Development and validation of a prediction model for glucocorticoid-associated osteonecrosis of the femoral head by targeted sequencing.

Objective To develop and validate a prediction model based on targeted sequencing for glucocorticoid (GC)-associated osteonecrosis of the femoral head (GA-ONFH) in GC-treated adults. Methods This two-centre retrospective study was conducted between July 2015 and April 2019 at Zhongshan Hospital (training set) and the Sixth People's Hospital (test set) in Shanghai, China. All patients had a history of GC therapy, with a dose exceeding 2000 mg equivalent prednisone within 6 weeks. Patients were divided into two groups according to whether they were diagnosed with GA-ONFH within 2 years after GC initiation. Blood or saliva samples were collected for targeted sequencing of 358 single nucleotide polymorphisms and genetic risk score (GRS) calculating for developing GA-ONFH prediction model. Receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA) were performed to evaluate and validate the model. Results. The training set comprised 117 patients, while the test set comprised 30 patients for external validation. Logistic regression analysis showed that GRS was significantly associated with GA-ONFH (OR 1.87, 95% CI: 1.48, 2.37). The ROC and DCA curves showed that the multivariate model considering GRS, age at GC initial, sex and underlying diseases had a discrimination with area under the ROC curve (AUC) of 0.98 (95% CI: 0.96, 1.00). This model was further externally validated using the test set with an AUC of 0.91 (95% CI: 0.81, 1.00). Conclusion Our prediction model comprising GRS, age, sex and underlying diseases yields valid predictions of GA-ONFH incidence. It may facilitate effective screening and prevention strategies of GA-ONFH. [ABSTRACT FROM AUTHOR]