research. In addition, the effects of polysaccharides and other phytochemicals with low bioavailability, which are difficult to predict using only compound-based pharmacological studies because of the nature of natural products, cannot be overlooked. Therefore, the development of botanical drugs using crude materials in EAHM research remains one of the most important topics.

Possible Pharmacological Mechanisms of Four
The materials derived from the core herbs in this study were Rehmannia glutinosa (Gaertn.) DC., Isatis tinctoria subsp. athoa (Boiss.) Papan., Paeonia × suffruticosa Andrews, and Scrophularia ningpoensis Hemsl., an EAHM that has been widely used for inflammatory and autoimmune diseases for a long time owing to its broad immunomodulatory effects. It can also reduce oxidative stress and inhibit apoptotic responses [221,222]. One study regarding the topical application of Rehmannia glutinosa (Gaertn.) DC. extracts to an atopic dermatitis animal model reported that skin inflammation was suppressed through the inhibition of dermal infiltration by inflammatory cells and the suppression of chemokine production by keratinocytes [223]. Isatis tinctoria subsp. athoa (Boiss.) Papan. has recently been attracting attention for its antiviral activity against various viral infections, including COVID-19, and its related effects, such as a wide range of anti-endotoxin, anti-inflammatory, and immune regulation activities, have also been confirmed [224,225]. Recently, it was reported that the anti-aging effect of this material is based on its antioxidant and antiinflammatory effects [226].
Paeonia × suffruticosa Andrews is the most widely used material for various skin inflammatory reactions and is expressed in redness and dryness [227]. As for its direct effect on psoriasis, it has been reported that the improvement of psoriasis-like skin lesions in a mouse model occurs through the inhibition of the mRNA expression of IL-23 and the activation of dendritic cells via actions on the toll-like receptor 7/8 signaling pathway [228]. Furthermore, it is well recognized that the Paeonia × suffruticosa Andrews has a variety of benefits for atopic dermatitis-related inflammation, dryness, and UV-induced skin aging [229][230][231]. Scrophularia ningpoensis Hemsl. is a material known for its role in various inflammatory diseases based on its pharmacological effects across anti-inflammatory, immune-enhancing, anti-apoptotic, and anti-allergic effects [232,233]. Scrophularia ningpoensis Hemsl. extract effectively inhibited inflammatory cell infiltration and the expression of IL-5, IL-13, IL-17, and immunoglobulin E, as well as reduced NF-κB phosphorylation, thereby exhibiting an anti-inflammatory effect [234].
As the four core herbs have a wide range of anti-inflammatory and immunomodulatory properties, it may be a promising candidate material for successfully reducing a variety of skin inflammatory consequences. At the same time, the assumption of the four core herbs' identification in this study will be a methodology to identify materials that play a key role in the manifestation of clinical effects found throughout IM clinical trial data which are also convincing.

