1.Expert Consensus on Clinical Application of Qinbaohong Zhike Oral Liquid in Treatment of Acute Bronchitis and Acute Attack of Chronic Bronchitis
Jian LIU ; Hongchun ZHANG ; Chengxiang WANG ; Hongsheng CUI ; Xia CUI ; Shunan ZHANG ; Daowen YANG ; Cuiling FENG ; Yubo GUO ; Zengtao SUN ; Huiyong ZHANG ; Guangxi LI ; Qing MIAO ; Sumei WANG ; Liqing SHI ; Hongjun YANG ; Ting LIU ; Fangbo ZHANG ; Sheng CHEN ; Wei CHEN ; Hai WANG ; Lin LIN ; Nini QU ; Lei WU ; Dengshan WU ; Yafeng LIU ; Wenyan ZHANG ; Yueying ZHANG ; Yongfen FAN
Chinese Journal of Experimental Traditional Medical Formulae 2025;31(4):182-188
The Expert Consensus on Clinical Application of Qinbaohong Zhike Oral Liquid in Treatment of Acute Bronchitis and Acute Attack of Chronic Bronchitis (GS/CACM 337-2023) was released by the China Association of Chinese Medicine on December 13th, 2023. This expert consensus was developed by experts in methodology, pharmacy, and Chinese medicine in strict accordance with the development requirements of the China Association of Chinese Medicine (CACM) and based on the latest medical evidence and the clinical medication experience of well-known experts in the fields of respiratory medicine (pulmonary diseases) and pediatrics. This expert consensus defines the application of Qinbaohong Zhike oral liquid in the treatment of cough and excessive sputum caused by phlegm-heat obstructing lung, acute bronchitis, and acute attack of chronic bronchitis from the aspects of applicable populations, efficacy evaluation, usage, dosage, drug combination, and safety. It is expected to guide the rational drug use in medical and health institutions, give full play to the unique value of Qinbaohong Zhike oral liquid, and vigorously promote the inheritance and innovation of Chinese patent medicines.
2.Compound Xishu Granules Inhibit Proliferation of Hepatocellular Carcinoma Cells by Regulating Ferroptosis
Yuan TIAN ; Yuxi WANG ; Zhen LIU ; Yuncheng MA ; Hongyu ZHU ; Xiaozhu WANG ; Qian LI ; Jian GAO ; Weiling WANG ; Wenhui XU ; Ting WANG
Chinese Journal of Experimental Traditional Medical Formulae 2025;31(2):37-45
ObjectiveTo study the mechanism of compound Xishu granules (CXG) in inhibiting the proliferation of hepatocellular carcinoma cells by regulating ferroptosis. MethodsThe transplanted tumor model of human Huh7 was established with nude mice and the successfully modeled mice were randomized into model, Fufang Banmao (0.21 g·kg-1), low-dose (1.87 g·kg-1) CXG, medium-dose (3.74 g·kg-1) CXG, and high-dose (7.49 g·kg-1) CXG groups. Mice were administrated with drinking water or CXG for 28 days, and the body weight and tumor volume were measured every 4 days. Hematoxylin-eosin staining was employed to observe the histopathological changes of tumors. The cell-counting kit-8 (CCK-8) was used to examine the survival rate of Huh7 cells treated with different concentrations (0, 31.25, 62.5, 125, 250, 500, 1 000 mg·L-1) of CXG for 24 h and 48 h. CA-AM, DCFH-DA, and C11-BODIPY581/591 fluorescent probes were used to determine the intracellular levels of ferrous ion (Fe2+), reactive oxygen species (ROS), and lipid peroxide (LPO), respectively. The colorimetric method was employed to measure the levels of glutathione (GSH) and superoxide dismutase (SOD). Western blot was employed to determine the protein levels of glutathione peroxidase 4 (GPX4), transferrin receptor 1 (TFR1), and ferritin heavy chain 1 (FTH1), respectively. ResultsIn the animal experiment, compared with the model group, the drug treatment groups showed reductions in the tumor volume from