LIVER TRANSPLANTATION 21:547–553, 2015
ORIGINAL ARTICLE
Pediatric Liver Transplant Portal Vein Anastomotic Stenosis: Correlation Between Ultrasound and Transhepatic Portal Venography C. Matthew Hawkins,1 Dennis W. W. Shaw,1 Patrick J. Healey,2 Simon P. Horslen,3 Andre A. S. Dick,2 Seth Friedman,1 and Giridhar M. Shivaram1 Divisions of 1Interventional Radiology and 2Pediatric Transplant Surgery and 3Liver and Intestinal Transplantation, Division of Gastroenterology and Hepatology, Seattle Children’s Hospital, University of Washington, Seattle, WA
The objective of this study was to determine which transabdominal ultrasound parameters correlate with portal vein stenosis (PVS) on percutaneous transhepatic portal venography in pediatric liver transplant patients. A retrospective review was performed of percutaneous transhepatic portal venograms performed between 2005 and 2013. The findings were compared to those from ultrasounds performed before venography and at the baseline. Patients were stratified on the basis of the presence of significant PVS (group 1, >50% stenosis; group 2, 50% stenosis) on portal venography. Findings were compared to those for age-matched controls. Twenty portal venograms were performed for 12 pediatric patients. Thirteen of the 20 patients (65%) demonstrated significant PVS (>50%). The mean peak anastomotic velocity (PAV) was 253.6 6 96 cm/s in group 1, 169.7 6 48 cm/s in group 2, and 51.3 6 20 cm/s in the control group. PAV (r 5 0.672, P 5 0.002) was the only ultrasound variable that correlated with the presence of significant PVS. A receiver operating characteristic curve was generated from PAV and PVS data (area under the curve 5 0.75, P 5 0.08). A threshold velocity of 180 cm/s led to a sensitivity of 83% and a specificity of 71% in predicting significant PVS on portal venography. At the baseline, the mean PAV was 155.8 6 90 cm/s for group 1 and 69.5 6 33 cm/s for group 2 (P 5 0.08); for control subjects, it was 78.9 6 53 cm/s (P 5 0.06). PAV is the only measured ultrasound parameter that correlates with significant PVS on portal venography in pediatric liver transplant patients. An elevated baseline PAV may increase the risk of developing PVS. Liver Transpl C 2015 AASLD. 21:547-553, 2015. V Received September 11, 2014; accepted January 2, 2015. Portal vein stenosis (PVS) is a relatively uncommon complication associated with liver transplantation.1-4 However, the incidence of PVS is more common in pediatric patients and has been reported to occur in
as many as 27% of segmental pediatric liver transplants.5-8 Gray-scale ultrasound and Doppler ultrasound (DUS) have become the primary imaging screening
Abbreviations: DAV, change in velocity across the portal vein anastomosis; DUS, Doppler ultrasound; PAV, peak anastomotic velocity; PTPV, percutaneous transhepatic portal venography; PVS, portal vein stenosis; ROC, receiver operating characteristic. Potential conflict of interest: Nothing to report. No grants were used to fund this study. All authors meaningfully contributed to the article and meet the criteria for authorship. C. Matthew Hawkins is currently affiliated with the Department of Radiology and Imaging Sciences, Emory University Hospital, Atlanta, GA Address reprint requests to C. Matthew Hawkins, MD, Department of Radiology and Imaging Sciences, Emory University Hospital, 1364 Clifton Road NE, Suite D112, Atlanta, GA 30322. Telephone: 517-927-6839; FAX: 404-712-7387; E-mail: [email protected] DOI 10.1002/lt.24077 View this article online at wileyonlinelibrary.com. LIVER TRANSPLANTATION.DOI 10.1002/lt. Published on behalf of the American Association for the Study of Liver Diseases
C 2015 American Association for the Study of Liver Diseases. V
548 HAWKINS ET AL.
Figure 1. Digitally subtracted portal venogram for a 6-year-old male liver transplant recipient with PVS (arrow). A 4F diagnostic catheter (Cook Group Inc., Bloomington, IN), passed through a 6F transhepatic sheath (Terumo Corporation, Tokyo), is seen crossing the stenosis and terminating in the superior mesenteric vein. The degree of stenosis is measured by the division of the difference between the maximum diameter of the recipient portal vein and the minimal diameter at the portal vein anastomosis by the maximum diameter of the recipient portal vein.
tools for evaluating the vascular complications associated with liver transplants.9,10 However, the sensitivity and specificity of sonographic parameters in predicting PVS in pediatric patients have not been delineated. Previous studies in adults have suggested that various portal vein velocities (as measured by DUS) can offer reliable predictive capabilities.1,11 However, liver transplantation in children differs in many ways from liver transplantation in adults. Most notably, children receive a larger number of segmental liver transplants rather than whole liver transplants, which are more conventional in adults. Also, the portal vein size mismatch, which is commonly associated with the transplantation of segmental liver transplants in children, may cause additional vascular complications that are not typically encountered in most adult liver transplant patients.8,12,13 The objective of this study was to determine which transabdominal gray-scale ultrasound and DUS parameters correlate with significant PVS in pediatric liver transplants as determined by percutaneous transhepatic portal venography (PTPV). Additionally, we aimed to evaluate whether the peak anastomotic velocity (PAV) on baseline DUS and the change in velocity from baseline DUS over time were associated with the development of PVS.
