Artificial intelligence for predicting the efficacy of Tuina in patients with knee osteoarthritis

Artificial intelligence for predicting the efficacy of Tuina in patients with knee osteoarthritis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Artificial intelligence for predicting the efficacy of Tuina in patients with knee osteoarthritis Hua Xing, San Zheng, Lijun Yao, Helin Wang, Yang Chen, Yangyang Fu, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7117693/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Background Knee osteoarthritis (KOA), a prevalent condition impacting middle-aged and older adults' quality of life, is increasing globally. Tuina, a Traditional Chinese Medicine technique, shows efficacy in reducing KOA pain and improving function, but response varies. This study aimed to develop a supervised machine learning classifier to predict Tuina efficacy for KOA, aiding personalized treatment planning. Methods This retrospective, registry-based, single-center prognostic study enrolled 355 KOA patients from the Tuina Department at Yueyang Hospital (Shanghai, China) between February 1, 2016, and December 31, 2023. All received standardized Tuina therapy (20-min sessions, 5×/week for 2 weeks). Efficacy was assessed via ΔVAS (post-treatment minus baseline VAS), categorized as high (ΔVAS = 4–6) or low efficacy (ΔVAS = 0–3). Eight machine learning models (e.g., Random Forest, SVM) were trained using 80% of the data (demographics, medical history, imaging assessments, physical exam findings, baseline VAS) to predict efficacy, validated on the remaining 20%. Statistical analysis used T-tests and Chi-square tests; model performance was evaluated via F1-score and AUC. Data analysis was performed from January 2024 to March 2025. Results The average reduction in VAS scores was 3.74. Among the eight trained machine learning models, the Random Forest-based model achieved the best predictive performance for the efficacy of Tuina treatment. The top six features influencing the model included the grinding test, knee joint range of motion, body mass index (BMI), height, imaging examination, and disease course. Conclusions Artificial intelligence models can reliably predict the efficacy of Tuina therapy in KOA patients. This study provides a valuable reference for Tuina practitioners in scientifically evaluating the effectiveness of KOA treatments. Trial registration: This retrospective study was approved by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine (No.2025-056). Knee osteoarthritis Manual therapy Machine learning Predictive model Pain management Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Knee osteoarthritis (KOA) is a common joint disease affecting middle-aged and older adults, characterized primarily by pain and knee dysfunction 1 , 2 . A systematic review conducted in 2020 found that the overall prevalence of KOA in China was 14.6%, with rates of 10.9% in men and 19.1% in women 3 . Globally, the number of KOA cases increased by 122.42%, rising from 163.91 million in 1990 to 364.58 million in 2019 4 . As a worldwide health issue, KOA severely impacts patients’ physical and psychological well-being, as well as their quality of life 5 , and brings substantial economic and social burdens 6 . Consequently, there is a growing consensus among patients and physicians regarding the need for safe and effective treatments for KOA. Tuina is a Traditional Chinese Medicine (TCM) physical therapy and a common non-surgical treatment for KOA, as recommended in the clinical guidelines 7 . Extensive experience in China has demonstrated its safety and effectiveness, supported by numerous rigorous clinical trials 8 , 9 , 10 . Research has shown that Tuina therapy benefits KOA by regulating muscle activation patterns, improving ligament and cartilage function, optimizing knee joint mechanics, reducing inflammation, alleviating joint pain, and promoting overall therapeutic outcomes 11 , 12 . However, the effectiveness of Tuina therapy was influenced by various endogenous factors, including the stage of KOA, the patient’s primary condition, the mechanical structure of the knee joint, and so on 13 . A systematic review also found that baseline effusion and synovitis, knee joint effusion, disease severity, drug injections, and more severe symptoms were associated with symptom relief in KOA patients following interventions 14 . These findings indicate that the baseline conditions significantly impact Tuina’s efficacy for KOA, raising an intriguing question: Can Tuina’s efficacy for KOA be predicted based on the patient’s initial condition? Machine learning (ML), a branch of artificial intelligence, offers the potential to develop predictive models based on existing data 15 . Most KOA prediction models have primarily focused on predicting disease progression 16 , with various well-performed ML algorithms varied in different research 17 – 19 . For instance, a logistic regression (LR) model was used to classify the radiographic progression of KOA 17 , and a support vector machine (SVM) model demonstrated superior performance, achieving approximately 93% accuracy in predicting the structural progression of KOA 18 . These findings indicated that ML had made significant strides in predicting KOA progression. However, its application in predicting treatment efficacy for KOA remains largely unexplored. For instance, Lin T used Lasso regression and LR models to predict the improvement in knee pain for KOA patients undergoing Vitamin D treatment 20 . However, no research has been conducted to assess the efficacy of Tuina therapy using ML. Therefore, exploring an optimal ML model to predict the efficacy of Tuina for KOA is of great practical value and importance. This study collected clinical data from KOA patients who underwent Tuina therapy and applied statistical and artificial intelligence methods to achieve three objectives: (1) evaluate the improvements in knee function and pain following Tuina therapy; (2) develop an optimal machine learning model to predict Tuina treatment outcomes; (3) use the SHApley Additive exPlanations (SHAP) method to identify the key factors affecting the model’s predictions. The refined model aims to assist Tuina practitioners in forecasting patient responses to treatment, enabling better patient stratification, identifying candidates most likely to benefit, and creating personalized treatment plans. By improving treatment efficiency and outcomes, this approach could reduce the personal and social financial burden caused by KOA. Methods Study Population This retrospective study focused on the treatment of KOA with Tuina therapy. Clinical data were collected from KOA patients hospitalized in the Tuina Department of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine between February 1, 2016, and December 31, 2023. After data anonymization by the hospital's Information Department, 355 eligible cases were identified. All patients received Tuina therapy following a unified protocol during hospitalization, supplemented by routine nursing care. The Tuina treatment involved the following steps: The patient was positioned supine while the physician treated the affected limb. Techniques included rolling the quadriceps femoris and pressing and kneading specific acupoints such as HeDing (EX-LE2), XueHai (SP10), FuTu (ST32), and LiangQiu (ST34). The patient was then prone, and the physician performed rolling manipulation along the back of the leg, including grasping manipulation at WeiZhong (BL40) and ChengShan (BL57). The treatment concluded with rubbing around the knee joint until a warm sensation penetrated internally. Acupoint locations were based on the World Health Organization (WTO) standards 21 (Fig. 1 A). Rolling manipulations were conducted at a frequency of 140 times/min. Each session lasted 20 min, with five sessions per week over 2 weeks. The inclusion criteria for KOA required patients to meet the clinical classification criteria established by the American College of Rheumatology (ACR) 22 , while the exclusion criteria included: 1) knee injuries or previous surgical treatment; 2) psoriasis, syphilitic neuropathy, metabolic bone disease, or acute trauma; 3) participation in other clinical trials within the previous three months; 4) use of nonsteroidal drugs during or within one month prior to treatment; 5) incomplete medical history or missing physical examination or imaging data; 6) concurrent treatments during hospitalization. This study has been reviewed and approved by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine (Approval No.2025-056). All data collected were only used for research purposes and did not include clinical specimens or human genetic information. This study adheres to the Checklist for Artificial Intelligence in Medical Imaging (CLAIM) for ML research 23 . Predictors Patient information for this study were collected from the hospital’s medical record system to establish a KOA database. The database includes four main categories: demographic features (gender, age, height, body weight, and BMI), medical history (hypertension, hyperlipidemia, diabetes, nephrosis, sleep disorder, smoking history, drinking history and the course of the disease), physical examination (knee joint range of motion, knee joint swelling, tenderness around the patella, medial and lateral collateral ligament tenderness, collateral ligament test, McMurray’s sign, drawer test, floating patella test and grinding test) and imaging assessments (Q angles, lateral tibiofemoral joint space, medial tibiofemoral joint space and tibiofemoral joint space ratio). Detailed statistical information for the selected features is presented in Table 1 . Table 1 Features of KOA patients involved in the study Feature No. (%) High △VAS a (n = 221) Low △VAS (n = 134) Total / (n = 355) Gender female 182 (82.35%) 113 (84.39%) 295 (83.10%) male 39 (17.65%) 21 (15.61%) 60 (16.90%) Age, mean (SD), y 68.38 ± 9.41 66.81 ± 10.50 67.78 ± 9.87 Height, mean (SD), cm 160.94 ± 5.97 161.01 ± 5.83 160.97 ± 5.92 Body weight, mean (SD), kg 66.63 ± 10.98 64.75 ± 8.93 65.92 ± 10.29 BMI b , mean (SD) 25.68 ± 3.64 24.99 ± 3.40 25.42 ± 3.56 Hypertension yes 125 (56.52%) 71 (52.94%) 196 (55.21%) no 96 (43.48%) 63 (47.06%) 159 (44.79%) Hyperlipidemia yes 23 (10.45%) 15 (11.27%) 38 (10.70%) no 198 (89.55%) 119 (88.73%) 317 (89.30%) Diabetes yes 32 (14.41%) 23 (17.16%) 55 (15.49%) no 189 (85.59%) 111 (82.84%) 300 (84.51%) Nephrosis yes 5 (2.27%) 6 (4.48%) 11 (3.10%) no 216 (97.73%) 128 (95.52%) 344 (96.90%) Sleep disorder yes 46 (20.83%) 40 (29.85%) 86 (24.23%) no 175 (79.17%) 94 (70.15%) 269 (75.77%) Smoking history yes 0 0 0 no 221 (100.00%) 134 (100.00%) 355 (100.0%) Drinking history yes 3 (1.36%) 1 (0.74%) 4 (1.13%) no 218 (98.64%) 133 (99.26%) 351 (98.87%) Course of the disease, mean (SD), y 4.14 ± 5.39 3.96 ± 5.51 4.07 ± 5.44 Imaging examination unilateral 120(54.30%) 72(53.73%) 192(54.08%) bilateral 101(45.70%) 62(46.27%) 163(45.92%) Q angle, mean (SD), ° 8.98 ± 2.67 8.60 ± 1.90 8.84 ± 2.42 Lateral tibiofemoral joint space, mean (SD), mm 6.00 ± 1.21 6.06 ± 1.19 6.02 ± 1.20 Medial tibiofemoral joint space, mean (SD), mm 4.60 ± 0.96 4.56 ± 0.85 4.59 ± 0.92 Tibiofemoral joint space ratio (lateral/medial) 1.34 ± 0.33 1.40 ± 0.43 1.37 ± 0.37 Knee joint range of motion, mean (SD) 103.14 ± 22.96 106.57 ± 20.22 104.44 ± 22.03 Knee joint swelling yes 119 (53.87%) 46 (34.33%) 165 (46.48%) no 102 (46.13%) 88 (65.67%) 190 (53.52%) Tenderness around the patella yes 183 (82.75%) 115 (85.82%) 298 (83.94%) no 38 (17.25%) 19 (14.18%) 57 (16.06%) Medial collateral ligament tenderness yes 118 (53.41%) 61 (45.52%) 179 (50.42%) no 103 (46.59%) 73 (54.48%) 176 (49.58%) Lateral collateral ligament tenderness yes 76 (34.32%) 47 (35.07%) 123 (34.65%) no 145 (65.68%) 87 (64.93%) 232 (65.35%) Collateral ligament test yes 124 (56.15%) 60 (44.78%) 184 (51.83%) no 97 (43.85%) 74 (55.22%) 171 (48.17%) McMurray’s sign yes 137 (61.94%) 75 (55.97%) 212 (59.72%) no 84 (38.06%) 59 (44.03%) 143 (40.28%) Drawer test yes 21 (9.43%) 10 (7.41%) 31 (8.73%) no 200 (90.57%) 124 (92.59%) 324 (91.27%) Floating patella test yes 46 (20.83%) 21 (15.67%) 67 (18.87%) no 175 (79.17%) 113 (84.33%) 288 (81.13%) Grinding test yes 116 (52.48%) 52 (38.81%) 168 (47.32%) no 105 (47.52%) 82 (61.19%) 187 (52.68%) Abbreviations: a. VAS indicates the Visual Analog Scale, a tool used to measure the intensity of certain pain along a continuous line, typically ranging from 0 (no pain) to 10 (maximum pain). b. BMI indicates body mass index calculated as weight in kilograms divided by height in square meters. Physical examination data were collected from the admission and discharge records of KOA patients to assess the effectiveness of Tuina therapy. Positive rates were used to describe the physical examination data, except for knee joint range of motion and Visual Analogue Scale (VAS) scores. All patients underwent X-ray examinations in our hospital’s radiology department. The model of the X-ray photography system was Samsung DGR-C28U2B/CN. Anteroposterior and lateral radiographs of the knee joint in a weight-bearing position were obtained. Two specialists independently measured all image data, and the results were averaged. The specific measurement methods are as follows: the tibiofemoral joint space (TJS) includes the medial tibiofemoral joint space (mTJS) and lateral tibiofemoral joint space (lTJS). According to Tourville et al. 24 , 25 , TJS was measured with the following detailed steps: (1) firstly, draw Line C, the axis of the tibial shaft through the tibial plateau; (2) secondly, draw Line D and Line E, the two parallel lines with Line C respectively through lateral and medial intercondylar tubercle; (3) then, draw the widest diameter line of the tibia platform and labeled Line F; (4) lastly, identify the mid-point between intercondylar tubercle and the ipsilateral tibial platform border, and through the two mid-point draw Line A and Line B, respectively paralleled with Line E and Line D. The medial and lateral tibiofemoral joint spaces were measured on the two reference lines parallel to Line F (Fig. 1 B). The quadriceps angle (Q-angle), as defined by Brattstrom, measures the angle between two lines: the one from the anterior-superior iliac spine to the mid-patella and the other from the mid-patella to the tibial tubercle 26 . The system automatically calculated the Q-angle using X-ray images based on the points and lines identified (Fig. 1 C). Tuina department residents documented all medical records at Yueyang Hospital, who specialized in Tuina therapy and underwent standardized training. Experienced chief physicians provided training and oversight for the physical examination procedures to ensure accuracy and consistency. Outcomes The VAS is a validated tool for measuring both acute and chronic pain. In this study, we evaluated the treatment effect of Tuina therapy by calculating the difference in VAS before and after treatment (△VAS). Notably, if patients experienced bilateral knee pain before treatment, the VAS score was recorded for the more severely affected knee, and so was △VAS. Patients assessed their pain levels with guidance from residents. We utilized a 10 cm sliding ruler marked with ten scales, ranging from 0 to 10, where 0 indicates no pain, and a 10 denotes unbearable pain. Patients selected the corresponding scale according to their feelings. In this study, a high △VAS was defined as a difference of 4 to 6 between pre-and post-treatment VAS scores, while a low △VAS was defined as a difference of 0 to 3. A High △VAS