Hepatic Perfusion as a Predictor of Mortalityafter Transjugular Intrahepatic PortosystemicShunt Creation in Patients with RefractoryAscitesEric Walser, MD, Orhan S. Ozkan, MD, Syed Raza, MD, Roger Soloway, MD, and Leka Gajula, MD
PURPOSE: To determine whether hepatic perfusion patterns predict mortality after transjugular intrahepatic porto-systemic shunt (TIPS) creation in patients with severe ascites.
MATERIALS AND METHODS: This retrospective study included 22 patients who had enhanced cine magneticresonance (MR) imaging performed immediately before TIPS creation in the angled coronal plane including the leftkidney, liver, and main portal vein. Regions of interest were centered over the liver and kidney, and perfusion curveswere generated and reviewed before the standard TIPS procedure was performed. Four patients did not undergo TIPScreation as a result of very poor hepatic perfusion by MR. All patients were followed clinically and by ultrasoundsurveillance of their shunt.
RESULTS: Eleven patients died within 6 months, including all four patients who did not have a TIPS because of MRevidence of poor hepatic perfusion. Of these 11 patients, eight (73%) had unfavorable liver flow consisting ofdiminished enhancement compared to the kidney and early peak enhancement of less than 50 seconds. The survivingpatients all showed a delayed peak enhancement of greater than 50 seconds.
CONCLUSIONS: In patients undergoing TIPS creation for refractory ascites, blunted arterial-type hepatic enhance-ment is a poor prognostic sign. Cine MR imaging with evaluation of hepatic perfusion can be performed and reviewedbefore the TIPS procedure. Alternative techniques for ascites reduction may be preferred for patients with unfavor-able hepatic perfusion.
J Vasc Interv Radiol 2003; 14:1251–1257
Abbreviations: MELD � model for end-stage liver disease, TIPS � transjugular intrahepatic portosystemic shunt, TTP � time to perfusion peak
PATIENTS with ascites are often re-ferred for transjugular intrahepaticportosystemic shunt (TIPS) creationafter optimum medical managementfails to control fluid accumulation or
jeopardizes renal function. As op-posed to patients with variceal hemor-rhage, patients with refractory ascitestolerate TIPS poorly, with publishedmortality rates of approximately 50%at 6 months (1,2). This diminished sur-vival likely results from the advancedstage of cirrhosis and the hepatic atro-phy seen in patients with severe as-cites. Malinchoc et al (3) subjected 231patients undergoing the TIPS proce-dure to a multivariate analysis andfound that patients with refractory as-cites had a 66% greater risk of dyingafter TIPS creation compared with pa-tients with variceal hemorrhage (P �.01). Recent reports have focused onclinical and laboratory parameters toidentify patients at high risk of hepatic
failure and death after TIPS creation(3–5). However, many patients withascites and acceptable liver functionand clinical status die of acceleratedliver failure after TIPS creation. Be-cause TIPS drastically alters hepaticperfusion by diverting the portalstream into the right atrium, it is pos-sible that TIPS-related mortality relieson preexisting function and perfusionof the cirrhotic liver. We believe thatseverely diminished portal perfusionnot compensated for by hepatic arte-rial hypertrophy predicts poor sur-vival after TIPS creation. We evalu-ated hepatic perfusion with magneticresonance (MR) imaging immediatelybefore TIPS creation to identify suchhigh-risk patients.
From the Departments of Radiology (E.W., O.S.O.,S.R.) and Internal Medicine (R.S., L.G.), Universityof Texas Medical Branch, 301 University Boulevard,Galveston, Texas 77555-0709. Received February 18,2003; revision requested May 1; final revision re-ceived June 11; accepted June 12. From the 2003 SIRAnnual Meeting. Address correspondence to E.W.;E-mail: [email protected]
None of the authors have identified a potential con-flict of interest.
This study is a retrospective reviewof 22 patients (five women, 17 men;mean age, 51 years) with refractoryascites or hydrothorax who were re-ferred to the interventional radiologyservice for TIPS consideration between2001 and 2002. The study design wasapproved by our institutional reviewboard before data collection andanalysis.