day 12 (P<0.01). After treatment, the Fufang Banmao and low-, medium-, and high-dose CXG groups had lower tumor volume, relative tumor volume, and tumor weight than the model group (P<0.05), with tumor inhibition rates of 48.99%, 79.93%, 91.38%, and 97.36%, respectively. Moreover, the CXG groups had lower tumor volume and relative tumor volume (P<0.05 in all the three dose groups) and lower tumor weight (P<0.05 in medium-dose and high-dose groups) than the Fufang Banmao group. Compared with the model group, the drug treatment groups showed reduced number of tumor cells, necrotic foci with karyopyknosis, nuclear fragmentation, and nucleolysis, and the high-dose CXG group showed an increase in the proportion of interstitial fibroblasts. In the cell experiment, compared with the blank group, CXG reduced the survival rate of Huh7 cells in a dose-dependent manner after incubation for 24 h and 48 h (P<0.05). Compared with the blank group, the RSL3 group and the low-, medium-, and high-dose CXG groups showed a decrease in the relative fluorescence intensity of CA-AM and increases in the fluorescence intensity of DCFH-DA and fluorescence ratio of C11-BODIPY581/591, which indicated elevations in the levels of Fe2+ (P<0.01), ROS (P<0.05), and LPO (P<0.01), respectively. Compared with the blank group, the RSL3 and low-, medium-, and high-dose CXG groups showed lowered levels of GSH and SOD (P<0.05). In addition, the RSL3 group and the medium- and high-dose CXG groups showed down-regulated expression of GPX4 and FTH1 (P<0.05), and the low- and high-dose CXG groups presented up-regulated expression of TFR1 (P<0.05). ConclusionCXG suppresses the proliferation of hepatocellular carcinoma cells by inducing ferroptosis via downregulating the GSH-GPX4 signaling axis and increasing intracellular Fe2+and LPO levels.
3.Efficacy and Mechanism of Shuanghua Drink in Treating Primary Dysmenorrhea Based on COX-2/NF-κB Signaling Pathway
Yuncheng MA ; Yuanyuan SHI ; Zhen LIU ; Yuxi WANG ; Yuan TIAN ; Qian LI ; Xiaozhu WANG ; Cheng HE ; Wenhui XU ; Weiling WANG ; Jian GAO ; Ting WANG
Chinese Journal of Experimental Traditional Medical Formulae 2025;31(12):72-80
ObjectiveTo evaluate the efficacy of Shuanghua drink in treating primary dysmenorrhea in the rat model and explore its mechanism of action. MethodsAn oxytocin-induced writhing mouse model was established to evaluate the analgesic effect of Shuanghua drink. Forty-eight non-pregnant female institute of cancer research (ICR) mice were randomly divided into six groups, including a blank group, a model group, an ibuprofen group (85.00 mg·kg-1), a low-dose group of Shuanghua drink (7.14 mL·kg-1), a medium-dose group of Shuanghua drink (14.28 mL·kg-1), and a high-dose group of Shuanghua drink (28.57 mL·kg-1). Each group consisted of eight mice. All treatment groups received daily intragastric administration at corresponding doses for 10 consecutive days. One hour after the final administration, 2 U of oxytocin was intraperitoneally injected per mouse. The writhing latency and number of writhing within 20 minutes were recorded. A primary dysmenorrhea rat model was established by using estradiol benzoate and oxytocin to evaluate the inhibitory effect of Shuanghua drink on the contraction of uterine smooth muscle. Forty-eight non-pregnant female Sprague-Dawley (SD) rats were divided into six groups, including a blank group, a model group, an ibuprofen group (51.00 mg·kg-1), a low-dose group of Shuanghua drink (4.28 mL·kg-1), a medium-dose group of Shuanghua drink (8.57 mL·kg-1), and a high-dose group