You are on page 1of 5

CT Signs of Hepatofugal Portal

Venous Flow in Patients with Cirrhosis

Thomas J. Bryce1 OBJECTIVE. We investigated whether CT signs can be used to predict hepatofugal flow
Benjamin M. Yeh1 in the main portal vein in patients with cirrhosis.
Aliya Qayyum1 MATERIALS AND METHODS. We retrospectively identified 36 patients with cirrho-
Preeyacha Pacharn2 sis, 18 with hepatopetal and 18 with hepatofugal flow in the main portal vein, who underwent
Nathan M. Bass3 contemporaneous abdominal sonography and CT. Two independent observers evaluated the
following features on the randomized CT studies: diameter of the portal, splenic, and superior
Ying Lu1
mesenteric veins; spleen size; and the presence of ascites, varices, or arterial phase portal
Fergus V. Coakley1 venous enhancement. These data were correlated with the flow direction seen on sonography.
RESULTS. A small main portal vein was the only sign significantly (p ≤ 0.05) predictive
of hepatofugal flow by univariate and multivariate analyses. Observers 1 and 2 recorded a
portal vein diameter of less than 1 cm in eight (44%) and seven (39%) of the 18 patients with
hepatofugal flow compared with one (6%) and none of the 18 patients with hepatopetal flow,
respectively (p < 0.02). Receiver operating characteristic analysis using the size of the portal
vein to predict flow direction revealed an area under the curve of 0.83 for observer 1 and 0.74
for observer 2.
CONCLUSION. A diameter of less than 1 cm for the main portal vein is highly specific,
although not sensitive, for hepatofugal portal venous flow in patients with cirrhosis. This sign
may be useful when sonography is limited, or this sign may prompt sonographic assessment
in patients not known to have hepatofugal flow.

epatofugal portal venous flow in vein [9, 10], although MRI [11–13] and con-
patients with cirrhosis indicates ventional angiography [14, 15] can also be
advanced portal hypertension used. To our knowledge, the role of CT in
[1], and cirrhotic patients with hepatofugal determining the direction of portal venous
flow have greater hepatic dysfunction [2], a flow has not been systematically examined,
higher incidence of hepatic encephalopathy although patients with cirrhosis frequently
[3], an increased risk of variceal bleeding undergo CT, particularly when hepatocellu-
[2, 4, 5], poorer response of varices to endo- lar carcinoma is a concern. The establish-
scopic ligation [6], and higher mortality [2, ment of CT criteria for the detection of
Received April 10, 2003; accepted after revision 7], compared with cirrhotic patients with hepatofugal flow in the main portal vein
June 26, 2003. hepatopetal flow. The detection of hepatofu- could be helpful in the evaluation of such pa-
Department of Medicine and Department of Radiology, gal flow in the main portal vein implies the tients. Therefore, we undertook this study to
Abdominal Imaging Section, University of California at San liver is perfused solely by the hepatic artery, determine whether CT signs can be used to
Francisco, Box 0628, 505 Parnassus Ave., San Francisco,
CA 94143-0628. Address correspondence to which is of therapeutic importance in plan- predict hepatofugal portal venous flow in pa-
F. V. Coakley ( ning chemoembolization of a hepatocellular tients with cirrhosis.
2 carcinoma or placement of a transjugular in-
Department of Radiology, Mahidol University, 2 Prannok
Rd., Bangkok, 10700 Thailand. trahepatic portosystemic shunt (TIPS). CT Materials and Methods
Department of Medicine, Division of Gastroenterology, arterioportography and portal venous an- Patients
University of California at San Francisco, Box 0538, 505 giography may be ineffective and therefore
Parnassus Ave., San Francisco, CA 94143-0538. This study was a retrospective single-institution
contraindicated in the presence of hepatofu- study approved by our institutional review board.
AJR 2003;181:1629–1633
gal portal venous flow [8]. Doppler sonogra- Informed consent was not required. We searched
0361–803X/03/1816–1629 phy is the principal technique used to our radiology information system (IDXrad [soft-
© American Roentgen Ray Society determine the direction of flow in the portal ware version 9.7.1], IDX Systems, Burlington, VT)