suggests that Tuina therapy has a significant impact on reducing pain in KOA patients. Machine learning-based modeling We used machine learning technologies to build classifiers that can predict the efficacy of Tuina therapy for treating KOA. The model-building process involved four key steps: data preprocessing, feature selection, algorithm selection, and parameter tuning. The prediction performance of the obtained models was evaluated, leading to selecting the best classifier, as illustrated in Fig. 2 . The method was described in our previous study 27 . We utilized Scikit-learn, a widely used Python-based machine learning library, to train our predictive models (refer to the website: https://scikit-learn.org/stable/ ) 28 . Eight representative machine learning algorithms were chosen for training: Random Forest, XGBoost, K-Nearest Neighbors (KNN), Decision Tree, Support Vector Machine (SVM), Logistic Regression, AdaBoost, and Artificial Neural Network (ANN). For each algorithm, we aim to identify a set of optimal parameters. Based on the training and validation datasets, we used a grid search approach to explore the parameter space by selecting a finite set of possible values for each parameter. This method iterated through each combination of parameters, and its predictive performance was evaluated. The parameters yielding the best average prediction performance were then retained. We adopted the following four metrics to quantify the predictive performance of each model: precision, recall, F1 score, and the AUC. Precision measures the proportion of actual high △VAS samples among those predicted as high △VAS by the model. Recall indicates the proportion of correctly predicted high △VAS samples among actual high △VAS samples. The F1 score, ranging from 0 to 1, represents the harmonic mean of precision and recall, with higher values indicating better overall model performance. AUC quantifies the model's ability to distinguish between positive and negative classes, with values ranging from 0 to 1; values closer to 1 reflect better predictive performance. The final classifier was determined by comparing the prediction performance of each trained model. Statistical Analysis Two investigators independently entered all data to ensure its completeness and accuracy, while a third investigator reviewed the data. Any identified anomalies were addressed by returning to the original records for correction. An independent professional radiologist conducted the measurement of X-ray images. Python programming software was used for statistical analysis. Numerical variables were expressed as mean ± standard deviation (SD), while categorical variables were described as counts and percentages. The 95% CI for the mean difference or proportions was calculated, a p -value of less than 0.05 was considered statistically significant. We employed the independent sample t-test for normally distributed data, while Welch's t-test was used for data that did not meet normal distribution criteria. Count data were compared using the Chi-Square test. To enhance the prediction model's interpretability, we quantified each feature’s importance in the prediction model using the SHAP method. SHAP is a representative method for interpreting the predictions of supervised machine learning classifiers. We obtained feature importance values using the metric mean (|SHAP value|), which is the average of the absolute SHAP values across all patients. When |SHAP value| is high, this feature contributes more to the prediction model. Code Availability The executable code for reproducing the analysis is available at [https://github.com/Helin-Wang/VAS_Prediction] under MIT License. Due to ethical restrictions, the original clinical dataset cannot be publicly shared but may be requested from the corresponding author upon reasonable approval. The repository includes synthetic data demonstrating the code functionality. Results Patient characteristics A total of 355 KOA patients were included in this study. After Tuina treatment, 221 (62.25% ) patients experienced a reduction in VAS scores of 4 to 6 (High △VAS group), while 134 (37.75%) patients reported a decrease of 0 to 3 (Low △VAS group). 295 (83.10%) were females, and 60 (16.90%) were males. The ages of the participants range from 39 to 89 years, with a mean age of 67.78 years (SD = 9.87). The average course of KOA among the patients was 4.07 years SD = 5.44). Regarding comorbidities, 196 (55.21%) patients had hypertension, 38 (10.70%) had hyperlipemia, 55 (15.49%) had diabetes, 11 (3.10%) had nephrosis, and 86 (24.23%) reported experiencing insomnia. All patients denied smoking, and only 4 patients (1.13%) had a history of alcohol consumption. Additional detailed information concerning the number, percentage, and p -value of each feature is presented in Table 1 . The average Q-angle of the knee joint was 8.84 degrees (SD = 2.42). The lateral and medial tibiofemoral joint spaces were 6.02 mm (SD = 1.20) and 4.59 mm (SD = 0.92). The tibiofemoral joint space ratio was 1.37 (SD = 0.37). Before treatment, the average knee flexion and extension angle was 104.44 degrees (SD = 22.03). Among the patients, 165 patients (46.48%) presented with swollen joints. Tenderness assessments revealed that 298 patients (83.94%) had peripatellar knee tenderness, 179 (50.42%) had medial tenderness, and 123 (34.65%) had lateral tenderness. The positive rate of the collateral ligament test, McMurray’s sign, drawer test, floating patella test, and grinding test were 184 (51.83%), 212 (59.72%), 31 (8.73%), 67(18.87%), and 168 (47.32%), respectively (Table 1 ). Effects of Tuina therapy on physical examination in patients with KOA After Tuina treatment, the flexion and extension angle of the patient’s knee joint improved to 116.46 degrees (SD = 0.86), representing a 10.03% increase from the baseline. The average VAS score decreased to 2.82 (SD = 0.05), which was 3.74 points lower than the pre-treatment score (Table 2 ). Additionally, comparisons of other physical examination results before and after treatment revealed significant reductions in the positive rate of knee joint swelling, tenderness around the patella, medial and lateral collateral ligament tenderness, collateral ligament test, McMurray’s sign, floating test, and grinding test. These findings indicate that Tuina therapy significantly improves patients’ knee joint function and alleviates knee pain. Table 2 Changes in physical examinations of KOA patients before and after Tuina treatment Physical examination parameter Positive No./total (%) Positive No./total (%) MD a /PR b change (%) P value c 95% CI baseline Tuina Knee joint range of motion, mean (SD d ), ° 106.43 ± 1.11 116.46 ± 0.86 -10.03 < 0.0001*** [6.46, 11.91] Knee joint swelling 165/355 (46.48) 99/355 (27.89) 18.59 < 0.0001*** [0.12, 0.26] Tenderness around the patella 298/355 (83.97) 235/355 (66.20) 17.77 < 0.0001*** [0.11, 0.24] Medial collateral ligament tenderness 179/355 (50.42) 125/355 (35.21) 15.21 < 0.0001*** [0.08, 0.22] Lateral collateral ligament tenderness 123/355 (34.65) 81/355 (22.82) 11.83 0.0005*** [0.05, 0.18] Collateral ligament test 184/355 (51.83) 106/355 (29.86) 21.97 < 0.0001*** [0.15, 0.29] McMurray’s sign 212/355 (59.72) 146/355 (41.13) 18.59 < 0.0001*** [0.11, 0.26] Drawer test 31/355 (8.73) 20/355 (5.63) 3.10 0.1099 [0.00, 0.07] Floating patella test 67/355 (18.87) 30/355 (8.45) 10.42 < 0.0001*** [0.05, 0.15] Grinding test 168/355 (47.32) 128/355 (36.06) 11.26 0.0023** [0.04, 0.19] VAS e , mean (SD) 6.56 ± 0.49 2.82 ± 0.05 3.74 < 0.0001*** [3.61, 3.88] Abbreviations: a. MD: mean difference. b. PR: positive rate. c. calculated by using Holm-Bonferroni correction. Special symbols are used to denote statistical differences in P values, such as *: P < 0.05 was considered statistically significant; **: P < 0.01; ***: P < 0.001. d. SD: standard deviation. e. VAS: visual analog scale. Machine-learning based Tuina efficacy prediction We employed supervised machine learning to develop an effective binary classifier for assessing the degree of knee pain relief following Tuina therapy. For this purpose, we selected eight machine learning algorithms: Random Forest, XGBoost, KNN, Decision Tree, SVM, Logistic Regression, AdaBoost, and ANN. These models were trained and validated using Dataset 1 and tested with Dataset 2 (Fig. 2 ). Our results showed that the Random Forest-based model outperformed the others, achieving an F1 score of 0.82 and an AUC value of 0.85, respectively. The results of the other models are summarized in Table 3 . The ROC curves for each trained model, reflecting their ability to predict changes in VAS scores resulting from Tuina treatment, are presented in Fig. 3 . Our results suggest that artificial intelligence can accurately predict the effectiveness of Tuina treatment for KOA patients based on their demographic characteristics, medical history, physical examinations, and imaging assessments. Table 3 the performance of each trained model in predicting VAS improvement after Tuina treatment Algorithm Precision Recall F1-Score AUC Random Forest 0.81 0.83 0.82 0.85 XGBoost 0.79 0.79 0.79 0.82 KNN 0.83 0.74 0.79 0.79 Decision Tree 0.84 0.81 0.83 0.74 SVM 0.81 0.74 0.78 0.82 Logistic Regression 0.77 0.77 0.77 0.72 AdaBoost 0.81 0.79 0.80 0.75 ANN 0.79 0.71 0.75 0.71 Feature importance Our dataset includes 28 features used to develop our models, each with varying significance. We calculated each feature’s SHAP value in the trained Random Forest-based classifier (Fig. 4 ). According to the SHAP value, the “Grinding test” ranked first among all features in this model. It means that the “Grinding test” result contributed the most to predicting the efficacy of Tuina treatment. “Knee joint range of motion”, “BMI”, “Height”, “Imaging examination”, and “Course of the disease” also ranked top. Then, we compared the top six features of KOA patients with different knee pain improvements after Tuina treatment (Fig. 5 ). We found that a KOA patient had a positive grind test, and Tuina was more effective in reducing knee pain. Besides, “Knee joint range of motion”, “BMI”, “Height”, and “Course of disease” also had a substantial predictive value for the efficacy of Tuina. In addition, we found that patients with knee KOA on both sides had worse Tuina efficacy than those with unilateral knee KOA. Discussion This study is the first attempt to use artificial intelligence to predict the effectiveness of Tuina therapy in patients with KOA. We developed a comprehensive KOA database encompassing 28 clinical characteristics. Our statistical analyses demonstrated that Tuina therapy significantly alleviated knee pain and improved knee joint function in the patients examined. By employing various machine learning algorithms, we constructed predictive models to evaluate the response of KOA patients to Tuina treatment, with the Random Forest algorithm-based model achieving the best predictive performance. Key factors such as grinding test, knee joint range of motion, and BMI significantly contributed to the model’s predictive accuracy. These findings provide valuable insights into the application of Tuina therapy in TCM and support a shift towards a more predictive, preventive, and personalized approach in TCM 29 . Our study demonstrated that Tuina therapy significantly reduced knee pain and improved joint function in patients with KOA (Table 2 ). The VAS score decreased from 6.56 ± 0.49 to 2.82 ± 0.05, with an average reduction of 3.74. These results align with existing literature, further supporting Tuina’s efficacy as a non-pharmacological treatment for KOA. For instance, a randomized controlled trial by Liu et al. showed that Tuina therapy substantially decreased knee pain, stiffness, and limitations in patient’s daily activities compared to health education 8 . However, they did not assess other aspects of knee function. Xu et al. reported that a six-week Tuina program significantly alleviated knee pain and improved patient’s mood and daily functioning 10 . Zhu et al. observed that the improved joint function was likely linked to the increased strength in knee extensors and flexors, though they noted no improvement in the range of motion after a two-week, three-times-per-week regimen 30 . Our study also showed that Tuina improved physical examination results in most KOA patients, including reductions in the incidence of knee joint swelling, tenderness around the patella, medial and lateral collateral ligament tenderness, collateral ligament tests, McMurray's sign, floating patella test, and grinding test (Table 2 ). However, these improvements do not necessarily indicate underlying knee joint structure changes, such as meniscal tears, ligament damage or cartilage degeneration. The observed benefits primarily reflect reduced pain and improved function, consistent with previous clinical studies. Further research, including imaging studies like MRI, is needed to assess whether Tuina therapy can impact joint structure. In summary, Tuina therapy, a TCM technique involving manual manipulation of the body surface, joints, and acupoints, improves local circulation and relieves muscle tension and spasms 31 . While manual therapy shows promise in reducing pain associated with KOA, yet it still has its limitations 9 . The effectiveness of Tuina for KOA necessitates further comprehensive evaluation, considering various influencing factors. In this study, we pioneered machine learning to model the therapeutic effects of Tuina therapy for KOA. By comparing eight different machine learning algorithms, we identified that the model based on the random forest algorithm best predicted the efficacy of the Tuina treatment, achieving an F1 score of 0.82 and an AUC value of 0.85 (Table 3 & Fig. 3 ). The application of machine learning in KOA research is expanding, encompassing areas such as disease diagnosis, efficacy prediction, risk assessment, and progression prediction et al. 16 , 32 . However, research specifically focused on predicting the therapeutic effects of TCM treatments like Tuina therapy remains limited. The random forest algorithm effectively integrates various factors influencing Tuina treatment for KOA by constructing multiple decision trees and synthesizing their predictive outcomes 33 . These factors include patient demographics, medical history, imaging results, and physical examinations at admission. The algorithm excels in capturing complex, nonlinear interactions among these factors, thus improving the accuracy of efficacy prediction. Additionally, the random forest algorithm possesses excellent robustness and generalization ability, minimizing sensitivity to noise and outliers and ensuring high predictive accuracy even with new data, which gives it a distinct advantage when dealing with diverse datasets 34 . Compared to traditional methods, which rely on observational studies and randomized controlled trials to evaluate the efficacy of Tuina, our study offers significant advantages. Traditional clinical research often fails to capture the complex interactions between individual patient differences and treatment parameters. In contrast, machine learning models integrate multidimensional patient data effectively to create more accurate predictive models 35 . This capability allows for anticipating potential treatment outcomes, providing a more reliable basis for clinical decision-making. In summary, this study demonstrated the potential of machine learning in predicting the effects of TCM treatments and also provides valuable insights for future research. It paves the way