关键词: 结核病, 病原学阴性, 列线图模型

Abstract:

Objective To establish a tuberculosis（TB）predictive model based on the multi-center laboratory data of patients with pathogen-negative initially treated TB and conduct validation，and to provide a new auxiliary tool for the diagnosis of pathogen-negative TB. Methods Totally，101 patients with pathogen-negative initially treated TB（disease group）from Qingpu Branch of Zhongshan Hospital of Fudan University from January 2020 to December 2024 were enrolled. They were matched with 303 patients with pneumonia or pulmonary infection（control group）at a ratio of 1︰3. The data of about 2/3 of the cases in disease group and control group were randomly selected as modeling set，and the remaining data were used as validation set. The data from 56 TB cases from Songjiang Hospital Affiliated to Shanghai Jiao Tong University School of Medicine and Tongren Hospital Affiliated to Shanghai Jiao Tong University School of Medicine were provided，including 31 cases in disease group and 25 cases in control group，all of which were used as validation data. General information and laboratory determination data of all the study subjects at admission were collected. The differences in various indicators between disease group and control group were compared. LASSO regression was used to screen potential factors for predicting TB，and a nomogram model was established based on the results of multivariate Logistic regression analysis. Delong test and T-cell enzyme-linked immunospot assay（T-SPOT）was used to compare the efficacy of the model for diagnosing TB. The efficacy of the nomogram model in predicting TB risk was evaluated using receiver operating characteristic（ROC）curve and precision-recall（P-R）curve. Results LASSO regression identified 6 potential predictive factors，namely age，erythrocyte sedimentation rate（ESR），white blood cell（WBC）count，platelet-to-C-reactive protein ratio（PCR），tuberculosis antibody（TB-Ab）and T-SPOT. The areas under curves（AUC）of ROC curve and P-R curve of the nomogram model for predicting TB risk were 0.954 and 0.908，respectively. The AUC of ROC curve of T-SPOT for predicting TB risk was 0.895. The AUC of ROC curve and P-R curve of the validation data from the other 2 hospitals were 0.940，0.954 and 0.926，0.961，respectively，and the AUC of T-SPOT were 0.767 and 0.803，respectively. The validation results from all the 3 hospitals showed that the efficacy of the nomogram model in predicting TB risk was superior to T-SPOT（P<0.05）. Conclusions The established nomogram model can be used as an auxiliary diagnostic tool for pathogen-negative initial TB.

Key words: Tuberculosis, Pathogen-negative, Nomogram model