PATIENTS AND METHODS After institutional review board approval, a retrospective electronic medical record and imaging archive review was performed to identify all PTPVs performed between 2005 and 2013 at a single tertiary-care pediatric hospital. Multiple patient parameters were evaluated, and they included the following: etiology of
LIVER TRANSPLANTATION, April 2015
primary liver disease, type of transplant received, age at transplant, age at PTPV, baseline DUS data (defined as nadir in PAV from available ultrasounds obtained during the first week after transplant), data from DUS within 1 month before PTPV, and PTPV findings and interventions. Imaging and medical charts were reviewed by 2 radiologists. Indications for PTPV included elevated transaminases/alkaline phosphatase, concerning findings from DUS, and/or clinical concerns for portal hypertension. PTPVs were stratified into 2 groups on the basis of the presence (group 1, n 5 13) or absence (group 2, n 5 7) of significant PVS (defined as >50% on PTPV). This parameter has been used to define significant PVS by other investigators.1,14 We measured PVS by dividing the difference between the maximum diameter of the recipient portal vein and the minimal diameter at the portal vein anastomosis by the maximum diameter of the recipient portal vein (Fig. 1). Pressure gradients across the stenosis were measured in only 10 of 20 PTPVs, and this consequently limited the analysis of this variable. Cross-sectional imaging (magnetic resonance imaging and/or computed tomography) was performed immediately before only 5 PTPVs, the findings of which were not analyzed. Each PTPV was compared to the most recent DUS (obtained within 1 month before PTPV). All DUS examinations were performed by ultrasound technologists. The liver transplant DUS protocol for portal vein imaging at our institution entails velocity and waveform measurement of the native portal vein, the portal vein anastomosis, and the donor portal vein. Additionally, the right and left branches of the portal vein (for whole liver transplants) and the left portal vein (for segmental transplants) are imaged to document the direction of flow and patency. The protocol also includes multiple planes of Doppler imaging of the hepatic artery, celiac axis, splenic artery, hepatic veins, inferior vena cava, superior mesenteric vein, and biliary ducts. Velocities and waveforms are obtained for each vessel. Although equipment has changed over time, we currently use Philips iU22 and EPIQ ultrasound machines (Philips Healthcare, Amsterdam, Netherlands) with 8- to 5-MHz curved array probes for infants and with 5- to 2-MHz curved array probes for older children. For patients who have undergone liver transplantation, we perform DUS examinations on postoperative days 1 and 5. Thereafter, annual surveillance ultrasounds are not performed unless clinical suspicion or abnormal laboratory tests warrant an unscheduled DUS examination. On DUS, PAV was defined as the maximum measured velocity at the portal vein anastomosis in the absence of arterial flow artifact (Fig. 2). Maximum prestenotic and poststenotic velocities were also recorded. Velocity data were not recorded on 1 pre-PTPV DUS because the portal vein was interpreted to be occluded at the time of the ultrasound. (This patient was found to have 65% PVS on PTPV.) The PVS percentage was not measured for 4 (3 in group 1 and 1 in group 2) of the 20 pre-PTPV
LIVER TRANSPLANTATION, Vol. 21, No. 4, 2015
HAWKINS ET AL. 549
Figure 2. DUS of a 4-year-old female liver transplant recipient. The maximum portal vein velocity, without arterial flow artifact, was recorded as the PAV.
ultrasounds. Baseline DUS data were available for 6 patients in group 1, 5 patients in group 2, and 9 control subjects. Data from each group were also compared to a control group (n 5 10) of randomly selected, age-matched liver transplant patients with no prior interventions, normal liver function tests, no clinical concern for portal hypertension, and gray-scale ultrasound and DUS findings that were interpreted as normal. Data were transferred to SPSS 19 (IBM, Chicago, IL) for parametric and nonparametric analysis. Correlations were evaluated through the calculation of Pearson product-moment correlation coefficients. To evaluate the predictive potential of PAV on DUS, a receiver operating characteristic (ROC) curve was generated. Additionally, mean values between groups, at the time of PTPV and at the baseline, were compared with a 2-tailed t test. P < 0.05 was considered statistically significant.
RESULTS Patient demographics for group 1 (>50% stenosis), group 2 (50% stenosis), and control subjects are detailed in Table 1. During the study period, 179 liver transplants were performed in 157 patients at our institution. In sum, 20 PTPVs were performed in 12 of these patients (7.6%). Two patients had 2 PTPVs. Three patients had 3 PTPVs. Thirteen PTPVs showed >50% PVS, and 7 PTPVs showed 50% PVS. Interventions were performed during 17 PTPVs (15 venoplasties and 2 stents). The mean age at the time of transplant was not significantly different between the 2 groups (group 1, 24 6 22 months; group 2, 45 6 36 months; P 5 0.22). The median age (and range) at the time of transplant was 18.9 months (range 5 4.3-72.7 months) in group 1 and 37.1 months (range 5 7.2-
91.5 months) in group 2. The mean PAV before PTPV was 253.6 6 96 cm/s in group 1, 169.7 6 48 cm/s in group 2 (P 5 0.04), and 51.3 6 20 cm/s for the agematched-control subjects (F 5 24.52, P < 0.001; group 1 > controls, P 5 0.001; group 2 > controls, P < 0.001; group 1 > group 2, P 5 0.04). Additional pre-PTPV DUS parameters, as detailed in Table 2, were not significantly different between group 1 and group 2 (all F < 0.924, P 5 not significant). An analysis of ultrasound findings and each PTPV revealed that PAV (r 5 0.672, P 5 0.002) was the only DUS variable that correlated with the presence of >50% PVS (Fig. 3). Specifically, the change in velocity across the portal vein anastomosis (DAV; r 5 0.13, P 5 0.59) and the diameter of the stenosis as measured by DUS (r 5 –0.19, P 5 0.51) showed a poor correlation. The ROC curve generated by data from the PAV correlation analysis (Fig. 4) showed an area under the curve of 0.75 (trend level P 5 0.08). Notably, there was complete nonoverlap of PAVs in patients subjected to PTPV with PAVs measured in control subjects. When we evaluated the ability of PAV on DUS to predict the presence of >50% PVS on PTPV, PAV < 100 cm/s showed 100% sensitivity, and PAV > 267 cm/s was 100% specific. Applying a PAV threshold of 180 cm/s led to 83% sensitivity and 71% specificity for the detection of significant PVS; this was the best combination of sensitivity and specificity for a given DUS PAV according to our data. Pressure gradients were measured during only 3 of 13 PTPVs in group 1 (>50% PVS) and in all 7 PTPVs in group 2 (50% stenosis). The mean pressure gradient across the area of stenosis was 7.5 6 3.9 mm Hg in group 1 and 3.9 6 4.8 mm Hg in group 2 (P 5 0.17). A single patient in group 2 had a pressure gradient greater than 5 mm Hg (15 mm Hg). Angioplasty was performed for this patient despite PVS on PTPV 50%