Patient Population
Patients were assessed clinically bythe gastroenterology and interven-tional radiology services. Encephalop-athy was graded with use of a stan-dard four-level classification (6) andpatients with active or recent enceph-alopathy were excluded regardless ofserum ammonia levels. Patients werethen assigned a modified Child-Pughscore and a model for end-stage liverdisease (MELD) score based on clinicaland laboratory parameters. The MELDscore was calculated with use of a for-mula available on the Mayo ClinicWeb site (www.mayoclinic.org/gi-rst/mayomodel5.html). Eight patients wereclassified as Child-Pugh class B and 14were class C. Cause of liver diseasewas chronic hepatitis C in all but two
patients, who had alcohol-related cir-rhosis. Three patients had hepatic hy-drothorax and the remainder of thepatients had severe ascites. Signifi-cant laboratory data (means � SD)were: International Normalized Ra-tio, 1.6 � 0.4; total bilirubin, 2.7mg/dL � 2.5; albumin, 2.3 g/dL �0.6; alanine aminotransferase, 58.5IU/L � 47.2; and serum creatinine,1.7 mg/dL � 1.7. Two patients hadhigh creatinine levels because ofchronic hemodialysis.
Dynamic MR Procedure
Before TIPS creation, a 5-F shortcannula was advanced into the rightinternal jugular vein with use of ultra-sound (US) guidance. The patient wasthen taken to the MR imaging suite(LX Echo speed 1.5 T magnet; GEMedical Systems, Milwaukee, WI) andpositioned in a torso phased-arraycoil, in which axial localizing imagesthrough the upper abdomen wereused to select an angled 2-cm coronalslab that included the liver, portalvein, and left kidney (Fig 1). The MRimaging sequence used was a coronalmultiphase sequential spoiled-gradi-ent acquisition (spoiled gradient-re-called acquisition in steady state; echo
time, 3.1 msec; flip angle, 60°C) ob-tained at a rate of one image per sec-ond over a 3-minute time interval (180images) after intravenous gadoliniuminjection at a rate of 4 mL/sec throughthe jugular venous catheter. Perfusionto the abdominal organs and vesselswas evaluated with use of the same-sized regions of interest and enhance-ment profiles generated by imagepostprocessing (Fig 2). Enhancementwithin the regions of interest was in-terpreted as relative rather than abso-lute MR intensity values (organ inten-sity at time 0 was assigned zeroenhancement). Perfusion curves of theliver and kidney were obtained in allpatients and evaluated for two charac-teristics. First, the time to perfusionpeak (TTP) was recorded. A long TTPrepresented portal-type perfusionwhereas a short TTP (�50 seconds)represented an arterial-type perfusion.The specific TTP value of 50 secondswas chosen after reviewing dynamicimages from these patients obtainedwith use of our MR and contrast ma-terial injection technique, which al-ways produced renal peaks at or be-fore 50 seconds. Second, the level ofhepatic enhancement was graded asless than, equal to, or greater than re-nal perfusion.
Figure 1. (a) Axial MR image through the mid-abdomen shows the angled coronal slab used to acquire dynamic hepatic perfusionimages. The acquisition volume includes the left kidney, portal vein, and hepatic parenchyma. (b) Angled coronal MR image throughthe upper abdomen. Equal-area regions of interest are centered over the right lobe of the liver and the left renal cortex.
1252 • Hepatic Perfusion as a Predictor of Mortality after TIPS October 2003 JVIR
TIPS Procedure
After MR imaging and review ofthe perfusion curves, the patient wassent to the angiography suite, wherethe 5-F cannula was exchanged for a10-F sheath and a TIPS was createdwith 8 –12-mm shunts created betweenthe right hepatic and right portal veinsto achieve a target portosystemic gra-dient of less than 14 mm Hg. Therewere no procedural complicationsfrom the procedure and all were tech-nically successful. Patients were fol-lowed clinically and by US examina-tion 1 day, 3 months, 6 months, andthen yearly after TIPS creation. Fourpatients did not undergo TIPS creationafter MR imaging showed extremelypoor liver perfusion (TTP � 50 sec-onds with very low peak enhance-ment). Their ascites was managed byalternative means (tunneled ports or
peritoneal dialysis catheters) untiltheir death, which was the primarystudy endpoint.
Statistical Analysis
Kaplan-Meier analysis generatedsurvival curves between the two per-fusion groups, with the log-rank testused to test for significance. Labora-tory values and Child-Pugh andMELD scores were compared betweenpatients surviving greater or less than6 months and between patients withfavorable or unfavorable perfusion.These data were analyzed for signifi-cance with the Wilcoxon signed-ranktest.
