Derivation And Validation Of A Prediction Model For Fragility Fracture Risk In Patients With Osteoporosis

Objective:The purpose of this study was to explore the influencing factors of fragility fractures in patients with osteoporosis in Jiangxi Province and to build a predictive model in order to select stronger interventions and management when appropriate.Methods:In this retrospective study,we selected 1083 patients with osteoporosis who were admitted to Jiangxi Provincial People’s hospital from January 2019 to December 2022.The patient records were reviewed through the osteoporotic patient data management platform and HIS system were used to review patients’ medical records,screen patients who met the inclusion criteria and collect the following information:age,sex,weight,height;laboratory parameters including:alkaline phosphatase(ALP),creatinine,blood calcium,blood phosphorus,carbon dioxide binding,type I collagen carboxy-terminal peptide、propeptide of total type I procollagen、vitamin D3、parathormone,bone mineral density of the lumbar spine,total hip,and femoral neck;and the patient’s past year fall history,parental hip fracture history,past medical history,exercise,sun exposure,dairy product intake,smoking,alcohol consumption,and glucocorticoids use,and supplemented by face-to-face or telephone inquiries to complete the missing information.Finally,912 patients were included,including 595 in the fracture group and 317 in the non-fracture group.SPSS.26 software was used for single factor analysis to screen statistically significant variables(P<0.05).Multifactor analysis was used to explore the factors influencing fragility fractures,and a binary logistic regression model was established.Meanwhile,the model efficacy was tested by Hosmer-Lemeshaugh(H-L)goodness and Receiver operating characteristic curve(ROC).R software was used to validate the model internally using the Bootstrap method,plot nomogram and calibration curve.Enter the relevant parameters in the FRAX scoring system and calculate the 10-year major osteoporotic fracture probability(MOFP)and 10-year hip fracture probability(HFP)and then compare the prediction model with FRAX.Results:1.The results of univariate analysis showed that age,history of falls in the past year,parental history of hip fracture,exercise,hypertension,hyperlipemia,chronic obstructive pulmonary disease(COPD),stroke,ALP,lumbar spine BMD,lumbar spine T value,femoral neck BMD,femoral neck T value,hip BMD,and hip T value were statistically significant.2.Binary logistic regression analysis of variables significant in the univariate analysis revealed that age(OR=1.050,95%Cl:1.023-1.078),history of falls in the past year(OR=5.738,95%Cl:3.119-10.558),parental history of hip fracture(OR=2.521,95%Cl:1.021-6.224),ALP(OR=1.010,95%Cl:1.003-1.017)was an independent risk factor for the development of fragility fracture,and lumbar spine BMD value(OR=0.061,95%Cl:0.009-0.396)was a protective factor.3.Based on the results of multifactor analysis,a prediction model was established using five factors influencing fragility fracture,including age,falls,parental history of hip fracture,ALP,and lumbar spine BMD,and the final formula for calculating the probability of fragility fracture risk in patients with osteoporosis,P=ex/(1+ex)×100%,e is an exponential function with a value of 2.718 and X=0.057xage+1.780xpresence of past 1 year falls(0 or 1)+1.009xhistory of parental hip fracture(0 or 1)+0.009xALP-2.686xlumbar spine BMD-2.778.4.The results of internal validation of the model by Bootstrap method showed that the predicted calibration curve fitted well with the standard curve,with a C-index of 0.753 and a H-L goodness-of-fit test result of P>0.05,suggesting that the model has good calibration and discrimination in predicting fracture risk.5.The MOFP and HFP were obtained by calculating the FRAX scores of the subjects participating in the study,and the two scores were combined using SPSS.26 to obtain a simple FRAX score model,and their ROC curves were plotted separately and found that:the Area Under Curve(AUC)of FRAX was 0.719(95%CI:0.684-0.755,P<0.001)and the Jorden Index 0.322,and the AUC of MOFP and HFP were 0.719(95%CI:0.684-0.754,P<0.001)and 0.708(95%CI:0.6720.744,P<0.001),respectively.Compared with FRAX,the AUC of the new model is 0.753,which is significantly higher than FRAX(P=0.0376),MOFP(P=0.039)and HFP(P=0.006)Conclusion:In this study,age,history of falls in the past year,parental history of hip fracture,and ALP were found to be risk factors for fragility fracture in patients with osteoporosis,and lumbar spine BMD was a protective factor.A predictive model for fragility fracture risk that based on these five influencing factors was developed and validated to help identify subgroups of patients with osteoporosis at high fracture risk for more intense and targeted interventions.