550 HAWKINS ET AL.
LIVER TRANSPLANTATION, April 2015
TABLE 1. Patient Demographics
Number of patients Number of PTPVs Mean age at transplant (months) Indication for transplant Biliary atresia Acute liver failure Autoimmune hepatitis Neonatal hemochromatosis Idiopathic cirrhosis Cryptogenic cirrhosis Progressive familial intrahepatic cholestasis Carbamoyl phosphatase synthetase deficiency Transplant type Mean age at PTPV (months) Interventions performed during PTPV
Group 1 (>50% PVS)
Group 2 (50% PVS)
Control Group
7 13 24
5 7 45
10 — 38
5 1 1 — — — —
3 2 — — — — —
3 2 — 1 1 1 1
—
—
1
4/7 segmental (57%) 51 11 angioplasty 2 angioplasty/stent
4/5 segmental (80%) 80 4 angioplasty
6/10 segmental (60%) — —
TABLE 2. DUS Data
Pre-PTPV DUS Mean PAV (cm/s) Mean pre-anastomotic velocity (cm/s) Mean postanastomotic velocity (cm/s) Mean DAV (cm/s) Mean % PVS Baseline DUS Mean PAV (cm/s) Mean D in PAV from baseline to time of PTPV (cm/s)
Group 1 (>50%
Group 2 (50%
PVS on PTPV)
PVS on PTPV)
Control Group
253.6 6 96 (n 5 12) 39.5 6 16 (n 5 12)
169.7 6 48 (n 5 7) 37.4 6 11 (n 5 7)
51.3 6 20 (n 5 10) 47.3 6 21 (n 5 10)
119.8 6 58 (n 5 12)
103.2 6 32 (n 5 7)
46.9 6 22 (n 5 10)
80.3 6 62 (n 5 12) 71.8 6 15 (n 5 10)
65.9 6 35 (n 5 7) 74.5 6 6 (n 5 6)
11.2 6 20 (n 5 10) —
155.2 6 90 (n 5 6) 93.2 6 106 (n 5 6)
69.5 6 33 (n 5 5) 113.9 6 40 (n 5 5)
78.9 6 53 (n 5 9) —
NOTE: Pre-PTPV mean PAV: F 5 24.52, P < 0.001; group 1 > controls, P 5 0.001; group 2 > controls, P < 0.001; and group 1 > group 2, P 5 0.04. All other pre-PTPV DUS parameters were not significantly different between group 1 and group 2 (all F < 0.924, P 5 not significant).
(45%). A single patient in group 1 had a pressure gradient less than 5 mm Hg (4 mm Hg). Angioplasty was performed because this patient had 66% PVS on PTPV. On baseline DUS, patients in group 1 had a mean PAV of 155.2 6 90 cm/s. This was higher than the baseline mean PAV seen in group 2 (69.5 6 33 cm/s, P 5 0.08) and in control subjects (78.9 6 53 cm/s, P 5 0.056). The change in PAV from the baseline to the time of intervention was 93.2 6 107 cm/s in group 1 and 114 6 40 cm/s in group 2 (P 5 0.69).
DISCUSSION The objective of this retrospective analysis was to determine which gray-scale ultrasound and DUS
parameters correlate with PVS detected by percutaneous portal venography in pediatric liver transplant recipients. Additionally, baseline ultrasounds were evaluated for variables correlating with the development of PVS in this patient population. In our analysis, we demonstrate that PAV is the only DUS parameter that correlates (r 5 0.672) with PVS on PTPV, which is defined as >50% PVS. Specifically, the DAV (r 5 0.13) and the diameter of stenosis measured by DUS (r 5 0.19) correlated poorly. These data also produced various PAV thresholds that could predict the likelihood of PVS in pediatric liver transplant patients. Specifically, PAV < 100 cm/s and >267 cm/ s had 100% sensitivity and specificity, respectively. A PAV threshold of 180 cm/s led to 83% sensitivity and
LIVER TRANSPLANTATION, Vol. 21, No. 4, 2015
Figure 3.
PVS versus PAV measured on DUS.
Figure 4. ROC curve generated by data from a correlation analysis of PAV on DUS and PVS on PTPV (area under the curve 5 0.75, P 5 0.076). There was complete nonoverlap of PAVs in patients subjected to PTPV with PAVs measured in control subjects.
