膝骨关节炎-肌少症共病风险预测模型的构建与验证OA

Objective:To construct and validate a risk prediction model for the comorbidity of knee osteoarthritis(KOA)and sarcope-nia.Methods:Elderly individuals recruited from the Yuyuan Community Health Service Center in Huangpu District of Shanghai from Janu-ary to December 2024 were selected as the study subjects.The included subjects were randomly divided into a training set and a validation set at a ratio of 7∶3.The KOA and sarcopenia were diagnosed based on the KOA diagnostic criteria proposed by the American College of Rheumatology and the sarcopenia diagnostic criteria established by the Asian Working Group for Sarcopenia,respectively.Information including gender,age,body mass index(BMI),smoking history,alcohol consumption history,comorbid diabetes,comorbid coronary heart disease,co-morbid hypertension,comorbid osteoporosis,physical activity level,dietary status,Kellgren-Lawrence grade,knee pain visual analogue scale(VAS)score,grip strength,gait speed,appendicular skeletal muscle index(ASMI),and maximum circumference of leg was collected.Pa-tients in the training set were divided into a comorbidity group and a non-comorbidity group based on the presence or absence of KOA-sar-copenia comorbidity.The general data and relevant indicators of the two groups were compared.The Gtsummary package in R language was used to perform univariate and multivariate logistic regression analyses on the relevant variables of the two groups.Based on the results of univariate and multivariate analyses and clinical experience,after excluding variables that might cause information leakage and multicol-linearity,the variables included in the model construction were determined.The models were constructed and compared based on the Baye-sian information criterion(BIC),and the five models with the smallest BIC values were selected.Performance evaluation indicators such as McFadden's R2,discrimination,and calibration were calculated for each model.The optimal model was selected based on the model's sim-plicity,interpretability,discrimination,calibration,and other performances,and a nomogram risk prediction model for KOA-sarcopenia co-morbidity was constructed based on the optimal model.Based on the data of the training set and validation set subjects,the receiver operat-ing characteristic(ROC)curve was used to evaluate the discrimination of the KOA-sarcopenia comorbidity risk prediction model,and the calibration curve and Hosmer-Lemeshow test were used to evaluate the calibration of the KOA-sarcopenia comorbidity risk prediction model.Results:A total of 640 subjects were included,with 448 in the training set and 192 in the validation set.In the training set,there were 56 cases in the comorbidity group and 392 cases in the non-comorbidity group.The differences between the two groups in gender,age,BMI,comorbid diabetes,physical activity level,knee pain VAS score,grip strength,gait speed,ASMI,and maximum circumference of leg were statistically significant(x2=5.853,P=0.016;Z=-4.210,P=0.000;Z=-5.630,P=0.000;x2=21.863,P=0.000;x2=7.382,P=0.007;Z=-2.700,P=0.007;Z=-3.892,P=0.000;Z=-2.977,P=0.003;Z=-4.624,P=0.000;Z=-4.955,P=0.000).Univariate logistic regression analysis showed that gender(female),age,BMI,comorbid diabetes,low physical activity level,knee pain VAS score,grip strength,gait speed,ASMI,and maximum circumference of leg were significantly associated with KOA-sarcopenia comorbidity(OR=2.23,95％CI(1.22,4.28),P=0.012;OR=1.13,95％CI(1.06,1.20),P=0.000;OR=0.72,95％CI(0.64,0.80),P=0.000;OR=3.90,95％CI(2.18,6.98),P=0.000;OR=2.63,95％CI(1.45,5.00),P=0.002;OR=1.91,95％CI(1.59,2.32),P=0.000;OR=0.92,95％CI(0.88,0.96),P=0.000;OR=0.03,95％CI(0.00,0.17),P=0.000;OR=0.22,95％CI(0.13,0.33),P=0.000;OR=0.49,95％CI(0.40,0.57),P=0.000).Multivariate logistic regression analysis showed that gender(female),fast gait speed,high ASMI,and large maximum circumference of leg were independent protective factors for KOA-sarcopenia comorbidity(OR=0.08,95％CI(0.01,0.68),P=0.025;OR=0.02,95％CI(0.00,0.38),P=0.011;OR=0.10,95％CI(0.02,0.39),P=0.002;OR=0.50,95％CI(0.37,0.65),P=0.000),while comorbid diabetes,low physical activity level,and high knee pain VAS score were inde-pendent risk factors(OR=6.28,95％CI(2.10,20.40),P=0.001;OR=3.13,95％CI(1.06,10.00),P=0.045;OR=2.10,95％CI(1.57,2.91),P=0.000).Seven variables including age,BMI,gender,maximum circumference of leg,comorbid diabetes,physical activity level,and knee pain VAS score were included to construct the KOA-sarcopenia comorbidity risk prediction model.The KOA-sarcopenia co-morbidity risk prediction model constructed based on 5 variables of BMI,comorbid diabetes,physical activity level,knee pain VAS score,and maximum circumference of leg was determined as the optimal model(BIC value was 156.73,McFadden's R2 value was 0.644,sensitiv-ity was 0.875,specificity was 0.911,accuracy was 0.906,F1 score was 0.700,Matthews correlation coefficient was 0.666,Brier score was 0.071).Based on the training set data,the area under the ROC curve of the KOA-sarcopenia comorbidity risk prediction model was 0.971(P=0.000,95％CI(0.953,0.990)),the calibration intercept was 0.000,the calibration slope was 1.000,the E/O value was 1.000,and the MAE value was 0.077(x2=1.584,P=0.991).Based on the validation set data,the area under the ROC curve of the KOA-sarcopenia comorbidity risk prediction model was 0.938(P=0.000,95％CI(0.900,0.975)),the calibration intercept was-0.588,the calibration slope was 0.625,the E/O value was 1.211,and the MAE value was 0.119(x2=17.866,P=0.022).Conclusion:The KOA-sarcopenia comorbidity risk prediction model constructed based on 5 variables of BMI,comorbid diabetes,physical activity level,knee pain VAS score,and maximum circumference of leg demonstrates good discrimination and accuracy,and can be used for clinical risk prediction of KOA-sar-copenia comorbidity.