RESULTS
All patients who had a TIPS proce-dure attempted (n � 18) had the pro-
cedure successfully completed. Threepatients had shunt revisions at 1 day, 4months, and 10 months after TIPS cre-ation. All patients received 12-mmWallstents (Boston Scientific, Natick,MA). Two patients had maximum bal-loon dilation of the stents to 12 mm,with the remainder of shunts dilatedto 10 mm except for one that was di-lated to only 8 mm. Balloon dilation ofthe stents was sequentially increasedfrom 8 to 12 mm on the basis of athreshold portosystemic gradient of 14mm Hg. Final portosystemic gradientswere 4 –20 mm Hg, with an averagegradient of 11.2 mm Hg (SD, 3.5 mmHg).
No procedural complications oc-curred except in the single patientwhose shunt thrombosed within 24hours. A transjugular portogramshowed incomplete coverage of thehepatic parenchymal tract, and theshunt was revised by placement of asecond stent without further incident.Analysis of the dynamic MR perfusionscans showed that only three of the 22patients had greater enhancement ofthe liver compared with the kidney.Two of these three patients had end-stage renal disease and were undergo-ing chronic hemodialysis treatment.Four patients had equal liver and renalenhancement and the remaining 15patients had reduced hepatic enhance-ment compared with the kidney. Thir-teen patients had portal-type perfu-sion (TTP � 50 seconds) and ninepatients had arterial-type hepatic per-fusion (TTP � 50 seconds). Eight of thenine patients with arterial hepaticwaveforms had enhancement belowrenal levels (Fig 3).
When the group of patients withdecreased arterial-type perfusion (en-hancement less than in the kidney andTTP � 50 seconds, n � 8) were com-pared with the other patients with bet-ter enhancement and/or portal-typeperfusion (n � 14), there was a signif-icant difference in mortality rate at 6months. All eight of the patients in theunfavorable-perfusion group diedwithin 6 months (half the patients ac-tually had TIPS created) versus threeof the 14 patients (21%) with favorableliver perfusion (all had TIPS created;Fig 4). The average survival of patientswith unfavorable perfusion (arterialwaveform with peak enhancementless than in the kidney) was 34.3 daysin the four patients who had TIPS cre-
Figure 2. Hepatic (1) and renal (2) enhancement curves obtained from MR images. Notethe slower TTP (�50 seconds) and higher peak of the liver curve compared with the renalcurve. This patient did well after TIPS creation and experienced significant reduction inascites. The y axis is relative enhancement of a defined region of interest (from time 0) andthe x axis is time in milliseconds.
Walser et al • 1253Volume 14 Number 10
ated versus 32.8 days in the patientswho did not have TIPS created andhad conservative management of theirascites instead. The 6-month mortalitydifferences between patients with ar-terial versus portal hepatic perfusioncurves (TTP less than or greater than50 seconds, respectively), was 80%(eight of 10) in the arterial group and33% (four of 12) in the portal group.However, the two 6-month survivorsin the arterial group had equal orhigher peak enhancement comparedwith the renal curve and an average of77 units of relative hepatic enhance-ment compared with the eight nonsur-vivors who all had peak enhancementsignificantly less than in the kidney.The average relative enhancement ofthe liver in these eight patients was 19units.
US examination of the liver andDoppler analysis of the hepatic vesselswas performed in all patients beforeTIPS creation. Although no patientshad hepatofugal flow, six of the eight
patients with unfavorable hepatic per-fusion on MR had partial portal veinthrombosis and/or sluggish portalvein flow (back-and-forth flow or noDoppler waveform identified). Pa-tients with favorable perfusion all ex-hibited hepatopedal flow on Doppleranalysis. The differences in laboratoryvalues, MELD score, and Child classbetween survivors and nonsurvivorsand between favorable and unfavor-able perfusion groups were generallynot significant (Tables 1, 2). However,there were significant differences increatinine levels and MELD scores be-tween the groups with favorable andunfavorable perfusion on MR. Six pa-tients had creatinine levels greaterthan 1.5 mg/dL, and five died within 6months (all had unfavorable hepaticperfusion). The surviving patient withrenal dysfunction had favorable he-patic perfusion. The two patients whohad end-stage renal disease (undergo-ing hemodialysis) had favorable per-fusion. One patient died within 2
months (MELD score, 27) and theother lived almost 2 years with resolu-tion of ascites (MELD score, 13).