71% specificity in the detection of PVS. Furthermore, an analysis of baseline ultrasounds revealed that patients who developed PVS had a higher mean baseline PAV (155.2 cm/s) than the patients without PVS (69.5 cm/s, P 5 0.08) and the control subjects (78.9 cm/s, P 5 0.06), although these differences were not statistically significant. Similar investigations have been performed in adult patients. Mullan et al.1 demonstrated that using a portal vein velocity threshold of 80 cm/s on DUS yielded a sensitivity of 100% and a specificity of 84% in its ability to predict significant PVS. Chong et al.11 found that a portal vein PAV of 125 cm/s was 95% sensitive and 73% specific for PVS. Finally, Lee et al.15 demonstrated PAVs of 110 to 300 cm/s in 11
HAWKINS ET AL. 551
cases of confirmed PVS in pediatric transplant patients. (Lee et al. defined PVS as a diameter < 2.5 mm on PTPV, an acceleration of flow at the stricture, a poststenotic jet, and/or a portal vein gradient >5 mm Hg. The degree of PVS was not correlated with PAV or DAV.) Our threshold velocity of 180 cm/s is higher than those identified in the prior studies by Mullan et al. and Chong et al., but it is within the range of velocities identified by Lee et al. This is most likely related to variable portal vein flow dynamics in pediatric patients after liver transplantation, which is largely attributable to the surgical size mismatch resulting from the larger size of the donor portal veins.13 Consequently, even mild PVS in children may produce greater increases in PAV than those seen in adults. Two patients in our study had PAV > 180 cm/s with PVS on PTPV 50%. One patient had PAV of 259 cm/s on DUS. PTPV revealed 45% PVS and a pressure gradient of 15 mm Hg across the stenosis. Angioplasty was performed successfully in this patient. An additional patient had PAV of 203 cm/s on DUS. PTPV revealed 39% PVS with a pressure gradient of 2 mm Hg across the stenosis. No intervention was performed. Additionally, portal vein velocities in the pediatric age range (0-18 years) may have a wider dynamic range and be influenced (to a greater degree) by other anatomic abnormalities. For example, 1 patient in our study had 66% PVS despite having a lower PAV (101 cm/s). However, this patient also had hepatic venous outflow stenosis, which likely contributed to decreased portal vein velocity. Another patient had PVS of 64% with a PAV of 112 cm/s in the setting of marked splenomegaly, which may have affected portomesenteric flow. These results suggest that cross-sectional imaging may be beneficial in patients with PAV < 180 cm/s when there is a high clinical suspicion for posttransplant vascular complications. There are a number of limitations to this study. First, this study is limited by a small patient population at a single institution and by the retrospective nature of the analysis. PVS, although more common in children than adults, remains a relatively rare complication of pediatric liver transplantation, and it is thus difficult to robustly study at a single children’s hospital. The small patient population limits the strength of our threshold (180 cm/s), which has a 17% false-negative rate and a 29% false-positive rate. As more patients are studied, this threshold will continue to be refined to improve positive and negative predictive values of PAV on DUS. At the time of this writing, after instituting this threshold in practice, we have subsequently performed 3 PTPVs on 3 patients with PAV > 180 cm/s. All 3 had >50% PVS. Similarly, although the age at the time of transplant was not statistically significant in our study, the differences between the 2 groups (group 1, mean 5 24 months, standard deviation 5 22 months, median 5 18.9 months, range 5 4.3-72.7 months; group 2, mean 5 45 months, standard deviation 5 36 months, median 5 37.1 months, range 5 7.2-91.5 months)
552 HAWKINS ET AL.
suggest that the age at the time of the transplant may eventually prove to be a statistically significant predictor for the development of PVS as the study population increases. Second, the diagnosis of “significant” PVS remains elusive. Although our criterion (>50% on PTPV) has been used by other investigators, there are no available data to suggest that a smaller degree of stenosis is, in fact, not significant in this patient population with widely variable clinical conditions. Third, PAV may be influenced by a variety of factors, in addition to the degree of anastomotic stenosis, such as bowel metabolism, Non per os (NPO) status, height, and weight.16 These variables were not reliably recorded at the time of DUS. Finally, some measurements were inconsistently obtained during ultrasound and PTPV examinations. For example, the portal vein diameter was not measured on 4 ultrasounds before PTPV, and portal vein gradients (across the stenosis) were measured in only 10 of 20 PTPVs. Although not reliably measuring portal vein pressure gradients during PTPVs may be perceived as a limitation, prior studies have suggested that this parameter can be misleading when it is used to diagnose PVS.1,15,17 Varices and other portosystemic collaterals can spuriously decrease pressure gradients across stenoses even in the setting of significant PV narrowing. Additionally, other investigations have reported low portal vein gradients in patients with significant PVS15 and in patients with low gradients who have improved after portal vein stent placement.17 In contradistinction, Cho et al.18 showed a decrease in the mean portal vein anastomosis pressure gradient after either angioplasty or stent placement from 8.75 to 1.75 mm Hg in a study evaluating 4 pediatric patients with late-onset PVS who were successfully treated. In our study, the average gradient in patients with >50% PVS was 7.5 6 3.9 mm Hg. In patients with 50% PVS, the mean gradient was 3.9 6 4.8 mm Hg (P 5 0.17). This trend toward significance lends support to using a pressure gradient > 5 mm Hg as an additional definition of significant PVS, which is a criterion that has been used by prior authors.14 Pressure gradients were measured during only 3 of 13 PTPVs in patients with >50% PVS. This is likely because pressure measurements were unlikely to change management (angioplasty) in the most severe cases of PVS. However, in patients with 50% PVS on PTPV, 6 of 7 patients had a pressure gradient
