| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ultrasonography |
1 From the Departments of Radiology (R.G.S., T.V., D.P.B.) and Neurology (B.J.R.), Beth Israel Deaconess Medical Center and Harvard Medical School, 330 Brookline Ave, Boston, MA 02215. Received August 9, 2001; revision requested September 19; revision received November 5; accepted January 7, 2002. Address correspondence to R.G.S. (e-mail: rsheiman@caregroup.harvard.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Baseline and prerevision US images obtained in 15 patients with venographically confirmed TIPS malfunction were retrospectively examined for spectral waveform pulsation. Cardiac pulsatility was quantified by using the venous pulsatility index (VPI), the venous equivalent of resistive index. VPIs were obtained at four locations from the main portal vein to the stent–hepatic venous junction. Baseline and follow-up examination results in 11 patients with functional TIPS acted as controls and were evaluated similarly. Baseline and follow-up mean VPIs at all four locations were compared for both sets of patients by using the Newman-Keuls pairwise multiple sample comparison test. The 
2 test was used to determine if a VPI threshold that would result in an acceptable sensitivity and specificity for shunt dysfunction existed.
RESULTS: One hundred twenty mean VPIs were obtained in the study group, and 88 mean VPIs were obtained in the control group. Prerevision VPIs at each location were significantly lower (P < .01) than all baseline values and than the follow-up values in the control group. A VPI less than 0.16 was 94% sensitive and 87% specific for shunt dysfunction.
CONCLUSION: The VPI, a quantitative measure of cardiac pulsation obtained with Doppler US, may be a useful parameter for assessing TIPS function.
© RSNA, 2002
Index terms: Hepatic veins, stenosis or obstruction, 957.721, 959.721 • Hepatic veins, US, 957.12983, 957.12984, 957.12989, 959.12983, 959.12984, 959.12989 • Shunts, portosystemic, 957.1268, 959.1268 • Ultrasound (US), Doppler studies, 957.12983, 957.12984, 957.12989, 959.12983, 959.12984, 959.12989
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Other investigators have recommended temporal changes in peak stent velocities rather than absolute velocity thresholds as sensitive indicators of impending shunt failure. Dodd et al (5) found that a temporal decline in midstent velocities of 50 cm/sec or greater was 93% sensitive and 77% specific for shunt stenosis. These authors also found no correlation between temporal changes in peak main portal venous velocity and shunt dysfunction, the exact opposite of findings obtained by Kanterman et al (2) and Zemel et al (6). Additional criteria advocated for assessment of TIPS with Doppler US have included the congestive index (3), which is the ratio of portal venous cross-sectional area and mean flow velocity, and, because of the reciprocal relationship between hepatic arterial and portal volumetric blood flow, hepatic arterial velocities (4).
Overall, there are no clear Doppler US velocity criteria that have consistently been useful in predicting TIPS malfunction. Some authors (7) dispute the use of Doppler US criteria for TIPS screening entirely. As indicated by Haskal et al (8), some of the controversy originates from the varying definitions of shunt dysfunction: Some authors define stenosis solely by means of portal venous and shunt venography, whereas others define it by means of portosystemic pressure gradients. The complexity of the hemodynamics of a TIPS adds to the problem, since portal venous outflow may occur through three routes: the TIPS, native liver, and varices. There have also been studies indicating pulsatile portal venous flow in healthy subjects (9) and significant effects of a patient’s respiratory state on TIPS velocities (10), which make measurement of peak systolic velocities more confusing.
