Nonspecific benign pathological results on computed tomography-guided lung biopsy: A predictive model of true negatives

Abstract

Objective: The aim of this study is to develop a predictive model for identifying true negatives among nonspecific benign results on computed tomography-guided lung biopsy. Materials and Methods: This was a single-center retrospective study. Between December 2013 and May 2016, a total of 126 patients with nonspecific benign biopsy results were used as the training group to create a predictive model of true-negative findings. Between June 2016 and June 2017, additional 56 patients were used as the validation group to test the constructed model. Results: In the training group, a total of 126 lesions from 126 patients were biopsied. Biopsies from 106 patients were true negatives and 20 were false-negatives. Univariate and multivariate logistic regression analyses were identified a biopsy result of “chronic inflammation with fibroplasia” as a predictor of true-negative results (P = 0.013). Abnormal neuron-specific enolase (NSE) level (P = 0.012) and pneumothorax during the lung biopsy (P = 0.021) were identified as predictors of false-negative results. A predictive model was developed as follows: Risk score = −0.437 + 2.637 × NSE level + 1.687 × pneumothorax - 1.82 × biopsy result of “chronic inflammation with fibroplasia.” The area under the receiver operator characteristic (ROC) curve was 0.78 (P < 0.001). To maximize sensitivity and specificity, we selected a cutoff risk score of −0.029. When the model was used on the validation group, the area under the ROC curve was 0.766 (P = 0.005). Conclusions: Our predictive model showed good predictive ability for identifying true negatives among nonspecific benign lung biopsy results.