Portal biliopathy refers to abnormalities of the entire biliary tract, including intrahepatic and extrahepatic bile ducts, cystic duct, and gallbladder in patients with PHT [3]. On cholangiogram, dilated intrahepatic biliary radicles, smooth indentations or impressions on extrahepatic bile duct, ductal angulations, ectasias, strictures, stones, filling defects, and pericholecystic and/or pericholedochal varices are suggestive findings [4]. Patients with features of PB only on cholangiogram were considered asymptomatic, and along with corroborative symptoms, like jaundice and/or cholangitis, symptomatic. Mere presence of gall stones was not enough to establish PB.

Postoperative complications were classified according to the Clavien–Dindo system. Clavien–Dindo > 3 was considered major complication.

Ninety-day postoperative mortality refers to death within 90 d after surgery, as either an inpatient or outpatient [5].

Long-term outcome was classified according to the persistence of symptoms related to disease or any surgical or non-surgical intervention required. Patients with favourable outcome were supposed to have none, while those with unfavourable ones have either.

EHPVO patients with failed secondary prophylaxis for bleeding and patients symptomatic with PB, hypersplenism, splenomegaly, growth retardation, and biliary lithiasis were operated. Patients on secondary prophylaxis, if developed gastric varices (GV), were also considered for surgery. Patients were again subdivided into favourable and unfavourable long-term outcome group to look for risk association and outcome difference.

Operative indications were defined by a multidisciplinary team of gastroenterologist, radiologist, and GI surgeon (Fig. 2). US Doppler was the most commonly performed investigation. A contrast enhanced computed tomography (CECT) was often added to precisely delineate a shuntable vein, as being absent might offer little or no benefit, particularly in the cases of symptomatic PB (Fig. 3A). Magnetic resonance cholangiopancreatography (MRCP), not a routine, was performed in a few cases of suspected PB (Fig. 3B).

Apart from planned mesocaval shunt, all patients underwent laparotomy through left subcostal with midline extension. Lesser sac was opened, and splenic artery ligated in continuity. Highly vascular splenorenal ligament was divided using harmonic scalpel, bipolar electrocautery, and ligatures. Splenic hilum was then dissected, polar tributaries of splenic vein (SV) were ligated, and divided. Splenectomy was then completed, and SV separated from dorsal surface of pancreas for 4−6 cm. Left renal vein was then exposed for 4−5 cm. A proximal splenorenal shunt (PSRS) was performed using 6-0 or 7-0 polypropelene suture, ensuring maximum possible shunt diameter (Fig. 4A). If there was no shuntable vein, esophagogastric devascularization involving proximal half of stomach and distal (7−8 cm) of esophagus was contemplated. In emergency situation with ongoing bleeding, particularly from an isolated gastric varix (IGV), often required a gastrostomy with direct ligation of varices (Fig. 4B). Long midline incision was preferred for mesocaval shunt. For interposition graft, we used right internal jugular vein (Fig. 4C) or engorged right colic vein; in the later, we could avoid an anastomosis with superior mesenteric vein (SMV). Some patients with symptomatic cholelithiasis without any varices or biliopathy were safely treated by open cholecystectomy.

All patients were followed up 3 monthly for 2 years, and then 6 monthly for the next 3 years. An imaging, either US Doppler or CECT abdomen, was performed 6 months after surgery to look for shunt patency. An upper GI endoscopy was performed if clinically indicated, or in the presence of a shunt thrombosis. Endoscopic retrograde cholangiography (ERC) with stenting and/or hepaticojejunostomy was/were needed for PB patients with persistent symptoms.

Data was retrieved from our prospectively maintained GI surgery database. Age, sex, age at index presentation, duration of illness, index presentations, bleeding episodes, blood transfusions, number, compliance and effectiveness of different endoscopic manoeuvre with findings, clinical presentations, complete blood count, liver function test, and associated conditions related to PHT were noted as preoperative data. Operative data included main indications for surgery, type of surgery, shunt size, utilisation of shuntable vein, operative blood loss and transfusion requirement, and duration of surgery. Splanchnic and systemic venous blood pressure were not measured routinely. Postoperative data included, length of hospital stay, morbidity, and mortality. In follow-up, incidence of re-bleeding, shunt patency, non-surgical or surgical interventions if required and long-term outcome were noted.