ORIGINAL ARTICLE
Pediatric Liver Transplant Portal Vein Anastomotic Stenosis: Correlation Between Ultrasound and Transhepatic Portal Venography C. Matthew Hawkins,1 Dennis W. W. Shaw,1 Patrick J. Healey,2 Simon P. Horslen,3 Andre A. S. Dick,2 Seth Friedman,1 and Giridhar M. Shivaram1 Divisions of 1Interventional Radiology and 2Pediatric Transplant Surgery and 3Liver and Intestinal Transplantation, Division of Gastroenterology and Hepatology, Seattle Children’s Hospital, University of Washington, Seattle, WA
The objective of this study was to determine which transabdominal ultrasound parameters correlate with portal vein stenosis (PVS) on percutaneous transhepatic portal venography in pediatric liver transplant patients. A retrospective review was performed of percutaneous transhepatic portal venograms performed between 2005 and 2013. The findings were compared to those from ultrasounds performed before venography and at the baseline. Patients were stratified on the basis of the presence of significant PVS (group 1, >50% stenosis; group 2, 50% stenosis) on portal venography. Findings were compared to those for age-matched controls. Twenty portal venograms were performed for 12 pediatric patients. Thirteen of the 20 patients (65%) demonstrated significant PVS (>50%). The mean peak anastomotic velocity (PAV) was 253.6 6 96 cm/s in group 1, 169.7 6 48 cm/s in group 2, and 51.3 6 20 cm/s in the control group. PAV (r 5 0.672, P 5 0.002) was the only ultrasound variable that correlated with the presence of significant PVS. A receiver operating characteristic curve was generated from PAV and PVS data (area under the curve 5 0.75, P 5 0.08). A threshold velocity of 180 cm/s led to a sensitivity of 83% and a specificity of 71% in predicting significant PVS on portal venography. At the baseline, the mean PAV was 155.8 6 90 cm/s for group 1 and 69.5 6 33 cm/s for group 2 (P 5 0.08); for control subjects, it was 78.9 6 53 cm/s (P 5 0.06). PAV is the only measured ultrasound parameter that correlates with significant PVS on portal venography in pediatric liver transplant patients. An elevated baseline PAV may increase the risk of developing PVS. Liver Transpl C 2015 AASLD. 21:547-553, 2015. V Received September 11, 2014; accepted January 2, 2015. Portal vein stenosis (PVS) is a relatively uncommon complication associated with liver transplantation.1-4 However, the incidence of PVS is more common in pediatric patients and has been reported to occur in
as many as 27% of segmental pediatric liver transplants.5-8 Gray-scale ultrasound and Doppler ultrasound (DUS) have become the primary imaging screening
Abbreviations: DAV, change in velocity across the portal vein anastomosis; DUS, Doppler ultrasound; PAV, peak anastomotic velocity; PTPV, percutaneous transhepatic portal venography; PVS, portal vein stenosis; ROC, receiver operating characteristic. Potential conflict of interest: Nothing to report. No grants were used to fund this study. All authors meaningfully contributed to the article and meet the criteria for authorship. C. Matthew Hawkins is currently affiliated with the Department of Radiology and Imaging Sciences, Emory University Hospital, Atlanta, GA Address reprint requests to C. Matthew Hawkins, MD, Department of Radiology and Imaging Sciences, Emory University Hospital, 1364 Clifton Road NE, Suite D112, Atlanta, GA 30322. Telephone: 517-927-6839; FAX: 404-712-7387; E-mail: [email protected] DOI 10.1002/lt.24077 View this article online at wileyonlinelibrary.com. LIVER TRANSPLANTATION.DOI 10.1002/lt. Published on behalf of the American Association for the Study of Liver Diseases
C 2015 American Association for the Study of Liver Diseases. V
548 HAWKINS ET AL.
Figure 1. Digitally subtracted portal venogram for a 6-year-old male liver transplant recipient with PVS (arrow). A 4F diagnostic catheter (Cook Group Inc., Bloomington, IN), passed through a 6F transhepatic sheath (Terumo Corporation, Tokyo), is seen crossing the stenosis and terminating in the superior mesenteric vein. The degree of stenosis is measured by the division of the difference between the maximum diameter of the recipient portal vein and the minimal diameter at the portal vein anastomosis by the maximum diameter of the recipient portal vein.
tools for evaluating the vascular complications associated with liver transplants.9,10 However, the sensitivity and specificity of sonographic parameters in predicting PVS in pediatric patients have not been delineated. Previous studies in adults have suggested that various portal vein velocities (as measured by DUS) can offer reliable predictive capabilities.1,11 However, liver transplantation in children differs in many ways from liver transplantation in adults. Most notably, children receive a larger number of segmental liver transplants rather than whole liver transplants, which are more conventional in adults. Also, the portal vein size mismatch, which is commonly associated with the transplantation of segmental liver transplants in children, may cause additional vascular complications that are not typically encountered in most adult liver transplant patients.8,12,13 The objective of this study was to determine which transabdominal gray-scale ultrasound and DUS parameters correlate with significant PVS in pediatric liver transplants as determined by percutaneous transhepatic portal venography (PTPV). Additionally, we aimed to evaluate whether the peak anastomotic velocity (PAV) on baseline DUS and the change in velocity from baseline DUS over time were associated with the development of PVS.