There is a clear need to develop simple quantitative US criteria that can be used to accurately distinguish between a functional and a failing shunt but that are not based solely on the volumetric flow of blood through the TIPS and, hence, absolute velocity measurements. Because the creation of a TIPS places the portal vein and shunt in direct continuity with the right side of the heart, we questioned whether cardiac pulsations are visible on spectral waveforms during Doppler US evaluation of TIPS. If visible pulsations are transmitted through patent shunts and pulsation magnitude declines with shunt stenosis, pulsation measurement may provide an additional criterion to assess during US screening of TIPS, and this criterion would not be dependent on absolute velocity measurements. Thus, the purpose of this study was to determine if cardiac pulsations are visible and quantifiable with spectral waveforms during Doppler US evaluation of TIPS, and if so, whether their magnitude declines with shunt dysfunction.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
All 15 patients underwent our standard US protocol for shunt evaluation, which includes spectral analysis for peak and minimal velocities in the main portal vein and proximal (portal venous–stent junction), middle, and distal (shunt–hepatic venous junction) portions of the stent; interval change (development, increase, or decrease) in ascites; and documentation of the direction of flow within the left portal vein. All US examinations (HDI 3000, Advanced Technology Laboratories, Bothell, Wash or Elegra, Siemens Medical Systems, Issaquah, Wash) were performed by using a 2.5- or 3.5-MHz transducer with the patient in a supine or decubitus position. Examination results were reviewed by one of four experienced staff radiologists (including R.G.S.) prior to patient discharge from the US suite.
All examinations were performed with patients having taken nothing by mouth for a minimum of 4 hours. Spectral analysis was performed in patients during quiet respiration for a minimum of 5 seconds. The angle of interrogation was less than or equal to 60°, with the pulse repetition frequency, wall filter, and Doppler gain optimized for each patient.
The resistive index of the portal vein and proximal, middle, and distal portions of the shunt was chosen as an indicator of cardiac variation. The resistive index should quantify the difference between the minimum (Vmin) and maximum (Vmax) velocities at these locations and also normalize these values for each patient because it is calculated as follows: (Vmax - Vmin)/Vmax (Fig 1). Additionally, it is a known index that can be directly and easily determined with currently available US units. The resistive index of the main portal vein has been termed the VPI (9). Similarly for our study, we prefer to use VPI rather than resistive index to refer to this measurement.
|
TIPS revision was performed by means of the right internal jugular approach. Venography of the splenic vein, main portal vein, and TIPS was performed in two oblique projections, and absolute pressures were obtained from the main portal vein through the hepatic vein, inferior vena cava, and right atrium. Revision was performed when a portosystemic gradient greater than 15 mm Hg was identified and consisted of either dilation or dilation and repeat stent placement because of narrowing that was thought to be the cause of the elevated gradient.
Data Analysis and Control Patients
For all 15 patients, baseline VPIs for each location were compared with each other and with those obtained immediately before TIPS revision. Comparison was performed by using the Newman-Keuls pairwise multiple sample comparison test, with a P value less than or equal to .05 considered to indicate statistical significance (11).
Portosystemic pressure gradients obtained immediately after TIPS creation and at revision were also documented for each patient, with a gradient of 15 mm Hg used as the threshold to discern between functional and failing TIPS, which is similar to the threshold used by others (2,8). We chose this threshold rather than narrowing of the TIPS at venography as an absolute indicator of shunt function because narrowing alone may not correlate with dysfunction. However, venograms were also reviewed (D.P.B.) to determine the location of the stenosis that was thought to be the cause of shunt dysfunction. Absolute shunt velocity measurements obtained from baseline and prerevision examination results were not formally evaluated because our only goal was to determine if spectral waveform cardiac variation was present anywhere along the course of a functional TIPS and, if so, if its loss was associated with dysfunction.
During the period of this study, 11 patients (seven men, four women; mean age, 57 years; age range, 47–79 years) underwent TIPS creation, had continually functional shunts on the basis of clinical grounds, and had US surveillance results that were interpreted as normal. These additional 11 patients served as control subjects. VPIs at each location, determined from baseline and routine follow-up US examination results, were calculated and compared with each other and with baseline and prerevision values in the group with failing shunts. This comparison was also performed by using the Newman-Keuls pairwise multiple sample comparison test. The portosystemic gradient at TIPS creation was also documented in all 11 patients to ensure adequate initial function and also compared with the initial gradients in the study group by using the Student t test. The inclusion of control patients allowed us to avoid selection bias in assessing only functional TIPS that would later fail—that is, to determine if waveform variation existed in all patients after TIPS creation and, if so, whether it persisted in patients with clinically functional TIPS.
