Identification of the Most Important Subset of Doppler, Laboratory, and Clinical Parameters for Serial TIPS...

OBJECTIVE. The purpose of the present study was to identify the subset of a wide range of serial Doppler, laboratory, and clinical parameters most predictive (both individually and in combination) of TIPS dysfunction in a large patient sample. MATERIALS AND METHODS. The medical records of 189 patients who had undergone TIPS procedures were analyzed. The patients (mean age, 52 years; 62% of whom were men) had undergone 1139 Doppler studies and 323 portovenograms. Laboratory parameters included model for end-stage liver disease (MELD) scores, serum albumin levels, presence of ascites, and time since last intervention. Doppler parameters included intrashunt velocities, temporal change in intrashunt velocities, main portal vein velocity, direction of flow in the left portal hepatic vein, and venous pulsatility index. Statistical analysis used ROC, univariate, and multivariate regression models to assess the parameters both individually and in combination. Shunt dysfunction was defined by a portosystemic gradient of more than 12 mm Hg. RESULTS. The laboratory and clinical parameters of greatest predictive value included the MELD score and the time since the last intervention. The Doppler parameters that were of greatest predictive value included the change in velocity at the hepatic venous end and the left portal vein flow direction. Multivariate models produced an AUC of 0.74. Differences between functional and dysfunctional shunts were also statistically significant for absolute velocity at the hepatic venous end, the change in velocity within the stent, and the temporal change in the mid shunt velocity. CONCLUSION. The subset of serial parameters most predictive of TIPS dysfunction are the temporal change in the velocity at the hepatic venous end, the absolute velocity at the hepatic venous end, the direction of flow in the left portal venous branch, and changes in the MELD score.