Cranial stent position is independently associated with the development of TIPS dysfunction

Complications of portal hypertension can be treated with transjugular intrahepatic portosystemic shunt (TIPS) in selected patients. TIPS dysfunction is a relevant clinical problem. This study investigated the prognostic value of two-dimensional (2D) TIPS geometry for the development of TIPS dysfunction. Three hundred and seven patients undergoing TIPS procedure between 2014 and 2019 were analyzed in this monocentric retrospective study. 2D angiograms from the patients with and without TIPS dysfunction were reviewed to determine geometric characteristics including insertion and curve angles and the location of the stent. Primary outcome was the development of TIPS dysfunction. A total of 70 patients developed TIPS dysfunction and were compared to the dysfunction-free (n = 237) patients. The position of the cranial stent end in the hepatic vein and the persistence of spontaneous portosystemic shunts were significantly associated with the development of TIPS dysfunction. Among significant parameters in univariable regression analysis (portal vein-pressure after TIPS, Child–Pugh Score before TIPS, MELD before TIPS and white blood cell count before TIPS), multivariable models showed cranial stent position (p = 0.027, HR 2.300, 95% CI 1.101–4.806) and SPSS embolization (p = 0.006, HR 0.319, 95% CI 0.140–0.725) as the only predictors of TIPS dysfunction. This monocentric study demonstrates that the position of the cranial stent end is independently associated with the development of TIPS dysfunction. The distance of the cranial stent end to the IVC at the time of TIPS placement should be less than 1 cm in 2D angiography.

www.nature.com/scientificreports/ association of the landing zone of TIPS stent in the portal or hepatic vein with the development of TIPS dysfunction. The placement of the TIPS stent in the hepatic vein to IVC junction seems important to reduce the risk of hepatic venous stenosis or occlusion 17,[22][23][24][25][26][27][28] . Moreover, interventional embolization of spontaneous portosystemic shunts (SPSS) during TIPS procedure was associated with less episodes of hepatic encephalopathy 29 . Larger cohort data on predicting the development of TIPS dysfunction as early as at the time of TIPS placement is still scarce, given the potentially devastating effects of recurrent variceal bleeding and ascites. Therefore, the aim of this study was to determine whether the 2D TIPS geometrical characteristics at the time of TIPS creation can predict TIPS dysfunction in a large cohort.