AJR:181, December 2003 1629

Bryce et al.

and medical records for the period of October 1998 time lag between CT and sonography for the two server. All p values of 0.05 or less were considered
to March 2002 to identify patients meeting the fol- patients with hepatopetal flow was 3 and 31 days, significant. Interobserver agreement for categori-
lowing sequential criteria: abdominal sonography and for the two patients with hepatofugal flow, the cal data was measured with kappa statistics [16].
report containing the term “cirrhosis”; contempora- lag was 37 and 62 days. Additional studies were Interobserver agreement for continuous data was
neous (within 5 months) abdominal CT; direction not available for the remaining three hepatopetal assessed using Bland-Altman regression [17] with
of flow in the main portal vein described in the and two hepatofugal patients. All additional sono- Bradley-Blackwood p values [18]. Multivariate
sonography report; absence of TIPS or thrombosis graphic studies evaluated confirmed the flow di- analysis was performed using logistic regression
of the main portal vein; and histologic diagnosis of rection seen on the initial study. with a generalized linear model to account for the
cirrhosis (n = 11) or a clear clinical diagnosis of presence of two observers. Parameters were se-
cirrhosis documented in the medical record (n = Imaging Technique lected in a stepwise fashion with a type 3 signifi-
25). We identified 18 patients with hepatofugal All CT examinations were performed on mul- cance level of 0.05 or less required for factors to
flow in the main portal vein who met these criteria. tidetector scanners (LightSpeed or HiSpeed, Gen- remain in the model. Receiver operating character-
We then randomly selected an additional 18 pa- eral Electric Medical Systems, Milwaukee, WI). istic (ROC) analysis and area under the ROC curve
tients with hepatopetal flow who also met these cri- Thirty-four of the 36 patients received 150 mL of (Az) calculations were performed to evaluate pre-
teria to form the final study population of 36 IV iohexol (Omnipaque 350, Nycomed Amersham, dictive models.
patients. None of the patients was assigned a Princeton, NJ), and images were acquired in the
Child-Pugh class because not all required data portal venous phase of enhancement (70-sec scan
were available retrospectively. delay with 5-mm slice collimation). Images were Results
The mean age of patients in the hepatopetal also acquired in the arterial phase of enhancement The CT signs recorded by each observer
group was 59 years (range, 44–80 years) versus 51 (45-sec scan delay with 2.5-mm slice collimation) showed high interobserver consistency for
years (range, 17–73 years) in the hepatofugal in 20 of these patients. Two patients (one with all measurements other than the presence of
group. The hepatopetal group comprised 10 men hepatopetal and one with hepatofugal flow) did not varices and the size of the spleen. A small
and eight women, and the hepatofugal group com- receive IV contrast material. All patients received main portal vein correlated strongly (p <
prised seven men and 11 women. In the hepato- oral diatrizoate meglumine (Hypaque, Nycomed
0.01) with hepatofugal flow by univariate
petal group, the diagnosis of cirrhosis was Amersham). All images were contiguous. Color
and multivariate analyses (Fig. 1). No other
established by histology in six patients and on Doppler sonography of the direction of flow in the
clinical grounds in 12; in the hepatofugal group, main portal vein was performed using a scanner signs achieved significance by either
cirrhosis was established by histology in five pa- (Sequoia 512, Acuson Solutions, Mountain View, method. ROC analysis using the diameter of
tients and on clinical grounds in 13. Of those pa- CA) with a 1.75- to 4-MHz sector transducer (4V1, the main portal vein to predict hepatofugal
tients without histologic confirmation, cirrhosis Acuson Solutions) or a 2.5- to 4-MHz sector trans- flow revealed an Az of 0.83 for observer 1
was diagnosed by a gastroenterologist in all but ducer (4V2, Acuson Solutions). and 0.74 for observer 2 (Fig. 2).
two, both of whom had hepatofugal portal venous Observer 1 found that the main portal vein