of Shuanghua drink (17.10 mL·kg-1). Each group consisted of eight rats. Rats received subcutaneous injections of estradiol benzoate for 10 consecutive days to enhance uterine sensitivity. On the eleventh day, oxytocin (2 U/rat) was intraperitoneally administered to induce abnormal uterine contractions for establishing the primary dysmenorrhea model. All treatment groups received daily intragastric administration from the second day of modeling for 10 days. The effects of Shuanghua drink were evaluated by using parameters including uterine motility and the variation rate of uterine motility. The mechanism of action was investigated in rats with primary dysmenorrhea. The content of prostaglandin F2α (PGF2α), prostaglandin E2 (PGE2), thromboxane B2 (TXB2), prostacyclin metabolite (6-keto-PGF1α), and β-endorphin (β-EP) in uterine tissue of rats was detected by using enzyme-linked immunosorbent assay (ELISA). The changes in the content of nitric oxide (NO) and inducible nitric oxide synthase (iNOS) were analyzed via colorimetric assay. Western blot was performed to determine the content of phosphorylated inhibitor of kappa B kinase beta (p-IKKβ)/IKKβ, phosphorylated inhibitor of kappa B alpha (p-IκBα), IκBα, phosphorylated p65 (p-p65), p65, and cyclooxygenase-2 (COX-2) proteins in uterine tissue of rats. ResultsIn the oxytocin-induced writhing mouse model, the model group exhibited significantly shortened writhing latency and increased writhing frequency compared to the control group (P<0.01). Both the ibuprofen group and the high-dose group of Shuanghua drink displayed prolonged writhing latency (P<0.05), while the ibuprofen group and the low-dose, medium-dose, and high-dose groups of Shuanghua drink exhibited reduced writhing frequency (P<0.01). In the primary dysmenorrhea rat model, the uterine motility and its variation rate in the model group were significantly higher than those in the blank group (P<0.01). These parameters were markedly suppressed by ibuprofen and Shuanghua drink at all tested doses (P<0.01). For the mechanism of action, the model group showed significantly increased PGF2α/PGE2, TXB2/6-keto-PGF1α, NO, and iNOS in uterine tissue (P<0.05, P<0.01) and significantly decreased β-EP (P<0.01). These parameters were significantly attenuated in the ibuprofen group and the low-dose, medium-dose, and high-dose groups of Shuanghua drink. The PGF2α/PGE2 (P<0.01), TXB2/6-keto-PGF1α (P<0.01), NO (medium-dose group P<0.05), and iNOS (P<0.01) were reduced, and the β-EP (medium-dose group P<0.05) was up-regulated. Compared to the model group, the ibuprofen group and medium-dose group of Shuanghua drink showed significantly increased content of β-EP in the serum of rats (P<0.05). Compared to the blank group, the model group showed significantly elevated expressions of COX-2, p-IKKβ/IKKβ, p-IκBα/IκBα, and p-p65/p65 proteins (P<0.01) and significantly reduced anti-inflammatory protein IκBα (P<0.05). Compared to the model group, the ibuprofen group and the low-dose, medium-dose, and high-dose groups of Shuanghua drink showed significantly reduced expressions of COX-2 (P<0.01), p-IKKβ/IKKβ (P<0.01), p-IκBα/IκBα (P<0.05, P<0.01), and p-p65/p65(P<0.01) and up-regulated expression of IκBα protein (P<0.05, P<0.01). ConclusionShuanghua drink effectively alleviates primary dysmenorrhea through analgesia and suppression of abnormal contractions of uterine smooth muscle. Its mechanism may be mediated by reduced levels of PGF2α/PGE2, TXB2/6-keto-PGF1α, iNOS, and NO, elevated β-EP level, and inhibited COX-2/NF-κB signaling pathway.
