Abstract
OBJECTIVES: To determine whether international normalized ratio (INR), bilirubin, and creatinine predict bleeding risk following percutaneous liver biopsy.
METHODS: A total of 870 consecutive patients (age 53 ± 14 years; 53% (459/870) male) undergoing non-targeted, ultrasound-guided, percutaneous liver biopsy at a single tertiary center from 01/2016 to 12/2019 were retrospectively reviewed. Results were analyzed using descriptive statistics and logistic regression models to evaluate the relationship between individual and combined laboratory values, and post-biopsy bleeding risk. Receiver operating characteristic (ROC) curves and area under ROC (AUC) curves were constructed to evaluate predictive ability.
RESULTS: Post-biopsy bleeding occurred in 2.0% (17/870) of patients, with 0.8% (7/870) requiring intervention. The highest INR within 3 months preceding biopsy demonstrated the best predictive ability for post-biopsy bleeding and was superior to the most recent INR (AUC = 0.79 vs 0.61, p = 0.003). Total bilirubin is an independent predictor of bleeding (AUC = 0.73) and better than the most recent INR (0.61). Multivariate regression analysis of the highest INR and total bilirubin together yielded no improvement in predictive performance compared to INR alone (0.80 vs 0.79). The MELD score calculated using the highest INR (AUC = 0.79) and most recent INR (AUC = 0.74) were similar in their predictive performance. Creatinine is a poor predictor of bleeding (AUC = 0.61). Threshold analyses demonstrate an INR of > 1.8 to have the highest predictive accuracy for bleeding.
CONCLUSION: The highest INR in 3 months preceding ultrasound-guided percutaneous liver biopsy is associated with, and a better predictor for, post-procedural bleeding than the most recent INR and should be considered in patient risk stratification.
CLINICAL RELEVANCE STATEMENT: Despite correction of coagulopathic indices, the highest international normalized ratio within the 3 months preceding percutaneous liver biopsy is associated with, and a better predictor for, bleeding and should considered in clinical decision-making and determining biopsy approach.
KEY POINTS: • Bleeding occurred in 2% of patients following ultrasound-guided liver biopsy, and was non-trivial in 41% of those patients who needed additional intervention and had an associated 23% 30-day mortality rate. • The highest INR within 3 months preceding biopsy (AUC = 0.79) is a better predictor of bleeding than the most recent INR (AUC = 0.61). • The MELD score is associated with post-procedural bleeding, but with variable predictive performance largely driven by its individual laboratory components.