flow. In these two patients, cirrhosis was diag- CT Interpretation
measured less than 1 cm in eight (44%) of 18
nosed on the basis of clinical evidence including a Two radiologists independently reviewed the patients with hepatofugal flow, but in only
history of severe alcohol abuse; gross findings of randomized CT images of all 36 patients on a
one (6%) of 18 patients with hepatopetal
cirrhosis on imaging studies; and large varices, as- PACS (picture archiving and communication sys-
flow (p < 0.02). Thus, the sensitivity of this
cites, hypoalbuminemia, and coagulopathy. The tem) workstation (Impax DS 3000 [release 4.1],
causes for cirrhosis in the group with hepatopetal Agfa, Mortsel, Belgium). Observers were unaware sign for predicting hepatofugal flow was
flow and in the group with hepatofugal flow, re- of clinical and sonographic findings. Both observ- 44% and the specificity was 94%. Observer 2
spectively, included the following: chronic viral ers recorded the following CT signs: short-axis di- found that the main portal vein measured less
hepatitis, 11 and four patients; viral hepatitis and ameter of the main portal vein, which was than 1 cm in seven (39%) of 18 patients with
ethanol abuse, two patients and one patient; etha- measured midway between the splenoportal con- hepatofugal flow and 0 of 18 patients with
nol abuse alone, 0 and six patients; autoimmune fluence and the portal vein bifurcation in the porta hepatopetal flow (p < 0.01), for a sensitivity
hepatitis, one and three patients; nonalcoholic ste- hepatis; maximum short-axis diameter of the supe- of 39% and a specificity of 100%.
atotic hepatitis, one patient and 0 patients; primary rior mesenteric vein, measured on the first image The 1-cm measurement was the largest
sclerosing cholangitis, one and three patients; and that was clearly inferior to the splenoportal conflu-
(most sensitive) threshold that could be used
cystic fibrosis, 0 and one patient. Cirrhosis was ence; diameter of the splenic vein, measured adja-
to predict hepatofugal flow without signifi-
cryptogenic in two patients with hepatopetal flow. cent to the midportion of the pancreatic tail;
The median time between sonography and CT presence or absence of ascites; presence or ab- cantly reducing specificity, given that both
was 23 days (range, 0–148 days) in the hepato- sence of varices; and spleen size, recorded as the observers measured the main portal vein at
petal group and 3 days (range, 0–97 days) in the maximum axial diameter. In addition, for mul- or slightly above 1 cm in several patients
hepatofugal group. If sonography had been per- tiphase CT examinations (n = 20), the presence or with hepatopetal flow (Fig. 1). A lower
formed more than 48 hr before or after the CT ex- absence of early (i.e., arterial phase) enhancement threshold would reduce sensitivity without
amination (hepatopetal group, n = 14; hepatofugal of the main portal vein was recorded. significantly increasing specificity, and a
group, n = 9), we identified a second Doppler higher threshold would greatly reduce speci-
sonographic study of flow direction in the main Data Analysis
ficity with only mildly increasing sensitivity
portal vein for each patient so that sonography oc- Statistical analysis was performed using statis-
(Fig. 2). A representative case illustrates the
curred before and after the CT examination tical analysis software (SAS version 8.1, SAS,
finding of a small portal vein in a patient
(hepatopetal group, n = 9; hepatofugal group, n = Cary, NC). Continuous data (vessel diameters and
5). Alternatively, if CT had not been performed be- spleen size) were examined by univariate analysis with hepatofugal flow (Fig. 3). Arterial phase
tween the sonographic examinations, we identified using the two-tailed Cochran t test. Noncontinuous enhancement of the portal vein was seen in
the two sonographic examinations closest in time data (all other categories) were examined using only one patient; in that patient, flow in the
to the CT examination (hepatopetal group, n = 2; Fisher’s exact test. The univariate analyses were portal vein was hepatopetal on sonography 1
hepatofugal group, n = 2). In these patients, the performed separately for the data for each ob- day before CT (Fig. 4).