Intuitively, one would expect the VPI to decline with increasing distance from the heart in patients with widely patent TIPS. If so, the potential to misinterpret low main portal venous pulsation as a sign of TIPS malfunction exists. We therefore compared (by using the Student t test) VPIs from the main portal vein with those from the stent–hepatic venous junction in all patients with either clinically or venographically confirmed functioning shunts to determine the effect of location, if any, on VPIs. Finally, because an associated portosystemic pressure gradient was obtained within 24 hours in 41 of the 52 Doppler US examinations of this study, linear regression was used to determine if any association existed between pressure gradients and VPIs.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
All 15 patients with shunt dysfunction demonstrated shunt stenosis at venography and a portosystemic pressure gradient greater than 15 mm Hg (mean, 20.1 mm Hg ± 3.2). Shunt stenoses were located at the stent–hepatic venous junction in 10 patients, in the middle of the stent in two patients, and at both the stent–hepatic venous junction and middle of the stent in three patients. All 15 patients underwent successful TIPS revision by means of either dilation or dilation and additional stent placement at the site of stenosis. All postrevision gradients were less than 15 mm Hg (mean, 9.5 mm Hg ± 1.6).
Overall, 52 Doppler US examinations were performed—22 in the control group of 11 patients and 30 in the group of 15 patients with shunt dysfunction. The mean VPIs obtained in each of the four locations in the initial and follow-up Doppler US studies are listed separately in the Table. The 16 mean VPIs yielded 120 separate paired combinations that were examined for significant differences.
|
|
|
|
|
2 analysis of all 208 VPIs was performed to determine if a threshold VPI could be found below which a satisfactory sensitivity and specificity for dysfunction could be established. One hundred forty of 148 VPIs in the functional shunts were greater than a threshold VPI of 0.16, as compared with only seven of 60 VPIs from failing shunts, which yielded a sensitivity and specificity of 94% and 87%, respectively, for dysfunction. Four of the seven VPIs greater than this threshold in failing shunts were obtained from the middle of the stent and stent–hepatic venous junctions in the two patients with isolated stenosis in the middle of the stent. Increasing the VPI threshold to 0.19 improved specificity to 93% but with a slight decline in sensitivity to 85%. Initial and follow-up Doppler US studies in the 11 control patients yielded 22 VPIs each at the main portal vein and the stent–hepatic venous junction, and initial examinations in the 15 study patients yielded 15 VPIs at each of these locations. Hence, 37 separate VPIs from functional shunts were obtained at the main portal vein and at the stent–hepatic venous junction and were compared. The VPIs in these functional shunts were location dependent (Student t test, P < .002); those at the main portal vein (mean, 0.28 ± 0.12) were significantly lower than those at the stent–hepatic venous junction (mean, 0.35 ± 0.10). Five (14%) of the 37 VPIs from the main portal vein were less than 0.16 but accounted for 63% (five of eight) of all the VPIs less than this threshold in functional shunts.
In 41 of the 52 Doppler US examinations performed, associated portosystemic pressure gradients were obtained within 24 hours. Linear regression analysis revealed no relationship between VPIs and the portosystemic pressure gradient.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Our study data also indicate that the VPI in functional shunts is location dependent and declines significantly as one moves from the stent–hepatic venous junction to the main portal vein. We specifically evaluated for a location-dependent decrease in VPI because it makes sense intuitively and we believe its presence helps to validate our hypothesis that cardiac pulsations that are detectable at Doppler US are transmitted across patent shunts and reach the main portal vein. Because of this, however, the potential to misinterpret a low VPI at the main portal vein in a functional shunt as a sign of shunt dysfunction exists. In our series, this occurred in five of 37 Doppler US examinations performed in patients with functional shunts. We recommend that an isolated finding of low (VPI < 0.16) pulsatility in the main portal vein not be considered an indicator of shunt dysfunction. A low VPI at a location other than the main portal vein must also be documented before shunt failure is considered.
Our hypothesis concerning the transmission of cardiac pulsations through a TIPS is supported by the results of a study conducted by Chong et al (1). These investigators state in their article that transmitted cardiac pulsations were usually present in patent shunts in their TIPS population. This is reiterated in the caption of the first figure of their article, in which they state that the wavy appearance of the spectral waveform of a functional shunt is due to transmitted cardiac pulsations. Shunt pulsatility is a direct result of creating a patent conduit between the right side of the heart and the portal vein. It logically follows that shunt stenosis would inhibit this transmission distal to the stenosis location.