DISCUSSION
The initial indication for TIPS pro-cedures was refractory variceal hem-orrhage. As clinical experience accu-mulated, it was noted that manypatients with ascites who underwentthe TIPS procedure experienced reso-lution of their peritoneal fluid accu-mulation. In current practice, manypatients are TIPS candidates solely be-cause of refractory ascites. However,clinical outcomes analysis reveals thatpatients with cirrhosis who experiencesevere ascites are different from pa-tients with cirrhosis who experiencevariceal hemorrhage in that the formertend to have worse hepatic reserveand tend to tolerate the TIPS proce-dure poorly, with increased mortalitycompared with patients with varices(7). Additionally, many patients withsevere ascites do not respond to theTIPS procedure and have a high6-month mortality rate (8). Many in-vestigators have attempted to clarifythis issue and identify a subset of pa-tients with cirrhosis who respondpoorly or not at all to the TIPS proce-dure. These studies evaluated the syn-thetic ability of the remaining liver(with use of, for instance, bilirubin lev-els and dye-clearance models) or clin-ical parameters (eg, Apache scores,MELD score, Child-Pugh classifica-tion) (4). We believe that another piecein this equation may, in fact, be thestatus of perfusion to the liver in thepatient with cirrhosis. Therapeutic im-plements to decrease portal pressure,such as TIPS or surgical portocavalshunts, carry a significant risk. Al-though the efficacy is high with pa-tients experiencing variceal hemor-rhage, treatment of ascites withportosystemic shunt creation has vari-able success rates (3). Because patientswith severe ascites have higher mor-tality rates after TIPS procedures, it isimportant from a public health stand-point to identify patients who will re-spond to this intervention.
This study builds on earlier inves-tigations that used nuclear medicinefirst-pass techniques before and afterthe TIPS procedure. One initial study(9) composed of 27 patients found asignificant increase in early mortality
Figure 3. Unfavorable hepatic perfusion curve (1). The peak occurs before 50 secondsand there is significantly reduced enhancement compared with the kidney (2). Thispatient died of liver failure 3 months after the TIPS procedure.
1254 • Hepatic Perfusion as a Predictor of Mortality after TIPS October 2003 JVIR
in patients with specific types of flowto the liver, namely arterialized flowthat was diminished in comparison torenal blood flow. Patients with thistype of hepatic perfusion before TIPScreation had a mean survival of 2months compared with a mean sur-vival of 23 months in the patients witha more favorable hepatic perfusionprofile (9). Liver blood flow was quan-tified in a similar fashion by Van Beers
et al (10) with use of computed tomog-raphy (CT) and region-of-interestanalysis. They found that CT nicelyshowed the diminishing portal flowthat comes with advancing cirrhosis.Our study supports these findings butuses an abdominal first-pass techniquewith MR imaging, a modality with su-perior resolution compared with nu-clear medicine imaging. Region-of-in-terest analysis in areas such as the
portal vein and inferior vena cava isnot possible with nuclear medicine im-aging but is simple with MR imaging.Additionally, MR imaging allows im-proved anatomic evaluation of the cir-rhotic liver. Therefore, perfusion MRof the liver can be easily combinedwith other sequences to check the pa-tency and flow direction of the portalvein and to rule out hepatoma or otherabdominal mass lesions.
Because patients were sent for TIPSimmediately after MR imaging, theuse of gadolinium rather than iodin-ated contrast material was also an im-portant advantage for this modality.The cine MR imaging images were re-viewed on a GE workstation duringpatient preparation for the TIPS pro-cedure, and four cases were cancelledwhen perfusion curves indicated unfa-vorable hepatic perfusion. Initially, weperformed TIPS procedures in fourpatients with such unfavorable he-patic perfusion profiles and createdsmaller-caliber shunts (8 –10 mm) andaccepted higher postshunt portosys-temic gradients (average of 12.3 mmHg in these four patients) in an effortto control ascites and avoid hepaticfailure after TIPS creation. However,early mortality was universal in thissmall group of four patients, and wesubsequently treated patients withsimilar perfusion patterns with alter-native methods of fluid control (tun-neled peritoneal catheters or ports).Even so, overall survival in patientswith unfavorable hepatic perfusionwas less than 6 months with or with-out TIPS.