1630 AJR:181, December 2003

CT Signs of Hepatofugal Portal Venous Flow

Diameter of Main Portal Vein (cm)


True-Positive Fraction
1.5 0.6

1 0.4


Hepatopetal Hepatofugal Hepatopetal Hepatofugal 0
Observer 1 Observer 2 0 0.2 0.4 0.6 0.8 1
False-Positive Fraction

Fig. 1.—Diagram shows diameters of main portal vein in patients with hepatopetal Fig. 2.—Diagram shows receiver operating characteristic (ROC) values achieved using
flow versus hepatofugal flow. Note that main portal vein diameter of less than 1 cm small main portal vein diameter to predict hepatofugal flow in main portal vein in pa-
is highly specific for hepatofugal flow. tients with cirrhosis. Operating points achieved by predicting hepatofugal flow when
main portal vein diameter was less than 1 cm are shown for observer 1 (×) and observer
2 (Δ). Note that at appropriate operating points, this sign is highly specific for hepatofu-
gal flow. Area under ROC curve (Az) was 0.83 for observer 1 and 0.74 for observer 2.


Fig. 3.—73-year-old woman with cirrhosis due to alcohol abuse.

A, Axial CT scan obtained during portal venous phase shows small (8 mm in diameter) main portal vein (arrow); this finding strongly correlates with hepatofugal flow.
B, Sonogram shows hepatofugal flow in main portal vein.

AJR:181, December 2003 1631

Bryce et al.

Discussion stages of portal hypertension and the possi- Our study has several limitations. The sam-
Our finding that a relatively small main bility that portal hypertension may provide a ple size was small and may have restricted the
portal vein in patients with cirrhosis indi- continuing impetus for venous enlargement, power of the study to detect differences be-
cates hepatofugal flow is consistent with the despite a reduction or reversal in flow. tween the hepatopetal and hepatofugal
physiology of hepatofugal flow in the cir- Arterial phase enhancement of the portal groups. For example, the prevalence of ascites
rhotic liver. The average size of the main vein has been reported as a sign of hepatofu- and varices was higher in patients with
portal vein in healthy adults has been found gal flow [1, 22]. In our study, this sign was hepatofugal flow, but these differences did not
to be 1.0 cm [19]; a main portal vein of less present in only one of 20 patients undergoing achieve statistical significance. These factors
than 1 cm can be considered relatively small CT with arterial phase contrast enhancement, might exhibit an association with hepatofugal
in a cirrhotic patient because portal hyper- and Doppler sonography of this patient 1 day flow in a larger study. CT and sonography
tension tends to increase portal vein size before CT showed hepatopetal flow. The were not immediately contemporaneous. The
[20]. Normally, the portal vein receives flow finding of arterial phase enhancement of the lag between the examinations may have low-
from the superior mesenteric vein and the portal vein on CT in a patient with hepato- ered the sensitivity of the study for additional
splenic vein. In patients with cirrhosis and petal portal venous flow on sonography is predictors of hepatofugal flow, although the
hepatofugal flow in the main portal vein, the puzzling but might be explained by transient possible effects of the time difference would
portal vein is supplied only by the hepatic changes in flow direction in the portal be expected to apply to both hepatopetal and
artery, which also supplies the hepatic veins. venous system. Such flow changes can occur hepatofugal groups of patients. The lag be-
This decrease in flow volume could explain spontaneously in cirrhotic patients [23], in tween studies would not be expected to gener-
the decreased diameter of the portal vein. (A the postprandial state [24, 25], and during ate spurious associations and should not
small main portal vein in cirrhotic patients hepatic arteriography [26]. Disturbance of detract from the finding that a small portal
has been associated with spontaneous sple- baseline physiology during abdominal CT, vein size is associated with hepatofugal portal
norenal shunts [21], a phenomenon that such as due to breath-hold technique or bolus venous flow. Analysis of additional sonogra-
would also be expected to reduce hepatic administration of IV contrast material, may phy examinations of the study population pro-
portal venous perfusion.) Although specific alter portal venous hemodynamics. For ex- vided evidence that the direction of portal
for hepatofugal flow, this sign was of only ample, contrast administration might elevate venous flow in our subjects was stable. There-
moderate sensitivity in our study group. right heart and hepatic vein pressures, and fore, it is reasonable to expect that the direc-
Possible explanations for the limited sensi- the subsequent increased shunting of hepatic tion of flow at the time of CT would be
tivity may include a tendency for the vein to arterial blood into the portal venous system accurately predicted by the contemporaneous
remain patulous after enlarging in earlier may result in temporary hepatofugal flow. sonographic examination that was evaluated.