4.Efficacy and Mechanism of Shuanghua Drink in Treating Primary Dysmenorrhea Based on COX-2/NF-κB Signaling Pathway
Yuncheng MA ; Yuanyuan SHI ; Zhen LIU ; Yuxi WANG ; Yuan TIAN ; Qian LI ; Xiaozhu WANG ; Cheng HE ; Wenhui XU ; Weiling WANG ; Jian GAO ; Ting WANG
Chinese Journal of Experimental Traditional Medical Formulae 2025;31(12):72-80
ObjectiveTo evaluate the efficacy of Shuanghua drink in treating primary dysmenorrhea in the rat model and explore its mechanism of action. MethodsAn oxytocin-induced writhing mouse model was established to evaluate the analgesic effect of Shuanghua drink. Forty-eight non-pregnant female institute of cancer research (ICR) mice were randomly divided into six groups, including a blank group, a model group, an ibuprofen group (85.00 mg·kg-1), a low-dose group of Shuanghua drink (7.14 mL·kg-1), a medium-dose group of Shuanghua drink (14.28 mL·kg-1), and a high-dose group of Shuanghua drink (28.57 mL·kg-1). Each group consisted of eight mice. All treatment groups received daily intragastric administration at corresponding doses for 10 consecutive days. One hour after the final administration, 2 U of oxytocin was intraperitoneally injected per mouse. The writhing latency and number of writhing within 20 minutes were recorded. A primary dysmenorrhea rat model was established by using estradiol benzoate and oxytocin to evaluate the inhibitory effect of Shuanghua drink on the contraction of uterine smooth muscle. Forty-eight non-pregnant female Sprague-Dawley (SD) rats were divided into six groups, including a blank group, a model group, an ibuprofen group (51.00 mg·kg-1), a low-dose group of Shuanghua drink (4.28 mL·kg-1), a medium-dose group of Shuanghua drink (8.57 mL·kg-1), and a high-dose group of Shuanghua drink (17.10 mL·kg-1). Each group consisted of eight rats. Rats received subcutaneous injections of estradiol benzoate for 10 consecutive days to enhance uterine sensitivity. On the eleventh day, oxytocin (2 U/rat) was intraperitoneally administered to induce abnormal uterine contractions for establishing the primary dysmenorrhea model. All treatment groups received daily intragastric administration from the second day of modeling for 10 days. The effects of Shuanghua drink were evaluated by using parameters including uterine motility and the variation rate of uterine motility. The mechanism of action was investigated in rats with primary dysmenorrhea. The content of prostaglandin F2α (PGF2α), prostaglandin E2 (PGE2), thromboxane B2 (TXB2), prostacyclin metabolite (6-keto-PGF1α), and β-endorphin (β-EP) in uterine tissue of rats was detected by using enzyme-linked immunosorbent assay (ELISA). The changes in the content of nitric oxide (NO) and inducible nitric oxide synthase (iNOS) were analyzed via colorimetric assay. Western blot was performed to determine the content of phosphorylated inhibitor of kappa B kinase beta (p-IKKβ)/IKKβ, phosphorylated inhibitor of kappa B alpha (p-IκBα), IκBα, phosphorylated p65 (p-p65), p65, and cyclooxygenase-2 (COX-2) proteins in uterine tissue of rats. ResultsIn the oxytocin-induced writhing mouse model, the model group exhibited significantly shortened writhing latency and increased writhing frequency compared to the control group (P<0.01). Both the ibuprofen group and the high-dose group of Shuanghua drink displayed prolonged writhing latency (P<0.05), while the ibuprofen group and the low-dose, medium-dose, and high-dose groups of Shuanghua drink exhibited reduced writhing frequency (P<0.01). In the primary dysmenorrhea rat model, the uterine motility and its variation rate in the model group