The Key Therapeutic Targets and Pathways Associated with the Four Core Herbs
It is widely known that the mechanism of action of natural products, including East Asian herbal medicine, is mainly based on the action of multi-compound multitarget/signaling pathways [235][236][237]. In addition, the synergistic effect generated in this action process further enhances the potential of natural product pharmacology [238]. However, it is not easy to interpret and predict such a wide range of action unless a sepa-rate research methodology is used for such multilayered and multifaceted pharmacology. Therefore, in this study, a network pharmacology technique was additionally used for an exploratory analysis of the range of mechanisms of action for psoriasis of the derived four core herbs. The active compounds of the four core herbs that treat psoriasis were predicted to act on 19 gene targets, including STAT3, CASP3, PTGS2, BCL2, MMP9, EGFR, ESR1, CCND1, and STAT1, by PPI analysis. Based on this, as a result of KEGG pathway enrichment analysis, eight pathways were identified to be related to the psoriasis treatment effect of four core herbs. Among them, the pathways expected to be most relevant to important therapeutic gene targets are the "pathways in cancer" and "AGE-RAGE signaling pathway in diabetic complications". These include a number of very complex partial signaling pathways. Thus, the detailed relationship between the target and the pathway was additionally explored using the KEGG mapper to more specifically clarify where the psoriasis treatment effect occurs.
The major gene targets of four core herbs, MMP9, BCL2, STAT1, and STAT3, each as a subset of pathways in cancer, were confirmed to be involved in the MAPK pathway, the mTOR pathway, and cytokine-cytokine receptor interaction. These are all pathways involved in proliferation. Pharmacological action on these pathways may be related to the alleviation of psoriasis. For example, previous studies in which metformin was applied to the MAPK and mTOR pathways showed the significant inhibition of cell proliferation and inflammatory responses in human immortalized keratinocytes [239,240]. In view of this, it is not difficult to predict that various natural products that act on the MAPK pathway will show significant psoriasis treatment effects, and consistent progress and multiple previous studies have reported this [241,242]. Several recent studies on the mTOR signaling pathway have identified it as an important regulatory target for the treatment of the epidermal and functional structure in dermatologic diseases, including psoriasis, and relatedly, there are reports that natural products that can modulate this signaling pathway have shown good efficacy in ameliorating animal models of psoriasis with dyslipidemia [243,244]. Meanwhile, STAT1 and STAT3 have been shown to be involved in the cytokine-cytokine receptor interaction pathway, and indeed, these targets are involved in suppressing the pathology of immune-mediated inflammation in psoriasis. For example, blocking the STAT1 function is known to suppress the activity of pathogenic Th1 cells and induce the hyperactivation of Th17 cells, thereby suppressing psoriasis caused by them [245]. On the other hand, in the case of STAT3, previous research reports have been made stating it to be a mediating pathway of alleviating skin inflammatory damage of psoriasis through the anti-inflammatory and antioxidant activities of natural products [246]. CASP3 and CASP9 have been shown to be involved in the regulation of the apoptotic pathway, and given that impaired apoptosis is an important contributor to keratinocyte proliferation abnormalities in psoriasis pathology, it is conceivable that the inhibition of these pathways may be related to similar mechanisms of action as in proliferative pathologies [247]. Based on the above, it is possible to interpret that the activity of the four core herbs on pathways in cancer is mainly exerted through the mechanism of the inhibition of the abnormal proliferation of keratinocytes. This is also the pharmacological mechanism of many natural products that have been shown to alleviate psoriasis, a finding that is relatively consistent with the results of the clinical trial data [248].
The major gene targets of the four herbs in relation to AGE-RAGE pathways include JNK, BLC2, CASP3, NFKB1, JAK2, STAT1, and STAT3, which have been implicated in the regulation of PI3K-AKT, JAK/STAT, and MAPK pathways as partial signaling pathways. Among these, the PI3K-AKT pathway is important as it is involved in the development of general immune-mediated inflammatory diseases. Notably, mTOR kinase, a downstream effect of this pathway, is over-activated during the course of psoriasis pathology, promoting keratinocyte proliferation and suppressing differentiation [249]. Additionally, when the immune-mediated inflammatory state of psoriasis is dysregulated, mTOR signaling is activated in keratinocytes. Hence, the PI3K-AKT-mTOR pathway is an important component of psoriasis pathology in which immune and inflammatory pathologies are linked to keratin hyperplasia. In this context, among the natural products and phytochemicals effective in psoriasis, a number of cases showing a correlation with the PI3K-AKT-mTOR pathway have already been reported in previous studies [250]. The JAK/STAT pathway has also been implicated in the pathophysiology of psoriasis, involving the proliferation of Th17 cells, keratinocytes, and γ-δ T cells [251]. It is similar to the PI3K-AKT pathway above in that it is activated by cytokines such as IL-17 and IL-23 and serves as a link between immune-mediated inflammatory pathology and the overproduction of epidermal keratinocytes. Therefore, the JAK/STAT pathway can be considered as one of the other major targets for the treatment of psoriasis with natural products [252].
In summary, the multiple pathways of action of the four core herbal targets extracted from the herbal prescription data of 126 clinical trials are all closely related to the actual pathogenesis of psoriasis, and their effects appear to be centered on modulating pathologic keratinocyte proliferation and inhibiting immune-mediated inflammation. This is meaningful because it is consistent with the inhibitory effects of PASI and various inflammatory cytokines seen in clinical trial results. Therefore, it may be worthwhile to conduct intensive non-clinical and clinical follow-up studies on the multi-component, multi-pathway mechanisms of the four core herbs in suppressing psoriasis.

