We performed an ethics-approved retrospective imaging and chart review at two quaternary pediatric liver transplant referral centers in two countries (hereafter referred to as site M and site E). The requirement for informed consent was waived. We included all children from birth to 16 years old who underwent liver transplantation for any indication at either site from 2001 to 2011 and who had available postoperative Doppler ultrasound images and reports, operative reports, and clinical follow-up. Patients were excluded if there were no ultrasound images from postoperative day 0, 1, or 2 or if clinical data were insufficient to assign a clear outcome. We anonymously pooled data from these two sites for analysis to increase sample size and widen the applicability of results. This resulted in 110 patients: 44 from site M, and 66 from site E. Fifty-nine (54%) of these patients were girls. The median age at transplantation was 1.3 years (site M, 2.3 years; site E, 0.9 years; p = 0.004 by Mann-Whitney U test), with youngest patient 25 days old and oldest 14 years old. The average length of follow-up was 3.5 years with a range of 0.1–9.4 years and less than 1 year of follow-up in 93 of 110 patients. The most common indication for liver transplantation was biliary atresia (57/110 patients). Eighteen patients had metabolic disorders, most commonly urea cycle defects. Ten patients had hepatic failure of uncertain cause; eight had nonneoplastic acquired diseases, such as hepatitis or primary sclerosing cholangitis; seven had familial hypercholesterolemia; five had unresectable hepatic tumors, specifically hepatoblastoma in three patients, fibrolamellar hepatocellular carcinoma in one, and a large adenoma in one; and five had liver failure associated with intestinal atresias or gastroschisis. Ultimately, 37 of 110 (34%) patients required a revision, laparotomy, or second transplantation, 20 of which were for vascular complications, including hepatic artery thrombosis (n = 7), portal vein thrombosis (n = 6), later portal vein stenosis (n = 3), hepatic venous stenosis (n = 3), and inferior vena cava (IVC) thrombosis (n = 1). There was also one portal vein thrombosis that was recognized at intraoperative ultrasound and treated with embolectomy during the transplantation surgery, showing normal portal venous flow on the postoperative day 1 ultrasound included in our study. Five patients died after transplantation in our follow-up period, with three of these deaths attributed to vascular complications (hepatic artery thrombosis, hepatic artery stenosis, and portal vein thrombosis). An endpoint of death or revision surgery was thus reached for vascular complications in 24 of 110 (22%) patients. Rejection was diagnosed and treated within 1 year of transplantation in 13 of 110 (12%) patients.

We focused our imaging review on the first Doppler ultrasound performed after the patient left the operating room from liver transplantation. This was generally on postoperative day 1 (POD1), with six of 110 done the same day (POD0) and another is done on POD2. Patients at both centers were scanned with available ultrasound units (mostly, iU22 units, Philips Healthcare) with probes including L12–5, C9–3, and C8–5 MHz. In accordance with our usual clinical practice, all examinations were initially performed by technologists and checked by radiologists, with additional focused scanning performed by the interpreting radiologist when required.

We evaluated the ultrasound Doppler flow parameters of vessels, including common hepatic artery, intrahepatic arteries, hepatic veins, main and lobar portal veins, and nearby native vessels including the IVC and celiac trunk. Numerically, for arteries we recorded peak flow velocities and resistive indexes and for other vessels, the flow direction (antegrade or retrograde) and velocities. Qualitatively, we assessed waveforms for mono-versus polyphasicity, high-versus low-resistive patterns, and degree and presence of diastolic flow.

The static images saved to the hospital archive were reviewed in consensus by two pediatric radiologists with 6 and 10 years of imaging experience who were blinded to clinical outcomes until after image review. Because age and sex were embedded in several locations in the available images, it was impractical to blind the reviewers to this information. Images with flow parameters likely to be inaccurate for reasons such as inappropriate beam steering or choice of vessel were flagged during review and excluded from numeric analysis. Values of peak velocity and resistive index were recorded directly as measured by the sonographer or, if not measured at time of the examination, manually approximated in consensus from the saved images.

Clinical charts were reviewed by a pediatric radiologist at each institution. From the operative reports, we recorded the surgical indication, source of the transplant, intraoperative complications, and type of vascular anastomoses. From follow-up clinical charts, we recorded outcomes including the need for any surgical revisions, clinical diagnosis of transplant rejection, vascular complication (thrombosis; stenosis; aneurysm; or leak from a regional artery, portal vein, or vein), nonvascular complication (such as biliary stricture or leak, rejection, or sepsis), death, or clinically well patient. Perihepatic collections were not considered a complication unless surgical intervention was required (e.g., due to abscess formation or interference with organ function).