Methods
Study population. This is a retrospective analysis of our observational monocenter NEPTUN and NEP-TUN 2 cohorts (Non-invasive Evaluation Program for TIPS and Follow Up Network) (clinicaltrials.gov identifier: NCT03628807 and NCT04393519) of patients undergoing TIPS procedure between January 1, 2014, to December 31, 2019 in our institution. The patients were regularly followed up clinically every 3 to 6 months using non-invasive imaging such as ultrasound and computer tomography (CT) as well as standard laboratory biochemical blood analyses to evaluate TIPS function (Fig. 1).
Inclusion criteria were patient age older than 18 years, first-time treatment with a TIPS, implantation of a PTFE-covered TIPS stent (Viatorr) and available digital subtraction angiography (DSA) studies from the time of TIPS placement. Exclusion criteria were previous TIPS revisions including balloon dilatation, stent-in-stent placement, and local lysis of the TIPS tract as well as reduction of TIPS stent. Patients with non-cirrhotic splanchnic venous thrombosis due to prothrombotic vascular liver disease have a different clinical trajectory. Thus, they were excluded from the study. All patients received anticoagulation with heparin or low molecular weight heparin (Partial Thromboplastin Time 2-3 times above normal) for 7 days without subsequent anticoagulation according to our institutions standard protocol.
Primary outcome was first revision due to TIPS dysfunction, defined as abnormal duplex sonographic measurement (reduction of flow velocity of more than 50% or missing flow), ascites, bleeding, or progression of esophageal varices, with resulting invasive revision of TIPS.
All patients signed written informed consent. This study was performed according to the guidelines of the Helsinki Declaration. Local ethics committee (Ethikkomitee der Medizinischen Fakultät, Universität Bonn) approved this study (Lfd. Nr. 038/20). TIPS procedure. The TIPS procedure was performed by a team of experienced radiologists and hepatologists under fluoroscopic and ultrasound guidance as previously described 14 . All contraindications were precluded beforehand. The procedure was performed under analgesia with pethidine. Initial portosystemic pressure gradient (PSPG) was recorded, then the 8-10 mm nominal diameter covered TIPS stent (Gore Viatorr endoprothesis, W.L. Gore Medical) was implanted. The TIPS stent was dilated according to PSPG at interventionalist's discretion. Post-TIPS PSPG targets were PSPG < 12 mmHg for variceal bleeding or 50% PSPG reduction for refractory ascites. Length of the stent was calculated by 2D angiogram with measuring pig tail catheter. SPSS, if present, were embolized with coils or histoacryl according to the interventionalist's discretion. TIPS patients were followed by routine follow ups including ultrasound examinations in our outpatient clinic every 3-6 months.  Cranial TIPS stent end. The distance between cranial stent end and inferior vena cava (IVC) in cm was measured with the initially placed radio-opaquely marked pigtail catheter as reference. Distance of more than 1 cm from the IVC was defined as the cranial TIPS stent ending in the hepatic vein. ROC Analysis was performed for the distance of cranial stent end to IVC (AUC = 0.592), and Youden Index showed an optimal cut off between 0.9 cm and 1.1 cm. Thus, we chose 1 cm as cut off.
α and γ angle. Supplement angles between two straight lines: the first runs orthogonally to and centrally through the beginning of the covered stent part (defined by the radiopaque gold ring of the TIPS stent). The second straight line models the course of the portal vein and runs centrally through a section of the vessel chosen to be as long as possible, in the middle of which the TIPS stent debouches. β angle. Angle between a straight line projecting the cranial course of the TIPS stent and oriented to radiopaque markings of the stent and a straight line extending to behind the cranial stent entry site and representing the course of the hepatic vein by a centrally located line. δ angle. Angle between the projected course of the lower TIPS tract, represented by a straight line passing centrally through the stent section that lies immediately cranial to the inferior confluent vessel, and a straight line modeling the course of the portal vein as described in α and γ angle.
IVC reflux. Reflux of contrast media after TIPS stent implantation was analyzed in angiogram loops.
Retrograde intrahepatic portal venous flow. Angiogram loops were analyzed for retrograde flow in the intrahepatic portal venous branches after TIPS implantation.
Caudal TIPS stent end. Determined on the final images after TIPS implantation and classified as insertion in portal vein, exact intersection of portal vein and liver parenchyma or liver parenchyma.
All measurements were performed manually by both an expert radiologist and trained hepatologist.

Statistical analysis.
Descriptive statistics were run for all variables. Continuous variables are shown as median (range), categorial variables as percentage or absolute cases. To compare the dysfunction and the dysfunction-free groups, non-parametric testing was used. Uni-and multivariable regression models were used to identify predictors of TIPS dysfunction, p values below 0.05 were considered statistically significant. Interobserver agreement was determined by Intraclass Correlation Coefficient (ICC) and Cohen's Kappa. ICC was used for variables with quantitative measurement scales and Cohen's kappa for variables with categorical measure- Figure 2. Schematic presentation of the measured two-dimensional TIPS stent geometry parameters and angles: 1: IVC; 2: liver vein; 3: cranial TIPS end; 4: PV, 5: distal TIPS end 6: distance between cranial TIPS stent end and IVC measured in cm (stent ends that extended into the IVC were noted as ≤ 0 cm); α-and γ-angle: angles measured between a straight line drawn at right angles to the beginning of the covered stent and a second straight line passing through the middle of the PV; β-angle: angle between the hepatic vein and the cranial TIPS stent end continued course; δ-angle: angle measured between the course of the PV and the course of the lower TIPS tract. www.nature.com/scientificreports/ ment scales; values above 0.8 were considered good and values above 0.9 were considered excellent interobserver agreement. Analysis of all was performed with Statistical Package for Social Sciences (SPSS version 24, IBM,).