for exploring the KOA pathogenesis, risk prediction, and the development of personalized treatment plans using this technology. To enhance the model's interpretability and determine which features significantly influence prediction outcomes, we applied the SHAP method to sort the importance of the 28 features (Fig. 4 ). The top three features identified were the grinding test, knee joint range of motion, and BMI, highlighting their importance in the model’s predictive performance. Other factors also contribute, albeit with relatively minor contributions. The grinding test is a physical examination used to evaluate knee function, often associated to patella dysfunction, which is common in KOA 36 . A positive grinding test, typically indicating patellofemoral grind, may be linked to synovitis in KOA, making it a reliable predictor of KOA symptoms and pain improvement 37 . A reduction in knee joint range of motion is a significant clinical indicator for KOA patients 38 , 39 . Our study demonstrated that Tuina therapy increased the average knee joint range of motion by 10 degrees, effectively improving knee function (Table 2 ). This factor ranks high in the model, suggesting that early assessment and intervention of knee joint range of motion are crucial for managing and treating KOA. BMI is a significant risk factor for KOA. High BMI values not only significantly increase the incidence and progression of KOA but are also associated with more severe knee pain and functional impairment 40 , 41 . In addition, imaging examinations revealing unilateral or bilateral osteoarthritis can predict the efficacy of Tuina treatment for KOA. Compared to patients with bilateral KOA, those with unilateral KOA patients often had more asymmetrical inter-limbs foot posture, which was significantly associated with the K/L grade in KOA 42 . This finding supports the opinion that the disease severity is an important factor affecting the prognosis of KOA. Although clinical guidelines commonly incorporate K/L grade ≥ 2 combined with clinical symptoms as diagnostic criteria for KOA, the K/L grading system was not adopted as an objective outcome measure for treatment efficacy in this study due to its well-documented inter-observer variability in radiographic interpretation. Limitations Our study has some limitations. Firstly, although we selected features from inpatient medical records that might relate to the efficacy of Tuina therapy based on published literature and our long-term experience, many other potentially influencing features were not included. For instance, gait analysis results were excluded due to the limited number of patients undergoing this assessment, leading to insufficient data for further analysis. Additionally, certain biomarkers that may be associated with KOA were not included in our investigation. Regarding MRI data, we only used the results derived from MRI scans, not the full images. Moreover, our sample size remains limited, though we are actively working to expand the number of patients in our database. Lastly, it is noteworthy that most participants in this study were female, which may introduce a bias when applying the predictive model to male populations. Conclusion Our study demonstrated that Tuina therapy significantly alleviated knee pain and improved knee joint function in patients with KOA. Based on these findings, we developed and validated an artificial intelligence model to predict the efficacy of Tuina therapy for KOA by calculating patients' demographic information, medical history, imaging data, and physical examination results. This research offers Tuina practitioners a valuable tool for scientifically assessing the therapeutic effects of the treatment, thereby improving the precision of clinical decision-making. Abbreviations KOA: Knee Osteoarthritis TCM: Traditional Chinese Medicine ML: Machine Learning LR: Logistic Regression SVM: Support Vector Machine SHAP: SHApley Additive exPlanations WTO: World Health Organization ACR: the American College of Rheumatology BMI: Body Mass Index VAS: Visual Analogue Scale TJS: Tibiofemoral Joint Space KNN: K-Nearest Neighbors ANN: Artificial Neural Network AUC: Area Under the Curve SD: Standard Deviation MRI: Magnetic Resonance Imaging Declarations Ethics approval and consent to participate This study has been reviewed and approved by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine (Approval No.2025-056). All procedures adhered to the Declaration of Helsinki to ensure the rights and welfare of participants. The requirement for informed consent was waived by the Ethics Committee because: (1) The study involved only retrospective analysis of anonymized clinical data; (2) All data were processed in strict compliance with China's Regulations on the Management of Human Genetic Resources and the Guidelines for Ethical Review of Clinical Research; (3) This exemption is consistent with national regulations permitting waiver of consent for retrospective studies posing minimal risk. Consent for publication NA Data availability statement The original datasets cannot be shared publicly due to patient privacy restrictions. Researchers may apply for access by contacting the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine ( [email protected] ) or the corresponding author Li Gong ( [email protected] ). Conflict of interest The authors declare that they have no competing interests. Role of the funding source This work was supported by the National Natural Science Foundation of China (Grant Number: 8197151584); Shanghai Key Clinical Specialty Construction Project (Grant Number: Shslczdzk04001); the Sailing program of Shanghai Science and Technology Commission (Grant Number: 22YF1444300); Shanghai "Science and Technology Innovation Action Plan" Social Development Science and Technology Research Project (23DZ1204004); Shanghai "Science and Technology Innovation Action Plan" Medical Innovation Research Special Project (23Y11921700); Clinical Research Project of Shanghai Municipality Health Commission (20234Y0077, 202140037). Author Contributions H.X. and L.G. participated in the study design; W.S, L.Y., and L.G. participated in the supervision; H.W. and Y.C. participated in data analysis and modeling; S.Z. and Y.F. participated in data collection; Z.K. and X.S. participated in data analysis; H.X., S.Z., and L.Y. contributed to the manuscript writing; All authors have read and approved the final version of the manuscript. Acknowledgements We also gratefully acknowledge Figdraw (www.figdraw.com) for providing the anatomical human model template in Figure 1A (used for acupoint labeling) and the schematic framework in Figure 2. The final figures were adapted and annotated by the authors. References Felson DT. Clinical practice. Osteoarthritis of the knee. N Engl J Med. 2006;354(8):841-848. doi:10.1056/NEJMcp051726.1. Zhang S, Huang R, Guo G, et al. Efficacy of traditional Chinese exercise for the treatment of pain and disability on knee osteoarthritis patients: a systematic review and meta-analysis of randomized controlled trials. Front Public Health. 2023;11:1168167. doi:10.3389/fpubh.2023.1168167.2. Li D, Li S, Chen Q, Xie X. The Prevalence of Symptomatic Knee Osteoarthritis in Relation to Age, Sex, Area, Region, and Body Mass Index in China: A Systematic Review and Meta-Analysis. Front Med (Lausanne). 2020;7:304. doi:10.3389/fmed.2020.00304.3. Long H, Liu Q, Yin H, et al. Prevalence Trends of Site-Specific Osteoarthritis From 1990 to 2019: Findings From the Global Burden of Disease Study 2019. Arthritis Rheumatol. 2022;74(7):1172-1183. doi:10.1002/art.42089.4. Vitaloni M, Botto-van Bemden A, Sciortino Contreras RM, et al. Global management of patients with knee osteoarthritis begins with quality of life assessment: a systematic review. BMC Musculoskelet Disord. 2019;20(1):493. doi:10.1186/s12891-019-2895-3.5. Cross M, Smith E, Hoy D, et al. The global burden of hip and knee osteoarthritis: estimates from the global burden of disease 2010 study. Ann Rheum Dis. 2014;73(7):1323-1330. doi:10.1136/annrheumdis-2013-204763.6. Chen WH, Liu XX, Tong PJ, Zhan HS. Expert consensus on traditional Chinese medicine diagnosis and treatment of knee osteoarthritis (2015 edition). The Journal of Traditional Chinese Orthopedics and Traumatology. 2015;27(07):4-5. Liu K, Zhan Y, Zhang Y, et al. Efficacy and safety of Tuina (Chinese Therapeutic Massage) for knee osteoarthritis: A randomized, controlled, and crossover design clinical trial. Front Med (Lausanne). 2023;10:997116. doi:10.3389/fmed.2023.997116.48. Zhu B, Ba H, Kong L, et al. The effects of manual therapy in pain and safety of patients with knee osteoarthritis: a systematic review and meta-analysis. Syst Rev. 2024;13(1):91. doi:10.1186/s13643-024-02467-7.39. Xu H, Zhao C, Guo G, et al. The Effectiveness of Tuina in Relieving Pain, Negative Emotions, and Disability in Knee Osteoarthritis: A Randomized Controlled Trial. Pain Med. 2023;24(3):244-257. doi:10.1093/pm/pnac127.8. Wang M, Liu L, Zhang CS, et al. Mechanism of Traditional Chinese Medicine in Treating Knee Osteoarthritis. J Pain Res. 2020;13:1421-1429. doi:10.2147/jpr.S247827.10. Tang C, Kong LJ, Yao CJ, et al. Research Progress of Tuina in the Treatment of Knee Osteoarthritis from the Perspective of Mechanical Homeostasis. Guiding Journal of Traditional Chinese Medicine and Pharmacy. 2021;27(09):159-163. doi:10.13862/j.cnki.cn43-1446/r.2021.09.035.11. Carr AJ, Robertsson O, Graves S, et al. Knee replacement. Lancet. 2012;379(9823):1331-1340. doi:10.1016/s0140-6736(11)60752-6.12. Maricar N, Callaghan MJ, Felson DT, O'Neill TW. Predictors of response to intra-articular steroid injections in knee osteoarthritis--a systematic review. Rheumatology (Oxford). 2013;52(6):1022-1032. doi:10.1093/rheumatology/kes368.13. Binvignat M, Pedoia V, Butte AJ, et al. Use of machine learning in osteoarthritis research: a systematic literature review. RMD Open. 2022;8(1). doi:10.1136/rmdopen-2021-001998.14. Jamshidi A, Pelletier JP, Martel-Pelletier J. Machine-learning-based patient-specific prediction models for knee osteoarthritis. Nat Rev Rheumatol. 2019;15(1):49-60. doi:10.1038/s41584-018-0130-5.53. Dunn CM, Sturdy C, Velasco C, et al. Peripheral Blood DNA Methylation-Based Machine Learning Models for Prediction of Knee Osteoarthritis Progression: Biologic Specimens and Data From the Osteoarthritis Initiative and Johnston County Osteoarthritis Project. Arthritis Rheumatol. 2023;75(1):28-40. doi:10.1002/art.42316.16. Bonakdari H, Pelletier JP, Blanco FJ, et al. Single nucleotide polymorphism genes and mitochondrial DNA haplogroups as biomarkers for early prediction of knee osteoarthritis structural progressors: use of supervised machine learning classifiers. BMC Med. 2022;20(1):316. doi:10.1186/s12916-022-02491-1.17. Tiulpin A, Klein S, Bierma-Zeinstra SMA, et al. Multimodal Machine Learning-based Knee Osteoarthritis Progression Prediction from Plain Radiographs and Clinical Data. Sci Rep. 2019;9(1):20038. doi:10.1038/s41598-019-56527-3.18. Lin T, Peng S, Lu S, et al. Prediction of knee pain improvement over two years for knee osteoarthritis using a dynamic nomogram based on MRI-derived radiomics: a proof-of-concept study. Osteoarthritis Cartilage. 2023;31(2):267-278. doi:10.1016/j.joca.2022.10.014.19. Huang LX, Gang WJ, Li SY, Huang YM, Tan YS. World Health Organization standard acupuncture and moxibustion location (Western Pacific region): Chinese and English comparison. People's Medical Publishing House. 2010.64. ISBN: 9787117123327. https://books.google.com.hk/books?id=4TCjYgEACAAJ. Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association. Arthritis Rheum . 1986;29(8):1039-49. doi: 10.1002/art.1780290816. Mongan J, Moy L, Kahn CE, Jr. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): A Guide for Authors and Reviewers. Radiol Artif Intell . 2020;2(2):e200029. doi: 10.1148/ryai.2020200029. Tourville TW, Johnson RJ, Slauterbeck JR, Naud S, Beynnon BD. Assessment of early tibiofemoral joint space width changes after anterior cruciate ligament injury and reconstruction: a matched case-control study. Am J Sports Med. 2013;41(4):769-778. doi:10.1177/0363546513477838.20 Dupuis DE, Beynnon BD, Richard MJ, Novotny JE, Skelly JM, Cooper SM. Precision and accuracy of joint space width measurements of the medial compartment of the knee using standardized MTP semi-flexed radiographs. Osteoarthritis Cartilage. 2003;11(10):716-724. doi:10.1016/s1063-4584(03)00158-4.21. Milovanović D, Begović N, Bukva B, et al. The Influence of the Q-Angle and Muscle Strength on Idiopathic Anterior Knee Pain in Adolescents. Medicina (Kaunas). 2023;59(6). doi:10.3390/medicina59061016.22. Zhou S, Ma Q, Lou Y, et al. Machine learning to predict clinical remission in depressed patients after acute phase selective serotonin reuptake inhibitor treatment. J Affect Disord. 2021;287:372-379. doi: 10.1016/j.jad.2021.03.079.37. Wei HL, Yu YS, Wang MY, et al. Exploring potential neuroimaging biomarkers for the response to non-steroidal anti-inflammatory drugs in episodic migraine. J Headache Pain. 2024;25(1):104. doi:10.1186/s10194-024-01812-4.38. Barrett M, Boyne J, Brandts J, et al. Artificial intelligence supported patient self-care in chronic heart failure: a paradigm shift from reactive to predictive, preventive and personalised care. Epma j. 2019;10(4):445-464. doi:10.1007/s13167-019-00188-9.24. Zhu Q, Li J, Fang M, Gong L, Sun W, Zhou N. [Effect of Chinese massage (Tui Na) on isokinetic muscle strength in patients with knee osteoarthritis]. J Tradit Chin Med. 2016;36(3):314-320. doi:10.1016/s0254-6272(16)30043-7.50. Lee NW, Kim GH, Heo I, et al. Chuna (or Tuina) Manual Therapy for Musculoskeletal Disorders: A Systematic Review and Meta-Analysis of Randomized Controlled Trials. Evid Based Complement Alternat Med. 2017;2017:8218139. doi:10.1155/2017/8218139.51. Lee LS, Chan PK, Wen C, et al. Artificial intelligence in diagnosis of knee osteoarthritis and prediction of arthroplasty outcomes: a review. Arthroplasty. 2022;4(1):16. doi:10.1186/s42836-022-00118-7.54. Breiman L. Random Forests. Machine Learning. 2001;45(1):5-32. doi:10.1023/A:1010933404324.55. Mohapatra N, Shreya K, Chinmay A. Optimization of the Random Forest Algorithm. Paper presented at: Advances in Data Science and Management; 2020//, 2020; Singapore.56. Rajkomar A, Dean J, Kohane I. Machine Learning in Medicine. N Engl J Med. 2019;380(14):1347-1358. doi:10.1056/NEJMra1814259.57. Iversen MD, Price LL, von Heideken J, Harvey WF, Wang C. Physical examination findings and their relationship with performance-based function in adults with knee osteoarthritis. BMC Musculoskelet Disord. 2016;17:273. doi:10.1186/s12891-016-1151-3.33. Deng H, Wu Y, Fan Z, Tang W, Tao J. The association between patellofemoral grind and synovitis in knee osteoarthritis: data from the osteoarthritis initiative. Front Med (Lausanne). 2023;10:1231398. doi:10.3389/fmed.2023.1231398.34. Steultjens MPM, Dekker J, van Baar ME, Oostendorp RAB, Bijlsma JWJ. Range of joint motion and disability in patients with osteoarthritis of the knee or hip. Rheumatology. 2000;39(9):955-961. doi:10.1093/rheumatology/39.9.955.58. Epskamp S, Dibley H, Ray E, et al. Range of motion as an outcome measure for knee osteoarthritis interventions in clinical trials: an integrated review. Physical Therapy Reviews. 2020;25(5-6):462-481. doi:10.1080/10833196.2020.1867393.59. Blagojevic M, Jinks C, Jeffery A, Jordan KP. Risk factors for onset of osteoarthritis of the knee in older adults: a systematic review and meta-analysis. Osteoarthritis Cartilage. 