Value of preoperative computed tomography for meso-Rex bypass in children with extrahepatic portal vein obstruction

Background Extrahepatic portal vein obstruction (EHPVO) is the most important cause of hematemesis in children. Intrahepatic left portal vein and superior mesenteric vein anastomosis, also known as meso-Rex bypass (MRB), is becoming the gold standard treatment for EHPVO. We analyzed the value of preoperative computed tomography (CT) in determining whether MRB is feasible in children with EHPVO. Results We retrieved data on 76 children with EHPVO (50 male, 26 female; median age, 5.9 years) who underwent MRB (n = 68) or the Warren procedure (n = 8) from 2013 to 2019 and retrospectively analyzed their clinical and CT characteristics. The Rex recess was categorized into four subtypes (types 1–4) depending on its diameter in CT images. Of all 76 children, 7.9% had a history of umbilical catheterization and 1.3% had leukemia. Sixteen patients (20 lesions) had associated malformations. A total of 72.4% of Rex recesses could be measured by CT, and their mean diameter was 3.5 ± 1.8 mm (range 0.6–10.5 mm). A type 1, 2, 3, and 4 Rex recess was present in 9.2%, 53.9%, 11.8%, and 25.0% of patients, respectively. MRB could be performed in patients with types 1, 2, and 3, but those with type 4 required further evaluation. The sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy of CT were 100%, 83.8%, 42.1%, 100%, and 85.5%, respectively. Conclusions Among the four types of Rex recesses on CT angiography, types 1–3 allow for the performance of MRB.

• A type 4 Rex recess requires further examination before MRB.

Background
Extrahepatic portal vein obstruction (EHPVO) is defined as thrombosis of the extrahepatic portal vein (PV) with or without extension to the intrahepatic PVs [1]. It is the cause of portal hypertension in 70% of pediatric patients and the most common cause of upper gastrointestinal bleeding in children [2,3]. Underlying etiologies of PV thrombosis include sepsis, dehydration, intra-abdominal/ pelvic infection, omphalitis, umbilical vein catheterization, a hypercoagulable state, biliary atresia, and chronic liver disease [4]. In up to 50% of children and young adults with EHPVO, the underlying etiology of the PV thrombosis remains unknown [2,5]. Patients may present with splenomegaly, ascites, encephalopathy, or cardiopulmonary complications. The Rex recess is the remnant of the embryonic umbilical vein. It is the space between hepatic segments III and IV, where the intrahepatic left PV (LPV) is conveniently placed for mesentericoportal anastomosis to restore hepatopetal flow. Meso-Rex bypass (MRB), also known as the Rex shunt (Fig. 1a), is the definitive treatment for EHPVO [6]. This procedure restores physiological portal liver reperfusion via a venous autograft connection from the superior mesenteric vein (SMV) to the intrahepatic LPV. MRB is therefore the gold standard treatment in children with favorable anatomy.
Preoperative imaging is essential because it provides anatomical information for surgical planning and excludes diseases for which MRB is not recommended; it is especially useful to determine the patency and size of the SMV and Rex recess. Preoperative ultrasound with color Doppler has been utilized to examine the patency of the LPV and SMV for MRB [7,8]. However, this examination technique is difficult because of the low flow velocity and small vascular caliber in pediatric patients. Magnetic resonance imaging (MRI) and computed tomography (CT) have been described as effective modalities in the evaluation of children with EHPVO [9,10]. They can provide an anatomical road map of the splanchnic and portal venous anatomy and create three-dimensional reconstructions to display spatial relationships. MRI has higher contrast resolution, but the longer acquisition times make the sequences sensitive to motion. It is also difficult for children to breath-hold during MRI. CT has higher spatial resolution, and CT angiography has been described as an effective modality. Notably, however, Superina et al. [6] successfully performed MRB in patients whose LPV was not visible by preoperative imaging. Bertocchini et al. [11] reported that wedged hepatic vein portography (WHVP) was an effective tool for preoperative assessment of the Rex recess. WHVP involves the performance of venous puncture for catheter access to the liver and can provide information on Rex recess patency. However, WHVP is an invasive examination that requires general anesthesia and endotracheal intubation.
No previous CT study or published classification has addressed the accuracy of using the characteristics of the Rex recess compared with the findings of direct visualization to identify pediatric patients with EHPVO who are good candidates for MRB. Our goal was to elucidate the technique and value of preoperative CT in evaluating the feasibility of MRP in children with EHPVO. Further, we calculated the proportion of patients whose LPV was not visible by preoperative imaging but who successfully underwent MRB, and we proposed follow-up management for these patients.