Data entry was done in MS Excel sheet. All statistical analyses were carried out by SPSS version 11.0 (SPSS Inc.). All continuous variables were assessed for normality using Shapiro–Wilk’s test. If the data were normally distributed, they were expressed as the mean ± standard deviation, otherwise median (interquartile range). All the categorical variables were expressed either as percentage or proportion. Comparison of all the normally distributed continuous variables was done by either independent sample t-test or analysis of variance based on the groups. For the non-normally distributed variables, statistical comparisons were made by Mann–Whitney U test or Kruskal–Wallis H test. Comparisons of categorical variables were made by chi-square test or Fisher’s exact test, based on the number of observations. Binary logistic regression in Wald methodology was used to determine the predictive factors responsible for unfavourable outcome. All p-values < 0.05 were considered as statistically significant.

EHPVO is a childhood vascular disorder of the liver. There is chronic blockage of the extra-hepatic PV with or without the involvement of intra-hepatic PV radicles or splenic or SMV, resulting in portal cavernoma and varices. Isolated splanchnic venous or intra-hepatic PV thrombosis alone is not considered as EHPVO [6]. Liver function is mostly well preserved, unless compromised by reduced hepatic cell mass and synthetic dysfunction, as rarely seen with advanced age, prolonged duration of disease, and development of SBC from PB [7]. Neonatal exchange transfusion through umbilical catheterisation, omphalitis, congenital abnormalities, childhood sepsis, and prothrombotic disorders, like deficiencies in the naturally occurring anticoagulants (protein c, protein s, and antithrombin III) and factor V leiden and prothrombin gene mutations, are some of the suspected reasons [2].

In developing nations, EHPVO is a major cause of PHT (54%) and upper GI bleeding (68%−84%) in children [8,9]. In the West, EHPVO, though rare, is the second most frequent cause of PHT in adults [10], whereas in children, it is less common (11%) [11]. Patients with EHPVO mostly present with variceal bleeding in their 1st or 2nd decade of life [12]. In our series, mean age of onset was 12.88 ± 10.71 years, and age at the time of surgery was 20.30 ± 9.96 years, the difference roughly corroborating the onset of PB. Adults are often asymptomatic, or present with PB [13]. PB as a sequel of EHPVO usually affects people at an interval of 6−14 years in 80%−100% cases, but becomes symptomatic only in 5%−30% cases [4,14], as found in the current study (symptomatic PB, 28.2%). Hypersplenism, symptomatic splenomegaly, biliary lithiasis, and growth retardations are other presentations (Table 1) [15].

EST or EVL is treatment of choice in acute variceal haemorrhage, though nearly 5% of cases may require surgical intervention in acute setting following failed endo and pharmacotherapy [1]. Endoscopy has largely replaced surgery, the only available options until the mid-20th century, in managing acute bleeding and preventing recurrence [16]. Nearly 90% cases of acute variceal bleeding may be controlled by EST or EVL [6]. As secondary prophylactic measure, EST or EVL or their combinations may ensure variceal eradication in 80%−90% of patients [16]. Recurrence rate after EVL/EST or combination may be lower than 10% [17,18]. But endoscopic management does not deal with underlying PHT; that may increase the incidences of gastric and ectopic varices and portal gastropathy, which often bleed torrentially, and are difficult to manage endoscopically [19,20]. Moreover, the ideal situation is not always available, due to poor socioeconomic condition of the patient, remote place, unavailability of blood products, and poor access to modern endoscopy services; the recurrence may reach 40% [21].

Ultrasound Doppler of splenoportal axis (SPA) is the investigation of choice, and it can detect the presence of portal cavernoma replacing variable length of PV, SMV, and SV with sensitivity and specificity > 95% [13]. It also shows condition of the liver and the presence of biliary lithiasis. Ninety-nine percent of patients in our series have preoperative ultrasound Doppler. The status of a shuntable vein and precise anatomy of SPA, particularly in cases of PB, have considerable impact on surgical decision, and here, CT/MR portography plays an important role [22].