2010;18(1):24-33. doi:10.1016/j.joca.2009.08.010.60. Zheng H, Chen C. Body mass index and risk of knee osteoarthritis: systematic review and meta-analysis of prospective studies. BMJ Open. 2015;5(12):e007568. doi:10.1136/bmjopen-2014-007568.61. Chen Z, Ye X, Shen Z, et al. Comparison of the Asymmetries in Foot Posture and Properties of Gastrocnemius Muscle and Achilles Tendon Between Patients With Unilateral and Bilateral Knee Osteoarthritis. Front Bioeng Biotechnol. 2021;9:636571. doi:10.3389/fbioe.2021.636571.35. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 04 Sep, 2025 Reviewers agreed at journal 21 Aug, 2025 Reviewers invited by journal 14 Aug, 2025 Editor assigned by journal 11 Aug, 2025 Editor invited by journal 17 Jul, 2025 Submission checks completed at journal 17 Jul, 2025 First submitted to journal 17 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7117693","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":500398185,"identity":"d5ba1824-0678-4b13-90ac-bd042276c73e","order_by":0,"name":"Hua Xing","email":"","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Hua","middleName":"","lastName":"Xing","suffix":""},{"id":500398186,"identity":"e583f82a-d37f-4218-b073-47a521766986","order_by":1,"name":"San Zheng","email":"","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"San","middleName":"","lastName":"Zheng","suffix":""},{"id":500398187,"identity":"5211b410-869b-4ed3-ba0c-9496e82c93b6","order_by":2,"name":"Lijun Yao","email":"","orcid":"","institution":"Tongji University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Lijun","middleName":"","lastName":"Yao","suffix":""},{"id":500398191,"identity":"30befe16-27e5-47b7-a7ef-5fce8028a7b3","order_by":3,"name":"Helin Wang","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Helin","middleName":"","lastName":"Wang","suffix":""},{"id":500398192,"identity":"c58288ec-5a2b-47f2-a75c-339547ef2919","order_by":4,"name":"Yang Chen","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Chen","suffix":""},{"id":500398193,"identity":"113b1627-fe60-472f-80e2-3787fe2ff51b","order_by":5,"name":"Yangyang Fu","email":"","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yangyang","middleName":"","lastName":"Fu","suffix":""},{"id":500398194,"identity":"5b945c3b-d768-4192-910e-fdfa96e86202","order_by":6,"name":"Zhiran Kang","email":"","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Zhiran","middleName":"","lastName":"Kang","suffix":""},{"id":500398195,"identity":"7ab77672-f0a2-41af-8796-6909b8dd1d26","order_by":7,"name":"Xiaojie Su","email":"","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xiaojie","middleName":"","lastName":"Su","suffix":""},{"id":500398196,"identity":"34c47063-77d7-4548-9504-ca2c6c9ea908","order_by":8,"name":"Wuquan Sun","email":"","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Wuquan","middleName":"","lastName":"Sun","suffix":""},{"id":500398197,"identity":"2403cdc0-c64f-4afc-a978-fe6d6f69641e","order_by":9,"name":"Li Gong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8ElEQVRIiWNgGAWjYDADPmbmA8wwjgRRWtiY2RKAWgxI0cLAY0CcFoMbOWaSXyru2LWx83z+XNj2R163gfngbR4Guzx8WqRlzjxLbmPm3SY9s83AcNsBtmRrHobkYrxaJNsOJ7MBtTDzthkkmB3gMZPmYTiQ2EBYC8/jzxAt/N8IapH82HbYDqiFQRpqCxteLZJnnhVbM5w5nAAMZKB7zhkbbjvMZmw5xyAZpxa+48kbb/6oOGzPz3/48WeeMjl5s+PND2+8qbDDqUXhQoYJ0BkMSArAacAAh3ogkO8//vjjDwYGe9xKRsEoGAWjYMQDAAyKUtGe6ufAAAAAAElFTkSuQmCC","orcid":"","institution":"Yueyang Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":true,"prefix":"","firstName":"Li","middleName":"","lastName":"Gong","suffix":""}],"badges":[],"createdAt":"2025-07-14 06:38:57","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7117693/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7117693/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89642434,"identity":"abcb4e76-82f0-4966-9119-4f3659d95334","added_by":"auto","created_at":"2025-08-22 08:18:45","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":63946,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAcupoint locations and imaging measurement in the knee joint\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA: acupoints used in Tuina treatment for KOA; B: the measurement of tibiofemoral joint space (TJS); C: the measurement of quadriceps angle (Q-angle).\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7117693/v1/b255324cfe054669218a7604.jpg"},{"id":89643525,"identity":"c03eb380-661c-4ba2-bf2c-992b5c8de369","added_by":"auto","created_at":"2025-08-22 08:34:45","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":69531,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe flow-chart of data processing and machine learning based modeling\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study involved 355 KOA patients and collected demographic information, medical history, physical examinations, and imaging assessments of each patient. The entire dataset was divided into a training and validation dataset (80%) and a testing dataset (20%). Utilizing the training and validation datasets, we selected eight machine learning algorithms for training, resulting in well-optimized predictive models following parameter tuning. The final classifier was determined by comparing the predictive performance of each trained model. Furthermore, the SHAP method assessed each feature's importance within the best model, enhancing its interpretability.\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7117693/v1/292700b6a512fc76efd7e5f3.jpg"},{"id":89641511,"identity":"b4118940-88f5-47ce-a7a4-5a113d7becf8","added_by":"auto","created_at":"2025-08-22 08:10:44","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":127869,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eROC curves for each trained model in predicting VAS improvement after Tuina treatment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eROC curve for each trained model to predict the improvement of KOA patients after Tuina treatment. Eight different machine learning algorithms, namely Random Forest, XGBoost, KNN, Decision Tree, SVM, Logistic Regression, AdaBoost, and ANN, were selected for training. The area under the curve value for each model was presented in the lower right corner of the graph. KNN: K-Nearest Neighbors; SVM: Support Vector Machine; ANN: Artificial Neural Network.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7117693/v1/aff2f1d0d18e31eb04a93829.jpg"},{"id":89641523,"identity":"f10e4f9b-7c4b-4497-baaf-b2726c17dc75","added_by":"auto","created_at":"2025-08-22 08:10:45","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":72304,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP summary plot of the Random Forest-based classifier in distinguishing patients with different VAS improvement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe relative importance for each feature in the Random Forest-based classifier is obtained by taking the average absolute value of each feature’s SHAP value.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7117693/v1/b9375f68bbc1c0bf999f3709.jpg"},{"id":89642430,"identity":"41ab8cfe-81de-4a0a-961e-6d57fcf90e55","added_by":"auto","created_at":"2025-08-22 08:18:44","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":62707,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComparison the top six features of KOA patients with different knee pain improvement based on graph metrics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA: the grinding test; B: knee joint range of motion (°); C: BMI; D: height; E: imaging examination; F: course of the disease.\u003c/p\u003e\n\u003cp\u003eCeladon column or line: KOA patients with low VAS improvement after Tuina treatment; Yellow column or line: KOA patients with high VAS improvement after Tuina treatment.\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7117693/v1/6432852991bac952e02052fa.jpg"},{"id":89643535,"identity":"484abbbb-8a7a-4b00-9f1b-921812e8002e","added_by":"auto","created_at":"2025-08-22 08:34:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1629334,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7117693/v1/21d61701-bce1-495f-870d-3467fd7e3d41.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Artificial intelligence for predicting the efficacy of Tuina in patients with knee osteoarthritis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eKnee osteoarthritis (KOA) is a common joint disease affecting middle-aged and older adults, characterized primarily by pain and knee dysfunction \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. A systematic review conducted in 2020 found that the overall prevalence of KOA in China was 14.6%, with rates of 10.9% in men and 19.1% in women \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Globally, the number of KOA cases increased by 122.42%, rising from 163.91\u0026nbsp;million in 1990 to 364.58\u0026nbsp;million in 2019 \u003csup\u003e4\u003c/sup\u003e. As a worldwide health issue, KOA severely impacts patients’ physical and psychological well-being, as well as their quality of life \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e, and brings substantial economic and social burdens \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Consequently, there is a growing consensus among patients and physicians regarding the need for safe and effective treatments for KOA.\u003c/p\u003e\u003cp\u003eTuina is a Traditional Chinese Medicine (TCM) physical therapy and a common non-surgical treatment for KOA, as recommended in the clinical guidelines \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Extensive experience in China has demonstrated its safety and effectiveness, supported by numerous rigorous clinical trials \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Research has shown that Tuina therapy benefits KOA by regulating muscle activation patterns, improving ligament and cartilage function, optimizing knee joint mechanics, reducing inflammation, alleviating joint pain, and promoting overall therapeutic outcomes \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. However, the effectiveness of Tuina therapy was influenced by various endogenous factors, including the stage of KOA, the patient’s primary condition, the mechanical structure of the knee joint, and so on \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. A systematic review also found that baseline effusion and synovitis, knee joint effusion, disease severity, drug injections, and more severe symptoms were associated with symptom relief in KOA patients following interventions \u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. These findings indicate that the baseline conditions significantly impact Tuina’s efficacy for KOA, raising an intriguing question: Can Tuina’s efficacy for KOA be predicted based on the patient’s initial condition?\u003c/p\u003e\u003cp\u003eMachine learning (ML), a branch of artificial intelligence, offers the potential to develop predictive models based on existing data \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Most KOA prediction models have primarily focused on predicting disease progression \u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, with various well-performed ML algorithms varied in different research \u003csup\u003e\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e–\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. For instance, a logistic regression (LR) model was used to classify the radiographic progression of KOA \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e, and a support vector machine (SVM) model demonstrated superior performance, achieving approximately 93% accuracy in predicting the structural progression of KOA \u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. These findings indicated that ML had made significant strides in predicting KOA progression. However, its application in predicting treatment efficacy for KOA remains largely unexplored. For instance, Lin T used Lasso regression and LR models to predict the improvement in knee pain for KOA patients undergoing Vitamin D treatment \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. However, no research has been conducted to assess the efficacy of Tuina therapy using ML. Therefore, exploring an optimal ML model to predict the efficacy of Tuina for KOA is of great practical value and importance.\u003c/p\u003e\u003cp\u003eThis study collected clinical data from KOA patients who underwent Tuina therapy and applied statistical and artificial intelligence methods to achieve three objectives: (1) evaluate the improvements in knee function and pain following Tuina therapy; (2) develop an optimal machine learning model to predict Tuina treatment outcomes; (3) use the SHApley Additive exPlanations (SHAP) method to identify the key factors affecting the model’s predictions. The refined model aims to assist Tuina practitioners in forecasting patient responses to treatment, enabling better patient stratification, identifying candidates most likely to benefit, and creating personalized treatment plans. By improving treatment efficiency and outcomes, this approach could reduce the personal and social financial burden caused by KOA.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cb\u003eStudy Population\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis retrospective study focused on the treatment of KOA with Tuina therapy. Clinical data were collected from KOA patients hospitalized in the Tuina Department of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine between February 1, 2016, and December 31, 2023. After data anonymization by the hospital's Information Department, 355 eligible cases were identified. All patients received Tuina therapy following a unified protocol during hospitalization, supplemented by routine nursing care. The Tuina treatment involved the following steps: The patient was positioned supine while the physician treated the affected limb. Techniques included rolling the quadriceps femoris and pressing and kneading specific acupoints such as HeDing (EX-LE2), XueHai (SP10), FuTu (ST32), and LiangQiu (ST34). The patient was then prone, and the physician performed rolling manipulation along the back of the leg, including grasping manipulation at WeiZhong (BL40) and ChengShan (BL57). The treatment concluded with rubbing around the knee joint until a warm sensation penetrated internally. Acupoint locations were based on the World Health Organization (WTO) standards \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e(Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). Rolling manipulations were conducted at a frequency of 140 times/min. Each session lasted 20 min, with five sessions per week over 2 weeks. The inclusion criteria for KOA required patients to meet the clinical classification criteria established by the American College of Rheumatology (ACR) \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e, while the exclusion criteria included: 1) knee injuries or previous surgical treatment; 2) psoriasis, syphilitic neuropathy, metabolic bone disease, or acute trauma; 3) participation in other clinical trials within the previous three months; 4) use of nonsteroidal drugs during or within one month prior to treatment; 5) incomplete medical history or missing physical examination or imaging data; 6) concurrent treatments during hospitalization. This study has been reviewed and approved by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine (Approval No.2025-056). All data collected were only used for research purposes and did not include clinical specimens or human genetic information. This study adheres to the Checklist for Artificial Intelligence in Medical Imaging (CLAIM) for ML research \u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003cb\u003ePredictors\u003c/b\u003e\u003c/p\u003e\u003cp\u003ePatient information for this study were collected from the hospital’s medical record system to establish a KOA database. The database includes four main categories: demographic features (gender, age, height, body weight, and BMI), medical history (hypertension, hyperlipidemia, diabetes, nephrosis, sleep disorder, smoking history, drinking history and the course of the disease), physical examination (knee joint range of motion, knee joint swelling, tenderness around the patella, medial and lateral collateral ligament tenderness, collateral ligament test, McMurray’s sign, drawer test, floating patella test and grinding test) and imaging assessments (Q angles, lateral tibiofemoral joint space, medial tibiofemoral joint space and tibiofemoral joint space ratio). Detailed statistical information for the selected features is presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eFeatures of KOA patients involved in the study\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eFeature\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e\u003cp\u003eNo. (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHigh △VAS \u003csup\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sup\u003e (n = 221)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLow △VAS (n = 134)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTotal / (n = 355)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003efemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e182 (82.35%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e113 (84.39%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e295 (83.10%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e39 (17.65%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (15.61%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e60 (16.90%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge, mean (SD), y\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e68.38 ± 9.41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66.81 ± 10.