Abstract
TIPS is commonly used to treat complications of portal hypertension. Hernia incarceration is a complication of the TIPS procedure that is poorly understood and not well documented. We examine this complication in a large, single-center cohort of TIPS patients and compared it to a cohort of patients with refractory ascites managed with serial large-volume paracentesis alone to determine whether TIPS itself increases incarceration risk. We identified 259 adults who underwent TIPS and 644 adults with refractory ascites managed with serial large-volume paracentesis at the University of California, San Francisco (UCSF) between 2015 and 2024. We extracted structured variables and unstructured documentation from the UCSF Clinical Data Warehouse. We used OpenAI's GPT-4o, a large language model, and manual chart review to identify post-TIPS hernia incarceration and clinical information from unstructured notes. We calculated time to events and used LASSO-Cox regression to identify risk factors associated with incarceration. Of 259 patients, 12.7% (33) developed post-TIPS incarceration with a median time to event of 31 days (IQR: 10-109). In comparison, only 0.9% (6/644) of large-volume paracentesis patients developed hernia incarceration. Within the TIPS cohort, patients with hernia incarcerations were more likely to have existing umbilical hernias (87.9% vs. 29.2%, p <0.01) and higher serum albumin (3.2 vs. 2.8 g/dL, p <0.01) compared to those without incarceration. Multivariate Least Absolute Shrinkage and Selection Operator-Cox regression showed the presence of an umbilical hernia pre-TIPS (HR=3.1, 95% CI: 1.8-5.4, p <0.01), serum creatinine at TIPS (HR=1.3, 95% CI: 1.0_1.6, p =0.04), serum albumin at TIPS (HR=1.7, 95% CI: 1.0-2.8, p =0.04) were predictors of post-TIPS incarcerated hernia development. In the incarcerated hernia group, 36.4% underwent hernia repair at a median of 0.5 days (IQR: 0-259), 18.2% received liver transplant at 55 days (IQR: 6-152), and 24.2% died at 422 days (IQR: 261-747), all measured from the time of incarceration diagnosis. Of the patients who did not develop incarcerated hernias, 6.2% underwent hernia repair (for non-incarcerated hernias), 23.0% received a transplant, and 16.4% died after TIPS. Kaplan-Meier analysis showed no significant difference in mortality between the incarceration and no-incarceration group (2-y probability of 13.7% vs. 12.8%, log rank p =0.32). In sensitivity analyses including abdominal binder use, post-TIPS abdominal binder (HR=3.1, 95% CI: 1.7-5.6, p <0.01) and pre-TIPS umbilical hernia (HR=2.8, 95% CI: 1.6-4.9, p <0.01) remained significant predictors, while serum creatinine dropped out as a significant variable, suggesting binder use may be a marker of ascites or hernia severity. TIPS substantially increases hernia incarceration risk compared to serial paracentesis alone, with 12.7% of patients in this single-center cohort developing incarceration at a median of 1 month after TIPS. The presence of a known umbilical hernia prior to TIPS had the strongest association with this complication. Serum creatinine and albumin association also suggests ascites severity as a contributor. Identifying pre-TIPS risk factors could facilitate the development of targeted strategies to mitigate post-TIPS complications.