Results
General patient characteristics at baseline. This study included three hundred and seven patients (n = 184 (60%) male) who underwent TIPS procedure. In 90% of our patients, TIPS was created between the right hepatic and right portal veins. Median age at TIPS procedure was fifty-nine (18-87) years. The most frequent indication for TIPS was refractory ascites in one hundred and ninety-one cases (63%); one hundred and twelve (37%) TIPS were implanted for variceal bleeding. The two main causes of cirrhosis were alcohol-related (n = 184, 60%) and chronic viral hepatitis (n = 37, 12%). Seventy (23%) patients developed TIPS dysfunction (Fig. 1). Median time to TIPS dysfunction was six (0-84) months. The most frequent indications for revision were signs of dysfunction on duplex-sonography and/or clinical reoccurrence of ascites (n = 54, 82%). There were no significant differences in procedure-related complication rates (Supplementary Table 1).
For local factors such as hepatocellular carcinoma (HCC) and for hepatic hemodynamics before and after TIPS procedure, there was no significant difference between the dysfunction and dysfunction-free group ( Table 1). Model of end-stage liver disease (MELD) score before TIPS (12 (6-27) dysfunction-free group vs 11 (6-21) dysfunction group; p = 0.036) and Child-Pugh score before TIPS (9 (5-14) dysfunction-free group vs 9 (5-13) dysfunction group; p = 0.041) showed a significant difference between the two groups.

Predictors of TIPS dysfunction.
In univariable regression analysis, the only TIPS geometry parameters associated with the development of TIPS dysfunction were the position of the cranial stent end in the hepatic vein and the distance of the cranial TIPS stent end from the IVC (Table 3). In multivariable regression models, only the position of the cranial stent end in the hepatic vein showed to be an independent predictor of development of TIPS dysfunction (Table 4). Moreover, the embolization of competing SPSS was also significant in univariable and multivariable regression analysis. None of the measured angles (α, β, γ, δ) or the nominal stent diameter showed a significant association with the development of TIPS dysfunction as well as local processes such as HCC (Table 3). In a subgroup analysis, cases with a stent end < 1 cm from the IVC were divided into those stents ending in the hepatic vein and those ending in the IVC (n = 10 (42%) group with dysfunction vs. n = 48 (45%) group without dysfunction, p = 0.749), showing no significant difference between the two subgroups (univariable regression analysis: p = 0.748, HR 0.863, 95% CI 0.352-2.117).

Discussion
This monocentric study demonstrates that the position of the cranial stent end in the hepatic vein, measured at the time of TIPS procedure, can predict TIPS dysfunction.
In recent years, several studies evaluated predictors of TIPS dysfunction. Besides obviously identifying the use of bare metal stents as predictors 10,12,16 , other factors such as liver function (MELD) 15 , indication for TIPS www.nature.com/scientificreports/ procedure 8,12 , interventionalist's experience 18 and portal venous flow post TIPS 19 have been described. Few studies evaluated TIPS geometry, such as stent position 17,20,21,30 , or stent-to-PV angle as predictive parameters of shunt dysfunction. Of all the proposed parameters, our study highlights the importance of the cranial stent position. However, we acknowledge only fair discrimination in our ROC analysis, which may indicate that development of TIPS dysfunction is a multifactorial process and may include the persistence of SPSS as well as other factors. All other suggested parameters of TIPS geometry and angles do not add predictive value, which is in line with smaller series that could not confirm those TIPS geometry parameters as predictors for TIPS dysfunction either 28,31 . Our results did not show increased portal blood flow by embolization of SPSS, which would have been a possible explanation for the influence of embolized SPSS on the lower rate of TIPS dysfunction. Since SPSS were embolized at the discretion of the interventionalist in this study, this decision may have been influenced by preferential flow through the SPSS during completion portography. At this point, it should be mentioned that, in order to make a more general recommendation on embolization of SPSS, further studies are needed.
Other not geometry-related factors such as white blood cell count, PV pressure after TIPS procedure, MELD or Child-Pugh Score were not significant in multivariable regression analysis.
Several reasons for the seemingly contrary results can be attributed to a non-standardized TIPS procedure performed across the world 32 .
First, the clinical practice of anticoagulation during and after TIPS procedure is still debated and no general consensus exist. However, this could play an important factor for TIPS dysfunction by in-stent thrombosis.
Second, the imaging techniques between the studies is inhomogeneous 20 , which might contribute to this result. Importantly, CT or magnetic resonance imaging (MRI) scans, for three-dimensional reconstruction, are usually not indicated immediately after TIPS procedure. Therefore, in clinical routine, angiography from the TIPS procedure often is the only available imaging. For this reason, our study focuses on the two-dimensional angiography data. Nevertheless, 3D geometry might reveal significant angular variations that are not visible in two-dimensional angiography, thus further studies are still needed.
Even though the geometric angles of the stent usually cannot be influenced, the cranial stent end position can be easily influenced by the interventionalist's choice of stent length. Given that the cranial stent position was the only predictive geometric parameter of TIPS dysfunction in our study, the simple advice for interventionalists is to choose a stent length long enough to cover the entire hepatic vein to the IVC. Practically, our study suggests that the distance between the cranial stent end and the IVC in 2D portogram should be less than 1 cm. In making this recommendation however, it must be kept in mind, that a cranial stent end projecting deep into the IVC Table 2. TIPS geometry parameters of angiography. TIPS a : Transjugular intrahepatic portosystemic shunt; IVC b : inferior vena cava; Distance cranial TIPS stent end to IVC (cm) c : negative values indicate that the stent end extends into the IVC; Cranial TIPS stent end in hepatic vein d : defined as > 1 cm distance from IVC in DSA; PV e : portal vein; α Angle f : left TIPS-tract angle beginning at covered stent part to PV in DSA (degrees); β Angle g : Angle of cranial TIPS stent end to hepatic vein/ IVC in DSA (degrees); γ Angle h : right TIPS-tract angle beginning at covered stent part to PV in DSA (degrees); δ i : Distal TIPS-tract angle to PV in DSA (degrees); SPSS j : spontaneous portosystemic shunt. *p < 0.05, **p < 0.01, ***p < 0.001.