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e67.78 ± 9.87\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHeight, mean (SD), cm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e160.94 ± 5.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e161.01 ± 5.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e160.97 ± 5.92\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBody weight, mean (SD), kg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e66.63 ± 10.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e64.75 ± 8.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e65.92 ± 10.29\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI \u003csup\u003e\u003cb\u003eb\u003c/b\u003e\u003c/sup\u003e, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e25.68 ± 3.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e24.99 ± 3.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e25.42 ± 3.56\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e125 (56.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71 (52.94%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e196 (55.21%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e96 (43.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e63 (47.06%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e159 (44.79%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHyperlipidemia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e23 (10.45%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15 (11.27%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e38 (10.70%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e198 (89.55%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e119 (88.73%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e317 (89.30%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e32 (14.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23 (17.16%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e55 (15.49%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e189 (85.59%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e111 (82.84%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e300 (84.51%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNephrosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5 (2.27%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6 (4.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e11 (3.10%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e216 (97.73%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e128 (95.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e344 (96.90%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSleep disorder\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e46 (20.83%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e40 (29.85%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e86 (24.23%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e175 (79.17%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e94 (70.15%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e269 (75.77%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmoking history\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e221 (100.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e134 (100.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e355 (100.0%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrinking history\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3 (1.36%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (0.74%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e4 (1.13%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e218 (98.64%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e133 (99.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e351 (98.87%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCourse of the disease, mean (SD), y\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.14 ± 5.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.96 ± 5.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e4.07 ± 5.44\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImaging examination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eunilateral\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e120(54.30%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e72(53.73%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e192(54.08%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ebilateral\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e101(45.70%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e62(46.27%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e163(45.92%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eQ angle, mean (SD), °\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8.98 ± 2.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.60 ± 1.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8.84 ± 2.42\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLateral tibiofemoral joint space, mean (SD), mm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.00 ± 1.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.06 ± 1.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6.02 ± 1.20\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedial tibiofemoral joint space, mean (SD), mm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.60 ± 0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.56 ± 0.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e4.59 ± 0.92\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTibiofemoral joint space ratio (lateral/medial)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.34 ± 0.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.40 ± 0.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.37 ± 0.37\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKnee joint range of motion, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e103.14 ± 22.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e106.57 ± 20.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e104.44 ± 22.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKnee joint swelling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e119 (53.87%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e46 (34.33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e165 (46.48%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e102 (46.13%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e88 (65.67%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e190 (53.52%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTenderness around the patella\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e183 (82.75%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e115 (85.82%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e298 (83.94%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e38 (17.25%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19 (14.18%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e57 (16.06%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedial collateral ligament tenderness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e118 (53.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e61 (45.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e179 (50.42%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e103 (46.59%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e73 (54.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e176 (49.58%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLateral collateral ligament tenderness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e76 (34.32%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e47 (35.07%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e123 (34.65%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e145 (65.68%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e87 (64.93%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e232 (65.35%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCollateral ligament test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e124 (56.15%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60 (44.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e184 (51.83%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e97 (43.85%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e74 (55.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e171 (48.17%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMcMurray’s sign\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e137 (61.94%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e75 (55.97%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e212 (59.72%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e84 (38.06%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e59 (44.03%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e143 (40.28%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrawer test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21 (9.43%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10 (7.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e31 (8.73%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e200 (90.57%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e124 (92.59%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e324 (91.27%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFloating patella test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e46 (20.83%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (15.67%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e67 (18.87%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e175 (79.17%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e113 (84.33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e288 (81.13%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGrinding test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eyes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e116 (52.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e52 (38.81%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e168 (47.32%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eno\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e105 (47.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e82 (61.19%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e187 (52.68%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eAbbreviations:\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003ea. VAS indicates the Visual Analog Scale, a tool used to measure the intensity of certain pain along a continuous line, typically ranging from 0 (no pain) to 10 (maximum pain).\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eb. BMI indicates body mass index calculated as weight in kilograms divided by height in square meters.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003ePhysical examination data were collected from the admission and discharge records of KOA patients to assess the effectiveness of Tuina therapy. Positive rates were used to describe the physical examination data, except for knee joint range of motion and Visual Analogue Scale (VAS) scores. All patients underwent X-ray examinations in our hospital’s radiology department. The model of the X-ray photography system was Samsung DGR-C28U2B/CN. Anteroposterior and lateral radiographs of the knee joint in a weight-bearing position were obtained. Two specialists independently measured all image data, and the results were averaged. The specific measurement methods are as follows: the tibiofemoral joint space (TJS) includes the medial tibiofemoral joint space (mTJS) and lateral tibiofemoral joint space (lTJS). According to Tourville et al. \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, TJS was measured with the following detailed steps: (1) firstly, draw Line C, the axis of the tibial shaft through the tibial plateau; (2) secondly, draw Line D and Line E, the two parallel lines with Line C respectively through lateral and medial intercondylar tubercle; (3) then, draw the widest diameter line of the tibia platform and labeled Line F; (4) lastly, identify the mid-point between intercondylar tubercle and the ipsilateral tibial platform border, and through the two mid-point draw Line A and Line B, respectively paralleled with Line E and Line D. The medial and lateral tibiofemoral joint spaces were measured on the two reference lines parallel to Line F (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). The quadriceps angle (Q-angle), as defined by Brattstrom, measures the angle between two lines: the one from the anterior-superior iliac spine to the mid-patella and the other from the mid-patella to the tibial tubercle \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. The system automatically calculated the Q-angle using X-ray images based on the points and lines identified (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC).\u003c/p\u003e\u003cp\u003eTuina department residents documented all medical records at Yueyang Hospital, who specialized in Tuina therapy and underwent standardized training. Experienced chief physicians provided training and oversight for the physical examination procedures to ensure accuracy and consistency.