The study of Kliewer et al (10) has relevance to our work, because it too was based on the concept of creating communication between the right side of the heart and the portal vein by means of a TIPS. These authors hypothesized that because of this communication afforded by a TIPS, a patient’s respiratory state may alter shunt flow. A significant decline in peak shunt velocity occurred with deep versus quiet inspiration in both functional and failing shunts. A pressure change within a TIPS, regardless of whether it is cardiac or respiratory in origin, should influence the portosystemic gradient and, hence, the velocity within the shunt.
We believe the work of Kliewer et al (10) complements and supports our work. It is also because of this work that we chose to perform TIPS spectral analysis during the patient’s quiet respiration, which posed no difficulty for the patient population in our study; in doing so, we standardized any influence of respiration on our data. Our findings, therefore, apply to Doppler US evaluation of TIPS only during quiet respiration of the patient. Deep inspiration would probably blunt the transmission of cardiac pulsations and globally lower VPIs, although this result is only speculative.
Most investigators studying velocity criteria for TIPS evaluation do not report the exact sites of narrowing within the TIPS that are thought to be the cause of dysfunction and portosystemic gradient elevation. The sites of stenosis affect velocity parameters, as alluded to by Dodd et al (5). These researchers found that peak stent velocity decreased in eight of nine patients with hepatic venous stenosis but increased in two of three patients with isolated stent stenosis and either decreased or increased in cases of concomitant stenosis within the stent and hepatic vein. Clearly, the location of shunt stenosis affects the Doppler US velocity measurements. Similarly, one would expect the VPI obtained at a specific site within a failing TIPS to depend on its location relative to the site of stenosis. We documented the site of stenosis on the basis of portographic results in the 15 patients in our study and, as found with velocity measurements, believe our results correlate with the stenosis site. Ten (67%) cases of shunt dysfunction were isolated to the stent–hepatic venous junction, similar to the 56% (nine of 16) reported by Dodd et al (5) and the 60% (15 of 25) reported by Murphy et al (12). In eight of these 10 patients, pulsatility was low (<0.16) at all four locations interrogated and low at three of four locations in the remaining two patients. Viewed differently, of the seven of 60 VPIs above 0.16 in the patients with failing shunts in our study, two were found at the midstent level and two were found at the stent–hepatic venous junction, all in the two patients with an isolated stenosis within the stent.
Although the number of cases was small, it appears that pulsatility is maintained proximally, or more centrally, to an isolated stenosis. Hence, the most likely location for detecting shunt dysfunction with VPI measurements appears to be at the stent–hepatic venous junction, the most common site for stenosis to develop.
The lack of consensus concerning velocity criteria for identifying potential shunt dysfunction stems from multiple issues. The definition of shunt dysfunction used in previous articles has not been standardized: Some authors require a portosystemic gradient greater than 12 mm Hg (10), and others require 15 mm Hg (5). In the work of Chong et al (1), portosystemic gradients were obtained but not reported or correlated with dysfunction. Still others (4) have used a shunt stenosis of 50% or greater at venography as an indicator of dysfunction, regardless of the gradient.
Murphy et al (12), by using a 15-mm Hg portosystemic gradient as a cutoff, identified 25 failing shunts that they correlated with US. These authors concluded that no combination of TIPS velocity parameters could be used to predict the portosystemic pressure gradient and, hence, shunt dysfunction; this finding was confirmed by the results of an earlier in vitro study performed by Daniel et al (13).
To correlate velocity parameters with a portosystemic gradient would require that TIPS hemodynamics, in their simplest form, follow the law of Hagan-Poiselle (14)—that is, maximum shunt velocity is directly related to the pressure gradient and the square root of the conduit radius and inversely related to shunt length. There has been a general failure by investigators to correlate shunt velocities with shunt size and length, a correlation that would be required to establish a relationship between velocity and pressure. We were unable to find any published studies on Doppler US evaluation of TIPS that take into account the size or length of stents used for TIPS creation.
