Evaluating clinical outcomes and adverse events: risk stratification of endoscopic ultrasound–guided through-the-needle biopsy for pancreatic cyst in a large cohort

BACKGROUND AND AIMS: Despite concerns regarding its safety and the incidence of adverse events (AEs), EUS-guided through-the-needle biopsy (EUS-TTNB) has emerged as a promising diagnostic tool. This study evaluated the clinical effectiveness of EUS-TTNB, assessed procedural AEs, and categorized risks based on predictive factors using recursive partitioning analysis (RPA). METHODS: This retrospective study included 301 patients who underwent EUS-TTNB at Asan Medical Center, Seoul, South Korea, using permutation-based accuracy and RPA. RESULTS: EUS-TTNB exhibited high diagnostic adequacy (80.3%, 241/301), showed strong histologic concordance among surgically resected cases (87.5%, 42/48), and led to changes in diagnosis and management in 16.3% of cases (49/301). AEs occurred in 20.9% of cases, with acute pancreatitis being the most common (15%, 45/301). Multivariable analysis identified intraductal papillary mucinous neoplasm (IPMN) (odds ratio [OR], 2.24; 95% CI, 1.26-3.98; P = .006) and complete cyst aspiration (OR, 0.50; 95% CI, 0.28-0.90; P = .02) as significant risk factors for AEs. In RPA, a random forest model and permutation-based accuracy were used to create 3 risk categories: low (AE rate: 11.9%, non-IPMN with complete cyst aspiration or ≥4 through-the-needle biopsies); high (AE rate: 37.0%, IPMN with age ≥70 or without complete cyst aspiration); and moderate (AE rate: 20.6%, not classified as low or high risk). CONCLUSIONS: EUS-TTNB demonstrated notable clinical utility in revising diagnoses and influencing therapeutic decisions but poses a considerable risk of AEs. By applying this risk categorization model with careful consideration and patient selection, its clinical utility may be enhanced.