\u003c/p\u003e\u003cp\u003e\u003cb\u003eOutcomes\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe VAS is a validated tool for measuring both acute and chronic pain. In this study, we evaluated the treatment effect of Tuina therapy by calculating the difference in VAS before and after treatment (△VAS). Notably, if patients experienced bilateral knee pain before treatment, the VAS score was recorded for the more severely affected knee, and so was △VAS. Patients assessed their pain levels with guidance from residents. We utilized a 10 cm sliding ruler marked with ten scales, ranging from 0 to 10, where 0 indicates no pain, and a 10 denotes unbearable pain. Patients selected the corresponding scale according to their feelings. In this study, a high △VAS was defined as a difference of 4 to 6 between pre-and post-treatment VAS scores, while a low △VAS was defined as a difference of 0 to 3. A High △VAS suggests that Tuina therapy has a significant impact on reducing pain in KOA patients.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMachine learning-based modeling\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe used machine learning technologies to build classifiers that can predict the efficacy of Tuina therapy for treating KOA. The model-building process involved four key steps: data preprocessing, feature selection, algorithm selection, and parameter tuning. The prediction performance of the obtained models was evaluated, leading to selecting the best classifier, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The method was described in our previous study \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. We utilized Scikit-learn, a widely used Python-based machine learning library, to train our predictive models (refer to the website: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://scikit-learn.org/stable/\u003c/span\u003e\u003cspan address=\"https://scikit-learn.org/stable/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)\u003csup\u003e28\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eEight representative machine learning algorithms were chosen for training: Random Forest, XGBoost, K-Nearest Neighbors (KNN), Decision Tree, Support Vector Machine (SVM), Logistic Regression, AdaBoost, and Artificial Neural Network (ANN). For each algorithm, we aim to identify a set of optimal parameters. Based on the training and validation datasets, we used a grid search approach to explore the parameter space by selecting a finite set of possible values for each parameter. This method iterated through each combination of parameters, and its predictive performance was evaluated. The parameters yielding the best average prediction performance were then retained.\u003c/p\u003e\u003cp\u003eWe adopted the following four metrics to quantify the predictive performance of each model: precision, recall, F1 score, and the AUC. Precision measures the proportion of actual high △VAS samples among those predicted as high △VAS by the model. Recall indicates the proportion of correctly predicted high △VAS samples among actual high △VAS samples. The F1 score, ranging from 0 to 1, represents the harmonic mean of precision and recall, with higher values indicating better overall model performance. AUC quantifies the model's ability to distinguish between positive and negative classes, with values ranging from 0 to 1; values closer to 1 reflect better predictive performance. The final classifier was determined by comparing the prediction performance of each trained model.\u003c/p\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003eTwo investigators independently entered all data to ensure its completeness and accuracy, while a third investigator reviewed the data. Any identified anomalies were addressed by returning to the original records for correction. An independent professional radiologist conducted the measurement of X-ray images. Python programming software was used for statistical analysis. Numerical variables were expressed as mean ± standard deviation (SD), while categorical variables were described as counts and percentages. The 95% CI for the mean difference or proportions was calculated, a \u003cem\u003ep\u003c/em\u003e-value of less than 0.05 was considered statistically significant. We employed the independent sample \u003cem\u003et-test\u003c/em\u003e for normally distributed data, while Welch's \u003cem\u003et-test\u003c/em\u003e was used for data that did not meet normal distribution criteria. Count data were compared using the Chi-Square test. To enhance the prediction model's interpretability, we quantified each feature’s importance in the prediction model using the SHAP method. SHAP is a representative method for interpreting the predictions of supervised machine learning classifiers. We obtained feature importance values using the metric mean (|SHAP value|), which is the average of the absolute SHAP values across all patients. When |SHAP value| is high, this feature contributes more to the prediction model.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eCode Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe executable code for reproducing the analysis is available at [https://github.com/Helin-Wang/VAS_Prediction] under MIT License. Due to ethical restrictions, the original clinical dataset cannot be publicly shared but may be requested from the corresponding author upon reasonable approval. The repository includes synthetic data demonstrating the code functionality.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cb\u003ePatient characteristics\u003c/b\u003e\u003c/p\u003e\u003cp\u003eA total of 355 KOA patients were included in this study. After Tuina treatment, 221 (62.25% ) patients experienced a reduction in VAS scores of 4 to 6 (High △VAS group), while 134 (37.75%) patients reported a decrease of 0 to 3 (Low △VAS group). 295 (83.10%) were females, and 60 (16.90%) were males. The ages of the participants range from 39 to 89 years, with a mean age of 67.78 years (SD\u0026thinsp;=\u0026thinsp;9.87). The average course of KOA among the patients was 4.07 years SD\u0026thinsp;=\u0026thinsp;5.44). Regarding comorbidities, 196 (55.21%) patients had hypertension, 38 (10.70%) had hyperlipemia, 55 (15.49%) had diabetes, 11 (3.10%) had nephrosis, and 86 (24.23%) reported experiencing insomnia. All patients denied smoking, and only 4 patients (1.13%) had a history of alcohol consumption. Additional detailed information concerning the number, percentage, and \u003cem\u003ep\u003c/em\u003e-value of each feature is presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eThe average Q-angle of the knee joint was 8.84 degrees (SD\u0026thinsp;=\u0026thinsp;2.42). The lateral and medial tibiofemoral joint spaces were 6.02 mm (SD\u0026thinsp;=\u0026thinsp;1.20) and 4.59 mm (SD\u0026thinsp;=\u0026thinsp;0.92). The tibiofemoral joint space ratio was 1.37 (SD\u0026thinsp;=\u0026thinsp;0.37). Before treatment, the average knee flexion and extension angle was 104.44 degrees (SD\u0026thinsp;=\u0026thinsp;22.03). Among the patients, 165 patients (46.48%) presented with swollen joints. Tenderness assessments revealed that 298 patients (83.94%) had peripatellar knee tenderness, 179 (50.42%) had medial tenderness, and 123 (34.65%) had lateral tenderness. The positive rate of the collateral ligament test, McMurray\u0026rsquo;s sign, drawer test, floating patella test, and grinding test were 184 (51.83%), 212 (59.72%), 31 (8.73%), 67(18.87%), and 168 (47.32%), respectively (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eEffects of Tuina therapy on physical examination in patients with KOA\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAfter Tuina treatment, the flexion and extension angle of the patient\u0026rsquo;s knee joint improved to 116.46 degrees (SD\u0026thinsp;=\u0026thinsp;0.86), representing a 10.03% increase from the baseline. The average VAS score decreased to 2.82 (SD\u0026thinsp;=\u0026thinsp;0.05), which was 3.74 points lower than the pre-treatment score (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Additionally, comparisons of other physical examination results before and after treatment revealed significant reductions in the positive rate of knee joint swelling, tenderness around the patella, medial and lateral collateral ligament tenderness, collateral ligament test, McMurray\u0026rsquo;s sign, floating test, and grinding test. These findings indicate that Tuina therapy significantly improves patients\u0026rsquo; knee joint function and alleviates knee pain.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eChanges in physical examinations of KOA patients before and after Tuina treatment\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ePhysical examination parameter\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePositive No./total (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePositive No./total (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eMD \u003csup\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sup\u003e/PR \u003csup\u003e\u003cem\u003eb\u003c/em\u003e\u003c/sup\u003e change (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e value \u003csup\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e95% CI\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003ebaseline\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTuina\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKnee joint range of motion, mean (SD \u003csup\u003e\u003cb\u003ed\u003c/b\u003e\u003c/sup\u003e), \u0026deg;\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e106.43\u0026thinsp;\u0026plusmn;\u0026thinsp;1.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e116.46\u0026thinsp;\u0026plusmn;\u0026thinsp;0.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-10.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[6.46, 11.91]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKnee joint swelling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e165/355 (46.48)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e99/355 (27.89)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e18.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.12, 0.26]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTenderness around the patella\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e298/355 (83.97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e235/355 (66.20)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e17.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.11, 0.24]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedial collateral ligament tenderness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e179/355 (50.42)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e125/355 (35.21)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e15.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.08, 0.22]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLateral collateral ligament tenderness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e123/355 (34.65)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e81/355 (22.82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e11.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.0005***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.05, 0.18]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCollateral ligament test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e184/355 (51.83)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e106/355 (29.86)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e21.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.15, 0.29]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMcMurray\u0026rsquo;s sign\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e212/355 (59.72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e146/355 (41.13)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e18.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.11, 0.26]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrawer test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e31/355 (8.73)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20/355 (5.63)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.1099\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.00, 0.07]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFloating patella test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e67/355 (18.87)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e30/355 (8.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e10.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.05, 0.15]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGrinding test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e168/355 (47.32)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e128/355 (36.06)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e11.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.0023**\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[0.04, 0.19]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVAS \u003csup\u003e\u003cb\u003ee\u003c/b\u003e\u003c/sup\u003e, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.0001***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e[3.61, 3.88]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eAbbreviations:\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003ea. MD: mean difference.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eb. PR: positive rate.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003ec. calculated by using Holm-Bonferroni correction. Special symbols are used to denote statistical differences in \u003cem\u003eP\u003c/em\u003e values, such as *: \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant; **: \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01; ***: \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003ed. SD: standard deviation.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003ee. VAS: visual analog scale.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eMachine-learning based Tuina efficacy prediction\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe employed supervised machine learning to develop an effective binary classifier for assessing the degree of knee pain relief following Tuina therapy. For this purpose, we selected eight machine learning algorithms: Random Forest, XGBoost, KNN, Decision Tree, SVM, Logistic Regression, AdaBoost, and ANN. These models were trained and validated using Dataset 1 and tested with Dataset 2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Our results showed that the Random Forest-based model outperformed the others, achieving an F1 score of 0.82 and an AUC value of 0.85, respectively. The results of the other models are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The ROC curves for each trained model, reflecting their ability to predict changes in VAS scores resulting from Tuina treatment, are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Our results suggest that artificial intelligence can accurately predict the effectiveness of Tuina treatment for KOA patients based on their demographic characteristics, medical history, physical examinations, and imaging assessments.