Negative Biopsy of Focal Hepatic Lesions: Decision Tree Model for Patient Management.

OBJECTIVE: The purpose of this study was to investigate patient- and procedure-related variables affecting the false-negative rate of ultrasound (US)-guided liver biopsy and to develop a standardized patient-tailored predictive model for the management of negative biopsy results. MATERIALS AND METHODS: We retrospectively included 389 patients (mean age ± SD, 62 ± 12 years old) who had undergone US-guided liver biopsy of 405 liver lesions between January 1, 2013, and June 30, 2015. We collected multiple patient- and procedure-related variables. By comparing pathology reports of biopsy and the reference standard (further histology or imaging follow-up), we were able to categorize the biopsy results as true-positive, true-negative, and false-negative. Diagnostic accuracy and diagnostic yield were measured. Univariate and multivariate analyses were performed to identify variables predicting false-negative results. A standardized patient-tailored predictive model of false-negative results based on a decision tree was fitted. RESULTS: Diagnostic accuracy and diagnostic yield were 93.8% (380/405) and 89.4% (362/405), respectively. The false-negative rate was 6.5% (25/387). Predictive variables of false-negative results at univariate analysis included body mass index, lesion size, sample acquisition techniques, and immediate specimen adequacy. The only independent predictors at multivariate analysis were patient age and Charlson comorbidity index. By combining lesion size and location with patient age and history of malignancy, we developed a decision tree model that predicts false-negative results with high confidence (up to 100%). CONCLUSION: False-negative results are not negligible at US-guided liver biopsy. The combination of selected lesion- and patient-specific variables may help predict when aggressive management is warranted in patients with likely false-negative results.

Duke Scholars

Author Daniele Marin Radiology, Abdominal Imaging