Fig. 4.—49-year-old man with cirrhosis due to chronic hepatitis C infection.

A, CT scan obtained during arterial phase shows contrast enhancement of portal vein. This sign is thought to indicate hepatofugal portal venous flow.
B, Doppler sonogram obtained 1 day earlier than A, however, reveals hepatopetal flow.

1632 AJR:181, December 2003

CT Signs of Hepatofugal Portal Venous Flow

High interobserver consistency was not sonographic assessment in patients not 15. Nelson RC, Lovett KE, Chezmar JL, et al. Com-
seen for the measurement of spleen size and known to have hepatofugal flow. parison of pulsed Doppler sonography and an-
giography in patients with portal hypertension.
assessment for the presence or absence of va-
AJR 1987;149:77–81
rices. The discrepancy in identification of va- References 16. Landis JR, Koch GG. The measurement of ob-
rices between observers could be explained by 1. Wachsberg RH, Bahramipour P, Sofocleous CT, server agreement for categorical data. Biometrics
a greater sensitivity of observer 2 for varices in Barone A. Hepatofugal flow in the portal venous 1977;33:159–174
borderline cases, given that all cases identified system: pathophysiology, imaging findings, and di- 17. Bland JM, Altman DG. Statistical methods for as-
as positive by observer 1 were also called posi- agnostic pitfalls. RadioGraphics 2002;22:123–140 sessing agreement between two methods of clini-
2. Tarantino L, Giorgio A, de Stefano G, et al. Re- cal measurement. Lancet 1986;1:307–310
tive by observer 2, but not vice versa. The in-
verse flow in intrahepatic portal vessels and liver 18. Bradley EL, Blackwood LG. Comparing paired
terobserver variation regarding spleen size function impairment in cirrhosis. Eur J Ultra- data: a simultaneous test of means and variances.
may have resulted from the complex and sound 1997;6:171–177 Am Stat 1989;43:234–235
partly subjective methodology used, which in- 3. Smith-Laing G, Camilo ME, Dick R, Sherlock S. 19. Niederau C, Sonnenberg A, Muller JE, Erckenbre-
volved identifying the maximal single dimen- Percutaneous transhepatic portography in the as- cht JF, Scholten T, Fritsch WP. Sonographic mea-
sion of the spleen on any axial image. sessment of portal hypertension: clinical correla- surements of the normal liver, spleen, pancreas, and
tions and comparison of radiographic techniques. portal vein. Radiology 1983;149:537–540
This study did not seek to measure the
Gastroenterology 1980;78:197–205 20. Subramanyam BR, Balthar EJ, Raghavendra BN,
prevalence of hepatofugal flow in the entire 4. Wachsberg RH, Simmons MZ. Coronary vein di- Lefleur RS. Sonographic evaluation of patients
population of cirrhotic patients, although this ameter and flow direction in patients with portal with portal hypertension. Am J Gastroenterol
information would be required to determine hypertension: evaluation with duplex sonography 1983;78:369–373
predictive values. This is important given that and correlation with variceal bleeding. AJR 21. Brancatelli G, Federle MP, Pealer K, Geller DA.
our sign, although highly specific, is of lim- 1994;162:637–641 Portal venous thrombosis or sclerosis in liver
5. Komatsuda T, Ishida H, Konno K, et al. Color transplantation candidates: preoperative CT find-
ited sensitivity. The prevalence of hepatofu-
Doppler findings of gastrointestinal varices. Ab- ings and correlation with surgical procedure. Ra-
gal flow in unselected cirrhotic patients has dom Imaging 1998;23:45–50 diology 2001;220:321–328
been previously estimated at 3.1–3.4% [2, 6. Matsumoto A, Hamamoto N, Ohnishi A, et al. 22. Park CM, Cha SH, Kim DH, et al. Hepatic arteri-
27], rising to 9% or more in patients with ad- Left gastric vein hemodynamics and variceal re- oportal shunts not directly related to hepatocellular