Clinical data
A database of children aged ≤ 14 years (range 0.7-14 years) with EHPVO was reviewed to identify MRB candidates who had undergone preoperative CT imaging at our institution from 2013 to 2019 (Fig. 1b). The inclusion criteria were noncirrhotic prehepatic portal hypertension caused by extrahepatic PV obstruction, which was defined by classic symptoms of portal hypertension (esophageal varices, splenomegaly, and hypersplenism with or without hyperammonemia, coagulopathy, or ascites); no evidence of associated liver disease (normal liver function test results and normal appearance); an abnormal PV trunk on imaging; and cavernous venous collaterals at the porta hepatis. The exclusion criteria were hepatic and post-hepatic portal hypertension, cirrhosis, and a history of having undergone MRB. In total, 76 patients were finally included in the study.
The clinical data included the age at surgery, sex, disease course, symptoms, associated malformations, history of umbilical vein intubation, and treatment history. All procedures were performed by the senior surgeon (Z.W.) or under his direct supervision. The left internal jugular vein/left gastric (coronary) vein was used as the vein graft. The Rex recess was examined by direct intraoperative observation or direct intraoperative Rex recess angiography. If the Rex recess was aplastic, selective portosystemic shunt-distal splenorenal shunt (Warren procedure) was considered. The Warren procedure is a nonphysiological and nonpermanent [1,12] shunt surgery that can maintain some portopetal flow and avoid encephalopathy. A successful Rex operation is indicated by disappearance of the clinical symptoms, normalization of the laboratory indicators related to hypersplenism, and a smooth anastomotic opening on ultrasound examination 3 days postoperatively. All 76 children were followed up after surgery. Follow-up involved assessment for any recurrence of clinical symptoms, assessment of laboratory test results, and performance of regular ultrasound and CT examinations. Nine (9/76, 11.8%) patients had anastomotic stenosis. The follow-up time ranged from 7 days to 6.4 years (median, 1.7 years).

Examination methods
All 76 patients underwent plain and contrast-enhanced abdominal CT. Before the examinations, the children fasted for 3-4 h, and those aged ≤ 5 years were orally administered chloral hydrate (0.5 ml/kg) for sedation. A 64-slice spiral CT scanning protocol was used with the following settings: tube voltage, 120 kV; automatic current conditions; thickness and layer spacing, 0.8 mm; matrix, 512 × 512; and standard algorithm reconstruction image with 1-mm layer thickness. CT angiography was performed with 2 ml/kg of contrast medium (Ultravist 300) during the arterial phase, portal venous phase, and delayed phase at 23 s, 45-52 s, and 90 s, respectively. Image post-processing included multiplanar reformation, maximum intensity projection, shaded surface display, and volume rendering. The duration from CT to MRB ranged from 1 to 194 days (median, 33 days).

Image analysis
The CT images were reassessed by two radiologists with 6 and 10 years of experience in pediatric radiology, respectively, who were blinded to the surgery. The radiologists reached a consensus regarding the imaging results.
Gastric fundic varices, the left gastric (coronary) vein [13], and cavernous transformation of the PV were defined as common portocaval shunts. Other portocaval shunts were defined as rare shunts. We also evaluated other abdominal findings including gallstones, the biliary tree, and ascites.
The maximum diameters of the SMV, splenic vein (SV), spleen, and Rex recess were measured twice in the portal phase, and the mean was used for analysis. The spleen diameter was measured in the coronal plane, and all vascular diameters were measured in the axial plane. We retrospectively assessed the CT findings of children with EHPVO and herein propose a system for further classification of the Rex recess according to the CT features. The CT pattern of the Rex recess was categorized into four subtypes: • Type 1 (Fig. 2a): The diameter of the Rex recess is ≥ 5 mm. • Type 2 ( Fig. 2b): The diameter of the Rex recess is 2 to < 5 mm. • Type 3 (Fig. 2c, d): The Rex recess is faintly visible and < 2 mm in diameter, or the border with the side branches is unclear but segment III can be distinguished.
(see figure on next page.)