were significantly higher than those in the blank group (P<0.01). These parameters were markedly suppressed by ibuprofen and Shuanghua drink at all tested doses (P<0.01). For the mechanism of action, the model group showed significantly increased PGF2α/PGE2, TXB2/6-keto-PGF1α, NO, and iNOS in uterine tissue (P<0.05, P<0.01) and significantly decreased β-EP (P<0.01). These parameters were significantly attenuated in the ibuprofen group and the low-dose, medium-dose, and high-dose groups of Shuanghua drink. The PGF2α/PGE2 (P<0.01), TXB2/6-keto-PGF1α (P<0.01), NO (medium-dose group P<0.05), and iNOS (P<0.01) were reduced, and the β-EP (medium-dose group P<0.05) was up-regulated. Compared to the model group, the ibuprofen group and medium-dose group of Shuanghua drink showed significantly increased content of β-EP in the serum of rats (P<0.05). Compared to the blank group, the model group showed significantly elevated expressions of COX-2, p-IKKβ/IKKβ, p-IκBα/IκBα, and p-p65/p65 proteins (P<0.01) and significantly reduced anti-inflammatory protein IκBα (P<0.05). Compared to the model group, the ibuprofen group and the low-dose, medium-dose, and high-dose groups of Shuanghua drink showed significantly reduced expressions of COX-2 (P<0.01), p-IKKβ/IKKβ (P<0.01), p-IκBα/IκBα (P<0.05, P<0.01), and p-p65/p65(P<0.01) and up-regulated expression of IκBα protein (P<0.05, P<0.01). ConclusionShuanghua drink effectively alleviates primary dysmenorrhea through analgesia and suppression of abnormal contractions of uterine smooth muscle. Its mechanism may be mediated by reduced levels of PGF2α/PGE2, TXB2/6-keto-PGF1α, iNOS, and NO, elevated β-EP level, and inhibited COX-2/NF-κB signaling pathway.
5.Ginkgo biloba extract EGb761 mitigates ischemic stroke via metabolic pathway modulation
Xiaoyao Liu ; Yuan Tian ; Wenhui Xu ; Yuxi Wang ; Zhen Liu ; Yuncheng Ma ; Weiling Wang ; Jian Gao ; Ting Wang
Journal of Traditional Chinese Medical Sciences 2025;2025(1):13-23
Objective:
To confirm the therapeutic efficacy of the ginkgo biloba extract EGb761 on ischemic stroke and elucidate its underlying mechanism.
Methods:
Male Sprague-Dawley rats were divided into three groups: sham, model, and EGb761 (ginkgo biloba extract). Ischemic stroke was then simulated in rats via embolic middle cerebral artery occlusion surgery, with the extract administered half an hour before surgery. Neurological deficit scores, infarct volume, cerebral edema rate, and inflammatory factors served as the primary metrics for drug efficacy. Serum metabolites were analyzed using 1H-nuclear magnetic resonance to elucidate the operative mechanism.
Results:
Treatment with the ginkgo biloba extract EGb761 significantly ameliorated the neurological deficit scores (P = .0343), diminished the cerebral infarct volume (P = .0001) and cerebral edema rate (P = .0030), and alleviated neuroinflammation (all P < .05) in middle cerebral artery occlusion rats. In addition, it significantly altered the contents of various metabolites, such as 2-hydroxybutyrate, isoleucine, isopropanol, isobutyric acid, N6-acetyllysine, glutamate, glutamine, methionine, and N,N-dimethylglycine (all P < .05). Enrichment analysis of the differential metabolites indicated that EGb761 may be involved in the regulation of amino acid metabolism, betaine metabolism, glucose-alanine cycle, Warburg effect, and urea cycle.
Conclusion
The ginkgo biloba extract EGb761 demonstrates anti-ischemic stroke effect on ischemic stroke model rats by regulating amino acids and amino acid derivatives, such as isoleucine, N6-acetyllysine, glutamate, methionine, and N,N-dimethylglycine.