Basic descriptive statistics were obtained. Univariate analysis of variance (ANOVA) was performed comparing cohorts with and without complications. Bivariate Pearson correlations were assessed among Doppler ultrasound findings and between these findings, demographics including age, and clinical outcomes. Significance of differences between means was computed using Student t tests, Mann-Whitney U test, or Fisher exact test when appropriate.

The ranges of velocities and resistive indexes observed in all patients and those with either no complication or selected complications are shown in Table 1. Limiting this analysis to patients less than 4 years old provided similar results (not shown). The distribution of common hepatic artery velocities and resistive indexes is shown in Figure 1. The common hepatic artery velocity (mean ± SD, 103 ± 69 cm/s) varied widely but was significantly lower in patients with hepatic artery complications (50 vs 107 cm/s, p = 0.03, r = 0.21). High common hepatic artery velocities of greater than 200 cm/s were uncommon, and three of five patients with these velocities had a complication (venous thrombosis, pancreatitis); the other two were uncomplicated. The common hepatic artery resistive index averaged 0.74 ± 0.18 and was significantly lower in patients with hepatic artery complications (resistive index = 0.50 vs 075, p = 0.001, r = 0.33). The common hepatic artery resistive index was also lower in older patients (i.e., age at transplantation was significantly negatively correlated with common hepatic artery resistive index [r = −0.25, p < 0.01]). The distribution of normal resistive index was skewed toward high values (Fig. 1B), but at the upper extreme, six of 10 patients with a very high resistive index of greater than 0.94, representing nearly absent diastolic flow, had complications (two venous, four nonvascular).

The common hepatic artery anastomosis velocity was widely variable at 152 ± 106 cm/s, and unlike the common hepatic artery resistive index, it was significantly positively correlated to patient age in that it was higher in older patients (r = 0.34, p < 0.01). The intrahepatic artery velocity (71 ± 43 cm/s) dropped significantly from POD0 to POD2 from 101 to 42 cm/s (p = 0.046). The hepatic vein mean velocity (36 ± 29 cm/s) was significantly lower with any vascular complication (25 vs 40 cm/s, p = 0.015, r = 0.23).

The main portal vein velocity (67 ± 46 cm/s) was also significantly lower with any vascular complication (30 ± 44 cm/s, p = 0.037, r = 0.20). Our technologists had difficulty imaging the portal veins, which were often small and obliquely oriented, and we omitted velocity values from calculations in 12 of 110 patients because of images showing clearly incorrect beam steering. This error tended to underestimate portal vein velocities.

The 13 patients who developed rejection showed statistically significantly higher velocities than those without rejection, at intrahepatic arteries (99 ± 40 cm/s, p = 0.02), main portal vein (99 ± 59 cm/s, p = 0.006), and intrahepatic portal veins (56 ± 39 cm/s, p = 0.04). Although significant, these differences were small in magnitude.

Of 110 patients, 82 (75%) underwent a left-lobe transplantation and 28 (25%), a whole-liver transplantation. The left-lobe transplantation patients were significantly younger (mean ± standard error of the mean, 2.3 ± 0.3 years vs 4.7 ± 0.8 years; p < 0.001). Of the 28 patients older than 4 years, 14 (50%) underwent whole-liver transplantations versus 14 of 82 (17%) less than 4 years (p = 0.002). The left lobe transplantation patients had significantly higher common hepatic artery resistive indexes (0.77 ± 0.16 vs 0.63 ± 0.18, p < 0.001) and intrahepatic resistive index (0.71 ± 0.17 vs 0.56 ± 0.14, p < 0.001). This could relate to age because age at transplantation was significantly negatively correlated with hepatic artery resistive index (r = −0.25, p < 0.01); the younger the patient, the higher the resistive index. However, for common hepatic artery and hepatic artery resistive index, ANOVA showed that there was no signifi-cant relation between patient age and transplantation type (left lobe or whole liver). Other velocity parameters were not significantly different between the two types of transplantations at ANOVA.

The normal common hepatic artery typically showed intermediate-resistance arterial waveform after transplantation, with brisk up-stroke, crisp systolic peak, and antegrade diastolic flow (Fig. 2). A minor blunting of the systolic peak had no correlation with vascular complication. Intrahepatic artery waveforms showed a systolic “double peak” in 28% of patients (Fig. 3). There was a nonsignificant trend toward this double peak, which was more common in patients without than with complications (33% vs 16%, p = 0.11).

The normal postoperative main and intrahepatic portal veins showed antegrade flow, either homogeneous or pulsatile (Fig. 4). Seemingly, retrograde flow was present at times on images showing pulsatile flow. Dynamic imaging was not available in this retrospective study. These waveforms were also similar in patients who developed vascular complications, except that portal vein thrombosis showed absent portal vein flow. We had one patient in whom no portal venous flow could be found at POD1 ultrasound, but a repeat examination the next day showed flow and no complication was identified.