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ethe performance of each trained model in predicting VAS improvement after Tuina treatment\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAlgorithm\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eF1-Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandom Forest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.82\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDecision Tree\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.82\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogistic Regression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAdaBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eANN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.71\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eFeature importance\u003c/b\u003e\u003c/p\u003e\u003cp\u003eOur dataset includes 28 features used to develop our models, each with varying significance. We calculated each feature\u0026rsquo;s SHAP value in the trained Random Forest-based classifier (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). According to the SHAP value, the \u0026ldquo;Grinding test\u0026rdquo; ranked first among all features in this model. It means that the \u0026ldquo;Grinding test\u0026rdquo; result contributed the most to predicting the efficacy of Tuina treatment. \u0026ldquo;Knee joint range of motion\u0026rdquo;, \u0026ldquo;BMI\u0026rdquo;, \u0026ldquo;Height\u0026rdquo;, \u0026ldquo;Imaging examination\u0026rdquo;, and \u0026ldquo;Course of the disease\u0026rdquo; also ranked top. Then, we compared the top six features of KOA patients with different knee pain improvements after Tuina treatment (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). We found that a KOA patient had a positive grind test, and Tuina was more effective in reducing knee pain. Besides, \u0026ldquo;Knee joint range of motion\u0026rdquo;, \u0026ldquo;BMI\u0026rdquo;, \u0026ldquo;Height\u0026rdquo;, and \u0026ldquo;Course of disease\u0026rdquo; also had a substantial predictive value for the efficacy of Tuina. In addition, we found that patients with knee KOA on both sides had worse Tuina efficacy than those with unilateral knee KOA.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study is the first attempt to use artificial intelligence to predict the effectiveness of Tuina therapy in patients with KOA. We developed a comprehensive KOA database encompassing 28 clinical characteristics. Our statistical analyses demonstrated that Tuina therapy significantly alleviated knee pain and improved knee joint function in the patients examined. By employing various machine learning algorithms, we constructed predictive models to evaluate the response of KOA patients to Tuina treatment, with the Random Forest algorithm-based model achieving the best predictive performance. Key factors such as grinding test, knee joint range of motion, and BMI significantly contributed to the model\u0026rsquo;s predictive accuracy. These findings provide valuable insights into the application of Tuina therapy in TCM and support a shift towards a more predictive, preventive, and personalized approach in TCM \u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eOur study demonstrated that Tuina therapy significantly reduced knee pain and improved joint function in patients with KOA (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The VAS score decreased from 6.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.49 to 2.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05, with an average reduction of 3.74. These results align with existing literature, further supporting Tuina\u0026rsquo;s efficacy as a non-pharmacological treatment for KOA. For instance, a randomized controlled trial by Liu et al. showed that Tuina therapy substantially decreased knee pain, stiffness, and limitations in patient\u0026rsquo;s daily activities compared to health education \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. However, they did not assess other aspects of knee function. Xu et al. reported that a six-week Tuina program significantly alleviated knee pain and improved patient\u0026rsquo;s mood and daily functioning \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Zhu et al. observed that the improved joint function was likely linked to the increased strength in knee extensors and flexors, though they noted no improvement in the range of motion after a two-week, three-times-per-week regimen \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. Our study also showed that Tuina improved physical examination results in most KOA patients, including reductions in the incidence of knee joint swelling, tenderness around the patella, medial and lateral collateral ligament tenderness, collateral ligament tests, McMurray's sign, floating patella test, and grinding test (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). However, these improvements do not necessarily indicate underlying knee joint structure changes, such as meniscal tears, ligament damage or cartilage degeneration. The observed benefits primarily reflect reduced pain and improved function, consistent with previous clinical studies. Further research, including imaging studies like MRI, is needed to assess whether Tuina therapy can impact joint structure. In summary, Tuina therapy, a TCM technique involving manual manipulation of the body surface, joints, and acupoints, improves local circulation and relieves muscle tension and spasms \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. While manual therapy shows promise in reducing pain associated with KOA, yet it still has its limitations \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. The effectiveness of Tuina for KOA necessitates further comprehensive evaluation, considering various influencing factors.\u003c/p\u003e\u003cp\u003eIn this study, we pioneered machine learning to model the therapeutic effects of Tuina therapy for KOA. By comparing eight different machine learning algorithms, we identified that the model based on the random forest algorithm best predicted the efficacy of the Tuina treatment, achieving an F1 score of 0.82 and an AUC value of 0.85 (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e \u0026amp; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The application of machine learning in KOA research is expanding, encompassing areas such as disease diagnosis, efficacy prediction, risk assessment, and progression prediction et al. \u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. However, research specifically focused on predicting the therapeutic effects of TCM treatments like Tuina therapy remains limited.\u003c/p\u003e\u003cp\u003eThe random forest algorithm effectively integrates various factors influencing Tuina treatment for KOA by constructing multiple decision trees and synthesizing their predictive outcomes \u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. These factors include patient demographics, medical history, imaging results, and physical examinations at admission. The algorithm excels in capturing complex, nonlinear interactions among these factors, thus improving the accuracy of efficacy prediction. Additionally, the random forest algorithm possesses excellent robustness and generalization ability, minimizing sensitivity to noise and outliers and ensuring high predictive accuracy even with new data, which gives it a distinct advantage when dealing with diverse datasets \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Compared to traditional methods, which rely on observational studies and randomized controlled trials to evaluate the efficacy of Tuina, our study offers significant advantages. Traditional clinical research often fails to capture the complex interactions between individual patient differences and treatment parameters. In contrast, machine learning models integrate multidimensional patient data effectively to create more accurate predictive models \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. This capability allows for anticipating potential treatment outcomes, providing a more reliable basis for clinical decision-making. In summary, this study demonstrated the potential of machine learning in predicting the effects of TCM treatments and also provides valuable insights for future research. It paves the way for exploring the KOA pathogenesis, risk prediction, and the development of personalized treatment plans using this technology.\u003c/p\u003e\u003cp\u003eTo enhance the model's interpretability and determine which features significantly influence prediction outcomes, we applied the SHAP method to sort the importance of the 28 features (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The top three features identified were the grinding test, knee joint range of motion, and BMI, highlighting their importance in the model\u0026rsquo;s predictive performance. Other factors also contribute, albeit with relatively minor contributions. The grinding test is a physical examination used to evaluate knee function, often associated to patella dysfunction, which is common in KOA \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. A positive grinding test, typically indicating patellofemoral grind, may be linked to synovitis in KOA, making it a reliable predictor of KOA symptoms and pain improvement \u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. A reduction in knee joint range of motion is a significant clinical indicator for KOA patients \u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. Our study demonstrated that Tuina therapy increased the average knee joint range of motion by 10 degrees, effectively improving knee function (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This factor ranks high in the model, suggesting that early assessment and intervention of knee joint range of motion are crucial for managing and treating KOA. BMI is a significant risk factor for KOA. High BMI values not only significantly increase the incidence and progression of KOA but are also associated with more severe knee pain and functional impairment \u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. In addition, imaging examinations revealing unilateral or bilateral osteoarthritis can predict the efficacy of Tuina treatment for KOA. Compared to patients with bilateral KOA, those with unilateral KOA patients often had more asymmetrical inter-limbs foot posture, which was significantly associated with the K/L grade in KOA \u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. This finding supports the opinion that the disease severity is an important factor affecting the prognosis of KOA. Although clinical guidelines commonly incorporate K/L grade\u0026thinsp;\u0026ge;\u0026thinsp;2 combined with clinical symptoms as diagnostic criteria for KOA, the K/L grading system was not adopted as an objective outcome measure for treatment efficacy in this study due to its well-documented inter-observer variability in radiographic interpretation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eLimitations\u003c/b\u003e\u003c/p\u003e\u003cp\u003eOur study has some limitations. Firstly, although we selected features from inpatient medical records that might relate to the efficacy of Tuina therapy based on published literature and our long-term experience, many other potentially influencing features were not included. For instance, gait analysis results were excluded due to the limited number of patients undergoing this assessment, leading to insufficient data for further analysis. Additionally, certain biomarkers that may be associated with KOA were not included in our investigation. Regarding MRI data, we only used the results derived from MRI scans, not the full images. Moreover, our sample size remains limited, though we are actively working to expand the number of patients in our database. Lastly, it is noteworthy that most participants in this study were female, which may introduce a bias when applying the predictive model to male populations.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur study demonstrated that Tuina therapy significantly alleviated knee pain and improved knee joint function in patients with KOA. Based on these findings, we developed and validated an artificial intelligence model to predict the efficacy of Tuina therapy for KOA by calculating patients' demographic information, medical history, imaging data, and physical examination results. This research offers Tuina practitioners a valuable tool for scientifically assessing the therapeutic effects of the treatment, thereby improving the precision of clinical decision-making.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eKOA: Knee Osteoarthritis\u003c/p\u003e\n\u003cp\u003eTCM: Traditional Chinese Medicine\u003c/p\u003e\n\u003cp\u003eML: Machine Learning\u003c/p\u003e\n\u003cp\u003eLR: Logistic Regression\u003c/p\u003e\n\u003cp\u003eSVM: Support Vector Machine\u003c/p\u003e\n\u003cp\u003eSHAP: SHApley Additive exPlanations\u003c/p\u003e\n\u003cp\u003eWTO: World Health Organization\u003c/p\u003e\n\u003cp\u003eACR: the American College of Rheumatology\u003c/p\u003e\n\u003cp\u003eBMI: Body Mass Index\u003c/p\u003e\n\u003cp\u003eVAS: Visual Analogue Scale\u003c/p\u003e\n\u003cp\u003eTJS: Tibiofemoral Joint Space\u003c/p\u003e\n\u003cp\u003eKNN: K-Nearest Neighbors\u003c/p\u003e\n\u003cp\u003eANN: Artificial Neural Network\u003c/p\u003e\n\u003cp\u003eAUC: Area Under the Curve\u003c/p\u003e\n\u003cp\u003eSD: Standard Deviation\u003c/p\u003e\n\u003cp\u003eMRI: Magnetic Resonance Imaging\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study has been reviewed and approved by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine (Approval No.2025-056). All procedures adhered to the Declaration of Helsinki to ensure the rights and welfare of participants.\u0026nbsp;The requirement for informed consent was waived by the Ethics Committee because:\u0026nbsp;(1) The study involved only retrospective analysis of anonymized clinical data; (2) All data were processed in strict compliance with China\u0026apos;s Regulations on the Management of Human Genetic Resources and the Guidelines for Ethical Review of Clinical Research;\u0026nbsp;(3) This exemption is consistent with national regulations permitting waiver of consent for retrospective studies posing minimal risk.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNA\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe original datasets cannot be shared publicly due to patient privacy restrictions. Researchers may apply for access by contacting the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine ([email protected]) or the corresponding author Li Gong ([email protected]).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRole of the funding source\u003c/strong\u003e\u003c/p\u003e\n\u003cp skip=\"true\"\u003eThis work was supported by the National Natural Science Foundation of China (Grant Number: 8197151584); Shanghai Key Clinical Specialty Construction Project (Grant Number: Shslczdzk04001); the Sailing program of Shanghai Science and Technology Commission (Grant Number: 22YF1444300); Shanghai \u0026quot;Science and Technology Innovation Action Plan\u0026quot; Social Development Science and Technology Research Project (23DZ1204004); Shanghai \u0026quot;Science and Technology Innovation Action Plan\u0026quot; Medical Innovation Research Special Project (23Y11921700); Clinical Research Project of Shanghai Municipality Health Commission (20234Y0077, 202140037).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp skip=\"true\"\u003eH.X. and L.G. participated in the study design; W.S, L.Y., and L.G. participated in the supervision; H.W. and Y.C. participated in data analysis and modeling; S.Z. and Y.F. participated in data collection; Z.K. and X.S. participated in data analysis; H.X., S.Z., and L.Y. contributed to the manuscript writing; All authors have read and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp skip=\"true\"\u003eWe also gratefully acknowledge Figdraw (www.figdraw.com) for providing the anatomical human model template in Figure 1A (used for acupoint labeling) and the schematic framework in Figure 2. The final figures were adapted and annotated by the authors.