Promising Therapeutic Targets in Psoriasis
Unexplained by the Pharmacology of Four Core Herbs As described above, the four core herbs in this study exert modulatory effects on key pathologies of psoriasis through broad multi-target, multi-pathway pharmacological modulation. However, recent studies have reported promising therapeutic targets for psoriasis beyond the scope of the four core herbs.
Notable among these are the leukotriene B4 (LTB4) receptors, BLT1 and BLT2. BLT1 blockade has been shown to treat immune-mediated inflammatory diseases such as rheumatoid arthritis and asthma, and should be strongly considered in the treatment of psoriasis [253]. Recent studies in this regard include a report on the most widely used EAHM with potent anti-inflammatory activity, Coptis chinensis Franch. [254]. This study showed that Coptis chinensis Franch. can inhibit the pathology of immunosuppressive cytokine secretion by tumor-associated macrophages that cause immune checkpoint blockade (ICB) treatment for cancer to fail. In particular, the blockade of LTB4 signaling and inhibition of ICB were observed as one of the important mechanisms, and this study confirmed that the combination of Coptis chinensis Franch. and anti-cancer immunotherapy may lead to better results. Although this herb is not used in the drug data included in this study, it can be considered as a promising candidate for psoriasis treatment considering the above mechanisms of action. On the other hand, BLT2 is a low affinity receptor and agonists against it are considered to be able to treat a number of skin diseases. On a related note, another EAHM, branches of Morus alba L. (Mori ramulus), has been reported to exert antioxidant and anti-inflammatory activity by blocking BLT2-dependent NADPH oxidase 1 and inhibiting the production of IL-6 [255]. It is worth noting that this is a mechanism that may also be useful for inhibiting inflammation-mediated skin damage in psoriasis, and indeed, the drug data in this study include a prescription for the leaves of Morus alba L. (Mori folium) of the same plant. Actually, a number of Korean references also report its use for skin conditions such as hyperpigmentation [256].
In addition to the BLT4s discussed above, other factors important in the pathogenesis of psoriasis are being studied, including cytoplasmic phospholipase A2, glycerophosphodiester phosphodiesterase domain containing 3, arachidonate 12-lipoxygenase R type, phospholipase B-like 1, sphingomyelin phosphodiesterase 3, ganglioside GM2 activator, and serine palmitoyltransferase long chain subunit, which have been suggested as important gene targets to consider for future studies [257]. Immunomodulatory and antiinflammatory activities via these targets have been reported in EAHMs, but to date, there appears to be a paucity of data. However, based on the above discussion, two directions for further research can be suggested. First, as we have seen, the more complex the disease, the more important the action on key gene targets and pathways involved in the pathology, so it is necessary to conduct efficacy studies for specific indications of EAHM by targeting modulatory actions on newly identified important pathologies. In this case, the multi-target, multi-pathway effects of EAHM, a characteristic and advantage of EAHM, can be reduced in advance by actively using the network pharmacology methodology. In the present study, there were 26 drugs that showed significant patterns in terms of prescription frequency for psoriasis. Further research could be conducted to determine the extent to which these drugs are involved in the important targets of psoriasis discussed above and which pathological endotypes they modulate, which would be a worthwhile endeavor from a drug discovery perspective. Overall, the value of this study is that it provides insights that can be used to narrow down hypotheses in this large EAHM dataset for further research.

Conclusions
The present study showed that the combination therapy of EAHM and CM for the improvement of inflammatory skin lesions in psoriasis patients is a potentially more effective and safer option compared to CM monotherapy. Further integrated data mining and network pharmacology analysis of the drug data in this study revealed that the four core herbs, Rehmannia glutinosa (Gaertn.) DC., Isatis tinctoria subsp. athoa (Boiss.) Papan., Paeonia × suffruticosa Andrews, and Scrophularia ningpoensis Hemsl., exerted curative effects on psoriasis by inhibiting keratinocyte proliferation and immune-mediated inflammation based on their effects on multiple targets and pathways directly related to psoriasis. However, the quality of the clinical trial data underlying this study is considered suboptimal in terms of both study design and quality control of the intervention. Therefore, the results should be interpreted with caution, bearing in mind that reproducibility is an issue. Further preclinical efficacy studies of the four core herbs with promising pharmacological properties, as well as data from clinical trials with substantially improved designs that meet CONSORT criteria, are needed to validate the information explored in this study. In this respect, this study may be of value in providing summary hypotheses regarding the value and direction of conducting such follow-up studies.  Table S1: Search terms used in each databases; Table S2: Detailed information on East Asian herbal medicine utilized as a component of integrative medicine interventions; Table  S3: Prestige centrality of each of the East Asian herbs used in the clinical trials of 5 percent or more; Table S4: Potential gene targets for bioactive compounds from four core herbs; Table S5: Detailed information on potential psoriasis targets from the GeneCard database.

Conflicts of Interest:
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.