Normal hepatic venous flow was dynamic and polyphasic, especially in younger children (Fig. 5A). We observed a more monotonous monophasic flow pattern (Fig. 5B) in 14 of 99 patients in whom we had hepatic vein Doppler tracings. The monophasic pattern was significantly more common in older children (7/26 [27%] ≥ 4 years vs 7/73 [10%] < 4 years, p = 0.03). The monophasic pattern was also significantly more common in the few patients with a venous complication (2/3 vs 12/96, p = 0.04). Only one of 13 patients who developed transplant rejection had a monophasic hepatic venous waveform at POD1 ultrasound. There were no statistically significant correlations between the waveform patterns and nonvascular complications.

In this study, the pediatric transplanted liver was initially hyperdynamic with high velocities in all vessels on POD1. We found that normal peak common hepatic artery velocities not associated with vascular complications averaged ~105 cm/s, whereas patients with hepatic artery thrombosis had low common hepatic artery velocities averaging 50 cm/s. The peak velocities we observed were somewhat higher than reported in adults. A recent adult study found mean peak common hepatic artery velocities of 84 ± 40 cm/s in uncomplicated cases versus 46 ± 27 cm/s in patients with hepatic artery stenosis [8]. However, that study considered ultrasound at hospital discharge, later than POD1, and velocities tend to decrease as time passes postoperatively. In this retrospective study, it was difficult to reliably assess for tardus parvus waveform, but our 50 cm/s average common hepatic artery velocity in cases with arterial complications agrees well with the adult study [8]. Furthermore, as in adults [9], high common hepatic artery velocities of greater than 200 cm/s were associated with hepatic artery complications, but these high velocities were only seen in five of our patients (Fig. 2).

The resistive index is generally 0.5–0.7 in adults after liver transplantation [7], with a higher resistive index of up to 0.8 acceptable in the first 72 hours after liver transplantation because of postoperative factors, such as vaso-spasm and reperfusion edema [4]. In our pediatric population, an even higher resistive index was normal in the common hepatic artery on POD1, especially in very young patients or those with a left-lobe transplant. The incidence of complications was low for a resistive index of less than 0.95. We found that the intrahepatic resistive index was usually slightly lower than the common hepatic artery resistive index. At the other extreme, a low resistive index of less than 0.5 in either the common hepatic artery or intrahepatic artery was strongly associated with vascular complications. This 0.5 threshold is similar to the adult criterion [9].

In this study, a frequent apparently normal variation in the hepatic artery waveform was a systolic double peak (Fig. 3) seen in nearly one third of patients. This finding has been termed “pulsus bisferiens,” which is Latin for “beat twice,” representing two prominent systolic peaks with an interposed mid systolic retraction. In carotid Doppler ultrasound, this can be seen normally in low-resistance waveforms but is present in ~50% of patients with aortic valvular disease and is also associated with hypertrophic obstructive cardiomyopathy [10]. In our patients, pulsus bisferiens was not associated with vascular complications and, in fact, showed a nonsignificant trend toward being more common in uncomplicated transplantations. This likely reflects normal hemodynamics and is not a cause for concern. To our knowledge, this finding has not previously been specifically reported or evaluated in liver transplant recipients.

In this study, hepatic artery thrombosis was the most common vascular complication, occurring in seven (6%) of 110 children. This rate is lower than the 9–18% reported in other pediatric studies and closer to the 4–12% incidence of arterial thrombosis seen in adult liver transplant recipients (4–12%) [4, 7, 11]. Even though thrombosis is an acute complication most common in the perioperative period, our study of a single POD1 ultrasound examination likely underestimates the thrombosis rate by not capturing all patients who had thrombosis requiring revision intraoperatively or on the day of surgery.

We found that the velocity in the largest portal vein (portal vein main or left, depending on transplant type) averaged ~65 cm/s. This is comparable to the mean ~58 cm/s reported in adults [12]. Although the ranges of portal vein velocities were wide, a decrease of the main portal vein velocity to 30 cm/s was associated with vascular complications (whether arterial, venous, or portal venous). This is intuitive given that in the extreme case of portal vein thrombosis, no flow is seen. In our population, there were nine portal venous complications, including six of 110 (5.5%) portal vein thrombosis and three of 110 (2.7%) with later portal vein stenosis, similar to reported rates of 1–3% for portal vein thrombosis and 4% for portal vein stenosis [2, 4]. Portal veins are small, tortuous, and difficult to evaluate, but the lack of definite portal venous flow on careful inspection by the technologist and the radiologist was a reliable indicator of thrombosis, with only one false-positive imaging result in this study. In adults, the ratio of preanastomotic to anastomotic portal vein velocities has been used as an indicator of portal vein stenosis [12]; however, we found this anastomosis difficult to visualize reliably in our pediatric patients, limiting the usefulness of this ratio in practice.