Zizka et al (3), by using a portosystemic gradient greater than 12 mm Hg as an indicator of shunt failure, found that the mean volumetric flow in the main portal vein in 110 patients with a TIPS declined from a baseline of 2,134.0 to 1,713.8 mL/min at the time of shunt failure. They also found that immediately after shunt creation in 144 patients, the volumetric flow in the portal vein increased from 1,149.6 to 2,263.2 mL/min. Because the volumetric flow through a failing TIPS declines, exact correlation of a portosystemic gradient with velocity changes becomes even more difficult. A TIPS is not an isolated system, but rather it exists in parallel with the liver and possibly varices, each having its own resistance to flow; this relationship indicates that the overall hemodynamics closely resemble a parallel circuit. On the basis of this information, the lack of correlation of VPIs with the portosystemic pressure gradient in the failing shunts was not surprising.
It was not our goal to confirm or disprove previously reported velocity criteria or to develop additional velocity criteria to help identify clinically silent shunt dysfunction. Our single goal was to see if loss of TIPS pulsatility could be used as a simple sign of shunt dysfunction. This is why we did not formally examine the peak velocities in the group of patients with documented shunt failure in our study. The 15 patients with shunt failure were retrospectively identified on the basis of venographic results and portosystemic pressure measurements for the purpose of this study. Prospectively, no specific velocity criteria, but rather a combination of clinical, Doppler, and gray-scale US findings, were applied in deciding if venography should be performed.
There were some shortcomings to our study. First, although our study results showed that VPIs at each specific location did not significantly change over time in the control group of patients, portosystemic pressure measurements were not obtained and venography was not performed to definitively establish adequate shunt function. However, because there was no clinical suggestion of dysfunction, such as interval variceal bleeding and/or development of or worsening ascites, and follow-up Doppler US depicted little change from baseline results (velocity measurements and color Doppler US appearance), we are confident that these patients had functional TIPS and little change in their portosystemic pressures from the baseline values.
Second, the mean follow-up in the control patients, though similar to that in the group with shunt failure, was 101 days. How VPIs would change with longer follow-up in these patients is not known. We believe this follow-up is relatively adequate, given that the primary patency of TIPS is only 50%–55% at 1 year (15).
We also wish to point out that no formal attempt to evaluate the cardiac status of the patients was made in our study, and this may have directly affected the VPIs. However, normal right atrial pressure is a prerequisite for TIPS creation, and revision and was documented in all the patients in our study.
Furthermore, VPIs can be evaluated for either a temporal decline or a decline to less than an absolute value (threshold). We chose the latter because most of the 15 patients in our study underwent only baseline and prerevision Doppler US. Determination of exactly how the VPIs changed as shunt stenosis progressed would have required interval Doppler US studies, which were not performed. We surmise that a temporal decline in VPI occurs as shunt stenosis develops and progresses, but an additional study would be warranted to confirm this hypothesis. Finally, as with any retrospective study with a small number of patients, the true accuracy of the VPI to predict shunt dysfunction remains to be determined by means of a larger prospective study.
In conclusion, TIPS formation creates a conduit through which cardiac pulsations can be transmitted to the portal vein and visualized at spectral analysis. The VPI, a quantitative measure of cardiac pulsation obtained by means of Doppler US, may be a useful parameter for assessing shunt function. Pulsatility throughout a TIPS and within the portal vein appears to be present and persistent in functional shunts and significantly declines in magnitude with shunt failure. Our early experience indicates that the VPI may be a simple, sensitive, and specific noninvasive indicator of TIPS dysfunction.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, R.G.S., T.V.; study concepts, R.G.S., T.V.; study design, R.G.S.; literature research, R.G.S., T.V., D.P.B.; clinical studies, R.G.S., D.P.B.; data acquisition, R.G.S., B.J.R.; data analysis/interpretation, R.G.S., T.V., B.J.R.; statistical analysis, B.J.R.; manuscript preparation, R.G.S., T.V.; manuscript definition of intellectual content, R.G.S.; manuscript editing and revision/review, all authors; manuscript final version approval, R.G.S., B.J.R.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