PATIENTS AND METHODS After institutional review board approval, a retrospective electronic medical record and imaging archive review was performed to identify all PTPVs performed between 2005 and 2013 at a single tertiary-care pediatric hospital. Multiple patient parameters were evaluated, and they included the following: etiology of
LIVER TRANSPLANTATION, April 2015
primary liver disease, type of transplant received, age at transplant, age at PTPV, baseline DUS data (defined as nadir in PAV from available ultrasounds obtained during the first week after transplant), data from DUS within 1 month before PTPV, and PTPV findings and interventions. Imaging and medical charts were reviewed by 2 radiologists. Indications for PTPV included elevated transaminases/alkaline phosphatase, concerning findings from DUS, and/or clinical concerns for portal hypertension. PTPVs were stratified into 2 groups on the basis of the presence (group 1, n 5 13) or absence (group 2, n 5 7) of significant PVS (defined as >50% on PTPV). This parameter has been used to define significant PVS by other investigators.1,14 We measured PVS by dividing the difference between the maximum diameter of the recipient portal vein and the minimal diameter at the portal vein anastomosis by the maximum diameter of the recipient portal vein (Fig. 1). Pressure gradients across the stenosis were measured in only 10 of 20 PTPVs, and this consequently limited the analysis of this variable. Cross-sectional imaging (magnetic resonance imaging and/or computed tomography) was performed immediately before only 5 PTPVs, the findings of which were not analyzed. Each PTPV was compared to the most recent DUS (obtained within 1 month before PTPV). All DUS examinations were performed by ultrasound technologists. The liver transplant DUS protocol for portal vein imaging at our institution entails velocity and waveform measurement of the native portal vein, the portal vein anastomosis, and the donor portal vein. Additionally, the right and left branches of the portal vein (for whole liver transplants) and the left portal vein (for segmental transplants) are imaged to document the direction of flow and patency. The protocol also includes multiple planes of Doppler imaging of the hepatic artery, celiac axis, splenic artery, hepatic veins, inferior vena cava, superior mesenteric vein, and biliary ducts. Velocities and waveforms are obtained for each vessel. Although equipment has changed over time, we currently use Philips iU22 and EPIQ ultrasound machines (Philips Healthcare, Amsterdam, Netherlands) with 8- to 5-MHz curved array probes for infants and with 5- to 2-MHz curved array probes for older children. For patients who have undergone liver transplantation, we perform DUS examinations on postoperative days 1 and 5. Thereafter, annual surveillance ultrasounds are not performed unless clinical suspicion or abnormal laboratory tests warrant an unscheduled DUS examination. On DUS, PAV was defined as the maximum measured velocity at the portal vein anastomosis in the absence of arterial flow artifact (Fig. 2). Maximum prestenotic and poststenotic velocities were also recorded. Velocity data were not recorded on 1 pre-PTPV DUS because the portal vein was interpreted to be occluded at the time of the ultrasound. (This patient was found to have 65% PVS on PTPV.) The PVS percentage was not measured for 4 (3 in group 1 and 1 in group 2) of the 20 pre-PTPV
LIVER TRANSPLANTATION, Vol. 21, No. 4, 2015
HAWKINS ET AL. 549
Figure 2. DUS of a 4-year-old female liver transplant recipient. The maximum portal vein velocity, without arterial flow artifact, was recorded as the PAV.
ultrasounds. Baseline DUS data were available for 6 patients in group 1, 5 patients in group 2, and 9 control subjects. Data from each group were also compared to a control group (n 5 10) of randomly selected, age-matched liver transplant patients with no prior interventions, normal liver function tests, no clinical concern for portal hypertension, and gray-scale ultrasound and DUS findings that were interpreted as normal. Data were transferred to SPSS 19 (IBM, Chicago, IL) for parametric and nonparametric analysis. Correlations were evaluated through the calculation of Pearson product-moment correlation coefficients. To evaluate the predictive potential of PAV on DUS, a receiver operating characteristic (ROC) curve was generated. Additionally, mean values between groups, at the time of PTPV and at the baseline, were compared with a 2-tailed t test. P < 0.05 was considered statistically significant.
RESULTS Patient demographics for group 1 (>50% stenosis), group 2 (50% stenosis), and control subjects are detailed in Table 1. During the study period, 179 liver transplants were performed in 157 patients at our institution. In sum, 20 PTPVs were performed in 12 of these patients (7.6%). Two patients had 2 PTPVs. Three patients had 3 PTPVs. Thirteen PTPVs showed >50% PVS, and 7 PTPVs showed 50% PVS. Interventions were performed during 17 PTPVs (15 venoplasties and 2 stents). The mean age at the time of transplant was not significantly different between the 2 groups (group 1, 24 6 22 months; group 2, 45 6 36 months; P 5 0.22). The median age (and range) at the time of transplant was 18.9 months (range 5 4.3-72.7 months) in group 1 and 37.1 months (range 5 7.2-
91.5 months) in group 2. The mean PAV before PTPV was 253.6 6 96 cm/s in group 1, 169.7 6 48 cm/s in group 2 (P 5 0.04), and 51.3 6 20 cm/s for the agematched-control subjects (F 5 24.52, P < 0.001; group 1 > controls, P 5 0.001; group 2 > controls, P < 0.001; group 1 > group 2, P 5 0.04). Additional pre-PTPV DUS parameters, as detailed in Table 2, were not significantly different between group 1 and group 2 (all F < 0.924, P 5 not significant). An analysis of ultrasound findings and each PTPV revealed that PAV (r 5 0.672, P 5 0.002) was the only DUS variable that correlated with the presence of >50% PVS (Fig. 3). Specifically, the change in velocity across the portal vein anastomosis (DAV; r 5 0.13, P 5 0.59) and the diameter of the stenosis as measured by DUS (r 5 –0.19, P 5 0.51) showed a poor correlation. The ROC curve generated by data from the PAV correlation analysis (Fig. 4) showed an area under the curve of 0.75 (trend level P 5 0.08). Notably, there was complete nonoverlap of PAVs in patients subjected to PTPV with PAVs measured in control subjects. When we evaluated the ability of PAV on DUS to predict the presence of >50% PVS on PTPV, PAV < 100 cm/s showed 100% sensitivity, and PAV > 267 cm/s was 100% specific. Applying a PAV threshold of 180 cm/s led to 83% sensitivity and 71% specificity for the detection of significant PVS; this was the best combination of sensitivity and specificity for a given DUS PAV according to our data. Pressure gradients were measured during only 3 of 13 PTPVs in group 1 (>50% PVS) and in all 7 PTPVs in group 2 (50% stenosis). The mean pressure gradient across the area of stenosis was 7.5 6 3.9 mm Hg in group 1 and 3.9 6 4.8 mm Hg in group 2 (P 5 0.17). A single patient in group 2 had a pressure gradient greater than 5 mm Hg (15 mm Hg). Angioplasty was performed for this patient despite PVS on PTPV 50%