The main indications of surgery in EHPVO are acute variceal bleeding not controlled by endoscopy and secondary prophylaxis of GI bleeding, particularly with failed eradication. Other indications are PB, symptomatic hypersplenism, symptomatic splenomegaly, and growth retardation, as these are not reversed by endoscopy [15], whereas PSS may help [23]. In a study of 20 patients with GV undergoing PSRS, only 2 patients experienced rebleeding [24]. Complications of endoscopic procedures were some associated important issues in current surgical cohort, like post-EST oesophageal stricture (n = 5), post-EVL ulcers (n = 5), ERC papillotomy bleeding (n = 3), and haemobilia (n = 1) in some symptomatic PB patients, and GV rupture during glue injection (n = 1). Surgical options are PSS, like PSRS or mesocaval shunt or Meso-Rex shunt, and esophagogastric devascularization with or without esophagogastric transection. Surgery may be the preferred approach for secondary prophylaxis, as it does not require stringent follow-up as required in endoscopic management [15]. A PSS is preferred to devascularization procedure wherever feasible for lower re-bleeding rate [25,26]. In the present series, 94.3% patients (n = 166) with shuntable vein (n = 176) underwent PSS. PSRS had been our preferred shunt procedure, if possible. Four patients required mesocaval shunt. In Meso-Rex shunt, an autologous internal jugular or great saphenous vein interposition graft is used to bridge the SMV and intrahepatic left PV within the Rex recessus of the liver [27]. A few comparative studies between meso-Rex shunt and PSS revealed better outcome with the former, in terms of reversing coagulopathy and growth retardation [27-29]. But these studies mostly considered a distal splenorenal shunt as PSS procedure. The existing literature also did not compare PSRS with Rex shunt. Patients of meso-Rex shunt were also of a much younger age group at 2−12 years, compared to our cohort at 20.30 ± 9.96 years. A PSRS patient has to sacrifice spleen, but at the same time gets rid of symptomatic hypersplenism and splenomegaly [16]. Post-shunt encephalopathy is unlikely with meso-Rex shunt, as it does not divert liver from splanchnic circulation. Some large studies with PSRS have shown no post-shunt encephalopathy as well [30-33]. In a study with 31 patients with EHPVO, Superina et al. [34] found that meso-Rex shunt provided excellent relief symptoms along with liver growth and the normalization of coagulation parameters, and concluded that this surgery should be done at as early an age as possible. A Rex shunt was not performed in the current study, as we were not familiar with it, and patients were of older age group. In our series, patients having no shuntable vain or inadvertently injured ones, and those with unstable hemodynamic condition during surgery, underwent splenectomy and esophagogastric devascularization. Shunt size varied 4−40 mm. We have never encountered encephalopathy with larger shunts, due to well preserved hepatic function. Major postoperative complications and operative mortality are low with surgical candidates. Operative mortality was 0.5% in our series, comparable with the existing literature [35,36]. In those studies, emergency PSS had higher mortality (~10%). In emergency setting, we therefore preferred devascularization procedure to tedious PSS hemodynamically unstable patient.

In univariate analysis, unfavourable outcome was significantly associated with poor compliance to endotreatment, cholangitis, biliary lithiasis, associated SAA, PB, IGV, need for ERC, higher bilirubin and alkaline phosphatase level, higher operative blood loss, and shunt thrombosis. In multivariate analysis, associated SAA, IGV, need for ERC, and shunt patency showed significant difference in favourable and unfavourable outcome.

Probably in the presence of PHT, large natural PSS reduce SV resistance, thus triggering an increase of splenic arterial flow to increase SV flow and maintain portal pressure to perfuse liver [37]. This hyperkinetic splenic circulatory state may induce splenic arterial dilatation along with arterial medial layer degeneration, resulting in the formation of aneurysms in 7%−20% of patients with PHT [38,39].