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFelson DT. Clinical practice. Osteoarthritis of the knee. \u003cem\u003eN Engl J Med. \u003c/em\u003e2006;354(8):841-848. doi:10.1056/NEJMcp051726.1.\u003c/li\u003e\n\u003cli\u003eZhang S, Huang R, Guo G, et al. Efficacy of traditional Chinese exercise for the treatment of pain and disability on knee osteoarthritis patients: a systematic review and meta-analysis of randomized controlled trials. \u003cem\u003eFront Public Health. \u003c/em\u003e2023;11:1168167. doi:10.3389/fpubh.2023.1168167.2.\u003c/li\u003e\n\u003cli\u003eLi D, Li S, Chen Q, Xie X. The Prevalence of Symptomatic Knee Osteoarthritis in Relation to Age, Sex, Area, Region, and Body Mass Index in China: A Systematic Review and Meta-Analysis. \u003cem\u003eFront Med (Lausanne). \u003c/em\u003e2020;7:304. doi:10.3389/fmed.2020.00304.3.\u003c/li\u003e\n\u003cli\u003eLong H, Liu Q, Yin H, et al. Prevalence Trends of Site-Specific Osteoarthritis From 1990 to 2019: Findings From the Global Burden of Disease Study 2019. \u003cem\u003eArthritis Rheumatol. \u003c/em\u003e2022;74(7):1172-1183. doi:10.1002/art.42089.4.\u003c/li\u003e\n\u003cli\u003eVitaloni M, Botto-van Bemden A, Sciortino Contreras RM, et al. Global management of patients with knee osteoarthritis begins with quality of life assessment: a systematic review. \u003cem\u003eBMC Musculoskelet Disord. \u003c/em\u003e2019;20(1):493. doi:10.1186/s12891-019-2895-3.5.\u003c/li\u003e\n\u003cli\u003eCross M, Smith E, Hoy D, et al. The global burden of hip and knee osteoarthritis: estimates from the global burden of disease 2010 study. \u003cem\u003eAnn Rheum Dis. \u003c/em\u003e2014;73(7):1323-1330. doi:10.1136/annrheumdis-2013-204763.6.\u003c/li\u003e\n\u003cli\u003eChen WH, Liu XX, Tong PJ, Zhan HS. Expert consensus on traditional Chinese medicine diagnosis and treatment of knee osteoarthritis (2015 edition). \u003cem\u003eThe Journal of Traditional Chinese Orthopedics and Traumatology. \u003c/em\u003e2015;27(07):4-5.\u003c/li\u003e\n\u003cli\u003eLiu K, Zhan Y, Zhang Y, et al. Efficacy and safety of Tuina (Chinese Therapeutic Massage) for knee osteoarthritis: A randomized, controlled, and crossover design clinical trial. \u003cem\u003eFront Med (Lausanne). \u003c/em\u003e2023;10:997116. doi:10.3389/fmed.2023.997116.48.\u003c/li\u003e\n\u003cli\u003eZhu B, Ba H, Kong L, et al. The effects of manual therapy in pain and safety of patients with knee osteoarthritis: a systematic review and meta-analysis. \u003cem\u003eSyst Rev. \u003c/em\u003e2024;13(1):91. doi:10.1186/s13643-024-02467-7.39.\u003c/li\u003e\n\u003cli\u003eXu H, Zhao C, Guo G, et al. The Effectiveness of Tuina in Relieving Pain, Negative Emotions, and Disability in Knee Osteoarthritis: A Randomized Controlled Trial. \u003cem\u003ePain Med. \u003c/em\u003e2023;24(3):244-257. doi:10.1093/pm/pnac127.8.\u003c/li\u003e\n\u003cli\u003eWang M, Liu L, Zhang CS, et al. Mechanism of Traditional Chinese Medicine in Treating Knee Osteoarthritis. \u003cem\u003eJ Pain Res. \u003c/em\u003e2020;13:1421-1429. doi:10.2147/jpr.S247827.10.\u003c/li\u003e\n\u003cli\u003eTang C, Kong LJ, Yao CJ, et al. Research Progress of Tuina in the Treatment of Knee Osteoarthritis from the Perspective of Mechanical Homeostasis. \u003cem\u003eGuiding Journal of Traditional Chinese Medicine and Pharmacy. \u003c/em\u003e2021;27(09):159-163. doi:10.13862/j.cnki.cn43-1446/r.2021.09.035.11.\u003c/li\u003e\n\u003cli\u003eCarr AJ, Robertsson O, Graves S, et al. Knee replacement. \u003cem\u003eLancet. \u003c/em\u003e2012;379(9823):1331-1340. doi:10.1016/s0140-6736(11)60752-6.12.\u003c/li\u003e\n\u003cli\u003eMaricar N, Callaghan MJ, Felson DT, O\u0026apos;Neill TW. Predictors of response to intra-articular steroid injections in knee osteoarthritis--a systematic review. \u003cem\u003eRheumatology (Oxford). \u003c/em\u003e2013;52(6):1022-1032. doi:10.1093/rheumatology/kes368.13.\u003c/li\u003e\n\u003cli\u003eBinvignat M, Pedoia V, Butte AJ, et al. Use of machine learning in osteoarthritis research: a systematic literature review. \u003cem\u003eRMD Open. \u003c/em\u003e2022;8(1). doi:10.1136/rmdopen-2021-001998.14.\u003c/li\u003e\n\u003cli\u003eJamshidi A, Pelletier JP, Martel-Pelletier J. Machine-learning-based patient-specific prediction models for knee osteoarthritis. \u003cem\u003eNat Rev Rheumatol. \u003c/em\u003e2019;15(1):49-60. doi:10.1038/s41584-018-0130-5.53.\u003c/li\u003e\n\u003cli\u003eDunn CM, Sturdy C, Velasco C, et al. Peripheral Blood DNA Methylation-Based Machine Learning Models for Prediction of Knee Osteoarthritis Progression: Biologic Specimens and Data From the Osteoarthritis Initiative and Johnston County Osteoarthritis Project. \u003cem\u003eArthritis Rheumatol. \u003c/em\u003e2023;75(1):28-40. doi:10.1002/art.42316.16.\u003c/li\u003e\n\u003cli\u003eBonakdari H, Pelletier JP, Blanco FJ, et al. Single nucleotide polymorphism genes and mitochondrial DNA haplogroups as biomarkers for early prediction of knee osteoarthritis structural progressors: use of supervised machine learning classifiers. \u003cem\u003eBMC Med. \u003c/em\u003e2022;20(1):316. doi:10.1186/s12916-022-02491-1.17.\u003c/li\u003e\n\u003cli\u003eTiulpin A, Klein S, Bierma-Zeinstra SMA, et al. Multimodal Machine Learning-based Knee Osteoarthritis Progression Prediction from Plain Radiographs and Clinical Data. \u003cem\u003eSci Rep. \u003c/em\u003e2019;9(1):20038. doi:10.1038/s41598-019-56527-3.18.\u003c/li\u003e\n\u003cli\u003eLin T, Peng S, Lu S, et al. Prediction of knee pain improvement over two years for knee osteoarthritis using a dynamic nomogram based on MRI-derived radiomics: a proof-of-concept study. \u003cem\u003eOsteoarthritis Cartilage. \u003c/em\u003e2023;31(2):267-278. doi:10.1016/j.joca.2022.10.014.19.\u003c/li\u003e\n\u003cli\u003eHuang LX, Gang WJ, Li SY, Huang YM, Tan YS. World Health Organization standard acupuncture and moxibustion location (Western Pacific region): Chinese and English comparison. \u003cem\u003ePeople\u0026apos;s Medical Publishing House. \u003c/em\u003e2010.64. ISBN: 9787117123327. https://books.google.com.hk/books?id=4TCjYgEACAAJ.\u003c/li\u003e\n\u003cli\u003eAltman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association.\u003cem\u003e Arthritis Rheum\u003c/em\u003e. 1986;29(8):1039-49. doi: 10.1002/art.1780290816.\u003c/li\u003e\n\u003cli\u003eMongan J, Moy L, Kahn CE, Jr. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): A Guide for Authors and Reviewers.\u003cem\u003e Radiol Artif Intell\u003c/em\u003e. 2020;2(2):e200029. doi: 10.1148/ryai.2020200029.\u003c/li\u003e\n\u003cli\u003eTourville TW, Johnson RJ, Slauterbeck JR, Naud S, Beynnon BD. Assessment of early tibiofemoral joint space width changes after anterior cruciate ligament injury and reconstruction: a matched case-control study. \u003cem\u003eAm J Sports Med. \u003c/em\u003e2013;41(4):769-778. doi:10.1177/0363546513477838.20\u003c/li\u003e\n\u003cli\u003eDupuis DE, Beynnon BD, Richard MJ, Novotny JE, Skelly JM, Cooper SM. Precision and accuracy of joint space width measurements of the medial compartment of the knee using standardized MTP semi-flexed radiographs. \u003cem\u003eOsteoarthritis Cartilage. \u003c/em\u003e2003;11(10):716-724. doi:10.1016/s1063-4584(03)00158-4.21.\u003c/li\u003e\n\u003cli\u003eMilovanović D, Begović N, Bukva B, et al. The Influence of the Q-Angle and Muscle Strength on Idiopathic Anterior Knee Pain in Adolescents. \u003cem\u003eMedicina (Kaunas). \u003c/em\u003e2023;59(6). doi:10.3390/medicina59061016.22.\u003c/li\u003e\n\u003cli\u003eZhou S, Ma Q, Lou Y, et al. Machine learning to predict clinical remission in depressed patients after acute phase selective serotonin reuptake inhibitor treatment. \u003cem\u003eJ Affect Disord. \u003c/em\u003e2021;287:372-379. doi: 10.1016/j.jad.2021.03.079.37.\u003c/li\u003e\n\u003cli\u003eWei HL, Yu YS, Wang MY, et al. Exploring potential neuroimaging biomarkers for the response to non-steroidal anti-inflammatory drugs in episodic migraine. \u003cem\u003eJ Headache Pain. \u003c/em\u003e2024;25(1):104. doi:10.1186/s10194-024-01812-4.38.\u003c/li\u003e\n\u003cli\u003eBarrett M, Boyne J, Brandts J, et al. Artificial intelligence supported patient self-care in chronic heart failure: a paradigm shift from reactive to predictive, preventive and personalised care. \u003cem\u003eEpma j. \u003c/em\u003e2019;10(4):445-464. doi:10.1007/s13167-019-00188-9.24.\u003c/li\u003e\n\u003cli\u003eZhu Q, Li J, Fang M, Gong L, Sun W, Zhou N. [Effect of Chinese massage (Tui Na) on isokinetic muscle strength in patients with knee osteoarthritis]. \u003cem\u003eJ Tradit Chin Med. \u003c/em\u003e2016;36(3):314-320. doi:10.1016/s0254-6272(16)30043-7.50.\u003c/li\u003e\n\u003cli\u003eLee NW, Kim GH, Heo I, et al. Chuna (or Tuina) Manual Therapy for Musculoskeletal Disorders: A Systematic Review and Meta-Analysis of Randomized Controlled Trials. \u003cem\u003eEvid Based Complement Alternat Med. \u003c/em\u003e2017;2017:8218139. doi:10.1155/2017/8218139.51.\u003c/li\u003e\n\u003cli\u003eLee LS, Chan PK, Wen C, et al. Artificial intelligence in diagnosis of knee osteoarthritis and prediction of arthroplasty outcomes: a review. \u003cem\u003eArthroplasty. \u003c/em\u003e2022;4(1):16. doi:10.1186/s42836-022-00118-7.54.\u003c/li\u003e\n\u003cli\u003eBreiman L. Random Forests. \u003cem\u003eMachine Learning. \u003c/em\u003e2001;45(1):5-32. doi:10.1023/A:1010933404324.55.\u003c/li\u003e\n\u003cli\u003eMohapatra N, Shreya K, Chinmay A. Optimization of the Random Forest Algorithm. Paper presented at: Advances in Data Science and Management; 2020//, 2020; Singapore.56.\u003c/li\u003e\n\u003cli\u003eRajkomar A, Dean J, Kohane I. Machine Learning in Medicine. \u003cem\u003eN Engl J Med. \u003c/em\u003e2019;380(14):1347-1358. doi:10.1056/NEJMra1814259.57.\u003c/li\u003e\n\u003cli\u003eIversen MD, Price LL, von Heideken J, Harvey WF, Wang C. Physical examination findings and their relationship with performance-based function in adults with knee osteoarthritis. \u003cem\u003eBMC Musculoskelet Disord. \u003c/em\u003e2016;17:273. doi:10.1186/s12891-016-1151-3.33.\u003c/li\u003e\n\u003cli\u003eDeng H, Wu Y, Fan Z, Tang W, Tao J. The association between patellofemoral grind and synovitis in knee osteoarthritis: data from the osteoarthritis initiative. \u003cem\u003eFront Med (Lausanne). \u003c/em\u003e2023;10:1231398. doi:10.3389/fmed.2023.1231398.34.\u003c/li\u003e\n\u003cli\u003eSteultjens MPM, Dekker J, van Baar ME, Oostendorp RAB, Bijlsma JWJ. Range of joint motion and disability in patients with osteoarthritis of the knee or hip. \u003cem\u003eRheumatology. \u003c/em\u003e2000;39(9):955-961. doi:10.1093/rheumatology/39.9.955.58.\u003c/li\u003e\n\u003cli\u003eEpskamp S, Dibley H, Ray E, et al. Range of motion as an outcome measure for knee osteoarthritis interventions in clinical trials: an integrated review. \u003cem\u003ePhysical Therapy Reviews. \u003c/em\u003e2020;25(5-6):462-481. doi:10.1080/10833196.2020.1867393.59.\u003c/li\u003e\n\u003cli\u003eBlagojevic M, Jinks C, Jeffery A, Jordan KP. Risk factors for onset of osteoarthritis of the knee in older adults: a systematic review and meta-analysis. \u003cem\u003eOsteoarthritis Cartilage. \u003c/em\u003e2010;18(1):24-33. doi:10.1016/j.joca.2009.08.010.60.\u003c/li\u003e\n\u003cli\u003eZheng H, Chen C. Body mass index and risk of knee osteoarthritis: systematic review and meta-analysis of prospective studies. \u003cem\u003eBMJ Open. \u003c/em\u003e2015;5(12):e007568. doi:10.1136/bmjopen-2014-007568.61.\u003c/li\u003e\n\u003cli\u003eChen Z, Ye X, Shen Z, et al. Comparison of the Asymmetries in Foot Posture and Properties of Gastrocnemius Muscle and Achilles Tendon Between Patients With Unilateral and Bilateral Knee Osteoarthritis. \u003cem\u003eFront Bioeng Biotechnol. \u003c/em\u003e2021;9:636571. doi:10.3389/fbioe.2021.636571.35.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Knee osteoarthritis, Manual therapy, Machine learning, Predictive model, Pain management","lastPublishedDoi":"10.21203/rs.3.rs-7117693/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7117693/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eKnee osteoarthritis (KOA), a prevalent condition impacting middle-aged and older adults' quality of life, is increasing globally. Tuina, a Traditional Chinese Medicine technique, shows efficacy in reducing KOA pain and improving function, but response varies. This study aimed to develop a supervised machine learning classifier to predict Tuina efficacy for KOA, aiding personalized treatment planning.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis retrospective, registry-based, single-center prognostic study enrolled 355 KOA patients from the Tuina Department at Yueyang Hospital (Shanghai, China) between February 1, 2016, and December 31, 2023. All received standardized Tuina therapy (20-min sessions, 5\u0026times;/week for 2 weeks). Efficacy was assessed via ΔVAS (post-treatment minus baseline VAS), categorized as high (ΔVAS\u0026thinsp;=\u0026thinsp;4\u0026ndash;6) or low efficacy (ΔVAS\u0026thinsp;=\u0026thinsp;0\u0026ndash;3). Eight machine learning models (e.g., Random Forest, SVM) were trained using 80% of the data (demographics, medical history, imaging assessments, physical exam findings, baseline VAS) to predict efficacy, validated on the remaining 20%. Statistical analysis used T-tests and Chi-square tests; model performance was evaluated via F1-score and AUC. Data analysis was performed from January 2024 to March 2025.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe average reduction in VAS scores was 3.74. Among the eight trained machine learning models, the Random Forest-based model achieved the best predictive performance for the efficacy of Tuina treatment. The top six features influencing the model included the grinding test, knee joint range of motion, body mass index (BMI), height, imaging examination, and disease course.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eArtificial intelligence models can reliably predict the efficacy of Tuina therapy in KOA patients. This study provides a valuable reference for Tuina practitioners in scientifically evaluating the effectiveness of KOA treatments.\u003c/p\u003e\u003ch2\u003eTrial registration:\u003c/h2\u003e\u003cp\u003e This retrospective study was approved by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, affiliated with Shanghai University of Traditional Chinese Medicine (No.2025-056).\u003c/p\u003e","manuscriptTitle":"Artificial intelligence for predicting the efficacy of Tuina in patients with knee osteoarthritis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-22 08:10:39","doi":"10.21203/rs.3.rs-7117693/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2025-09-05T03:53:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"321693145414371166436219229332322032930","date":"2025-08-21T07:13:28+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-14T08:35:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-11T19:21:42+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-07-17T15:25:54+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-17T12:22:23+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2025-07-17T08:48:35+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0b24e7bc-2564-494c-9e44-ae46d1da8dc0","owner":[],"postedDate":"August 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-08-22T08:10:39+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-22 08:10:39","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7117693","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7117693","identity":"rs-7117693","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}