In this study, the hepatic vein velocity normally averaged ~40 cm/s, with wide variability. A low hepatic vein velocity (average, 25 cm/s) was associated with vascular complications (whether arterial, venous, or portal venous). This nonspecific finding likely implies poor hepatic outflow. Of our patients, two of 110 eventually had hepatic venous stenoses requiring angioplasty, and one had an IVC thrombosis. IVC stenosis is thought to be more common in children than adults after liver transplantation, particularly in those with a partial left lobe transplantation or a size discrepancy between donor and recipient vessels [2]. In our population, because of the focus on immediate postoperative imaging, none of these narrowings were directly seen on POD1 ultrasound—but two of these three patients had monophasic hepatic vein flow. Young children normally show dynamic polyphasic hepatic venous flow, reflecting changes related to cardiac activity: forward flow during diastole and reversed flow during right atrial contraction [13].

Monotonous monophasic hepatic vein flow was often present in older children (≥ 4 years) in our patient population but was also significantly associated with venous complications and appears to be a concerning finding in a younger child. This corresponds well with the published finding that monophasic hepatic vein flow in children is sensitive but nonspecific for acute liver transplant graft rejection and is also associated with IVC thrombosis and other conditions, including compression by adjacent collection, Budd-Chiari syndrome, systemic vascular collapse, and renal insufficiency [14]. The difference in waveforms has been quantified by computer software assessment of the venous pulsatility index in adults [12]. The hepatic vein and IVC waveforms are likely more important than the flow velocity in assessment for complications.

Because of the small caliber of the pediatric intrahepatic vessels, there were several POD1 studies in which an intrahepatic vessel could not be identified. On subsequent ultrasound examinations and surgical revision, these nonvisualized vessels were confirmed to be thrombosed in all but one case, stressing the need to raise the possibility of a vascular thrombosis if a vessel is not visualized.

Flow velocity and resistive index can be lower than expected because of several factors, including poor beam steering, which effectively decreases the Doppler frequency and hence the measured flow velocity [15]. There is also an expected decrease in common hepatic artery resistive index with patient age and an expected global decrease in velocities and resistive indexes on POD2 and beyond. Whole-liver transplants had lower arterial resistive index than left-lobe transplants, but this difference might be due to the older age of patients receiving whole-liver transplants; our study did not show interaction between these factors but was not powered or designed to fully test these interactions.

In the 13 of 110 (12%) patients treated for acute or early chronic rejection in this study, velocities were slightly higher in the intrahepatic arteries and portal veins than in patients without complications. This suggests relatively hyperdynamic flow, but the magnitude of difference in velocities was small, limiting clinical utility of this observation. A change to damped or monophasic hepatic venous flow on serial ultrasound was previously shown to suggest rejection [3]. In our study, limited to the first postoperative ultrasound, monophasic hepatic venous flow did not correlate with rejection as expected given that it usually requires more than 1 day to develop.

Limitations of our study were primarily related to its retrospective nature. Although sonographers and radiologists endeavor to ensure that standard images are obtained for every patient, the number and quality of images available varied. For example, images of the IVC were not always included, likely due to technical difficulty in visualizing this deep vessel perpendicular to the ultrasound probe, especially with bandages narrowing the available sonographic windows. Several images were not included in data analysis due to technical concerns, such as inappropriate beam steering when imaging portal veins. We were also limited to reliance on the documented clinical outcomes in available medical records, with the risk that we may have missed some complications. However, both sites have regional electronic medical records covering catchment areas of several million people, and follow-up was readily available. The sample size, although substantial for a study of this type, was still limited, particularly with respect to subgroup analyses (e.g., three patients with hepatic vein or IVC complications). The patient populations were somewhat different between sites, with older patients on average at site M. However, the diversity of the study population is arguably a strength of the study rather than a limitation. Finally, the time of initial postoperative ultrasound varied, generally on POD1 but varying between late on POD0 and on POD2. This may affect some vascular flow patterns that can change rapidly in the recovery period. Future studies could review the serial evolution of normal changes in the first weeks postoperatively.

It should be noted that despite the statistically significant differences observed, we saw wide and overlapping SDs of all velocities, limiting the predictive ability of a single numeric value in any individual patient. Some complications do not become evident until serial ultrasound studies show changes in flow patterns, which is one reason that daily ultrasound studies in at least the few days after liver transplantation are recommended [1]. To determine whether a complication is likely, the physician interpreting the ultrasound study must consider the Doppler velocities and flow patterns in all relevant vessels together as well as considering the pattern of changes over any prior ultrasound studies. Our most consistent finding was that on the initial POD1 ultrasound the care team should be particularly aware of low velocities and resistive indexes, which are concerning for complications.