Clinical predictive model for identifying high-risk factors in Pleurisy Tuberculoma patients

Abstract

Introduction : Pleurisy tuberculoma(PTM) manifests as a tumor-like lesion in either the parietal or visceral pleura. The majority of PTM cases arise during the management of tuberculous pleural effusion, underscoring a strong association between the two conditions, despite the precise pathogenesis remaining elusive. To explore the high-risk factors for the development of PTM, we constructed a clinical predictive model aimed at providing more powerful clues for PTM treatment. Methods A retrospective study was conducted on patients with PTM or TPE who were treated at Nanjing Chest Hospital and Nanjing Second Hospital between March 2013 and April 2024. The demographic information and clinical characteristics were sorted out, and the model was constructed by logistic regression, lasso regression, and best subset selection. ROC curves, calibration curves, DCA curves, and clinical impact curves were analyzed for all constructed models to determine the final clinical prediction model. In the internal validation, the performance of the selected model was further evaluated. Results The final predictive model incorporated two risk factors associated with PTM development: Ki-67+CD4+T and Ki-67+CD8+T. Despite its limited calibration, the model has a strong discriminating ability (c-statistic = 0.987; 95% confidence interval: 0.969-1.000) and holds significant clinical value (as indicated by DCA analysis, maintaining a high net benefit across a wide threshold range). During internal validation, the model exhibited excellent predictive performance (AUC: 0.98; Accuracy: 0.96; Kappa: 0.93; Sensitivity: 0.95; Specificity: 0.96). Conclusion A predictive model based on two indicators may help clinicians make early predictions of PTM based on laboratory data. The results of this study may motivate healthcare providers to further investigate the immune role of T lymphocytes in PTM individuals, enhancing comprehension of the development of PTM.