550 HAWKINS ET AL.
LIVER TRANSPLANTATION, April 2015
TABLE 1. Patient Demographics
Number of patients Number of PTPVs Mean age at transplant (months) Indication for transplant Biliary atresia Acute liver failure Autoimmune hepatitis Neonatal hemochromatosis Idiopathic cirrhosis Cryptogenic cirrhosis Progressive familial intrahepatic cholestasis Carbamoyl phosphatase synthetase deficiency Transplant type Mean age at PTPV (months) Interventions performed during PTPV
Group 1 (>50% PVS)
Group 2 (50% PVS)
Control Group
7 13 24
5 7 45
10 — 38
5 1 1 — — — —
3 2 — — — — —
3 2 — 1 1 1 1
—
—
1
4/7 segmental (57%) 51 11 angioplasty 2 angioplasty/stent
4/5 segmental (80%) 80 4 angioplasty
6/10 segmental (60%) — —
TABLE 2. DUS Data
Pre-PTPV DUS Mean PAV (cm/s) Mean pre-anastomotic velocity (cm/s) Mean postanastomotic velocity (cm/s) Mean DAV (cm/s) Mean % PVS Baseline DUS Mean PAV (cm/s) Mean D in PAV from baseline to time of PTPV (cm/s)
Group 1 (>50%
Group 2 (50%
PVS on PTPV)
PVS on PTPV)
Control Group
253.6 6 96 (n 5 12) 39.5 6 16 (n 5 12)
169.7 6 48 (n 5 7) 37.4 6 11 (n 5 7)
51.3 6 20 (n 5 10) 47.3 6 21 (n 5 10)
119.8 6 58 (n 5 12)
103.2 6 32 (n 5 7)
46.9 6 22 (n 5 10)
80.3 6 62 (n 5 12) 71.8 6 15 (n 5 10)
65.9 6 35 (n 5 7) 74.5 6 6 (n 5 6)
11.2 6 20 (n 5 10) —
155.2 6 90 (n 5 6) 93.2 6 106 (n 5 6)
69.5 6 33 (n 5 5) 113.9 6 40 (n 5 5)
78.9 6 53 (n 5 9) —
NOTE: Pre-PTPV mean PAV: F 5 24.52, P < 0.001; group 1 > controls, P 5 0.001; group 2 > controls, P < 0.001; and group 1 > group 2, P 5 0.04. All other pre-PTPV DUS parameters were not significantly different between group 1 and group 2 (all F < 0.924, P 5 not significant).
(45%). A single patient in group 1 had a pressure gradient less than 5 mm Hg (4 mm Hg). Angioplasty was performed because this patient had 66% PVS on PTPV. On baseline DUS, patients in group 1 had a mean PAV of 155.2 6 90 cm/s. This was higher than the baseline mean PAV seen in group 2 (69.5 6 33 cm/s, P 5 0.08) and in control subjects (78.9 6 53 cm/s, P 5 0.056). The change in PAV from the baseline to the time of intervention was 93.2 6 107 cm/s in group 1 and 114 6 40 cm/s in group 2 (P 5 0.69).
DISCUSSION The objective of this retrospective analysis was to determine which gray-scale ultrasound and DUS
parameters correlate with PVS detected by percutaneous portal venography in pediatric liver transplant recipients. Additionally, baseline ultrasounds were evaluated for variables correlating with the development of PVS in this patient population. In our analysis, we demonstrate that PAV is the only DUS parameter that correlates (r 5 0.672) with PVS on PTPV, which is defined as >50% PVS. Specifically, the DAV (r 5 0.13) and the diameter of stenosis measured by DUS (r 5 0.19) correlated poorly. These data also produced various PAV thresholds that could predict the likelihood of PVS in pediatric liver transplant patients. Specifically, PAV < 100 cm/s and >267 cm/ s had 100% sensitivity and specificity, respectively. A PAV threshold of 180 cm/s led to 83% sensitivity and
LIVER TRANSPLANTATION, Vol. 21, No. 4, 2015
Figure 3.
PVS versus PAV measured on DUS.
Figure 4. ROC curve generated by data from a correlation analysis of PAV on DUS and PVS on PTPV (area under the curve 5 0.75, P 5 0.076). There was complete nonoverlap of PAVs in patients subjected to PTPV with PAVs measured in control subjects.