Fig. 2
Computed tomography classification of the Rex recess. a Type 1. The Rex recess is widened, and the diameter is ≥ 5 mm (13.8 mm in this figure). b Type 2. The Rex recess is clear and can be measured, and the diameter is ≥ 2 to < 5 mm (3.8 mm in this figure). c, d Type 3. The Rex recess is faintly visible and < 2 mm in diameter, or the border with the side branches is unclear, but the segment III branch can be distinguished by computed tomography. We calculated the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of CT to identify the feasibility of MRB compared with laparotomy.
A diagram of the pediatric portal hypertension procedure is shown in Fig. 2g.

Statistical analysis
The statistical analysis was performed using SPSS 26.0 (IBM Corp.). Statistical significance was achieved at a 0.05 level.
The categorical variables (sex, associated malformations, intraductal bile duct dilatation, common portocaval shunt, rare portocaval shunt, and ascites) were analyzed by Pearson's Chi-square test. The Kolmogorov-Smirnov normality test was used to assess continuous variables (age at surgery, disease course, and diameters of Rex recess, SMV, SV, and spleen). The continuous variables were not normally distributed, and the Kruskal-Wallis test was used.
Successful MRB was considered a true positive, and aplasia of the Rex recess was considered a true negative.
The sensitivity, specificity, PPV, NPV, and accuracy were calculated.
Next, we used binary logistic regression to analyze multiple variables (sex, age at surgery, disease course, and diameters of Rex recess, SMV, SV, and spleen) for their effect on predicting MRB surgery. with severe esophageal varices. The mean platelet count was 106 ± 60 × 10 9 (range 22-290 × 10 9 ; reference range 140-440 × 10 9 ). One patient underwent surgical repair of splenic rupture, two underwent the Warren procedure, and three underwent splenectomy before CT.

Patient
Sixteen patients (20 lesions) had associated malformations. Seven patients had common bile duct cysts (six postoperatively, one preoperatively), four had congenital heart disease, three had renal malformations (two horseshoe kidney, one polycystic kidney), two had intestinal malformations (hypertrophic pyloric obstruction, intestinal malrotation), two had spinal deformity, one had gallstones, and one had an inguinal hernia. Sixty-eight patients successfully underwent the Rex procedure, and eight underwent the Warren procedure when the surgeon found that the Rex recess was dysplastic.
The continuous variables were not normally distributed (Table 1). Children who underwent the Warren procedure had a higher proportion of malformations than children who underwent MRB (p = 0.006). There were no other significant differences between MRB and the Warren procedure.
No patients had an intraluminal PV thrombus. We did not measure the RPV because it was difficult to distinguish the RPV from the collateral veins.
A significant difference was found in the diameter of the SV (p = 0.018). Further two-two comparisons showed that the SV diameter was shorter in type 4 than  Table 3).
All patients with a type 1, 2, or 3 Rex recess successfully underwent MRB. Among 19 patients with type 4, 57.9% (11/19) were proven suitable to undergo MRB; the remaining 42.1% (8/19) were found to have Rex recess aplasia and underwent the Warren procedure. We assigned the patients with a type 1, 2, and 3 Rex recess into the MRB group and those with type 4 into the potential MRB group. The sensitivity, specificity, PPV, NPV, and diagnostic accuracy of CT were 100%, 83.8%, 42.1%, 100%, and 85.5%, respectively.

Impact of time and treatment on CT classification
Six patients (Table 4) underwent CT re-examination before MRB (mean, 2.8 years). As the disease progressed, the spleen diameter increased. Five patients had no change in their CT classification, and one of them underwent the Warren procedure (Fig. 5). One patient changed from type 3 to 4 with conservative treatment (Fig. 6).

