Long-term Follow-up after TIPS: Use of Doppler Velocity Criteria for Detecting Elevation of the Portosystemic Gradient

Abstract

To evaluate the performance of Doppler ultrasound as a screening test for detecting elevated portosystemic gradients in failing transjugular intrahepatic portosystemic shunts (TIPS). Twenty-seven of 61 patients who underwent TIPS creation between November 1991 and March 1996 were studied. At routine intervals, angle-corrected velocity measurements of portal venous and intrashunt blood flow (at the portal venous, middle, and hepatic venous levels of the shunt) were obtained. These were compared with portal hemodynamics for diagnostic accuracy in predicting clinically significant elevation of the portosystemic gradient. Venographic and manometric correlations were obtained on all patients available for follow-up and were not limited to those with symptoms or "abnormal" Doppler studies. Receiver-operating characteristic (ROC) curves were done. Linear regression was done to study correlation of shunt velocities with portal pressure, and logistic regression was done to predict shunt stenosis with use of shunt velocities. The most accurate location for shunt velocity measurement was the main portal vein, but this had an area under the ROC curve of only 0.70. Accuracy of any velocity threshold (including maximum shunt velocity) was no greater than 70%. Maximum shunt velocity of less than 60 cm/sec was 93% specific for detecting shunt restenosis, but only 25% sensitive, for an overall accuracy of 64%. High sensitivity (90%) could only be achieved with poor specificity (< 33%). Linear regression revealed poor correlation between shunt or portal vein velocity measurements and portal pressure (/r/ < 0.23 for all). Intrashunt and portal venous Doppler velocities alone do not accurately predict elevation of the portosystemic gradient on long-term follow-up after TIPS.