Although our conservative ap-proach to shunt creation in patientswith unfavorable hepatic perfusionwas different than that in patientswith better perfusion, the bias doesnot diminish the conclusion of thisstudy: that patients with unfavorablehepatic perfusion experience verypoor survival. It is possible that evensmaller shunts and the acceptance ofhigher final portosystemic gradientswould improve survival in these pa-tients, but we are skeptical given theirhigh mortality regardless of whetherTIPS was created. The hepatic fibrosisand tenuous blood flow in such pa-tients precludes any intervention andcan be considered terminal disease un-less expeditious hepatic transplanta-tion is available. We found, as haveothers (11), that patients who experi-
Figure 4. Kaplan-Meier survival curve of patients with favorable versus unfavorablehepatic perfusion on MR imaging. The standard error for patients with unfavorableperfusion is 0.12– 0.18 and that for the patients with favorable perfusion is 0.07– 0.12.
Table 1Laboratory/Clinical Comparison between Favorable and Unfavorable HepaticPerfusion
* Statistically significant.Note.—Values presented as means � SD. INR � International Normalized Ratio.
Walser et al • 1255Volume 14 Number 10
enced longer-term survival after TIPShad good ascites control, whereasnonresponders usually died withinmonths.
Several studies that examined mor-tality after TIPS creation centered onpreprocedural clinical and laboratoryevaluation of patients with cirrhosis.Ferral et al (4) evaluated a group of 14patients undergoing TIPS creation forrefractory ascites and suggested that aChild-Pugh score of 11 or greater in-dicated a poor prognosis. Other stud-ies concluded that bilirubin levelsgreater than 3.0 mg/dL and creatininelevels of 1.5 mg/dL or higher wereassociated with treatment failure andearly mortality (1,2). Malinchoc et al(3) derived a nomogram to quantifymortality after TIPS creation from theexperience at four medical centers in-volving 231 patients. Their model re-lied on abnormalities in creatinine lev-els, bilirubin levels, InternationalNormalized Ratio, and albumin levels(MELD score). Ferral et al (12) usedthis model to evaluate the survival of72 patients after TIPS creation andfound that patients with a MELD scoregreater than 18 had a significantlylower 3-month survival rate of 55%,versus 85% in patients with a MELDscore less than 18. However, the au-thors noted that the predicted mortal-ity rate was higher than the observedmortality rate, suggesting that somepatients could be deprived the clinicalbenefits of TIPS if their caregiverswere to base treatment decisions en-tirely on MELD classification. In fact,review of the study population of Fer-ral et al (12) shows a sensitivity of 50%and sensitivity of 93% in predictingdeath within 6 months in patients withMELD scores �18.
Although our series of patients is
small, we found a sensitivity of 73%and sensitivity of 100% in predictingdeath within 6 months in patients withunfavorable MR perfusion scan re-sults. These numbers remained fairlyconstant at 70% and 100%, respec-tively, when we combined our previ-ous findings from nuclear medicineperfusion scans (total of 49 patients).We did not find significant differencesin laboratory values, Child class, orMELD scores between survivors andnonsurvivors, but there were signifi-cant differences in serum creatininelevels and MELD scores between thepatients with favorable versus unfa-vorable hepatic perfusion. Therefore,the status of hepatic perfusion beforeTIPS creation probably correlates withthese laboratory parameters, which, aspreviously mentioned, are also usefulpredictors of post-TIPS mortality.Nevertheless, if we had offered or de-nied TIPS to patients on the basis ofthe aforementioned laboratory andclinical cutoff values (MELD � 18 orcreatinine level � 1.5 mg/dL), five ofour 11 patients who responded to TIPScreation would have been excludedand three of 11 patients who diedearly after TIPS procedures wouldhave been considered to be at low riskfor TIPS. If MELD scores alone wereconsidered, one of our responders andlong-term survivors would have beenexcluded, and three patients who diedwithin 3 months would have beendeemed good TIPS candidates (MELDscores of 8, 17, and 17). It seems rea-sonable, therefore, to evaluate patientswith refractory ascites with use ofMELD scoring before attempting TIPS.Patients with very low or very highMELD scores should or should not un-dergo TIPS procedures, respectively,and patients with borderline MELD
scores can be evaluated further withMR perfusion analysis. Additionally,patients with cirrhosis who undergoMR abdominal imaging can also havetheir hepatic perfusion evaluated withthe simple addition of a dynamic MRhepatic sequence. With these perfu-sion, laboratory, and clinical data, theinterventionalist can make reasonabledecisions about the application of TIPSto particular patients. To evaluate allpotential TIPS patients with MR he-patic perfusion scans would be pro-hibited by cost, and we are currentlylimiting this evaluation to those pa-tients with borderline MELD scores.
