Non-malignant pathological results from CT-guided biopsy for pulmonary nodules: a predictive model for identifying false-negative results.

ABSTRACT

BACKGROUND: Computed tomography (CT)-guided biopsy (CTB) procedures are commonly used to aid in the diagnosis of pulmonary nodules (PNs). When CTB findings indicate a non-malignant lesion, it is critical to correctly determine false-negative results. Therefore, the current study was designed to construct a predictive model for predicting false-negative cases among patients receiving CTB for PNs who receive non-malignant results. MATERIALS AND METHODS: From January 2016 to December 2020, consecutive patients from two centers who received CTB-based non-malignant pathology results while undergoing evaluation for PNs were examined retrospectively. A training cohort was used to discover characteristics that predicted false negative results, allowing the development of a predictive model. The remaining patients were used to establish a testing cohort that served to validate predictive model accuracy. RESULTS: The training cohort included 102 patients with PNs who showed non-malignant pathology results based on CTB. Each patient underwent CTB for a single nodule. Among these patients, 85 and 17 patients, respectively, showed true negative and false negative PNs. Through univariate and multivariate analyses, higher standardized maximum uptake values (SUVmax, P = 0.001) and CTB-based findings of suspected malignant cells (P = 0.043) were identified as being predictive of false negative results. Following that, these two predictors were combined to produce a predictive model. The model achieved an area under the receiver operating characteristic curve (AUC) of 0.945. Furthermore, it demonstrated sensitivity and specificity values of 88.2% and 87.1% respectively. The testing cohort included 62 patients, each of whom had a single PN. When the developed model was used to evaluate this testing cohort, this yielded an AUC value of 0.851. CONCLUSIONS: In patients with PNs, the predictive model developed herein demonstrated good diagnostic effectiveness for identifying false-negative CTB-based non-malignant pathology data.