Triphasic CT and Doppler ultrasound usually pick up SAA, but may miss small and leaking aneurysms [40]. Angiography, though a gold standard to localize and characterize the lesion, is used therapeutically before coil/glue embolization and stent exclusion [41,42]. In patients with PHT, associated SAA should be treated irrespective of size for higher incidence of spontaneous rupture. So, it is better for patients with EHPVO, when requiring surgery, to undergo simultaneous excision of the aneurysm, or ligation if multiple [43]. But a PSS procedure is often failed, due to tedious dissection, inability to get a good shuntable vein, and associated high per-operative bleeding [44]. We had 5 patients with SAA (2.5%) with average size of 4.5 cm, having shuntable veins in 4, PSRS being performed in 3, and follow-up shunt thrombosis in 2, further procedure required in 3, and 1 death. All of them underwent excision, and 3 of them, distal pancreatectomy. SAA probably reflects an advanced disease, difficult surgery, and therefore, high complications and worse outcome. Yet such patients should be offered surgery in an experienced centre with minimal morbidity and good outcome [45].

Natural history of IGV is poorly understood as to whether it may be an initial presentation or a later development, but it may be logical to predict its appearance in long-standing cases with multiple times endoscopic treatment of esophageal and gastroesophageal varices. IGV1s affect fundal region of stomach, are more common in EHPVO, and are likely to bleed much more than IGV2s, the gastric antral ectopic variety [46]. Gastric veins from IGV1 are predominantly anastomose with left inferior phrenic vein (IPV) in postero-superior bare area of stomach, from where they decompress in IVC through gastrorenal shunt, often along with left adrenal vein (80%−85%) or a transverse gastrocaval shunt (10%−15%), according to difference in drainage of IPV, either in left renal vein or IVC [47]. IGV2 involves the gastroepiploic and pancreaticoduodenal venous system [48]. Sarin et al. [49] found IGV1 in 39.4% patients of EHPVO, and 78% of them had significant bleeding. A study from GB Pant Hospital, India, found that the amount (3.69 ± 1.91 mL) and sessions for glue injection (2.11 ± 0.97, range [1−7]) to eradicate IGV in EHPVO are significantly higher than that of cirrhotics [46]. In the present study, 22 patients had IGV, of which 20 had IGV1. Control of bleeding often seemed difficult, and emergency surgery was required in 7 cases (31.8 %), significantly higher than patients without IGV (6.1%, p ≤ 0.001). For the same reason, the incidence of PSS was significantly less in patients with IGV than those without (59.1% vs. 85%, p = 0.003). In our understanding, prolonged disease state, greater need for emergency surgery, and less percentage of PSS in patients with IGV offered poorer outcome (Fig. 3D).

The need for ERC and stenting surrogates symptomatic PB. Irregular strictures and dilatations of the extrahepatic biliary tree in patients with portal cavernoma favour biliary lithiasis in long-standing cases, resulting in pain, jaundice, and intermittent cholangitis [4,50]. Despite EHPVO being a childhood disease, and bleeding being the most common first presentation, in our series, for 22 patients who required preoperative ERC and stenting, the average age was 32.6 years, and most common index presentation was PB (n = 9, 50%), with age at index presentation being 24.9 years. Average duration of illness in this ERC group was 92.1 months in the present study, but it should be considered much more, as PB was the major index presentation, and the onset of EHPVO is supposed to be a decade earlier than PB. Surgical candidates for PB may have poorer outcome if requiring ERC and stenting in the preoperative period, as the biliary strictures are initially due to vascular compression; but later on, ischaemia, chronic inflammation, and fibrosis predominate, rendering strictures more resistant to reverse after PSS [50]. Ten out of 22 patients in the current study had bile duct stones, 20 had jaundice, and 12 had cholangitis, and it clearly reflected a long-standing nature. In the present study, it seemed that rather than preoperative ERC and stenting resulting in worse outcome, the group of patients of EHPVO who required preoperative ERC and stenting were probably the worst kind of symptomatic PB patients, and therefore contributed significantly to unfavourable outcome. Twenty patients underwent PSS, and had postoperative major complications (Clavien–Dindo > 3) in 5 (25.0%), including 1 operative mortality. Three patients in this ERC group had shunt thrombosis, 4 required subsequent ERC, and 1, hepaticojejunostomy. Three patients died in the long run due to unresolved cholangitis, with 1 having developed cirrhosis. Though shunt surgery was possible in the majority of cases (20 out of 22, 90.9%) of PB who had prior ERC in our study, outcome was worse in this subset, as shunt surgery might fail to relieve biliary strictures, due to fibrotic nature in long-standing cases [51].