Selective use of this technique canalso be helpful to evaluate patients de-nied TIPS on the basis of poor clinicalor laboratory assessment, even thoughour data do not clearly support this.One patient in this study had a MELDscore of 23 but responded well to TIPSand lived almost 2 years after the pro-cedure. With such an approach andimproved patient selection, mortalityafter TIPS creation may decrease andinterventional radiology and resourceallocation may improve. From our re-sults, we surmise that TIPS likely has-tens death in patients with poor he-patic perfusional reserve and that theirsurvival is limited without transplan-tation. We have adopted conservativemeasures to reduce the fluid burden inthese patients (tunneled catheters orports) during their limited remaininglifespan.
Why patients with reduced arterial-type hepatic perfusion respond poorlyto TIPS is not clear. One possibility isthat some patients with advanced cir-rhosis have poorly formed or obliter-ated transsinusoidal or transplexalcommunications between the hepaticarterioles and portal venous system(sinusoids). In this situation, the he-patic arterial bed could not undergosufficient hypertrophy in response tothe decreasing portal flow experiencedin all cirrhotic livers. The MR perfu-sion curve in such livers would be ar-terial (early TTP) as a result of the lossof portal flow, but would be dimin-ished because of lack of a significantarterial-to-portal shunting with re-duced sinusoidal perfusion. AfterTIPS creation, the resulting systemicshunt of portal blood would over-whelm the meager arterial compensa-tory mechanism and lead to early andprofound hepatic failure. Patients with
Table 2Laboratory/Clinical Comparison between Survivors and Patients Who Died Early
Note.—Values presented as means � SD. INR � International Normalized Ratio.
1256 • Hepatic Perfusion as a Predictor of Mortality after TIPS October 2003 JVIR
a higher and/or delayed hepatic per-fusion peak may have retained thesearterial–portal communications, andthey may have enlarged, allowing bet-ter sinusoidal blood flow after TIPScreation. Whatever the reasons are, wefound that patients who retain portal-type perfusion (TTP � 50 seconds) orcompensated arterial-type perfusion(TTP � 50 seconds but with enhance-ment equal or greater than in thekidney) have improved survivalcompared with those patients withvery characteristic early but severe-ly diminished hepatic parenchymalperfusion.
Although Doppler US measuresgross flow within major hepatic ves-sels rather than actual parenchymalperfusion (as measured by dynamicMR perfusion scanning), it is notewor-thy that six of the eight patients withunfavorable perfusion on MR hadvery poor portal flow and/or nonoc-clusive portal thrombus on US exami-nation. Such Doppler findings mayalso indicate poor hepatic compensa-tion for decreasing portal blood flow.In this study and in our earlier studyinvolving nuclear medicine hepaticperfusion scans, we identified a groupof patients with strong, early enhance-ment (low TTP) of the liver with he-patic curves similar to or higher thanthe renal perfusion curves. We sur-mise that these patients have excellenthepatic arterial compensation for re-duced portal perfusion and do wellafter TIPS procedures. Interestingly,this perfusion pattern appears moreprevalent in patients with varicealhemorrhage than in those with refrac-tory ascites.
Although use of renal perfusioncurves as a standard of reference inthis study is confounded by the in-volvement of patients with renal dys-function, we believed that it was a use-ful method to gauge the type anddegree of hepatic enhancement. Nowthat we have experience with this tech-nique, use of absolute values of rela-tive enhancement is reasonable, par-ticularly in patients with renaldysfunction. A relative hepatic en-hancement of 30 units or less with aTTP � 50 seconds would be character-istic of a high-risk TIPS candidate.
Our study is limited by the smallsample size obtained retrospectively.Additionally, several patients did notundergo a TIPS procedure because of
poor prognostic signs on MR imaging.Although it is possible that the pa-tients deferred TIPS would have re-sponded and experienced long-termsurvival after TIPS creation, we thinkthis is highly unlikely, given that theydied of hepatic failure within 6 monthsof our initial TIPS evaluation. Our se-ries is small because of its inclusion ofonly those patients referred for TIPSfor refractory ascites. However, we de-cided to focus on this population be-cause of its fairly high postproceduralmortality rates.