6.Autologous platelet-rich plasma treatment for infected deep second-degree burn wounds: a case report
Hao HUANG ; Jun HUANG ; Qun LIANG ; Jian LIU ; Ting LI ; Yang ZHAO
Chinese Journal of Blood Transfusion 2024;37(1):95-100
【Objective】 To explore the feasibility of using autologous platelet-rich plasma (PRP) in the treatment of deep second-degree burns complicated with wound infection. 【Methods】 A retrospective analysis was conducted on the treatment process of a patient with deep second-degree burns and bacterial infection on the wound using autologous PRP. Clinical treatment highlights and outpatient follow-up were combined to discuss the feasibility and clinical effects of using autologous PRP in the treatment of burn wounds complicated with infection. 【Results】 The patient had a deep second-degree burn with a coagulase-negative Staphylococcus infection on the left lower limb. After one week of conventional wound dressing and antibiotic treatment, the patient's body temperature returned to normal. However, wound healing was slow and yellow secretion persisted. Subsequently, the burn wound was treated combined with topical autologous PRP. The wound pain score gradually decreased from 8 to 1. After 2 weeks, the bacterial culture of the wound secretion was negative, and the wound healed completely after 18 days. The wound scar score decreased from 5 to 2 at 1, 3 and 6 months after PRP treatment, and no obvious scar formation was observed. In the course of PRP treatment, there were no adverse reactions such as increased wound inflammation, abnormal blood routine and liver and kidney function test results. 【Conclusion】 For deep second-degree burn patients with localized wound bacterial infections who either refuse surgery or are not suitable for surgery, autologous PRP is a safer alternative that can effectively promote tissue regeneration and wound healing. The patient in this case achieved a curative effect in a short period of time.
7.Construction and validation of predictive models for intravenous immunoglobulin–resistant Kawasaki disease using an interpretable machine learning approach
Linfan DENG ; Jian ZHAO ; Ting WANG ; Bin LIU ; Jun JIANG ; Peng JIA ; Dong LIU ; Gang LI
Clinical and Experimental Pediatrics 2024;67(8):405-414
Background:
Intravenous immunoglobulin (IVIG)-resistant Kawasaki disease is associated with coronary artery lesion development.Purpose: This study aimed to explore the factors associated with IVIG-resistance and construct and validate an interpretable machine learning (ML) prediction model in clinical practice.
Methods:
Between December 2014 and November 2022, 602 patients were screened and risk factors for IVIG-resistance investigated. Five ML models are used to establish an optimal prediction model. The SHapley Additive exPlanations (SHAP) method was used to interpret the ML model.
Results:
Na+, hemoglobin (Hb), C-reactive protein (CRP), and globulin were independent risk factors for IVIG-resistance. A nonlinear relationship was identified between globulin level and IVIG-resistance. The XGBoost model exhibited excellent performance, with an area under the receiver operating characteristic curve of 0.821, accuracy of 0.748, sensitivity of 0.889, and specificity of 0.683 in the testing set. The XGBoost model was interpreted globally and locally using the SHAP method.
Conclusion
Na+, Hb, CRP, and globulin levels were independently associated with IVIG-resistance. Our findings demonstrate that ML models can reliably predict IVIG-resistance. Moreover, use of the SHAP method to interpret the established XGBoost model's findings would provide evidence of IVIG-resistance and guide the individualized treatment of Kawasaki disease.
8.Construction and validation of predictive models for intravenous immunoglobulin–resistant Kawasaki disease using an interpretable machine learning approach
Linfan DENG ; Jian ZHAO ; Ting WANG ; Bin LIU ; Jun JIANG ; Peng JIA ; Dong LIU ; Gang LI
Clinical and Experimental Pediatrics 2024;67(8):405-414
Background:
Intravenous immunoglobulin (IVIG)-resistant Kawasaki disease is associated with coronary artery lesion development.Purpose: This study aimed to explore the factors associated with IVIG-resistance and construct and validate an interpretable machine learning (ML) prediction model in clinical practice.
Methods:
Between December 2014 and November 2022, 602 patients were screened and risk factors for IVIG-resistance investigated. Five ML models are used to establish an optimal prediction model. The SHapley Additive exPlanations (SHAP) method was used to interpret the ML model.