Shunt thrombosis is one of the important post PSS complications, seen in nearly 5.6%–47% of cases [36,52,53]. In early postoperative period, probable causes of shunt thrombosis are small diameter shunt, poor SV backflow, improper lie, pathological SV wall, and difficulty in anastomosis; whereas at late phase, procoagulant disorder and post-splenectomy thrombocytosis play important negative role [52,54-57]. According to Bismuth, to avoid thrombosis, the shunt size should be > 10 mm [58], though in paediatric group of patients, it is often not feasible, and shunt size > 4 mm may show a good patency rate of 84%−96% [30]. In the present series, shunt diameter varied 4−40 mm. We tried to make the shunt aperture as large as possible, as we have never encountered post-shunt hepatic encephalopathy with non-selective shunts in EHPVO, as evidenced in the literature [30,31]. Of 11 patients who had shunt thrombosis, 1 had mesocaval shunt, while the other 10 had PSRS. One patient with symptomatic PB developed biliary cirrhosis, ascites, and cholangitis after PSRS thrombosis, and died. The other 9 patients with shunt thrombosis were followed-up. On follow-up US doppler/ CECT, if shunt thrombosis was found, patients underwent endoscopy to look for varices. Out of 9 patients, 3 (33.3%) remained asymptomatic. Four patients developed non-bleeding varices. One had rebleeding. All varices (n = 5, 55.6%) required endoscopic surveillance for at least 2 years post-detection. ERC was required in 1 case of newly developed PB after shunt thrombosis. A successful PSS mostly offers a one-time solution for EHPVO patients. The single RCT by Wani et al. [59] from India showed comparable mortality and treatment failure with EST and shunt surgery, but less rebleeding rate and blood transfusions in later group. Chaudhary et al. [16] recommended shunt surgery for secondary prophylaxis of all EHPVO patients in resource-limited developing countries where this entity is common. EHPVO patients with good shuntable veins, having surgical expertise available, may preferably be considered for PSS, as it may reduce the development of gastric and ectopic varices, and worsening of PB [6,13,60].

This study has some strengths and limitations. The strength is that it is one of the largest case series of PSS for EHPVO with low mortality, acceptable postoperative complications, and excellent long-term functional outcomes (Table 6), as compared to some other relevant studies [30-33,36,61,62]. But there are limitations. The analysis was made retrospectively. It was a pure surgical series, and therefore it cannot reflect the actual burden and morbidity of EHPVO, and no comparison could be made between endoscopic management and PSS. Data regarding growth retardation was not kept uniformly, therefore being omitted from outcome analysis. As PSS is a difficult procedure, over time the perfection in surgery and therefore the outcome improved, lacking uniformity. Lastly, it is a single-centre experience. This series lacks the experience of Rex shunt.

PSS affords excellent short- and long-term outcome in patients with symptomatic EHPVO, and may be considered for secondary prophylaxis if there are shuntable veins, and surgical expertise is available. Endoscopically non-reversible complications of EHPVO, like PB, hypersplenism, and huge splenomegaly, are better managed surgically. Even splenectomy and devascularization procedure offers good results in patients where PSS is not possible, except in patients with PB. SAA, IGV, and preoperative ERC probably reflect long-standing symptomatic EHPVO with worse surgical outcome, so early surgery in symptomatic cases may be recommended. A patent PSS mostly offers a one-time solution.