Additionally, TIPS creation is anonemergent procedure, and less-in-vasive methods for fluid managementare available. Therefore, after clinicaland laboratory evaluation of these pa-tients, we perform perfusional MR im-aging just before TIPS creation. Theunfavorable perfusion pattern of lowarterial-type flow justifies an alterna-tive therapy for ascites or hydrothoraxbecause we believe these patients willhave poor survival with or withoutthis expensive and highly invasiveprocedure.
CONCLUSION
TIPS creation is a good option forpatients with refractory ascites pro-vided they are chosen carefully. Asubset of patients experience earlymortality, and the 6-month mortalityrate approaches 50%. Interventionalradiologists rely on clinical and labo-ratory parameters to identify thesepoor responders. We found that pre-TIPS MR imaging evaluation of he-patic perfusion provides anothermethod to exclude high-risk TIPS can-didates. Patients with arterial hepaticperfusional waveforms (peak en-hancement � 50 seconds) and sub-re-nal enhancement levels did not re-spond to TIPS and experienced high6-month mortality rates (with or with-out TIPS) compared with patients withslow portal-type hepatic enhance-ment. When patients have unfavorablehepatic perfusion profiles, alternativetherapies for ascites are preferred.
Acknowledgment: The authors thankVicki McDowell for her valuable time andeffort in preparing this manuscript.
References1. Nazarian GK, Bjarnason H, Dietz CA,
et al. Refractory ascites: midterm re-
sults of treatment with a transjugularintrahepatic portosystemic shunt. Ra-diology 1997; 205:173–180.
2. Crenshaw WB, Grodon FD, McEniffNJ, et al. Severe ascites: efficacy of thetransjugular intrahepatic portosys-temic shunt in treatment. Radiology1996; 200:185–192.
3. Malinchoc M, Kamath PS, Gordon FD,Peine CJ, Rank J, terBorg PCJ. Amodel to predict poor survival in pa-tients undergoing transjugular intrahe-patic portosystemic shunts. Hepatol-ogy 2000; 31:864 – 871.
4. Ferral H, Bjarnason H, Wegryn SA, etal. Refractory ascites: early experi-ence in treatment with transjugular in-trahepatic portosystemic shunt. Radi-ology 1993; 189:795– 801.
5. Salerno F, Merli M, Cazzaniga M, et al.MELD score is better than Child-Pughscore in predicting 3-month survival ofpatients undergoing transjugular intra-hepatic portosystemic shunt. J Hepatol2002; 36:494 –500.
6. LaMont JT, Koff RS, Isselbacher KJ.Cirrhosis. In: Petersdorf RG, AdamsRD, Braunwald E, Isselbacher KJ,Martin JB, Wilson JD, eds. Harrison’sprinciples of internal medicine, 10thedition. New York: McGraw-Hill, 1983;1804 –1816.
7. Peron JM, Barange K, Otal P, et al.Transjugular intrahepatic portosys-temic shunts in the treatment of refrac-tory ascites: results in 48 consecutivepatients. J Vasc Interv Radiol 2000; 11:1211–1216.
8. Deschenes M, Dufresne MP, Bui B, etal. Predictors of clinical response totransjugular intrahepatic portosys-temic shunt (TIPS) in cirrhotic patientswith refractory ascites. Am J Gastroen-terol 1999; 94:1361–1365.
9. Walser EM, De La Pena R, Villaneuva-Meyer J, Ozkan O, Soloway R. He-patic perfusion before and after theTIPS procedure: impact on survival. JVasc Interv Radiol 2000; 11:913–918.
10. Van Beers BE, Leconte I, Materne R,Smith AM, Jamart J, Horsmans Y.Hepatic perfusion parameters inchronic liver disease: dynamic CT mea-surements correlated with disease se-verity. AJR Am J Roentgenol 2001; 176:667– 673.
11. Spencer EB, Cohen DT, Darcy MD.Safety and efficacy of transjugular in-trahepatic portosystemic shunt cre-ation for the treatment of hepatic hy-drothorax. J Vasc Interv Radiol 2002;13:385–390.
12. Ferral H, Vasan R, Speeg K, et al.Evaluation of a model to predict poorsurvival in patients undergoing elec-tive TIPS procedure. J Vasc Interv Ra-diol 2002; 13:1103–1108.