Parameter
No dysfunction (n = 237) TIPS dysfunction (n = 70) p www.nature.com/scientificreports/ may complicate liver transplantation; yet, to date, we have not experienced the impossibility of transplantation due to stent placement. The suggested passive expansion of nitinol stents may cause changes in the stents geometry over time 33 . With the introduction of controlled expansion stents, there may be further differences in TIPS geometry depending on the type of covered stents 34 . Importantly, the cranial position of the stent should be unaffected by potential changes of stent geometry over time.
Even though this is a comprehensive analysis of the largest cohort on this topic so far, there are some limitations. The main limitation is the retrospective and monocentric character of the study limiting its generalizability. The importance of the stent position in the hepatic vein to IVC junction has been proposed in smaller series 22,23 . However, this study is the largest evaluating 2D angiography data from the time of TIPS placement.

Conclusion
In conclusion, this study demonstrates that the position of the cranial TIPS stent end measured in two-dimensional angiography imaging at the time of TIPS implantation is an independent predictor of the development of TIPS dysfunction. Our study suggests that the distance between the cranial stent end and the IVC in 2D angiogram should be less than 1 cm. Table 3. Univariable regression analysis with TIPS dysfunction as endpoint. OR a : odds ratio; 95%-CI b : 95% confidence interval; TIPS c : Transjugular intrahepatic portosystemic shunt; Cranial TIPS stent end in hepatic vein d : defined as > 1 cm distance from IVC in DSA; SPSS e : spontaneous portosystemic shunt; IVC f : inferior vena cava; Distance cranial TIPS stent end to IVC (cm) g : negative values indicate that the stent end extends into the IVC; PV-Pressure h : portal venous pressure (mmHg); PSPG i : portosystemic pressure gradient (mmHg); MELD j : model of end-stage liver disease; PV k : portal vein; α Angle l : left TIPS-tract angle beginning at covered stent part to PV in DSA (degrees); β Angle m : Angle of cranial TIPS stent end to hepatic vein/ IVC in DSA (degrees); γ Angle n : right TIPS-tract angle beginning at covered stent part to PV in DSA (degrees); δ Angle o : Distal TIPS-tract angle to PV in DSA (degrees); INR p : international normalized ratio; ALT q : alanine aminotransferase; WBC r : white blood cell count; HCC s : hepatocellular carcinoma. *p < 0.05, **p < 0.01, ***p < 0.001.