Discussion
In the present study, EHPVO was the most important cause of portal hypertension, and the patients' age at the time of diagnosis and surgery was slightly older because of the long disease course. Portal hypertension was caused by EHPVO in 90.9% of children, by biliary atresia in 1.1%, and by cirrhosis in 8.0%. According to the literature, EHPVO is typically diagnosed between the ages of 2 and 4 years [7]; the median age in our study was 5.9 years. Because of the unbalanced level of medical development in China, not all regions can perform MRB for children, and the history of umbilical vein intubation was unclear in most children of our study. Notably, of seven children with combined common bile duct cysts, six patients' cysts developed within a few years after common bile duct surgery. However, the relationship between the history of common bile duct surgery and EHPVO remains unclear. All of these data suggest that the condition of the PV should be carefully evaluated at the time of the first diagnosis of a choledochal cyst and after completion of surgical treatment of the choledochal cyst. Additionally, the children with Rex dysplasia in our study had a higher incidence of combined malformations. However, we were unable to identify the specific types of deformities because of the small sample.
CT is an effective modality in the evaluation EHPVO [14,15] with the characteristic findings of cavernous transformation and portal hypertension. No thrombi were observed because most patients had a long history. Gastroesophageal varices and the left gastric (coronary) vein occurred in 97.4% of children, followed by cavernous transformation of the PV in 90.8%. This collateral was cavernous transformation of the extrahepatic PV that meandered into the hilar region and supplied the right liver, and it was difficult to distinguish from the trunk of the RPV. The rare portocaval shunts were splenorenal veins, gallbladder veins, and paravertebral veins. In the ideal presurgical candidate for MRB, both the intrahepatic LPV and the SMV will be visible, and the feasibility of completing the proximal and distal bypass attachments can be evaluated. The positive rate of the SMV is 100%. The Rex recess is an ideal location for placement of the shunt because it is rarely involved with cavernous transformations and collaterals [5].
Cárdenas et al. [16] reported that a Rex recess of ≥ 2 mm is the main indication for MRB and accounted for 63.2% of the cases. A total of 26.3% of children were able     In clinical practice, the Rex recess may be confused by collaterals running alongside it or a dilated left hepatic artery, especially in patients with a type 3 or 4 Rex recess. The 45-to 52-s portal venous phase may no longer be appropriate in children with EHPVO in our experience because of the poor display of the PVs in some of these children. We consider that the delayed phase may replace the portal venous phase in the future to reduce the amount of radiation.
In addition to the Rex recess and SMV, other relevant venous anatomical structures must be evaluated simultaneously. First, adequately sized splanchnic collateral vessels should be identified because they may be utilized for the meso-Rex graft instead of the internal jugular vein [11]. The left gastric (coronary) vein, inferior mesenteric vein, gastroepiploic vein, recanalized umbilical vein, and saphenous vein have been described as graft material [18][19][20]. Second, inspection of systemic veins, including the inferior vena cava and renal veins, is required. If an MRB candidate is found at surgery to have inadequate intrahepatic portal venous anatomy to support the graft, the next surgical option is typically the Warren procedure.

Limitations
The neonatal history of most children was unclear. Additionally, the relationship between the long-term surgical effect (clinical symptoms and/or anastomotic stenosis) Fig. 6 Male patient aged 2 years 6 months with a 6-month history of black stool. a The Rex recess was faintly visible and the diameter of the Rex recess (white arrow) was 1.8 mm (type 3). b Computed tomography angiography after 18 months. Complete loss of vascular landmarks of the Rex recess (white arrow), which was type 4. b3 Volume rendering showed an increase in cavernous transformation Table 4 Re-examination CT angiography before MRB in six patients  and type of Rex recess remains unknown. Finally, the study was retrospective in nature; we should use these diagnostic criteria for prospective studies.

Conclusion
CT is a reliable method for visualization of the Rex recess in children with EHPVO. We recommend four categories of the Rex recess based on CT angiography. The sensitivity, specificity, PPV, NPV, and diagnostic accuracy of CT in evaluating the success of MRB were 100%, 83.8%, 42.1%, 100%, and 85.5%, respectively. Among the four types of Rex recesses on CT angiography, types 1 to 3 allow for the performance of MRB. Patients with a type 4 Rex recess should undergo WHVP before MRB.