Development and validation of a predictive model for pulmonary hemorrhage in computed tomography–guided percutaneous lung biopsy

Abstract

Abstract Purpose This study aimed to identify risk factors for pulmonary hemorrhage (PH) and higher-grade PH that complicate computed tomography (CT)-guided percutaneous lung biopsy (CT-PNLB) and establish predictive models to quantify the risk. Methods A total of 2653 cases of CT-PNLB were enrolled. Multivariate logistic regression was used to identify independent risk factors to develop a nomogram prediction model. The model was assessed using the area under the curve (AUC) of the receiver operator characteristic (ROC) and calibration curves and validated in the validation group. Results PH occurred in 23.52% (624/2653) of cases, and higher-grade PH occurred in 7.09% (188/2653) of cases. The parameters of lesion size, puncture depth, and contact to pleura were identified as risk factors of PH and higher-grade PH in the logistic regression model, besides the position as a risk factor for PH. The AUC of the PH prediction model was 0.776 [95% confidence interval (CI): 0.752–0.800], whereas that of the validation group was 0.743 (95% CI: 0.706–0.780). The AUC of the higher-grade PH prediction model was 0.782 (95% CI: 0.742–0.832), whereas that of the validation group was 0.769 (95% CI: 0.716–0.822). The calibration curves of the model showed good agreement between the predicted and actual probability in the development and validation groups. Conclusion We identified risk factors associated with PH and higher-grade PH after PNLBs. Furthermore, we developed and validated two risk prediction models for PNLB-related PH and higher-grade PH risk prediction and clinical decision support. Key messages What is already known on this topic Pulmonary hemorrhage (PH) and other hemorrhagic complications are the most common complication in CT-guided percutaneous lung biopsy (CT-PNLB), except pneumothorax. However, the risk factors associated with PH remain controversial, and research on models of PH and higher-grade PH is also limited. What this study adds The parameters of lesion size, puncture depth, and contact to pleura were identified as risk factors of PH and higher-grade PH in the logistic regression model, besides the position as a risk factor for PH. In addition, we developed and validated two risk prediction models for PNLB-related PH and higher-grade PH risk prediction and clinical decision support. How this study might affect research, practice, or policy Of all the predictors, the position is the key factor to be considered by the operator. Moreover, two risk prediction models show good discrimination and calibration characteristics to identify patients at high risk of hemorrhage and higher-grade PH after PNLB, so these could assist clinicians in avoiding risk factors in advance.