vanced cirrhosis [2, 28]. Finally, the diagno- currence in patients undergoing prophylactic en- carcinoma: findings on CT during hepatic arteriog-
sis of cirrhosis was confirmed histologically doscopic ligation of high-risk esophageal varices. raphy, CT arterial portography and dual phase spi-
in only about one third of the study subjects. Gastrointest Endosc 1999;50:768–774 ral CT. Clin Radiol 2000;55:465–470
7. Finucci G, Bellon S, Merkel C, et al. Evaluation 23. Hoevels J, Lunderquist A, Tylen U. Spontaneous
However, most patients with a diagnosis of
of splanchnic angiography as a prognostic index intermittent reversal of blood flow in intrahepatic
cirrhosis do not undergo histologic confirma- of survival in patients with cirrhosis. Scand J portal vein branches in cirrhosis of the liver. Car-
tion. Restricting the analysis to only those pa- Gastroenterol 1991;26:951–960 diovasc Radiol 1979;2:267–273
tients with a histologic diagnosis would have 8. Oliver JH 3rd, Baron RL, Dodd GD 3rd, Peterson 24. de Vries PJ, de Hooge P, Hoekstra JB, van Hat-
selected for substantially more diseased or MS, Carr BI. Does advanced cirrhosis with porto- tum J. Postprandial reversal of the portal venous
symptomatic patients, possibly harming the systemic shunting affect the value of CT arterial flow in a patient with liver cirrhosis. Neth J Med
portography in the evaluation of the liver? AJR 1995;47:235–240
generalizability of our results. Given that all
1995;164:333–337 25. Tochio H, Kudo M, Nishiuma S, Okabe Y. Intra-
patients in this study had significant risk fac- 9. Ralls PW. Color Doppler sonography of the he- hepatic spontaneous retrograde portal flow in pa-
tors for cirrhosis, had imaging features con- patic artery and portal venous system. AJR tients with cirrhosis of the liver: reversal by food
sistent with cirrhosis, and had either 1990;155:517–525 intake. AJR 2001;177:1109–1112
histologic confirmation of cirrhosis or had 10. Johansen K, Paun M. Duplex ultrasonography of the 26. Ohnishi K, Saito M, Sato S, et al. Direction of
been assigned the diagnosis of cirrhosis at a portal vein. Surg Clin North Am 1990;70:181–190 splenic venous flow assessed by pulsed Doppler
11. Burkart DJ, Johnson CD, Morton MJ, Ehman RL. flowmetry in patients with a large splenorenal
tertiary care center (by a gastroenterologist in
Phase-contrast cine MR angiography in chronic shunt: relation to spontaneous hepatic encephalop-
all but two cases), we believe that the proba- liver disease. Radiology 1993;187:407–412 athy. Gastroenterology 1985;89:180–185
bility of noncirrhotic patients existing within 12. Imazu H, Matsui T, Noguchi R, et al. Magnetic 27. Gaiani S, Bolondi L, Li Bassi S, Zironi G, Siringo
the study population is low. resonance angiography for monitoring prophylac- S, Barbara L. Prevalence of spontaneous hepatofu-
In conclusion, our data suggest that a main tic endoscopic treatment of high risk esophageal gal portal flow in liver cirrhosis: clinical and endo-
portal vein diameter of less than 1 cm is varices. Endoscopy 2000;32:766–772 scopic correlation in 228 patients. Gastroenterology
highly specific for hepatofugal portal venous 13. Finn JP, Kane RA, Edelman RR, et al. Imaging of 1991;100:160–167
the portal venous system in patients with cirrho- 28. von Herbay A, Frieling T, Haussinger D. Color
flow in cirrhotic patients, although the sensi-
sis: MR angiography vs duplex Doppler sonogra- Doppler sonographic evaluation of spontaneous
tivity of this sign is limited. If validated in phy. AJR 1993;161:989–994 portosystemic shunts and inversion of portal
prospective studies, this sign may be useful 14. Herlinger H. Arterioportography. Clin Radiol venous flow in patients with cirrhosis. J Clin Ul-
when sonography is limited or may prompt 1978;29:255–275 trasound 2000;28:332–333

AJR:181, December 2003 1633