Evaluation of Incidence, Clinical and 3D Geometric Predictors of TIPS Revision and Thrombosis

Purpose: To evaluate occurrence and timing of TIPS reasons for revision and potential predictors such as clinical characteristics, intra-operative measures and CT determined 3D geometric TIPS measures.

Materials and Methods: Consecutive adults (n=98) who underwent TIPS between January 2011 and January 2024 and with a CT scan within 30 days after TIPS were included. Median age at TIPS was 57.8 (range 21.7-80.3), 61% were male, and 57% were Hispanic, 18% white, 13% black, 11% other. Median MELD 3.0 score prior to TIPS was 14 (range 7-34). TIPS stents and IVC and PV were isolated in a semi-automated fashion with TotalSegmentator {1}, a validated artificial intelligence anatomic segmentation tool, and determined Hounsfield unit (HU) thresholds for the TIPS stent within the segmented liver. From isolated stent, IVC and PV, multiple 3D geometric measures were calculated including stent curvature, distance of stent from center of IVC and nearest PV, and angles of insertion of stent into IVC and PV. Univariable and multivariable tests for revision and survival were explored for predictors with and without adjusting for clinical and operative data (pre- and post- portosystemic gradients).

Results: The incidence of TIPS revision was 32.7% (n=32): thrombosis (11), recurrent ascites (7), stenosis or partial occlusion (6), hepatic encephalopathy (4) and recurrent varices (4). Crude post-TIPS mortality was 35.7% (n=35) with Kaplan-Meier survival at 1y 76% (95%CI 66-83%); 3y, 70% (60-79%); and 5y, 57% (44-69%). Median (and mean) time to revision for thrombosis was 0.4 months (1.3) and for ascites was 1 month (10.4) (T-test p=0.09). One-year TIPS non-thrombosis rate was 87% (95%CI 78-93%) with all 11 such events occurring within the first year. One-year, 3y and 5y revision-free survivals for all revisions (n=32) were 73% (62-81), 65% (53-75), and 57% (44-69%) respectively. TIPS stent length predicted the 11 thrombotic events (p=0.006, Fisher Exact) and in logistic regression (p=0.01 when adjusted for age, race and MELD score). It did not predict for the other 21 revision events. Other 3D variables were not predictive for outcomes. Survival at 3y was 61% after revision for thrombosis (n=11) and 54% after revision for other events (n=21).

Conclusion: A third of TIPS patients require revision, especially for thrombosis and recurrent ascites. Only TIPS stent length was predictive for this. Distances, angles and curvature were not predictive. Post-placement portosystemic gradient was not predictive, differing from the literature {2}. Survival after revision was acceptable.