Results:
Na+, hemoglobin (Hb), C-reactive protein (CRP), and globulin were independent risk factors for IVIG-resistance. A nonlinear relationship was identified between globulin level and IVIG-resistance. The XGBoost model exhibited excellent performance, with an area under the receiver operating characteristic curve of 0.821, accuracy of 0.748, sensitivity of 0.889, and specificity of 0.683 in the testing set. The XGBoost model was interpreted globally and locally using the SHAP method.
Conclusion
Na+, Hb, CRP, and globulin levels were independently associated with IVIG-resistance. Our findings demonstrate that ML models can reliably predict IVIG-resistance. Moreover, use of the SHAP method to interpret the established XGBoost model's findings would provide evidence of IVIG-resistance and guide the individualized treatment of Kawasaki disease.
9.Construction and validation of predictive models for intravenous immunoglobulin–resistant Kawasaki disease using an interpretable machine learning approach
Linfan DENG ; Jian ZHAO ; Ting WANG ; Bin LIU ; Jun JIANG ; Peng JIA ; Dong LIU ; Gang LI
Clinical and Experimental Pediatrics 2024;67(8):405-414
Background:
Intravenous immunoglobulin (IVIG)-resistant Kawasaki disease is associated with coronary artery lesion development.Purpose: This study aimed to explore the factors associated with IVIG-resistance and construct and validate an interpretable machine learning (ML) prediction model in clinical practice.
Methods:
Between December 2014 and November 2022, 602 patients were screened and risk factors for IVIG-resistance investigated. Five ML models are used to establish an optimal prediction model. The SHapley Additive exPlanations (SHAP) method was used to interpret the ML model.
Results:
Na+, hemoglobin (Hb), C-reactive protein (CRP), and globulin were independent risk factors for IVIG-resistance. A nonlinear relationship was identified between globulin level and IVIG-resistance. The XGBoost model exhibited excellent performance, with an area under the receiver operating characteristic curve of 0.821, accuracy of 0.748, sensitivity of 0.889, and specificity of 0.683 in the testing set. The XGBoost model was interpreted globally and locally using the SHAP method.
Conclusion
Na+, Hb, CRP, and globulin levels were independently associated with IVIG-resistance. Our findings demonstrate that ML models can reliably predict IVIG-resistance. Moreover, use of the SHAP method to interpret the established XGBoost model's findings would provide evidence of IVIG-resistance and guide the individualized treatment of Kawasaki disease.
10.Construction and validation of predictive models for intravenous immunoglobulin–resistant Kawasaki disease using an interpretable machine learning approach
Linfan DENG ; Jian ZHAO ; Ting WANG ; Bin LIU ; Jun JIANG ; Peng JIA ; Dong LIU ; Gang LI
Clinical and Experimental Pediatrics 2024;67(8):405-414
Background:
Intravenous immunoglobulin (IVIG)-resistant Kawasaki disease is associated with coronary artery lesion development.Purpose: This study aimed to explore the factors associated with IVIG-resistance and construct and validate an interpretable machine learning (ML) prediction model in clinical practice.
Methods:
Between December 2014 and November 2022, 602 patients were screened and risk factors for IVIG-resistance investigated. Five ML models are used to establish an optimal prediction model. The SHapley Additive exPlanations (SHAP) method was used to interpret the ML model.
Results:
Na+, hemoglobin (Hb), C-reactive protein (CRP), and globulin were independent risk factors for IVIG-resistance. A nonlinear relationship was identified between globulin level and IVIG-resistance. The XGBoost model exhibited excellent performance, with an area under the receiver operating characteristic curve of 0.821, accuracy of 0.748, sensitivity of 0.889, and specificity of 0.683 in the testing set. The XGBoost model was interpreted globally and locally using the SHAP method.
Conclusion
Na+, Hb, CRP, and globulin levels were independently associated with IVIG-resistance. Our findings demonstrate that ML models can reliably predict IVIG-resistance. Moreover, use of the SHAP method to interpret the established XGBoost model's findings would provide evidence of IVIG-resistance and guide the individualized treatment of Kawasaki disease.


Result Analysis
Print
Save
E-mail