71% specificity in the detection of PVS. Furthermore, an analysis of baseline ultrasounds revealed that patients who developed PVS had a higher mean baseline PAV (155.2 cm/s) than the patients without PVS (69.5 cm/s, P 5 0.08) and the control subjects (78.9 cm/s, P 5 0.06), although these differences were not statistically significant. Similar investigations have been performed in adult patients. Mullan et al.1 demonstrated that using a portal vein velocity threshold of 80 cm/s on DUS yielded a sensitivity of 100% and a specificity of 84% in its ability to predict significant PVS. Chong et al.11 found that a portal vein PAV of 125 cm/s was 95% sensitive and 73% specific for PVS. Finally, Lee et al.15 demonstrated PAVs of 110 to 300 cm/s in 11
HAWKINS ET AL. 551
cases of confirmed PVS in pediatric transplant patients. (Lee et al. defined PVS as a diameter < 2.5 mm on PTPV, an acceleration of flow at the stricture, a poststenotic jet, and/or a portal vein gradient >5 mm Hg. The degree of PVS was not correlated with PAV or DAV.) Our threshold velocity of 180 cm/s is higher than those identified in the prior studies by Mullan et al. and Chong et al., but it is within the range of velocities identified by Lee et al. This is most likely related to variable portal vein flow dynamics in pediatric patients after liver transplantation, which is largely attributable to the surgical size mismatch resulting from the larger size of the donor portal veins.13 Consequently, even mild PVS in children may produce greater increases in PAV than those seen in adults. Two patients in our study had PAV > 180 cm/s with PVS on PTPV 50%. One patient had PAV of 259 cm/s on DUS. PTPV revealed 45% PVS and a pressure gradient of 15 mm Hg across the stenosis. Angioplasty was performed successfully in this patient. An additional patient had PAV of 203 cm/s on DUS. PTPV revealed 39% PVS with a pressure gradient of 2 mm Hg across the stenosis. No intervention was performed. Additionally, portal vein velocities in the pediatric age range (0-18 years) may have a wider dynamic range and be influenced (to a greater degree) by other anatomic abnormalities. For example, 1 patient in our study had 66% PVS despite having a lower PAV (101 cm/s). However, this patient also had hepatic venous outflow stenosis, which likely contributed to decreased portal vein velocity. Another patient had PVS of 64% with a PAV of 112 cm/s in the setting of marked splenomegaly, which may have affected portomesenteric flow. These results suggest that cross-sectional imaging may be beneficial in patients with PAV < 180 cm/s when there is a high clinical suspicion for posttransplant vascular complications. There are a number of limitations to this study. First, this study is limited by a small patient population at a single institution and by the retrospective nature of the analysis. PVS, although more common in children than adults, remains a relatively rare complication of pediatric liver transplantation, and it is thus difficult to robustly study at a single children’s hospital. The small patient population limits the strength of our threshold (180 cm/s), which has a 17% false-negative rate and a 29% false-positive rate. As more patients are studied, this threshold will continue to be refined to improve positive and negative predictive values of PAV on DUS. At the time of this writing, after instituting this threshold in practice, we have subsequently performed 3 PTPVs on 3 patients with PAV > 180 cm/s. All 3 had >50% PVS. Similarly, although the age at the time of transplant was not statistically significant in our study, the differences between the 2 groups (group 1, mean 5 24 months, standard deviation 5 22 months, median 5 18.9 months, range 5 4.3-72.7 months; group 2, mean 5 45 months, standard deviation 5 36 months, median 5 37.1 months, range 5 7.2-91.5 months)
552 HAWKINS ET AL.
suggest that the age at the time of the transplant may eventually prove to be a statistically significant predictor for the development of PVS as the study population increases. Second, the diagnosis of “significant” PVS remains elusive. Although our criterion (>50% on PTPV) has been used by other investigators, there are no available data to suggest that a smaller degree of stenosis is, in fact, not significant in this patient population with widely variable clinical conditions. Third, PAV may be influenced by a variety of factors, in addition to the degree of anastomotic stenosis, such as bowel metabolism, Non per os (NPO) status, height, and weight.16 These variables were not reliably recorded at the time of DUS. Finally, some measurements were inconsistently obtained during ultrasound and PTPV examinations. For example, the portal vein diameter was not measured on 4 ultrasounds before PTPV, and portal vein gradients (across the stenosis) were measured in only 10 of 20 PTPVs. Although not reliably measuring portal vein pressure gradients during PTPVs may be perceived as a limitation, prior studies have suggested that this parameter can be misleading when it is used to diagnose PVS.1,15,17 Varices and other portosystemic collaterals can spuriously decrease pressure gradients across stenoses even in the setting of significant PV narrowing. Additionally, other investigations have reported low portal vein gradients in patients with significant PVS15 and in patients with low gradients who have improved after portal vein stent placement.17 In contradistinction, Cho et al.18 showed a decrease in the mean portal vein anastomosis pressure gradient after either angioplasty or stent placement from 8.75 to 1.75 mm Hg in a study evaluating 4 pediatric patients with late-onset PVS who were successfully treated. In our study, the average gradient in patients with >50% PVS was 7.5 6 3.9 mm Hg. In patients with 50% PVS, the mean gradient was 3.9 6 4.8 mm Hg (P 5 0.17). This trend toward significance lends support to using a pressure gradient > 5 mm Hg as an additional definition of significant PVS, which is a criterion that has been used by prior authors.14 Pressure gradients were measured during only 3 of 13 PTPVs in patients with >50% PVS. This is likely because pressure measurements were unlikely to change management (angioplasty) in the most severe cases of PVS. However, in patients with 50% PVS on PTPV, 6 of 7 patients had a pressure gradient
