Development and external validation of a predictive model of nondiagnostic results in patients undergoing CT-guided percutaneous transthoracic needle lung biopsy.

Abstract

Objectives: To develop and validate a nomogram to predict nondiagnostic results in patients undergoing CT-guided percutaneous transthoracic needle biopsy (PTNB) of the lung.

Methods: A total of 954 PTNBs in the training cohort and 186 PTNBs in the external validation cohort were retrospectively included. PTNB results were categorized as diagnostic or nondiagnostic. Logistic regression was used to identify independent predictors of nondiagnostic results. C-statistic, calibration curve, and decision curve analysis were used to evaluate discrimination, calibration, and clinical usefulness, respectively.

Results: Of 954 PTNBs in the training cohort, 280 (29.4%) were nondiagnostic results. The model included six independent predictors: age at biopsy, lesion size, lobulation sign, air bronchogram, the number of samples, and pre-test probability. The C-statistics for the training and external validation cohorts were 0.752 and 0.734, respectively. Two risk groups were identified with low (<40%) and high (≥40%) probabilities of nondiagnostic results. For lesions of low risk with the number of samples ≤ 2, ≤2 samples should be obtained; for lesions of low risk with the number of samples ≥3, more samples should be obtained when appropriate; for lesions of high risk with the number of samples ≥3, PTNB needs to be reconsidered.

Conclusions: The nomogram showed good performance in predicting the nondiagnostic results of PTNB of the lung. Suggestions for each risk group may facilitate clinical practice.

Advances in knowledge: Pretest probability was a significant factor to predict nondiagnostic results of PTNB. The number of samples of PTNB should be different for different risk groups to avoid nondiagnostic results.